Fire exposure and vulnerability analysis in a wildland urban interface of Northern Sardinia, Italy

The goal of this study is to analyze fire probability and vulnerability of urban settings in a study area of about 46,000 ha located in a coastal area of North-East Sardinia. The study was carried out in a fire-prone area characterized by predominant and dense shrubland vegetation, by extensive rural-urban interfaces and by a relevant touristic pressure in the summer season. The analysis was performed using both the characterization and mapping of housing units, and the modeling of fire behavior by a probabilistic approach. The first part of the work was based on the analysis of the aerial photographs and field surveys. With this approach, we evaluated the vulnerability to fire of about 13,000 housing units through the evaluation and characterization of five factors: type of vegetation, percentage of the housing units in contact with the vegetation within a buffer area of 10 meters, presence of flammable structures, distance of flammable structures from the houses, and percentage of the flammable structures in contact with the vegetation, within a buffer area of 10 meters. The results highlighted that about 58% of the housing units were located in urban settings, while the remaining 42% were classified as rural-urban interfaces. About the latter, only 25% of units showed a reasonable distance from the surrounding vegetation, and 30% of units presented more than half of their perimeters in contact with the vegetation; in addition, 7% of units were close to highly burnable structures. Afterwards, we used simulation modeling based on the MTT algorithm of FlamMap in order to estimate the spatial variation of the main fire probability and severity indicators. Then we selected 32 different scenarios able to represent the most common and critical conditions for fire occurrence and spread in the study area. In more details, we combined two wind directions (210° and 290°), two wind speeds (32 and 20 km h-1), two fuel moisture conditions (arid and extreme), two ignition patterns (historically based and random), and two active fire spread durations (1 and 3 hours). For each scenario, we simulated 10,000 fires. Overall, the analysis of the fire simulations showed that fuel moisture and ignition patterns were the main factors affecting burn probability. Furthermore, the combination of extreme fuel moisture and strong wind speed represented the most critical scenario. The maps of conditional flame length were strongly affected by vegetation types and topography, particularly in some specific areas and nearby the coastline. Also the fire size maps highlighted hot-spot areas able to originate very large fires, particularly when wind direction and slope effects were compounded. We finally combined the results of the WUI mapping and characterization with the fire simulations, and we identified and quantified the level of exposure and risk of the housing units. As expected, the houses surrounded by shrublands and complex topography, particularly if isolated, showed very high fire exposure profiles, while the lowest exposure was observed in compact villages and towns.