Injecting vegetation-based spatialization in the hydrogeological framework for erosion modelling

Erosion processes and landslide are widespread across Italy and frequently cause significant damage to people, infrastructure, and ecosystems. These processes are primarily triggered by rainfall events, whose impact depends on multiple interacting factors, including geomorphology, soil properties, land use, and vegetation. Among these, vegetation plays an essential role in regulating hillslope hydrology by controlling interception, infiltration, and runoff. However, current hydro-geomorphological modelling framework lacks systematic methods to incorporate vegetation variability at the sub-grid scale, typically relying on coarse land cover (LC) classification and simplified class-based parametrization. This work presents a computational framework to explicitly incorporate sub-grid vegetation variability into soil erosion modelling, with a focus on its impact on Curve Number (CN) estimation for runoff and potential erosion assessment. Vegetation is represented as a spatially distributed categorical variable and modelled through an indicator-based geostatistical approach, allowing the propagation of spatial uncertainty into hydrological condition (hc) evaluation and CN assessment. The proposed approach integrates (1) a GIS-based pipeline for geospatial and watershed analysis, (2) a stochastic spatial model of vegetation based on point observations and structured mesh support, and (3) erosion modelling within a unified workflow. The methodology is designed to be transferable across different study areas and data availability conditions. The entire approach is tested on a real case study in Monte Pisano (Tuscany, Italy), an area affected by wildfire in 2018. Prior to erosion modelling, the stochastic-based vegetation modelling is validated under different sampling configurations to assess robustness and sensitivity. Results show that the stochastic framework produces erosion estimates broadly consistent with deterministic approaches, while enabling the explicit representation of sub-grid variability and associated uncertainty. Although a slight underestimation is observed, the probabilistic characterization enhances the interpretability of model outputs and supports more informed decision-making. By reducing reliance on subjectively defined parameters, as the hydrologic condition, adn embedding uncertainty into the modelling chain, the proposed framework represents a shift toward advancing environmental modelling practice, promising direction for further research in joining biodiversity and hydro-geomorphological applications.