Evaluation of the Performance of a Finite Volume Physics-Based Model for Soil Erosion Simulation

Having reliable tools for characterizing rainfall-induced soil erosion is fundamental to the effective management of agroforestry systems in order to increase resilience against climate change. Physics-based models provide a robust, comprehensive and widely applicable framework to quantify runoff generation and soil erosion during intense rainfall events in agroforestry catchments. In this work, we propose a novel hydro-erosive model to simulate hydrodynamical flow and bed mobilization, movement and deposition. This hydro-erosive model solves the two-dimensional shallow water equations (SWE-2D) with hydrological source terms for runoff generation, coupled with the 2D depth-averaged solid transport and the soil surface evolution equations. The partial differential system is solved using a finite volume method. Alternative Integral/Differential Bed Slope and explicit upwind/implicit pointwise friction term discretization options can be used to improve performance in terms of numerical stability and conservation. The behavior of different discretization options in this hydro-erosive model is evaluated through an analytical hillslope verification, a benchmark V-catchment rainfall–runoff test and a laboratory dam-break experiment over an erodible bed. The results show that the Differential Bed Slope formulation combined with the upwind friction discretization provides the most accurate and conservative predictions. Also, an Upwind Bed Updating method for integrating soil surface elevation change is compared with the cell-centered integration of the bed change term by suppressing non-physical oscillations without compromising computational efficiency. Overall, the proposed open-source hydro-erosive model provides a reliable and computationally efficient framework for high-resolution simulations of rainfall-induced soil erosion and represents a valuable tool for environmental and agroforestry applications, but appropriate calibration and mesh resolution are required to ensure reliable predictions.