Covered bridges represent a significant component of global cultural heritage, with a prevalence across Asia, Europe, and North America. Nevertheless, the current fire prevention strategies for these timber structures are notably insufficient, particularly in the critical area of fire risk assessment, where significant research gaps persist. In particular, timber covered bridges are confronted with considerable challenges in terms of fire resistance, due to the inherent low fire resistance of conventional timber materials. This issue is further compounded by the diverse structural types and complex environments of numerous timber covered bridges in southeastern China. This research collates and analyses statistical data on fire incidents involving timber covered bridges in ancient Southeast China. This is achieved by combining local gazetteers, transportation gazetteers, genealogies and various survey reports from different periods. By conducting a comprehensive cross-analysis of the natural and social environments surrounding these bridges and the causes of the fires, a historical fire prediction model is established using machine learning. This is done in an attempt to uncover the correlations between fire incidents and their environmental and human contexts. Ultimately, the research seeks to provide a theoretical foundation and essential insights for improving fire risk assessment methodologies for timber covered bridges in China.