Performance Evaluation of Burn Area Indices for Effective Fire Detection Using Sentinel-2 Satellite Imagery

In recent years, different spectral indices have been adapted or proposed for burn area (BA) extraction from satellite imagery. Many such indices have been particularly designed for specific satellite sensors, which could limit their applicability to other platforms. This research aims to explore the performance of spectral indices for burn area detection and post-fire recovery evaluation tasks in forest ecosystems. For this purpose, nine vegetation and burn area indices, commonly used in the current literature, were chosen to perform different experiments using Sentinel-2 images collected from three study areas characterised by large fire events. A separability analysis using the Spectral Discrimination index (SDI) led us to determine that A New Burned Area Index (ABAI), the Normalised Burn Radio Plus (NBR+), and the Normalised Burn Radio (NBR) indices were capable of discriminating burn areas when clouds and shadows were present in the imagery. Moreover, a short-term time series analysis allowed the identification of particular spectral index methods that could be useful for post-fire recovery evaluation in forest ecosystems.