A forest fire susceptibility map generated with the fire susceptibility model is the basis of fire prevention resource allocation. A more reliable susceptibility map helps improve the effectiveness of resource allocation. Thus, further improving the prediction accuracy is always the goal of fire susceptibility modeling. This paper developed a forest fire susceptibility model based on an ensemble learning method, namely light gradient boosting machine (LightGBM), to produce an accurate fire susceptibility map. In the modeling, a subtropical national forest park in the Jiangsu province of China was used as the case study area. We collected and selected eight variables from the fire occurrence driving factors for modeling based on correlation analysis. These variables are from topographic factors, climatic factors, human activity factors, and vegetation factors. For comparative analysis, another two popular modeling methods, namely logistic regression (LR) and random forest (RF) were also applied to construct the fire susceptibility models. The results show that temperature was the main driving factor of fire in the area. In the produced fire susceptibility map, the extremely high and high susceptibility areas that were classified by LR, RF, and LightGBM were 5.82%, 18.61%, and 19%, respectively. The F1-score of the LightGBM model is higher than the LR and RF models. The accuracy of the model of LightGBM, RF, and LR is 88.8%, 84.8%, and 82.6%, respectively. The area under the curve (AUC) of them is 0.935, 0.918, and 0.868, respectively. The introduced ensemble learning method shows better ability on performance evaluation metrics.

Forest fires create burned and unburned areas on a spatial scale, with the boundary between these areas known as the fire boundary. Following an analysis of forest fire boundaries in the northern region of Yangyuan County, located in the Liangshan Yi Autonomous Prefecture of Sichuan Province, China, several key factors influencing the formation of fire boundaries were identified. These factors include the topography, vegetation, climate, and human activity. To explore the impact of these factors in different spaces on potential results, we varied the distances between matched sample points and built six fire environment models with different sampling distances. We constructed a matched case-control conditional light gradient boosting machine (MCC CLightGBM) to model these environment models and analyzed the factors influencing fire boundary formation and the spatial locations of the predicted boundaries. Our results show that the MCC CLightGBM model performs better when points on the selected boundaries are paired with points within the burned areas, specifically between 120 m and 480 m away from the boundaries. By using the MCC CLightGBM model to predict the probability of boundary formation under six environmental models at different distances, we found that fire boundaries are most likely to form near roads and populated areas. Boundary formation is also influenced by areas with significant topographic relief. It should be noted explicitly that this conclusion is only applicable to this study region and has not been validated for other different regions. Finally, the matched case-control conditional random forest (MCC CRF) model was constructed for comparison experiments. The MCC CLightGBM model demonstrates potential in predicting fire boundaries and fills a gap in research on fire boundary predictions in this area which can be useful in future forest fire management, allowing for a quick and intuitive assessment of where a fire has stopped.