FireRisk-Zone-LR: A Logistic Regression-Based Wildfire Hazard Zoning Framework for Mediterranean Forests in Tangier, Morocco

Abstract

Wildfires increasingly threaten Mediterranean forests in northern Morocco, particularly around Tangier, where rugged terrain and expanding peri-urban zones amplify fire hazards. Existing risk maps often rely on expert-weighted overlays or coarse meteorological indices, limiting their usefulness for local-scale planning. This study develops an open-access, reproducible geospatial workflow that integrates satellite-derived indicators with logistic regression to assess wildfire susceptibility. We combined Sentinel-2 vegetation indices, SRTM topography, MODIS land-surface temperature anomalies, and burned-area masks to generate a dataset of 734 labeled pixels. The logistic regression model identified NDVI as a significant predictor of fire probability, while slope was not statistically significant. The model achieved an AUC of 0.72, demonstrating reasonable discriminatory power. Overall, the workflow shows how freely available satellite data and cloud-based tools can provide a scalable and transparent approach to fire-risk assessment in Morocco and other Mediterranean forest systems.