1 4 0 WAffKR.

RKSOURCES ENGINEERING

5.2 Recurrence Interval

The recurrence interval1 is defined as the average interval in years between th
occurrence of a flood of specified magnitude and an equal or larger flood. Th
mth largest flood in a data series has been equaled or exceeded m times in t*
period of record. N years and an estimate of its recurrence interval Tr as given
the Weilbull formula is
N+ 1
T = (5.
m
Several other formulas have been suggested for the calculation of recurren
interval or return period. The disagreement between the various formulas is limit
to the larger floods, where n%is small. If m equals 5 or more, the calculated valu
of Tr by all methods are almost identical. Equation (5.1) can be used to defi
plotting positions (Fig. 5.3), which provide a goód estimate of flood flows wi
return periods of less than 20 yr.

1 Recurrence interval is also referred to as return period. There is no implication that floods with
return period of Tr will recur precisely T r years apart. For example, one would expect the 5-yr fl
to be equaled or exceeded approximately 20 times in a 100-yr period. The recurrence could Occur
successive years or there might be a span of considerably more than 5 yr between recurrences.
 PROBABILITY CONCEPTS IN PLANNING 141

Values of b

frequency curve of annual floods for the Susquehanna River at Harrisburg, Pennsylvania (1874^-1949).

If an event has a true recurrence interval of Tr years, then the probability P

that it will be equaled or exceeded in any one year is
i
( 5.2)
,:T ; ■

lince the only possibilities áre that the event will or will nót occur in any year,
the probability that it Will not occur in a given year is 1 — P. From the principles
f probability, the probability J that at least one event that equals or exceeds
i Tr-year event will occur in any series of N years is
y = i - (i - P)N , , , (5.3)
ills equation is derived as follows:
P is the probability of the occurrence of an event,
' 1 —P is the probability that the event will not occur. . '
(1 — P)(l —P) is the probability the event will not occur in two successive
m. ;.: ‘ 'J -" C ■
(1 —P)3 is the probability, that the event will hot occur in three successive
M. ¡- /i
.!■ (1 —p)N is the probability that the event will nbt occur during a span of N
Ccessive years. * >
Hence J = 1 —(1 —P)N is the probability that the event will occur during
ipan of N years.
 142 WATERf RESOURCES ENGINEERING

TABLE 5.2
Probability that an event of given recurrence interval will be equaled or
exceeded during periods o f various lengths
Probability /fo r Various Periods

TPy yr 1 yr Syr 10 yr 25 yr 50 yr 100 yr 200 yr SOO yr

" - 5i 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0

: . . ..... - j
0.5 0.97 0.999 . * 4 * *

% 0.2 0.67 0.89 0.996 ■ I* * ■ * *

10 0.1 0.41 0.65 0.93 0.995 *• ■■ * ■ ....* ■•

50 0.02 0.10 0.18 0.40 0.64 0.87 0'98 *

to o 0.01 0 05 0.10 0.22 0.40 0.63 0,87 0.993

26o 0.005 0.02 0.05 0.12 0.22 : 0.39 - í 0.63 0.92

* In these cases J can never be exactly 1, but for all practical purposes its value may be'taken as unity.

Table 5.2, which has been computed from Eq. (5.3), shows that there are 4 i
chances in 10 that the 100-yr flood (or greater) will occur in any 50-yr period and^j
even a 22 percent probability that the 200-yr flood (or greater) might occur in the
50-yr period. On the other hand, there are 36 chances in 100 that the 50-yr flood?;
will not occur ih any 50-yr period. Equation 5.3 (or Table 5.2) may be used to!
estimate the risk of failure during the lifetime of a project when using different ;
design criteria. j
Table 5,2 illustrates also that there -can be no inference that the “ A-year \
flood” will be equaled or exceeded exactly once in every period of N years. All?
that is meant is that in a lortg period, say 10,000 years, there will be 10,000/AT
floods equal to or greater than the A-year flood. All such floods might occur in*j
consecutive years, but this is not very probable. i
If the design flood for a particular project is to have a recurrence Jnterva'lj
much shorter than the period of record, its value may be determined by, plottings
peak flows versus Tr as computed from Eq. (5.1) and sketching a curve through^
the plotted points (Fig. 5.3). Because of inaccuracies in the plotted positions of the*
larger floods, a line sketched to conform to these floods may depart substantially
from the location of the true frequency curvé. ;
' ■
■ 1 / ■■
' . ■ ■
■: \ ■r ■: .r'' ■
. •. " - ■■ ... v-n t ■>-<.,

5.3 Statistical M ethods for Estim ating the

Frequency o f Rare Events j
With an extremely long period of record it would be possible to use a smaller^
class interval, and Fig. 5.1 might approach a smoqth frequency distribution such;
as Fig. 5.4. The ordinates of Fig. 5.4 are probability density and the abscissas ar$
the magnitudes of the floods. The ratio of the area under the curve above any
magnitude X x to the area under the entire curve is the probability that X x ,\yiH béj
equaled or exceeded in any year. Jj
 PROBABILITY CONCEPTS IN PLANNING J43

FIGURE 5.4
Idealized flood frequency,distribution.

Many kinds of évents conform to one of several standard frequency distribu­

tions that have been studied at length and the equation of the distribution well
established. The probability of such events can be determined quite easily. Only
U very large number of samples (i.e., a long record length) will permit accurate
definition of a distribution, and no streamflow records are long enough to
positively establish the appropriate distribution. It is known that X must be greater
than zero and that future floods will exceed those that have been observed.
Several distributions have been suggested1 as appropriate for streamflow,
but there is no real proof of their validity. Fisher and Tippett2 showed that if one
selected the largest event from each of many large samples; the distribution of
these extreme values was independent of the original distribution and conformed
to a limiting function. Gumbel3 suggested that this distribution of extreme values
Was appropriate for flood analysis since the annual flood could be assumed to be
tjie largest of a saiñple of 365 possible valúes each year. Based on the argument
that the distribution Of floods is unlimited, i.e., that there is no physical limit to
the maximum flood, he proposed that the probability P of the occurrence of a
Value equal to or greater than any X be expressed as
P = 1 - e~9’b (5.4)
Where e is the base of Natural logarithms and b is given by

b- ^ ( X - X + 0.45<r) (5.5)
0.7797o- ' /

*. A. Foster, Theoretical Frequency Curves, Trans. ASCE, Vol. 87, pp. 142-173, 1924; Allen Hazen,
flood Flows,” Wiley, New York, 1930; and L. R. Beard, Statistical Analysis in Hydrology, Trans.
5CE, Vol. 103, pp. 1110-1160, 1943.
, A. Fisher and L. H. C. Tippett, Limiting Fonns of the Frequency Distribution of the Largest or
llllest Member of a Sample, Proc. Cambridge Philos. Soc., Vol. 24, pp. 180-190, 1928.
A#J. Gumbel, FJoods Estimated by the Probability Method, Eng. News^Record, Vol. 134, pp. 833^-837,
IMS.
144 WATER-RESOURCES ENGINEERING *

In Eq. (5.5), X is the flood magnitude with the probability P, X is the arithmetic
average of all floods in the series, and a is the standard deviation of the series
computed from
~Y,(x - ^ 2_ i l/2
( 5,6)

where N is the number of items in the series (the number of years of record),
The probability P is related to the recurrence interval Tr f>y Eq. (5.2). Válues of b
corresponding to various return periods are given in Appendix A-4.

Example 5.1. Using the data of Table 5.1, find the theoretical recurrence interval for
a flood flow of 700,000 cfs using the Gumbel approach.

Solution. Expressing all flows in thousands of cfs, from the table, X = 287.8 and a —
\(962^|67/^f5.= K.' u ;;

1
[700 - 288 + 0.45(113.3)] = 5.24
0.7797 x 113.3
The recurrence interval for X = 700 is, from Eqs. (5.2) and (5.4),
7^:
T r- = 189 yr
1
D l:.. .J
By the same method Tr =f= 1.28 yr when X = 200 and 6.89 yr when X ^ 400. TheseJ
.points are, shown, on Fig. 5.3 by the large circles. * a
>3
The plotting paper used for Fig. 5.3 is constructed by laying out on ¡
linear scale o f b the corresponding values of Tr = \/P from Eq. (5.4). Thus th
computed line will be straight, and it is sufficient to calculate the return period
corresponding to two flows. A third point is a convenient check. |
In 1967, the U.S. Water Resources Council1 adopted the log Pearson Type
III distribution (of which the lognormal is a special case) as a standard for use by
federal agencies. The purpose was to achieve standardization of procedures. The
recommended procedure 12 to conyert the series to flogarithms apd compute th |
mean, standard deviation, and skew coefficient <7, which is
N ^ (log X —log X)?
9 = ( 5.7)
(N - 1)(N - 2)(aloeX)3
The values of X for various periods are computed from
log X = log X + K<Jlogx (5 -3)

1 A Uniform Technique for Determining Flood Flow Frequencies, U.S. Water ¿Resources Counci.
H ydrol Comm. Bull. 15, December 1967, Revised June 1977. 1
2 Subcommittee on Hydrology, Methods o f Flow Frequency Analysis, Interagency Comm. W atk
Resources Bull.'13, U.S. Government Printing Office, Washington, D.C., April 1966.
 PROBABILITY CONCEPTS IN PLANNING 145

where K is selected from Table A-5 (Appendix) for the computed value of g and
the desired return period. Lognormal probability paper should be used for
graphical display of the curves. A straight line will result only if g = 0.

Example 5.2. Using the data of Table 5.1, find the magnitude of the 10- and 100-yr
floods using the log Pearson Type III distribution.

Solution, Expressing all flows in thousands of cfs, from the table, log X — 2.432,
<TlogX= (1.641/75)0*5 = 0.148, and g = (76 x 0.1619)/[75 x 74 x (0.148)3] = 0.682,
Obtaining K values from the table in the Appendix we get
10-yr flood: log* = 2.432 + (1.332X0.148) = 2.628
Hence, X i0 = 429,000 cfs
100-yr flood: log* = 2.432 + (2.810X0.148).= 2.828
Hence, X100 = 705,000 cfs
Comparing these results with the Gumbel, plot of Fig. 5.3 we note the following:

Gumbel ; Log Péarson typ e III

10-yr flood 435.000 cfs . 429,000 cfs

100-yr flood 640.000 cfs 705,000 cfs

Figure 5.3 is typical of frequency plots ip that the computed line conforms
Well to most of the data but diverges from some of the largest values. The discussion
of recurrence interval in Sec. 5.2 points out that these higher points may be
Incorrectly located. Hence, to force a distribution to conform J o these points ipay
Only perpetuate th^ error,- Until much longer records are available* there is no
proof •o f’thé adequacyw ith -which a theoretical distribution;: fits the actual
distribution of floods. There are logical (grounds for arguing that no single
theoretical distribution can be expected to fit all streams.1

&4 Partial Duration

The annual series is sometimes criticized on the basis that the second highest flood
in some years will exceed annual floods that are included in the series. The partial
juration series, all floods above some arbitrary base value of flow, is sometimes
Suggested as a substitute. The two series give nearly the same recurrence intervals
for the larger floods, but the partial duration series will show higher flows for the
Shorter recurrence intervals. The partial duration series should not be used to

1 R. F. Ott, Streamflow Frequency Using Stochastically Generated Hourly Rainfall, Technical Report
151, Department of Civil Engineering, Stanford University, December 1971,
1 46 WATER-RESOURCES ENGINEERING

determine the frequency of rare events. Where the lesser floods are of specie
interest and particularly where a recurrence interval less than 2 yr is desired, th
partial series should be used by simply plotting» flood peaks versus Tr arid sketching
a curve by eye.

5.5 Flood Form ulas

If the highest floods of record at a group of gaging stations within a limited arei
are plotted on a graph such as Fig. 5.5, where peak flow Qp in cubic feet pe:
second per square mile of drainage área is plotted against drainage area Ad\ a fey
of the higher floods seem to define an upper-limit line, or enveloping curve (sotó
line). On logarithmic paper the slope of the line is the exponent n, and the jntercep
where area is unity is the coefficient c in an equation of the form i
' ■ ^ (5.9
If exponent n is often taken as —0.5, indicating that flood peaks vary inverse^
with* the square root of drainage area (dashed line, Fig. 5.5). It Can be seen fron
the figure that the slope of the line and the intercept are poorly defined by the data.
\ Most flood formulas assume that flood-peak magnitude is a function o
drainage krea, and many formulas include such factors as basin shape and meat
annual precipitation in an attempt to minimize the variations in the coefficient c

Drainage area in km
100! 200 500 1000 2Ó0Ó 5000 10,000

HÉHMMUiÉÉiÉáNl
1000 "C--------- ! Record - 10.0
y k- ■ •; ■ * ■ ‘f No. Station , years*
800
r , 23
in i'U ii.' - h Kerrs Cr. nr Lexington,, Va.<
600 t 2 Mechym R, nrjvy, 7
• 500, •V:, ■ *'' 3 Sandy R. hr Danville, Va. 20 5.0
■V
400; ■
*; s . r i ' «nin iii-0 .5 , 4-" Smith R; at Martinsville; Va. 20
s. ■
■
■
, d
5 Rivanna R. at Ralmyrai, Va. 15
i 300 . "V
. s 6 Roarioke R. at Altavista, Va., 1*9
■V f s ■;.y
•l i; Roanoke R.'at Clover, Va: 20
\ 200 . <1
s 8
7
Roanoke R. at Clarkesville, Va. 15 V 2.0
• , Q = 1660. S* ,
* A-d0 *2
4 - >
«i «. *
100 m m•.• 1•
• •
> 1.0 ‘
80 . • \
60 . .• r
■■ ¿ 6-
f
,'50 ’"f1 - • ■ - 0.5 -A
40 • 8
* r?-v; V!; (} ¿i.
■
• A-
30
... ■ ■i

20 i 0.2
Equaitions are expressed
í
' * f n English units M

10 1L 1 I I i-L
10 20 30 40 60 80100 200 300 500 700 1000 2000 6000
Drainage area in sqflare miles

FIGURE 5.5
Record floods on coastal streams of Virginia prior to 1949.
 PROBABILITY CONCEPTS IN PLANNING J$7

Early engineers were forced to use equations like Eq. (5.9) because they had no
Other data with which to work. Such a simple formula cannot possibly describe
the complex phenomena involved in a flood. Moreóver, such formulas offer ho
glue to the probability of the flow computed from them. With relatively abundant
data and far better understanding of hydrology* the modern engineer has little
excuse for utilizing empirical formulas. Since critical decisions concerning a project
arc often made on the balsis of preliminary studies, approximations are not evfn
Justified in such studies. )

S.6 Flood Frequency at Point? without Streamflow

Records ! V ' 1 ; ; j i j

Few projects are built at the exact spot where a streamflow record has been
Obtained. Many projects are buili on streams where no record exists. Several
alternative methods have been used to estimate flood frequency, in the absence of
Itreamflow data., If hodrly rainfall recqrds are available, one method is to simulate
the hydrographs of storm runoff for major storms using synthetic unit hydrographs
(Sec. 3.13) and adding base flow to obtain an annual flood series that can then be
lubjected to a frequency analysis using the Gumbbl or log Pearson Type III
distribution. Because of inherent inaccuracy oHhé ^nthetic unit hydrograph, the
results óf such an analysis are questionable. Ótliérv approaches to frequency
analysis of ungaged streams include regional streamflow analysis (Sec.; 5.7) and
Computer simulation (Sec. 5.8). \

5.7 Regional Streamflow Analysis

In a regional streamflow analysis the flood frequency at an ungaged point is
estimated from data at nearby gating stations on either the same catchment or
nearby catchments with similar characteristics. Frequency curves for two gáging
Stations can be identical only when the two basins are quite ’s t e l l a r ^
Should have geometric similarity in terms of area, shape, slope, and topography;
hydrologic similarity in terms of rainfall, snowfall, soils, and valley storage; and
geologic similarity with regard to those items that affect groundwater flow. The
frequency curves in Fig. 5.6 represent six basins of the same general size in the
Puyallup River basin,1 which drains from Mt. Rainier in western Washington,
^ th o u g h the stations are close together, a considerable divergence of the curves
evident. Note that flood magnitudes are expressed in terms of the ratio to the
$iean annual flood. This ratio eliminates some of the differences caused by
lifferences in basin size and by rainfall variations between basins! The character-
Í Itics of Eq. (5.4) are such that the mean annual flood has a recurrence interval
2.33 yí. "

,* p . L. Bodhaine andW. H. Robinson, Floodsipi Western Wa5hington, US. Geol. Surv. Circ. 191,1952.
148 WATER-RESOURCES ENGINEERING

FIGURE 5.6
Comparison of flood frequency curves for adjacent basins.

A substantial error might result from the Use of one of these frequency curved
on any other basin. The' basins áre in a mountainous krea, and their flood flowk
áre influenced by orographic rainfall variations and t>y the varying role of snow]
on the hydrology of the basin. Some differences may be due to Chance variations*)
Which can be importánt when the length of record is short (in this case, Í8 yr).j ]
Figure 5.7 shows the relation between drainage area and mean annual flood;!
for the six basins of Fig. 5.6. Note that one of the points plots off the curve. This
point is known as an outlier, a data point that possesses some peculiarity that
causes it to diverge substantially from the other data points. In this case, this bastó
may have been at a higher elevation than the Others arid subjected to much higher!^
 PROBABILITY CONCEPTS IN PLANNING 149

Drainage area in km2 ,

100 200 500 1000 2000

10,000
8000
,f 200
Ú <' , .! ■ \
6000
5000
0/
4000 C>
4' 100
3000 ■!

Mean annual (2.33-y) flood in m3/s

/o
«
.6 2000
9* -50
O

1000
800 ,
20
600
1 500
400
l
300
' '<M' ;; m
200
‘-f ' ■SM.
: ' ■' ■11,9 . ‘ :!
S■’ )9 :S '\
100
10 20 30 40 60 80 100
Drainage area in mi2 íw í ,,
' l U ; , : . - , n\

Variation of mean annual flood with drainage area. ,

tm 9m
Tflinfall or it fright have had a relatively S
small time of concentration that would
result in higher flows, Figures 5.6 and 5.7 will permit on^ to estimate flood flows
on nearby watersheds that have nearly the same characteristics as those whose
data is plotted in the figures. > r •<, ,

Example 5.3 Estimate the 100-yr flood at an ungaged point on a catchment in, the
Puyallup River basin if the drainage area of the catchment is 152 mi2 and it possesses
characteristics that conform most closely to basin 4 of Fig. 5.6.

Solution. From Fig. 5.7 the mean annual flood equals 2700 cfs and front Fig. 5.6,
(Gp)ioo/«2p)2.33 = 2.1. Hence,
'mu\'
(Gp)ioo = 21 X 2700 = 5670 cfs
 WATER-RESOURCES ENGINEERING *

This is an estimate of the 100-yr flood. However, it could easily be in error by as

much as 20 percent. ,

5.8 Computer Simulation o f Ungaged

Streams » ‘ >. j
If flow data atid hourly rainfall data for nearby basins with characteristics i
similar to the ungaged basin and hourly rainfall data representative of the
ungaged;basin are available, computer simulation (Sec. 3.21) is possible. Param a i
eters determined through calibration of the nearby gaged basins provide guid­
ance for the selection of parameters for the ungaged basin. By simulation, an entire:
frequency series may be generated and a frequency curve constructed. The physical j
characteristics and the rainfall record of the ungaged watershed can thus be utilized
in a rational way rather than by gross comparison with some other basin.
The ¡choice óf methods for frequency analysis of floods on ungaged basins J]
depends on* what type of data is available. Budgetary considerations may also
govern, r ; ; 7- ; ' 4 - -f : - ■- - i

5.9 Rainfall Frequency

In developing the relation between intensity, duration, and frequency of ráinfall \
at a given gage, the data set is derived by examining the recorded rainfall for each j
year and noting the maximum rainfall amount for each duration of interest such
as at 30 ^iin, 1 hr, 2“hr, 6 hr, 24 hr, etc. Rain gage records are required for short
durations. The datá set consists of the maximum rainfall amount for the given ;
duration for each year over the period of record. The data set4s then subjected !
to a frequency analysis usually employing either the Gumbel or log Pearson Type4
III distribution. Data sets for each duration are analyzed and combined into a^
family of curves (Fig. 5.8). In addition tó presenting the results in the form of
rainfall duration-frequency curves as in Fig. 5.8, the results are often presented as^j
intensity-durationrfrequency (IDF) curves. The curves of Fig. 5.8 can be readily ;
convérted to such a form: The ID F curves are useful as they provide information ^
that is directly applicable to the rational formula (Sec. 3:11). I
The precediiig procedurevrelátes to point rdinfall characteristics. Whefl^
dealing-with large watersheds, the frequency of average rainfall over large areal
is of interest. The best method for estimating the frequency of average rainfall over"
large ¿tea's is to calculate Of estimate thé average rainfall (Sec. 2.6) on the area^
during important stonhs arid make á frequency analysis of these average values.:!
It is incorrect to average the individual AT-year values at a number of rainfall!
stations to determine the Af-year average value over an area. Unless the several
events cári occur simuitanebüsly, the average value has a return period greater!
than that of the individual values (see Sec. 5.1). -
 PROBABILITY CONCEPTS IN PLANNING 151

Figure 5.9 is one of a series of rainfall intensity maps1 based ¡on analysis of all
available recording gage records' The maps cover a range of durations from 30
Min to 24 hr and frequencies from 1 to 100 yr. The maps are reliable in areas of
n e g l i g i b l e relief but may be inaccurate in mountainous areas. Actually only the
U and 24-hr maps are constructed from observed data, the others being inter-
Pplated by relationships presented in the report.
^ The relation between rainfall intensity i and duration tR has often been
pressed by formulas such as

t
* ■ . ■ ,
ft
■} ' (5,10)

¡«ere the constants k and n are regional characteristics.

Rftinfall Frequency Atlas for the United States, U.S. Weather Bur. Tech. Paper 40, May 1961.
 FIGURE 5.9
1" g j E j t t ? N^ a l Weather Service) ^ „,
i »— — -
 PROBABILITY CONCEPTS IN PLANNING 153

It is always preferable to utilize actual rainfall data in a frequency analysis

Instead of using maps or formulas. Sometimes, however* it is convenient to express
the rainfall intensity-frequency relations for a station in terms of a formula such
II Eq. (5.10). The ¡constants in the equation can be determined by plotting intensity
Vlfsus duration op logarithmic paper. If these data plot as a straight line, k is the
Intensity where tg is unity and n is the slope of the linei

0. 11 C onditional or J o in t P rob ab ility

If two events are entirely independent (unrelated in cause) and their probabilities
Of occurrence áre Px and P2, respectively, the probability that they will occur at
the same time is PxP2: Since both Px and Pz aré less than J , the probability of
thilir joint occurrence is less thán ttie probability of either event independently.
Numerous hydraulic design problems involve the simultaneous occurrence of
different events. It has been cbmmqn practice to transform a design rainfall of
known frequency to streamflow by the methods of Ghap. 3 and to assume that
the resulting flow has the same frequency as the rainfall. Since the transformation
Involves an assumption as to antecedent conditions, it is a problem in conditional
probability. Storm rainfall appears to be substantially independent of antecedent
conditions. Unless the selected runoff ; coefficient or infiltration index has a
probability of 1.0, the resulting flood has a return period different than that of
The rainfall—usually greater. This procédure should never be Used, except for very
imall impervious areas.
The spillway design flood for a dam depends on the assumed flood flows
Into the reservoir and the available storage space in the reservoir at the time these
flows occur. The worst possible condition is that the reservoir be full when the
Illum ed inflow occurs. However, if the probability of the reservoir being full at
the time the design flood occurs is less than 1* the computed spillway design flood
fill a greater recurrence interval and hence a smaller probability of occurrence
than the assumed reservoir inflow.
Because most hydrologic events are not strictly independent, it is usually
flCCessary to solve problems of joint frequency by direct analysis rather than by
<Uie of the simple product rule.
1, Figure 5.10 illustrates a joint frequency analysis applied to the problem of
«the simultaneous flooding of two streams above their junction. Curves ^ and 2?
jpf the figure are the separate frequency curves of the two streams. Curve C is a
pquency curve for the sum of the two flows computed oh the assumption of
complete dependence, i.e., events of equal. probability, are..assumed to occur
jllmultaneously and the total iflow below the junction is the sum of the flows on
\ht two streams having the same return periods. Curve D assumes complete
Independence so that the probability of any two flows occurring simultaneously
ll the product of the probability that they will occur independently. Curve E is
Obtained by adding the flows that actually occurred at the same time and
performing conventional frequency analysis pn the sums. It is evident that the
flows on the two streatns are partially but not wholly dependent.
 tW WATER-RBSOURCES ENGINEERING

Example of joint-frequency analysis.

Instead of attempting to combine probabilities, it is often more satisfactory

to work with synthetic record. Using rainfall-runoff relations and unit hydros
graphs, one may synthesize1 the flood peaks for years without streamflow record
and subject the computed peaks to frequency analysis. Computer simulation can
serve the same purpose.

5.12 Probable M axim um F loods

Since about 1940, the spillways of many major dams have been designed to
discharge the probable maximum flood . The magnitude of this flood is determined
by meteorologic estimate2 of the physical limit of rainfall over the drainage basin.

1 J; L. H. Paulhus and J. F. Miller, Flood Frequencies Derived from Rainfall Data, J. Hydraul. Div„
ASCE,December 1957.
2 World Meteorological Organization, Manual for Estimation of Probable Maximum Precipitation,
Opera. Hydrol. Rept. li WMO No. 332, Geneva; 1973. . ^ v
 PROBABILITY CONCEPTS IN PLANNING 155

FIGURE 5.11
Probable maximum precipitation in inches for areas of 10 mi2 and duration of 6 hr. (U.S. National
Weather Service; Corps o f Engineers) *

This rainfall is used to compute the probable maximum flood flow by the methods
outlined in Chap. 3. Figure 5,11 shows the estimated probable maximum rainfall1
for a duration of 6 hr and an area of 10 mi2. Values for other areas and durations
Can be determined from the reference. Although these values may seem large*
several .storms have exceeded 80 percent of the estimated probable maximum
values in their region of occurrence. The consequences of project failure must be
very serious to justify design against the próbable maximum flood. Such a
condition exists where the failure of a dam would result in heavy loss of life
downstream.

1 Generalized Estimates of Maximum Possible Pirecipitatipn in the United; States, U.S. Weather Bur.
Hydrometeorol. Rept. 23, June 1947.
156 WATER-RESOURCES ENGINEERING

5.13 Drought
Drought is often defined in terms of a fixed period of time with less than
some minimum amount of rainfall. Even when applied to a specified arda and
crop* such a definition is far from adequate since the critical time and rainfall
depend on the stage of crop development, initial moisture content of the soil,
temperature and wind during the drought period, and other factors. If drought is
defined in terms of inadequate rainfall for crop production, most of the ^western
United States has a drought every year, since rainless summers are common in
much of the West. Since this is a normal occurrence, provision has been made to
store or divert water from streams for crops and other needs. Under such
conditions drought is defined in terms of inadequate water availability. If irrigation
is accomplished by direct diversion of water without storagé, a single winter of low
precipitation may cause a water shortage. If storage reservoirs are designed to
carry water over from one year to the next, a single year of low runoff may not
be critical. , •
In general terms a drought is a lack of water ibr some purpose: Mofe
specific definitions are possible only when local conditions are specified. The
following sections deal with methods for expressing the probability oM ów
streamflow. Methods of reservoir design are covered in Chap. 7. \

5.14 Duration Curves

The natural streamflow characteristics of a river are frequently summarized
in a flow duration curve. Such a curve (Fig. 5.12) shows the percentage
time that flow is equal to or less than various rates during the period o
study. The same data may also be plotted to show the percentage of timé
that various flows are equaled or exceeded (Fig. 5.13). A duration cutve Í1
constructed by counting the number of days, months, or yeárs with flow i
various class intervals. As the length of the time unit increases, the range
the curve decreases. The selection of the time unit depends on the purpos
of the curve. If a project for diversion without storage is under study, th
time unit should be the day so that absolute minimum flows will be indicated.
The main defect of the flow duration curve as a, design tool, is that
does not present flow in naturál sequence. It is not possible to tell whethe
the lowest flows occurred in consecutive periods or were scattered throug'
out the record. Duration curves are moist useful , for preliminary studies
for comparisons between streams. Fig. 5.13 compares the Cherry Creek d u ratr
curve with one for a stream with much more stable flow characteristics.* Cher
Creek offers no chance of successful development without provision fpr stora
to provide water during periods of low natural flow. Hat Creek coulc), howeve
provide at least 100 cfs on a continuous basis for direct diversion. Storage wouí
be required on both streams to meet a demand of 140 cfe, bat the volume requir^
for Hat Creek (ABC) is much less than for Cherry Cieek (ERD). Cherry Cre
produces considerably more runoff than Hat Creek and with proper stora''
 T

PROBABILITY CONCEPTS IN PLANNING t 1§7

FIGURE 5.12 i.. ;l;' ' ; r'. 'v /,:,

Flow duration curves for Cherry Creek near Hetch Hetchy, California (1941-1950).

facilities could provide a much higher yield. The exact storage requirements are
dependent on the actual sequence of flow (Chap. 7) and cannot be accurately
estimated from duration curves.

5.15 Drought Frequency

If drought can be defined in specific terms for a particular project, drought
frequency can be analyzed in the same manner as flood frequency. It is also
possible to prepare generalized frequency curves of low flow (Fig. 5.14). These
curves were constructed by determining the minimum flow each year during
periods of various lengths, and the data for each period length were plotted
as a frequency curve.1 Because of the wide range of flow values, a logarithmic

1 Data for this figure are from W. P. Cross and E. E. Webber, Ohio Stream-flow Characteristics,
Ohio Dep. Mat: Resources Bull: 13, Part 2, table 1, December 1950. See also J. B. Stall and J.
C. Neill, Partial Duration Series for Low-Flow Analysis, J. Geophys. Res., Vol. 66, pp. 4219-4225,
December 1961, for a description of the procedure for assembling data for figures such as Fig. 5.14.
158 WATER-RESOURCES ENGINEERING

Per cent of time indicated flows are equaled or exceeded

F IG U R E 5.13
Comparison of flow duration curves for two streams.

scále was chosen to permit more accurate plotting of the low flows. Curves!
of this type cán serve many purposes in connection with design. For example,
a small water-supply project requiring 0.9 c fs (0.025 m 3/s) might pump directly1
from the stream of Fig. 5.14. Once in 4 yr, flow would be inadequate to meet the
demand (point A in Fig. 5.14), but if storage were provided for (O.9-0.3) cfs-day
0.6 sfd = 1.2 acre-ft = 390,000 gal, a shortage would occur only once every
10 yr. This analysis assumes that the demand for water//emajns constant With
time. The 0.3 cfs in the preceding domes from point B in Fig. 5.14. ^
The degree of treatment required by wastewater is often dependent oni
the amount of water available for dilution of the treated wastewater when
it is discharged into a stream (Chap. 19). The critical flow is not necessarily
the absolute minimum, since an economic arialysis may show a substantial
saving in plant costs if moderate nuisance (without danger to health) is tolerable
at infrequent intervals. Figure 5.14 provides the data necessary for the analysis of
such a problem. ¿

546 Synthetic Streamflow ¿

It is often important to know something of the probability of floods or drougíxtsí
more severe than anything observed on a stream. The methods discussed in Sepi
5.3 provide some estimate of the probability of extreme floods. Because of thé^
 PROBABILITY CONCEPTS IN PLANNING 159

F IG U R E 5.14
.Frequency of minimum flows for Yellow Creek near Hammondsyille, Ohio (1915^1935).

difficulty of defining a drought (Sec. 5lll) and because of the few cases of
tl0ng¡-period drought in a short record, the procedures of Sec. 5.13 are really hot
Adequate for defining recurrence intervals equál to or greater than the period of
record., V ^ V vr;'v7 ; •'^ --■■ ■ - -f-
, : On the assumption that streamflow is essentially a random variable, it is
possible to develop a synthetic flow record by statistical methods. Such a record
 WATER-RESOURCES ENGINEERING

can be of any desired length and may well include flow sequences more critical
than any in the observed record. A random series m$y be generated by the
equation1
<7;+i = <Zj+ 1 + b f a - qj) + t¡<7,.+ 1(l - rj)1'2 ( 5. 11)

where qt and #i+1 are the flows in the ith and (i + l)th months from the
start of the synthetic sequence, qj and qj+ 1 are the mean monthly flows in the jth
and 0 + l)th month of the annual cycle, and bj is the regression coefficient
for estimating flows in the (y + l)th month from the flows in the yth month.
The first two terms on the right-hand side pf Eq. (5.11) .represent a Simple
regression between months as determined from the historic data. In the final
term, tt is a random normal deviate with a zero méan and unit variance,
aj+1 is the standard deyiation, of the flows for the (j + ljth month, and r3 is
the correlation coefficient between flows in the jth and ( / + l)th months. This
term of the equation imparts a random variation to the flows, but this variation
is constrained by the known characteristics óf (lid record, a and r. The procedure
assumes that the flows (or their transform) are normally distributed.
Stochastic analysis should be used to generate a number of synthetic
flow traces of length equal to the expected useful life of the project under
study. Each trace can then be analyzed to determine the required storage
for the project demand (Sec. 7.4). If demand is expected to grow over time,
the analysis of each trace should assume a variable demand. Approximately
one thousand traces should be generated and analyzed. The required storages
for all of the traces will form a probability distribution conforming to the
extreme-value theory. A design, ¡storage can be selected from this distribution
with a specified prpbábility that it will prove adequáte during the project life’2
The procedure does not allow for time trends or cycles, if any exist.
Present evidence suggests that no; significant trends can be expected within
reasonable design periods and1 that the apparent cyclic variations in weather
may be entirely random. The procedure also assumes that the statistical character­
istics of the streamflow regime are adequately defined by the observed record—an
assumption that should be reasonable if the observdd record is 50 yr or more but
may not be valid if the record is quite short. »i
Stochastic methods may be employed to generate a synthetic record of ;
rainfall, which could be transformed to streamflow by an appropriate method.
Such a procedure could be employed if a streamflow record were too short tot
provide a basis for stochastic generation or if more detail were desired in’ the

1 H. A. Thomas, Jr., and Myron Fiering, Mathematical Synthesis of Streamflow Sequences for.
the Analysis of River Basins by Simulation, chap. 12 in Arthur Maass and others, “Design o f
Water-Resource Systems,” Harvard University Press, Cambridge, Mass., 1962.-
2 Stephen J. Burges, Use of Stochastic Hydrolqgy to Determine Storage Requirements for Single* '
Purpose Storage Reservoirs—A Critical Analysis, Engineering-Economic Planning Report 34, Depart­
ment of Civil Engineering, Stanford University, September 1970.
 PROBABILITY CONCEPTS IN PLANNING 161

streamflow than could be obtained from monthly flows. Pattison1 demonstrated

the feasibility of using a Markov process to generate a sequence of rainfall data,
and Fran¿2 employed multivariate normal analysis to generate compatible hourly
data at several rainfall stations.

PR O B L EM S
5.1. Tabulated in what follows are the annual flood series for two gaging stations in the
Delaware; River basin. The station at Delhi is upstream of the station at Hale Eddy.
A large reservoir was constructed on the West Branch Delaware River between the
two stations sometime during the period of overlapping records (1937 to 1974).
Determine the year when this inhomogeneity in the Hale Eddy record began. You
may find it helpful to construct a double-mass curve for the two stations, that is, a
plot of cumulative peak flow for one station against cumulative peak flow for the
other station for the period of overlap. After eliminating the inhomogeneous portion
of the Hale Eddy record after the reservoir was constructed, prepare histograms of
the two records using the same number of class intervals for each. Are there
significant differences in the shapes of the two histograms?

Peak flow, cfs Peak flow, cfs

West Branch West Branch

West Branch Delaware River West Branch Delaware River
Water Delaware River at Hale Eddy, Water Delaware River at Hale Eddy,
year at Delhi, NY NY year at Delhi, NY NY

1913 25,000 r 1930 V.... /r 6,520

1914 ,, v ;/.. , 21,300 1931 '} y ; V; 8,340
1915 20,000 1932 f.r; I. ',v 11,100,
1916 — , J 17,400 1933 ..... 16,000
1917 ' — ’:' L . 11,800 ! 1934 '-.íL L t-L.' L I • L. 13,000
1918 1Í8,30Ó 1935 :; ■ ■'■ 19,000 (
, , - - ;
1919 5,420 1936 25,900
1920 18,100 1937 2,180 10,100
1921 11,200 1938 8,940 25,600J
1922 19,000 1939 4,520 14,300
1923 — 11,400 1940 6,430 I 23,400
1924 ■■■ • .. . 26,500 1941 1,780 8,210
1925 "V' — ' , 22,000 1942 6,090 21,900
1926 . — ■. 12,200 1943 4,870' 23,300
1927 16,600 1944 3,450 14,200
1928 ' . — • 14,000 1945 3,240 11,000
1929 17,800 1946 2,720 11,000
(c o n tin u e d )

1 Allan Pattison, Synthesis of Rainfall Data, Téchnical Report 40, Department of Civil Engineering,
Stanford University, July 1964.
1 Delberi Franz, Hourly Rainfall Synthesis for a Network of Stations, Technical Report 126, Depart­
ment of Civil Engineering, Stanford University, March 1970.
162 WATER-RESOURCES ENGINEERING

Peak flow, cfs Peak flow, c f s l'

West Branch West Branch.

West Branch Delaware River \ West Branch Delaware River
Water Delaware River at Hale Eddy, \ Water Delaware River at Hale Eddy,
year at Delhi, NY NY year at Delhi, NY NY

1947 3,400 12,800 1966 1,660 2,970

1948 4,680 28,900 1967 1,750 3,450
1949 4,250 16,000 1968 2,870 6,610
1950 4,430 i 5,500 1969 2,680 3,900
1951 6,700 19,500 1970 4,410 10,700
1952 2,910 10,500 1971 7,500
1953 3,740 16,600: 1972 4,320 6,200
1954 2,560 10,200 1973 3,970 .13,200
.1955 4,660 16,000 . 1974 6,070 . 4,420
1956 4,800 15,300 1975 .. —^ 9,740
1957 2,980 9,230 1976 — -■ 11,900
1958 , 2,830 14,100 1977 . — y 14,200
- - ...
1959 5,500 20,100 1978 8,620
;,r.. ;
1960 4,900 18,200 1979 14,800
1961 3,130 17,300 1980 — 5,620
1962 3,490 14,200 1981 6,960
1963 4,840 12,300 1982 -yy.,';';-’' " 3,900
1964 ,, , 6,330 9,000 V-
1983 *
1965, ,2,840, 2,700 1984 * 8,700

5.2. After ranking the data of Prob. 5.1, compute the Weilbull plotting positions for each |
data point and plot the data for each station on normal probability plotting paper I
(analogous to the plotting paper shown in Fig. 5.3, only constructed so that normal !<
distributions plot as straight lines). Repeat for lognormal probability paper. Both ¡j
normal and lognormal probability paper are available commercially. Sketch in your ]
best estimate of a frequency curve on all plots.
5.3. Why do you think that the Weilbull plotting position formula [Eq. (5.1)] uses N + 1 j
in the numerator rather than A? s ^ j
5.4. What is the probability of having a flood equal to or greater than the 25-yr flood j
during this next 4 yr? During the next 25 yr? |
5.5. What is the probability that a flood equal to or greater than the 10-yr flood will not j
occur in any 10-yr period? Any 25-yr period? ,
5.6. (a) What is the probability of having a flood equal to or greater than the 5-yr flood J
next year? : /Uy .y m ' ■ .:yy * "4
(b) What is the probability of having at least one flood equal to or greater than
the 5-yr flood during the next 3 yr? ij
5.7. The life expectancy of a person born in 1984 is approximately 75 yr. What is the return ,
period of the flood for which the probability of experiencing a flood of that(magnitude ]
,or larger for the average person born in 1984 is 50 percent? ’
5.8. The life expectancy of a woman born in 1984 is approximately 78 yr; while the life i
expectancy of a man born in the same year is approximately 71 yr. How much more \
likely is it that an average woman will experience a flood equal to or greater than a
50-yr flood during her lifetime than an average man will experience such a flood during
his lifetime?
 PROBABILITY CONCEPTS IN PLANNING 163

5.9. For how long must a gaging station be operated so that the probability of observing
a flood equal to or greater than a 20-yr flood is at least 90 percent?
5.10. Analysis of the annual flood series covering the period 1920 to 1989 at a gaging
station on a river shows that the 100-yr flood has a magnitude of 425,000 cfs and the
10-yr flood a magnitude of 245,000 cfs. Assuming that the flood peaks are distributed
according to the theory of extreme values, answer the following questions:
(a) What is the probability of having a flood as great as or greater than 350,000
cfs next year? *
(b) What is the magnitude of the flood having a recurrence interval of 20 yr?
(c) What is the probability of having at least one 10-yr flood in the next 8 yr?
(d) Find X, the mean of the annual floods;
(e) Find <r, the standard deviation of the annual floods.
5*11. Determiné X , <r, and b for the data of Table 511 using the years 1874 to 1911. Assuming
the flood peaks follow the extreme-value distribution, compute >the magnitude of the
10- and 100-yr floods. Repeat using data from Table 5.1 for the years 1912 to 1949.
5.12. Deteritiine log ^f; <xlogJ¿, and g fórthe data of Table 5.1 using the years 1874 to 1911.
Assuming the flood peaks follow a log Pearson Type III distribution, compute the
magnitude of the 10- and 100-yr floods. Repeat using data from Table 5.1 for the
years 1912 to 1949.
1.13. Fit the data for each station of Prob. 5.1 to the Gumbel distribution by determining
X , <t, and hence b. Coihpute the magnitude of the 50-, 100-, and 500-yr floods,
9.14. Repeat Prob. 5.13 using the log Péarson Type III distribution. How much would
your estimates of the magnitudes of the 50-, 100-, and 500-yr floods change if the
skew coefficient g were zero instead of the calculated value?
5.15. Referring to the data of Prob. 5.1, in water year 1903 a flood peak of magnitude
46,000 cfs Was observed at the Hale Eddy gage location. Estimate its return period
based on the analyses of Prob. 5.14.
5.16. Listed in what follows are all pesak flows observed on a river over a 27-yr period of
record." ■ 1 -’ " "

Date Flow, m3/s Date Flow, m3/s

June 3, 1964; 458 June 13, 1975 > 382 , ‘

May 1, 1965 184 May 29, 1976 374
Jan. 14, 1966 560 May 20, 1977 481
June 4, 1966 574 June 8, 1978 362
June 24, 1966 413 May 25, 1979 439 '
Jan. 1, 1967 416 May 31, 1979 4505
Jan. 31, 1968 580 June 5, 1979 563.
June 13, 1968 447 May 16, 1980 385,
June 18,1968 464 May 9, 1981- iio '
May 30, 1969 351 May 29, 1982 320
May 23, 1970 204 May 25, 1983 328
Jan. 25, 1971 1690 Feb. 18,1984 543 '
May 30, 1971 422 - May 17,4984 396
June 1, 1972 518 May 15, 1985 192
June 9, 1972 439 June 16,* 1,986 416
Jan. 17, 1973 V 1285 V May 28, 1987 238 , '
May 5, 1973 ii-;- 439> Dec. 28, 1988 569
June 9, 1973 .... 461 \ May 18, 1989 362 /
June 17, 1973 450 June 15, 1990 408
Feb. 22, 1974 507
1 64 WATER-RESOURCES ENGINEERING 1

(a) Arrange the “annual peaks” in order of magnitude as ip Table 5.1. Tabulate]
the values of m, tp, X —X, and X —X 2. I
(tb) Plot a frequency histogram using class intervals of 100 m3/s. |
(c) Plot an integrated histogram of annual flood peaks. :1
(d) Plot a frequency curve similar to Fig. 5.3 using values of tp as determined by>
Eq. (5.1). Draw your best estimate of the curve by eye.
(e) Compute and plot the position of the theoretical frequency curve using Gumbel’s
method. What is the magnitude of the 100-yr flood peak? i
(/) Plot a frequency curve for the partial duration series. What differences are there!
between the line defined by the partial series and that for the annual flood series? 3
5.17. Using the annual peaks of Prob. 5.16 compute the magnitude of the 10- and 100-yr
floods assuming a log PearsOn Type III idistribution.
5.18. Obtain data on annual flood peaks for a stréam-gaging station selected by you!
instructor and plot a frequency curve similar to Fig. 5.3. Draw your bést estimate ol
the curve by eye. What effect would floods that, have occurred subsequent to thé
publication of your source have on the position of the curve you have sketched? }
5.19. Compute the position of the theoretical frequency curve for the data of Prob. 5.18
using Gumbel’s method. What differences are there between the theoretical curve and
the one drawn by eye? What is the percentage difference in estimated flows at a
recurrence interval of 5 yr? At 100 yr? »- , , ^ jj
5.20. The analysis of annual flood peaks on a stream with 60 yr of record indicates thaj
the mean, value of the annual peaks is 700 m3/s. The standard deviation of the peakj
is 110, m3/s. Assume flood peaks on this stream follow an extreme-value distribution. ¡
(a) Find the probability of having a flood next year of magnitude equal to oi
, greater than 990 m3/s. , j
(b) What is the probability that at least one flood of such magnitude or greater
will occur during the next 30 yr?
(c) What is the magnitude of a flood with a recurrence interval of 15 yr? .i
5.21. The mean of the annual flood peaks on a given river is 1420 m3/s, and the standarc
deviation of these flood peaks is 270 m3/s. What is the probability of a flood equaling
or exceeding 1650 m3/s within the next 5 yr on this river? Assume that flood peaks
on this river follow a Gumbel distribution.
5.22. A frequency analysis of Dry Creek using lognormal assumptions gives the following
results: v.\ ; v f ■ ;|

Recurrence interval (yr) Peak flow (cfs)

2 30,000
50 90,000

(a) What peak flow will have a return period of 200 yr? i
(b) What is the probability that a flow of 77,000 cfs or greater will not occur ii
the next 25 yr? |
(c) A dam will be built on this creek. Flood damage will result only if the reservoii
of the proposed dam is full and a flood peak of 60*000 cfs or more occurs. The;
reservoir will be full 20 percent of the time. What is the return period of flood
damage if reservoir storage and flood-peak probabilities are independent? |
 PROBABILITY CONCEPTS IN PLANNING 165

(d) During construction of the dam a diversion tunnel is to be used. Find a design
flow for this tunnel such that there is a 20 percent probability of surcharge above
the top of the tunnel entrance during a 2-yr construction period.
5.23. The analysis of annual flood peaks on a stream with 60 yr of record indicates that
the mean value of the logarithms of the annual peaks is 5.279. The standard deviation
of the logarithms of the peaks is 0.110 and the skew coefficient is 0.6, Assume the
annual peak floods on this stream conform to the log Pearson Type III distribution.
(a) Estimate the probability of having a flood next year Of magnitude equal to
or greater than 300,000 cfs.
(b) What is the magnitude of a flood with a recurrence interval of 10 yr?
(c) What is the probability of having a flood equal to or greater than 350,000 cfs
during the next 100 yr? r
5.24. Repeat (e) in Prob. 5.16 assuming that the flood peak of January 25,1971 is an outlier
and should be discarded as unreliable. How large a difference-in the 100-yr flood
estimate results from discarding this data point?
5.25. Prepare plots analogous to Fig. 5.6 for the two stations of Prob. 5.1 using normal
probability paper. Can the flood frequency distribution for the two catchments be
represented by a single curve when scaled by the mean annual flood in this manner?
5.26. Using the results from Prok 5.25, predict the 100-yr flood for a location on the West
Branch Delaware River with a mean annual flood of 10,000 qfs.
5.27. The magnitudes of annual precipitation at a given point during the 80-yr period of
record are distributed as follows: +

Less than 6.00 in. 6

Between 6.00 and 9.99 in. 12
Between 10.00 and 13.99 in. 27
Between 14.00 and 17.99 in. 28
Over 18.00 in. 7

(а) On the basis of these data, what is the probability of háving-^^reeipitation in

excess of 18 in. in any one year?
(б) What is the probability of having three successive years in which the precipita­
tion exceeds 18^ in./yr? , , -
(c) What is the probability of having less than 12 in. of precipitation ne$t year? Give
your best estimate on the basis of the given data.
5.28. In a tropical area where the precipitation consists of short storms moré or less
uniformly distributed throughout the year, there is a 70 percent chance of rain on
any brie day; What is the probability that there will be no rain bn any two successive
days? What is the probability that it will rain on both days? That it will rain on only
one of the two days?
5.29. The following data were obtained from a recording rain gage in San Jose, California.
Shown are the maximum rainfall depths observed each year for five different durations
(in minutes). Construct the family of rainfall duration-frequiency curves for this gage.
Usé either the Gumbel or log Pearson Type III distribution, depending on which
distribution appears to fit the data more satisfactorily. Afeo present your family of
curves in the form of intensity-duration-frequency curves.;'
166 WATFR-RFSOURCES FNGINF.KRING

Year t 5 min 10 min 15 min 30 min 60 min

1908 0.06 in. 0.11 in, 0.15 in. 0.19 in. 0.31 in.
1909 0.09 0.14 0.19 0.35 0.46
1910 0.10 0.16 0.19 0.20 0.28
1911 0.18 0.19 0,22 0.34 0.65
1912 0.08 0.14 0.14 0.14 0.20
1913 0.17 0.18 0.34 0.36 , 0.39
1914 0.10 0.15 0.19 0.31 0,36
1915 0.17 0.22 0.23 0.27 0.40
1916 0.11 0,15 0.17 0.24 0.39
1917 0.09 0.15 0.18 0.29 0.45
1918 0.13 0.17 0.21 0.33 0.57
1919 0.08 0.11 0.12 0.13 0.23
1920 0.14 0.20 0.20 0.24 0.31
1921 0.13 0.16 0.19 ,0.24 0.37.
1922, 0.11 0.17 0.19 0.25 0.36
1923 0.07 0.13 0.19 0.29 0.35
1924 6.11 0.16 0.17 0.22 0.37
1925 0.13 0.19 0.24 0.26 0.31
1926 0.11 0.13 0.16 0.22 0.35
1927 0.10 0.14 0.17 0.21 0.32
1928 ! 0.15 0.22 0.27 0.30 0.34
1929 0.14 0.22 0.28 0.42 0.50
1930 0.12 0.18 0.20 0.25 0.36
1931 0.09 0.14 0.19 0.27 0.30
1932 0.14 0.21 0.26 0.33 0.36
1933 0.10 0.14 0.16 0.19 0.29
1934 0.22 0.41 0.47 0.57 0.67
1935 0.08 0.13 0.17 0.28 0.41
1936 0.09 0.15 0.19 0.25 0.35
1937 0.10 0.17 0.23 0.34 0.48
1938 0.17 0.24 0.27 0.31 0.37
1939 0.08 0.12 0.13 0.18 0.26
1940 0.15 0.18 0.23 0.33 0.50
1941 0.18 0.24 0.29 0.35 6.38
1942 0.13 0.27 0.33 0.50 0.85
1943 0.18 0.24 0.29 0.35 0.41
1944 0.14 0.18 0.20 0.30 0.47
1945 0.08 0.11 0.13 . 0.19 0.34
1946 0.05 0.10 0.15 0.15 0.25
1947 0.09 0.16 , 0.18 0.20 , , 0.31
1948; 0.10 0.16 0.21 0.26 0.31
1949 0.07 0.10 0.12 0.19 t. 0.24,
1950 0.30 0.34 0.35 0.38 * 0.40
1951 0:12 6.20 0.20 0.26 0.33
1952 0.25 0.28 0.31 0.36 0.42
1953 0.11 0.17 0.25 0.36 0.41
3954 0.16 0.24 0.38 0.46 0.47
1955 0.09 0.17 0.24 0.36 0.52 -
1956 0.13 0.21 0.27 0.40 ir 0.53
1957 0.25 0.28 0.38 0.36 0.67
1958 0.07 0.14 0.16 0.22 0.36
(c o n tin u e d )
 PROBABILITY CONCEPTS IN PLANNING 167

Year 5 min 10 min 15 min 30 min 60 min

1959 0.07 0.11 0.15 0.24 0.41

1960 , 0,07 0.13 0,15 . 0.24 0.36
1961 0.07 0.12 0.14 0.21 0.30
1962 0.09 0.16 . 0.19 0.28, , 0.50
1963 0.14 0.21 0.25 ,0.31 0.51
1964 0.14 0.24 0.33 . 0.52 0.79
1965 0.09 0.11 0.13 0.19 t 0.27 .
1966 0.08 0.11 0.13 0.Í7 0.30
1967 0.30 0.32 0.34 0.40 0.57
1968 0.35 0.67 0.88 , . 1.39 1.53 .
1969 0.20 0.23 0.24 , 0.26 ¿38
1970 0.15 0.19 0.24 0.35 , 0.66
1971 0.10 0.11 0.13 0.16 0.20
1972 0.16 0.30 0.41 0.63 0.70 ,
1973 0.14 0.25 0.33 0.35 0.40
1974 0.08 , 0.11 0.15 0.23 041
1975 0.16 0.24, 0.26 0.37, 0.44

5.30. Using the data of Prob. 5.29, what is your best estimate of the return period of the
largest 1-hr rainfall observed during the historic 68-yr period? Of the largest 5-min
rainfall observed during the historic period? What is the probability of observing a
30-min rainfall exceeding 0.40 in. at this location?
5.31. Fit an equation of the form of Eq. (5.10) to the data of Prob. 5.29 for return periods
of 2, 10, and 50 yr. Do the constants appear to be a function of return period?
5.32. Obtain data on rainfall at a station selected by your instructor. Plot intensity-
frequency curves for durátions of 5, 15, 30, 60, and 120 min. Compare your results
with Figs. 5.8 and 5.9.
5.33. Fit an equation of the form oFEq. (5.10) to the data of Prob. 5.32 for several selected
return periods.
5.34. Refer to Fig. 5.8. How often on the average might one expect a storm of intensity
2.3 in./hr or more with a duration of 30 min to occur in Cleveland, Ohio?
5.35. Make an analysis of the frequency of the average daily rainfalls at two stations about
20 mi apart. Draw the individual frequency curves for the two stations and compute
the frequency of average values on the assumptions of complete dependence and
complete independence. Does the degree of dependence between the two stations
appear to be greater at low or high rainfalls? Why?
5.36. Make an analysis of frequency of low flows for periods of 1, 7, 14, and 30 days at a
streamflow station designated by your instructor. Use a period of record of at least
20 yr for the analysis.
5.37. Construct a flow duration curve using daily streamflow values for a station designated
by your instructor. Use a period of about 3 yr for the curve, Superimpose on this
curve one constructed on the basis of monthly average flows for the same period.
Why is the range of variation of the latter curve less than that for daily data? If
members of your class used different periods for their duration curves, what differences
resulted?
5.38. Refer to Fig. 5.14. How often on the average might one expect the flow over the driest
30-day period to be about equal to or less than 120 acre-ft?
168 WATER-RESOURCES ENGINEERING

B IB L IO G R A P H Y
Beard, L. R.: “Statistical Methods in Hydrology,” U.S. Army Corps of Engineers, Sacramento District!
1962. ■ -S'-r - ! r i ‘;; r;
Benjamin, J: R., and C. A. Cornell: “ Probability, Statistics and Decision for Civil Engineers;*!
McGraw-Hill, New York, 1969.
Chow, Ven Te: Statistical and Probability Analysis of Hydrologic Data, sec. 8-1 in V. T. Chow (Ed.|
“Handbook of Applied Hydrology,” McGraw-Hill, New York, 1964.
Fiering, Myron, and Barbara Jackson: “Synthetic Streamflows,” Water Resources Monograph l |
American Geophysical Union, Washington, D.C., 1971.
“Guidelines for Determining Flood Flow Frequency,” U.S. Water Resources Council, Washington
D.C., 1976. , "V;.. -' ' ' —^ ■
'
Gumbel, E. J.: “Statistics of Extremes,” Columbia University Press, New York, 1958.
Haan, C. T.: “Statistical Methods in Hydrology,” University of Iowa Press, Iowa City, 1977.
Jarvis, C. S:, and others: Floods in the United States; Magnitude and Frequency, U.S. Géol Surveyg
Water Supply Paper 771, 1936.
Linsley, R. K., M. A. Kohler, and J. L. H. Paulhus: “ Hydrology for Engineers,” 3d ed., McGraw-Hill^
New York, 1982.
McCuen, Richard H., and Willard M. Snyder: “Hydrologic Modeling: Statistical Methods andf
Applications,” Prentice-Hall; Englewood Cliffs, New Jersey, 1986.
 CHAPTER

WATER LÁW

In regions where water supply is inadequate to meet the needs of póténtiál users,
Water is a commodity of consideráble value. A system of laws has been developed
to determine who has the right to water when shortages occur. Water rights play
an important role in determining the availability of water in scime regions. It is
not the purpose of this chapter to prepare the reader to argue a water-law cáse
in court but rather to provide an understanding of the legal problems th at may
be encountered in Water-resources engineering; Water law ¿an alsof play a major
role in the economic aspects of water development since limitations on who máy
develop water often control how it is developed and utilized.

6.1 Riparian Rights

The doctrine of riparian rights was taken from French civil law by two American
jurists, Story and Kent. During the early nineteenth century, English courts
adopted the doctrine as part of their common law. Subsequently, American
jurisdictions that adopted the English common law also accepted the riparian
doctrine. Sómé form of riparianism was brought to Mexico by the Spaniards; and
Texas courts hold that riparian doctrine has existed in Texas since Mexican days.
lender the concept of riparian rights the owner of land adjacent to a stream
(riparian land) is entitled to receive the full natural flow of the stream without
change in quality oí* quantity. The riparian owner is protected against the diversion
Of Watets upstreám from his or her property or from the diversion 6f excess
floodwaters toward the property. In other words, no upstream owner may
materially lessen or increase the natural flow of a stream to the disadvantage of
a downstream owner.

169