Evaluating the Uncertainties in the SWAT Model Outputs due to DEM Grid Size and Resampling Techniques in a Large Himalayan River Basin

Uncertainty in hydrological and nutrient modeling is a challenging task over large snow-fed and complex Himalayan watersheds. Digital elevation models (DEMs) are an important input for hydrological models. At times, DEM grid size needs to be altered for hydrological simulation in large basins. Therefore, appropriate grid sizes need to be investigated for modeling of hydrological variables. Many studies have investigated the effect of DEM grid size on flow and nutrient modeling. However, to the best of the authors' knowledge, no one has attempted to understand the uncertainties in SWAT model outputs due to DEM grid size and resampling methods over large, snow-affected, mountainous river basins. In this study, the Gandak River Basin, a large snow-covered river basin in the Himalayas, was examined at monthly and annual time steps using 40-1,000-m grid sizes resampled with nearest neighborhood, bilinear, and cubic convolution methods. Results showed that sediment and flow are greatly affected by DEM resolutions and by the choice of resampling method. Flow and sediment are overestimated for DEM grid sizes > 300 and > 150 m, respectively. However, total nitrogen (TN) and total phosphorous (TP) change at DEM grid sizes of >= 150 m via slope and volume of flow. T-test statistics indicate the significance of changes in SWAT outputs due to DEM resolution and resampling. They are significant for SWAT outputs at > 500-m grid sizes at a yearly time step. Results also showed significant increases in relative difference (RD) with changes in DEM grid size during snowmelt periods, which indicates changes in the SWAT snowmelt parameter due to the resampled DEM. The proposed results can be applied to flow and nutrient modeling over any snow-affected, large, mountainous river basin. (C) 2017 American Society of Civil Engineers.