Mapping climate change induced forest fire susceptibility using tree-based machine learning algorithms in Bangladesh

• Tree-based ML models analyzed forest fire susceptibility in CHT, highlighting biodiversity threats and climate risks. • AdaBoost (with highest AUC = 0.82) identified 4.31 % as highly and 66.74 % as moderately susceptible. • Temperature, precipitation, wind speed, and height were identified as key influencing factors. • Projections under RCP 8.5 indicate a 15.39% increase in high-susceptibility areas by 2080. • Findings support climate-adaptive fire management strategies and sustainable forest conservation. Forest fires in forested areas are an environmental disaster resulting from either natural phenomena or human activities. Forest fires pose a substantial risk in the forested and grassland areas of the Chittagong Hill Tracts. Therefore, mitigating the danger of forest fires is crucial for lessening their potential negative effects. This study employs tree-based machine learning (ML) algorithms to assess Forest Fire Susceptibility in the Chittagong Hill Tracts (CHT) of Bangladesh under current and anticipated climate change scenarios. Nineteen environmental and climatic variables were analyzed, with temperature, precipitation, wind speed, and height identified as the most relevant factors. This research analyses 267 historical instances of forest fires. Historical fire data and environmental attributes were utilized to train and validate eight ML models: Random Forest (RF), LightGBM, XGBoost, Decision Tree (DT), Gradient Boosting Machine (GBM), Adaptive Boosting (AdaBoost), Categorical Boosting (CatBoost), and Bagging. AdaBoost had the highest predictive accuracy (AUC: 0.82), classifying 4.31% of the study area as very susceptible, 66.74% as moderately susceptible, and 28.94% as having low Susceptibility. Future fire Susceptibility were projected under Representative Concentration Pathways (RCP 2.6, RCP 4.5, RCP 6.0, and RCP 8.5) for respective years of 2040, 2060, 2080, and 2100, indicating an increase in fire-prone regions, with high-susceptible areas expected to grow by as much as 15.39% under RCP 8.5 by 2080. These findings underscore the growing vulnerability of forested regions due to climate change and highlight the necessity of implementing proactive fire management strategies. The results of this research give significant novel insights for authorities in forestry administration and legislators, facilitating the development of climate-adaptive fire mitigation policies to strengthen environmental resilience and safeguard natural resources, while also supporting local governments in implementing sustainable forest resource management practices that improve the welfare of residents in Bangladesh.