COST-EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

by Jennifer Oxley Bruce Corben Sjaanie Koppel Brian Fildes Nisha Jacques Mark Symmons Ian Johnston

Project sponsored by: April 2004 Report No. 217

ii

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE REPORT DOCUMENTATION PAGE

Report No. 217 Title and sub-title: Date April, 2004 ISBN 0 7326 1727 8 Pages 157

Cost-effective infrastructure measures on rural roads

Author(s):

Jennifer Oxley, Bruce Corben, Sjaanie Koppel, Brian Fildes, Nisha Jacques, Mark Symmons, Ian Johnston
Sponsoring Organisation(s): Swedish National Road Administration (SNRA) _______________________________________________________________________________

Abstract:
Rural road safety accounts for a considerable share of the total road safety problem, with up to twothirds of all road fatalities occurring on rural roads. This is mainly due to higher operating speeds, hazardous roadsides and generally poorer road geometry compared to urban roads, multi-functionality and lower enforcement levels. The physical features of the road play a major role in determining mobility and crash and injury risk, and infrastructural improvements have the potential to create a safer travel environment whilst maintaining mobility. This report provides a review of current international literature in regard to road infrastructure and how it can be improved to reduce the frequency and severity of rural road crashes. The findings were used to compile a set of recommendations for cost-effective infrastructure measures. The following features of rural roads were shown to increase the risk of serious injury: the presence of roadside hazards such as trees and poles; intersections and their design and operational features; alignment and design speeds; and road surface conditions. Single-vehicle crashes are common on rural roads and mostly involve vehicles leaving the roadway and colliding with rigid objects or overturning. Measures to address single -vehicle crashes include: i) treatments to reduce speeds and speeding; ii) improved road and shoulder surfaces; iii) lane widening; iv) bridge or culvert widening; v) improved delineation of edge-lines, centre-lines and lane-lines; vi) geometric improvements to curves; vii) roadside barriers; and, viii) clearing of roadside hazards (including establishment of clear-zones). Multi-vehicle collisions are also an important source of road trauma in rural areas and include collisions at intersections, head-on impacts and rear-end collisions. Measures to address multi-vehicle crashes include: i) treatments to reduce speeds and speeding; ii) geometric improvements to intersections including conversion to roundabouts, grade-separation, improved channelisation, sight distance, and medians; iii) geometric improvements to road lengths including use of crashworthy barrier systems or medians and delineation; iv) conversion of undivided roads to divided roads; v) geometric improvements to curves; and, vi) improved road and shoulder surfaces. Fundamental changes to the inherent safety of the rural road system are essential. A system-wide and comprehensive application to key crash problems in rural areas appears to offer the most effective and cost-effective solutions. The most strategically important measures to reduce crash and injury risk are: i) introduction of grade-separated intersections; ii) construction of roundabouts; iii) installation of crashworthy barrier systems; and, iv) introduction of speed reduction measures. Key Words: Infrastructure; Road design; Rural roads; Transportation; Crash risk; Injury risk; Countermeasures Reproduction of this page is authorised
Monash University Accident Research Centre, Building 70, Wellington Road, Clayton, Victoria, 3800, Australia. Telephone: +61 3 9905 4371, Fax: +61 3 9905 4363

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

iii

iv

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

Preface
Project Manager
Mr. Bruce Corben, Senior Research Fellow

Research Team
Dr. Jennifer Oxley, Research Fellow Prof. Brian Fildes, Chair of Road Safety Ms. Sjannie Koppel, Research Fellow Ms. Nisha Jacques, Research Assistant Mr. Mark Symmons, Research Fellow Prof. Ian Johnston, Director

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

vi

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

Contents
EXECUTIVE SUMMARY............................................................................................ XIII 1 INTRODUCTION ....................................................................................................... 1 1.1 DEFINING RURAL ROADS ......................................................................................... 3 1.2 CRASH AND INJURY RISK ON RURAL ROADS ............................................................ 3 1.3 TYPES OF CRASHES OCCURRING ON RURAL ROADS ................................................. 5 1.3.1 Single-vehicle run-off-road crashes .............................................................. 5 1.3.2 Multi-vehicle collisions.................................................................................. 6 1.4 OVERVIEW .............................................................................................................. 8 2 CONTRIBUTING FACTORS ................................................................................... 9 2.1 OVERVIEW .............................................................................................................. 9 2.1.1 Road user factors ........................................................................................... 9 2.1.2 Vehicle factors ............................................................................................. 10 2.1.3 Road factors........................................................................................................ 10 2.1.3.1 Speed and speed limits............................................................................. 10 2.1.3.2 Road Lengths ........................................................................................... 11 2.3.1.3 Horizontal Curves .................................................................................... 11 2.3.1.4 The Roadside ........................................................................................... 12 2.3.1.5 Intersections ............................................................................................. 12 2.3.1.6 Other Factors ........................................................................................... 13 2.2 ROAD USERS ......................................................................................................... 13 2.2.1 Road user inattention and fatigue................................................................ 13 2.2.2 Driver age and risk-taking behaviour ......................................................... 15 2.2.3 Vulnerable road users.................................................................................. 16 2.3 VEHICLE FEATURES .............................................................................................. 17 2.3.1 Vehicle crashworthiness .............................................................................. 17 2.3.2 Heavy vehicles ............................................................................................. 19 2.3.2.1 Aggressivity of heavy vehicles toward smaller vehicles ......................... 19 2.3.2.2 Single-vehicle heavy vehicle crashes ...................................................... 20 2.4 ROAD FEATURES ................................................................................................... 21 2.4.1 Speed and speed limits................................................................................. 21 2.4.1.1 Speed limits on rural roads ...................................................................... 23 2.4.1.2 Speed limits on approaches to provincial cities and towns ..................... 24 2.4.2 Road lengths ................................................................................................ 25 2.4.2.1 Road type ................................................................................................. 25 2.4.2.2 Pavement type.............................................................................................. 28 2.4.2.3 Overtaking lanes ...................................................................................... 29 2.4.3 Horizontal curves......................................................................................... 30 2.4.3.1 Speed on curves ....................................................................................... 32 2.4.3.2 Radius of curve / degree of curvature ...................................................... 34 2.4.3.3 Curve transition ....................................................................................... 35 2.4.3.4 Super-elevation........................................................................................ 36 2.4.3.5 Lateral placement ..................................................................................... 37 2.4.3.6 Trucks ...................................................................................................... 37 2.4.3.7 Weather conditions .................................................................................. 38 2.4.4 The roadside ................................................................................................ 38 2.4.4.1 Contributing factors ................................................................................. 39

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

vii

2.4.5 Intersections .................................................................................................41 2.4.5.1 Poor or inappropriate intersection control and signage ............................43 2.4.5.2 Poor or restricted sight distance ...............................................................44 2.4.5.3 Poor or inappropriate delineation or alignment ........................................45 2.4.5.4 Poor or inadequate lighting ......................................................................47 2.4.5.5 Poor design of freeway interchange ramps ..............................................47 2.4.5.6 Speeds.......................................................................................................49 2.4.5.7 Rail-road crossings .......................................................................................49 2.4.6 Other factors.................................................................................................50 2.4.6.1 Vertical curves ..........................................................................................51 2.4.6.2 Bridges and Culverts ................................................................................51 2.4.6.3 Construction zones ...................................................................................53 3 COUNTERMEASURES ............................................................................................57 3.1 OVERVIEW .............................................................................................................57 3.1.1 Innovative philosophies and visions .............................................................57 3.1.2 Treatments to reduce speeding .....................................................................57 3.1.3 Treatments to road features..........................................................................58 3.1.3.1 Treatments on road lengths ......................................................................59 3.1.3.2 Treatme nt on horizontal curves ................................................................60 3.1.3.3 Treatments on roadsides ...........................................................................61 3.1.3.4 Treatments at intersections .......................................................................61 3.1.3.5 Other improvements .................................................................................63 3.2 PHILOSOPHIES OF ROAD SAFETY PROGRAMS..........................................................65 3.2.1 Swedens Vision Zero.................................................................................66 3.2.2 The Dutch philosophy ...................................................................................67 3.2.3 EuroRAP.......................................................................................................67 3.2.4 Some examples of integrated use of countermeasures to treat long sections of the rural road network .............................................................................................69 3.3 TREATMENTS TO REDUCE SPEEDING AND THE INJURY CONSEQUENCES OF SPEEDING 73 3.3.1 Speed limits...................................................................................................74 3.3.2 Traffic calming measures on approaches to and within rural towns...........76 3.3.3 Perceptual countermeasures to reduce speeding .........................................79 3.3.3.1 The unobtrusive nature of PCMs..............................................................79 3.3.3.2 PCMs treatment evaluation ......................................................................80 3.3.3.3 Limitations with PCMs.............................................................................81 3.3.4 Technologies to enhance speed limit compliance ........................................82 3.4 TREATMENTS TO IMPROVE THE SAFETY OF ROAD LENGTHS ...................................83 3.4.1 Provision of additional lanes........................................................................83 3.4.1.1 Conversion of two- lane undivided roads to four- lane undivided roads ...84 3.4.1.2 Implementation of 2+1 roads .................................................................84 3.4.1.3 Provision of overtaking lanes ...................................................................85 3.4.2 Conversion of undivided roads to divided roads..........................................86 3.4.2.1 Median type ..............................................................................................87 3.4.3 Improved road surface .................................................................................88 3.4.4 Improved delineation....................................................................................90 3.5 MEASURES TO IMPROVE THE SAFETY OF HORIZONTAL CURVES .............................92 3.5.1 Measures to reduce speed on curves ............................................................93 3.5.2 Curve flattening............................................................................................94 3.5.3 Improved alignment design of curves...........................................................95

viii

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

3.5.4 Spiral Transitions ........................................................................................ 96 3.5.5 Provision of signing and road marking ....................................................... 97 3.6 MEASURES TO IMPROVE THE SAFETY OF ROADSIDES ............................................. 98 3.6.1 Provision of a forgiving roadside .............................................................. 100 3.6.1.1 Removal or relocation of trees and poles .............................................. 100 3.6.1.2 Removal or relocation of other objects.................................................. 104 3.6.2 Installation of safety barrier systems......................................................... 105 3.6.3 Shoulder improvements ............................................................................. 108 3.6.3.1 Shoulder pavement and width ............................................................... 108 3.6.3.2 Shoulder slope ....................................................................................... 110 3.6.3.3 Edge- line treatments .............................................................................. 110 3.7 MEASURES TO IMPROVE THE SAFETY OF INTERSECTIONS .................................... 112 3.7.1 Provision of grade-separated intersections ............................................... 112 3.7.2 Provision of appropriate traffic control at at-grade intersections ............ 113 3.7.2.1 Provision of traffic signals with fully- or partially-controlled turning phases 114 3.7.2.2 Provision, placement and form of stop and give-way signs .............. 114 3.7.2.3 Provision of roundabouts ....................................................................... 115 3.7.3 Provision of adequate sight distance......................................................... 116 3.7.4 Provision of adequate delineation and alignment ..................................... 117 3.7.4.1 Provision of channelisation and delineation.......................................... 117 3.7.4.2 Provision of medians ............................................................................. 118 3.7.4.3 Provision of turning lanes at median breaks .......................................... 119 3.7.5 Provision of warning signs at intersections............................................... 120 3.7.6 Provision of adequate lighting at intersections ......................................... 121 3.7.7 Countermeasures for rural rail-road intersections .......................................... 122 3.7.7.1 Slowing the train........................................................................................ 122 3.7.7.2 Removing the crossing .............................................................................. 122 3.7.7.3 Modifying the crossing .............................................................................. 122 3.7.7.4 Influencing road vehicle drivers ................................................................ 124 3.7.7.5 Other countermeasures and issues ............................................................. 125 3.8 OTHER IMPROVEMENTS ...................................................................................... 125 3.8.1 Measures to improve the safety of vertical curves..................................... 126 3.8.2 Measures to improve the safety of bridges and culverts............................ 126 3.8.2.1 Widening of bridge and shoulder width ................................................ 126 3.8.2.2 Installation of guard-rails....................................................................... 127 3.8.2.3 Provision of signage and pavement marking......................................... 127 3.8.3 Measures to improve the safety of construction zones .............................. 128 3.8.3.1 Speed reduction measures...................................................................... 128 3.8.3.2 Measures to warn of lane closures ......................................................... 128 3.8.3.3 Measures to guide drivers through construction zones.......................... 129 3.8.4 Measures to improve the safety of pedestrians and cyclists...................... 130 4 SUMMARY.............................................................................................................. 131 4.1 KEY CRASH TYPES ............................................................................................. 131 4.1.1 Single-vehicle Crashes............................................................................... 131 4.1.2 Multi-vehicle Crashes ................................................................................ 131 4.1.3 Vulnerability Considerations ..................................................................... 131 4.2 INFRASTRUCTURE COUNTERMEASURES .............................................................. 132 4.2.1 Single-vehicle Crash Countermeasures..................................................... 132 4.2.2 Multi-vehicle Crash Countermeasures ...................................................... 132

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

ix

4.3 MOST PROMISING COUNTERMEASURES ...............................................................133 4.3.1 Single-vehicle crashes ................................................................................133 4.3.2 Intersection Crashes ...................................................................................133 4.3.3 Head-on crashes .........................................................................................134 4.4 INCREMENTAL VERSUS FUNDAMENTAL IMPROVEMENTS .....................................134 4.5 THE FUTURE ........................................................................................................135 REFERENCES .................................................................................................................137

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

Figures
FIGURE 1: TYPICAL CURVE TREATMENT .............................................................................. 81 FIGURE 2: TYPICAL INTERSECTION TREATMENT .................................................................. 81

Tables
TABLE 1: EFFECT OF CHANGE IN SPEED ON CRASH AND INJURY RISK .................................... 23 TABLE 2: EFFECTS OF TREATMENT TYPE ON CRASH RISK AND CRASH COST AT VICTORIAN RURAL BLACK- SPOT SITES . ............................................................................................ 72 TABLE 3: P RIORITISED TREATMENTS TO IMPROVE SAFETY FOR MOTORCYCLISTS ................ 89

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

xi

xii

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

EXECUTIVE SUMMARY

Introduction
The physical features of the road play a major role in determining the levels of mobility and safety, particularly for rural roads. Improvements in the infrastructure have the potential to create a safer travel environment, particularly a more crashworthy environment. Traditional safety models attempt to strike a balance between safety and mobility. More recent models, however, such as Vision Zero in Sweden and Sustainable Safety in The Netherlands view safety as the prevailing consideration. These models argue that the road-transport system can only be safe when the road i nfrastructure is designed and operates in a way that explicitly recognises both human tolerance to violent forces and normal human error so that death and serious injuries can be prevented. Effectively, this means reducing travel speed and providing road infrastructure that is forgiving of human error. Road features play a vital role in determining not only the risk of crashing but, more importantly, the severity of injuries sustained in a crash. The following features are critical: the presence of roadside hazards such as trees and poles; intersections and their design and operational features; the geometry of roads and their design speeds; and road surface conditions. Rural roads, more so than their urban counterparts, contribute disproportionately to serious injury crashes, due to higher travel speeds and are therefore a high priority for infrastructure improvements.

The Project
A literature review was undertaken to assess the current state of knowledge in regard to road infrastructure and how it can be improved to reduce the frequency and severity of rural road crashes. In particular, the review addressed: Creating a crashworthy system; Improving the infrastructure to provide safe or safer roadways, including: o Higher design standards, such as divided travel paths, barrier systems, sealed shoulders, skid-resistant surfaces, improved delineation, improved intersection design, and More forgiving roadside settings;

Reducing travel speeds to levels where road users will not experience life-threatening injuries in the event of a crash, including: o o o Lower speed limits, Traffic calming measures, and Perceptual countermeasures;

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

xiii

The findings were used to compile a set of recommendations for cost-effective infrastructure measures.

The Problem
Rural road trauma has been, for many decades, a major concern worldwide. For example, in OECD countries, traffic crashes in rural areas account for some 60% of all road deaths. Travel on rural roads usually occurs at high-speeds. Single-vehicle crashes typically account for 30-40% of injury-producing crashes. Single-vehicle crashes mostly involve vehicles leaving the roadway and colliding with rigid objects or overturning. It is common for errant vehicles to strike roadside trees, poles, embankments or a variety of man-made structures. Vehicles overturn when roadsides are uneven, too steep, or both. Crashes in which vehicles overturn or strike roadside hazards tend to result in severe injuries because of the rigid nature and often narrow dimensions of the objects struck, as well as the high impact speeds. Modern-day vehicles, with current standards of crashworthiness, offer only limited protection to their occupants from the high levels of intrusion and deceleration experienced in these types of crashes. Multi-vehicle collisions are also an important source of road trauma in rural areas. Collisions at intersections, head-on impacts and rear-end collisions (or similar) are the main multi-vehicle crash categories in rural areas. Again, injuries tend to be severe because of the high impact speeds and the inability of vehicles to adequately protect their occupants in many of the common crash types at these speeds. Head-on impacts and intersection crashes involving side-impacts occur frequently and with high severity. Pedestrian and bicyclist injuries are, in general, relatively infrequent, although in some countries (e.g., South Africa and Poland), they represent a large share of rural road trauma. Pedestrians, bicyclists and motorcyclists are especially vulnerable in high-speed rural settings, where specific facilities such as footpaths and dedicated lanes are rarely provided.

Single-vehicle Crashes
Straight Roadway Sections Single-vehicle crashes occur primarily away from intersections. While over-represented on rural curves, the majority of single-vehicle crashes occur on straight sections of roadway. Driver fatigue, alcohol and inattention play an important role in crash risk, as does high travel speed. When a driver falls asleep, loses concentration or is otherwise experiencing some deficit in driving performance, it is common for vehicles to leave the formed roadway and enter the roadside at speeds of around 100-120 km/h. In many cases, such speeds are within the legal limits designated for roads in rural areas. At these speeds, it is frequently impossible for drivers (or riders) to slow significantly before colliding with any of the numerous objects lining the rural roadside. In other cases, steep and/or uneven roadsides precipitate vehicle rollover. Given the nature of roadside hazards and the speed of departure of errant vehicles, modern vehicles cannot protect their occupants adequately from the violent forces and decelerations that result. The consequences of these types of crashes are even more concerning where motorcyclists are involved. While the frequency of single-vehicle crashes is often higher, because of traffic volumes, on so-called high-standard rural roads the risk per vehicle-kilometre of travel is highest where roadways and lanes are narrow, shoulders are unsealed, paved edges are broken and

xiv

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

differences exist in the level between the paved and unpaved surfaces. Crash risk also increases with higher speed limits and with higher vehicle operating speeds. At high speed, vehicles travelling in narrow lanes are more likely to leave their lane and, in many cases, the paved surface than when lanes are wider. Having left the paved surface, the opportunity for the driver to slow and regain directional control is diminished where only gravel and/or narrow shoulders are provided. Hazardous objects, hazardous terrain, or both, greatly increase the likelihood of severe injuries to vehicle occupants. Vertical curves also increase crash risk, largely because of the shorter stopping sight distances available when a hazard, such as an intersection or curve, is located beyond a crest. Bridges and culverts are significant contributors to rural road trauma, especially for singlevehicle crashes. Narrow bridges and/or narrow lanes, high vehicle speeds, and rigid or otherwise hazardous bridge and culvert structures are the chief contributors to crash and injury risk. Of considerable importance is the relatively high cost of constructing or upgrading bridges. Consequently, older low-standard bridges are often not upgraded when road widening takes place to meet increasing traffic demands in rural areas. Over time, more and more vehicle occupants and riders are exposed to these hazards, typically at closer spacing and higher travel speeds. In addition to bridge width and the absence of shoulders, bridges and culverts located on combinations of curves and downgrades experience a particularly high crash risk. Poor bridge railing design is also a contributing factor to injury risk. Curves Single-vehicle crashes are over-represented on curves, relative to t heir proportion by length of the road network because of the natural tendency for vehicles to continue in a straight line, unless subject to sideways forces that enable the vehicle to track safely on a curved path. The safe tracking of a vehicle around a c urve at relatively high speed requires greater skill by drivers than when driving along a straight section of roadway. The demands on the driver to stay safely within his/her lane are amplified if there is inconsistency and lack of predictability in the alignment ahead. The effects of inconsistency are of greater concern when curves are severe, especially when sharper than expected and travel speeds are high. The crash risk is further heightened when speed limits and/or vehicle operating speeds on approach to a curve are markedly greater than the design speed of the curve. The designspeed concept was created to provide a consistent operating speed profile along an alignment using two principles; first, all curves on an alignment should be designed for the same speed and secondly, the design speed should reflect the speed that a high proportion of drivers desire to maintain. As with single-vehicle crashes on straight sections of roadway, narrow lanes and unsealed shoulders further increase crash risk, and injury risk rises with the presence of roadside hazards and/or non-traversable terrain. Due to the greater difficulty in perceiving the true road alignment ahead, the risk of single-vehicle crashes on curves also rises with prevailing light and weather conditions such as darkness, rain or other adverse weather. In safely negotiating a curve at high speed, a vehicle must not exceed the sideways force available through the friction generated between the vehicle tyres and the roadway. Deficient delineation of a curve, adverse or inadequate superelevation within curves and poor skid resistance of paved surfaces heighten the risk of single-vehicle crashes on curves. In the same way that drivers leave the roadway on straight sections of rural road, due to

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

xv

fatigue, alcohol, inattention, etc., so too will these deficits in human performance add to crash risk on curves. Both changing vertical alignment and the presence of bridges or culverts on curved sections of rural roadway also contribute to higher crash and injury risk. Increased crash risk occurs through the greater demands on the driver to choose the appropriate speed when entering a curve, which may be of unknown sharpness to the driver, as well as being able to see far enough ahead to stay within the travelling lane at generally high speed. On narrow curves, drivers tend to position their vehicles closer to the centreline and thereby increase their risk of head-on collisions. In the case of single-vehicle collisions with bridges or culverts, the severity of injuries to vehicle occupants appears to be higher on curves than on straight sections of roadway. This may be due to the generally higher impact angles involved, compared with straight sections. Multi-vehicle crashes Multi-vehicle crashes commonly occur either at intersections and driveways, or in the form of head-on or rear-end crashes away from intersections. To a lesser but significant extent, serious multi-vehicle crashes also occur in the vicinity of grade-separated intersections, namely, near entry and exit ramps forming part of, sometimes complex, freeway interchanges. Intersections Intersections, by definition, are locations within the road network where the paths of vehicles cross. Intersecting paths can vary from the simple crossing of paths at (approximately) right angles to a large number possible points of conflict when turning movements and multiple approaches take place. The nature of terrain in some rural areas can make the identification of intersections more difficult. Terrain that is mountainous, hilly or undulating may obscure the presence of intersections from the view of an approaching driver, while flat featureless areas may offer few visual cues to the presence of an intersecting road. This problem is more acute where simple forms of intersection control apply, such as stop or give way (yield, in the USA) signs. Especially in rural areas, the successful negotiation of intersections requires the selection of safe gaps by drivers who are not afforded priority through the intersection. The selection of safe gaps can be a particularly challenging task in rural settings where vehicle approach speeds are typically high. The task is made more difficult when: Sight distances are restricted, perhaps by roadside vegetation or by changes in the horizontal geometry and/or vertical alignment of the approach roads; Traffic volumes are high thereby reducing the availability of safe gaps; Intersection layout is complex; Intersection control is inadequate for the circumstances; Driver performance deficits, temporary or permanent, such as alcohol, fatigue or agerelated deficits are present.

xvi

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

When at least one driver fails in the gap selection task at a rural intersection a number of outcomes commonly arise. First, two or more vehicles will collide in a side-impact type collision, at or near a right-angle. Such crash scenarios are not only frequent at rural intersections, but they tend to produce severe injuries to vehicle occupants. There are several reasons for these high-severity injury outcomes, namely: Impact speeds are more likely to be high; Most vehicle types offer only limited protection to their occupants in a side-impact, a major problem being the lack of energy absorbing (crush) space available between the side of a passenger vehicle and its occupants; Incompatibility between vehicle types involved in side-on impacts. Passenger vehicles struck by larger vehicles of incompatible geometric design (e.g., higher, stiffer structures), such as four-wheel-drive vehicles, both rigid and articulated trucks, and buses, are not capable of protecting their occupants from serious injury at more moderate speeds (as, for example, in urban areas) and so the occupant protection task at high-speed rural intersections is particularly challenging. It is commonplace on rural roads throughout the world for heavy vehicles, with their mismatched design features, to mix with passenger vehicles.

Secondly, another frequently occurring gap selection error at rural intersections involves a vehicle turning across the path of a vehicle approaching from the opposite direction (i.e., along the same road). As with cross-traffic type collisions described above, turn-against crashes occur because of difficulties in judging the speed and distance of approaching traffic and then executing the turn in a safe manner. The gap selection task is made even more difficult by local obstructions to sight distance (other vehicles, crests, curves, vegetation, etc.), the high speeds and accelerations of approaching traffic, wide, multi-lane approaches and generally complex intersection layouts. Once again, high prevailing speeds, head-on or near head-on impact angles and incompatible vehicle types combine to raise injury risk when such crashes occur. Thirdly, rear-end (or rear-end type) crashes commonly occur at rural intersections and driveways, often as a result of left or right turning vehicles being struck by following vehicles. While injury risk may be lower than for cross-traffic and turn-against crash types, rear-end collisions represent a common source of serious injury. Clearly, crash and injury risk are both higher in rural than in urban areas because of the higher travel speeds involved. Obstructions to sight distances, low-conspicuity intersections and driveways, and turns that must be made from a lane shared with through-traffic also contribute to the risk of rear-end crashes at rural intersections. Multi-vehicle crashes at intersections will result in more severe injuries when one or more of the vehicles is a motorcyclist, a bicyclist, a heavy vehicle or when older people are involved (their greater frailty places them at heightened risk when a crash occurs). Head-on Collisions One of the major categories of serious trauma on rural roads involves head-on collisions away from intersections. Head-on crashes typically occur:

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

xvii

When drivers, due to fatigue, alcohol, inattention, or other (temporary) declines in driving performance, drift from their lanes into the path of traffic travelling in the opposite direction. Such crashes occur on undivided roads, as well as on divided roads provided errant vehicles are not prevented by median barriers, roadside objects or non-traversable median surfaces from reaching the opposite carriageway; When drivers drift from their lane towards the roadside (rather than the centre of the road), enter an unsealed shoulder or the roadside beyond and, in attempting to regain vehicle control, travel into the path of traffic from the opposite direction. The presence of gravel or unsealed shoulders, and significant differences in height between the paved and unpaved surfaces increase the risk of this type of crash; When overtaking other, slower moving traffic, drivers or riders misjudge the safety of available gaps in oncoming traffic; When negotiating rural curves, drivers leave their lanes either towards the roadside or to the centre of the road. The former potentially presents many risks where hazardous roadsides exist, while the latter presents opportunities for high-speed, high-severity collisions with opposing traffic.

In all of the above scenarios, passenger vehicles, even those with high standards of crashworthiness, are unable to protect their occupants at speeds above about 70 km/h. Survivable impact speeds will be greatly reduced if head-on collisions involve cars hitting trucks, buses or other vehicles of incompatible design, or where the more vulnerable road users are involved (e.g., older people as vehicle occupants, bicyclists, motorcyclists or occupants of small vehicles). Construction Zones Construction zones experience as much as a doubling of crash rate compared with the period prior to construction. The increased risk of crashes near and within construction zones has been partly attributed to the high volumes of traffic travelling at high-speeds in rural environments. In addition, there is often a need for drivers to rapidly lower speed and merge with other traffic on approach to construction zones, where lane closures or diversions have been temporarily introduced. Rear-end crashes are common. These added demands on the driver, together with the occasional requirement for the drivers of through-vehicles to avoid construction workers on foot, to give way to heavy construction vehicles and to avoid temporary roadside hazards, such as changes in surface levels, and the absence, or inappropriate use, of concrete barriers make construction zones on rural roads especially hazardous.

The Infrastructure-Based Solutions

Of strategic significance is whether infrastructure countermeasures exert their effects on crash risk, injury risk, or both. It is also important to understand whether they act on all crash types or only on a particular subset of crashes. Treatments also range from relatively inexpensive, making them highly suited to addressing problems that are common throughout the road-transport system, to major improvements requiring a large capital investment, which effectively limits their potential use to only a small proportion of the network.

xviii

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

Combining the notions of the effect of a countermeasure on crash and injury risk, with the whole-of-life costs of the countermeasure, leads to the concept of cost-effectiveness and how to measure it. Reliable estimates of cost-effectiveness facilitate rational decisionmaking and, ultimately, the achievement of the best safety gains from available resources. This summary of infrastructure countermeasures specifically addresses crash type, where this is relevant. Some countermeasures, such as speed limit reduction have the potential to affect both crash and injury risk for all crash types, while other countermeasures have their principal effect on a particular category of crash. This summary endeavours to reflect these important differences, while also addressing the key crash categories of single-vehicle crashes on both straight and curved sections of roadway, multi-vehicle crashes at intersections, and head-on crashes occurring generally along stretches of rural roadway between intersections. In addition, attention is given to the needs of the more vulnerable road user categories, namely, pedestrians, motorcyclists and bicyclists. Finally, this summary endeavours to make important distinctions between infrastructure measures that lead to fundamental changes in safety and infrastructure measures that offer incremental improvements only.

Single-vehicle Crashes
On straight or curved sections of rural road, a substantial proportion of drivers leave the roadway as a result of temporary deficits in functional performance. Common reasons for reduced functional performance among drivers and riders include fatigue, alcohol, other drugs, distraction or loss of attention. On curved sections, the added challenge for drivers and riders of negotiating, in a high-speed environment, a road that changes its alignment, especially in unpredictable ways, further increases crash risk. The following infrastructure countermeasures have been used to address single-vehicle crashes: Treatments to reduce speeds and speeding: o o Lower speed limits; Traffic calming on the approaches to and within rural towns, including pavement narrowing, refuge islands, road surface alterations, raised road surfaces, hazard marker posts, countdown signs, Wundt illusions, chicanes and transverse lines; Perceptual countermeasures, comprising transverse pavement markings, Drenthe treatments (visual and tactile), physical roadway design to enforce operating speeds, peripheral lane-edge and herringbone markings, centre-line marking, other edge-line markings, enhanced post curvature treatments (with and without ascending height posts) and hatched median treatments; Vehicle-activated signs to advise of travel speed.

Improved road and shoulder surfaces: o o Skid resistant pavements; Wake-up pavements;

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

xix

o o

Shoulder sealing; Shoulder rumble-bars.

Lane widening; Bridge or culvert widening where narrow clearances exist for passing traffic; Delineation: o o o o Edge-lines; Centre-lines; Lane-lines; Profile edge-lines.

Single-vehicle Crashes on Curves Typical infrastructure improvements used specifically to address single-vehicle crashes on curves include: Treatments to reduce speeds and speeding: o Speed limit reduction or advisory speed warning signs, especially where a better match can be achieved between vehicle operating speeds and the design speed of the curve(s); Perceptual countermeasures, particularly enhanced post curvature treatments as described above.

Geometric improvements to curves, including curve flattening, improvements to curve transition design, lane widening and the provision of appropriate curve superelevation. Curve delineation, including post-mounted delineators, painted edge-lining (regular or audible/tactile), chevron hazard markers and raised reflective hazard markers.

Single-vehicle crashes and Reduction of Injury Risk The following infrastructure treatments have been used to reduce injury risk in singlevehicle crashes: Roadside barriers, spanning a range of barrier types such as wire rope safety barriers, steel-beam guard-rail and concrete, as well as crash cushions of various forms. Two distinctly different types of practice have been found, namely: o o Localised use of roadside barriers to protect passing traffic from specific hazards, or Use of roadside barriers over extended lengths of roadway (i.e., sections extending over tens or hundreds of kilometres in length);

xx

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

Flattening of roadside slopes. Clearing of roadside hazards, including the establishment of clear-zones. Replacement of man-made roadside hazards with structures of more forgiving design, such as frangible lighting columns, sign supports or lower profile culvert designs.

Single-vehicle crashes - Summary Of the numerous measures outlined above, the greatest potential for reducing serious injuries resulting from single-vehicle crashes in rural areas comes from: Speed limit reductions especially along roads of curved alignment and/or with hazardous roadsides; because of the powerful relationship that exists between vehicle speeds and crash and injury risk for all crash types, especially single-vehicle crashes; Crashworthy roadside barriers erected over extended lengths of rural roadway; because of the extreme difficulty in preventing a substantial number of drivers and riders in high-speed rural environments from leaving roadways; and Skid resistant pavements and shoulder sealing with tactile edge-lining; because they assist a sizable proportion of the drivers and riders of errant vehicles to regain control or their vehicles without entering the roadside and, desirably, without striking, unnecessarily, roadside barriers.

Multi-vehicle Crashes
Intersections A major target for reducing crash and injury risk in rural areas is intersections in highspeed settings, especially intersections with four or more approaches. Common crash types include cross-traffic, turn-against oncoming traffic and rear-end collisions. High travel speeds on rural roads are a major contributor to crash risk, as is the difficulty experienced by crash-involved drivers in selecting safe gaps in approaching traffic streams. The following infrastructure countermeasures have been used at rural intersections: Treatments to reduce speeds and speeding: o o Speed limit reduction on the approach to intersections; Perceptual countermeasures, particularly transverse peripheral lane-edge and herringbone markings. pavement markings,

Conversion to roundabouts. Traffic signals. Grade-separation, with various alternative designs for exit and entry ramps.

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

xxi

Channelisation of various types, including closure of selected approaches, restrictions to various movements and approach islands to better define the presence of intersections. Removal of obstructions to sight distance. Signing to provide directional priority. Medians on major roads and/or exclusive/protected turn lanes at intersections. Skid resistant pavements. Limited access from private driveways and minor side-roads.

Most of these are effective, to varying degrees, in reducing crash occurrence. However, few also substantially limit the risk of serious injury when crashes occur (e.g., traffic signals or sign-controlled intersections). For intersection crashes in high-speed rural settings, roundabouts have been shown to not only reduce crash frequency by some 70 to 80%, but to reduce crash costs (reflecting injury severity) by around 90%. That is, roundabouts in rural areas have been found to reduce, in a fundamental way, both crash risk and the risk of serious injury to the occupants of vehicles colliding at intersections. This is true for the main crash types occurring at rural intersections, namely, cross-traffic, turn-against and, in some cases, rear-end crashes. Grade-separation, while virtually eliminating major conflicts, still allows high-speed crashes to occur where entry/exit ramps interact with the major flow of traffic. For some forms of grade-separation, there may be no ramps, a limited number of ramps or full provision of ramps as part of the intersection treatment. Head-on Crashes Other multi-vehicle crash types for which infrastructure countermeasures have been used and, in some cases, evaluated include: Delineation: o o o o Centre-lines (double lines pavement markers, etc.); Edge-lines; Lane-lines; Chevron hazard markers and/or post-mounted delineators on curves. (regular or audible/tactile), raised reflective

Treatments to reduce speeds and speeding: o Speed limit reduction or advisory speed warning signs, especially where a better match can be achieved between vehicle operating speeds and the design speed of the curve(s); Perceptual countermeasures, but specifically treatments and hatched median treatments. enhanced post curvature

xxii

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

Improved roadway cross-section: o o o Two-lane undivided to four-lane undivided roads; Overtaking lanes; Conversion of undivided roads to divided roads, with variations to median type, covering the use of central islands, medians with either mound or swale profiles, wide or narrow medians, flush medians or medians which include two-way turning lanes.

Geometric improvements to curves, including curve flattening, improvements to curve transition design, lane widening, and the provision of appropriate curve superelevation. Improved road and shoulder surfaces: o o o o Skid resistant pavements; Wake-up pavements; Shoulder sealing; Shoulder rumble-bars.

Crashworthy barrier systems on medians or to separate opposing vehicle directions along otherwise undivided rural roads (e.g., the 2+1 road configuration introduced in recent years in Sweden).

Measures that make a fundamental improvement in the inherent safety of rural roads, with respect to head-on crashes, are favoured. These include the use of crashworthy barrier systems on medians or to separate opposing vehicle directions along undivided rural roads, reductions in speed limits to substantially reduce crash risk and to magnify the injury risk reductions offered by crashworthy barrier systems in separating opposing directions of high-speed traffic. Crashworthy barrier systems, because of the difficulty in predicting where head-on crashes will occur, should be installed over extended lengths (i.e., tens or hundreds of kilometres) of rural roadway. Geometric improvements to curves also have the potential to substantially reduce the incidence of head-on crashes, however, these treatments are both costly and unlikely to reduce, fundamentally, either the risk of head-on collisions or the serious injury consequences of such crashes in high-speed rural settings.

Incremental versus Fundamental Change to Crash and Injury Risk

Many studies in the literature investigate rural crash problems but relatively few include rigorous evaluations of the effectiveness or cost-effectiveness of specific infrastructure measures. Some infrastructure countermeasures affect crash risk, injury risk, or both, incrementally. That is, they have the potential to eliminate a relatively minor proportion of crashes and leave a major residual crash problem. Other countermeasures have the potential to eliminate a major proportion of crashes by changing, in a fundamental way, the

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

xxiii

inherent safety of the road system, or elements of it. Relatively few infrastructure measures developed and used to date have the potential to improve, fundamentally, crash and injury risk. Having regard to the pattern of key crash types on rural roads, the most strategically important measures found from the literature review, were: 1. Grade-separated intersections can virtually eliminate intersection crashes by completely separating in space, conflicting vehicle paths. That is, they have the potential to be 100% effective but in terms of cost-effectiveness, the high cost of grade-separation may make them less attractive than some other alternatives. Roundabouts can reduce casualty crash risk at intersections by between 70 and 80%, crash costs by around 90% and have been found to result in Benefit-Cost Ratios (BCRs) of around 19 when constructed at rural intersections with a high crash record. Crashworthy barrier systems when used over extended lengths of high-speed rural road, barrier systems have the potential to reduce fatal and serious injuries to the occupants of errant vehicles by around 90%, with conservatively estimated BCRs of around eight. Flexible barrier systems can address two major rural crash categories, namely single-vehicle and head-on crashes, on straight or curved road sections, without the need for costly road duplication and/or geometric improvements to rural infrastructure. Speed reduction speed has a fundamental relationship with both crash and injury risk. Substantial improvements are achievable for all crash types from minor reductions in travel speed and even smaller reductions in impact speed. However, while these effects span the full range of crash types that occur in rural areas, reductions in rural speeds tend to be viewed as of limited effectiveness, as both travel and impact speeds in rural areas are often high (i.e., 100-120 km/h). At impact speeds often found in rural crashes, modern-day vehicles, even those fitted with the best available safety features, offer only limited protection from serious injury to their occupants. For particular forms of infrastructure, it appears that lower impact speeds can be especially effective in reducing the risk of severe injuries in rural settings. The main reason for this relates to the combined, or synergistic, effects of lower impact speeds and more forgiving (or crashworthy) infrastructure, where the latter limits the transfer of kinetic energy at impact to the vehicle and its occupants. That is, the overall benefits can be greater that the sum of the individual benefits of more forgiving infrastructure and speed reduction. Put another way, there are substantial synergistic benefits when speed and forgiving infrastructure interact in an optimal/compatible way. Infrastructure countermeasures that eliminate conflict or combine, in an optimal way, the management of speed and energy transfer at impact appear to offer the most promising options for enhancing rural road safety. A system-wide and comprehensive application to key crash problems in rural areas appears to offer the most effective and cost-effective solutions to rural road trauma in the foreseeable future.

2.

3.

4.

xxiv

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

COST-EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

1 INTRODUCTION

Rural road safety accounts for a considerable share of the total road safety problem, with approximately 60 percent of all road fatalities in OECD countries occurring on rural roads (see Section 1.1 for an operational definition). The risk of being killed is generally higher on rural roads than on urban roads and is four to six times higher than on motorways (OECD, 1999). The key contributing elements to the increased severity of rural crashes compared to urban crashes include higher operating speeds, hazardous roadsides and generally poorer road geometry (many roads now no longer perform the function for which they were constructed), multi-functionality and lower enforcement levels. The nature of rural roads, more-so than for their urban counterparts, plays a major role, not only in terms of the risk of crashing but more importantly in terms of the severity of injuries sustained. The following features of rural roads have been shown to increase the risk of serious injury: the presence of roadside hazards such as trees and poles, intersections and their design and operational features, alignment and design speeds, and road surface conditions. Moreover, the effect of these features is magnified by high travel speeds. The Monash University Accident Research Centre (MUARC) was commissioned by Vgverket the Swedish National Road Administration (SNRA) to undertake an extensive literature review to assess the key issues and current state of knowledge in regard to rural road safety and the impact of infrastructure on the frequency and severity of crashes and to make recommendations on the effectiveness of measures to reduce crash and serious injury risk in these environments. The literature search was undertaken on the Australian Transport Index, which contains over 135,000 records of publications from throughout the world, on roads, transport and related fields. Records cover books, reports, journal articles and conference papers. The database is produced by the ARRB Transport Research Library and is Australias major transport database. The TRIS database, produced by the US National Transport Library, and the ITRD database, produced by the OECD, were also searched for relevant references. Three separate searches were conducted under the following headings: Crashes on rural roads, Crash and injury risk on rural roads, Effectiveness of countermeasures to reduce crash and injury risk on rural roads.

This introductory Chapter outlines the problem by defining rural roads, reporting on their crash and injury risk, and, examining the types of crashes that commonly occur.

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

Chapter two examines the literature on the contributing factors. Crashes are complex in nature, often involving several contributing factors. Nevertheless, a number of factors have been identified that influence crash and injury risk, including road user factors, vehicle factors, and road factors. Road user and vehicle factors are discussed briefly before turning to the influence of road infrastructure on crash and injury risk. This section forms the main part of the review and includes extensive discussions of the following issues: The contribution of road features to crash and injury risk in the following settings: o o o o o Speed, speeding and speed limits, Road geometry (straight sections, horizontal curves and passing zones), Roadsides (sealed, unsealed shoulders, presence of trees, poles and other roadside furniture), Intersections, Other geometric features.

Chapter three presents the literature on countermeasures that aim to improve safety on rural roads. First, a number of innovative philosophies are discussed. Secondly, because both speed and speeding are major factors in rural road crashes, treatments to reduce speeding and the injury consequences of speeding are separately discussed. Thirdly, measures to improve safety in the following settings are described: Measures to improve the safety on road lengths such as lane configurations, 2+1 roadways, medians, road surface and delineation, Measures to improve the safety of horizontal curves such as speed reductions, curvature, super-elevation, flattening, transitions, pavement and signing, Measures to improve the safety of roadsides such as provision of clear-zones, barriers, guard-rail and shoulder sealing), Measures to improve safety at intersections such as intersection design and operation, signage, signals, roundabouts, channelisation, traffic calming measures, etc., Interactive effects of measures, and Other improvements (this includes measures to improve the safety of approaches to provincial towns and cities, measures to improve the safety of bridges and culverts, and measures to improve the safety of vulnerable road users, namely pedestrians, bicyclists and motorcyclists).

It should be noted here that the approach used in this literature review is not a critical examination of studies, but to report studies as equally robust and to review the range of countermeasures used world-wide to reduce crash and injury risk on rural roads.

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

1.1

DEFINING RURAL ROADS

There is no clear and simple way of defining rural roads. Every country has a system of road classification. For example, in England, roads are classified as A, B, and C class roads while in Portugal there is a larger range (EN, IP, IC, etc). These nomenclatures do not necessarily reflect the features of roads such as traffic speed, traffic volume, degree of congestion, road environment, etc. Some countries define roads by the area surrounding them. For instance, a rural road could be any road outside an urban area (often defined as h aving a population of at least 50,000 people). Others define rural roads by the type of trip made; for example, long-distance, high speed, intercity trips are generally made on motorways through rural areas, yet these roads may not be considered rural roads. Rural roads also can be defined according to the Average Annual Daily Traffic (AADT). Henderson (1995) distinguished between urban and rural based on population distribution, density and degree of remoteness. He defined urban crashes as having occurred on roads in major population centres, and rural crashes occurring on the remainder of roads, divided into four categories. These categories are defined as boundary (crashes in zones between urban centres and rural regions), non-remote (crashes in locations less than 100 km from a capital city or a city of more than 100,000 population, semi-remote (crashes in locations more than 100 km from a capital city or a city of more than 100,000 population, and remote (crashes in locations more than 100 km from a town of at least 5,000 population). For the purposes of this review, we have taken the broadest definition of rural roads in order to fully examine the safety problems and suggested countermeasures. For this review, a rural road is any road outside an urbanized or built-up area. In effect, we have had to accept a given authors definition where it has not been made explicit. This is not crucial to the main purpose of this review although it does mean that estimates of frequency and rate are relatively imprecise. 1.2 CRASH AND INJURY RISK ON RURAL ROADS

Both crash and injury risk on rural roads differ from those on urban roads. While crash rates are generally higher in urban areas (because of the greater number of intersections and higher traffic volumes), crashes on rural roads tend to be more severe (because of the greater speeds and diversity of road conditions). Wegman (1995) argued that the high crash and injury rates can be explained by a number of factors including high speeds, high speed variation, and different types of road users occupying the same space. European countries, the United States and Australia all report high rates of death and injury on rural roads. Many road-related deaths in Europe (40%) occur on major roads outside built-up areas, most on single-carriageway roads with speed limits of more than 80 km/h. In England, for example, nine percent of deaths on major roads outside built-up areas are on the motorways, 19% on dual carriageways, 38% on single carriageways of national and regional importance and 34% on other single carriageways (Schouten, 2002). In Denmark, approximately 30% of all traffic crashes occur in rural areas. More importantly, rural crashes comprise 44% of all casualties and 64% of all fatalities (Nielsen, 2000). The Danish road network consists predominantly of two-lane roads with a speed limit of 80 km/h and it is here that the largest proportion of crashes in rural areas occurs. Swiss data,

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

similarly, indicate high crash and injury risk on rural roads. The death rate is estimated to be five times higher on rural roads than on motorways and about 4% higher than in built-up areas (Hehlen, 1992). Hughes (1994) noted that crashes on rural roads in Cambridgeshire (UK) accounted for 44% of all traffic crashes and 51% of all traffic crash casualties in 1992. Moreover, rural road crashes accounted for 89% of fatal crashes and 90% of all fatalities, 47% of all serious crashes and 55% of all serious casualties, and 42% of all slight crashes and 49% of all slight casualties. He also noted that between 1983 and 1992, rural traffic crashes averaged 1.6 casualties per crash compared to urban traffic crashes which averaged 1.2 casualties per crash. Australian data report similar crash rates. In his overview of the rural road safety problem in Australia, Henderson (1995) noted that nearly half of all fatal crashes occurred on rural roads and another 14% in regional towns. Lydon (1997), similarly, reported that approximately 48% of all fatal and 40% of casualty crashes occur in rural areas. In the State of Victoria alone (one which has higher density metropolitan regions than other Australian states), Ogden (1997) reported that about one-third of fatal crashes and one-fifth of casualty crashes take place on open roads in rural areas. In the USA, too, crash and fatality rates on low-volume rural roads are higher than other highways (Stamatiadis, Jones & Aultman-Hall, 1999). Tessmer (1996) undertook a comparative analysis of rural and urban crashes in the USA and noted the following: There are approximately 40% more crashes, vehicles involved, individuals involved and deaths in rural areas than in urban areas, even though there are fewer vehicle kilometres travelled in rural areas compared to urban areas, While fatal crashes occur mostly on roads with 55 mph (89 km/h) speed limits in both rural and urban areas, rural roads with this speed limit account for almost 70% of rural fatal crashes, while urban roads with this speed limit account for 22% of urban fatal crashes, Rural fatal crashes result in multiple deaths 21% of the time, whereas urban fatal crashes result in multiple deaths seven percent of the time, A larger proportion of rural fatal crashes involve trucks compared to urban fatal crashes (22% and 10% respectively), The proportion of fatal rural crashes that involve head-on collisions (25%) is higher than the proportion of fatal urban crashes that involve head-on collisions (15%), A larger proportion of individuals in fatal rural crashes are passengers (40%) than in urban crashes (32%), but fewer pedestrian fatalities occur in rural areas compared to urban areas (4% compared to 11%), Crashes involving a single vehicle striking a fixed object or a vehicle rollover are more prevalent in rural areas than urban areas, while vehicle striking another vehicle occur with greater frequency in urban crashes than in rural crashes, The proportion of males involved in rural crashes is higher than the corresponding proportion of males involved in urban crashes,

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

Rural crashes result in a more severe injury outcome than urban crashes an individual involved in a crash is up to three times more likely to die as a result of a rural crashes than to die from an urban crash. TYPES OF CRASHES OCCURRING ON RURAL ROADS

1.3

The literature has identified four major types of crashes that occur on rural roads. These are: run-off-road, head-on, rear-end and intersection crashes (Henderson, 1995; Nielsen, 2000; McLean, Baldcock & Kloeden, 2002), with run-off-road and head-on crashes generally resulting in most injuries and death. Rural road crashes frequently occur on two-lane sealed roads, both on straight sections and on curves, involving loss of control. Henderson (1995) reported that fatal run-off-road crashes where the driver initially lost control on the left shoulder of the road (equivalent to the right shoulder in the USA and most European countries) accounted for 29% of rural crashes. About 40% of these vehicles departed the road to the left and crashed into a fixed object. The remainder re-entered the road and either hit another vehicle head-on, veered to the other side of the road or overturned on the road. 1.3.1 Single-vehicle run-off-road crashes

Single-vehicle run-off-road crashes comprise a large proportion of rural road trauma and a number of studies world-wide have examined the circumstances of these types of crashes, most highlighting the extreme hazards associated with collisions with trees, poles and other fixed objects on the roadside (Armour & Cinquegrana, 1990; Haworth, Vulcan, Bowland & Pronk, 1997; Henderson, 1995; Lydon, 1997; McLean et al., 2002; Nielsen, 2000). Wegman (1995) reported that 32% of rural crashes in The Netherlands involved a singlevehicle hitting a fixed object. Likewise, Toivonen and Niskanen (1998) showed that, on semi-motorways in Finland (particularly lower volume roads), single-vehicle crashes accounted for a little over a third of all crashes resulting in injury or death. Haworth et al. (1997) estimated that single-vehicle crashes comprised approximately 30% of road trauma in Victoria, Australia. Gelston (1998), too, reported that a high proportion of rural fatal crashes (60-65%) and rural casualty crashes (60%) in South Australia involved a singlevehicle running off the road (compared with only 25% and 15-20% for metropolitan areas respectively). A number of causes for these types of crashes have been identified, including: excessive speed, wheels on the verge or soft shoulder, fatigue and alcohol (Henderson, 1995; Nielsen, 2000). Single-vehicle run-off-road crashes generally occur following loss of control (often on sealed roads with unsealed shoulders). An early, detailed study of 155 individual single-vehicle crashes in Victoria, Australia, identified the contributing factors for these crashes (Armour & Cinquegrana, 1990). The main factors included: unsealed shoulders (contributing to between 25 and 40% of crashes), roadside objects (contributing to between 20 and 34% of crashes), batter slopes (contributing to between 21 and 35% of crashes), driver fatigue (contributing to between 25 and 40% of crashes), and high speed (contributing to between 18 and 32% of crashes). Other factors included weather conditions, traffic lanes, shoulder condition, drains and alcohol. This study clearly identified the important roles of roadside objects (most

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

commonly trees), unsealed shoulders, poor shoulder condition and driver fatigue in rural crashes. Haworth et al. (1997) later investigated a sample of 127 fatal single-vehicle crashes in metropolitan and rural Victoria (Australia) to determine the circumstances and factors contributing to them and found that, of the crashes investigated, 40% occurred on rural roads. Of these, most crashes occurred on two-way undivided roads with speed limits at or above 100 km/h, almost 75% of crashes involved an impact with a tree or pole or both, crashes were most common at night and on weekends, most involved young male drivers, over 50% of drivers and riders had BAC levels over the legal limit (>0.05), and 40% occurred on curves (mostly with radii between 100-200m). More recently, McLean et al. (2002) investigated rural crashes in South Australia and found that the most common type of crash on rural roads in South Australia involved a single vehicle (comprising over 44% of the sample). These crashes typically occurred on sealed roads with an unsealed shoulder and no edge-lining, involved right-curved sections of road, and young, inexperienced or drunk drivers. Other common causal factors identified included problems with the road surface (either a deterioration of the surface or the presence of foreign material), a lack of edge-lining, and problems with the road layout. Furthermore, when those crashes that were considered to involve loss of vehicular control were further investigated, it was found that the most commonly hit object was a tree, followed by a rollover, a collision with a fence, a collision with an embankment and a collision with a utility pole. The most severe crashes were those with trees and utility poles. In the USA, a number of studies have examined the effect of road features on singlevehicle rural crashes. Ivan, Pasupathy and Ossenbruggen (1999) found that single-vehicle crashes in Connecticut occurred when traffic volume was low and on roads with less forgiving geometry, arguing that under low traffic volumes, there is less likely to be another vehicle to hit and the less than ideal geometry (sharp horizontal curves and narrow shoulders) reduces the manoeuvring room for a driver to recover. Likewise, Stamatiadis, Jones and Aultman-Hall (1999) examined the effects of a number of road features along with driver and vehicle factors on crash risk on low-volume rural roads in Kentucky and North Carolina. They found that, for single-vehicle crashes, young drivers were least safe at night on roadways with higher speeds, narrowest lanes, both narrowest and widest shoulders, and sharpest curves. Interestingly, they also found low crash rates on roads with the worst conditions, i.e., roads with no shoulder and sharp curves. While i t is intuitive that crash risk would decrease with increasing shoulder width and curvature, this finding suggests otherwise. Stamatiadis and his colleagues argued that the driving task may require more attention when there are no shoulders and curves are s harp, and drivers may be more cautious in these conditions and drive in a more careless manner on roads with wide shoulders and less sharp curves. McLean (1996), too, hypothesised that there is a relationship between road geometry and perceived road environment, and that drivers are inclined to drive more slowly when road conditions are more constrained. 1.3.2 Multi-vehicle collisions

Multi-vehicle collisions are also common on rural roads and occur at both intersections and on road lengths, and are significantly related to the number of vertical curves with insufficient passing-sight distance, and the number of intersections (Al-Masaeid, AlSuleiman, Hammed & Halawa, 1994; Chang, Chen, & Carter, 1993).

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

Wegman (1995) reported that 24% of rural crashes involved collisions with intersecting vehicles, 20% involved two-vehicle rear-end collisions, and a further 19% involved twovehicles colliding head-on. In Finland, Toivonen and Niskanen (1998) reported that intersection crashes formed the largest category of crashes on main roads with two carriageways, with head-on collisions accounting for approximately 25% of crashes on semi-motorways carrying over 1500 vehicles per day. In the State of Connecticut, USA, Ivan et al. (1999) found that multi-vehicle crashes occurred in the presence of intense traffic conflicts such as at intersections, or roads serving local trips, or when there are high volumes of trucks in the traffic stream. Nielsen (2000) also identified three major causes for head-on collisions in Denmark. These are: high speed, overtaking manoeuvre, and wheels on verge or soft shoulder. Other causes included: alcohol, trailer out of control, fatigue, and missing head lamps. In their sample of rural crashes in South Australia, McLean et al. (2002) reported that, of those crashes in which more than one vehicle was involved, the most common crash type was a mid-block collision. Curved roads and young, inexperienced drivers were again over-represented. Head-on collisions comprised half of the sample and were generally severe. A majority of the head-on collision cases resulted from one vehicle running onto the unsealed shoulder on the left and then over-correcting and veering back across the road out of control. Again, crashes were caused by unsealed shoulders, particularly on curved sections of road and problems with the road surface. Head-on collisions can also occur when one vehicle is overtaking another on an undivided two-lane road. Research on overtaking crashes is comparatively rare, despite the frequency and severity of these types of crashes. Clarke, Ward and Jones (1998) reported that overtaking crashes accounted for eight percent of fatal crashes on rural roads in Nottinghamshire, England and that their crash severity index (the proportion of cases resulting in death or serious injury) is over 20%. In Australia, too, Armour (1984) found that overtaking accounts for about ten percent of rural casualty crashes. These rates are considerably higher than the 3 -4% of such crashes reported in the USA (Khasnabis, 1986), which probably reflects the much greater length and usage of multi-lane, divided roads in that country. Crashes at rural intersections are also common, occurring at cross roads and T -junctions in equal proportions. The Transportation Research Board (1987, cited in Leonard, Bilse & Recker, 1994) suggested that approximately one-third of all rural crashes occur at intersections. In McLean et al.s (2002) Australian sample, the most common crash type at intersections was when one vehicle was executing a right turn (equivalent to a left turn in the USA and most European countries) and drivers were more often elderly. Turning across traffic at an intersection is one of the most complex driving tasks, requiring the selection of safe gaps in the traffic and is made even more difficult with poor sight distance, high speeds and high traffic volume. McLean et al. found that the most common factor contributing to this type of crash was a sight restriction, often caused by vegetation or crests. Overall, these findings tend to support the view that the nature of the road network and the way in which it varies between countries along rural roads is a key determinant of crash and injury profiles.

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

1.4

OVERVIEW

Crashes on rural roads are a major road safety problem, accounting for up to two-thirds of deaths and serious injuries. Most importantly, crashes on rural roads tend to be more severe than crashes on urban roads because of a combination of high operating speeds, poor road environment and enforcement levels. There are four major types of crashes that occur on rural roads: single-vehicle run-off-road crashes and multi-vehicle crashes involving head-on, intersection or rear-end collisions. Single-vehicle run-off-road crashes account for the majority of fatal and serious injury crashes, typically because of the speeds at which rigid objects (particularly trees and poles) on the roadside are hit. Head-on collisions, too, account for a large proportion of fatal and serious injury crashes, again, because of the speeds at which vehicles collide. Collisions at intersections on rural roads also result in severe injuries and most commonly occur when a driver executing a turn across traffic makes a poor gap selection. Compounding problems here are high speeds, high traffic volumes, and restricted sight distance. While driver error (e.g., loss of control, poor gap selection) is a contributing factor in most crashes, it is the road environment itself that is one of the most important factors determining crash and injury risk. Speeding is a major contributing factor and the combination of high speed with poor road and roadside design increases crash and injury risk.

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

CONTRIBUTING FACTORS

This Chapter presents a discussion of the nature of road trauma in rural areas and the factors that have been associated with crash risk. The literature concerning road user and vehicle factors are discussed first, followed by an extensive discussion of the various features of the road and road environment are discussed extensively. Given that the discussion provides a large amount of detail, an initial overview of the contributing factors is provided. The reader is referred to relevant sections for more detail. 2.1 OVERVIEW

Crashes are complex, often involving several contributing factors. Many studies have attempted to establish relationships between crash frequency and severity, road user characteristics, vehicle factors, road features and other possible contributory factors. A landmark in-depth study of rural road crashes in Australia (McLean, Ryan, Wright & Hinrichs, 1988) identified various combinations of driver, vehicle and environmental factors in rural road crashes. The Queensland Parliamentary Travelsafe Committee (1999) in Australia noted the following: Higher speeds on rural roads result in more severe crashes, Fatigue and the combination of drink-driving with failure to wear seat-belts are more common in rural crashes, High speed, single vehicle run-off-road crashes are more frequent in rural areas, Loss of control is a frequent factor in rural crashes. This often results from vehicles running off roads on curves, particularly on roads with unsealed shoulders, First aid is less readily available, and response times of emergency and trauma recovery services are much longer in rural areas compared to urban areas, and Enforcement of traffic laws by traditional means is very difficult and expensive on remote and low volume rural roads. Road user factors

2.1.1

A range of road user characteristics and behaviours that contribute to increased crash and injury risk on rural roads was identified. For the most part, driving on straight sections of rural roads may be a relatively easy task, requiring low levels of attention. However, the road network is a dynamic one with geometric complexities and inconsistencies that requires drivers to react quickly and appropriately to these changes, particularly as travel is usually at high speed. Fatigue and inattention play an important role in crash risk along straight sections of roadway. When a driver falls asleep or loses concentration, it is common for vehicles to leave the formed roadway and enter the roadside at speeds of around 100-120 km/h (see Section 2.2.1). Risk-taking also heightens crash and injury risk. Here, the combination of speeding and alcohol increases the risk of losing control of vehicles, particularly on curves, or overtaking in unsafe gaps, commonly resulting in severe run-off-road or head-on collisions (see Section 2.2.2).

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

Speeding is a major contributory factor in rural road crashes. There are many factors that perpetuate adoption of high travel speeds including a lack of awareness of the relationship between speed and road trauma, an under-estimation of crash and injury risk and overestimation of what is a safe speed. Young, inexperienced drivers are over-represented in single-vehicle run-off-road crashes on rural roads. Pedestrians and cyclists are especially vulnerable to injury in high-speed environments. While crashes involving pedestrians and cyclists have previously been treated as an urban problem, it seems that there are high death and injury rates on rural roads in some parts of the world (see Section 2.2.3). 2.1.2 Vehicle factors

The crash protection capabilities of vehicles greatly affect injury outcome. It seems that in real-world crashes, particularly on high-speed rural roads, even the best vehicles cannot protect their occupants or other road users in many common crash types. Safe impact speeds generally fall below 30 to 50 km/h for the following crashes: Side impacts at intersections, Collisions with trees, poles, and other roadside hazards, Collisions with pedestrians and other vulnerable road users, and Crashes between incompatible vehicle types (see Section 2.3.1).

In addition, there are problems associated with heavy vehicles on rural roads. These are the result of the mix of vehicles on these roads, the aggressivity of heavy vehicles toward smaller vehicles and the difficulties experienced by drivers of heavy vehicles in manoeuvering the vehicle on roads in poor condition, with poor alignment, curvature and shoulders (see Section 2.3.2). 2.1.3 Road factors The road environment is thought to be one of the most important factors determining crash and injury risk and many strong relationships between adverse road elements, high speed and high crash locations have been identified. 2.1.3.1 Speed and speed limits Speed and speeding are major contributors to crashes on rural roads. Although powerful relationships between speed, crash and injury risk are well-established, drivers, by and large, lack awareness of or respect for the true consequences of speeding. Speed limits, arguably, have the single largest influence on driver choice of speed, however, due to the many inconsistencies on rural roads, drivers cannot safely drive at high speeds all of the time and throughout their entire journey, and changes in the road environment require constant speed adaptation. The requirement to adapt speed to suit the environment can increase the opportunity for human error and lead to higher crash and injury risk (see Section 2.4.1).

10

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

2.1.3.2 Road Lengths Three major features of road lengths were identified that appear to contribute to increased crash and injury risk. The literature clearly shows that undivided two-lane rural roads are the most dangerous roads in the rural network. Even though they generally carry low levels of traffic, travel speeds are often high on these roads and the design of many of these roads is not appropriate for such high speeds. For instance, these roads t end to have carriageway widths, curves and shoulders that are not appropriate for high-speed travel, frequent minor and uncontrolled intersections and many roadside hazards (see Section 2.4.2.1). The literature addressing the effect of road surface on crash and injury risk is more difficult to interpret. The differences seem to be related to type of crash and events leading to the crash. Road roughness seems to decrease the rate of single-vehicle crashes. Given that single-vehicle crashes are often related to high speed and loss of control, it is suggested that rough roads reduce travel speed and therefore the potential for these types of crashes. In contrast, road roughness seems to increase the rate of multi-vehicle crashes. It is suggested that this is related to lateral variation and abrupt changes in travel speed, rather than high travel speed (see Section 2.4.2.2). It appears that the rate of overtaking crashes is related to the provision and geometric design of passing lanes. When passing lanes are n ot provided on long sections of rural road lengths, there is increased potential for risky or misjudged overtaking manoeuvres, particularly when sight distance is short. Further, it seems that design practices for passing lanes (in terms of passing zone lengths and number of passing zones) may not be appropriate for many drivers to pass slow traffic or multiple vehicles in a safe manner (see Section 2.4.2.3). 2.3.1.3 Horizontal Curves The literature reports an over-representation of single- and multi-vehicle crashes at or near curves, particularly on two-lane rural roads. Operating speeds on the approach to and through curves that are too high to safely negotiate the curve account for the majority of crashes, most of which are run-off-road or head-on collisions. Abrupt changes in operating speeds, too, contribute to crashes on curves and on the approach to curves, particularly rear-end type crashes (see Section 2.4.3.1). Trucks are also over-involved in crashes on curves, particularly rollover and head-on collisions. In addition, lateral placement errors or misjudgements contribute to crash risk and these operations are closely linked to geometric features of the curve (see Section 2.4.3.6). The problems associated with geometric design of horizontal curves are many. It seems that design speed of many curves is much below that of operating speeds. Degree of curvature seems to be the primary contributing factor to crash risk on curves, with crash risk increasing with sharpness of the curve (particularly those over 5). Some of the literature has addressed the design of the approach to curves and shown that poorly designed transitions contribute to increased crash risk. The alignment of curves, too, has been shown to affect crash risk. Here, it seems to be a problem of deficient super-elevation, resulting in a higher demand for side friction, thus increasing crash risk (see Sections 2.4.3.2 to 2.4.3.6). One of the less obvious oversights in the professional practices of road design and operation of (rural) roads concerns the apparent lack of dialogue between road engineers and vehicle designers. For example, SUVs may interact differently with road geometry

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

11

features, such as curves and super-elevation, than the design standard car. Also, gradual changes in vehicle characteristics over time do not seem to be comprehensively reflected in changes in road design standards. 2.3.1.4 The Roadside Single-vehicle crashes are common on the rural road network and they typically involve vehicles leaving the roadway and colliding with fixed objects on the roadside or overturning. These crashes generally result in serious injury outcomes. In combination with high speeds, an unforgiving roadside plays a major role in the frequency and severity of run-off-road crashes. This is largely because of the positioning and rigid nature of fixed objects, and because shoulders are unsealed, uneven, or too steep. These conditions frequently result in serious injury or death (See Section 2.4.4.1). 2.3.1.5 Intersections Rural intersections are dangerous locations, particularly at-grade intersections that are uncontrolled or controlled by stop or give-way signs, often found on low-volume, singlecarriageway roads. At these locations, conflicts occur at high impact angles, often at high speeds and result in high injury severity. Some geometric features of at-grade intersections have been shown to increase crash and injury risk. Intersections controlled by stop or giveway signs are more dangerous than intersections controlled by traffic signals because of the conspicuity of the intersection and the requirement for drivers to make gap selection decisions unassisted. While intersections controlled by traffic signals are safer than those controlled by stop or give-way signs, there is also some suggestion that those that do not provide fully-controlled turning phases also pose difficulties for turning manoeuvres. Also, traffic signals operating in high-speed rural areas exhibit poorer safety records than in lower-speed environments (see Section 2.4.5.1). Sight distance, too, is an important feature of intersection design. Locations where there is restricted sight distance experience higher crash and injury risk than those with good sight distance (see Section 2.4.5.2). Other features of intersection design, such as provision of channelisation, delineation and alignment, can also affect crash risk, where poor design or maintenance of these features can increase crash and injury risk (see Section 2.4.5.3). While grade-separated intersections are provided on some highways and freeways and are considered to improve safety and operational performance, there is some suggestion that poor design of entry and exit ramps can increase the potential for crashes while merging and exiting freeways, and crashes resulting from wrong-way manoeuvres (see Section 2.4.5.5). Intersections between roads and rail lines in rural areas are typically passively controlled that is, warning signs are placed on the road approaches and the driver must decide whether it is safe to cross. Only rarely active elements such as flashing lights or boom gates in place at rural crossings to alert an approaching driver that a train is approaching. One explanation for the relatively high rate of train-road vehicle crashes as such intersections is the infrequency of trains drivers do not expect a train and so they do not slow to an appropriate speed to allow them time to properly scan for trains in both directions. Additionally, the poor physical condition of many crossings requires the driver to attend to the roadway rather than looking in both directions for trains. Sight lines are also often obscured by foliage. The cost of upgrading more than a small percentage of rural rail

12

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

crossings to include active devices to warn if a train is actually approaching is prohibitive (see Section 2.4.5.7). 2.3.1.6 Other Factors In addition to problems associated with road lengths, horizontal curves, roadsides and intersections, a number of other road factors contributing to crash risk on rural roads are identified. These include vertical curves, bridges and culverts and construction zones. It has been suggested that the problems at vertical curves (especially those on two-lane roads), are related to restricted stopping sight distance, particularly when a major hazard such as an intersection or sharp horizontal curve exists, as well as absence of shoulder, and excessive speeding for the conditions (see Section 2.4.6.1). Crashes on the approach to, and on, bridges and culverts also constitute a substantial problem on rural roads, particularly as these crashes are generally severe in nature. The problems here are similar to those of roadsides where, single-vehicle run-off-road crashes and striking a fixed object are common. The features of bridges that can increase crash and injury risk include: bridge structures such as entrance posts and wing walls, bridge railings; width of bridge (there are significantly more crashes on narrow bridges); lack of shoulders; vertical and horizontal alignment; and entrances of intersections adjacent to the bridge (see Section 2.4.6.2). Construction zones also create hazards for drivers on rural roads, with estimates of a twofold increase in crashes at these locations. The combination of high travel speed, high traffic volume, the requirement for drivers to make behaviour adjustments such as reducing speed, merging and/or changing lanes result in a heightened potential for multi-vehicle and rear-end crashes on the approach to and through the work zone, collisions with fixed objects through the work zone as well as pedestrian crashes. Geometric features of these sites also increase crash and injury risk. These include: changes in elevation at the edge of lanes; absence of or improper use of temporary concrete barriers; and, inadequate lane closure treatments (see Section 2.4.6.3). 2.2 ROAD USERS

Road user characteristics are contributing factors in most crashes. Behavioural characteristics such as inattention, fatigue, inexperience and risk-taking behaviour (speeding, drunk-driving and failure to wear a seat-belt) have all been identified as factors that significantly contribute to increased crash and injury risk on rural roads. Further, there is a growing awareness within the road safety community that vulnerable road users may have their own particular needs and difficulties in using the road system and that this should be considered in designing the system. 2.2.1 Road user inattention and fatigue

Driving, riding, cycling and walking in the traffic environment are everyday tasks that require an individual to constantly monitor stimuli both on and off the road. These tasks involve a range of sensory, perceptual, attentional and cognitive skills and can be demanding and complex activities. Road users must attend to a wide range of information including the roadway and roadside geometry, traffic control devices, and traffic conditions and make complex decisions quickly, often in a hazardous environment. An increase or

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

13

sudden change in any of these information sources places increasing demands on a road users processing ability. If a road users processing system is overloaded at any stage while using the road system, then crash risk is increased (Shinar, 1978). For example, driving on straight lengths of rural roads is, intuitively, a relatively easy task requiring low levels of attention and information processing. However, driver workload increases with increasing complexity and, consequently, a drivers level of performance can decrease markedly. It has been argued that driver workload increases on rural roads with increasing geometric complexity and inconsistency, particularly when unusual or unexpected geometric features are encountered, such as a very long or very sharp curve, or a one-lane bridge (Krammes & Glascock, 1992; Steyer, Sossoumihen & Weise, 2000). Therefore, crash and injury risk at these locations can increase. Driver fatigue is a significant contributory factor to road crashes, particularly those on country roads, in most developed countries. Estimates of the contribution of driver fatigue vary from four percent (Treat, 1980, cited in Fairclough, 1997) to 25% (Maycock, 1995, cited in Fairclough, 1997). An Australian study (Armour, Harrison & South, 1986) estimated that 27% of single-vehicle crashes in rural areas were fatigue-related. Furthermore, Ryan, Spittle and Cooper (1998) investigated truck crashes in Western Australia and reported that rear-end crashes were the most common type of truck crash and that fatigue was associated with around 20% of these crashes. Although a detailed discussion of the contribution of fatigue to rural crashes is outside the scope of this review, it is worth noting the following findings from Haworth and Rechnitzer (1993). Fatiguerelated crashes often involve longer trips and are more likely than other crashes to: Occur between midnight and 6am, but less likely to occur between noon and midnight, Occur in rural areas, on highways and in the absence of street lighting, Be single-vehicle crashes and involve fewer vehicles on average than other crashes, and Be off path, on straight roads and less often involve pedestrians or vehicles from adjacent direction.

Sleep deprivation has been found to impair performance on a wide range of tasks including vigilance, reaction time, and tracking, all of which are involved in driving. Although few studies have investigated the effects of sleep loss on driving performance, there is evidence that driving performance such as lateral position, speed and ability to respond to peripheral stimuli is also impaired by sleep loss (Lenn, 1997). Others have found that increasing levels of fatigue reduce the precision of a drivers ability to self-monitor performance, for example, to predict ability to continue driving and to estimate the appropriate point at which to discontinue driving (Fairclough, 1997; Hartley, Horberry, Mabbott & Krueger, 2000). Apart from sleep disturbance, other causes of driver fatigue include monotony and predictability of the road environment. This has been termed highway hypnosis which is described as a lowered state of alertness leading to the development of drowsiness and failure to react adequately to changes in the road situation (Wertheim, 1978, p. 111).

14

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

2.2.2

Driver age and risk-taking behaviour

The literature clearly shows that young, inexperienced drivers who are either speeding, drunk or both are over-represented in rural road crashes (Haworth et al., 1997; McLean et al., 2002; Stamatiadis et al., 1999). Young novice drivers constitute a large and, to date, intractable road safety problem world-wide. Road trauma continues to be the primary cause of death for young adults aged 15 to 24 years (Cavallo & Triggs, 1996; Macdonald, 1994). Across Australia, and internationally, the crash involvement rates of young drivers are typically two to four times those of drivers who are older and who have been driving for longer. In Australia, 16 to 24 year olds comprise about 15% of the driving population, but account for around 35% of fatal and 50% of injury crashes (Macdonald, 1994). Young novice drivers are over-represented in single-vehicle run-off-road crashes on both curves and straight sections (mainly as a result of risk-taking) and those in which they fail to respond successfully to conflicts created by unexpected changes in the road environment and unexpected actions of other road users (Catchpole, 1997). In addition, this group show elevated hostility, high levels of driving to reduce tension, high levels of driving-related aggression, competitive speed and sensation-seeking (Deery, Kowadlo, Westphal-Wedding & Fildes, 1998; Wilson, 1991). Young novice driver crashes are generally attributed to two factors: age (youthfulness) and inexperience. In relation to age, researchers have argued that over-representation in crashes of young drivers reflects the fact that they are teenagers and are simply willing to take more risks (Macdonald, 1994). With respect to inexperience, research on skilled performance suggests that, compared to more experienced drivers, novice drivers have under-developed perceptual and cognitive skills as a result of inexperience, namely underdeveloped perceptual (i.e., detecting and assessing traffic hazards), attentional control (i.e., dealing with changing workloads), time-sharing (i.e., dealing with multiple tasks) and calibration (i.e., matching ones driving performance with task demands) skills (Regan, Triggs & Godley, 1999). As indicated previously, speed is a major contributory factor to young novice driver crashes. The choice of speed involves assessing the risk involved and deciding on how much risk one wishes to accept clearly this is an issue for young drivers who are willing to take more risks than older drivers. The literature suggests that speeding, except at extreme levels, is not regarded by drivers or riders as a particularly dangerous activity. Unsafe speed-related behaviours are consistently reinforced by many positive consequences, and rarely result in negative consequences. From a driver/riders viewpoint, there is almost certainly no subjective change in their crash risk when they drive or ride a little faster than the posted speed limit. In their review of the literature on speeding, Oxley and Corben (2002) noted that there are a number of attitudes and beliefs that can adversely affect the speed at which individual drivers/riders travel and, hence, the inherent safety of the road-traffic system. These are: Most of us drive/ride without crashing and this reinforces the attitude that speeding is not risky and results in an under-estimation of the association between speeding and probability of serious injury or death in a crash, Most drivers/riders over-estimate the speed of other drivers and want to drive at a similar speed to others,

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

15

 2.2.3

Most drivers/riders over-estimate their ability to control their vehicle in an emergency, Most drivers/riders drive at a speed within a perceived tolerance or margin where speed enforcement will not occur, and Many drivers/riders over-estimate what is an appropriate or safe speed. Vulnerable road users

At collision speeds above 30 km/h, the probability that a pedestrian will be fatally injured rises rapidly, with almost certain death at impact speeds of 55-60 km/h (Anderson, McLean, Farmer, Lee & Brooks, 1997; Ashton & Mackay, 1979). Yeates (2001) argued that pedestrians and cyclists are only likely to be relatively safe in areas with traffic speeds of 30-40 km/h subject to traffic volumes. At these speeds, most potential collision situations can be recognised and avoided. Speeds on rural roads, however, are invariably much higher than this and pose great danger to pedestrians and cyclists. Pedestrian and cyclist crashes are comparatively rare on rural roads, and are predominantly an urban phenomenon in many countries such as the USA, Australia, and most European countries. Nevertheless, there are some reports that indicate a disproportionately high fatality rate of pedestrian and bicyclist crashes on rural roads in some countries (Ribbens, 1986; Toivonen & Niskanen, 1998). Ribbens (1986) found that an alarming 40% of all rural fatalities were pedestrians and bicyclists in South Africa. The contributing factors to these crashes seemed to be gender and ethnicity (the majority of pedestrians and cyclists killed were young, black males), location (at or near intersections, bus stops, shops and housing compounds), on paved roads in rural townships, time of day (late afternoon/early evening), and day of week (most occurring on Fridays and weekends). The most common types of crashes for pedestrians were pedestrian walking or running across road and pedestrian walking on road or shoulder. For bicyclists, the most common type of crash was where the cyclist swerved in front of a vehicle travelling in the same direction. Ribbens concluded that crashes were related to three major causative factors, i.e., human behaviour and error (taking unsafe gaps, stepping out from behind buses, cyclists swerving across the road, drivers speed), inadequate geometric design with regard to pedestrians and cyclists (lack of proper road shoulders and paths, narrow bridge structures and cuttings, lack of proper bus stops, narrow road sections), and road maintenance (chipped road edges, poor grading of road shoulders, drainage problems). In Finland, too, pedestrian and bicycle traffic crashes account for a large share of all traffic fatalities in rural areas (Toivonen & Niskanen, 1998). Between 1992 and 1996 pedestrians were involved in 13% of fatal crashes occurring outside built-up areas in Finland, while bicyclists were involved in ten percent of fatal crashes occurring outside built-up areas. The crash risk for vulnerable road users on 100 km/h roads is 36% greater than on 80 km/h roads. Toivonen and Niskanen reported that pedestrian crashes were more frequent in winter months and occurred in the dark or semi-dark. The most common type of crash occurred when a pedestrian unexpectedly crossed the road (50%). The next most common type was caused by the pedestrian stopping in the middle of the road (13%), followed by darting out from behind a parked car (4%). Bicyclist crashes, in contrast, occurred in the spring and autumn and usually happened during the day and in good weather. Half occurred at intersections, particularly when drivers approached from different directions

16

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

and crashed when they drove ahead (24%), or approached a bicyclist from behind (17%). In 13% of the crashes a car ran over a bicyclist travelling along the edge of the road. In these cases the crash was usually caused by inattentiveness or an error on the part of the driver or the bicyclist. Toivonen and Niskanen (1998) suggested that the high rate of pedestrian and cyclist fatalities in rural areas is probably due to the Finns general tendency to live away from built-up areas. They also conceded that because of the long distances to be travelled on rural roads, speed limits need to remain relatively high and it is therefore difficult to improve pedestrian and cyclist safety in all sparsely inhabited rural areas. Polish data suggest that pedestrian fatalities in both urban and rural locations are also high. Gaca and Tracz (2000) reported that, in 1997 in Cracow, crashes involving pedestrians accounted for 54% of all crashes, but fatal pedestrian crashes represented as much as 77% of all road fatalities. They found that a high proportion of these fatalities (about 40%) occurred on rural roads. These generally occurred on national roads through small towns and villages. On national roads outside small rural towns and villages, the share of crashes involving pedestrians was also found to be high at around 20-25% of all crashes. In comparison, they noted that on similar German rural roads, the proportion of pedestrian crashes was only two percent (i.e., around 10 times lower than in Poland). Alcohol was considered a major contributing factor to fatal pedestrian crashes on road sections outside built-up areas, accounting for 44% of crashes. Most crashes occurred because of incorrect behaviour of pedestrians (often alcohol-induced), the presence of pedestrians in places unexpected by drivers, presence of pedestrians in conditions of poor lighting and visibility, unfavourable walking conditions (soft shoulders, snow, and rain), insufficient protection of pedestrians on shoulders (narrow and poorly maintained) and lack of footpaths, incorrect location of bus stops and lack of walkways leading to bus stops, high vehicle speeds and drunk drivers. Garber and Lienau (1996) compared urban and rural pedestrian crashes in the State of Virginia, USA. They found that, for urban crashes, the zone just outside the intersection was the most dangerous location for pedestrians in all age group, while for rural crashes, the mid-block zone was slightly more dangerous than intersections for pedestrians in all age groups (except those aged 55-64 years old). They attributed the high involvement rates at the mid-block zone in rural locations to the relatively higher speeds in rural areas compared to in urban areas. Moreover, they found that the injury and fatal involvements rates are generally higher (particularly for older pedestrians) in rural areas compared to urban areas and that alcohol was a contributing factor in more rural crashes than urban crashes. 2.3 VEHICLE FEATURES

A number of vehicle factors can influence crash and injury risk, particularly during highspeed travel on rural roads. These factors include vehicle crashworthiness, type of vehicle, performance, handling and power/weight ratio and vehicle incompatibility. 2.3.1 Vehicle crashworthiness

Humans are limited in their capacity to sustain impact forces and vehicles can provide a level of protection. New passenger vehicles sold in most western countries are required to

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

17

meet vehicle secondary safety standards. For instance, the Euro NCAP tests vehicles in a frontal impact at 64 km/h, in a side impact at 50 km/h and a pedestrian impact at 40 km/h. While vehicles in the last two decades have become more crashworthy, and are equipped with an increasing number of safety features such as airbags, there are, finite physical limitations on the extent to which vehicles can protect occupants. Some vehicles have lower speed thresholds than other vehicles in respect to their abilities to protect their occupants or other road users in crashes. It is possible to quantify the crashworthiness of vehicles involved in real-life crashes. A series of studies have been undertaken at MUARC, investigating the level of protection afforded occupants of current generation passenger cars available in the Australian market when involved in frontal- or side-impact crashes (Fildes, Lane, Lenard & Vulcan, 1991; 1994). While it is difficult to determine the precise speed at which vehicles were travelling in retrospective crash investigations, the speed zone and/or delta-V level can give good approximations for travel and impact speed. In their mass data analysis of injuries to occupants in vehicles, Fildes et al. (1991) reported that 71% of vehicle crashes were in speed zones over 76 km/h. They also reported that occupants of small cars (especially mini-sized vehicles) were over-represented in urban environment crashes (equal to or less than 75 km/h posted speed), while intermediate and large car occupants were overrepresented in higher speed (> 75 km/h) rural crashes. Fildes et al. (1991) also reported on in-depth crash investigations of occupants who were admitted to hospital following a crash in a passenger car that occurred in urban and rural Victoria after 1989. The mean estimated delta-V value in frontal-impacts was 45.4 km/h, ranging from 3 km/h to over 111 km/h, but with 70% equal to or below 48 km/h. For side-impacts, this value was between 36 and 42 km/h, ranging from 12 km/h to 96 km/h, with more than 90% equal to or below 55 km/h and 26% equal to or below 27 km/h. Side-impacts present a difficult problem in crash protection as there is little crushable structure between the occupant and the impacting vehicle or object. The front structure of a car can absorb two to five times as much energy as the side structure (Cesari & Bloch, 1984). The recently introduced US requirements for side-impact testing (FMVSS 214) specifies a 30 mph (48 km/h) crabbed impact velocity of the bullet vehicle. For vehicles of equal mass this equates to a 27 km/h perpendicular impact velocity of the struck vehicle. At this value, one-quarter of the vehicle occupants observed in Fildes et al.s (1991) crashed vehicle study were injured sufficiently to require hospitalisation. Swedish data suggest that, in a vehicle that affords good crash protection to its occupants, it is possible to experience a frontal crash with another vehicle at 70 km/h without long term injuries. However, for impacts with poles or trees when only 20-25% of the frontal structure is directly contacted, any speed above 30 km/h is likely to result in serious injury to the car occupants. For side-impact crashes, a car can only offer occupant protection at speeds under 50 km/h (Vgverket, personal communication). Although there are clearly limitations in the ability of vehicles to protect occupants and other vulnerable road users in a crash, it is possible that many drivers are either unaware of these limitations, or over-estimate the protective capabilities of their vehicles. Silcock, Smith, Knox and Beuret (1999) found that many drivers linked choice of speed to invehicle safety features such as air-bags and side-impact bars and many drivers were more confident in driving faster when these features were present in vehicles.

18

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

2.3.2

Heavy vehicles

In addition to the protection capabilities of vehicles to occupants, there are problems associated with heavy vehicles on rural roads. First, there is the problem of the mix of vehicles on rural roads and the risk of injury to vehicle occupants involved in a crash with a heavy vehicle. Secondly, there is the problem of single-vehicle crashes involving heavy vehicles and the contribution of road features to these crashes. Although crashes involving trucks are less common than those involving smaller vehicles, they tend to be more severe than those involving smaller vehicles. Haworth and Vulcan (1991) noted that about 75% of fatal crashes involving articulated trucks occurred outside of capital cities, with 23% of these being single-vehicle crashes. Significantly, freight transport is an important factor for continuing economic growth in most countries and is likely to increase in the years ahead. Most societies including Australia and the USA predict a continued trend of increased road freight transport in the next decade (Johnston, 2001; Mohamedshah, Paniati & Hobeika, 1993). This increase in truck and heavy vehicle travel on the rural road network may impact on a number of operational and safety problems on these roads. Heavy goods vehicles and buses have characteristics that are quite different from those of passenger cars. Essentially, problems relating to heavy vehicles on rural roads result from three characteristics: i) heavy vehicles are much heavier and larger in dimension compared with passenger cars, therefore experience instability and manoeuvrability problems, ii) heavy vehicles have less effective acceleration capabilities than passenger cars and have greater difficulty maintaining speeds on upgrades and this speed variation generates more instances of overtaking and the potential for head-on collisions with oncoming vehicles, and iii) heavy vehicles have a lower deceleration in response to braking than do passenger cars, which increases the potential for severe rear-end crashes. 2.3.2.1 Aggressivity of heavy vehicles toward smaller vehicles There have been numerous studies investigating the aggressivity of vehicles to other road users, particularly to vulnerable road users such as pedestrians, cyclists and motorcyclists (Cameron, Newstead & Le, 1998; Newstead, Cameron & Le, 2000). Vehicle aggressivity has been defined as the extent to which a vehicle transfers collision energy to the struck object rather than absorbing the collision energy itself (Fildes et al., 1993). Another definition used by Newstead et al. (2000) relates to injury outcome. Here, aggressivity is defined as .the fatality or injury risk to occupants of a vehicle or other more vulnerable road users colliding with the subject vehicle (pp. 4). In Western Australia in 1988, 3.2% of motor vehicles involved in all injury crashes were trucks and12% of vehicles involved in fatal crashes were trucks (Western Australian Police Department, 1989, cited in Ryan et al., 1998). Similarly, in the USA, the Department of Transportation (NHTSA, 1992) found that trucks accounted for eight percent of the vehicles involved in fatal crashes but only three percent in property damage only crashes. These findings suggest that when a truck is involved in a crash it is more likely to result in an injury or fatality than crashes involving only other vehicles, presumably because the larger mass of the truck produces a higher risk of injury to the occupants of other vehicles. Geometry, mass and stiffness are considered to be the three most important vehicle properties influencing aggressivity. Les, Morris, Olsson, Pettersson and Holmqvist (2001) identified a number of physical features that contribute to vehicle aggressivity including

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

19

frontal bonnet height, frontal bonnet length and wheelbase. Due to the height and mass of heavy vehicles, any impact with a passenger car is likely to result in massive head and chest injuries to the occupants of that vehicle (Grzebieta, Rechnitzer, Tingvall, Judd & Powell, 2000). Road trains (articulated trucks towing two or three trailers) are a common form of road transport in Australia, where they operate primarily n i rural and remote regions. Ryan et al. (1998) found that the crash involvement rate per 100 million vehicle miles travelled by road trains increased with the number of trailers in tow. The rate for three-trailer road trains was over twice that for two-trailer road trains. This is most likely related to the inherent instability of a second trailer. About one-quarter of road train casualty crashes involved a fatality, compared with fewer than ten percent for semi-trailers. Opposite direction and overtaking crashes involving road trains had a high proportion of fatalities in which the occupants of the other vehicle were at much greater risk than the road train occupants. 2.3.2.2 Single-vehicle heavy vehicle crashes While multi-vehicle crashes involving heavy vehicles generally result in injuries to the occupants of vehicles struck rather than the heavy vehicle occupants, single-vehicle crashes involving heavy vehicles are also common on rural roads. Ryan et al. (1998) noted that when truck drivers were fatally injured, this was usually due to rollover, ejection or contact with interior surfaces. In their examination of Western Australian single-vehicle casualty crashes involving trucks, Ryan et al. (1998) found that 44% occurred in rural regions and a further 35% in remote regions. The most common type of single-vehicle crash was a run-off-the-road crash, with 25% running off to the left (equivalent to running off to the right in the USA and most European countries) and 15% veering to the right (into medians or across oncoming traffic). Thirty five percent of the crashes resulted in a rollover, with a high proportion of them occurring at night (36%). Other injuries were caused by occupant ejection or contact with interior surfaces. It is argued that road factors are relevant in these types of crashes (Federal Highway Administration, 1989; 1990; Ryan et al., 1998; Sweatman, Ogden, Haworth, Pearson & Vulcan, 1990). In a study of heavy vehicle truck crashes in New South Wales, Australia, road condition was considered to be a factor in 80% of the crashes. Other factors considered important were road alignment, especially horizontal and vertical curvature (48%), roadside objects (33%), road shoulder (24%), and delineation (21%) (Sweatman et al., 1990). The US Federal Highway Administration (FHWA) examined the critical geometric features of roads affecting truck crashes (FHWA, 1989) and identified six major deficiencies for interchanges causing rollover and jack-knifing truck crashes. Examples include: abrupt changes in compound curves, short deceleration lanes on tight radius, and steep downgrade at the exit ramp. Another FHWA report (FHWA, 1990) identified a range of geometric features of roads that contribute to truck or heavy vehicle crashes, including number of lanes, lane width, shoulder width, median width, alignment, surface condition, and pavement condition. Trucks are also over-represented in crashes on horizontal curves, primarily because of their high centres of gravity and the possibility of loads shifting while negotiating a curve. On two-lane roads especially, trucks may encroach on the opposite lane while negotiating a

20

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

curve, posing a hazard for opposing vehicles (Mohamedshah et al., 1993). A more detailed discussion on this issue is provided in Section 2.4.3.4). 2.4 ROAD FEATURES

The road environment, which differs between urban and non-urban regions, is thought to be one of the most important factors determining crash risk. The design, construction, condition and signage of the road may play a large role in the chain of events that lead to a crash and to the severity of injury. Indeed, strong relationships between adverse road elements, high speed and high crash locations have been identified. The effects of roadway characteristics on traffic safety are substantial, particularly on rural roads. For example, in the USA in 1988, the fatality rates on rural Interstates, other rural Federal-aid primary arterials, and rural non-Federal-aid arterials were 9.7, 21.7 and 50.9 fatalities per billion vehicle kilometres, respectively (Miaou & Lum, 1993). Potential factors that make vehicle crash rates different from one roadway class to another include the physical nature of the roadway, such as geometric design, roadway markings, and traffic signs, as well as the type of travel, traffic control and traffic conditions. Numerous studies have drawn relationships between geometric features of rural roads and safety. For instance, narrow pavement and shoulder widths, severe or sharp horizontal curves, restricted sight distance and signage at intersections, and significantly narrower bridge width than the approach roadway have all been associated with increased crash risk. Indeed, Tziotis (1993) identified seven major geometric features of rural roads zoned at 100 km/h in Victoria, Australia, that contributed to crash risk. These were: poor curve delineation (54% of crashes); trees that were in hazardous locations or that were unprotected (49%); poor sight distance for overtaking (40%); guard fencing not provided (40%); partially sealed (40%) and unsealed (37%) shoulders; and, poor sight distance at intersections and other access points (both 28%). Many studies have shown that two-lane rural roads pose the greatest risk to drivers and riders. Crashes on two-lane rural roads in both Germany and the USA represent between 60 and 70% of the total number of fatalities on rural roads (Lamm, Guenther & Chouieri, 1999). Likewise, in the UK, rural two-lane roads have a higher crash involvement rate than motorways (54 per 100 million vehicle kilometres versus 23 per 100 million vehicle kilometres) and the highest crash involvement rate for fatal crashes compared to all other road types (2.0 fatal crashes per 100 million vehicle kilometres, compared to 0.5 on motorways and 1.0 on built up roads) (Comte & Carsten, 1998). Moreover, adapting roads to the landscape through mountainous or coastal areas is more difficult than in flat country. Mountainous roads have at least four particular characteristics that increase hazards on the road. These include steep inclines, difficult topography, inclement weather conditions, and restricted road use due to the closing of stretches of road during winter months (Hehlen, 1992). 2.4.1 Speed and speed limits

As indicated previously, high speed on rural roads is a major issue, as maximum legal speeds tend to be relatively high in these environments (usually between 80 km/h and 120 km/h). Rural road speed-related crashes are some four times more severe than those in the urban area based on the relative difference in serious crash proportions (Kidd & Willett,

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

21

2002). In other words, although the risk of being involved in a speed-related crash relative to a non-speed-related crash is about the same in both types of road environments, once involved in such a crash the consequences are more severe on rural roads due to higher speeds. Higher speeds increase both the probability of crash involvement and the seriousness of the consequences. Higher driving speed reduces predictability for other road users and reduces a drivers ability to control the vehicle, negotiate curves or manoeuvre around obstacles on the roadway, therefore increasing the chance of running off the road or into an oncoming vehicle. Higher speed also increases the distance a vehicle travels w hile the driver reacts to a hazard, thereby reducing the time available to avoid a collision. One of the key problems on rural roads (particularly single-carriageway roads) is that they often do not have consistent design characteristics over their total length. This means that drivers cannot drive at high speeds and safely all of the time and everywhere, and changes in the road environment require constant speed adaptation. The requirement to adapt speed to suit the environment can increase the opportunity for human error and lead to a higher crash and injury risk. Driving too fast for the conditions is a major factor in crash causation and injury severity. High speeds in conjunction with the varying geometric conditions common on rural single-carriageway roads result in a fatal crash rate that is higher than that for any other type of road (Pyne, Dougherty, Carsten & Tight, 1995). More importantly, higher speed increases the severity of the impact in a collision. Even small increases in speed can result in a dramatic increase in the forces experienced by crash victims and it is argued that the probability of sustaining an injury in a crash increases exponentially rather than linearly with vehicle speed (European Transport Safety Council [ETSC], 1995; Nilsson, 1984; Taylor, Baruya & Kennedy, 2002). Taylor et al. (2002) examined the relationship between speed and crashes on rural single-carriageway roads. They identified four relatively homogeneous road groups defined by crash rate, mean speed, minor intersection density, bend density, access density and hilliness (reflecting the operational characteristics of the road, or road quality) within which the speed-crash relationship could be investigated. The analyses showed marked increases in crash and injury risk with increases in mean speed and decreases in road quality, i.e., increases in the density of sharp bends and the density of minor crossroad intersections (crashes increased by 13% and 33% respectively, per additional bend/crossroad per kilometre and single-vehicle crashes were particularly strongly affected by the density of sharp bends, 34% increase in crash frequency per additional sharp bend per kilometre). Table 1 provides a summary of the literature that has examined the effect of change in speed on crash and injury risk.

22

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

Table 1: Effect of change in speed on crash and injury risk Change in speed Increase in mean speed Effect on crash and injury risk Reference

Rapid increase in frequency of injury crashes to Taylor et al. the power of approximately 2.5. A 10% (2002) increase in mean speed resulted in a 26% increase in the frequency of all injury crashes. For every 1 km/h increase in mean speed, the Nilsson (1984) number of injury crashes will rise by around 3%. An increase of 10 km/h would result in a 30% increase in injury crashes. Increase in crashes at intersections with Taylor et al. increases of mean speed, roughly proportional (2002) to the 5th power.

Increases in vehicle speed

The proportion of fatal crashes rises Nilsson (1984) exponentially to the 4th power. An increase of 10% in vehicle speed results in an increased risk of fatal crashes by 46%. The proportion of serious injury crashes rises Nilsson (1984) exponentially to the 3rd power. An increase of 10% in vehicle speed results in an increased risk of serious injury crashes by 33%. The proportion of casualty crashes rises Nilsson (1984) exponentially to the 2nd power. An increase of 10% in vehicle speed results in an increased risk of casualty crashes by 21%.

Increase in impact speed

Estimated that the likelihood of death with ETSC (1995) impact speed of 80 km/h is 20 times higher than at an impact speed of 30 km/h

2.4.1.1 Speed limits on rural roads Speed limits are considered to be the most powerful road feature that determines speed choice and therefore play a pivotal role in determining overall crash and injury risk on rural roads. The primary reason for setting speed limits, at least in some jurisdictions, is said to be to strike an appropriate balance between travel times and crash risk for specific roadway sections. The posted speed limit aims to inform drivers and riders of maximum driving speeds that the authorities consider reasonable and safe for the roadway. Most countries have a general rural speed limit of between 80 km/h and 100 km/h on rural arterials and highways, with speed limits between 110 km/h and 130 km/h on motorways and higher standard rural freeways. Many studies have examined the effect of raising or lowering speed limits in both rural and urban environments (e.g., Brown, Maghsoodloo & Ardle, 1990; Freedman & Esterlitz, 1990; Mace & Heckard, 1991; Parker, 1997; Pfefer, Stenzel & Lee, 1991). Studies consistently show that crash incidence or crash severity decline whenever speed limits have
COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS 23

been reduced (Engel, 1990; Finch, Kompfner, Lockwood & Maycock, 1994; Newstead & Mullan, 1996; Nilsson, 1990; Peltola, 1991; Sliogeris, 1992;). Conversely, studies in the USA and Australia show that the number of crashes and crash severity generally increase when speed limits are raised, especially on freeways (McKnight & Klein, 1990; Newstead & Mullan, 1996; NHTSA, 1989; Parker, 1997; Sliogeris, 1992). A number of studies examined the effect of raising speed limits on rural highways from 55 mph (89 km/h) to 65 mph (105 km/h) on crash risk in the USA (e.g., Baum, Wells & Lund, 1990; Garber & Graham, 1990; Godwin, 1992). While the size of the effects differed across the States, there was a general increase in fatalities with the increase in speed limit, with increases in the number of fatalities in the order of 15 to 29%. Godwin (1992) added that there was evidence that motorists had responded to this change by driving faster on the highways that had an increased speed limit, but also they were driving faster on highways without the change in speed limit. Parker (1997) found increases in speeds ranging from 1 to 4 mph when speed limits on freeways were increased from 55 mph (89 km/h) to 65 mph (105 km/h). Parker also evaluated the effects of raising and lowering speed limits by various amounts at nonfreeway sites. Before-and-after free-flow speeds were measured at sites where speed limits were altered and at control sites where speed limits were not altered. Parker found little effect of raising or lowering speed limits o n free-flowing speed the average change in the th mean and 85 percentile speeds was less than 1 mph and similar to the sites that were not changed. 2.4.1.2 Speed limits on approaches to provincial cities and towns One of the major issues in setting rural speed limits is the transition from the rural road network to the urban road network, i.e., the approach to and speed zones through rural townships. Indeed, some studies have shown that the crash and injury rate on major roads into and through country towns is substantially higher than on the open road (Steinbrecher, 1992; Victorian Parliamentary Road Safety Committee, 2002). In his investigation of the casualty crashes on the approaches to, or on the immediate outskirts of, major provincial cities and towns in the State of Victoria, Australia (roads extending from urban areas (60 km/h), through a speed transition zone (or partially developed area) (75 km/h) to exclusively rural areas (100 km/h), Tziotis (1993) reported a greater incidence of casualty crashes in partially developed speed transition zones (75 km/h) compared to rural zones (100 km/h). Specifically, in the first instance, Tziotis reported 45 casualties/100 million km/year compared to 27 casualites/100 million km/year, and in the second instance, 1.3 casualties/km/year compared to 0.4 casualties/km/year. Furthermore, the number of rearend casualty crashes and casualty crashes involving either rigid trucks or motorcycles peaked within the transitional/partially developed zone, as did the proportion of casualty crashes occurring during the dusk/dawn period. Tziotis also noted that the changes in the nature of crashes that prevailed along these feeder roads had occurred within a relatively short length of the transition zone. While speed limits of up to 100 km/h are generally permitted on Australian rural roads, the speed limit within urban areas can be as low as 50 km/h (Steinbrecher, 1992; Tziotis, 1993). It is desirable for drivers to lower their travelling speed during the transition from 100 km/h to urban speeds, and before they reach the town entrance, and that they do not exceed the speed limit while they travel in and through the town. Steinbrecher (1992), noted, however, that in reality, most drivers do not initiate speed reduction until the last minute. Furthermore, he argued that the proportion of through-traffic in rural towns is usually quite high, and motorists tend to view the town as an obstacle to their trip.

24

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

Consequently, the willingness of these motorists to adjust to the speed limits on their way through the town is very low. There is some suggestion that the characteristics of road crashes experienced on the approach to, or nearing provincial cities, differ substantially from those crashes occurring in either urban or rural road environments. For instance, Tziotis (1993) reported an increase in cross-traffic crashes, right-turn-against crashes (equivalent to left-turn against in the USA and most European countries), and pedestrian crashes close to and within provincial cities. Not surprisingly, the number of intersection casualty crashes increased as the proximity to the provincial city increased, however the frequency of night-time casualty crashes decreased. In contrast, as the distance out from provincial cities increased, the severity of casualty crashes, and the proportions of head-on and out-of-control crashes increased significantly. Interestingly, this increase commenced as the speed transition zone was entered. Tziotis (1993) also investigated the site conditions and various road environment factors to determine whether or not they had contributed to, or had the potential to, contribute to road crashes. The major road features in the transitional zone that were identified as contributing factors according to their relative importance were: poor curve delineation (40%); traffic islands not provided, in particular in right-turn lanes (33%); unprotected/hazardous poles and trees (both 33%); unsealed (19%) and inadequate shoulder sealing (14.0%); unsatisfactory entry to carriageway (i.e., from minor roads, service roads, private driveways and public facilities, 30%); edge-lines not provided (12%), or poorly maintained (9%). Tziotis concluded that road design on approaches generally fails to respond to the changes in roadside development, particularly as they relate to the interaction of locally generated traffic, and non-local, through traffic travelling at generally higher speeds. 2.4.2 Road lengths

In this section, the effects of geometric features of road lengths on crash and injury risk are discussed. First, the influence of road type is discussed, particularly the risks associated with single-carriageway roads. Secondly, the influence of pavement type is discussed, particularly the risks associated with unsealed pavements. Lastly, the issues relating to provision of and geometric features of overtaking lanes on single-carriageway roads are discussed. 2.4.2.1 Road type As noted previously, the rural network is made up of different road types and road type has been shown to affect crash risk. Generally, motorways and dual-carriageways are the best performing roads in terms of crash rate, because they preclude some of the crashes that otherwise account for the majority of fatal and serious injuries. Schmidt (1996) noted that, although vehicles tend to travel at higher speeds on road stretches with a high design standard, these roads also have lower crash rates. This is because these types of roads have a high standard of sealing, opposing traffic flows separated by median safety fences, wide carriageways, well-designed or grade-separated intersections, pedestrians and cyclists are excluded, and there are safe roadside areas or barriers to prevent injury when vehicles run off the road. These types of roads carry high traffic volumes and, because of increased exposure, can show a high frequency of crashes. While the frequency of crashes is an important issue, it has not been adequately covered in the literature reviewed. The most important factor identified in the literature is that the high-risk roads are generally single-

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

25

carriageways, often carrying low levels of traffic travelling at relatively high speeds (Benekohal, 1990; Kalakota, Seneviratne & Islam, 1992; Schouten, 2000). Of the few studies that have examined crash risk on dual-carriageway and multi-lane roads, Karlaftis and Golias (2002) noted that average annual daily traffic (AADT) was the most important variable affecting crash risk. Other important factors included median width, access control, and pavement condition. The authors noted that, on roads with heavy traffic, the importance of access control increases because of the need for vehicle manoeuvring arrangements in these traffic situations. In contrast, on roads with low flow conditions, the existence of a median becomes more important, presumably because of higher operating speeds. Walmsley, Summersgill and Payne (1998) also reported that dual-carriageway roads with a higher proportion of obstructions on the median had a higher crash frequency. Specifically, for each increase of ten percent in the proportion of the rural link with an obstruction, the crash frequency increased by 15% overall, with greater increases for crash groups involving a vehicle overtaking or stopping on the carriageway. Turning to the effect o f single-carriageway roads on crash risk, it is noted that many singlecarriageway highways were constructed for slower moving vehicles and at present are generally not appropriate for faster vehicles and heavy traffic volumes. In addition, these roads tend to have frequent minor and uncontrolled intersections and have many roadside hazards. On these roads, the risk of head-on collisions, collisions at intersections and runoff-road crashes are consistently associated with high injury risk. Pyne, Carsten and Tight (1995) reported that rural single-carriageway arterials in England had the highest rate of involvement in fatal crashes of any category of road (2.1 involvements in fatal crashes per 100 million vehicle kilometres as compared to 0.6 on motorways and 1.5 on urban arterials). Two recent studies in England (Walmsley et al., 1998; Walmsley & Summersgill, 1998) confirmed the high risk on single-carriageway roads compared to dual-carriageway roads, reporting 10.4 injury crashes per 100 vehicle kilometres on dual-carriageway roads, excluding intersections, and 14.3 injury crashes per 100 vehicle kilometres on these roads, including crashes at intersections. Rates on singlecarriageway roads were appreciably higher at 12.5 injury crashes per 100 vehicle kilometres and 22.4 injury crashes per vehicle kilometres, excluding and including crashes at intersections, respectively. Kalakota et al. (1992) also noted that approximately 2.5 million miles or 63% of highways in the USA are two-lane rural highways and 50% of fatalities occur on these roads, noting that this rate is four to seven times higher than that on rural interstate highways. Likewise, Baruya (1998) found that European roads with a low classification (those that are highly congested, with average to high flow, narrow to medium width, low to average traffic speed and speed limits between 90 and 100 km/h) had higher average crash frequencies than roads with a high classification (those with low flow, very wide roads, low congestion, high traffic speed and speed limits of 110 km/h). Despite these high crash rates on single-carriageway roads, Walmsley et al. (1998) did observe that the latest modern single-carriageway roads are safer than those built earlier, with those constructed since the mid-1980s being 14% safer than those constructed before this time. They concluded that safety had improved because of improvements in road design, an increase in the use of hardstrips (i.e., sealed shoulders), safety fences, wide carriageways and fewer and better designed intersections.

26

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

Nevertheless, there have been numerous studies that have attempted to identify the features of single-carriageway roads that contribute to high crash and injury risk, and have identified a number of factors that are related to crash occurrence and injury severity (Benekohal, 1991; Walmsley et al., 1998). These include: AADT, section length, lane and shoulder width, shoulder type, slideslope, horizontal and vertical alignment, and presence of accesses and intersections, (see Sections 2.4.4 and 2.4.5 for detailed discussions of the contribution of roadside features and intersections to crash and injury risk). With regard to the influence of intersections on crash risk, Walmsley and Summersgill (1998) reported that the presence of intersections (both major and minor) was a major determinant of the numbers of crashes on the rural road network as a whole, both on dualcarriageway and single-carriageway roads. On a typical dual-carriageway road, each major junction contributed to crash risk by approximately 22%, and on a typical rural singlecarriageway, this was about 11%. On both types of road each minor junction contributed an increase of about six percent. This finding is consistent with Hunt and Mahdis (1996) research in which it was concluded that crashes often occur close to intersections and the frequency may be related to the turning movements between vehicles (refer Section 2.4.5). With regard to pavement edges, Walmsley et al. (1998) observed that the main feature of carriageway design which affected crash rates on modern rural trunk roads was the edge treatment (the presence or otherwise of kerbs and hardstrips [i.e., paved shoulders]). On both single- and dual-carriageway roads, crash rates on roads with hardstrips were consistently lower than roads without. Dual-carriageway roads without hardstrips recorded approximately 16% more crashes than roads with hardstrips (most with hardstrips on both sides and some with hardstrips only one side of the carriageway). Similarly, singlecarriageway roads without hardstrips recorded approximately 18% more crashes than those with. With regard to carriageway width, Walmsley and Summersgill (1998) reported that crash rates on standard-width, single-carriageway roads (e.g., 7.3m wide) were higher than the rates on wide roads (e.g., 10m wide), by approximately 22%. Interestingly, a standardwidth carriageway combined with edge treatment (hardstrips) resulted in similar crash rates to those on a wider carriageway without edge treatment, suggesting that it is the edge treatment, rather than carriageway width, that affected crash rates. In addition, there was some indication that three-lane, dual-carriageway roads had a lower crash rate than twolane dual carriageway roads. The authors concluded that narrower roads are less safe than wider single-carriageways, and probably wider dual-carriageways, and roads without hardstrips are less safe. They noted, however, that this may not be a robust finding due to the small sample size. Karlaftis and Golias (2002) recent investigation of the roadway characteristics that effect crash rates on both rural two-lane and multi-lane roadways in the US state of Indiana showed that, for both rural two-lane and multi-lane roadways, the AADT was the most important variable. Looking at the effect of other variables normalized on AADT (i.e., statistically cancelling out the effect of AADT), Karlaftis and Golias found that lane width and pavement condition factors (such as serviceability index, pavement type and friction) were the most important variables affecting crash rates on two-lane rural roads. Interestingly, they noted that the effect of lane width on crash risk appeared to increase with increased traffic flows. In contrast, the importance of pavement condition on crash risk appeared to increase with lower traffic flows, presumably because of the higher speeds that can occur when traffic volumes are low.

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

27

2.4.2.2 Pavement type The limited literature addressing the effect of road surface on rural road crashes has provided results that, to date, are not entirely consistent, and a clear relationship between road roughness and crash rates is yet to be established (Al-Masaeid, 1997). One of the problems is the fact that many rural roads include both sealed and unsealed sections, and crash trends specifically for unpaved sections are generally not available (Caldwell & Wilson, 1996). There are some studies that report that unsealed roads occupy a large proportion of the rural road network in many countries and experience high levels of crash and injury risk (Cairney, 1995; Caldwell & Wilson, 1996; Calvert & Wilson, 1999; Jordaan & Mashiri, 1994), with levels increasing as road roughness increases (Jacobs, 1976; Kamel & Garshore, 1982). Caldwell and Wilson (1996) reported that the number of injury crashes per mile travelled on Wyoming (USA) unsealed roads is five to six times as high as on the states sealed road network. Calvert and Wilson (1999) later added that m any of these unpaved rural roads worldwide have geometric deficiencies that do not conform with recognized standards and guidelines. In many instances, roadway improvements are not being completed because of the inability to fund improvements to meet these standards and guidelines. Moreover, conditions of unsealed roads can deteriorate rapidly, particularly after long periods of wet weather. Cairney (1995) reported that many of the crashes on rural unsealed roads in Australia are run-off-road and head-on crashes. In South Africa, too, Jordaan and Mashiri (1994) reviewed the collision patterns of approximately 570 crashes that had occurred on unsealed roads. Their analysis revealed that most gravel road collisions were single-vehicle crashes. Furthermore, these crashes typically occurred on straight sections of the road, and to a smaller extent on curves, but generally away from intersections. The authors identified the six most important causes of collisions: loss of control over vehicle (20-24%); animals on the roadway (8-30%); turning movements of vehicles (9-15%); tyre burst (6-12%); skid on road surface (2-17%); and pedestrian on the roadway (3-11%). Other studies have failed to find increased crash and injury rates on unsealed roads compared to sealed roads. Indeed, Halawa (1993) reported that pavement conditions had n o effect on the total traffic crash occurrence within the frontage distance of rural towns. Cleveland (1987) further reported that on two-lane rural roads with traffic volumes in the range of 1000 to 8000 vehicles per day, crash rates increased significantly with the smoothness of the road surface. This finding is consistent with the view that smooth surfaces allow higher travel speeds, which result in higher crash rates. The most recent study investigating the potential effects of road surface as well as v arious road geometry and traffic conditions on rural road crashes is by Al-Masaeid (1997). For each road section examined, the pavement condition was measured in terms of international roughness index (IRI) and the present serviceability rating (PSR). The IRI parameter summarises the longitudinal surface profile of the road in a wheel track (Paterson, 1986), while the PSR parameter is also an indicator of riding quality or road condition. Al-Masaeid (1997) found some interesting differences in the effect of road surface on single- and multi-vehicle crash rates. For single-vehicle crashes, a high correlation between crash rate and road condition was found, and expressed in terms of IRI and PSR, in
28 MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

addition to correlations with number of horizontal curves and presence of roadside hazards. Like Cleveland (1987), Al-Masaeid reported that an increase in the road roughness level tended to reduce the single-vehicle crash rate. He explained this finding by suggesting that rougher roads reduce the quality of driving, which reduces the speed of traffic and consequently reduces the likelihood of a crash. This explanation is consistent with a recent OECD (1994) finding that when the road deteriorates and roughness increases, vehicle speeds are progressively reduced in the interest of comfort and perceived safety. Al-Sharaf (1995), too, reported that the mean speeds of passenger cars and light trucks decreased significantly as roughness or IRI increased. For multi-vehicle crashes, the reverse was true. Multi-vehicle crashes increased significantly with road roughness. Al-Masaeid suggested that the increase in multi-vehicle crash rates on rougher roads may be associated with two factors. First, the lateral variation in a vehicles path along two-lane roads increases with the increase in road roughness. This factor might reduce the clearance distance between vehicles and therefore increase multivehicle crash rates. Secondly, rougher roads tend to have road defects such as potholes and improper patching. These defects appear to force drivers to change their travelling speed abruptly. Previous studies have shown that the necessity to change speeds abruptly may increase the probability of multiple-vehicle crashes (Lamm, Hayward & Cargin, 1986). Finally, Al-Masaeid (1997) reported that the total crash rate (single- and multi-vehicle crashes combined) correlated highly with the number of vertical and horizontal curves, the number of intersections, and roadside hazards. However, pavement condition variables did not significantly contribute to the total crash rate. Al-Masaeid concluded that this finding may be due to the fact that the increase in multi-vehicle crashes on rougher roads was compensated for by the decrease in single-vehicle crashes on rougher roads. 2.4.2.3 Overtaking lanes Research on overtaking crashes is comparatively rare, despite the frequency and severity of these types of crashes on rural roads. Crashes resulting from risky or misjudged overtaking manoeuvres are typically fairly serious. They generally result in head-on, sideswipe or rear-end collisions and occur mainly on two-lane rural roads. Clarke, Ward and Jones (1998) reported that overtaking crashes accounted for eight percent of fatal crashes in Nottinghamshire, England, and that their crash severity index (the proportion of cases resulting in death or serious injury) is over 20%. In Australia, too, Armour (1984) found that overtaking is involved in about ten percent of rural casualty crashes. These rates are much higher than the reported 3-4% of such crashes in the USA (Khasnabis, 1986; Mohamedshah, 1992), which probably reflects the much greater length and usage of multilane, divided roads in that country. The overtaking manoeuvre on two-lane rural roads without the assistance of additional passing lanes is a complex driving task, requiring critical information processing and decision-making skills and a lengthy section of road to complete the manoeuvre (enough distance to allow a sufficiently large gap in the oncoming traffic, plus the distance travelled by that vehicle, plus a safety margin). The ability to see oncoming traffic clearly from a maximum distance is important for safe overtaking. Extended roadway sections with severe sight restrictions and inadequate passing opportunities pose a real safety problem on rural roads because of the difficulty passing a slow moving vehicle, increased delay, conflict and driver frustration. The ability to pass a slower moving vehicle safely without the assistance of passing lanes is influenced

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

29

by a variety of factors. These include the volume of through and opposing traffic, the difference in speed between the passing and passed vehicle, driver characteristics such as reaction times and risk perception, and the road geometry particularly the available sight distance (Fildes, Oxley & Corben, in press). For many years, most countries have provided overtaking lanes on unsustainable grades where some vehicles (particularly trucks and buses) experience a problem maintaining desired speed. While overtaking lanes are becoming more common at locations other than on unsustainable grades, there are still many two-lane rural roads that do not provide adequate passing opportunities. Moreover, questions which have been raised about current design practices of passing lanes suggest that the geometric features of passing lanes affect crash risk. Hughes, Joshua and McGee (1992) questioned the minimum length of 122 m (400 ft) for a passing zone in the USA, stating that, for design speeds above 48 km/h (30 mph), the distance required for one vehicle to pass another is much longer than 122 m. Weaver and Glennon (1972) reported that most drivers cannot pass even within a 244 m (800 ft) section and that use of passing zones remains very low when their length is shorter than 274 m (900 ft). Further, Polus, Livneh and Frischer (2000) argued that, in todays traffic where 25% or more of passing manoeuvres are multiple passings (i.e., in which more than one vehicle is overtaken in one manoeuvre), longer distances are required and therefore longer sight distances than specified in current guidelines. Hughes et al. (1992) also noted that current minimum passing distances are inadequate for the abortive manoeuvre, and that they do not consider the length of the vehicle being overtaken. 2.4.3 Horizontal curves

Horizontal curves are one of the most dangerous parts of the rural road network and have attracted a substantial amount of attention in the safety literature. The literature consistently reports an over-representation of single- and multi-vehicle crashes at, or near curves, particularly on two-lane rural roads. Crash rates on curves are estimated to range from 1 to four times higher than on tangents (Johnston, 1982; 1983; Neuman, 1992; Troxel, Ray & Carney, 1994; Zegeer, Stewart, Reinfurt, Council, Neuman, Hamilton, Miller & Hunter, 1991) with an even higher rate of 4 for truck crashes (Leonard, Bilse & Recker, 1994). In their investigation of geometric variables of road lengths on crash and injury risk, Walmsley et al. (1998) estimated t hat a rural link with twice the average number of curves per kilometre would have around ten percent more crashes than straight road lengths in total, and up to 60% more crashes involving a vehicle leaving the carriageway. In Germany, Steyer et al. (2000) noted that nearly half of crashes on non-built-up roads (i.e., rural roads) occurred on curved roadway sections. In the UK, Taylor and Barker (1992) found that, of all crashes on rural two-lane roads, 18.5% occurred on curves. Similar rates are found in Denmark, with 20% of all personal injury crashes and 13% of all fatalities occurring on horizontal curves in rural areas (Nielsen & Griebe, 1998). In France, the situation is worse, where 21% of all fatalities occurred on rural curves. In the Australian State of New South Wales, a reported 48% of all fatal crashes on rural highways occurred either on a curve or on straight road sections where nearby curves were partially responsible for the crash. Of these fatal crashes, 70% occurred on acute curves where the radius of the curve was less than 300 m (Moses, 1990).

30

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

A recent study of fatal single-vehicle crashes in Victoria, Australia (Haworth et al., 1997) revealed that about one-third of the crashes investigated occurred on curves (27% occurring in metropolitan areas and 36% occurring in rural areas). Puvanachandran (1995) similarly reported that 36% of all crashes on rural roads in Papua New Guinea occurred on curves. Likewise, Lamm, Guenther and Choueiri (1995) noted that at least 30% of the fatalities that occur on the rural road network system in both Germany and the USA, occur on curved roadway sections. The incidence of crashes on curves as a proportion of all crashes is clearly related to general road alignment, which in turn is influenced by the local terrain. Regardless, it is evident that higher crash risk, especially for single-vehicle crashes, exists on curves. More severe curves are associated with higher crash rates. Crashes are more likely to occur on curves than on straight segments of a roadway because of increased demands placed on the driver and the vehicle. Negotiating a curve constitutes a more difficult driving task than driving along a straight section of road. A vehicle entering or departing a horizontal curve must safely undergo a change in steering angle and a resulting change in side friction forces. Successful curve negotiation, therefore, depends upon the choice of appropriate approach speed, appropriate deceleration and adequate lateral positioning through the curve and loss of control crashes result from an inability to maintain lateral position through the curve because of excessive speed (not corresponding to the alignment of the roadway) and inadequate deceleration in the approach zone. It appears that perceptual factors play an important role in crashes on curves. These factors include poor anticipation of vehicle control requirements on approach and within the curve, inadequate perception of the demands of the curve, and inadequate appreciation of the degree of hazard associated with a given curve. There are a number of crash types that are over-represented on curves as compared to tangents. These include: i) single-vehicle run-off-road, ii) multiple-vehicle collision between vehicles travelling in the opposite direction (head-on collision), and iii) multiplevehicle collision between vehicles travelling in the same direction (rear-end collision) (Anderson, Bauer, Harwood & Fitzpatrick, 1999; Zegeer et al., 1991). Undoubtedly, crashes on curves result from complex relationships between the driver, the vehicle and the road. Nevertheless, given that high rates of crashes continue to occur at these locations, it may be reasonable to assume that unfavourable road or roadside conditions exist. There have been a number of studies that have attempted to examine the relationship between crash risk and operational measures (such as encroachments and speed reduction) and non-operational measures (degree of curvature, grade, shoulder width, distance since last traffic event, super-elevation and roadside conditions) (e.g., Lamm, Beck & Zumkeller, 2000; Lamm et al., 1995; Lamm, Choueiri, Hayward & Paluri, 1988; Terhune & Parker, 1986; Zegeer et al., 1991). While the causes of the high rates of crashes on curves are not yet fully understood, some have found relationships between crash and injury risk and geometric features of curves and attributed the high frequency of crashes on curves to a range of factors listed below (Council, 1998; Hall & Zador, 1981; McBean, 1982; Lamm et al., 2000; Steyer et al., 2000; Taylor & Foody, 1966; Voigt, 1996; Voigt & Krammes, 1998; Wright & Zador, 1982): Curve features (including a decrease in the degree of curvature, curvature change rate, ratio between curve angle and arc length, super-elevation and side friction, and design of curve transition section),

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

31

Sharpness of curve relative to standard of adjacent geometry (i.e., unexpected sharpness), Cross-sectional curve elements (lane width, shoulder width, shoulder type, shoulder slope), Roadside hazards on the curve (clear-zone, sideslope, rigidity and types of obstacles), Traffic volume on the curve and traffic mix (e.g., percentage of trucks), Stopping sight distance on curve or approach to curve, Presence/distance of curve to other horizontal and vertical features (such as adjacent curves, intersections, driveways, bridges, etc.), and Presence and type of traffic control devices (signs and delineation).

In general, previous research on geometric measures as predictors of crash experience on horizontal curves have focussed primarily on measures of curve features, often in isolation of other contextual variables. Degree of curvature has consistently been found to be the primary contributing factor to crash risk. Other factors such as shoulder width, presence of other hazardous features and traffic are less well researched, but some studies indicate a moderate effect on crash risk. The effects of these factors on crash risk are discussed in the following sections. 2.4.3.1 Speed on curves Vehicles travelling at high speeds on curved road sections not only risk departure to the outside of the curve, but also transgressions into the opposite lane (Persaud, Retting & Lyon, 2001; Puvanachandran, 1995). Crashes on curves primarily occur when drivers exceed the critical design speed of a curve by substantial amounts, especially at lower design speed levels, and either lose control of their vehicle or are forced into a cornercutting movement in order to maintain control of the vehicle (Comte & Carsten, 1998; Lamm et al., 1995). 2.4.3.1.1 Design speed

It has been suggested that safety problems occur on horizontal curves on two-lane rural roads because many of these roads are old and pre-date the design speed concept, tend to be sub-standard and have wide variation in the maximum speed at which different elements can be safely negotiated, despite the fact that the speed limit is constant (Comte & Carsten, 1998). While more recent roads, constructed using the design speed concept, may be safer than older roads, there still appear t o be some problems associated with these procedures resulting in increased crash and injury risk on horizontal curves. Current horizontal curve design procedures in most western countries employ the designspeed concept, created to provide a consistent operating speed profile along an alignment using two principles. First, all curves on an alignment should be designed for the same speed and secondly, the design speed should reflect the speed that a high proportion of drivers desire to maintain. However, this concept does not recognise that a discrepancy between operating and design speed may exist and therefore provides no guidance to eliminate potentially dangerous inconsistencies in alignment. Voigt (1996) noted that this

32

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

concept works well when the design speed selected for a roadway adequately represents the desired speeds of drivers. However, he also added that if the selection of the design speed is in error (i.e., set too low), undesirable operating speed/design speed disparities become apparent. Indeed, there have been numerous reports of both 85th percentile speeds exceeding design speeds on curves with design speeds less than 90-100 km/h; and 85th percentile speeds lower than design speeds on curves with design speeds greater than about 100 km/h (see Voigt, 1996). The desired speeds of drivers are undoubtedly influenced by the speed limit. If the speed limit is set well above the design speed of curves, variations in speed along a road will be accentuated, with consequent increases in crashes and injury in the vicinity of, and within, curves. 2.4.3.1.2 Operating speed and speed reductions

Several studies have evaluated operating measures as predictors of crash rates on horizontal curves on rural two-lane highways. Most, however, have been limited to small databases and the findings are inconsistent. For example, Taylor et al. (cited in Krammes, Brackett, Shafer, Otteson, Anderson, Fink, Collins, Pendleton & Messer, 1993) found that the speed-based measure with the best (albeit not statistically significant) correlation with crash rate was the deceleration value from the beginning to the mid-point of the curve. Puvanachandran (1995) argued that it is prevailing approach speeds, rather than geometric features of curves that determine the speeds of a vehicle on curved road sections. In contrast, Datta, Perkins, Taylor and Thompson (1983) and Terhune and Parker (1986) found no correlation between operational measures and crash risk, only degree of curvature. Zegeer et al. (1991) re-examined the data from these studies and concluded that high rates of speed reduction (i.e. deceleration) and vehicle encroachments, as well as geometric features, most affected crash risk. Krammes et al. (1994), too, concluded that both mean speed reduction and mean degree of curvature were good predictors of the mean crash rate on horizontal curves. Since the 1960s, many researchers have recognised the fact that abrupt changes in operating speeds lead to crashes on two-lane rural roads, and that these inconsistencies or variations may be largely attributed to abrupt changes in horizontal alignment (Lamm et al., 1995; Leisch & Leisch, 1977). Anderson, Bauer, Harwood and Fitzpatrick (1999) found a strong relationship between the speed reduction on a horizontal curve and crash frequency on that curve. In other words, the greater the speed reduction experienced by drivers on a horizontal curve, the greater the curves crash experience. They suggested that speed reduction appears very promising as a design consistency measure. The issue of speed variability and its effect on crash risk has been addressed in considerable detail over the last decades and many have concluded that an increase in speed variation increases crash risk (see Cirillo, 1968; Cumming & Croft, 1971; Hauer, 1971; Solomon, 1964; West & Dunn, 1971). Lindeman and Ranft (1978, cited in Collins, Fitzpatrick, Bauer & Harwood, 1999) examined geometry effects of speed distribution measures at the midpoint of horizontal curves with radii ranging from 33 to 1000 m. Their findings suggested that standard deviation of speed increased as curve radius increased for small radius curves, and thus with increasing speed, and remained constant for larger radius curves.

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

33

One of the important factors in crash risk on horizontal curves is approach speed. In his examination of driver speed on horizontal curves, McLean (1981) found that speeds depended not only on the radius of curvature but also on the speed environment which is defined as the 85th -percentile speed on level tangent sections reflecting the drivers perception of the overall speed standard of a road. Lamm et al. (1995) examined the safety criteria for evaluating curved roadway sections with respect to the differences in 85th percentile operating speeds by classifying roadway sections as good, fair or poor designs in Washington, USA. They found that the crash rate was highest for horizontal curves in the poor category, with 2.76 crashes per million vehicle kilometres driven and lowest for the horizontal curves in the good category, with 0.46 crashes per million vehicle kilometres driven (the crash rate for fair curves was 1.44 crashes/million veh-km). These results suggest that horizontal curves that require drivers to make greater speed reductions from the approach tangent are likely to have higher crash rates than horizontal curves requiring lower speed reduction. These findings seem intuitively sound. Collins et al. (1999) also examined the effect of roadway geometry of curves on speed distribution and hypothesized that increased speed variance can be used as an indicator of design inconsistency. First, they found no relationship between geometric features (curve length, super-elevation, lane and pavement width, grade) and speed variance, with one exception curves with radii below 100 m resulted in smaller standard deviations compared to larger radius curves. That is, for horizontal curves with radii smaller than 100m, higher-speed drivers entering the curve reduced their speeds more than lower-speed drivers. Secondly, they found that higher posted speed limits generally resulted in higher standard deviations, and that higher standard deviations are associated with higher speeds. They concluded that speed variance is not an appropriate measure of design consistency for horizontal curves and that, while an increase in speed variance may be an indicator of potential safety problems for some geometric design features or traffic situations, it is not useful in explaining safety differences between tangents and horizontal curves on two-lane highways. 2.4.3.2 Radius of curve / degree of curvature The average radius expresses the sharpness of curves that drivers typically encounter on a given section of the roadway. A large average radius would indicate curves that are typically not sharp and it would be expected that higher speeds would exist on these types of curves. In contrast, a curve with a small average radius, indicates that the curves are quite sharp and lower speeds would be expected in these conditions. In general, degree of curvature is clearly the geometric feature that most affects crash experience and vehicle operations on horizontal curves, where operating speeds decrease with increasing curvature and crash rates increase with increasing degree of curvature. As indicated previously, negotiating curves constitutes a more difficult task than driving on straight roads. This difficulty increases as degree of curvature increases. Krammes, Brackett, Shafer, Otteson, Anderson, Fink, Collins, Pendleton and Messer (1994) examined operations, geometric features, driver workload and crashes on horizontal curves in the USA and found a linear relationship between driver workload (measured by occluded vision tests) and degree of curvature, with workload increasing as degree of curvature increases.

34

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

Cairney (1998) noted that sections with curvature of between 5 -10 have at least twice the crash rate of sections with a curvature of 1 -5, and sections with curvature of between 1015 have crash rates four times as great. In t erms of curve radius, 200 m seems to be the point below which crash rate greatly increases. Lin (1990) examined curves with gentle curvature and which were free of bridge structures, driveways and intersections located on 33 State highways in the USA and found that crash rates increased as a function of curvature, noting that as curvature exceeds 9, there is a large increase in the probability of being hazardous. He also noted that this curvature coincides with the curvature at which a speed reduction of 24 km/h is expected of a driver travelling from a tangent section to a curve and, more importantly, that 85th percentile speeds exceed the maximum attainable design speed when the curvatures are greater than 10, that is, the majority of drivers drive at inappropriately high speeds on sharp curves. Anderson and Krammes (2000) examined the relationship between mean crash rate and mean degree of curvature, showing that horizontal curves that require speed reductions (generally, curves sharper than about 4, which corresponds to design speeds less than 100 km/h and estimated 85th percentile speeds less than drivers desired speeds on long tangents) had higher crash rates than curves that do not require speed reductions. More specifically, they found that mean crash rates were similar for degrees of curvature from 0.25 to 4. They described 4 as a breakpoint, after which crash rates increased linearly for the remaining intervals 5 and over. They also noted that when curve sites were grouped into speed-reduction intervals, there was a significant relationship between the intervals mean crash rate and mean speed reduction. In short, the mean crash rate increased approximately linearly with the mean speed reduction. As a final note, it is clear that sharp curves result in significantly increased rates of crashes, as well as high rates of speed reduction and vehicle encroachments. However, Wegman (1995) also notes that it is not just the sharpness of a curve in isolation crash rates on curves are only relatively high when the average curvature of the whole alignment is low and when a bend follows a long straight tangent. Moreover, Wegman cites some studies that show that curve-related factors also have important effects, especially on curves with a small or medium average radius of curvature. The main problem here is irregularity of the curvature throughout the bend. He therefore argues that consistency is vital to create predictability and so to prevent driver error and therefore crashes. 2.4.3.3 Curve transition The design of the transition from tangents to curves can affect crash risk because it is here that drivers make decisions about the appropriate speed to travel through a curve and adjust their speed accordingly. Council (1998) noted that a curve transition section has two important safety-related functions. First, it acts to direct a driver into a safe path on the approach to a curve while (s)he is changing the steering wheel position from a relatively straight position on the tangent to another fixed p osition on the curve. Secondly, it provides a space for super-elevation to change from the normal design on the tangent to the full super-elevation required by the curve, thus minimising side friction forces. There are three forms of curve transition including: i) no transition, where the tangent abuts the curve in this situation, the driver must choose his/her own transition path, ii) a compound transition, where a section of less sharp curve is placed between the tangent and

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

35

the main curve, and iii) a spiral transition, which begins as a tangent and ends with the same degree of curvature as the main curve. Council (1998) argued that optimally designed curve transitions are an important safety feature, as in approximately 62% of fatal crashes and 49% of injury crashes on curves, the first manoeuvre in the crash sequence was at the beginning or end of the curve, rather than in the centre. Lamm et al. (2000) noted that the dominant influence on crash rate is curvature change rate. They showed a significant increase in crash and injury risk with increasing curvature change values, particularly for rates with values greater than 200 gon/km (a measure of the absolute sum of curvature change rates) that correspond to radii less than 320 m without regarding transition curves. 2.4.3.4 Super-elevation The negotiation of curves, especially more severe curves along high-speed roadways, can be facilitated by the provision of roadway super-elevation through the curve. That is, by introducing into the design of a curved section of roadway a lateral slope towards the centre of the curve, the side friction required by a vehicle to safely negotiate the curve is markedly reduced. A number of studies have found significant relationships between super-elevation deficiency and crash experience, and between side friction and crash rates (Voigt, 1998; Voigt & Krammes, 1998; Zador, Stein, Hall & Wright, 1985; Zegeer et al., 1991). Zegeer et al. (1991) revealed a significant relationship between total crash rate and degree of curvature, width, presence of spiral, and super-elevation deficiency. They estimated the crash reduction corresponding to a 0.02 super-elevation deficiency on a 3 curve with a width of 30 ft, no spiral, and 0.3 million vehicle miles to be about 10.6%. They also concluded that too much super-elevation was not associated with higher crash rates. Voigt and Krammes (1998) examined in detail the relationships between 85th -percentile speed on horizontal curves, and super-elevation and crash experience on horizontal curves with super-elevation deficiencies on rural two-lane highways, noting the disparity between operating and design speed. They found that super-elevation had significant effects on 85th percentile speed on rural two-lane horizontal curves. In addition, they found that operating speed reduction and super-elevation deficiency were significant crash predictors, and that implied side friction demand (based upon the actual radius or degree of curvature and super-elevation rate) was the strongest crash surrogate. The design standards for setting super-elevation rates along curved sections of roads are based on design speed, the consideration of passenger vehicle driver comfort, the limiting co-efficient of friction of slow-moving vehicles on wet roadway surfaces, the tendency of slow vehicles to track to the middle of the curve, the stability of high-profile vehicles, and the length available to introduce super-elevation (AASHTO, 1990). Super-elevation rate may be as high as 12%, depending on the curve radius and design speed. However, there seem to be several flaws in the application and distribution of superelevation and side friction that can weaken the physical relationship between vehicles and the roadway (Kanellaidis, 1991; Krammes et al., 1994; Voight & Krammes, 1998). For instance, it has been argued that the methods described in AASHTO guidelines use a flawed assumption that drivers will operate at the design speed even on curves where they

36

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

feel comfortable operating at higher speeds (McLean, 1981; V oigt, 1996), and fail to take into account the tendency of vehicles to rollover on a curve, even at relatively low speeds, particularly heavy vehicles with a higher centre of gravity than the average passenger vehicle (Krammes et al., 1994). In addition, t hese guidelines encourage the use of aboveminimum design values on horizontal curves, which may encourage operating speeds greater than the design speed of the controlling element. Krammes et al. (1994) and Krammes (1998) further argued that the lack of uniformity in US practice among different jurisdictions and consequent variation in maximum super-elevation rates complicates the drivers task of selecting the appropriate speed on curves. 2.4.3.5 Lateral placement One of the operating measures identified as a contributing factor to crash risk on horizontal curves is vehicle lateral placement (Taylor et al., (undated) & Stimpson et al., (undated) both cited in Krammes et al., 1993; Steyer, 2000). Steyer et al. (2000) noted that one of the important safety-related features in curve negotiation is vehicle lateral placement and they argue that the driving path should be considered when investigating crashes on curves. They make the distinction between right and left curves and offer some suggestions as to why centre-line encroachments occur. Centre-line encroachments on curves to the drivers near-side (that is, left curves in the USA and most European countries and right curves in Australia, New Zealand and the UK) may be considered controlled or intentional encroachments where some drivers intentionally cut the corner or straighten on the curve if they cannot detect any opposing traffic. They argue that these types of encroachments are mainly associated with the radius of the curve, curve length, grade and available sight distance. Centre-line encroachments on curves on the drivers far-side (that is right curves in the USA and most European countries and left curves in Australia, New Zealand and the UK) may result because of excessive speed and a tendency for drivers to steer away from roadside hazards. 2.4.3.6 Trucks While there is little research that addresses the effect of road and geometric factors on truck crashes along curved sections of roads, there is some evidence that articulated trucks are particularly sensitive to the characteristics of the roadway and are vulnerable to crashes due to loss-of-control at these locations (Leonard et al., 1994; Mohamedshah et al., 1993). Mohamedshah et al. (1993) noted that trucks are over-represented in crashes on horizontal curves arguing that horizontal curves can contribute to rollover problems for trucks with high centres of gravity or when the load shifts while negotiating a curve. Moreover, they can encroach on the opposite lane while negotiating a curve, posing a hazard for opposing vehicles. In their analysis of crashes in Los Angeles in the USA, Leonard et al. (1994) showed that typical truck crashes on curves occurred in clear weather (72%), on dry roads (78%), on weekdays (88%), during daylight hours (64%), on roads without unusual features (63%) and are most associated with speeding or other hazardous violations (78%). Leonard et al. also examined the effect on crash risk of reducing mainline super-elevation at intersections on curves. This practice is used to facilitate vehicles turning onto the major road from the minor intersecting cross road. While this practice was found to decrease accelerations because of the slope of the major road and, therefore, increase safety for vehicles entering the major roadway, there was some suggestion that safety for vehicles (trucks in particular) travelling on the major roadway was slightly reduced because of the effect of greater lateral acceleration.

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

37

2.4.3.7 Weather conditions In addition to geometric features of horizontal curves, adverse weather conditions can exacerbate the problems experienced at these locations. Shankar, Mannering and Barfield (1995) examined the interaction of weather and geometric elements and their consequent impact on crash likelihood and found that exposure to rainy and snow conditions (with associated wet pavements, low visibility traction and lane-marking-related problems) played a significant role in crash occurrence. They argued that in areas that frequently experience adverse weather conditions, design criteria should be extended to avoid steep grades and horizontal curves with low design speed. In particular, they recommended that eliminating all horizontal curves with a design speed less than 96.5 km/h on a roadway section that experiences at least 5.1 cm of snowfall on one or two days a month can reduce the monthly crash frequency by 47.3%. 2.4.4 The roadside

The roadside environment has a significant influence on crash and injury risk. Sideslope and roadside ditch geometry influence a drivers ability to recover from an unanticipated excursion off the paved surface. More importantly, roadside objects are the most notorious cause of serious injury in crashes on rural roads and single-vehicles leaving the roadway and striking a fixed object have been identified as the most common crash type in rural areas in most countries (Hall, 1991; Ogden, 1997; Sanderson & Fildes, 1984). Utility poles, trees, and non-yielding signs are some of the objects that cause serious problems on rural roads such objects have been and continue to be inconsistent with the notion of a forgiving roadside and create substantial trauma and costs for individuals and society (Corben, Deery, Mullan & Dyte, 1997). Since the mid-seventies, there have been reports of over-representation of serious injury crashes involving a single vehicle running off the road. In the USA, Agent and Deen (1975), Hall and Zador (1981), and Wright and Zador (1981) reported that run-off-road crashes account for between 15 and 53% of all road crashes, depending on the area studied, the type of road and the time of day. In Canada, Cooper (1980) reported that at least onethird of all fatal and other casualty crashes in rural areas involved single vehicles running off the road. In New Mexico, 45% of fatal crashes were single-vehicle crashes, with a high proportion of these being run-off-road crashes in sections of rural highways. The most recent US study reported that single-vehicle, run-off-road crashes in the USA account for around 30% of all highway fatalities, resulting in 0.5 million people injured per year and cost the society around US$80m per annum (Griffith, 1999). Sanderson and Fildes (1984) analysed run-off-road crashes in Victoria, Australia. They reported that 62% of vehicles struck a fixed object, and that these are commonly severe crashes, with nine percent resulting in a fatality, 58% requiring hospitalisation of at least one occupant, and 33% requiring medical attention. Interestingly, the severity outcomes were similar for the 38% of crashes where vehicles ran-off-the-road but did not strike an object. Crashes of this type may cause serious injury as the vehicle encounters steep or uneven terrain and rolls over or the occupants are flung around and strike other occupants or elements of the vehicles interior. Of those crashes in which a vehicle struck a fixed object, almost half (47%) of these crashes involved striking trees or shrubs, more than onequarter (27%) involved striking essential features (e.g., bridges, fences, embankments or walls), and 20% involved hitting introduced features (e.g., poles, guideposts or safety rails).

38

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

A more recent study in the same State found a similar proportion of single-vehicle run-offroad crashes (Corben et al., 1997). Of the 5,184 serious casualty crashes in Victoria during 1994, 1,175 involved a single vehicle striking a fixed roadside object. These crashes were found to be severe in nature, accounting for 23% of all serious casualties on Victorias roads, with 1,454 people either s eriously injured or killed in 1994 alone. The most common objects struck were trees, poles and embankments. 2.4.4.1 Contributing factors As with most crashes, driver, vehicle and road factors have been identified as contributing to crash and injury risk. Driver factors include inattention and fatigue, inexperience, alcohol/drug impairment and speeding (Adamo & Wray, 1992; Bahar, 1987; Ogden, 1997). Ogden (1997) considered driver fatigue to be a contributing factor in 33% of all rural crashes, and in 25% of crashes, excessive or inappropriate speed contributed to crashes. Adamo and Wray (1992) observed that drowsy drivers tend to drift to the right more than to the left (note that in this study driving occurs on the right-hand-side of the road) and typical vehicle departure angles for drowsy drivers who have drifted off the roadway are around three degrees. Moreover, those studies that have examined the time of day of single-vehicle run-off-road crashes on rural roads show that the majority occur in the evenings and at night, suggesting that fatigue and light conditions are factors (Mintsis & Pitsiava-Latinopoulou, 1990; OECD, 1975; Sanderson & Fildes, 1984). As indicated in Section 1.3.1, speeding and drink-driving are major contributing factors to single-vehicle crashes. Sanderson and Fildes (1984) reported that, in non-fatal, run-offroad crashes, the majority of drivers claimed that they were travelling within the speed limit. On the other hand, in fatal run-off-road crashes, 26% of drivers appeared to have exceeded the speed limit by more than 15 km/h. They therefore suggested speed could significantly contribute to run-off-road crashes. Moreover, as reported in Section 2.3.1, vehicle factors play a role in crash severity because of the limited ability of vehicles to protect their occupants from serious injury, given high impact speeds with fixed objects. The prevalence and unforgiving nature of roadside hazards play a major role in the severity of run-off-road crashes. The presence of fixed roadside objects, poor shoulder design, narrow lanes and poor pavement are some of the features of roadsides that contribute to crash and injury risk (Ogden, 1997). Cairney (1998) and Cairney and McGann (1997) added that the following features contribute to crash risk: traffic flow; road cross-section and shoulder conditions such as change in cross-section; lane widths; width and surface quality of shoulder; presence and width of overtaking lanes; posted advisory speeds; presence of barrier lines presence of structures narrowing the carriageway; width and nature of clear zone beyond shoulder; presence of objects within clear zone; configuration; and quality of delineation. 2.4.4.1.1 Roadside objects

The presence of hazardous objects on the roadside and the ability of drivers of errant vehicles to safely negotiate roadsides are major factors contributing to the severity of single-vehicle crashes in rural areas, with many reports of fatalities and severe injury crashes associated with fixed roadside objects (Mintsis & Pitsiava-Latinopoulou, 1990; OECD, 1975; Schnerring, 1995). In rural areas particularly, there are often high numbers of hazards positioned close to the edge of the carriageway and many studies have reported

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

39

the dangers associated with the positioning of fixed objects close to the edge of the carriageway. Bahar (1987) reported that 47% of single-vehicle crashes in South Africa involved vehicles striking hazards in either the median (43%) or roadside (57%), with the majority colliding with objects located within four metres of the roadway. Bahar also noted that distances travelled by errant vehicles after leaving the roadway ranged from 10 m up to 300 m and he concluded that many roadside objects are too close to allow drivers to control or stop their vehicle. Furthermore, Huelke and Gikas (1967, cited in Adamo & Wray, 1992) found that, in 80% of crashes involving vehicles running off the road, the vehicle collided with an object within 27 feet (8 m) of the edge of the roadway. Mintsis and Pitsiava-Latinopoulou (1990) investigated the contributing factors in 477 crashes with fixed roadside obstacles on the Greek rural network within a five-year study period from 1979 to 1983. They reported that these crashes constituted approximately eight percent of the total number of crashes on the rural network and resulted in 11% of the total fatalities and eight percent of the injuries over the whole study period. The authors concluded that trees, cuts and fills near unprotected ditches constitute the most dangerous roadside objects. Some roadside hazards are more threatening to vehicle occupants than others because of their positioning and structure, and it seems that collisions with trees and poles make up a major part of the run-off-road crash and injury problem in rural areas. Indeed, McLean (1996) noted that, because of the narrowness of many off-road structures, collisions with these objects usually result in severe vehicle crush and localised penetration of the structure into the vehicles passenger compartment. He concluded that the intrusion and the contact with these structures is highly likely to cause death to vehicle occupants. The US Federal Highway Administration (FHWA, 1992) provided ratings of the severity of injuries caused as a result of impacts with fixed roadside objects. The ratings indicate that the objects causing the most severe injuries were wooden utility poles, trees, bridge or overpass entrances, field approaches (i.e., ditches created by driveways), culverts and embankments. The least severe injuries arose from collisions with guard-rails, small signposts and fences. In discussing the role of man-made structures in fatal run-off-road crashes, Rattenbury and Gloyns (1992) highlighted property line fences on the roadside as particularly aggressive in impacts with passenger cars, as their (horizontal) rails and (vertical) posts can penetrate the passenger compartments. Such fences account for around 1.5% of all fatal crashes involving car occupants. 2.4.4.1.2 Shoulder design

In addition to the presence of fixed objects close to carriageways, there is some suggestion that the design and condition of shoulders can contribute to increased crash risk on rural roads. Intuitively, a wide shoulder will provide more space and levels of friction in which to maintain or regain control of a vehicle. McLean (1996) reviewed the relationships between crash frequencies on two-lane rural roads and cross-section elements and roadside design in Australia, and found that lane and shoulder width were key contributing factors to crash risk, noting an increase in crash rates when narrow shoulders were present. In his study of single-vehicle crashes on rural roads in South Africa, Bahar (1987) found that 53% of the crashes involved vehicles over-turning without striking a roadside object and suggested that the main contributing factor in these crashes was the lack of drivable
40 MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

roadsides and medians. Interestingly, Ivan et al. (1999) noted the benefits of wide shoulders for run-off-road crashes, but also found a positive correlation on right shoulder width in multi-vehicle crashes and suggested that with wide shoulders, drivers are more likely to drive too fast and collide with other vehicles exiting or entering the carriageway. Embankments have also been identified as contributing factors to increased crash risk on rural roads. Moore (1998) reported that in 1992, approximately 21% of all fatal crashes and 14% of all injury crashes on rural two-lane highways in California involved vehicles driving over the embankment, with the vast majority overturning, and approximately 48% of fatal crashes and 36% of injury crashes involving collisions with trees or sinking in water. In most of these crashes, the vehicle overturned. He also reported that these crashes generally occurred at random locations, with only 11% occurring at concentrated locations (i.e., less than 0.5 mile or 0.8 km apart). Interestingly, nearly half of the over-embankment crashes occurred on the opposite side of the roadway on which the vehicle was travelling, many of them involving over-correction by the driver. Moore noted that drunk drivers were over-involved in over-embankment crashes, making up 19% of total crashes, 43% of fatal crashes and 23% of injury crashes, noting that these proportions are substantially higher than the proportions for all highway crashes (7%, 25% and 10%, respectively). For all types of over-embankment crashes the percentage of total drink driving crashes involving fatalities was approximately 2.6-4.6 times greater than the percentage of total non-drink driving crashes involving fatalities. 2.4.5 Intersections

Intersections are an important part of the roadway system. At-grade intersections are generally found on all classes of rural road other than freeways and motorways and include four-way cross roads, T-intersections and private accesses. At-grade intersections are potentially the most dangerous parts of the road network because they present a driver with many potential points of conflict with other road users. Many intersections on low-volume, single-carriageway roads and roads approaching towns are either uncontrolled or controlled by stop or give-way signs. Moreover, in rural settings, these types of intersection are even more dangerous because these conflicts often occur at high speeds. Grade-separated intersections include overpasses and entrance and exit ramps, and are generally provided on highways, freeways and motorways with high design speeds and carrying high traffic volumes. This discussion will focus primarily on the problems of at-grade intersections as most rural intersection crashes occur at these types of intersections. The issues surrounding problems of grade-separated intersections form only a small part of this discussion as, generally, they are much safer than at-grade intersections. At-grade intersections include non-signalised and signalised intersections and roundabouts. Drivers negotiating at-grade, signalised or non-signalised, intersections must often travel across the path of other vehicles, particularly when making a turn across oncoming traffic (i.e., a left-turn in the USA and most European countries or a right-turn in England, Australia and New Zealand). While the aims of intersection design are to improve traffic flow, reduce the number of points of conflict and reduce the likelihood of crashes, poor design of intersections can mean an increase in the risk of collision and/or the risk of severe injury for all road users.

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

41

The safety performance of an intersection is strongly influenced by decisions about spacing between intersections, type of control, hierarchy, size and geometry. It is important when designing intersections to ensure that drivers: Are aware of the presence of an intersection and clearly perceive the actual conflict areas, Are made aware of other vehicles within and approaching the intersection, Have confidence in the course required to negotiate the intersection correctly and safely, including which driver, rider or pedestrian has right of way in the event of a potential conflict, Encounter uniformity in the application of traffic devices and procedures at intersections, and Are provided with adequate reaction and decision time to move safely through the intersection.

The estimated proportion of rural crashes occurring at intersections ranges from around 20% in some countries with low-density populations to over 40% in more densely populated countries. For instance, Barker, Farmer and Taylor (1999) noted that, on British roads in 1994 and 1995, 22% of fatalities occurred within 20 m of a rural intersection. In New Zealand, Wilson and Dunn (1994) estimated that over 33% of injury crashes in 1992 occurred at intersections on the rural road network. Likewise, the Helsinki Roads and Waterways Administration (cited in Kulmala, 1991) found that 42% of all injury crashes on the Finnish rural highway system in 1986 occurred at intersections. In contrast, Ryan (1993, cited in Lydon, 1997) found a smaller proportion of rural crashes in Western Australia occurring at intersections compared to crashes on other road lengths in rural (24%) and remote areas (19%). However, this finding may be confounded by the size and remoteness of rural areas in Western Australia compared with other more densely populated and smaller countries such as in Europe. Nevertheless, crashes a t intersections in this State of Australia still constitute a substantial proportion of all rural crashes. A more recent study of rural crashes in another, more densely populated and smaller State of Australia (Victoria), found that over 30% of rural crashes resulting in death or serious injury (Road Safety Committee, Parliament of Victoria, 2002) occurred at intersections. Not surprisingly, this study also noted that there was a higher proportion of crashes at intersections on local roads and within country towns compared with intersections on arterial roads and major highways. In his study of crash risk by intersection type in England, Hughes (1994) found that over half of the crashes at rural intersections occurred at T or staggered layouts, with 60% occurring on strategic roads, 53% on secondary single carriageway roads and 36% on strategic dual carriageways. Crashes at rural intersections have been attributed to many factors including driver characteristics (e.g., the ability to negotiate the intersection and perception of possible risk, speed behaviour, fatigue, and alcohol use), vehicle factors, and road factors, with road design being identified as the major contributing factor to crash and injury risk (Hughes & Amis, 1996; Stackhouse & Cassidy, 1996; VicRoads, 1998;). The following geometric features have been shown to contribute to increased crash and injury risk on rural roads. The problems associated with these factors are discussed in the following sections.

42

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

Poor or inappropriate intersection control and signage, Poor or restricted sight distance, Poor or inappropriate delineation or alignment, Inadequate street lighting, Poor design of freeway interchange ramps, and Inappropriate design speed.

2.4.5.1 Poor or inappropriate intersection control and signage The type of at-grade intersection control seems to greatly influence crash and injury risk, primarily because of the requirement for drivers to make gap selection decisions when turning across or into traffic at intersections that do not provide turning assistance. Intersection control ranges from signs (give-way, stop) through to signals (providing either no control, partial control or full control of turns across traffic for left-turns in the USA and most European countries and right-turns in England, Australia and New Zealand) to roundabouts. In general, intersections on low-volume single-carriageway roads are controlled by stop or give-way signs. At such intersections, priority is invariably given to the major road by the installation of stop or give-way signs and provision of holding lines on the minor road approaches. However, because of the relatively high speed of cross traffic, crashes caused by failure to stop or give way can be severe (Russell, Stokes & Rys, 1999). Indeed, there is evidence to suggest that these are the most dangerous types of intersections. A study in Iowa, USA, (Iowa Department of Transportation, 1991) showed that during 1986-1990, running stop signs was a contributing circumstance in 297 crashes in which 352 people were killed. Stokes, Rys, Russell, Robinson and Budke (2000) noted that crashes that are caused by drivers failing to stop or give way to traffic on the priority road are becoming increasingly frequent at some rural intersections on the Kansas State highway system. Although they acknowledged that some of the crashes investigated may have been caused by drivers entering the intersection but failing to accelerate fast enough to avoid being hit, they argued that many of the crashes were probably due to signage inadequacies. For instance, crashes may have occurred because of inappropriate placement of signs, resulting in drivers experiencing difficulty seeing and responding to the signs, or signs being inappropriate for the conditions of the intersection. Stokes et al. (2000) concluded that, if stop or give-way signs are inappropriate for the intersection, then traffic signals should be considered. They added that traffic signals should also be considered in other situations. For instance, when there are excessive delays at the intersection, when there are problems caused by turning traffic, an overrepresentation of angle collisions, and an over-representation of pedestrian crashes. Fildes, Corben, Morris, Oxley, Pronk, Brown & Fitzharris (2000) found that a high proportion of crashes in both rural and urban areas involving older drivers occurred at intersections, most of which were controlled by stop or give-way signs (65%) with the remaining 35% intersections controlled by traffic signals. The majority of these crashes

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

43

(76%) occurred when the older driver failed to select a safe gap in conflicting traffic when turning across or crossing traffic at intersections. This basic problem manifests itself mainly at intersections controlled by stop or give-way signs, or at signalised intersections that provided either no control or partial-control of right-turn manoeuvres. Of those crashes that occurred at signalised intersections, Fildes et al. found that a lack of fully-controlled right-turn signals (where making a turn across oncoming traffic is controlled by green, amber and red signals) contributed to increased crash risk. Crashes could be a result of obscured traffic lights or inappropriate phases of lights that may cause people to intentionally or unintentionally run red lights. Greibe & Nielsen (1996) also found that the higher the speed limit, the higher the crash risk and signalised intersections on 80 km/h roads have a significantly higher risk than all other intersections. If the crash risk is based on the number of people fatally and seriously injured, then the relationship between speed limit and crash risk is even stronger, especially for intersections in rural areas. Fildes et al., like Stokes et al. (2000), concluded that the use of fully-controlled right-turns should be considered at busy intersections where there is high potential for turning conflicts. Roundabouts are another form of intersection control that is becoming increasingly popular in both urban and rural settings. Roundabouts are generally installed at intersections with a high crash and injury history and are considered a safety feature. For this reason, roundabouts are discussed in more detail in Section 3.7.2.3. Notwithstanding the many safety benefits of roundabouts, there is some evidence that motorcyclist, bicyclists and pedestrians experience some problems at roundabouts (Hughes, 1994). A study in the United Kingdom found that crash involvement rates of cyclists and motorcyclists, in reported crashes, was 14 to 16 times higher than of car occupants (Allot & Lomax, 1991 cited in Robinson, 1998). Robinson (1998) also found that 6 percent of those injured at cross intersections in NSW were cyclists compared with 18 percent at roundabouts. This over-representation of motorcyclists and cyclists in crashes at roundabouts in NSW, especially in urban areas, is possibly due to a higher volume of cyclists in this area and excessive speeds, especially on the approach. Due to the difference in travel speeds between urban and rural areas there is a significant difference in the level of safety gains in those areas. Types of crashes on roundabouts, due to excessive speeds while entering or travelling on the roundabout, include single-vehicle crashes where vehicles fail to stay on course running off the road into a parked car or object Brewer et al. (2001) also found that while roundabouts are effective in reducing the severity of crashes, they are a less effective form of intersection when the number of entry legs exceeds four. This is mainly because of the size of the intersection and the higher circulating speeds that can be achieved. Brewer et al. (2001) and Robinson (1998) also noted that while roundabouts can reduce the severity of crashes, the reduction in the number of overall crashes is sometimes not as large. 2.4.5.2 Poor or restricted sight distance The provision of adequate sight distance on approaches to and entering intersections is fundamental to safe intersection design. Sight distance is defined as the distance at which an attentive driver can see a specified object ahead of them, given clear, well-lit conditions, good visual acuity and the object centrally located in their field of vision (Stackhouse &

44

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

Cassidy, 1996). Unrestricted sight distance assists a drivers ability to judge the s peed and distance of oncoming vehicles and therefore make appropriate decisions to negotiate the intersection. In contrast, a sight distance problem occurs when the sight distance available to drivers to manoeuvre through the intersection (whether turning across traffic or crossing priority lanes) is less than the sight distance required to complete these manoeuvres. A restricted sight distance can hamper a drivers ability to see traffic movements at the intersection and on its approaches, and perceive gaps in the traffic, therefore increasing the potential for crash and injury risk. This is especially a problem on high-speed highways where lengthy sight distances are necessary (Bonneson, McCoy & Truby, 1993). There is some evidence to suggest that intersections with short sight distances generally have higher crash rates. Current road design guidelines in the USA, Australia and other countries use values of driver reaction times, approach speed and driver eye and object height to set intersection sight distance. Sight distance values generally range from 1.5 to 3.5 seconds, depending on the road environment, with higher values for normal un-alerted driving situations on rural roads. However, it has been argued that sight distance values of 2.5s may not be sufficient in all situations and may vary up to 5.0s depending on driver characteristics, the complexity of the intersection and the function of the road (Fildes et al., in press; Neuman, 1989). Fildes et al. (2000) found that sight distances of less than 2.5s in both urban and rural environments contributed to up to 23% of crashes involving older drivers. They suggested that, at the intersections inspected, older (and other aged drivers) experienced difficulty selecting gaps in the traffic and that limited sight distance contributed substantially to this difficulty. Sight distances were inadequate due to roadside obstructions, vegetation, other vehicles, curves, and dips and crests, resulting in difficulty for drivers in detecting potential conflicts and acting appropriately. 2.4.5.3 Poor or inappropriate delineation or alignment Channelisation, alignment and delineation are treatments that can enhance the safety of atgrade intersections. However, there is some concern regarding the design and appropriateness of some of these features, with some suggestion that poor design of these features can contribute to increased crash and injury risk. For instance, the reduction of super-elevation at intersections on curves may cause stability problems, especially for heavy vehicles; poor design of turning lanes, traffic islands and medians may create operational problems and increase the incidence of certain types of crashes; and skewed intersections may result in misperceptions of safe gaps. 2.4.5.3.1 Super-elevation

Generally, when intersections are located on a curved section of roadway, the priority road super-elevation height is reduced to facilitate vehicles entering the priority carriageway from the intersecting road. Leonard, Bilse and Recker (1994) argued that this practice can generate increased crash rates for vehicles travelling on the priority road, particularly trucks and heavy vehicles, as vehicles travelling along the curved section of the major roadway experience a greater lateral acceleration due to the reduced counterbalance of the super-elevation. 2.4.5.3.2 Turning lanes and traffic islands

There are a number of measures to facilitate traffic turning from a minor road onto a priority road. Turning lanes (or slip lanes) are provided to minimise delays and ease the

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

45

turn movement (right-turn in the USA and most European countries and a left-turn in countries such as England and Australasia) to a merging manoeuvre at busy intersections. In addition, some intersections provide facilities such as traffic islands for turning traffic from a minor road onto a priority road, that has to cross traffic on the major road (a leftturn in the USA and most European countries, and a right-turn in countries such as England and Australasia). These facilities are widely used in urban areas where traffic density is high. However, on rural roads, particularly low-volume two-lane roads, many intersections do not have these facilities, or if they are present, may not be well designed or maintained. Indeed, in their examination of crashes in the county of Cambridgeshire in England, Hughes and Amis (1996) found that ghost island facilities were associated with an increase of crashes where the vehicle entering the major roadway is struck on the off-side. They concluded that the problem is due to the extra carriageway width taken up by the turning lane of the right-turn facility (left-turn in the USA and most European countries), the increased distance between the through traffic on the major roadway and the turning vehicle, and problems seeing the traffic (and therefore judging speed and distance) by drivers entering the major road. 2.4.5.3.3 Median width

Intersections between multi-lane divided highways and undivided minor roads are complex and present many problems for drivers intending entering the intersection from a minor road. Medians are usually installed as a safety measure on busy and high-speed rural roads and highways, however, it seems that median width can affect crash risk. Harwood, Pietrucha, Wooldridge, Brydai and Fitzpatrick (1995) investigated the effect of median width on driver behaviour at intersections on divided highways and argued that crash risk increased as median width decreased. They found undesirable driver behaviour at intersections through narrow medians. Drivers tended to line their vehicles side by side rather than single file, stop their vehicles in the median at an angle to the major road, and encroach on through-lanes of major roads. Wide medians, too, have been shown to increase crash risk. There is evidence to suggest that wrong-way movements when wide medians are provided are common. This is generally because wide medians can increase confusion for drivers entering the intersection from the minor road. Vaswani (1977) found that approximately half of the incidents that involved wrong-way driving on multi-lane divided highways occurred at intersections with highway exits and secondary roads. Scifres and Loutzenheister (1975) found that wrongway movements most often occurred under conditions of low traffic volume, low visibility, and low lane-use density. They also noted that almost 70% of the wrong-way drivers were drunk, older or fatigued. More recently, McKnight and Urpuijo (1993) found evidence of older driver difficulties likely to result in wrong-way movements in these situations. Crowley and Seguin (1986) reported that there are significantly more incidents of wrongway driving than there are crashes. They conducted a comprehensive survey of US policies and practices for signing intersections to inform drivers of travel direction and to prevent wrong-way movements to address these problems. They found considerable variability in the location, placement and types of signs used to prevent wrong-way movements from occurring in the 48 States investigated and reported that the greatest variability in practice occurred in locations where a median divider existed. They pointed out that median width is a key factor in the incidence of wrong-way movements, stating that when medians are

46

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

extremely narrow, there appears to be little confusion that the intersecting roadway is twoway, but when wide medians are provided there is more confusion. 2.4.5.3.4 Intersection angle

There is broad agreement that 90 intersections are the preferred design, and that skewed intersections can cause problems for drivers. These include: a longer distance to traverse while crossing the intersecting roadway; difficulties for drivers with physical limitations in turning head or upper body for an adequate line of sight down an acute angle approach; drivers making turns around an acute-angle radius may encroach on lanes intended for oncoming traffic; skewed intersections require a larger intersection area to accommodate paths of larger vehicles through the acute angle; through-roadway drivers making turns across an obtuse angle may attempt to maintain a higher than normal turning speed and cut across the oncoming traffic lane on the intersecting road; and, the line of sight of drivers with an acute-angle approach might be obstructed by the body of large vehicles (Gattis & Low, 1998; Parsonson, Isler & Hansson, 1996). Skewed intersections have been identified as a particular problem for older drivers with impaired head and neck movement (Hunter-Zaworski, 1990; Staplin et al., 1997). In their examination of older driver crashes in Australasia, Fildes et al. (2000) estimated that highly skewed intersections could have contributed to around 2 percent of the crashes investigated. In their survey of older drivers in the USA, Staplin et al. (1997) found that, apart from vision problems, difficulty with head turning was placed first among all concerns of older drivers. Older drivers in this study preferred 90 intersections, noting that it is easier to see traffic approaching from both left and right because the required head turn is not extreme in either direction. 2.4.5.4 Poor or inadequate lighting Adequate lighting at and around intersections is essential to increase visibility of the roadway and its immediate environment to enable drivers to manoeuvre more safely and efficiently through an intersection thereby enhancing vehicle and pedestrian flows. There are many intersections, particularly minor intersections on low-volume single-carriageway roads, where inadequate or no lighting is provided. Indeed, Stackhouse and Cassidy (1996) noted that there were more night-time crashes at rural intersections compared to day-time crashes and that they tended to be more severe. Likewise, the Commission Internationale de lEclairage (1990) reported that road crashes at night are disproportionately higher in number and severity compared with crashes during the day-time. Data from 13 OECD countries showed that the proportion of fatal night-time crashes ranged between 25 and 59% (average value 48.5%). This increased risk of night-time crash may also be attributed to high blood alcohol concentrations (BAC), high speeds and other risk-taking factors. Nevertheless, poor lighting appears to play some a role in crash risk. 2.4.5.5 Poor design of freeway interchange ramps Freeways are becoming more common within and between major cities, and are constructed to accommodate increasing traffic demand. As indicated in Section 2.4.2.1, freeways are often characterised by a higher safety level when compared with other road types in rural locations. Grade-separated intersections (i.e., interchanges) may be provided when one or more of the intersecting roads are freeways or expressways. At these locations, access to freeway carriageways is limited to selected intersecting roads at which

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

47

an interchange is constructed with specially designed ramps to carry turning traffic and allow traffic to enter and leave the freeway with minimum interference to through traffic. While it is argued that this provision, in conjunction with other factors such as control of access and the generally high standard of freeways, largely contributes to the substantially better safety and operational performance of these roads (Austroads, 1988), there is some concern that geometric features of entry and exit ramps at grade-separated intersections can result in significant safety and operational problems (Harkey, Huang & Zegeer, 1996; Knoblauch, Nitzburg, Reinfurt, Council, Zegeer & Popkin, 1995; Kolle, 2000; Malfetti & Winter, 1987). Exit and e ntry areas of freeways are transitional areas that may require a substantial change in tracking. A driver (especially in an unfamiliar location) must process a large amount of information during a short period of time and at high speeds, while maintaining or modifying his/her position within the traffic stream. When weaving actions are required, the information processing task demands for freeway interchange manoeuvres are further magnified. Merging manoeuvres should take place at the speed of the freeway traffic stream, therefore, entry acceleration ramps need to be long enough for drivers to accelerate to the speed of the freeway traffic. There is some evidence, however, that crash rates are higher on shorter entry ramps than on longer ramps (Cirillo, 1970; Lundy, 1967; Reilly, Pfefer, Michaels, Polus & Schoen, 1989). The design of exit ramps can also increase crash risk. Taylor and McGee (1973) investigated driver behaviour at exit gore areas of interchanges to determine the characteristics of driving manoeuvres. In their interviews with drivers who made erratic manoeuvres, many of these respondents indicated that they were not adequately prepared for the exit. These drivers indicated that the signs lacked needed information or that the information was misleading. Interviews with drivers who made erratic movements and continued through showed that approximately half of these drivers had difficulty identifying their desired direction. Approximately 30% stated the signing was not clear, 21% responded they could not clearly distinguish the location of the exit ramp, and 34% thought the road markings were inadequate. Finally, the design of cross-road approaches to an exit ramp of a freeway can potentially lead to wrong-way movements, i.e., vehicles entering an exit ramp at the cross-road terminal. Although wrong-way crashes on highways and freeways account for only a small proportion of crashes, they tend to be severe crashes and likely to result in death or severe injury. Tamburri and Theobald (1965) reported that out of 100 wrong-way crashes, 63 result in an injury or fatality, compared to 44 out of 100 for all freeway crashes. More recently, Copelan (1989) found that the number of fatal wrong-way crashes in California averaged about 35 per year in the 1970s and 1980s. Wrong-way crashes accounted for approximately 2.9% of the fatal, 0.3% of the injury and 0.1% of the property damage crashes on Californian freeways. It seems that some ramp designs are more problematic than others. Copelan (1989) argued that the full diamond interchange design can pose problems for drivers, because an off-ramp can be mistaken for a frontage road located parallel to the ramp. Parsonson and Marks (1979) also determined that several ramp designs were particularly susceptible to wrong-way movements. These include the halfdiamond, partial cloverleaf, trumpet and buttonhook ramp interchanges. Partial cloverleaf ramps are susceptible to wrong-way movements because the entrance and exit ramps are in close proximity, while half-diamond designs are susceptible because they are an incomplete interchange and drivers may make unintentional wrong-way entries.

48

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

2.4.5.6 Speeds As noted previously, at-grade intersections are inherently dangerous parts of the road system, but are even more dangerous on rural roads because conflicts often occur at high speeds. Griebe and Nielsen (1996) examined the effect of speed limit on crash risk at signalised intersections. They found that the higher the permitted speed along the priority road, the higher the crash and injury risk is at intersections, noting that many of the crashes were rear-end and turning across traffic crashes. Fildes et al. (2000), too, noted that high speed contributed to the frequency and severity of intersection crashes involving older drivers. Hughes and Amis (1996) also noted that many of the crashes investigated involved skidding on a dry road surface, suggesting that vehicle speed may have been a contributing factor. 2.4.5.7 Rail-road crossings In contrast with the other road safety issues considered in this review, at-grade rail crossings (or highway-rail intersections [HRIs]) are unique in that road-based transport and trains generally only interact at set locations. Accordingly, countermeasures need only address these locations or the behaviour of individuals at these locations. The literature describes two types of HRI active crossings and passive crossings. The active-passive distinction is essentially a question of whether the crossing has a device activated by the approach of a train (an active crossing) or not (a passive crossing) (Russell, 2002). Active rail crossing measures include booms/gates/barriers (the terminology depends on the jurisdiction) that block either half the road (i.e. just the approach lane) o n each side of the crossing, or they block the whole road. The other principal active countermeasure relates to flashing lights and bells that are set off by the approaching train. Assuming that the protective infrastructure is functioning properly, the primary safety issue at actively protected crossings relate to drivers driving around barriers and ignoring the bells and flashing lights if they believe that they have sufficient time to cross the tracks before the train arrives. Passive countermeasures are simply signs. An approaching driver need not see the looming train in order to avoid a collision at an active crossing; they need only take heed of the flashing lights or booms. For their own safety, motorists approaching a passive crossing must search for the train themselves before deciding whether to proceed across the railway line. Generally, active HRIs are more likely to be found in urban areas, at busy intersections (i.e., a large number of trains and road vehicles use the intersection daily) or on very fast train routes. Passive crossings are more predominant in rural areas where traffic volumes (for both trains and road vehicles) are light. For example, in the US active HRIs see on average 13.7 trains and 4,003 vehicles per day, compared with 6.2 t rains and 849 vehicles per day for passive crossings (Weiland & Woll, 2002). Passive crossings are the focus of this section of the review. Passive HRIs are a significant safety issue. In excess of 50% of any particular jurisdiction's HRIs are likely to be passive crossings. For example: Sweden 58% of all crossings are passive (3,800 crossings) (Mornell, 1997) New Zealand 60% (1,136 crossings) (Allan, 1992) Finland 82% (4,219 crossings) (Pajunen, 2002)

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

49

US 60% (around 94,800 crossings) (Railroad Safety Statistics, 1999, cited in Wigglesworth, 2002) Australia has approximately 6,000 passive crossings (Wigglesworth, 2002)

Despite the low traffic volumes at passive crossings, the sheer number of them means that they contribute significantly to crashes: In the US, over 50% of crashes at HRIs occur at passive crossings (Hall, 1999), and 60% of these are fatal (NTSB, 1998). Around 400 deaths annually result from passive crossing crashes in the US (Russell, 2002) In Finland 74% of crashes at HRIs occur at passive crossings (Pajunen, 2002) In Australia 44% of rail crossing fatalities occur at passive HRIs, and a further 41% occur at crossings with some active device other than boom gates (ATSB, 2001, cited in Wigglesworth, 2002).

Rail crossings are a unique road safety issue for a number of reasons. Almost all motorists would be aware that a train always has the right of way, and this is reinforced at active crossings where boom gates are lowered to halt road traffic while trains pass through. Signs at passive crossings also generally indicate that motorists (and pedestrians, bicyclists, etc.) should only proceed across the tracks if there are no trains within sight. Additionally, all motorists can be expected to realise that in any crash the train will always suffer less damage and that the only avoidance manoeuvre that a train driver can take is to brake a train can not, of course, swerve (and trains do not stop smartly). Perhaps the most critical element in regards to rural HRIs is the infrequency of trains. A local resident may cross a rail line (whether as a motorist, cyclist or pedestrian) many times (indeed most times) without seeing a train. This reinforces an expectancy in the motorist of the safety of crossing the line and it is likely that the caution exercised will decrease until only the roughness of the physical crossing dictates the crossing speed. Without an active warning device a complacent motorist is unlikely to make more than a cursory glance in each direction as they approach the crossing, particularly as they become increasingly familiar with the particular crossing. Unless sight lines and distances are good (i.e., the rail line and the road cross at right angles, there is no foliage to block the drivers vision, there are no bends in the approaching road or the intersecting rail track, etc) and there are no distractions, the motorist is unlikely to see an approaching train until it is too late. The complacency of motorists is reflected in the fact that a majority of crashes occur close to the drivers home (Hall, 1999; Johnstone, 2002; Pajunen, 2002; Russell, 2002). This suggests that the driver is very often quite familiar with the crossing. They simply do not expect a train to be coming. 2.4.6 Other factors

There are a number of other road factors contributing to crash risk on rural roads. These include: vertical curves, bridges and culverts and construction zones.

50

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

2.4.6.1 Vertical curves The literature identified vertical curves as a risk factor for crashes on rural roads. Crests on vertical curves on rural highways can severely restrict stopping sight distance and there are some reports of increased crash rates at these locations. The most recent and extensive study of crashes on vertical curves with limited stopping sight distances was conducted by Fitzpatrick, Fambro and Stoddard (2000). They concluded that the shorter the stopping sight distance, the greater the crash risk, particularly when a major hazard (such as an intersection or sharp horizontal curve) exists beyond the crest. Fitzpatrick et al. (2000) examined the effect of limited stopping sight distances on crash characteristics in three rural road conditions where vertical curves existed (multi-lane, twolane with shoulders and two-lane without shoulders). Not surprisingly, they found that the multi-lane highway sites had a lower average crash rate (0.53 crashes per million vehicle kilometres) than the two-lane sites, with the two-lane highway without shoulder sites experiencing the highest average crash rate (1.04 crashes per million vehicle kilometres). In addition, using subjective ratings, they found that sites with a lower design speed (i.e., those several closely spaced curves with shorter stopping sight distances) had higher crash rates than those with higher design speeds. However, only small proportions of crashes on limited sight distance roadways, in which a limited stopping sight distance was noted as a possible contributing factor, were found (3% of crashes on two-lane without shoulder sites, 5% of crashes on two-lane with shoulder sited and none of the crashes on multi-lane sites). The factors that contributed to these crashes were also reviewed (Fitzpatrick et al., 2000). On two-lane roads, the most frequent crash type was striking an object in the roadway (24% on roads without shoulder and 40% on roads with shoulder) and the second most frequent crash type was travelling too fast for the conditions (17% on both with and without shoulders). On multi-lane roads, the most frequent crash type was travelling too fast for the conditions (44%), with only 8 percent of crashes involving collisions with an object in the roadway. Other crash types included icy pavement (20% on multi-lane roads and 10% for the two-lane roads), and driver impairment (20% on multi-lane roads and 8% on the two-lane roads). In conclusion, Fitzpatrick and his colleagues concluded that limited sight distance was not the key problem. Rather, it seemed there was a major problem of vehicles stopping in the roadway to make a turn either into a driveway, access or turning at an intersection. 2.4.6.2 Bridges and Culverts Crashes at bridges, particularly narrow bridges, and culverts appear to pose a significant safety problem on rural roads (Mak & Sicking, 1994; Evans, 1997; Ogden & Howie, 1989a; 1989b). For example, Mak and Sicking (1994) estimated that approximately 43% of bridge crashes in North America occurred in rural areas, and that 82% of cases where a vehicle travelled through or over a bridge rail occurred on a rural highway. Mak and Sicking argued that, since bridges occur with less frequency in rural areas, and tend to carry lower traffic volumes than bridges in urban areas, this represents a significant cause for concern. In addition, the crashes that occur at bridges and culverts tend to be more severe than other crashes (Bowman & Brinkman, 1988; Hilton, 1973; Ogden & Howie, 1989a). Hilton (1973) argued that rural bridge-related crashes are roughly twice as severe as crashes generally. Likewise, Agent and Deen (1976) attributed a severity index of 3.24 to bridge crashes compared with 2.75 for all crashes. These findings are supported by more recent work (Abed-Al-Rahim & Johnston, 1993; Bowman & Brinkman, 1988).

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

51

The significance of the role of bridges in rural crashes was illustrated in the early eighties by Hollingworth (1983) who demonstrated that the number of narrow bridges present was the best predictor of crash rate on a given section of the road. Bowman and Brinkman (1988) propose that one reason for the high crash rate at rural bridges and culverts is that these structures are fairly expensive components of the road network, and often when rural roads are upgraded, the bridges associated with them are not. Indeed, Evans (1997) suggests that a major problem on North American and Australian rural highways is that many were built before new design standards were introduced in 1964 and 1970 respectively. Consequently, bridges and culverts built before upgraded design standards were introduced have a significantly higher crash rate than newer structures. Like for crashes on other parts of the rural road system, a number of contributing factors to crash and injury risk associated with the design and operation of bridges and culverts have been identified (Evans, 1997; Ogden & Howie, 1989a). First, it a ppears that most crashes involving vehicles impacting bridge or culvert features are single-vehicle crashes in which the vehicle impacts the bridge or its associated structures, and, in the case of multi-vehicle crashes, the presence of a bridge or culvert is not directly involved, suggesting that the problem is similar to that of an unforgiving roadside (Ogden & Howie, 1989a; Evans, 1997). Ogden and Howie suggested that the relatively high rate of crashes at bridges and culverts is more than coincidental. The majority of crashes at these structures fall into a limited number of categories suggesting that, rather than random events, crashes at bridges and culverts are triggered by common characteristics of these sites. Secondly, bridge crashes in rural areas tend to involve actual bridge structures more than in urban areas. For example, Agent and Deen (1976) reported that collisions involving bridge entrance posts and wing walls are more often associated with fatalities than any other bridge feature. On the other hand, the lowest severity occurs when a vehicle collides with bridge guard-rails (see also Zegeer & Council, 1995). Ogden and Howie (1989a) also reported a higher proportion of crashes on bridges in rural areas than urban areas involved the bridge or bridge railing being struck (39% compared to 27%). On the other hand, urban bridge crashes tend to be the result of failed attempts at overtaking. Several authors have attempted to determine the geometric features of bridges that increase crash and injury risk (Kloeden, McLean, Baldock & Cockington, 1999). The most significant contributing factor to bridge- and culvert-related crashes is the width of the bridge and this can refer to absolute width, lane width or relative width. Lane widths on bridges are often reduced because railings and piers associated with the structure are positioned within the shoulder space on either side of the pavement and there are reports of significantly more crashes on narrow bridges than on wide bridges (Evans, 1997; Hollingworth, 1983; Ogden & Howie, 1989a). One of the main concerns associated with reduced width or absent shoulders on bridges is that the bridge railings are very close fixed roadside objects, and in the event of a driver straying off the road or losing control of their vehicle, there is very little room allowing for control to be regained (Ganhi et al., 1984). Consequently, on narrow bridges there tends to be an increased risk of striking a fixed object (Ivey et al., 1979). Bridge sites are particularly hazardous when the roadway on the bridge is more than 20% narrower than the approach roadway (Ivey et al., 1979; Turner, 1984; Brown & Foster, 1966). For example, Brown and Foster (1966) found that 70% of bridge crashes featured a bridge-to-approach ratio of less than 0.79, suggesting that such relatively narrow bridge lanes caused drivers to collide with abutments, railings, guard-rails and other traffic.

52

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

A related factor to the width of the bridge is the railing that is positioned along the length of the structure. Where railing is absent, a vehicle is placed at significant risk if the drop off the bridge is substantial (Agent & Dean, 1976). However, impacts with an unprotected bridge rail, particularly at high speed, can also contribute to increased crash severity (RACV, 1995). It is important to note that many rural bridges do not have protection installed to their rails (Smith, 1982). Where the drop is not particularly deep or steep, or traffic volume is very low, safety benefits may be gained by removing bridge rails altogether. It may also be appropriate to install guard-rails at sites other than these, in order to protect vehicles from impacting on the bridge rails (Evans, 1997). Other important geometric features of bridges and culverts that can contribute to crashes are vertical alignment (usually combined with ice and snow), horizontal alignment of the bridge, particularly left curve (note, these data were collected in the USA), and entrances of intersections adjacent to the bridge or its approach (Hilton, 1973; Ogden & Howie, 1989b). Hilton also noted that the road geometry appears to become particularly significant in bridge crashes when there is a combination of a downhill approach and a horizontal curve in the road. Ogden and Howie (1989b) found that while only 19% of bridge crashes occurred at bridges on curves, these crashes accounted for 33% of bridge-related fatalities. Driver behaviour in response to a perceived hazard has also been identified as a significant factor in collisions that occur either on the bridge itself or on its approach (Agent & Dean, 1976). Specifically, on a bridge that is narrow and/or lacks shoulders, the driver tends to react by laterally displacing the vehicle towards the centre of the bridge (Ivey et al., 1979). The average movement towards the centre line varies according to the width of the bridge. This movement is reported to range from 0.7m on a 4.5 m wide bridge, to no noticeable displacement on a bridge that is 10.8m wide (Ivey et al.). This reaction has the potential to place the driver in the path of oncoming traffic. Ogden and Howie (1989b) also observed that drivers tend to slow down around corners, but not on crests or dips in the road and attributed this to the fact that drivers only see horizontal alignment as a hazard but not vertical alignment. Many researchers have also noted that bridge and culvert related crashes occur at much higher frequency during darkness and adverse environmental conditions (Evans, 1997). For example, Mak and Sicking (1994) found that 48% of collisions with bridge railings occurred at night, and 84% of crash sites had no street lighting, suggesting that this may be due to problems associated with drivers visual perception of bridge sites. However, to date this suggestion has not been fully explored and other factors such as differences in speed and alcohol use between day-time and night-time may also be important. Weather appears to be another significant factor in crashes at bridges and culverts. Hilton (1973) reported that 50% of crashes at bridges occurred in wet, snowy or icy conditions, while for all crashes, this figure was 31%. While Ogden and Howie (1989a) found that environmental conditions were not significant factors, differences in weather patterns between sites at which the data were collected need to be considered, as generally, Australian conditions are not as extreme as other countries (North America and many European countries). 2.4.6.3 Construction zones Vehicle travel through roadway construction or maintenance work zones is known to be hazardous. Crashes on the approach to or through construction zones are particularly severe on rural highways, probably due to high travel speeds and/or high volume traffic,

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

53

and mainly involve multi-vehicle crashes, rear-end impacts and often with heavy vehicles. Indeed, it is suggested that severe injury work zone crashes occur most frequently on freeways and major arterials on weekdays (Sorock, Ranney & Lehto, 1996; Wang, Hughes, Council & Paniati, 1995). Crash analyses consistently show t hat crash rates increase about two-fold on road segments containing work zones than on the same road segment before the zones were implemented (Doege & Levy, 1977; Garber & Woo, 1990; Graham, Aulsen & Glennon, 1977; Hawkins, Kacir & Ogden, 1992; Pigman & Agent, 1990; Hall, 1988). Hall (1988) found that the crash rate in areas of road reconstruction in New Mexico increased by 17 to 33%. Hall also found that, while nine percent of fatalities occur on highways, 34% of fatalities on highways occur at or near construction zones, and often at night. Moreover, the Federal Highway Administration (Morgan, 1987 cited in Hall, 1988) reported that tractor-trailers, which are normally involved in about ten percent of fatal crashes on rural roads, accounted for 20% of fatal crashes in construction zones. In the USA, the number of annual vehiclerelated fatalities in construction zones averaged about 770 between 1995 and 1999, and the average number of persons injured each year was about 40,000 (www.wzsafety.tamu.edu). Generally, problems occur when two or more lanes of traffic must be warned sufficiently in advance of the presence of a construction zone and closure of lanes or sections of lanes so that drivers may travel safely through the one lane passing through the work zone. Whenever a work zone is created on an existing road, it represents an unexpected change in visual features, path, speed, and, occasionally detouring off the normal route. Most importantly, the closure of a lane during construction or maintenance work means that drivers must make behaviour adjustments such as reducing speed and/or changing lanes. A number of studies have investigated the characteristics of crashes in work zones. Certain crash types, such as multiple-vehicle crashes and rear-end collisions, are over-represented in construction zones, and have been related to large deviations in speed. Construction zones have also been found to have a higher proportion of sideswipe crashes and crashes into fixed objects than other sections of highways (Fildes et. al., in press). Rear-end collision account for 35-40% of crashes leading up to the work site, however, at the actual work site 30-40% of crashes involve a collision with a fixed object (McGee, 1982 cited in Hall, 1988). Pain, Hanscom and McGee (1983) suggested that this range of crash types in construction zones reflects the change in crash rate at different places within the work zone, for example, many rear-end crashes occur on the approach to the construction zone and where tapers are marked for lane closure where there is large speed variation and merging manoeuvres, while side-swipe and multi-vehicle crashes occur within the work zone and exit area. Hall (1988) also noted that pedestrian crashes increase during construction work. This, he argues, is due to the fact that construction zone workers are classified as pedestrians and, with high numbers of workers at these locations, pedestrian exposure increases. The literature identified a number of predominant behavioural factors contributing to construction zone crashes. These include: driver inattention; failure to drive within a single lane; speed (both failure to reduce speed and driving too slow); failure to give way, failure to drive within the designated lane (Garber & Patel, 1995; Sorock et al., 1996); following too closely; and, improper lane changing (Hall & Lorenz, 1989). Alexander and Lunefeld (1986) emphasised that driver expectancy is a key factor affecting the safety and efficiency of all aspects of the driving task and this is especially so for areas that change unexpectedly because of road works. Indeed, Pigman and Agent (1990) examined crashes at 20 work

54

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

zone locations and found that the most frequently listed contributing factors were driver attention errors and failure to give right-of-way. Others have noted operational and geometric design factors contributing to crashes in construction zones. Pain et al. (1983) noted the following factors: abrupt changes in elevation at the edge of through traffic lanes; failure to use, or hazardous application of, temporary concrete barriers; inadequate or inconsistent use of devices and methods in closing roadways and establishing lane closure tapers; traffic control system design that fails to consider roadway geometric design; and, inadequate, improper and/or inconsistent use of advance warning signs.

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

55

56

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

COUNTERMEASURES

This Chapter reviews in detail the many improvements to rural road infrastructure worldwide aimed to reduce the frequency and severity of crashes. In addition, it provides a review of innovative philosophies and visions and discusses some examples of integrated and innovative use of treatments on rural road sections. Like in the previous Chapter, an overview of the contents of this Chapter is provided first, followed by detailed discussions of these treatments and technologies and, where available, effectiveness of the treatments. 3.1 3.1.1 OVERVIEW Innovative philosophies and visions

The notion of providing a crashworthy road network is one that is now receiving attention world-wide. Traditional safety models have stressed a balance between safety and mobility. More recent models, however, such as Vision Zero in Sweden and Sustainable Safety in The Netherlands view safety as the most important determinant in the provision of the road-transport system. These models argue that the road-transport system can only be safe when the road infrastructure is designed and operated in a way that explicitly recognises that humans do make errors and designs crashes that are within the human tolerance to violent forces so that death and serious injuries can be prevented. Effectively, this means reducing travel speed or providing improved protection of road users through infrastructure and other changes. The review of philosophies and visions reveals that, in order for the road system to operate efficiently and safely, design criteria must match the characteristics of the road users and vehicles using it. However, it seems that, for the most part, the system is an unforgiving one that does not take into account the combination of high speed, human error, and vehicle protection limitations (see Sections 3.2.1 to 3.2.3). A number of programs world-wide that innovatively combined both high and low cost treatments on long sections of rural roads are discussed and all show promising safety benefits. In general, treatments aim to reduce the types of crashes commonly occurring on rural roads including single-vehicle crashes (run-off-road and striking a fixed object) and multi-vehicle crashes (including head-on, rear-end and sideswipe crashes). The treatments range from provision of visible traffic signs, edge-line and centre-line treatments, signing on curves, provision of overtaking lanes, realignment improvements of curves, sealing and widening of shoulders and pavements, installation of roundabouts and reductions in speed limits (see Section 3.2.4). 3.1.2 Treatments to reduce speeding

Speed and speeding are major underlying factors in crash and injury risk on rural roads and need to be addressed effectively if there is to be any reduction in crash and injury risk on the rural road system. Even small reductions in travel speeds can result in significant crash and injury reductions. The important issue of speed and its impact on crash and injury risk is discussed and measures aimed to reduce the trauma associated with high speed are reviewed (see Section 3.3). There are many forms of speed control including conventional enforcement, education and publicity, and engineering traffic management approaches, and it seems that the physical

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

57

road infrastructure plays a major role in determining the speeds at which drivers and riders choose to travel. The following four effective speed management approaches on rural roads are: Setting of speed limits appropriate for the environment, specifically recognizing the following: o o o Trade-offs between death and serious injury and travel time savings may be unacceptable, A powerful relationship exists between crash and injury risk and travel and impact speed, There are many limitations with the current criteria used for setting speed zones such as 85th percentile speeds, presence of roadside hazards and vulnerable road users, etc. These criteria are often insensitive to factors that influence crash and injury risk. M uch of the literature indicates that drivers and riders are not good judges of what is a safe speed, especially in changing traffic and road environment conditions. Thus, speed zone setting practices that are based on measurement of 85th percentile speeds may well be flawed in logic and therefore ought to be cognizant of other factors that influence speeding (see Section 3.3.1).

Implementation of traffic calming measures on approaches to and within rural townships: o Treatments such as speed humps, pavement narrowing, median islands and raising of the road surface can result in significant speed reductions and reductions in speed variation (see Section 3.3.2).

Implementation of selected low-cost perceptual countermeasure treatments can bring about an unobtrusive change in speed behaviour and may be appropriate at hazardous locations such as curves and intersections: o Treatments such as transverse lines, lane edge and median treatments, chevrons and enhanced post spacings can lead to speed reductions of up to five percent at some locations (see Section 3.3.3).

Technologies to enhance speed limit compliance such as dynamic messaging in the form of active speed warning signs and/or variable message signs. Implementation of speed indicator displays (SIDs) on approaches to rural towns (estimated speed reductions in the order of 4 to 8 km/h with associated BCRs predicted to range for around 7 to 45, depending on the particular crash circumstances (see Section 3.3.4). Treatments to road features

3.1.3

Treatments to various road features known to increase risk on rural roads are reviewed. In general, these fell into the categories of treatments to improve the safety of road lengths, horizontal curves, roadsides, intersections, vertical curves, bridges and culverts, construction zones and treatments to improve the safety of vulnerable road users. These are:

58

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

3.1.3.1 Treatments on road lengths Crashes with high injury outcomes on road lengths generally consist of run-off-road crashes and head-on crashes, and the following measures to reduce these types of crashes include: Provision of additional lanes (see Section 3.4.1): o Conversion of two-lane undivided roads to four-lane undivided roads provides limited safety benefits (up to around 20% reduction in crashes, but possibly also increased crash risk on low volume roads because of resulting higher travel speeds), Implementation of 2+1 roads provides substantial safety benefits (good safety experience reported in European countries and more cost-effective than conversion of two-lane to four-lane roads), Provision of overtaking lanes (reported crash reductions of between 34 and 40% in all crashes, 60% for fatal crashes, and 20% for injury crashes),

Physical separation of traffic directions (see Section 3.4.2): o o o o Conversion of two- to four-lane divided sections provides substantial safety benefits (crash reductions ranging from 40 to 60%), Raised island medians provide greatly enhanced safety particularly in combination with barrier fencing and guard-rails, performance,

Wide medians with widths of 60 to 80 ft with flat surfaces should be considered where feasible, Flush medians also provide cost-effective safety benefits (similar crash rates as highways with added turn lanes),

Improved road surface (see Section 3.4.3): o Improvements to roughness of road surface provide some safety benefits (reductions in multi-vehicle crashes, but possibly also an increase in singlevehicle crashes due to resulting higher operating speeds), Provision of wake-up pavements reduce fatigue-related crashes (reported large reductions in crashes attributed to fatigue), Provision of skid resistant paving provides some safety benefits (reported reductions in crash occurrence and severity of injuries).

o o

Improved delineation (See Section 3.4.4): o Addition of edge-lines and centre-lines provides some safety benefits (reported crash reduction ranging from no effect up to 36%, but suggestion of crash increases on narrow roads carrying low volumes of traffic therefore possibly only effective on roads with high traffic volumes, fewer lane-keeping errors with wide and high contrast line markings).

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

59

3.1.3.2 Treatment on horizontal curves Given that many severe injury crashes occur on or near horizontal curves, much effort has been given to developing and implementing treatments at these sites, including: Speed reduction (see Section 3.5.1): o Consistency in design to achieve low speed variance on the approach to and through the curve, and low variance between passenger cars and heavy vehicles (recommended maximum differentials between 15 and 25 km/h), Updating of design guidelines to include checks of operating speeds, Advanced technology solutions on approaches to curves such as speed limiters or intelligent speed adapters (found to significantly reduce speeds by approximately 6 km/h and particularly effective in small radius curves), and Road markings to reduce speed and speed variance (the most effective being transverse lines with reduced spacing, chevrons with increasing angles and hatched areas at the edge of the road).

o o

Curve flattening (see Section 3.5.2): o Flattening of sharp curves, particularly those over 5 can reduce crash and injury risk, vehicle operating costs, travel time and maintenance costs, however, can be an expensive treatment and unlikely to be cost-effective for less sharp curves (not more than 11), and low AADT.

Improved alignment (see Section 3.5.3): o Updating of design guidelines for maximum super-elevation rates and side friction factors to allow for high operating speeds and side friction demand.

Transitions (see Section 3.5.4): o Spiral transitions are the preferred design, however, both benefits and disbenefits are reported, depending on degree of curvature, lane and shoulder width, roadside (estimated crash reductions of 2 to 9%), and It is suggested that spiral transitions are beneficial on sharp curves in level terrain with good sight distance, wider carriageways and shoulders, but result in increased crash probability on flatter curves in rolling terrain).

Signing and road marking (see Section 3.5.5): o Consistency in provision of signing and road markings on the approach to sharp and/or sub-standard curves. For example, on sharp curves where operating speed is markedly higher than design speed, advance curve warning signs, curve advisory speed signs, multi chevrons throughout the outer side of the curve, enhanced centre-lines and edge-lines through the curve with retroreflective pavement markers, centre barrier lines, are recommended (estimated reduction of day-time crashes by 5% and night-time crashes by 25%, estimated cost/benefit of at least 2:1 for 5-year life of reflective pavement markers), and

60

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

Potential safety effects of a dynamic curve warning system on sharp curves (moderate effect on vehicle speeds, but estimated improvement on ability of high-speed vehicles to successfully navigate through curves).

3.1.3.3 Treatments on roadsides Single-vehicle crashes are common on rural roads and often result in severe injury to car occupants. Treatments on roadsides generally fall into two categories, those that aim to protect car occupants from serious injury in the event that they leave the carriageway and those that aim to keep vehicles on the carriageway. Measures include: The provision of safe or forgiving roadsides. There are many opportunities available to protect car occupants from serious injury including the provision of a clear-zone. The larger the clear-zone the safer the road, however, there are often practical constraints on how wide this zone can be and there is general consensus that the minimum width for a clear-zone to effectively reduce severe injury crashes is 3m. While much of the literature focuses on the effect of removal or trees and poles (with reported crash reductions between approximately 13% to 60%) there is also suggestion that removal or relocation of other roadside hazards such as culvert headwalls, signs, mailboxes, fences, gates and poorly placed guard-rails can reduce crash and injury risk (see Section 3.6.1). Safety barrier systems have proved to be an effective means of protecting occupants from striking rigid objects in the event that they leave carriageway and to prevent vehicles reaching hazardous roadside objects (see Section 3.6.2). There are also a number of shoulder improvements that have been shown to reduce crash and injury risk (see Section 3.6.3). These are: o o o Provision of sealed shoulders with minimum width of 1m (with estimated crash reduction rates in the order of 10 to over 40%) (see Section 3.6.3.1), Flattening of side-slopes (with estimated crash reduction rates of at least onethird) (see Section 3.6.3.2), Edge-line treatments, particularly those that alert a driver to the imminent departure of their vehicle from the roadway and those that reduce the danger once they have actually left the paved surface. These include shoulder rumble strips (estimated to reduce run-off-road crashes by between 13% and 50%) and indented shoulder strips (estimated to reduce crashes by approximately 30%) (see Section 3.6.3.3).

3.1.3.4 Treatments at intersections Many multi-vehicle crashes on the rural road network occur at intersections because of the increased potential for conflicts. Treatments that have been introduced or suggested to reduce the incidence and severity of crashes at these locations include: Provision of grade-separated intersections (see Section 3.7.1):

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

61

Considered to greatly enhance safety, however, have lower BCRs due to their high capital cost, and are rarely considered for intersections on low-volume two-lane rural roads, Improvements to geometric features of grade-separated intersections include: provision of longer entrance ramps (minimum lengths of between 240 and 280 m); provision of multiple advance warning signs before exit ramps; and, provision of traffic control devices (signs and pavement markings) to prohibit wrong-way movements on exit ramps from the minor intersecting road,

Provision of appropriate traffic control at at-grade intersections: o o fully-controlled turning phase signals, as part of overall signalisation of an intersection. Benefits are greater on high-volume roads (see Section 3.7.2.1), Provision of stop or give-way signals that are retro-reflective and conspicuous, holding lines and advance warning signs to alert drivers of the presence of the intersection on low-volume rural roads (see Section 3.7.2.2), Provision of new roundabouts which have been found to result in significant reductions in crash frequency and crash severity (up to 80 and 90%, respectively) with BCRs of up to 19 for hazardous rural intersections (see Section 3.7.2.3),

Provision of adequate sight distance on approaches to and entering intersections, and when turning across oncoming traffic (a minimum perception-reaction time value of around 5 s is recommended for rural intersections). Substantial crash reductions of up to 67% are estimated (see Section 3.7.3), Provision of adequate delineation and alignment: o Associated benefits of channelisation and delineation include crash reductions between 30 and 60% and BCRs of 2.3. Improvements to channelisation and delineation include: the use of raised channelisation with raised reflective pavement markers for turning lanes; provision of features to prevent wrong lane-use when turning such as lane-use control signs, lane-use arrow markers, pavement markings ad delineation of median noses; provision of painted edgelines when unsealed shoulders are present; and provision of semi-mountable kerbs that are painted on the slope and at least a portion of the top surface (see Section 3.7.4.1), Provision of wide medians with geometric features such as provision of divided highway crossing sign to prevent wrong-way movements, adequate width to accommodate crossing manoeuvres of buses and trucks, and TWLTLs (with estimated crash reductions of approximately 20 to 30%, particularly rearend and sideswipe crashes) (see Section 3.7.4.2),

Provision of warning signs to alert drivers of the presence of intersections including flashing beacons, variable message signs and vehicle-activated signs (see Section 3.7.5).

62

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

Provision of lighting at intersections (estimated significant reductions in crashes and benefits that outweigh cost by a ratio of 15 to 1) (see Section 3.7.6). Measures to improve safety at rail-road intersections. Due to the prohibitive cost of upgrading passive rail-road intersections with active-warning devices, countermeasures must focus on improving sight lines by removing foliage, maintaining the crossing so that drivers can devote a maximum amount of their attention to scanning for trains, and further encouraging drivers to slow down and actually look, regardless of how infrequent trains are at the particular crossing (See Section 3.7.7).

3.1.3.5 Other improvements Other improvements included measures to improve the safety on vertical curves, on bridges and culverts, construction zones, and the safety of pedestrians and bicyclists. The main measures to alleviate these problems include: Vertical curves (see Section 3.8.1): o o Minimum stopping sight distance between 95 and 110 m, Review of placement of driveways and intersections near crest vertical curves,

Bridges and culverts: o Widening of bridge and shoulders with a minimum of 2 m road width and 1 m shoulder width (estimated reduction in crashes ranging from 15 to 85%) (see Section 3.8.2.1); Installation of guard-rails, particularly when widening is not feasible (while crash frequency may be reduced, estimates of severity reductions of between 80 and 90%) (see Section 3.8.2.2); Approach-related measures such as reducing speed limits, provision of signage to warn of narrow bridges and delineation such as hazard markers, reflective guide posts, chevrons, raised reflective pavement markers, edge- and centre-line treatments and rumble strips (crash reduction estimates of between 15 and 18% with installation of raised reflective pavement markers) (see Section 3.8.2.3);

Construction zones: o Posting of lower regulatory speed limit on approach to and through construction zones (with added flaggers and variable message signing) (see Section 3.8.3.1); Supplementary signs and devices for lane closure, e.g., flashing arrows at beginning of lane closure taper, flashing warning signs, symbol signs, variablemessage signs, (significant reductions in number of vehicles in wrong lane and reductions in speed variances) (see Section 3.8.3.2);

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

63

Devices to guide drivers safely through zones including safety barriers to separate opposing traffic, raised pavement markers to increase delineation, temporary rumble strips, chevrons and wing barricades (see Section 3.8.3.3).

Pedestrians and cyclists (see Section 3.8.4): o o o o o Implementing speed-reducing measures where there are high numbers of pedestrians and cyclists; Provision and maintenance of wide road shoulders, particularly near bus stops; Improved bridge design to incorporate pathways for pedestrians and cyclists; Construction of separate bicycle and pedestrian lane networks; Improvement of visibility of pedestrians and cyclists, placement of fencing and convenient places to cross, provision of refuge islands, signs and markings. ****************

International literature has identified the many issues that confront safe travel on rural roads and the measures that have been undertaken to alleviate the safety issues. Unfortunately, however, the OECD still argues that these problems have been neglected in comparison with those in urban areas (OECD website, cited in Grigg, 2000). This may, in part, be attributed to the fact that rural roads often require different countermeasures and certainly a different approach (Queensland Parliamentary Travelsafe Committee, 1999). Despite the fact that many urban and rural road crashes appear similar in character, tackling the rural problem is more complex. Most crashes occur where exposure to risk is highest and black-spot sites are readily identifiable in urban areas. However, in rural areas that are more sparsely populated and lightly travelled, fewer crash clusters are identifiable and the black-spot approach is less efficient. The majority of rural crash sites are widely dispersed across the road network and this makes it difficult to find viable, cost-effective, engineering-based solutions. Difficult decisions about priority need to be made. However, given the severity of crashes when they do occur on rural roads and the large proportion of the serious injury problem occurring on rural roads, considerable effort must be made. Driver factors (including human error, risk-taking behaviour, inattention and fatigue) as well as vehicle factors contribute to play a major role in rural crashes. While there is clearly much effort worldwide to change road user behaviour, and to design safer vehicles, it is clear that these strategies have met with limited success and may not effectively reduce trauma on rural roads in the years ahead. Improved infrastructure, on the other hand, can provide an immediate and effective countermeasure to reduce the frequency and severity of rural crashes. However, it must first be recognised that human operators are prone to make errors. These errors often occur at speeds that exceed the combined abilities of the road infrastructure and the vehicles that are used on it, to protect vehicle occupants or to avoid serious injuries to others who may be struck. Improving the infrastructure means creating a safer travel environment. Infrastructure improvements that tolerate human error and take account of vehicle safety limitations, can provide high levels of mobility while preventing crashes or serious injuries. Indeed, many physical improvements to the road have been shown to be highly effective and it seems that innovative use of existing treatments known to reduce crash and injury risk may prove
64 MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

to greatly enhance the safety of the rural road system. The list of potential infrastructure upgrades or modifications to reduce the frequency and severity of crashes is extensive, however a fundamental principle is the responsibility of the road authority in any jurisdiction to build a safe road system. Such a responsibility would make the road authority, and therefore the government of the day, accountable for the safety p erformance of the network. Good roads are those which have no unexpected or unnecessary hazards, either related to the road geometry, or to features of the roadside such as roadside objects. The literature has identified a range of treatments that can be applied on rural roads to reduce the incidence and severity of crashes (Ogden, 1990; Austroads, 1995; Nielsen, 2000): Increasing the seal or carriageway width (lane or shoulder), Improving the condition of the road (surface improvements, shoulder edges), Improving delineation (painted lane and edge lines, post mounted delineators, raised reflective pavement markers, chevron alignment markers at curves, width markers at bridges and culverts, etc), Providing a forgiving roadside (increased clear areas on the roadside for vehicles leaving the roadway by removal or relocation of roadside objects when they are close to the pavement or, where they cannot be removed, provision of guard fencing, provision of appropriate shoulders), Treatment of substandard horizontal or vertical curves (flattening of curves, widening of pavement width, appropriate super-elevation, effective and uniform warning signs and advisory speed signs at curves), Intersection treatments (improved sight distance, provision of warning signs, hazard markers, splitter islands, provision of turn lanes and clearly defined paths, installation of roundabouts), and Speed reducing measures (setting of speed limits, implementation of traffic calming measures and perceptual countermeasures).

In what follows, a number of innovative philosophies and visions are outlined, then measures to improve safety on rural roads are identified and discussed. As speeding is possibly the major contributor to road trauma on rural roads, measures to reduce speeds are reviewed at some length, followed by the measures to improve safety on road lengths, horizontal curves, roadsides and intersection design and operation. 3.2 PHILOSOPHIES OF ROAD SAFETY PROGRAMS

Traditionally, road safety programs aimed at treating hazardous road locations have concentrated on the identification of black-spots. Some have argued that this concept works well in urban areas but is more difficult to apply on rural roads because crashes on rural roads are rarely clustered, rather they are usually spread along a length of the road (Lydon, 1997; Ogden, 1990). Nevertheless, there are road features which can be expected to be associated with poor safety and innovative application of proven countermeasures

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

65

rather than the development of specific site-related countermeasures would seem appropriate for rural road safety. More recent philosophies introduced in some countries have emphasized innovative safety initiatives. For instance, some European countries have adopted general philosophies that enhance community values and integrate roadways into communities and the environment and design roadways for specific purposes and address safety in a way that considers all users. In particular, the Swedish Vision Zero and the Dutch Sustainable Safety philosophies have received much attention worldwide. These philosophies include some ethical viewpoints that guide a broader view of what is to be done to reduce road trauma. 3.2.1 Swedens Vision Zero

The SNRA has adopted the Vision Zero philosophy, with the primary goal to ensure a socio-economically efficient transport system that is sustainable in the long term for individuals and industry throughout the country. To achieve this goal, five sub-goals have been identified, including high accessibility, high transport quality, a good fit in the environment, promotion of regional development and, most importantly, no fatalities or serious injuries (zero vision). Sweden is widely recognised as a world leader in road safety, and often displays what is considered worlds best practice with regard to road safety and road trauma reduction. Since 1997, Sweden has based its road trauma reduction efforts on Vision Zero, a radical and ethically-based philosophy which holds a long term goal that no-one will be killed or seriously injured within the Swedish road transport system, provided they behave legally while using the system. A unique and driving force in Vision Zero is the strong ethical foundation to the way in which the road transport system is designed and operates and one that accepts that humans will make errors (SNRA, 2002). A new philosophy towards road trauma reduction is vital to assist social change among those who design, operate and use the road transport system so that safe choices become easy choices, and health is no longer traded for mobility (Ministry of Transport and Communications 1997). Vision Zero appears to offer such a philosophy, so its potential to strengthen road trauma reduction is high. The Vision centres around an explicit goal, and develops into a highly pragmatic and scientifically-based strategy which challenges the traditional approach to road safety. This means that mobility and safety cannot be traded against each other, and so mobility becomes a function of safety, not vice versa. Speed must be limited to a level commensurate with the inherent safety of the road system, which is a true paradigm shift and contrasts to the more conventional approaches, where human life, mobility and other impacts are valued in financial terms and compared with each other. Speed limits within the road transport system should be determined by the t echnical standard of vehicles and roads so as not to exceed the level of violence that the human body can tolerate. The safer the roads and vehicles, the higher the speed that can be accepted. Whilst the philosophy acknowledges that it may never be possible to achieve the ultimate goal of zero deaths or serious injuries, only by striving for this vision will the energy needed for action and innovation be generated.

66

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

3.2.2

The Dutch philosophy

The Dutch concept of sustainable road safety is similar to the approach taken in Swedens Vision Zero. It is based on the idea that humans are the reference standard. Humans are capable of many things, but traffic makes heavy demands and humans make mistakes. The transport system therefore needs to be adapted to humans so that they can behave safely (Wegman, 1995). A sustainably safe traffic system is based on three key safety principles, i.e., functionality, homogeneity and predictability, and has: An infrastructure that is adapted to the limitations of human capacity through proper road design, Vehicles fitted with ways to simplify the task of man and constructed to protect the vulnerable human as effectively as possible, and A road user who is adequately educated, informed and, where necessary, controlled.

The sustainable safety approach focuses on reduced alcohol use, increased use of seatbelts, speed management, separation of cyclists and vehicles, improving hazardous locations, addressing issues regarding heavy vehicles, and providing a network infrastructure that is self-explaining. The ultimate objective is development of a uniform roadway network where similar roadways will look and drive alike. Van Schagen and Janssen (2000) noted that the most important measures along rural distributor road stretches are: i) introduction of parallel or alternative facilities for slow traffic and local traffic, so that only motorised through traffic drives on the main carriageway, ii) separation of driving directions in combination with a speed limit of 80 km/h creating a more homogeneous driving speed and eliminating overtaking manoeuvres, and iii) improvements to roadside safety by creating obstacle free zones. In addition, they argue that speed reduction measures such as roundabouts are the most suitable measure to increase safety at crossroads of rural distributor roads and recommend that only turning right should be permitted on crossroads of three branches. Van Schagen and Janssen estimated that if these measures are implemented by 2010 the potential injury crash reduction on rural distributor roads would be 43% for road stretches and 45% for crossroads compared to 1998. Further, because of reduced speed and increased roadside safety, the crash severity would be reduced by 20%. Wegman (1995) argues, however, that currently many non-motorway rural roads fail to meet these requirements. Outside built-up areas, very little has been done to design nonmotorway roads in accordance with the principles of sustainable safety. For example, he argues that the majority of present non-motorway rural roads fail to separate slow and fast moving traffic and speeds are still too high for the conditions. 3.2.3 EuroRAP

EuroRAP is the European Road Assessment Programme for rating road-related crash risk in an attempt to mobilise government action through community demand. It systematically tests risk on roads, awards star safety ratings and identifies problems that can be addressed by practical road improvement measures. The philosophy behind this program stems from vehicle occupant safety programs such as the EuroNCAP car crash test programme. It acknowledges that, just as vehicles have different protection capabilities, so roads should not be designed that give little safety margin in the event of a simple driving error. It also

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

67

promotes the recognition of the need to reduce death and serious injury, rather than simply reducing crashes. This programme began in 2000 and is led by the AA Foundation for Road Safety Research based in the UK and is currently supported by 14 other leading motoring organizations across Europe, assisted by governments in the UK, Netherlands and Sweden. By the end of 2004 it is likely that the work will be extended into 12 European countries including Spain, France, Italy and Ireland. EuroRAP shows the safety performance of routes in relation to the amount of traffic they carry, not just the number of crashes that occur on any given stretch. The scoring system includes separation of opposing traffic flows, checks for roadside protection, intersection design and frequency, and checks for facilities for pedestrians and cyclists. A motorway, for example, will achieve the highest 4-star rating if: The central reserve and sealed shoulder are at least three metres wide; The land adjacent to the road is clear to a minimum width of ten metres or there are barriers to all aggressive features; It is generally flat and straight without being featureless; Merging traffic can only use major intersections; There is no access permitted for pedestrians; and There are safety areas for cars to stop in an emergency.

A road will receive the lowest safety assessment if: It is undivided; It is less than nine metres wide; It has no sealed shoulder or roadside clear zone; It has a high frequency of bends and crests, uncontrolled cross roads at major intersections, cross- and side-road access; and Has no provision for pedestrians.

Using this method, 833 roads have been assessed in the UK and 23 achieved no stars in the ratings, 90 received only one star, 213 two stars, 415 three stars, and 92 gained the top four star rating. Roads with no stars had crash rates ten times higher than the best performing roads in the four star category (www.eurorap.org). EuroRAP information is useful at several levels. First, it will generate consumer information, providing road users with information about the relative risk of roads and understanding of the features that make roads safe and unsafe. For instance, a risk rate map has been published for selected road sections in England, presenting the colour coded crash risk bands from low to high, with road sections in black having a risk rate more than 10 times higher than those coloured in dark green. The ratings highlight road sections where there is a lack of appropriate balance between the speeds at which a vehicle might be
68 MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

involved in a crash and the protection provided by the road and vehicle design to minimise the severity of any resulting injury. Secondly, it provides important messages for road designers and planners. It demonstrates where the deaths and severe injuries occur on the road network and can initiate more detailed analyses of crashes. It also provides a strategic view of the safety of routes and what to do about what causes injury on these routes. It will also show how and where measures must be implemented to increase crash protection, separate vehicles moving in opposite directions and reduce conflicts where possible, and provide investment in other elements that improve the safety of the road. Thirdly, it provides important information for government agencies on which to base strategies. For instance, the information can be used to balance vehicle design, road engineering and behaviour crash reduction priorities, to promote the understanding of crash and injury risk, and to select appropriate speed limits for roads. In England alone, the identification of dangerous major roads and simple improvements to those stretches that perform worse than average is estimated to save 2300 people from death or serious injury each year. This is equivalent to more than a third of the fatal and serious injury crashes that occur on these roads (www.eurorap.org). 3.2.4 Some examples of integrated use of countermeasures to treat long sections of the rural road network

A number of reports of integrated use of countermeasures to treat long sections of the rural road network were reviewed and are discussed in this section. Lydon (1997) pointed out that many improvements such as road widening, curve realignment or intersection design can involve high costs but lower cost countermeasures such as improved delineation, improved shoulder maintenance and improved signing can lead to significant benefits. Many of the programs reviewed here have combined both high and low cost measures to significantly improve the safety of these road lengths. A combination of low cost engineering measures and enforcement were used to improve the safety on a single carriageway rural road length that connects the Portuguese coastal area and Spain and were reported by De Almeida Roque & Cardoso (2001). This road length has very high average daily traffic volumes, high volumes of heavy vehicles, a design speed of between 80 km/h and 100 km/h, and a high proportion of curves (approximately 55% of the road length). It had a high crash history (35 fatalities and 37 serious injuries in 1997). The most common crash types included head-on collisions and run-off-the-road crashes, predominantly on curves, approaches to interchanges, and on zones with additional climbing lanes. These crashes seemed to be mostly related to speeding and irregular overtaking. A range of low-cost measures were implemented including improvements of the surface water drainage, improvement of traffic sign visibility, definition of 90 km/h speed zone throughout the section, provision of edge rumble strips, suppression of overtaking possibilities in dangerous locations, signing of curves, compulsory use of day-light running lights, mounting of plastic position markers on the centre-line on curves, reinforcement of deceleration lane markings, installation of prohibitory sign panels in descending lanes with no passing zone for trucks only. Enforcement consisted of elimination of tolerance levels and increased overall activity of traffic police by more than 75% in the first month and by 25% thereafter. The impacts on driver behaviour, traffic volumes and on crash frequencies were examined using a beforeand-after study. Overall, the engineering treatments resulted in lower speeds under night conditions and on difficult road curves, however, speeds in daylight hours remained the
COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS 69

same. Combined with the enforcement campaign, lower speeds were found in daylight hours, however, the effect during night time was not significant. The engineering measures resulted in significant reductions in speeds at curves (in the order of between 5 km/h and 11 km/h), especially in the inside lane, better lateral placement of vehicles in lanes, a reduction in vehicle paths over the inside shoulder and the distance to the right edge line for vehicles in the inside lane significantly increased. On tangents, no significant change in speed was found due to the engineering measures, however, speed reductions were associated with the enforcement campaign. Encouragingly, the use of daylight running lights increased from under 11% before treatment and enforcement to over 90% after treatment and enforcement. Truck driver queuing and headway on climbing lanes did not change as a result of treatments and/or enforcement. Preliminary crash figures showed that the number of crashes remained fairly constant over the time period studied (1997 1999), however, given that AADT increased 9% in 1998 and an additional 6% in 1999, comparisons of crash risk to 1997 figures showed a modest decrease in fatal and serious injury crash rates. In Australia, a 1992 survey revealed that the major highway between the Australian Capital Territory and the New South Wales coast was substandard. The road provided limited overtaking opportunities (over 60% of the overall route), a winding alignment, reduced speed limits, poor delineation and sections of steep grade and the most prominent crash type occurring on the road was run-off-road crashes, particularly on curves (McDonald & DeSanti, 1998). Subsequent improvements were made to the road, including increased number of lanes on steep grades and alignment improvements to horizontal curves. These improvements were associated with a 33% reduction in the crash rate and a 47% reduction in the casualty rate (i.e., fatalities and injuries). However, McDonald and DeSanti (1998) acknowledged that whilst some good improvements have been completed, the road remains unforgiving. Numerous roadside hazards, limited overtaking opportunities, hazards such as unprotected drop offs in close proximity t o the road edge, unprotected structures on the end of narrow bridges, poor guard-rail terminals and twisting alignments still leave very little margin for error. They further suggested that, where possible, run-off-road crashes may be minimised by further provision of good delineation and wide even shoulders clear from roadside hazards, provision of overtaking lanes and realignment of substandard curves. Walker and Lines (1991) identified the contribution made to crash reduction by improvements to the rural road network at 85 sites in three regions in England using a before-and-after experimental design. The treatments included regional and national schemes using route treatments (for example, bend re-alignment) and spot treatments (for example, intersections). They found: An overall crash reduction of approximately 17% after treatments on regional schemes, however, if downstream effects were taken into consideration, this figure dropped to around eight percent, Crash remedial treatments at intersections (where the majority of conflicts occur) produced net savings in the order of 50%, while non-intersection crash remedial treatments indicated a (non-significant) net savings of approximately 13%, An overall crash reduction of 11% after treatments on national road schemes with the rural schemes, as a group, producing a highly significant reduction estimated at 32% of the before crashes.

70

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

The greatest benefits of national road schemes in terms of crash reductions, were produced by grade-separated intersections (57%), rural road re-alignment (42%) and rural bypasses (32%) to both single and dual carriageway standard. Corben, Newstead, Diamantopoulou and Cameron (1996) reported the results of an evaluation of the accident black spot program funded by the Transport Accident Commission (TAC) in the state of Victoria, Australia. The program consisted of $75 million of black spot treatments implemented during the 1992/93 and 1993/94 financial years. The evaluation included over 430 treatments throughout Victoria, including rural areas of the state. Effectiveness was measured in terms of the reductions in casualty crash frequencies and casualty crash costs in the treatment group, and the economic worth of the program of treatments. The evaluation also examined the effectiveness of the program by treatment type. Of the treatment types with potential for use in rural areas, the following showed statistically reliable reductions (at the 95% confidence level) in both casualty crash frequencies and casualty crash costs: New roundabouts (91% and 75%, respectively); Intersection channelisation (42% and 80%); Tactile edge marking (9% and 19%); and Shoulder sealing (25% and 39%)

While most of the signal remodel treatments involving fully controlled right-turn phases were undertaken in urban settings, it is worth noting that the corresponding reductions found in casualty crash frequencies and casualty crash costs were 51% and 90%, respectively. As a final note, given the relatively short duration of the after period, the authors advised that the results should be regarded with due care. In a more comprehensive and updated study of the Transport Accident Commission (TAC) funded Accident Black Spot Program, implemented during 1992 to 1996, more than 550 individual treatments throughout Victoria were evaluated (Newstead & Corben, 2001). As with previous MUARC evaluations of this type, the purpose of this study was to evaluate the effectiveness of the program in terms of reductions in casualty crashes and casualty crash costs at treated locations, and the economic worth of treatments. The evaluation also addressed the economic impacts on the community and examined effectiveness by treatment type. The results showed the Program to be effective overall, yielding positive outcomes for the community. In summary, the Program reduced reported casualty crash numbers by 26.4%, and casualty crash costs to the community by 29.6%. Over its full life, the program was estimated to deliver safety benefits to the community of four to five times the program costs. Expressed differently, the Program was estimated to achieve a Net Present Worth of more than AUS$270m for the community. Evaluation of the program identified some treatment types that were highly successful, while others seem to have been unsuccessful or have not been shown to be successful because of insufficient numbers of sites having undergone these treatment types.

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

71

With regard to treatment types that have the potential to be applied in rural settings, Newstead and Corben (2001) found the results reported in Table 2 from evaluating their effects on Victorian black spot locations. Table 2: Effects of treatment type on crash risk and crash cost at Victorian rural black-spot sites. Estimated Casualty Crash Reduction
70% Modify Existing Roundabout New Roundabout Signal Remodel Fully Controlled Right Turn New Signals Other Intersection Improvements Street Lighting Channelisation Splitter Islands Staggered T Left Turn Slip Lane Pavement Resealing (Route) Roadway Delineation (Route) Median Construction Edge Line Marking Curve Delineation Curve Realignment Shoulder Sealing Shoulder Sealing with Regular Line Marking Shoulder Sealing with Obstacle Clearing and Regular Line Marking Pavement Widening 57% 73% 25% 32% 32% 16% 29% 36% 27% 86% 30% 44% 28% 61% 24% 51% 46% 31% 29% 51% 31%

Treatment Type
Roundabout

Sub-Treatment Type

Estimated Casualty Crash Cost Reduction

83% 58% (NS)* 87% 34% 40% 17% (NS) 26% 30% (NS) 55% 40% 98% (NS) 25% (NS) 66% 31% 66% (NS) 31% 20% (NS) 32% (NS) 29% 29% 50% 37% (NS)

* Non-significant results at p<0.05 level. The most impressive of the intersection treatments, applicable to rural settings, was the construction of a new roundabout, where average reductions in casualty crash frequencies of 73% and in casualty crash cost of 87% were achieved. The estimated BCR for new roundabouts was 5.0. Traffic signal treatments also showed statistically reliable reductions

72

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

in crash frequency and cost, but traditionally have been regarded as less s uitable for highspeed rural settings, where injury severity tends to be high when collisions occur. For road lengths, route-based pavement resealing was found to cut casualty crash frequencies by 44% and casualty crash costs by 66%. The estimated BCR of pavement resealing was 18.3. Shoulder sealing, especially when accompanied by obstacle clearing and edge-lining also showed impressive results, with reductions of 51% and 50%, respectively, in casualty crash frequencies and casualty crash costs in the latter case. Average BCR estimates of 4.3, ranging up to 8.6, were also found for shoulder sealing treatments. Route-based roadway delineation was also among the better performing treatments in rural areas. Reductions in casualty crash frequencies of 28% and in casualty crash cost of 31% were found for improved roadway delineation. The estimated average BCR for roadway delineation was 32.8. (All of the above results were statistically reliable at better than 95% confidence). 3.3 TREATMENTS TO REDUCE SPEEDING AND THE INJURY CONSEQUENCES OF SPEEDING

As indicated previously, speeding remains a major contributor to trauma on rural roads, and is held to be a major factor in fatal and serious injury crashes. While there have been many efforts to manage the speed at which people travel on the road, there is still a substantial amount of speed-related trauma. This suggests that speeding and speed behaviour is a continuing problem and one of the last frontiers in road safety. Excessive or inappropriate speed compromises safety and is one of the key areas where effective management, enforcement and education measures are required to minimise crash frequency and severity. In their review of speed and speeding behaviour, Oxley and Corben (2002) argued that it is important t o ensure that speeds are safe when errors are made and to match speed choice with the intrinsic safety of the road-traffic system. They added that such an approach should recognise the: Crucial role of speed in determining injury outcomes, Inherent vulnerability of unprotected road users (e.g., pedestrians, motorcyclists, bicyclists and older road users), Major contribution of road and roadside hazards to road trauma (e.g., presence of trees and poles close to the roadside, unsealed or gravel road surfaces, horizontal curves, and selected operational features of intersections), Inherently low standard of some roads (such roads often do not afford safe travel much above 50-60 km/h, yet are commonly zoned at 100 km/h), Risks associated with high-speed, high-volume operation, Growing problems of vehicle incompatibility, and Capacity of speed enforcement.

Road characteristics greatly influence the speed at which drivers and riders operate their vehicles. The European Transport Safety Council (1995) pointed out that road

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

73

characteristics determine what is physically possible for a vehicle but they also influence what seems to be the appropriate speed to a driver/rider. These characteristics include speed limit, curvature, grade, length of grade, number of lanes, surface conditions, sight distance, lateral clearance, number of intersections, advisory and warning signs, traffic density and composition, and speed of the traffic and the presence of road lighting (European Transport Safety Council, 1995; Fildes et al., 1991; Warren, 1982). There are a number of ways to manage speed and these are discussed in the following sections. 3.3.1 Speed limits

The primary reason for setting speed limits is to strike a balance between travel time and crash risk for specific roadway sections. The posted speed limit aims to inform divers and riders of maximum driving speeds that the authorities consider reasonable and safe for the roadway. As noted previously, many studies have examined the effect of raising or lowering speed limits and the evidence unequivocally shows that the number of crashes and level of crash severity increase when speed limits are raised (and vice versa). The MASTER (MAnaging Speeds of Traffic on European Roads) is one such project that aims to determine appropriate and acceptable ranges of speed for different kinds of road in the European Union. One of the tasks under the MASTER project is to investigate the speed-crash relationships on different kinds of rural single-carriageway roads. Taylor et al. (2002) examined the influence of speed on rural crashes along 171 road links in Sweden, The Netherlands, Portugal and England with speed limits ranging from 70 km/h to 110 km/h, grouping roads according to quality. Low quality roads were defined as those that were hilly, had high bend density and low traffic speed, while high quality roads were those that had low bend density, intersections and accesses, and had a high traffic speed. As noted in Section 2.4.1, they found that crash frequency was highest on low quality roads and that crash frequency in all categories increased rapidly with increases in vehicle speed on all road categories, suggesting that crash rates were related to speed limits. The practice for setting speed limits in most western countries is similar and most guidelines state that the main aim is to achieve a balance between safety, mobility and amenity (AASHTO, 1994; Austroads, 1988). However, for such a vital safety issue as speed, transparency is needed as to how this balance is reached and whether it results in an acceptable safety outcome. It is also unclear as to how these three factors are weighed up to reach this balance. Trade-offs that favour mobility and compromise safety relative to other goals may not be accepted in our present-day society. Striking such a balance implies the trading of the lives and health of humans for other societal benefits such as reduced travel times, lower emissions, and in some cases, improved amenity of areas. It seems that, in many countries, a high level of importance or weight is given to travel time savings expected to result from higher vehicle speeds. While these travel time savings are estimated to be moderate in rural settings, there remains the ethical dilemma of trading human lives and health of other benefits in society. Some societies are explicitly declaring that such trade-offs are ethically unacceptable. Specifically, the notion of trading-off a number of deaths or long-term injuries for moderate travel time savings accrued by a large number of r oad users has been questioned in recent years and, indeed, is no longer acceptable in some societies. The appropriateness of placing monetary value on human lives and then weighing it against time savings has also been called into question. For instance, Miller (1993, cited in Haworth, Ungers,

74

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

Vulcan & Corben, 2001) warned of the danger of making decisions based on conflicting travel time and cost values. He concluded that by using monetary crash costs in resource allocation highway engineers inadvertently increased mobility by sacrificing lives (p.605). Current speed limit setting criteria do not adequately account for the factors that influence driver/rider speed choice and the consequences of human limitations on this choice. While current guidelines purport to meet driver/rider expectations, it is essential that human limitations such as poor risk perception, estimation of speed (over-estimation of others speed and under-estimation of own speed), and motivations for speeding are understood and included among the criteria for speed limit setting. There is clear evidence that drivers and riders consistently drive at speeds above the posted speed limit and that in many environments the average travel speed is not appropriate or safe relative to the road environment, particularly in rural environments where there are unforgiving roadsides, poor road conditions, or sharp curves. Further, in some countries the importance placed on the 85th percentile driving speed means that new proposed limits will always be compared with travel speeds that are most likely excessive for the inherent safety of the road environment. Most speed management guidelines place emphasis on the notion of achieving credible speed limits, arguing that if speed limits are set too low they will not be respected and obeyed. By seeking to match expectation it is assumed that drivers and riders consistently and accurately assess both crash and injury risk and drive/ride accordingly. The evidence, however, suggests that they are not good at t his; they over-estimate safe driving speeds and under-estimate the role of speed in crash risk, but more importantly in injury risk. This tends to perpetuate the setting of speed limits at the higher end of the range. Likewise, current criteria are largely insensitive to increases in injury risk due to the physical environment. Factors such as presence of roadside trees and poles, barriers, the mix of traffic, the diversity and incompatibility of vehicles, and presence of vulnerable road users are largely o verlooked in speed limit setting criteria. Most of our roads zoned at speeds above 60 km/h have trees, poles or other hazards within the roadside, including in clear-zones. Even the best vehicles will not be able to adequately protect occupants in impacts with these objects at speeds around 50-70 km/h, and higher. The historical basis for clear zones, still in use today, is that up to 85% of drivers will recover vehicle control within the clear zone, and by inference, 15% will not be able to recover control. The clear zone concept has been questioned recently (Corben, Tingvall, Fitzharris, Newstead, Les & Johnston, 2001). Corben et al. (2001) suggested that the existing Australian guidelines requiring medians to be less than 15m wide before a barrier is considered warranted are inadequate in todays society because i) they focus only on approximately 20% of the deaths and serious injuries resulting from median encroachments, ii) for head-on crashes, median widths (in the range typically found on Victorian roads) have not been found to be an effective criterion for deciding where median barriers should be installed, and iii) it would be necessary to provide excessively wide medians, in order to prevent errant vehicles from reaching the opposing carriageway in crash scenarios which are quite realistic for Victorian roads such requirements would be uneconomical and generally impractical. In their review of the speed management approach in Victoria, Australia, Oxley and Corben (2002) identified ways to improve the setting of speed limits. They argued that speed zoning policies should be reviewed and strengthened, recognising the powerful relationship that exists between crash and injury risk, and travel and impact speed.

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

75

Moreover, they argued that policies should be set with a view to creating environments that promote safe behaviour rather than relying on drivers/riders to decide what is a safe and appropriate speed. In particular, they noted that the following important factors should be recognised: The injury risk for roadsides containing large numbers of rigid objects, The crash and injury risk associated with roads of lower design standards (including roads with gravel shoulders, unsealed surfaces, inconsistent horizontal and/or vertical alignment, etc.). Such roads often do not afford safe travel much above 50-60 km/h, yet are commonly zoned at 100 km/h, The high crash and injury risk to pedestrians associated with travel speeds above about 40 km/h (much lower speeds apply for older pedestrians, for example, due to their greater frailty), and The crash and injury risk associated with driver/rider behaviour at intersections, particularly the ability to (or willingness to) respond to changing traffic signals in high-speed zones. Traffic calming measures on approaches to and within rural towns

3.3.2

In addition to setting appropriate speed limits, there are a number of other ways to achieve speed reduction in hazardous locations. For instance, several traffic calming countermeasures have been suggested to induce more moderate speeds on the approach to and within towns. Traffic calming techniques have been used extensively in urban areas to reduce the number and speed of vehicles in local streets and in areas where there is high pedestrian activity. They act to make drivers more attentive to their surroundings and drive more slowly or appropriately for the surroundings and this created safety, equity and an enhanced environment. One such concept is the Woonerf concept developed for use in urban areas in The Netherlands whereby drivers are encouraged to drive slowly by installation of roundabouts, speed humps, width restrictions and gateway treatments. Although traffic calming countermeasures have not generally been considered for rural settings, there may be s cope for these techniques to reduce speeds on the approach to and through rural townships. The following techniques have been considered for these environments: Pavement narrowing one approach for improving safety on these roads is the use of narrower lanes. This countermeasure can be implemented either by physically creating narrower travel ways or by visually decreasing the available width. The more widely used techniques in optically narrowing the road are painting wider edge lines or eliminating centre-line striping. It should be noted that this optical narrowing is a concept more often used for low-volume rural roads than roads that carry higher volumes of traffic (TranScan, 2001). Installation of refuge islands refuge islands that are constructed i n the middle of the road may slow motorists as they approach towns by creating a change in the physical environment. These features allow the approaching motorist to clearly see a constructed physical element in the middle of the road from far away. Previous research has shown that refuge islands with plants are most visible to motorists. Speed reduction is forcibly achieved at the refuge island. This is because of lane width reduction (in the order of 3 to 3.5m) and the resulting curve in the carriageway.

76

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

It is noted, however, that lane width reduction alone does not appear to affect speed, it is the extent to which the refuge island laterally diverts the lanes that has a significant effect on drivers speeds. For example, if the lane is designed to cause sharp, lateral deflection in vehicle paths, definite steering movements are necessary to drive through the structure. Alteration of the road surface this countermeasure is also used to achieve a perceived optical narrowing of the road. Raising of the road surface the road surface can be raised by installation of speed humps. While speed humps are effective in reducing speed, care should be taken in designing speed humps in a rural environment because problems may rise if the ramp is too steep. Indeed, Pyne et al. (1995) noted that traffic calming techniques that have been developed for urban areas (such as speed humps and chicanes) had the potential to cause more crashes on rural roads because of the consequent radical vertical deflection of vehicles with the potential for drivers to lose control of their vehicle while driving over a hump at high speed. Pyne et al. (1995) suggested that alternative approaches such as narrowing of carriageways or removing lane space from the edge of the road, various module lengths and mark gap ratios for edge and centre lines, transverse lines, etc. may be more effective than installation of speed humps. Alternatively, road authorities in Germany have begun installing raising of the road in asphalt, where a raising of the road surface made of asphalt at the narrowing is intended to ensure the reduction of speed.

While many of these techniques are fairly well established and widely used in urban areas, there has been little scientific evaluation of their effectiveness in altering behaviour or improving safety, particularly in rural settings. One study in the district of Neuss, Germany, assessed the effectiveness of traffic calming measures on the approaches to 13 rural towns by examining crash occurrence, speed behaviour of motorists, and residents opinion before-and-after installation of treatments (Steinbrecher, 1992). In the first phase of the reconstruction, the widths of the lanes were reduced to approximately 5.5 m, and strips of pavement stones were positioned to optically narrow the road further. Refuge islands were included in the second phase, as well as raising the road in asphalt. The raised road sections were signed and had a horizontal plane with ramps of 8 cm on a length of 3 m with an inclination corresponding to a proportion of about 1:35. The raisings could be negotiated at 50 km/h without difficulty. Overall, the results were encouraging, with a reduction in the number of crashes after installation of treatments at all 13 town entrances. Specifically, Steinbrecher (1992) reported that prior to the reconstruction the average crash rate per year was 1.1, and following the reconstruction the average crash rate decreased to 0.6. Furthermore, it should be noted that traffic density at the town entrance increased up to ten percent per year. Therefore, Steinbrecher (1992) concluded that on the whole, the restructuring measures were effective in increasing road safety. In terms of speed reduction, average speeds driven on these roads were between 75 and 85 km/h. After installation of treatments, speed reduction occurred in every town, with average speeds reduced to between 60 and 75 km/h. Steinbrecher (1992) noted that speed reduction varied across the sites examined between 2 and 14 km/h, with an average reduction of 9 km/h. Most importantly, Steinbrecher noted that the most significant innovation in the reconstruction of these roads was the road raising aspect and, in
COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS 77

combination with installation of refuge islands, achieved the greatest effect of speed reduction. Similarly, Pyne et al. (1995) evaluated the effectiveness of speed-reducing measures on the approaches to villages in England by measuring 85th percentile speeds and speed variance in a simulated road environment. They found that the most effective treatments were hazard marker posts, countdown signs, Wundt illusion and chicanes. They also noted that installation of transverse lines resulted in a reduction of 85th percentile speed, however, this was not effective in reducing speed variance. Moreover, they found that the 85th percentile speed was not maintained through the village. They also noted that many of these measures could be combined because they are not located in the same places relative to the carriageway. Possible combinations include hazard marker posts, countdown signs, and chicane, to which the Wundt illusion or transverse lines could be added. While they argued that such a combination may exploit the strengths of all the measures, this was not tested. Al-Masaeid (1994) investigated the effects of the existence of towns on or near two-lane primary rural roads, examining the effects on both the residents of these towns and the through-traffic. He found that crash rates, in general, were affected by the number of horizontal and vertical curves, the speed of through-traffic, the number of paved accesses and employment level within the towns. With regard to the effect of through-traffic speed, factors such as the existence of a town along/adjacent to a primary rural road, the length of its frontage, the spatial distribution of the town around the road, the town population and the towns setback from the road all significantly influenced through-traffic speeds. From this analysis, Al-Masaeid (1994) recommended a number of safety precautions be developed for pedestrian crashes and for traffic crashes in general in rural towns. For new towns, it was recommended that towns be located entirely on one side of the primary rural road, and set back from the roadway to reduce the possibility of conflict with throughtraffic. This strategy was especially relevant for locating schools within the residential areas of such towns. Only one or two direct accesses between the road and the town should be provided. In relation to existing towns, several measures were recommended to reduce crash risk, especially for pedestrians, and to minimise the reductions in the speed of through-traffic. Measures included: restrictions of frontage distance of individual properties; closing or redirection of direct accesses, together with the use of parallel service roads; prevention or restriction of future development on the side of the roadway with lesser development; introduction of speed reduction measures; relocation of schools close to the roadway; and, finally, both improvements to sight distances and widening of shoulders throughout town frontages. Tziotis (1993) concluded that several road design elements failed to respond to the conflicts experienced by motorists in partially developed outer urban areas. The most pronounced deficiencies of the road design within a transitional environment related to: inadequate curve delineation, turning lanes not available, the absence of acceleration/deceleration lanes, inadequate shoulder seals and poles located in hazardous positions. He recommended appropriate measures designed to improve the safety on roads in this environment (i.e., protected turning lanes, acceleration/deceleration lanes, improved shoulder sealing practices, increased curve delineation, and better pole placement/protection).

78

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

3.3.3

Perceptual countermeasures to reduce speeding

Perceptual countermeasures (PCMs) have been identified as potentially effective treatments to reduce speeding in rural areas. PCMs refer to low-cost road or road setting treatments that are likely to influence travel speed by motorists by altering how drivers perceive the road or roadside. As indicated earlier, there are a number of factors that influence speed choice, and the physical road environment and road surface represent a primary source of information for speed perception on which drivers choose a safe and appropriate speed. Relative coding of the moving environment on the retinal surface of the eye is an extremely important cue for the perception of speed (Calvert, 1954; Gibson, 1950, 1 958, 1968; Harrington, Harrington, Wilkins & Koh, 1980; Lee & Lishman, 1977; Moore, 1968). PCMs aim to modify the relatively subconscious visual information arriving at the drivers eye. An early study of the perception of speed on urban and rural roads by Fildes, Fletcher and Corrigan (1987) showed there was potential for manipulating the environment to produce changes in speed behaviour. In particular, it was argued that changes to the road surface have the greatest potential for influencing a driver's perception of speed, although roadside environment effects can also influence speed perception. The effectiveness of the treatment, however, was likely to be a function of the overall level of safety a driver perceived for a particular road and environment location. Denton (1973) conducted an early experiment in England, showing the effectiveness of transverse lines on the approach to an intersection and Helliar-Symons (1981) followed up with an evaluation of a number of yellow-bar carriageway marking installations in that country. Agent (1980) and Agent and Creasey (1986) reported on driver speed effects from transverse pavement markings and from improved delineation of horizontal curves in the USA. De Waard, Jessurun, Steyvers, Raggatt and Brookhuis (1995) developed and evaluated a trial section of road in the Netherlands (the Drenthe treatment) comprising a visual and tactile treatment on an 80 km/h section of highway. Others (e.g.: Emerson & West, 1985; Enustun, 1972) have reported speed reductions from transverse pavement stripes and rumble bars on the approach to a hazardous location and on bridges. Lum (1984) showed the benefits of using road markings on narrow lanes for controlling speed in residential areas while Hall (1991) listed a number of innovative treatments to counter runoff-the-road crashes in the USA. In their review of European procedures and practices in roadway geometric design, Brewer, German, Krammes, Movassaghi, Okamoto, Otto, Ruff, Sillan, Stamatiadis and Walters (2001) noted that countries such as Denmark, Sweden, The Netherlands, Germany and England relied on the physical roadway design to enforce operating speeds. They noted that this philosophy could be considered as a speed management approach in which the objective is not simply to reduce speeds but to provide a roadway planned and designed in such a way that an appropriate speed is obtained. They also noted that this approach allows the establishment of speed limits close to the expected operating speeds, thus avoiding higher travel speeds. 3.3.3.1 The unobtrusive nature of PCMs Unlike conventional enforcement or traffic management approaches to speed reduction, PCMs attempt to bring about a change in behaviour unobtrusively (i.e., without the driver being aware necessarily of any change in his or her behaviour). This more subtle approach to speed control has several advantages over the traditional methods. First, by influencing

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

79

the visual information on display to the driver, it is attempting to address the underlying problem. If drivers subconsciously perceive a particular road situation to be safe, then applying cognitive restrictions (involving conscious thought processes) will have only a marginal effect on their behaviour. This is evident from the fact that police enforcement is only effective in safe environments, for the most part, as long as the deterrent is obvious to the driver. In addition, modifying the perceptual environment is less likely to annoy or frustrate the driver and, therefore, more likely to be of benefit long-term. A change in the visual input to create an illusion of less safe will probably go by unnoticed and, therefore, not lead to crash migration by forcing speed deviants onto other roads. It could be argued that subtle changes to the road or the environment may be the only effective long- term means of influencing drivers who blatantly refuse to obey the law by driving at excessive speeds. In any event, removing restraints which people believe to be unnecessary should result in safer driving for the total driving population. Finally, PCMs, by definition, do not involve introducing additional hazards on the roads in the same way as LATM devices have in the past. Most of these treatments simply involve painted lines or additional plastic or gravel surfaces applied to the road surface to create the desired effect. Apart from the obvious road safety benefit, this can also mean that the measure is likely to be relatively inexpensive, may be easier to justify in terms of cost/benefit effectiveness, and may enable more treatments per budget than other methods. 3.3.3.2 PCMs treatment evaluation MUARC conducted an evaluation trial of a range of PCMs to speeding using a driver simulator (Godley, Fildes, Triggs and Brown, 1999). Treatments tested included transverse road markings, lane-edge and herringbone markings, the Drenthe treatment from the Netherlands, centre-line and other edge-line markings, and several enhanced curvature treatments. Drivers drove a series of test tracks in the simulator that had been previously validated for use in this environment (Fildes, Godley, Triggs and Jarvis, 1997). They reported that a number of these treatments were effective at reducing travel speed in the simulator, namely: full-width transverse lines, peripheral transverse lines and lane edge herringbone treatments, hatched median treatments, especially with lanes narrower than 3 m, and with and without intermittent gravel edge-lines, enhanced post spacings for curves, with and without ascending height posts.

Subsequently, two of these treatments have been applied at a number of sites in Victoria and New South Wales in Australia to assess their on-road effectiveness. These treatments include peripheral transverse lines on the approach to intersections and enhanced post spacings for curves with ascending height posts (see Figures 1 and 2).

80

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

Figure 1: Typical curve treatment

Figure 2: Typical intersection treatment

A preliminary evaluation showed some benefits in reduced speed immediately after treatment and 12 months later compared to matched control sites (Macaulay, Tziotis & Fildes, 2002). More detailed analysis is currently underway on the level of effectiveness for these treatments. 3.3.3.3 Limitations with PCMs While PCMs may have several advantages over other forms of speed control, it would be unrealistic to expect these treatments to solve all speeding problems on the road. The different findings between the perceptual results and the free speed measurements in Fildes et al. (1987), as well as the more recent study reported here, show the subtle relationships that exist between speed perception and behaviour, and especially the likely limitations in the effectiveness of PCMs.

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

81

When drivers' perceptions of speed at particular road sites were in the too slow range of the response scale (i.e., drivers generally felt quite safe), free speeds at these road sites tended to be above the speed limit, but the pattern of the results were generally less sensitive to the effects of the independent variables. However, when perceptions were less safe (responses were in the too fast range of the scale), free speed effects almost mirrored the perceptual effects. This suggests that countermeasures aimed at reducing the driver's perception of safety at particular road sites may only be effective if perceptions of speed are not overly safe to begin with. More work is still required to determine the precise speed reduction that can be attributed to these low cost road treatments. However, preliminary findings suggest that carefully selected treatments can lead to speed reductions of up to 5% in some locations. While these changes may appear to be small, nevertheless, they can have significant crash reduction and injury mitigation benefits. These treatments therefore do offer road designers and traffic engineers another approach to speed reduction for existing roadways and at high crash locations. 3.3.4 Technologies to enhance speed limit compliance

Corben, Lenn, Regan and Triggs (2001) reviewed the technologies available world-wide to enhance speed limit compliance. There are a number of out-of-vehicle and in-vehicle technologies with the potential to enhance speed compliance. This review focused on the out-of-vehicle approaches. Traditional speed limit and warning signs tend to be static, which can reduce their impact on road users. Dynamic messaging, in the form of speed warning signs, variable message signs, or a combination of the two, is more vehiclespecific and so has the potential to have a much greater effect on the driver. Active speed warning signs are predominantly used to make the driver aware of his/her own speed as a means of reinforcing the speed limit. Speed Indicator Displays (SIDs) were noted by Corben et al. (2001) as a promising technology to reduce vehicle speeds. The authors reported that such technology has been shown to reduce average vehicle speeds in European trials by between 4 and 8 km/h. SIDs can record the number and speed of vehicles, as well as vehicle headways for analysis, and data can be retrieved remotely. While mostly used to display drivers speeds, permanently mounted SIDs have also been used to display mandatory speed limits up to 90 km/h. SIDs are primarily used in permanently mounted positions. Detector loops are embedded in the roadway 70 metres before the SID. This provides a firm distance from which to calculate the leeway time, which is the time between detection of the vehicle on the detector loops and when the vehicle passes the SID. Unlike radar technology, the detector loops will only detect each vehicle once and so is more suited to displaying the speed for individual vehicles. More portable trailer-mounted SIDs can also be deployed for temporary use such as during periods of road works, although use of radars rather than embedded detection loops is recommended for ease of temporary installation. The BCRs for using SIDs in Victoria (Australia) were predicted for a range of common scenarios in Victoria, namely, Case 1 which involved a National Highway on the outskirts of the Melbourne metropolitan area; Case 2, a State Highway passing through a township in regional Victoria; and Case 3, a State Highway passing through a large provincial city. Baseline crash data, together with the results of research on the relationship between crash risk, personal injury consequences and travel speed (Nilsson, 1984), for these sites were used to predict reductions in crash severity resulting from reduced vehicle speeds.

82

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

These calculations showed that very attractive BCRs can be expected from reductions in vehicle speeds in both urban and rural settings, through the use of SIDs in such circumstances. The benefits were also considerably greater for Case 1, by virtue of the fact that a much higher number of crashes had occurred along that length of roadway. The resultant BCRs for 4 km/h speed reductions were 45.6 (Case 1), 7.9 (for Case 2), and 7.7 (Case 3). If the 8 km/h reduction were assumed, the BCR values approximately double. 3.4 TREATMENTS TO IMPROVE THE SAFETY OF ROAD LENGTHS

Generally, motorways and dual-carriageway roads experience low crash and injury risk (per unit of exposure). In contrast, single-carriageway roads experience high levels of severe injury crashes, particularly those resulting from run-off-road and head-on crashes. The high crash and injury risk on these roads has been attributed to low design standards, a lack of median separators, poor road surface, and inadequate provision for safe overtaking manoeuvres. In this section a number of countermeasures to improve the safety of road lengths are discussed. 3.4.1 Provision of additional lanes

It could be argued that the head-on collision problem is under control on divided roads because traffic directions are separated and there are provisions for safe overtaking manoeuvres (Toivonen & Niskanen, 1998). In contrast, the problem of head-on collisions on single-carriageway undivided roads is high because there is nothing separating driving directions and any vehicle encroachment onto the opposing lane (either during an overtaking manoeuvre or loss of vehicle control) may result in a head-on collision. One solution is to provide additional lanes or widen roads. This provides increased capacity by reducing congestion and delay. Moreover, added lanes provide safety benefits by improving the quality of service, i.e., they enable safe overtaking manoeuvres, avoid disruptions to through-traffic by separating slowing turning traffic, reduce driver frustration and stress, and provide more room for drivers to re-gain control of their vehicle in the event of loss of control. There are a number of configurations identified in the literature that provide additional lanes including conversion of two-lane undivided roads to four-lane undivided roads, provision of 2+1 roadways, in which the erection of a (flexible) mid-barrier to separate high-speed, opposing traffic is the principal treatment, and provision of overtaking lanes at selected locations. The FHWA (1992) noted that the total width of the roadway, including travel lanes, shoulders and medians, is one of the most important cross-section considerations in the safety performance of two-lane rural roads, noting that wider roadways generally result in fewer crashes. They evaluated the safety effects of design alternatives (passing lanes, short four-lane sections, shoulder use sections, turnout lanes, and two-way left-turn lanes) for use along two-lane undivided highway sections. The evaluation noted the potential for wide variation in crash experience as a consequence of differing site and/or traffic characteristics, as well as the possibilities for creating new safety problems as a result of treating two-lane highways with generally low-cost design alternatives. Notwithstanding these cautionary remarks, they found that two-way left-turn lanes reduced the number of fatal and injury crashes by between 70 and 85%. The use of short four-lane sections and turnout lanes were each found to cut fatal and injury crash numbers by 40%, while for passing lanes, the corresponding reduction was 30%. Shoulder use sections showed no statistically reliable effect on crash reductions. They also found crash reductions in related

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

83

crash types (i.e., run-off-road, head-on, and opposite and same direction side-swipe crashes) ranging from 12, 23, 32 and 40% for corresponding i ncreases in lane width from 10 feet (3 metres), in one, two, three and four foot intervals. VTT (1992) conducted a study to investigate how the expected number of crashes on a highway section depends on the road and traffic characteristics of that section, with particular interest in the possible effects of private road junctions, pavement width and road geometry. The results indicated that the effect of hilliness seems to depend on the paved width of the road. That is, for narrow roads, the crash rate increased slowly with increasing hilliness, however, for wider roads, the crash rate was much more pronounced. More specifically, when the hilliness was more than 20 m/km, crash rates were higher on wide roads than on narrow roads. The authors suggested t hat this result could be connected with the relationship between overtaking behaviour and road width in different road alignments (see Section 3.4.1.3 on the provision of overtaking lanes). 3.4.1.1 Conversion of two-lane undivided roads to four-lane undivided roads Conversions of undivided two-lane roads to undivided four-lane roads are usually undertaken to improve the flow of increased volumes of traffic. Such conversions should also, intuitively, attract safety benefits. For instance, multiple-lane roads provide continuous opportunities for safe overtaking and can therefore have a large effect on traffic operations by decreasing the interactions between faster and slower vehicles. Despite this, there are surprisingly few studies that have attempted to estimate the safety benefits of this treatment, moreover, there is little convincing evidence to suggest that this results in crash reduction. Indeed, Council and Stewart (1999) predicted crash reductions for the conversion of a two-lane undivided road to a four-lane undivided road, ranging from no effect to 20%. Rogness, Fambro, and Turner (1982) used a before-and-after study design to assess the effect of converting a two-lane rural road with paved shoulders to four-lanes without paved shoulders (i.e., converting the shoulders to travel lanes) in Texas. A total of 60 sites were evaluated, and crash data at each site were recorded over a four-year period (two years before and two years after the conversion). The sites were divided into three AADT categories low volume roads (fewer than 3,000 vehicles per day), moderate volume roads (3,000-5,000 vehicles per day), and high volume roads (5,000-7,000 vehicles per day). The only significant finding for total crashes was on moderate volume roads, in which total crashes decreased by 9.1%. For non-intersection crashes only, significant decreases of 19 and 28% were observed for moderate and high volume roads, respectively. Interestingly, they found a significant increase of 12.6% in non-intersection crashes on low v olume roads and it could be argued that the addition of lanes on low volume roads resulted in higher travel speeds and consequently increased the potential for crash occurrence. 3.4.1.2 Implementation of 2+1 roads The conversion of long stretches of two-lane road lengths to four-lane road lengths is an expensive measure. Other treatments are also available that can be more economical and provide similar safety benefits, one of which is the implementation of 2+1 roadways. To simultaneously improve capacity and safety on the rural network, Sweden, Denmark, England and Germany have begun to implement the Swedish concept of 2+1 roadways instead of 4-lane carriageways on high volume rural highways. This design has been shown to be more economical than conversion to four-lane roadways and effectively reduces the

84

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

severity and incidence of crashes related to overtaking. In this configuration, a passing lane is created in the centre of the road between the opposing travel lanes, and is used as a passing lane that alternates about every 1 to 1.25 km between the two directions of travel. Such roadways can be built within existing roadway rights of way. For example, in Germany, the passing lane is 3.25 m wide, the adjacent same-direction travel lane is 3.5m wide, and the opposing-direction travel lane is 3.75 m (paved shoulders on both sides of the road). All four countries have reported capacity gains and safety improvements associated with conversion to 2+1 roadways (TranScan, 2001). Each country has customised this design to conform to its design guidelines and safety goals, including the use of varied roadway widths, lengths of passing lanes, flexible barriers on medians, and end-treatment of passing lanes (Bergh & Carlsson, 2000). For example, Sweden and England have retro-fitted their roads (within the same right-of-way) to wider cross sections of 13 m and 12 m respectively. The passing lane has a width of 3.5 m and the remaining travel width is evenly spread between the other two lanes. In Germany, a slightly wider right-of-way is used and a 15.5 m cross-section is utilised. In Sweden, a cable barrier is used to separate the two directions of travel. Capacity gains and safety improvements associated with conversion to 2+1 roadways have been reported by TranScan (2001). 3.4.1.3 Provision of overtaking lanes An alternative, potentially less expensive solution to added lanes on long road stretches, is to provide overtaking lanes at selected intervals (Harwood, Hoban & Warren, 1988). While overtaking lanes are generally provided on unsustainable grades where some vehicles experience a problem maintaining desired speed, the provision of overtaking lanes is becoming more common at locations other than on unsustainable grades where maintaining desired speed is not a vehicle performance characteristic. Several studies in the USA have assessed the safety effects of overtaking lanes. A Californian study in the 1970s, reported by Kaub and Berg (1988), examined the conversion of a two-lane rural highway to a three-lane highway incorporating an overtaking lane. Before-and-after comparisons indicated that fatal crashes were reduced by about 60%, and injury crashes and property damage crashes each by about 20%. Based on this study, Kaub and Berg (1988) estimated that, if overtaking lanes were constructed on all US two-lane primary highways, the total cost of overtaking-related crashes on such highways would be reduced by an estimated 38%. McLean (1996) also reported on the findings of US studies of the effects of overtaking lanes on crash rates. The US studies found reductions in crash rates of 25-38% when overtaking lanes were constructed along two-lane highways. McLean believes that these estimates may actually understate the true safety benefit of overtaking lanes. Harwood and St. John (1985) examined the effectiveness of adding short four-lane passing sections to two-lane facilities at ten locations, using a comparative evaluation of case and control sites in which the former were the short four-lane undivided road segments and the latter were the sections of two-lane road immediately before-and-after the treated segment. They found a non-significant decrease of 34% in total crash rate, a significant decrease of 50% in cross-centre-line crashes and a non-significant decrease across all other crash types. Importantly, there was no increase in crash rates for any treated section of the roadway. Before-and-after data were available for only one site, which showed an overall crash rate decrease of 40% after the lanes were added a reduction from 1.34 to 0.81 crashes per

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

85

million vehicle kilometres of travel. While these reductions appear high, it should be noted that these results were based on relatively small samples of locations. Fildes et al. (in press) also noted that the provision of passing zones can alleviate some of the problems associated with overtaking crashes on two-lane roads and added that the geometric features of overtaking lanes should accommodate the needs of all drivers, including older drivers. Some of their recommendations on how these crashes can be reduced or avoided are: Passing zones should be provided frequently and regularly on rural roads, with a maximum of 5 km intervals between passing zones for each direction of traffic. A minimum passing zone length of 500 m should be provided to allow those drivers who drive slowly, require longer headways, accept longer gaps and/or require more time to make decisions (e.g., older drivers) to pass a slower vehicle safely. It is essential that oncoming traffic can be seen clearly from a maximum distance to safely overtake slower vehicles, both at merge and diverge tapers of passing zones. At the merge, a lack of clear visibility of the actual merge can lead to conflicting behaviour with slow vehicles merging too soon and faster vehicles attempting one more pass. At the diverge, good visibility provides an opportunity for the following drivers to move up and prepare to overtake as soon as the added lane is available. To increase safety of these manoeuvres, a minimum continuation sight distance of 300 m should be provided for roads with a speed zone of 60 km/h. Conversion of undivided roads to divided roads

3.4.2

The provision of added lanes has been shown to generally improve safety on rural roads, however, they have limited potential to reduce conflicts between directions of traffic in the event of opposing lane encroachments. One way of reducing severe head-on collisions on rural roads is to physically separate driving directions from each other by converting undivided roads to divided roads. The most common form of traffic separation is the provision of a median. Medians are used on both rural and urban highways to physically separate opposing traffic streams, improve traffic safety by reducing the potential for head-on collisions and provide an area for errant or out-of-control vehicles to recover before entering oncoming traffic lanes (Balke & Fitzpatrick, 1993; Toivonen & Niskanen, 1988). Depending on their width, medians can also improve safety by reducing the headlight glare, and providing an area out of the traffic stream for disabled vehicles to stop in case of an emergency. Further, median areas designated for turning traffic can be used to accelerate or decelerate when turning either onto or off of a highway. Council and Stewart (1999) examined the safety effects of conversions of two- to four-lane and divided to undivided roads and reported differences in crash rates in four US States (California, Washington, Michigan, and North Carolina). Predicted crash reductions for conversion from most typical two- to four-lane divided sections ranged from 40 to 60%. In their examination of safety on dual- and single-carriageway roads, Walmsley et al. (1998) found that dual carriageways with a median safety fence had 57% fewer access crashes than dual-carriageways without a median safety fence and 33% fewer crashes which involved a vehicle crossing the central reservation. These findings are consistent

86

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

with other findings that reported that dual-carriageways with a median safety fence along most of their length would be approximately three percent safer than those without (Walmsley & Summersgill, 1998). However, it should be noted that median safety fences are usually located in places where the consequences of a vehicle crossing the central reservation would otherwise be particularly severe, due to the presence of hazards and obstructions. 3.4.2.1 Median type There are several types of median that can be installed on rural highways, and the type of median used depends on a number of factors, including the amount of traffic, the functional classification of the highway, the availability of right-of-way, the operating speed of the highway, and the type and intensity of the development alongside the highway. The first type is a median island and this is generally used on higher functional classes of highway, where it is desirable to maintain as little interruption to the movement of traffic as possible (Balke & Fitzpatrick, 1993). This treatment uses a physical island to separate opposing directions of traffic and greatly enhance the safety performance of these roads, particularly in combination with barrier fencing or guard-rails. Median widths can range from as little as 4 ft (1.2 m) in highly developed areas, where right-of-way is extremely limited, to 76 ft (23.8 m) in suburban and rural areas, where right-of-way is typically less constrained. Turning access is usually controlled through the placement of established breaks in the median and, usually, with the provision of added turn lanes. The US Federal Highway Administration (FHWA, 1992) noted that median width, slope, type (raised or depressed) or the presence or absence of a median will affect crash frequency or severity, including the potential for head-on crashes, roll-over crashes and single-vehicle crashes into fixed objects. While they noted that installation of median barrier systems will most likely increase the incidence of crashes (into median barriers), they emphasized the beneficial effect of median barriers on reducing crash severity, especially for head-on crashes. Moreover, while they pointed out that wider medians reduce the risk of head-on collisions, there are usually practical constraints on the available right-of-way. The safety performance of mound medians with swale medians was compared in this study and it was found that mildly depressed medians provided a greater opportunity for encroaching vehicles to return safely to the roadway. The FHWA noted: Similar safety performance can be expected from wide medians (i.e., 84 ft or 26 m) with mildly depressed profiles (i.e., depressed by 4 ft [1.2 m] with 8:1 down-slopes) as from mound medians with 3:1 up-slopes, For narrower medians (i.e., 20 to 30 ft, or 6 to 12 m), slopes of 6:1 or flatter are important for safety, and There is a higher incidence of overturning crashes where deeply depressed medians with slopes of 4:1 or steeper are provided.

In conclusion, the FHWA (1992) noted that crash relationships for medians of 20 ft (6 m) or less are not well defined, however, it is clear that wider medians are generally safer, with widths of 60 to 80 ft or more, with flat slopes being considered where feasible. They

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

87

recommended installation of medians of at least 30 to 40 ft (9 to 12 m) and slopes of 6:1 or flatter. Other median treatments include the addition of two-way left-turn lanes in the centre of the road (i.e., left-turn in the USA and most European countries and right-turn in countries such as England, Australia and New Zealand) to maintain low levels of interruption to traffic flows and to facilitate safe turning operations. While these can also be considered as intersection treatments and are discussed in more detail in Section 3.7.1, some argue that these act as an adequate form of separating traffic flows. Flush medians are another type of median that combine many of the attributes and features of median islands and added turning lanes in the centre of the road. With the flush median design, the area between the travel lanes is at-grade. The median is marked with either a single solid or double solid (yellow) line. Left turn access to adjacent properties is provided at left-turn bays that have been striped at established locations. Since the median area is atgrade, it can easily be traversed by drivers turning into and out of adjacent accesses (Balke & Fitzpatrick, 1993). Balke and Fitzpatrick (1993) examined the differences in the operations and safety between four-lane rural highways on approaches to towns marked with added turning lanes and flush medians. They compared total crash rates, mid-block crash rates and crash severity over a three-year period and found no difference in crash and severity rates for highways with added turn lanes and flush medians when access densities were low (i.e., less than nine accesses per mile [14.5 accesses per kilometre]). Due to the limited number of flush median sites, however, the authors conceded that it was not possible to determine if there is a difference in safety between the two median treatments at higher levels of development (i.e., with access densities greater than nine accesses per mile [14.5 accesses per kilometre]). Additional field studies also showed no difference in the way that drivers used highways marked with added turning lanes and those marked with flush medians. At both treatment types, drivers used the added lanes as storage and acceleration lanes, suggesting that flush medians and added turning lanes function in a similar way on four-lane highways. 3.4.3 Improved road surface

The road surface itself can affect crash risk and there are a number of treatments to pavement conditions that can reduce crash and injury risk. Skid resistant paving has been shown to be successful in improving safety (both in terms of crash occurrence and severity of injuries) on both urban and rural roads (Cairney, 1997; Andreassen, 1992). It has the ability to reduce the braking distance of vehicles, thereby reducing the probability of colliding with other road users, as well as reducing the impact speed if a collision occurs (de Silva, 2001). Previous studies have reported that approximately eight percent of crashes in dry weather involve skidding, whereas the corresponding figure for wet weather is 27% (PPK, 1998). One of the major causes of crashes on rural roads is fatigue and a substantial amount of the safety literature has addressed ways to reduce the frequency of crashes caused by drivers falling asleep at the wheel. Yamashita and Takata (1991) noted that this problem manifests itself on specific sections of inter-urban motorways that often have a moderate down slope or are moderately curved. In an attempt to wake drowsy drivers, wake-up pavements were installed on sections of the motorway so as to change the sound and

88

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

vibration of the tyres as the driver travels over them. Yamashita and Takata reported that after the introduction of wake-up pavements, there was a drastic reduction in the number of crashes that were due to drivers falling asleep at the wheel, and in some cases the number of crashes was reduced to zero in sections that had previously had two to three crashes a month. While many have suggested that unsealed or rough roads increase crash risk, Al-Masaeid (1997) found a reduction of single-vehicle crashes (along with an increase of multi-vehicle crashes) on rough roads and suggested that smooth roads may encourage higher speeds, and therefore increased potential for run-off-road loss of control type crashes. To overcome this inconsistency, Al-Masaeid (1997) recommended that unsealed roads should be sealed, but also recommended that some innovative techniques be adopted to control speeds on smooth surface roads. Furthermore, he suggested that roadsides should also be improved to minimise single-vehicle crashes such as flattening of roadside slopes and relocation of roadside objects or obstructions (see Section 3.6 for details of improvements to roadsides). He also suggested that designers of rural roads should minimise the number of sharp horizontal curves and the number of vertical curves with insufficient sight distance. In a study that aimed to develop and trial some low cost treatments to improve safety for motorcyclists on Victorian roads, Pronk and Corben (1999) conducted a national and international review of existing and innovative countermeasures. Specifically, the study aimed to prioritise and recommend three innovative countermeasures, mainly involving road engineering, for possible implementation as a trial. In order to systematically identify priority countermeasures for implementation, a matrix was developed, linking common motorcycle crash types with measures assessed as having the potential to reduce the incidence and/or severity of each crash type. Five priority countermeasures were identified, each of which has potential for application along rural roads. They are listed below (Table 3), together with an estimate of the proportion of all reported motorcycle-involved casualty crashes that can potentially be addressed by the countermeasure. Table 3: Prioritised treatments to improve safety for motorcyclists Countermeasure Skid-resistant surfaces Shoulder sealing Tactile centre-lining Reduction of speed limits along high-risk sections of road Speed cameras targeting motorcycles Potential Crash Target (% of all Mcycle Crashes) up to 58% up to 63% up to 44% up to 100% up to 100%

Pronk and Corben (1999) did not indicate the expected effectiveness of these measures specifically on motorcycle crash or injury risk. They did, however, cite the results of other studies which have found or have reported reductions of between ten percent and 60% in all crashes as a result of skid-resistant surfaces, and between 23% and 50% reduction in all crashes from shoulder sealing in combination with edge-lining. No estimates are available

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

89

for the use of tactile centre-lining, as it was proposed as a new measure, yet to be trialled in such a way. 3.4.4 Improved delineation

Delineation is an important component of the geometric design feature of roads, which assists drivers manoeuvre their vehicles along road lengths and through intersections safely. It consists of a system of reflectors, posts and painted lines along the road network (Cairney, 1995). For instance, painted edge-lines and reflectorised posts delineate the edge of the travelled way and are marked to discourage travel on road shoulders and to make driving safer and more comfortable. Painted edge-lines also act as a marker to enhance the conspicuity of hazards and as a guide past objects that are close to the pavement edge that may constitute a hazard. As for delineation of road edges, painted delineation of raised surfaces such as traffic islands, kerbs and medians is also an important roadway feature to assist drivers to visually separate these features from the roadway and to manoeuvre vehicles safely. Centre-lines, too, help oncoming traffic to avoid collisions (Miller, 1993). Double centre-lines prohibit overtaking. Further, reflectorised pavement markings improve visibility particularly at night and convey a warning message to the driver without diverting his/her attention from the roadway (Al-Masaeid & Sinha, 1994). While the importance of good delineation is often emphasized in the safety literature, the effectiveness of pavement markings, particularly in terms of conspicuity and line width, however, is not clear-cut. One of the earliest studies investigating the effectiveness of pavement marking on two-lane rural roads in the USA was conducted by Bali and colleagues (1978, cited in Miller, 1993). In this ten-State study, Bali et al. included more than 500 sites, with each site having either delineation that remained unchanged and adequately maintained over two to three years, or having no delineation at all (matched control sites). Data were obtained on crash experience for two to three years at each site. The authors reported that the addition of edge-lines and centre-lines reduced the average crash rate by 36%. A recent study (McKnight, McKnight & Tippetts, 1998) examined the combined effect of lane line width and line-pavement contrast upon lane-keeping in terms of heading error, position error, lane excursions and road excursions in a simulated driving environment. They found that lane line width and contrast had large effects upon lane-keeping performance at extremely low levels of contrast, with many more errors made with narrow and low contrast lines compared to wide and high contrast lines. They also noted that low contrast levels are present at night, on wet roads and when light snow covers the pavement adding to the risk of crashing. In contrast, Glennon (1979) found that centre-line markings were not cost-effective on rural roads with traffic volumes below 300 vehicles per day. In a later study, Glennon (1985) indicated that dashed centre-line markings may only be beneficial with respect to traffic safety for pavement widths of 20 ft or more and traffic volumes of 500 or more vehicles per day. Moreover, studies conducted by the FHWA (1981, 1982) provided mixed results. They found that, for all AADTs on rural roads and for a range of lane widths, the derived crash reduction factors due to pavement markings varied from 13% to +35% (note that the negative sign indicated an increase in crash experience after implementation of markings). Al-Masaeid and Singa (1994) used a probabilistic procedure to estimate a crash reduction factor associated with pavement marking on undivided rural roads having traffic volumes

90

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

in the range of 1000-4000 vehicles per day. They found that, when all pavement marking sites were considered, pavement marking was not effective in reducing crash rates, with a small reduction rate of about 3.4%. However, the authors did concede that pavement marking of hazardous sites provided a significant crash reduction factor of 13.5%, with the expected range of 6.5% and 21.5%. This is not an unexpected finding, given the fact that the hazardous sites had a relatively high crash history, with a substantial proportion of crashes most likely resulting from bad visibility. Al-Masaeid and Singa concluded that the marking of hazardous sites is warranted, and recommended pavement markings for rural undivided highways that have an average daily traffic volume in the range of 1000 4000 vehicles per day. They also suggested that future studies should investigate the impact of pavement markings on crash severity at different traffic levels. Miller (1993) conducted a benefit-cost analysis of edge-lines, centre-lines, and lane-lines, considering the applications of fast-drying paint or thermoplastic which are the most frequently used marking materials in the USA. Miller reported that striping with fastdrying paint costs $0.035/linear-ft in rural areas and $0.07 linear-ft in urban areas. He added that while the application of thermoplastic lines initially costs more than painted lines, they can have lower life-cycle costs and, more importantly, in areas where snowploughing is unnecessary, they have longer lives. Miller (1993) also noted that previous studies report crash reductions of 21% with existing longitudinal pavement markings and crash reductions of approximately eight percent with painted edge-lines on rural two-lane highways, and concluded that edge-lines would be cost-effective on a mile of rural two-lane highway if one crash a year occurred outside the roadway every 15 miles (25 km). Cairney (1995), too, noted that new pavement markings such as profile edge-line and larger reflective glass beads hold out the possibility of delineation materials which will perform better than conventional paint and last much longer, thus providing road users with a safer travelling environment and reducing the frequency with which relatively high-risk re-marking operations have to be undertaken. In their report on a second stage of the study by Pronk and Corben (1999) on measures to improve motorcyclist safety on rural roads, Corben, Gelb, Pronk and Fitzharris (1999) examined more closely two of the recommended innovative approaches to addressing motorcyclist safety, namely, tactile centre-line marking and speed enforcement technology for use on police motorcycles. This study reviewed the likely effectiveness of using tactile centre-line markings to improve safety at locations that have a poor safety record involving motorcyclists losing control on curves or colliding with oncoming vehicles on curves. While it was concluded that tactile centre-line markings may increase the risk of motorcycle loss-of-control on curves, the possibility of using transverse bar markings on the approach to hazardous curves was identified. Much of the potential use of these types of engineering treatments for motorcycle crashes appears to be in rural settings. Curvilinear routes, which are common in some rural areas and are known to have a poor safety record in terms of the target crash types, were identified and assessed. A subset of routes considered suitable for the trial of transverse bar markings were defined. However, in relation to such a trial, further work on the development of the concept, in terms of placement of bars, separation, height, material, colour, etc., was recommended before on-road trials are undertaken.

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

91

3.5

MEASURES TO IMPROVE THE SAFETY OF HORIZONTAL CURVES

As indicated previously, horizontal curves are one of the most dangerous parts of the rural road network, with high proportions of single- and multi-vehicle crashes at or near curves, particularly on two-lane rural roads. In the last decade or so, it has become apparent that three safety criteria should be applied: i) harmonising design speed and operating speed, especially on wet pavements, ii) achieving consistency between successive design elements, and iii) providing adequate dynamic safety of driving (Lamm et al., 1995; Leisch & Leisch, 1977). The 1990 AASHTO geometric design guidelines recommend: Consistent alignment should always be sought, Sharp curves should not be introduced, and Sudden changes from areas of flat curvature to areas of sharp curvature should be avoided.

There are a number of potential improvements to geometric features of horizontal curves that have been identified in the literature that can affect crash risk at these locations. These include measures to reduce speed on the approach to and through horizontal curves, measures to improve curvature, super-elevation, transitions, pavement and signing and the effectiveness of these measures are discussed in the following sections. Shankar et al. (1993) investigated the impacts of various geometric features of curves on specific crash types and found: An increase in the number of horizontal curves designed below 96.5 km/h would increase sideswipe and rear-end collisions but decrease parked vehicle collisions as well as fixed object collisions. They suggested that drivers tend to slow down on under-designed curves with design speeds less than 96.5 km/h because of signing and visual perception and therefore avoid severe crashes such as fixed object collisions. However, this possible speed reduction does not appear to be enough to avoid sideswipe or rear-end crashes caused by lane violations or speed differentials due to braking on the curve. An increase in the number of horizontal curves designed between 96.5 km/h and 112.6 km/h would increase same direction and fixed object collisions. They suggested that curves under-designed between 96.5 km/h and 112.6 km/h do not create the visual impact on drivers to decrease speed, resulting in an increase in both lane violations and vehicles running off the roadway and colliding with fixed objects. They recommended that consideration should be given to upgrading marginally under-designed curves if fixed object collisions show increasing trends. An increase in the number of curves in sections would increase same direction and fixed object collisions because, they argued, that when there are many curves, speeds do not decrease enough to avoid lane violations. The presence of guard-rail prevents a more severe type of crash from occurring. They recommended that it is preferable to design longer but fewer horizontal curves where the terrain makes construction of straight sections impossible. An increase in the average spacing of curves would increase overturn collisions because drivers are more likely to speed as a result of lower caution. While Shankar

92

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

et al. (1993) agreed that decreasing the spacing of curves may be counter-intuitive, they suggested that more advance warning signs should be placed in sections with longer curve spacing. Low curve radius would increase sideswiping collisions and decrease overturn collisions. Measures to reduce speed on curves

3.5.1

A successful curve treatment is considered to be one that produces a low curve entry speed, maintenance of an appropriate speed throughout the curve and a small range of lateral position. There have been various attempts to decrease speeds on curves and approaches to curves and there is general agreement that the provision of timely information to the driver is of paramount importance, and that the effectiveness of this information is enhanced with associated enforcement. Over 20 years ago, Leisch and Leisch (1977) recommended that the design speed on horizontal curves should be as consistent as possible and that procedures should be based on three principles: i) average speeds along an alignment should not vary by more than 16 km/h, ii) design speed reduction should not exceed 16 km/h, and iii) average truck speed should not differ from passenger car speed by more than 16 km/h. The Transportation Research Board (1987) noted that some countries, notably Germany and Australia, have addressed the disparity between design speed and operating speed by updating their design procedures to include checks of actual driver speed behaviour. These findings provide support for the adoption of an operating-speed based design procedure for two-lane rural highways. Several guidelines have been recommended for maximum speed reductions from tangents to horizontal curves and for maximum differentials between design and operating speeds on curves. The recommended maximum differentials range between 15 and 25 km/h (Anderson & Krammes, 2000). In addition, Puvanachandran (1995) recommended that approach speed should be reduced to a level appropriate for existing radii for road sections and that traffic management measures could be used to address this issue. These may include warning signs or physical measures such as ramps on low speed roads. Comte and Carsten (1998) suggested that there is a role for advanced technology solutions (such as speed limiters or intelligent speed adapters) to increase safety on the approach to and on curves. They evaluated the effectiveness of four speed management systems on drivers speed choice and performance measures (braking, steering performance and lateral placement) as they approached and negotiated horizontal curves in a driving simulator. The systems tested included advice systems (either using traditional methods such as transverse bars, or an in-car advice system), a system that conveyed the threat of punishment (variable message signal [VMS]), and a fully automated speed control system. Of all the systems, automatic speed control was significantly more effective than the other systems at reducing speed on approach to curves and consequently having additional positive effects on lateral control in curve negotiation. All the advisory systems (both traditional measures and advanced technology treatment solutions) significantly reduced speeds in the order of 6 km/h when activated and were particularly effective in smaller radius curves. They concluded that the provision of advice, in any format, could be effective in reducing speeds. The VMS sign was particularly effective in lowering speeds early in the approach to the curves, however, this early advantage was not maintained and eventually paralleled

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

93

the speed reduction curves of the other advice systems. While this study also showed that there were no obvious adverse effects of providing information in terms of driver distraction and workload, the long-term performance of these systems was not able to be determined. In a simulator-based study, Pyne et al. (1995) investigated the effectiveness of measures to reduce speeds on rural single-carriageway roads. They examined three treatments; those that reduce speed and speed variance on straight roads, those that reduce curve entry speeds for sharp bends, and those that reduce speeds on the approach to and through villages. Treatments included road markings to reduce lane width or produce horizontal deflection, signs both on posts and on the road surface, and PCMs. Substantial reductions in speeds and speed variance were obtained by some of the treatments evaluated. For curves, the most effective treatments were transverse lines with reduced spacing, a Wundt illusion (a series of chevrons with increasing angles but constant spacing, pointing towards the driver), and hatched areas at the edge of the road, combined with sign treatments on both posts and the road surface. For approaches to villages, the most effective combination of treatments was the chicane without hatching, yellow or white transverse lines and countdown speed limit signs. For the general treatments, all those which involved lane narrowing produced speeds significantly less than the control. Shoulders delineated by continuous lines were more effective than those delineated by broken line. Shoulder width was not important, but carriageway width was. For central hatching, type of delineation and width of hatched area was not important. In addition, the location and type of the narrowing was not important. 3.5.2 Curve flattening

Sharp curves (particularly those over 5) have high crash risk. Given that the primary factor for increased crash risk on curves is sharp curvature, there have been some suggestions that curve flattening may be an appropriate measure to reduce this risk. Curve flattening can reduce crash and injury risk, vehicle operating costs, travel time and maintenance costs, and shorten the length of a highway. As a curve becomes flatter, a smoother transition between the tangent and the curved section of a highway is provided. The evidence suggests that curve flattening is one of the most effective means of decreasing crash rates on curves (Cairney, 1998; Talarico & Morrall, 1994; Zegeer et al., 1992). Lin (1990), however, also noted that curve flattening cannot always be justified, arguing that these beneficial impacts should be weighed against capital outlays such as road construction, right-of-way acquisition, and access provision, and that the benefits and costs of flattening can vary substantially from one case to another. Generally, curve flattening costs depend on the original curve geometry, and costs increase with increases in either the central angle or the change in degree of curve. Lin argued that flattening curves with curvatures of not more than 11 is unlikely to be cost-effective even if no additional rightof-way has to be purchased. For sharper curves, a minimum reduction of about 7 of curvature is needed to achieve cost-effective new designs. For example, to compensate the net costs of curve flattening through crash reduction, a 13 curve would have to be replaced by a 5 curve. For curvatures less than 15, it is difficult to justify curve flattening based on crash reduction alone. Moreover, when service life (approximately 25 years) is taken into account, Lin argued that the flattening of a 9 curve to a 2 curve would be cost effective, however, if growth in average daily travel was also taken into consideration (a growth of up to 85%), flattening of curves of less than 9 in curvature difficult to justify.

94

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

Talarico and Morrell (1994) subsequently argued that the model suggested by Lin may not be valid for analysing the cost-effectiveness of flattening a curve by a large amount. Using construction cost models developed by Glennon et al. (1985), and including time savings, Talarico and Morrell determined the relationships between costs per crash eliminated for a range of AADTs (although it should be noted that the notion of valuing small travel time changes, especially for private travel, has been increasingly called into question in recent years (see Haworth et al., 2001)). Talarico and Morrell found that the cost-effectiveness of curve flattening increases exponentially with AADT, since greater vehicle operating savings, time savings, and crash savings are achieved. They also noted that the costeffectiveness of flattening decreases with increasing degree of curvature because very limited savings are realised, but substantial construction costs are incurred by flattening the curve by a small amount. As the curve is flattened more, the benefits increase at a faster rate than the construction cost, thereby decreasing the cost per crash eliminated. They concluded that an optimum degree of flattening exists for curves with central angles less than or equal to 45, regardless of the construction cost model used. This indicates that, at some point, the costs associated with flattening these types of curves outweigh the benefits. Furthermore, the cost-effectiveness of flattening these types of curves decreases with increasing central angle. This is due to the fact that the increase in construction costs of flattening curves with larger central angles outweighs the benefits. In addition, they concluded that the cost-effectiveness of flattening curves with central angles greater than 60 increases with central angle and that the curve should be flattened as much as possible in order to achieve maximum cost-effectiveness. They developed the following guidelines for curve flattening: Increase the rate of super-elevation of the curve when the design speed of the curve is below the 85th percentile speed of the approaching vehicles, and the existing superelevation is less than the maximum rate suggested by AASHTO (1990) design guidelines; and Investigate the cost-effectiveness of curve flattening if the improved super-elevation is unable to reduce the difference between the design speed of the curve and the 85th percentile speed of the approaching vehicles below 24 km/h, and the AADT is greater than 750 vehicles per day.

Considering that trucks experience particular problems on horizontal curves, Mohamedshah et al. (1993) suggested that the significant degree of curvature on Interstate highways was = 2.5, whereas on two-lane rural roads it was = 6. 3.5.3 Improved alignment design of curves

The Transportation Research Board (1987) made the following recommendations for upgrading existing horizontal curves on rural two-lane highways in the USA based on the disparity between design and operation speeds: Highway agencies should increase the super-elevation of horizontal curves when the design speed of an existing curve is below the running speeds of approaching vehicles and existing super-elevation is below the allowable maximum specified by AASHTO new construction policies. The Board also reported that German guidelines incorporate the idea that when differences between operating speeds and design speeds exist, super-elevation and stopping-sight

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

95

distances should be based on the normally higher operating speeds. German maximum super-elevation rates are typically 0.08 (compared to 0.10 to 0.12 in the USA). German guidelines also have limited side friction factors to 40% of the maximum allowable tangential values for rural roadway design. By using only 40% of the side friction there is more than 90% available for friction in the tangential direction (see Voigt, 1996). This limitation is important because it leaves adequate tangential friction forces available for sudden acceleration, deceleration, or evasive actions. Australian guidelines also provide an iterative method of horizontal alignment design. Considering side friction, Australian guidelines recognise that the differences between the roadway and vehicle conditions means that side friction values used in design may not be derived directly from the known inventory of pavement resistance. In contrast, the side friction values are derived from observations of driver speed behaviour. Side friction values for design speeds less than 90 km/h are those used by vehicles travelling at the 85th percentile speed. Maximum side friction values for design speeds higher than 90 km/h are more than those side friction values used by vehicles operating at the 85th percentile speed. This is an example of somewhat prescriptive road engineering practices that are often based on judgments of experienced road designers rather than rigorous scientific analyses of road geometry on curves over a range of operating speeds. Voigt and Krammes (1998) and Voigt (1998) recommended that the USA follow the lead of other countries (in particular, Germany and Australia) by adopting a two-dimensional approach to ensure the design of a safe alignment. First, they noted that problems only arise when the speed chosen for the design is lower than the expected operating speed and recommended that the use of 85th percentile speeds on curves should be used to design curve parameters, including super-elevation. Secondly, they recommended incorporating a feedback loop in rural horizontal alignment design to check for and address operation speed inconsistencies. Specifically, they recommended that a set of operating speed reduction ranges, similar to one set of criteria in the German method of consistency checks that are based on operating speed reduction and side friction demand would be appropriate in US design procedures. Likewise, Cafiso (2000) examined the relationship between alignment inconsistency and road hazard and noted that safety was adversely affected when the operating speed was inconsistent with the design speed. He developed analytical methods for integrating geographical and alphanumeric data to enable accurate identification of road defects both in the design and re-design of Italian rural roads. In addition to providing increased super-elevation levels on curves, Lamm et al. (1995) noted that sufficient friction supply should be a main safety consideration in designing, redesigning or resurfacing roadways and suggested that pavement skid resistance treatments can decrease the side friction factor, therefore decrease crash and injury risk. 3.5.4 Spiral Transitions

The approach to curves is an important transitional length where drivers make decisions about their speed on a curve. It also provides an area for super-elevation to change from the normal design of the road length to the full super-elevation required on the curve. Therefore, the geometric design of the transition can greatly affect crash and injury risk. There are a number of transition types and, in general, spiral transitions are the preferred

96

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

design, however, the literature suggests that there are both benefits and disbenefits of this design. Glennon et al. (1985) concluded that spiral transitions were beneficial, since almost all drivers tend to overshoot curve entrances when no spiral transition existed. Zegeer et al. (1991) also showed that the presence of a spiral reduced total crashes by two to nine percent, depending on degree of curve and central angle. In contrast, a study by Tom (1995) in the USA compared crash severity and rates on matched curves with and without spiral transition curves and found that curves with spiral transitions had generally higher serious injury crash rates than standard non-spiral locations. Councils (1998) analysis of the safety effect of spiral transitions in level, rolling and mountainous terrain took into account other confounding variables such as degree of curvature, lane and shoulder width, roadside, etc., and showed both benefits and disbenefits of spiral transitions, depending on other features. Council concluded that these results seem to indicate that, in level terrain (which might be assumed to have significant preview distance), spirals can be beneficial on sharper curves (those > 3). In rolling terrain, they appear beneficial at flatter curves (those = 8) on minor roads and on principal arterials designed for higher average daily travel (i.e., greater than approximately 4000 ADT). In mountainous areas, the general disbenefit finding (crash probability increase) indicates that spirals should be used very seldom, and only on roads with wider lanes and shoulders. 3.5.5 Provision of signing and road marking

Given the evidence that excessive speed on the approach to and through a horizontal curve increases crash risk, there may be potential safety effects of signing at these locations, particularly on sub-standard horizontal curves. Signing and road marking can provide the driver with the information needed to mentally prepare for a more complex driving task and to select the most appropriate speed before entering and while on the curve. While most EU countries have guidelines stating how to apply signing measures at horizontal curves, the existing criteria and practice used for signing and marking leave much to be desired (Persaud et al., 2001). Moreover, practice often differs from the guidelines. Therefore, drivers do not get the same information every time they enter similar curves (Nielsen & Griebe, 1998). Nielsen and Griebe argue that road authorities need to classify curves on rural roads into a number of danger categories (expressed by the difference between the approach speed and the design speed of the curve) in a structured and uniform way. By modelling five categories of danger, Nielsen and Griebe demonstrated that the limits of each danger category followed an elliptic cylindrical curve reaching a level close to the limit no signing required at very high approach speeds. This method gives a qualified estimation of which danger category a certain curve belongs to, however, they noted that there may be other conditions that may need to be considered to shift the curve to a more dangerous category. These include: bad visibility of the curve from the approach; bad readability of the curve; a curve with small radii after a long straight section; if the curve radius is not regular; massive obstacles on the edge of the road; and a high number of traffic crashes. They recommended that the following signing and marking concepts should be used in a uniform way so that road users always encounter the same basic signing and marking in curves belonging to the same danger category: delineators (marker posts), centre-lines and

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

97

edge-lines, pre-warning signs, advisory speed signs, and chevron signs. For example, a curve that is classified in Danger Category A (where approach speed is marginally higher than design speed) the following signs and markings should be considered: delineators through the curve on the outer side of the curve, ordinary centre-lines between the carriageways, and ordinary edge-lines through the curve, both at the inner and outer side of the curve. A curve that is classified in Danger Category E (where approach speed is markedly higher than design speed) the following signs and markings should be considered: advance curve warning sign, curve advisory speed sign, multi chevrons throughout the outer side of the curve, profile centre-lines between carriageways, and profile edge-lines through the curve both on the inner and outer side of the curve. Persaud et al. (2001) developed an effective model for estimating the expected safety of curves on two-lane, paved rural roads, based on actual crash history as well as geometric and roadside characteristics (such as pavement and shoulder width, curvature, length, terrain, and AADT). Preston and Schoenecker (1999) explored the potential safety effects of a dynamic curve warning system at a 4 curve along a rural two-lane highway. Their findings confirmed that the initial speed of a vehicle prior to entering a curve does have a significant effect o n the probability of successfully navigating through the curve. They also found that the overall effect of the dynamic curve warning system on vehicle speeds was relatively small. However, they noted that the system had a much higher effect on high-speed v ehicles than a static curve warning sign and that the dynamic system significantly improved the ability of the high-speed vehicles to successfully navigate through the curve. Moses (1990) evaluated a number of engineering treatments on sharp rural horizontal curves with operating speeds at least 15 km/h deficient of the zoned speed. Treatments aimed to highlight the curves, especially for the benefit of night-time drivers and included yellow retro-reflective raised pavement markers over the length of the curve to complement centre barrier lines, and enhanced edge-lines. Crash rates involving vehicles failing to take the bend were compared before-and-after treatment implementation. The results were promising, with a reduction of 25% of night-time crashes and a marginal reduction in day-time crashes of five percent. In terms of the cost/benefit of these treatments, Moses estimated a cost/benefit of at least 2:1 over a five-year marker life, bearing in mind the high cost of run-off-road crashes. 3.6 MEASURES TO IMPROVE THE SAFETY OF ROADSIDES

As indicated in Section 2.4.4, the roadside environment has a significant influence on crash and injury risk. Single-vehicle run-off-road crashes where vehicles strike rigid fixed objects or overturn result in severe injury to drivers and car occupants. Vehicles cannot protect occupants sufficiently when impact speed and intrusion into the vehicle by objects such as trees, poles or other rigid structures is high. Given that these types of crash constitute a large problem in rural road safety, there have been many calls to give priority to improving the design of roadsides, particularly reviewing policies on the provision of safe roadsides such as retaining trees and poles adjacent to the highway network and providing a safe zone (FHWA, 1992; Rattenbury & Gloyns, 1992; Sanderson & Fildes, 1984). In exploring innovative treatments for single-vehicle, run-off-the-road crashes in the USA, Hall (1991) remarks on the inability of the more traditional countermeasures, such as

98

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

roadside clearing, less-rigid fixed objects, and older standards of barriers, to completely address the situation of single-vehicle crashes. Attention is also drawn to the role and responsibility of the highway engineer in preventing such crashes or in minimising their severity. Hall cites Representative John A Blatnik (D-MN): It is the height of cynicism to contend that drivers should never have left the road or that many of them must have been drunk, or that somehow the driver was at fault. Why or how he left the road is not the issue. Whether he left because he was drunk, or stealing a kiss, or because he suffered a bee sting, dozed, had a blowout, was sideswiped, or was forced off is irrelevant to road builders. What is relevant is that those who are responsible for road construction recognize that the roadside is as vital to safe operation of a vehicle as the pavement itself, and that the duty to make the roadside safe is a very real one. This line of argument highlights very powerfully the importance of the attitudes and culture of road agencies, road builders and traffic engineering professions in determining the inherent safety of roadsides. While not itself a countermeasure, institutional attitudes are a most important determinant of safety in both rural and urban settings (Hall, 1991). There are essentially two categories of roadside crash and injury protection measures (Mintsis & Pitsiava-Latinopoulou, 1990). Injury reduction countermeasures include: the removal or relocation of roadside objects, (i.e., setting objects away from the kerb, thereby creating a clear-zone); reduction of number of roadside objects by pole-sharing and the use of frangible, breakaway or flexible materials in the construction of the street furniture; and protection of off-road objects with crash barriers and energy absorbing systems. Crash reduction countermeasures include: removal of culvert headwalls and replacement with end sections and grates, end treatments and guard-rails as well as widening or upgrading of the shoulder, realignment and delineation. The types of roadside improvements (both traditional and innovative) will be discussed in detail in the following sections. McLean (1996) provided crash prediction models allowing for variations in seal width, unsealed shoulder width, a subjective roadside hazard rating and a regional calibration factor to predict crash rates on two-lane rural roads. From these models, McLean provided general guidance to improve the safety of roadsides, namely that i) the removal of obvious and easily addressed roadside hazards can result in an approximate 18% reduction in crashes, and ii) reductions of 35 to 45% in crash rate can be expected from projects which specifically target roadside safety improvements, such as removing hazards and providing roadside barriers. He also conceded that there are practical constraints on the extent to which roadsides can be made safer, with a factor of around two being the maximum achievable result in most circumstances. Moore (1998) suggested that countermeasures that might reduce the potential for run-offroad crashes, particularly those that are over-embankment crash types include: installing new guard-rails; lengthening existing guard-rails; extending embankment guard-rail near cut-slope areas and burying guard-rail into cut-slopes; providing additional signage and striping; improving the shoulder; installing rumble strips; installing audible edge stripes; and/or cleaning up the roadside environment to provide a clear recovery area. He also noted that over-embankment crashes on rural two-lane conventional highways have a drink driving involvement nearly three times higher than that of all road types and suggested a program of enforcement should be initiated which would concentrate on locations having the highest drink driving involvement.
COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS 99

3.6.1

Provision of a forgiving roadside the safety of roadsides that has received much attention worlda clear-zone as large as practically possible by removing or provision of a clear-zone facilitates drivers maintaining or vehicle, or an opportunity to substantially reduce speed before

One treatment to improve wide is the provision of relocating obstacles. The regaining control of their they strike rigid structures.

3.6.1.1 Removal or relocation of trees and poles As indicated previously, objects causing the most severe injuries in run-off-road crashes are trees and poles, bridges, culverts and embankments, with collisions with trees making up the major part of the injury problem in rural areas. Removal or relocation of trees and poles to provide clear-zone distances of between 5 and 30 ft (1.5 to 10 m) can reduce crash and injury risk substantially (estimates range from 13% to approximately 60%). McLean (1996) noted the benefits of clear zones of at least 9 m on the sides of rural roads having a speed limit of 80 km/h or higher. Schnerring (1995) also argued that a 9m clearzone is sufficient space for most drivers to either return to the road, or to bring their vehicle to a safe stop. Notwithstanding this, Schnerring did point out that a 9m clear-zone is not always possible, in which case roadside objects need to be modified or protected (see following sections for detailed discussions on treatments to modify or protect roadside objects). Early work by Huelke and Gikas (1967, cited in Adamo & Wray, 1992) highlighted the role of hazardous roadside objects, particularly noting that 80% of crashes involving vehicles running off the road resulted in vehicles striking an object within 8m of the edge of the roadway. They suggested that there are many opportunities available to improve safety through the clearing of obstacles within roadsides as well as reducing roadside slopes and improving roadside ditch design. Rattenbury and Gloyns (1992) predicted that suitable modifications of the roadside environment have the potential to halve car occupant fatalities in run-off-road crashes, leading to an overall nine percent reduction in car occupant fatalities. They noted three main countermeasure options involving modifications to the road infrastructure. These are: Eliminate as many hazardous roadside objects as possible, thereby maximising recovery space for errant vehicles, Position man-made structures such as road signs, lamp posts and utility poles as far away from the road edge as possible, and Use crash barriers or impact alternating devices where hazardous objects cannot be removed or sited in a safe location. They also noted that care should be taken in the design and placement of barriers to avoid unintended hazards.

The FHWA (1992) estimated that, by clearing trees for 10 ft (3 m), from 8 to 18 ft (2.4 to 5.5 m), the incidence of tree crashes would fall by around 57%. This estimate assumes that a steep side-slope is not present to increase the likelihood of errant vehicles striking trees beyond the cleared range. Likewise, for utility poles, the FHWA (1992) found that increasing the lateral offset of utility poles located beside rural roads can reduce the incidence of crashes involving run-off-road vehicles. An approximate 50% reduction in utility pole crashes was predicted for a 3 m increase in lateral offset of utility poles (from 5 ft to 15 ft) from the edge of the roadway. (Note: this case was based on a road with an
100 MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

AADT of 10,000 vehicles per day, and a linear density of poles along the road of 60 per mile, approximately 40 per km). The FHWA (1992) also examined the effect of pole offset in a range of rural scenarios using nomographs developed from studies of utility pole crash experience. These nomographs cover i ncreases in pole offset of between 3 and 15 ft (1 and 5 m), for a range of lateral offsets in the before condition. The percent reduction in utility pole crashes ranged from 18% for minor increases in lateral offset of the poles (say 3 ft, or 1 m), to 75% for 15 ft (about 5 m) increases in offset when poles were originally placed within 2 ft (0.6 m) of the edge of the roadway. Corben, Deery, Mullan and Dyte (1997) conducted a study to evaluate treatments implemented to address the problem of single vehicles crashing into fixed roadside objects, in terms of casualty crash occurrence, crash costs, and economic worth. The results of the evaluation were intended for use in setting priorities to further improve the return on road safety investments. Over 400 engineering treatments undertaken in Victoria since 1989 to address crashes into fixed roadside objects (drawn from State and Federally funded programs) were initially examined. Of those, 254 were retained for evaluation, and the balance excluded due to incomplete records. Overall, casualty crash frequencies and costs were reduced by nine percent and 16%, respectively an estimated BCR of 4.1. With regard to particular treatment types, road/roadside geometry improvements reduced casualty crash frequencies by 23%, with a subset of these treatments (changes to horizontal road geometry), found to reduce casualty crashes by 44%. Road surface improvements reduced casualty crashes and costs by 30% and 36%, respectively, and achieved a BCR of 5.6. A subset of these treatments (largescale shoulder sealing) reduced casualty crashes and costs by 32% and 37%, respectively. Corben et al. (1994) recommended that both the immediate and medium-term strategies for roadside safety give increased emphasis to the highly effective treatments designed to improve the road surface and road geometry, such as large-scale shoulder sealing, improved horizontal road alignment and skid resistant pavements. Due to the absence of statistically reliable reductions in casualty crashes from guard-rail end treatments, street lighting and pavement widening, it was also recommended that further critical assessment be undertaken of these treatments before they form a significant part of future roadside safety strategies. At a broad level, and given that the program of treatments in rural areas was found to be less effective than that in metropolitan Melbourne, it was proposed that more strenuous efforts be made to enhance Victorias rural program, through an increased emphasis on the most effective treatments. Corben and Deery (1998) focussed on those treatments that led to an increase in crash occurrence and/or severity in the earlier research (Corben et al., 1997), with a detailed examination of the crash history and treatment features with the aim of identifying possible explanations for their poor performance. Closer examination of individual treatments showed that the outcomes were often site specific. Some site treatments appeared unsuccessful because they were implemented pro-actively, as part of black length treatments, at bridges with no, or very low, crash frequencies of the type targeted by bridge end post and bridge guard-rail treatments. That is, when implemented at bridges with no significant crash problem, improvements cannot be reliably demonstrated. At best, the good crash record will be maintained. It is the authors belief that implementing treatments pro-actively as part of a black length program is an appropriate approach to harm

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

101

reduction, but that such an approach makes it difficult to demonstrate a significant treatment benefit as part of a scientific evaluation. Secondly, when lengths of road having a record of run-off-road crashes are treated with guard-rails on bridges or on the approaches to bridge end posts, the remaining s ections of the road length can continue to expose traffic to relatively high crash and injury risks. More comprehensive treatments, covering longer road lengths, are needed in such circumstances. In this regard, the results point to the most effective treatment types, such as road and roadside geometry improvements (e.g., horizontal and vertical alignment) and road surface improvements (e.g., shoulder sealing). Finally, notwithstanding the results of the evaluation, it remains the view of the authors that guard-rail treatments at bridges are likely to reduce injury risks should a crash occur. However, unless the guard-rail treatments are implemented at bridges with a substantial record of crashes into end posts or other parts of the bridge structure, they can only be regarded as pro-actively addressing a potential crash problem, rather than being a proven treatment at black spot bridge locations. As a result, justifying treatments of this type will be largely subjective, rather than being supported by rigorous quantitative evaluation. For these treatments, both the actual effectiveness and the reliability of estimates of effectiveness may well be increased if more rigorous procedures are employed in the future to identify candidate locations for treatment. In the most recent MUARC study on crashes involving vehicles leaving the roadway, Delaney, Langford, Corben, Newstead and Jacques (2002) noted this crash type as a large and severe contributor to road trauma in Australia and a source of major concern internationally. Their study presented an important opportunity to consolidate, in a single document, vital information on the problem and its possible solutions, as well as defining state-of-the-art approaches being taken both nationally and internationally to address roadside safety concerns. Delaney et al. (2002) addressed the incidence, causes and extent of road crashes involving vehicles leaving the road and colliding with roadside objects, particularly trees and utility poles; liability and accountability issues; world best practice on managing the risk of roadside crashes and how Australian/Victorian practice compares; the extent and nature of design guidelines available; potential improvements (e.g., infrastructure, design, legislative, professional education, etc.), and which authority might be responsible for implementation. The key conclusions were: A long-term, proactive and system-wide strategy is necessary to address the high frequency and severity of run-off-the-road crashes. Treatments need to address both the risk of a crash occurring and the severity of a crash if one does occur. Different approaches to roadside safety and crashes with fixed roadside objects have been adopted internationally. The US approach is essentially a step-by-step or incremental approach whereas Sweden and the Netherlands have adopted more systemic approaches.

102

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

In Sweden, the use of wire-rope barriers and the removal of solid objects and other engineering treatments along more than 200 km of road has proven to be a low-cost approach to improving safety, compared to other options. However, some difficulties continue. A range of engineering treatments implemented in Victoria from 1989/90 to 1993/4 aimed at reducing crashes into fixed roadside objects have been shown to effectively reduce all casualty crashes by 8.6% on average. Crash costs also declined by 15.5%, indicating that the treatments were also effective in reducing the severity of crashes. Evaluation of the (Victorian) TAC funded Accident Black Spot Program from 1992 to 1996 indicates that a number of treatments aimed at reducing the frequency and/or severity of run-off-the-road casualty crashes were effective in reducing casualty crash frequency. The most effective was shoulder sealing with obstacle clearing and regular line marking that reduced casualty crash frequency and costs by approximately 50%. Despite early doubts about the effectiveness of safety barriers, recent US research indicates that well designed and maintained safety barriers are effective overall in reducing casualty crash severity. The use of wire-rope safety barriers in Sweden has also been effective. The major treatment types that may reduce the frequency of run-off-the-road crashes are: o o o o o o Traffic barriers; Roadway and roadside design; Improvements to the road surface (skid resistance and shoulder sealing); Improved delineation and sight distances; Improved roadside slopes and other hardware; and, Improved drainage structures.

Other treatment types such as clear zones and pole modifications may also contribute to reductions in casualty crash severity.

Delaney et al. (2002) report, however, that the adequacy and validity of the clear zone concept as a means of addressing roadside safety is increasingly being questioned. The current clear zones were defined after a US study conducted approximately 30 years ago found that up to 85% of vehicles travelling along highways could recover within 9 m from the edge-line of the road. A recent study of median encroachments by the drivers of vehicles travelling on high-speed roads concluded that the Victorian guidelines for the median width above which no median barrier is required (15m) frequently did not prevent a vehicle passing through the median and potentially colliding with vehicles travelling in the opposite carriageway (Corben et al., 2001). Therefore, 9 m (either to the left or right of the carriageway) will often not be adequate for a person to regain control of their vehicle before reaching the far boundary of the clear zone.

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

103

Also, from a first principles assessment that, many vehicles leaving the roadway would be still travelling at high speeds by the time they reach the far boundary of a 9 m clearzone. In many cases, the contact speeds with rigid objects beyond the clear-zone will dramatically exceed the capacity of vehicles with the highest standards of crashworthiness to protect their occupants. A key recommendation by Delaney et al. (2002) was to urge government to establish a new, over-riding philosophy for the various agencies with responsibility (formal or informal, direct or indirect) for roadside safety. Such a philosophy would provide the fundamental underpinning required to ensure safe roadsides are an essential goal of developing and operating the road-transport system. In effect, this would require a formal accountability for designing and building safe roads (analogous, for example, to the position of the SNRA), rather than treating safety as a by-product of transport, to be managed. Specifically, this new philosophy should adopt the following objectives of providing: 1. Vigorous, committed leadership across the diverse range of organisations for whom road safety is not the core business or represents only one of several organisational goals. There is a clear need to instil a very strong sense of organisational ownership that goes well beyond the fear of litigation. Natural incentives to attack the problem. Systematic, evidence-based programs for eliminating infrastructure deficiencies, in terms of crash risk, injury risk or both. New and comprehensive road-transport system standards to prevent new hazards being created or built into the infrastructure.

2. 3. 4.

This requires a shift in government emphasis in many countries from preventing crashes to providing crashworthy infrastructure. While outside of the scope of Delaney et al.s (2002) study to recommend a specific mechanism by which this new over-riding philosophy could be adopted, the authors envisaged a strengthening and broadening of relevant system operation and road design standards to ensure that both new road improvements and existing roads are properly and fully addressed in terms of roadside safety. Through the various agencies of government, the private sector, in the form of road builders, utility companies and tram operators, will also be required to assure that new roadside hazards are not introduced to the system and that existing hazards are eliminated within a reasonable time frame. 3.6.1.2 Removal or relocation of other objects While the removal or relocation of trees and poles are regarded as the most important in providing a forgiving roadside, there are other obstacles that have been identified. For example, the FHWA (1992) suggested that crashes with culvert headwalls can be reduced by around 40% if drainage facilities within a 10 ft (3 m) range between 5 and 15 ft (i.e., between 1.5 and 4.6 m), are reconstructed so that they are flush with roadside terrain. The FHWA also notes that sign placement and the positioning of mailboxes represent another source of roadside hazards in rural areas and estimated crash frequency reductions ranging from 14% for lateral relocations of 3 feet (1 m), to 40% for lateral relocations of 10 feet (3 m). They also pointed out that such relocations need to be cognisant of the needs of drivers
104 MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

to see the signs and of other practical considerations associated with the use of mailboxes. In addition, the FHWA estimated that impacts with roadside fences or gates in rural areas can be reduced by around 20% for a 3 ft (1 m) increase in lateral offset to about 52% for a 10 ft (3 m) increase. With this countermeasure, it was noted that both the cost of such measures and the extent of the problem must be considered relative to the potential road safety benefits, as such relocation of gates and fences would most likely require the purchase of additional right-of-way land. Last, the FHWA discussed the potential benefits of guard-rail relocation and estimated reductions ranging from 36% to 78% are achievable from increases in lateral offset of guard-rails from between 3 and 10 ft (1 to 3 m). 3.6.2 Installation of safety barrier systems

Safety barrier systems are proving to be an effective means of reducing the incidence and severity of run-off-road crashes. If roadside hazards cannot be removed, or adequate clearzones created, consideration should be given to the installation of some physical barrier (McLean et al., 1998; Schnerring, 1995; Ydenius, Kullgren & Tingvall, 2001). Schnerring also suggested that crash cushions can be used to absorb the energy of the vehicle so that injuries are prevented or greatly reduced. Ydenius et al. (2001) examined the safety effects of three different types of barrier systems (flexible wire rope barrier, semi-rigid W-bean barrier and a rigid concrete barrier) and associated vehicle structural integrity and restraint systems. Vehicles ran into barriers at two speeds, 80 km/h and 110 km/h and at two different impact angles, 45 and 20. The tests demonstrated high variance between barrier types on crash severity and physical behaviour. Rigid barriers gave moderate or no deflection and redirected vehicles away from the hazard area, however, resulted in a higher crash severity to the vehicle. The authors suggested that concrete barriers would not be useful on a three-lane highway where impact angles and speed would be high because of high crash severity. The semi-rigid barrier had a lower deflection than the flexible barrier but was more flexible that the rigid barrier and may be preferable when the expected impact angle is great and the situation only admits small deflection. The authors suggested that W-bean barriers could be preferred when the expected impact angle is great and the situation only admits small deflection and that they could be suitable both as a roadside barrier and as a mid-barrier, as long as the combination of expected crash angles and travel speed does not create a situation leading to high injury risks. Flexible barriers gave a longer duration for deceleration and therefore a lower crash severity, however, there was high deflection on the barrier (dependent on the impact angle). The authors suggested that wire rope barrier systems may be useful for a wide range of impact angles, and is appropriate for mid-barrier use. In sum, Ydenius et al. suggested that the choice of barrier system should depend largely on the road environment itself. From an examination of the techniques used to predict crash rates along sections of rural roads, Troutbeck (1995) proposed an approach for ranking sites for the installation of safety barriers. He found that the hazardousness of objects in the roadside can be used to predict crash rates. Other factors used in the crash prediction method are curve radius, lateral distance of hazard within roadside, running speed of vehicles along the road and traffic volume. Techniques developed by the American Association of State Highway and Transportation Officials (AASHTO, 1969) were also assessed.

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

105

In their review of the adequacy of existing guidelines for use of barrier systems throughout the State of Victoria, Australia, Corben, Tingvall, Fitzharris, Newstead, Les and Johnston (2001) quantified the frequency, nature and spatial distribution of injury-related crashes involving median encroachments. In addition, the adequacy of the existing guidelines as a means of preventing death and serious injury on Victorias divided roads was assessed. Over the decade of the 1990s in Victoria, over 3,600 reported casualty crashes occurred as a result of vehicles encroaching into medians on all classes of road. While head-on collisions account for 18% of these casualty crashes and only 12% on high-speed roads the major part of the median crash problem involves errant vehicles striking hazards in the median, such as trees and poles. Corben et al. (2001) argued that the guidelines for the use of median barriers need to be strengthened. The current Victorian guidelines require medians to be less than 15 metres wide before a barrier is considered warranted, though height/slope or fixed object criteria, or there is a history of cross-median accidents may also be taken into account. The analysis of crashes conducted in this study suggests that the existing guidelines are inadequate for the following broad reasons: The use of median widths of less than 15 metres to justify median barrier installation focuses on less than 20% of deaths and serious injuries resulting from median encroachments. The current median barrier guidelines do not address 80% of the road trauma problem, namely, death and serious injuries resulting from collisions with roadside hazards within the median (e.g., trees and poles); For head-on crashes, median widths (in the range typically found on Victorian roads) were not found to be an effective criterion for deciding where median barriers should be installed. That is, there did not appear to be a strong relationship between median width and crash occurrence. Median width is even less relevant if considered in the broader context of all median-related injuries, not just head-on crashes; The results of a first principles analysis indicated that it would be necessary to provide excessively wide medians, in order to prevent errant vehicles from reaching the opposing carriageway in crash scenarios that are quite realistic for Victorian roads. Such requirements would be uneconomic and generally impractical.

Indicative BCRs for the installation of median barriers are, in some typical cases, as high as 8:1. This is twice the estimate for current black spot programs in Victoria and would be considerably higher if less conservative estimates of barrier project life and/or barrier maintenance costs were to be used in the evaluation of economic impacts. It is of considerable practical importance that a large proportion of the deaths and serious injuries occur on a relatively small proportion of Victorias road network. This finding was regarded as facilitating the practical and cost-effective treatment of the problem. It was recommended that a program of retro-fitting median barriers to existing high-speed divided roads be introduced and that priorities for such a program be driven principally by speed, traffic volume and the presence of roadside hazards. Furthermore, median barriers should be installed on all new, high-speed divided roads. Corben and his colleagues noted that in the future, especially for new roads, careful consideration should be given to both the choice of barrier system used and also to adopting an optimum median width, which:

106

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

Allows median barriers to be set back an appropriate distance so that minor median encroachments and subsequent recovery can occur without barriers being struck; Enables trees and other vegetation to remain (or be planted) within medians (but shielded by barriers), without the risk of death and serious injuries to the occupants of errant vehicles; Provides sufficient working width behind barriers to enable barrier deformation to occur without contact with fixed hazards in the median; Minimises the land acquisition costs and environmental impacts of road development and construction.

While conventional barrier systems have performed well for the occupants of passenger cars, their effects on the safety of other road user groups, especially motorcyclists, are not well understood. Duncan, Corben, Truedsson and Tingvall (2000) conducted a feasibility study with the aim to recommend a research method for investigating the interactions between motorcycles and road safety barriers. A review of the relevant national and international literature was conducted, revealing a relative lack of published material regarding the nature of motorcycle collisions both with roadside barriers as well as motorcycle crashes in general. Various features of barrier systems were identified in the literature as providing a significant safety risk to fallen motorcyclists, particularly barrier posts. There have been numerous strategies employed, mostly in Europe, to better protect motorcyclists from impacts with barriers, including the installation of additional W-Beams, using impact attenuators to cover exposed barrier posts and substituting traditional IPE posts with more forgiving sigma posts. In addition, there have been several new barrier designs and/or modifications that have been developed and tested in Europe with promising results. Guidelines developed for the conduct of physical crash-tests with motorcycles were reviewed, and alternative methods, such as computer/mathematical simulations and component testing, were also considered. Based on the information from the literature review and subsequent consultation with several experts and stakeholders, recommendations for a multi-stage research program were made. Duncan, Corben, Truedsson and Fitzharris (2001) studied the problem of injuries resulting from motorcyclist impacts with roadside barriers, especially wire-rope safety barriers (WRSBs). This issue is of particular significance given the strong body of evidence emerging from recent Swedish trials of installing flexible barrier systems along 13 m wide, high-speed rural roads to separate opposing traffic and, where side barriers were installed, to prevent errant vehicles from entering (hazardous) roadsides. Several research issues were examined, including the extent and nature of the run-off-road crash problems along two major high-speed Victorian routes; an indication of the theoretical relative risk for severity of injury for motorcyclists colliding with wire rope safety barrier (WRSB) compared to concrete barrier and guard-rail; and priorities, having regard to the overall outcome of reduced road trauma in aggregate, of attention to barriers (relative to other road safety initiatives) for motorcyclists. The key conclusions and recommendations from this study were: 1. In light of the current lack of information regarding the interaction between motorcyclists and barriers, there is a need to explicitly address this highly vulnerable road user group in the development, testing and installation of roadside barriers. Any future barrier designs which are more motorcyclist-friendly, in order to be costCOST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS 107

effective, may need to be limited to use at motorcyclist crash black-spots. To effectively implement any form of motorcyclist countermeasure program, more information is required regarding motorcyclist/barrier crash characteristics, as well as the safety performance of barriers for motorcyclists, specifically a comparison of the different barrier types currently in use. 2. In the meantime, however, given the demonstrated safety benefits of WRSBs to the vast majority of road users and the fact that motorcyclist impacts into WRSBs have, to date, been extremely infrequent, the question of whether it is ethically responsible to deny the vast majority of vehicle-based road user groups the safety benefits of WRSB over other available barrier types must be raised. To resolve these technical and ethical concerns, recommendations for a multi-stage research program were made. Shoulder improvements

3.6.3

In addition to the provision of clear-zones there are suggestions that improvements to the road shoulder can reduce the frequency and severity of single-vehicle run-off-road crashes. Indeed, Sanderson and Fildes (1994) concluded that run-off-road crashes which did not strike a fixed object are as much a problem as those which did strike a fixed object and argued that treatments that aim to keep vehicles on the carriageway are as important as providing a forgiving roadside environment. Moreover, they argued that, because there is an equal chance of vehicles running off the road both to the left and to the right, effective countermeasures will need to be able to cope with both directional run-off-road crashes. Improvements to shoulders include provision of adequate widths, pavement type (sealing), minimum slope, and installation of edge-line treatments such as rumble strips. In an Australian study of black spot treatments, Tziotis (1993) evaluated the effectiveness of mid-block treatments undertaken in the State of Victoria between 1980 and 1989. In relation to specific measures that may be suitable for use in rural settings, Tziotis (1993) found that treatments for roadside hazards reduced pole and other fixed roadside object casualty crashes by 68% (statistically significant at 95% level of confidence), while pole and other hazard removal treatments reduced fixed roadside object casualty crashes by 71% (statistically significant at 90% level of confidence). Limited specific information was provided by Tziotis (1993) on many of the types of treatments evaluated. That is, treatment types tended to be r eported in quite broad categories only. However, he did report highly significant reductions of 52% in relation to off-path/head-on treatments, and 42% in all crash types. In particular, delineation treatments were found to have reduced offpath/head-on casualty crashes by 41% (statistically significant at 90% level of confidence), while road works, including pavement/curve reconstruction, road-widening and channelisation, reduced both off-path/head-on and all casualty crashes by a highly significant 57% and 56%, respectively. Based on economic assessment, BCRs of over 7 were estimated remedial treatment of these types. 3.6.3.1 Shoulder pavement and width Given the significance of both the magnitude and nature of run-off-road crashes, many have evaluated the safety effect of paving and widening shoulders on rural roads, especially two-lane roads (McLean et al., 1998; Ogden, 1997; Polus, Craus, Livneh & Katznelson, 1999; Schnerring, 1995; Turner, Fambro & Rogness, 1981). By widening and

108

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

paving shoulders, drivers will have more space and levels of friction in which to maintain or regain control of their vehicle. Ogden (1997) reported that the presence or condition of unpaved shoulders contributed to at least 30% of crashes in the State of Victoria, Australia. He compared the effect of shoulder paving ranging from 600 to 1200 mm, with widths of 600 to 800 mm being most common and found an overall 41% reduction in casualty crashes per vehicle kilometre at treated locations, particularly in run-off-road to left, run-off-road to right into fixed objects, rear-end, and overtaking-out-of-control crash types. This was equivalent to a reduction of 0.071 casualty crashes per million vehicle kilometres. Ogden therefore concluded that it would be economically worthwhile to pave shoulders along roads with at least 360 vehicles per day. This conclusion was based on BCRs of 2.8 times the AADT in thousands. For the example of 4,000 vehicles per day, a BCR of 11.2 was predicted. Other factors associated with increased crash rates included narrow lanes, curves, unpaved shoulders and low skidresistance pavements. McLean (1996) examined the relationships between crash frequencies on two-lane rural roads and cross-section elements and road design and noted that the size of the effect of roadside design on crash frequency is similar to the combined effects of lane and shoulder widths. He predicted the following beneficial effects of shoulder improvements on twolane rural roads: A 20% reduction in crash rate for every one metre increase in seal width of an existing road formation (the increase in width can take the form of either increased lane width or part shoulder seals); A 9% reduction in crash rate for every one metre increase in road formation width (where increases in formation width occur through additional gravel shoulders).

McLean did caution the application of the predictions from his modelling, noting that, first, the predictions should not be extrapolated beyond current design practice, and secondly, that some have found increased crash rates where wide surfaced shoulders have been applied and suggested that this is because wide shoulders may encourage higher speeds and/or shoulder driving. In a later article, McLean et al. (1998) confirmed these predictions, noting that, in light of the frequency with which unsealed shoulders contributed to the causation of single vehicle crashes, and multi-vehicle head-on collisions, there appears to be strong justification for the shoulders of all highways and major rural roads to be sealed to a width of at least half a metre, with priority given to sealing the shoulders on the outside of curves and with edgelining introduced in conjunction with the shoulder sealing. Polus et al. (1998) examined the operational benefits (flow and safety) of paved shoulders on heavily-trafficked, two-lane rural roads, identifying the need for safety specific analyses on this issue, especially of crashes occurring directly as a result of the use by slow-moving vehicles of paved shoulders. They found that crash rates fell at seven of the ten sites (typically about 6-7 km long) where paved shoulders were constructed and utilised by slow-moving vehicles, such as trucks. The FHWA (1992) noted that, in addition to widening of lanes and medians on two-lane rural roads, widening of paved or unpaved shoulders can also increase safety benefits. They provided crash reduction estimates for a wide array of possible combinations of

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

109

increases in lane width and in shoulder width, for both paved and unpaved shoulders. Of relevance here, they estimated that widening of shoulders could reduce run-off-road, headon, and opposite and same direction side-swipe crashes by 16% for a 2 ft (0.6 m) increase in shoulder width from a base width of 2 ft. For an 8 ft (2.4 m) new paved shoulder, the reduction in crashes is predicted to be around 49%. For unpaved shoulders, the reductions in related crashes were reported as ranging from 13% for a 2 ft increase in width to 43% for an 8 ft increase in width. They recommended that, in addition to the provision of wide medians, paved shoulders of around 12 ft (3.7 m) should be provided along roadway sections with guard-rails installed. These estimates apply when roadside characteristics along two-lane rural roads, such as side-slope and clear-zones, are re-established to the same condition as before the lane and/or shoulder widening occurred. Several other US studies reported in Adamo and Wrays (1992) review of rural road improvements have demonstrated a marked reduction in the number of single-vehicle run-off-road crashes from the use of full width chip seal shoulder treatments, significant decreases and reductions of up to 80% where grooved sections had been used, while untreated sections showed slight increases in these crash types. Adamo and Wray also reported mixed opinion in relation to the use of wide edge-lines to better define the outer edge of the roadway, however, gave no quantitative measure of effect. While Adamo and Wray were unable to develop a useful mathematical model for predicting future single-vehicle run-off-road crashes, they recommended the implementation of shoulder treatments to reduce the incidence of crashes of this type and that further work should investigate the most appropriate design features for the shoulder treatments as well as the provision of rest areas as a means of reducing driver drowsiness. Moreover, they proposed that, in view of the lack of reliable measures of effectiveness of past shoulder treatments, future monitoring of crash data at field sites should be undertaken in order to reliably evaluate their effectiveness. 3.6.3.2 Shoulder slope The probability of an errant vehicle recovering after encroaching into the roadside is determined, in part, by the steepness of the side-slope of the roadside and the steeper the slope, the more likely a vehicle will rollover. Rollover crashes are typically severe but their incidence can be reduced though flattening of side-slopes. There is general agreement that side-slopes of 5:1 or flatter are needed to significantly reduce rollover crash risk. US studies reported by FHWA (1992) reveal that single-vehicle crash rates (as a ratio of crashes on a 7:1 slope) on two-lane rural roads can be expected to fall from almost 1.4 to around 1.0 (i.e., an approximate reduction of at least one-third) as a result of flattening side-slopes from around 2:1 to 7:1. In another example of the effects on safety of flattening side-slopes, the study reports reductions of 21% in single-vehicle crashes and 12% in total crashes when roadsides are flattened from 2:1 to 6:1. These reductions assume that the newly flattened roadsides will be relatively free of rigid objects. The report also includes a tabulation of a range of combinations of side-slope flattening options and their predicted crash reduction effects for both single-vehicle crashes and on all crash types and recommends the most cost-effective ratio for side-slopes should not exceed 4:1. 3.6.3.3 Edge-line treatments Road engineers attempt to improve the visibility of highways by delineating the road ahead (Sanderson & Fildes, 1984) and there are a number of edge-line treatments that can reduce

110

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

the incidence and severity of run-off-road type crashes, particularly those that have been used to either alert a driver to the imminent departure of their vehicles from the roadway and/or to reduce the danger once they have actually left the paved surface. An early study by OHanlon and Kelley (1974, cited in Adamo & Wray, 1992) examined the effect of three different types of textured shoulder treatments, comprising corrugated concrete strips, on psychophysiological factors of drivers, however, did not extrapolate these findings to crash frequency or injury severity. Nevertheless, this study did report effects of treatments on driver performance. The raised rib treatments resulted in high levels of driver arousal, however, this effect lasted only around five minutes. They also found that shoulder treatments were more effective if placed as close as possible to the right driving lane and recommended that 10 to 20 mile lengths of highway (16 to 32 kilometres) with a high incidence of single-vehicle run-off-road crashes should be treated in this way. More recently, Griffith (1999) argued that continuous shoulder rumble strips installed on freeways would alert drowsy drivers who are beginning to drift off the road. In a beforeand-after comparison with control groups Griffith reported that all single-vehicle run-offroad crashes fell by 18.3% on all freeways and 21% on rural freeways alone, and that injury single-vehicle run-off-road crashes on all freeways fell by 13%. While the study did not include benefit-cost analyses, Griffith concluded that the potential adverse effects of continuous shoulder rumble strips such as causing startled or panic responses in drivers, or causing crash migration were insignificant. Grder and Alexander (1995) also noted that driver fatigue is a serious safety problem and a substantial contributing factor to many of the fatal crashes reported on Maines interstate highways (USA), with some 80% of fatal crashes occurring on straight road sections. Many of the highways in Maine have been treated by installation of continuous rumble strips in an effort to address this problem, and the effects of continuous shoulder rumble strips on crash and injury risk on rural roads were investigated by Grder and Alexander (1995). They demonstrated that run-off-road crashes can be reduced by between 20 and 50% by the use of continuous shoulder rumble strips. Conservative estimates of BCRs indicated that values of 5 to 20, probably higher, would result from the installation of continuous rumble strips throughout the rural Interstate system in Maine. Taking into consideration the potential adverse effects on driver performance, motorcyclists and bicyclists, noise and vibration, durability, maintenance and road conditions, etc., Grder and Alexander supported a recommendation for widespread implementation of continuous rumble strips, noting the safety benefits of this treatment, not only on road edges but also on centre-lines to address the problem of head-on collisions. Preliminary guidelines for the use of continuous shoulder rumble strips were outlined. Another treatment identified in the literature is the use of indented strips on the roadway shoulder. Studies report that this treatment is effective in reducing, but not eliminating the incidence of single-vehicle, run-off-road crashes. Furthermore, indented shoulder strips appear to be cost-effective and are predicted to cut crashes by approximately 30%. Possible problems associated with snow ploughing and bicycling can be readily addressed. The author is unaware of alternative treatments of equivalent potential. Indented shoulder strips should be constructed as part of rural resurfacing projects and are believed to be most suited to extended lengths of straight roadway, where vehicle departure angles from the roadway are relatively low, shoulders are wide, and roadsides beyond the indented shoulder strip are traversable and reasonably free of rigid hazards.

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

111

Other delineation treatments include centre-lines and guide-posts with reflectors, increased number of posts on bends, and installation of raised pavement markers. Triggs, Harris and Fildes (1979) and Fildes (1979) demonstrated that centre-lines and guide-posts with reflectors enhanced static direction judgement at night. Furthermore, Triggs, Meehan and Harris (1980) reported that increasing the number of posts on the outside of the bend appeared to improve the drivers judgement. Witt and Hoyos (1976) and Johnston (1983) observed marginal advantages in driving performance with edge-lines. However, Sanderson and Fildes (1984) argued that it is difficult to establish how edge-lines affect driving from these studies. Triggs and Wisdom (1979) and Triggs (1980) concluded that edge marking is more useful for short-term lane position manoeuvres rather than for longer-term direction assessment. A study of a 550 km rural two-lane highway in Victoria showed that a 15% reduction in the casualty crash rate occurred following the installation of raised reflective pavement markers (RCA, 1983). Johnston (1983) generally confirmed the efficacy of line-markings and posts in curve delineation for both sober and drinking drivers. However, he did report an overall advantage for chevron markings and suggested that an optimum delineation treatment for curves at night is chevron signing with wide edge-lines. 3.7 MEASURES TO IMPROVE THE SAFETY OF INTERSECTIONS

The literature considers a range of treatments to improve the safety of rural intersections: grade-separation; staggering cross intersections; provision of traffic signals installation of roundabouts; and the provision of adequate sight distances, signing, pavement markings, delineation, channelisation and lighting. Generally, all of these treatments are used to attract a drivers attention on their approach to the intersection and to improve guidance through the intersection. At-grade four-way intersections are dangerous parts of the rural main road network (Austroads, 1995; Lydon, 1997; Toivonen & Niskanen, 1998). One measure suggested, even though it is a high-cost countermeasure, is the realignment of four-way intersections to create two three-way intersections (Austroads, 1995; Lydon, 1997). In addition, safety of rural intersections can be improved by increasing visibility, and by using an intersection layout that forces drivers to lower their speed. There are other suggestions such as conversion of busy and high speed at-grade intersections to grade-separated intersections (Toivonen & Niskanen, 1998), or at least provision of signalized intersections where the permitted speed limit is 80 km/h or above (Greibe & Nielsen, 1996). Moreover, there have been many calls to install roundabouts near built-up areas and on high-volume roads (Corben, Ambrose & Foong, 1990; Lydon, 1997; Toivonen & Niskanen, 1998). 3.7.1 Provision of grade -separated intersections

Grade-separated intersections remove the points of conflict and are generally provided on highways, freeways and motorways with high design speed and carrying high traffic volumes (Austroads, 1988; Hughes Amis & Walford, 1997; Kolle, 2000). However, they are an expensive treatment with low BCRs and are rarely considered on low volume, twolane, rural highways. Kolle (2000) stated that, to make grade-separated intersections costeffective, taking into account the high construction and land acquisition costs, they should only be considered on roads where the distance between intersections is at least 2 km and the travel speeds of vehicles do not exceed 70 km/h. At higher speeds (80 km/h and over), he suggested that grade-separated intersections should be considered only when separated by 3 km.

112

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

Where grade-separated intersections are provided, attention needs to be given to geometric features of entry and exit ramps, particularly for older drivers (Fildes et al., in press ; Staplin, Lococo & Byington, 1998). These include: Provision of longer entrance ramps (a minimum length of 280 m followed by a merge taper with a minimum length of 100 m for safe vehicle acceleration and operation). Lundy (1967) demonstrated reduced crash rates for both exit and entrance ramps when lengths of ramps were increased (274 m for exit ramps and 244 m for entrance ramps). Cirillo (1970) also found that increasing the length of acceleration lanes reduced crash rates, but only if merging vehicle constituted more than six percent of the mainline volume, Provision of multiple advance warning signs at least 3.8 km (2 miles) before exit ramps to convey information to drivers and prevent wrong-way movements and wrong lane choice, and Provision of traffic control devices (signs and pavement markings) to prohibit wrongway movements on exit ramps. Wrong-way driving on freeways occurs when drivers enter an exit ramp from the minor road. Copelan (1989) argued that, while the halfdiamond, partial cloverleaf, trumpet and buttonhook ramp interchanges can be very susceptible to wrong-way movements, full cloverleaf interchanges are the most desirable type of interchange to avoid wrong-way movements. He also suggested that larger, highly reflective signs may be helpful for confused drivers. Slip ramps, too, have been shown to pose relatively few problems for drivers. Woods, Rowan and Johnson (1970) suggested that positive signing such as routemarkers, trailblazers and a freeway entrance sign, to indicate the correct path or turning manoeuvre to drivers, may help minimize confusion. A white pavement arrow placed at off-ramps pointing in the direction of the right-way movement can be effective in reducing the number of wrong-way movements (Parsonson & Marks, 1979; Tamburri, 1969). Provision of appropriate traffic control at at-grade intersections

3.7.2

Traffic control measures provide safe and efficient operation. They aim to clearly define the conflict area, which driver has priority, as well as provide advanced warning of the intersection, reducing the potential for traffic conflict and delays by sharing time between opposing movements. There are different control types that may be suitable for one intersection but unsuitable for another. Moreover, there are decisions to be made by road engineers as to which type of control is most cost-effective, particularly on low-volume rural roads. For instance, the provision of fully-controlled turning phases at traffic signals can work well for major roads as they prohibit filter turns, thereby reducing turning conflicts, however, they can also exacerbate existing delays on minor roads. In addition, roundabouts operate well at locations where there is high volume traffic on both approaches but significant delays may be experienced when an incompatible combination of conflicting traffic flows occurs. Wilson and Dunn (1994) developed an expert system to determine the most appropriate form of intersection treatment and control. Expert systems, or knowledge-based systems (KBS) are a relatively new generation of computer programs that are used to solve specific problems in a similar fashion to human specialists. Some of the measures that are most commonly applied to rural intersections to reduce crash and injury risk include installation of traffic signal control, advance warning signs and roundabouts.
COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS 113

3.7.2.1 Provision of traffic signals with fully- or partially-controlled turning phases As indicated, many of the problems at intersections occur when drivers attempt a turn across oncoming traffic. Specifically, this safety problem arises because of the conflict between turning vehicles and oncoming traffic, both when vehicles are turning from a minor road onto a priority road and have to cross oncoming traffic, and when vehicles are turning from a priority road onto a minor road, again across traffic, and the associated difficulties in determining acceptable gaps when there is no protective turning phase. The provision of fully-controlled turning phases that prevent filter turns can dramatically reduce the potential for crashes due to inappropriate judgments of gaps in oncoming traffic (Bui, Cameron & Foong, 1991; Hughes & Amis, 1996; Taylor, 1991). Bui et al. (1991) reported the results of evaluating the effectiveness of installing fullycontrolled right-turn phases at traffic signals in urban settings. They found reductions of 45 to 65% in all casualty crashes, depending on whether or not there was a partially-controlled right-turn phase already installed, and corresponding reductions of 82 to 93% in the target crash type, namely right-turn-against crashes. While these findings were for lower speed urban settings, they indicate that there is a high potential to reduce the incidence of serious crashes at traffic signals through the introduction of fully-controlled turn phases (right turns in England, Australia and New Zealand, left turns in the USA, Canada and many countries of Europe). On many low-volume rural roads, provision of signalised intersections with full-control of right-turns is often not considered justified by traffic engineers. Nevertheless, this type of traffic control does reduce the problems associated with gap selection and can dramatically reduce the incidence and severity of these types of crashes. Fildes et al. (in press) argued that, when full-control of turns is either not considered justified or where it is not achievable or practical, partial-control may be beneficial, especially if advance warning signs are also provided. They argued that, for this type of phasing to be effective in reducing crashes between turning vehicles and oncoming traffic, drivers need to understand the different operational modes of the signals and phasing. For this reason, they suggested supplementary traffic signs may be necessary for drivers to understand and assess the conditions under which they can safely complete the turn. 3.7.2.2 Provision, placement and form of stop and give-way signs Most minor intersections, particularly on low-volume, two-lane, rural roads are unsignalised and place greater demands on drivers, being more difficult to detect, recognize and respond to the traffic control device. While one option is the installation of signals to improve safety, it is clearly not practical to signalize every minor intersection on low-volume rural roads. These intersections should be controlled by stop or give-way signs and by holding lines on the minor road approaches (Fildes et al., in press; Staplin et al., 1998). Fildes and his colleagues argued that it is essential to use conspicuous and well-placed retro-reflective signs and pavement markings to enhance awareness of the presence of the intersection and provide enough time for drivers to respond to the message. Mace and Pollack (1983), too, noted that sign conspicuity is important and developed a model to derive the retro-reflective levels required for adequate visibility distance, taking into account driver visual performance, as well as driver response requirements to the information presented on a given sign when encountered in a given situation with an assumed operating speed for signs of varying size and placement. Taking speed and sign application into account, they recommended that

114

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

retro-reflectivity for stop signs ranged between 10 cd/m2 /lux up to 24 cd/m2 /lux with significantly higher values for give-way signs (between 24 and 39 cd/m2 /lux). 3.7.2.3 Provision of roundabouts Roundabouts are becoming increasingly popular as safe traffic control devices. They are designed to control the traffic flow at intersections without the use of stop signs or traffic signals. Hughes (1994) compared crash and injury risk by intersection type and found the lowest crash severity index occurring for intersections controlled by roundabouts (0.13). Persaud, Retting, Garder and Lord (2001) also examined the use of roundabouts in the USA and found that there was a significant reduction in crash severity at intersections that had been converted from other forms of traffic control (signals, stop and give-way signs) to roundabouts. They noted that the reported crash reductions at roundabouts could primarily be attributed to two factors: reduced traffic speeds, and elimination of specific types of motor vehicle conflicts that frequently occur at angular intersections. A study by Brewer et al. (2001) of a number of European countries found that roundabouts were extensively used, with two main design concepts. The first design is the flared, tangential approach that allows comfortable travel without significant loss of speed, while the second design is the radial approach where lower entry speeds force drivers to slow down when entering the intersection. The second design for roundabouts is mainly being used for speed reduction and traffic calming. Corben et al. (1990) evaluated 116 accident black spot treatments undertaken at intersections following the commencement of a Statewide program in 1979. The study evaluated the effectiveness of the program and its individual types of treatments for both metropolitan and non-metropolitan regions of the state of Victoria, Australia. While the overall program was successful in cutting intersection casualty crashes by a statistically reliable 33%, it was more effective outside of the metropolitan area with a 69% reduction in casualty crash frequency, compared to a 30% reduction in metropolitan Melbourne. New roundabouts were the most impressive of the intersection treatments, showing casualty crash reductions of 81%. Corben et al. estimated BCRs for all intersection treatments and found an overall BCR of 8.8, with a BCR of 19.1 for intersection treatments in rural areas. In the case of roundabouts constructed at black spot intersections (including nonmetropolitan intersections), the estimated BCR was 7.5. No other treatment types, other than new traffic signals or remodelled traffic signals, were found to reduce crashes in a statistically reliable way. In their more recent evaluation of blackspot treatments, Newstead and Corben (2001) also found that the most impressive of the intersection treatments, applicable to rural settings, was the construction of a new roundabout, where average reductions in casualty crash frequencies of 73% and in casualty crash cost of 87% were achieved. The estimated BCR for new roundabouts was 5.0. Though not specifically focussing on rural roads, Corben (1989) recommended replacing existing traffic signals with roundabouts, based on the proven, superior safety performance of roundabouts when compared to intersection signals. It is likely that this superior performance would be further accentuated in higher speed rural settings. Corben (1989) concluded that such a program had a sound economic base (favourable values of Net Present Worth were predicted for a range of typical crash history/construction cost scenarios) and would be well-suited to a mass action type of approach.

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

115

Providing roundabouts means that drivers need only select a gap in one direction of traffic at a time. This is a fundamentally simpler and safer task than choosing a coincident gap in two streams of traffic. Moreover, in the event of a crash, the injury consequences will be less severe because of the greatly reduced impact speeds and the lower angles of impact experienced under this form of intersection control. There are some problems, however, associated with the construction of roundabouts that should be noted. Persaud et al. (2001) conceded that roundabouts are not feasible or appropriate at every intersection, as a sufficient road space must be available (diameter of about 30 m). More importantly, there is some concern with regard to potential increased crash risk for motorcyclists, bicyclists and pedestrians at roundabouts, with reports of over-representation of these road users in crashes at intersections controlled by roundabouts when compared to other intersection types (Hughes, 1994). An important criterion for comparison between roundabouts and other control types is the severity profile for various crash or road use types. For roundabouts, severity profiles may be superior due to the more favourable angles and speeds on impact. 3.7.3 Provision of adequate sight distance

As indicated in Section 2.4.5.2, the provision of adequate sight distance on approaches to and entering intersections is fundamental to safe intersection design, however, many rural roads have features that can severely restrict sight distances, such as physical obstructions, severe grades and poor horizontal alignment. The provision of unrestricted sight distances is especially pertinent at intersections on rural roads where speeds are generally high. Indeed, studies have shown reductions of up to 67% in crashes where obstructions that inhibited sight distance are removed (David & Norman, 1979; Mitchell, 1972; both cited in Staplin et al., 1998; Strate, 1980). The research generally recommends that minimum standard intersection sight distance should be provided on both major and minor roads to ensure drivers will always have an unobstructed view of the entire intersection and sufficient lengths of the intersecting road. Current US and Australian guidelines suggest a perception-reaction time of 2s for rural roads (which is higher than the value of 1.5s for urban roads). Neuman (1989), however, suggested that these values may be too low, particularly on rural roads, and recommended that perception-reaction time values of up to 5s should be considered in some situations, particularly considering the complexity of the driving task and the function of the road. Likewise, Fildes et al. (in press) recommended increased perception-reaction time values with a minimum of 3.5s for approach sight distance on rural roads (and 2.5s for urban situations). Adequate sight distances for turning traffic on the priority road making a left-turn (in the USA and most European countries and equivalent to a right-turn in England and Australasia) across oncoming traffic also on the priority road are essential to safe intersection design. The available sight distance in these situations depends on the degree to which the drivers line of sight is obstructed by opposite turning vehicles and the extent to which it is limited by the alignment of the turning lanes. Adequate sight distances for this manoeuvre will ensure turning traffic will always have an unobstructed view of the entire intersection and sufficient lengths of the road ahead and this can be achieved through offsetting opposing turning lanes (Fildes et al., in press; Staplin et al., 1998).

116

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

3.7.4

Provision of adequate delineation and alignment

There are a number of geometric design measures that have been suggested to improve the safety of intersections. These include channelisation and delineation of turning lanes, and provision of medians and traffic islands. 3.7.4.1 Provision of channelisation and delineation Channelisation is used to delineate paths of travel for turning movements at intersections and there are a number of safety benefits associated with the provision of channelisation at intersections, particularly the provision of raised channelisation. These include separation of turning vehicles from the through traffic stream, increased proper use of travel lanes at intersections and reduction in the frequency and severity of crashes. In addition, enhanced signing and delineation are also considered as effective measures to reduce crashes at intersections (Bonneson et al., 1993; Lydon, 1997). A number of studies in the USA and Australia have examined the effectiveness of channelisation at intersections from a safety perspective and the majority have found associated benefits. While not all of the studies examined these treatments for rural intersections exclusively, there is potential for these treatments to have a beneficial effect for these locations. For example, Strate (1980) showed that channelisation produced an average BCR of 2.31, and Hagenauer, Upchurch, Warren and Rosenbaum (1982) found that the channelisation of intersections reduced crashes by 32% and injury crashes by 50%. McFarland, Griffin, Rollins, Stockton, Phillips and Dudek (1979) compared the effect of painted channelisation and raised channelisation separating the left-turn lane from the through-traffic lane at non-signalised intersections. They found that painted channelisation resulted in reductions of crashes by 50%, 30% and 15% for rural, suburban and urban areas, respectively. When raised channelisation devices were used, the crash reductions were more substantial at 60% in rural areas, 65% in urban areas and 70% in urban areas. Likewise, Morris (1992) argued that well maintained road markings that provide visual cues for safe lane position are equally important as traffic signals. Most of these US studies are now quite dated, however, a recent evaluation of the Transport Accident Commission (TAC) funded crash black-spot program in Victoria, Australia, (Newstead & Corben, 2001) showed that channelisation was highly effective in reducing casualty crash frequencies and costs. An estimated 36% reduction in casualty crashes was found, along with an estimated 55% reduction in casualty crash cost. There are associated features that can improve the detection of the proper path and enhance the safety benefits of channelisation including lane-use control signs, lane-use arrow markers, pavement markings and delineation of edge lines and median noses. Edge-lines that clearly delineate the edge of the travelled way at intersections are marked to discourage travel on road shoulders, to make driving safer a nd more comfortable, and can reduce crashes resulting from lane-keeping errors. McKnight, McKnight and Tippetts (1998) examined the effect of edge-lines on driving performance, examining the combined effect of lane line width and line-pavement contrast upon lane-keeping in terms of heading error, position error, lane excursions and road excursions in a simulated driving environment. They showed that lane line width and contrast have large effects upon lane-keeping performance at extremely low levels of contrast and that low contrast levels are present at night, on wet roads and when light snow covers the pavement.

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

117

As for delineation of road edges, delineation of kerbs is important for drivers to manoeuvre their vehicle through intersections. Proper delineation of kerbs enhances the conspicuity of traffic islands, medians and visually separates shoulders from the roadway. Fildes et al. (in press) made a number of recommendations to intersection design with regard to channelisation and delineation. While the emphasis of this report was on providing a safe environment for older drivers, these improvements apply to all-aged drivers: The use of raised channelisation with raised reflective pavement markers maintained at a luminance contrast level of 3.0 for both left- and right-turn lanes, Provision of features to prevent wrong lane-use when making a right-turn (equivalent to a left-turn in the USA and most European countries). These features include laneuse control signs, lane-use arrow markers, pavement markings and delineation of median noses using reflectorised paint, For at-grade intersections with unsealed shoulders, provision of a minimum contrast level between the painted edge of the roadway and the roadway surface is recommended. The minimum levels of contrast recommended are 2.0 for intersections with overhead lighting and 3.0 for intersections without overhead lighting, All kerbs at at-grade intersections should be semi-mountable and painted on the slope and at least a portion of the top surface. Proper delineation of kerbs enhances the conspicuity of traffic islands, medians and visually separates footpaths from the roadway, thereby reducing problems associated with lane-use as well as avoiding collisions with the raised surface.

Rumble strips are also suggested as appropriate measures to reduce crash and injury risk at and on the approaches to hazardous rural intersections. They are areas of coarse or grooved pavement surfacing, often intermittently spaced in the direction of vehicle travel. This technique uses sound and vibration that are transmitted into the car when straying from path and effectively warn the drivers of straying vehicles, as well as indicate to drivers to slow down when approaching hazardous intersections. Stackhouse and Cassidy (1996) argued that these are an effective measure as drivers tend to react more to audible than visual signals. 3.7.4.2 Provision of medians As indicated in Section 3.4.2.1, wide medians are considered an effective measure to reduce head-on crashes resulting from vehicles encroaching onto the opposite travel lanes and naturally form part of intersection design when minor roads and/or adjacent properties intersect with major divided roads. Wide kerbed medians are a form of intersection treatment developed specifically for intersections between priority divided roads with minor two-way undivided roads. The aim of the treatment is to prevent drivers on the minor road crossing the divided road at high speed or mistaking the first carriageway as a two-way road and making a wrong-way movement. While wide medians are usually installed as a safety measure on busy and high-speed rural roads, there are potential disbenefits when minor roads intersect. Drivers on minor roads need to be made aware of the permitted movements through these types of intersections and it is generally noted that the provision of regulatory or warning signs can increase the conspicuity of medians, alert

118

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

drivers to the hazards of the intersection, advise them of conditions that require caution and guide traffic into appropriate lanes, therefore reducing conflicting movements. Crowley and Seguin (1986) pointed out that median width is a key factor in the incidence of wrong-way movements, stating that, when medians are extremely narrow, there appears to be little confusion that the intersecting road is two-way, but when wide medians are provided there is more confusion. They also reported that some jurisdictions recommend the use of optional signs, i.e., DO NOT ENTER, KEEP RIGHT, WRONG WAY, but noted that these signs are not helpful to drivers on the approaching minor road because they are only detected when a driver begins a wrong-way movement upon reaching the intersection. In this regard, they recommended the use of a DIVIDED HIGHWAY CROSSING sign that has direct impact on the decision process of drivers on a minor road approaching a divided highway median. Harwood et al. (1995) also examined widths of medians at rural unsignalised intersections and recommended that medians should be wide enough to accommodate turning and crossing manoeuvres for at least a bus or truck. They also noted that wider medians at signalised intersections were associated with increased crash frequency and increased delay, and recommended that narrower medians be u sed in these situations. Bonneson et al. (1993) suggested that medians of 40 to 50 ft (12 to 15 m), which are most commonly used for rural expressways, have generally been found to provide a good balance between overall right-of-way width and safe traffic operations. They also noted that treatments to provide left-turn bays (equivalent to right-turn bays in England and Australasia) and indirect turns improved drivers field of view. 3.7.4.3 Provision of turning lanes at median breaks Two-way, left-turn lanes (TWLTLs) (equivalent to two-way right-turn lanes in England and Australasia) are at the other end of the spectrum of medians in terms of the amount of control that can be exercised over turns across oncoming traffic on the priority road into minor roads or adjacent properties. With TWLTLs, turning vehicles have unlimited access to adjacent properties and they can be used by turning vehicles from either direction of the highway. Aside from the benefits of providing a longer sight distance for turning traffic, these lanes can be used as a storage area for turning vehicles waiting for appropriate gaps in the opposing traffic stream. Since the vehicle is physically removed from the main traffic stream, it is argued that both safety and flow can be dramatically improved with the installation of a TWLTL on rural highways (Thakkar, 1984). Previous research has shown that crash rates decrease by approximately 20% or more after installing a TWLTL on previously undivided highways (Thakkar, 1984; Walton, Machemehl, Horne, & Fung, 1979). Furthermore, Thakkar reported that TWLTLs can also reduce particular types of crashes, such as rear-end and sideswipes, by as much as 30%. Thakkar proposes that this reduction is due to the fact that TWLTLs provide an area for left-turning vehicles to queue outside of the through travel lanes while waiting to turn. TWLTLs also provide a refuge and merging area for vehicles turning left out of minor roads and adjacent properties. Bypass lanes are a viable alternative to exclusive left-turn lanes (equivalent to right-turns in England and Australasia) at T-intersections because they are less costly. Preston et al. (1999b) estimated the costs for installation of bypass lanes in the range of US$50,000 for roads with paved shoulders to more than US$200,000 for a full set of left-turn lanes.

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

119

3.7.5

Provision of warning signs at intersections

Warning signs are provided to alert drivers of hazardous conditions that may not be apparent. Hazards may include road geometry or environmental conditions, such as intersections with poor sight distance. The provision of signing on the approach to an intersection, particularly rural intersections with high operating speeds, warning of the presence of the intersection and traffic control devices can reduce problems associated with detecting the presence of the intersection, thereby reducing crash and injury risk. There is general consensus that a combination of traffic signals and advance warning flashing beacons are an effective corrective measure at rural highway at-grade intersections and are regularly used in the USA. Stackhouse and Cassidy (1997) evaluated the effectiveness of flashing beacons at rural intersections in the USA and found that driver alertness or awareness to potential hazards were enhanced by the installation of flashers. Recently there has been a shift toward variable message signs and vehicle-activated signs. Vehicle-activated signs have the advantage of being blank when not activated and limiting the drivers visual intrusion, and lighting up only if excess speeds are detected for the conditions on that particular stretch of road. For this reason, vehicle-activated signs have been suggested as effective measures on approaches to intersections or on bends as they put the hazard into context (UK Department of Transport, 2001). While they may be at least initially effective, there are a number of disadvantages to standard signs and flashing lights at hazardous rural locations: they add to the visual clutter of the road environment (although this is less of an issue in rural areas than in more builtup areas); they may require a power source (although solar power may be an option); flashing lights that are sufficiently bright during the day may be too bright at night; and habituation may occur such that drivers cease to pay attention to them over time. A relatively recent innovation is the vehicle-activated sign, which only lights up when triggered by a passing vehicle exceeding a preset condition usually speed-related (although Hellier-Symons (1983, cited in Winnett & Wheeler, 2002) found that a warning sign that activated when drivers were travelling too close to the vehicle in front was effective in increasing following distances). Modern versions use LEDs so that power consumption is low, and their brightness can automatically vary according to the ambient light levels. Winnett and Wheeler (2002) evaluated 60 sign installations on single-carriageway roads throughout rural England. The signs were of four types: speed limit notifications within reduced limit areas (such as the approach to a township), bend warnings, junction warnings, and speed camera warning signs. The bend and junction signs lit up when an approaching vehicle exceeded what was considered to be a safe approach speed for the situation (set at the 50th percentile of free speeds at the location prior to installation). Winnett and Wheelers (2002) signs warning of a junction or bend did not specify a speed, only the message slow down and a figure of the hazard being approached (i.e. a bend or an intersection). There was a mean speed reduction of up to 4 mph (6.4 km/h) (the speed limit reminder signs yielded a speed reduction of up to 14 mph (22.4 km/h)). Across all sites in one of the counties there was also a statistically significant reduction of a third of the number of crashes than would have been expected without the signs. Other signs had not been in place for sufficient time to allow for adequately meaningful data.

120

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

According to Winnett and Wheeler (2002), operating costs of vehicle-activated signs are low and there seemed to be little or no habituation to the warnings such they ceased to have any effect, at least up to three years after their installation. This is consistent with HellierSymons (1983, Winnett & Wheeler, 2002) finding that the effect of their following distance warning sign was still present five years later. Vehicle-activated signs could be used on the approach to any hazard where inappropriate speed significantly increases the likelihood that the driver will not be able to negotiate the situation safely especially where the prevailing speed limit for the approach road is too fast for the hazard. The technology can also be used for signs triggered by other events. For example, signs designed to sense rain, ice or darkness can activate when the conditions are sufficiently dangerous. However, any installations using such technology will be more expensive to install and maintain compared with standard signs. 3.7.6 Provision of adequate lighting at intersections

Adequate lighting is essential to increase visibility of the roadway and its immediate environment at night, thereby enabling the driver to manoeuvre safely and efficiently through an intersection. Current guidelines, particularly those in the USA and Australasia (AASHTO, 1990; Austroads, 1988), however, take a pragmatic approach for the provision of street lighting, taking into account the significant costs associated with installation, operation and maintenance of lighting, particularly in the current climate of increased demands for energy conservation. This is especially so in rural areas on low-volume twoway roads where it may not be practical to light every intersection. Despite this approach, research clearly shows the benefits of providing street lighting at intersections in terms of crash incidence and severity reduction. The Commission International de lEclairage (1990) evaluated the proportion of crashes at intersections of differing lighting levels at 62 sites in 13 OECD countries and found lighting to be beneficial, with approximately one-third of the results statistically significant. According to Preston and Schoenecker (1999a) the installation of street lighting at rural intersections is a low cost and very effective strategy for reducing night-time crashes. In their examination of the effect of installation of street lights at rural intersections on crash risk, they found a reduction in night-time crashes at intersections where lights were installed and noted that the benefits associated with the installation of street lighting at rural intersections outweighed the cost by a ratio of 15 to one. They also argued that increased lighting, together with channelisation, rumble strips and overhead flashing beacons can substantially reduce the number and severity of crashes. With regard to the provision of lighting on freeway interchanges, Gramza, Hall and Sampson (1980) investigated the effects of lighting on crash experience at interchanges and found that the presence of high-mast lighting significantly reduced total crash rates, total crashes involving fatalities and injuries, and crashes involving fatalities and injuries other than vehicle-to-vehicle and vehicle-to-fixed-object categories (e.g., crashes caused by striking pedestrians). Moreover, increases in the illumination level at interchanges were associated with significant reductions in vehicle-to-fixed-object crashes involving property damage and vehicle-to-vehicle crashes involving fatalities and injuries. In addition, Janoff, Freedman and Decina (1982) compared the effectiveness of partial lighting of interchanges with complete interchange lighting systems. They found that complete interchange lighting provided a better traffic operating environment than did

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

121

partial interchange lighting and that any interchange lighting performed better than no lighting. 3.7.7 Countermeasures for rural rail-road intersections There are essentially four ways to significantly improve the levels of safety at rail crossings: 1. 2. 3. 4. Require trains to slow or stop as they approach HRIs Remove the at-grade crossing Modify the crossing (e.g., convert passively controlled crossings to active crossings); Convince road vehicle drivers to pay more attention and actively search for trains as they approach and set out to cross the rail line; and if they see a train they should immediately give way to the train.

3.7.7.1 Slowing the train In terms of issues such as fossil fuel use and greenhouse gases, train travel is a more environmentally friendly mode of transport than most road-based travel. Increasing rail travel times by requiring trains to slow significantly or stop at crossings is likely to discourage rail patronage. The large amount of energy required to overcome the inertia to accelerate a train again is also very energy intensive (particularly for high speed crosscountry trains without regenerative braking). The additional time for an accelerating train to clear a crossing is also likely to increase the frustration of road users waiting to cross the tracks. 3.7.7.2 Removing the crossing While there is a push to remove at-grade crossings, particularly for very high-speed trains (Ries, Carroll, & Warren, 2002), the crossings being replaced are most likely to be busy crossings. Accordingly, these crossings are likely to already be equipped with active warning devices or boom gates. Due to the large number of passive rural crossings and their lower levels of use, the cost of installing grade-separated crossings would be prohibitively expensive. 3.7.7.3 Modifying the crossing Allan (1992) quotes (but does not name) US research that says that flashing lights and bells can be expected to avert 65% of the crashes that would occur with crossbucks (also called a St Andrews Cross the X -shaped rail warning sign) alone, and that half-arm barriers can avert 88% of the crashes that would occur with cross bucks alone. While active crossings are considered safer than passive crossings (Wigglesworth, 2001), not all crashes will be prevented while the HRI is still at-grade it is a common occurrence for drivers to attempt to beat the lowering of barriers or to simply drive around them (Davis, Alexander & Brissis, 1997; Hall, 1999). Allan (1992) advocates applying a cost-benefit analysis to each crossing to determine the financial viability of additional protection. Benefits were in terms of lives saved (using a value of life in New Zealand in 1992 of two million dollars) by preventing an expected number of crossing crashes per year calculated using the crossings Product number.

122

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

The product number is an index derived from variables such as traffic volumes, road type, number of trains, vehicle speeds, etc. Costs were based on installation, maintenance, etc. Allan (1992) found that when crashes occur at least as often as one every 30 years, new installations of flashing lights and bells are justified. If crashes occur every 15 years or more often then the added cost of half-arm barriers is justified by the added safety benefit. Other systems exist for systematically determining a risk index for particular rail crossings. For example, Hughes (2002) discusses the Australian Queensland Risk Based Scoring System for level crossings, which also allows for a calculation of the reduction in risk of any particular countermeasure being considered for the crossing. This system also allows for types of accident mechanism to be incorporated into the equation. According to the author, the system has been favourably evaluated against crash histories at particular crossings. Due to their remoteness, many rural HRIs may not have a power source present and all active countermeasures require a power source. However, in some regions solar powered battery storage systems may be an option, particularly if light emitting diodes (LEDs) are used. LEDs are advantageous due to their increased brightness over incandescent sources (and so they are more effective during daylight hours) and low power usage. As a large number of individual LEDs are often used to make up a single light source, a number of individual bulbs can break down without significantly affecting the overall performance of the device this would be an additional advantage in rural areas as it would increase the time between maintenance checks. Mornell (1997) discusses an evaluation of a number of countermeasures. Rumble strips were found to be unsuccessful as drivers often used the other lane (for oncoming traffic exiting the crossing) in order to avoid the strips (strips were not placed on these lanes due to a concern for rear end crashes). Strips would therefore need to be combined with permanent barriers that did not allow vehicles to travel in the wrong lane on the approach to the crossing. Mornell also describes a field evaluation of additional elevated signals. Once residents were familiar with the time delay between the onset of the signals and the lowering of barriers, this countermeasure was found to actually encourage drivers to speed up in order to cross before the barriers lowered. After an extensive review of available and potential countermeasures, Wigglesworth, Corben and Triggs (1999) narrowed 18 potential devices to three that they considered appropriate for passive rail crossings. Their first recommendation was something akin to a standard traffic signal, so that there would be minimal learning required of drivers. Due to power considerations (they advocate the use of solar power) they point out that the device would have to be silent until activated. However, it would seem that such a device is functionally similar to the device used in Australia, except that a standard traffic light operates in a steady state while the current warning device flashes four red lights as a train approaches (at installations where power is available). Indeed it is generally accepted that flashing lights are superior in attracting attention, and a train installation needs only two states warning and all clear which correspond to lights flashing and lights silent. Another lighting device suggested by Wigglesworth, et al. (1999) is a sign that, when activated, would display lit wording, such as Stop train approaching. Such a sign would be unambiguous to the driver by providing a clear written instruction (i.e., Stop). However, it is not clear how such a sign would be otherwise advantageous over flashing lights.

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

123

Rechnitzer (2002) discusses a conceptual device that would act as a rumble strip only when a train approached (and so would technically be classed as an active measure). The system consists of a number of hydraulic tubes embedded in the approach to the crossing. An electrical device (which could be solar powered due to its low power consumption) maintains sufficient hydraulic pressure in the tubes so that they are flush with the road surface when a vehicles tyres roll over them. The approach of a train triggers the device to allow the pressure to be minimised, forming indentations in the road that perform the function of a rumble strip. An advantage of such a system is that in the event of a power outage or system malfunction the un-powered state is actually the active mode it is failsafe. As the device is at the conceptual stage, no data are available as to the propensity of the device to foul on the grit from road surfaces. Two other countermeasures relate to the maintenance of the crossing. As a driver approaching a passive HRI must see the train in order to avoid a collision, maximising sight lines and sight distance are critical. Russell (2002) suggests that a brush cutter may be the most cost-effective countermeasure that could be applied to passive HRIs. In addition to ensuring that drivers have good visibility of the rail approaches, unnecessary distractions should be removed so that drivers can focus a maximal amount of attention to determining whether a train is approaching. One potential source of distraction is the condition of the approach road and the crossing itself. Due to their remoteness and lack of traffic, many rural rail lines cross dirt or gravel roads. It would therefore be expected that drivers would focus attention on negotiating the crossing rather than more fully searching for approaching trains (Wrigglesworth, 2002). Also, if the HRI is at a significant elevation relative to the roadway, the motorist may not be able to see the approach or departure sections of road on the other side of the crossing. At a time when the motorist should be focussing most of their attention on whether a train is coming, they may also need to devote attention to whether there is oncoming traffic that also can not see them and whether there is stationary or slow moving traffic on the other side of the crossing, with the consequent risk of a rear-end crash. 3.7.7.4 Influencing road vehicle drivers According to Allan (1992), there is poor driver compliance with the duty to determine whether or not the way is clear before proceeding across a railway crossing. Due to the infrequency of trains in rural areas, many drivers barely slow down before crossing a railway line, and field studies have found that large proportions of drivers do not turn their head to look for trains in both directions or at all (Wigglesworth, 2001). Wigglesworth suggests that this may be due in part to some drivers expecting to be told that a train is approaching, as they would be at an active crossing a number of field studies yielded little difference in looking behaviour of motorists crossing active and passive HRIs and calls for advance warning signage to be different for active and passive crossings. Some jurisdictions have tried speed limits for rail crossings, but as Allan (1992) points out, arbitrary speed limits that can not be effectively enforced are not respected. The same would apply to placing stop signs at crossings (Wigglesworth, 2002). Wigglesworth (2002) also points out that the time required for a large, loaded truck to clear a crossing from a standing start at an approach stop sign would probably exceed the time taken for a train to traverse the sight distance from the crossing.

124

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

Education campaigns on a national level and at community level to target the crossings in particular areas may be needed. For example, Operation Lifesaver is a national funding program in the USA for local initiatives publicising the dangers of rail crossings. 3.7.7.5 Other countermeasures and issues Since most crashes between trains and vehicles (or pedestrians) occur because a driver/pedestrian has ignored the rules or not noticed the impending danger, it is a traffic problem rather than a rail problem per se (at least while trains have the right-of-way at these intersections). As such, the countermeasures canvassed thus far focus on the road aspect of the intersection and road vehicle drivers. However, there is some provision for increasing the safety at crossings through modifications of the rail aspect of the intersection or the train. Rechnitzer (2002) discusses the usefulness of an attenuating device fitted to the front of trains, such as an airbag that could be deployed by the train driver in the face of an impending collision. Such a device does nothing to prevent a collision, only decrease its severity. As most train/road vehicle crashes involve a train impacting the road vehicle (Wigglesworth, 2002; Russell, 2002), modifying the train has great potential. While most rail crossing crashes occur during daylight hours (ATSB, 2001, cited in Wigglesworth, 2002; Pajunen, 2002), opportunities exist for increasing the alerting or visibility properties of the train, such as fitting strobe lights (Johnstone, 2002) and using retro-reflective materials and paints on the front and sides of trains (Russell, 2002). Wigglesworth (2002) discussed the possibility of a louder horn, but points out that it often would need to exceed the threshold of pain at its source to potentially be of sufficient benefit obviously not practical from the point of view of the train driving crew or those living beside the rail line. Russell (2002) discussed the merits of using retro-reflective materials and paints on signs (back and front) and signposts at crossings. Russell also described a new sign developed by 3M that picks up the light from an approaching train and reflects it at a ninety degree angle, through a panel that can be mounted with lettering or symbols it is said to appear as though the sign is lit from within, but actually requires no power sources. According to Russell and Rys (1996, cited in Russell, 2002), the sign receives sufficient light from the approaching train to appear to light up at distances in excess of 650 m, and can be seen by a motorist several hundred metres away. Wigglesworth et al. (1999) also discussed a version of the standard crossbuck sign that incorporates a three-dimensional component that reflects the oncoming trains headlight out along the road approaches. However, it must be noted that with both this device and the one developed by 3M, the illumination properties are obviously not effective during daylight, although both would still function as a passive sign. An additional important shortcoming of these reflective devices is that if they are somehow rotated (unintentionally or otherwise), the angles required to reflect light from the train to the road driver can be altered and the devices effectiveness lost, or at least diminished. 3.8 OTHER IMPROVEMENTS

A number of improvements to other road features were identified. These include improvements to vertical curves, improvements to bridges and culverts, improvements to

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

125

construction zones, and improvements aimed to reduce pedestrian and cyclists crash and injury risk. 3.8.1 Measures to improve the safety of vertical curves

While vertical curves are identified in the literature as a problem on rural roads, surprisingly few studies have addressed potential measures to improve the safety at these locations. This may be due to the fact that there are other confounding problems at these sites, particularly the problems of short sight distances and vehicles stopping in the roadway to make a turn either into a driveway, access or turning at an intersection. Fitzpatrick et al. (2000) recommended that minimum stopping sight distance for safe operations on vertical curves with a speed limit of 90 km/h should be somewhere between 95 and 110 m and any longer (when other hazards are not present) would produce an unfavourable BCR. In addition, they recommended that placement of driveways and intersections near crest vertical curves should be carefully reviewed in the design of new roadways or in reconstruction of existing roadways. 3.8.2 Measures to improve the safety of bridges and culverts

It appears that road geometry is a significant contributor to rural bridge and culvert crashes (Evans, 1997). To date, research examining the effectiveness of countermeasures at bridges and culverts has been largely inconclusive (Bowman & Brinkman, 1988). Evans (1997) suggests that there are several reasons for this. First, crashes on individual bridges are rare events. Secondly, the exact date of installation of many countermeasures is not known, which leads to confusion regarding whether some crashes occurred before or after the installation of these treatments. Nevertheless, it appears that bridges may be amenable to a number of countermeasures, and that consideration of the ways in which bridge structures contribute to crash frequency could lead to better planning in the construction of new bridges (Kloeden et al., 1999). These countermeasures can be broadly split into two main categories: structural (e.g., width of approach and the bridge and installing guard-rails); and approach-related (e.g., signage and pavement markings). 3.8.2.1 Widening of bridge and shoulder width While it is noted that limited research exists on the crash effects of the geometric characteristics of bridges, it is reported that bridge width and the width of the bridge in relation to the approach width are the two m ost important features affecting crash rate at these locations. Widening is considered the best countermeasure of all, to improve bridge safety, as it has the capacity to reduce the frequency of crashes, not just the severity (Gunnerson, 1961, cited in Turner, 1984; Hilton, 1973). Smith (1982) reported that crash frequency could be reduced by as much as 15% when rural bridges were widened from 5.4 to 6 m and suggested that, in some cases, widening the bridge could be achieved by removing bridge railing. However, Smith also noted that this could only be done when there was no hazardous drop (greater than 2.7 m) at either side of the bridge. Likewise, Turner (1984) suggested that the minimum desirable bridge width should be 6 ft (2 m) greater than the width of the travelled way. He also suggested, as a minimum, shoulders of at least three ft (1 m) be provided on bridges. Interestingly, Ivey et al. (1979) found that when the approach roadway was widened, but not the bridge, the number of crashes increased, suggesting that the bridge-to-approach width ratio should be as small as

126

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

possible. Indeed, they recommended that all new bridges be built with a bridge-to-approach width ratio no greater than 0.8. Drawing upon the results of Turners study, the FHWA (1992) developed a table of predicted crash reduction effects of bridge widening, in the form of shoulder widening on either side of the travelled way. The predicted crash reduction effects for a bridge with effectively no shoulders, widened by either two ft (0.6 m) or eight ft (2 m) on each shoulder ranged from 23% to 85%, respectively. Numerous intermediate values were also tabulated for other possible combinations of shoulder width both before-and-after widening. Based on a the results of modelling, Zegeer and Council (1995) also reported that adding a shoulder of 60 cm on each side, where previously there was no shoulder, led to a crash reduction of 23%. 3.8.2.2 Installation of guard-rails Another structural countermeasure, involving the installation of guard-rails, has been discussed in detail in Section 3.6.2. Guard-rails effectively protect the occupants of errant vehicles from colliding into bridge end posts, and direct drivers back to the road (Evans, 1997). Unlike bridge widening, guard-rails will not decrease the number of crashes that occur, but can decrease the severity of these crashes (Hilton, 1973). Brown and Foster (1966) estimated that severity of crashes could be reduced by the presence of guard-rails in approximately 90% of fatal and 80% of serious bridge and culvert crashes, however it is important to note that there may be, in some circumstances, an increase in crash frequency. Therefore, Hilton concluded that where bridge widening to reduce crash frequency is not feasible, the installation of guard-rails is the best countermeasure to reduce the crash severity on narrow bridges. Furthermore, the combination of guard-rail with driver warning devices such as signs and pavement markings, a reduction in the number of crashes may also be achieved (see also Agent & Deen, 1976; Zegeer & Council, 1995). 3.8.2.3 Provision of signage and pavement marking Approach-related countermeasures include methods such as warning drivers of hazards (signage) and or improving the road delineation (hazard markers, rumble strips). As outlined previously, these countermeasures assist motorists by providing them with warning of, and information about upcoming hazards, including bridges (Bowman & Brinkman, 1988), without actually eliminating or reducing the extent of the hazard. Unlike the structural countermeasures discussed previously, approach-related countermeasures have the potential to reduce the frequency, but not the severity, of crashes that occur (Evans, 1997). Gandi, Lytton and Das (1984) recommended that approach-related countermeasures at bridge sites should include signs to inform drivers of reduced speed limits on approach to bridges, and that these signs should be paired with rumble strips that aid the driver in maintaining their course. Similarly, Hoque and Sanderson (1988) and Ogden and Howie (1989a) examined approach-related countermeasures and reported that reflective guideposts, edge- and centre-lines, raised reflective pavement markings, chevrons and bridge width markers were all effective measures to reduce bridge-related crashes. For example, Ogden and Howie (1989b), in their study of crashes at bridges and culverts, reported the results of various studies that found crash reductions of 15 to 18% where raised reflective pavement markings had been installed. Night-time crash reductions were estimated to be even higher.

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

127

3.8.3

Measures to improve the safety of construction zones

As indicated in Section 2.4.6.3, crashes on the approach to or through construction zones are particularly severe on rural highways, probably due to high travel speeds and/or high volume traffic, and mainly involve multi-vehicle crashes, rear-end impacts, and often heavy vehicles. There are a number of measures that have been suggested to improve safety at these locations. These include measures to provide warnings to drivers of the presence of the zone (and lane closures) and encourage drivers to reduce speeds on the approach to and through the construction zone, and measures to guide traffic through the zone safely. 3.8.3.1 Speed reduction measures One of the concerns raised in the literature is the high speed at which drivers drive through work zone areas, particularly in terms of safety of the workers around the area and because of changes to the road layout. Speeds need to be reduced on the approach to these zones so that drivers have the opportunity to negotiate travel through the zone safely, as well as make necessary changes to their travel if lanes are closed. Indeed, Richards and Dudek (1986, cited in Hall, 1988) argued that speed in construction zones should not exceed 15 mph (24 km/h). A common technique to reduce speeds on the approach to and through construction zones is the posting of a lower (regulatory) speed limit (Hall, 1988) and, while this may be an effective measure, as noted in Section 2.2.2, many drivers drive at speeds that exceed speed limits. Other forms of speed reduction, therefore need to be considered. For instance, some methods employed in the State of New Mexico include the use of flaggers, law enforcement personnel, variable message signs and lane width reduction (Hall, 1988). 3.8.3.2 Measures to warn of lane closures In many countries signs specify works on roads and other temporary situations as a general warning in advance of the works activity, and to positively guide drivers through, around or past the work site. They provide guidelines for use of a variety of supplementary signs and devices for lane closure. One such method is the provision of flashing arrows that convey the message that a lane is closed and drivers need to change lanes. Generally, they convey this message effectively and reduce confusion and non-compliance, thereby increasing the safety and efficiency of traffic in work zones (Fildes et al., in press). Faulkner and Dudek (1981) found that a supplemental arrowboard placed before the beginning of a lane closure taper was effective in eliciting early merging and argued that this occurred because the effective sight distance to the work zone was increased. In addition, Lyles (1981) evaluated the effectiveness of alternate signing sequences for providing warning to drivers of work zones that required a lane closure on two-lane rural roads. The signs tested included a standard MUTCD warning sequence, the same sequence augmented with continuously flashing warning lights on the signs, and a sequence of symbol signs. The most effective sign sequence was one that was flasher augmented. This treatment was twice as effective as similar signs with no warning lights in slowing vehicles in the vicinity of the lane closure. Hummer and Scheffler (1999) examined the effectiveness of fluorescent orange warning signs placed at the beginning of lane closure tapers on the approach to work zones. They found that, while most traffic conflicts (i.e., conflicts between two vehicles where one brakes or swerves to avoid hitting the other) occurred at the taper and most were rear-end

128

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

or lane-change conflicts, the number of conflicts were reduced at treated sites, compared to untreated sites. They also found a five percent reduction in the number of vehicles in the closed lane at the midpoint of the taper, and a reduction in speed variances. Another method used to convey to drivers the presence of work zones is the provision of upstream supplementary vehicle-mounted advance warning signs of lane closure. A number of studies have been performed to determine the effectiveness and driver comprehension of static signs and variable message signs for lane closures. Hanscom (1981) evaluated the effectiveness of variable-message signing in advance of highway construction areas to elicit appropriate lane-changing behaviours approaching left and right lane closures. Before-and-after studies conducted in three US states consistently demonstrated beneficial traffic operational effects of this application. Increased advance preparatory lane change activity, smooth lane change profiles and significantly fewer late exits (exit from closed lane within 100 ft) were observed. Moreover, Hanscom (1981) found that driver ratings of traffic control device adequacy were highest during the presence of a message conveying combined speed and lane closure. Drivers reported that this message was the most helpful, easiest to read, met all information needs and that they were most likely to change lanes early and reduce speed when the speed and closure message was displayed. It was recommended that these devices should be located 0.75 miles (1.2 km) in advance of lane closure and should be a supplement to the arrowboard. Dudek, Richards and Faulkner (1981) also evaluated the effects of variable message signs on lane changes at a work zone lane closure by measuring the percentage of vehicles that remained in the closed (median) lane as traffic progressed toward the taper. They found that the variable message sign encouraged drivers to vacate or avoid the closed lane, compared with driver responses at the same site without the use of a variable message sign. The percent volumes in the closed lane were significantly lower (around 46%) when a lane closure message was displayed than during periods when the sign was blank. 3.8.3.3 Measures to guide drivers through construction zones Driving in construction zones is a more complex task than driving on a regular section of road because more attention is required to maintain safe lane position and speed throughout the area. A number of effective devices to guide drivers around work zones safely and appropriately have been suggested and channelisation devices are effective. Portable safety barriers are widely used to separate opposing traffic and to separate traffic from long-term work activities and roadside hazards. The addition of raised pavement markers increases delineation by creating a brighter path for drivers to follow and results in better lateral positioning within lanes and fewer encroachments (Fildes et al., in press). Hall (1988) identified other methods used to improve guidance through construction zones. These include: Temporary rumble strips in advance of construction sites. This tactile method notifies drivers that there is a hazard ahead. Chevron signs can provide clear, unambiguous information about the path and speed of travel in a construction zone. Wing barricades, when placed at the beginning of a construction zone can give the illusion of road narrowing. This conveys the impression to the driver to be cautious, as the road ahead may not be as they expect.

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

129

Crash attenuators may not reduce crash frequency but can reduce crash severity through a reduction in energy transfer during a collision.

The New Mexico State Highway Departments Five Year Plan estimates that $3.3 billion will be needed to fund 468 separate improvement projects on the state rural highways. The purpose of the improvements is to enhance safety on the road system and thereby reduce the crash rate. Therefore, any increase in the number of crashes during the construction period must be out weighed by a reduction in crashes following the completion of the improvements (Hall 1988). 3.8.4 Measures to improve the safety of pedestrians and cyclists

Ribbens (1986) noted that the road shoulders are the only areas pedestrians and cyclists have along rural roads, and regarded provision and maintenance of wide road shoulders in densely populated rural areas as an essential interim measure to serve as a path for pedestrians and cyclists alongside roads. Ribbens also noted that the ideal solution would be to provide a pathway within the road reserve but separated from the carriageway. Given that many of the pedestrian and bicyclist crashes in South Africa occurred close to bus stops, shops and areas where there are large numbers of pedestrians and cyclists, Ribbens argued that provision should be made for such large numbers of vulnerable road users. Wide paved or gravel areas, so that buses can pull off the road to allow an unobstructed view of pedestrians and cyclists by oncoming drivers, were identified. In addition, Ribbens identified problems at bridges where there were inadequate pathways for pedestrians and cyclists thereby forcing them onto the carriageway and suggested improved bridge design with proper road shoulder, drainage, and signs to allow more room for pedestrians and cyclists. In rural areas, separation of slower road users (pedestrians and cyclists) from the motorised traffic should be considered. Construction of separate bicycle and pedestrian lane networks is an important safety factor (Hehlen, 1992). Gaca and Tracz (2000) suggested a range of countermeasures to improve pedestrian safety on rural roads in Poland. In addition to education and enforcement measures, they suggested land-use planning measures such as development of networks of access roads and walkways serving housing estates, location of facilities to minimise pedestrian traffic along and across main roads, engineering measures, such as building and separating pedestrian facilities along road (widening of shoulders, provision of separate walkways), improvement of visibility conditions (removing obstructions, lighting of critical road elements), placement of fencing and other obstacles directing pedestrians to safer and more convenient places to cross, provision of refuge islands, signs and marked crossings, and implementing speed reducing measures.

130

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

SUMMARY

Deaths and serious injuries from traffic crashes occur on an alarming scale throughout the world and have done so for many decades. At least one and as many as two of every three deaths occur in rural areas. This report presents the findings of a literature review of rural road trauma and the road infrastructure measures used, or being developed, to reduce this trauma. A particular focus is the identification of cost-effective infrastructure safety measures. To gain most from the application of the limited resources available for safe infrastructure, it is necessary to be strategic in addressing the overall problem. Thus, a focus of this report is on infrastructure that can fundamentally reduce crash risk, injury, or both, for the key crash categories in rural areas. 4.1 KEY CRASH TYPES

There is a relatively small number of crash categories predominating in rural areas. These are: single-vehicle, run-off-road crashes and multi-vehicle crashes, both head-on and rearend crashes. 4.1.1 Single-vehicle Crashes

Single-vehicle crashes account for around one-third to one-half of serious trauma. These crashes typically involve vehicles leaving the roadway, often at high-speed, and colliding with roadside objects, overturning on non-traversable terrain or otherwise leading to injury to riders or vehicle occupants. Crash risk is heightened on curves, while injury risk rises considerably where roadside trees, poles, embankments or narrow bridges are present. 4.1.2 Multi-vehicle Crashes

Multi-vehicle crashes also account for a major component of death and serious injury, w ith most occurring at intersections (or driveways); in head-on collisions; or as rear-end impacts, especially in the vicinity of intersections or construction zones. Common intersection crash types are cross-traffic, turn-against oncoming traffic and rear-end collisions. These crashes tend to occur when drivers fail to perceive the presence of intersections in high-speed settings, fail to select safe gaps in approaching traffic or approach intersections at speeds that will not allow them to avoid impacts with vehicles ahead, either slow-moving vehicles or stationary vehicles that are waiting to turn. It is common for head-on crashes to take place away from intersections, when drivers drift into opposing traffic lanes (on either undivided or divided roads), especially on curves, or otherwise lose control and encroach into opposing traffic lanes. 4.1.3 Vulnerability Considerations

Particular categories of road users that also warrant attention are pedestrians, cyclists, motorcyclists, and older drivers and passengers, because of their elevated vulnerability both in traffic and in traffic crashes, and heavy vehicles because of their greater threat to the safety of others in collisions with them. The greater threat posed by heavy vehicles relates to higher crash risk, because of longer stopping distances, and to reduced manoeuvrability. However, in the event of a crash with another vehicle, injury risk is also

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

131

elevated due to the substantially greater mass and incompatible design features of heavy vehicles, compared with other vehicles. Most vehicles, even with the highest standard safety features, can offer only limited protection to their occupants in many of the high-speed crashes prevalent in rural areas. 4.2 INFRASTRUCTURE COUNTERMEASURES

While many infrastructure countermeasures have been used to reduce road trauma in rural areas, relatively few have been comprehensively evaluated. In particular, quantification of cost-effectiveness has been conducted in only the minority of cases. 4.2.1 Single-vehicle Crash Countermeasures

To address single-vehicle crashes, a variety of infrastructure measures has been used and, in some cases, evaluated. Infrastructure measures include treatments to reduce speed and speeding, such as lower speed limits, traffic calming devices on the approach to and within towns, and a range of innovative measures designed to influence driver perception of speed. Other treatments address road surface, including the use of skid resistant pavement materials, sealing of gravel shoulders and other less conventional measures to alert drivers when they drift from their lane. Improved delineation in the form of centre-lines, lane-lines and edge-lines, increasingly, with audible and tactile qualities, being used to enhance the delineation effect. Lane widening and bridge widening are other examples of infrastructure treatments to reduce single-vehicle crash risk on rural roads. A category of single-vehicle crashes of particular concern occurs on curves, mainly because of the added challenge for drivers and riders of negotiating a curve at high speed. Infrastructure improvements for run-off-road crashes on curves include treatments to reduce speeds and speeding, especially if a better match can be achieved between vehicleoperating speeds and curve design speeds, speed perceptual measures, geometric improvements to curves such as curve flattening and provision of super-elevation, lane widening and curve transition design. Enhanced delineation of curves is also achieved through a variety of line-marking, signing and reflectorised devices. While it is always desirable to reduce crash risk as far as possible, realistically a significant residual risk will remain. For these cases, roadside barriers, flattening of roadsides, clearing of roadside hazards and use of forgiving objects within roadsides have all been used. Mixed effects have been found, however, some (e.g., barriers) are proving to be promising treatments. 4.2.2 Multi-vehicle Crash Countermeasures

For intersection crashes, the main infrastructure-based treatments used have been reduced speed limits on approach to intersections, speed perception measures, roundabouts, traffic signals, grade-separation, channelisation, signing to clarify priority, removal of sight distance obstructions, provision of medians (with and without turn-lanes), skid resistant pavements and limiting access from side-roads and driveways. Another important category of but also their high severity, solutions for this crash type vehicle lane-keeping, especially
132

multi-vehicle crashes, not only because of their frequency, comprises head-on collisions. The principal infrastructure include delineation of centre- and edge-lines to improve on curves, measures to reduce vehicle speeds on curves, as

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

noted above for single-vehicle crashes, and modifications to road cross-sections, such as additional lanes, over-taking lanes and duplication of carriageways. Geometric improvements and road surface improvements described above to reduce the risk of runoff-road crashes have also been used on curves to address the risk of head-on crashes. Crashworthy barrier systems have been highly effective, especially when used over extended lengths of high-speed roadway in addressing the severity consequences of vehicle departures towards opposing flows. 4.3 MOST PROMISING COUNTERMEASURES

In consolidating new knowledge and insights into infrastructure-based countermeasures for traffic crashes on rural roads, it is important to adopt a strategic perspective, especially in regard to the effects of infrastructure countermeasures on crash risk, injury risk, or both, whether they act on all crash types or only on a particular subset of crashes, the relative costs of various types of infrastructure and their potential large-scale, widespread application to key crash problems in rural areas. Of central importance is the notion of cost-effectiveness. Cost-effectiveness facilitates rational decision-making and, ultimately, the achievement of the best safety gains from available resources. 4.3.1 Single-vehicle crashes

Of the numerous measures outlined above, the greatest potential for reducing serious injuries resulting from single-vehicle crashes in rural areas comes from: Speed limit reductions especially along roads of curved alignment and/or with hazardous roadsides; because of the powerful relationship that exists between vehicle speeds, and crash and injury risk for all crash types, especially single-vehicle crashes; Crashworthy roadside barriers erected over extended lengths of rural roadway; because of the extreme difficulty in preventing a substantial number of drivers and riders in high-speed rural environments from leaving roadways; and Skid resistant pavements and shoulder sealing with tactile edge-lining; because they assist a sizable proportion of the drivers and riders of errant vehicles to regain control of their vehicles without entering the roadside and, desirably, without striking, unnecessarily, roadside barriers. Intersection Crashes

4.3.2

For intersections in high-speed settings, most of the identified countermeasures are effective, to varying degrees, in reducing crash occurrence. However, few of these infrastructure countermeasures also substantially limit the risk of serious injury when crashes occur (e.g., traffic signals or sign-controlled intersections). For intersection crashes in high-speed rural settings, roundabouts have been shown to not only reduce crash frequency by some 70 to 80%, but they also reduce crash costs (reflecting injury severity) by around 90%. That is, roundabouts in rural areas have been found to reduce, in a fundamental way, both crash risk and the risk of serious injury to the occupants of vehicles colliding at intersections. This is true for the main crash types occurring at rural intersections, namely, cross-traffic, turn-against and, in some cases, rear-end crashes.

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

133

Grade-separation, while virtually fully effective in eliminating most major conflicts, still allows high-speed crashes to occur where entry/exit ramps interact with the major flow of traffic. However, these conflicts tend to occur at more favourable (low) angles than conflicts that tend to occur at conventional intersections, where right-angle impacts are common. For some forms of grade-separation, there may be no ramps, a limited number of ramps or full provision of ramps as part of the intersection treatment. 4.3.3 Head-on crashes

As with previous countermeasure types, measures to address head-on crashes have the potential to reduce crash risk, injury risk or both. Measures that make a fundamental improvement in the inherent safety of rural roads, with respect to head-on crashes, are favoured. These measures include the use of crashworthy barrier systems on medians or to separate opposing vehicle directions along undivided rural roads, reductions in speed limits to substantially reduce crash risk and to magnify the injury risk reductions offered by crashworthy barrier systems in separating opposing directions of high-speed traffic. Due to the difficulty in predicting where head-on crashes will occur, crashworthy barrier systems should be installed over extended lengths (i.e., tens or hundreds of kilometres) of rural roadway. Geometric improvements to curves also have the potential to substantially reduce the incidence of head-on crashes, however, these treatments are both costly and unlikely to reduce, fundamentally, the risk of head-on collisions nor the serious injury consequences of such crashes in high-speed rural settings. 4.4 INCREMENTAL VERSUS FUNDAMENTAL IMPROVEMENTS

There are many studies reported in the literature that investigate rural crash problems but relatively few include rigorous evaluations of the effectiveness or cost-effectiveness of specific infrastructure measures. Some infrastructure countermeasures affect crash risk, injury risk, or both, incrementally. That is, they have the potential to eliminate a relatively minor proportion of crashes and leave a major residual crash problem. Other countermeasures have the potential to eliminate a major proportion of crashes by changing, in a fundamental way, the inherent safety of the road system, or elements of it. Relatively few infrastructure measures developed and used to date have the potential to improve, fundamentally, crash and injury risk. Having regard to the pattern of key crash types on rural roads, the most strategically important measures found from the literature review, were: 1. Grade-separated intersections can virtually eliminate intersection crashes by completely separating in space, conflicting vehicle paths. That is, they have the potential to be 100% effective but in terms of cost-effectiveness, the high cost of grade-separation may make them less attractive than some other alternatives. Roundabouts can reduce casualty crash risk at intersections by between 70 and 80%, crash costs by around 90% and have been found to result in BCRs of around 19 when constructed at rural intersections with a high crash record. Crashworthy barrier systems when used over extended lengths of high-speed rural road, barrier systems have the potential to reduce fatal and serious injuries to the occupants of errant vehicles by around 90%, with conservatively estimated BCRs

2.

3.

134

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

of around eight. Flexible barrier systems can address two major rural crash categories, namely single-vehicle and head-on crashes, on straight or curved road sections, without the need for costly road duplication and/or geometric improvements to rural infrastructure. 4. Speed reduction speed has a fundamental relationship with both crash and injury risk. Substantial improvements are achievable for all crash types from minor reductions in travel speed and even smaller reductions in impact speed. However, while these effects span the full range of crash types that occur in rural areas, reductions in rural speeds tend to be viewed as of limited effectiveness, as both travel and impact speeds in rural areas are often high (i.e., 100-120 km/h). At impact speeds often found in rural crashes, modern-day vehicles, even those fitted with the best available safety features, offer only limited protection from serious injury to their occupants. For particular forms of infrastructure, it appears that lower impact speeds can be especially effective in reducing the risk of severe injuries in rural settings. The main reason for this relates to the combined, or synergistic, effects of lower impact speeds and more forgiving (or crashworthy) infrastructure. More forgiving infrastructure substantially limits the transfer of kinetic energy at impact to vehicles and their occupants. That is, the overall benefits can be greater that the sum of the individual benefits of more forgiving infrastructure and speed reduction. Put another way, there are substantial synergistic benefits when speed and crashworthy infrastructure interact in a compatible way. Infrastructure countermeasures that eliminate conflict or combine, in an optimal way, the management of speed and energy transfer at impact appear to offer the most promising options for enhancing rural road safety. THE FUTURE

4.5

The system-wide and comprehensive application of measures such as speed reduction, grade separation, roundabouts at intersections, and extended use of crashworthy barrier systems to key crash problems in rural areas appears to offer the most effective and costeffective solutions to rural road trauma in the foreseeable future. Acceptance of these principles will be necessary for societies to commence the challenging journey towards a fundamentally safer road-transport system.

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

135

136

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

REFERENCES
AASHTO (1990). A policy on the geometric design of highways and streets. Washington DC: US Department of Transport. AASHTO (1994). A policy on geometric design of highways and transportation officials. Washington, DC: American Association of State Highway and Transportation Officials, US Department of Transportation. Abed-Al-Rahim, I.J., & Johnston, D.W. (1993). Estimating bridge-related traffic accident rates and costs. Transportation Research Record, 1392, 99-106. Adamo, A.F., & Wray, W.K. (1992). Analysis of single-vehicle run-off-the-road accidents on rural four lane or more divided highways in Texas. (Report, no.1257-1F). Austin, Texas: Texas Department of Transportation. Agent, K.R. (1980). Transverse pavement markings for speed control and accident reduction, Transportation Research Record, 773, 11-14. Agent, K. R., & Creasey T. (1986). Delineation of horizontal curves. (Report No. UKTRP-86-4). Kentucky: Kentucky Transportation Research Program, College of Engineering, University of Kentucky. Agent, K.R., & Deen, R.C. (1976). Relationship between roadway geometrics and accidents. Transportation Research Record, 541. Al-Masaeid, H.R. (1997). Impact of pavement condition on rural road accidents. Canadian Journal of Civil Engineering, 24, 523-531. Al-Masaeid, H.R., Al-Suleiman, T.I., Hammed, M.M., & Halawa, K.R. (1994). Impacts of small rural towns of road safety and traffic operation. Road and Transport Research, 3, 8588. Al-Masaeid, H.R., & Sinha, K.C. (1994). Analysis of accident reduction potentials of pavement markings. Journal of Transportation Engineering, 120, 723-736. Al-Sharaf, J.S. (1995). Effect of road roughness and pavement conditions on traffic speed. M.Sc. thesis, Department of Civil Engineering, Jordan University of Science and Technology, Irbid, Jordan. Allan, R. (1992). Risk management protection levels at road/rail crossings. Paper presented at the International Rail Safety Conference, Wellington, New Zealand. Alexander, G., & Lunefeld, H. (1986). Driver expectancy in highway design and traffic operations. (Report No. FHWA-TO-86-1). Washington DC: US Department of Transportation, Federal Highway Administration. Anderson, I., Bauer, K., Harwood, D., & Fitzpatrick, K. (1999). Relationship to safety of geometric design consistency measures for rural two-lane highways. Transportation Research Record, 1658, 43-51. Anderson, I., & Krammes, R. (2000). Speed reduction as a surrogate for accident experience at horizontal curves on rural two-lane highways. Transportation Research Record, 1701, 86-94. Anderson, R., McLean, A., Farmer, M., Lee, B., & Brooks, C. (1997). Vehicle travel speeds and the incidence of fatal pedestrian crashes. Accident Analysis and Prevention, 29, 667-674.

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

137

Andreassen, D. (1992). A guide to the use of road accident cost data in project evaluation and planning. (Report No. ARR 226). Melbourne, Australia: Australian Road Research Board. Armour, M. (1984). A description of casualty accidents on Australian rural highways. Report No. AIR 1120-3. Melbourne, Australia: ARRB Transport Research. Armour, M., & Cinquegrana, C. (1990). Victorian study of single vehicle rural accidents. Proceedings of the 15th Australian Road Research Board Conference (pp 79-92). Darwin, Australia: ARRB Transport Research. Armour, M., Harrison, W., & South, D. (1986). Random breath testing in Victoria. (Report No. RN/86/2). Melbourne, Australia: Road Traffic Authority. Ashton, S., & Mackay, G. (1979). Some characteristics of the population who suffer trauma as pedestrians when hit by cars and some resulting implications. Proceedings of the IRCOBI Conference. (pp 39-48). Goteburg, Sweden. Austroads (1988). Guide to traffic engineering practice. Part 9: Arterial road traffic management . (Report AP-11.9/88). Sydney, Australia: Austroads. Balke, K.N., & Fitzpatrick, K. (1993). An evaluation of flush medians and two-way, left turn lanes on four-lane rural highways. (Report No. 1293-1). Austin, Texas: Texas Transportation Institute. Barker, J., Farmer, S. & Taylor, M. (1999). The development of accident-remedial intervention levels for rural roads. (TRL Report, No.425). Crowthorne, UK: Transport Research Laboratory. Baruya, A. (1998). Speed-accident relationships on European roads. VTI Konferens, 10A(10): 3-17. Linkping, Sweden: Swedish National Road and Transport Research Institute. Baum, H., Wells, J., & Lund, A. (1990). Motor vehicle crash fatalities in the second year of 65 mph speed limits. Journal of Safety Research, 21, 1-8. Benekohal, R. (1991). Methods for estimating accident reductions from highway improvements. ITE Compendium of Technical Papers from 61st Annual Meeting (Publication No PP-023) (pp 419-423). Washington, DC: Institute of Transportation Engineers. Bergh, T., & Carlsson, A. (2000). 2+1 roads with and without cable barriers speed performance. Pacific Rim Conference on Highway Capacity June 27-July 1, 2000. Transportation Research Board, Committee A3A10, Highway Capacity and Quality of Service. Bester, C.J., & Makunje, J.A. (1998). The effect of rural road geometry on safety in Southern Africa. Proceedings International Symposium on Highway Geometric Design Practices, 1995. (Paper 15). Washington, DC: Transportation Research Board. Bonneson, J., McCoy, P., & Truby, J. (1993). Safety improvements at intersections on rural expressways: A survey of state departments of transportation. Transportation Research Record, 1385. Bowman, B.L., & Brinkman, P. (1988). The effect of low cost accident countermeasures on vehicle speed and lateral placement at narrow bridges. Transportation Research Record, 1185, 11-23. Brewer, J., German, J., Krammes, R., Movassaghi, K., Okamoto, J., Otto, S., Ruff, W., Sillan, S., Stamadiatis, N., & Walters, R. (2001). Geometric design practices for European

138

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

roads. (Report No. FHWA-PL-01-026). Administration, US Department of Transport.

Washington

DC:

Federal

Highway

Brinkman, C.P., & Smith, S.A. (1984). Two-lane rural highway safety. Public Roads, 48, 48-53. Brown, J.V., & Foster, J. (1966). Bridge accidents on rural highways in New Zealand: Analysis and appraisal. Proceedings of the 3rd ARRB Conference, 3, 638-646. Brown, D., Maghsoodloo, S., & Ardle, M. (1990). The safety impact of 65 mph speed limit: A case study using Alabama accident records. Journal of Safety Research, 21, 125139. Cafiso, S. (2000). Experimental survey of safety condition on road stretches with alignment inconsistencies. International Symposium on Highway Geometric Design. (pp. 377-387). Cologne, Germany: Road and Transportation Research Association. Cairney, P. (1995). The road in rural road safety. Proceedings from the Rural Road Safety Focus for the Future Conference. Canberra, Australia: National Road Trauma Advisory Council. Cairney, P. (1998). The relationship of crash risk to geometry and cross-section on Australian rural roads. Proceedings of the 19th ARRB Transport Research Conference (pp 133-155). Melbourne, Australia: ARRB Transport Research. Cairney, P.T., & McGann, A.P. (1997). Risk factors on Australian rural roads. Proceedings Road Safety Research and Enforcement Conference. Hobart, Australia: Tasmania Department of Transport. Caldwell, R.C., & Wilson, E.M. (1996). A safety improvement program for rural unpaved roads. (MPC Report No. 97-70). Fargo, ND: Mountain-Plains Consortium. Calvert, E.S. (1954). Visual judgements in motion, Journal of the Institute of Navigation, 7, 233-251. Calvert, E.C., & Wilson, E.M. (1996). A safety improvement program for rural unpaved roads. (Report, No.97-70). Fargo, North Dakota: Mountain-Plains Consortium. Calvert, E.C., & Wilson, E.M. (1999). Incremental safety improvements for unpaved rural roads. U.S. Department of Transportation Annual Report, Bureau of Transportation Statistics. Washington DC: U.S. Department of Transportation. Cameron, M., Newstead, S., & Le, C. (1998). Rating the aggressivity of Australian passenger vehicles towards other vehicle occupants and unprotected road users. Journal of Crash Prevention and Injury Control, 1(2), 129-141. Catchpole, J. (1997). An investigation of factors contributing to the over-representation of young drivers in traffic accidents. Unpublished Masters Thesis. Melbourne, Australia: Latrobe University. Cavallo, A., & Triggs, T. (1996). Young driver research strategy. (Report No. 85). Melbourne, Australia: Monash University Accident Research Centre. Cesari, D., & Bloch, J. (1984). The influence of car structures behaviour on occupant protection in car to car side impact. Proceedings International Conference on Vehicle Structures. (C163/84, pp. 7-10). London: Mechanical Engineering Publications. Chang, G., Chen, C., & Carter, E.C. (1993). Intervention analysis for the impacts of the 65 mph speed limit on rural interstate highway facilities. Journal of Safety Research, 24, 3335.

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

139

Cirillo, J. (1970). The relationship of accidents to length of speed-change lanes and weaving area on interstate highways. Highway Research Record, 312. Cirillo, J. (1968). Interstate system accident research Study II. (Interim Report II). Public Roads, 35(3). Clarke, D., Ward, P., & Jones, J. (1998). Overtaking road accidents: Differences in manoeuvre as a function of driver age. Accident Analysis and Prevention, 30, 455-467. Cleveland, D.E. (1987). Effect of resurfacing on highway safety: Relationship between safety and key highway features. In E. Crump. (Ed.). State of the Art Report 6, (pp 78-95). Washington, D.C: National Research Council. Collins, J., Fitzpatrick, K., Bauer, K.M., & Harwood, D.W. (1999). Speed variability on rural two-lane highways. Transportation Research Record, 1658, 60-69. Commission Internationale de lEclairage (1990). A guide for the design of road traffic lights. (Publication No. 79). Paris, France: Commission Internationale de lEclairage. Comte, S., & Carsten, O. (1998). The effectiveness of ATT ad non-ATT treatments for sharp curves on two-lane rural roads. VTI Konferens, 10A(9), 81-94. Linkping, Sweden: Swedish National Road and Transport Research Institute. Copelan, J. (1989). Prevention of wrong-way accidents on freeways. (Publication No. FHWA/CA-TE-89-2). Washington DC: Federal Highway Administration, US Department of Transportation. Corben, B.F., Ambrose, C., & Foong, C. W. (1990). Evaluation of accident Black Spot treatments. (Report No. 11). Melbourne, Australia: Monash University Accident Research Centre. Corben, B.F., & Deery, H.A. (1998). Ro ad engineering countermeasures: An evaluation of their effectiveness and in-depth investigations of the unsuccessful treatments. 9th Road Engineering Association of Asia and Australasia, Wellington, New Zealand, May, 1998. Corben, B.F., Deery, H., Mullan N., & Dyte, D. (1997). The general effectiveness of countermeasures for crashes into fixed roadside objects. (Report No. 111). Melbourne, Australia: Monash University Accident Research Centre. Corben, B., Gelb, K., Pronk, N., & Fitzharris, M. (1999). Motorcycle safety Stage 2. Contract report prepared for VicRoads. Melbourne, Australia: Monash University Accident Research Centre. Corben, B., Lenn, M., Regan, M., & Triggs, T. (2001). Technology to enhance speed limit compliance. Proceedings of the 2001: Road Safety Research, Policing and Education Conference. Melbourne, Australia: Monash University Conference Management Office. Corben, B.F., Newstead S.V., Diamantopoulou, K., & Cameron M.H. (1996). Results of an evaluation of TAC funded accident black spot treatments. Proceedings of the ROADS 96 - Joint 18th ARRB Transport Research Conference and Transit New Zealand Land Transport Symposium. Christchurch, N.Z. Corben, B.F., Tingvall, A., Fitzharris, M., Newstead, S., Les, M., & Johnston I.R. (2001). Crash analysis for the review of guidelines for median barriers. Contract report prepared for VicRoads. Melbourne, Australia: Monash University Accident Research Centre. Council, F. (1998). Safety benefits of spiral transitions on horizontal curves on two-lane rural roads. Transportation Research Record, 1635, 10-17.

140

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

Council, F.M., & Stewart, J.R. (1999). Safety effects of the conversion of rural two-lane to four-lane roadways based on cross-sectional models. Transportation Research Record, 1665, 35-43. Crowley, K., & Seguin, E. (1986). Wrong way traffic control at intersections. (Publication No. FHWA-RD-86-116). Washington DC: Federal Highway Administration, US Department of Transportation. Cumming, R., & Croft, P. (1971). A review of speed control in relation to road safety. Canberra, Australia: Australian Department of Transport. Datta, T., Perkins, J., Taylor, I., & Thompson, H. (1983). Accident surrogates for use in analysing high way safety hazards. Volume 2. Technical report . (Report FHWA/RD82/104). Washington DC: Federal Highway Administration, US Department of Transport. Davis, G., Alexander, B., & Brissie, R. (1997). A 15-year review of railway-related deaths in Jefferson County, Alabama. American Journal of Forensic Medicine and Pathology, 18, 363-368. de Almeida Roque, C., & Cardoso, J. (2001). Using low cost engineering measures and stricter enforcement to improve road safety on a dangerous rural route. Routes/Roads PIARC, 311, 71-82. de Silva, H. (2001). Road safety benefits of improved skid resistance: A valuation procedure. Proceedings ARRB Transport Research Conference, Melbourne, Australia: ARRB Transport Research. de Waard D., Jessurun M., Steyvers F.J.J.M., Raggatt P.T.F., & Brookhuis K.A. (1995). The effect of road layout and road environment on driving performance, drivers physiology and road appreciation, Ergonomics, 38, 1395-1407. Deery, H., Kowadlo, N., Westphal-Wedding, T., & Fildes, B. (1998). Identifying subtypes of young novice drivers: Implications for matching training to the needs of the driver. (Report No. 142). Melbourne, Australia: Monash University Accident Research Centre. Delaney, A., Langford, J., Corben, B.F., Newstead, S.V., & Jacques, N. (2002). Roadside environment safety. Contract report prepared for RACV. Melbourne, Australia: Monash University Accident Research Centre. Denton, G.G. (1971). The Influence of visual pattern on perceived speed. (Report No. LR409). Crowthorne, U.K.: Transportation and Road Research Laboratory. Denton, G.G. (1973). The Influence of visual pattern on perceived speed at Newbridge MB Midlothian. (Report No. LR531). Crowthorne, U.K.: Transportation and Road Research Laboratory. Doege, T., & Levy, P. (1977). Injuries, crashes and construction on the superhighway. American Journal of Public Health, 67, 147-150. Duncan, C., Corben, B., Truedsson, N., & Fitzharris, M. (2001). Motorcyclists and barriers. Contract report prepared for VicRoads. Melbourne, Australia: Monash University Accident Research Centre. Duncan, C.C., Corben, B.F., Truedsson, P.N., & Tingvall, C. (2000). Motorcycle safety barrier crash-testing: Feasibility study. (Report No. CR201). Canberra, Australia: Australian Transport Safety Bureau. Engel, U. (1990). Effects of reducing measures in Danish residential areas. VTI Rapport 363A. Linkping, Sweden: Swedish Road and Traffic Research Unit.

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

141

Emerson, J.W., & West, L.B. (1985). Shoulder rumble strips at narrow bridges. Proceedings Conference on the Effectiveness of Highway Safety Improvements, (March 2426) Tennessee: Highway Division of the American Society of Civil Engineers. Enustun, N. (1972). Three experiments with transverse pavement stripes and rumble bars. (Report No. TSD-RD-21672). Michigan, Indiana: Department of State Highways. European Transport Safety Council (1995). Reducing traffic injuries resulting from excess and inappropriate speed. Brussels: European Transport Safety Council. Evans, J. (1997). Road crashes involving bridges and culverts. (Report No. ARR 303). Melbourne, Australia: ARRB Transport Research. Fairclough, H. (1997). Monitoring driver fatigue via driving performance. In Y. Noy (Ed.), Driver fatigue: An experimental investigation, (pp. 363-379). Englewood Cliffs, New Jersey: Lawrence Erlbaum. Fildes, B., Corben, B., Morris, A., Oxley, J., Pronk, N., Brown, L., & Fitzharris, M. (2000). Road safety environment and design for older drivers. (Report AP-R169). Sydney, Australia: Austroads. Fildes, B.N., Fletcher, M.R., & Corrigan, J.M. (1987). Speed perception 1: Drivers' judgements of safety and speed on urban and rural straight roads. (Report No. CR 54). Canberra, Australia: Federal Office of Road Safety, Department of Transport & Communication. Fildes, B.N., Godley S.T., Triggs T.J., & Jarvis J. (1997). Perceptual countermeasures: Simulation Validation Study. (Report No. CR169). Canberra, Australia: Federal Office of Road Safety, Department of Transport & Communication. Fildes, B., Lane, J., Lenard, J., & Vulcan, P. (1991). Passenger cars and occupant injury. (Report CR 95). Canberra, Australia: Federal Office of Road Safety. Fildes, B., Lane, J., Lenard, J., & V ulcan, P. (1994). Passenger cars and occupant injury: Side impact crashes. (Report CR 134). Canberra, Australia: Federal Office of Road Safety. Fildes, B., Oxley, J., & Corben, B. (in press). Older Driver Road Design Manual: Recommendations and Guidelines. (Preliminary Draft). Melbourne, Australia: Monash University Accident Research Centre. Finch, D., Kompfner, P., Lockwood, C., & Maycock, G. (1994). Speed, speed limits and accidents. (Report No. 58). Crowthorne, U.K.: Transport Research Laboratory. Fitzpatrick, K., Fambro, D., & Stoddard, A. (2000). Safety effects of limited stopping sight distance on crest vertical curves. Transportation Research Record, 1701, 17-24. Friebele, J., Messer, C., & Dudek, C. (1971). State-of-the-art of wrong-way driving on freeways and expressways. (Report No. 139-7). Washington DC: US Department of Transportation, Texas Transportation Institute. Freedman, M., & Esterlitz, J. (1990). Human factors and safety research related to highway design and operation. Transportation Research Record, 1281, 52-61. Gaca, S., & Tracz, M. (2000). Pedestrian risk on rural roads. VTI Konferens, 13A, 5: 11929. Linkping, Sweden: Swedish National Road and Transport Research Institute. Gandhi, B.V.R., Lytton, R.L., & Das, B. (1984). An improved bridge safety index for narrow bridges. Transportation Research Record, 960, 40-48.

142

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

Garber, S., & Graham, J. (1990). The effects of the new 65 mph speed limit on rural highway fatalities: a state-by-state analysis. Accident Analysis and Prevention, 22, 137149. Garber, N., & Lienau, T. (1996). Traffic and highway geometric characteristics associated with pedestrian crashes in Virginia. (Report No. VTRC 96-R29). Charlottesville, Virginia: Virginia Transport Research Council. Garber, N., & Patel, S. (1995). Control of vehicle speeds in temporary traffic control zones (work zones) using changeable message signs with radar. Transport Research Record, 1509, 73-81. Garber, N., & Woo, T. (1990). Accident characteristics at construction and maintenance zones in urban areas. (Report No. VTRC90-R12). Charlottesville, VA: Virginia Transportation Research Council. Grder, P., & Alexander, J. (1995). Continued research on continuous rumble strips: Final report. (Report No.94-4). Augusta, Maine: Maine Department of Transportation. Gattis, J., & Low, S. (1998). Intersection angle geometry and drivers field of view. Transportation Research Record, 1612, 10-16. Gelston, P. (1998). Emerging issues in rural road safety a SA perspective. Proceedings of the National Road Safety Summit (pp 331-336). Canberra, Australia: Federal Office of Road Safety. Gibson, J.J. (1950). The Perception of the Visual World, Boston: Riverside Press. Gibson, J.J. (1958). Visually controlled locomotion and visual orientation in animals, British Journal of Psychology, 49, 182-194. Gibson, J.J. (1968). What gives rise to the perception of motion, Psychological Review, 75, 335-346. Gibson, J.J., & Crooke, L.E. (1938). A theoretical field analysis of automobile driving, American Journal of Psychology, 51, 453-471. Glennon, J.C. (1987). Effect of alignment on highway safety. State of the Art Report 6, (pp 48-63). Washington, D.C: Transport Research Board. Glennon, J., Neuman, T., & Leish, J. (1985). Safety and operational considerations for design of rural highway curves. (Report No. FHWA/RD-86/035). Washington DC: Federal Highway Administration, US Department of Transport. Godley S.T., Fildes B.N., Triggs T.J. & Brown L.J. (1999). Perceptual countermeasures: experimental research. (Report No. CR182). Canberra, Australia: Australian Transport Safety Bureau. Godwin, S. (1992). Effect of the 65 mph speed limit on highway safety in the USA (with comments and reply to comments). Transport Reviews, 12, 1-14. Gordon, D.A. (1966). Experimental isolation of drivers' visual input. Public Roads, 33, 266-273. Graham, J., Paulsen, R., & Glennon, J. (1977). Accidents and speed studies in construction zones. (Publication No. FHWA-RD-77-80). Washington DC: US Department of Transportation, Federal Highway Administration. Gramza, K., Hall, J., & Sampson, W. (1980). Effectiveness of freeway lighting. (Publication No. FHWA-RD-79-77). Washington DC: Federal Highway Administration, US Department of Transportation.
COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS 143

Griebe, P., & Nielsen, M.A. (1996). Safety at four-armed signalized junctions situated on roads with different speed limits. Proceedings Road Safety in Europe Conference. (VTI Konferens, no.7A, Part 2, pp. 151-163). Linkping, Sweden: Swedish National Road and Traffic Research Institute. Grigg, C. (2000). Rural road safety. Roadwise, 12(3), 17-20. Grzebieta, R., Rechnitzer, G., Tingvall, C., Judd, R., & Powell, C. (2000). Vehicle compatibility get the geometry and interfaces right first. Proceedings of the Road Safety Research, Policing and Education Conference. (pp. 507-513). Brisbane, Australia. Hagenauer, G., Upchurch, J., Warren, D., & Rosenbaum, M. (1982). Intersections: In synthesis of safety research related to traffic control and roadway elements. (Publication No. FHWA-TS-82-232). Washington DC: Federal Highway Administration, US Department of Transportation. Halawa, K.R. (1993). Impacts of residential communities along main roads on traffic safety. M.Sc. thesis, Department of Civil Engineering, Jordan University of Science and Technology, Irbid, Jordan. Hall, G. (1999). Statement of Gerri L. Hall to Hearing on Highway-Rail Grade Crossing safety. US Subcommittee on Surface Transportation and Merchant Marine. Hall, J. (1988). Traffic safety in construction zones (Report No: FHWA-HPR-NM-86-91). Santa Fe, New Mexico: New Mexico State Highway and Transportation Department. Hall, J. W. (1991). Innovative treatments for run-off-the-road accidents. (Report No.FHWA-NMSHTD-91-02). Washington DC: Federal Highway Administration. Hall, J., & Lorenz, V. (1989). Characteristics of construction zone accidents. Transportation Research Record, 1230, 20-27. Hall, J., & Zador, P. (1981). Survey of single-vehicle fatal rollover crash sites in New Mexico. Transportation Research Record, 819, 1-8. Harkey, D., Huang, H., & Zegeer, C. (1996). Examination of older driver problems on freeways using the highway safety information system . Washington DC: Transportation Research Board. Harrington, T.L., Harrington, M.K., Wilkins, C.A., & Koh, Y.O. (1980). Visual orientation by motion-produced blur patterns: Detection of divergence, Perception & Psychophysics, 28(4), 293-305. Hartley, L., Horberry, T., Mabbott, N., & Krueger, G. (2000). Review of fatigue detection and prediction technologies. Paper prepared for the National Road Transport Commission, Australia. Harwood, D., Hoban, C., & Warren, D. (1988). Effective use of passing lanes on two-lane highways. Transportation Research Record, 1195, 70-91. Harwood, D., Pietrucha, M., Wooldridge, M., Brydia, R. & Fitzpatrick, K (1995). Median intersection design. (Report No. 375). Washington D.C: National Cooperative Highway Research Program. Harwood, D.W., & St. John, A.D. (1985). Passing lanes and other operational improvements on two-lane highways. (Report No. FHWA/RD-85/028). W ashington DC: US Department of Transportation. Hauer, E. (1971). Accidents, overtaking and speed control. Accident Analysis and Prevention, 3, 1-13.

144

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

Hawkins, H., Kacir, K., & Ogden, M. (1992). Traffic control guidelines for urban arterial workzones: Volume 2. (Publication No. FHWA-TX-91-1161-5). Washington DC: US Department of Transportation, Federal Highway Administration. Haworth, N., Ungers, B., Vulcan, P., & Corben, B. (2001). Evaluation of a 50 km/h default urban speed limit for Australia. Melbourne, Australia: National Road Transport Commission. Haworth, N., & Vulcan, P. (1991). Testing of commercially available fatigue monitors. (Report No. 15). Melbourne, Australia: Monash University Accident Research Centre. Haworth, N., Vulcan, P., Bowland, L., & Pronk, N. (1997). Characteristics of fatal single vehicle crashes. (Report No. 120). Melbourne, Australia: Monash University Accident Research Centre. Hehlen, P. (1992). Safety of rural roads. Paper presented at the 4th European Workshop on Recent Developments in Road Safety Research, Bergisch-Gladbach, Germany. Helliar-Symons, R.D. (1981). Yellow bar experimental carriageway markings accident study. (Report No. Lr1010). Crowthorne, UK: Transportation and Road Research Laboratory. Henderson, M. (1995). An overview of crash data and countermeasure implications. Paper presented at the Rural Road Safety Seminar: Focus for the future. Canberra, Australia: National Road Trauma Advisory Council. Hilton, M.H. (1973). Some case studies of highway bridges involved in accidents. Highway Research Record, 432, 41-51. Hollingworth, G.H. (1983). Road crashes at bridge sites their nature and difficulties with their prediction. Institution of Engineers Australia, Queensland Division, Technical papers, 24(7), 11-18. Hoque, M. M., & Sanderson, J.T. (1988). Road safety countermeasures for rural roads. (Report No. TS88/3). Melbourne, Australia: Royal Automobile Club of Victoria. Hughes, W. (1994). Accidents on rural roads: A study in Cambridgeshire. Basingstoke, UK: AA Foundation for Road Safety Research. Hughes, P. (2002). A risk assessment system for passive level crossings. Proceedings Getting Active at Passive Crossings: Seventh International Symposium on RailroadHighway Grade Crossing Research and Safety. Melbourne, Australia: Monash University. Hughes, W., & Amis, G. (1996). Accidents on rural roads: Single carriageway "A" class roads. Basingstoke, United Kingdom: AA Foundation for Road Safety Research. Hughes, W., Amis, G., & Walford, A. (1997). Accidents on rural roads: Dual carriageway 'A' class roads. Basingstoke, United Kingdom: AA Foundation for Road Safety Research. Hughes, W., Joshua, S., & McGee, H. (1992). Study designs for passing sight distance requirements. (Publication No.FHWA-RD-91-078). Washington DC: US Department of Transportation, Federal Highway Administration. Hunt, J., & Mahdi, T. (1996). A model to predict accident frequency on rural single carriageway roads. Proceedings Road Safety in Europe and Strategic Highway Research Program Conference. (pp 55-71) Linkping, Sweden: Swedish National Road and Traffic Research Institute. Hunter-Zaworski, K. (1990). T-intersection simulator performance of drivers with physical limitations. Transportation Research Record, 1281.

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

145

Iowa Department of Transportation (1991). Iowa running stop signs as a contributing circumstance in fatal accidents 1986-1990. Des Moines, IA: Driver Safety and Improvement, Iowa Department of Transportation. Ivan, J., Pasupathy, R., & Ossenbruggen, P. (1999). Differences in causality factors for single and multi-vehicle crashes on two-lane roads. Accident Analysis and Prevention, 31, 695-704. Ivey, D.L., Olson, R.M., Walton, N.E., Weaver, G.D., & Whitehurst-Furr, L. (1979). Safety at narrow bridges. (Report No. 203). Washington D.C.: National Co-operative Highway Research Program, Highway Research Board. Janoff, M., Freedman, M., & Decina, L. (1982). Partial lighting of interchanges. (Report No. 256). Washington DC: National Co-operative Highway Research Program. Johansson, G. (1977). Studies on visual perception of locomotion, Perception, 6, 365-376. Johansson, G. (1985). About visual event perception. In W.H. Warren and R.E. Shaw (Eds), Persistence and Change: Proceedings of the First International Conference on Event Perception. (pp 29-454). Englewood Cliffs, New Jersey: Lawrence Erlbaum. Johnston, I. (1982). Modifying driver behaviour on rural road curves: A review of recent research. Proceedings of the 11th Australian Road Research Board Conference, (115-124). Melbourne, Australia: ARRB Transport Research. Johnston, I. (2001). Keynote address. Proceedings of the 2001 Road Safety Research, Policing and Education Conference. Melbourne, Australia. Johnston, I. (1983). The effects of roadway delineation on curve negotiation by both sober and drinking drivers. (Report No. ARR128). Melbourne, Australia: ARRB Transport Research. Johnston, I.R., White, G.R., & Cummings, R.W. (1973). The role of optical expansion patterns in locomotor control. American Journal of Psychology, 86, 311-324. Jordaan, P.W., & Mashiri, M.A.M. (1994). Geometric design standards for gravel roads. Transport Information Bulletin, 9(2), 22-25. Kalakota, K., Seneviratne, P., & Islam, N. (1992). Influence of geometric design variables on accident rates on two-lane rural highways. VTI Rapport, 380A, 207-226. Kanellaidis, G. (1991). Aspects of highway Transportation Engineering, 117, 624-632. super-elevation design. Journal of

Karlaftis, M.G., & Golias, I. (2002). Effects of road geometry and traffic volumes on rural roadway accident rates. Accident Analysis and Prevention, 34, 357-365. Katayama, T., Hashimoto, H., & Yang, Z. (2001). Traffic accidents in rural area and assistance system for traffic safety. Proceedings of 17th International Technical Conference on the Enhanced Safety of Vehicles. (Report No. DOT HS 809 220) (pp. 4-7). Washington, DC: National Highway and Traffic Safety Administration. Kaub, A., & Berg, W. (1988). Design guide for auxiliary passing lanes on rural two-lane highways. Transportation Research Record, 1195, 92-100. Kidd, B., & Willett, P. (2002). Crash patterns in Western Australia. Proceedings of the 2001 Road Safety Research, Policing and Education Conference. (CD-ROM). Adelaide, Australia: Causal Productions. Khasnabis, S. (1986). Operational and safety problems of trucks on no-passing zones on two-lane rural highways. Transportation Research Record, 1052, 36-44.

146

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

Kloeden, C.N., McLean, A.J., Baldock, M.R.J., & Cockington, A.J.T. (1999). Severe and fatal car crashes due to roadside hazards. Adelaide, Australia: NHMRC Road Accident Research Unit, University of Adelaide. Knoblauch, R., Nitzburg, M., Reinfurt, D., Council, F., Zegeer, C., & Popkin, C. (1995). Traffic operations control for older drivers. (Report No. FHWA-RD-93-177). Washington DC: Federal Highway Administration, US Department of Transportation. Krammes, R., Brackett, R., Shafer, M., Otteson, J., Anderson, I., Fink, K., Collins, K., Pendleton, O., & Messer, C. (1993). Horizontal alignment design consistency for rural two-lane highways. (Report No. FHWA-RD-94-034). Washington DC: Federal Highway Administration, US Department of Transport. Krammes, R.A., & Glascock, S.W. (1992). Geometric inconsistencies and accident experience on two-lane rural highways. Transportation Research Record, 1356, 1-10. Kulmala, R (1991). Accident models for rural highway junctions. Planning and, Transport Research and Computation (International) Co Ltd (PTRC) Summer Annual Meeting, University of Sussex, United Kingdom. Lamm, R., Beck, A., & Zumkeller, K. (2000). Analysis and evaluation of inter-relationship between traffic safety and highway geometric design on two-lane rural roads. Proceedings of the 2nd International Symposium on Highway Geometric Design. (pp 557-570). Cologne, Germany: Road and Transportation Research Association. Lamm, R., Choueiri, E., Hayward, J., & Paluri, A. (1988). Possible design procedure to promote design consistency in highway geometric design on two-lane rural roads. Transportation Research Record, 1195, 111-122. Lamm, R., Guenther, A., & Choueiri, E., (1995). Safety module for highway geometric design. Transportation Research Record, 1512, 7-15. Lamm, R., Hayward, J.C., & Cargin, J.G. (1986). Comparison of different procedures for evaluating speed consistency. Transportation Research Record, 1100, 10-20. Lee, D.N., & Lishman, R. (1977). Visual control of locomotion. Scandinavian Journal of Psychology, 18, 224-230. Leisch, J., & Leisch, J. (1977). New concepts in design speed application. Transportation Research Record, 631. Lenn, M. (1997). Sleep deprivation, alcohol and driving experience: Effect on driving performance across the day. Unpublished PhD Thesis. Melbourne, Australia: Department of Psychology, Monash University. Leonard, J., Bilse, D., & Recker, W. (1994). Super-elevation rates at rural highway intersections. (Report No. RTA-53P434). Irvine, CA: University of Carolina Institute of Transportation Studies. Les, M., Morris, A., Olsson, T., Pettersson, J., & Holmqvist, K. (2001). Vehicle properties determining aggressivity. Proceedings of the Road Safety Research, Policing and Education Conference. Melbourne, Australia: Monash University. Lin, F.B. (1990). Flattening of horizontal curves on rural two-lane highways. Journal of Transport Engineering, 116(2), 181-196. Lum, H.S. (1984). The use of road markings to narrow lanes for controlling speed in residential areas. Institute of Transportation Engineers, 54(6), 50-53.

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

147

Lundy, R. (1967). The effect of ramp type and geometry on accidents. Highway Research Record, 163. Lydon, M. (1997). Rural crashes and available countermeasures. Paper presented at the 14th State Municipal Engineering Conference. Perth, Australia. Macaulay J., Tziotis M., & Fildes B. (2002). Evaluation of perceptual counter-measure treatments. Proceedings of the 2001 Road Safety Research, Policing and Education Conference. (CD-ROM). Adelaide, Australia: Causal Productions. Macdonald, W. (1994). Young driver research program: A review of information on young driver performance characteristics and capacities. (Report No. CR 129). Canberra, Australia: Federal Office of Road Safety. Mace, W.M. (1985). Johansson's approach to visual event perception in Gibson's perspective. In W.H. Warren and R.E. Shaw (Eds), Persistence and Change: Proceedings of the First International Conference on Event Perception. Englewood Cliffs, New Jersey: Lawrence Erlbaum. Mace, D., & Heckard, R. (1991). Effect of the 65 mph speed limit on travel speeds and related crashes. (Report No. DOT-HS-807-764). Washington DC: US Department of Transport. Mace, D., & Pollack, L. (1983). Visual complexity and sign brightness in detection and recognition of traffic signs. Transportation Research Record, 904. Mak, K.K., & Sicking, D.L. (1994). Analysis of bridge railing accidents. Transportation Research Record, 1468, 19-24. Malfetti, J., & Winter, D. (1987). Safe and unsafe performance of older drivers: A descriptive study. Falls Church, VA: American Automobile Association, Foundation for Traffic Safety. McBean, P. (1982). Influence of road geometry at a sample of accident sites. (Report No. LR1053). Crowthorne, UK: Transportation and Road Research Laboratory. McDonald, N., & DeSanti, M. (1998). Kings Highway: NRMA road safety audit report a review of the road safety performance and influential factors along the Kings Highway. Canberra, Australia: New South Wales Royal Motorists Association. McFarland, W., Griffin, L., Rollins, J., Stockton, W., Phillips, D., & Dudek, C. (1979). Assessment of techniques for cost-effectiveness of highway accident countermeasures. (Report No. FHWA-RD-79-53). Washington DC: Federal Highway Administration, US Department of Transportation. McKnight, A., & Klein, T., (1990). Relationship of 65 mph limit to speeds and fatal accidents. Transportation Research Record, 1281. McKnight, S., McKnight, J., & Tippetts, S. (1998). The effect of lane line width and contrast upon lane-keeping. Accident Analysis and Prevention, 30, 617-624. McKnight, J., & Urpuijo, J. (1993). Sign of deficiency among elderly drivers. Transportation Research Record, 1405. McLean, J. (1981). Driver speed behaviour and rural road alignment design. Traffic Engineering and Control, 22(4), 208-211. McLean, J. (1996). Review of accidents and rural cross section elements including roadsides. (Report No. ARR 297). Melbourne, Australia: ARRB Transport Research.

148

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

McLean, A., Baldock, M., & Kloeden, C. (2002). Rural indepth crash investigation. Proceedings of the 2001 Road Safety Research, Policing and Education Conference. Melbourne, Australia: Monash University Conference Management Office. McLean, A., Ryan, G., Wright, J., & Hinrichs, R. (1988). In-depth study of rural road crashes. Paper presented at the Road Safety Researchers Conference. Canberra, Australia: Federal Office of Road Safety. Miaou, S., & Lum, H. (1993). Statistical evaluation of the effects of highway geometric design on truck accident involvements. Transportation Research Record, 1407, 11-23. Miller, T.R. (1993). Benefit-cost analysis of lane marking. Public Roads, 56(4), 153-163. Mintsis, G., & Pitsiava-Latinopoulou, M. (1990). Traffic accidents with fixed roadside obstacles: a study of the Greek rural road network. Traffic Engineering and Control, 31, 306-311. Mohamedshah, Y. (1992). Investigation of passing accidents using the HSIS data base. Public Roads, 56, 61-66. Mohamedshah, Y., Paniati, J., & Hobeika, A. (1993). Truck accident models for interstates and two-lane rural roads. Transportation Research Record, 1407, 35-41. Mornell, O. (1997). Reducing human errors at level crossings in Sweden. Paper presented at the International Rail Safety Conference, Lucerne, Switzerland. Moore, R.L. (1968). Some human factors affecting the design of vehicles and roads. Proceedings of a Symposium on Vehicle and Road Design for Safety. Crowthorne, U.K. Moore, R. (1998). Over-embankment accidents on rural 2-lane conventional highways. (Report No. FHWA/CA/TE-98/15). Sacramento, CA: California Department of Transportation. Morris, H. (1992). Road design and maintenance standards: The needs of older drivers. Proceedings of the Conference Road Safety in Europe (VTI Rapport No. 380A, Part 3), 3138. Moses, P. (1990). Safety improved on rural road curves. Western Roads, 15(2), 2-3. National Highway Traffic Safety Administration (NHTSA) (1989). Effects of the 65 mph speed limit through 1988: A report to Congress. Washington DC: US Department of Transport. Neuman, T. (1989). New approach to design for stopping sight distance. Transportation Research Record, 1208. Neuman, T. (1992). Chapter 6. Roadway geometric design. In J. Pline (Ed.), Traffic Engineering Handbook (4th ed). Englewood Cliffs, New Jersey: Institute of Transportation Engineers, Prentice Hall. Newstead, S., Cameron, M., & Le, C. (2000). Vehicle crashworthiness and aggressivity ratings and crashworthiness by year of vehicle manufacture: Victoria and NSW crashes during 1987-98, Queensland crashes during 1991-98. (Report No. 171). Melbourne, Australia: Monash University Accident Research Centre. Newstead, S., & Corben, B. (2001). Evaluation of the 1992-1996 Transport Accident Commission funded accident black spot treatment program in Victoria. (Report No. 182). Melbourne, Australia: Monash University Accident Research Centre.

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

149

Newstead, S., & Mullan, N. (1996). Evaluation of the crash effects of the changes in speed zones in Victoria during 1993-1994 (excluding 100 to 110 km/h). (Report No. 98). Melbourne, Australia: Monash University Accident Research Centre. Nielsen, M. (2000). Accidents on rural roads in Denmark. VTI Konferens, 13A,7. Linkping, Sweden: Swedish National Road and Transport Research Institute Nilsson, G. (1984). Speed, accident rates and personal injury consequences for different road types. (VTI Rapport 277). Linkping, Swedish National Road and Transport Research Institute. Nilsson, G. (1990). Reduction in the speed limit from 110 km/h to 90 km/h during summer 1989. (VTI Rapport 358A). Linkping, Sweden: Swedish National Road and Transport Research Institute NTSB (1998). Safety at passive grade crossings. Volume 1: Analysis. (Report No. PB9817004 NTSB/SS-98/02). Washington, D.C: National Transportation Safety Board. OECD (1999). Safety strategies for rural roads. Paris, France: Organisation for Economic Co-operation and Development ONeill, W., Ullah, K., & Wang, M. (1993). IVHS and rural road safety: A prototype ATIS. (Report No. 93-23). Fargo, North Dakota: Mountain-Plains Consortium. Ogden, K. (1990). Road environment factors rural. Paper presented at the RTA Road Hazards Conference, Wollongong, Australia. Ogden, K.W. (1997). The effects of paved shoulders on accidents on rural highways. Accident Analysis and Prevention, 29, 353-362. Ogden, K.W., & Howie, D. (1989a). Traffic safety: Can we be proactive? Proceedings 14th Australian Transport Research Forum, (vol 2, pp 869-891). Perth, Australia. Ogden, K.W., & Howie, D. (1989b). Road crashes at bridges and culverts. A pro-active safety study. Proceedings 5th World Conference on Transport Research. (vol 1, pp 452467). Yokahama, Japan. Oxley, J., & Corben, B. (2002). Effective speed management. Report prepared for VicRoads. Melbourne, Australia: Monash University Accident Research Centre. Pain, R., Hanscom, F., & McGee, H. (1983). Work-site traffic controls in the US: Exciting and new technologies. Traffic Management and Control, 24(10), 477-484. Pajunen, K. (2002). Safety at passive grade crossings in Finland. Proceedings Getting Active at Passive Crossings: Seventh International Symposium on Railroad-Highway Grade Crossing Research and Safety. Melbourne, Australia: Monash University. Parker, M. (1997). Effects of raising and lowering speed limits on selected roadway sections. (Report No. FHWA-RD-92-084). Washington DC, US Department of Transport. Parsonson, B., Isler, R., & Hansson, G. (1996). Driver behaviour at rural intersections. (Transit New Zealand Research Report No. 56). Hamilton, New Zealand: Traffic and Road Safety Research Group, University of Waikato. Parsonson, P., & Marks, J. (1979). Wrong-way traffic movements on freeway ramps. (Report No. FHWA-GA-79-001, Research Project No 7703). Washington, DC: Washington DC: Federal Highway Administration, US Department of Transportation, Georgia Department of Transportation. Paterson, W.D. (1986). International roughness index: Relationship to other measures of roughness and riding quality. Transportation Research Record 1084, 49-59.

150

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

Peltola, H. (1991). Drivers opinions about seasonally changing speed limits. The experiment of seasonally changing speed limits in 1987-89, phase 2. Helsinki, Finland: Technical Research Centre of Finland (VTT). Persaud, B.N., Retting, R.A., Garder, P.E., & Lord, D. (2001). Crash reductions following installation of roundabouts in the United States. Proceedings Road Safety on Three Continents 2000. (VTI Konferens, NO: 15A) (pp. 540-542). Linkping, Sweden: Swedish National Road and Transport Research Institute. Persaud, B., Retting, R., & Lyon, C. (2001). Guidelines for identification of hazardous highway curves. Transportation Research Record, 1717, 14-18. Pigman, J., & Agent, K. (1990). Highway accidents in construction and maintenance zones. Transportation Research Record, 1270, 22-31. Pfefer, R., Stenzel, W., & Lee, B. (1991). Safety impact of the 65 mph speed limit: A time series analysis. Transportation Research Record, 1318, 22-33. Polus, A., Craus, J., Livneh, M., & Katznelson, S. (1999). Analysis of flow, safety and warrants for paved shoulders on two-lane rural highways. Road and Transport Research, 8, 42-56. Polus, A., Livneh, M., & Frischer, B. (2000). Evaluation of the passing process on twolane rural highways. Transportation Research Record, 1701, 53-60. Preston, H., & Schenecker, T. (1999a). Potential safety effects of dynamic signing at rural horizontal curves. (Report No. MN/RC-2000-14). St Paul, Minnesota: Minnesota Department of Transportation. Preston, H., & Schenecker, T. (1999b). Bypass lane safety, operations, and design study. (Report No. MN/RC-2000-22). St Paul, Minnesota: Minnesota Department of Transportation. Pronk, N., & Corben, B.F. (1999). Innovative low cost measures for motorcycle safety. Contract report prepared for VicRoads. Melbourne, Australia: Monash University Accident Research Centre. Puvanachandran, V. (1995). Effect of road curvature on accident frequency: Determining speeds to improve local curves. Road and Transport Research, 4(3), 76-83. Pyne, H., Carsten, O., & Tight, M. (1995). Speed on rural arterial roads. VTI Konferens, 2A(4), pp 8-22. Linkping, Sweden: Swedish National Road and Transport Research Institute Pyne, H., Dougherty, M., Carsten, O., & Tight, M. (1995). A simulator based evaluation of speed reduction measures for rural arterial roads. (ITS working paper 434). Leeds, UK: Institute for Transport Studies. Queensland Parliamentary Travelsafe Committee (1999). Rural road safety in Queensland. (Issues Paper No. 4). Brisbane, Australia: Queensland Legislative Assembly. RACV (1995). Calder Highway: highway safety review. (Report No.TS94/8). Noble Park, Australia: Royal Automobile Club of Victoria, Traffic & Safety Department. Rattenbury, S., & Gloyns, P. (1992). Accident patterns in rural areas and scope for countermeasures: vehicles and highways. Traffic Engineering and Control, 33, 540-544. Rechnitzer, G. (2002). Innovative measures to reduce crash occurrence and severe injury risk at passive railroad crossings. Proceedings Getting Active at Passive Crossings:

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

151

Seventh International Symposium on Railroad-Highway Grade Crossing Research and Safety. Melbourne, Australia: Monash University. Regan, D., & Beverley, K.I. (1978). Illusionary motion in depth: After-effect of adaptation to changing size, Vision Research, 18, 209-212. Regan, D., & Beverley, K.I. (1982). How do we avoid confounding the direction we are looking and the direction we are moving, Science, 215, 194-196. Regan, M., Triggs, T., & Godley, S. (1999). Simulator-based evaluation of a novice driver CD-ROM based training product: Issues and methods. Proceedings of the Road Safety Research, Policing and Education Conference, Canberra, ACT, Volume 1: 877-886. Reilly, W., Pfefer, R., Michaels, R ., Polus, A., & Schoen, J. (1989). Speed change lanes: Final publication. (Report No. 3-35). Washington, DC: National Co-operative Highway Research Program. Ribbens, H. (1986). Pedestrian and pedal cycle accidents on rural roads in South Africa. (Technical Report No. RV/24). National Institute for Transport and Road Research, Pretoria, South Africa. Ries, R., Carroll, A., & Warren, J. (2002). Closure of US highway-railroad grade crossings: A status report. Proceedings Getting Active at Passive Crossings: Seventh International Symposium on Railroad-Highway Grade Crossing Research and Safety. Melbourne, Australia: Monash University. Robinson, D.L (1998). Accidents at Roundabouts in New South Wales. Road & Transport Research. Vol 7 No. 1 March 1998. Rogness, R.O., Fambro, D.B., & Turner, D.S. (1982). Before-After Accident Analysis for Two Shoulder Upgrading Alternative. In Transportation Research Record 855, TRB, National Research Council, Washington D.C., pp. 41-46. Russell, E. (2002). A review of studies to improve safety at passive rail-highway crossings at grade (HRI). Proceedings Getting Active at Passive Crossings: Seventh International Symposium on Railroad-Highway Grade Crossing Research and Safety. Melbourne, Australia: Monash University. Russell, E., Stokes, R., & Rys, M. (1999). Analysis of rural intersection crashes caused by stop sign violation and failure to yield the right-of-way. VTI Konferens, no.13A, Part 7. (pp 17-36). Linkping, Sweden: Swedish National Road and Transport Research Institute. Ryan, G., Spittle, J., & Cooper, L. (1998). Truck crashes in Western Australia, 1988-1992. (Report No. RR 69). Road Accident Prevention Research Unit, Western Australia, Australia. Salvatore, S. (1972). The perception of real motion: A literature review. (Report No. ICRL-RR-70-7). Providence, Rhode Island: US Public Health Service Injury Control Research Laboratory. Sanderson, J.T., & Fildes, B.N. (1984). Run off the road accidents in rural areas. (Report No.TS84/6). Noble Park, Australia: Royal Automobile Club of Victoria, Traffic and Safety Department. Schmidt, G. (1996). Results of speed and accident studies on two-lane rural roads in Germany. VTI Konferens, 4A(2), pp 2-17. Linkping, Sweden: Swedish National Road and Transport Research Institute Schnerring, F. (1995). Roadside hazard management. Wodonga Rural Road Safety Seminar. Canberra, Australia: National Road Trauma Advisory Council.

152

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

Schouten, S. (2002). Euro star: Evaluating the safety of Europes roads. Traffic Technology International, 4-5, 70-74. Scifres, P., & Loutzenheister, R. (1975). Wrong-way movements on divided highways. (Report No. JHRP-13-75, HS-017 293). Purdue University and Indiana State Highway Commission JHRP. Shankar, V., Mannering, F., & Barfield, W. (1995). Effect of roadway geometrics and environmental factors on rural freeway accident frequencies. Accident Analysis and Prevention, 27, 370-389. Shinar, D. (1978). Psychology on the road: The human factor in highway safety. New York, Wiley. Silcock, R., Smith, K., Knox, D., & Beuret, K. (1999). What limits speed? Factors that affect how fast we drive. Hampshire: AA Foundation for Road Safety Research. Sliogeris, J. (1992). 110 kilometre per hour speed limit: Evaluation of road safety effects. (Report No. GR 98-8). Melbourne, Australia: VicRoads. Smith, B.L. (1982). Case for removing bridge or culvert rails on low volume rural roads. Transportation Research Record, 875, 50-53. Solomon, D. (1964). Accidents on main rural highways related to speed, driver and vehicle. Washington DC: US Department of Commerce, Bureau of Public Roads. Sorock, G., Ranney, T., & Lehto, M. (1996). Motor vehicle crashes in roadway construction workzones: An analysis using narrative text from insurance claims. Accident Analysis and Prevention, 28, 131-138. Stackhouse, S., & Cassidy, P. (1996). Warning flashers at rural intersections. (Report No. MN/RC-1998/01). St Paul, Minnesota: Minnesota Department of Transportation. Stamatiadis, N., Jones, S., & Aultman-Hall, L. (1999). Causal factors for accidents on south-eastern low-volume rural roads. Transportation Research Record, 1652, 111-117. Staplin, L., Harkey, D., Lococo, K., & Tarawneh, M. (1997). Intersection geometric design and operational guidelines for older drivers and pedestrians. Volume 1, Final Report . (Report No. FHWA-RD-96-132). Washington DC: US Department of Transportation. Staplin, L., Lococo, K., & Byington, S. (1998). Older driver highway design manual. (Report No. FHWA-RD-97-135). Washington DC: US Department of Transportation. Steinbrecher, J. (1992). Restructuring of town entrances on roads classified as major road safety in Europe. VTI Rapport, no.380A, part 4, pp 17-31. Linkping, Sweden: Swedish National Road and Transport Research Institute. Steyer, R., Sossoumihen, A., & Weise, G. (2000). Traffic safety on two-lane rural roads: New concepts and findings. Proceedings of the 2nd International Symposium on Highway Geometric Design (pp 299-312). Cologne, Germany: Road and Transportation Research Association. Stokes, R., Rys, M., Russell, E., Robinson, R., & Budke (2000). Analysis of rural intersection accidents caused by stop sign violation and failure to yield the right-of-way. (Report No. K-TRAN: KSU-98-6). Kansas Department of Transportation, The Kansas State University and The University of Kansas. Strate, H. (1980). Making highways safe: A realistic approach. Institute of Transport Engineers Journal.

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

153

Sweatman, P., Ogden, K., Haworth, N., Pearson, R., & Vulcan, P. (1990). New South Wales heavy vehicle crash study: Final Technical Report . Canberra, Australia: Federal Office of Road Safety. Talarico, R.J., & Morrall, J.F. (1994). The cost-effectiveness of curve flattening in Alberta. Canadian Journal of Civil Engineering, 21, 285-296. Tamburri, T. (1969). Wrong-way driving accidents are reduced. Highway Research Record, 292. Tamburri, T., & Theobald, D. (1965). Wrong-way driving (Phase II). California Division of Highways, Department of Public Works. Taylor, M.C., & Barker (1992). Injury accidents on rural single-carriageway roads: An analysis of STATS19 data. (Report No. 365). Crowthorne, United Kingdom: Transport Research Laboratory. Taylor, M., Baruya, A., & Kennedy, J. (2002). The relationship between speed and accidents on rural single-carriageway roads. (Report No. TRL511). Crowthorne, UK: Transport Research Laboratory. Taylor, W., & Foody, T. (1966). Ohios curve delineation program: An analysis. Traffic Engineering, 36, 41-45. Taylor, J., & McGee, H. (1973). Improving traffic operations and safety at exit gore areas. (Publication No. 145). Washington DC: National Cooperative Highway Research Program Terhune, K., & Parker, M. (1986). An evaluation of accident surrogates for safety analysis of rural highways. (Report No. FHWA-RD-86/127). Washington DC: Federal Highway Administration, US Department of Transport. Tessmer, J. (1996). Rural and urban crashes: A comparative analysis. (Report No. DOTHS-808-450). Washington DC: National Highway Traffic Safety Administration, US Department of Transport. Thakkar, J.S. (1984). Study effect of two-way left-turn lanes on traffic-accidents. In Transportation Research Board 960, National Research Council, Washington D.C. Toivonen, S., & Niskanen, R. (1998). Safety problems of rural roads. Finlands material for RS8 Scientific Expert Group. (Report No. 17/1998). Helsinki, Finland: Finnish National Road Administration. Tom, G. (1995). Accidents on spiral transition curves. ITE Journal, 65(9), 49-53. TranScan (2001). Roads in proper context: Providing mobility, safety, and aesthetics. NCHRP Project 20-36, Highway Research and Technology International Information Sharing. Transportation Research Board (1987). Designing safer roads: Practices for resurfacing, restoration, and rehabilitation. (Special Report No. 214). Washington DC: Transportation Research Board, National Research Council. Triggs, T.J. (1986). Speed estimation. In G.A. Peters and B.J. Peters (Eds), Automotive Engineering and Litigation, Vol. l (Supplement). (pp 95-124). Garland Press. Troxel, L., Ray, M., & Carney, J. (1994). Accident data analysis of side-impact, fixedobject collisions. (Report No. SA-89-006). Washington DC: Federal Highway Administration, US Department of Transport.

154

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

Troutbeck, R.J. (1995). Predicting accident rates of rural roads. International Conference on Accident Investigation, Reconstruction, Interpretation and the Law. Brisbane, Australia: Queensland University of Technology. Turner, D. (1984). Prediction of bridge accident rates. ASCE Journal of Transportation Engineering, 116(1), 90-103. Tziotis, M. (1993). Crashes on the approaches to provincial cities. Highway Engineering in Australia, 25(4), 14-21. Tziotis, M. (1993). Evaluation of mid-block accident Black Spot treatments. (Report No. 48). Melbourne, Australia: Monash University Accident Research Centre. Van Schagen, I., & Janssen, T. (2000). Managing road transport risks: Sustainable safety in the Netherlands. IATSS Research, 24(2), 18-27. Vaswani, N. (1977). Virginias crash Transportation Research Record, 644. program to reduce wrong-way driving.

VicRoads (1998). Safety on rural roads. (Report No. 108). Melbourne, Australia: VicRoads, Road Safety Department. Victorian Parliamentary Road Safety Committee (2002). Inquiry into rural road safety and infrastructure. Melbourne, Australia: Victorian Government Printer. Voigt, A. (1996). Evaluation of alternative horizontal curve design approaches on rural two-lane highways. (Report No.TTI-04690-3). College Station, Texas: Texas Transportation Institute. Voigt, A., & Krammes, R. (1998). An operational and safety evaluation of alternative horizontal curve design approaches on rural two-lane highways. VTT (1992). Accidents as a function of road and traffic characteristics. Nordic Road and Transport Research, 1, 8-11. Walker, C., & Lines, C., (1991). Accident reductions from trunk road improvements. (Report No. 321). Crowthorne, UK: Transport and Road Research Laboratory. Walmsley, D.A., & Summersgill, I. (1998). The relationship between road layout and accidents on modern rural trunk roads. (Report, No. 334). Transport Research Laboratory (TRL), Crowthorne, United Kingdom. Walmsley, D.A., Summersgill, I., & Payne, A. (1998). Accidents on modern rural dualcarriageway trunk roads. (Report No. 335), Crowthorne, UK: Transport and Road Research Laboratory. Walton, C.M., Machemehl, R.B., Horne, T.W., & Fung, W.K. (1979). Accident and operation guidelines for continuous two-way left-turn median lanes. In Transportation Research Board 737, National Research Council, Washington D.C. Wang, J., Hughes, W., Council, F., & Paniati, J. (1995). Investigation of highway work zone crashes: What we know and what we dont know. Transportation Research Record, 1529, 54-62. Warren, D. (1982). Speed zoning and control. Synthesis of safety research related to traffic control and roadway elements, Volume 2. (Report No. FHWA-TS-82-233). Washington DC: US Department of Transport. Warren, W.H., & Shaw, R.E. (1985). Events and encounters as units of analysis for ecological psychology. In W.H. Warren and R.E. Shaw (Eds), Persistence and Change:

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

155

Proceedings of the 1st International Conference on Event Perception. Englewood Cliffs, New Jersey: Lawrence Erlbaum. Weaver, G., & Glennon, J. (1972). Design and striping for safe passing operations. Highway Research Record, 390, 36-39. Wegman, F. (1995). Influence of infrastructure design on road safety. Paper presented at the International Symposium on Traffic Safety: A Global Issue, Kuwait. (No. D-95-1). Leidschendam, Netherlands: Institute for Road Safety Research (SWOV) Weiland, R., & Woll, T. (2002). Challenges and opportunities for ITS standards at highway-rail intersections. Proceedings Getting Active at Passive Crossings: Seventh International Symposium on Railroad-Highway Grade Crossing Research and Safety. Melbourne, Australia: Monash University. Wertheim, A. (1978). Explaining highway hypnosis: Experimental evidence for the role of eye movements. Accident Analysis and Prevention, 10, 111-129. West, L., & Dunn, J. (1971). Accidents, speed deviation and speed limits. Traffic Engineering, 41(10), 17-50. Wilson, R. (1991). Subtypes of DUIs and high-risk drivers: Implications for differential intervention. Alcohol, Drugs and Driving, 7, 1-12. Wilson, D.J., & Dunn, R.C.M. (1994). Safer intersection design using an expert system. Proceedings International Conference on Safety and the Environment in the 21st Century. (pp. 397-408). Tel-Aviv, Israel: Israel. Ministry of Transport, Highway Safety Administration. Winnett, M., & Wheeler, A. (2002). Vehicle-activated signs: A large scale evaluation. (TRL Report No. TRL548). Crowthorne, UK: Transport Research Laboratory. Woods, D., Rowan, N., & Johnson, J. (1970). Significant points from the diagnostic field studies: Summary report . (Report No. 606-4). Texas Transportation Institute, Texas A & M University. Wrigglesworth, E. (2001). A human factors commentary on innovations at railroadhighway grade crossings in Australia. Journal of Safety Research, 32, 309-321. Wrigglesworth, E. (2002). Passive railroad-highway grade crossings: What are the research priorities? Proceedings Getting Active at Passive Crossings: Seventh International Symposium on Railroad-Highway Grade Crossing Research and Safety. Melbourne, Australia: Monash University. Wigglesworth, E.C., Corben, B.F., & Triggs, T.J. (1999). A trial and evaluation of cost effective measures to improve safety at passive railway level crossings. Contract report for VicRoads. Melbourne, Australia: Monash University Accident Research Centre. Wright, P., & Zador, D. (1981). Study of fatal rollover crashes i n Georgia. Transportation Research Record, 819, 8-17. Yamashita, M., & Takata, T. (1991). Wake-up tire noise for safety driving. Proceedings International Conference on Noise Control Engineering (vol.2, pp 831-834). Sydney, Australia: Australian Acoustical Society. Ydenius, A., Kullgren, A., & Tingvall, C. (2001). Development of a crashworthy system: Interaction between car structural integrity, restraint systems and guard-rails. (Paper No. 171). Proceedings of the 17th ESV Conference, Amsterdam, The Netherlands.

156

MONASH UNIVERSITY ACCIDENT RESEARCH CENTRE

Yeates, M. (2001). Can urban roads be safe for motorists if they are not safe for people walking or cycling? Proceedings of the National Speed and Road Safety Conference, Adelaide, Australia. Zador, P., Stein, H., Hall, P., & Wright, J. (1985). Effects of chevrons, post-mounted delineators and raised pavement markers on driver behaviour at roadway curves. Washington DC: Insurance Institute for Highway Safety. Zegeer, C.V., & Council, F.M. (1995). Safety relationships associated with cross sectional roadway elements. Transportation Research Record, 1512, 29-36. Zegeer, C., Stewart, J., Council, F., & Reinfurt, D. (1992). Safety effects of geometric improvements on horizontal curves. Transportation Research Record, 1356, 11-19. Zegeer, C., Stewart, J ., Reinfurt, D., Council, F., Neuman, T., Hamilton, E., Miller, T., & Hunter, W. (1991). Cost effective geometric improvements for safety upgrading of horizontal curves. (Report No. FHWA-RD-90-021). Washington DC: Federal Highway Administration, US Department of Transport.

COST -EFFECTIVE INFRASTRUCTURE MEASURES ON RURAL ROADS

157