Data source effects on forest fire prediction models in Northern China under multiple climate scenarios

This study focuses on the Greater Khingan Range in Inner Mongolia, a typical representative area of northern China's forests. It integrates dual-source fire point data from manual records and remote sensing to analyze the seasonal variations in the occurrence of forest fires influenced by various factors, including meteorology, topography, and vegetation, from 2001 to 2022. Additionally, the study further predicts the trends of fire risk evolution under different climate scenarios (SSP126, SSP245, SSP585) for the near-term (2021-2040), mid-term (2041-2060), and long-term (2081-2100).The results indicate that: (1) Both data sources demonstrate that the BRT model exhibits the best predictive performance throughout the year, while the XGB model performs best in summer, with notable seasonal differences in model effectiveness; (2) The manual data highlights the dominant role of meteorological factors, with high-risk areas located in the eastern and northern parts of the study area; remote sensing data provides a more comprehensive reflection of the combined effects of topography and vegetation, identifying high-risk zones primarily in the east, while the north is predominantly low-risk; (3) Under various future climate scenarios, the proportion of high and extreme fire risk areas is expected to decline, while the proportion of moderate risk areas is projected to increase.This research provides valuable insights for precise fire prevention and control in the region and serves as a reference for adaptive fire management in northern forests globally.