Enhancing seasonal fire predictions with hybrid dynamical and random forest models

Abstract This study presents an innovative approach to forecasting seasonal anomalies in burned areas (BA) by integrating process-based seasonal prediction and a random forest climate-fire model. The Standardized Precipitation Index (SPI), derived from observed precipitation, allows us to predict burned area anomalies a month before the start of the target fire season in ~68% of the burnable area. When utilizing seasonal predictions, the system maintains skillful results in ~46% of the burnable area. Given the availability of observational and forecast data in near-real-time, a prototype operational forecast for burned areas could be provided to enhance climate services.