Abstract: Fire is a common natural disturbance in forest ecosystems and has extensive and profound impacts on ecosystem structure and function. It is important to examine the spatial-temporal pattern of forest fire and its spatial controls for a better understanding on the feedback among climate change, fire regimes, and forest ecosystems. In this study, three forest fire regime components (fire occurrence density, burned rate, and median fire size) from 2003 to 2016 in China were extracted based on the Global Fire Atlas dataset. Three models (Boosted Regression Tree, Random Forest and Support Vector Machine) were used to systematically analyze the impacts of climate, human activities, vegetation, and topography on forest fire regimes in China. Finally, the effects of climate change on future forest fire regimes were evaluated. The results showed that forest fires were spatially heterogeneously distributed in China, and fire occurrence density, burned rate and median fire size were higher in Northeast China than other regions. Different forest fire regime components were controlled by different factors, with climate as the dominant driver at national level. Climate warming will increase the frequency, size, and burned area of forest fire. The occurrence density, burned rate and median size of forest fire will increase by 39.4%-95.8%, 85.2%-143.0%, and 220.4%-206.6% under SSP126 and SSP585 scenarios. Future research needs to focus on evaluating the effects of climate factors on multiple fire regime components to gain a better understanding of future trends of forest fire regimes.

Key words: forest fire, climate change, Boosted Regression Tree, forest.