Bonfring International Proceedings of AARCV 2012 - International Conference on Advances in Architecture and Civil

Engineering, 21st – 23rd June 2012, Paper ID HYD 110 , VOL. 1

`
Variable Source Area Rainfall – Runoff
Modeling: A Case Study On Poondi Basin -
Tamil Nadu
H.S.Govardhana Swamy 1, Dr.B.V.Mudgal 2

Abstract--- The central concept of the variable source area in the top layers of soil through large openings usually called
modeling is that, input generally infiltrate through undisturbed as macro pores created by biotic activates. ( Beven, K. and
soils, migrates down the slope and maintains saturation or German P. ,1982) However Harsh’s findings remained
near- saturation at lower slope regions. These lower slope relatively unrecognized, mainly because evidences to the
regions rapidly contribute subsurface flow to storm flow, as contrary out weighted his field observations and
the zone of saturated soil surface expands. In the present study measurements. But when series of Experiments carried out by
such a variable source area concept is used to develop a Hewlett (1961) at Coweeta research station, supported Hursh,
lumped conceptual Rainfall and Runoff model for Poondi the universality of the Horton’s theory began to be questioned.
basin, Tamil Nadu.The model developed has twelve Hewlett put forward an alternative explanation of the
parameters. The parameters are calibrated using Genetic Hydrological process. According to him, storm runoff is due
Algorithm optimization procedure. In order to check the to a combination of subsurface flow and runoff from rain
validity of the model for ungauged catchments, the optimized falling on those parts of the catchments, riparian of the down
parameter set obtained for that of Poondi basin were directly slope movement of soil water. Hewlett ( 1961), noted “ Under
used for estimating river flow of a Thevankariar basin .The prolonged and heavy rainfall, the storm flow- producing area
simulated model components show good correlation with contiguous to stream channels may grow wider and wider
observed values. depending on the nature and depth of the earth mantle” and
this area called as variable source area. . (Betson R. P. and
Keywords--- Rainfall and Runoff, variable source area, Marius J B, 1969) The model with a variable source area
Genetic Algorithm; concept, which was then totally new at that period, was
considered as one of the models which have changed the
fundamental ideas in surface hydrology. The central concept
of the variable source area is that, Input generally infiltrate
I. INTRODUCTION through undisturbed soils, migrates down slope and maintains
saturation or near- saturation at lower slope regions. These
lower slope regions rapidly contribute subsurface flow to
R Ainfall-runoff modeling is an important area of
hydrologic ststudies, and one in which research is still being
actively carried out. Watershed model is a mathematical
storm flow, as the zone of saturated soil surface expands. The
degree to which saturation and subsequent actual expansion
would occurs, for a given slope, varies as a function of
representation of the catchment processes capable of antecedent soil - moisture conditions, precipitation volume,
simulating stream flow and other outputs of the catchment and duration of input. Such a variable source area concept is
system, corresponding to any given values of the input mainly used to develop a conceptual Rainfall and Runoff model and is
precipitation. Hence the model is normally utilized either for applied for Poondi basin Tamil Nadu. In order to check the
generating stream flow or to determine how runoff is affected validity of the model for un-gauged catchments, the optimized
by factors such as afforestation (e.g. Aston and Dunin,1980; parameter set obtained for that of Poondi basin were directly
Eeles and Blackie 1993 ), Urbanization (Smith and used for estimating river flow from another basin called
Bedient,1981 ) or rainfall augmentation ( Lamb and Thevankariar basin in Tamil Nadu. For the visualization of
Linsley,1971 ) There are two-broad categories of rainfall- results, x-y graphs are plotted. The results obtained show high
runoff models, namely : Conceptual models :Which attempt correlation with observed datas.
to represent the physical processes which occur on the
catchment ( Franchini M. and Pacciani (1991)., Duan Q.
Sorooshian S. and Gupta V. K. (1992), Mathematical models
II. STUDY AREA
(which only consider the mathematical relationship between
rainfall and runoff without considering the physical process The study area Poondi is the sub-basin of kosathalayaru basin.
which occur on the catchment.) David R Dawdy, and Terence Kosasthalayaru is the interstate river of sub-continent
O ‘Donnell (1965)., R.T.Clarke (1973). Hursh and Hoover ( originating partly and flowing through states of AP and TN
1941) were the first to note that direct run off in stream to flow and draining in to Bay of Bengal. The basin map is as shown
in figure 1. The basin is located between 790 14’ E. and 790
1
Asst Professor Dept of Civil Engg, MSRIT, , B’lore – 560054,
54’ E longitudes and 120 52’N and 130 20’N latitudes. The
2
Associate Professor, Dept of Civil Engg, Anna University, Chennai -25 total catchment area of the sub basin upto poondi regulator is

ISBN XXXX – XXXX | © 2012 Bonfring

 Bonfring International Proceedings of AARCV 2012 - International Conference on Advances in Architecture and Civil
Engineering, 21st – 23rd June 2012, Paper ID HYD 110 , VOL. 1

`
2704 km2. This sub basin varies in altitude from 0 to 300 m respectively. Figure 3, shows time variation of rainfall and
above msl. Daily Mean temperature of the catchment is above observed river flow during calibration period.
27.5 c. The catchment is influenced by four separate seasons
of rainfall: 1. Cool Weather (dry) (Jan – Feb) 2. Hot weather Table 1: Annual values of Rainfall and Runoff of Poondi and Thavankariar basins
(dry) (Mar – May) 3.South – West monsoon (June – Sep) 4. MODEL STRUCTURE
North- East monsoon (Oct – Dec).The average total annual
rainfall is 925 mm and is spread over the seasons .Rain is not VSA Runoff model developed for the poondi basin is as
evenly spread throughout the year, but is concentrated, on an shown in fig 2.The model has three non linear storage zones:
average of 23 rainy days each during S-W and N-E monsoons. Interception storage, Soil moisture store and Ground water
Variations in rainfall are significant with the N-E monsoon store. Inflow and outflow of moisture through these zones
being considerable more erratic than the S-W monsoon. All takes place based on antecedent storage conditions and is also
these factors indicate the importance of irrigation to Tamil related to few parameters. This VSA model tries to simulate
Nadu. following hydrological processes:
1) Interception 2) Evapo-transpiration
3) Variable source area 4) Interflow
5) Macro-pore flow 6) Ground Water recharge
7) Base flow on daily basis.

III. METHODOLOGY
The data available for the study are daily rainfall and river
flow values. The Rainfall measured at eight rain gauge
stations are taken in to account. The catchment mean rainfall
is determined using Thessian polygon method. The daily
rainfall and river flow series used in the study has the time
period of 1st January 1991 to 31st December 2001. Model
uses all the data items in depth units i.e., mm. The model
Figure 1: basin map
parameters are obtained by minimizing the most commonly
DATA USED FOR THE VSA MODEL used objective function for hydrologic simulation models,
The data available for the study are daily rainfall and river (Diskin and Simon., 1977). i.e. the sum of squared deviations
flow values. Rain fall measured at rain gauge station, (1) as defined by equation:
Pallipattu (2) Ponnai (3) Waljapet, (4) Sholighur (5) N
Arakonam (6) Kesavaram, (7) Thiruvelangadu and (8) OF = ∑ (Qobs – Qsim )2
Poondi, is made available by W.R.O., PWD, T.N. . The daily t=1
rainfall series used has the time period of 1st January 1991 to where Q obs = observed runoff for the t th day;
31st December 2001. The Annual Rainfall and Runoff for Q sim = simulated runoff for the t th day;
Poondi and Thevankariaiar basin are as shown Table 1 N = number of days of simulation;
Catchment Parameters of the model are determined by fitting the model to
Poondi Thevankariaiar hydrologic data with an optimization technique called Genetic
Algorithm. (Franchini M. 1996). (Wang, Q.J., 1991). The
Annual Annual
year model parameter set finally selected is based on the results of
Rainfall Runoff Rainfall Runoff
many optimization runs. Various optimization runs were
mm mm mm mm conducted using different initial seed, so that for every
optimization run a new search space is initialized. The
1991 864.30 437.15
optimized parameter set is shown in table 2. The split record
test is used for model fitting technique, as it is a widely
1992 583.98 175.19 accepted test. The technique uses 1st January 1991 to 31st
1993 747.79 224.33 December 1995 rainfall data and river flow data to calibrate
1994 700.79 210.03 the model and then 1st January 1996 to 31st December 2001
1995 954.04 286.20 867.32 147.54 rainfall and river flow data for validation. The simulated river
1996 1436.00 574.39 788.54 115.20 flow is obtained by adding four runoff components namely
1997 870.96 243.28 1047.00 436.07 source area runoff, interflow, macro pore-flow and base flow.
1998 819.38 215.81 904.15 136.45 The set of model parameters obtained for the Poondi
1999 942.03 222.60 915.50 155.60 catchment was directly applied to the data of Thevankariar
2000 754.04 223.51 890.63 120.36 basin (Tamil Nadu state) for determining the model
2001 815.13 189.93 - - performance, as an example for use in un-gauged catchments.
For the visualization of results, x-y graphs are plotted for daily

ISBN XXXX – XXXX | © 2012 Bonfring

 Bonfring International Proceedings of AARCV 2012 - International Conference on Advances in Architecture and Civil
Engineering, 21st – 23rd June 2012, Paper ID HYD 110 , VOL. 1

`
values of observed runoff & simulated runoff and are as
shown in figures 3 and 4.

OBSERVED VS SIMULATED RIVERFLOW ANALYSIS

Figure 4, shows time series graph for observed river flow and
simulated river flow for Poondi basin for the years 1991 and
1995. Model output agrees well with the observed river flow.
Few data points show large difference between the observed
and simulated river flow. Main reasons for this discrepancy
are : 1) errors in the assumed initial conditions.2) From the
beginning of first monsoon spell , much of rainfall is made
available to building up of soil moisture level till field
capacity is reached 3) At the end of monsoon period
sometimes there are one or two trailing monsoon spells which
sometimes show mismatch

Sl. Optimized Bound

Parameter Description
No. Value Lower Upper
Maximum
1 CEPMAX interception 2.872 0.0 3.0
storage capacity
Source area
2. SAK 3.122 1.0 6.0
coefficient
Source area
3. SAE 0.712 0.4 0.8
exponent
4. SZWP Wilting Point 40.027 40 150
5. SZFC Filed Capacity 299.286 SZWP 350.00
6. SZPC Pore Capacity 447.890 SZFC 450.00
Soil Zone
Storage
7. SZRK 0.6 0.1 1.0
recession
coefficient
Interflow runoff
8. SZROK 0.05 0.01 0.1
coefficient
Base flow
9. GZK 493.518 200.0 700.0
coefficient
Base flow
10. GZE 6.448 3.0 9.0
exponent

Table 2: Optimized parameter set: range and actual values

OBSERVED VS SIMULATED RIVERFLOW ANALYSIS

Figure 4, shows time series graph for observed river flow and
simulated river flow for Poondi basin for the years 1991 and
1995. Model output agrees well with the observed river flow.
Few data points show large difference between the observed
and simulated river flow. Main reasons for this discrepancy
are : 1) errors in the assumed initial conditions.2) From the
beginning of first monsoon spell , much of rainfall is made
available to building up of soil moisture level till field
capacity is reached 3) At the end of monsoon period
sometimes there are one or two trailing monsoon spells which
sometimes show mismatch

ISBN XXXX – XXXX | © 2012 Bonfring

 Bonfring International Proceedings of AARCV 2012 - International Conference on Advances in Architecture and Civil
Engineering, 21st – 23rd June 2012, Paper ID HYD 110 , VOL. 1

`
MODEL CALIBRATION

Model calibration was done using a probability based search REFERENCES

technique namely Genetic Algorithm the technique is been Betson R. P.(1964). ‘What is watershed runoff?’. J. of
discussed. Model was applied to poondi catchment calibration Geophysics Res., Vol.69 No 8 pp.1541-1552.
data period for the catchment is five years from 1st Jan 1991 to
31st Dec 1995. Set of parameters obtained from model Betson R. P. and Marius J B (1969). ‘Source areas of storm
calibration is used for model validation. runoff’. Water Resource Res., Vol.5 No3 pp.574-582.

MODEL VALIDATION Beven, K. and German P. (1982). ‘Macropores and water

flow in soils’. Water Resource Res, Vol.18 No 5 pp.1311-
Verification period for poondi catchment is from 1st Jan 1997 1325.
to 31st Dec 2001. The parameter set obtained from calibration
is used to validate the data in order to verify the possibility of Black P. E. (1970). ‘Runoff from watershed models’. Water
model being used for un-gauged catchments. Resource Res., Vol.6 No 2 pp.465-477.

Bonell M. (1993). ‘Progress in the understanding of

generation dynamics in forests’. J. Hydrology., Vol.150, pp.
217-275.

Chow V. T. David T. M. and Larry W. M. (1988). Applied

Hydrology, McGraw Hill Book Company.

Clarke R. T. (1973). ‘A review of some mathematical models

used in hydrology with observations on their calibration and
use’. J. Hydrology., Vol.19, pp. 1-20.
IV. CONCLUSION
The model developed can be used where only daily rainfall Diskin M. H. and Simon E. (1977). ‘A Procedure for the
and river flow values are known. Further the model developed selection of objective functions for hydrologic simulation
can be applied where the infiltration excess runoff theory of models’. J. Hydrology., Vol.43, pp.129-149.
Horton is not valid and runoff is produced by other
mechanisms such as pipe flow or macro-pore flow, soil matrix Duan Q. Sorooshian S. and Gupta V. K. (1992). ‘Effective
flow or saturation excess. The usage of Genetic Algorithm for and efficient global optimization for conceptual rainfall runoff
model calibration yielded good results. From the results it can models’. Water Resources Res., Vol.28 No 4 pp.1015-1031.
be said that the variable source area runoff model works fine
for the catchments which do not have very less rainfall for the Dunne T. (1978). Field studies of hillslope flow processes. In
catchment. The model developed was applied to sub Kirby (ed) Hillslope Hydrology. John Wiley and Sons, pp.
catchments: Poondi and Thevankariar in Tamil Nadu. The two 227-293.
catchments are more or less similar. The Model was applied
using split record technique, for obtaining calibrated Dunne T. and Black R. D. (1970a). ‘An experimental
parameters. For the Poondi catchment: First five year data investigation of runoff production in permeable soils’. Water
(1991-1995) used for calibration and remaining six year data Resource Res. Vol. No 62 pp. 478-490.
(1996-2001) used for validation. Thevankariar catchment:
1995 year data is used for calibration and year 2000 used for Franchini M. (1996). ‘Use of a Genetic Algorithm combined
validation. In order to check the validity of the model for un- with a local search method for the automatic calibration of
gauged catchments, the optimized parameter set obtained for conceptual rainfall-runoff models’. Hydrological Sci. J.,
that of Poondi catchment was directly used for estimating river Vol.41 No 1 pp. 21-39.
flow from Thevankariar basin. The Model output consists of
all the model components and measures such as model Franchini M. and Galeati G. (1997). ‘Comparing several
efficiency, standard error, correlation coefficient between genetic algorithm schemes for the calibration of conceptual
observed and simulated river flows. The results obtained from rainfall-runoff models’. Hydrological Sci. J. Vol. 42 No 3 pp.
model calibration show high correlation with observed data 357-379.
for the basins studied. The model performance is as shown in
the table 3. Franchini M. and Pacciani (1991). ‘Comparative analysis of
several conceptual rainfall-runoff models’. J. Hydrology.,
Vol.122, pp.161-219.

ISBN XXXX – XXXX | © 2012 Bonfring

 Bonfring International Proceedings of AARCV 2012 - International Conference on Advances in Architecture and Civil
Engineering, 21st – 23rd June 2012, Paper ID HYD 110 , VOL. 1

`
Gan T. Y. and Biftu G. F. (1996). ‘Automatic calibration of
rainfall-runoff models : Optimization algorithms, catchment
conditions, and model structure. Water Resource Res.,
Vol.321 No 2 pp. 3513-3524.

Gupta S. K. and Solomon S. I. (1977). ‘Distributed numerical

model for estimating runoff and sediment discharge of
ungauged rivers- 1. The information system’. Water
Resource Res., Vol.13 No 3 pp. 613-618.

Gupta V. K. and Sorooshian S. (1985). ‘The Automatic

calibration of conceptual catchment model using derivative
based optimization algorithms’. Water Resource Res.,
Vol.214 No pp. 473-485.

Horton R. E. (1933). ‘The role of infiltration in the hydrologic

cycle’. Trans. A. Geophysics Union, Vol.14, pp. 446-460.

Jain S. K.(1993). ‘Calibration of conceptual models for

rainfall-runoff simulation’. Hydrological Sci. J. Vol. 38(5),
pp.431-441.

Singh V. P. and Birsoy Y. K. (1977). ‘Some statistical

relationships between rainfall and runoff’. J. Hydrology.,
Vol.34, pp. 251-268.

Sorooshian S. Duan, Q. and Gupta V. K. (1993). ‘Calibration

of rainfall runoff models : application of global optimization
to the Sacramento soil moisture accounting model’. Water
Resource Res., Vol.29 No 4 pp. 1185-1194.

Sorooshian S. and Gupta V. K. (1983). ‘Automatic calibration

of conceptual rainfall-runoff models: The question of
parameter observability and uniqueness’. Water Resource
Res., Vol.19 No 1 pp. 260-268.

Wang Q. J. (1991). The genetic algorithm and its application

to calibrating Conceptual rainfall-runoff models’. Water
Resource Res., Vol.2 No 9 pp. 2467-2471.

ISBN XXXX – XXXX | © 2012 Bonfring