APPLICATION OF SWAT MODEL TO THE STUDY AREA

Size: px
Start display at page:

Download "APPLICATION OF SWAT MODEL TO THE STUDY AREA"

Transcription

1 CHAPTER 5 APPLICATION OF SWAT MODEL TO THE STUDY 5.1 Introduction The Meenachil river basin suffers from water shortage during the six non-monsoon months of the year. The available land and water resources are to be effectively utilised to improve the livelihood and socio-economic conditions of the people living in the river basin. The land-water system of the area is adversely affected by the rapid growth of population and changes in the LU/LC. The need for hydrological investigations in the Meenachil river basin has been recognised with an aim to suggest improved catchment management programs for the conservation of soil and water resources in the area. The lack of decision support tools and limitation of hydrologic data significantly hindered the research activities in the area. In this chapter, the performance and feasibility of the SWAT 2012 model for prediction of flow in the Meenachil river basin of the humid tropics in south-west India has been tested and validated. 89

2 5.2 Model data inputs The SWAT is a comprehensive model requiring a diversity of information. The first step in setting up a SWAT river basin simulation is to partition the basin into sub units. The first level of sub division is the sub basin. The sub basin delineation is defined by surface topography so that the entire area within a sub basin flows to the sub basin outlet. The land area in a sub basin is divided into Hydrologic Response Units (HRUs). These portions of a sub watershed possess unique land use/management/soil attributes. The number of HRUs in a sub basin is determined by a threshold value for land use and soil delineation in the sub basin. The use of HRUs generally simplifies a simulation run because all similar soil and land use areas are lumped into a single response unit. The ArcGIS platform provides the user with a complete set of GIS tools for developing, running and editing hydrologic and management inputs and finally calibrating the model. The spatially distributed data required for ArcSWAT include the Digital Elevation Model (DEM), soil and land use data, either as shape files or grid data. The weather and measured streamflow data are also required as input for the calibration and prediction purposes Digital elevation model The DEM for the study area prepared to use with SWAT2012 is given in Figure 3.4. This DEM is used to delineate the river basin using automated delineation tool in SWAT. The entire river basin was divided into 17 sub basins (Figure 5.1), each of which was again divided into several HRUs. A total of 307 HRUs were created Climate data The climate data required are precipitation, maximum and minimum air temperature, wind speed, relative humidity and solar radiation. Values for these parametersmaybereadfromtherecordsofobserveddataorthesemaybegenerated. 90

3 Figure 5.1: SWAT delineated river basin map 91

4 The weather generator input file contains the statistical data needed to generate representative daily climate data for the sub basins. Climate data will be generated for two instances - when user specifies that simulated weather will be used or when measured data is missing. In the present study, a weather generator input file was created from the data record for 42 years from the weather station at Puthupally as given in Appendix 1. Daily observed data for precipitation from the four rain gauge stations - Kottayam, Erattupetta, Teekoy and Kozha were used for input data preparation. Daily observed data on maximum and minimum temperature, wind speed and relative humidity collected from Puthupally station were used for the climate input data preparation Streamflow data Daily streamflow values for Peroor and Cheripad collected from the Hydrology Division of Water Resources Department of Kerala State were used for preparing the observed data file for use in the calibration process Land use data Land use maps prepared from the satellite image for the year 1990 (Figure 3.15) was used for the calibration period. The SWAT land cover was appropriately selected from the in-built SWAT database for each land cover in the map and reclassified as given in Figure

5 Figure 5.2: SWAT reclassified land use map Soil data The digitised soil map was used in SWAT and the soil properties for different layers were fed as the input data for the soils in the user soils database of SWAT, as given in Appendix 2. Major soils of the study area are Muthur, Arpookara, Kooropada, Lakkattoor, Koduman, Nellappara and Mavady series as shown in Figure The Soil map was linked to the appropriate soil type from the soil data base and reclassified as showin in Figure

6 Figure 5.3: SWAT reclassified soil map 5.3 Model application In order to apply SWAT model to the Meenachil river basin, the major steps involved are: 1) data preparation, 2) river basin and sub basin delineation, 3) HRU definition, 4) sensitivity analysis, and 5) model calibration and validation. The precipitation and temperature data files were created for the observed data in the format specified in SWAT. The spatial data sets required were projected to the same projection, WGS 1984 UTM ZONE 43N using ArcGIS The DEM was used to delineate the watershed and to analyse the drainage pattern of the land surface terrain. The spatial data on LU/LC were reclassified into SWAT land cover/plant types. User defined soil types were added to the soil database and the spatial soil data were linked to the appropriate types. The multiple HRU definition suggested by the ArcSWAT User s Manual - 20 percent land use, 10 percent soil and 20 percent 94

7 slope threshold - was applied in the study. The parameter sensitivity analysis was done for the whole river basin. Eighteen hydrologic parameters pertinent to water flow (SWAT2005 User s Guide, 2007) were tested for sensitivity for the simulation of streamflow in the study area. The top ranked three parameters were used for calibrating the model. The data for the period 1990 to 2000 were used for calibrating the model for the observed flows at Peroor and Cheripad. An independent precipitation, temperature, wind speed, relative humidity and streamflow data set ( ) were prepared. Periods from 1987 to 1989 were taken as warm-up period for calibration. The warm-up period allows the model to get the hydrologic cycle fully operational. For the study area, an increasing trend in the area under rubber plantation and a decreasing trend in the area under mixed crop cultivation is observed. Hence, while simulating streamflow using SWAT model, land use update files has been incorporated. Since the gap from 1973 to 1990 is high to properly interpolate the land cover variation, simulation of the model was done for the period from 1990 to The land cover map for the year 1990 (Figure 5.2) was used with the SWAT model. SWAT allows a maximum of ten files for updating the land use. A spatial linear interpolation was applied for updating the land use. Also, the area under rubber plantation is found to be progressing downstream. Table 5.1 gives the area variation made on creating the land use update files. Seven land use update files were created considering a linear variation between the available year-span. The final land use map prepared after incorporating the LU/LC change in the model is given in Figure

8 Table 5.1: Year wise area conversion made for land use update Year Sub basins changed Land use New Land use ,4,5 AGRL RUBR ,8,9 AGRL RUBR ,6,10,11,13,14 AGRL RUBR ,6,7 RICE URMD ,16 RICE URMD AGRL RUBR ,14 AGRL URBN Figure 5.4: SWAT final land use/land cover map 96

9 5.4 SWAT-CUP5 software SWAT CUP5 software was used for the calibration of the model. Sequential Uncertainity Fitting(SUFI2) algorithm was used for calibration. The model was calibrated for the three top ranked parameters - alpha bf (base flow alpha factor in days), gw revap (ground water revap coefficient) and rchrg dp (deep aquifer percolation factor). 5.5 Evaluation of model performance Simulated data from the SWAT model can be compared statistically to observed data to evaluate the predictive capability of the model. 5.6 Uncalibrated model results The uncalibrated model results were obtained from a SWAT simulation using the default SWAT settings for parameter values before any calibration was performed. The uncalibrated simulation was performed for the period , with as warm-up period. The R 2 values for correlation between simulated and observed streamflow were relatively high (0.73 and 0.84) for the two stations, indicating a strong linear relationship between simulated and observed flows. Also, the N SE values were greater than 0.5, which shows that the model is suitable for this particular river basin. But the PBIAS for Peroor was -51.4% (ie., > ±25 %), which indicates an over-prediction. Considering this, the need for calibration of the model was recognised. 5.7 Model calibration For calibrating the model, a preliminary sensitivity analysis was performed on all the flow parameters based on the available climatic and hydrologic 97

10 input data for the period from 1987 to The first three ranked parameters were selected for calibration purpose. Guidance for identifying input parameters for manual calibration provided by Feyereisen et al. (2007) based on the study conducted by Van Liew et al. (2007) has been followed in this study for calibrating the streamflow from the two gauging sites in the Meenachil river basin. Looking to the uncalibrated model result, the two parameters, base flow recession constant (alpha bf) and groundwater revap coefficient(gw revap), were adjusted for the entire area, since the base flow is high for the simulated flows. The parameter rchrg dp is found to be the most sensitive parameter. So the model was calibrated with these three parameters for the observed streamflow values at Cheripad and Peroor. The study area of Meenachil river basin lies in the highland and midland regions. For the highland station at Cheripad the un-calibrated model results give -2.0% PBIAS. Hence, the parameter rchrg dp was calibrated for sub basins in the midland region alone to arrive at the best value for predicting the accurate streamflow. The SUFI2 algorithm in SWAT CUP was used for calibration. The calibrated values for the parameters are given in Table 5.2. Table 5.2: SWAT flow sensitive parameters and fitted values after calibration using SUFI2 Sl.no. Sensitivity Lower and Final fitted Parameter description parameters upper bounds values 1 alpha bf Baseflow alpha factor (days) 2 gw revap Ground water revap coefficient 3 rchrg dp (for Deep aquifer percolation midland) fraction sub basins) Comparison between the observed and calibrated streamflow values for eleven years of simulation indicated that there is a good agreement between the observed and simulated flows with higher values of Nash-Sutcliffe efficiency and lower values of RSR. The calibrated model predictive performance statistics for 98

11 monthly flows are summarised in Table 5.3. Table 5.4 gives the calibration statistics for individual years. Table 5.3: Streamflow calibration results for Cheripad and Peroor Station N SE R 2 RSR d PBIAS(%) Cheripad Peroor Table 5.4: Streamflow calibration statistics for each year Cheripad Peroor Year N SE R 2 d N SE R 2 d Model validation The streamflow for from the stations at Peroor and Cheripad were used for validating the predictive capability of the SWAT model with respect to Meenachil river basin. The comparison statistics for observed and simulated monthly streamflow for the validation period are shown in Table 5.5. Table

12 gives the statistics for individual years. Figures 5.5 to 5.6 give the time series of observed and simulated monthly streamflow during the calibration and validation period. The other graphical forms of model evaluation - double mass and scatter plot = are given in Figures It can be seen from the plot that the model underpredicts high values and overpredicts lower values. However, the overall statistics shows that the model is very good for predicting monthly streamflow in the Meenachil river basin. This calibrated model was then used for computing the impacts on hydrology due to land use/land cover change in the river basin. Table 5.5: Streamflow validation statistics ( ) Station N SE R 2 RSR d PBIAS(%) Cheripad Peroor Table 5.6: Streamflow validation statistics for each year Cheripad Peroor Year N SE R 2 d N SE R 2 d

13 Figure 5.5: Observed and simulated flow at Peroor ( ) 101

14 Figure 5.6: Observed and simulated flow at Cheripad ( ) 102

15 Figure 5.7: Scatter plot of observed and simulated streamflow - Cheripad 103

16 Figure 5.8: Scatter plot of observed and simulated streamflow - Peroor 104

17 Figure 5.9: Double mass curve for observed and simulated streamflow - Cheripad 105

18 Figure 5.10: Double mass curve for observed and simulated streamflow - Peroor 106

19 The average water balance components of the basin, as obtained by the model simulation, is represented in Figure Figure 5.11: Average annual basin values - SWAT model results 107