ABSTRACT
The increasing threat of high-severity wildfires in Mediterranean Wildland-Urban Interface (WUI) areas demands to develop effective fire risk assessment and management strategies. Simultaneously, the newfound accessibility of spaceborne hyperspectral data represents a significant potential for generating fire severity assessments, whereas National Forest Inventories (NFI) offer a vast dataset related to vegetation and fuel loads, which is essential for shaping the planning and strategies of forest services. This research work aims to advance the state-of-the-art in WUI fire risk mapping in the western Mediterranean Basin by combining PRISMA spaceborne hyperspectral data and Spanish NFI data. The proposed methodology had three main stages: (i) fire severity assessment at local scale (a wildfire) by using PRISMA hyperspectral data and Multi-Endmember Spectral Mixture Analysis (MESMA) leveraging field-based measurements of the Composite Burn Index (70 plots); (ii) development of a high fire severity probability map at regional scale from the extrapolation of a Random Forest predictive model calibrated from fire severity estimates, NFI data and topo-climatic variables at local scale (overall accuracy = 92 %; Kappa = 0.8); and (iii) identification and characterization of zones that concentrate WUIs with high probability of high fire severity if a fire event occurs (hot-spot WUIs) by crossing the information from the previous regional high fire severity probability map and a WUI cartography developed at regional scale. Study area was Castilla y León Autonomous Region (larger Spanish region, 94,226 km2), where the second-largest extreme Spanish wildfire event (28,000 ha) occurred. We identified hot-spot WUIs so that stakeholders and decision-makers could (i) prioritize resources and interventions for effective fire management and mitigation, (ii) allocate resources for prevention, and (iii) plan evacuation measures to safeguard lives and property. This study contributes to the development of next-generation fire risk assessment methods that combine remote sensing technologies with comprehensive ground-level datasets.
