You are on page 1of 12

HYDROLOGICAL PROCESSES

Hydrol. Process. 21, 1829 1840 (2007)


Published online 18 May 2007 in Wiley InterScience
(www.interscience.wiley.com) DOI: 10.1002/hyp.6318

Runoff estimation for an ungauged catchment using


geomorphological instantaneous unit hydrograph (GIUH)
models
Rakesh Kumar,1 C. Chatterjee,2 * R. D. Singh,1 A. K. Lohani1 and Sanjay Kumar1
1 National Institute of Hydrology, Jalvigyan Bhavan, Roorkee-247667, Uttaranchal, India
2 Agricultural and Food Engineering Department, Indian Institute of Technology Kharagpur, Kharagpur-721302, West Bengal, India

Abstract:
A geomorphological instantaneous unit hydrograph (GIUH) is derived from the geomorphological characteristics of a catchment
and it is related to the parameters of the Clark instantaneous unit hydrograph (IUH) model as well as the Nash IUH model
for deriving its complete shape. The developed GIUH based Clark and Nash models are applied for simulation of the direct
surface run-off (DSRO) hydrographs for ten rainfall-runoff events of the Ajay catchment up to the Sarath gauging site of
eastern India. The geomorphological characteristics of the Ajay catchment are evaluated using the GIS package, Integrated
Land and Water Information System (ILWIS). The performances of the GIUH based Clark and Nash models in simulating
the DSRO hydrographs are compared with the Clark IUH model option of HEC-1 package and the Nash IUH model, using
some commonly used objective functions. The DSRO hydrographs are computed with reasonable accuracy by the GIUH
based Clark and Nash models, which simulate the DSRO hydrographs of the catchment considering it to be ungauged. Inter
comparison of the performances of the GIUH based Clark and Nash models shows that the DSRO hydrographs are estimated
with comparable accuracy by both the models. Copyright 2007 John Wiley & Sons, Ltd.
KEY WORDS runoff estimation; geomorphological instantaneous unit hydrograph; geographic information system; ungauged
catchment
Received 28 July 2003; Accepted 24 February 2006

INTRODUCTION land-use and climate patterns, the model parameters are


required to be updated from time to time.
For planning, development and operation of various water
The rainfall pattern, in general is undergoing a change
resources schemes estimation of run-off response from
due to global changes in atmospheric conditions. Fur-
ungauged catchments is an important subject of research
in the sphere of surface water hydrology. Conventional ther, because of different activities in the catchment, its
techniques of unit hydrograph (UH) derivation require land-use is also undergoing a gradual change and this
historical rainfall-runoff data. Owing to obvious rea- has an impact on the run-off characteristics of the catch-
sons, adequate run-off data are generally not available ment. Thus, the coupling of the hydrologic characteristics
for many small and medium size catchments, particu- of the catchment with the geomorphologic parameters
larly in developing countries like India. Hence, indi- can provide an insight to the hydrologic behaviour of
rect inferences through regionalization are sought for the ungauged catchments in particular. As a first step
such ungauged catchments. In the process of region- in the direction of coupling the hydrologic characteris-
alization, the parameters of UH or instantaneous unit tics of a catchment with the geomorphologic parameters,
hydrograph (IUH) models are related with physiographic the concept of the geomorphologic instantaneous unit
and climatologic characteristics for gauged catchments hydrograph (GIUH) was introduced by Rodriguez-Iturbe
in hydro-meteorologically homogeneous regions. These and Valdes (1979). The GIUH approach is applicable for
relationships are then used for run-off estimation for ungauged or scantily gauged catchments wherein rainfall
the ungauged catchments of the hydro-meteorologically data are available but runoff data are not. The GIUH
homogeneous regions. This process of regionalization is approach is more advantageous than the conventional
a difficult task since it not only requires a good amount IUH methods such as the Clark IUH model (Clark, 1945)
of rainfall-runoff data for the gauged catchments, but and the Nash IUH model (Nash, 1957) since it avoids
the hydro-meteorological homogeneity of the region is the requirement of stream flow data. Also, the param-
also difficult to ascertain. Further, with the change of the eters of the Clark and Nash IUH models require to be
updated from time to time because of changing land-use
and climatic conditions. Further, the GIUH approach is
* Correspondence to: C. Chatterjee, Humboldt Research Fellow, Potsdam more advantageous than the regionalization techniques
University, Institute for Geo-Ecology (Hydrology/Climatology), 14415
Potsdam, Germany. as it does not require any information about the other
E-mail: cchat iit@yahoo.com; cchatnih@yahoo.com catchments in the hydro-meteorologically homogeneous

Copyright 2007 John Wiley & Sons, Ltd.


1830 R. KUMAR ET AL.

region. It also does not require stream flow data for the expressions were given as:
catchment for which it is to be applied. Another advan-
tage of the GIUH approach is its potential for deriving the qp D 131RL043 V/L 1
UH using the geomorphological characteristics obtain- tp D 044 L /VRB /RA  055 038
RL  2
able from topographic maps or remote sensing, possibly
linked with geographic information system (GIS) and dig- where qp is peak flow in units of inverse hour, tp is time
ital elevation model (DEM). However, the GIUH models to peak in hour, RB , RL , RA are bifurcation, length and
are applicable only for the direct surface runoff (DSRO) area ratios given by the Hortons laws of stream numbers,
component of the stream flow and hence, can be used to lengths and areas respectively, L is length in km of
generate only the DSRO hydrographs. In case of flood the highest order stream, and V is a dynamic velocity
events, the base flow contribution to the stream flow parameter in m s1 . On multiplying Equations (1) and
is very small as compared to the DSRO. For ungauged (2) a non-dimensional term qp tp as mentioned below
catchments the regional values of base flow per km2 of is obtained.
the catchment area are usually available (Central Water
Commission, 1983). These regional values of base flow qp tp D 05764 RB /RA 055 RL 005 3
may be added to the DSRO hydrographs computed by
the GIUH approach for estimating the stream flow hydro- The term qp tp is not dependent on the velocity and
graphs. thereby on the storm characteristics and hence, it is a
The objectives of the present study are (i) to derive function of only the geomorphological characteristics of
GIUH from the geomorphological characteristics of a the catchment. The dynamic velocity parameter in the
catchment by relating the GIUH to the parameters of the formulation of Geomorphological IUH (GIUH) incorpo-
Clark IUH model (Clark, 1945) and the Nash IUH model rates the effect of climatic variation. Rodriguez-Iturbe
(Nash, 1957) for deriving its complete shape, (ii) to eval- et al. (1979) showed that the dynamic velocity param-
uate the geomorphological characteristics of the catch- eter of the GIUH can be taken as the velocity at the
ment required for derivation of the GIUH by employing peak discharge time for a given rainfall-runoff event in
GIS, (iii) to compare the observed and simulated DSRO a catchment. Valdes et al. (1979) compared the GIUHs
hydrographs obtained using the GIUH based Clark and for some real world basins with the IUHs derived from
Nash models, the Clark IUH model option of the HEC-1 the discharge hydrograph produced by a physically based
rainfall-runoff model of the same basins and found them
package (HEC-1, 1990) and the Nash IUH model using
to be remarkably similar.
some commonly adopted objective functions, and (iv) to
Gupta et al. (1980) examined the above approach, and
compare the GIUH based Clark and Nash models for
reformulated, simplified and generalized it. In the deriva-
estimating DSRO hydrographs by convoluting the excess
tion of the GIUH, one of the greatest difficulties involved
rainfall with the derived UHs of unit volume and specific
is the estimation of peak velocity. This is a parameter that
duration.
must be evaluated for each flood event. Rodriguez-Iturbe
et al. (1982a,b) rationalized that velocity must be a func-
tion of the effective rainfall intensity and its duration. It
BACKGROUND OF GIUH BASED APPROACH leads to the development of geomorphoclimatic IUH. In
a further development, the Nash model scale and shape
The GIUH theory was introduced by Rodriguez-Iturbe parameters were related to Horton order ratios on the
and Valdes (1979) by relating IUH peak and time to basis of the GIUH model of catchment response (Rosso,
peak with geomorphologic characteristics of the catch- 1984); a hill slope velocity term was incorporated into the
ment and a dynamic velocity parameter. This pioneer-  GIUH (van der Tak and Bras, 1990); and the GIUH was
ing work of Rodriguez-Iturbe and Valdes (1979), which reformulated on a deterministic basis in contrast to the
explicitly integrates the geomorphology details and the Markov and the statistical mechanics approaches (Chutha
climatological characteristics of a basin, in the frame- and Dooge, 1990). Zelazinski (1986) gave a trial and
work of travel time distribution, is a boon for stream error procedure for estimating the flow velocity, which
flow synthesis in the basins having no or scanty infor- involves the development of the relationship between the
mation on flow data. This original formulation of GIUH velocity and corresponding peak discharge. Troutman and
is based upon the probability density function (pdf) of Karlinger (1984, 1985, 1986) and Karlinger and Trout-
the time history of a randomly chosen drop of effec- man (1985) endeavoured the topologic IUH, which is
tive rainfall arrived to the trapping state of a hypothetical based on the pattern of the finite number of topologic
basin, treated as a continuous Markovian process, where features, using linear schemes including translation, diffu-
the state is the order of the stream in which the drop is sion and the general linear routing. Rinaldo et al. (1991)
located at any time. The value at the mode of this pdf developed the hydraulic response of a channel network
produces the main characteristics of GIUH. Equations in the framework of inverse Gaussian function of channel
were derived by Rodriguez-Iturbe and Valdes (1979) to holding time, using the conceptual structure as used by
express peak and time to peak characteristics of the IUH Rodriguez-Iturbe et al. (1979), and assuming that the lin-
as functions of Hortons order ratios (Horton, 1945). The ear advectiondispersion relationship define the transport

Copyright 2007 John Wiley & Sons, Ltd. Hydrol. Process. 21, 1829 1840 (2007)
DOI: 10.1002/hyp
RUNOFF ESTIMATION USING GIUH MODELS 1831

of water through the individual channels. Snell and Siva- intensity is defined as the excess-rainfall depth per unit
palan (1994) used a contributing area-flow distance func- time.
tion extracted directly from DEM, without the assump- There are two different approaches for development
tions of Strahler stream ordering to derive the GIUH. of relationship between equilibrium velocity and excess-
They found that the geomorphological dispersion coef- rainfall intensity and these are described in detail by
ficient derived from the area-distance function expresses Kumar et al., (2002). A brief description of the Clark and
the natural dispersion within the catchment in a more fun- Nash IUH models as well as the procedure developed for
damental manner than the other methods. Corradini et al. derivation of UH using the GIUH based approaches is
(1995) established the variation of the GIUH with basin given below.
order reduction. Rinaldo and Rodriguez-Iturbe (1996),
and Rodriguez-Iturbe and Rinaldo (1997) expressed the Clark IUH model
pdf of travel times as a function of the basin forms.
The Clark IUH model is based on the concept that
Bhaskar et al. (1997) derived the GIUH from the water-
IUH can be derived by routing the unit excess rainfall in
shed geomorphological characteristics and then related
the form of a timearea diagram through a single linear
the parameters of the Nash IUH model for deriving its
reservoir. For the derivation of the IUH, the Clark model
complete shape. Al-Wagdany and Rao (1998) correlated
uses two parameters viz. time of concentration (Tc ) in
the velocity parameter of the exponential distribution
hour and storage coefficient (R) in hour of a single linear
GIUH, the linear routing GIUH and the geomorpho-
reservoir, in addition to the timearea diagram. The
logical dispersion GIUH. Lee (1998) linked the GIUH
governing equation of the Clark IUH model is expressed
with GIS to simulate the design hydrographs. Karvonen
as:
et al. (1999) used the GIUH model for different land-use
ui D CIi C 1  Cui1 4
areas and calibrated the GIUH routing parameters using
an optimization technique. Jain et al. (2000) derived a where, ui is ith ordinate of the IUH, C and (1  C) are
GIS supported GIUH for flood estimation in an ungauged the routing coefficients, and C is t/R C 05 t, t
basin and found that the peak characteristics of the design is computational interval in hour, Ii is ith ordinate of the
flood are more sensitive to the storm pattern. However, timearea diagram. A UH of desired duration (D) may
there were some limitations in their study due to the lack be derived using the following equation.
of data availability for velocity estimation and valida-
tion. Further, Kumar et al. (2002) studied the sensitivity 1
Ui D 05 uiN C uiNC1 C C ui1 C 05 ui  5
analysis of the GIUH based Clark model. N
Summarizing the above, it is observed that in the initial
GIUH approach a triangular IUH is assumed, notifying where, Ui is ith ordinate of the UH of D-hour duration
only the peak and time to peak of the GIUH. Later, and computational interval t hour, and N is number of
though there have been isolated efforts to couple the computational intervals in D-hour and is equal to D/t.
GIUH model with the Clark IUH and Nash IUH models
to derive the complete shape of the IUH, however, there Nash IUH model
has been no concerted effort to develop these models The Nash model (Nash, 1957) is based on the concept
as well as evaluate and compare their performances. that IUH can be derived by routing the instantaneous
Thus, this study is carried out to derive the GIUH from inflow through a cascade of linear reservoirs with equal
the geomorphological characteristics of a catchment by storage coefficient. The outflow from the first reservoir
relating the GIUH to the parameters of the Clark and is considered as inflow to the second reservoir, and so
Nash IUH models for deriving its complete shape; and to on. For derivation of IUH, the Nash model uses two
evaluate their performances and compare them with the parameters viz. number of linear reservoirs (n), which
Clark IUH and the Nash IUH models. is dimensionless and storage coefficient (k) in hour. The
governing equation of the Nash IUH model is expressed
DEVELOPMENT OF GIUH BASED CLARK AND as:
1
NASH MODELS u(t) D t/kn1 et/k 6
kn
Development of the GIUH based Clark and Nash models
involves derivation of relationship between equilibrium where, u (t) denotes IUH ordinates in hour1 , t is
velocity and excess-rainfall intensity as well as a proce- sampling time interval in hour, n and k are parameters of
dure for derivation of the UH using the GIUH based Clark Nash IUH model as described above and is  function.
and Nash models, respectively. Equilibrium velocity is A UH of desired duration (D) may be derived using
defined as the velocity corresponding to the flow resulting the following equation:
due to specific excess-rainfall intensity over the catch- 1
ment for infinite duration. It occurs when the duration of U(D,t) D I(n,t/k)  In, t  D/k 7
the excess rainfall exceeds the time of concentration of D
the catchment, and the entire catchment begins to con- where, U D, t denotes ordinates of UH of D-hour
tribute to the flow at the gauging site. Excess-rainfall duration in hour1 , t is the sampling time interval in

Copyright 2007 John Wiley & Sons, Ltd. Hydrol. Process. 21, 1829 1840 (2007)
DOI: 10.1002/hyp
1832 R. KUMAR ET AL.

hour, I n, t k 1  is the incomplete  function of order 10. Compute the next trial value of R using the following
n at (t k 1 ) and D is the duration of UH in hour. governing equations of NewtonRaphsons method:
FCN1
Derivation of UH using the GIUH based Clark model R D 12
FPN
The steps involved in derivation of a UH of a specific and
duration using the GIUH based Clark model approach are RNEW D R1 C R 13
as follows:
11. For the next trial, consider R1 D R2 and R2 D RNEW
1. Evaluate the geomorphological parameters like RA , and repeat steps (6) to (10) till one of the following
RB and RL for consecutive orders and the overall criteria of convergence is achieved.
values of these parameters for the catchment as a
whole. (a) FCN2 D 0000001
2. Compute the excess-rainfall hyetograph either by (b) No. of trials exceeds 200
uniform loss rate procedure or by soil conservation (c) ABS R/R1 D 0001
service (SCS) curve number method (USDA-SCS,
1972; Chatterjee et al., 2001) or by any other suitable 12. The final value of storage coefficient (R2 ) obtained
method. as above is the required value of the parameter R
3. For a given storm, estimate the peak velocity V for corresponding to the value of time of concentration
the highest rainfall excess by using the relationship (Tc ) for the Clark model.
between velocity and intensity of rainfall excess 13. Compute the IUH using the GIUH based Clark model
(Kumar et al., 2002). with the help of final values of storage coefficient (R)
4. Compute the time of concentration (Tc ) in hour using as computed in step (12), time of concentration (Tc )
the equation: and the timearea diagram.
14. Compute the D-hour UH using the relationship
Tc D 02778 L/V 8 between IUH and UH of D-hour as given by Equation
(5).
where, L is length of the main channel in km, and 15. For estimation of the DSRO hydrograph convolute,
V is peak velocity in m s1 . Considering this Tc as the excess-rainfall hyetograph with UH obtained in
the largest time of travel compute the ordinates of Step (14).
cumulative isochronal areas corresponding to integral
multiples of computational time interval with the Derivation of UH using the GIUH based Nash model
help of non-dimensional relation between cumulative
The complete shape of the GIUH can be obtained
isochronal area and the percent time of travel. This
by linking qp and tp of the GIUH with the scale (k)
describes the ordinates of the timearea diagram at
and shape (n) parameters of the Nash IUH model.
each computational time interval.
By equating the first derivative (with respect to t)
5. Compute the peak discharge (qp ) of IUH given by
of Equation (6) to zero, t becomes the time to peak
Equation (1).
discharge, tp . Thus, taking the natural logarithm of both
6. Assume two trial values of the storage coefficient of
sides of Equation (6), differentiating with respect to t and
GIUH based Clark model as R1 and R2 . Compute
simplifying, we get,
the ordinates of two IUHs by Clark model using  
time of concentration Tc as obtained in step (iv) and 1 n  1
two storage coefficients R1 and R2 , respectively using ln[u(t)] D  C 14
t k t
Equation (4). Compute the IUH ordinates at a very
small time interval say 01 or 005 h so that a better Equating (14) to zero results in the following expres-
estimate of peak value may be obtained. sion, where t is equal to tp
7. Find out the peak discharges Qpc1 and Qpc2 of  
1 n  1
the IUHs obtained for Clark model for the storage  C D0 15
k t
coefficients R1 and R2 , respectively at step (6).
8. Compute the value of objective function, using the and hence,
relation: t D tp D kn  1 16

FCN1 D qp  Qpc1 2 9 Substituting the value of tp from Equation (16) in
Equation (6) and simplifying, we get
FCN2 D qp  Qpc2 2 10
1 n1
qp D e n  1n1 17
9. Compute the first numerical derivative FPN of the kn
objective function FCN with respect to parameter R From Equations (16) and (17), we get,
as:
FCN1  FCN2 n  1 n1
FPN D 11 qp tp D e n  1n1 18
R1  R2 n

Copyright 2007 John Wiley & Sons, Ltd. Hydrol. Process. 21, 1829 1840 (2007)
DOI: 10.1002/hyp
RUNOFF ESTIMATION USING GIUH MODELS 1833

Equating Equation (18) with Equation (3), we get, where, Volo is observed runoff volume, Volc is computed
runoff volume.
n  1 n1
e n  1n1 Qop  Qcp 
n PEP D 100 25
 055 Qop
RB
D 05764 RL 005 19
RA where, Qop is observed peak discharge, Qcp is computed
peak discharge.
All the terms in the right-hand side of Equation (19)
are known. The only unknown term is the Nash model Top  Tcp
parameter n, which is obtained by solving Equation (19) PETP D 100 26
Top
using the Newton Rapshon method of non-linear opti-
mization. The Nash model parameter k for the given where, Top is time to peak of observed discharge, Tcp is
velocity V is obtained using Equations (2) and (16) and time to peak of computed discharge.
the known value of the parameter n as follows:
 055
044L RB 1 STUDY AREA
kD RL 038 20
V RA n  1
The GIUH based Clark and Nash models described in
The derived values of n and k are used to determine the previous section were applied to the Ajay river
the complete shape of GIUH based Nash model using catchment defined by the Sarath gauging site of eastern
Equation (6). Subsequently, the D-hour UH is obtained India. The Ajay catchment lies between latitudes 23 500 N
using the relationship between IUH and UH of D-hour and 24 60 N and longitudes 86 160 E and 87 50 E. The river
as given by Equation (7). The DSRO hydrograph is length and catchment area up to the Sarath gauging site
estimated by convoluting the excess-rainfall hyetograph are 8218 km and 11914 km2 , respectively. The Ajay
with the UH obtained above. river emerges from forest covered hills of Chakai block
in Munger district of Jharkhand state at an elevation
Objective functions used for evaluation of the computed of 400 m above mean sea level. The study area, in
DSRO hydrographs general, has a mild terrain having small patches of
The following objective functions were employed for steep, mountainous and low lands. The soils in the Ajay
evaluation of the DSRO hydrographs computed by the river catchment have been almost entirely derived from
GIUH based Clark and Nash models in comparison with granite gneiss and schists and are mostly yellow and
the observed DSRO hydrographs as well as with the red at the surface except in low lands where olive-to-
DSRO hydrographs estimated by the HEC-1 package olive grey soils may also be found. Most of these soils
and the Nash IUH model: (i) efficiency (EFF); (ii) are sandy and loamy sand especially on uplands. But
absolute average error (AAE); (iii) root mean square in lowlands, medium textured sandy loam-to-loam soils
error (RMSE); (iv) average error in volume (AEV); are encountered. About 7% of the total area is under
(v) percentage error in peak (PEP); and (vi) percentage forest and about 63% is under cultivation. The average
error in time to peak (PETP). annual rainfall in the catchment varies from 1280 mm to
1380 mm while the mean peak flow at Sarath gauging site

m 
m is about 440 m3 s1 . The rainfall in plains is more than
Qoi  Q2  Qoi  Qci 2 in the hilly catchment. The monsoon rainfall accounts for
iD1 iD1 about 7580% of the annual rainfall in the catchment.
EFF D 100 21

m
2
Qoi  Q
iD1 EVALUATION OF THE GEOMORPHOLOGIC
th CHARACTERISTICS USING GIS
where, Qoi is i ordinate of the observed discharge, Q
is average of the ordinates of observed discharge, Qci is In application of the GIUH approaches, some of the
computed discharge, and m is number of ordinates. important geomorphological characteristics are required
to be evaluated from the topomaps. It is extremely dif-

m
ficult to manually derive the geomorphological charac-
jQoi  Qci j
teristics from topomaps. Thus, it discourages the users
iD1
AAE D 22 from adopting the GIUH approach. But, nowadays GIS
m
 software like Integrated Land and Water Information Sys-
 m
 tem (ILWIS), Earth Resources Data Analysis System
 Qoi  Qci 2
 (ERDAS), ARC/INFO and GRASS etc. are available for
 iD1
RMSE D 23 derivation of these characteristics in a simplified man-
m ner. In this study, the geomorphologic characteristics of
Volo  Volc  the catchment under study are evaluated using the GIS
AEV D 24 software, ILWIS as described below.
m

Copyright 2007 John Wiley & Sons, Ltd. Hydrol. Process. 21, 1829 1840 (2007)
DOI: 10.1002/hyp
1834 R. KUMAR ET AL.

The boundary of the catchment, the stream network Table I. Details of number, length, mean length and mean area
and topographical contours are derived using the Survey for streams of various orders for Ajay catchment up to Sarath
of India topomaps in the scale of 1 : 50 000. These maps Order of Number Length Mean length Mean area
were converted to digital form for ILWIS. The digitized stream (km) (km) (km2 )
map was corrected for any errors due to digitization
such as overshoot, undershoot etc. After removal 1 1266 75384 060 09
of errors from the drainage network map, the network 2 260 32548 125 38
3 59 17957 304 161
ordering was carried out in ILWIS. In ILWIS, a table is 4 19 7668 404 676
generated which contains the order of each stream, the 5 3 3813 1271 2840
number of streams and the length of the streams. The 6 1 5967 5967 11914
Strahlers ordering scheme was followed for ordering of
the river network (Strahler, 1957). Figure 1 shows the
drainage network map of the study area for the sixth
order streams of the catchment. Table I provides the The Tc is computed using the Kirpich formula. Using
details of stream numbers, length, average length and the DEM of Ajay catchment up to Sarath gauging site
average areas for streams of various orders for the study in a GIS based on proportionality constant, lengths and
area. The geomorphological parameters viz. RB D 4199, slopes of various segments of the streams of the drainage
RL D 2375, RA D 4190 were computed graphically by network, relative times of travel at various locations over
plotting stream numbers, mean stream length and mean the catchment were progressively computed, beginning
stream area versus order of the streams and estimating from the gauging site of the catchment. All the values of
absolute values of slopes of the best fit lines as shown in time of travels for each stream were then marked on the
Figures 24. The length of the highest order stream i.e. map of the catchment. Then these points were transferred
the sixth order stream (L ) is measured as 5967 km. to the digital form. Using an interpolation technique,
The time of travel (T) is directly proportional to a map of time distribution was drawn through these
L S05 , where L is the length of stream and S is the points. Subsequently, a timearea diagram in the form
slope of the stream. The proportionality constant can be of cumulative catchment area versus time of travel was
computed using the time of concentration (Tc ) of the prepared as shown in Figure 5. This timearea diagram
catchment, length and average slope of the mainstream. forms an input to the GIUH based Clark model.

Figure 1. Drainage network map of Ajay catchment up to sarath gauging site

Copyright 2007 John Wiley & Sons, Ltd. Hydrol. Process. 21, 1829 1840 (2007)
DOI: 10.1002/hyp
RUNOFF ESTIMATION USING GIUH MODELS 1835

8 8
RB = 4.199 RA = 4.190
7 7

6
6

Ln (Mean Stream Area)


Ln (Stream Numbers)

5
5

4
4
3
3
2
2
1
1
0
0 1 2 3 4 5 6
0 Stream Order
0 1 2 3 4 5 6
Stream Order Figure 4. Variation of mean stream area with stream order for Ajay
catchment up to sarath
Figure 2. Variation of stream numbers with stream order for Ajay
catchment up to sarath
1200
5

1000
RL = 2.375
4
Cumulative Area (km2)

800
Ln (Mean Stream Length)

3
600

2
400

1
200

0
0
0 2 4 6 8 10 12 14 16
Time (h)
-1
0 1 2 3 4 5 6
Figure 5. Time of travel versus cumulative area diagram for Ajay
Stream Order catchment up to sarath
Figure 3. Variation of mean stream length with stream order for Ajay
catchment up to sarath
values of the parameters are computed each time for the
rest of the nine rainfall-runoff events out of ten events
ANALYSIS AND DISCUSSION OF RESULTS considering the remaining one as an independent event.
The parameters of the Clark IUH model option of HEC-1 These geometric mean values of the parameters are used
and the Nash IUH model are estimated using histor- for computing the DSRO hydrographs for each of the
ical data of ten hourly rainfall-runoff events. Initially, independent events.
the base flows are separated from the discharge hydro- The values of the parameters for all the ten individual
graphs for obtaining the DSRO hydrographs. The uniform rainfall-runoff events and the geometric mean parameter
loss rate model is used for separating the losses from values for the Clark and Nash IUH models derived from
the rainfall hyetographs for obtaining the excess-rainfall the historical data are given in Table II. The values of
hyetographs. Subsequently, the DSRO and excess-rainfall arithmetic mean parameters of HEC-1 package viz. Clark
hyetographs are analysed to derive the parameters of IUH model and the Nash IUH model derived from the
Clark and Nash IUH models. For estimating the parame- historical data and their 90 and 95% confidence limits are
ters of Clark and Nash IUH models, the geometric mean given in Table III.

Copyright 2007 John Wiley & Sons, Ltd. Hydrol. Process. 21, 1829 1840 (2007)
DOI: 10.1002/hyp
1836 R. KUMAR ET AL.

Table II. Values of parameters for individual storms and mean 25 m s1 , which is based on the information obtained
parameters for Clark and Nash IUH models derived from
from the field engineers about the normally prevailing
historical data
velocity at the gauging site under study, during the
Event Clark IUH model Nash IUH occurrence of the type of rainfall-runoff storms, which
number (HEC-1 package) model are considered in this study.
For Geometric For Geometric For evaluation of the performances of the GIUH based
individual meana individual meana Clark and Nash models, DSRO hydrographs of each
storm storm of the independent events reproduced from the Clark
IUH model option of the HEC-1 package (HEC-1, 1990)
Tc R Tc R n k n k
and Nash IUH model (Nash, 1957) as well as GIUH
1 878 607 1015 651 386 239 431 259 based Clark and Nash models are compared with the
2 424 966 1100 619 231 421 456 243 observed DSRO hydrographs. Some of the commonly
3 1157 717 984 639 398 286 430 254 used objective functions are evaluated for the GIUH
4 1027 552 997 658 463 218 422 262 based Clark model, GIUH based Nash model, Clark
5 1097 678 990 643 480 245 421 258
IUH of HEC-1 package and the Nash IUH model using
6 1201 668 980 644 461 264 423 256
7 1050 703 995 641 465 244 422 259 observed and the computed DSRO hydrographs.
8 1822 349 936 693 619 203 409 264 The values of Tc and R for the GIUH based Clark
9 1338 633 968 648 595 230 411 260 model were obtained as 913 and 845 while the values
10 672 785 1045 633 318 273 440 255 of n and k for the GIUH based Nash model were obtained
a Geometric mean values based on the parameters T and R and n and k as 344 and 310, respectively. The values of Tc and
c
of the remaining storms. R for the GIUH based Clark model and n and k for
the GIUH based Nash model are identical for all the
rainfall-runoff events because the parameter velocity is
Table III. Mean parameters of Clark and Nash IUH models considered constant for all the events in the absence of
derived from the historical data and their 90% and 95% con-
fidence limits cross sectional details of the gauging site and Mannings
roughness coefficient. As seen from Table III, the values
Parameter Mean Standard Confidence Confidence of Tc and R for the GIUH based Clark model and n
value deviation level D 90% level D 95% and k for the GIUH based Nash model are very close
Lower Upper Lower Upper to the mean values of these parameters derived from the
limit limit limit limit historical data of all the ten events and lie well within
90% confidence limits of these parameters values.
Parameters of Clark IUH model (HEC-1 package)
Tc 1067 338 584 1500 447 1686 Comparison of observed and computed DSRO
R 666 159 463 869 405 923
hydrographs
Parameters of Nash IUH model
n 442 117 292 591 250 634 The DSRO hydrographs computed by the GIUH based
k 262 061 184 341 162 363 Clark and Nash models are compared with the observed
DSRO hydrographs as well as the DSRO hydrographs
computed using the Clark IUH model option of the
The DSRO hydrographs are computed using the GIUH HEC-1 package and the Nash IUH model as shown in
based Clark and Nash models. For applying the GIUH Figures 610. The values of peak discharge and time to
models, excess-rainfall hyetographs may be separated peak of the DSRO hydrographs for the ten rainfall-runoff
from the observed rainfall by SCS method (Chatterjee events are given in Table IV.
et al., 2001) or any other suitable approach as the GIUH
approaches are applicable to the ungauged catchments. Comparison of objective functions used for evaluation of
However, in the present study, for comparing the DSRO the computed DSRO hydrographs
hydrographs computed by all the four approaches, the The values of following objective functions, as
same excess-rainfall hyetographs have been adopted for described earlier, computed for evaluation of the DSRO
the GIUH based Clark and Nash models as adopted for hydrographs for GIUH based Clark model, GIUH based
the Clark and Nash IUH models. Also, the geometric Nash model, Clark IUH model option of the HEC-1
properties of the gauging section and the value of package and the Nash IUH model are given in Table V:
Mannings roughness coefficient as well as the velocities (i) EEF; (ii) AAE; (iii) RMSE; (iv) AEV; (v) PEP; and
corresponding to discharges passing through the gauging (vi) PETP. The objective functions based on comparison
section at different depths of water flow were not of the observed and computed DSRO hydrographs for the
available for the Sarath gauging site. Thus, the two four methods, presented in Table V, show that the values
approaches of estimation of the velocity mentioned above of EFF for GIUH based Clark model are the highest
and discussed in Kumar et al. (2002) could not be among the four methods for five flood events out of the
applied in this study. Instead, the GIUH based Clark and ten events. The values of EFF for the GIUH based Nash
Nash models were run by adopting the peak velocity of model are the highest among the four methods in case of

Copyright 2007 John Wiley & Sons, Ltd. Hydrol. Process. 21, 1829 1840 (2007)
DOI: 10.1002/hyp
RUNOFF ESTIMATION USING GIUH MODELS 1837

Excess Rainfall (mm)


Excess Rainfall (mm)
400 8 Event No.3
Event No. 1 1000 24
6 20
16
4 900 12
8
2 4
800 0
300

Direct Surface Runoff (m3/s)


0
Direct Surface Runoff (m3/s)

0 1 2 3 4 5
0 1 2 3 4 5
700 Time (h)
Time (h)

Observed 600
Observed
200 GIUH (Clark) 500 GIUH (Clark)
GIUH (Nash) GIUH (Nash)
400
HEC-1 HEC-1
Nash 300 Nash
100
200

100

0
0 0 6 12 18 24 30
0 6 12 18 24 30 Time (h)
Time (h)

Excess Rainfall (mm)


Event No.4 20
Excess Rainfall (mm)

600 16
900
Event No. 2 16
12 12
800 8
8
4
500 4
0
700 0 1 2
0
Direct Surface Runoff (m3/s)

0 1 2 Time (h)
Direct Surface Runoff (m3/s)

Time (h) 600


400
500
Observed
Observed
300 400 GIUH (Clark)
GIUH (Clark)
GIUH (Nash)
GIUH (Nash)
300 HEC-1
200 HEC-1
Nash
Nash 200

100 100

0
0 6 12 18 24 30
0
Time (h)
0 6 12 18 24 30
Time (h) Figure 7. Comparison of observed and computed DSRO hydrographs for
event Nos. 3 and 4
Figure 6. Comparison of observed and computed DSRO hydrographs for
event Nos. 1 and 2

Table IV. Peak discharge and time to peak of observed and computed DSRO hydrographs for the various rainfall-runoff events

Event no. Observed GIUH (Clark) GIUH (Nash) HEC-1 Nash

Qop Top Qcp Tcp Qcp Tcp Qcp Tcp Qcp Tcp
m3 s1  (h) m3 s1  (h) m3 s1  (h) m3 s1  (h) m3 s1  (h)

1 3632 9 3187 11 3323 10 3612 10 3433 11


2 4684 6 4725 10 4848 9 5358 9 5158 10
3 7920 11 8299 9 8896 9 9751 8 9364 10
4 8084 9 7153 9 7311 8 8140 8 7585 9
5 1053 8 908 9 924 8 1054 8 972 9
6 5019 10 5108 9 5200 8 5931 7 5497 9
7 3990 9 3976 10 4053 9 4617 9 4247 10
8 3188 13 3068 11 3386 10 3452 9 3520 10
9 4583 10 4181 9 4256 8 4854 7 4498 9
10 1612 7 1693 9 1736 8 1967 8 1796 9

Copyright 2007 John Wiley & Sons, Ltd. Hydrol. Process. 21, 1829 1840 (2007)
DOI: 10.1002/hyp
1838 R. KUMAR ET AL.

Table V. Error functions computed for DSRO hydrographs

Excess Rainfall (mm)


120 Event No.5
4
estimated by the GIUH based Clark and Nash models, HEC-1
package and Nash IUH model for the various rainfall-runoff
events
100
Methods Error functions for DSRO hydrographs
0
0 1

Direct Surface Runoff (m3/s)


EFF AAE RMSE AEV PEP PETP Time (h)
80
Event 1
GIUH (Clark) 7838 4491 5808 14572 1227 2222
GIUH (Nash) 9604 2050 2735 15921 851 1111 Observed
60
HEC-1 9643 1941 2621 15440 055 1111 GIUH (Clark)
NASH 8746 3775 4667 16044 549 2222
GIUH (Nash)
Event 2
GIUH (Clark) 3854 7918 11323 19161 088 6667 40 HEC-1
GIUH (Nash) 6394 6358 8915 20216 349 5000 Nash
HEC-1 6039 6308 9322 19963 1439 5000
NASH 2898 9028 12346 20410 1013 6667
20
Event 3
GIUH (Clark) 9145 6462 7619 37823 478 1818
GIUH (Nash) 8585 8069 9901 40157 1232 1818
HEC-1 6738 11853 14947 39735 2312 2327 0
NASH 8589 8351 9890 40415 1823 909 0 6 12 18 24 30
Event 4 Time (h)
GIUH (Clark) 8640 7763 10362 33352 1152 0
GIUH (Nash) 9789 3132 4577 36247 957 1111
Event No. 6

Excess Rainfall (mm)


HEC-1 9198 6668 8220 35510 070 1111 600 20
NASH 9643 3718 3701 36514 618 0 16
Event 5 12
GIUH (Clark) 9321 688 808 3974 1381 1250 8

GIUH (Nash) 7812 1095 1459 4247 1225 0 500 4


0
HEC-1 6478 1540 1847 4164 009 0 0 1
Direct Surface Runoff(m3/s)

NASH 8572 938 1182 4285 768 1250 Time (h)


Event 6 400
GIUH (Clark) 9425 3342 3976 20631 177 1000
GIUH (Nash) 8035 5795 7389 21668 361 2000
HEC-1 5818 8917 10752 21550 1818 3000 Observed
300
NASH 8694 4650 6038 21836 952 1000 GIUH (Clark)
Event 7 GIUH (Nash)
GIUH (Clark) 9397 2655 3354 16833 034 1111
GIUH (Nash) 9134 3064 4069 17998 159 0 200 HEC-1
HEC-1 7988 5302 6161 17698 1572 0 Nash
NASH 9496 2607 3131 18164 643 1111
Event 8 100
GIUH (Clark) 7732 440 5307 14828 375 1538
GIUH (Nash) 5689 6336 7385 15898 62 2308
HEC-1 2597 8422 9640 15443 829 3077
NASH 6355 5817 6802 16041 1040 2308 0
0 6 12 18 24 30
Event 9
GIUH (Clark) 7845 5491 6843 16445 878 1000 Time (h)
GIUH (Nash) 4827 7885 10674 17187 714 2000
Figure 8. Comparison of observed and computed DSRO hydrographs for
HEC-1 2119 10494 13145 17016 592 3000 event Nos. 5 and 6
NASH 5618 7184 9835 17326 185 1000
Event 10
GIUH (Clark) 4841 2969 4067 8376 501 2857 of AAE and RMSE for HEC-1 package are the lowest
GIUH (Nash) 7800 1860 2788 9237 767 1429
HEC-1 7919 1962 2618 9029 2201 1429
for two events while the values of AAE and RMSE for
NASH 5568 3047 3863 9312 1140 2857 Nash IUH model are the lowest for only one event. It is
observed that the values of AEV are the lowest for GIUH
based Clark model for all the ten flood events. It is also
two flood events, whereas, the values of EFF for HEC-1 observed that the values of PEP are the lowest for GIUH
package are the highest in case of the other two events. based Clark model for six flood events. The values of PEP
However, the values of EFF for Nash IUH model are the are the lowest for HEC-1 package for three events; while,
highest in case of only one event. It is observed that the the values of PEP are the lowest for Nash IUH model for
values of AAE and RMSE for the GIUH based Clark only one event. Thus, it is found that the GIUH based
model are the lowest in five flood events out of the ten Clark model and the GIUH based Nash model which
events. The values of AAE and RMSE for GIUH based consider the catchment under study as ungauged, estimate
Nash model are the lowest for two events. The values the DSRO hydrographs with reasonable accuracy for the

Copyright 2007 John Wiley & Sons, Ltd. Hydrol. Process. 21, 1829 1840 (2007)
DOI: 10.1002/hyp
RUNOFF ESTIMATION USING GIUH MODELS 1839

Excess Rainfall (mm)


Event No.7 16

Excess Rainfall (mm)


500 500 E vent No. 9 20
12 16
8 450 12

4 8

400 4
0 400
0 1 2 3 0
Direct Surface Runoff (m3/s)

0 1

Direct Surface Runoff (m3/s)


Time (h)
350 Time (h)

300 300
Observed
Observed
GIUH (Clark) 250
GIUH (Clark)
GIUH (Nash)
200 200 GIUH (Nash)
HEC-1
HEC-1
Nash 150
Nash
100 100

50

0 0
0 6 12 18 24 30 0 6 12 18 24 30 36
Time (h) Time (h)

Event No. 10 8

Excess Rainfall (mm)


Event No.8 200
Excess Rainfall (mm)

400 8

350 4 175 4

150
Direct Surface Runoff (m3/s)

300 0 0
Direct Surface Runoff (m3/s)

0 1 2 3 4 5 0 1 2 3
Time (h) Time (h)
125
250
Observed
Observed
100 GIUH (Clark)
200 GIUH (Clark)
GIUH (Nash)
GIUH (Nash)
75 HEC-1
150 HEC-1
Nash
Nash
50
100

25
50

0
0 0 6 12 18 24
0 6 12 18 24 30
Time (h)
Time (h)
Figure 10. Comparison of observed and computed DSRO hydrographs
Figure 9. Comparison of observed and computed DSRO hydrographs for for event Nos. 9 and 10
event Nos. 7 and 8

Ajay catchment up to Sarath gauging site as compared to


the HEC-1 package and Nash IUH model.
RMSE for the GIUH based Clark model are lower than
Comparison of the GIUH based Clark model and GIUH GIUH based Nash model for six flood events out of the
based Nash model ten events; whereas the values of AAE and RMSE for
Comparison of the values of error functions of the GIUH based Nash model are lower than the GIUH based
observed and computed DSRO hydrographs for the Clark model for four events. It is observed that the values
GIUH based Clark model and GIUH based Nash model, of AEV are lower for GIUH based Clark model for all
presented in Table V, shows that the values of EFF for the flood events. It is also seen that the values of PEP
GIUH based Clark model are higher than the GIUH based are lower for GIUH based Clark model for six events;
Nash model for six flood events out of the ten events; whereas the values of PEP are lower for the GIUH based
while the values of EFF for the GIUH based Nash model Nash model for four events. In general, it is found that the
are higher than the GIUH based Clark model for four GIUH based Clark and Nash models estimate the DSRO
flood events. It is observed that the values of AAE and hydrographs with comparable accuracy.

Copyright 2007 John Wiley & Sons, Ltd. Hydrol. Process. 21, 1829 1840 (2007)
DOI: 10.1002/hyp
1840 R. KUMAR ET AL.

CONCLUSIONS Horton RE. 1945. Erosional development of streams and their drainage
basins: hydrophysical approach to quantitative morphology. Bulletin of
In the GIUH approaches, the catchment is considered Geological Society of America 56: 275 370.
Jain SK, Singh RD, Seth SM. 2000. Design flood estimation using
as ungauged and the parameters of the GIUH based GIS supported GIUH approach. Water Resources Management 14:
Clark and Nash models are estimated with reasonable 369376.
accuracy by using the geomorphological characteristics of Karlinger MR, Troutman BM. 1985. Assessment of the instantaneous unit
hydrograph derived from the theory of topologically random networks.
the catchment under study. Comparison of values of the Water Resources Research 21(11): 1693 1702.
error functions reveals that the GIUH based Clark model Karvonen T, Koivusalo H, Jauhiainen M, Palko J, Weppling K. 1999. A
and the GIUH based Nash model, estimate the DSRO hydrological model for predicting runoff from different land-use areas.
Journal of Hydrology 217: 253265.
hydrographs with reasonable accuracy as compared to Kumar R, Chatterjee C, Lohani AK, Kumar S, Singh RD. 2002.
the Clark IUH model option of the HEC-1 package and Sensitivity analysis of the GIUH based Clark model for a catchment.
Nash IUH model. Performances of Clark IUH model Water Resources Management 16: 263 278.
Lee KT. 1998. Generating design hydrographs by DEM assisted
option of HEC-1 package and the Nash IUH model for geomorphic runoff simulation: a case study. Journal of the American
estimation of the DSRO hydrographs may be considered Water Resources Association 34(2): 375 384.
as comparable. Nash JE. 1957. The form of instantaneous unit hydrograph. International
Association of Sciences and Hydrological Publications 45(3):
The geomorphological parameters required for deriva- 114121.
tion of the GIUH based Clark model and GIUH based Rinaldo A, Marani A, Rigon R. 1991. Geomorphological dispersion.
Nash model are evaluated for the Ajay catchment up Water Resources Research 27(4): 513 525.
Rinaldo A, Rodriguez-Iturbe I. 1996. Geomorphological theory of the
to Sarath gauging site of eastern India using the GIS hydrological response. Hydrological Processes 10(6): 803829.
package viz. ILWIS. Manual estimation of geomorpho- Rodriguez-Iturbe I, Devoto G, Valdes JB. 1979. Discharge response
logical parameters is a tedious and cumbersome process analysis and hydrologic similarity: the interrelation between the
geomorphologic IUH and the storm characteristics. Water Resources
and often discourages the field engineers from developing Research 15(6): 1435 1444.
regional methodologies for solving various hydrologi- Rodriguez-Iturbe I, Gonzalez-Sanabria M, Bras RL. 1982a. The geo-
cal problems of the ungauged catchments or in limited morphoclimatic theory of the instantaneous unit hydrograph. Water
Resources Research 18(4): 877886.
data situations. At times, it also leads to erroneous esti- Rodriguez-Iturbe I, Gonzalez-Sanabria M, Caamano G. 1982b. On the
mates. On the other hand, modern techniques like the GIS climatic dependence of the IUH: a rainfall-runoff analysis of the Nash
serve as an efficient approach for storage, processing and model and the geomorphoclimatic theory. Water Resources Research
18(4): 887 903.
retrieval of a large amount of database. Its spatial mod- Rodriguez-Iturbe I, Rinaldo A. 1997. Fractal River Basins: Chance and
eling and tabular databases constitute a powerful tool for Self Organization. Cambridge University Press: New York.
the data analysis. Also, the database created and stored Rodriguez-Iturbe I, Valdes JB. 1979. The geomorphologic structure of
hydrologic response. Water Resources Research 15(6): 1409 1420.
in GIS system may be updated as and when required. Rosso R. 1984. Nash model relation to Horton order ratios. Water
Resources Research 20(7): 914920.
Snell JD, Sivapalan M. 1994. On geomorphological dispersion in
REFERENCES natural catchments and the geomorphological unit hydrograph. Water
Resources Research 30(7): 2311 2323.
Al-Wagdany AS, Rao AA. 1998. Correlation of the velocity parameter Strahler AN. 1957. Quantitative analysis of watershed geomorphology.
of three geomorphological instantaneous unit hydrograph methods. Transactions American Geophysical Union 38: 913 920.
Hydrological Processes 12: 651659. Troutman BM, Karlinger MR. 1984. On the expected width function for
Bhaskar NR, Parida BP, Nayak AK. 1997. Flood estimation for ungauged topologically random channel networks. Journal of Applied Probability
catchments using the GIUH. Journal of Water Resources Planning and 21: 836884.
Management, ASCE 123(4): 228 238. Troutman BM, Karlinger MR. 1985. Unit hydrograph approximation
Central Water Commission. 1983. Flood Estimation Report for Upper assuming linear flow through topologically random channel networks.
Narmada and Tapi Subzone (Subzone 3c) (1983). Directorate of Water Resources Research 21(5): 743 754.
Hydrology (SC) CWC: New Delhi. Troutman BM, Karlinger MR. 1986. Averaging properties of channel
Chatterjee C, Jha R, Lohani AK, Kumar R, Singh R. 2001. Runoff curve networks using methods in stochastic branching theory. In Scale
number estimation for a basin using remote sensing and GIS. Asian- Problem in Hydrology, Gupta VK, Roriguez-Iturbe I, Wood EF (eds).
Pacific Remote Sensing and GIS Journal 14: 17. D. Reidel: Norwell, MA; 85216.
Chutha I, Dooge JCI. 1990. The shape parameters of the geomorphologic USDA-SCS. 1972. Hydrology. National Engineering Handbook, Section
unit hydrograph. Journal of Hydrology 117: 8197. 4 . US Department of Agriculture: Washington, DC.
Clark CO. 1945. Storage and the unit hydrograph. Transactions of the Valdes JB, Fiallo Y, Rodriguez-Iturbe I. 1979. A rainfall-runoff analysis
American Society of Civil Engineers 110: 1419 1446. of the geomorphologic IUH. Water Resources Research 15(6):
Corradini C, Melone F, Singh VP. 1995. Some remarks on the use of 1421 1434.
GIUH in the hydrological practice. Nordic Hydrology 26: 297 312. van der Tak LD, Bras RL. 1990. Incorporating hillslope effects into
Gupta VK, Waymir E, Wang CT. 1980. A representation of an the geomorphologic instantaneous unit hydrograph. Water Resources
instantaneous unit hydrograph from geomorphology. Water Resources Research 26(10): 2393 2400.
Research 16(5): 855 862. Zelazinski J. 1986. Application of the geomorphological instantaneous
HEC-1. 1990. In Flood Hydrograph Package Users Manual . US Army unit hydrograph theory to development of forecasting models in
Corps of Engineers: Hydrologic Engineering Center, USA. Poland. Hydrological Sciences Journal 31: 26.

Copyright 2007 John Wiley & Sons, Ltd. Hydrol. Process. 21, 1829 1840 (2007)
DOI: 10.1002/hyp

You might also like