Simulation of Sediment Yield Using SWAT Model: A case of Kulekhani Watershed

Abstract Soil erosion is the major cause of land degradation and reservoir sedimentation. Therefore, modelling of runoff and sediment yield at a watershed level is very important. A conceptual, distributed and continuous time, SWAT2012 (Soil and Water Assessment Tool) model was selected for the simulation of the runoff and sediment yield from Kulekhani watershed, in Bagmati river basin, Nepal. The main objective of the study was to examine the applicability of the SWAT model in Kulekhani watershed. To set up the model for simulation a 30 m DEM (Digital Elevation Model), 1 km spatial resolution of land use map and 10 km spatial resolution of soil map was used. The daily precipitation and daily minimum and maximum air temperature from 1972 to 2013 was used as input to the model. The stream flow data was available from 2007 to 2009 and for four months from 2004. The daily sediment record for four months from 2004 was available. The model was calibrated by using both automated calibration and manual calibration for daily stream flow by using the flow data from 2007 to 2008 and validated for 2009 and 2004. The calibration for sediment was conducted for the whole period for which the sediment data was available. The model was not validated for sediment yield as there was not enough length of data to do so. The performance of the model was evaluated by using a time series plots of observed and simulated value and the statistical measures of coefficient of determination (R2) and the Nash-Sutcliffe efficiency (NS). The statistical analysis of calibration results for Kulekhani watershed showed satisfactory agreement between observed and simulated daily values, with an R2 value of 0.6, and NS of 0.44 in discharge simulation; and an R2 value of 0.54, and NS of 0.53 at Palung Khola and an R2 value of 0.4, and NS of 0.07 at Chitlang Khola for sediment simulation. The R2 and NS value for flow validation period in 2009 was 0.59 and -0.59 respectively. The model was also validated for flow at Palung Khola for 2004 with R2 and NS value of 0.66 and 0.29 respectively; and at Chitlang Khola for 2004 with R2 and NS value of 0.81 and 0.74 respectively. In general, the model was capable of simulating runoff and sediment from Kulekhani watershed. But, the result from sediment simulation was not as good as the runoff. This is believed to be because of the inability of the SWAT model to realistically simulate the sediment from gully erosion, landslide and mass wasting. The result could have been improved by using distributed rainfall data, longer period of runoff and sediment record and better quality land use and soil data