Model parameter transfer for streamflow and sediment loss prediction with SWAT in a tropical watershed

Abstract

Distributed hydrological models are increasingly used to describe the spatiotemporal dynamics of water and sediment fluxes within basins. In data-scarce regions like Ethiopia, oftentimes, discharge or sediment load data are not readily available and therefore researchers have to rely on input data from global models with lower resolution and accuracy. In this study, we evaluated a model parameter transfer from a 100 hectare (ha) large subwatershed (Minchet) to a 4800 ha catchment in the highlands of Ethiopia using the Soil and Water Assessment Tool (SWAT). The Minchet catchment has long-lasting time series on discharge and sediment load dating back to 1984, which were used to calibrate the subcatchment before (a) validating the Minchet subcatchment and (b) through parameter transfer validating the entire Gerda watershed without prior calibration. Uncertainty analysis was carried out with the Sequential Uncertainty Fitting-2 (SUFI-2) with SWAT-Cup and ArcSWAT2012. We used a similarity approach, where the complete set of model parameters is transposed from a donor catchment that is very similar regarding physiographic attributes (in terms of landuse, soils, geology and rainfall patterns). For calibration and validation, the Nash-Sutcliff model efficiency, the Root Mean Square Error-observations Standard Deviation Ratio (RSR) and the Percent Bias (PBIAS) indicator for model performance ratings during calibration and validation periods were applied. Goodness of fit and the degree to which the calibrated model accounted for the uncertainties were assessed with the P-factor and the R-factor of the SUFI-2 algorithm. Results show that calibration and validation for streamflow performed very good for the subcatchment as well as for the entire catchment using model parameter transfer. For sediment loads, calibration performed better than validation and parameter transfer yielded satisfactory results, which suggests that the SWAT model can be used to adequately simulate monthly streamflow and sediment load in the Gerda catchment through model parameter transfer only.