Multi-spatial and multi-criteria calibration to guarantee a robust SWAT+ hydrological model in a large and heterogeneous catchment

• A multi-spatial and multi-criteria zonal calibration methodology was developed. • It was applied to a SWAT+ model of the Upper Tagus River basin, a complex one. • Streamflow performance and its components were evaluated in natural regime areas. • Model optimization required defining from 4 to 11 geological calibration regions. • A satisfactory and realistic streamflow simulation was achieved in most of the areas. Calibration is essential to ensure hydrological models’ reliability. While performance metrics for simulated variables are often used as the primary indicators of model adequacy, it is equally important to verify that the underlying hydrological processes are realistically represented. This consideration becomes even more critical in complex basins characterized by heterogeneous climatic or geological conditions, where processes variability may necessitate the use of zonal calibration approaches. This study presents a multi-spatial and multi-criteria calibration of a SWAT+ model for the Upper Tagus River basin, a large and geologically diverse catchment. This approach, which includes a sensitivity analysis, has been designed to ensure a realistic simulation of hydrological processes in geologically diverse areas. The daily streamflow simulation performance and its components have been evaluated at 22 gauged areas, as inflows to 12 reservoirs, collectively covering 44 % of the modelled area (all under natural flow regimes). The optimization of the model required a detailed zonation based on geological properties, and a comprehensive validation process, including a fine-tuning, was necessary to obtain a set of parameters resulting in an accurate (considering performance metrics) and realistic (considering the streamflow components) simulation in most of the areas evaluated. The global performance (evaluated aggregating streamflow of the different areas considered) demonstrated a very good daily streamflow simulation performance (NSE > 0.8, PBIAS < 5 %) and an accurate simulation of the daily streamflow components. The developed calibration approach, coded in R, can be effectively applied to other hydrologically complex regions.