Assessment of Burned Area Mapping Methods for Smoke Covered Sentinel-2 Data

Abstract

Wildfires become more frequent in the context of global warming and severe drought in several parts of the globe. Earth observation data can be used to provide information in such cases, but sometimes, when using optical satellite imagery, the evaluation of the effects produced by ongoing large scale forest fires, can be impeded by smoke. It can reduce the accuracy of the information required by disaster management authorities when allocating resources. To improve both the usability of optical remote sensing data and the quality of the obtained information we compare multiple feature extraction, classification, and visual enhancement methods and algorithms for land cover mapping of smoke covered Sentinel-2 data. The demonstration is performed for the 2019 forest fires in Australia.