Environ Monit Assess (2012) 184:3153–3163

DOI 10.1007/s10661-011-2179-3

Development of geomorphologic instantaneous

unit hydrograph for a large watershed
Abdul Razzaq Ghumman &
Muhammad Masood Ahmad &
Hashim Nisar Hashmi & Mumtaz Ahmad Kamal

Received: 8 November 2010 / Accepted: 8 June 2011 / Published online: 29 June 2011
# Springer Science+Business Media B.V. 2011

Abstract Hill torrents cause a lot of environmental phology of the basin. Geomorphologic parameters
and property damage in Pakistan every year. Proper were derived from satellite images of the basin and
assessment of direct runoff in the form of hill torrents ERDAS and ArcGIS were used for data processing.
is essential for protection of environment, property, Computer program was developed to systematically
and human life. In this paper, direct surface runoff estimate the dynamic velocity, its related parameters
hydrograph (DSRH) was derived for a large catch- by optimization and thereby to simulate the DSRH.
ment using the geomorphologic instantaneous unit The data regarding rainfall–runoff and satellite images
hydrograph concept. The catchment with hill torrent were collected from Punjab Irrigation and Power
flows in semi-arid region of Pakistan was selected for Department, Pakistan. Model calibration and valida-
this study. It was divided into series of linear cascades tion was made for 15 rainfall–runoff events. Ten
and hydrologic parameters required for Nash’s con- events were used for calibration and five for valida-
ceptual model, and were estimated using geomor- tion. Model efficiency was found to be more than
90% and root mean square error to be about 5%.
M. M. Ahmad Impact of variation in model parameters (shape
COMSATS, Sahiwal Campus, parameter and storage coefficient) on DSRH was
Sahiwal, Pakistan investigated. For shape parameter, the number of
linear cascades varied from 1 to 3 and it was found
M. M. Ahmad
e-mail: mmasood567@[Link] that the shaper parameter value of 3 produced the best
DSRH. Various values of storage coefficient were
A. R. Ghumman (*) : H. N. Hashmi : M. A. Kamal used and it was observed that the value determined
Civil Engineering Department,
from geomorphology and the dynamic velocity
University of Engineering and Technology Taxila,
Taxila, Pakistan produced the best results.
e-mail: abdulrazzaq@[Link]
H. N. Hashmi Keywords GIUH . Nash model . Storage coefficient .
e-mail: drhnh@[Link] Peak discharge . The dynamic velocity
M. A. Kamal
e-mail: kamal@[Link]
Introduction
Present Address:
A. R. Ghumman
Civil Engineering Department, Al Qassim University, Sustainable water resources planning and develop-
Al Qassim, Saudi Arabia ment is the need of the day. On the one hand there is
3154 Environ Monit Assess (2012) 184:3153–3163

acute shortage of water and on the other hand floods Rodriguez-Iturbe et al. (1979) expressed the
cause a lot of environmental damage every year. dynamic velocity as a function of excess rainfall
Solution to such a problem requires evaluation of the intensity. Kumar et al. (2004) used an arbitrary value
intensity of floods for various rainfall events. The of the dynamic velocity due to absence of gauge data
rainfall–runoff process is very complicated and the for GIUH analysis. Sahoo et al. (2006) showed that if
problem of finding the flood magnitudes becomes the rainfall excess is assumed to be distributed
more challenging when the data is limited or scantily uniformly over the entire watershed both with respect
available (Ahmad et al. 2010; Bekele and Knapp to space and time, average excess rainfall intensity
2010; Bahremand and De Smedt 2010; Ahmad 2009; corresponding to the event peak discharge (Qmax) can
Bahremand and De Smedt 2008). It requires a be calculated from the ratio of peak discharge Qmax
hydrologic model which may provide the flood and basin area Ac ðQmax =Ac Þ, and can be related to the
magnitudes using limited parameters and data. For dynamic velocity. Elaborative work was done to
such situations, geomorphologic instantaneous unit correlate the dynamic velocity with maximum dis-
hydrograph (GIUH) is a very useful tool to predict charge in the present study. Rodriguez-Iturbe et al.
hydrographs at the basin outlet especially due to (1979) and Zelazinski (1986) proposed this technique
recent developments of satellite images and data for estimation of maximum velocity. A watershed can
processing tools. The conceptual models like the be categorized as small, medium, and large watershed.
linear cascade model (Nash Model) and Clark Model A small watershed has area less than 2.5 km2 and the
are the most commonly used in this family of models. time of concentration less than 2 h. A medium
These models are very effective for operational watershed has area in the range of 2.5 to 25 km2
forecasting (Ahmad 2009; Ahmad et al. 2009; and the time of concentration less than 10 h. A large
Zelazinski 1986). watershed has area greater than 25 km2 and the time
Due to the above-mentioned facts, the use of of concentration greater than 10 h. In this way, the
geomorphology GIS, remote sensing GIUH for data processing in case of large watersheds is highly
rainfall–runoff models for poorly gauged basins challenging. The dimensionless parameters, Horton's
has attracted the attention of many researchers and Ratios, explained in the coming sections, describe the
presently a lot of work is being published in this geomorphology of the catchment. These parameters
regard (Ahmad et al. 2010; Dave et al. 2010; play an important role for rainfall–runoff process. The
Coskun et al. 2010; Cao et al. 2010; Ahmad 2009; other parameters which distinguish between large,
Rai et al. 2009; Nguyen 2008; Sarangi et al. 2007). medium, and small catchments are the number of
Bhaskar et al. (1997) showed that temporal distri- streams, average length of streams, average area of
bution of rainfall excess has no marked effect on streams, and the order of streams (see Rai et al. 2009,
GIUH simulation. Ramirez (2000) has done exten- for definition of these parameters). All these factors
sive research and has elaborated various areas of have comparatively higher values in case of large
GIUH. Singh (2004) proposed a simplified method catchments.
for developing GIUH. Jain et al. (2006) introduced a
third hydrologic parameter in IUH to get smooth
recession limb of the simulated hydrograph. Jena Geomorphology of watershed
and Tiwari (2006) correlated unit hydrograph param-
eters with geomorphologic parameters through re- Horton (1945) derived relationships in the form of
gression analysis. Such approach is applicable to a geometric progressions between number of streams,
particular region. Singh et al. (2007) developed stream lengths, and corresponding drainage areas.
synthetic unit hydrograph by concept of a number These relationships are known as Horton's laws. The
of linear cascades with different storage coefficient expressions for peak discharge parameter qp and time
“k” values. Ahmad et al. (2010) have done to peak tp were derived by Rodriguez-Iturbe et al.
extensive work on GIUH; however, the dynamic (1979) as:
velocity parameter was not addressed. In GIUH, the

dynamic velocity is one of the very important V
qp ¼ 1:31RL 0:43 ð1Þ
parameters. LΩ
Environ Monit Assess (2012) 184:3153–3163 3155

  0:55 not give direct value of n, rather numerical simulation

LΩ RB
t p ¼ 0:44 RL 0:38 ð2Þ are required for this. Rosso (1984) derived the
V RA
following expressions for estimation of Nash's model
Where LΩ is length of the highest order stream in parameter “n” and “k”:
kilometers; V is velocity in meters per second, RA, RB,  0:78
and RL are Horton's Ratios representing the area ratio, RB
n ¼ 3:29 RL 0:07 ð5Þ
bifurcation ratio, and length ratio, respectively. These RA
are given as:
  w1
1   w1
1  0:48

RA ¼ Aw
;
1
RB ¼ ½Nw  Ωw ; and RL ¼ Lw RA LΩ
A1 L1 k ¼ 0:7 ð6Þ
RB RL V
Here ωis order of the stream, Ω is order of the Where RA, RB, RL are the Horton's ratios; LΩ is
basin determined from systematic ordering of streams, length of highest order stream in kilometers; V is
Aw is the average area drained by the streams of order expected peak velocity in meters per second and k is
ω, N w is number of streams of order ω, Lw is average in hours. The term ½LΩ =V  is travel time in the highest
length of streams of order ω, and L1 is average length order stream, which shows that k is effective when
of streams of first order. surface runoff enters the highest order stream. Kumar
The units of qp and tp are in per hour and in hours, et al. (2004) showed that shape of GIUH is very
respectively. The product of the qp and tp is therefore sensitive to LΩ and V, the dynamic velocity. Zelazinski
a dimensionless impulse response (IR), the hydrologic (1986) reported a similar equation for estimation of
similarity coefficient. This IR value calculated from “k” in slightly different form as:
geomorphologic parameters of the watershed is used  0:55 0:36

for determination of hydrologic parameters (n, k) of RB RL LΩ
k ¼ 1:58 ð7Þ
the linear cascade model (Nash Model). RA ðn  1Þ V
The original Nash's model (Nash 1957, 1960) is where LΩ is the length of main stream in meters. All
based on linear reservoir theory for input and output the terms in Eq. 7 are derived from geomorphology of
in a watershed. The ordinates of Nash's instantaneous the watershed except velocity. Thus, an important
unit hydrograph are given as: aspect associated with Nash's Model based on GIUH

h in1
1 t is determination of maximum velocity corresponding
exp :ð k Þ
t
Un ðtÞ ¼ ð3Þ to certain hydrologic rainfall–runoff event.
kΓ ðnÞ k
Rodriguez-Iturbe et al. (1979) related the dynamic
Where n and k are hydrologic parameters of Nash velocity parameter of GIUH to the peak velocity
Model, t is time and Γ(n)=(n−1)!=[(n−1) factorial] corresponding to the peak discharge for a particular
for integer n. Equation 3 is a two-parameter gamma rainfall–runoff event. Zelazinski (1986) proposed this
function. The first parameter “n” is the shape factor or technique of velocity determination and showed that
degrees of freedom (number of linear cascades the representative stream velocity associated with the
attenuating the IUH peak) and second parameter “k” maximum discharge at a section where main stream
is the scale factor (storage coefficient, equal for all leaves the boundary of watershed can be represented
linear cascades). The parameters n and k are related as follows:
with one another as½t=k  ¼ n  1. The right hand side
(RHS) of Eq. 3 is actually the product of qp and tp. Vmax ¼ aðQmax Þb ð8Þ
Hence, Eq. 3 can be rewritten as follows:
Where α and β are constants determined from the

  0:55 set of values of measured discharge. The discharge
ðn  1Þn 1n RB
e ¼ 0:58 RL 0:05 ð4Þ was measured at a section where the main stream
Γ ðnÞ RA
leaves the watershed boundary. A higher value of β
The RHS of Eq. 4 is IR, the product of peak flow means rapid increase in velocity corresponding to
and time to peak. It is to be determined from discharge. The Eq. 8 is essentially an equation of
geomorphologic study of the watershed. This does hydraulic geometry with exponent beta being less
3156 Environ Monit Assess (2012) 184:3153–3163

than unity. Velocity can be determined if the flows. Local farmers harness moisture from flood
parameters α and β are identified by optimization, flows for cultivation by building various temporary
using a set of flood events data. This concept is diversions. The geomorphology of the area is very
elaborated in this paper. A subroutine for optimiza- complex due to rugged terrain and sharp bends in
tion of parameters α and β was developed on the natural streams. No detailed maps are available as
basis of Downhill Simplex method (Gill et al. 1981; the area is scarcely populated and is inaccessible due
William et al. 1986). to non-availability of proper road infrastructure.
However, presence of satellite images of the area
and GIS environment processing (ERDAS) has
Study area made possible to know the geomorphologic details
of this area. The 60–65% of catchment area consists
Pakistan is a developing country and depends of barren mountains without or with minor vegeta-
heavily on its water resources. Arid/semi-arid tion. Almost all the streams are ephemeral in nature,
regions in the country receive storms for very short flowing during monsoon or winter rains only. The
duration. So it was felt necessary to apply the catchment is poorly drained with floods carrying
concept of GIUH in Pakistan, as direct surface heavy sediment load to the River Indus. The land-
runoff hydrograph based on GIUH has not been use map of the area is shown in Fig. 1b. It is a space
developed for any catchment of Pakistan. The image taken by NASA World Wind (Oosterbaan
study area is located in arid/semi-arid region of (2010)). The Indus River is clearly visible in it. On
Pakistan (Fig. 1a) and is called “Kaha Watershed”. the right-hand side of Indus, some alluvial fans and
There are a few perennial irrigation schemes called spate irrigation (Rod-Kohi) practices can be visual-
karezes originating from aquifers having negligible ized but the left-hand side related to the study area is
flows like 0.02 m3/s and are charged by hill torrent mainly barren.

Fig. 1 a Map of Pakistan showing geographical location of the study area. b Land-use map of the area (Image from space obtained
from NASA World Wind; after Oosterbaan (2010))
Environ Monit Assess (2012) 184:3153–3163 3157

The watershed's stream order was determined to be 5

as highlighted in Fig. 2. The data of rainfall–runoff for
last 15 years was obtained from Irrigation and Power
Department, Pakistan. Although the number of rain
gauges needed to fully define rainfall distribution over
the drainage area is not limited, a minimum number is
normally required to arrive at a realistic conclusion. In
this study, data from 13 non-recording type rain gauges
was used out of which three stations were installed in
Kaha Watershed. The geographic location of rain gauges
is shown in Fig. 1a. Various colors in it represent
locations of various catchments (Ahmad 2009). Kaha is
the largest among these catchments. Each catchment has
a gauging station located approximately at the center of
the catchment. The Kaha catcment has more than one
Fig. 2 Drainage pattern and stream order in Kaha Watershed gauging stations evenly distributed over the area. The
total rainfall depth was converted to excess rainfall by
Digitization of catchment map and data processing subtracting the losses using the percentage runoff
technique (Linsley et al. 1982). The rainfall values
Data processing of the watershed in this study was done corresponding to runoff events are given in Table 1. The
using ArcGIS to obtain the required geomorphologic hill torrent flows are highly dangerous and it is not
information. The watershed hydrologic area was deter- advisable to measure velocity of flow for determination
mined as 5,597.80 km2. The total length of streams was of observed flows. So the observed discharge hydro-
calculated as 1,711.19 km. Strahler's Method was graphs were obtained by the use of rating curve. The
applied to determine the scheme for stream ordering cross-section of the main stream and its bed slope was
(Linsley et al. 1982). The drainage density which is the obtained with the help of survey data and the stage
stream's length per unit area of watershed was found to values were recorded by a non-recording gauge with
be 0.3 km/km2. The length of longest stream was respect to time for the demarcations at various levels of
estimated to be as 171.06 km and the length of highest the main stream for different flood events. All this data
order stream was found to be 53.72 km. was taken from Irrigation and Power Department,

Table 1 Rainfall data of

Kaha Watershed Event number Year Total rainfall (mm) Total rainfall excess (mm) Duration (h)

01 1977 009 03.83 06.0

02 1978 110 32.24 13.0
03 1980 050 15.54 08.0
04 1981 013 04.04 06.0
05 1982 077 23.09 10.0
06 1983 110 32.99 10.0
07 1984 082 23.29 10.0
08 1985 046 14.30 08.0
09 1988 050 15.54 08.0
10 1989 055 16.49 10.0
11 1990 058 17.20 08.0
12 2003 062 17.30 08.0
13 2004 057 17.10 08.0
14 2005 055 17.00 08.0
15 2006 037 11.58 06.0
3158 Environ Monit Assess (2012) 184:3153–3163

a b 4.00
c 8
7 y = -1.6455x + 6.8032
10 3.50 y = -0.8879x + 3.4421 6 R2 = 0.9954
8 3.00 R2 = 0.986
5
2.50
4
ln(Nω)

6 2.00
y = 1.5861x + 2.1427 1.50 3
4
R2 = 0.9961 1.00 2
2 1
0.50
- - -
- 2 4 6 0 2 4 6 0 2 4 6
Ω-ω Ω -1 ω -1

Fig. 3 a Plot for Horton's ratio RB; b. Plot for Horton's ratio RL; c Plot for Horton's ratio RA
( "  #) 12
Pakistan. It is worth mentioning here that there is some 1 X
NQ
2 Qoi þ Qo
noise in the data regarding observed flow for low stage Z¼ ðQoi  Qci Þ ð10Þ
NQ 1 2Qo
values due to which all storms have nearly the same
base length of 30 h. However, the maximum discharge Where “η” is model efficiency and “Z” is peak
for various events was estimated by also the authors weighted root mean square error. NQ is number of
from the highest flood marks to validate the peak flows ordinates of the hydrograph, i is index varying from 1 to
in current data. NQ, Qoi is ith ordinate of the observed hydrograph, Qci is
ith ordinate of the computed/simulated hydrograph, and
Qo is mean of the ordinates of the observed hydrograph.
Horton's geomorphologic descriptors

The Horton's Ratios are dimensionless parameters Parameters estimation for Nash's conceptual model
describing geomorphology of the catchment. These
are obtained from the best-fit lines of graphs of Shape parameter “n” (number of linear
number of streams, average length of streams, and cascades/reservoirs)
average area of streams vs their respective order. The
values of bifurcation ratio RB, length ratio RL, and The parameter “n” is related to Horton's geomorpho-
area ratio RA for Kaha watershed were found to be logic descriptors by Eq. 4. Equation 4 does not give a
4.8847, 2.43, and 5.18, respectively, from the plot of direct value of “n”. The right-hand side of Eq. 4 was
Horton's Ratios as shown in Fig. 3a to c. estimated from the geomorphologic descriptors of Kaha
catchment. A small computer program was developed
to solve Eq. 4 by hit and trial method. The value of “n”
Model performance test was varied each time by 0.01 and the absolute error (the
difference between right-hand side and left-hand side of
The following parameters were selected to check the Eq. 4) was checked. Zero error was obtained for n=3.3.
performance of the model (Nash and Sutcliffe 1970 A few values of n and corresponding error in the
and Ahmad 2009). solution of Eq. 4 are given in Table 2. However, to
facilitate the research in future the left-hand side of
2 3 Eq. 4 for various values of n has been plotted against
P
i¼NQ
2
6 ðQoi  Qci Þ 7 “n” and given in Fig. 4. The value of “n” can be
6 i¼1 7
h ¼ 61  i¼NQ 7  100 ð9Þ obtained directly from this figure against the right-hand
4 P  2 5
side of Eq. 4 estimated from the geomorphologic
Qoi  Qo
i¼1 descriptors of any catchment. The value of “n”
Table 2 Error in solution of
Eq. 4 for various n values n 1 1.5 2 2.5 3 3.3 3.5 4 4.5 5

ABS (error) 0.58 0.34 0.34 0.12 0.04 0.0 0.03 0.09 0.15 0.2
Environ Monit Assess (2012) 184:3153–3163 3159

5 variations in hydrologic inputs and outputs. The k

parameter is inversely related to velocity by Rosso

Nash's 'n'
4
3 (1984) as given by Eq. 6. The determination of
2 velocity for single rainfall–runoff event requires its
1
relationship to peak discharge. The velocity derived
0.0 0.2 0.4 0.6 0.8 1.0 from observed rainfall–runoff event is used to
Hydrologic Similarity determine “k” in this study.
Coefficient (IR)
Fig. 4 Parameter n and hydrologic similarity coefficient IR
Results and discussion
determined as 3.3 using geomorphologic descriptors of
Kaha watershed was rounded to the nearest integer as 3 Sensitivity analysis
in this paper for further use. The value of n=3 have
significant meanings. It is highly important parameter in In order to study the effect of Nash “n” and “k” on the
the rainfall–runoff processing. As it is related to the degree of error in simulated peak discharge, model was
shape of the watershed so it reflects the impact of size run for all the 15 observed rainfall–runoff events for
of the watershed as well. It is a catch-all-type parameter the values of n and k in the range their values obtained
with respect to the shape of the watershed and its value from the geomorphologic descriptors. The results are
increases by the increase of size of the watershed. The shown in Figs. 5 and 6. For a given value of k, the
parameter “n” represents the number of linear cascades value of n was varied from 1 to 3 which is an upper
attenuating the IUH peak; hence, large catchments will limit obtained from geomorphologic descriptors. It was
have higher values of n as compared to the small observed that the value of n equal to 3 produced the
catchments. As the simulated results are sensitive to the best results as the catchment area was large. This value
values of n parameter so rounding off for this parameter produced 1:1 slope of best-fit line between observed
introduces error in the simulated runoff. Figure 4 shows and simulated maximum discharge. This is in line with
the impact of changing the values of “n”. The n value the Ponce results who describes that the value of “n” is
equal to 1 does not incorporate all the features of a large normally 3 for surface runoff simulation in large
watershed and gives high error between simulated and catchments (Ponce 1989). The value of “k” was varied
observed runoff. The large watersheds should be as 3, 3.5, 4, and 4.5 h for a given value of n. The 4-
modeled by taking n=3. h value of “k” produced minimum error between
observed and computed maximum discharge. This
Scale parameter “k”—storage coefficient value of k corresponds to velocity of 3.7 m/s in the
highest order stream. The velocity in this range is
As explained earlier, the parameter “k” is related to normally observed during flood events. The runoff
velocity which is a dynamic characteristic of the diffusion phenomenon is dominant as compared to
watershed. It is a catch-all parameter and accounts for translation flow effects when evaluating hydrologic

a b y = 1.189x + 92.028
c
y = 1.9584x + 782 .89
2 y = 0.9907 x - 74. 763
Simulated Discharge (m /s)

2 R = 0.8024
10,000
Simulated Discharge (m /s)

R = 0.6102 2
Simulated Maximum

4,000
3

4,000 R = 0.8588
3
Discharge (m /s)

8,000
3

3,000 3,000
6,000
2,000 2,000
4,000
2,000 1,000 1,000

- - -
0 1000 2000 3000 4000 0 1000 2000 3000 4000 - 2,000 4,000
3 3 3
Observed Maximum Discharge (m /s) Observed Discharge (m /s) Observed Discharge (m /s)

Fig. 5 a Observed vs simulated discharge for n=1. b Observed vs simulated discharge for n=2. c Observed vs simulated discharge for
n=3
3160 Environ Monit Assess (2012) 184:3153–3163

Fig. 6 a Simulated and

observed peak discharges a y = 1.2141x + 32.95 b y = 1.094x - 37.413
2

Computed Discharge (m /s)

for k=3. b Simulated and 2 R = 0.8437

Computed Discharge (m /s)

R = 0.8193 4,000
5,000

3
3
observed peak discharges 3,500
for k=3.5. c Simulated and 4,000 3,000
observed peak discharges 2,500
3,000
for k=4. d Simulated and 2,000
observed peak discharges 2,000 1,500
for k=4.5 1,000
1,000
500
- -
0 2000 4000 0 1000 2000 3000 4000
Observed Discharge (m3/s) Observed Discharge (m3/s)

c y = 0.9906x - 74.443
d y = 0.9023x - 93.658

Computed Discharge (m /s)

2

Computed Discharge (m /s)

2
R = 0.8587 3,500 R = 0.8685

3
3 3,500
3,000
3,000
2,500
2,500
2,000 2,000
1,500 1,500
1,000 1,000
500 500
- -
- 2,000 4,000 0 1000 2000 3000 4000
3
Observed Discharge (m3/s) Observed Peak Discharge (m /s)

response of catchments of large size. Hence, the direct optimized values of α, β, V, and k. The velocity V in
runoff hydrograph shape is more sensitive to n value this table is simulated velocity and represents the
than that of k in large catchments. optimized value of velocity. It is observed that the
events 1 and 4 show comparatively larger values of
Calibration of the model velocity. It is due to the fact that small discharge values
are result of pocket flash floods in the vicinity of outlet
After having an idea about the sensitivity of parameters, and the velocity under discussion represents the peak
the optimization process was made to have global flows in the highest order stream. Hence, it should not
parameters of the catchment. Table 3 shows the be intermixed with the normal stream velocity appli-
cable to the entire watershed.
Table 3 Parameters of Nash-GIUH obtained by the calibration
of the model Validation of the model
Event no. Qo (observed) α β V k
(m³/s) (m/s) (h)
Using the values of “n” and “k” identified by
optimization of the velocity parameters, five rainfall
1 866 0.5820 0.4750 5.55 1.5 events were simulated to confirm the applicability of
2 3,356 0.1734 0.3021 2.60 3.2 the derived parameters of the Nash Model. The
3 1,671 0.1828 0.2875 2.17 3.9 observed and simulated hydrographs of direct surface
4 913 0.6600 0.4924 6.40 1.3 runoff for five events 11 to 15 are shown in Fig. 7.
5 2,364 0.1625 0.2924 2.16 3.9 Table 4 shows the results obtained for the validation
6 3,109 0.1321 0.2680 1.93 4.3 events 11 to 15 in terms of error estimators.
7 2,478 0.1689 0.2976 2.29 3.7 It is observed from Table 4 that Nash's model based
8 1,570 0.1897 0.3112 2.20 3.8 on geomorphologic details of watershed gives fairly
9 1,557 0.1742 0.2995 2.02 4.2 accurate results. The model efficiency varies from
10 2,265 0.2115 0.3263 2.78 3.0 87% to 95%, which is acceptable as unit hydrograph
Optimized 0.2233 0.3280 4.0
technique itself have some inherent approximations
(Linsley et al. 1982). Hydrologists, while reporting
Environ Monit Assess (2012) 184:3153–3163 3161

Fig. 7 a Hydrographs for Observed Hydrograph Observed Hydrograph

event no. 11. b Hydrographs a b Simulated Hydrograph
Simulated Hydrograph
for event no. 12. c 2,000 2,000
2.15 mm/h 2.16 mm/h
Hydrographs for event

Discharge (m³/s)

Discharge (m³/s)
no. 13. d Hydrographs for 1,500 1,500
event no. 14. e Hydrographs
for event no. 15 1,000 1,000

500 500

- -
0 10 20 30 0 10 20 30
Time (Hours) Time (Hours)

c Observed Hydrograph
Simulated Hydrograph
d Observed Hydrograph
Simulated Hydrograph
2,000 2.14 mm/h 2,000 2.12 mm/h

Discharge (m³/s)

Discharge (m³/s)
1,500 1,500
1,000 1,000
500 500
- -
0 10 20 30 0 10 20 30
Time (Hours) Time (Hours)

Observed Hydrograph
e Simulated Hydrograph
1,400 1.9 mm/h
Discharge (m³/s)

1,200
1,000
800
600
400
200
-
0 10 20 30
Time (Hours)

flood hydrograph peak, increase it by 5–20% before generated by the rainfall excess estimation or by
using unit hydrograph for the estimation of extreme missing input data.
floods (Linsley et al. 1982). The objective function Z As shown in Fig. 7, the peak discharge, time to
has values less than 5%. The percent error in runoff peak, and time base of the observed and computed
volume is around 10% except for the last event hydrographs match well. However, the rising limb of
(number 15) which itself is of low potential. This hydrograph is under estimated. This error is thought
error may be due to bias in data. The hydrograph peak to be due to estimation of parameter “n”. If n is
was observed to be slightly over estimated which is decreased, the time to peak will decrease with
expected to be due to rounding off the parameter “n” associated increase in peak discharge due to quick
on lower side. It was rounded-off from 3.3 to 3. response from the watershed. It was estimated with
Zelazinski (1986) reports that this deviation is the help of Fig. 3 that the length of maximum order

Table 4 Statistical
parameters obtained from Event no. η (percent; %) Percent error in Percent error in Percent error in Z
validation test of Nash's runoff volume peak discharge time to peak
Model
11 92 12 13 0 138
12 95 1 2 8 93
13 88 9 17 14 175
14 87 18 19 7 182
15 95 5 5 8 76
3162 Environ Monit Assess (2012) 184:3153–3163

stream is about 30% of the maximum hydraulic path. the watershed. It represents the number of linear
Therefore, runoff will be accumulated abruptly when cascades attenuating the IUH peak. Rounding off for
entering the stream of highest order, but before this the parameter n introduces error in the simulated
time, the response of the watershed will start slowing runoff. The n value equal to 1 does not incorporate all
down. the features of a large watershed. The best value of
Nash parameter n obtained from geomorphologic
descriptors may vary from catchment to catchment
Conclusion and is equal to 3 for a large catchment like Kaha. The
parameter k should be determined from velocity–
Nash's model of linear cascades based on Horton's discharge relationship that produces best results for
geomorphologic descriptors has been developed for a the observed peak discharge events. It is concluded
large watershed of area 5,598 km2 in a semi-arid that the value of k equal to four produces the best
region with complex geomorphology and hill torrent results. The runoff diffusion phenomenon is dominant
flows in Pakistan. The dimensionless parameters as compared to translation flow effects when evalu-
(Horton's Ratios) describing geomorphology of the ating hydrologic response of catchments of large size.
watershed play an important role for rainfall–runoff Hence, the direct runoff hydrograph shape is more
process in large watersheds These are dependent upon sensitive to n value than that of k in large catchments.
number of streams, average length of streams, average
area of streams, and their respective order. All these
parameters have comparatively higher values in case
References
of large catchments. The total length of streams was
calculated as 1,711.19 km. The drainage density
Ahmad, M. M. (2009). Investigation and application of new
which is the stream's length per unit area of watershed
techniques in determination of runoff for various
was found to be 0.3 km/km2. The length of longest conditions of humid and arid catchments. PhD thesis,
stream was estimated to be as 171.06 km and the University of Engineering and Technology, Taxila,
length of highest order stream was found to be Pakistan.
Ahmad, M. M., Ghumman, A. R., & Ahmad, S. (2009).
53.72 km. The watershed's stream order was deter-
Estimation of Clark's instantaneous unit hydrograph
mined to be 5. The values of bifurcation ratio RB, parameters and development of direct surface runoff
length ratio RL, and area ratio RA for Kaha watershed hydrograph. Journal of Water Resources Management,
were found to be 4.8847, 2.43, and 5.18, respectively. 23, 2417–2435.
Ahmad, M. M., Ghumman, A. R., Ahmad, S., & Hashmi, H. N.
A computer program has been developed for Nash
(2010). Estimation of a unique pair of Nash model
Model and for estimation of expected peak flow parameters: An optimization approach. Journal of Water
velocity. A subroutine based on Downhill Simplex Resources Management, 24(12), 2971–2989.
method for optimization of parameters of velocity has Bahremand, A., & De Smedt, F. (2008). Distributed
hydrological modeling and sensitivity analysis in Torysa
been developed. Sparse watershed hydrologic data Watershed, Slovakia. Water Resources Management, 22
has been used to develop Nash's-GIUH model which (3), 393–408.
is easy to use and requires minimum updating of the Bahremand, A., & De Smedt, F. (2010). Predictive analysis and
hydrologic parameters. The conceptual model gives simulation uncertainty of a distributed hydrological model.
Water Resources Management, 24(12), 2869–2880.
results of acceptable accuracy. It is established that Bekele, E. G., & Knapp, H. V. (2010). Watershed modeling to
hydrologic response of the watershed is closely assessing impacts of potential climate change on water
related to geomorphology. The GIUH theory can be supply availability. Water Resources Management, 24(13),
applied to large watersheds in arid to semi-arid 3299–3320.
Bhaskar, N. R., Parida, B. P., & Nayak, A. T. (1997). Flood
regions. The dynamic velocity parameter can be estimation for ungauged catchments using the GIUH.
related to peak flow of rainfall event. Two parameters Journal of Hydrology for Engineering ASCE, 123(4),
need to be optimized for this purpose. The parameter 228–238.
n is an important parameter. It is related to the shape Cao, S., Lee, K. T., Ho, J., Liu, X., Huang, E., & Yang, K.
(2010). Analysis of runoff in ungauged mountain
and size of the watershed. It is a catch-all-type watersheds in Sichuan, China using kinematic-wave-
parameter with respect to the shape of the watershed based GIUH model. Journal of Mountain Science, 7(2),
and its importance increases by the increase of size of 157–166.
Environ Monit Assess (2012) 184:3153–3163 3163

Coskun, H. G., Alganci, U., Eris, E., Agıralioglu, N., Oosterbaan R. (2010). Spate Irrigation: Water harvesting and
Cigizoglu, H. K., Yilmaz, L., et al. (2010). Remote agricultural land development options in the NWFR of
sensing and GIS innovation with hydrologic modelling Pakistan, Proceeding of International Policy Workshop
for hydroelectric power plant (HPP) in poorly gauged “Water Management and Land Rehabilitation, NW Frontier
basins. Journal of Water Resources Management, 24(14), Region, Pakistan”, Islamabad, December 6–8, 2010
3757–3772. Ponce, V. M. (1989). Engineering hydrology principles and
Deckers, D. L. E. H., Booij, M. J., Rientjes, T. H. M., & Krol, practices. NJ: Prin. Hall Inc.
M. S. (2010). Catchment variability and parameter Rai, R. K., Upadhyay, A., Sarkar, S., Upadhyay, A. M., &
estimation in multi-objective regionalisation of a rainfall– Singh, V. P. (2009). GIUH based transfer function for
runoff model. Journal of Water Resources Management, Gomti River Basin of India. Journal of Spatial Hydrology,
24(14), 3961–3985. 9(2), 24–50. Fall 2009.
Gill, P. E., Murray, W., & Wright, M. H. (1981). Practical Ramirez, J. A. (2000). Prediction and modelling of flood
optimization. London: Academic. hydrology and hydraulics, Chapter 11 of inland flood
Horton, R. E. (1945). Erosional development of streams and hazards: Human, riparian and aquatic communities Eds.
their drainage basins; hydrophysical approach to quanti- Ellen Wohl; Cambridge University Press, 2000.
tayive morphology. Bulletin Geological Society of Amer- Rodriguez-Iturbe, I., Devoto, G., & Valdes, J. B. (1979).
ica, 56, 275–370. Discharge response analysis and hydrologic similarity:
Jain, K. S., Singh, V. P., & Bhunya, P. K. (2006). The interrelation between the geomorphologic IUH and
Development of optimal and physically realizable unit the storm characteristics. Water Resources Research, 15(6),
hydrograph. Journal of Hydrology for Engineering 1435–1444.
ASCE, 11(6), 612–616. Rosso, R. (1984). Nash model relation to Horton order ratios.
Jena, S. K., & Tiwari, K. N. (2006). Modelling synthetic unit Water Resources Research, 20(7), 914–920.
hydrograph parameter with geomorphologic parameters of Sahoo, B., Chandranath, C., Narendra, S. R., Singh, R., & And,
watershed. Journal of Hydrology, 319, 1–14. K. R. (2006). Flood estimation by GIUH-based Clark and
Kumar, R., Chaterrjee, C., Singh, R. D., Lohani, A. K., & Nash models. Journal of Hydrology for Enginering ASCE,
Kumar, S. (2004). GIUH based Clark and Nash models for 11(6), 515–525.
runoff estimation for an ungauged basin and their Sarangi, A., Madramootoo, C. A., Enright, P., & Prasher, S.
uncertainty analysis. International Journal of River Man- O. (2007). Evaluation of three unit hydrograph models
agement, 2(4), 281–290. to predict the surface runoff from a Canadian water-
Linsley, R. K., Kohler, M. A., & Paulhus, J. L. H. (1982). shed. Journal of Water Resources Management, 21(7),
Hydrology for engineers (3rd ed.). New York: McGraw- 1127–1143.
Hill Book Co., Inc. Singh, S. K. (2004). Simplified use of gamma-distribution/Nash
Nash, J. E. (1957). The form of instantaneous unit hydrograph. model for runoff modeling. Journal of Hydrology for
International Association of Scienific Hydrologists. Publi- Engineering ASCE, 9(3), 240–243.
cation No. 51, 546–557, IAHS, Gentbrugge, Belgium. Singh, P. K., Bhunya, P. K., Mishra, S. K., & Chaube, U. C.
Nash, J. E. (1960). A unit hydrograph study with particular (2007). An extended hybrid model for synthetic unit
reference to British catchments. Proceeding of Interna- hydrograph derivation. Journal of Hydrology, 336, 347–
tional Civil Engineering, 17, 249–282. 360.
Nash, J. E., & Sutcliffe, J. V. (1970). River flow forecasting William H. P., Brain P. F., Saul A. T., & Wiliam T. V., (1986).
through conceptual models, Part-I: A discussion of princi- Minimization or maximization of functions. Numerical
ples. Journal for Hydrology, 10(3), 282–290. Proc.-Intl. Recipes, The Art of Scientific Computing. p. 274–334.
Civ. Eng., 17, 249–282. Zelazinski J. (1986). Application of the geomorphological
Nguyen, H. Q. (2008). Utilization of SRTM data for flood instantaneous unit hydrograph theory to development of
protection based on GIUH approach. Review Geographica forecasting models in Poland. Hydrol. Sci. Journal-des
Acadêmica, 2(2), 14–25. viii.2008. Sciences Hydrogiques, 31, 2, 6/1986.
Copyright of Environmental Monitoring & Assessment is the property of Springer Science & Business Media
B.V. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright
holder's express written permission. However, users may print, download, or email articles for individual use.