Jil Bournazel, Damon Rand and Padraic Harley
In 2018, Ecometrica Ltd. was tasked with identifying potential deforestation over 7000 sites spread across South America over a one year period. This required the analysis of more than 15 Terabytes of satellite imagery, and was equivalent to processing a total area of approximately 25 million km2 for the entire time sequence. Given the vast area and dense time series data required, we developed an approach that uses freely available data from Sentinel 1 (radar) and the Hansen et. al. Forest/Non-Forest map, and processed the imagery with using an open source stack including data preparation software (SNAP toolbox), Omnibus change detection algorithm, and Python 3.x GIS libraries (GDAL, Fiona and Rasterio etc).
None