Integrating deep learning with geospatial intelligence for real-time forest fire risk assessment

Climate change continues to intensify the frequency of forest fires and other natural disasters around the globe.Intense and recurring extreme weather conditions, particularly in sensitive areas such as the Mediterranean, which is highly susceptible to intense and frequent fires, amplify the frequency and severity of forest fires.This study applies an integrated approach combining machine learning (ML), remote sensing, and geographic information systems (GIS) to predict forest fire risk using the XGBoost algorithm.The model incorporates several factors including topographic features (slope, aspect, and elevation), meteorological features (relative humidity, temperature, dew point, solar radiation, wind speed, precipitation), anthropogenic influences (distance from urban centers, roads, and cultivated land), and the types of fuel available.The final output, the forest fire risk map, divides the region into three risk zones: high, moderate, and low, to enhance community and stakeholder participation and to advance preparedness measures in emergencies.The forecasting model demonstrated remarkable predictive capabilities, achieving an overall accuracy of 0.94 and an AUC of 0.99, underscoring its ability to improve fire management systems, response strategies, and emergency response readiness.