See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/276306170

Rainfall-Runoff Modeling Using MIKE 11 NAM Model for Vinayakpur Intercepted

Catchment, Chhattisgarh

Article in Indian Journal of Dryland Agricultural Research and Development · January 2014
DOI: 10.5958/2231-6701.2014.01206.8

CITATIONS READS
16 4,706

5 authors, including:

Ashutosh - Singh Surjeet Singh

Rani Lakshmi Bai Central Agricultural University Jhansi National Institute of Hydrology
125 PUBLICATIONS 1,358 CITATIONS 99 PUBLICATIONS 673 CITATIONS

SEE PROFILE SEE PROFILE

A.K. Nema Gaurav Singh

Banaras Hindu University Central Soil and Water Conservation Research and Training Institute
47 PUBLICATIONS 217 CITATIONS 35 PUBLICATIONS 49 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

IrisSeg: A Fast and Robust Iris Segmentation Framework for Non-Ideal Iris Images View project

Natural Resource Management View project

All content following this page was uploaded by Gaurav Singh on 20 September 2016.

The user has requested enhancement of the downloaded file.

Indian J. Dryland Agric. Res. & Dev. 2014 29(2) : 01-07

Rainfall-Runoff Modeling using MIKE11 NAM Model

for Vinayakpur Intercepted Catchment, Chhattisgarh

Ashutosh Singh1*, Surjeet Singh2, A.K. Nema1, Gaurav Singh, Anshu Gangwar1
1
Department of Farm Engineering, Institute of Agricultural Sciences, BHU, Varanasi-221 005 Uttar Pradesh
2
National Institute of Hydrology, Roorkee-247 667, Uttarakhand
Email: ashutosh11001@gmail.com

ABSTRACT: Rainfall-runoff estimation for a catchment is of vital importance in most of the hydrologic analysis for water resources
planning. This study envisages the rainfall-runoff modeling using MIKE 11 NAM model in Vinayakpur intercepted catchment in
Chhattisgarh state. The model was calibrated using measured stream flow data for the period 2001 to 2004 and then validated from
period 2005 to 2007. The calibration and validation procedures were carried out to provide a satisfactory estimation. The simulated
runoff occurred maximum in August (1681.63 cumecs) and minimum in April (84.14 cumecs). The outputs of the calibrated model
were used in water resources management model viz. MIKE BASIN as they normally work based on monthly flows with a large time
horizon. The optimum values of nine NAM model parameters obtained during calibration procedure were used for simulation. The
reliability of MIKE 11 NAM was evaluated based on Nash–Sutcliffe coefficient, correlation coefficient (r2) and root mean square
error (RMSE). The R2 value of model calibration and validation were observed to be 0.79 and 0.75, respectively.

Key words: Rainfall-runoff, NAM, Vinayakpur intercepted catchment, MIKE BASIN

Rainfall-runoff estimation plays a very important role for data appropriately. Lipiwattanakarn et al. (2004) compared the
purposes of water resources planning; flood forecasting, performance of ANN and MIKE 11 NAM models. They found
pollution control, inter-basin water transport, decision- that ANN model was more efficient in simulating the discharge
making and policy formulations etc. Precipitation distribution, peak while MIKE11 NAM was more capable in simulating
evaporation, transpiration, abstraction, watershed topography, the base flow or discharge. Liu et al. (2007) suggested a novel
and soil types are implicit and explicit factors which affect sensitive analysis scheme for MIKE 11 NAM rainfall-runoff
rainfall-runoff process in modeling (Dawson et al., 2000). model which indicated the sensitivity analysis problem in a
The runoff discharge and flow rate at a river site varies greatly general multi-objective framework. Model calibration is needed
throughout the course of a time, depending on seasonal rainfall, because the parameters of such models cannot be obtained
watershed characteristics and many other parameters. Several directly from measurable quantities of catchment characteristics.
models have been developed to simulate hydrological processes A trial and error parameter adjustment is made in the process
such as rainfall-runoff process, which can be divided into of manual calibration. In such cases, comparing the simulated
three categories viz. “physical”, “conceptual” and “black box” and the observed hydrographs based on a visual judgment is
models. Lumped conceptual models require significant amounts used as a basis for the calibration process. In auto-calibration,
of calibration data and also experience of user is necessary. modeling parameters are automatically adjusted according to a
Physical distribution based models become unsuitable as they specified search scheme and the resulting numerical measures of
need a large amount of data about topology, soil, vegetation the goodness of fit (Madsen, 2003). MIKE 11 NAM is a rainfall-
and geological characteristics of catchment area. The quality of runoff model which is a part of the MIKE 11 RR module. MIKE
observed data also plays important role in accuracy of empirical 11 NAM, MIKE SHE and WATBAL models were validated on
black box models and they are useful operational tools in three catchments in Zimbabwe for water resources decision-
cases where enough meteorological data are not available making (Refsgaard and Knudsen, 1996), where at least one year’s
(Bojkow, 2001). Because of non-linear and multi-dimensional data were available for calibration. The runoff hydrographs for
nature, rainfall-runoff modeling is extremely complicated the un-gauged Nzhelele river were simulated using MIKE 11
(Lipiwattanakarn et al., 2004). The widely known rainfall-runoff NAM model and the Australian Water Balance Model (AWBM)
models identified are of rational method (Mc Pherson, 1969), by Makungo et al. (2010). The simulated runoff hydrographs can
Soil Conservation Service-Curve Number Method (Maidment, be used in water resources planning and management, and water
1993) and Green Ampt Method (Green and Ampt, 1911). The resources systems operation. The rainfall-runoff relationship
hydrological model of NAM 11 is an integrated and conceptual in the Strymonas river catchment was studied by Doulgeris et
model of rainfall-runoff which is able to simulate surface flow, al. (2012) using the MIKE 11 NAM model. MIKE 11 NAM
subsurface and base flow; this model was developed by Danish was used for the simulation of rainfall-runoff process in the
Hydraulic Institute (DHI) in 1972 (DHI, 1999). Shamsudin and Strymonas river and Lake Kerkini by Doulgeris et al., (2008)
Hashim (2002) applied NAM model for predicting the runoff for water resources management aspects. In this paper, lumped
rate in Liang River located in northern part of Malaysia and they conceptual rainfall-runoff model is used for long term daily and
obtained satisfactory results and found that the predicted values monthly discharge calculation for the Vinayakpur intercepted
by the NAM model were in accordance with the historical catchment based on the available rainfall and evaporation data.
1
 Ashutosh Singh et al.
For this purpose, the MIKE 11 NAM model was adopted in order Structure of the NAM model
to do better calibration of NAM parameters. The calibration and The model structure is an imitation of the land phase of the
validation procedures of the model were carried out to provide a hydrological cycle. NAM simulates the rainfall-runoff process
satisfactory estimation. by continuously accounting for water content in four different
Material and Methods and mutually interrelated storages that represent different
physical elements of a catchment as shown in Figure 1.
Study area
The study area is the Vinayakpur intercepted basin which falls
under the Durg district of Chhattisgarh. The basin extends
between 20°35′10″ to 20°85′11″ N latitude and 81°23′05″ to
81°7′30″ E longitude. The total area of the basin is 1019.17 km2.
The Durg district is located at 317 m above mean sea level. It is
bounded by Bemetara district in the North, Rajnandgaon district
in the West, Balod district in the South and Raipur district in the
East. The annual average rainfall is 1052 mm. The study area
can be located on the Survey of India (SOI) toposheet no. 64
G/4, 64 G/8, 64 D/13, 64 H/1, 64 H/5, 64 D/13 64 H/2 and 64
H/6 on 1:50,000 scale. The toposheets were used to delineate
the study area for preparing the drainage network map. The
Tandula river is the main river with a total length of 51.01 km Fig. 1 : Structure of NAM model for rainfall-runoff simulation
and stream order is five. There is only one gauging site in the
study area which is located at Tandula Vinayakpur (owned by These mutually interrelated storages include snow storage,
Water Resource Department, Chhattisgarh). The distribution of surface storage, lower or root zone storage and groundwater
the study area as in Table 1 is comprised of six blocks namely- storage. In addition, NAM allows treatment of man-made
Balod, Dondilohara, Gunderdehi, Gurur, Durg and Doundi. interventions in hydrological cycle such as irrigation and
groundwater pumping.
Table 1 : Distribution of study area under different blocks
of Durg district Based on the meteorological input data, NAM produces
catchment runoff as well as information about other elements
Area of block falling Percentage area of the land phase of the hydrological cycle, such as the temporal
Blocks
under basin (km2) of basin variation of the evapotranspiration, soil moisture content,
Balod 250.49 24.57 groundwater recharge and groundwater levels. The resulting
catchment runoff is divided conceptually into overland flow,
Dondilohara 238.41 23.39 inter flow and base flow.
Durg 19.06 1.87 The application of MIKE 11 model for rainfall runoff estimation
Gunderdehi 473.24 46.43 can be divided into two stages. The first stage is calibration
Gurur 31.12 3.05 process to determine an optimum value of model parameters.
The second stage is stream flow simulation using estimated
Doundi 1.17 0.11 model parameter during calibration process.
Table 2 : Input data requirement by the model NAM module (MIKE11) model setup for Vinayakpur
intercepted catchment
Variable Type Unit TS Type
NAM module of MIKE11 was used to estimate all the basin
Daily Rainfall Rainfall Mm Step accumulated water balance components, i.e. runoff, actual evapotranspiration,
and groundwater recharge. For this purpose, seven years (2001
Daily potential Evaporation Mm Step accumulated to 2007) data of rainfall, evaporation and runoff were used. The
evapotranspiration rainfall of Admabad, Anda, Balod and Gondly stations was used.
The observed runoff data of Vinayakpur gauging site was used
Daily discharge Discharge m3/s Instantaneous for comparison of results. The catchment area of 1019.17 sq. km
was assigned to the model. The total data period of 07 years was
Rainfall-runoff modeling (MIKE 11 NAM model) divided into two parts; i.e. 2001 to 2004 for the calibration and
NAM is a part of rainfall-runoff module of MIKE 11 river 2005 to 2007 for validation purpose.
modeling system which simulates rainfall-runoff process
occurring at the catchment scale. This can either be applied NAM calibration
independently or used to represent one or more contributing During calibration, the catchment parameters are adjusted
catchments that generate lateral inflows to a river network. In this until a satisfactory fit between simulated flow contributions,
manner, it is possible to treat a single catchment or a large river (overland flow, inter flow and base flow) and observed stream
basin containing numerous catchments and a complex network flow is attained. The following objectives are usually considered
of rivers and channels within the same modeling framework. during model calibration:
2
 Rainfall-Runoff Modeling using MIKE11 NAM Model
i. A good agreement between the average simulated and Input data requirements
average observed catchment runoff. The data requirements for NAM MIKE 11 RR model are
ii. A good overall agreement of the shape of the hydrograph. meteorological data, stream flow data for model calibration and
iii. A good agreement of the peak flows with respect to timing, verification, definition of catchment parameters, and definition
rate and volume. of initial conditions. The basic meteorological data requirements
are precipitation time-series, potential evapotranspiration
iv. A good agreement for base flows. time-series, temperature and radiation time-series if snow
accumulation and melt is to be modelled. Table 3 describes the
In calibration process, different calibration objectives mentioned
required input data and the format.
above should be taken into account. If objectives are of equal
importance, one should seek to balance all objectives, whereas There are six rain gauge stations in basin namely Admabad,
in case of priority to a certain objective, that objective should Anda, Balod, Gondali, Kharkhara and Matia. The rainfall data
be favoured. at these stations is collected by Water Resources Department,
Govt. of Chhattisgarh. The rainfall of these stations was analyzed
For a general evaluation of calibrated model, simulated
station-wise. These rainfall data were used in rainfall-runoff
runoff is compared with observed runoff measurements. Both
modeling and estimation of water availability in the Vinayakpur
graphical and numerical performance measures should be
intercepted basin.
applied in calibration process. The graphical evaluation includes
comparison of simulated and observed hydrograph, and
comparison of simulated and observed accumulated runoff. The Results and Discussion
numerical performance measures include overall water balance Rainfall-runoff modeling (MIKE 11 NAM model)
error (i.e., difference between average simulated and observed
The simulated minimum and maximum runoff for a seven year
runoff), and a measure of overall shape of hydrograph based on
period (2001-2007) show that the annual runoff varies between
coefficient of determination and Nash-Sutcliffe coefficient.
68.6 mm to 611.6 mm. The runoff at the gauging site was
The model was calibrated for the period 2001 to 2004 using simulated with the help of NAM model. The simulated runoff
auto-calibration. In auto-calibration, model fixes surface and was maximum for the month of August (1681.63 m3/s) and
root zone parameters and groundwater parameters automatically. minimum for the month of April (84.14 m3/s).

Table 3 : Model parameter values and their range

Parameter Unit Description Model Parameter Change Effects

parameter range
values
Umax mm Maximum water content 19.2 5.76–20 Increase Peak runoff decreased
in surface storage Runoff volume reduced
Lmax mm Maximum water content 292 100–300 Increase Peak runoff decreased
in root zone storage Runoff volume reduced
CQOF Overland flow runoff 0.263 0.1–1 Increase Peak runoff decreased
coefficient Runoff volume increased
CKIF hrs Time constant for 534.4 200–1000 Increase -
routing interflow
CK1K2 hrs Time constant for 29.6 10–50 Increase Peak runoff decreased The triangular
routing overland flow shape expand horizontally
TOF Root zone threshold 0.0144 0-0.99 Increase Peak runoff decreased
value for overland flow Runoff volume reduced
TIF Root zone threshold 0.000784 0-0.99 Increase -
value for interflow
TG Root zone threshold 0.813 0-0.99 Increase -
value for GW recharge
CKBF hrs Time constant for 3707 500-10000 Increase Base flow decreased
routing base flow Lower
base flow/recharge to
lower reservoir
3
 Ashutosh Singh et al.
Model calibration Conclusions
The calibrated results of observed and simulated runoff are The results of NAM model for calculation of daily discharge
shown in Figure 2. The correlation coefficient during the from catchment rainfall is acceptable to use monthly discharges
calibration is 0.79 and Nash-Sutcliffe coefficient is 0.63 which in water management models like MIKE BASIN. The rainfall-
indicates a good match. The total water balance error during runoff models viz., MIKE 11 NAM is lumped conceptual type
calibration is 6.1% which is within acceptable limits. of model and need not require lot of data to calculate daily and
monthly discharges. Thus these are useful tools for use in water
management on the large scale modeling with medium and long
term simulation periods.

References
Bojkow VH. 2001. Runoff Modeling with Genetic Programming
and Artificial Neural Networks. D2K Technical Report
D2K TR 0401-1.
Dawson CW, Brown MR and Wilby RL. 2000. Inductive
learning approaches to rainfall-runoff modeling. Int. J.
Neural Syst., 10: 43-57.
Fig. 2 : Comparison of observed and simulated runoff during model DHI Water and Environment. 1999. MIKE 11 Reference Manual.
calibration (coefficient of determination = 0.63)
Doulgeris C, Georgiou P, Papadimos D and Papamichail D. 2012.
Model validation Ecosystem approach to water resources management using the
The model was validated for the period 2005 to 2007 using the MIKE 11 modeling system in the Strymonas River and Lake
model calibrated parameters. The validated results of observed Kerkini. J. Environ. Manage., 94: 132-143.
and simulated runoff are shown in Figure 3. The correlation Doulgeris C, Halkidis I and Papadimos D. 2008. Use of Modern
coefficient during validation was calculated as 0.75 and Nash- Technology for the Protection and Management of Water
Sutcliffe coefficient was 0.56 which indicates a good match. The Resources in Strymonas/Struma River Basin, Technical Report,
total water balance error during the validation was 4.7% which Greek Biotope/Wetland Centre (EKBY) Thermi Greece. 82 p.
is again within acceptable limits.
Green WH and Ampt GA. 1911. Studies on soil physics. Journal of
Agricultural Sciences, 4: 1-24.
Lipiwattanakarn S, Sriwongsitanon N and Saengsawang S. 2004.
Improving neural network model performance in runoff
estimation by using an antecedent precipitation index. Journal of
Hydrosci. Hydraul. Eng., 22(2): 141-154.
Liu HL, Chen X, Bao AM and Ling Wang. 2007. Investigation of
groundwater response to overland flow and topography using
a coupled MIKE SHE/MIKE 11 modeling system for an arid
watershed. Journal of Hydrology, 347: 448-459.
Madsen H. 2003. Parameter estimation in distributed hydrological
Fig. 3 : Comparison of observed and simulated runoff during model catchment modeling using automatic calibration with multiple
validation (coefficient of determination = 0.56) objectives. Adv Water Resour., 26: 205-216.
The difference between total accumulated observed and Maidment DR. 1993. Handbook of Hydrology, (1stedn), McGraw Hill,
simulated runoff during calibration and validation was observed NewYork. 1424 p.
to be 6.1% and 4.7%, respectively. This indicates that difference
Makungo R, Odiyo JO, Ndiritu and Mwaka. 2010. Rainfall-runoff
between total accumulated observed and simulated runoff is
modeling approach for ungauged catchments: A case study of
well within acceptable limits. The coefficient of determination
Nzhelele River sub-quaternary catchment. Physics and Chemistry
for calibration and validation period of model was 0.63 and
of the Earth, 35: 596-607.
0.56, respectively which indicates that the developed model
is performing well to simulate the runoff. The auto-calibration Mc Pherson MB. 1969. Some notes on the Rational Method of Storm
procedure for the NAM model parameters, a set of NAM model Drain Design, Tech. Memo. ASCE, Water Resources Research
parameters was calculated and then the simulated discharge Program. Harvard University, Cambridge, USA.
was compared with observations. The final values of NAM Refsgaard JC and Knudsen J. 1996. Operational Validation and
parameters that have been adjusted in the calibration process are Intercomparison of Different Types of Hydrological Models.
illustrated in Table 3. It shows that the final calculated values Water Resources Research, 32: 2189-2202.
of model parameters represented hydrologically the catchment’s
Shamsudin S and Hashim N. 2002. Rainfall runoff simulation using
characteristics in the range of predefined bandwidth during the
MIKE11 NAM. Jurnal Kejuruteraan Awam, 15(2): 26-38.
auto-calibration procedure.
4

View publication stats