Comparison of single-site, multi-site and multi-variable SWAT calibration strategies

This study compares single-site, multi-site and multi-variable SWAT calibration. The SWAT model was applied to a large basin (63 884 km(2)) and calibrated at a monthly time step with the SUFI-2 algorithm, using the Kling-Gupta efficiency (KGE) as the objective function. Multi-variable calibration was performed by combining streamflow and remote sensing-derived actual evapotranspiration data. Parameter transferability was also investigated, by daily time step validation. The KGE for the outlet ranged from 0.73 to 0.86, and the average KGE of all streamflow gauge stations ranged from 0.73 to 0.80, reflecting a good overall simulation performance for the monthly time step. In daily time step validation, KGE ranged from 0.62 to 0.68, and the Nash-Sutcliffe efficiency ranged from 0.40 to 0.60, for the average of all gauge stations. Multi-site and multi-variable calibrations did not significantly improve inner sub-basin simulation performance but improved streamflow uncertainty when compared to single-site calibration.