Optimising future scenarios of forest fire occurrence in Daxing'anling using long-term survey data and intelligent modelling

The Daxing'anling Mountains, as a climate-sensitive region, are experiencing forest fires that threaten the area's ecological security. Nevertheless, most of the existing fire prediction models are stationary. They do not have an all-embracing scheme for simultaneously managing fire ignition causes, dynamic fire scenarios and spatial targeting. Hence, the development of an accurate and efficient forest fire forecasting system is vital. This study establishes a prediction framework that integrates long-term survey data with multi-source remote sensing, incorporating spatiotemporal clustering, spatial autocorrelation and an optimised ensemble of LR–RF–SVM–GBDT algorithms. Among the 3368 recorded fire incidents, lightning-ignited fires accounted for 51.19%, making lightning storms the predominant cause of ignition. While the frequency of lightning-induced fires increased significantly (1.24 per year, p < 0.05), the total burned area remained relatively stable. The proposed framework outperformed individual models by achieving higher predictive metrics (accuracy = 0.89, AUC = 0.94, F1 = 0.89) and providing robust support for operational early warning and real-time management. The projections for future climate, based on the SSP126 and SSP585 scenarios, depict a notable geographical shift in fire-prone areas. Besides the traditionally known eastern areas of Xiaogenhe and Chabanhe, which are expected to see an increase in fire occurrences, new high-fire-risk areas are expected to emerge in the central–western regions, such as Huzhong and Wuyuan. Quantitative findings reveal that the divergence in forest fire probabilities between the high-emission SSP585 and SSP126 scenarios will increase over time. The expected increase ranges from 0.29% in the 2030s to 0.92% in the 2050s, then rises to 4.48% in the 2070s and reaches 6.48% by the 2090s. These figures highlight the urgency of implementing fire management practices that are not only adaptive but also specific to particular areas. The scenario-based forecasts represent a proactive approach to assisting forest fire governance under climate change, providing a basis for future decisions as quantitative evidence.