Spatio-temporal patterns and drivers of fire spread and severity: a case study of an extreme wildfire in China’s boreal forests

Abstract Background Wildfire behavior results from nonlinear interactions among weather, fuel, topography, and human activities. Understanding the spatio-temporal characteristics of fire events is essential for gaining a deeper understanding of fire behavior and improving model predictions. However, real-time data collection on large wildfire events remains challenging, particularly for fires that occurred in earlier years. This study examines the 2003 "Jinhe 5·5 Fire" (105,834 ha) in China’ s Greater Khingan Mountains, a critical ecological barrier and wildfire-prone region, to identify fire behavior patterns and influencing factors. Methods We retrieved burn dates from the MCD14DL active fire data and quantified fire severity using the Differenced Normalized Burn Ratio (dNBR) derived from Landsat imagery. Due to the limited availability of data, semivariogram analysis was employed to assess the spatial heterogeneity of burn dates and fire severity. Furthermore, we used a random forest model to identify the key factors influencing fire severity. Results The main findings include: 1) The wildfire initially spread rapidly along roads from the ignition point in the northwest to the southeastern end of the burned area, later expanding perpendicularly (NE-SW). 2) Fire severity was highest on steeper slopes and in areas with rapid fire spread. 3) Pre-fire climate moisture and under-fire wind, as the main drivers of fire severity, were further influenced by human activities and topography. Conclusions The spatial pattern of fire severity is jointly determined by fire spread dynamics and environmental factors, with faster-spreading fires exhibiting higher severity. These findings provide valuable insights into fire spread dynamics and spatial variability, informing fire prevention and post-disaster recovery strategies.