Forest Fire Severity Detection using AI

Forest fires pose significant threats to ecosystems and contribute to global warming. Assessing fire severity is crucial for effective firefighting and minimizing damage, typically relying on post-fire metrics. This study addresses real-time forest fire severity assessment, crucial for minimizing fire damage. We propose a method combining deep learning and weather data. Fire severity was then estimated by combining fire area (from image analysis), a visual fire distance rating, and the Fire Weather Index (FWI) from weather data. Various optimizing techniques are used on a formulated weighted average equation such as Genetic algorithm, Particle Swarm Optimization, and Quasinewton method (QN) for finding the optimized weights that minimizes the Mean Square Error (MSE) between assigned and calculated severity scores. Interestingly, the FWI received the highest weight, while the fire area received the least, highlighting the importance of weather data alongside image analysis.