Remote sensing for assessing the impact of forest fire severity on ecological and socio-economic activities in Kozan District, Turkey

Monitoring the ecological and socioeconomic impacts of wildfires using traditional approaches requires significant financial resources, time, and sampling expertise. However, not only are resources scarce, but the spatial and temporal extent of forest fires can also make it impractical to assess large areas over time. Thus, fire monitoring initiatives are often not realized. This has inevitably made the remote sensing approach an interesting option for fire protection managers and decision-makers due to its ability to measure large areas and its temporal capabilities. In this study, burn spectral indices derived from Landsat 8 (difference normalized vegetation index (dNDVI) and difference normalized burn ratio (dNBR)) were used to assess the ecological and socioeconomic impacts of forest fires based on an existing land use/land cover dataset. The relationships between estimated fire severity/area and environmental and anthropogenic factors were also evaluated. The results show that more than 700 hectares of forest and other land use categories were burned. Fires adversely affect high forests, thickets, degraded forests, and most cultivated and rural areas. The study also revealed a moderate positive relationship between burn severity and pre-fire vegetation (R2 = 0.48 and R2 = 0.49 for the dNDVI and dNBR, respectively). This result suggested that the fuel amount is the main driver of burn severity during the fire season in this particular ecosystem. Topography has been shown to affect fire behavior in the study area, where fires occur primarily at elevations averaging 400-800 meters above mean sea level. In contrast, there is a weak positive relationship between population density and burnt area. This phenomenon is commonly observed in specific regions, where the incidence of fire is directly proportional to the density of the population. However, the severity decreases when burning exceeds a threshold. This study has shown that Landsat 8 data-derived burn spectral indices (dNDVI and dNBR) have high potential for the spatial analysis of wildfires.