Bayesian Networks and GIS-based analysis for flood risk assessment in agriculture during heavy rainfall events

This article covers the growing challenge of heavy rainfall events becoming more frequent and intense under climate change, increasing flood and damage risks for agriculture while localized impacts remain difficult to predict. It contributes a replicable methodology that integrates open government geospatial data with probabilistic risk modelling to assess agricultural flood risk from heavy rainfall, explicitly accounting for uncertainty. The approach combines GIS-derived exposure and susceptibility indicators-such as proximity to rivers, road density, and proximity to forests-with official datasets on rainfall intensity, temperature, land use, soil type and moisture, slope and elevation, and river discharge within a Bayesian Network using conditional probability distributions. The model is then tested through scenario, sensitivity, and optimization analyses, as well as validation exercises. The methodology was applied to the 2 June 2024 heavy rainfall event in the Rems-Murr district of Germany, where it produced qualitatively convincing flood risk results for agricultural areas. Validation against official reference data yielded RMSE values of approximately 23% in Rudersberg and 30% in Miedelsbach, suggesting reasonable performance and transferability across sub-regions. Overall, the article offers a structured, data-driven approach that disaster risk and agriculture stakeholders can adapt for risk screening, preparedness planning, and prioritizing mitigation measures in farming areas using largely open data sources.