Modeling framework shows the influence of human factors on wildfire ignition across Mediterranean areas, suggesting practical applications for risk assessment.
Key Points
This research aims to develop a modeling framework to compare wildfire ignition drivers across Mediterranean regions.
Utilized harmonized 500 m predictors from global remote-sensing datasets.
Integrated indicators like vegetation condition, topography, climate, and human pressure.
Applied tree-based machine-learning models (Random Forest and Extreme Gradient Boosting).
Employed spatial cross-validation and cross-region transfer experiments.
Anthropogenic pressure dominates ignition susceptibility across all regions.
Night-time lights and human modification indices are crucial for model importance.
Models achieved high predictive performance (AUC > 0.90) and stable accuracy in cross-region transfer (mean transfer AUC ≈ 0.85).