Burn area mapping in Google Earth Engine (GEE) cloud platform: 2019 forest fires in eastern Australia

Abstract

Forest fires occur throughout the year in rainforests and deserts of Australia. The disastrous bush fire event occurred during November 2019, and lasted until February 2020, destroying more than 46 million acres of land. Burn area mapping is a major parameter in carrying out mitigation measures and regrowth activities by forest officials or fire managers post fire event. In this study, multi-temporal satellite datasets such as images acquired from Sentinel-2 (S2) and Landsat-8 (L8) missions are used to map the burn areas. Two thematic indices such as Differenced Normalized Burn Ratio (dNBR) and Relativized Burn Ratio (RBR) are implemented on the study area. The entire analysis, i.e., accessing the datasets, preprocessing, and calculation of indices for brunt area mapping is carried out on Google Earth Engine cloud platform. Rather than ground survey, the active fire product VIIRS product (VNP14IMGTDL) is used as a proxy for the actual fire indices in accuracy assessment. Results revealed that RBR showed better accuracy than dNBR for both the datasets (S2 and L8). S2 burn severity maps of dNBR and RBR showed better accuracy than L8 burn severity maps because of S2 having a higher spatial resolution. Thus, S2 datasets can be useful for rapid mapping of burn areas with improved spatial as well as temporal resolution.