Wildfire risk in a changing climate: Evaluating fire weather indices and their global patterns with CMIP6 multi-model projections

Abstract

This study investigates potential wildfire risks across different global warming scenarios through a comparative analysis of two prominent fire weather indices: the McArthur Forest Fire Danger Index (FFDI) and the Canadian Forest Fire Danger Index (FWI), leveraging the latest multi-model projections from the 6th phase of the Coupled Model Intercomparison Project (CMIP6). Utilizing the Extreme Gradient Boosting (XGBoost) algorithm and the Shapley value, we identify the impacts of meteorological variables on fire weather danger as represented by FFDI and FWI. Our findings reveal that under the Shared Socioeconomic Pathways (SSP) 5–8.5 high-emission scenario, both FFDI and FWI project significant intensification of fire weather, particularly in historically recognized high-risk hotspots, demonstrating robust inter-model consistency. Notably, the future projections of FFDI indicate the likely occurrence of wildfires with unprecedented severity. The comparative analysis using Shapley values highlights substantial regional and index-specific variations in the contribution of meteorological input variables to fire weather simulations. While these global patterns are generally retained as global warming leads to a systematic reinforcement of all variables, in-depth regional scale analyses further uncover a stark contrast of dominant factors controlling FFDI and FWI. These findings stimulate discussion on the potential adaptability and discrepancies of empirically derived fire models, highlighting the need for future research to advance fire weather modeling with enhanced flexibility and multi-factor consideration.