Epsilon Archive for Student Projects

Abstract

In May 2022, there was a mega-fire in the south-eastern part of ChEZ, caused by the Russian Invasion of Ukraine. It burned plantations established after the mega-fire in 1992 in the same area. Despite the mega-fire, some forest patches survived. A detailed analysis of forest management data from 2016 shows that the survived forest patches were comprised of broadleaved tree species with more than 80% of the canopy. This suggests that the dominance of broadleaves could be the factor that contributed to the survival of the forest patches (refugia). Consequently, this study investigates the characteristics of their surrounding forests that enhance the likelihood of these forests escaping large natural fires. We hypothesize that fuel discontinuity and structural heterogeneity in the buffers increase the probability of refugia occurrence. We used available local forest inventory datasets (geospatial polygons with attribute tables) from 2016 to map fire refugia and their adjacent areas (buffer zones) using QGIS and R. From the GEDI LiDAR footprints and Planet satellite images that were acquired before (2021) and after the mega-fire, we created wall-to-wall maps for canopy cover (CC), canopy height (CH), presence of broadleaves (POB) and forest cover (FC). Spectral bands were used as the predictor variables for predicting these attributes. CC map was used as a proxy for fuel discontinuity, and the POB map was used as a proxy for fuel type. We extracted the values of the variables from the maps for 10 m, 30 m, and 50 m buffers around survived and burned polygons. CH difference was calculated as the difference in CH between buffer and polygon, which was used as a proxy for structural heterogeneity. Using these variables as predictors, we developed Random Forest models for the 10 m, 30 m, and 50 m buffers that predict refugia occurrence using R. After analyzing the refugia occurrence with variable importance plots and partial dependence plots, we found that the fuel type, POB was the factor that saved some forests from the 2022 mega-fire. We could not find any evidence to support our hypothesis of structural heterogeneity and fuel discontinuity. It was observed that the decrease in the probability of refugia occurrence despite higher POB values was due to the adjacence of the non-refugia buffers to highly flammable Scots pine stands. These stands got burned, and the crown fire spread to surrounding patches and burned them down even if the POB values ranged between 75-81.5% in their buffers of 10 m width. This research recommends incorporating broadleaved species such as European oak, European aspen, and silver birch into the future production forest management regimes as they are fire-resistant and enhance forest resilience.