Projecting future wildfire risk in California from changing climate and vegetation composition

Changing wildfire regimes are of key concern for human health, ecosystem function, biodiversity, and climate mitigation. In California, a rapid increase in burned area over recent years has resulted in substantial release of carbon to the atmosphere, loss of human life, and economic costs in the billions of dollars per year. Understanding how burned area may change in the future is critical for informing land management and carbon policies, and for protecting human health and vulnerable ecosystems. In this study we used machine learning to investigate the human and environmental factors controlling historical (1990-2021) fire occurrence in California, and made projections for future expected annual burn probability for 2081-2100 under different climate and vegetation scenarios. We found that rising temperature and changing precipitation patterns could increase fire risk by approximately 35-50%, with particularly high future risk in scenarios with increased precipitation and shrub cover. Our results highlight the important role of precipitation as a driver of future risk in addition to temperature, with wetter climate scenarios leading to largest fuel loads. We also uniquely identify a potential positive fire-vegetation feedback, as fire drives conversion of forests to shrublands and further increases subsequent fire risk. Spatially, we found increased fire risk particularly in the northern coastal forests of California, which are currently adapted to mild climates and also host several forest carbon offset projects. These findings regarding risks, drivers, and uncertainties could directly inform policies to manage fuels, protect carbon stocks, and set realistic mitigation goals in the face of a changing climate.