Current model capabilities and future predictive pathways of wildfires under a changing climate
TL;DR
This Review compares land surface, Earth system, statistical, and machine learning model approaches that are used to project future fire activity across global and regional scales under multiple climate scenarios to show that different modelling approaches provide complementary strengths for distinct applications.
Abstract
Wildland fire activity is expected to change substantially under future climate warming, with important implications for ecosystems, air quality, and climate feedbacks. In this Review, we compare land surface, Earth system, statistical, and machine learning model approaches that are used to project future fire activity across global and regional scales under multiple climate scenarios. Despite broad agreement that fire risk will increase in many extra-tropical regions, the magnitude of projected changes remains highly variable and uncertain due to differences in climate sensitivity, vegetation responses, human influences, and fire process representation. Recent advances include improved fire-process representation, increased coupling of fire with Earth system processes, and emerging frameworks that integrate observations, process-based modelling, and machine learning. By synthesising results across model types, we show that different modelling approaches provide complementary strengths for distinct applications, and that the greatest confidence in future wildfire projections arises where independent modelling approaches converge. Historical model evaluation and multi-model intercomparison projects provide a critical basis for interpreting disagreements among projections and identifying priorities for improving next-generation wildfire models. Modelling future wildfires across the world is highly uncertain despite improvements in fire-process representation and the introduction of hybrid frameworks integrating observations, process-based modelling, and machine learning.