Hydropedological clustering: improving the representation of low streamflows in a semi-distributed hydrological model

Low streamflows are critical for sustaining water supply in Mediterranean regions, yet their simulation remains challenging due to the complex influence of soils on subsurface water storage and release. This study evaluates how different soil datasets and classification approaches affect the performance of the semi-distributed, physically based SWAT+ model in simulating low streamflows and soil water content (SWC). Using the Mediterranean Cauquenes catchment in central Chile, we compared two global soil datasets (HWSDv1.2, DSOLMap) and two locally derived products (CLSoilMapsTex, CLSoilMapsCl). The latter implements a new hydropedological clustering approach based on K_s[jls-end-space/], AWC[jls-end-space/], and α[jls-end-space/]. Results show that CLSoilMapsCl substantially improved low streamflow simulations (KGE_lf = 0.67, 44 % higher than HWSDv1.2) and reproduced hydrological signatures more accurately. These findings highlight that integrating hydropedological information enhances the representation of soil–water interactions in SWAT+, supporting more reliable low streamflow modeling and water-resource assessments in Mediterranean catchments.