Chapter 7 - Google Earth Engine and machine learning classifiers for obtaining burnt area cartography: a case study from a Mediterranean setting

Abstract

Climate change is linked to the increase of natural disasters occurrence over the world. The increasing frequency of wildfire events constitutes a threat toward ecology, economy, and human lives. Hence, accurate information extracted from detailed burnt area cartography plays a primary role in the ecosystems’ preparation. Geoinformation has enabled rapid and low-cost Earth Observation data acquisition and their processing in cloud-based platforms such as Google Earth Engine (GEE). In the present study, a GEE-based approach that exploits machine learning (ML) techniques and ESA’s Sentinel-2 imagery is developed with the purpose of automating the mapping of burnt areas. As a case study is used an area located in the outskirts of Athens in Greece, on which it was occurred one of the largest wildfires in the summer of 2021. A Sentinel-2 image, obtained from GEE immediately after the fire event, was combined with ML classifiers for the purpose of mapping the burnt area at the fire-affected site. Validation of the derived products was based on the error matrix and the Copernicus Rapid Mapping operational product available for this fire event. Our study results clearly demonstrated the importance of rapid and accurate burnt area mapping exploiting sophisticated algorithms such as ML, when applied to Sentinel-2 imagery. Our findings provide valuable information regarding random forests and support vector machines classifiers through comparisons during their implementation as well as the potential of automated mapping in cloud-based platforms such as GEE. All in all, burnt area mapping is very valuable toward preparation activities regarding postfire management in future fire incidents.