Geostatistical and Spatial Clustering Analysis for Wildfire Monitoring in Türkiye
Abstract
The wildfire situation has become increasingly dangerous to the eastern Mediterranean’s forest lands and human life. Among these are wildfires in Türkiye that have had one of the largest impacts on the country in the region. A pipeline consisting of five steps was applied using an integrated methodology to analyze MODIS Collection 6.1 data for the fires occurring in Türkiye during the three years of 2018 through 2020. The results were used to understand where, when and how much energy burned by wildfires throughout Türkiye. Five methods made up this process. The first method used to evaluate this information is FRP weighted Kernel Density Estimation (KDE) which provides the locations of highest concentration of fire in each year. The second method is Global Spatial Autocorrelation (Morans I) which examines whether fire occurs near other fires. Local Hotspot Detection (Getis-Ord Gi*) is the third method which identifies hotspots within individual grid cells. Interannual Difference Maps were produced using the fourth method to compare differences in FRP values from each year. Finally, the fifth method, Density-Based Spatial Clustering Analysis (DBSCAN), identified clusters of fire based on their geographic location. The southern central Anatolian belt (37°-38°N) continued to be identified as the area of greatest fire intensity, characterized by a large amount of FRP weighted density, statistically significant local hotspot concentrations (HS≥95%) in the same area in 2019 and was also located at the centroid of the DBSCAN cluster in all years studied. In addition to identifying areas with increased FRP, national average FRP values increased by about 22%, from 19MW in 2018 to 23MW in 2020. Additionally, the decline of 64% (from .233 to .083) in Global Morans I indicates an increase in both the intensity and dispersal of fire occurrence globally. The number of DBSCAN clusters increased from three in 2019 to seven in 2020; two additional clusters emerged in the northeast part of Anatolia and in eastern Thrace. Based on these findings it appears that there may be an early indication of changes in the structure of wildfires in Türkiye due to warming conditions in the eastern Mediterranean.