The risk assessment of forest wildfires in Guangxi Province

Forests are the largest carbon sinks on Earth and play a vital role in protecting biodiversity, regulating the climate, and maintaining soil health as an essential part of ecosystems. In recent years, due to the influence of extreme weather conditions, forest fires have occurred more frequently, causing significant damage. Therefore, conducting wildfire risk assessments to provide scientific and reliable disaster prevention and management recommendations is crucial. Previous studies on forest fire risk have often overlooked seasonal characteristics, wildfire occurrence patterns, and have had low accuracy. This study addresses these issues by building a wildfire risk model that incorporates seasonal features and performs computational analysis of wildfire occurrences. The scikit-optimize framework was used to optimize the model for the best results. The findings are as follows: (1) Incorporating seasonal features into the model better reflects the real wildfire situations. (2) The spatiotemporal patterns of wildfires show clear seasonal characteristics, with wildfires being more frequent in spring and winter compared to other seasons. (3) After optimization, the performance of all models improved, with the best-performing model being the seasonal FFS (Forest Fire Risk) model built using BO-XGBoost. The model's accuracy rates were: spring (0.915), summer (0.913), autumn (0.921), and winter (0.901). Visualization of the model results indicated that special attention should be paid to wildfire risks in northern Baise, southern Chongzuo, and western Fangchenggang during spring, summer, and autumn, and in northern Baise during winter, to inform fire prevention strategies.