Reliable wildfire projections and associated socioeconomic risk assessments are crucial for developing effective adaptation and mitigation strategies. This study presents a machine learning framework to constrain future fire carbon emissions simulated by 13 Earth system models from CMIP6 using historical observations. The observation-constrained ensemble projects a weaker increase in global fire carbon emissions but a higher increase in global wildfire exposure across population, GDP, and agricultural areas compared to the default ensemble. Elevated socioeconomic risks stem from the combined regional enhancement of future wildfire activity and socioeconomic development in western and central Africa, highlighting the urgent need for strategic wildfire preparedness in these regions.

Yan Yu, Jiafu Mao, Stan D. Wullschleger, Anping Chen, Xiaoying Shi, Yaoping Wang, Forrest M. Hoffman, Yulong Zhang, Eric Pierce