This dataset was produced using Landsat 8 Operational Land Imager and Landsat 7 Enhanced Thematic Mapper Plus surface reflectance data spanning 2013–2018 and Spectral Mixture Analysis for the identification of patterns of forest loss for each year. High-resolution Planet Dove (3m) and RapidEye (5m) imagery were used to validate the forest loss map. Overall Accuracy obtained for the forest loss map was 96%. Publication: https://doi.org/10.1088/1748-9326/ab57c3 Google Earth Engine code: https://code.earthengine.google.com/024b42f8eb3ab0c5fa8e0ad8fba86f36 For more information on SERVIR, visit http://www.servirglobal.net

Project Foresight was launched, in early 2019, as a continuous effort to develop machine-learning based deforestation and forest fire risk assessment for tropical forests, using increasing higher resolution satellite data and official country data on anthropogenic activity. Version 1.0 included maps of the Peruvian and Colombian Amazon using 18 years of official deforestation data and the open source release of Maxent, as the machine-learning algorithm. Newer versions have been developed using mutli-model ensembles in R and Google Earth Engine.