Predicting and mapping fire susceptibility is a top research priority in fire-prone forests worldwide. This study evaluates the abilities of the Bayes Network (BN), Naïve Bayes (NB), Decision Tree (DT), and Multivariate Logistic Regression (MLP) machine learning methods for the prediction and mapping fire susceptibility across the Pu Mat National Park, Nghe An Province, Vietnam. The modeling methodology was formulated based on processing the information from the 57 historical fires and a set of nine spatially explicit explanatory variables, namely elevation, slope degree, aspect, average annual temperate, drought index, river density, land cover, and distance from roads and residential areas. Using the area under the receiver operating characteristic curve (AUC) and seven other performance metrics, the models were validated in terms of their abilities to elucidate the general fire behaviors in the Pu Mat National Park and to predict future fires. Despite a few differences between the AUC values, the BN model with an AUC value of 0.96 was dominant over the other models in predicting future fires. The second best was the DT model (AUC = 0.94), followed by the NB (AUC = 0.939), and MLR (AUC = 0.937) models. Our robust analysis demonstrated that these models are sufficiently robust in response to the training and validation datasets change. Further, the results revealed that moderate to high levels of fire susceptibilities are associated with ~19% of the Pu Mat National Park where human activities are numerous. This study and the resultant susceptibility maps provide a basis for developing more efficient fire-fighting strategies and reorganizing policies in favor of sustainable management of forest resources.

Wildfires are one of the most common natural hazards worldwide. Here, we compared the capability of bivariate and multivariate models for the prediction of spatially explicit wildfire probability across a fire-prone landscape in the Zagros ecoregion, Iran. Dempster–Shafer-based evidential belief function (EBF) and the multivariate logistic regression (LR) were applied to a spatial dataset that represents 132 fire events from the period of 2007–2014 and twelve explanatory variables (altitude, aspect, slope degree, topographic wetness index (TWI), annual temperature, and rainfall, wind effect, land use, normalized difference vegetation index (NDVI), and distance to roads, rivers, and residential areas). While the EBF model successfully characterized each variable class by four probability mass functions in terms of wildfire probabilities, the LR model identified the variables that have a major impact on the probability of fire occurrence. Two distribution maps of wildfire probability were developed based upon the results of each model. In an ensemble modeling perspective, we combined the two probability maps. The results were verified and compared by the receiver operating characteristic (ROC) and the Wilcoxon Signed-Rank Test. The results showed that although an improved predictive accuracy (AUC = 0.864) can be achieved via an ensemble modeling of bivariate and multivariate statistics, the models fail to individually provide a satisfactory prediction of wildfire probability (EBF_AUC = 0.701; LR_AUC = 0.728). From these results, we recommend the employment of ensemble modeling approaches for different wildfire-prone landscapes.