To analyze the driving factors of forest fire in Tongren city and study the fire risk distribution, so as to provide reference for forest fire management in the study area.

Taking the fire point data of Tongren City from 2001 to 2020 as the research object, this study investigated 22 factors including weather, terrain, vegetation, and human activities related to forest fire occurrence in Tongren City, the time distribution of fire point was analyzed, and the main driving factors of forest fire occurrence were obtained based on Logistic regression model and random forest model. A probability model was established, and the probability and fire risk zoning map of forest fire occurrence in Tongren City was drawn.

In the past 20 years, the number of fire points data in Tongren City showed a downward trend and a peak occurred every 3-5 years. Besides, more than 70% of forest fires were concentrated from January to April. Population density, monthly average humidity, monthly average precipitation and distance from railway were selected as the main driving factors by the two models, all showed negative correlation. The AUC value and prediction accuracy of each sample in both models were greater than 0.750 and 70%. The occurrence probability of forest fire in Tongren City was highest in spring and lowest in winter. The high risk areas were mainly concentrated in Shiqian county, Yinjiang Tujia and Miao Autonomous County and Songtao Miao Autonomous County. The low risk areas were Yanhe Tujia Autonomous County, Dejiang County and Sinan County in northwest of Tongren City and Yuping Dong Autonomous County in south.

Climate factors are the main driving factors of forest fire occurrence in Tongren City. The prediction probability of random forest is higher than that of Logistic regression model, which is more suitable for predicting forest fire occurrence in Tongren City. The probability of forest fire occurrence and fire risk zoning map obtained from the study provide scientific support for the forest fire management department in Tongren city. In high-risk areas, patrol management should be strengthened, observation towers and monitoring equipment should be increased. Fire prevention education and publicity should be added in low-risk areas, and fire management should be strengthened during holidays to reduce the probability of fire occurrence.