Enhanced calibration of a distributed hydrological model in the Brazilian Semi-Arid: integrating spatiotemporal evapotranspiration and streamflow data

A hydrological model should be ideally calibrated and evaluated regarding both the representation of the streamflow and the intermediate hydrological processes. A challenging aspect is the poor spatial and temporal representativeness of in-situ measurements of these processes. This research evaluated the benefit of incorporating actual evapotranspiration (AET) estimates from two remote sensing (RS) data sources for calibrating a large-scale hydrologic model applied to a Brazilian Semi-Arid basin. Both spatially averaged and spatially distributed approaches were tested, and single-(only streamflow or only AET) and multi-variable calibration (streamflow and AET) procedures were used. Incorporating AET from MOD16 (MODerate-resolution Imaging Spectroradiometer product 16) did not help improve hydrological model performance independently of the calibration configuration. The introduction of AET estimates from GLEAM (Global Land Evaporation Amsterdam Model) as a calibration variable led the hydrological model to reproduce these AET values closer and simultaneously better represent the observed streamflow. The findings of this research highlight the benefit of incorporating AET as a calibration variable for improving the overall hydrologic model performance, which can even outperform the calibration based solely on streamflow. However, there is the need to test different calibration configurations and distinct RS data sources, which may result in very contrasting results.