Evaluating the performance of streamflow simulated by an eco-hydrological model calibrated and validated with global land surface actual evapotranspiration from remote sensing at a catchment scale in West Africa

Abstract

Study region: Republic of Benin. Study focus: In the Ou´em´e River Basin (48 292 km2), we investigate if the Soil and Water Assessment Tool (SWAT) model can satisfactorily simulate streamflow when observed data is not available and when SWAT is calibrated using monthly AET data from GLEAM. Thus, we compare the performance of SWAT by applying two different calibration and validation procedures (i) using time series of observed streamflow (Q-proc); (ii) using GLEAM (v3.0a) AET data (AETproc). New hydrological insight for the region: The streamflow simulations from AET-proc did not perform as well as with Q-proc. The highest NSE values from AET-proc for streamflow were obtained in 2 catchments; 0.45 and 0.66, with Q-proc these NSE values were 0.55 and 0.72, respectively. In fact, with Q-proc, acceptable streamflow NSE and R2 values were obtained at all four gauges. Yet, the Q-proc simulation of AET was poor, with only 3 and 1 gauges showing satisfactory PBIAS and KGEs, respectively. The AET-proc however simulated AET very well, with satisfactory KGE and PBIAS statistics at all four gauges, and satisfactory NSE and R2 values at 2 gauges. For streamflow, only the R2 values were satisfactory at all gauges. Comparing further simulated variables, such as the soil moisture, water yield, and crop yields from Q-proc and AET-proc, the result showed the use of GLEAM AET data for calibration can reproduce the temporal dynamics of the rainfall-runoff behaviour. Further research is required to fine-tune the AET-proc for improved streamflow simulation, conceivably by including soil moisture products.