Building NBR time series with cloud-free Landsat 7 and 8 images to determine burned area at subcontinental scale

SIONE, W.; ARMANDO RODRIGUEZ ; JESUS ADOLFO ANAYA ACEVEDO

Santiago

Conferencia; ForestSAT2016; 2016

Universidad Mayor Campus Manuel Montt

Large discrepancies have been found for burned area estimation using remote sensing data with data sets of medium resolution (500 m ? 1 km) as GBA2000, GlobScar, WFA, MCD64A1. Several aspects of remote sensing technology as spatial and temporal resolution has been identified as responsible for such inaccuracy. The present work considers the uncertainty generated by changes in reflectance at TOA before and after the fire using a 30 m pixel. TOA reflectance is affected by the AOD and vegetation phenology of each ecosystem, which in turn generates unique per pixel conditions to determine detection thresholds of burned area. Time series of Landsat reflectance bands and NBR were analyzed for burned areas at different ecosystems (natural forest, grasslands, and shrublands) and different regions of Colombia, Bolivia and Argentina. An analysis of the time series considering the quality assessment of Landsat 8 and a new method for defining thresholds for burned area identification is provided. This case study has been implemented on the Google Earth Engine platform which allows access to a vast amount of satellite imagery and tools for algorithm development. Results were validated given priority to those areas affected by fire indicating the potential to generate burned area products at 30-m resolution at a subcontinental scale.