Mapbiomas

A collection of Mapbiomas imagery as COGs with overviews

Contents

root

Details

Visibility: Public
Owner: Tristan Grupp
Created: 23 Jul 2026
Last Updated: 23 Jul 2026

Analytics

Contents

Details

Analytics

Mapbiomas

A collection of Mapbiomas imagery as COGs with overviews

Contents

root

Details

Visibility: Public
Owner: Tristan Grupp
Created: 23 Jul 2026
Last Updated: 23 Jul 2026

Analytics

28 days

Downloads

1,343

Data served

4.6 GB

Countries

28-day downloads

9 Jul 2026

5 Aug 2026

Analytics

README

MapBiomas Annual Land Cover and Land Use, South America (2019-2024)

Annual land cover and land use maps from the MapBiomas Project for ten South American countries, 2019-2024, as Cloud-Optimized GeoTIFFs at 30 m. One collection per country, one item per year. Pixel values are the raw national MapBiomas class codes - nothing was reclassified. Because each MapBiomas national initiative maintains its own legend, codes are not comparable across borders as published; the README documents a three-tier harmonisation (exact level-1 collapse, a pan-MapBiomas legend, and a corrected crosswalk onto MapBiomas Brazil Collection 10) and ships it as a machine-readable harmonization.csv.

Collections

MapBiomas Argentina — annual land cover and land use, 2019-2024

MapBiomas annual land cover and land use for Argentina, 2019-2024, at 30 m in EPSG:4326, one Cloud-Optimized GeoTIFF per year. Pixel values are the raw Argentina MapBiomas class codes with no recla...

MapBiomas Bolivia — annual land cover and land use, 2019-2024

MapBiomas annual land cover and land use for Bolivia, 2019-2024, at 30 m in EPSG:4326, one Cloud-Optimized GeoTIFF per year. Pixel values are the raw Bolivia MapBiomas class codes with no reclassif...

MapBiomas Brazil — annual land cover and land use, 2019-2024

MapBiomas annual land cover and land use for Brazil, 2019-2024, at 30 m in EPSG:4326, one Cloud-Optimized GeoTIFF per year. Pixel values are the raw Brazil MapBiomas class codes with no reclassific...

MapBiomas Chile — annual land cover and land use, 2019-2024

MapBiomas annual land cover and land use for Chile, 2019-2024, at 30 m in EPSG:4326, one Cloud-Optimized GeoTIFF per year. Pixel values are the raw Chile MapBiomas class codes with no reclassificat...

MapBiomas Colombia — annual land cover and land use, 2019-2024

MapBiomas annual land cover and land use for Colombia, 2019-2024, at 30 m in EPSG:4326, one Cloud-Optimized GeoTIFF per year. Pixel values are the raw Colombia MapBiomas class codes with no reclass...

MapBiomas Ecuador — annual land cover and land use, 2019-2024

MapBiomas annual land cover and land use for Ecuador, 2019-2024, at 30 m in EPSG:4326, one Cloud-Optimized GeoTIFF per year. Pixel values are the raw Ecuador MapBiomas class codes with no reclassif...

MapBiomas Paraguay — annual land cover and land use, 2019-2023

MapBiomas annual land cover and land use for Paraguay, 2019-2023, at 30 m in EPSG:4326, one Cloud-Optimized GeoTIFF per year. Pixel values are the raw Paraguay MapBiomas class codes with no reclass...

MapBiomas Peru — annual land cover and land use, 2019-2024

MapBiomas annual land cover and land use for Peru, 2019-2024, at 30 m in EPSG:4326, one Cloud-Optimized GeoTIFF per year. Pixel values are the raw Peru MapBiomas class codes with no reclassificatio...

MapBiomas Uruguay — annual land cover and land use, 2019-2024

MapBiomas annual land cover and land use for Uruguay, 2019-2024, at 30 m in EPSG:4326, one Cloud-Optimized GeoTIFF per year. Pixel values are the raw Uruguay MapBiomas class codes with no reclassif...

MapBiomas Venezuela — annual land cover and land use, 2019-2024

MapBiomas annual land cover and land use for Venezuela, 2019-2024, at 30 m in EPSG:4326, one Cloud-Optimized GeoTIFF per year. Pixel values are the raw Venezuela MapBiomas class codes with no recla...

Provenance — legends, harmonisation tables and build scripts

The machine-readable tables the catalog README refers to, and the scripts that build this catalog from the raw MapBiomas country exports. legends.json holds the official legend for each of the te...

Coverage

Spatial Extent

West: -92.0947, South: -57.2646, East: -32.3922, North: 14.0148

Temporal Extent

2019-01-01 to 2024-12-31

Source

https://mapbiomas.org

Processing Notes

What was done to the data

Source GeoTIFFs were exported per country and year from the MapBiomas platform and converted to Cloud-Optimized GeoTIFFs. Pixel values were not changed. No reclassification was applied to any raster. The harmonisation described below is documentation - apply it yourself, at analysis time, to the tier you need.

The conversion uses the Portolan CLI's shipped COG defaults with one deliberate deviation and three additions:

setting	value
compression	DEFLATE, predictor 2	Portolan default
internal tiles	512 x 512	Portolan default
overview resampling	`mode`	deviation - Portolan defaults to `nearest`; these are categorical maps, where the modal class is the meaningful downsample
data type	uint8	addition - every observed class code is below 256; the full-resolution maximum was verified on each non-uint8 source before casting, so the cast is lossless
nodata	0	addition - 0 is background/outside-country in every source raster
colour table	official MapBiomas palette for the country	addition - files render correctly on open in QGIS and ArcGIS
class names	TIFF tags `CLASS_<code>`	addition - the legend travels with the file

Provenance tags written into every COG: MAPBIOMAS_COUNTRY, MAPBIOMAS_YEAR, MAPBIOMAS_COLLECTION, MAPBIOMAS_LEGEND_SOURCE, SOURCE_FILE, OVERVIEW_RESAMPLING, and one CLASS_<code> tag per legend entry.

Which MapBiomas collection is each country?

MapBiomas is not one map. It is a federation of national initiatives, each publishing its own collection on its own schedule with its own legend. The collection version was not assumed - it was established for each country by testing which published legend contains the class codes that actually occur in the pixels.

country	years	MapBiomas collection	how it was established
Argentina	2019-2024	Collection 2	codes 63, 66, 73 and 77 are Collection 2 legend entries
Bolivia	2019-2023	Collection 2	carries 4, 23, 29, 66 and 68, none of which exist in the Collection 1 legend, and none of 39, 72, 81, 82, which are Collection 3 additions
Bolivia	2024	Collection 3	carries 39, 72, 81 and 82; agriculture code 18 is gone, split into 39 and 72
Brazil	2019-2024	Collection 10	carries 6 Floodable Forest, introduced in Collection 10, plus 62 Cotton and 75 Photovoltaic; code 13 is absent, as Collection 10 retired it
Chile	2019-2021	Collection 1	none of 59, 60, 63, 67 are present, and 21 Mosaic of uses is, which the Collection 2 legend does not define
Chile	2022	Collection 1, with mosaic artefacts	as 2019-2021, plus a few hundred stray pixels - see known issues
Chile	2023-2024

Two countries change collection mid-series, so their own years are not directly comparable to each other without harmonising first.

Why harmonisation is needed

The same integer means different things in different countries, and some integers exist in only one country:

4 is Savanna Formation in Brazil, Open forests in Argentina and Bolivia, Dry forest in Peru, Wooded savanna in Venezuela, Open Natural Woodlands in Paraguay.
50 is Herbaceous Sandbank Vegetation in Brazil and Colombia but Xerophytic grassland/shrubland in Venezuela - the same code for two unrelated cover types.
63 is Shrub and herbaceous mosaic in Argentina but Steppe in Chile.
Andean classes 81/82, shrubland 66, fog oasis 70, peatland 73, steppe 63, salt flat 61, glacier 34 and the Uruguayan plantation split 79/80/83 have no equivalent at all in the Brazilian legend.
Uruguay's legend documents agriculture as 19, but every Uruguayan raster encodes it as 18.

Three tiers of harmonisation

How much comparability you can get depends on how much detail you are willing to give up. Rather than one crosswalk, this catalog documents three, and every class code in every raster has an entry in all three.

Tier A - MapBiomas level 1 (exact)

Every national legend is a strict hierarchy under the same six parents. Collapsing to level 1 loses detail but never correctness, and needs no judgement calls. Use this for any statistic that has to be defensible across all ten countries.

level-1 code	class
1	Forest formation
10	Non-forest natural formation
14	Farming
22	Non-vegetated area
26	Water body
27	Not observed

Three caveats, each of which follows the country's own published hierarchy rather than being silently forced:

Peru files code 32 (Coastal salt flat) under level-1 22 Non-vegetated area, while Brazil, Colombia and Venezuela file the same code (Hypersaline tidal flat) under level-1 10 Non-forest natural formation. Tier A follows each country's own published hierarchy rather than forcing one of the two.
Uruguay's code 3 is 'Closed forest and closed shrubland', so Uruguayan level-1 Forest includes shrubland that other countries would place under level-1 10 Non-forest natural formation.
Code 4 is Savanna Formation in Brazil, Open forest in Argentina, Bolivia and Ecuador, Dry forest in Peru, Wooded savanna in Venezuela and Open Natural Woodlands in Paraguay. All six sit under level-1 1 Forest formation, so Tier A is unaffected; Tier B keeps the code with this caveat attached.

Tier B - pan-MapBiomas legend (~50 classes)

Most non-Brazilian codes already mean the same thing in every country that publishes them: 13, 29, 61, 66, 68, 81, 82, 34 are consistent across the Andean and Southern Cone legends. Tier B keeps those codes as published and only resolves genuine collisions:

Venezuela 50 (xerophytic shrubland) → 66 Shrubland, so it stops colliding with Brazil's and Colombia's 50 herbaceous sandbank vegetation.
Argentina 77 (open shrubland) → 66, merging Argentina's closed/open shrubland split into the single shrubland class the other legends use.
Argentina 63 (shrub-herbaceous mosaic) → 66, leaving 63 to mean Chile's Steppe unambiguously.
Uruguay 19 → 18, matching what the Uruguayan rasters actually encode.
Chile 7 and 16 → 27 Not observed; they are mosaic artefacts, not classes.

Tier B is the right target for continental work that needs real thematic resolution. Its cost is that no single country emits the full legend: Brazil never emits Andean classes, the Andean countries never emit Brazil's crop split.

Tier C - MapBiomas Brazil Collection 10 (corrected)

This is the target the Trazo Fields extraction pipeline (extract_hansen_mapbiomas_v16.Rmd) uses. The version of the crosswalk in that script drops 30 class codes into 27 Not Observed without saying so, and treats Colombia and Venezuela as Brazil-compatible passthroughs when seven of their codes have no Brazilian meaning. The table below is the corrected version.

Two rules were applied. No code falls through to 27 - 27 means "MapBiomas could not observe this pixel" and nothing else. No fold may change a pixel's Tier A class - that is what forces shrubland to 12 Grassland rather than 4 Savanna Formation (4 sits under level-1 Forest, so the other choice would reclassify shrubland as forest and inflate forest area across five countries), and glacier to 33 rather than 25.

Lossy folds, applied identically in every country:

Tier B code	class	→	Brazil C10	why, and what is lost
13	Other non-forest natural formation	→	12 Grassland	Brazil Collection 10 dropped 'Other non-forest natural formation'. Folded into 12 Grassland, which shares its level-1 parent (10).
34	Glacier, ice and permanent snow	→	33 River, Lake and Ocean	Brazil Collection 10 has no glacier class. Folded into 33 River, Lake and Ocean because every national legend files glacier under level-1 26 Water body; folding to 25 instead would move the pixels to level-1 22 Non-vegetated.
59	?	→	3 Forest Formation	Brazil Collection 10 does not separate primary from secondary forest. Folded into 3 Forest Formation; level-1 parent (1) is preserved.
60	?	→	3 Forest Formation	As for 59.
61	Salt flat (salar)	→	25 Other non Vegetated Areas	No salt flat class in Brazil. Folded into 25 Other non Vegetated Areas; level-1 parent (22) is preserved.
63	Steppe	→

Per-country tables

Argentina

code	national class	Tier A	Tier B	Tier C (Brazil C10)	Tier C loss
3	Closed forests	1 Forest formation	3 Forest	3 Forest Formation	exact
4	Open forests	1 Forest formation	4 Open / dry forest or savanna formation	4 Savanna Formation	exact
6	Flooded forests	1 Forest formation	6 Flooded forest	6 Floodable Forest	exact
9	Forest plantations	14 Farming	9 Forest plantation	9 Forest Plantation	exact
11	Flooded grasslands	10 Non-forest natural formation	11 Wetland / flooded herbaceous	11 Wetland	exact
12	Grasslands	10 Non-forest natural formation

Bolivia

code	national class	Tier A	Tier B	Tier C (Brazil C10)	Tier C loss
3	Forest	1 Forest formation	3 Forest	3 Forest Formation	exact
4	Open forest	1 Forest formation	4 Open / dry forest or savanna formation	4 Savanna Formation	exact
6	Flooded forest	1 Forest formation	6 Flooded forest	6 Floodable Forest	exact
11	Flooded grassland/shrubland	10 Non-forest natural formation	11 Wetland / flooded herbaceous	11 Wetland	exact
12	Grassland/shrubland	10 Non-forest natural formation	12 Grassland / herbaceous	12 Grassland	exact
13	Other non-forest natural formation	10 Non-forest natural formation

Brazil

code	national class	Tier A	Tier B	Tier C (Brazil C10)	Tier C loss
3	Forest Formation	1 Forest formation	3 Forest	3 Forest Formation	exact
4	Savanna Formation	1 Forest formation	4 Open / dry forest or savanna formation	4 Savanna Formation	exact
5	Mangrove	1 Forest formation	5 Mangrove	5 Mangrove	exact
6	Floodable Forest	1 Forest formation	6 Flooded forest	6 Floodable Forest	exact
9	Forest Plantation	14 Farming	9 Forest plantation	9 Forest Plantation	exact
11	Wetland	10 Non-forest natural formation	11 Wetland / flooded herbaceous

Chile

code	national class	Tier A	Tier B	Tier C (Brazil C10)	Tier C loss
3	Forest	1 Forest formation	3 Forest	3 Forest Formation	exact
4	Open forest (pan-MapBiomas standard code; not in the Chile Collection 2 legend)	1 Forest formation	4 Open / dry forest or savanna formation	4 Savanna Formation	exact
6	Flooded forest (pan-MapBiomas standard code; not in the Chile Collection 2 legend)	1 Forest formation	6 Flooded forest	6 Floodable Forest	exact
7	(not in the published national legend)	27 Not observed	27 Not observed	27 Not Observed	exact
9	Silviculture	14 Farming	9 Forest plantation	9 Forest Plantation	exact
11	Wetland

Colombia

code	national class	Tier A	Tier B	Tier C (Brazil C10)	Tier C loss
3	Forest	1 Forest formation	3 Forest	3 Forest Formation	exact
5	Mangrove	1 Forest formation	5 Mangrove	5 Mangrove	exact
6	Flooded forest	1 Forest formation	6 Flooded forest	6 Floodable Forest	exact
9	Forest plantation	14 Farming	9 Forest plantation	9 Forest Plantation	exact
11	Wetland	10 Non-forest natural formation	11 Wetland / flooded herbaceous	11 Wetland	exact
12	Grasslands / herbaceous	10 Non-forest natural formation	12 Grassland / herbaceous

Ecuador

code	national class	Tier A	Tier B	Tier C (Brazil C10)	Tier C loss
3	Forest	1 Forest formation	3 Forest	3 Forest Formation	exact
4	Open Forest	1 Forest formation	4 Open / dry forest or savanna formation	4 Savanna Formation	exact
5	Mangrove	1 Forest formation	5 Mangrove	5 Mangrove	exact
6	Floodable forest	1 Forest formation	6 Flooded forest	6 Floodable Forest	exact
9	Silviculture	14 Farming	9 Forest plantation	9 Forest Plantation	exact
11	Non forest wetland	10 Non-forest natural formation	11 Wetland / flooded herbaceous

Paraguay

code	national class	Tier A	Tier B	Tier C (Brazil C10)	Tier C loss
3	Closed Natural Woodlands	1 Forest formation	3 Forest	3 Forest Formation	exact
4	Open Natural Woodlands	1 Forest formation	4 Open / dry forest or savanna formation	4 Savanna Formation	exact
6	Flooded Natural Woodlands	1 Forest formation	6 Flooded forest	6 Floodable Forest	exact
9	Forest Plantation	14 Farming	9 Forest plantation	9 Forest Plantation	exact
11	Flooded grasslands	10 Non-forest natural formation	11 Wetland / flooded herbaceous	11 Wetland	exact
12	Grassland	10 Non-forest natural formation

Peru

code	national class	Tier A	Tier B	Tier C (Brazil C10)	Tier C loss
3	Forest	1 Forest formation	3 Forest	3 Forest Formation	exact
4	Dry forest	1 Forest formation	4 Open / dry forest or savanna formation	4 Savanna Formation	exact
5	Mangrove	1 Forest formation	5 Mangrove	5 Mangrove	exact
6	Flooded forest	1 Forest formation	6 Flooded forest	6 Floodable Forest	exact
9	Planted forest	14 Farming	9 Forest plantation	9 Forest Plantation	exact
11	Swamp or Flooded Grassland	10 Non-forest natural formation	11 Wetland / flooded herbaceous

Uruguay

code	national class	Tier A	Tier B	Tier C (Brazil C10)	Tier C loss
3	Closed forest and closed shrubland	1 Forest formation	3 Forest	3 Forest Formation	exact
11	Wetland	10 Non-forest natural formation	11 Wetland / flooded herbaceous	11 Wetland	exact
12	Grassland	10 Non-forest natural formation	12 Grassland / herbaceous	12 Grassland	exact
15	Pasture	14 Farming	15 Pasture	15 Pasture	exact
18	Agriculture (code used by the rasters in place of the legend's code 19; color from the pan-MapBiomas standard palette)	14 Farming	18 Agriculture	18 Agriculture	exact
22	Non-vegetated area	22 Non-vegetated area

Venezuela

code	national class	Tier A	Tier B	Tier C (Brazil C10)	Tier C loss
3	Forest	1 Forest formation	3 Forest	3 Forest Formation	exact
4	Wooded savanna	1 Forest formation	4 Open / dry forest or savanna formation	4 Savanna Formation	exact
5	Mangrove	1 Forest formation	5 Mangrove	5 Mangrove	exact
6	Flooded forest	1 Forest formation	6 Flooded forest	6 Floodable Forest	exact
9	Forest plantation	14 Farming	9 Forest plantation	9 Forest Plantation	exact
11	Flooded grassland/shrubland	10 Non-forest natural formation	11 Wetland / flooded herbaceous

Applying it

The full table is published as a machine-readable asset, provenance/harmonization.csv, with columns country, code, national_class, tier_a_code, tier_a_class, tier_b_code, tier_b_class, tier_b_note, tier_c_code, tier_c_class, tier_c_lossy, tier_c_note.

Anything not listed for a country stays 0, which is nodata - so an unexpected code shows up as a hole rather than as a silently wrong class.

Known issues and caveats

Legends differ between countries, and for Bolivia and Chile between years within the same country. Never compare raw codes across collections. See the collection table and the three harmonisation tiers above.
Chile 2022 contains mosaic artefacts. Codes 4, 6, 7, 13, 16 and 32 occur at 1-3 pixels each in a 1-in-16 sample, against hundreds of thousands of pixels for every real class. Their values fall between adjacent real codes (7 between 6 and 9, 16 between 15 and 18, 32 between 29 and 34), which is the signature of a non-nearest resampling during the mosaic step upstream of this catalog. Codes 7 and 16 exist in no MapBiomas legend at all. Treat all six as no-data for that year.
Uruguay encodes agriculture as 18, while the published Uruguay Collection 3 legend lists agriculture as 19. No Uruguayan raster contains any pixel with value 19.
The Paraguay Collection 2 legend PDF assigns #ffefc3 to class 9 (Forest Plantation) - the same colour it assigns to the class 14 farming aggregate. Every other national legend uses #7a5900. The published value is reproduced verbatim in the colour tables rather than corrected.
Paraguay coverage ends at 2023. MapBiomas Paraguay had not released a 2024 collection when these rasters were exported.
Codes 4, 6, 13, 21 and 32 occur in Chilean rasters but are absent from the published Chile legends. Names and colours come from the pan-MapBiomas standard palette shared by the other national legends.
0 is background, not a class. It is set as nodata in every COG.

Citation

MapBiomas Project — annual land cover and land use collections for Argentina, Bolivia, Brazil, Chile, Colombia, Ecuador, Paraguay, Peru, Uruguay and Venezuela. Accessed through https://mapbiomas.org. Republished as Cloud-Optimized GeoTIFFs with the raw national legends preserved.

Attribution

MapBiomas Project (mapbiomas.org), CC-BY-SA 4.0. Each country map is produced by the corresponding MapBiomas national initiative.

License

CC-BY-SA-4.0

Generated by Portolan from STAC metadata and .portolan/metadata.yaml

README

MapBiomas Annual Land Cover and Land Use, South America (2019-2024)

Collections

MapBiomas Argentina — annual land cover and land use, 2019-2024

MapBiomas Bolivia — annual land cover and land use, 2019-2024

MapBiomas Brazil — annual land cover and land use, 2019-2024

MapBiomas Chile — annual land cover and land use, 2019-2024

MapBiomas Colombia — annual land cover and land use, 2019-2024

MapBiomas Ecuador — annual land cover and land use, 2019-2024

MapBiomas Paraguay — annual land cover and land use, 2019-2023

MapBiomas Peru — annual land cover and land use, 2019-2024

MapBiomas Uruguay — annual land cover and land use, 2019-2024

MapBiomas Venezuela — annual land cover and land use, 2019-2024

Provenance — legends, harmonisation tables and build scripts

Coverage

Spatial Extent

West: -92.0947, South: -57.2646, East: -32.3922, North: 14.0148

Temporal Extent

2019-01-01 to 2024-12-31

Source

https://mapbiomas.org

Processing Notes

What was done to the data

The conversion uses the Portolan CLI's shipped COG defaults with one deliberate deviation and three additions:

setting	value
compression	DEFLATE, predictor 2	Portolan default
internal tiles	512 x 512	Portolan default
overview resampling	`mode`	deviation - Portolan defaults to `nearest`; these are categorical maps, where the modal class is the meaningful downsample
data type	uint8	addition - every observed class code is below 256; the full-resolution maximum was verified on each non-uint8 source before casting, so the cast is lossless
nodata	0	addition - 0 is background/outside-country in every source raster
colour table	official MapBiomas palette for the country	addition - files render correctly on open in QGIS and ArcGIS
class names	TIFF tags `CLASS_<code>`	addition - the legend travels with the file

Which MapBiomas collection is each country?

country	years	MapBiomas collection	how it was established
Argentina	2019-2024	Collection 2	codes 63, 66, 73 and 77 are Collection 2 legend entries
Bolivia	2019-2023	Collection 2	carries 4, 23, 29, 66 and 68, none of which exist in the Collection 1 legend, and none of 39, 72, 81, 82, which are Collection 3 additions
Bolivia	2024	Collection 3	carries 39, 72, 81 and 82; agriculture code 18 is gone, split into 39 and 72
Brazil	2019-2024	Collection 10	carries 6 Floodable Forest, introduced in Collection 10, plus 62 Cotton and 75 Photovoltaic; code 13 is absent, as Collection 10 retired it
Chile	2019-2021	Collection 1	none of 59, 60, 63, 67 are present, and 21 Mosaic of uses is, which the Collection 2 legend does not define
Chile	2022	Collection 1, with mosaic artefacts	as 2019-2021, plus a few hundred stray pixels - see known issues
Chile	2023-2024

Two countries change collection mid-series, so their own years are not directly comparable to each other without harmonising first.

Why harmonisation is needed

The same integer means different things in different countries, and some integers exist in only one country:

4 is Savanna Formation in Brazil, Open forests in Argentina and Bolivia, Dry forest in Peru, Wooded savanna in Venezuela, Open Natural Woodlands in Paraguay.
50 is Herbaceous Sandbank Vegetation in Brazil and Colombia but Xerophytic grassland/shrubland in Venezuela - the same code for two unrelated cover types.
63 is Shrub and herbaceous mosaic in Argentina but Steppe in Chile.
Andean classes 81/82, shrubland 66, fog oasis 70, peatland 73, steppe 63, salt flat 61, glacier 34 and the Uruguayan plantation split 79/80/83 have no equivalent at all in the Brazilian legend.
Uruguay's legend documents agriculture as 19, but every Uruguayan raster encodes it as 18.

Three tiers of harmonisation

Tier A - MapBiomas level 1 (exact)

level-1 code	class
1	Forest formation
10	Non-forest natural formation
14	Farming
22	Non-vegetated area
26	Water body
27	Not observed

Three caveats, each of which follows the country's own published hierarchy rather than being silently forced:

Peru files code 32 (Coastal salt flat) under level-1 22 Non-vegetated area, while Brazil, Colombia and Venezuela file the same code (Hypersaline tidal flat) under level-1 10 Non-forest natural formation. Tier A follows each country's own published hierarchy rather than forcing one of the two.
Uruguay's code 3 is 'Closed forest and closed shrubland', so Uruguayan level-1 Forest includes shrubland that other countries would place under level-1 10 Non-forest natural formation.
Code 4 is Savanna Formation in Brazil, Open forest in Argentina, Bolivia and Ecuador, Dry forest in Peru, Wooded savanna in Venezuela and Open Natural Woodlands in Paraguay. All six sit under level-1 1 Forest formation, so Tier A is unaffected; Tier B keeps the code with this caveat attached.

Tier B - pan-MapBiomas legend (~50 classes)

Venezuela 50 (xerophytic shrubland) → 66 Shrubland, so it stops colliding with Brazil's and Colombia's 50 herbaceous sandbank vegetation.
Argentina 77 (open shrubland) → 66, merging Argentina's closed/open shrubland split into the single shrubland class the other legends use.
Argentina 63 (shrub-herbaceous mosaic) → 66, leaving 63 to mean Chile's Steppe unambiguously.
Uruguay 19 → 18, matching what the Uruguayan rasters actually encode.
Chile 7 and 16 → 27 Not observed; they are mosaic artefacts, not classes.

Tier C - MapBiomas Brazil Collection 10 (corrected)

Lossy folds, applied identically in every country:

Tier B code	class	→	Brazil C10	why, and what is lost
13	Other non-forest natural formation	→	12 Grassland	Brazil Collection 10 dropped 'Other non-forest natural formation'. Folded into 12 Grassland, which shares its level-1 parent (10).
34	Glacier, ice and permanent snow	→	33 River, Lake and Ocean	Brazil Collection 10 has no glacier class. Folded into 33 River, Lake and Ocean because every national legend files glacier under level-1 26 Water body; folding to 25 instead would move the pixels to level-1 22 Non-vegetated.
59	?	→	3 Forest Formation	Brazil Collection 10 does not separate primary from secondary forest. Folded into 3 Forest Formation; level-1 parent (1) is preserved.
60	?	→	3 Forest Formation	As for 59.
61	Salt flat (salar)	→	25 Other non Vegetated Areas	No salt flat class in Brazil. Folded into 25 Other non Vegetated Areas; level-1 parent (22) is preserved.
63	Steppe	→

Per-country tables

Argentina

code	national class	Tier A	Tier B	Tier C (Brazil C10)	Tier C loss
3	Closed forests	1 Forest formation	3 Forest	3 Forest Formation	exact
4	Open forests	1 Forest formation	4 Open / dry forest or savanna formation	4 Savanna Formation	exact
6	Flooded forests	1 Forest formation	6 Flooded forest	6 Floodable Forest	exact
9	Forest plantations	14 Farming	9 Forest plantation	9 Forest Plantation	exact
11	Flooded grasslands	10 Non-forest natural formation	11 Wetland / flooded herbaceous	11 Wetland	exact
12	Grasslands	10 Non-forest natural formation

Bolivia

code	national class	Tier A	Tier B	Tier C (Brazil C10)	Tier C loss
3	Forest	1 Forest formation	3 Forest	3 Forest Formation	exact
4	Open forest	1 Forest formation	4 Open / dry forest or savanna formation	4 Savanna Formation	exact
6	Flooded forest	1 Forest formation	6 Flooded forest	6 Floodable Forest	exact
11	Flooded grassland/shrubland	10 Non-forest natural formation	11 Wetland / flooded herbaceous	11 Wetland	exact
12	Grassland/shrubland	10 Non-forest natural formation	12 Grassland / herbaceous	12 Grassland	exact
13	Other non-forest natural formation	10 Non-forest natural formation

Brazil

code	national class	Tier A	Tier B	Tier C (Brazil C10)	Tier C loss
3	Forest Formation	1 Forest formation	3 Forest	3 Forest Formation	exact
4	Savanna Formation	1 Forest formation	4 Open / dry forest or savanna formation	4 Savanna Formation	exact
5	Mangrove	1 Forest formation	5 Mangrove	5 Mangrove	exact
6	Floodable Forest	1 Forest formation	6 Flooded forest	6 Floodable Forest	exact
9	Forest Plantation	14 Farming	9 Forest plantation	9 Forest Plantation	exact
11	Wetland	10 Non-forest natural formation	11 Wetland / flooded herbaceous

Chile

code	national class	Tier A	Tier B	Tier C (Brazil C10)	Tier C loss
3	Forest	1 Forest formation	3 Forest	3 Forest Formation	exact
4	Open forest (pan-MapBiomas standard code; not in the Chile Collection 2 legend)	1 Forest formation	4 Open / dry forest or savanna formation	4 Savanna Formation	exact
6	Flooded forest (pan-MapBiomas standard code; not in the Chile Collection 2 legend)	1 Forest formation	6 Flooded forest	6 Floodable Forest	exact
7	(not in the published national legend)	27 Not observed	27 Not observed	27 Not Observed	exact
9	Silviculture	14 Farming	9 Forest plantation	9 Forest Plantation	exact
11	Wetland

Colombia

code	national class	Tier A	Tier B	Tier C (Brazil C10)	Tier C loss
3	Forest	1 Forest formation	3 Forest	3 Forest Formation	exact
5	Mangrove	1 Forest formation	5 Mangrove	5 Mangrove	exact
6	Flooded forest	1 Forest formation	6 Flooded forest	6 Floodable Forest	exact
9	Forest plantation	14 Farming	9 Forest plantation	9 Forest Plantation	exact
11	Wetland	10 Non-forest natural formation	11 Wetland / flooded herbaceous	11 Wetland	exact
12	Grasslands / herbaceous	10 Non-forest natural formation	12 Grassland / herbaceous

Ecuador

code	national class	Tier A	Tier B	Tier C (Brazil C10)	Tier C loss
3	Forest	1 Forest formation	3 Forest	3 Forest Formation	exact
4	Open Forest	1 Forest formation	4 Open / dry forest or savanna formation	4 Savanna Formation	exact
5	Mangrove	1 Forest formation	5 Mangrove	5 Mangrove	exact
6	Floodable forest	1 Forest formation	6 Flooded forest	6 Floodable Forest	exact
9	Silviculture	14 Farming	9 Forest plantation	9 Forest Plantation	exact
11	Non forest wetland	10 Non-forest natural formation	11 Wetland / flooded herbaceous

Paraguay

code	national class	Tier A	Tier B	Tier C (Brazil C10)	Tier C loss
3	Closed Natural Woodlands	1 Forest formation	3 Forest	3 Forest Formation	exact
4	Open Natural Woodlands	1 Forest formation	4 Open / dry forest or savanna formation	4 Savanna Formation	exact
6	Flooded Natural Woodlands	1 Forest formation	6 Flooded forest	6 Floodable Forest	exact
9	Forest Plantation	14 Farming	9 Forest plantation	9 Forest Plantation	exact
11	Flooded grasslands	10 Non-forest natural formation	11 Wetland / flooded herbaceous	11 Wetland	exact
12	Grassland	10 Non-forest natural formation

Peru

code	national class	Tier A	Tier B	Tier C (Brazil C10)	Tier C loss
3	Forest	1 Forest formation	3 Forest	3 Forest Formation	exact
4	Dry forest	1 Forest formation	4 Open / dry forest or savanna formation	4 Savanna Formation	exact
5	Mangrove	1 Forest formation	5 Mangrove	5 Mangrove	exact
6	Flooded forest	1 Forest formation	6 Flooded forest	6 Floodable Forest	exact
9	Planted forest	14 Farming	9 Forest plantation	9 Forest Plantation	exact
11	Swamp or Flooded Grassland	10 Non-forest natural formation	11 Wetland / flooded herbaceous

Uruguay

code	national class	Tier A	Tier B	Tier C (Brazil C10)	Tier C loss
3	Closed forest and closed shrubland	1 Forest formation	3 Forest	3 Forest Formation	exact
11	Wetland	10 Non-forest natural formation	11 Wetland / flooded herbaceous	11 Wetland	exact
12	Grassland	10 Non-forest natural formation	12 Grassland / herbaceous	12 Grassland	exact
15	Pasture	14 Farming	15 Pasture	15 Pasture	exact
18	Agriculture (code used by the rasters in place of the legend's code 19; color from the pan-MapBiomas standard palette)	14 Farming	18 Agriculture	18 Agriculture	exact
22	Non-vegetated area	22 Non-vegetated area

Venezuela

code	national class	Tier A	Tier B	Tier C (Brazil C10)	Tier C loss
3	Forest	1 Forest formation	3 Forest	3 Forest Formation	exact
4	Wooded savanna	1 Forest formation	4 Open / dry forest or savanna formation	4 Savanna Formation	exact
5	Mangrove	1 Forest formation	5 Mangrove	5 Mangrove	exact
6	Flooded forest	1 Forest formation	6 Flooded forest	6 Floodable Forest	exact
9	Forest plantation	14 Farming	9 Forest plantation	9 Forest Plantation	exact
11	Flooded grassland/shrubland	10 Non-forest natural formation	11 Wetland / flooded herbaceous

Applying it

Anything not listed for a country stays 0, which is nodata - so an unexpected code shows up as a hole rather than as a silently wrong class.

Known issues and caveats

Legends differ between countries, and for Bolivia and Chile between years within the same country. Never compare raw codes across collections. See the collection table and the three harmonisation tiers above.
Chile 2022 contains mosaic artefacts. Codes 4, 6, 7, 13, 16 and 32 occur at 1-3 pixels each in a 1-in-16 sample, against hundreds of thousands of pixels for every real class. Their values fall between adjacent real codes (7 between 6 and 9, 16 between 15 and 18, 32 between 29 and 34), which is the signature of a non-nearest resampling during the mosaic step upstream of this catalog. Codes 7 and 16 exist in no MapBiomas legend at all. Treat all six as no-data for that year.
Uruguay encodes agriculture as 18, while the published Uruguay Collection 3 legend lists agriculture as 19. No Uruguayan raster contains any pixel with value 19.
The Paraguay Collection 2 legend PDF assigns #ffefc3 to class 9 (Forest Plantation) - the same colour it assigns to the class 14 farming aggregate. Every other national legend uses #7a5900. The published value is reproduced verbatim in the colour tables rather than corrected.
Paraguay coverage ends at 2023. MapBiomas Paraguay had not released a 2024 collection when these rasters were exported.
Codes 4, 6, 13, 21 and 32 occur in Chilean rasters but are absent from the published Chile legends. Names and colours come from the pan-MapBiomas standard palette shared by the other national legends.
0 is background, not a class. It is set as nodata in every COG.

Citation

Attribution

MapBiomas Project (mapbiomas.org), CC-BY-SA 4.0. Each country map is produced by the corresponding MapBiomas national initiative.

License

CC-BY-SA-4.0

Generated by Portolan from STAC metadata and .portolan/metadata.yaml

Tier C is lossy by construction for the Andes and Patagonia. Andean grassland 81 alone is a large share of the Bolivian, Peruvian, Ecuadorian and Colombian highlands, and Tier C folds it into 12 Grassland. Highland analyses should use Tier A or Tier B.

These rasters are a mirror, not the authoritative source. MapBiomas releases new collections regularly; check mapbiomas.org before assuming these are current.

1import csv
2import numpy as np
3import rasterio
4
5TIER = "tier_c_code"          # or tier_a_code / tier_b_code
6COUNTRY = "bolivia"
7
8lut = np.zeros(256, dtype="uint8")        # 0 stays 0 (nodata)
9for row in csv.DictReader(open("provenance/harmonization.csv", encoding="utf-8")):
10    if row["country"] == COUNTRY:
11        lut[int(row["code"])] = int(row[TIER])
12
13path = "mapbiomas-bolivia/mapbiomas_bolivia_2024/mapbiomas_bolivia_2024.tif"
14with rasterio.open(path) as src:
15    harmonised = lut[src.read(1)]

1import csv
2import numpy as np
3import rasterio
4
5TIER = "tier_c_code"          # or tier_a_code / tier_b_code
6COUNTRY = "bolivia"
7
8lut = np.zeros(256, dtype="uint8")        # 0 stays 0 (nodata)
9for row in csv.DictReader(open("provenance/harmonization.csv", encoding="utf-8")):
10    if row["country"] == COUNTRY:
11        lut[int(row["code"])] = int(row[TIER])
12
13path = "mapbiomas-bolivia/mapbiomas_bolivia_2024/mapbiomas_bolivia_2024.tif"
14with rasterio.open(path) as src:
15    harmonised = lut[src.read(1)]