Accurate prediction of forest fires is crucial for enhancing regional fire prevention and control. Existing models frequently rely on static factors such as weather and terrain, while insufficiently taking into account the Fuel Moisture Content (FMC), a critical internal factor that directly determines fire behavior. Instead, proxies like the Normalized Difference Vegetation Index (NDVI) are commonly employed, which weakens the physical foundation of predictions. This study assesses the marginal contribution of integrating dynamic FMC into fire prediction models. Concentrating on California, we developed a random-forest-based model that incorporates high-resolution FMC products retrieved by our team, along with meteorological, topographic, vegetation, and anthropogenic data. Through comparative experiments and SHapley Additive exPlanations (SHAP) analysis, we evaluated model improvements and the contribution mechanisms of key drivers. The results indicated that: (1) Incorporating FMC significantly enhanced model performance, with precision and specificity increasing by 3.93% and 3.60%, respectively, and the Area Under the Curve (AUC) showing improvements, suggesting heightened sensitivity in detecting actual fire occurrences. (2) SHAP analysis disclosed nonlinear effects and threshold dynamics: temperature was the dominant positive driver (the fire risk soared above 20 °C); FMC demonstrated a negative correlation with fire risk, with 100% serving as a potential threshold; elevation presented an inverted U-shaped pattern (the peak risk occurred at 1000–1500 m); and population density exhibited a shifting influence from positive to negative. (3) The monthly risk maps for California in 2023 captured the seasonal progression of fire risk and spatial patterns consistent with historical fire points. The fire risk map for 9 September 2020 also demonstrated consistency with the spatial distribution of the actual fire points on that day. This study validates that the integration of dynamic FMC strengthens the mechanistic foundation and early-warning capacity of fire prediction models, providing scientific backing for targeted fire management.
Li et al. (Tue,) studied this question.