https://orcid.org/0000-0001-8680-1244
![Sample our Environment & Agriculture journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5934/TOC-Subject-Sample-2021-Environment-Agriculture.jpg"})
ABSTRACT
Coarse spatial resolution of remote sensing imagery still hampers a comprehensive representation of long-term fire patterns at the regional level, in particular in areas characterized by small and sparse fire scars. The Visible Infrared Imaging Radiometer Suite (VIIRS) sensor launched in 2011 upgrades the spatial resolution (375 m) and gives continuity to the Earth long-term monitoring initiated by Advanced Very High-Resolution Radiometer (AVHRR) and Moderate Resolution Imaging Spectroradiometer (MODIS) sensors. Therefore, aiming to assess VIIRS 375 m imagery capabilities to improve the accuracy and reliability of fire scars mapping over the Brazilian Cerrado, we developed a burned area detection algorithm (VIIRS-SVM) based on machine learning techniques. For this purpose, the (V, W) burnt index adjusted to VIIRS near-infrared and middle-infrared channels and the One-Class Support Vector Machine algorithm were used for burned area identification. The VIIRS-SVM algorithm was applied over the Brazilian Cerrado and evaluated against reference scars from 15 Landsat-8 scenes during the fire season of 2015, covering a large area with substantial variability in terms of fire scars characteristics. We also performed a comparison with the MCD64A1 collection-6 product over the validation sites. Relying on VIIRS 375 m imagery, the VIIRS-SVM algorithm allows an enhancement of 25% in discrimination of small and medium fire scars (25 to 1000 ha), when compared to the MODIS-derived product. Results have demonstrated that the enhancement of medium and small fire scars mapping over the Cerrado is possible using VIIRS sensor capabilities.
Disclosure statement
No potential conflict of interest was reported by the authors.