Segment-level modeling of wildfire susceptibility in Iranian semi-arid oak forests: Unveiling the pivotal impact of human activities

Abstract

Iranian semi-arid oak (Quercus brantii) forests are characterized by their sparse canopy cover and an increasing risk of wildfire. In Lorestan Province, west of Iran (28,300 km²) wildfire incidents have surged, prompting this study to perform segment-level modeling of Wildfire Susceptibility (WS) distribution. The forest cover of the province was categorized into five density classes using an object-oriented classification method applied to a composite Landsat image, ranging from Forest Edge (with less than 10 trees per hectare) to Dense Forest (exceeding 150 trees per hectare). An ensemble of models provided by Biomod2 was utilized to generate the WS layer, revealing that the Normalized Difference Vegetation Index (NDVI) serves as a crucial indicator of WS. Dense, foliage-rich forests exhibited a heightened vulnerability to fire. Moreover, the study revealed that human activities near road networks and urban centers significantly shape wildfire patterns, accentuating the anthropogenic nature of wildfires in the central zone of the Lorestan Province. The findings pinpoint over 1600 km2 of highly susceptible areas, predominantly coinciding with dense and high-moderate forest classes, emphasizing the need for prioritized conservation efforts. The study recommends community involvement in forest protection and the promotion of alternative income sources as strategies to safeguard the diminishing forests of Lorestan.

Keywords

• Biomod2
• Landsat
• Wildfire susceptibility
• Object-oriented classification
• Iran