Building patterns and fuel features drive wildfire severity in wildland-urban interfaces in Southern Europe

Abstract

Fire danger analysis is crucial for landscape planning, which is particularly relevant in vulnerable areas such as the wildland-urban interface (WUI). The aim of our work is to investigate the capacity of fuel characteristics to predict burn severity in different WUI typologies classified according to nearby building density criteria. To achieve this goal, we selected 23 wildfires across Southern Europe, in which we differentiated non WUI areas, isolated, scattered, dense and very dense WUIs. Moreover, we spatialized burn severity and different fuel metrics using multispectral and radar satellite imagery. This information was used to analyze burn severity, and to identify its drivers in the different WUI typologies through analysis of variance, correlation analysis and machine learning models. Our results indicate that burn severity is lower in WUIs of clustered buildings than in non-WUI areas, which were also the most homogeneous in terms of vegetation cover. Moreover, we found that vegetation biophysical properties related to the amount of living fuel (fraction of vegetation cover and fraction of absorbed photosynthetically active radiation) showed the highest influence on burn severity in all WUI typologies. We also found that burn severity is less predictable in dense and very dense WUIs than in the rest, which can be attributed to their greater landscape complexity, presence of artificial structures and efficiency of extinction efforts. Our results serve to guide landscape management strategies, and impulse next-generation fire danger models, which should be based not only on synoptic fire weather indices but also on landscape-scale susceptibility to severe fires.