Using satellite-based weather data as input to SWAT in a data poor catchment

Abstract

Satellite-based weather estimates are spatially and temporally continuous data that seamlessly capture weather variability across the globe. A need was identified to assess the usefulness and reliability of satellite-based weather estimates in the lower Vaal River Catchment due to the scarcity of weather data. We explored the utility of Climate Forecast System Reanalysis (CFSR) data for streamflow simulation in an important agricultural catchment in South Africa, using the Soil and Water Assessment Tool (SWAT). First, we compared CFSR precipitation against rain gauge measurements at an annual time step. Two models were set up, one with conventional gauge weather data obtained from the South African Weather Services (SAWS) and the Agricultural Research Council (ARC), and the other with CFSR data. Calibration and validation of the models were performed using Sequential Uncertainty Fitting ver.2 (SUFI-2) algorithm of the SWAT Calibration Uncertainties Program (SWAT- CUP) at five stream gauge sites, with the data split between 2003 to 2008 and 2009 to 2013 for calibration and validation respectively. A low statistical agreement between CFSR precipitation and rain gauge data was obtained. The model with gauge data slightly outperformed CFSR model before calibration and during validation, whereas CFSR performed better mainly during calibration. Overall, both models achieved fair simulations of streamflow and long-term monthly streamflow regime at both the main catchment outlet and interior sub-catchments. While conventional gauge data remain the most accurate and reliable data source, CFSR can play a role as an alternative data input where there is inadequate gauge data.