Exploratory: generated automatically from the UmaTerra platforms' data; not independently validated or peer reviewed.

UmaTerra · story Português
Forest · fire · climate · Brazil · Canada

What the satellite saw and what is ahead: forest loss, fire and climate in the same Brazil and Canada places

Forest loss, fire and climate in the same places of Brazil (Pará and Paragominas) and British Columbia: what the satellite has recorded since 1985 and what projections say is ahead, with measured and projected always kept apart. This is a report showing how the data can be used, not a peer-reviewed publication. It reads exposure, not prediction.

This is the second story of the series: the first, Who decides about the forest, covers tenure, rules and the right to cut in the same places.

How to cite: Biondo, Leandro M. (2026). What the satellite saw and what is ahead: forest loss, fire and climate in the same Brazil and Canada places. UmaTerra. https://umaterra.com.br/en/stories/forests-fire-climate/

Loading the numbers…

—places read closely, in both countries measured
—ha of tree-cover loss in the places, 2001–2025 (Hansen) measured
—fire-weather models compared, with each place's range estimated

measured by satellite, official outline or count. estimated a reading that depends on a choice, a threshold or a projection. missing a part that was not done.

Why look at Brazil and Canada

Two of the world's largest forests. More than half of the world's forests (54%) are in five countries: Russia, Brazil, Canada, the United States and China. Brazil has 497 million hectares of forest, 12% of the world total, and Canada has 347 million, 9%. Together they hold about a fifth of the planet's forests (FAO, 2020). What each one decides about protecting, cutting or restoring forest matters far beyond its borders.

Climate. Forests take carbon out of the air: on average 1.1 billion tonnes of carbon a year, roughly one-sixth of fossil fuel emissions (Pan et al., 2011). That balance can change. In the Amazon, measurements linked the forest's shift to a carbon source to deforestation and climate change (Gatti et al., 2021). In British Columbia, a mountain pine beetle outbreak was studied as a forest carbon feedback to climate (Kurz et al., 2008).

Protecting and restoring biomass. The boreal forest, which covers much of Canada, is one of the largest biomes on Earth, and about two-thirds of its area is under some form of management, mostly for wood production (Gauthier et al., 2015). In Brazil, restoring is part of the answer: secondary forests of the American tropics build biomass back (Poorter et al., 2016), and the much-reduced Atlantic Forest is treated as a restoration opportunity (Rezende et al., 2018).

Water. Forest is also water. Trees return moisture to the air, and that moisture becomes rain somewhere else (Ellison et al., 2017); in the Amazon, the rain the forest recycles helps buffer droughts (Staal et al., 2018). In British Columbia, the valleys' water depends on the snow that builds up in the mountains, the subject of another story.

How the places were chosen. The places are not a random sample and do not represent the two countries. The choice was partly arbitrary and partly guided by similarities. The regions came first: British Columbia, because forest law in Canada is provincial and that is where the tools and the open data reach furthest; Pará and Maranhão, for the Amazon and its forest concessions; Paraná and Espírito Santo, for the Atlantic Forest. Inside them we looked for one place of each tenure type (public production forest, protected area and private land), preferably near places the portal had already studied, such as the state capitals and the Okanagan. We then paired them by tenure type, not by biome or size: a national forest with a concession beside a licence to cut on public land, a reserve beside a park, a mosaic of properties beside a block of private land. The pairs differ in size, climate and forest; what they share is the question of who decides. Other places would give other numbers.

What the two stories do. This series does not measure carbon or water. It shows, in the same places of both countries, who decides about the forest and what the satellite recorded and what is ahead. The numbers in this box come from the sources cited below, not from our own measurements.

Sources for this box: FAO (2020). Global Forest Resources Assessment 2020: Key findings. doi:10.4060/ca8753en; Pan et al. (2011). A large and persistent carbon sink in the world’s forests. Science. doi:10.1126/science.1201609; Gatti et al. (2021). Amazonia as a carbon source linked to deforestation and climate change. Nature. doi:10.1038/s41586-021-03629-6; Kurz et al. (2008). Mountain pine beetle and forest carbon feedback to climate change. Nature. doi:10.1038/nature06777; Gauthier et al. (2015). Boreal forest health and global change. Science. doi:10.1126/science.aaa9092; Poorter et al. (2016). Biomass resilience of Neotropical secondary forests. Nature. doi:10.1038/nature16512; Rezende et al. (2018). From hotspot to hopespot: An opportunity for the Brazilian Atlantic Forest. Perspectives in Ecology and Conservation. doi:10.1016/j.pecon.2018.10.002; Ellison et al. (2017). Trees, forests and water: Cool insights for a hot world. Global Environmental Change. doi:10.1016/j.gloenvcha.2017.01.002; Staal et al. (2018). Forest-rainfall cascades buffer against drought across the Amazon. Nature Climate Change. doi:10.1038/s41558-018-0177-y.

1

The places of Who decides about the forest, now through fire and climate

This story returns to the places of the first one, read side by side in Brazil and in British Columbia: — places of public tenure, protection and private property. There the question was who decides; here it is what the satellite recorded in them and what projections say is ahead.

It is a report showing how the data can be used, not a peer-reviewed publication. It reads exposure, not prediction: nothing here says what will happen in a place.

2

Caxiuanã and TFL 49, from Landsat to Sentinel-2

The slider moves between an annual Landsat composite near — and a true-colour Sentinel-2 image of 2025. Caxiuanã National Forest, in Pará, and TFL 49, in British Columbia, are at the same scale.

The image measures nothing: it shows what the numbers of the next steps measure. TFL 49 fills —% of the window; Caxiuanã, —%. The window clips about — km at each end, north and south, of TFL 49.

3

The other concession and the Revelstoke licence

Saracá-Taquera National Forest shares the window with the Rio Trombetas Biological Reserve. Beside it, TFL 56, held by Revelstoke Community Forest Corporation, a corporation set up by the City of Revelstoke.

4

Rio Trombetas Biological Reserve and Okanagan Mountain Park

The two panels have their own boxes and scale bars, flagged in the legend: the Brazilian reserve fits a — km window and the Canadian park one of — km. Size is not to be compared by eye.

5

The Paragominas window and the E&N block

On the left, the Paragominas window, with properties registered in the CAR. On the right, the E&N block, labelled “Managed forest land (descriptive label)”: a descriptive stand-in, rebuilt from the parcel fabric (private parcels of — ha or more), not the list of lands under the private-forest-land act.

6

The same legend in both countries

Hansen's map marks, in 30 m pixels, where the tree cover of 2000 was lost in each period from 2001 to 2025, with one legend for both countries. Tree-cover loss is not deforestation: legal cutting, fire and management also count. Sooretama, Irati and Saint-Hilaire/Lange have numbers but no maps: they appear in the tables only.

7

Inside the unit and in the 10 km ring

For each place the chart shows each year's loss inside the unit and in the 10 km ring around it. The largest annual loss inside TFL 49 was — ha, in —; in Caxiuanã, — ha, in —.

8

Harvest or fire, with each source's limits

The global drivers layer gives one dominant class per 1.1 km pixel for the whole of 2001 to 2023, not per year. In TFL 49 the dominant class is —, with —% of Hansen loss. Hectares per class are Hansen loss (30 m) falling inside 1.1 km pixels of that class, not a cause area measured at 30 m.

NTEMS separates harvest and fire by year, from 1985 to 2020. The open cutblocks (FTEN) complete the picture, without holders flagged First Nations or uncertain. After 2020 there is only Hansen, with no cause: loss from 2021 to 2025 stays “cause not attributed”.

9

Primary and secondary vegetation, and regeneration

MapBiomas records, by year, the clearing events of primary and of secondary vegetation, and regrowth. In Caxiuanã, the total cleared from 1987 to 2025 (the product's period) was — ha; in Paragominas, — ha. The meaning of the event classes was read from the code of the Yvynation and muda-clima tools, and must be checked against MapBiomas documentation before citing.

For contrast, the same global drivers layer, with the same caveat: one class per pixel, for the whole period. Hectares per class are Hansen loss (30 m) falling inside 1.1 km pixels of that class.

10

Two sets of causes, not to be added

In Canada the chart stacks harvest and fire up to 2020, from NTEMS. In Brazil it stacks clearing of primary and of secondary vegetation, from MapBiomas. Each source's limits are written in the legend, and the two stacks are not comparable bar for bar.

11

Age in 2019 in Canada, secondary vegetation in Brazil

In British Columbia, forest age is NTEMS's for 2019, a single year. In Brazil, age comes from MapBiomas Collection 10.1 secondary vegetation, to 2024: Collection 11 does not have this product. They are two definitions, and each map has its own legend. Along the rivers the Brazilian product shows a lot of secondary vegetation (floodplain): read with care.

Mean forest age is — years in TFL 49; in Caxiuanã, the oldest age class of secondary vegetation is the —-year class.

12

One product for both countries

Above-ground biomass comes from ESA CCI, version 6, the same product in both countries. Inside the unit the mean is — Mg/ha in Caxiuanã and — Mg/ha in TFL 49; in the 10 km ring, — and —.

13

The comparison is descriptive, not causal

In each pair, loss and cover change inside the unit are set beside those of the 10 km ring. The ratio of loss inside to loss in the ring (as % of the 2000 forest) is — in TFL 49 and — in Caxiuanã.

Inside and outside differ in soil, relief, access and history. The difference describes; it does not say what the unit caused.

14

Caxiuanã and Saracá-Taquera, side by side

They are two concessions of different size, conceded share and contract year: —% of Caxiuanã National Forest is conceded, against —% of Saracá-Taquera's. The maps set loss and regrowth side by side in each.

The story does not conclude which is more profitable or more accessible: neither was measured.

15

Burned area, last fire and times burned

MapBiomas Fogo, Collection 5, gives each year's burned area, the year of the last fire and how many times each pixel burned, at 30 m. In Paragominas, —% of the window burned at least once; in Caxiuanã, —%.

For a year of fire at biome scale, see the Cerrado in 2024.

16

NBAC polygons to 2023 and severity to 2020

NBAC polygons run to 2023 and give each fire's year and area. In Okanagan Mountain Park NBAC records — ha burned since 1972.

NTEMS fire severity (dNBR) goes only to 2020 and only where there is data: in TFL 49 it covered —% of NTEMS fire pixels. The NTEMS year can differ from NBAC's by up to two years; fire year and area come from NBAC.

17

The only burned-area map the same in both countries

MODIS MCD64A1 has the same definition, the same resolution (500 m) and the same period (since November 2000) in both countries. It is therefore the only record that can be compared across the two sides.

The limit is the pixel: burns smaller than it escape. In the Paragominas window, adding the years 2001 to 2025 of each product, MODIS sees —% of the area MapBiomas Fogo sees. In the E&N block MODIS sees more than NBAC: MODIS 500 m also counts small and prescribed burns; the cause is not established here.

18

Each source in its own chart, never added

For each place, each year's burned area appears source by source: MapBiomas Fogo in Brazil, NBAC in Canada and MODIS in both. Products are not added, and a difference between them is not one's or the other's error.

The worst year in TFL 49 by NBAC was —, with — ha; in Paragominas, by MapBiomas Fogo, —, with — ha.

19

Recurrence, from 1985 or 2001 to the last year

The map counts, per pixel, how many times it burned: since 1985 in MapBiomas Fogo, since 1972 in the NBAC polygons and since 2001 in MODIS. The periods differ and each pair's legend says which applies.

—% of Okanagan Mountain Park burned more than once; in Paragominas, —% of the window.

20

The park and TFL 49, and two stories already written

Okanagan Mountain Park has fires recorded by NBAC in —; in TFL 49, NBAC adds up to — ha.

We do not repeat what is already told: see the West Kelowna fire and the Okanagan dry winters.

21

In Paragominas, fire and clearing on one map

In the Paragominas window the map puts burned area and cleared vegetation in the same frame. Within it, —% of the area burned from 1985 to 2025 (sum of the years) fell on pasture, the class of the burned pixel itself.

It is coincidence in space, not order in time nor cause: the maps do not say who set fire, nor whether fire came before or after the cutting.

22

What the report may and may not say

The climate part follows six muda-clima rules. One: name the scale, by drawing the model cell. Two: keep measured and projected apart. Three: agree before asserting. muda-clima's rule is by share: a change is asserted when at least 2 of every 3 models share the sign of the median; with the 9 models of this story that is 6 of 9 (the “majority” class), and 8 of 9 or more is “robust”.

Four: count area, not damage. Five: degrade, do not invent, and say what is missing. Six: say what it is not: a reading of probable exposure, not a forecast for a property nor an assessment.

23

A 0.25° cell over each place

The map draws, together, the model's 0.25° cell, the unit and the analysis circle. The cell is about — km tall, and its width shrinks with latitude: — km at Caxiuanã and — km at TFL 49. In Brazil the service caps the circle radius at 25 km: in the three large units the circle is smaller than the unit. In all four British Columbia places the circle extends beyond the 0.25° cell.

A model value belongs to the cell, not to a point inside it. TFL 49 crosses — cells of 0.25°; muda-clima reads the cell at the centre of the circle.

24

What lies under each cell's climate

Inside each circle, land use is read as muda-clima reads it: MapBiomas Collection 11 (2025) in Brazil and GLC-FCS30D 2022, a global map, in Canada, each with its own groups. AAFC is not used, and MapBiomas applies only in Brazil. In Caxiuanã, —% of the circle is forest; in TFL 49, —%.

25

Rain, water deficit and temperature

Before any projection, the measured: CHIRPS and TerraClimate rain, water deficit and temperature. Where CHIRPS does not reach, in British Columbia, ERA5-Land is used. A trend counts as “clear” only when the two rain sources agree. Measured temperature is MODIS land-surface temperature (not air temperature), and only in Brazil; there is no measured temperature for British Columbia in this story.

In Paragominas, the rain trend is —; in the two sources, — and — mm per decade.

26

Only what at least 2 of every 3 models assert

For each indicator the chart shows the change between 1985–2014 and 2041–2070 under scenarios SSP2-4.5 and SSP5-8.5, with the range of the 9 models and how many share the sign of the median. Only what at least 2 of every 3 models (6 of 9) support is asserted; where they do not agree, the chart says there is no agreement.

In Caxiuanã, median warming is — °C under SSP2-4.5 and — °C under SSP5-8.5; in TFL 49, — and — °C.

27

Hectares by hazard and land-use group, never damage

For each hazard on which the models agree, the step adds up the hectares of the land-use groups it acts on. Under SSP5-8.5, fire conditions reach — ha of forest in the TFL 49 circle and — ha in Caxiuanã's.

It is area, not loss of timber, carbon or income: that would need growth and management models the tool does not have.

28

Few models and thresholds that are choices

Nine models, one per centre, describe a range, not a probability; “2 of every 3” (6 of 9) is a minimum, and a larger ensemble could change the medians and the agreement counts.

Each hazard's thresholds are choices stated on the page, not values calibrated against observed damage. Land use is held at 2025: nothing is said about land-use change between the two windows.

29

A condition of the atmosphere, not a burned area

The Canadian Fire Weather Index (FWI) sums up temperature, humidity, wind and rain into a measure of how easily vegetation burns and how fast fire spreads. A day with FWI above 30 is a day of high to extreme danger. The index describes weather, not fire: it is a condition; burned area is an outcome.

The source is provisional: NASA marks the data as subject to change and the product's paper is under review. Each 30-year window is missing years: it has 21 to 29 of 30, and the mean uses those that exist.

30

The 0.25° grid over the places

The Brazil and Canada grids, from muda-clima, have 0.25° cells and nine models. The map shows the cells that cross each place: — in TFL 49, — in Caxiuanã, — in Okanagan Mountain Park.

A place's value is the mean across cells, weighted by the unit's area in each; the source is provisional and each window has 21 to 29 years of 30.

31

Today and in 2041–2070, with the models' range

For each place the chart shows the days per year above FWI 30 in the baseline and in 2041–2070, under SSP2-4.5 and SSP5-8.5, with the nine models' range and how many agree. In TFL 49 the median goes from — to — days under SSP5-8.5; in Caxiuanã, from — to —. Agreement is muda-clima's class: “majority” when at least 6 of the 9 models share the sign of the median, “robust” with 8 or more.

In the Amazon, in TFL 56 and in the E&N block, days above FWI 30 are about zero today and in 2041–2070, and the models split: the chart shows them as “about zero, models split”. That is why the next step measures each place against its own 95th percentile.

32

Days above the place's own 95th percentile

Each cell's 95th percentile, in the product's reference period, defines what is extreme there. In the baseline the threshold is exceeded on —% of days in TFL 49 and —% in Caxiuanã; in 2041–2070, under SSP5-8.5, it becomes —% and —% (median of the models). It is the indicator that informs where FWI 30 is rare.

33

Hectares where models agree on more risk days

Inside the cells where at least 6 of the 9 models agree on more days above the place's own 95th percentile (SSP5-8.5), the step adds up the hectares of the forest that is there today: — ha in TFL 49 and — ha in Caxiuanã.

It is exposure, not burned area: the number says where the condition worsens, not how much will burn.

34

Yesterday's burned area, tomorrow's fire weather

For each place, the burned area measured in Act 3 and the projected fire weather sit side by side, in separate panels with their own legends. One is never multiplied by the other: there is no “projected burned area” in this story.

Fire also depends on ignition, fuel, management and suppression, which the index does not contain.

35

Products, end years and cuts

Each product ends in a year: NTEMS in 2020 (harvest, fire and severity), VLCE2 in 2022, NBAC in 2023 and MapBiomas in 2025. Legends are not mixed, and loss after 2020 in Canada stays without a cause. The drivers layer gives one class per pixel for 2001–2023, not per year.

No Indigenous land or First Nations data appears, in either country. The E&N block is a descriptive stand-in and the CAR is self-declared, simplified and “registered”. Projections are exposure, not prediction; the FWI is provisional.

36

The same places, in the apps and in another story

Each place can be read in Yvynation, Naturametrics, Camposcope and muda-clima; the map's buttons lead to each. Who decides about the forest in these places is in the first story: Who decides about the forest.

Other stories with the same instruments: West Kelowna, the Okanagan and the Cerrado. In Maranhão, another story covers the Indigenous lands Porquinhos and Bacurizinho, which are not part of this one.

umaterra.com.br · What the satellite saw
Loading…measured
Loading the numbers…

What this story does not say

Read first: Who decides about the forest

Written with Claude, from Anthropic, from the numbers the data returned: no fact or number comes from the model. Português.

Notes and references

  1. muda-clima, an open climate-exposure tool (clima.umaterra.com.br), and the six rules of exposure reporting described in the manuscript presenting the tool, still in preparation. The references marked for verification in that manuscript are not used here.
  2. Gorelick, N. et al. (2017). Google Earth Engine: planetary-scale geospatial analysis for everyone. Remote Sensing of Environment 202, 18–27. doi:10.1016/j.rse.2017.06.031. The annual Landsat composites, the Sentinel-2 images of steps 2 to 5 and the raster layers were read in Earth Engine.
  3. Government of British Columbia, news release 2021FLNRO0032-000836: “TFL 56 is held by the Revelstoke Community Forest Corporation”. The holder's corporate nature was not read in a primary registry; the text repeats only the release's sentence and what the corporation is according to local press: a company set up by the City of Revelstoke. news.gov.bc.ca.
  4. SICAR/CAR, state databases of registered rural properties (ES, MA, PA and PR), copied into our own table; geometries simplified to 12 m and declared by the owner, not validated. The text speaks of registered properties, never of regular or irregular ones.
  5. Contains information licensed under the Open Government Licence – British Columbia (TFL boundaries, FADM_TFL; authorized cutblocks, FTEN_CUT_BLOCK_POLY_SVW; parcel fabric, PMBC_PARCEL_FABRIC_POLY_SVW) and the Open Government Licence – Canada (CPCAD protected areas; NRCan). The E&N block was rebuilt from the parcel fabric and has First Nations overlaps subtracted in memory; no First Nations geometry is published, only the area removed.
  6. Hansen, M. C. et al. (2013). High-resolution global maps of 21st-century forest cover change. Science 342, 850–853. doi:10.1126/science.1244693. Tree cover in the year 2000 (treecover2000), 30 m pixels.
  7. Sims, M. J. et al. (2025). Global drivers of forest loss at 1 km resolution. Environmental Research Letters 20, 074027. doi:10.1088/1748-9326/add606; WRI/Google DeepMind layer (CC BY 4.0), 2001 to 2023: one dominant class per 1.1 km pixel for the whole period, not per year, and it does not tell natural forest from plantation.
  8. White, J. C. et al. (2017). A nationwide annual characterization of 25 years of forest disturbance and recovery for Canada using Landsat time series. Remote Sensing of Environment 194, 303–321. doi:10.1016/j.rse.2017.03.035; NTEMS, harvest and wildfire maps 1985 to 2020 and fire magnitude (dNBR), Open Government Licence – Canada. Severity coverage is partial (see the step).
  9. MapBiomas, Collection 11 Deforestation and Secondary Vegetation (clearing events of primary and secondary vegetation, regeneration and secondary-vegetation age), mapbiomas.org. The legend is MapBiomas's own and is not added to NTEMS's.
  10. Maltman, J. C. et al. (2023). Estimating and mapping forest age across Canada's forested ecosystems. Remote Sensing of Environment 290, 113529. doi:10.1016/j.rse.2023.113529; NTEMS, forest age in 2019 (a single year).
  11. Santoro, M. et al. (2021). The global forest above-ground biomass pool for 2010 estimated from high-resolution satellite observations. Earth System Science Data 13, 3927–3950. doi:10.5194/essd-13-3927-2021 (paper of the series); the story uses the ESA CCI Biomass product, version 6, one product for both countries. The version 6 record page was not read: check before publishing.
  12. Brazilian Forest Service, ArcGIS layers Limites_UMFs and unidades_concessoes_florestais (forest management unit polygons), read on 15 Nov 2026; areas computed from the polygon in an equal-area projection (ESRI:102033); the SFB area field was not used. Contracts, signing dates and volumes: SFB open data (CSV, CC-BY) and the Public Forest Management Report 2025, tables 5, 6, 7, 22 and 25.
  13. Alencar, A. A. C. et al. (2022). Long-term Landsat-based monthly burned area dataset for the Brazilian biomes using deep learning. Remote Sensing 14, 2510. doi:10.3390/rs14112510; MapBiomas Fogo, Collection 5 (burned area, year of last fire and times burned), mapbiomas.org. 30 m pixels; it maps burning, not the cause of the fire.
  14. Hall, R. J. et al. (2020). Generating annual estimates of forest fire disturbance in Canada: the National Burned Area Composite. International Journal of Wildland Fire 29, 878–891. doi:10.1071/WF19201; NBAC, Canadian Forest Service (CWFIS), Open Government Licence – Canada, polygons 1972 to 2023 read from the Earth Engine community catalog. The CWFIS version runs to 2025 and was not used.
  15. Giglio, L. et al. (2018). The Collection 6 MODIS burned area mapping algorithm and product. Remote Sensing of Environment 217, 72–85. doi:10.1016/j.rse.2018.08.005; MCD64A1 product (collection 061), 500 m, from November 2000. The only record common to both countries; 500 m pixels miss small burns.
  16. Thrasher, B. et al. (2022). NASA Global Daily Downscaled Projections, CMIP6. Scientific Data 9, 262. doi:10.1038/s41597-022-01393-4; NEX-GDDP-CMIP6, 0.25°, bias-corrected; scenarios SSP2-4.5 and SSP5-8.5; windows 2041–2070 against 1985–2014.
  17. Souza, C. M. et al. (2020). Reconstructing three decades of land use and land cover changes in Brazilian biomes with Landsat archive and Earth Engine. Remote Sensing 12, 2735. doi:10.3390/rs12172735; MapBiomas Project, Collection 11 (1985 to 2025), mapbiomas.org.
  18. Funk, C. et al. (2015). The climate hazards infrared precipitation with stations — a new environmental record for monitoring extremes. Scientific Data 2, 150066. doi:10.1038/sdata.2015.66; CHIRPS, 0.05°, from 1981; it does not reach beyond 50° N.
  19. Abatzoglou, J. T. et al. (2018). TerraClimate, a high-resolution global dataset of monthly climate and climatic water balance from 1958–2015. Scientific Data 5, 170191. doi:10.1038/sdata.2017.191; climatic water deficit and rain, about 4 km, monthly.
  20. Muñoz-Sabater, J. et al. (2021). ERA5-Land: a state-of-the-art global reanalysis dataset for land applications. Earth System Science Data 13, 4349–4383. doi:10.5194/essd-13-4349-2021; used where CHIRPS does not reach (Canada), monthly, 0.1°.
  21. NASA, NEX-GDDP-FWI (The NASA Earth eXchange Global Daily Downscaled Projections Fire Weather Index), GDDP-IMPACT portal, data.nas.nasa.gov/gddpimpact/FWI, 00ReadMe file read on 15 Nov 2026: the data are “considered provisional and subject to change”. The paper describing the product is under review and is not cited as published. We use nine models and the muda-clima grids, built on 31 Oct 2026.
  22. Hermosilla, T. et al. (2018). Disturbance-informed annual land cover classification maps of Canada's forested ecosystems for a 29-year (1984–2012) period. Canadian Journal of Remote Sensing 44, 67–87. doi:10.1080/07038992.2018.1437719; NTEMS VLCE2 product (1984 to 2022), Open Government Licence – Canada. Contains information licensed under the Open Government Licence – Canada.
  23. No Indigenous land, First Nations licence or title area, reserve or treaty land enters any map or number of this story, in either country (the OCAP principle of data ownership); only the area removed from the E&N block is kept.