Integrating remote sensing and GIS for advanced forest fire hazards vulnerability and risk assessment in Gironde, France

Abstract

Abstract Background Forest ecosystems play a crucial role in maintaining biodiversity and controlling the climate. In recent years, these systems have been increasingly threatened by forest fires, particularly in Mediterranean regions. Traditional fire prevention and suppression methods often lack precision, adaptability, and timely updates due to the complex and diverse nature of forest ecosystems. To address this, we used a geospatial modeling methodology modified from techniques already used in comparable contexts to create a forest fire risk map for the Gironde region in southwestern France. We employed a methodological framework derived from the fire risk assessment model first proposed by Erten et al. for Turkey, which is particularly suitable for Mediterranean-type ecosystems, to identify the regions in France most susceptible to forest fires. Several geospatial datasets must be integrated in this model: The characteristics of vegetation and land cover were determined using Sentinel-2 satellite imagery; slope and aspect information were obtained from digital elevation models (DEM), road networks and human settlements were extracted using OpenStreetMap (OSM) data; and the spatial layers were validated using recorded fire locations from the Fire Information for Resource Management System (FIRMS). To create a composite forest fire risk, these thematic layers were processed in a GIS context and weighed based on their proportional influence on fire incidence. To make sure that the findings consider both current circumstances and the seasonal variations of fire threats, the analysis focused on the summer fire season throughout the 2020–2022 timeframe. Results The results of this study revealed a wide range of fire risk levels, with 52% of the study area categorized as high to very high risk. Increased human presence, busy road networks, steep south-facing slopes, and flammable flora were all highly linked to these risks. The risk map was validated by using recorded fire locations from the Fire Information for Resource Management System (FIRMS). The model showed an overall accuracy of 0.84. Additionally, the area under the ROC curve (AUC) was 0.753, confirming the model’s effectiveness in predicting vulnerable zones. Conclusions This spatial risk assessment offers important information for Gironde’s forest fire management. These findings can be used by local authorities, planners, and fire departments to improve emergency response, prioritize intervention zones, and strengthen forest protection tactics. The approach shows how geospatial tools can make a significant contribution to adaptive fire risk mitigation and applied fire research.