关键词: 驱动因子, Logistic回归, 预测模型, 火险区划,  贵州省

Abstract: The forest area of southwest China has complex fire sources and human interference, most of which are karst landform and in the ecotone of agriculture and forestry, leading to the most serious forest fire disaster area in China. The area has high mountains and steep slopes, making it extremely difficult to extinguish a fire once it occurs. It is of great significance to analyze the driving factors of forest fire in this region and to carry out fire risk zoning for rational forest fire management. Based on the forest fire point data, geospatial data, meteorological data, vegetation data and human activity data from 2011 to 2020, the spatiotemporal pattern of forest fire distribution in Guizhou in the past decade was analyzed, and the driving factors and probability prediction model of forest fire occurrence were obtained. The forest fire occurrence probability and forest fire risk zoning map of Guizhou in different seasons were analyzed. The results showed that in the past 10 years, the number of fire points in Guizhou showed a downward trend year by year. Fire points were mainly concentrated from January to March, accounting for 61% of the total number of fire points in a whole year. Distance from residential areas, distance from railways, population density, monthly average air temperature, monthly average relative humidity and monthly cumulative rainfall had significant effects on the probability of forest fire in Guizhou. A Logistic regression model was established. The prediction accuracy of the model was 81.9%, and the area under the curve was 0.904. The occurrence probability of forest fire in spring was higher than that in other seasons. The high-risk areas of forest fires in spring, autumn, and winter was mainly concentrated in western Guizhou, while the high-risk areas in summer were mainly distributed in eastern Guizhou. Research on the driving factors of forest fire occurrence and fire risk zoning maps based on the seasons is of great significance to the scientific management of forest fires in Guizhou. Forest fires in western Guizhou mainly occur in remote areas with a high probability of fire occurrence. Observation towers and video surveillance equipment should be added, patrol should be strengthened, and monitoring scope and timeliness should be improved. Fire prevention publicity and the control of human activities should be strengthened in eastern Guizhou in summer to reduce the probability of fire occurrence.

Key words: driving factor, Logistic regression, prediction model, fire risk rating, Guizhou Province