Assessment of Forest Damage in Croatia using Landsat-8 OLI Images

Background and Purpose: Rapid assessments of forest damage caused by natural disasters such as ice-break, wind, flooding, hurricane, or forest fires are necessary for mitigation and forest management. Forest damage directly impacts carbon uptake and biogeochemical cycles, and thus, has an impact on climate change. It intensifies erosion and flooding, and influences socio-economic well-being of population. Quantification of forest cover change represents a challenge for the scientific community as damaged areas are often in the mountainous and remote regions. Forested area in the western Croatia was considerably damaged by ice-breaking and flooding in 2014. Satellite remote sensing technology has opened up new possibilities for detecting and quantifying forest damage. Several remote sensing tools are available for rapid assessment of forest damage. These include aerial photographic interpretation, and airborne and satellite imagery. This study evaluates the capability of Landsat-8 optical data and a vegetation index for mapping forest damage in Croatia that occurred during the winter of 2014. Materials and Methods: The change detection analysis in this study was based on the Normalized Difference Vegetation Index (NDVI) difference approach, where pre- and post- event Landsat-8 images were employed in the ENVI image change workflow. The validation was done by comparing the satellite-generated change detection map with the ground truth data based on field observations and spatial data of forest management units and plans. Results: The overall damage assessment from this study suggests that the total damaged area covers 45,265.32 ha of forest. It is 19.20% less than estimated by Vuletić et al. [3] who found that 56,021.86 ha of forest were affected. Most damage was observed in the mixed, broadleaf and coniferous forest. The change errors of commission and omission were calculated to be 35.73% and 31.60%, respectively. Conclusions: Landsat-8 optical bands are reliable when detecting the changes based on the NDVI difference approach. The advantage of Landsat-8 data is its availability to acquire data and detect changes within a few days after an event. The data are publicly available and free of charge. The spatial resolution of 30 m is fine enough for a rapid assessment of forest damage. Merging different optical sensors (e.g. Landsat and Sentinel-2), or, considering active and/or thermal remote sensing satellite imagery would be necessary for monitoring damaged areas during winter time