Wildfire Forecasting
Overview
Wildfire forecasting estimates fire occurrence, spread, size, intensity, and related impacts such as smoke, helping support preparedness, resource allocation, and evacuation decisions. Methods combine satellite observations with weather, terrain, vegetation, and fuel data using models such as CNNs and U-Nets, transformers, graph neural networks, and cellular automata; physics-guided and probabilistic models aim to represent fire dynamics and uncertainty. Evaluation spans satellite-based benchmarks, held-out fire seasons and regions, and comparisons with simulations, using measures of detection, spatial overlap, calibration, and false alarms. A central challenge is making forecasts reliable across places and changing conditions while aligning metrics with operational needs, especially when fires are rare and missed events and false alarms have different costs.
Latest papers
27 papers6 papers in the last four weeks, up 100% on the four weeks before. 0.0% of all new papers.