Maps for Good

Maps for Good is Impact Observatory's free, open-access series of annual global land use and land cover (LULC) maps. It is aimed at governments, NGOs, finance and industry that need consistent, comparable information on how land is changing, for example to track urbanization, forest loss and gain, and changes in cropland, rangeland, snow and water.

Each map classifies the whole land surface at 10 m resolution into nine classes: water, trees, flooded vegetation, crops, built area, bare ground, snow/ice, clouds and rangeland. Impact Observatory trained a deep-learning model on billions of human-labelled pixels (curated by the National Geographic Society) and applies it to each year's Copernicus Sentinel-2 imagery, aggregating a full year of classifications into one annual composite. The provider reports 85% overall accuracy and publishes a methodology and accuracy summary and an IEEE paper. The time series starts in 2017 and a new year is added annually; the Esri Living Atlas layer currently covers 2017 to 2025.

The maps are released under a CC BY 4.0 licence in partnership with Esri and Microsoft (the dataset is also known as "Esri Land Cover"). A map clipped to any area of interest can be downloaded from the IO Store as a Cloud Optimized GeoTIFF, a PNG, a CSV of area per class and a GeoJSON of the area. The same data is published on Esri ArcGIS Living Atlas, Microsoft Planetary Computer and the AWS Registry of Open Data, with a public STAC endpoint for tile queries.