Extreme wildfire events are increasingly shaping Mediterranean fire regimes under the combined influence of climatic variability, vegetation stress, and growing anthropogenic pressure, posing critical risks to ecosystem stability and landscape resilience. Understanding the spatial determinants of wildfire susceptibility is therefore essential for advancing evidence-based fire risk assessment in fire-prone Mediterranean environments. This ongoing doctoral research investigates the environmental controls of wildfire occurrence in Muğla Province (Türkiye) through the integration of multi-source remote sensing data, geospatial analysis, and interpretable machine learning techniques. Burned-area reference data for 2021–2024 were derived from the MODIS MCD64A1 product within the Google Earth Engine environment and represented using a binary burned/non-burned classification. Predictor variables include ERA5-Land meteorological indicators, SRTM-derived topographic parameters, MODIS-based NDVI vegetation condition, and WorldPop population density as a proxy for human exposure, harmonized at a common 1 km spatial resolution. A Random Forest model was implemented to examine wildfire susceptibility patterns with emphasis on model interpretability and variable contribution rather than predictive optimization. Preliminary results indicate that vegetation condition, wind-related dynamics, and population density are dominant contributors to wildfire occurrence, reflecting coupled ecological vulnerability and human influence. Extreme fire conditions observed during 2021 are intentionally reserved for subsequent validation and stress-testing analyses. The proposed framework provides a transparent and transferable methodological basis for analyzing extreme wildfire susceptibility in Mediterranean landscapes and supports future development of interpretable, data-driven wildfire risk assessment approaches.