<p>Join Land &#038; Carbon Lab&#8217;s upcoming webinar on February 11, 2025 to discover how DIST-ALERT’s integration into Global Forest Watch can enhance your work and help shape the future of land monitoring. <a href="https://gfw.global/4hBlNJF">Learn more</a>.</p>


WEBINAR: Monitoring Possibilities with DIST-ALERT

Data &amp; Tools

Data &amp; Tools | Global Forest Watch Blog

forest change

forest change | Global Forest Watch Blog

mapping

mapping | Global Forest Watch Blog

Michelle Sims

Liz Goldman

Elise Mazur

<p>It’s well established that forests are important for people, nature and climate, but not all forests are created equal. Natural forests tend to <a href="https://zenodo.org/records/3553579#.ZFU8GnbMJwI" target="_blank" rel="noreferrer noopener">harbor more biodiversity</a> and offer a wider array of ecosystem services than planted forests, and therefore keeping natural forests standing is particularly critical.</p>



<p>The tree cover data available on Global Forest Watch (GFW) includes all types of tree cover, including natural forests and planted trees such tree crops and wood fiber plantations. Although data on primary forests are available for the humid tropics, distinguishing natural forests globally from other types of tree cover previously required overlaying multiple data sets to exclude areas that are not natural forests.</p>



<p>Now, new <a href="https://gfw.global/4fQfl03">natural forest data on GFW</a> from the <a href="https://sciencebasedtargetsnetwork.org/" target="_blank" rel="noreferrer noopener">Science Based Target Network (SBTN)</a> Natural Lands Map can help shed light on where the world’s natural forests are located and provide a baseline for monitoring loss within them since 2020.</p>



<h2 class="wp-block-heading">What is the Science Based Target Network Natural Lands Map?</h2>



<p>Land & Carbon Lab collaborated with WWF and Systemiq to develop the <a href="https://landcarbonlab.org/insights/natural-lands-map-companies-no-conversion-targets/" target="_blank" rel="noreferrer noopener">Natural Lands Map</a> to support companies who have committed to eliminate conversion of natural lands in their supply chains through targets set with the SBTN land sector framework. The map provides a 2020 baseline of natural lands that companies can use to see where production activities have caused conversion since 2020, as well as identify and prevent future risk of conversion.</p>



<p>The Natural Lands Map includes several classes of natural land cover. The natural forest layer on GFW is a subset of the SBTN Natural Lands Map that includes the natural forest class and relevant subclasses, including wetland forests, mangroves and peat forests.</p>



<p>The <a href="https://sciencebasedtargetsnetwork.org/wp-content/uploads/2023/05/Technical-Guidance-2023-Step3-Land-v0.3-Natural-Lands-Map.pdf" target="_blank" rel="noreferrer noopener">Natural Lands Map</a> was created by combining the best available global and local land cover and land use data sets into a single harmonized map at 30-meter resolution circa the year 2020. This includes data sets such as the University of Maryland GLAD Lab’s 2020 <a href="https://www.frontiersin.org/journals/remote-sensing/articles/10.3389/frsen.2022.856903/full" target="_blank" rel="noreferrer noopener">land cover data set</a>, various data sets on cropland and <a href="https://www.wri.org/research/spatial-database-planted-trees-sdpt-version-2" target="_blank" rel="noreferrer noopener">planted trees</a>, and local data sets such as <a href="https://brasil.mapbiomas.org/en/" target="_blank" rel="noreferrer noopener">MapBiomas</a> in South America and Indonesia, among others. The various input data sets were selected based on accuracies reported from the data producers when available and alignment with the definitions used by SBTN. Input data sets were combined to create the final Map.</p>



<h2 class="wp-block-heading">How does the Map define natural forests?</h2>



<p>SBTN uses the Accountability Framework Initiative (AFi) <a href="https://accountability-framework.org/use-the-accountability-framework/definitions/" target="_blank" rel="noreferrer noopener">definitions</a> of natural ecosystems and natural forests to align with other standard setting initiatives. AFi adopts the Food and Agriculture Organization (FAO) definition of forests: land spanning more than 0.5 hectares with trees higher than 5 meters and a canopy cover of more than 10%, or trees able to reach these thresholds in situ, not including land that is predominantly under agricultural or other land use. AFi then differentiates <em>natural</em> forests as those where the species composition, structure and ecological function do not show effects of major human impacts.</p>



<p>Because satellite data cannot directly map all the elements of these definitions, the data development team operationalized them using proxies based on available data to exclude tree cover that would not meet the natural forest definition. Non-natural tree cover may include tree crops (e.g., palm oil, orchards, cocoa), planted forests (e.g., eucalyptus, teak, or pine plantations), patches of trees under 0.5 hectares, or trees within areas predominantly under agricultural or urban land use (e.g., trees planted within agricultural fields or within small urban parks). Natural forests may include primary, secondary or naturally regenerating forests, and may include some forms of low-intensity management.</p>



<p>AFi guidance states that quantitative thresholds established by national and sub-national forest definitions may take precedence over the global definitions, so local maps were given priority over global data in select regions.</p>



<h4 class="wp-block-heading">The natural forest map distinguishes natural forests from other tree cover</h4>



<div class="flourish-embed flourish-photo-slider" data-src="visualisation/21139033?163031"><script src="https://public.flourish.studio/resources/embed.js"></script><noscript><img decoding="async" src="https://public.flourish.studio/visualisation/21139033/thumbnail" width="100%" alt="photo-slider visualization" /></noscript></div>



<h2 class="wp-block-heading">What data is available on Global Forest Watch?</h2>



<p>The natural forest layer now available on the GFW map includes all natural forest subclasses from the SBTN Natural Lands Map (<a href="https://sciencebasedtargetsnetwork.org/wp-content/uploads/2023/05/Technical-Guidance-2023-Step3-Land-v0.3-Natural-Lands-Map.pdf" target="_blank" rel="noreferrer noopener">version 0.3</a>). This natural forest layer has an overall accuracy of 88%.*</p>



<iframe loading="lazy" width="670" height="490" frameborder="0" src="https://www.globalforestwatch.org/embed/map/?map=eyJkYXRhc2V0cyI6W3siZGF0YXNldCI6Im5hdHVyYWwtZm9yZXN0cyIsIm9wYWNpdHkiOjEsInZpc2liaWxpdHkiOnRydWUsImxheWVycyI6WyJuYXR1cmFsLWZvcmVzdHMtMjAyMCJdfSx7ImRhdGFzZXQiOiJwb2xpdGljYWwtYm91bmRhcmllcyIsImxheWVycyI6WyJkaXNwdXRlZC1wb2xpdGljYWwtYm91bmRhcmllcyIsInBvbGl0aWNhbC1ib3VuZGFyaWVzIl0sIm9wYWNpdHkiOjEsInZpc2liaWxpdHkiOnRydWV9XX0%3D"></iframe>



<p>Previously, GFW users could analyze tree cover loss within and outside of the <a href="https://gfw.global/2Qbv7fM">Spatial Database of Planted Trees</a>, which helped identify which losses were more likely to represent plantation harvesting or tree crop rotation. Now, the GFW dashboard allows users to separate tree cover loss in natural forests and non-natural tree cover from 2021 onwards using the natural forest map, meaning users can specifically monitor loss of natural forests around the world. This provides users with a more comprehensive and improved method to estimate loss in natural forests that is more aligned with widely accepted definitions.</p>



<iframe loading="lazy" width="630" height="460" frameborder="0" src="https://www.globalforestwatch.org/embed/widget/treeLossPlantations/country/BRA"></iframe>



<p></p>



<p>Users can also now calculate the area** of natural forest, non-natural tree cover and other land cover within administrative boundaries and other land use categories, such as protected areas.</p>



<iframe loading="lazy" width="315" height="460" frameborder="0" src="https://www.globalforestwatch.org/embed/widget/naturalForest/global"></iframe>



<p></p>



<h2 class="wp-block-heading">How can this data be used?</h2>



<p>The natural forest layer can be used to assess the extent of natural forest in 2020, and can be combined with additional data, such as tree cover loss, to monitor loss of natural forests from 2021 onward in a user’s area of interest. For example, companies can use the data to assess whether their production or sourcing activities have caused deforestation since 2020, or if their future plans pose a risk of deforestation, for tracking progress under voluntary commitments or the European Union Deforestation Regulation (EUDR).</p>



<p>For any use case however, users should consider the definitions used for the natural forest layer and determine whether these definitions align with their monitoring needs. For example, there is a key difference in how AFi and EUDR define deforestation that users must consider: while AFi defines deforestation as conversion of natural forest to non-natural tree cover or another land cover, <a href="https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:32023R1115&qid=1686611894470" target="_blank" rel="noreferrer noopener">EUDR defines deforestation</a> as the conversion of any type of forest — including planted forest — to agricultural use. Because the natural forest layer does not include planted forest, users may need to use additional data to include planted forests in their forest baseline to monitor conversion of natural and planted forest to agriculture for EUDR purposes.</p>



<p>For a full description of how the natural forest map relates to EUDR, see <a href="https://landcarbonlab.org/insights/natural-lands-map-companies-no-conversion-targets/" target="_blank" rel="noreferrer noopener">this blog</a>. For more information on EUDR definitions and how GFW can support EUDR monitoring, read <a href="https://www.globalforestwatch.org/blog/data-and-tools/satellite-data-eu-regulation-deforestation-free-supply-chains/">part one</a> and <a href="https://www.globalforestwatch.org/blog/data-and-tools/monitoring-forest-degradation-eudr/">two</a> of our blog series.</p>



<p>Whenever using the SBTN Natural Lands Map, including the natural forests map on GFW, users should also be aware of a few limitations of the data:</p>



<ul type="1">
<li>Overestimation of natural forest in agroforestry systems and in temperate and boreal forests: Agroforestry systems, especially those where crops are grown under shade trees (such as some types of coffee and cocoa systems), are difficult to differentiate from natural forests in satellite imagery, and therefore available data on their extent is limited. Similarly, publicly available data on planted forests is limited in many temperate and boreal regions. As a result, these areas may be included as natural forest in the Map.</li>



<li>Inconsistent forest definitions in certain regions: Because the SBTN Natural Lands Map team prioritized local and regional input data over global data during the compilation and harmonization process when the Map was developed, there may be inconsistencies in the height and canopy density thresholds used to define forests in some regions where local data was used, including Brazil, the Amazon, Atlantic Forest, Pampa, Chaco, Indonesia, South Africa and New Zealand.</li>



<li>Temporal inconsistencies: The Map is as close to the year 2020 as possible, but due to limited data availability, some input data is from earlier time periods (e.g., 2015).</li>



<li>Resolution inconsistencies: Most input data used to create the Map is at least 30-meter resolution, but in some cases lower resolution data was used when higher resolution data was unavailable. These inconsistencies create data artifacts (sharp boundary lines that are only due to resolutions) in some areas.</li>
</ul>



<p>See more information on these limitations in the <a href="https://sciencebasedtargetsnetwork.org/wp-content/uploads/2023/05/Technical-Guidance-2023-Step3-Land-v0.3-Natural-Lands-Map.pdf" target="_blank" rel="noreferrer noopener">technical documentation</a>.</p>



<h2 class="wp-block-heading">What’s next for the natural forest data on GFW?</h2>



<p>Currently, <a href="https://sciencebasedtargetsnetwork.org/wp-content/uploads/2023/05/Technical-Guidance-2023-Step3-Land-v0.3-Natural-Lands-Map.pdf" target="_blank" rel="noreferrer noopener">version 0.3</a> of the natural forest map is available on GFW. In the coming months, we will update GFW to <a href="https://sciencebasedtargetsnetwork.org/wp-content/uploads/2024/09/Technical-Guidance-2024-Step3-Land-v1-Natural-Lands-Map.pdf" target="_blank" rel="noreferrer noopener">version 1</a>, which is now available on the <a href="https://datasets.wri.org/datasets/sbtn-natural-lands-map?map=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" target="_blank" rel="noreferrer noopener">WRI Open Data Portal</a>, <a href="https://developers.google.com/earth-engine/datasets/catalog/WRI_SBTN_naturalLands_v1" target="_blank" rel="noreferrer noopener">Google Earth Engine</a> and <a href="https://pro-news.globalforestwatch.org/news-events/how-gfw-pro-supports-eudr-compliance" target="_blank" rel="noreferrer noopener">GFW Pro</a>. We plan to continue to update the data on GFW as future versions of the SBTN Natural Lands Map become available.</p>



<p>We also plan to release additional analysis capabilities with the data, such as the option to filter deforestation and fire alerts to within natural forests. Stay tuned for future updates!</p>



<p></p>



<p>* <em>Note that the validation points were labeled based on definitions of natural forests from v1 of the SBTN Natural Lands Map, which includes more regional data sets and their corresponding definitions, as well as an additional filtering of local savanna classes to remove vegetation shorter than 5 meters from the forest class. This may affect the validation of the v0.3 natural forest layer because of variation in definitions of natural forest where local data was used. See the <a href="https://sciencebasedtargetsnetwork.org/wp-content/uploads/2024/09/Technical-Guidance-2024-Step3-Land-v1-Natural-Lands-Map.pdf" target="_blank" rel="noreferrer noopener">v1 technical note change log</a> to see the full list of changes.</em></p>



<p><em>** We generally </em><a href="https://landcarbonlab.org/insights/natural-lands-map-companies-no-conversion-targets/" target="_blank" rel="noreferrer noopener"><em>caution against</em></a><em> using the SBTN Natural Lands Map to quantify the area of natural lands due to data limitations, including overestimation of natural lands. However, data on natural and non-natural forest cover is much more widely available than other land covers, such as natural grassland and pasture, so we consider the natural forest classes to provide a reasonable approximation of natural forest area. Users should still be aware of the limitations of the data when using it to quantify area.</em></p>



<style type="text/css"></style>

New Data on Global Forest Watch Distinguishes Natural Forests from Other Tree Cover

<p>It’s well established that forests are important for people, nature and climate, but not all forests are created equal. Natural forests tend to <a href="https://zenodo.org/records/3553579#.ZFU8GnbMJwI" target="_blank" rel="noreferrer noopener">harbor more biodiversity</a> and offer a wider array of ecosystem services than planted forests, and therefore keeping natural forests standing is particularly critical.</p>



<p>The tree cover data available on Global Forest Watch (GFW) includes all types of tree cover, including natural forests and planted trees such tree crops and wood fiber plantations. Although data on primary forests are available for the humid tropics, distinguishing natural forests globally from other types of tree cover previously required overlaying multiple data sets to exclude areas that are not natural forests.</p>



<p>Now, new <a href="https://gfw.global/4fQfl03">natural forest data on GFW</a> from the <a href="https://sciencebasedtargetsnetwork.org/" target="_blank" rel="noreferrer noopener">Science Based Target Network (SBTN)</a> Natural Lands Map can help shed light on where the world’s natural forests are located and provide a baseline for monitoring loss within them since 2020.</p>



<h2 class="wp-block-heading">What is the Science Based Target Network Natural Lands Map?</h2>



<p>Land &amp; Carbon Lab collaborated with WWF and Systemiq to develop the <a href="https://landcarbonlab.org/insights/natural-lands-map-companies-no-conversion-targets/" target="_blank" rel="noreferrer noopener">Natural Lands Map</a> to support companies who have committed to eliminate conversion of natural lands in their supply chains through targets set with the SBTN land sector framework. The map provides a 2020 baseline of natural lands that companies can use to see where production activities have caused conversion since 2020, as well as identify and prevent future risk of conversion.</p>



<p>The Natural Lands Map includes several classes of natural land cover. The natural forest layer on GFW is a subset of the SBTN Natural Lands Map that includes the natural forest class and relevant subclasses, including wetland forests, mangroves and peat forests.</p>



<p>The <a href="https://sciencebasedtargetsnetwork.org/wp-content/uploads/2023/05/Technical-Guidance-2023-Step3-Land-v0.3-Natural-Lands-Map.pdf" target="_blank" rel="noreferrer noopener">Natural Lands Map</a> was created by combining the best available global and local land cover and land use data sets into a single harmonized map at 30-meter resolution circa the year 2020. This includes data sets such as the University of Maryland GLAD Lab’s 2020 <a href="https://www.frontiersin.org/journals/remote-sensing/articles/10.3389/frsen.2022.856903/full" target="_blank" rel="noreferrer noopener">land cover data set</a>, various data sets on cropland and <a href="https://www.wri.org/research/spatial-database-planted-trees-sdpt-version-2" target="_blank" rel="noreferrer noopener">planted trees</a>, and local data sets such as <a href="https://brasil.mapbiomas.org/en/" target="_blank" rel="noreferrer noopener">MapBiomas</a> in South America and Indonesia, among others. The various input data sets were selected based on accuracies reported from the data producers when available and alignment with the definitions used by SBTN. Input data sets were combined to create the final Map.</p>



<h2 class="wp-block-heading">How does the Map define natural forests?</h2>



<p>SBTN uses the Accountability Framework Initiative (AFi) <a href="https://accountability-framework.org/use-the-accountability-framework/definitions/" target="_blank" rel="noreferrer noopener">definitions</a> of natural ecosystems and natural forests to align with other standard setting initiatives. AFi adopts the Food and Agriculture Organization (FAO) definition of forests: land spanning more than 0.5 hectares with trees higher than 5 meters and a canopy cover of more than 10%, or trees able to reach these thresholds in situ, not including land that is predominantly under agricultural or other land use. AFi then differentiates <em>natural</em> forests as those where the species composition, structure and ecological function do not show effects of major human impacts.</p>



<p>Because satellite data cannot directly map all the elements of these definitions, the data development team operationalized them using proxies based on available data to exclude tree cover that would not meet the natural forest definition. Non-natural tree cover may include tree crops (e.g., palm oil, orchards, cocoa), planted forests (e.g., eucalyptus, teak, or pine plantations), patches of trees under 0.5 hectares, or trees within areas predominantly under agricultural or urban land use (e.g., trees planted within agricultural fields or within small urban parks). Natural forests may include primary, secondary or naturally regenerating forests, and may include some forms of low-intensity management.</p>



<p>AFi guidance states that quantitative thresholds established by national and sub-national forest definitions may take precedence over the global definitions, so local maps were given priority over global data in select regions.</p>



<h4 class="wp-block-heading">The natural forest map distinguishes natural forests from other tree cover</h4>



<div class="flourish-embed flourish-photo-slider" data-src="visualisation/21139033?163031"><script src="https://public.flourish.studio/resources/embed.js"></script><noscript><img src="https://public.flourish.studio/visualisation/21139033/thumbnail" width="100%" alt="photo-slider visualization" /></noscript></div>



<h2 class="wp-block-heading">What data is available on Global Forest Watch?</h2>



<p>The natural forest layer now available on the GFW map includes all natural forest subclasses from the SBTN Natural Lands Map (<a href="https://sciencebasedtargetsnetwork.org/wp-content/uploads/2023/05/Technical-Guidance-2023-Step3-Land-v0.3-Natural-Lands-Map.pdf" target="_blank" rel="noreferrer noopener">version 0.3</a>). This natural forest layer has an overall accuracy of 88%.*</p>



<iframe width="670" height="490" frameborder="0" src="https://www.globalforestwatch.org/embed/map/?map=eyJkYXRhc2V0cyI6W3siZGF0YXNldCI6Im5hdHVyYWwtZm9yZXN0cyIsIm9wYWNpdHkiOjEsInZpc2liaWxpdHkiOnRydWUsImxheWVycyI6WyJuYXR1cmFsLWZvcmVzdHMtMjAyMCJdfSx7ImRhdGFzZXQiOiJwb2xpdGljYWwtYm91bmRhcmllcyIsImxheWVycyI6WyJkaXNwdXRlZC1wb2xpdGljYWwtYm91bmRhcmllcyIsInBvbGl0aWNhbC1ib3VuZGFyaWVzIl0sIm9wYWNpdHkiOjEsInZpc2liaWxpdHkiOnRydWV9XX0%3D"></iframe>



<p>Previously, GFW users could analyze tree cover loss within and outside of the <a href="https://gfw.global/2Qbv7fM">Spatial Database of Planted Trees</a>, which helped identify which losses were more likely to represent plantation harvesting or tree crop rotation. Now, the GFW dashboard allows users to separate tree cover loss in natural forests and non-natural tree cover from 2021 onwards using the natural forest map, meaning users can specifically monitor loss of natural forests around the world. This provides users with a more comprehensive and improved method to estimate loss in natural forests that is more aligned with widely accepted definitions.</p>



<iframe width="630" height="460" frameborder="0" src="https://www.globalforestwatch.org/embed/widget/treeLossPlantations/country/BRA"></iframe>



<p></p>



<p>Users can also now calculate the area** of natural forest, non-natural tree cover and other land cover within administrative boundaries and other land use categories, such as protected areas.</p>



<iframe width="315" height="460" frameborder="0" src="https://www.globalforestwatch.org/embed/widget/naturalForest/global"></iframe>



<p></p>



<h2 class="wp-block-heading">How can this data be used?</h2>



<p>The natural forest layer can be used to assess the extent of natural forest in 2020, and can be combined with additional data, such as tree cover loss, to monitor loss of natural forests from 2021 onward in a user’s area of interest. For example, companies can use the data to assess whether their production or sourcing activities have caused deforestation since 2020, or if their future plans pose a risk of deforestation, for tracking progress under voluntary commitments or the European Union Deforestation Regulation (EUDR).</p>



<p>For any use case however, users should consider the definitions used for the natural forest layer and determine whether these definitions align with their monitoring needs. For example, there is a key difference in how AFi and EUDR define deforestation that users must consider: while AFi defines deforestation as conversion of natural forest to non-natural tree cover or another land cover, <a href="https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:32023R1115&amp;qid=1686611894470" target="_blank" rel="noreferrer noopener">EUDR defines deforestation</a> as the conversion of any type of forest — including planted forest — to agricultural use. Because the natural forest layer does not include planted forest, users may need to use additional data to include planted forests in their forest baseline to monitor conversion of natural and planted forest to agriculture for EUDR purposes.</p>



<p>For a full description of how the natural forest map relates to EUDR, see <a href="https://landcarbonlab.org/insights/natural-lands-map-companies-no-conversion-targets/" target="_blank" rel="noreferrer noopener">this blog</a>. For more information on EUDR definitions and how GFW can support EUDR monitoring, read <a href="https://www.globalforestwatch.org/blog/data-and-tools/satellite-data-eu-regulation-deforestation-free-supply-chains/">part one</a> and <a href="https://www.globalforestwatch.org/blog/data-and-tools/monitoring-forest-degradation-eudr/">two</a> of our blog series.</p>



<p>Whenever using the SBTN Natural Lands Map, including the natural forests map on GFW, users should also be aware of a few limitations of the data:</p>



<ul type="1">
<li>Overestimation of natural forest in agroforestry systems and in temperate and boreal forests: Agroforestry systems, especially those where crops are grown under shade trees (such as some types of coffee and cocoa systems), are difficult to differentiate from natural forests in satellite imagery, and therefore available data on their extent is limited. Similarly, publicly available data on planted forests is limited in many temperate and boreal regions. As a result, these areas may be included as natural forest in the Map.</li>



<li>Inconsistent forest definitions in certain regions: Because the SBTN Natural Lands Map team prioritized local and regional input data over global data during the compilation and harmonization process when the Map was developed, there may be inconsistencies in the height and canopy density thresholds used to define forests in some regions where local data was used, including Brazil, the Amazon, Atlantic Forest, Pampa, Chaco, Indonesia, South Africa and New Zealand.</li>



<li>Temporal inconsistencies: The Map is as close to the year 2020 as possible, but due to limited data availability, some input data is from earlier time periods (e.g., 2015).</li>



<li>Resolution inconsistencies: Most input data used to create the Map is at least 30-meter resolution, but in some cases lower resolution data was used when higher resolution data was unavailable. These inconsistencies create data artifacts (sharp boundary lines that are only due to resolutions) in some areas.</li>
</ul>



<p>See more information on these limitations in the <a href="https://sciencebasedtargetsnetwork.org/wp-content/uploads/2023/05/Technical-Guidance-2023-Step3-Land-v0.3-Natural-Lands-Map.pdf" target="_blank" rel="noreferrer noopener">technical documentation</a>.</p>



<h2 class="wp-block-heading">What’s next for the natural forest data on GFW?</h2>



<p>Currently, <a href="https://sciencebasedtargetsnetwork.org/wp-content/uploads/2023/05/Technical-Guidance-2023-Step3-Land-v0.3-Natural-Lands-Map.pdf" target="_blank" rel="noreferrer noopener">version 0.3</a> of the natural forest map is available on GFW. In the coming months, we will update GFW to <a href="https://sciencebasedtargetsnetwork.org/wp-content/uploads/2024/09/Technical-Guidance-2024-Step3-Land-v1-Natural-Lands-Map.pdf" target="_blank" rel="noreferrer noopener">version 1</a>, which is now available on the <a href="https://datasets.wri.org/datasets/sbtn-natural-lands-map?map=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" target="_blank" rel="noreferrer noopener">WRI Open Data Portal</a>, <a href="https://developers.google.com/earth-engine/datasets/catalog/WRI_SBTN_naturalLands_v1" target="_blank" rel="noreferrer noopener">Google Earth Engine</a> and <a href="https://pro-news.globalforestwatch.org/news-events/how-gfw-pro-supports-eudr-compliance" target="_blank" rel="noreferrer noopener">GFW Pro</a>. We plan to continue to update the data on GFW as future versions of the SBTN Natural Lands Map become available.</p>



<p>We also plan to release additional analysis capabilities with the data, such as the option to filter deforestation and fire alerts to within natural forests. Stay tuned for future updates!</p>



<p></p>



<p>* <em>Note that the validation points were labeled based on definitions of natural forests from v1 of the SBTN Natural Lands Map, which includes more regional data sets and their corresponding definitions, as well as an additional filtering of local savanna classes to remove vegetation shorter than 5 meters from the forest class. This may affect the validation of the v0.3 natural forest layer because of variation in definitions of natural forest where local data was used. See the <a href="https://sciencebasedtargetsnetwork.org/wp-content/uploads/2024/09/Technical-Guidance-2024-Step3-Land-v1-Natural-Lands-Map.pdf" target="_blank" rel="noreferrer noopener">v1 technical note change log</a> to see the full list of changes.</em></p>



<p><em>** We generally </em><a href="https://landcarbonlab.org/insights/natural-lands-map-companies-no-conversion-targets/" target="_blank" rel="noreferrer noopener"><em>caution against</em></a><em> using the SBTN Natural Lands Map to quantify the area of natural lands due to data limitations, including overestimation of natural lands. However, data on natural and non-natural forest cover is much more widely available than other land covers, such as natural grassland and pasture, so we consider the natural forest classes to provide a reasonable approximation of natural forest area. Users should still be aware of the limitations of the data when using it to quantify area.</em></p>



<style type="text/css"></style>

New natural forest data on GFW reveals where the world’s natural forests are located and provides a baseline for monitoring loss within them.

New Data Distinguishes Natural Forests from Other Tree Cover

Comment

Global Forest Watch Content

https://content.globalforestwatch.org/wp-content/uploads/2018/06/gfwGlobal.png

https://secure.gravatar.com/avatar/be19f67e91ba81d93d32ae47706dcbb8?s=96&d=mm&r=g

Paige Johnson

Oil palm plantation in Sentabai Village, West Kalimantan, Indonesia.

Jessica Richter

Nancy Harris

<p>Determining the boundaries of different forest types allows us to better understand how forests are changing over time. A longstanding limitation of the tree cover data available on Global Forest Watch (GFW) has been the inability to differentiate planted forests and tree crops from primary and secondary natural forests. There is no globally consistent data available on the location of planted trees, making it difficult to determine whether tree cover change occurred in natural forest or if it was simply harvesting or management activities in planted trees.</p>



<p>In 2019, GFW and World Resources Institute published the original Spatial Database of Planted Trees (SDPT) to fill this data gap by compiling data sources on planted tree boundaries across the globe. This effort allowed GFW users to view the location of planted trees on the GFW map and focus tree cover loss and gain statistics on change inside or outside planted tree areas.</p>



<p>Now, a new version of the SDPT is <a href="https://gfw.global/2Qbv7fM">available on GFW</a>, which provides broader geographic coverage and several key updates that improve its application to global forest monitoring and management.</p>



<h2 class="wp-block-heading">What is the Spatial Database of Planted Trees (SDPT)?</h2>



<p>The SDPT aims to spatially differentiate planted forests and tree crops from natural and seminatural forests on a global scale. We conducted extensive research and outreach to governments, civil society organizations and researchers to compile, synthesize and integrate national and regional data sources of the world’s planted trees into the SDPT.</p>



<figure class="wp-block-image size-large"><img decoding="async" loading="lazy" width="2048" height="1495" src="https://content.globalforestwatch.org/wp-content/uploads/2024/11/23_GRAPHIC-GFW-blog-SPDT-spatial-database-planted-trees_V2-2048x1495.png" alt="Table showing definitions of planted trees, planted forest and tree crops as defined in the Spatial Database of Planted Trees" class="wp-image-18224" srcset="https://content.globalforestwatch.org/wp-content/uploads/2024/11/23_GRAPHIC-GFW-blog-SPDT-spatial-database-planted-trees_V2-2048x1495.png 2048w, https://content.globalforestwatch.org/wp-content/uploads/2024/11/23_GRAPHIC-GFW-blog-SPDT-spatial-database-planted-trees_V2-1024x748.png 1024w, https://content.globalforestwatch.org/wp-content/uploads/2024/11/23_GRAPHIC-GFW-blog-SPDT-spatial-database-planted-trees_V2-768x561.png 768w, https://content.globalforestwatch.org/wp-content/uploads/2024/11/23_GRAPHIC-GFW-blog-SPDT-spatial-database-planted-trees_V2-1536x1121.png 1536w" sizes="(max-width: 2048px) 100vw, 2048px" /></figure>



<p>The SDPT includes key information on each integrated data set, including the planted tree type (planted forest vs. tree crop), original data source information and carbon removal factors — or the rate at which planted trees remove atmospheric carbon dioxide and store it in biomass. Where available, the SDPT also includes species information, as well as the method and year(s) of imagery used to determine the planted tree boundary.</p>



<p>The SDPT was established as a living database that will continue to evolve as new data becomes available. The recent publication of an updated version of SDPT (<a href="https://www.wri.org/research/spatial-database-planted-trees-sdpt-version-2" target="_blank" rel="noreferrer noopener">v2.0</a>) reflects this aim and provides significant updates to the initial (<a href="https://www.wri.org/research/spatial-database-planted-trees-sdpt-version-10" target="_blank" rel="noreferrer noopener">v1.0</a>) database.</p>



<p>Users can download the v2.0 database <a href="https://gfw2-data.s3.amazonaws.com/plantations/sdpt/sdpt_v2.zip">here</a>. The SDPT is also available as the <a href="https://gfw.global/2Qbv7fM">tree plantation</a> layer on the GFW map and is included in several analyses on the GFW dashboard.</p>



<style>
.css-19vc5o2 aside {
   width: 100% !important;
float: none !important;
margin-left:
0px
!important;
}
</style>
<aside id="textbox"><em><b>Comparing the SDPT to Forest Resources Assessment Reporting</b><p></p>
The record of planted forest statistics provided in the <a href="https://www.globalforestwatch.org/blog/data-and-research/global-forest-watch-and-the-forest-resources-assessment-explained-in-5-graphics-2/" target="_blank" rel="noopener noreferrer">Food and Agriculture Organization of the United Nations’ (FAO)</a> <a href="https://www.fao.org/3/ca9825en/ca9825en.pdf" target="_blank" rel="noopener noreferrer">Forest Resources Assessment (FRA)</a> every five years serves as an important baseline of comparison to the SDPT. However, data sets included in the SDPT may not adhere to the definitions used by FAO as there are no universal definitions for the terms “planted trees,” “planted forests” and “tree crops.” In the SDPT, we focused on including data sets, or classes within data sets, that used the terms “plantation,” “plantation forest” or “planted forest,” which we classified under the term planted forest. The SDPT also includes maps of agroforestry systems and various perennial tree crops such as rubber, oil palm, coffee, coconut, cocoa and orchards, which we classified under the term tree crop. FAO instead uses the terms planted forest and plantation, defining planted forest as “forest predominantly composed of trees established through planting and/or deliberate seeding” and plantation as “planted forest that is intensively managed and meet ALL the following criteria at planting and stand maturity: one or two species, even age class, and regular spacing.” FAO’s reporting of planted forests does not include tree crop systems, so we compared FAO’s planted forest statistics to SDPT data classified as planted forest. The SDPT is not intended to resolve definitional debate surrounding these terms but rather to synthesize data.
</em> </aside>



<h2 class="wp-block-heading">What’s new in SDPT version 2.0?</h2>





<p>A range of <a href="https://www.wri.org/research/spatial-database-planted-trees-sdpt-version-2" target="_blank" rel="noreferrer noopener">improvements have been made</a> to the SDPT in v2.0 based on new global and national data sources. These include:</p>



<p><strong>1. Near-global coverage</strong></p>



<p>Coverage in v2.0 has expanded from 82 countries in v1.0 to a total of 158 countries represented. Based on the newly compiled data, v2.0 reports approximately 329 million hectares (Mha) of planted trees worldwide in 2020, made up of 264 Mha of planted forests and 65 Mha of tree crops. This represents 8% of total global tree cover (an area roughly the size of India).</p>



<div class="flourish-embed flourish-map" data-src="visualisation/19737652?163031"><script src="https://public.flourish.studio/resources/embed.js"></script><noscript><img decoding="async" src="https://public.flourish.studio/visualisation/19737652/thumbnail" width="100%" alt="map visualization" /></noscript></div>



<div class="flourish-embed flourish-chart" data-src="visualisation/16052609?163031"><script src="https://public.flourish.studio/resources/embed.js"></script><noscript><img decoding="async" src="https://public.flourish.studio/visualisation/16052609/thumbnail" width="100%" alt="chart visualization" /></noscript></div>



<p><strong>2. Tree species information</strong></p>



<p>SDPT v2.0 added or updated tree species information for 39 countries and now includes species-level information for 52 out of the 158 countries covered. Users can view available species information by <a href="https://gfw2-data.s3.amazonaws.com/plantations/sdpt/sdpt_v2.zip">downloading the database</a> and referencing the <em>sciName</em> data attribute.</p>



<p><strong>3. Planting year information</strong></p>



<p>A new planting year attribute has been added to the database that provides the latest approximate year that an individual area was planted. Researchers at Tsinghua University were able to detect the year of planting in SDPT v2.0 using the <a href="https://www.nature.com/articles/s41597-022-01260-2" target="_blank" rel="noreferrer noopener">LandTrendr algorithm</a> on a time series of Landsat images from 1982 to 2020. Results from this analysis were added to each planted tree area and are available by <a href="https://gfw2-data.s3.amazonaws.com/plantations/sdpt/sdpt_v2.zip">downloading the database</a> and referencing the <em>plantedYear </em>data attribute.</p>



<p><strong>4. Carbon removal factors</strong></p>



<p>Carbon removal factors, or carbon sequestration rates, have been assigned to planted trees represented in SDPT v2.0 that were not included in v1.0. In addition to carbon removal factors for aboveground biomass, belowground biomass and the standard deviations of removal factors, the database includes notes on methods and sources for each removal factor. Users can access carbon removal factor information by <a href="https://gfw2-data.s3.amazonaws.com/plantations/sdpt/sdpt_v2.zip">downloading the database</a> and referencing attributes such as <em>removalFactorAGCBGCMgHaYr</em>, <em>removalFactorAGCBGCstdevMgHaYr</em>, <em>removalFactorSource</em> and <em>removalFactorNotes</em>.</p>



<h2 class="wp-block-heading">How can the SDPT be used?</h2>



<p>The SDPT exists first and foremost to differentiate the geographic extent of planted trees from natural forests on a global scale, leading to easier identification of “natural forest” areas. Already, SDPT v2.0 is being used as an input layer into the Science Based Targets Network (SBTN) <a href="https://landcarbonlab.org/insights/natural-lands-map-companies-no-conversion-targets/" target="_blank" rel="noreferrer noopener">Natural Lands Map</a> to filter out planted forests and produce a natural forest class.</p>



<p>Specific SDPT attributes, such as species information, also have important global applications. Enforcement actors, such as United States government customs agencies, can use SDPT species information to expedite inspections of timber shipments. The database can help confirm whether there are other planted timber forests of the same species within the stated country of origin, as this impacts the legality of the import. The removal factors can be used to estimate carbon sequestration rates over large areas of planted trees where plot-level carbon measurements are not available.</p>



<p>Beyond these specific applications, the SDPT has implications for global forest monitoring and management. It can aid in tracking where and when planted trees may be replacing natural forests and help monitor the impact of these planted forests across diverse landscapes. This can lead to more effective tracking of national and global progress towards major international commitments associated with forests, climate and biodiversity.</p>



<h2 class="wp-block-heading">Future plans for the SDPT</h2>



<p>While major challenges remain in mapping planted trees, we expect to see continued, significant advances in large-scale planted tree mapping as well as improvements to national- and regional-scale data over time. The SDPT will continue to integrate these sources across all scales to provide users with the most detailed spatial information available. Incorporating more species-specific data into the SDPT is a key part of this effort. This includes species information taken directly from image-interpreted data sets as well as integrating non-spatial species information derived from literature reviews.</p>



<p>Users can <a href="https://www.wri.org/research/spatial-database-planted-trees-sdpt-version-2" target="_blank" rel="noreferrer noopener">explore the tree plantations layer on GFW</a> and can <a href="https://gfw2-data.s3.amazonaws.com/plantations/sdpt/sdpt_v2.zip">download the data</a><em>. </em>For more information on the SDPT v2.0 update, <a href="https://www.wri.org/research/spatial-database-planted-trees-sdpt-version-2">read the technical note here</a>.</p>



<p><em>The SDPT compiles data from a variety of sources. If you have spatial data on planted trees that you think should be incorporated into the data set, </em><a href="https://forms.gle/pUzJy72N7szVMydg9" target="_blank" rel="noreferrer noopener"><em>please let us know here</em></a><em>.</em></p>



<p><em>Additional minor updates have been made to the SDPT in a version 2.1 incorporating new data in Brazil, Cambodia, Cote d’Ivoire and Ghana. </em><a href="https://gfw-files.s3.amazonaws.com/plantations/SDPT_v2.1/sdpt_v21_v09152024_public.gdb.zip"><em>Access the latest version here</em></a><em>.</em></p>



<style type="text/css"></style>

Updated Planted Trees Map Reaches Near-Global Coverage

<p>Determining the boundaries of different forest types allows us to better understand how forests are changing over time. A longstanding limitation of the tree cover data available on Global Forest Watch (GFW) has been the inability to differentiate planted forests and tree crops from primary and secondary natural forests. There is no globally consistent data available on the location of planted trees, making it difficult to determine whether tree cover change occurred in natural forest or if it was simply harvesting or management activities in planted trees.</p>



<p>In 2019, GFW and World Resources Institute published the original Spatial Database of Planted Trees (SDPT) to fill this data gap by compiling data sources on planted tree boundaries across the globe. This effort allowed GFW users to view the location of planted trees on the GFW map and focus tree cover loss and gain statistics on change inside or outside planted tree areas.</p>



<p>Now, a new version of the SDPT is <a href="https://gfw.global/2Qbv7fM">available on GFW</a>, which provides broader geographic coverage and several key updates that improve its application to global forest monitoring and management.</p>



<h2 class="wp-block-heading">What is the Spatial Database of Planted Trees (SDPT)?</h2>



<p>The SDPT aims to spatially differentiate planted forests and tree crops from natural and seminatural forests on a global scale. We conducted extensive research and outreach to governments, civil society organizations and researchers to compile, synthesize and integrate national and regional data sources of the world’s planted trees into the SDPT.</p>



<figure class="wp-block-image size-large"><img src="https://content.globalforestwatch.org/wp-content/uploads/2024/11/23_GRAPHIC-GFW-blog-SPDT-spatial-database-planted-trees_V2-2048x1495.png" alt="Table showing definitions of planted trees, planted forest and tree crops as defined in the Spatial Database of Planted Trees" class="wp-image-18224"/></figure>



<p>The SDPT includes key information on each integrated data set, including the planted tree type (planted forest vs. tree crop), original data source information and carbon removal factors — or the rate at which planted trees remove atmospheric carbon dioxide and store it in biomass. Where available, the SDPT also includes species information, as well as the method and year(s) of imagery used to determine the planted tree boundary.</p>



<p>The SDPT was established as a living database that will continue to evolve as new data becomes available. The recent publication of an updated version of SDPT (<a href="https://www.wri.org/research/spatial-database-planted-trees-sdpt-version-2" target="_blank" rel="noreferrer noopener">v2.0</a>) reflects this aim and provides significant updates to the initial (<a href="https://www.wri.org/research/spatial-database-planted-trees-sdpt-version-10" target="_blank" rel="noreferrer noopener">v1.0</a>) database.</p>



<p>Users can download the v2.0 database <a href="https://gfw2-data.s3.amazonaws.com/plantations/sdpt/sdpt_v2.zip">here</a>. The SDPT is also available as the <a href="https://gfw.global/2Qbv7fM">tree plantation</a> layer on the GFW map and is included in several analyses on the GFW dashboard.</p>



<style>
.css-19vc5o2 aside {
   width: 100% !important;
float: none !important;
margin-left:
0px
!important;
}
</style>
<aside id="textbox"><em><b>Comparing the SDPT to Forest Resources Assessment Reporting</b><p></p>
The record of planted forest statistics provided in the <a href="https://www.globalforestwatch.org/blog/data-and-research/global-forest-watch-and-the-forest-resources-assessment-explained-in-5-graphics-2/" target="_blank" rel="noopener noreferrer">Food and Agriculture Organization of the United Nations' (FAO)</a> <a href="https://www.fao.org/3/ca9825en/ca9825en.pdf" target="_blank" rel="noopener noreferrer">Forest Resources Assessment (FRA)</a> every five years serves as an important baseline of comparison to the SDPT. However, data sets included in the SDPT may not adhere to the definitions used by FAO as there are no universal definitions for the terms “planted trees,” “planted forests” and “tree crops.” In the SDPT, we focused on including data sets, or classes within data sets, that used the terms “plantation,” “plantation forest” or “planted forest,” which we classified under the term planted forest. The SDPT also includes maps of agroforestry systems and various perennial tree crops such as rubber, oil palm, coffee, coconut, cocoa and orchards, which we classified under the term tree crop. FAO instead uses the terms planted forest and plantation, defining planted forest as “forest predominantly composed of trees established through planting and/or deliberate seeding” and plantation as “planted forest that is intensively managed and meet ALL the following criteria at planting and stand maturity: one or two species, even age class, and regular spacing.” FAO’s reporting of planted forests does not include tree crop systems, so we compared FAO’s planted forest statistics to SDPT data classified as planted forest. The SDPT is not intended to resolve definitional debate surrounding these terms but rather to synthesize data.
</em> </aside>



<h2 class="wp-block-heading">What’s new in SDPT version 2.0?</h2>





<p>A range of <a href="https://www.wri.org/research/spatial-database-planted-trees-sdpt-version-2" target="_blank" rel="noreferrer noopener">improvements have been made</a> to the SDPT in v2.0 based on new global and national data sources. These include:</p>



<p><strong>1. Near-global coverage</strong></p>



<p>Coverage in v2.0 has expanded from 82 countries in v1.0 to a total of 158 countries represented. Based on the newly compiled data, v2.0 reports approximately 329 million hectares (Mha) of planted trees worldwide in 2020, made up of 264 Mha of planted forests and 65 Mha of tree crops. This represents 8% of total global tree cover (an area roughly the size of India).</p>



<div class="flourish-embed flourish-map" data-src="visualisation/19737652?163031"><script src="https://public.flourish.studio/resources/embed.js"></script><noscript><img src="https://public.flourish.studio/visualisation/19737652/thumbnail" width="100%" alt="map visualization" /></noscript></div>



<div class="flourish-embed flourish-chart" data-src="visualisation/16052609?163031"><script src="https://public.flourish.studio/resources/embed.js"></script><noscript><img src="https://public.flourish.studio/visualisation/16052609/thumbnail" width="100%" alt="chart visualization" /></noscript></div>



<p><strong>2. Tree species information</strong></p>



<p>SDPT v2.0 added or updated tree species information for 39 countries and now includes species-level information for 52 out of the 158 countries covered. Users can view available species information by <a href="https://gfw2-data.s3.amazonaws.com/plantations/sdpt/sdpt_v2.zip">downloading the database</a> and referencing the <em>sciName</em> data attribute.</p>



<p><strong>3. Planting year information</strong></p>



<p>A new planting year attribute has been added to the database that provides the latest approximate year that an individual area was planted. Researchers at Tsinghua University were able to detect the year of planting in SDPT v2.0 using the <a href="https://www.nature.com/articles/s41597-022-01260-2" target="_blank" rel="noreferrer noopener">LandTrendr algorithm</a> on a time series of Landsat images from 1982 to 2020. Results from this analysis were added to each planted tree area and are available by <a href="https://gfw2-data.s3.amazonaws.com/plantations/sdpt/sdpt_v2.zip">downloading the database</a> and referencing the <em>plantedYear </em>data attribute.</p>



<p><strong>4. Carbon removal factors</strong></p>



<p>Carbon removal factors, or carbon sequestration rates, have been assigned to planted trees represented in SDPT v2.0 that were not included in v1.0. In addition to carbon removal factors for aboveground biomass, belowground biomass and the standard deviations of removal factors, the database includes notes on methods and sources for each removal factor. Users can access carbon removal factor information by <a href="https://gfw2-data.s3.amazonaws.com/plantations/sdpt/sdpt_v2.zip">downloading the database</a> and referencing attributes such as <em>removalFactorAGCBGCMgHaYr</em>, <em>removalFactorAGCBGCstdevMgHaYr</em>, <em>removalFactorSource</em> and <em>removalFactorNotes</em>.</p>



<h2 class="wp-block-heading">How can the SDPT be used?</h2>



<p>The SDPT exists first and foremost to differentiate the geographic extent of planted trees from natural forests on a global scale, leading to easier identification of “natural forest” areas. Already, SDPT v2.0 is being used as an input layer into the Science Based Targets Network (SBTN) <a href="https://landcarbonlab.org/insights/natural-lands-map-companies-no-conversion-targets/" target="_blank" rel="noreferrer noopener">Natural Lands Map</a> to filter out planted forests and produce a natural forest class.</p>



<p>Specific SDPT attributes, such as species information, also have important global applications. Enforcement actors, such as United States government customs agencies, can use SDPT species information to expedite inspections of timber shipments. The database can help confirm whether there are other planted timber forests of the same species within the stated country of origin, as this impacts the legality of the import. The removal factors can be used to estimate carbon sequestration rates over large areas of planted trees where plot-level carbon measurements are not available.</p>



<p>Beyond these specific applications, the SDPT has implications for global forest monitoring and management. It can aid in tracking where and when planted trees may be replacing natural forests and help monitor the impact of these planted forests across diverse landscapes. This can lead to more effective tracking of national and global progress towards major international commitments associated with forests, climate and biodiversity.</p>



<h2 class="wp-block-heading">Future plans for the SDPT</h2>



<p>While major challenges remain in mapping planted trees, we expect to see continued, significant advances in large-scale planted tree mapping as well as improvements to national- and regional-scale data over time. The SDPT will continue to integrate these sources across all scales to provide users with the most detailed spatial information available. Incorporating more species-specific data into the SDPT is a key part of this effort. This includes species information taken directly from image-interpreted data sets as well as integrating non-spatial species information derived from literature reviews.</p>



<p>Users can <a href="https://www.wri.org/research/spatial-database-planted-trees-sdpt-version-2" target="_blank" rel="noreferrer noopener">explore the tree plantations layer on GFW</a> and can <a href="https://gfw2-data.s3.amazonaws.com/plantations/sdpt/sdpt_v2.zip">download the data</a><em>. </em>For more information on the SDPT v2.0 update, <a href="https://www.wri.org/research/spatial-database-planted-trees-sdpt-version-2">read the technical note here</a>.</p>



<p><em>The SDPT compiles data from a variety of sources. If you have spatial data on planted trees that you think should be incorporated into the data set, </em><a href="https://forms.gle/pUzJy72N7szVMydg9" target="_blank" rel="noreferrer noopener"><em>please let us know here</em></a><em>.</em></p>



<p><em>Additional minor updates have been made to the SDPT in a version 2.1 incorporating new data in Brazil, Cambodia, Cote d’Ivoire and Ghana. </em><a href="https://gfw-files.s3.amazonaws.com/plantations/SDPT_v2.1/sdpt_v21_v09152024_public.gdb.zip"><em>Access the latest version here</em></a><em>.</em></p>



<style type="text/css"></style>

A new version of the Spatial Database of Planted Trees provides near-global coverage and other key updates.

Forest Insights

Forest Insights | Global Forest Watch Blog

deforestation

deforestation | Global Forest Watch Blog

policy

policy | Global Forest Watch Blog

Viviana Zalles

Fred Stolle

Matt Hansen

How Rethinking ‘What Is a Forest’ Can Result in More Effective Conservation

How we define "what is a forest?" can have major implications for how forests — and more importantly, changes to forests — are measured.

How Rethinking ‘What Is a Forest’ Can Result in More Effective Conservation | Forest Insights | Global Forest Watch Blog

Ripe Cocoa pods from a cocoa farm in Ghana.

Cocoa

Cocoa | Global Forest Watch Blog

commodities

commodities | Global Forest Watch Blog

<p>Martina Schneider is a GIS Analyst with Global Forest Watch. She supports the GFW team in preparing data and developing content for the GFW platform and web application.</p>



<p>As a graduate student, Martina specialized in applying GIS in the fields of environmental science and natural resources management. Specifically, she studied wildfire ecology and environmental factors that contribute to risk of ignition. In addition, Martina worked at Ducks Unlimited, the world's leader in wetland and waterfowl conservation, where she was part of a team updating the National Wetlands Inventory in Minnesota.</p>



<p>Martina holds an M.S. in Geographic Information Systems from Eastern Michigan University and a B.S. in Geography from Texas A&amp;M University.</p>



<p>She lives in Baltimore, MD with her husband and dog.</p>

Martina Schneider

Caroline Winchester

Sofia Tenorio Fenton

<p>The global demand for chocolate has had a significant impact on West African forests. Approximately 70% of the global supply of cocoa originates from Ghana and Côte d’Ivoire, where production sites have replaced millions of hectares of forest. Between 2001 and 2015 alone, <a href="https://research.wri.org/gfr/forest-extent-indicators/deforestation-agriculture#how-much-forest-has-been-replaced-by-cocoa" target="_blank" rel="noreferrer noopener">cocoa was linked to</a> the loss of a third of Ghana’s forest area and a quarter of Côte d’Ivoire’s.</p>



<p>The <a href="https://www.worldcocoafoundation.org/initiative/cocoa-forests-initiative/" target="_blank" rel="noreferrer noopener">Cocoa & Forests Initiative (CFI)</a>, facilitated by World Cocoa Foundation (WCF) and the <a href="https://www.idhsustainabletrade.com/initiative/cocoa-and-forests/" target="_blank" rel="noreferrer noopener">Sustainable Trade Initiative (IDH)</a>, was established in 2017 in an effort to end cocoa-driven deforestation and restore forest areas. Companies that joined the CFI agreed to map their supply chains, align methodologies on deforestation risk assessment and contribute to transparent monitoring and reporting. To support these efforts, World Resources Institute (WRI) partnered with WCF and 19 major cocoa and chocolate companies to create two new resources:*</p>



<ul>
<li>The <strong>West Africa Cocoa dataset (WAC)</strong>, a database of mapped cocoa plot boundaries in the direct supply chains of participating companies that provides an extensive view of mapped cocoa production in Côte d’Ivoire and Ghana. The public version of the dataset is <a href="https://gfw.global/4bEEES2">available on Global Forest Watch (GFW)</a> as a cocoa plot heat map that further anonymizes data to protect the privacy of farmers and companies.</li>



<li>The <strong>Cocoa Deforestation Risk Assessment (Cocoa DRA)</strong>, a map that provides a standardized view of cocoa-driven deforestation risk across the cocoa growing region of these two countries. The <a href="https://data.globalforestwatch.org/documents/gfw::cocoa-deforestation-risk-assessment/about">Cocoa DRA map layer</a> is available for download on GFW.</li>
</ul>



<p>These resources are the result of an unprecedented collaboration among companies to share highly sensitive cocoa plot data, and an extensive ethics and legal review to ensure the protection of farmers’ data throughout the process. For the first time, companies and other stakeholders in the cocoa sector have a shared view of priority areas in Ghana and Côte d’Ivoire for coordinated efforts toward a deforestation-free cocoa sector.</p>



<style>
.css-19vc5o2 aside {
   width: 100% !important;
float: none !important;
margin-left:
0px
!important;
}
</style>
<aside id="textbox"><em><b>From Palm Oil to Cocoa</b><p></p>
The development of the WAC and the Cocoa DRA builds on the success of similar products developed to support the palm oil sector: the <a href="https://www.globalforestwatch.org/blog/commodities/improved-universal-mill-list-expands-partnerships-improves-transparency-in-palm-oil-industry/" target="_blank" rel="noopener noreferrer">Universal Mill List (UML)</a>, a harmonized database of palm oil mill locations, and the <a href="https://www.wri.org/research/palm-risk-assessment-methodology-prioritizing-areas-landscapes-and-mills" target="_blank" rel="noopener noreferrer">Palm Risk Tool</a>, a method of identifying deforestation risks in their vicinity. However, differing supply chain dynamics between palm oil and cocoa were considered, such as:<li>
The locations of palm oil mills in the UML are represented as points while the cocoa plots in the WAC are represented by polygons. Further, the UML maps industrial scale infrastructure, whereas the WAC maps smallholder production. This raised additional privacy concerns, and steps were taken to avoid violating farmer privacy and safeguarding farmer livelihoods.<li>
The Palm Risk Tool analyzes deforestation risk by creating a buffer around each mill point to estimate a potential sourcing area. Oil palm fruit must be delivered to a mill quickly or the quality of the oil will decrease. This places a geographic constraint on how far oil palm fruit can travel to a mill, and stakeholders in the sector rely on a standardized sourcing radius of 50-km for their risk assessments. Cocoa does not have a geographic constraint nor an agreed upon estimation of a sourcing area, meaning that cocoa collected at a central point of purchase could have been grown anywhere in these two countries.<li>
The Palm Risk Tool was developed to assess deforestation risks in the vicinity of a specific type of infrastructure: palm oil mills. The Cocoa DRA was developed to be more flexible, enabling analysis of multiple areas of interest such as the plots in the WAC, cooperatives or other types of collection points and/or warehouses.<ul>

</em> </aside>



<figure class="wp-block-image size-full"><img decoding="async" loading="lazy" width="1175" height="783" src="https://content.globalforestwatch.org/wp-content/uploads/2024/02/Cocoa-pod.jpg" alt="Cocoa pod" class="wp-image-16723" srcset="https://content.globalforestwatch.org/wp-content/uploads/2024/02/Cocoa-pod.jpg 1175w, https://content.globalforestwatch.org/wp-content/uploads/2024/02/Cocoa-pod-1024x682.jpg 1024w, https://content.globalforestwatch.org/wp-content/uploads/2024/02/Cocoa-pod-768x512.jpg 768w" sizes="(max-width: 1175px) 100vw, 1175px" /><figcaption class="wp-element-caption">Cocoa pod. Credit: Caroline Winchester, WRI.</figcaption></figure>



<h2 class="wp-block-heading"><strong>What is the West Africa</strong> <strong>Cocoa dataset?</strong></h2>



<p>The West Africa Cocoa dataset (WAC) aggregates data from 2021 on cocoa plot locations (a cocoa plot is a subset of a cocoa farm, and a single farm may be represented by multiple plots) in the direct supply chains of 19 cocoa and chocolate companies operating in Ghana and Côte d’Ivoire. Project participants developed and agreed on a data collation protocol and data sharing agreement that outlined submission requirements and the procedure for cleaning and aggregating individual company data sets. To be included in the WAC, each plot had to be current, actively growing cocoa and occupying a single parcel of land. WRI then reviewed, cleaned and aggregated the data.</p>



<p>While the cocoa plot polygon data is considered sensitive and will not be available publicly (see more below), we visualized the data as a heat map, <a href="https://gfw.global/4bEEES2">available on GFW</a>, which avoids disclosing the precise boundaries of the plots in the WAC while giving users access to information on the approximate location of cocoa plots. Additionally, we distributed the plot data to impartial research institutions for use as training data to create derivative products; specifically, public cocoa production extent maps.</p>



<p>These strategies make the data publicly available to support companies and other users in more robust cocoa mapping and monitoring while safeguarding the privacy of individual farmers.</p>



<style>
.css-19vc5o2 aside {
   width: 100% !important;
float: none !important;
margin-left:
0px
!important;
}
</style>
<aside id="textbox"><em><b>Protecting the Privacy of Cocoa Farmers and Companies</b><p></p>

The raw data used to create the WAC included sensitive supply chain information from both companies and farmers. Cocoa is primarily grown by smallholder farmers who live near the cocoa plots they manage. A public release of this raw data would show the boundaries of specific cocoa plots and could allow individual farmers and their homes to be identified. Further, since there are over 2 million farmers across West Africa, it is difficult to know whether they are all fully aware of how data pertaining to them is being used. <br><br>
To protect farmer and company privacy, the WAC data collection protocol and data sharing agreement were informed by an ethics review to ensure that potentially sensitive information was respected and handled properly, and by a legal review to ensure the data privacy laws of Ghana and Côte d’Ivoire were followed. Based off these reviews, and in order to protect farmer privacy, no personal data was included in the WAC or shared with WRI, and a secure method of transferring data among all parties was established. With farmer privacy in mind, the parties involved also decided that precise locations of cocoa plot boundaries would not be made publicly available.

</em> </aside>



<h2 class="wp-block-heading">What is the Cocoa Deforestation Risk Assessment?</h2>



<p>The Cocoa Deforestation Risk Assessment (Cocoa DRA) is a map that assesses the risk of future deforestation due to new or expanding cocoa production. The Cocoa DRA aligns various methods for assessing risk to provide a standardized view that companies and other stakeholders can use to make decisions around reducing deforestation. It can be used to proactively and efficiently identify deforestation risk across a supply chain that includes tens of thousands of geographically dispersed producers. Once risk has been identified, Cocoa DRA users can implement interventions for promoting sustainable production, protecting remaining forest cover and improving farmer livelihoods.</p>



<p>To create the map, we analyzed spatial variables that influence cocoa-linked deforestation including recent forest loss, proximity to cocoa, accessibility and suitability for cocoa production. The present-day extent of natural forest was approximated using the University of Maryland’s <a href="https://www.science.org/doi/10.1126/science.1244693" target="_blank" rel="noreferrer noopener">tree canopy cover</a> data set and removing cocoa plots mapped in the WAC, <a href="https://essd.copernicus.org/articles/13/1211/2021/" target="_blank" rel="noreferrer noopener">other data sets</a> that <a href="https://www.wri.org/research/spatial-database-planted-trees-sdpt-version-10" target="_blank" rel="noreferrer noopener">map planted trees</a> and historic tree cover loss.</p>



<p>Primary and intact forests in West Africa are limited, so to ensure the Cocoa DRA comprehensively assesses where future cocoa-linked deforestation is likely to occur in <em>all</em> natural forests, its scope includes natural forests both inside and outside of protected areas as well as naturally regenerating forests.</p>



<p>The Cocoa DRA translates risk into five priority rankings at a 1-km resolution (high, mid-high, medium, mid-low and low priority) indicating the likelihood of future deforestation associated with cocoa production. Cocoa DRA users can assess risk at different scales, including groups of farms, supply sheds around cocoa bean collection points and other types of landscapes.</p>



<p>Users can download the <a href="https://data.globalforestwatch.org/documents/gfw::cocoa-deforestation-risk-assessment/about">Cocoa DRA map layer</a> on GFW.</p>



<h2 class="wp-block-heading">What have we learned from the WAC and Cocoa DRA?</h2>



<p>Across Ghana and Côte d’Ivoire, the WAC maps 840,000 unique cocoa plots in 2021. Though cocoa plots tend to be small (most plots were between 1 to 2 hectares), the plots included in the WAC cover approximately 1.5 million hectares of planted cocoa area.</p>



<p>Cocoa plot data is most comprehensive in western Côte d’Ivoire as well as near the shared border between Côte d’Ivoire and Ghana. The forests in these areas have been <a href="https://research.wri.org/gfr/forest-extent-indicators/deforestation-agriculture#how-much-forest-has-been-replaced-by-cocoa" target="_blank" rel="noreferrer noopener">particularly impacted by cocoa production</a>. We found that the Cocoa DRA showed that these same regions had a significant amount of high priority areas (areas with higher likelihood of future cocoa-related deforestation). This is most evident in Côte d’Ivoire near the border of Liberia, while in western Ghana the landscape is more mixed.</p>



<figure class="wp-block-image size-full is-resized"><img decoding="async" loading="lazy" width="1430" height="868" src="https://content.globalforestwatch.org/wp-content/uploads/2024/02/Cocoa-Deforestation-Risk-Assessment-Priority-2023.png" alt="Cocoa deforestation risk assessment priority map for 2023" class="wp-image-16731" style="aspect-ratio:1.6474654377880185;width:840px;height:auto" srcset="https://content.globalforestwatch.org/wp-content/uploads/2024/02/Cocoa-Deforestation-Risk-Assessment-Priority-2023.png 1430w, https://content.globalforestwatch.org/wp-content/uploads/2024/02/Cocoa-Deforestation-Risk-Assessment-Priority-2023-1024x622.png 1024w, https://content.globalforestwatch.org/wp-content/uploads/2024/02/Cocoa-Deforestation-Risk-Assessment-Priority-2023-768x466.png 768w" sizes="(max-width: 1430px) 100vw, 1430px" /></figure>



<p>While creating the Cocoa DRA, we found that proximity to recent loss was an important predictor of future deforestation risk. The influence of this variable is most evident in several high-interest protected areas where, presumably due to effective legal protection, forest loss has been prevented in their boundaries. Because these forests have experienced less loss compared to the surrounding landscape, this resulted in a low priority ranking for the areas.</p>



<p>A good example of this is Taï National Park in western Côte d’Ivoire, which contains the only remaining intact forest within the study area. This is notable because the park is located within a region of Côte d’Ivoire where forests have been significantly impacted by cocoa production. A similar example is Bia National Park in western Ghana, which contains primary forest that has experienced low rates of forest loss compared to the surrounding landscape.</p>



<figure class="wp-block-image size-large"><img decoding="async" loading="lazy" width="2048" height="1448" src="https://content.globalforestwatch.org/wp-content/uploads/2024/02/tai-and-bia_labels-1-2048x1448.png" alt="Cocoa deforestation risk assessment priority ranking map (2023) showing protected areas" class="wp-image-16750" srcset="https://content.globalforestwatch.org/wp-content/uploads/2024/02/tai-and-bia_labels-1-2048x1448.png 2048w, https://content.globalforestwatch.org/wp-content/uploads/2024/02/tai-and-bia_labels-1-1024x724.png 1024w, https://content.globalforestwatch.org/wp-content/uploads/2024/02/tai-and-bia_labels-1-768x543.png 768w, https://content.globalforestwatch.org/wp-content/uploads/2024/02/tai-and-bia_labels-1-1536x1086.png 1536w" sizes="(max-width: 2048px) 100vw, 2048px" /></figure>



<p>That being said, additional data on legal status and conservation value provides important contextual information and we recommend users consider this data when using the Cocoa DRA for prioritizing engagement activities and supply chain interventions. &nbsp;</p>



<h2 class="wp-block-heading">Driving collective action: How companies and others can use the WAC and Cocoa DRA</h2>



<p>Used together, the WAC and the Cocoa DRA can inform decision making and coordinated action to reduce cocoa-driven deforestation. These resources also have utility beyond the scope of the CFI. For example, the public WAC heat map of cocoa plots may be used alongside near-real-time alerts to identify deforestation in cocoa production landscapes, or as a resource to collectively assess the industry’s progress towards ending deforestation. Use of the WAC as training data to develop public cocoa extent maps will also shed light on production across West Africa. Such accessible and holistic extent data is essential for stakeholders striving to tackle cocoa-driven deforestation.</p>



<p>For companies and farmers seeking to comply with the EUDR, the WAC and Cocoa DRA can also be valuable tools. The risk map can enable companies and other stakeholders to identify high-priority landscapes and prevent deforestation before it happens. To achieve this, engagement with farmers in high priority areas will be needed, enabling them to protect remaining forests while safeguarding their livelihoods. The risk map and WAC can also help companies determine where to prioritize further cocoa plot data collection efforts. High priority areas where the WAC shows a lack of mapped cocoa plots is a good indication of where such efforts are needed for enabling compliance with the EUDR’s geolocation requirement. This data will also be key for enabling inclusion of smallholder farmers in the EU market.</p>



<p>As public resources that provide stakeholders with a shared understanding of where it’s most urgent to address deforestation while respecting farmer privacy, the WAC and Cocoa DRA can help realize a more sustainable cocoa sector. Through this collective knowledge, stakeholders can more easily form collaborative approaches to eliminate cocoa-driven deforestation. Furthermore, through ongoing updates based on best available data, both resources can remain a public good and contribute to transparent monitoring of the cocoa sector.</p>



<small><em>* Development of these products was also informed by preliminary research carried out by Climate Focus on other methods of assessing deforestation risk associated cocoa.</small></em><br><br>



<style type="text/css"></style>

Ending Deforestation from Cocoa in West Africa with New Data-Driven Resources

<p>The global demand for chocolate has had a significant impact on West African forests. Approximately 70% of the global supply of cocoa originates from Ghana and Côte d’Ivoire, where production sites have replaced millions of hectares of forest. Between 2001 and 2015 alone, <a href="https://research.wri.org/gfr/forest-extent-indicators/deforestation-agriculture#how-much-forest-has-been-replaced-by-cocoa" target="_blank" rel="noreferrer noopener">cocoa was linked to</a> the loss of a third of Ghana’s forest area and a quarter of Côte d’Ivoire’s.</p>



<p>The <a href="https://www.worldcocoafoundation.org/initiative/cocoa-forests-initiative/" target="_blank" rel="noreferrer noopener">Cocoa &amp; Forests Initiative (CFI)</a>, facilitated by World Cocoa Foundation (WCF) and the <a href="https://www.idhsustainabletrade.com/initiative/cocoa-and-forests/" target="_blank" rel="noreferrer noopener">Sustainable Trade Initiative (IDH)</a>, was established in 2017 in an effort to end cocoa-driven deforestation and restore forest areas. Companies that joined the CFI agreed to map their supply chains, align methodologies on deforestation risk assessment and contribute to transparent monitoring and reporting. To support these efforts, World Resources Institute (WRI) partnered with WCF and 19 major cocoa and chocolate companies to create two new resources:*</p>



<ul>
<li>The <strong>West Africa Cocoa dataset (WAC)</strong>, a database of mapped cocoa plot boundaries in the direct supply chains of participating companies that provides an extensive view of mapped cocoa production in Côte d’Ivoire and Ghana. The public version of the dataset is <a href="https://gfw.global/4bEEES2">available on Global Forest Watch (GFW)</a> as a cocoa plot heat map that further anonymizes data to protect the privacy of farmers and companies.</li>



<li>The <strong>Cocoa Deforestation Risk Assessment (Cocoa DRA)</strong>, a map that provides a standardized view of cocoa-driven deforestation risk across the cocoa growing region of these two countries. The <a href="https://data.globalforestwatch.org/documents/gfw::cocoa-deforestation-risk-assessment/about">Cocoa DRA map layer</a> is available for download on GFW.</li>
</ul>



<p>These resources are the result of an unprecedented collaboration among companies to share highly sensitive cocoa plot data, and an extensive ethics and legal review to ensure the protection of farmers’ data throughout the process. For the first time, companies and other stakeholders in the cocoa sector have a shared view of priority areas in Ghana and Côte d’Ivoire for coordinated efforts toward a deforestation-free cocoa sector.</p>



<style>
.css-19vc5o2 aside {
   width: 100% !important;
float: none !important;
margin-left:
0px
!important;
}
</style>
<aside id="textbox"><em><b>From Palm Oil to Cocoa</b><p></p>
The development of the WAC and the Cocoa DRA builds on the success of similar products developed to support the palm oil sector: the <a href="https://www.globalforestwatch.org/blog/commodities/improved-universal-mill-list-expands-partnerships-improves-transparency-in-palm-oil-industry/" target="_blank" rel="noopener noreferrer">Universal Mill List (UML)</a>, a harmonized database of palm oil mill locations, and the <a href="https://www.wri.org/research/palm-risk-assessment-methodology-prioritizing-areas-landscapes-and-mills" target="_blank" rel="noopener noreferrer">Palm Risk Tool</a>, a method of identifying deforestation risks in their vicinity. However, differing supply chain dynamics between palm oil and cocoa were considered, such as:<li>
The locations of palm oil mills in the UML are represented as points while the cocoa plots in the WAC are represented by polygons. Further, the UML maps industrial scale infrastructure, whereas the WAC maps smallholder production. This raised additional privacy concerns, and steps were taken to avoid violating farmer privacy and safeguarding farmer livelihoods.<li>
The Palm Risk Tool analyzes deforestation risk by creating a buffer around each mill point to estimate a potential sourcing area. Oil palm fruit must be delivered to a mill quickly or the quality of the oil will decrease. This places a geographic constraint on how far oil palm fruit can travel to a mill, and stakeholders in the sector rely on a standardized sourcing radius of 50-km for their risk assessments. Cocoa does not have a geographic constraint nor an agreed upon estimation of a sourcing area, meaning that cocoa collected at a central point of purchase could have been grown anywhere in these two countries.<li>
The Palm Risk Tool was developed to assess deforestation risks in the vicinity of a specific type of infrastructure: palm oil mills. The Cocoa DRA was developed to be more flexible, enabling analysis of multiple areas of interest such as the plots in the WAC, cooperatives or other types of collection points and/or warehouses.<ul>

</em> </aside>



<figure class="wp-block-image size-full"><img src="https://content.globalforestwatch.org/wp-content/uploads/2024/02/Cocoa-pod.jpg" alt="Cocoa pod" class="wp-image-16723"/><figcaption class="wp-element-caption">Cocoa pod. Credit: Caroline Winchester, WRI.</figcaption></figure>



<h2 class="wp-block-heading"><strong>What is the West Africa</strong> <strong>Cocoa dataset?</strong></h2>



<p>The West Africa Cocoa dataset (WAC) aggregates data from 2021 on cocoa plot locations (a cocoa plot is a subset of a cocoa farm, and a single farm may be represented by multiple plots) in the direct supply chains of 19 cocoa and chocolate companies operating in Ghana and Côte d’Ivoire. Project participants developed and agreed on a data collation protocol and data sharing agreement that outlined submission requirements and the procedure for cleaning and aggregating individual company data sets. To be included in the WAC, each plot had to be current, actively growing cocoa and occupying a single parcel of land. WRI then reviewed, cleaned and aggregated the data.</p>



<p>While the cocoa plot polygon data is considered sensitive and will not be available publicly (see more below), we visualized the data as a heat map, <a href="https://gfw.global/4bEEES2">available on GFW</a>, which avoids disclosing the precise boundaries of the plots in the WAC while giving users access to information on the approximate location of cocoa plots. Additionally, we distributed the plot data to impartial research institutions for use as training data to create derivative products; specifically, public cocoa production extent maps.</p>



<p>These strategies make the data publicly available to support companies and other users in more robust cocoa mapping and monitoring while safeguarding the privacy of individual farmers.</p>



<style>
.css-19vc5o2 aside {
   width: 100% !important;
float: none !important;
margin-left:
0px
!important;
}
</style>
<aside id="textbox"><em><b>Protecting the Privacy of Cocoa Farmers and Companies</b><p></p>

The raw data used to create the WAC included sensitive supply chain information from both companies and farmers. Cocoa is primarily grown by smallholder farmers who live near the cocoa plots they manage. A public release of this raw data would show the boundaries of specific cocoa plots and could allow individual farmers and their homes to be identified. Further, since there are over 2 million farmers across West Africa, it is difficult to know whether they are all fully aware of how data pertaining to them is being used. <br><br>
To protect farmer and company privacy, the WAC data collection protocol and data sharing agreement were informed by an ethics review to ensure that potentially sensitive information was respected and handled properly, and by a legal review to ensure the data privacy laws of Ghana and Côte d’Ivoire were followed. Based off these reviews, and in order to protect farmer privacy, no personal data was included in the WAC or shared with WRI, and a secure method of transferring data among all parties was established. With farmer privacy in mind, the parties involved also decided that precise locations of cocoa plot boundaries would not be made publicly available.

</em> </aside>



<h2 class="wp-block-heading">What is the Cocoa Deforestation Risk Assessment?</h2>



<p>The Cocoa Deforestation Risk Assessment (Cocoa DRA) is a map that assesses the risk of future deforestation due to new or expanding cocoa production. The Cocoa DRA aligns various methods for assessing risk to provide a standardized view that companies and other stakeholders can use to make decisions around reducing deforestation. It can be used to proactively and efficiently identify deforestation risk across a supply chain that includes tens of thousands of geographically dispersed producers. Once risk has been identified, Cocoa DRA users can implement interventions for promoting sustainable production, protecting remaining forest cover and improving farmer livelihoods.</p>



<p>To create the map, we analyzed spatial variables that influence cocoa-linked deforestation including recent forest loss, proximity to cocoa, accessibility and suitability for cocoa production. The present-day extent of natural forest was approximated using the University of Maryland’s <a href="https://www.science.org/doi/10.1126/science.1244693" target="_blank" rel="noreferrer noopener">tree canopy cover</a> data set and removing cocoa plots mapped in the WAC, <a href="https://essd.copernicus.org/articles/13/1211/2021/" target="_blank" rel="noreferrer noopener">other data sets</a> that <a href="https://www.wri.org/research/spatial-database-planted-trees-sdpt-version-10" target="_blank" rel="noreferrer noopener">map planted trees</a> and historic tree cover loss.</p>



<p>Primary and intact forests in West Africa are limited, so to ensure the Cocoa DRA comprehensively assesses where future cocoa-linked deforestation is likely to occur in <em>all</em> natural forests, its scope includes natural forests both inside and outside of protected areas as well as naturally regenerating forests.</p>



<p>The Cocoa DRA translates risk into five priority rankings at a 1-km resolution (high, mid-high, medium, mid-low and low priority) indicating the likelihood of future deforestation associated with cocoa production. Cocoa DRA users can assess risk at different scales, including groups of farms, supply sheds around cocoa bean collection points and other types of landscapes.</p>



<p>Users can download the <a href="https://data.globalforestwatch.org/documents/gfw::cocoa-deforestation-risk-assessment/about">Cocoa DRA map layer</a> on GFW.</p>



<h2 class="wp-block-heading">What have we learned from the WAC and Cocoa DRA?</h2>



<p>Across Ghana and Côte d’Ivoire, the WAC maps 840,000 unique cocoa plots in 2021. Though cocoa plots tend to be small (most plots were between 1 to 2 hectares), the plots included in the WAC cover approximately 1.5 million hectares of planted cocoa area.</p>



<p>Cocoa plot data is most comprehensive in western Côte d’Ivoire as well as near the shared border between Côte d’Ivoire and Ghana. The forests in these areas have been <a href="https://research.wri.org/gfr/forest-extent-indicators/deforestation-agriculture#how-much-forest-has-been-replaced-by-cocoa" target="_blank" rel="noreferrer noopener">particularly impacted by cocoa production</a>. We found that the Cocoa DRA showed that these same regions had a significant amount of high priority areas (areas with higher likelihood of future cocoa-related deforestation). This is most evident in Côte d’Ivoire near the border of Liberia, while in western Ghana the landscape is more mixed.</p>



<figure class="wp-block-image size-full is-resized"><img src="https://content.globalforestwatch.org/wp-content/uploads/2024/02/Cocoa-Deforestation-Risk-Assessment-Priority-2023.png" alt="Cocoa deforestation risk assessment priority map for 2023" class="wp-image-16731" style="aspect-ratio:1.6474654377880185;width:840px;height:auto"/></figure>



<p>While creating the Cocoa DRA, we found that proximity to recent loss was an important predictor of future deforestation risk. The influence of this variable is most evident in several high-interest protected areas where, presumably due to effective legal protection, forest loss has been prevented in their boundaries. Because these forests have experienced less loss compared to the surrounding landscape, this resulted in a low priority ranking for the areas.</p>



<p>A good example of this is Taï National Park in western Côte d’Ivoire, which contains the only remaining intact forest within the study area. This is notable because the park is located within a region of Côte d’Ivoire where forests have been significantly impacted by cocoa production. A similar example is Bia National Park in western Ghana, which contains primary forest that has experienced low rates of forest loss compared to the surrounding landscape.</p>



<figure class="wp-block-image size-large"><img src="https://content.globalforestwatch.org/wp-content/uploads/2024/02/tai-and-bia_labels-1-2048x1448.png" alt="Cocoa deforestation risk assessment priority ranking map (2023) showing protected areas" class="wp-image-16750"/></figure>



<p>That being said, additional data on legal status and conservation value provides important contextual information and we recommend users consider this data when using the Cocoa DRA for prioritizing engagement activities and supply chain interventions. &nbsp;</p>



<h2 class="wp-block-heading">Driving collective action: How companies and others can use the WAC and Cocoa DRA</h2>



<p>Used together, the WAC and the Cocoa DRA can inform decision making and coordinated action to reduce cocoa-driven deforestation. These resources also have utility beyond the scope of the CFI. For example, the public WAC heat map of cocoa plots may be used alongside near-real-time alerts to identify deforestation in cocoa production landscapes, or as a resource to collectively assess the industry’s progress towards ending deforestation. Use of the WAC as training data to develop public cocoa extent maps will also shed light on production across West Africa. Such accessible and holistic extent data is essential for stakeholders striving to tackle cocoa-driven deforestation.</p>



<p>For companies and farmers seeking to comply with the EUDR, the WAC and Cocoa DRA can also be valuable tools. The risk map can enable companies and other stakeholders to identify high-priority landscapes and prevent deforestation before it happens. To achieve this, engagement with farmers in high priority areas will be needed, enabling them to protect remaining forests while safeguarding their livelihoods. The risk map and WAC can also help companies determine where to prioritize further cocoa plot data collection efforts. High priority areas where the WAC shows a lack of mapped cocoa plots is a good indication of where such efforts are needed for enabling compliance with the EUDR’s geolocation requirement. This data will also be key for enabling inclusion of smallholder farmers in the EU market.</p>



<p>As public resources that provide stakeholders with a shared understanding of where it’s most urgent to address deforestation while respecting farmer privacy, the WAC and Cocoa DRA can help realize a more sustainable cocoa sector. Through this collective knowledge, stakeholders can more easily form collaborative approaches to eliminate cocoa-driven deforestation. Furthermore, through ongoing updates based on best available data, both resources can remain a public good and contribute to transparent monitoring of the cocoa sector.</p>



<small><em>* Development of these products was also informed by preliminary research carried out by Climate Focus on other methods of assessing deforestation risk associated cocoa.</small></em><br><br>



<style type="text/css"></style>

Two new data-driven resources provide a shared view of priority areas in West Africa and can help realize a a deforestation-free cocoa sector.

Ending Deforestation from Cocoa in West Africa | GFW Blog

Sarah Carter

Jan Hooijmans

<p>Global Forest Watch (GFW) and the Global Forest Resources Assessment (FRA) from the Food and Agricultural Organization of the United Nations (FAO) both provide global data sets with information on the world’s forests. The latest FRA was released in 2020, and when compared to 2020 data on forest area from the University of Maryland (UMD), available on GFW, we see similarities and differences. So how similar are the numbers and what could explain the differences? This blog compares the forest extent from these two data sets, discusses why there are differences and explains what this means for users. To learn more about general differences between data on GFW and FRA data, <a href="https://www.globalforestwatch.org/blog/data-and-research/global-forest-watch-and-the-forest-resources-assessment-explained-in-5-graphics-2/">check out this blog</a>.</p>



<h2 class="wp-block-heading">What was the area of trees and forests in 2020 according to UMD and FRA data?</h2>



<style>
.css-19vc5o2 aside {
   width: 100% !important;
float: none !important;
margin-left:
0px
!important;
}
</style>
<aside id="textbox"><em><b>Tree Cover and Forest Area Definitions on GFW and in the FRA At a Glance </b><p></p>
Some of the primary distinctions between GFW and FRA are in their data sources and definitions (<a href="https://www.globalforestwatch.org/blog/data-and-research/global-forest-watch-and-the-forest-resources-assessment-explained-in-5-graphics-2/" target="_blank" rel="noopener noreferrer">read more here</a>).<br><br>
For 2020, GFW uses the UMD <a href="https://gfw.global/2Mei4Z0" target="_blank" rel="noopener noreferrer">2020 tree cover height data</a> to report on the extent of tree cover — a purely biophysical definition related to height and extent of trees. This satellite data picks up all trees with a minimum canopy height, which users can set when visualizing the data. For this analysis, we use a minimum height threshold of 5 meters. GFW also has a tree cover extent data set for 2000 and 2010 from UMD based on the percent of canopy density, which is used in many functions on GFW, but this was largely not used in this blog. <br><br>
The FRA relies on country-reported data based on a definition of forest that includes both a biophysical and land use component. In the FRA data, forests have a tree canopy height of at least 5 meters and canopy cover of at least 10% or are capable of reaching this threshold. Additionally, the land use criterion requires that land be officially or legally designated for a “forest use,” such as conservation or timber production.<br><br>
This means that in the FRA data, land that is temporarily devoid of trees — for example, recently logged or burned areas — would still be classified as forest if the official land use stipulates that trees will be allowed to regenerate in the future. Conversely, logged and burned areas are not captured as tree cover by the UMD data. Land that contains trees but is designated for agricultural or urban uses — for example, an oil palm plantation, shade grown cocoa (agroforestry) or a city park — would not be considered a forest by the FRA, but would be included in UMD tree cover data. 

</em> </aside>



<h4 class="wp-block-heading">Extent of trees and forests globally and by region in 2020 (in million hectares — Mha)</h4>



<table>
	<tbody>
		<tr>
			<td style="width:180px;height:20px;">
			<p>&nbsp;</p>
			</td>
			<td style="width:80px;height:20px;">
			<p><span style="font-size:10px;"><strong>Africa</strong></span></p>
			</td>
			<td style="width:95px;height:20px;">
			<p><span style="font-size:10px;"><strong>Asia including Russian Federation</strong></span></p>
			</td>
			<td style="width:95px;height:20px;">
			<p><span style="font-size:10px;"><strong>Europe, excluding Russian Federation</strong></span></p>
			</td>
			<td style="width:95px;height:20px;">
			<p><span style="font-size:10px;"><strong>North and Central America</strong></span></p>
			</td>
			<td style="width:80px;height:20px;">
			<p><span style="font-size:10px;"><strong>South America</strong></span></p>
			</td>
			<td style="width:80px;height:20px;">
			<p><span style="font-size:10px;"><strong>Oceania</strong></span></p>
			</td>
			<td style="width:80px;height:20px;">
			<p><span style="font-size:10px;"><strong>Global </strong>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;</span></p>
			</td>
		</tr>
		<tr>
			<td style="width:92px;height:10px;">
			<p><span style="font-size:10px;"><strong>UMD tree cover extent 2020</strong></span></p>
			</td>
			<td style="width:69px;height:10px;">
			<p><span style="font-size:10px;">674 Mha</span></p>

			<p>&nbsp;</p>
			</td>
			<td style="width:65px;height:10px;">
			<p><span style="font-size:10px;">1,483 Mha</span></p>

			<p>&nbsp;</p>
			</td>
			<td style="width:83px;height:10px;">
			<p><span style="font-size:10px;">204 Mha</span></p>
			</td>
			<td style="width:76px;height:10px;">
			<p><span style="font-size:10px;">675 Mha</span></p>
			</td>
			<td style="width:68px;height:10px;">
			<p><span style="font-size:10px;">834 Mha</span></p>
			</td>
			<td style="width:71px;height:10px;">
			<p><span style="font-size:10px;">147 Mha</span></p>
			</td>
			<td style="width:112px;height:10px;">
			<p><span style="font-size:10px;">4,018 Mha</span></p>

			<p>&nbsp;</p>
			</td>
		</tr>
		<tr>
			<td style="width:92px;height:10px;">
			<p><span style="font-size:10px;"><strong>FRA forest area 2020</strong></span></p>
			</td>
			<td style="width:69px;height:20px;">
			<p><span style="font-size:10px;">637 Mha</span></p>
			</td>
			<td style="width:65px;height:20px;">
			<p><span style="font-size:10px;">1,438 Mha</span></p>
			</td>
			<td style="width:83px;height:20px;">
			<p><span style="font-size:10px;">202 Mha</span></p>
			</td>
			<td style="width:76px;height:20px;">
			<p><span style="font-size:10px;">753 Mha</span></p>
			</td>
			<td style="width:68px;height:20px;">
			<p><span style="font-size:10px;">844 Mha</span></p>
			</td>
			<td style="width:71px;height:20px;">
			<p><span style="font-size:10px;">184 Mha</span></p>
			</td>
			<td style="width:112px;height:20px;">
			<p><span style="font-size:10px;">4,059 Mha</span></p>

			<p>&nbsp;</p>
			</td>
		</tr>
		<tr>
			<td style="width:92px;height:10px;">
			<p><span style="font-size:10px;"><strong>Difference of UMD extent as compared to FRA</strong></span></p>
			</td>
			<td style="width:69px;height:20px;">
			<p><span style="font-size:10px;"><strong>+5.6%</strong></span></p>

			<p>&nbsp;</p>
			</td>
			<td style="width:65px;height:20px;">
			<p><span style="font-size:10px;"><strong>+3.0%</strong></span></p>

			<p>&nbsp;</p>
			</td>
			<td style="width:83px;height:20px;">
			<p><span style="font-size:10px;"><strong>+1.0%</strong></span></p>

			<p>&nbsp;</p>
			</td>
			<td style="width:76px;height:20px;">
			<p><span style="font-size:10px;"><strong>-11.5%</strong></span></p>

			<p>&nbsp;</p>
			</td>
			<td style="width:68px;height:20px;">
			<p><span style="font-size:10px;"><strong>-1.2%</strong></span></p>

			<p>&nbsp;</p>
			</td>
			<td style="width:71px;height:20px;">
			<p><span style="font-size:10px;"><strong>-24.8%</strong></span></p>

			<p>&nbsp;</p>
			</td>
			<td style="width:112px;height:20px;">
			<p><span style="font-size:10px;"><strong>-1.02%</strong></span></p>

			<p>&nbsp;</p>
			</td>
		</tr>
	</tbody>
</table>



<div class="imagecenter"><img class=" wp-image-3569 alignright" style="padding-bottom: 10px;" width="100%"><small><em>*The regional figures compare the same list of countries, with some adjustments made to accommodate different country boundaries in the Database of Global Administrative Areas (GADM) and FRA. </em></small></div>



<p>Globally and regionally, tree/forest cover as estimated by UMD data and the FRA are similar. Global estimates of forest area in 2020 only vary by about 1% (4,059 million hectares — Mha — for the FRA and 4,018 Mha for the UMD data).</p>



<p>At the regional level there are many similarities, with very small differences for Asia, Europe and South America. Oceania shows the largest difference at the regional scale, which is almost entirely attributable to the FRA’s greater estimate of forest area in Australia — Australia’s FRA forest area is 52% greater than UMD’s tree cover extent. This may be because Australia’s definition of “forest” specifies a minimum vegetation height of 2 meters, which is used in the FRA estimate. The majority of countries in the FRA follow the FAO definition of a minimum vegetation height of 5 meters, which is also the height threshold used in this blog for analysis of the UMD data.</p>



<h4 class="wp-block-heading">Forest area/tree cover 2020 according to UMD data on GFW and the FRA</h4>



<div class="flourish-embed flourish-map" data-src="visualisation/15325037?240776"><script src="https://public.flourish.studio/resources/embed.js"></script></div>



<p>In central Africa and Southeast Asia, UMD’s tree cover estimates are typically greater than the FRA’s forest area estimations. In Malaysia, for example, UMD’s estimate is 33% higher than the FRA’s (28.6 Mha for UMD vs. 19.1 Mha for FRA). This may be explained by the difference in definitions, for example Southeast Asia typically contains large areas of oil palm plantations, which would be considered tree cover according to UMD but not forest according to the FRA.</p>



<p>For a better comparison, it would be useful subtract the area of tree plantations recorded in the <a href="https://gfw.global/2Qbv7fM">Spatial Database of Planted Trees</a>, which incorporates a number of different tree crops and planted trees, from the UMD tree cover estimate to better match the UMD data with the FAO forest definition. In the instance of Malaysia for example, excluding oil palm plantations from UMD tree cover results in an area considerably closer to the forest area reported in the FRA.</p>



<p>For some countries, the UMD tree cover extent matches the FRA forest extent very closely. For example, FRA’s forest area estimate for Chile is only 0.1% less than UMD’s. Another example is Nigeria, where the FRA forest area is just 5% less than UMD’s tree cover (21.6 Mha for FRA and 22.7 Mha for UMD).</p>



<p>The forest area according to the FRA and UMD data have not always matched so well though — the FRA estimate for forest extent in 2000 (reported in the FRA 2015) was 46% less than UMD’s tree cover area (for the year 2000). Nigeria significantly changed their forest area figures for 2000 as reported in the 2020 FRA, based on the availability of more and better data, resulting in the FRA and UMD figures becoming much more similar (3% difference).</p>



<style>
.css-19vc5o2 aside {
   width: 100% !important;
float: none !important;
margin-left:
0px
!important;
}
#textbox table {
    width: 100% !important;
    margin-left: auto;
}
</style>
<aside id="textbox"><em><b>Improvements in Monitoring Mean More Similar Numbers</b><p></p>
Due to the improvements in data quality from both the FRA and UMD, the data is becoming more and more similar. 
A <a href="https://www.sciencedirect.com/science/article/pii/S0048969722048872" target="_blank" rel="noopener noreferrer">scientific study</a> found that in many countries, the quality of data sources used for forest area estimates as reported to the FRA have increased over time. Specifically, the percentage of forest area monitored by satellites with a quality indication of “good” or “very good” in the FRA has increased. To incorporate these data advances, older data points are recalculated if methodologies/data have been revised — for example, the forest extent in 2000 presented in the FRA 2020 is not the same as was presented in the FRA 2000.<br><br>
At the same time, the data on GFW has also improved. The original UMD tree cover data on GFW, which is still used as a baseline for the annual tree cover loss data, is based on canopy density (<a href="https://www.science.org/doi/10.1126/science.1244693" target="_blank" rel="noopener noreferrer">Hansen et al. 2013</a>) for 2000 and 2010 (thresholded in this case to 30% canopy cover). New data, also from UMD, is based on tree height (<a href="https://www.sciencedirect.com/science/article/abs/pii/S0034425720305381?via%3Dihub" target="_blank" rel="noopener noreferrer">Potapov et al. 2021</a>) for 2000 and 2020. The new data incorporates updated methods, and new data such as GEDI which better captures the structure of trees.<br><br>
Thanks to these improvements, the difference estimates of the FRA forest area and UMD height-based tree cover extent has shrunk to less than 1% (when comparing UMD’s height-based data with the most recent FRA estimate).

<br><br><Strong>GLOBAL FOREST AREA/TREE COVER EXTENT FOR 2000 AND 2020 IN THE FRA AND UMD DATA*</strong>

<table border="1" cellpadding="0" cellspacing="0">
	<tbody>
		<tr>
			<td style="width:157px;height:20px;">
			<p><strong>FRA 2000</strong></p>
			</td>
			<td style="width:157px;height:20px;">
			<p><strong>FRA 2020</strong></p>
			</td>
			<td style="width:157px;height:20px;">
			<p><strong>UMD canopy density data (2000)</strong></p>
			</td>
			<td style="width:157px;height:20px;">
			<p><strong>UMD tree height data (2020)</strong></p>
			</td>
		</tr>
		<tr>
			<td style="width:157px;height:20px;">
			<p>3,869 Mha</p>
			</td>
			<td style="width:157px;height:20px;">
			<p>4,158 Mha</p>
			</td>
			<td style="width:157px;height:20px;">
			<p>3,993 Mha</p>
			</td>
			<td style="width:157px;height:20px;">
			<p>4,122 Mha</p>
			</td>
		</tr>
	</tbody>
</table>
*These figures are the globally reported figures

</em> </aside>



<p>Some countries showed higher estimates for FRA forest areas than UMD tree cover extent, for instance, many of the Nordic European countries and Canada. Again, the difference between the definitions of tree cover used in UMD and forest area used by FRA likely explains the differences. According to the FRA 2020, a substantial proportion of forested areas in these countries are production forests, meaning that vast areas of land temporarily unstocked due to timber harvesting are included as forest area for the FRA. UMD does not include these regions, nor areas that have been regenerated but have not yet reached the UMD height threshold, as tree cover.</p>



<h2 class="wp-block-heading">What does this mean for those using UMD and FRA data?</h2>



<p>Estimates by UMD and the FRA of tree cover and forest can vary, in some cases substantially within a country. The same patterns are not seen for all countries; some countries have higher estimates according to UMD, and for others, FRA estimates are higher. In many cases, these can be explained when taking into account the different forest definitions used. </p>



<p>Because of their differences, both data sets can be useful for different stakeholders. Read our blog <a href="https://www.globalforestwatch.org/blog/data-and-research/global-forest-watch-and-the-forest-resources-assessment-explained-in-5-graphics-2/"><em>Global Forest Watch and the Forest Resources Assessment, Explained in 5 Graphics</em></a><em> </em>to learn more about different ways the UMD and FRA data can be used. In the end, it’s important for users to know the differences between the data sets to help them decide which data will best meet their needs.</p>



<hr> 

<div class="attribute" style="”padding-bottom:" 10px;”=""> We acknowledge and thank the FAO for their helpful comments on this blog.

 

</div> 

 

<hr>



<style type="text/css"></style>



<p></p>



<p></p>

Comparing Forest Extent in 2020 from Global Forest Watch and the Forest Resources Assessment

<p>Global Forest Watch (GFW) and the Global Forest Resources Assessment (FRA) from the Food and Agricultural Organization of the United Nations (FAO) both provide global data sets with information on the world's forests. The latest FRA was released in 2020, and when compared to 2020 data on forest area from the University of Maryland (UMD), available on GFW, we see similarities and differences. So how similar are the numbers and what could explain the differences? This blog compares the forest extent from these two data sets, discusses why there are differences and explains what this means for users. To learn more about general differences between data on GFW and FRA data, <a href="https://www.globalforestwatch.org/blog/data-and-research/global-forest-watch-and-the-forest-resources-assessment-explained-in-5-graphics-2/">check out this blog</a>.</p>



<h2 class="wp-block-heading">What was the area of trees and forests in 2020 according to UMD and FRA data?</h2>



<style>
.css-19vc5o2 aside {
   width: 100% !important;
float: none !important;
margin-left:
0px
!important;
}
</style>
<aside id="textbox"><em><b>Tree Cover and Forest Area Definitions on GFW and in the FRA At a Glance </b><p></p>
Some of the primary distinctions between GFW and FRA are in their data sources and definitions (<a href="https://www.globalforestwatch.org/blog/data-and-research/global-forest-watch-and-the-forest-resources-assessment-explained-in-5-graphics-2/" target="_blank" rel="noopener noreferrer">read more here</a>).<br><br>
For 2020, GFW uses the UMD <a href="https://gfw.global/2Mei4Z0" target="_blank" rel="noopener noreferrer">2020 tree cover height data</a> to report on the extent of tree cover — a purely biophysical definition related to height and extent of trees. This satellite data picks up all trees with a minimum canopy height, which users can set when visualizing the data. For this analysis, we use a minimum height threshold of 5 meters. GFW also has a tree cover extent data set for 2000 and 2010 from UMD based on the percent of canopy density, which is used in many functions on GFW, but this was largely not used in this blog. <br><br>
The FRA relies on country-reported data based on a definition of forest that includes both a biophysical and land use component. In the FRA data, forests have a tree canopy height of at least 5 meters and canopy cover of at least 10% or are capable of reaching this threshold. Additionally, the land use criterion requires that land be officially or legally designated for a “forest use,” such as conservation or timber production.<br><br>
This means that in the FRA data, land that is temporarily devoid of trees — for example, recently logged or burned areas — would still be classified as forest if the official land use stipulates that trees will be allowed to regenerate in the future. Conversely, logged and burned areas are not captured as tree cover by the UMD data. Land that contains trees but is designated for agricultural or urban uses — for example, an oil palm plantation, shade grown cocoa (agroforestry) or a city park — would not be considered a forest by the FRA, but would be included in UMD tree cover data. 

</em> </aside>



<h4 class="wp-block-heading">Extent of trees and forests globally and by region in 2020 (in million hectares — Mha)</h4>



<table>
	<tbody>
		<tr>
			<td style="width:180px;height:20px;">
			<p>&nbsp;</p>
			</td>
			<td style="width:80px;height:20px;">
			<p><span style="font-size:10px;"><strong>Africa</strong></span></p>
			</td>
			<td style="width:95px;height:20px;">
			<p><span style="font-size:10px;"><strong>Asia including Russian Federation</strong></span></p>
			</td>
			<td style="width:95px;height:20px;">
			<p><span style="font-size:10px;"><strong>Europe, excluding Russian Federation</strong></span></p>
			</td>
			<td style="width:95px;height:20px;">
			<p><span style="font-size:10px;"><strong>North and Central America</strong></span></p>
			</td>
			<td style="width:80px;height:20px;">
			<p><span style="font-size:10px;"><strong>South America</strong></span></p>
			</td>
			<td style="width:80px;height:20px;">
			<p><span style="font-size:10px;"><strong>Oceania</strong></span></p>
			</td>
			<td style="width:80px;height:20px;">
			<p><span style="font-size:10px;"><strong>Global </strong>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;</span></p>
			</td>
		</tr>
		<tr>
			<td style="width:92px;height:10px;">
			<p><span style="font-size:10px;"><strong>UMD tree cover extent 2020</strong></span></p>
			</td>
			<td style="width:69px;height:10px;">
			<p><span style="font-size:10px;">674 Mha</span></p>

			<p>&nbsp;</p>
			</td>
			<td style="width:65px;height:10px;">
			<p><span style="font-size:10px;">1,483 Mha</span></p>

			<p>&nbsp;</p>
			</td>
			<td style="width:83px;height:10px;">
			<p><span style="font-size:10px;">204 Mha</span></p>
			</td>
			<td style="width:76px;height:10px;">
			<p><span style="font-size:10px;">675 Mha</span></p>
			</td>
			<td style="width:68px;height:10px;">
			<p><span style="font-size:10px;">834 Mha</span></p>
			</td>
			<td style="width:71px;height:10px;">
			<p><span style="font-size:10px;">147 Mha</span></p>
			</td>
			<td style="width:112px;height:10px;">
			<p><span style="font-size:10px;">4,018 Mha</span></p>

			<p>&nbsp;</p>
			</td>
		</tr>
		<tr>
			<td style="width:92px;height:10px;">
			<p><span style="font-size:10px;"><strong>FRA forest area 2020</strong></span></p>
			</td>
			<td style="width:69px;height:20px;">
			<p><span style="font-size:10px;">637 Mha</span></p>
			</td>
			<td style="width:65px;height:20px;">
			<p><span style="font-size:10px;">1,438 Mha</span></p>
			</td>
			<td style="width:83px;height:20px;">
			<p><span style="font-size:10px;">202 Mha</span></p>
			</td>
			<td style="width:76px;height:20px;">
			<p><span style="font-size:10px;">753 Mha</span></p>
			</td>
			<td style="width:68px;height:20px;">
			<p><span style="font-size:10px;">844 Mha</span></p>
			</td>
			<td style="width:71px;height:20px;">
			<p><span style="font-size:10px;">184 Mha</span></p>
			</td>
			<td style="width:112px;height:20px;">
			<p><span style="font-size:10px;">4,059 Mha</span></p>

			<p>&nbsp;</p>
			</td>
		</tr>
		<tr>
			<td style="width:92px;height:10px;">
			<p><span style="font-size:10px;"><strong>Difference of UMD extent as compared to FRA</strong></span></p>
			</td>
			<td style="width:69px;height:20px;">
			<p><span style="font-size:10px;"><strong>+5.6%</strong></span></p>

			<p>&nbsp;</p>
			</td>
			<td style="width:65px;height:20px;">
			<p><span style="font-size:10px;"><strong>+3.0%</strong></span></p>

			<p>&nbsp;</p>
			</td>
			<td style="width:83px;height:20px;">
			<p><span style="font-size:10px;"><strong>+1.0%</strong></span></p>

			<p>&nbsp;</p>
			</td>
			<td style="width:76px;height:20px;">
			<p><span style="font-size:10px;"><strong>-11.5%</strong></span></p>

			<p>&nbsp;</p>
			</td>
			<td style="width:68px;height:20px;">
			<p><span style="font-size:10px;"><strong>-1.2%</strong></span></p>

			<p>&nbsp;</p>
			</td>
			<td style="width:71px;height:20px;">
			<p><span style="font-size:10px;"><strong>-24.8%</strong></span></p>

			<p>&nbsp;</p>
			</td>
			<td style="width:112px;height:20px;">
			<p><span style="font-size:10px;"><strong>-1.02%</strong></span></p>

			<p>&nbsp;</p>
			</td>
		</tr>
	</tbody>
</table>



<div class="imagecenter"><img class=" wp-image-3569 alignright" style="padding-bottom: 10px;" width="100%"><small><em>*The regional figures compare the same list of countries, with some adjustments made to accommodate different country boundaries in the Database of Global Administrative Areas (GADM) and FRA. </em></small></div>



<p>Globally and regionally, tree/forest cover as estimated by UMD data and the FRA are similar. Global estimates of forest area in 2020 only vary by about 1% (4,059 million hectares — Mha — for the FRA and 4,018 Mha for the UMD data).</p>



<p>At the regional level there are many similarities, with very small differences for Asia, Europe and South America. Oceania shows the largest difference at the regional scale, which is almost entirely attributable to the FRA’s greater estimate of forest area in Australia — Australia’s FRA forest area is 52% greater than UMD’s tree cover extent. This may be because Australia’s definition of “forest” specifies a minimum vegetation height of 2 meters, which is used in the FRA estimate. The majority of countries in the FRA follow the FAO definition of a minimum vegetation height of 5 meters, which is also the height threshold used in this blog for analysis of the UMD data.</p>



<h4 class="wp-block-heading">Forest area/tree cover 2020 according to UMD data on GFW and the FRA</h4>



<div class="flourish-embed flourish-map" data-src="visualisation/15325037?240776"><script src="https://public.flourish.studio/resources/embed.js"></script></div>



<p>In central Africa and Southeast Asia, UMD’s tree cover estimates are typically greater than the FRA’s forest area estimations. In Malaysia, for example, UMD's estimate is 33% higher than the FRA's (28.6 Mha for UMD vs. 19.1 Mha for FRA). This may be explained by the difference in definitions, for example Southeast Asia typically contains large areas of oil palm plantations, which would be considered tree cover according to UMD but not forest according to the FRA.</p>



<p>For a better comparison, it would be useful subtract the area of tree plantations recorded in the <a href="https://gfw.global/2Qbv7fM">Spatial Database of Planted Trees</a>, which incorporates a number of different tree crops and planted trees, from the UMD tree cover estimate to better match the UMD data with the FAO forest definition. In the instance of Malaysia for example, excluding oil palm plantations from UMD tree cover results in an area considerably closer to the forest area reported in the FRA.</p>



<p>For some countries, the UMD tree cover extent matches the FRA forest extent very closely. For example, FRA’s forest area estimate for Chile is only 0.1% less than UMD’s. Another example is Nigeria, where the FRA forest area is just 5% less than UMD's tree cover (21.6 Mha for FRA and 22.7 Mha for UMD).</p>



<p>The forest area according to the FRA and UMD data have not always matched so well though — the FRA estimate for forest extent in 2000 (reported in the FRA 2015) was 46% less than UMD’s tree cover area (for the year 2000). Nigeria significantly changed their forest area figures for 2000 as reported in the 2020 FRA, based on the availability of more and better data, resulting in the FRA and UMD figures becoming much more similar (3% difference).</p>



<style>
.css-19vc5o2 aside {
   width: 100% !important;
float: none !important;
margin-left:
0px
!important;
}
#textbox table {
    width: 100% !important;
    margin-left: auto;
}
</style>
<aside id="textbox"><em><b>Improvements in Monitoring Mean More Similar Numbers</b><p></p>
Due to the improvements in data quality from both the FRA and UMD, the data is becoming more and more similar. 
A <a href="https://www.sciencedirect.com/science/article/pii/S0048969722048872" target="_blank" rel="noopener noreferrer">scientific study</a> found that in many countries, the quality of data sources used for forest area estimates as reported to the FRA have increased over time. Specifically, the percentage of forest area monitored by satellites with a quality indication of “good” or “very good” in the FRA has increased. To incorporate these data advances, older data points are recalculated if methodologies/data have been revised — for example, the forest extent in 2000 presented in the FRA 2020 is not the same as was presented in the FRA 2000.<br><br>
At the same time, the data on GFW has also improved. The original UMD tree cover data on GFW, which is still used as a baseline for the annual tree cover loss data, is based on canopy density (<a href="https://www.science.org/doi/10.1126/science.1244693" target="_blank" rel="noopener noreferrer">Hansen et al. 2013</a>) for 2000 and 2010 (thresholded in this case to 30% canopy cover). New data, also from UMD, is based on tree height (<a href="https://www.sciencedirect.com/science/article/abs/pii/S0034425720305381?via%3Dihub" target="_blank" rel="noopener noreferrer">Potapov et al. 2021</a>) for 2000 and 2020. The new data incorporates updated methods, and new data such as GEDI which better captures the structure of trees.<br><br>
Thanks to these improvements, the difference estimates of the FRA forest area and UMD height-based tree cover extent has shrunk to less than 1% (when comparing UMD’s height-based data with the most recent FRA estimate).

<br><br><Strong>GLOBAL FOREST AREA/TREE COVER EXTENT FOR 2000 AND 2020 IN THE FRA AND UMD DATA*</strong>

<table border="1" cellpadding="0" cellspacing="0">
	<tbody>
		<tr>
			<td style="width:157px;height:20px;">
			<p><strong>FRA 2000</strong></p>
			</td>
			<td style="width:157px;height:20px;">
			<p><strong>FRA 2020</strong></p>
			</td>
			<td style="width:157px;height:20px;">
			<p><strong>UMD canopy density data (2000)</strong></p>
			</td>
			<td style="width:157px;height:20px;">
			<p><strong>UMD tree height data (2020)</strong></p>
			</td>
		</tr>
		<tr>
			<td style="width:157px;height:20px;">
			<p>3,869 Mha</p>
			</td>
			<td style="width:157px;height:20px;">
			<p>4,158 Mha</p>
			</td>
			<td style="width:157px;height:20px;">
			<p>3,993 Mha</p>
			</td>
			<td style="width:157px;height:20px;">
			<p>4,122 Mha</p>
			</td>
		</tr>
	</tbody>
</table>
*These figures are the globally reported figures

</em> </aside>



<p>Some countries showed higher estimates for FRA forest areas than UMD tree cover extent, for instance, many of the Nordic European countries and Canada. Again, the difference between the definitions of tree cover used in UMD and forest area used by FRA likely explains the differences. According to the FRA 2020, a substantial proportion of forested areas in these countries are production forests, meaning that vast areas of land temporarily unstocked due to timber harvesting are included as forest area for the FRA. UMD does not include these regions, nor areas that have been regenerated but have not yet reached the UMD height threshold, as tree cover.</p>



<h2 class="wp-block-heading">What does this mean for those using UMD and FRA data?</h2>



<p>Estimates by UMD and the FRA of tree cover and forest can vary, in some cases substantially within a country. The same patterns are not seen for all countries; some countries have higher estimates according to UMD, and for others, FRA estimates are higher. In many cases, these can be explained when taking into account the different forest definitions used. </p>



<p>Because of their differences, both data sets can be useful for different stakeholders. Read our blog <a href="https://www.globalforestwatch.org/blog/data-and-research/global-forest-watch-and-the-forest-resources-assessment-explained-in-5-graphics-2/"><em>Global Forest Watch and the Forest Resources Assessment, Explained in 5 Graphics</em></a><em> </em>to learn more about different ways the UMD and FRA data can be used. In the end, it’s important for users to know the differences between the data sets to help them decide which data will best meet their needs.</p>



<hr> 

<div class="attribute" style="”padding-bottom:" 10px;”=""> We acknowledge and thank the FAO for their helpful comments on this blog.

 

</div> 

 

<hr>



<style type="text/css"></style>



<p></p>



<p></p>

This blog compares the forest extent in 2020 for data from UMD on GFW and the FAO Forest Resources Assessment and explains the differences.

Forest Extent from GFW and the Forest Resources Assessment

Side by side comparison of UMD tree cover loss and JRC Tropical Moist Forest data

forest quality

forest quality | Global Forest Watch Blog

Peter Potapov

Christelle Vancutsem

Clément Bourgoin

Fréderic Achard

<p>The wide availability of forest and land cover maps based on satellite data provides more information on the world’s forests than ever before. However, this plethora of information can also be challenging for people trying to understand differences in data sources and decide which data to use for a specific purpose.</p>



<p>In the past, we’ve compared the <a href="https://gfw.global/3BfwQWz">annual tree cover loss data</a> from the University of Maryland (UMD) and available on Global Forest Watch (GFW) with national data from <a href="https://www.globalforestwatch.org/blog/data-and-research/2022-tree-cover-loss-data-explained/">Brazil and Indonesia</a>, as well as with data from the United Nations Food and Agriculture Organization’s (FAO) <a href="https://www.globalforestwatch.org/blog/data-and-research/global-forest-watch-and-the-forest-resources-assessment-explained-in-5-graphics-2/">Global Forest Resources Assessment</a>, to shed light on some of these differences.</p>



<p>Another data set that tracks forest change in the tropics is the <a href="https://forobs.jrc.ec.europa.eu/TMF" target="_blank" rel="noreferrer noopener">Tropical Moist Forest (TMF) data set</a> from the European Commission Joint Research Centre (JRC). Like the UMD tree cover loss data, the JRC TMF data is based on Landsat satellite imagery and is updated annually. Yet while both data sets map forest disturbances, the maps and numbers look a little different.</p>



<p>In this blog, we explain some key differences in the definitions and methods used for each data set that impact the resulting maps and discuss how the disturbances mapped by both data sets compare (while the data sets also map forest gain or regrowth, this blog focuses on comparing their mapping of forest disturbances). While understanding these differences can help determine the contexts for which each data set might be useful, note that this analysis does not evaluate the quality or accuracy of either data set.</p>



<h2 class="wp-block-heading">What areas do the JRC and UMD data sets cover, and how do they define forests?</h2>



<p>The UMD data set maps forests globally based on <a href="https://www.science.org/doi/10.1126/science.1244693" target="_blank" rel="noreferrer noopener">tree cover extent</a> in 2000. <a href="https://www.science.org/doi/10.1126/science.1244693" target="_blank" rel="noreferrer noopener">Tree cover</a> is defined as woody vegetation with a minimum height of at least five meters at 30-meter resolution, which includes all types of tree cover — including natural or managed forest and plantations — and is not limited to a specific forest type. The GFW map shows tree cover with a 30% canopy density, which we use in our analyses and in this comparison, although users can adjust the threshold based on their needs.</p>



<p>The JRC data set, on the other hand, <a href="https://www.science.org/doi/10.1126/sciadv.abe1603" target="_blank" rel="noreferrer noopener">maps forests</a> beginning in 1990. The data includes all closed primary and secondary forests in the humid tropics and does not apply a minimum height or canopy density threshold to delineate forest extent.</p>



<p>As a result of these definitional differences, some areas included in the 2000 UMD tree cover extent in the tropics like dry forests, young secondary forests and plantations are not classified as forests in the JRC data set in 2000 — a total of 655 million hectares (Mha), or 34%, of the UMD tree cover extent area that overlaps with the JRC data. Likewise, the JRC forest extent in 2000 includes some areas that do not meet the height and 30% canopy density threshold in the UMD data or are otherwise not mapped as tree cover by the UMD data — 17.2 Mha, or 1% of the JRC forest extent.</p>



<p>Because of the differences in geographic coverage and scope between the two data sets, this blog only compares forest disturbances from 2001-2022 for areas that, in the year 2000, are mapped both as tree cover by UMD and as forest by JRC.</p>



<h2 class="wp-block-heading">What types of forest disturbances do the UMD and JRC data map?</h2>



<h3 class="wp-block-heading"><em>The JRC Tropical Moist Forest data set maps degradation and deforestation in the humid tropics</em></h3>



<p>The JRC TMF data set maps disturbances to tropical moist forests — including degradation and deforestation — from 1990 to 2022. The JRC team maps these changes using an algorithm that detects forest disruptions, meaning an absence of tree foliage cover within a 30-meter pixel for each available Landsat image, and then classifies the status of forest based on how long these disruptions are visible.</p>



<p>The TMF data set <a href="https://www.science.org/doi/10.1126/sciadv.abe1603" target="_blank" rel="noreferrer noopener">defines degradation</a> as temporary disturbances occurring for up to 2.5 years, while it defines deforestation as longer-term disturbances where disruptions to tree cover are visible for more than 2.5 years and no tree regrowth is detected over the last three years. In any given year, a forest is considered disturbed if there is a disruption detected in at least one Landsat image during that year.</p>



<p>Thus, pixels classified as degraded are still considered forest, with existing or regrowing trees, while pixels classified as deforested are undergoing a longer-term loss of forest. Because disruptions are recorded and analyzed for the entire time series, the data set classifies dynamics over time; for example, a forest pixel may be classified as degraded in one year after a short-term disturbance, then after a period of recovery may later be classified as deforested after ongoing disturbances with no tree regrowth. The JRC deforestation class therefore includes direct deforestation (deforestation of undisturbed forest) or deforestation after degradation (deforestation of forest that was previously classified as degraded).</p>



<p>The data set also analyzes the frequency, duration of disruptions and time of forest recovery in between disruption events to identify deforestation following degradation versus multiple degradation events.</p>



<h3 class="wp-block-heading"><em>The UMD tree cover loss data set on GFW maps all types of tree cover loss globally</em></h3>



<p>The UMD tree cover loss data set maps annual tree cover loss from 2001 to 2022. The UMD team uses an algorithm that detects loss using change detection metrics based on comparison of Landsat images between the present and past years. Tree cover loss, <a href="https://www.science.org/doi/10.1126/science.1244693" target="_blank" rel="noreferrer noopener">defined as</a> a clearing of tree cover of at least a half of a 30-meter Landsat pixel, includes both temporary and permanent losses without defining or differentiating which loss is considered degradation or deforestation. Once classified as loss, pixels are masked from further updates, meaning that any following forest change dynamics — such as regrowth or subsequent disturbances — are not recorded.</p>



<p>The UMD tree cover loss data can be combined with contextual data layers, such as <a href="https://gfw.global/3kqcoqa">UMD’s map of humid tropical primary forest</a>, to see loss within specific forest types.</p>



<h4 class="wp-block-heading">Key differences between the JRC TMF and UMD tree cover loss data</h4>



<table border="1" cellpadding="0" cellspacing="0">
	<tbody>
		<tr>
			<td style="width:165px;height:20px;">
			<p>&nbsp;</p>
			</td>
			<td style="width:320px;height:20px;">
			<p><strong>JRC TMF</strong></p>
			</td>
			<td style="width:320px;height:20px;">
			<p><strong>UMD tree cover loss</strong></p>
			</td>
		</tr>
		<tr>
			<td style="width:165px;height:20px;">
			<p>Temporal coverage</p>
			</td>
			<td style="width:320px;height:20px;">
			<p>1990-2022</p>
			</td>
			<td style="width:320px;height:20px;">
			<p>&nbsp;2000-2022</p>
			</td>
		</tr>
		<tr>
			<td style="width:165px;height:20px;">
			<p>Geographic coverage and extent</p>
			</td>
			<td style="width:320px;height:20px;">
			<p><strong>Tropical moist forest:</strong> closed primary and secondary forests in the humid tropics in 1990.</p>
			</td>
			<td style="width:320px;height:20px;">
			<p><strong>Global tree cover: </strong>woody vegetation with a minimum height of at least five meters at 30-meter resolution in 2000.</p>
			</td>
		</tr>
		<tr>
			<td style="width:165px;height:20px;">
			<p>Definition of disturbance</p>
			</td>
			<td style="width:320px;height:20px;">
			<p>An absence of tree foliage cover within a 30-meter pixel.</p>
			</td>
			<td style="width:320px;height:20px;">
			<p>Clearing of at least half of tree cover within a 30-meter pixel.</p>
			</td>
		</tr>
		<tr>
			<td style="width:165px;height:20px;">
			<p>Disturbance classes</p>
			</td>
			<td style="width:320px;height:20px;">
			<p>Disturbances are classified as <strong>degradation </strong>if they are visible for less than 2.5 years, and <strong>deforestation </strong>if they are visible for more than 2.5 years with no vegetative regrowth over the last three years.</p>
			</td>
			<td style="width:320px;height:20px;">
			<p>All disturbances classified as <strong>tree cover loss. </strong>The length of disturbance is not considered.</p>
			</td>
		</tr>
		<tr>
			<td style="width:165px;height:20px;">
			<p>Methods</p>
			</td>
			<td style="width:320px;height:20px;">
			<p>Classifies valid single-date Landsat images and derives disturbance classes by analyzing the single-date classifications. Dynamics of disturbance events over time are recorded.</p>
			</td>
			<td style="width:320px;height:20px;">
			<p>Classifies change using change detection metrics derived from all valid Landsat images over the present and past years. Only the first loss event that is detected is recorded.</p>
			</td>
		</tr>
	</tbody>
</table>



<h2 class="wp-block-heading">What do the UMD and JRC data show about forest disturbances in the tropics?</h2>



<p>Overall, the JRC data detects a larger area of disturbances across the tropics from 2001-2022 compared with the UMD data. The JRC data detects 258 Mha of disturbances, including 113 Mha of degradation and 145 Mha of deforestation, while the UMD data detects 143 Mha of tree cover loss. However, trends in total deforestation mapped by JRC and tree cover loss mapped by UMD are generally aligned: They are largely similar for the first decade; then from 2013-2015, JRC detects a larger area as deforestation than UMD detects as loss, and after 2016, UMD detects a larger area.</p>



<div class="flourish-embed" data-src="story/2112582?163031"><script src="https://public.flourish.studio/resources/embed.js"></script></div>



<p>We further separated out the disturbance types mapped by the JRC data and overlaid the data sets to evaluate how the two data sets compare at each pixel.</p>



<p>Among all the JRC data’s disturbance classes, the JRC direct deforestation class aligned most closely with UMD tree cover loss. Across the tropics from 2001-2022, 46.8 Mha, or 68%, of the area classified as direct deforestation by the JRC data was also classified as tree cover loss in the UMD data. This varied by region and was highest in Asia/Oceania and the Americas (66% and 75% respectively) and lowest in Africa (50%). The UMD data less often classified areas that JRC identified as deforestation after degradation as tree cover loss (38.1 Mha, or 50% across the tropics).</p>



<p>In some areas, disturbances classified as degradation by the JRC TMF data set were classified as tree cover loss by the UMD data. Across the tropics from 2001-2022, 32.6 Mha, or 29%, of the area the JRC data classified as degradation was also classified as tree cover loss in the UMD data. The JRC degradation class includes three types of degradation: short-duration (less than one year), long-duration (1-2.5 years) and short-duration over 2-3 periods (short-duration degradation that occurs two to three separate times throughout the time period). The JRC data also measures “intensity” of degradation based on the total number of Landsat images where disruptions in tree cover were visible. We found that, in general, the UMD data:</p>



<ul>
<li>More often detected longer duration degradation as tree cover loss. Short-duration degradation (including events occurring just once or over 2-3 periods), which represents 85% of the forest area mapped as degraded by JRC since 2001, is usually not shown as tree cover loss by the UMD data.</li>



<li>Classified degradation events as tree cover loss that had an intensity of greater than six images on average, while those UMD did not classify as loss had approximately three images or less on average.</li>



<li>Detected a larger proportion of the area JRC classified as degradation from 2015 onwards, which may be <a href="https://www.globalforestwatch.org/blog/data-and-research/tree-cover-loss-satellite-data-trend-analysis/">due to the algorithm changes</a> in the UMD data that increase forest loss detection sensitivity in case of fires, selective logging and other disturbance types that JRC classified as degradation.</li>
</ul>



<p>In other areas across the tropics, the UMD data detected tree cover loss for 6.4 Mha, or 7%, of areas where there the JRC data detected no disturbance events.</p>



<figure class="wp-block-image size-full is-resized"><img decoding="async" loading="lazy" width="1222" height="1163" src="https://content.globalforestwatch.org/wp-content/uploads/2024/01/Comparison-JRC-TMF-and-UMD-table.png" alt="Comparison of JRC TMF and UMD tree cover loss disturbances across the tropics" class="wp-image-16572" style="aspect-ratio:1.0507308684436802;width:771px;height:auto" srcset="https://content.globalforestwatch.org/wp-content/uploads/2024/01/Comparison-JRC-TMF-and-UMD-table.png 1222w, https://content.globalforestwatch.org/wp-content/uploads/2024/01/Comparison-JRC-TMF-and-UMD-table-1024x975.png 1024w, https://content.globalforestwatch.org/wp-content/uploads/2024/01/Comparison-JRC-TMF-and-UMD-table-768x731.png 768w" sizes="(max-width: 1222px) 100vw, 1222px" /></figure>



<h4 class="wp-block-heading">Forest disturbances related to shifting cultivation in Mongala, DRC</h4>



<div class="flourish-embed flourish-photo-slider" data-src="visualisation/16055613?163031"><script src="https://public.flourish.studio/resources/embed.js"></script></div>



<h4 class="wp-block-heading">Large scale deforestation along BR-230 in Amazonas, Brazil</h4>



<div class="flourish-embed flourish-photo-slider" data-src="visualisation/16055698?163031"><script src="https://public.flourish.studio/resources/embed.js"></script></div>



<h4 class="wp-block-heading">Selective logging in Pará, Brazil</h4>



<div class="flourish-embed flourish-photo-slider" data-src="visualisation/16365261?163031"><script src="https://public.flourish.studio/resources/embed.js"></script></div>



<h2 class="wp-block-heading">Why does the UMD and JRC data align in some ways and differ in others?</h2>



<p>The two data sets align more closely for more abrupt disturbances that are likely to meet the definitions used in both data sets, such as intense fires (which the JRC data may classify as long-duration degradation) or clearance of tree cover for agricultural expansion (usually classified as deforestation in the JRC data). The two data sets are less likely to detect the same disturbance when it is followed by a rapid recovery (such as short-duration degradation in the JRC data, including disturbances like drought, selective logging or low-intensity fires, which may only be detected in as few as one Landsat image) or disturbances resulting in gradual reduction of tree cover over a longer time period (such as degradation followed by deforestation in the JRC data). In these cases, the disturbance may not meet the UMD definition of tree cover loss and therefore is less likely to be detected in the UMD data.</p>



<p>Both data sets have different strengths. The UMD data is global in scope, allowing for comparative analysis across regions due to global consistency in definitions and methods. Within the tropics, the UMD data includes broader coverage of forest types beyond tropical moist forests, including tropical dry forests.<a href="https://gfw.global/3GEmSQh"></a> The UMD data also provides flexibility for users to adjust the canopy density threshold according to their preference or needs, which can be used to align with a specific forest definition or forest characteristics of a specific biome.</p>



<p>Meanwhile, the JRC data detects degradation and less intense disturbances, which can be helpful for users who want to understand where forests have experienced various types of stresses that have not resulted in complete loss. The JRC data also monitors forests after initial disturbances to track forest recovery and subsequent disturbances, which is helpful for users who would like to understand yearly change dynamics and identify where there are repeated disturbances.</p>



<p>However, it is important to note that we are not evaluating the accuracy of either data set. Assessing which data set is more accurate for a specific use case or geography would require a formal validation exercise.</p>



<p>Ultimately, these two data sets present complementary information that can help us better understand the various types of changes occurring in tropical forests.</p>



<hr> 

<div class="attribute" style="”padding-bottom:" 10px;”="">
 

Note: This blog investigates the JRC TMF map, which is a separate product from the JRC global map of forest cover for year 2020, released on December 7, 2023.

 

</div> 

 

<hr>



<style type="text/css"></style>

Differences Between Global Forest Watch’s Tree Cover Loss Data and JRC’s Tropical Moist Forest Data Explained

<p>The wide availability of forest and land cover maps based on satellite data provides more information on the world’s forests than ever before. However, this plethora of information can also be challenging for people trying to understand differences in data sources and decide which data to use for a specific purpose.</p>



<p>In the past, we’ve compared the <a href="https://gfw.global/3BfwQWz">annual tree cover loss data</a> from the University of Maryland (UMD) and available on Global Forest Watch (GFW) with national data from <a href="https://www.globalforestwatch.org/blog/data-and-research/2022-tree-cover-loss-data-explained/">Brazil and Indonesia</a>, as well as with data from the United Nations Food and Agriculture Organization’s (FAO) <a href="https://www.globalforestwatch.org/blog/data-and-research/global-forest-watch-and-the-forest-resources-assessment-explained-in-5-graphics-2/">Global Forest Resources Assessment</a>, to shed light on some of these differences.</p>



<p>Another data set that tracks forest change in the tropics is the <a href="https://forobs.jrc.ec.europa.eu/TMF" target="_blank" rel="noreferrer noopener">Tropical Moist Forest (TMF) data set</a> from the European Commission Joint Research Centre (JRC). Like the UMD tree cover loss data, the JRC TMF data is based on Landsat satellite imagery and is updated annually. Yet while both data sets map forest disturbances, the maps and numbers look a little different.</p>



<p>In this blog, we explain some key differences in the definitions and methods used for each data set that impact the resulting maps and discuss how the disturbances mapped by both data sets compare (while the data sets also map forest gain or regrowth, this blog focuses on comparing their mapping of forest disturbances). While understanding these differences can help determine the contexts for which each data set might be useful, note that this analysis does not evaluate the quality or accuracy of either data set.</p>



<h2 class="wp-block-heading">What areas do the JRC and UMD data sets cover, and how do they define forests?</h2>



<p>The UMD data set maps forests globally based on <a href="https://www.science.org/doi/10.1126/science.1244693" target="_blank" rel="noreferrer noopener">tree cover extent</a> in 2000. <a href="https://www.science.org/doi/10.1126/science.1244693" target="_blank" rel="noreferrer noopener">Tree cover</a> is defined as woody vegetation with a minimum height of at least five meters at 30-meter resolution, which includes all types of tree cover — including natural or managed forest and plantations — and is not limited to a specific forest type. The GFW map shows tree cover with a 30% canopy density, which we use in our analyses and in this comparison, although users can adjust the threshold based on their needs.</p>



<p>The JRC data set, on the other hand, <a href="https://www.science.org/doi/10.1126/sciadv.abe1603" target="_blank" rel="noreferrer noopener">maps forests</a> beginning in 1990. The data includes all closed primary and secondary forests in the humid tropics and does not apply a minimum height or canopy density threshold to delineate forest extent.</p>



<p>As a result of these definitional differences, some areas included in the 2000 UMD tree cover extent in the tropics like dry forests, young secondary forests and plantations are not classified as forests in the JRC data set in 2000 — a total of 655 million hectares (Mha), or 34%, of the UMD tree cover extent area that overlaps with the JRC data. Likewise, the JRC forest extent in 2000 includes some areas that do not meet the height and 30% canopy density threshold in the UMD data or are otherwise not mapped as tree cover by the UMD data — 17.2 Mha, or 1% of the JRC forest extent.</p>



<p>Because of the differences in geographic coverage and scope between the two data sets, this blog only compares forest disturbances from 2001-2022 for areas that, in the year 2000, are mapped both as tree cover by UMD and as forest by JRC.</p>



<h2 class="wp-block-heading">What types of forest disturbances do the UMD and JRC data map?</h2>



<h3 class="wp-block-heading"><em>The JRC Tropical Moist Forest data set maps degradation and deforestation in the humid tropics</em></h3>



<p>The JRC TMF data set maps disturbances to tropical moist forests — including degradation and deforestation — from 1990 to 2022. The JRC team maps these changes using an algorithm that detects forest disruptions, meaning an absence of tree foliage cover within a 30-meter pixel for each available Landsat image, and then classifies the status of forest based on how long these disruptions are visible.</p>



<p>The TMF data set <a href="https://www.science.org/doi/10.1126/sciadv.abe1603" target="_blank" rel="noreferrer noopener">defines degradation</a> as temporary disturbances occurring for up to 2.5 years, while it defines deforestation as longer-term disturbances where disruptions to tree cover are visible for more than 2.5 years and no tree regrowth is detected over the last three years. In any given year, a forest is considered disturbed if there is a disruption detected in at least one Landsat image during that year.</p>



<p>Thus, pixels classified as degraded are still considered forest, with existing or regrowing trees, while pixels classified as deforested are undergoing a longer-term loss of forest. Because disruptions are recorded and analyzed for the entire time series, the data set classifies dynamics over time; for example, a forest pixel may be classified as degraded in one year after a short-term disturbance, then after a period of recovery may later be classified as deforested after ongoing disturbances with no tree regrowth. The JRC deforestation class therefore includes direct deforestation (deforestation of undisturbed forest) or deforestation after degradation (deforestation of forest that was previously classified as degraded).</p>



<p>The data set also analyzes the frequency, duration of disruptions and time of forest recovery in between disruption events to identify deforestation following degradation versus multiple degradation events.</p>



<h3 class="wp-block-heading"><em>The UMD tree cover loss data set on GFW maps all types of tree cover loss globally</em></h3>



<p>The UMD tree cover loss data set maps annual tree cover loss from 2001 to 2022. The UMD team uses an algorithm that detects loss using change detection metrics based on comparison of Landsat images between the present and past years. Tree cover loss, <a href="https://www.science.org/doi/10.1126/science.1244693" target="_blank" rel="noreferrer noopener">defined as</a> a clearing of tree cover of at least a half of a 30-meter Landsat pixel, includes both temporary and permanent losses without defining or differentiating which loss is considered degradation or deforestation. Once classified as loss, pixels are masked from further updates, meaning that any following forest change dynamics — such as regrowth or subsequent disturbances — are not recorded.</p>



<p>The UMD tree cover loss data can be combined with contextual data layers, such as <a href="https://gfw.global/3kqcoqa">UMD’s map of humid tropical primary forest</a>, to see loss within specific forest types.</p>



<h4 class="wp-block-heading">Key differences between the JRC TMF and UMD tree cover loss data</h4>



<table border="1" cellpadding="0" cellspacing="0">
	<tbody>
		<tr>
			<td style="width:165px;height:20px;">
			<p>&nbsp;</p>
			</td>
			<td style="width:320px;height:20px;">
			<p><strong>JRC TMF</strong></p>
			</td>
			<td style="width:320px;height:20px;">
			<p><strong>UMD tree cover loss</strong></p>
			</td>
		</tr>
		<tr>
			<td style="width:165px;height:20px;">
			<p>Temporal coverage</p>
			</td>
			<td style="width:320px;height:20px;">
			<p>1990-2022</p>
			</td>
			<td style="width:320px;height:20px;">
			<p>&nbsp;2000-2022</p>
			</td>
		</tr>
		<tr>
			<td style="width:165px;height:20px;">
			<p>Geographic coverage and extent</p>
			</td>
			<td style="width:320px;height:20px;">
			<p><strong>Tropical moist forest:</strong> closed primary and secondary forests in the humid tropics in 1990.</p>
			</td>
			<td style="width:320px;height:20px;">
			<p><strong>Global tree cover: </strong>woody vegetation with a minimum height of at least five meters at 30-meter resolution in 2000.</p>
			</td>
		</tr>
		<tr>
			<td style="width:165px;height:20px;">
			<p>Definition of disturbance</p>
			</td>
			<td style="width:320px;height:20px;">
			<p>An absence of tree foliage cover within a 30-meter pixel.</p>
			</td>
			<td style="width:320px;height:20px;">
			<p>Clearing of at least half of tree cover within a 30-meter pixel.</p>
			</td>
		</tr>
		<tr>
			<td style="width:165px;height:20px;">
			<p>Disturbance classes</p>
			</td>
			<td style="width:320px;height:20px;">
			<p>Disturbances are classified as <strong>degradation </strong>if they are visible for less than 2.5 years, and <strong>deforestation </strong>if they are visible for more than 2.5 years with no vegetative regrowth over the last three years.</p>
			</td>
			<td style="width:320px;height:20px;">
			<p>All disturbances classified as <strong>tree cover loss. </strong>The length of disturbance is not considered.</p>
			</td>
		</tr>
		<tr>
			<td style="width:165px;height:20px;">
			<p>Methods</p>
			</td>
			<td style="width:320px;height:20px;">
			<p>Classifies valid single-date Landsat images and derives disturbance classes by analyzing the single-date classifications. Dynamics of disturbance events over time are recorded.</p>
			</td>
			<td style="width:320px;height:20px;">
			<p>Classifies change using change detection metrics derived from all valid Landsat images over the present and past years. Only the first loss event that is detected is recorded.</p>
			</td>
		</tr>
	</tbody>
</table>



<h2 class="wp-block-heading">What do the UMD and JRC data show about forest disturbances in the tropics?</h2>



<p>Overall, the JRC data detects a larger area of disturbances across the tropics from 2001-2022 compared with the UMD data. The JRC data detects 258 Mha of disturbances, including 113 Mha of degradation and 145 Mha of deforestation, while the UMD data detects 143 Mha of tree cover loss. However, trends in total deforestation mapped by JRC and tree cover loss mapped by UMD are generally aligned: They are largely similar for the first decade; then from 2013-2015, JRC detects a larger area as deforestation than UMD detects as loss, and after 2016, UMD detects a larger area.</p>



<div class="flourish-embed" data-src="story/2112582?163031"><script src="https://public.flourish.studio/resources/embed.js"></script></div>



<p>We further separated out the disturbance types mapped by the JRC data and overlaid the data sets to evaluate how the two data sets compare at each pixel.</p>



<p>Among all the JRC data’s disturbance classes, the JRC direct deforestation class aligned most closely with UMD tree cover loss. Across the tropics from 2001-2022, 46.8 Mha, or 68%, of the area classified as direct deforestation by the JRC data was also classified as tree cover loss in the UMD data. This varied by region and was highest in Asia/Oceania and the Americas (66% and 75% respectively) and lowest in Africa (50%). The UMD data less often classified areas that JRC identified as deforestation after degradation as tree cover loss (38.1 Mha, or 50% across the tropics).</p>



<p>In some areas, disturbances classified as degradation by the JRC TMF data set were classified as tree cover loss by the UMD data. Across the tropics from 2001-2022, 32.6 Mha, or 29%, of the area the JRC data classified as degradation was also classified as tree cover loss in the UMD data. The JRC degradation class includes three types of degradation: short-duration (less than one year), long-duration (1-2.5 years) and short-duration over 2-3 periods (short-duration degradation that occurs two to three separate times throughout the time period). The JRC data also measures “intensity” of degradation based on the total number of Landsat images where disruptions in tree cover were visible. We found that, in general, the UMD data:</p>



<ul>
<li>More often detected longer duration degradation as tree cover loss. Short-duration degradation (including events occurring just once or over 2-3 periods), which represents 85% of the forest area mapped as degraded by JRC since 2001, is usually not shown as tree cover loss by the UMD data.</li>



<li>Classified degradation events as tree cover loss that had an intensity of greater than six images on average, while those UMD did not classify as loss had approximately three images or less on average.</li>



<li>Detected a larger proportion of the area JRC classified as degradation from 2015 onwards, which may be <a href="https://www.globalforestwatch.org/blog/data-and-research/tree-cover-loss-satellite-data-trend-analysis/">due to the algorithm changes</a> in the UMD data that increase forest loss detection sensitivity in case of fires, selective logging and other disturbance types that JRC classified as degradation.</li>
</ul>



<p>In other areas across the tropics, the UMD data detected tree cover loss for 6.4 Mha, or 7%, of areas where there the JRC data detected no disturbance events.</p>



<figure class="wp-block-image size-full is-resized"><img src="https://content.globalforestwatch.org/wp-content/uploads/2024/01/Comparison-JRC-TMF-and-UMD-table.png" alt="Comparison of JRC TMF and UMD tree cover loss disturbances across the tropics" class="wp-image-16572" style="aspect-ratio:1.0507308684436802;width:771px;height:auto"/></figure>



<h4 class="wp-block-heading">Forest disturbances related to shifting cultivation in Mongala, DRC</h4>



<div class="flourish-embed flourish-photo-slider" data-src="visualisation/16055613?163031"><script src="https://public.flourish.studio/resources/embed.js"></script></div>



<h4 class="wp-block-heading">Large scale deforestation along BR-230 in Amazonas, Brazil</h4>



<div class="flourish-embed flourish-photo-slider" data-src="visualisation/16055698?163031"><script src="https://public.flourish.studio/resources/embed.js"></script></div>



<h4 class="wp-block-heading">Selective logging in Pará, Brazil</h4>



<div class="flourish-embed flourish-photo-slider" data-src="visualisation/16365261?163031"><script src="https://public.flourish.studio/resources/embed.js"></script></div>



<h2 class="wp-block-heading">Why does the UMD and JRC data align in some ways and differ in others?</h2>



<p>The two data sets align more closely for more abrupt disturbances that are likely to meet the definitions used in both data sets, such as intense fires (which the JRC data may classify as long-duration degradation) or clearance of tree cover for agricultural expansion (usually classified as deforestation in the JRC data). The two data sets are less likely to detect the same disturbance when it is followed by a rapid recovery (such as short-duration degradation in the JRC data, including disturbances like drought, selective logging or low-intensity fires, which may only be detected in as few as one Landsat image) or disturbances resulting in gradual reduction of tree cover over a longer time period (such as degradation followed by deforestation in the JRC data). In these cases, the disturbance may not meet the UMD definition of tree cover loss and therefore is less likely to be detected in the UMD data.</p>



<p>Both data sets have different strengths. The UMD data is global in scope, allowing for comparative analysis across regions due to global consistency in definitions and methods. Within the tropics, the UMD data includes broader coverage of forest types beyond tropical moist forests, including tropical dry forests.<a href="https://gfw.global/3GEmSQh"></a> The UMD data also provides flexibility for users to adjust the canopy density threshold according to their preference or needs, which can be used to align with a specific forest definition or forest characteristics of a specific biome.</p>



<p>Meanwhile, the JRC data detects degradation and less intense disturbances, which can be helpful for users who want to understand where forests have experienced various types of stresses that have not resulted in complete loss. The JRC data also monitors forests after initial disturbances to track forest recovery and subsequent disturbances, which is helpful for users who would like to understand yearly change dynamics and identify where there are repeated disturbances.</p>



<p>However, it is important to note that we are not evaluating the accuracy of either data set. Assessing which data set is more accurate for a specific use case or geography would require a formal validation exercise.</p>



<p>Ultimately, these two data sets present complementary information that can help us better understand the various types of changes occurring in tropical forests.</p>



<hr> 

<div class="attribute" style="”padding-bottom:" 10px;”="">
 

Note: This blog investigates the JRC TMF map, which is a separate product from the JRC global map of forest cover for year 2020, released on December 7, 2023.

 

</div> 

 

<hr>