Global burned area mapping from Sentinel-3 Synergy and VIIRS active fires

• Published in: Remote sensing of environment 2022-12, Vol.282, p.113298, Article 113298
• Main Authors: Lizundia-Loiola, Joshua, Franquesa, Magí, Khairoun, Amin, Chuvieco, Emilio
• Format: Article
• Language: English
• Subjects: Active fires; Burned area mapping; Sentinel-3; Synergy; VIIRS
• ISSN: 0034-4257; 1879-0704
• DOI: 10.1016/j.rse.2022.113298

After more than two decades of successful provision of global burned area data the MODIS mission is near to its end. Therefore, using alternative images to generate moderate resolution burned area maps becomes critical to guarantee temporal continuity of these products. This paper presents the development of a hybrid algorithm based on Copernicus Sentinel-3 (S3) Synergy (SYN) data and Visible Infrared Imaging Radiometer Suite (VIIRS) 375 m active fires for global detection of burned areas. Using the synergistic and co-located measurements of OLCI and SLSTR instruments on board S3A and S3B, the SYN product offers global, near-daily surface reflectance data at 300 m for both sensors. Our algorithm relied on SYN shortwave infrared (SWIR) bands to compute a multi-temporal separability index that enhanced the burn signal. Active fires from the VIIRS sensor were used to generate spatio-temporal clusters for determining local detection thresholds. Active fires were filtered from those thresholds to obtain the seeds from which a contextual growing was applied to extract burned patches. The algorithm was processed globally for 2019 data to generate a new burned area product, named FireCCIS310. Based on a stratified random sampling, error estimates showed an important reduction of omission errors versus other global burned area products while keeping the commission errors at a similar level (Oe = 41.2% ± 3.0%, Ce = 19.2% ± 1.7%). The new FireCCIS310 dataset included 4.99 million km2 for the year 2019, which implied around 1 million more than the precursor FireCCI51 product, based on MODIS 250 m reflectance values. Temporal reporting accuracy was improved as well, detecting 53% of the burned pixels within a 0–1 day difference. Besides, the new product was much less affected by the border effects than FireCCI51, as a result of an improved active fire filtering process. The FireCCIS310 product is accessible through the CCI Open Data Portal (https://climate.esa.int/es/odp/#/dashboard, last accessed on July 2022). •A burned area algorithm based on S3 SYN data and VIIRS active fires is presented.•The algorithm was used to create a new global BA dataset at 300 m.•The new product estimates 1 million km2 more BA than previous global products.•Omission error of current MODIS-based BA products was improved.•The temporal reporting accuracy and border effects were greatly improved.