[A-CG34] Global Carbon Cycle Observation and Analysis

[ACG34-09] The burned area extracting in Chernobyl Exclusion Zone using random forest

*JUN HU¹, Shunji Kotsuki¹, Yasunori IGARASHI², Mykola TALERKO³, Kazuhito Ichii¹ (1.Center for Environmental Remote Sensing, Chiba University, Chiba, Japan, 2.Institute of Environmental Radioactivity, Fukushima University, Fukushima, Japan, 3.Institute for Safety Problems of Nuclear Power Plants, National Academy of Sciences of Ukraine, Kyiv, Ukraine)

Keywords:Wildfire, Chernobyl Exclusion Zone, Random forest, Satellite observation, Landsat, MODIS

The Chernobyl Nuclear Power Plant (CNPP) accident that happened in 1986 is the largest source of anthropogenic radionuclides released into the environment in history. In recent 20 years, the climate and land-use changes have increased the frequency of large forest fires in and around the Chernobyl Exclusion Zone. It is critical to extract the burned areas accurately, because they are the basis to estimate the biomass burning emission and then analyze the second diffusion of radioactive residue released from the CNPP accident. In this study, we established a burned area extracting method based on the random forest (RF) algorithm using the Moderate Resolution Imaging Spectroradiometer (MODIS) MOD09GA / MYD09GA and LANDSAT -7 ETM+ /-8 OLI images. The field observation in 2015 and MODIS MOD14A1 (thermal anomaly data) product were adopted to generate sampling points for RF. The reflectance difference spectroscopy of near-infrared band and difference in vegetation indices (NDVI, NBR, NDWI) between pre- and post-fire imagery were used as input data for the RF classifier. Subsequently, the historical burned area in 2015 and 2020 were detected using the trained RF classifier. The preliminary results of the identified burned area show good consistency with the MODIS MCD64A1.006 product of NASA and FireCCI51product of ESA. It should be noted that our RF algorithm can even detect the relatively small fire scars compared to the two existing products due to the usage of high-resolution LANDSAT image.