Forest fires significantly impact on forest ecosystems,causing extensive ecological damage and significant economic losses. This underscores the urgent need for accurate and reliable forecasting models to support effective forest fire prevention and management strategies. This study compares the performance of Logistic Regression and Spatial Generalized Additive Models in predicting forest fire occurrences and accessing fire risk levels,thus providing a solid scientific basis for informed decision-making in fire prevention efforts. Using comprehensive forest fire data from Heilongjiang Province,collected over a 15-year period (2006-2020),the study integrated multiple influencing factors,including meteorological conditions,topography,and vegetation characteristics,to assess the predictive capabilities of the Logistic Regression Model and four different basis functions of the Spatial Generalized Additive Models. The SGAMs tested in this study included models with Gaussian process smooth spline (GP),Cubic regression spline (CR),Thin plate regression spline (TP),and Adaptive smooth spline (AD). The results clearly demonstrated the significant advantages of Spatial Generalized Additive Models over the conventional Logistic Regression Model,with Spatial Generalized Additive Models consistently achieving superior model fit and predictive accuracy. Among these models,the Adaptive smooth spline (AD) model exhibited the best overall performance,with a 4.2% increase in test set accuracy and a 0.053 improvement in the area under the curve (AUC) compared to its counterparts. Further spatial analysis identified high-risk forest fire areas in Heilongjiang Province,primarily concentrated in the northwest and central-southern regions. These findings closely align with the province's existing fire prevention layout,underscoring the practical relevance of Spatial Generalized Additive Models-based models for real-world applications. Notably,the study highlights the critical role of spatial data in forest fire forecasting. By incorporating spatial data,the understanding of fire risk patterns across diverse landscapes is significantly enhanced,enabling region-specific prevention strategies and optimized resource allocation. Additionally,the unique advantages of the AD model stand out. Its ability to provide segmented linear insights into the effects of influencing variables allows for a more nuanced interpretation of how factors such as meteorology,topography,and vegetation interact to influence fire risk. This capability supports the development of more precise and targeted fire prevention measures. In conclusion,this research underscores the potential of Spatial Generalized Additive Models,particularly the AD model,as highly effective tools for advancing forest fire management. By adopting these advanced models,regions like Heilongjiang Province can strengthen fire prevention strategies,better protect forest ecosystems,and achieve sustainable forest management goals while minimizing the risks of future catastrophic disasters.