Exploiting Medium-Resolution Sentinel Data in Google Earth Engine for Burned Area Reflectance Classification

Toanalyze and forecast changes triggered by a fire in a natural ecosystem, it is critical to look into the extent of specific traceable physical indicators, such as fire intensity, fire severity, and burn severity. Previous research has shown that Earth observation data plays an important role in quickly assessing burnt regions. Despite several technological constraints, especially related to geospatial and temporal resolution, these data enable the development of burned area reflectance categorization maps with a strong correlation with ground-truth fire severity. Thus, this research is aimed at investigating the damages induced by f ire in two Apulian municipalities (Bitonto and Vico del Gargano) by exploiting the potentialities of Sentinel images. To accomplish this purpose, a custom JavaScript code was created in the Google Earth Engine environment to calculate two indices: the Delta Normalized Burn Ratio (dNBR) and the Relativized Burn Ratio (RBR). The resulting thematic maps’ accuracy was assessed using both the ground truth given by the European Forest Fire Information System (EFFIS) and those obtained by applying a photointerpretation technique. Both indices performed well in extracting burnt regions using the two ground-truth datasets, even though EFFIS perimeters had low-quality categorization, indicating poordataaccuracy.Nonetheless,inbothsituations,RBRoutperformeddNBR, and hence may be suggested as a reliable method for swiftly quantifying burnt zones. Finally, the Javascript code could be easily transferred to handle data from multiple research zones based on geographical location and size. Google Earth Engine appears to be a feasible option for storing and processing a large amount of geospatial big data.