A multi-scale analysis of vegetation and soil burn severity drivers during the 2025 wildfire season in NW Spain

The 2025 wildfire season in Spain was characterized by numerous large wildfires, with 401,266 ha burned under extreme climatic conditions. Fifteen wildfires, covering 115,604 ha, predominantly in shrubland areas, were analyzed to assess the relative importance of climate, fire history, vegetation and topography in explaining vegetation burn severity (VBS) and soil burn severity (SBS). SBS was assessed only in areas with high VBS (dNBR > 0.66), where vegetation removal allowed the spectral response to be primarily driven by soil and ash. Vegetation and soil burn severity varied markedly among wildfires and vegetation classes. Shrublands exhibited significantly higher values of the differenced normalized burn ratio (dNBR) values, used to assess VBS, whereas no significant differences among vegetation classes were observed for the blue normalized difference index (BNDVI), used to assess SBS. A multiscale random forest approach was used to analyze burn severity at wildfire and pixel scales. At the wildfire level, VBS was mainly associated with antecedent drought conditions, fine fuel moisture, and previously burned area, whereas SBS was primarily associated with fire weather index on the day of ignition. At the pixel level, vegetation type, pre-fire vegetation greenness and topographic variables emerged as the main drivers of burn severity. The influence of these predictors varied among wildfire events, limiting the transferability of pixel-level models across independent wildfires and highlighting the importance of accounting for event-specific variability. Explicitly distinguishing between VBS and SBS improves post-fire assessment and supports the identification of areas at higher erosion risk, facilitating timely and targeted management interventions under increasingly extreme fire regimes.