Spatial-statistical analysis of forest fire distribution in the Republic of Serbia

Fires represent a significant ecological disturbance which affects the structure and dynamics of forest ecosystems, whereby their spatial distribution depends on the combination of climatic, vegetational and anthropogenic factors. The objective of this research is the analysis of the spatial distribution of fire detections in the territory of the Republic of Serbia, at the level of administrative units (municipalities) in the period from 2012 to 2026, restricted to areas classified as forest-related by the CORINE Land Cover classification (classes 2, 3, 4 and 5). The research was conducted using the data obtained through VIIRS sensors available through NASA FIRMS system of satellite data on active fires, in combination with CLC+ Backbone data. Also, Fire Radiative Power (FRP) was analyzed in order to estimate variability of the radiation power of fires and their classification. The analysis of FRP included 504 forest fire detections. The results show pronounced variability of radiation power of forest fire detections, whereby the majority of detections belong to the lower and interquartile ranges of the FRP distribution, while a smaller number of events reaches extreme values (up to 300 MW). For identification of spatial patterns, the methods of spatial statistics were used, including Moran's I index, as well as Getis-Ord Gi* statistics, which enables identification of local clusters with statistically significantly high (hot spot) and low (cold spot) values. The distribution of fire detections is characterized by pronounced spatial heterogeneity, whereby the majority of municipalities records a small number of fire detections, while in a smaller number of municipalities increased frequency occurs. Statistically significant spatial clusters are identified, with the most pronounced zones in southern and southeastern parts of Serbia. The results obtained indicate that the fire risk assessment should not be based solely on their frequency, but also on FRP values. The application of integrated approach, which combines satellite data on occurrence of fire detections and classification of land cover, enables reliable identification of risk areas and it represents the basis for improving the monitoring system and planning of fire protection measures.