Impact of Soil Data Representation on Streamflow Modelling: A Contiguous-Object Connectivity Approach Case Study in Switzerland

Joana Eichenberger, Bettina Schaefli, Tatenda Lemann

Hydrological models often rely on soil data to represent water partitioning, storage and release within a catchment. However, differences in soil data representation may be obscured by calibration when performance is evaluated primarily against streamflow. This study investigates how alternative soil datasets influence streamflow performance, internal hydrological processes, and parameter sensitivity in a Contiguous-Object Connectivity Approach (COCOA) Soil and Water Assessment Tool Plus (SWAT+) model of an agricultural catchment in Switzerland. Three otherwise identical model setups use a) a global low-resolution soil dataset (World-Soil), b) a national medium-resolution dataset (Swiss-Soil), and c) a high-resolution local soil dataset (Local-Soil). COCOA maintains identical spatial modelling units across scenarios, separating soil-data effects from changes in model discretisation. Before calibration, the three soil scenarios produce pronounced differences in water partitioning: World-Soil produces the greatest surface runoff, Swiss-Soil the greatest subsurface-flow contributions, and Local-Soil the highest evapotranspiration. After calibration, streamflow performance converges, with comparable relative Nash–Sutcliffe efficiency values across scenarios and higher Kling–Gupta efficiency for Swiss-Soil. Nevertheless, contrasting water balance partitioning, soil-water storage, and groundwater-release behaviour persist. Soil-data effects are also more apparent under median- and low-flow conditions than during high flows. These results show that similar outlet-streamflow performance can mask substantial differences in simulated catchment functioning. Soil datasets should therefore be selected according to the intended modelling objective, and internal hydrological diagnostics should complement conventional streamflow-performance metrics.