MapBiomas Brazil

MapBiomas is a multi-institutional network of universities, NGOs and technology companies that maps land cover and land use change. MapBiomas Brasil classifies Landsat imagery pixel by pixel with machine learning on Google Earth Engine to produce annual 30-metre land cover and land use maps of Brazil for every year from 1985 to the present (Collection 11 covers 1985–2025), plus a 10-metre series from Sentinel-2 imagery starting in 2016.

The platform combines this forty-year baseline with thematic monitoring layers used by environmental agencies, public prosecutors, banks, researchers and companies managing agricultural supply chains. Mapping is organised the way Brazil is: teams led by the institutions that know each biome and cross-cutting theme best, among them Imazon, IPAM, Fundação SOS Mata Atlântica, SOS Pantanal, LAPIG/UFG, UEFS, UFRGS and Solved, produce separate classifications that are integrated into a single national map.

-Brazilian Class Legend: Around thirty classes tuned to national vegetation and cropping, including savanna formation, restinga, mangrove, and separate soy, sugarcane, coffee, citrus, rice, cotton and oil palm classes. -Biome-Level Calibration: Random Forest and other models trained separately for each of the six biomes, then merged with cross-cutting maps of pasture, agriculture, forest plantation, coastal zone, mining, urban areas, wind farms and solar plants. -Thematic Layers: Deforestation, secondary vegetation, pasture, fire (annual and monthly maps since 1985, plus a monthly Sentinel-2 fire monitor since 2019), mining, surface water, soil carbon, degradation, urban extent and heat islands, coastal reefs, air quality and climate risk over the same series. -Deforestation Alerts: MapBiomas Alerta compiles alerts from detection systems such as INPE's DETER and Imazon's SAD, validates them visually on high-resolution Planet imagery, publishes weekly, and issues a report for every confirmed deforestation and for each CAR-registered rural property it overlaps. A GraphQL API is available. -Territorial Query: Automatic statistics and charts for biomes, states, municipalities, watersheds, conservation units and Indigenous lands, with cross-territory and cross-theme analysis. -Open Method and Data: Maps, statistics, ATBD methodology, accuracy assessments and source code published under CC BY 4.0 via the platform, GeoTIFF download, Google Earth Engine assets and a QGIS plugin.

MapBiomas methods are reviewed by an independent Scientific Committee of international experts from The Nature Conservancy, GFW, Clark University and FAO. The data are used well beyond research. According to MapBiomas, BNDES avoided R$728 million in rural credit to landholders with illegal-deforestation alerts on the basis of MapBiomas data. The Monitor do Crédito Rural links public farm credit to land-use change, and the platform had 702,880 users by Collection 9.