The MapBiomas land use and land cover classification dataset are the most consistent and comprehensive annual maps of Brazil of the last 40 years. Iain Woodhouse explores why this dataset is essential for Earth Blox customers working on nature-based solutions and regulatory compliance.
The MapBiomas land use and land cover classification dataset are the most consistent and comprehensive annual maps of Brazil of the last 40 years. Iain Woodhouse explores why this dataset is essential for Earth Blox customers working on nature-based solutions and regulatory compliance.
MapBiomas is a multi-institutional initiative involving NGOs, universities, and technology companies that maps land cover and land use changes in Brazil. Unlike official government datasets that often focus solely on deforestation, MapBiomas reconstructs the entire history of the landscape, covering forests, pastures, agriculture, water, and urban areas, on an annual basis.
The latest release, Collection 10, covers a remarkable 40-year period from 1985 to 2024. It utilises the entire Landsat satellite archive and machine learning algorithms on Google Earth Engine to distinguish between 37 different land cover classes.
MapBiomas is one of the most comprehensive historical land cover datasets available for any individual country in the world. For Earth Blox users, its value lies in its depth: it doesn't just tell you what is there now, but what was there every year for the last four decades.
This makes it the critical tool for the financial services sector. By overlaying this historical data with asset locations, banks and investors can assess "nature in agriculture", verifying if a property has a history of illegal deforestation or if a funded pasture is degrading in quality. It remains vital for carbon baselines and regulatory compliance, but its capacity to audit rural credit portfolios against environmental liabilities is a standout feature.
✅ Unrivalled history: Continuous annual data from 1985 to 2024 allows for deep historical analysis of national land use change.
✅ Thematic richness: Distinguishes between 37 classes, including specific crop types (soy, sugarcane), pasture quality, and secondary vegetation stages.
✅ Open science: The methodology is transparent, peer-reviewed, and constantly refined by a network of experts in the specific Brazilian biomes.
✅ EUDR ready: The historical data allows users to explicitly verify land use status before and after the December 31, 2020 cutoff date.
❌ Resolution: At 30m (Landsat), it may miss small-scale degradation or narrow riparian features detectable by 10m Sentinel-2 or 3m commercial data.
❌ "Alternation" noise: Pixel-based classification can result in "flickering" (e.g., Forest -> Pasture -> Forest), which may overestimate annual change rates if not accounted for correctly.
❌ Complexity: The sheer volume of classes and layers can be overwhelming for new users compared to binary "Forest/Non-Forest" maps.
❌ Processing lag: Annual collections are typically released in the second half of the following year (e.g., 2024 data in August 2025).
MapBiomas is not a single map but a living database. The vision for this product is focused on a shift from centralised government mapping to a "networked intelligence" model. Instead of one agency mapping the whole country, MapBiomas uses local experts for each biome. For example, Imazon maps the Amazon, SOS Mata Atlântica maps the Atlantic Forest, and so on.
They process petabytes of Landsat imagery using Google Earth Engine. A key feature is iterative reprocessing: every time they release a new Collection, they reanalyse the entire archive back to 1985. This ensures that any change you see is a real change in the landscape, not just a change in the algorithm.
You get a semantic history of every pixel with a high level of granularity on the classes. Whereas the ESRI landcover is only 9 classes, and the ESA Worldcover is 11 classes, MapBiomas brings 37 classes with a specificity designed around the biomes of Brazil.
The global accuracy for Collection 10 is reported at approximately 91.5% for Level 1 classes (Forest, Farming, Non-Vegetated).
However, users must be aware of the "Alternation" effect. A recent study by Matos et al. (2025) noted that MapBiomas can exhibit more annual class transitions ("flickering") than reference datasets. This is a side effect of pixel-based machine learning. For Earth Blox users, this means you should be cautious about calculating deforestation rates by simply subtracting one year's map from the next, but instead, look at the longer-term trends.
The 30m resolution is the standard for long-term land use change monitoring. It is excellent for identifying landscape-scale changes, farm-level conversions, and verifying supply chains for large commodities like soy and beef.
It is not designed for counting individual trees or monitoring small-scale selective logging . For users requiring higher precision for recent years, MapBiomas has released a 10m Beta Collection based on Sentinel-2 (2016-2023), which is better at resolving narrow features like riparian buffer zones. If you would like access to this data within Earth Blox, please contact us.
Iain H Woodhouse is Knowledge and Outreach Lead at Earth Blox and Professor of Applied Earth Observation at the University of Edinburgh. He specialises in active remote sensing, with over 27 years experience in academia and industry, and more than 100 publications. Iain has advised multiple UK government agencies on EO strategy and is former Chair of the UK Space Agency’s EO Advisory Committee.
.png)