Hydrological calibration of the SWAT+ model using precipitation and evapotranspiration data derived from remote sensing in data-scarce basins

Abstract

The limited availability of observed hydrometeorological data, such as precipitation and streamflow, represents one of the main challenges for hydrological modeling and water resources management in river basins. This study evaluated the performance of precipitation data from the CHIRPS dataset for the Ribeirão Claro watershed (São Paulo State, Brazil) against in situ observations from two rain gauge stations over a 21-year period (2000–2020), using statistical metrics and extreme precipitation indices defined by the Expert Team on Climate Change Detection and Indices (ETCCDI). The semi-distributed hydrological model SWAT+ was calibrated and validated at a monthly scale using actual evapotranspiration data derived from the MODIS product, aiming to evaluate its applicability as an alternative to streamflow-based calibration. The results indicated that CHIRPS performs well in representing mean precipitation at monthly and seasonal scales. However, extreme precipitation events were underestimated, while months with low accumulated precipitation tended to be overestimated. The SWAT+ model, driven by CHIRPS data, satisfactorily reproduced the monthly MODIS evapotranspiration series, with NSE values ranging from 0.53 to 0.68, KGE from 0.62 to 0.80, and PBIAS from 3.28 to 4.53%. These findings highlight the potential of integrating remote sensing products for hydrological model calibration in data-scarce basins, thereby expanding the applicability of these tools in regions with limited hydrological monitoring.