Explainable wildfire susceptibility assessment using fuzzy-evolutionary rules and Monte Carlo sensitivity analysis

Understanding the joint influence of geoenvironmental and anthropogenic variables on wildfire occurrence is crucial for developing predictive models that balance accuracy and explainability. Here, we introduce a wildfire susceptibility assessment framework that integrates fuzzy logic with multi-objective evolutionary algorithms (NSGA-II and ENORA) to derive interpretable rules describing how environmental variables are associated with wildfire events. Cohen's d effect-size analysis was used to identify key discriminative variables, while Monte Carlo–based sensitivity analysis (±10% perturbation, 95% confidence) assessed model responsiveness and rule uncertainty. Applied to a fire-prone region in southern Iran, the framework produced fuzzy rules linking wildfire likelihood to specific geoenvironmental and anthropogenic conditions. Effect-size analysis identified rainfall (d = 3.83), topographic wetness index (d = 2.86), slope (d = −2.18), distance to roads (d = 2.11), drought magnitude (d = −2.07), land use/cover (d = −1.79), soil type (d = −1.74), NDVI (d = 1.32), and distance to rivers (d = −1.03) as the most discriminative predictors, whereas curvature, valley depth, aspect, and distance to settlements showed limited influence. The Monte Carlo sensitivity analysis indicated that temperature (0.168), distance to settlements (0.136), drought magnitude (0.116), soil type (0.104), and distance to roads (0.102) were the most dynamically sensitive variables, suggesting that small perturbations in these parameters may substantially influence predicted fire susceptibility. Collectively, the results indicated that wildfire occurrence arose from nonlinear interactions among climatic, hydrologic, and anthropogenic drivers. While the proposed framework is fully reproducible and transferable in methodology, the derived fuzzy rules and calibrated parameters are context-dependent and needed to be recalibrated when applied to other regions or ecosystems.