New hybrid transition map at 10m (beta version)

• A hybrid transition map is now available at a resolution of 10m for the period from 1990 to 2022. This beta version combines the recently updated TMF map for the year 2022, derived from Landsat images at 30m, with the detections of forest disturbances by Sentinel 2 for the year 2022 at 10m. The integration of Sentinel-2 data allows a better identification of degraded forests, and a refinement of the disturbance edges and linear disturbances, such as logging roads and small rivers within the forest. The hybrid transition map is available for visualisation and download via ftp or GEE, along with a map detailing the data source of each disturbance detection. This feature allows users to identify whether disturbances were detected using Landsat, Sentinel 2, or both sensors in query mode by clicking on the map . Historical Sentinel 2 data for the previous years (2016-2021) will be used to produce an enhanced beta version of this hybrid transition map when the back processing of the Sentinel 2 archive, notably the improvement of the geometric accuracy, will be completed by ESA. Part of the small-scale forest disturbances that have been detected in the new beta map through the use of Sentinel 2 data in 2022 may have occurred in previous years but were not detected by 30m Landsat imagery. The use of the full Sentinel 2 archive for the final planned version of the hybrid map is expected to enhance the overall quality of the hybrid map and mitigate the issue of timing of the disturbance event. The final product combining the full archive of Landsat and Sentinel 2 data up to year 2022 is intended to be delivered with an accuracy assessment.

List of updates integrated in the TMF 2022 products

• Integration for year 2021 and year 2022 of Landsat Collection 2 imagery which is designed, amongst other issues, to improve the geometric and radiometric accuracy of the imagery (https://www.usgs.gov/landsat-missions/landsat-collection-2). The finalisation of the whole Landsat archive’s (1990-2020) reprocessing and integration to the TMF dataset is foreseen for the end of 2023. This will not only result in better quality satellite image data, but increase the overall number of available imagery for analysis. For year 2021 and at the pantropical level, we quantified an increase of 13% in the detection of forest disturbances (new degradation or direct deforestation) from Collection-2 (C2) compared to Collection-1 data (C1). 66% of new forest disturbances from C2 are detected by C1, while 18% are within 60m (~ 2 Landsat sized pixels) from forest disturbances detected by C1 (most probably linked to the increased geometric accuracy in C2 data) and 16% of C2- forest disturbances are located beyond this buffer areas (most probably linked to the increased number of Landsat scenes and valid observations). You can find below a country-level comparison of 2021 detection of forest disturbances between C1 and C2 data:
  NewForestDisturbances_2021_C1_C2_byCountry.xlsx
• Integration for year 2022 of Landsat 9 (L9) imagery (https://www.usgs.gov/landsat-missions/landsat-9) along with Landsat 7 and 8 (resp. L7, L8). For the pantropical region, this resulted in a 35% increase of forest disturbances detection (new degradation or direct deforestation) in 2022. 43% of forest disturbances are detected by both L7/L8 and L9, while 31% are only detected by L7/L8 and 26% are only detected by L9. You can find below a country-level comparison of 2022 detection of forest disturbances between L7/L8 and L9 data:
  NewForestDisturbances_2022_L7_L8_L9_byCountry.xlsx
• Integration of the global closed-canopy coconut map from Descals et al. 2023 (https://essd.copernicus.org/preprints/essd-2022-463/) which helps refine the dynamics of forest conversion to plantations.
• Integration of the global mangrove watch 2020 map (https://data.unep-wcmc.org/datasets/45) to refine the maximum extent mask of mangroves which was originally defined from the 1996-2016 global mangrove watch data.
• Delineation of new tree plantation mainly in Central, West Africa, Indonesia and Malaysia.
• Improvements and corrections of errors in the Annual Change collection in the sequence of values for deforestation of old regrowth forest (areas of deforestation after regrowth can now be identified using a dedicated GEE asset).
• Afforestation can now be monitored in the Annual Change collection as the change of values 6 (other land cover) to 4 (regrowth).

List of updates integrated in the TMF 2021 products

• We improved the identification of forest conversion to commodities (classes 81-86 in the Transition map- Sub types) as follows:
  • We visually delineated missing plantation areas (mainly industrial plantations of oil palm, rubber, tea…) in Indonesia, Malaysia, Cambodia, Nigeria, Ghana and Brazil on the basis of the transition map 2020 for all areas with specific geometric shapes corresponding to plantations and very high resolution images available from Google Earth Engine (GEE) and Planet monthly composites. We also used the latest version of the Spatial Database of Planted Trees from the Global Forest Watch (accessed in February 2022) to complement missing large-scale industrial plantation from the TMF dataset.
  • We fully incorporated the smallholder and industrial closed-canopy oil palm plantations dataset from Descals et al. 2021 (https://doi.org/10.5194/essd-13-1211-2021) within the commodities mask.
  • We excluded tree plantations from the commodities mask and reclassified them into other land cover or afforestation (classes 91-93). To do so, we used the Spatial Database of Planted Trees database and more specifically using the plantation classes of acacia, eucalyptus, pine, teak or araucaria but also the class forest plantation identified in the Level-2 of the MapBiomass dataset 1985-2018 (https://www.mdpi.com/2072-4292/12/17/2735/htm). This reclassification was visually checked and completed using high-resolution images available in GEE and Planet monthly composites.
  • We used a buffer of 60m to ensure that the commodities mask did not overlap with the Open Street Map dataset (accessed in 2019) and the built-up areas (from 1975-2014 epochs) from the Global Human Settlement Layers P2016 (https://ghsl.jrc.ec.europa.eu/ghs_bu.php) (where TMF products are classified as non-forest).

• We improved the distinction between deforestation without prior degradation, deforestation occurring after degradation and multiple events of degradation. Up to TMF version 2020, we applied two conditions to consider that deforestation occurred after degradation: “a recurrence value lower than 58% or a recurrence value lower than 70% with at least 6 years without any disruption events between the degradation and the deforestation disturbances”. In this new version, we now analyze the full sequence of disruptions (1982-2021) to better characterize complex trajectories such as direct deforestation, several events degradation (up to 4 events of degradation) or deforestation after degradation (single or multiple short events). Degradation is still characterized by a 2.5 years maximum duration while deforestation can be observed for a longer time. We required a minimum period of two years between two disturbance events (with no disruption detection).

  The figure below displays the values of Transition Map (2021), Annual Change (1982-2021) and disruption observation (1982-2021) for a given pixel (long:-58.57, lat: -11.40) The sequence of disruptions shows three distinct disturbance events of less than 2.5 years duration and separated by at least 2 years with no disruption observation. Previously this pixel corresponded to a tropical moist forest that has been degraded in 2005 and deforested in 2008. After corrections, this pixel is now classified as degraded forest with three stages of short duration degradation in 2005, 2008 and 2019.

  

  The update described above consequently improved the detection of the year of deforestation occurring after degradation which also concerns the date of forest conversion to plantation when a prior degradation occurred. We now provide this information as a GEE asset (see the Google Earth Engine below).

  The figure below shows the information of a pixel (long:-47.76, lat: -3.49) of tropical moist forest that has been degraded and deforested. The deforestation year in TMF version 2020 was 2001 and corresponded to the second time a disruption was observed. After corrections, the year of deforestation after degradation has been modified to 2012 as we detect a disturbance event of more than 3 years duration. We also improved the detection of the several events of short duration degradation prior deforestation (in 1998, 2001 and 2008). This improvement modifies the Transition Map but also the values of the annual change collection.

  
• We improved the separation between degradation and deforestation that started the last year (2021). Up to TMF version 2020, a threshold of 10 disruptions was used to define a deforested land. We now calculate a ratio between the annual number of disruption and the annual number of valid observation. We defined a threshold of 45% (or 23% if the disturbance started after the second half of the year) on the basis of visual identification over new logging, burning or deforestation events in 2021.
• We now integrate deforestation of regrowth forest of at least 10 years old and reclassify this trajectory in the Transition Map as a deforested land. To distinguish it from deforestation of undisturbed or degraded forest, we provide a mask of deforestation of regrowth as a GEE asset (see the Google Earth Engine below).

All of the datasets that have been produced to document the Tropical Moist Forest cover and changes over the past three decades are being made freely available using the following delivery mechanisms: Tropical Moist Forest Explorer, Country level statistics, Data Download, Google Earth Engine and Web Map Service. These are described in the following sections.

Country-level statistics 1990-2022

We provide the statistics of annual forest cover changes for the period 1990-2022 at the country level for countries with more than 1 Mha forest area in 1990. These statistics have been extracted directly from the different TMF products. Please select the country of interest in this dropdown menu and you will retrieve the annual area in million hectares of undisturbed TMF, forest degradation, deforestation, forest regrowth as well as many other sub-classes. More information on the different classes reported are provided in the FAQ.

or download the CSV file with all countries' statistics.

Country factsheet on tropical forest status and dynamics of deforestation and forest degradation

The fact sheets provide country-level information on land cover status, main forest types and recent (2001-2022) dynamics of forest cover change in humid and dry tropical domains. They include automated charts and descriptions that report the distribution of the main land cover types from the Copernicus Global Land Cover map of 2019 (herbaceous, cultivated, other land cover and forest corresponding to pixels in the Discrete Land cover classes of closed and open forest [111-125] where the Forest Fractional Cover is at least 30%). The fact sheets also report the trends and rates of deforestation, forest degradation from the JRC Tropical Moist Forest dataset. Tropical forest cover losses outside the TMF domain are reported from the University of Maryland (UMD) Global Forest Change products (using a minimum of 10% tree cover percentage in 2000, a minimum mapping unit of 0.5 ha and excluding areas under forestry operations as defined by Curtis et al. 2018). These fact sheets are only produced for tropical countries with more than 1 Mha forest area in 1990.

Data Download

The following spatial datasets are available for download: Undisturbed and degraded tropical moist forest, Transition Map - Sub types, Transition Map - Main Classes, Annual change collection (1990-2022), Degradation year and Deforestation year.

The Version of the Product appears in the file Names that are available for download e.g. JRC_TMF_TransitionMap_Subtypes_v1_1982_2021[...].tif

Other layers (Intensity, Duration, Annual disruption observations, Annual valid observations, Start monitoring period) are available in Google Earth Engine.

Download process

Individual 10°x10° files

The Tropical Moist Forest data are available to download in tiles 10°x10° from the map shown below. Click on the tile to show a list of the available datasets. Each one of these datasets is a hyperlink to the *.tif file.

Granule with top-left corner at: {{ lon_label }}, {{ lat_label }}

Transition Map - Sub types : download

Transition Map - Main Classes : download

Degradation year : download

Deforestation year : download

Undisturbed and degraded tropical forest : download

Annual change : {{ year }},

Annual change collection (one zip file including all years between 1990-2022) : download

Hybrid Transition Map - Sub types : download

Supporting files

Symbology

Each of the downloadable files can be displayed in desktop GIS tools (such as QGIS or ArcGIS) using a symbology that contains the colormap and the labels for the values. These can be added to the files by using the following symbology files.

Metadata

The downloadable files do not contain any metadata information and so it is provided here for each of the datasets. You may need to right click and Download Linked file.

Google Earth Engine

The data can also be accessed and used in the Google Earth Engine platform - for more information see here. The following asset ids are used in Google Earth Engine:

Web Map Service

The Tropical Moist Forest data can also be used within other websites or GIS clients as WMS (Web Map Service). These service provide a direct link to the images that are used in the Tropical Moist Forest Explorer and is the best option if you simply want to map the data and produce cartographic products.
They are not suitable for analysis as the data are represented only as RGB images.
The WMS url is: https://ies-ows.jrc.ec.europa.eu/iforce/tmf_v1/wms.py?

License

All data here are produced under studies funded by the Directorate-General for Climate Action of the European Commission through the Roadless-For pilot project and the Lot 2 (Tropical moist Forest Monitoring) of the ForMonPol project (Forest Monitoring for Policies). All data are provided free of charge, without restriction of use. For the full license information see the Copernicus Regulation of the European Commission.

Publications, models and data products that make use of these datasets must include proper acknowledgement, including citing datasets and the journal article as in the following citation.

Disclaimer

The designations employed and the presentation of material on this map do not imply the expression of any opinion whatsoever on the part of the European Union concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries.