<p>A new high-resolution dataset on Global Forest Watch reveals where forest roads are opening up in the Congo Basin. Using satellite imagery and AI, it maps roads first detected from 2019 onward and combines them with historical data back to 1976, offering an unprecedented view of road expansion across the region. Explore the <a href="https://gfw.global/82nKz4">dataset</a> and read <a href="https://www.globalforestwatch.org/blog/data-and-tools/congo-basin-forest-road-mapping/?utm_medium=notification&#038;utm_source=gfwnoticeboard&#038;utm_campaign=loggingroads">our blog</a> to learn about the data and how tracking these roads can help monitor forest usage and fragmentation.</p>


New on GFW: Congo Basin forest roads update

Data &amp; Tools

Data &amp; Tools | Global Forest Watch Blog

Congo Basin

Congo Basin | Global Forest Watch Blog

logging

logging | Global Forest Watch Blog

mapping

mapping | Global Forest Watch Blog

Anika Berger

Bart Slagter

Sarah Carter

<p>In Congo Basin forests, a new road can cause big changes. What may start as a single narrow path can start a cascade of disturbances by opening up access to intact forest, making it vulnerable to agricultural expansion, mining, logging, poaching and other threats. Tracking these roads is an important step in reducing their impact on forests, yet these roads — often unpaved, temporary and quickly overgrown — can be difficult to detect.</p>



<figure class="wp-block-image size-full"><img decoding="async" loading="lazy" width="2048" height="1363" src="https://content.globalforestwatch.org/wp-content/uploads/2026/03/CIFOR-ICRAF-Wood-truck-Cameroon_Ollivier-Girard.jpg" alt="Wood truck near the village of Ngon. District of Ebolowa, Cameroon. Photo credit: Ollivier Girard/CIFOR." class="wp-image-20073" srcset="https://content.globalforestwatch.org/wp-content/uploads/2026/03/CIFOR-ICRAF-Wood-truck-Cameroon_Ollivier-Girard.jpg 2048w, https://content.globalforestwatch.org/wp-content/uploads/2026/03/CIFOR-ICRAF-Wood-truck-Cameroon_Ollivier-Girard-1024x682.jpg 1024w, https://content.globalforestwatch.org/wp-content/uploads/2026/03/CIFOR-ICRAF-Wood-truck-Cameroon_Ollivier-Girard-768x511.jpg 768w, https://content.globalforestwatch.org/wp-content/uploads/2026/03/CIFOR-ICRAF-Wood-truck-Cameroon_Ollivier-Girard-1536x1022.jpg 1536w" sizes="(max-width: 2048px) 100vw, 2048px" /><figcaption class="wp-element-caption"><em>Wood truck near the village of Ngon. District of Ebolowa, Cameroon. Photo credit: Ollivier Girard/CIFOR.</em></figcaption></figure>



<p>A <a href="https://gfw.global/kkkquy">new, high-resolution data set</a> on Global Forest Watch (GFW) helps to close this gap by providing the most comprehensive view to date of forest road development in the Congo Basin. This updated data set builds on recent research that uses <a href="https://www.sciencedirect.com/science/article/pii/S0034425724004061#bb0225" target="_blank" rel="noreferrer noopener">satellite imagery and AI to map forest roads</a> in unprecedented detail, identifying roads first detected from 2019 onward and integrating them with a historical archive of road data extending from 2018 back to 1976. Together, this offers decades of insight into how roads have expanded across one of the world’s most vital tropical forest regions.</p>



<p>This release comes at a critical moment. Road expansion, particularly associated with selective logging, can lead to serious <a href="https://gfr.wri.org/forest-condition-indicators/forest-degradation" target="_blank" rel="noreferrer noopener">forest degradation</a> by fragmenting habitats, increasing carbon emissions and opening previously intact areas to further clearing. Data tracking road development can support verification of logging legality, as well as monitor mining encroachment and agricultural expansion.</p>



<p>In the Congo Basin, there is strong and growing demand for this type of data from government ministries, civil society organizations and forest monitoring initiatives. With technological developments such as this AI-powered approach, we are able to automatically track in high detail where and when road networks expand, meeting demands for valuable information needed by those who monitor forest change in near-real-time.</p>



<figure class="wp-block-image size-full"><img decoding="async" loading="lazy" width="1920" height="1080" src="https://content.globalforestwatch.org/wp-content/uploads/2026/03/congobasinroads_fast.gif" alt="Congo Basin forest roads data set overlaid on satellite imagery in Republic of the Congo" class="wp-image-20078"/><figcaption class="wp-element-caption"><em>Forest roads visible in Sentinel-2 satellite imagery can later disappear as vegetation regrows, but the Congo Basin forest roads data set&nbsp;retains&nbsp;a record of these roads even after they are no longer visible in imagery.</em></figcaption></figure>



<h2 class="wp-block-heading">Why do forest roads matter?</h2>



<p>Forest roads are often the first physical footprint of industrial activity in tropical forests. Most of the roads in the Congo Basin are unpaved, and only temporarily used for selective logging, which entails harvesting valuable and mature trees while leaving the majority of the forest intact. While forest products benefit regional economies, some extractive practices are more detrimental than others. In tropical forests, selective logging can be part of a sustainable management regime if done with care, compared to clear cutting and establishing timber plantations.</p>



<p>However, the impact of logging roads<a href="https://ethz.ch/content/dam/ethz/special-interest/usys/ites/ecosystem-management-dam/documents/EducationDOC/TRF_DOC/Reading/Laurance_2009.pdf" target="_blank" rel="noreferrer noopener"> can be long-lasting</a>. Clearing forest corridors for road construction is the <a href="https://www.researchgate.net/publication/331479711_Reduced-impact_logging_for_climate_change_mitigation_RIL-C_can_halve_selective_logging_emissions_from_tropical_forests" target="_blank" rel="noreferrer noopener">largest source of carbon emissions</a> associated with selective logging in Congo Basin countries, and also contributes to habitat fragmentation, soil disturbance, edge effects and increased fire risk.</p>



<p>The impacts of logging roads often extend well beyond the initial harvest. When roads remain accessible after logging, they can leave remote forests vulnerable to <a href="https://www.researchgate.net/publication/317802710_Impacts_of_logging_roads_on_tropical_forests" target="_blank" rel="noreferrer noopener">unauthorized activities</a> such as agricultural expansion, mining, hunting and unregulated logging. In recent decades, road networks have increasingly expanded into <a href="https://www.wri.org/insights/worlds-last-intact-forests-increasingly-fragmented" target="_blank" rel="noreferrer noopener">intact forest landscapes</a>, which are identified as having little or no prior human influence. The <a href="https://www.wur.nl/en/news/road-networks-stretch-congo-basin-forests-satellites-and-ai-help-monitor" target="_blank" rel="noreferrer noopener">latest data indicates</a> that almost a quarter of Congo Basin forest roads are located in forests that were mapped as intact as of 2016.</p>



<h2 class="wp-block-heading">What does the new Congo Basin forest roads data show?</h2>



<p>Monitoring forest roads is important but challenging. Without knowing where these roads are and how they’re changing over time, it’s impossible to assess their impacts or know what actions need to be taken. Many logging roads are narrow and vegetation grows back quickly after harvest, making them difficult or impossible to detect without the use of radar and optical satellite imagery. Especially in the densely cloud-covered Congo Basin, transient roads going unnoticed pose a major risk.</p>



<p>The new forest roads data set provides highly detailed mapping of road development in tropical rainforests across six Congo Basin countries: Cameroon, Central African Republic, the Democratic Republic of the Congo, Equatorial Guinea, Gabon and Republic of Congo.</p>



<p>Key features include:</p>



<ul>
<li><strong>Updated road detection from 2019 onward</strong>, using AI and both optical and radar-based satellite imagery to map transient forest roads that often go undetected in conventional road data sets, with roads attributed by detection year and viewable by time range on GFW.</li>



<li><strong>A long-term historical archive</strong>, combining the new data with earlier road maps created from satellite imagery going from 2018 back to 1976.</li>



<li><strong>Improved coverage and consistency</strong>, offering users a single, integrated data set rather than multiple sources.</li>
</ul>



<p>Prior to this change, GFW hosted a Congo Basin logging roads data set that was last updated in 2014. That data set was produced through a crowdsourcing effort from the Logging Roads initiative led by Moabi, GFW and the European Commission’s Joint Research Centre. Volunteers used satellite imagery and OpenStreetMap-based tools to manually trace roads and estimate when they appeared. While groundbreaking at the time, the data set covered a limited time period and was difficult to update.</p>



<p>In subsequent years, researchers <a href="https://www.nature.com/articles/s41893-019-0310-6" target="_blank" rel="noreferrer noopener">manually updated</a> the data set through 2018 and integrated it with an earlier road inventory that had <a href="https://www.science.org/doi/10.1126/science.1141057" target="_blank" rel="noreferrer noopener">digitized roads</a> from more than 300 Landsat images collected between 1976 and 2003, creating a comprehensive archive of roads data.</p>



<p>The new data combines the hand-digitized data with a new AI-based approach to offer a more complete and current picture of forest road expansion across the region.</p>



<p>Users can view the Congo Basin forest roads data on the GFW map by selecting the <a href="https://www.globalforestwatch.org/map/?map=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%3D&mapMenu=eyJtZW51U2VjdGlvbiI6ImRhdGFzZXRzIiwiZGF0YXNldENhdGVnb3J5IjoibGFuZFVzZSJ9">Congo Basin forest roads layer</a> in the Land Use data sets section under Infrastructure.</p>



<figure class="wp-block-image size-full"><img decoding="async" loading="lazy" width="1310" height="788" src="https://content.globalforestwatch.org/wp-content/uploads/2026/03/congobasinroads.png" alt="Screenshot of Forest roads in the Congo Basin (1976 to 2024)" class="wp-image-20079" srcset="https://content.globalforestwatch.org/wp-content/uploads/2026/03/congobasinroads.png 1310w, https://content.globalforestwatch.org/wp-content/uploads/2026/03/congobasinroads-1024x616.png 1024w, https://content.globalforestwatch.org/wp-content/uploads/2026/03/congobasinroads-768x462.png 768w" sizes="(max-width: 1310px) 100vw, 1310px" /><figcaption class="wp-element-caption"><em>Forest roads in the Congo Basin (1976 to 2024), with older data shown in darker brown and newly detected roads highlighted in bright gold.</em></figcaption></figure>



<p>Additionally, the data will be integrated into national <a href="https://www.wri.org/initiatives/forest-atlases" target="_blank" rel="noreferrer noopener">Forest Atlases</a>, which combine government information with independent monitoring tools to support transparent forest management.&nbsp;</p>



<h2 class="wp-block-heading"><strong>How was the data created?</strong></h2>



<p>We mapped roads from 2019 onward using an advanced deep learning model <a href="https://www.wur.nl/en/news/road-networks-stretch-congo-basin-forests-satellites-and-ai-help-monitor" target="_blank" rel="noreferrer noopener">developed by Wageningen University & Research and The Nature Conservancy</a>, based on both Sentinel-1 and Sentinel-2 satellite imagery. The imagery uses radar sensors that can “see through” clouds, which allows for road detections even during the Congo Basin’s frequent rainy seasons. The automated algorithm processed thousands of satellite images and was designed to be continuously updated.</p>



<p>To extend the data set further back in time, the new road data was merged with <a href="https://www.nature.com/articles/s41893-019-0310-6" target="_blank" rel="noreferrer noopener">historical maps</a> that include the <a href="https://www.science.org/doi/10.1126/science.1141057" target="_blank" rel="noreferrer noopener">manually digitized roads</a> from Landsat imagery covering 1976–2003 and crowdsourced mapping efforts. While detection methods differ, together they provide a valuable long-term archive of forest access and fragmentation.</p>



<p>As with most large-scale data sets, there are important caveats that users should be aware of:</p>



<ul>
<li><strong>Not all forest roads are associated with selective logging</strong>, although many mapped roads are linked to logging activities. The <a href="https://gfw.global/OQbmTK">logging concessions layer</a> on the GFW map can provide further context.</li>



<li>The data set does <strong>not indicate the legality</strong> of roads or associated activities.</li>



<li><strong>Road widths are not shown to scale</strong>, and the data does <strong>not show whether a road is still in use</strong> (for example, some roads may have overgrown since they were detected).</li>



<li>Additionally, overlap between older and newer roads may indicate the reopening of older roads. This may be of interest to users investigating selective logging activities because a road reopening can indicate that a new harvest cycle has started at a location where old, overgrown logging roads are being reopened and reused (which generally has <a href="https://besjournals.onlinelibrary.wiley.com/doi/epdf/10.1111/1365-2664.12661" target="_blank" rel="noreferrer noopener">less environmental impact</a> than constructing entirely new ones). However, note that <strong>reopened roads may not perfectly align with new roads</strong> on the map, since new roads (2019 onward) were mapped with more spatial precision than old roads.</li>
</ul>



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



<p>The updated forest roads data set supports a wide range of applications:</p>



<ul>
<li><strong>Forest monitoring and law enforcement:</strong> Users can track where new roads appear inside or outside <a href="https://gfw.global/OQbmTK">logging concessions</a>, helping to flag potential illegal activity and support patrol planning.</li>



<li><strong>Concession oversight:</strong> Governments and civil society organizations can assess whether road development aligns with approved logging plans and schedules. Detailed logging road maps can also be used to evaluate and potentially reduce the impact of selective logging activities.</li>



<li><strong>Research and analysis:</strong> Researchers can use the long-term archive to study logging trends, infrastructure persistence and the carbon implications of new versus reopened roads, as well as the relationship between roads, degradation and deforestation.</li>



<li><strong>Conservation and land-use planning:</strong> By revealing patterns of access and fragmentation, the data set can inform broader efforts to protect intact forest landscapes and protected areas.</li>
</ul>



<figure class="wp-block-image size-full"><img decoding="async" loading="lazy" width="1310" height="788" src="https://content.globalforestwatch.org/wp-content/uploads/2026/03/congobasinroads_v2.png" alt="Screenshot of Forest roads in the Congo Basin in Republic of Congo" class="wp-image-20080" srcset="https://content.globalforestwatch.org/wp-content/uploads/2026/03/congobasinroads_v2.png 1310w, https://content.globalforestwatch.org/wp-content/uploads/2026/03/congobasinroads_v2-1024x616.png 1024w, https://content.globalforestwatch.org/wp-content/uploads/2026/03/congobasinroads_v2-768x462.png 768w" sizes="(max-width: 1310px) 100vw, 1310px" /><figcaption class="wp-element-caption"><em>The forest roads data set provides context to forest change data such as near-real-time disturbance alerts, informing users what may be driving forest change. For example, recent disturbance alerts occurring near forest roads in intact forest landscapes highlight areas where access infrastructure and potential logging operations may be contributing to the reduction of previously intact forest.</em></figcaption></figure>



<h2 class="wp-block-heading"><strong>What’s next for forest road monitoring data?</strong></h2>



<p>By combining cutting-edge AI with decades of historical mapping, this new forest roads data set offers an unprecedented view of how access has shaped the Congo Basin’s forests over time. As road expansion continues to impact forests, having timely, transparent and accessible data is essential for effective monitoring and action.</p>



<p>Moving forward, we plan to update the data set annually to support continued monitoring of road expansion across the region. In the future, we may apply these methods to map roads in other regions such as the Amazon and Southeast Asian rainforests.</p>



<hr class="wp-block-separator has-alpha-channel-opacity"/>



<p><em>Access the data: The data set is </em><a href="https://gfw.global/kkkquy"><em>now available on GFW</em></a><em> as well as the </em><a href="https://www.wri.org/initiatives/forest-atlases" target="_blank" rel="noreferrer noopener"><em>Forest Atlases</em></a><em> where it <strong>complements existing forest disturbance data</strong>, such as </em><a href="https://gfw.global/qfkEaG"><em>drivers of disturbance alerts</em></a><em> and </em><a href="https://gfw.global/RIZqAn"><em>tree cover loss by dominant driver</em></a><em>, by<strong> adding spatial detail on logging-related access and infrastructure</strong>, helping users better understand forest change in the Congo Basin.</em></p>



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

Mapping Forest Roads in the Congo Basin in Unprecedented Detail with AI

<p>In Congo Basin forests, a new road can cause big changes. What may start as a single narrow path can start a cascade of disturbances by opening up access to intact forest, making it vulnerable to agricultural expansion, mining, logging, poaching and other threats. Tracking these roads is an important step in reducing their impact on forests, yet these roads — often unpaved, temporary and quickly overgrown — can be difficult to detect.</p>



<figure class="wp-block-image size-full"><img src="https://content.globalforestwatch.org/wp-content/uploads/2026/03/CIFOR-ICRAF-Wood-truck-Cameroon_Ollivier-Girard.jpg" alt="Wood truck near the village of Ngon. District of Ebolowa, Cameroon. Photo credit: Ollivier Girard/CIFOR." class="wp-image-20073"/><figcaption class="wp-element-caption"><em>Wood truck near the village of Ngon. District of Ebolowa, Cameroon. Photo credit: Ollivier Girard/CIFOR.</em></figcaption></figure>



<p>A <a href="https://gfw.global/kkkquy">new, high-resolution data set</a> on Global Forest Watch (GFW) helps to close this gap by providing the most comprehensive view to date of forest road development in the Congo Basin. This updated data set builds on recent research that uses <a href="https://www.sciencedirect.com/science/article/pii/S0034425724004061#bb0225" target="_blank" rel="noreferrer noopener">satellite imagery and AI to map forest roads</a> in unprecedented detail, identifying roads first detected from 2019 onward and integrating them with a historical archive of road data extending from 2018 back to 1976. Together, this offers decades of insight into how roads have expanded across one of the world’s most vital tropical forest regions.</p>



<p>This release comes at a critical moment. Road expansion, particularly associated with selective logging, can lead to serious <a href="https://gfr.wri.org/forest-condition-indicators/forest-degradation" target="_blank" rel="noreferrer noopener">forest degradation</a> by fragmenting habitats, increasing carbon emissions and opening previously intact areas to further clearing. Data tracking road development can support verification of logging legality, as well as monitor mining encroachment and agricultural expansion.</p>



<p>In the Congo Basin, there is strong and growing demand for this type of data from government ministries, civil society organizations and forest monitoring initiatives. With technological developments such as this AI-powered approach, we are able to automatically track in high detail where and when road networks expand, meeting demands for valuable information needed by those who monitor forest change in near-real-time.</p>



<figure class="wp-block-image size-full"><img src="https://content.globalforestwatch.org/wp-content/uploads/2026/03/congobasinroads_fast.gif" alt="Congo Basin forest roads data set overlaid on satellite imagery in Republic of the Congo" class="wp-image-20078"/><figcaption class="wp-element-caption"><em>Forest roads visible in Sentinel-2 satellite imagery can later disappear as vegetation regrows, but the Congo Basin forest roads data set&nbsp;retains&nbsp;a record of these roads even after they are no longer visible in imagery.</em></figcaption></figure>



<h2 class="wp-block-heading">Why do forest roads matter?</h2>



<p>Forest roads are often the first physical footprint of industrial activity in tropical forests. Most of the roads in the Congo Basin are unpaved, and only temporarily used for selective logging, which entails harvesting valuable and mature trees while leaving the majority of the forest intact. While forest products benefit regional economies, some extractive practices are more detrimental than others. In tropical forests, selective logging can be part of a sustainable management regime if done with care, compared to clear cutting and establishing timber plantations.</p>



<p>However, the impact of logging roads<a href="https://ethz.ch/content/dam/ethz/special-interest/usys/ites/ecosystem-management-dam/documents/EducationDOC/TRF_DOC/Reading/Laurance_2009.pdf" target="_blank" rel="noreferrer noopener"> can be long-lasting</a>. Clearing forest corridors for road construction is the <a href="https://www.researchgate.net/publication/331479711_Reduced-impact_logging_for_climate_change_mitigation_RIL-C_can_halve_selective_logging_emissions_from_tropical_forests" target="_blank" rel="noreferrer noopener">largest source of carbon emissions</a> associated with selective logging in Congo Basin countries, and also contributes to habitat fragmentation, soil disturbance, edge effects and increased fire risk.</p>



<p>The impacts of logging roads often extend well beyond the initial harvest. When roads remain accessible after logging, they can leave remote forests vulnerable to <a href="https://www.researchgate.net/publication/317802710_Impacts_of_logging_roads_on_tropical_forests" target="_blank" rel="noreferrer noopener">unauthorized activities</a> such as agricultural expansion, mining, hunting and unregulated logging. In recent decades, road networks have increasingly expanded into <a href="https://www.wri.org/insights/worlds-last-intact-forests-increasingly-fragmented" target="_blank" rel="noreferrer noopener">intact forest landscapes</a>, which are identified as having little or no prior human influence. The <a href="https://www.wur.nl/en/news/road-networks-stretch-congo-basin-forests-satellites-and-ai-help-monitor" target="_blank" rel="noreferrer noopener">latest data indicates</a> that almost a quarter of Congo Basin forest roads are located in forests that were mapped as intact as of 2016.</p>



<h2 class="wp-block-heading">What does the new Congo Basin forest roads data show?</h2>



<p>Monitoring forest roads is important but challenging. Without knowing where these roads are and how they’re changing over time, it’s impossible to assess their impacts or know what actions need to be taken. Many logging roads are narrow and vegetation grows back quickly after harvest, making them difficult or impossible to detect without the use of radar and optical satellite imagery. Especially in the densely cloud-covered Congo Basin, transient roads going unnoticed pose a major risk.</p>



<p>The new forest roads data set provides highly detailed mapping of road development in tropical rainforests across six Congo Basin countries: Cameroon, Central African Republic, the Democratic Republic of the Congo, Equatorial Guinea, Gabon and Republic of Congo.</p>



<p>Key features include:</p>



<ul>
<li><strong>Updated road detection from 2019 onward</strong>, using AI and both optical and radar-based satellite imagery to map transient forest roads that often go undetected in conventional road data sets, with roads attributed by detection year and viewable by time range on GFW.</li>



<li><strong>A long-term historical archive</strong>, combining the new data with earlier road maps created from satellite imagery going from 2018 back to 1976.</li>



<li><strong>Improved coverage and consistency</strong>, offering users a single, integrated data set rather than multiple sources.</li>
</ul>



<p>Prior to this change, GFW hosted a Congo Basin logging roads data set that was last updated in 2014. That data set was produced through a crowdsourcing effort from the Logging Roads initiative led by Moabi, GFW and the European Commission’s Joint Research Centre. Volunteers used satellite imagery and OpenStreetMap-based tools to manually trace roads and estimate when they appeared. While groundbreaking at the time, the data set covered a limited time period and was difficult to update.</p>



<p>In subsequent years, researchers <a href="https://www.nature.com/articles/s41893-019-0310-6" target="_blank" rel="noreferrer noopener">manually updated</a> the data set through 2018 and integrated it with an earlier road inventory that had <a href="https://www.science.org/doi/10.1126/science.1141057" target="_blank" rel="noreferrer noopener">digitized roads</a> from more than 300 Landsat images collected between 1976 and 2003, creating a comprehensive archive of roads data.</p>



<p>The new data combines the hand-digitized data with a new AI-based approach to offer a more complete and current picture of forest road expansion across the region.</p>



<p>Users can view the Congo Basin forest roads data on the GFW map by selecting the <a href="https://www.globalforestwatch.org/map/?map=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%3D&amp;mapMenu=eyJtZW51U2VjdGlvbiI6ImRhdGFzZXRzIiwiZGF0YXNldENhdGVnb3J5IjoibGFuZFVzZSJ9">Congo Basin forest roads layer</a> in the Land Use data sets section under Infrastructure.</p>



<figure class="wp-block-image size-full"><img src="https://content.globalforestwatch.org/wp-content/uploads/2026/03/congobasinroads.png" alt="Screenshot of Forest roads in the Congo Basin (1976 to 2024)" class="wp-image-20079"/><figcaption class="wp-element-caption"><em>Forest roads in the Congo Basin (1976 to 2024), with older data shown in darker brown and newly detected roads highlighted in bright gold.</em></figcaption></figure>



<p>Additionally, the data will be integrated into national <a href="https://www.wri.org/initiatives/forest-atlases" target="_blank" rel="noreferrer noopener">Forest Atlases</a>, which combine government information with independent monitoring tools to support transparent forest management.&nbsp;</p>



<h2 class="wp-block-heading"><strong>How was the data created?</strong></h2>



<p>We mapped roads from 2019 onward using an advanced deep learning model <a href="https://www.wur.nl/en/news/road-networks-stretch-congo-basin-forests-satellites-and-ai-help-monitor" target="_blank" rel="noreferrer noopener">developed by Wageningen University &amp; Research and The Nature Conservancy</a>, based on both Sentinel-1 and Sentinel-2 satellite imagery. The imagery uses radar sensors that can “see through” clouds, which allows for road detections even during the Congo Basin’s frequent rainy seasons. The automated algorithm processed thousands of satellite images and was designed to be continuously updated.</p>



<p>To extend the data set further back in time, the new road data was merged with <a href="https://www.nature.com/articles/s41893-019-0310-6" target="_blank" rel="noreferrer noopener">historical maps</a> that include the <a href="https://www.science.org/doi/10.1126/science.1141057" target="_blank" rel="noreferrer noopener">manually digitized roads</a> from Landsat imagery covering 1976–2003 and crowdsourced mapping efforts. While detection methods differ, together they provide a valuable long-term archive of forest access and fragmentation.</p>



<p>As with most large-scale data sets, there are important caveats that users should be aware of:</p>



<ul>
<li><strong>Not all forest roads are associated with selective logging</strong>, although many mapped roads are linked to logging activities. The <a href="https://gfw.global/OQbmTK">logging concessions layer</a> on the GFW map can provide further context.</li>



<li>The data set does <strong>not indicate the legality</strong> of roads or associated activities.</li>



<li><strong>Road widths are not shown to scale</strong>, and the data does <strong>not show whether a road is still in use</strong> (for example, some roads may have overgrown since they were detected).</li>



<li>Additionally, overlap between older and newer roads may indicate the reopening of older roads. This may be of interest to users investigating selective logging activities because a road reopening can indicate that a new harvest cycle has started at a location where old, overgrown logging roads are being reopened and reused (which generally has <a href="https://besjournals.onlinelibrary.wiley.com/doi/epdf/10.1111/1365-2664.12661" target="_blank" rel="noreferrer noopener">less environmental impact</a> than constructing entirely new ones). However, note that <strong>reopened roads may not perfectly align with new roads</strong> on the map, since new roads (2019 onward) were mapped with more spatial precision than old roads.</li>
</ul>



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



<p>The updated forest roads data set supports a wide range of applications:</p>



<ul>
<li><strong>Forest monitoring and law enforcement:</strong> Users can track where new roads appear inside or outside <a href="https://gfw.global/OQbmTK">logging concessions</a>, helping to flag potential illegal activity and support patrol planning.</li>



<li><strong>Concession oversight:</strong> Governments and civil society organizations can assess whether road development aligns with approved logging plans and schedules. Detailed logging road maps can also be used to evaluate and potentially reduce the impact of selective logging activities.</li>



<li><strong>Research and analysis:</strong> Researchers can use the long-term archive to study logging trends, infrastructure persistence and the carbon implications of new versus reopened roads, as well as the relationship between roads, degradation and deforestation.</li>



<li><strong>Conservation and land-use planning:</strong> By revealing patterns of access and fragmentation, the data set can inform broader efforts to protect intact forest landscapes and protected areas.</li>
</ul>



<figure class="wp-block-image size-full"><img src="https://content.globalforestwatch.org/wp-content/uploads/2026/03/congobasinroads_v2.png" alt="Screenshot of Forest roads in the Congo Basin in Republic of Congo" class="wp-image-20080"/><figcaption class="wp-element-caption"><em>The forest roads data set provides context to forest change data such as near-real-time disturbance alerts, informing users what may be driving forest change. For example, recent disturbance alerts occurring near forest roads in intact forest landscapes highlight areas where access infrastructure and potential logging operations may be contributing to the reduction of previously intact forest.</em></figcaption></figure>



<h2 class="wp-block-heading"><strong>What’s next for forest road monitoring data?</strong></h2>



<p>By combining cutting-edge AI with decades of historical mapping, this new forest roads data set offers an unprecedented view of how access has shaped the Congo Basin’s forests over time. As road expansion continues to impact forests, having timely, transparent and accessible data is essential for effective monitoring and action.</p>



<p>Moving forward, we plan to update the data set annually to support continued monitoring of road expansion across the region. In the future, we may apply these methods to map roads in other regions such as the Amazon and Southeast Asian rainforests.</p>



<hr class="wp-block-separator has-alpha-channel-opacity"/>



<p><em>Access the data: The data set is </em><a href="https://gfw.global/kkkquy"><em>now available on GFW</em></a><em> as well as the </em><a href="https://www.wri.org/initiatives/forest-atlases" target="_blank" rel="noreferrer noopener"><em>Forest Atlases</em></a><em> where it <strong>complements existing forest disturbance data</strong>, such as </em><a href="https://gfw.global/qfkEaG"><em>drivers of disturbance alerts</em></a><em> and </em><a href="https://gfw.global/RIZqAn"><em>tree cover loss by dominant driver</em></a><em>, by<strong> adding spatial detail on logging-related access and infrastructure</strong>, helping users better understand forest change in the Congo Basin.</em></p>



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

In Congo Basin forests, roads can cause big changes. A new, high-res data set on GFW provides a detailed view of these forest roads.

Mapping Forest Roads in the Congo Basin Powered by AI

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

Integrated disturbance alerts layer on GFW

Deforestation Alerts

Deforestation Alerts | Global Forest Watch Blog

Teresa Schofield

Amy Pickens

Johannes Reiche

<p>Global Forest Watch (GFW) provides near-real-time information on where vegetation disturbances are happening around the world. The <a href="https://gfw.global/3CrDkAs"><strong>integrated disturbance alerts layer</strong></a> — originally introduced in 2022 and previously called “integrated deforestation alerts” — brings together four independent alert systems, optical and radar-based, to offer the earliest, most confident and most frequent signal of vegetation loss available on the GFW platform. In addition to GLAD-L, GLAD-S2 and RADD alerts, the integrated disturbance alerts layer now includes global vegetation disturbance alerts (DIST-ALERT), which expands the alert coverage to all vegetation both inside and outside of forests globally and detects repeated disturbances. By combining these data sources, users can monitor deforestation and other disturbances more effectively than by relying on any one system alone.</p>



<p>This layer is designed for users who need fast, reliable information for monitoring forests and other ecosystems, responding to illegal activity or prioritizing areas for follow-up.</p>



<h2 class="wp-block-heading"><strong>Which alert systems are included in the integrated disturbance alerts layer?</strong></h2>



<p>Previously, the integrated alerts layer showed disturbances only within tropical forests. The addition of the 30-meter resolution <a href="https://landcarbonlab.org/insights/new-monitoring-system-land-ecosystem-disturbances/" target="_blank" rel="noreferrer noopener"><strong>global vegetation disturbance alerts (DIST-ALERT)</strong></a> system from the University of Maryland (UMD) and NASA expands monitoring beyond forests to all vegetation types — capturing all vegetation disturbances, not only in tree cover — and brings global coverage to the integrated alerts layer for the first time.</p>



<p>With the addition of DIST-ALERT, the integrated alerts layer now includes four alert systems: GLAD-L, GLAD-S2, RADD and DIST-ALERT.</p>



<p><a href="https://www.globalforestwatch.org/blog/data-and-research/glad-deforestation-alerts/">GLAD-L</a> is GFW’s longest-standing alert product from the UMD Global Land Analysis and Discovery (GLAD) lab and uses imagery from NASA’s Landsat satellites to detect tree cover loss at 30-meter resolution.</p>



<p>UMD introduced the higher resolution (10-meter) <a href="https://www.globalforestwatch.org/blog/data-and-research/glad-s2-offers-high-resolution-deforestation-alerts/">GLAD-S2</a> alerts in May 2021, which use a similar methodology to GLAD-L but are sourced from the European Space Agency’s Sentinel-2 satellites. Radar for Detecting Deforestation (<a href="https://www.globalforestwatch.org/blog/data-and-research/radd-radar-alerts/">RADD</a>) alerts produced by Wageningen University use 10-meter Sentinel-1 radar data. These alerts have the additional capability of detecting forest change through cloud cover that often blocks the view of the other satellites.</p>



<p>One of the greatest strengths of higher resolution (10-meter) RADD and GLAD-S2 alerts is their ability to capture smaller changes in the forest than the 30-meter products. This information has been vital for detecting small canopy gaps from logging activities.</p>



<p>We integrated DIST-ALERT in January 2026, which combines Landsat and Sentinel-2 data to cover disturbances both inside forests as well as in other vegetation. Users who want to focus exclusively on forests can still toggle the option on the map legend to display only alerts within areas of tree cover. This flexibility allows the integrated alerts layer to support both monitoring forests specifically, as well as the broader landscape.</p>



<div class="flourish-embed flourish-table" data-src="visualisation/26576189?163031"><script src="https://public.flourish.studio/resources/embed.js"></script><noscript><img decoding="async" src="https://public.flourish.studio/visualisation/26576189/thumbnail" width="100%" alt="table visualization" /></noscript></div>



<h2 class="wp-block-heading"><strong>What does the integrated alerts layer offer?</strong></h2>



<figure class="wp-block-image size-full"><img decoding="async" loading="lazy" width="1575" height="1079" src="https://content.globalforestwatch.org/wp-content/uploads/2026/01/alert_extent.png" alt="" class="wp-image-19907" srcset="https://content.globalforestwatch.org/wp-content/uploads/2026/01/alert_extent.png 1575w, https://content.globalforestwatch.org/wp-content/uploads/2026/01/alert_extent-1024x702.png 1024w, https://content.globalforestwatch.org/wp-content/uploads/2026/01/alert_extent-768x526.png 768w, https://content.globalforestwatch.org/wp-content/uploads/2026/01/alert_extent-1536x1052.png 1536w" sizes="(max-width: 1575px) 100vw, 1575px" /><figcaption class="wp-element-caption"><em>Map shows the geographic coverage of the integrated disturbance alerts, with brighter pink representing where multiple alert systems are available within tree cover.</em></figcaption></figure>



<p>These systems use satellite imagery of different spatial resolution and are updated at varying frequencies. The integrated option (resampled to 10-meter resolution) enables users to derive meaningful summary statistics that incorporate each of the individual systems. To achieve this, the layer indicates where multiple systems have reported an alert in any given location and displays it through confidence levels:</p>



<figure class="wp-block-image size-full"><img decoding="async" loading="lazy" width="1575" height="716" src="https://content.globalforestwatch.org/wp-content/uploads/2026/01/alert_confidence.png" alt="" class="wp-image-19908" srcset="https://content.globalforestwatch.org/wp-content/uploads/2026/01/alert_confidence.png 1575w, https://content.globalforestwatch.org/wp-content/uploads/2026/01/alert_confidence-1024x466.png 1024w, https://content.globalforestwatch.org/wp-content/uploads/2026/01/alert_confidence-768x349.png 768w, https://content.globalforestwatch.org/wp-content/uploads/2026/01/alert_confidence-1536x698.png 1536w" sizes="(max-width: 1575px) 100vw, 1575px" /><figcaption class="wp-element-caption"><em>Note that the low confidence category is not currently available for DIST-ALERT.</em></figcaption></figure>



<p>Each individual alert system has its own method of determining confidence based on the number of satellite observations that register a loss of vegetation or tree cover. Across all systems, an alert begins as low confidence when the first signs of potential disturbance appear and is upgraded to high confidence only after additional satellite observations provide further evidence or probability of loss. Highest confidence is only reached when multiple alerts systems detect a disturbance in the same pixel location. To learn more about what the confidence levels mean, <a href="https://gfw.global/3CrDkAs">click the “Layer info” icon here</a>.</p>



<p>In an important change from the previous versions of the layer, the addition of DIST-ALERT means the integrated alerts layer now detects repeated disturbances. In contrast to many other alert systems available on GFW, DIST-ALERT continues to monitor pixels where it has identified an alert in the past. The DIST-ALERT layer on GFW shows the date of the most recent alert, keeping users informed of the most up-to-date losses of forest and vegetation cover, meaning repeat alerts are captured in the integrated alerts layer for the first time. Alerts from multiple systems that occur within 180 days (six months) of each other are considered a single event and reach highest confidence. In this case, the earliest date within that 180-day window will appear as the alert date in the integrated alerts layer. The date of the alert remains until another new alert outside of the six-month timeframe is registered in the same pixel.</p>



<p>Capturing repeat alerts means that areas where tree cover has regrown (such as areas where restoration has taken place) can be monitored for further disturbances, and areas that increasingly face recurring disturbances like fire can also be tracked.</p>



<h2 class="wp-block-heading"><strong>What are the benefits of combining alerts?</strong></h2>



<strong>1. Users will always have the earliest information available on new alerts</strong>



<p>By combining&nbsp;multiple&nbsp;alert systems into one layer, we are harnessing&nbsp;the&nbsp;different&nbsp;timing of&nbsp;the&nbsp;satellites to increase the chances of getting&nbsp;an unobscured&nbsp;look&nbsp;at tropical forest and therefore detect change faster, as well as inform users of vegetation loss anywhere on the planet. <a href="https://iopscience.iop.org/article/10.1088/1748-9326/ad2d82#erlad2d82s4"﷟HYPERLINK "https://iopscience.iop.org/article/10.1088/1748-9326/ad2d82" target="_blank" rel="noreferrer noopener">A study on alerts</a> found that integrating alert systems leads to quicker detection of new disturbances, sometimes by days or even months. For example, users&nbsp;looking at integrated alerts from 2021 in the Amazon basin&nbsp;were alerted&nbsp;to forest loss&nbsp;on average nine days earlier than by referring to any one system alone. This improvement in timeliness varies by region and across seasons, as the alert systems complement each other by providing their own strengths. For example, in consistently cloud-covered areas, the radar-based alert system provides an advantage of detecting&nbsp;forest loss&nbsp;through clouds. However, the higher revisit time of the optical alert systems gives a more frequent opportunity to detect disturbances, every one-four days, particularly in areas with a dry season; whereas the radar system&nbsp;may take up to 12 days to detect change. This research was done with just three systems integrated, so we expect even more improved results with a fourth system integrated.</p>



<p><strong>2. The overlap of multiple systems increases confidence to prioritize alerts</strong></p>



<p>Confidence levels help forest monitors prioritize alerts for follow up since satellite-derived data is subject to errors including false alerts. If two or more alert systems detect a change in the same location, we are more confident (“highest confidence”) that these alerts are real vegetation disturbances, rather than errors in the data (known as false positives). For individual systems, there is a delay before a first detection can be verified by additional satellite passes and thus reach “high confidence.” The integrated layer displays where multiple systems overlap, in some cases providing increased confidence faster than waiting for individual systems to reach high confidence through additional satellite images, which can take weeks or months. False positives are effectively eliminated in the highest confidence class as it’s uncommon for two systems to commit the same error since they use different data streams and algorithms.&nbsp;</p>



<p><strong>3. Different loss types are captured across a variety of landscapes </strong></p>



<p>Because of DIST-ALERT, the integrated layer covers not only forests globally but also disturbances in other ecosystems such as grasslands, savannas, shrublands, wetlands and crops. Users who are interested only in forest disturbances can use the tree cover filter to show only alerts within tree cover (defined as all vegetation greater than <a href="https://www.globalforestwatch.org/map/?map=eyJjZW50ZXIiOnsibGF0IjoxNi45MjQ1NzkwMDE2NTkzOSwibG5nIjotMTUuNzE3NTU5NzU4MjQ4NjE5fSwiZGF0YXNldHMiOlt7ImRhdGFzZXQiOiJ1bWQtdHJlZS1oZWlnaHQiLCJvcGFjaXR5IjoxLCJ2aXNpYmlsaXR5Ijp0cnVlLCJsYXllcnMiOlsidW1kLXRyZWUtaGVpZ2h0LTIwMjAiXSwicGFyYW1zIjp7ImhlaWdodCI6IjUifX0seyJkYXRhc2V0IjoicG9saXRpY2FsLWJvdW5kYXJpZXMiLCJsYXllcnMiOlsiZGlzcHV0ZWQtcG9saXRpY2FsLWJvdW5kYXJpZXMiLCJwb2xpdGljYWwtYm91bmRhcmllcyJdLCJvcGFjaXR5IjoxLCJ2aXNpYmlsaXR5Ijp0cnVlfV19&mapMenu=eyJkYXRhc2V0Q2F0ZWdvcnkiOiJsYW5kQ292ZXIifQ%3D%3D&modalMeta=umd_tree_cover_height_2020">5 meters in height (2020)</a> with greater than <a href="https://www.globalforestwatch.org/map/?map=eyJjZW50ZXIiOnsibGF0IjoxNi45MjQ1NzkwMDE2NTkzOSwibG5nIjotMTUuNzE3NTU5NzU4MjQ4NjE5fSwiZGF0YXNldHMiOlt7ImRhdGFzZXQiOiJwb2xpdGljYWwtYm91bmRhcmllcyIsImxheWVycyI6WyJkaXNwdXRlZC1wb2xpdGljYWwtYm91bmRhcmllcyIsInBvbGl0aWNhbC1ib3VuZGFyaWVzIl0sIm9wYWNpdHkiOjEsInZpc2liaWxpdHkiOnRydWV9LHsiZGF0YXNldCI6InRyZWUtY292ZXIiLCJsYXllcnMiOlsidHJlZS1jb3Zlci0yMDEwIl0sIm9wYWNpdHkiOjEsInZpc2liaWxpdHkiOnRydWV9XX0%3D&mapMenu=eyJkYXRhc2V0Q2F0ZWdvcnkiOiJsYW5kQ292ZXIifQ%3D%3D&modalMeta=umd_tree_cover_density_2000">30% canopy cover (2010)</a> and where no <a href="https://www.globalforestwatch.org/map/?map=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%3D&mapMenu=eyJtZW51U2VjdGlvbiI6ImRhdGFzZXRzIiwiZGF0YXNldENhdGVnb3J5IjoiZm9yZXN0Q2hhbmdlIn0%3D&modalMeta=umd_tree_cover_loss">tree cover loss</a> has been recorded since 2021, and may take the form of natural forests or plantations).</p>



<figure class="wp-block-image size-full"><img decoding="async" loading="lazy" width="1310" height="788" src="https://content.globalforestwatch.org/wp-content/uploads/2026/01/alerts_screenshot_wildfires.png" alt="" class="wp-image-19909" srcset="https://content.globalforestwatch.org/wp-content/uploads/2026/01/alerts_screenshot_wildfires.png 1310w, https://content.globalforestwatch.org/wp-content/uploads/2026/01/alerts_screenshot_wildfires-1024x616.png 1024w, https://content.globalforestwatch.org/wp-content/uploads/2026/01/alerts_screenshot_wildfires-768x462.png 768w" sizes="(max-width: 1310px) 100vw, 1310px" /><figcaption class="wp-element-caption"><em>Now that alerts are available globally, they can capture disturbances beyond the tropics, such as&nbsp;wildfire activity&nbsp;in Alberta, Canada.</em><strong>&nbsp;</strong></figcaption></figure>



<p>In the tropics, where multiple alert systems are available, the alert systems can be used in concert to identify illegal activities quickly and with increased confidence, a critical tool for both law enforcement and Indigenous Peoples and local communities working to combat deforestation.</p>



<figure class="wp-block-image size-full"><img decoding="async" loading="lazy" width="1310" height="788" src="https://content.globalforestwatch.org/wp-content/uploads/2026/01/alerts_screenshot.png" alt="" class="wp-image-19910" srcset="https://content.globalforestwatch.org/wp-content/uploads/2026/01/alerts_screenshot.png 1310w, https://content.globalforestwatch.org/wp-content/uploads/2026/01/alerts_screenshot-1024x616.png 1024w, https://content.globalforestwatch.org/wp-content/uploads/2026/01/alerts_screenshot-768x462.png 768w" sizes="(max-width: 1310px) 100vw, 1310px" /><figcaption class="wp-element-caption"><em>The integrated disturbance alerts layer demonstrates the ability of multiple alert systems to detect small changes, such as tree removal along logging roads in the Peruvian Amazon.</em></figcaption></figure>



<p>Note that the benefits of increased confidence and more rapid detection resulting from multiple monitoring systems are still limited to tropical forests, as areas outside of forests and outside of the tropics are only captured by one system (DIST-ALERT).</p>



<h2 class="wp-block-heading"><strong>What’s next?</strong></h2>



<p>Information on <a href="https://gfw.global/Jioe3c">drivers of alerts is now available on GFW</a> for the tropics, and is based on the integrated alert layer. This information supports users in prioritizing alerts by distinguishing natural from human-caused disturbances since users may have more interest in investigating human-caused disturbances, such as to assess the legality of forest clearing. In the future, this layer will be updated monthly, based on the last month’s integrated alerts, giving users a quicker picture of the cause of alerts. Stay tuned for future updates to the integrated alert layer and other advancements in the field of forest monitoring.</p>



<h2 class="wp-block-heading"><strong>Get started with integrated alerts</strong> </h2>



<p>With several alert systems to choose from, we aim to simplify users’ workflows in identifying priority alerts. The integrated layer provides the most information on recent forest disturbances for users who want to be notified of forest loss with the highest confidence and provides up-to-date information on other vegetation losses. Though we see these systems as complimentary to each other, users may still opt to select individual systems for alert notifications and analysis if these suit their needs.</p>



<p><br><strong><em>For an in-depth look at how the integrated layer was designed and how to use it, see the </em></strong><a href="https://www.globalforestwatch.org/help/forest-watcher/webinars/visualizing-analyzing-integrated-deforestation-alerts/"><strong><em>Help Center</em></strong></a><strong><em>.</em></strong></p>



<hr> 

<div class="attribute" style="”padding-bottom:" 10px;”=""> 
Amy Pickens is an assistant research professor at University of Maryland and operates the GLAD alert systems.<br><br>

Johannes Reiche is an associate professor at Wageningen University and operates the RADD alert system.<br><br>

This piece was originally published in 2022. It was updated in January 2026 to reflect the latest developments with the integrated alerts. 
</div>



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

Looking for the Quickest Signal of Deforestation or Vegetation Disturbance Globally? Turn to GFW’s Integrated Disturbance Alerts

<p>Global Forest Watch (GFW) provides near-real-time information on where vegetation disturbances are happening around the world. The <a href="https://gfw.global/3CrDkAs"><strong>integrated disturbance alerts layer</strong></a> — originally introduced in 2022 and previously called “integrated deforestation alerts” — brings together four independent alert systems, optical and radar-based, to offer the earliest, most confident and most frequent signal of vegetation loss available on the GFW platform. In addition to GLAD-L, GLAD-S2 and RADD alerts, the integrated disturbance alerts layer now includes global vegetation disturbance alerts (DIST-ALERT), which expands the alert coverage to all vegetation both inside and outside of forests globally and detects repeated disturbances. By combining these data sources, users can monitor deforestation and other disturbances more effectively than by relying on any one system alone.</p>



<p>This layer is designed for users who need fast, reliable information for monitoring forests and other ecosystems, responding to illegal activity or prioritizing areas for follow-up.</p>



<h2 class="wp-block-heading"><strong>Which alert systems are included in the integrated disturbance alerts layer?</strong></h2>



<p>Previously, the integrated alerts layer showed disturbances only within tropical forests. The addition of the 30-meter resolution <a href="https://landcarbonlab.org/insights/new-monitoring-system-land-ecosystem-disturbances/" target="_blank" rel="noreferrer noopener"><strong>global vegetation disturbance alerts (DIST-ALERT)</strong></a> system from the University of Maryland (UMD) and NASA expands monitoring beyond forests to all vegetation types — capturing all vegetation disturbances, not only in tree cover — and brings global coverage to the integrated alerts layer for the first time.</p>



<p>With the addition of DIST-ALERT, the integrated alerts layer now includes four alert systems: GLAD-L, GLAD-S2, RADD and DIST-ALERT.</p>



<p><a href="https://www.globalforestwatch.org/blog/data-and-research/glad-deforestation-alerts/">GLAD-L</a> is GFW’s longest-standing alert product from the UMD Global Land Analysis and Discovery (GLAD) lab and uses imagery from NASA's Landsat satellites to detect tree cover loss at 30-meter resolution.</p>



<p>UMD introduced the higher resolution (10-meter) <a href="https://www.globalforestwatch.org/blog/data-and-research/glad-s2-offers-high-resolution-deforestation-alerts/">GLAD-S2</a> alerts in May 2021, which use a similar methodology to GLAD-L but are sourced from the European Space Agency's Sentinel-2 satellites. Radar for Detecting Deforestation (<a href="https://www.globalforestwatch.org/blog/data-and-research/radd-radar-alerts/">RADD</a>) alerts produced by Wageningen University use 10-meter Sentinel-1 radar data. These alerts have the additional capability of detecting forest change through cloud cover that often blocks the view of the other satellites.</p>



<p>One of the greatest strengths of higher resolution (10-meter) RADD and GLAD-S2 alerts is their ability to capture smaller changes in the forest than the 30-meter products. This information has been vital for detecting small canopy gaps from logging activities.</p>



<p>We integrated DIST-ALERT in January 2026, which combines Landsat and Sentinel-2 data to cover disturbances both inside forests as well as in other vegetation. Users who want to focus exclusively on forests can still toggle the option on the map legend to display only alerts within areas of tree cover. This flexibility allows the integrated alerts layer to support both monitoring forests specifically, as well as the broader landscape.</p>



<div class="flourish-embed flourish-table" data-src="visualisation/26576189?163031"><script src="https://public.flourish.studio/resources/embed.js"></script><noscript><img src="https://public.flourish.studio/visualisation/26576189/thumbnail" width="100%" alt="table visualization" /></noscript></div>



<h2 class="wp-block-heading"><strong>What does the integrated alerts layer offer?</strong></h2>



<figure class="wp-block-image size-full"><img src="https://content.globalforestwatch.org/wp-content/uploads/2026/01/alert_extent.png" alt="" class="wp-image-19907"/><figcaption class="wp-element-caption"><em>Map shows the geographic coverage of the integrated disturbance alerts, with brighter pink representing where multiple alert systems are available within tree cover.</em></figcaption></figure>



<p>These systems use satellite imagery of different spatial resolution and are updated at varying frequencies. The integrated option (resampled to 10-meter resolution) enables users to derive meaningful summary statistics that incorporate each of the individual systems. To achieve this, the layer indicates where multiple systems have reported an alert in any given location and displays it through confidence levels:</p>



<figure class="wp-block-image size-full"><img src="https://content.globalforestwatch.org/wp-content/uploads/2026/01/alert_confidence.png" alt="" class="wp-image-19908"/><figcaption class="wp-element-caption"><em>Note that the low confidence category is not currently available for DIST-ALERT.</em></figcaption></figure>



<p>Each individual alert system has its own method of determining confidence based on the number of satellite observations that register a loss of vegetation or tree cover. Across all systems, an alert begins as low confidence when the first signs of potential disturbance appear and is upgraded to high confidence only after additional satellite observations provide further evidence or probability of loss. Highest confidence is only reached when multiple alerts systems detect a disturbance in the same pixel location. To learn more about what the confidence levels mean, <a href="https://gfw.global/3CrDkAs">click the “Layer info” icon here</a>.</p>



<p>In an important change from the previous versions of the layer, the addition of DIST-ALERT means the integrated alerts layer now detects repeated disturbances. In contrast to many other alert systems available on GFW, DIST-ALERT continues to monitor pixels where it has identified an alert in the past. The DIST-ALERT layer on GFW shows the date of the most recent alert, keeping users informed of the most up-to-date losses of forest and vegetation cover, meaning repeat alerts are captured in the integrated alerts layer for the first time. Alerts from multiple systems that occur within 180 days (six months) of each other are considered a single event and reach highest confidence. In this case, the earliest date within that 180-day window will appear as the alert date in the integrated alerts layer. The date of the alert remains until another new alert outside of the six-month timeframe is registered in the same pixel.</p>



<p>Capturing repeat alerts means that areas where tree cover has regrown (such as areas where restoration has taken place) can be monitored for further disturbances, and areas that increasingly face recurring disturbances like fire can also be tracked.</p>



<h2 class="wp-block-heading"><strong>What are the benefits of combining alerts?</strong></h2>



<strong>1. Users will always have the earliest information available on new alerts</strong>



<p>By combining&nbsp;multiple&nbsp;alert systems into one layer, we are harnessing&nbsp;the&nbsp;different&nbsp;timing of&nbsp;the&nbsp;satellites to increase the chances of getting&nbsp;an unobscured&nbsp;look&nbsp;at tropical forest and therefore detect change faster, as well as inform users of vegetation loss anywhere on the planet. <a href="https://iopscience.iop.org/article/10.1088/1748-9326/ad2d82#erlad2d82s4&quot;﷟HYPERLINK &quot;https://iopscience.iop.org/article/10.1088/1748-9326/ad2d82" target="_blank" rel="noreferrer noopener">A study on alerts</a> found that integrating alert systems leads to quicker detection of new disturbances, sometimes by days or even months. For example, users&nbsp;looking at integrated alerts from 2021 in the Amazon basin&nbsp;were alerted&nbsp;to forest loss&nbsp;on average nine days earlier than by referring to any one system alone. This improvement in timeliness varies by region and across seasons, as the alert systems complement each other by providing their own strengths. For example, in consistently cloud-covered areas, the radar-based alert system provides an advantage of detecting&nbsp;forest loss&nbsp;through clouds. However, the higher revisit time of the optical alert systems gives a more frequent opportunity to detect disturbances, every one-four days, particularly in areas with a dry season; whereas the radar system&nbsp;may take up to 12 days to detect change. This research was done with just three systems integrated, so we expect even more improved results with a fourth system integrated.</p>



<p><strong>2. The overlap of multiple systems increases confidence to prioritize alerts</strong></p>



<p>Confidence levels help forest monitors prioritize alerts for follow up since satellite-derived data is subject to errors including false alerts. If two or more alert systems detect a change in the same location, we are more confident (“highest confidence”) that these alerts are real vegetation disturbances, rather than errors in the data (known as false positives). For individual systems, there is a delay before a first detection can be verified by additional satellite passes and thus reach "high confidence." The integrated layer displays where multiple systems overlap, in some cases providing increased confidence faster than waiting for individual systems to reach high confidence through additional satellite images, which can take weeks or months. False positives are effectively eliminated in the highest confidence class as it's uncommon for two systems to commit the same error since they use different data streams and algorithms.&nbsp;</p>



<p><strong>3. Different loss types are captured across a variety of landscapes </strong></p>



<p>Because of DIST-ALERT, the integrated layer covers not only forests globally but also disturbances in other ecosystems such as grasslands, savannas, shrublands, wetlands and crops. Users who are interested only in forest disturbances can use the tree cover filter to show only alerts within tree cover (defined as all vegetation greater than <a href="https://www.globalforestwatch.org/map/?map=eyJjZW50ZXIiOnsibGF0IjoxNi45MjQ1NzkwMDE2NTkzOSwibG5nIjotMTUuNzE3NTU5NzU4MjQ4NjE5fSwiZGF0YXNldHMiOlt7ImRhdGFzZXQiOiJ1bWQtdHJlZS1oZWlnaHQiLCJvcGFjaXR5IjoxLCJ2aXNpYmlsaXR5Ijp0cnVlLCJsYXllcnMiOlsidW1kLXRyZWUtaGVpZ2h0LTIwMjAiXSwicGFyYW1zIjp7ImhlaWdodCI6IjUifX0seyJkYXRhc2V0IjoicG9saXRpY2FsLWJvdW5kYXJpZXMiLCJsYXllcnMiOlsiZGlzcHV0ZWQtcG9saXRpY2FsLWJvdW5kYXJpZXMiLCJwb2xpdGljYWwtYm91bmRhcmllcyJdLCJvcGFjaXR5IjoxLCJ2aXNpYmlsaXR5Ijp0cnVlfV19&amp;mapMenu=eyJkYXRhc2V0Q2F0ZWdvcnkiOiJsYW5kQ292ZXIifQ%3D%3D&amp;modalMeta=umd_tree_cover_height_2020">5 meters in height (2020)</a> with greater than <a href="https://www.globalforestwatch.org/map/?map=eyJjZW50ZXIiOnsibGF0IjoxNi45MjQ1NzkwMDE2NTkzOSwibG5nIjotMTUuNzE3NTU5NzU4MjQ4NjE5fSwiZGF0YXNldHMiOlt7ImRhdGFzZXQiOiJwb2xpdGljYWwtYm91bmRhcmllcyIsImxheWVycyI6WyJkaXNwdXRlZC1wb2xpdGljYWwtYm91bmRhcmllcyIsInBvbGl0aWNhbC1ib3VuZGFyaWVzIl0sIm9wYWNpdHkiOjEsInZpc2liaWxpdHkiOnRydWV9LHsiZGF0YXNldCI6InRyZWUtY292ZXIiLCJsYXllcnMiOlsidHJlZS1jb3Zlci0yMDEwIl0sIm9wYWNpdHkiOjEsInZpc2liaWxpdHkiOnRydWV9XX0%3D&amp;mapMenu=eyJkYXRhc2V0Q2F0ZWdvcnkiOiJsYW5kQ292ZXIifQ%3D%3D&amp;modalMeta=umd_tree_cover_density_2000">30% canopy cover (2010)</a> and where no <a href="https://www.globalforestwatch.org/map/?map=eyJjZW50ZXIiOnsibGF0IjoxNi45MjQ1NzkwMDE2NTkzOSwibG5nIjotMTUuNzE3NTU5NzU4MjQ4NjE5fSwiZGF0YXNldHMiOlt7ImRhdGFzZXQiOiJ0cmVlLWNvdmVyLWxvc3MiLCJvcGFjaXR5IjoxLCJ2aXNpYmlsaXR5Ijp0cnVlLCJsYXllcnMiOlsidHJlZS1jb3Zlci1sb3NzIl0sInRpbWVsaW5lUGFyYW1zIjp7InN0YXJ0RGF0ZSI6IjIwMjEtMDEtMDEiLCJlbmREYXRlIjoiMjAyNC0xMi0zMSIsInRyaW1FbmREYXRlIjoiMjAyNC0xMi0zMSJ9fSx7ImRhdGFzZXQiOiJwb2xpdGljYWwtYm91bmRhcmllcyIsImxheWVycyI6WyJkaXNwdXRlZC1wb2xpdGljYWwtYm91bmRhcmllcyIsInBvbGl0aWNhbC1ib3VuZGFyaWVzIl0sIm9wYWNpdHkiOjEsInZpc2liaWxpdHkiOnRydWV9XX0%3D&amp;mapMenu=eyJtZW51U2VjdGlvbiI6ImRhdGFzZXRzIiwiZGF0YXNldENhdGVnb3J5IjoiZm9yZXN0Q2hhbmdlIn0%3D&amp;modalMeta=umd_tree_cover_loss">tree cover loss</a> has been recorded since 2021, and may take the form of natural forests or plantations).</p>



<figure class="wp-block-image size-full"><img src="https://content.globalforestwatch.org/wp-content/uploads/2026/01/alerts_screenshot_wildfires.png" alt="" class="wp-image-19909"/><figcaption class="wp-element-caption"><em>Now that alerts are available globally, they can capture disturbances beyond the tropics, such as&nbsp;wildfire activity&nbsp;in Alberta, Canada.</em><strong>&nbsp;</strong></figcaption></figure>



<p>In the tropics, where multiple alert systems are available, the alert systems can be used in concert to identify illegal activities quickly and with increased confidence, a critical tool for both law enforcement and Indigenous Peoples and local communities working to combat deforestation.</p>



<figure class="wp-block-image size-full"><img src="https://content.globalforestwatch.org/wp-content/uploads/2026/01/alerts_screenshot.png" alt="" class="wp-image-19910"/><figcaption class="wp-element-caption"><em>The integrated disturbance alerts layer demonstrates the ability of multiple alert systems to detect small changes, such as tree removal along logging roads in the Peruvian Amazon.</em></figcaption></figure>



<p>Note that the benefits of increased confidence and more rapid detection resulting from multiple monitoring systems are still limited to tropical forests, as areas outside of forests and outside of the tropics are only captured by one system (DIST-ALERT).</p>



<h2 class="wp-block-heading"><strong>What's next?</strong></h2>



<p>Information on <a href="https://gfw.global/Jioe3c">drivers of alerts is now available on GFW</a> for the tropics, and is based on the integrated alert layer. This information supports users in prioritizing alerts by distinguishing natural from human-caused disturbances since users may have more interest in investigating human-caused disturbances, such as to assess the legality of forest clearing. In the future, this layer will be updated monthly, based on the last month's integrated alerts, giving users a quicker picture of the cause of alerts. Stay tuned for future updates to the integrated alert layer and other advancements in the field of forest monitoring.</p>



<h2 class="wp-block-heading"><strong>Get started with integrated alerts</strong> </h2>



<p>With several alert systems to choose from, we aim to simplify users’ workflows in identifying priority alerts. The integrated layer provides the most information on recent forest disturbances for users who want to be notified of forest loss with the highest confidence and provides up-to-date information on other vegetation losses. Though we see these systems as complimentary to each other, users may still opt to select individual systems for alert notifications and analysis if these suit their needs.</p>



<p><br><strong><em>For an in-depth look at how the integrated layer was designed and how to use it, see the </em></strong><a href="https://www.globalforestwatch.org/help/forest-watcher/webinars/visualizing-analyzing-integrated-deforestation-alerts/"><strong><em>Help Center</em></strong></a><strong><em>.</em></strong></p>



<hr> 

<div class="attribute" style="”padding-bottom:" 10px;”=""> 
Amy Pickens is an assistant research professor at University of Maryland and operates the GLAD alert systems.<br><br>

Johannes Reiche is an associate professor at Wageningen University and operates the RADD alert system.<br><br>

This piece was originally published in 2022. It was updated in January 2026 to reflect the latest developments with the integrated alerts. 
</div>



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

GFW's integrated alerts layer combines the analytical power of four alert systems for an improved view deforestation and other vegetation disturbances globally.

Explore GFW's Updated Integrated Disturbance Alerts

deforestation

deforestation | Global Forest Watch Blog

forest change

forest change | Global Forest Watch Blog

Nancy Harris

Crystal Davis

Jan Hooijmans

<p>Today, we have more data about forests than ever before, but we still can’t seem to agree on where, when and why forests are changing around the world. Even two prominent global data sources appear to disagree, at least on the surface. “Deforestation continues to decline” was <a href="https://openknowledge.fao.org/server/api/core/bitstreams/2dee6e93-1988-4659-aa89-30dd20b43b15/content/FRA-2025/key-findings.html" target="_blank" rel="noreferrer noopener">a headline</a> of the <a href="https://www.fao.org/forest-resources-assessment/en/" target="_blank" rel="noreferrer noopener">2025 Global Forest Resources Assessment (FRA)</a> published by the United Nations Food and Agriculture Organization (FAO). At the same time, Global Forest Watch (GFW) <a href="https://gfr.wri.org/latest-analysis-deforestation-trends" target="_blank" rel="noreferrer noopener">reported that</a> “the tropics lost a record-shattering 6.7 million hectares of primary rainforest in 2024 … nearly double that of 2023.”</p>



<p>What gives? Unfortunately, the answer is not simple. Both GFW and FRA provide valuable information on the state of global forests, but they differ substantially in terms of purpose, scope and approach. By answering questions on how and why these two global data sources differ, we can begin to see them as complementary rather than contradictory. Ultimately, our knowledge of global forests remains incomplete, and we need all the information we can get.</p>



<div class="textboxright"><strong>WHAT ARE GLOBAL FOREST WATCH (GFW) AND THE GLOBAL FOREST RESOURCES ASSESSMENT (FRA)?</strong> <br><br>GFW is an independent online forest monitoring platform launched in 2014 by a consortium of partners led by World Resources Institute. GFW seeks to empower forest managers, law enforcement officers and other forest stakeholders with free access to timely and high-resolution data about the current status of forests and recent forest change. GFW uses satellite technology to monitor tree cover loss worldwide with weekly and annual updates, as well as to provide longer term tree cover gain and net change data. Gain, net change and annual loss data are GFW’s core data sets and are provided by the University of Maryland. Other data sets on GFW provide useful contextual information related to forests and forest change.<br><br>

The FRA, which the FAO has published at 5- to 10-year intervals since 1946, seeks to provide a consistent and comprehensive approach to describing the world’s forests and how they are changing. Countries provide reports to FRA based on data from National Forest Inventories, satellite data and extrapolation based on past data. The FAO combines the data submitted by countries into global reports and fills in any gaps where data is not available from the countries. The most recent edition of the FRA in 2025 included national-level statistics on around 50 broad ecological and socio-economic variables reported by 236 countries and territories. </div>



<h2 class="wp-block-heading">What is considered a “forest”?</h2>



<p>GFW and FRA both talk about “forests,” but what are they actually measuring and monitoring? There are nearly&nbsp;<a href="https://www.nrs.fs.usda.gov/pubs/gtr/gtr_wo077/gtr_wo077_011.pdf" target="_blank" rel="noreferrer noopener">800 definitions</a>&nbsp;of “forest” in use, which only adds to the confusion.</p>



<p>Every five years, FRA publishes forest area globally and per country. These area measurements are based on a <a href="https://openknowledge.fao.org/server/api/core/bitstreams/a6e225da-4a31-4e06-818d-ca3aeadfd635/content" target="_blank" rel="noreferrer noopener">standardized&nbsp;definition</a>&nbsp;of “forest” adopted by the FAO in 2000, which takes into account biophysical and land use criteria. The biophysical criterion establishes minimum thresholds for the height, canopy cover and extent of trees. The land use criterion requires that land be officially or legally designated for a “forest use,” such as conservation or harvest. 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. At the same time, 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.</p>



<p>GFW provides data on “tree cover,” which is derived from satellite data and therefore cannot discern intended land use. Thus, GFW’s monitoring systems rely entirely on biophysical criteria — related to height, canopy cover and extent of trees. GFW does not adopt a specific definition of forest, but rather monitors all forms of tree cover including natural forests and tree plantations. Using GFW’s interactive online platform, users can filter this data based on their preferred definition of forest. Users can select by canopy cover percent, as well as tree height. Users can also view and derive statistics using other data sets that give additional context on the land use, such as <a href="https://gfw.global/3jyrlta">primary forests</a>, <a href="http://gfw.global/4fQfl03">natural forests</a> and <a href="https://gfw.global/2Qbv7fM">planted trees</a>.</p>


<div class="wp-block-image">
<figure class="aligncenter size-large"><img decoding="async" loading="lazy" width="2048" height="1867" src="https://content.globalforestwatch.org/wp-content/uploads/2023/07/23_GFW_vs_FAO_graphics_What-is-considered-a-forest-1-2048x1867.png" alt="What is considered a "Forest" by GFW and FRA?" class="wp-image-16209" srcset="https://content.globalforestwatch.org/wp-content/uploads/2023/07/23_GFW_vs_FAO_graphics_What-is-considered-a-forest-1-2048x1867.png 2048w, https://content.globalforestwatch.org/wp-content/uploads/2023/07/23_GFW_vs_FAO_graphics_What-is-considered-a-forest-1-1024x934.png 1024w, https://content.globalforestwatch.org/wp-content/uploads/2023/07/23_GFW_vs_FAO_graphics_What-is-considered-a-forest-1-768x700.png 768w, https://content.globalforestwatch.org/wp-content/uploads/2023/07/23_GFW_vs_FAO_graphics_What-is-considered-a-forest-1-1536x1401.png 1536w" sizes="(max-width: 2048px) 100vw, 2048px" /></figure></div>


<h2 class="wp-block-heading">How is “forest change” reported?</h2>



<p>Both FRA and GFW aggregate monitoring data into global statistics about forest change. Like the different ways they measure forests, GFW and FRA apply different lenses to monitor <em>changes to</em> forests, which subsequently influence how both forest expansion and deforestation are reported. GFW reports annual gross loss of tree cover, as well as gross gain for four 5-year periods and net change for a 20-year period. FRA reports on deforestation, net change and forest expansion as separate statistics, though many countries only report forest area, from which net change can be derived.</p>



<h3 class="wp-block-heading">Deforestation and tree cover loss</h3>



<p>Using its land use criterion, the FAO’s definition of deforestation requires permanent land use change to occur. Losses of tree cover considered to be temporary (e.g., forests that have been clear cut but will be left to regrow) are not counted as deforestation.</p>



<p>Conversely, GFW detects and reports all instances of tree cover loss, regardless of whether the loss will be temporary or permanent (e.g., clear cut harvest followed by agriculture). As with forests, GFW does not seek to define deforestation but rather allows users to filter tree cover loss data based on their own preferred definition.</p>



<p>Though GFW data includes all forms of tree cover loss, data on certain drivers of loss is available on GFW, such as <a href="https://gfw.global/3eO7ZjW">tree cover loss due to fires</a> and <a href="https://gfw.global/39qbPdC">tree cover loss by dominant driver,</a> the latter of which can be used to better align with the definition of deforestation used by FRA. GFW uses the tree cover loss by dominant driver data to create a “<a href="https://research.wri.org/gfr/data-methods#forestchange1" target="_blank" rel="noreferrer noopener">deforestation proxy</a>” consisting of tree cover loss due to shifting agricultural expansion into primary forests, commodity-driven deforestation and urbanization. GFW users can also filter the tree cover loss data by primary forest and natural forest to view and download statistics for just these areas.</p>



<p>FRA deforestation data is available for the 5- or 10-year reporting periods, while GFW tree cover loss data is available annually.</p>



<h3 class="wp-block-heading">Forest expansion, tree cover gain and reforestation</h3>



<p>As with the deforestation data, information on forest expansion is available for the FRA reporting periods of 5 or 10 years. In FRA, forest expansion can be due to afforestation or a natural expansion of forest area.</p>



<p>GFW data on tree cover gain is available for 5-year time periods between 2000-2020. Like with the tree cover data, GFW gain data includes regrowth in plantations following harvesting, which would not be included in forest expansion according to FRA.</p>



<p>FRA also provides reforestation data, which does not influence forest area or their net change figure. FRA defines reforestation as the planting of trees in areas already considered as forests by FAO’s land use criterion.</p>



<h3 class="wp-block-heading">Forest area and tree cover net change</h3>



<p>FRA and GFW report net change in forest and tree cover extent respectively. GFW net change data currently only compares 2000 to 2020. FRA data is available for their 5- or 10-year reporting periods.</p>


<div class="wp-block-image">
<figure class="aligncenter size-large"><img decoding="async" loading="lazy" width="1583" height="2048" src="https://content.globalforestwatch.org/wp-content/uploads/2025/10/25_Forest_types_table-1-1583x2048.png" alt="" class="wp-image-19767" srcset="https://content.globalforestwatch.org/wp-content/uploads/2025/10/25_Forest_types_table-1-1583x2048.png 1583w, https://content.globalforestwatch.org/wp-content/uploads/2025/10/25_Forest_types_table-1-791x1024.png 791w, https://content.globalforestwatch.org/wp-content/uploads/2025/10/25_Forest_types_table-1-768x994.png 768w, https://content.globalforestwatch.org/wp-content/uploads/2025/10/25_Forest_types_table-1-1187x1536.png 1187w" sizes="(max-width: 1583px) 100vw, 1583px" /></figure></div>


<h2 class="wp-block-heading">Where is the data from?</h2>



<p>FRA compiles official government statistics on forest-related information. This data is self-reported by countries and can come from different sources using a variety of methods, including national forest inventories (NFIs), academic studies, government registries or district forest offices. Remote sensing (RS) is becoming more frequently used as an input into some nationally reported data. Although FAO provides guidelines on what and how to report, in reality, many countries employ their own definitions and methods.</p>



<p>GFW features data that is based on satellite data, providing an independent, transparent and globally consistent record of Earth’s land surface and how it is changing. This record enables consistent comparisons across countries and over time.</p>


<div class="wp-block-image">
<figure class="aligncenter size-large"><img decoding="async" loading="lazy" width="2048" height="1612" src="https://content.globalforestwatch.org/wp-content/uploads/2023/07/23_GFW_vs_FAO_graphics_Where-is-data-from-1-2048x1612.png" alt="Where is the data from for GFW and FRA?" class="wp-image-16211" srcset="https://content.globalforestwatch.org/wp-content/uploads/2023/07/23_GFW_vs_FAO_graphics_Where-is-data-from-1-2048x1612.png 2048w, https://content.globalforestwatch.org/wp-content/uploads/2023/07/23_GFW_vs_FAO_graphics_Where-is-data-from-1-1024x806.png 1024w, https://content.globalforestwatch.org/wp-content/uploads/2023/07/23_GFW_vs_FAO_graphics_Where-is-data-from-1-768x605.png 768w, https://content.globalforestwatch.org/wp-content/uploads/2023/07/23_GFW_vs_FAO_graphics_Where-is-data-from-1-1536x1209.png 1536w" sizes="(max-width: 2048px) 100vw, 2048px" /></figure></div>


<h2 class="wp-block-heading">When is the data collected and reported?</h2>



<p>FRA reports are released every 5 years based on countries’ reported information. Some countries update their forest information systems frequently, while others update them only every 10-20 years, meaning some FRA reports are based on older data. When country reports are unavailable, the FAO fills in gaps using available literature and expert estimates. <a href="https://iopscience.iop.org/article/10.1088/1748-9326/abd81b/meta" target="_blank" rel="noreferrer noopener">A study examining the 2020 FRA data sources</a> showed the most recent years in which RS and NFI data were collected and found that although many countries use data that are only a few years old, there are a number of countries using data over 30 years old.</p>



<p>GFW’s tree cover loss data set is updated each year with the previous year’s data. GFW’s net change and gain data covers the 20-year period from 2000 to 2020, with plans underway for more frequent updates.</p>


<div class="wp-block-image">
<figure class="aligncenter size-large"><img decoding="async" loading="lazy" width="2048" height="1188" src="https://content.globalforestwatch.org/wp-content/uploads/2025/10/25_GFW_vs_FAO_graphics_When-is-data-reported-2048x1188.png" alt="" class="wp-image-19766" srcset="https://content.globalforestwatch.org/wp-content/uploads/2025/10/25_GFW_vs_FAO_graphics_When-is-data-reported-2048x1188.png 2048w, https://content.globalforestwatch.org/wp-content/uploads/2025/10/25_GFW_vs_FAO_graphics_When-is-data-reported-1024x594.png 1024w, https://content.globalforestwatch.org/wp-content/uploads/2025/10/25_GFW_vs_FAO_graphics_When-is-data-reported-768x445.png 768w, https://content.globalforestwatch.org/wp-content/uploads/2025/10/25_GFW_vs_FAO_graphics_When-is-data-reported-1536x891.png 1536w" sizes="(max-width: 2048px) 100vw, 2048px" /></figure></div>


<h2 class="wp-block-heading">What is the unit of analysis?</h2>



<p>FRA collects statistics at the national scale, which are aggregated to provide insights about forest change dynamics regionally and globally.</p>



<p>Satellite-based data on GFW records changes in forest area by dividing Earth’s land surface into pixels. These pixels — which for most data on forest area and change are 30 m by 30 m in size — can then be aggregated to show changes over larger areas of interest. This modular format means that in addition to understanding forest dynamics at global, regional and national scales (as with the FRA), users can also monitor changes within custom areas of interest, such as a specific national park in their country, a provincial district, an Indigenous territory or even their own backyards.</p>


<div class="wp-block-image">
<figure class="aligncenter size-large"><img decoding="async" loading="lazy" width="1275" height="1223" src="https://content.globalforestwatch.org/wp-content/uploads/2023/07/GFW_vs_FAO_graphics_final-05.png" alt="What is the unit of analysis for GFW vs FRA?" class="wp-image-16213" srcset="https://content.globalforestwatch.org/wp-content/uploads/2023/07/GFW_vs_FAO_graphics_final-05.png 1275w, https://content.globalforestwatch.org/wp-content/uploads/2023/07/GFW_vs_FAO_graphics_final-05-1024x982.png 1024w, https://content.globalforestwatch.org/wp-content/uploads/2023/07/GFW_vs_FAO_graphics_final-05-768x737.png 768w" sizes="(max-width: 1275px) 100vw, 1275px" /></figure></div>


<h2 class="wp-block-heading">What does it all mean?</h2>



<p>How one defines forest, deforestation, reforestation and forest expansion is at the heart of the confusion and controversies surrounding these two data sources. Critics of FRA point out that country-reported data may be outdated, inconsistent, incomplete or based on non-transparent methods. In addition, FAO’s net forest change statistic (which is the only forest change variable presented for many countries) provides an overly optimistic view of global forest trends since by their definition, the loss of natural forests — rich in biodiversity and carbon — can be offset by expansion of tree plantation monocultures. The same applies for GFW’s net change information when viewed without the contextual data sets.</p>



<p>Meanwhile, critics of GFW claim that satellite-based monitoring has fundamental limitations on what can be observed from satellites, and that tree cover loss creates an overly pessimistic view since these data lump together permanent and temporary loss of tree cover within natural forests and tree plantations. In the case of FRA, temporary loss is not included in deforestation information.</p>



<p>Ultimately, neither GFW nor FRA is perfect or complete, but they both have important, and complementary, applications.</p>



<h2 class="wp-block-heading">When might I use GFW or FRA?</h2>



<p>Bottom-up approaches like FRA benefit from locally appropriate methodologies, customization of data and use of country definitions — although many countries follow the FAO’s standard definitions. FRA is more comprehensive in scope, ranging from biophysical to socio-economic and legislative variables. FRA’s data will be relevant for those that have a broad interest in the state of forests in a country and how they are being managed.</p>



<p>However, FRA’s data is only reported on a national level and cannot be used for assessing regional differences within countries. GFW’s spatially precise updates on tree cover loss allow users to track where changes in tree cover are happening on a sub-national level. GFW also offers more consistency in methods and definitions across geographies for making comparisons across countries.</p>



<p>GFW currently provides more frequent tree cover loss information (annually, and in near-real-time) than FRA does on deforestation, while FRA currently provides more up-to-date information on forest expansion than GFW does on tree cover gain — though GFW will eventually provide annual tree cover gain data. FRA’s figures also reflect tree planting initiatives that are too recent to be picked up by satellite data, although not all of these projects are successful.</p>



<p>In the end, the differences between GFW and FRA reveal strengths and shortcomings of both sources that make them appropriate for different applications, meaning that both are useful to understand what is happening with the world’s forests.</p>



<p></p>



<p><em>This blog was originally published in 2023. It was updated in October 2025 to reflect the latest updates to GFW and the FRA.</em> <em>Rachael Petersen and Octavia Payne contributed to a previous version of this blog.</em></p>



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



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

Global Forest Watch and the Forest Resources Assessment, Explained in 5 Graphics

<p>Today, we have more data about forests than ever before, but we still can’t seem to agree on where, when and why forests are changing around the world. Even two prominent global data sources appear to disagree, at least on the surface. “Deforestation continues to decline” was <a href="https://openknowledge.fao.org/server/api/core/bitstreams/2dee6e93-1988-4659-aa89-30dd20b43b15/content/FRA-2025/key-findings.html" target="_blank" rel="noreferrer noopener">a headline</a> of the <a href="https://www.fao.org/forest-resources-assessment/en/" target="_blank" rel="noreferrer noopener">2025 Global Forest Resources Assessment (FRA)</a> published by the United Nations Food and Agriculture Organization (FAO). At the same time, Global Forest Watch (GFW) <a href="https://gfr.wri.org/latest-analysis-deforestation-trends" target="_blank" rel="noreferrer noopener">reported that</a> “the tropics lost a record-shattering 6.7 million hectares of primary rainforest in 2024 ... nearly double that of 2023.”</p>



<p>What gives? Unfortunately, the answer is not simple. Both GFW and FRA provide valuable information on the state of global forests, but they differ substantially in terms of purpose, scope and approach. By answering questions on how and why these two global data sources differ, we can begin to see them as complementary rather than contradictory. Ultimately, our knowledge of global forests remains incomplete, and we need all the information we can get.</p>



<div class="textboxright"><strong>WHAT ARE GLOBAL FOREST WATCH (GFW) AND THE GLOBAL FOREST RESOURCES ASSESSMENT (FRA)?</strong> <br><br>GFW is an independent online forest monitoring platform launched in 2014 by a consortium of partners led by World Resources Institute. GFW seeks to empower forest managers, law enforcement officers and other forest stakeholders with free access to timely and high-resolution data about the current status of forests and recent forest change. GFW uses satellite technology to monitor tree cover loss worldwide with weekly and annual updates, as well as to provide longer term tree cover gain and net change data. Gain, net change and annual loss data are GFW’s core data sets and are provided by the University of Maryland. Other data sets on GFW provide useful contextual information related to forests and forest change.<br><br>

The FRA, which the FAO has published at 5- to 10-year intervals since 1946, seeks to provide a consistent and comprehensive approach to describing the world’s forests and how they are changing. Countries provide reports to FRA based on data from National Forest Inventories, satellite data and extrapolation based on past data. The FAO combines the data submitted by countries into global reports and fills in any gaps where data is not available from the countries. The most recent edition of the FRA in 2025 included national-level statistics on around 50 broad ecological and socio-economic variables reported by 236 countries and territories. </div>



<h2 class="wp-block-heading">What is considered a “forest”?</h2>



<p>GFW and FRA both talk about “forests,” but what are they actually measuring and monitoring? There are nearly&nbsp;<a href="https://www.nrs.fs.usda.gov/pubs/gtr/gtr_wo077/gtr_wo077_011.pdf" target="_blank" rel="noreferrer noopener">800 definitions</a>&nbsp;of “forest” in use, which only adds to the confusion.</p>



<p>Every five years, FRA publishes forest area globally and per country. These area measurements are based on a <a href="https://openknowledge.fao.org/server/api/core/bitstreams/a6e225da-4a31-4e06-818d-ca3aeadfd635/content" target="_blank" rel="noreferrer noopener">standardized&nbsp;definition</a>&nbsp;of “forest” adopted by the FAO in 2000, which takes into account biophysical and land use criteria. The biophysical criterion establishes minimum thresholds for the height, canopy cover and extent of trees. The land use criterion requires that land be officially or legally designated for a “forest use,” such as conservation or harvest. 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. At the same time, 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.</p>



<p>GFW provides data on “tree cover,” which is derived from satellite data and therefore cannot discern intended land use. Thus, GFW’s monitoring systems rely entirely on biophysical criteria — related to height, canopy cover and extent of trees. GFW does not adopt a specific definition of forest, but rather monitors all forms of tree cover including natural forests and tree plantations. Using GFW’s interactive online platform, users can filter this data based on their preferred definition of forest. Users can select by canopy cover percent, as well as tree height. Users can also view and derive statistics using other data sets that give additional context on the land use, such as <a href="https://gfw.global/3jyrlta">primary forests</a>, <a href="http://gfw.global/4fQfl03">natural forests</a> and <a href="https://gfw.global/2Qbv7fM">planted trees</a>.</p>



<figure class="wp-block-image aligncenter size-large"><img src="https://content.globalforestwatch.org/wp-content/uploads/2023/07/23_GFW_vs_FAO_graphics_What-is-considered-a-forest-1-2048x1867.png" alt="What is considered a &quot;Forest&quot; by GFW and FRA?" class="wp-image-16209"/></figure>



<h2 class="wp-block-heading">How is “forest change” reported?</h2>



<p>Both FRA and GFW aggregate monitoring data into global statistics about forest change. Like the different ways they measure forests, GFW and FRA apply different lenses to monitor <em>changes to</em> forests, which subsequently influence how both forest expansion and deforestation are reported. GFW reports annual gross loss of tree cover, as well as gross gain for four 5-year periods and net change for a 20-year period. FRA reports on deforestation, net change and forest expansion as separate statistics, though many countries only report forest area, from which net change can be derived.</p>



<h3 class="wp-block-heading">Deforestation and tree cover loss</h3>



<p>Using its land use criterion, the FAO’s definition of deforestation requires permanent land use change to occur. Losses of tree cover considered to be temporary (e.g., forests that have been clear cut but will be left to regrow) are not counted as deforestation.</p>



<p>Conversely, GFW detects and reports all instances of tree cover loss, regardless of whether the loss will be temporary or permanent (e.g., clear cut harvest followed by agriculture). As with forests, GFW does not seek to define deforestation but rather allows users to filter tree cover loss data based on their own preferred definition.</p>



<p>Though GFW data includes all forms of tree cover loss, data on certain drivers of loss is available on GFW, such as <a href="https://gfw.global/3eO7ZjW">tree cover loss due to fires</a> and <a href="https://gfw.global/39qbPdC">tree cover loss by dominant driver,</a> the latter of which can be used to better align with the definition of deforestation used by FRA. GFW uses the tree cover loss by dominant driver data to create a “<a href="https://research.wri.org/gfr/data-methods#forestchange1" target="_blank" rel="noreferrer noopener">deforestation proxy</a>” consisting of tree cover loss due to shifting agricultural expansion into primary forests, commodity-driven deforestation and urbanization. GFW users can also filter the tree cover loss data by primary forest and natural forest to view and download statistics for just these areas.</p>



<p>FRA deforestation data is available for the 5- or 10-year reporting periods, while GFW tree cover loss data is available annually.</p>



<h3 class="wp-block-heading">Forest expansion, tree cover gain and reforestation</h3>



<p>As with the deforestation data, information on forest expansion is available for the FRA reporting periods of 5 or 10 years. In FRA, forest expansion can be due to afforestation or a natural expansion of forest area.</p>



<p>GFW data on tree cover gain is available for 5-year time periods between 2000-2020. Like with the tree cover data, GFW gain data includes regrowth in plantations following harvesting, which would not be included in forest expansion according to FRA.</p>



<p>FRA also provides reforestation data, which does not influence forest area or their net change figure. FRA defines reforestation as the planting of trees in areas already considered as forests by FAO’s land use criterion.</p>



<h3 class="wp-block-heading">Forest area and tree cover net change</h3>



<p>FRA and GFW report net change in forest and tree cover extent respectively. GFW net change data currently only compares 2000 to 2020. FRA data is available for their 5- or 10-year reporting periods.</p>



<figure class="wp-block-image aligncenter size-large"><img src="https://content.globalforestwatch.org/wp-content/uploads/2025/10/25_Forest_types_table-1-1583x2048.png" alt="" class="wp-image-19767"/></figure>



<h2 class="wp-block-heading">Where is the data from?</h2>



<p>FRA compiles official government statistics on forest-related information. This data is self-reported by countries and can come from different sources using a variety of methods, including national forest inventories (NFIs), academic studies, government registries or district forest offices. Remote sensing (RS) is becoming more frequently used as an input into some nationally reported data. Although FAO provides guidelines on what and how to report, in reality, many countries employ their own definitions and methods.</p>



<p>GFW features data that is based on satellite data, providing an independent, transparent and globally consistent record of Earth’s land surface and how it is changing. This record enables consistent comparisons across countries and over time.</p>



<figure class="wp-block-image aligncenter size-large"><img src="https://content.globalforestwatch.org/wp-content/uploads/2023/07/23_GFW_vs_FAO_graphics_Where-is-data-from-1-2048x1612.png" alt="Where is the data from for GFW and FRA?" class="wp-image-16211"/></figure>



<h2 class="wp-block-heading">When is the data collected and reported?</h2>



<p>FRA reports are released every 5 years based on countries’ reported information. Some countries update their forest information systems frequently, while others update them only every 10-20 years, meaning some FRA reports are based on older data. When country reports are unavailable, the FAO fills in gaps using available literature and expert estimates. <a href="https://iopscience.iop.org/article/10.1088/1748-9326/abd81b/meta" target="_blank" rel="noreferrer noopener">A study examining the 2020 FRA data sources</a> showed the most recent years in which RS and NFI data were collected and found that although many countries use data that are only a few years old, there are a number of countries using data over 30 years old.</p>



<p>GFW’s tree cover loss data set is updated each year with the previous year’s data. GFW’s net change and gain data covers the 20-year period from 2000 to 2020, with plans underway for more frequent updates.</p>



<figure class="wp-block-image aligncenter size-large"><img src="https://content.globalforestwatch.org/wp-content/uploads/2025/10/25_GFW_vs_FAO_graphics_When-is-data-reported-2048x1188.png" alt="" class="wp-image-19766"/></figure>



<h2 class="wp-block-heading">What is the unit of analysis?</h2>



<p>FRA collects statistics at the national scale, which are aggregated to provide insights about forest change dynamics regionally and globally.</p>



<p>Satellite-based data on GFW records changes in forest area by dividing Earth’s land surface into pixels. These pixels — which for most data on forest area and change are 30 m by 30 m in size — can then be aggregated to show changes over larger areas of interest. This modular format means that in addition to understanding forest dynamics at global, regional and national scales (as with the FRA), users can also monitor changes within custom areas of interest, such as a specific national park in their country, a provincial district, an Indigenous territory or even their own backyards.</p>



<figure class="wp-block-image aligncenter size-large"><img src="https://content.globalforestwatch.org/wp-content/uploads/2023/07/GFW_vs_FAO_graphics_final-05.png" alt="What is the unit of analysis for GFW vs FRA?" class="wp-image-16213"/></figure>



<h2 class="wp-block-heading">What does it all mean?</h2>



<p>How one defines forest, deforestation, reforestation and forest expansion is at the heart of the confusion and controversies surrounding these two data sources. Critics of FRA point out that country-reported data may be outdated, inconsistent, incomplete or based on non-transparent methods. In addition, FAO’s net forest change statistic (which is the only forest change variable presented for many countries) provides an overly optimistic view of global forest trends since by their definition, the loss of natural forests — rich in biodiversity and carbon — can be offset by expansion of tree plantation monocultures. The same applies for GFW’s net change information when viewed without the contextual data sets.</p>



<p>Meanwhile, critics of GFW claim that satellite-based monitoring has fundamental limitations on what can be observed from satellites, and that tree cover loss creates an overly pessimistic view since these data lump together permanent and temporary loss of tree cover within natural forests and tree plantations. In the case of FRA, temporary loss is not included in deforestation information.</p>



<p>Ultimately, neither GFW nor FRA is perfect or complete, but they both have important, and complementary, applications.</p>



<h2 class="wp-block-heading">When might I use GFW or FRA?</h2>



<p>Bottom-up approaches like FRA benefit from locally appropriate methodologies, customization of data and use of country definitions — although many countries follow the FAO’s standard definitions. FRA is more comprehensive in scope, ranging from biophysical to socio-economic and legislative variables. FRA's data will be relevant for those that have a broad interest in the state of forests in a country and how they are being managed.</p>



<p>However, FRA's data is only reported on a national level and cannot be used for assessing regional differences within countries. GFW’s spatially precise updates on tree cover loss allow users to track where changes in tree cover are happening on a sub-national level. GFW also offers more consistency in methods and definitions across geographies for making comparisons across countries.</p>



<p>GFW currently provides more frequent tree cover loss information (annually, and in near-real-time) than FRA does on deforestation, while FRA currently provides more up-to-date information on forest expansion than GFW does on tree cover gain — though GFW will eventually provide annual tree cover gain data. FRA’s figures also reflect tree planting initiatives that are too recent to be picked up by satellite data, although not all of these projects are successful.</p>



<p>In the end, the differences between GFW and FRA reveal strengths and shortcomings of both sources that make them appropriate for different applications, meaning that both are useful to understand what is happening with the world’s forests.</p>



<p></p>



<p><em>This blog was originally published in 2023. It was updated in October 2025 to reflect the latest updates to GFW and the FRA.</em> <em>Rachael Petersen and Octavia Payne contributed to a previous version of this blog.</em></p>



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



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

Both GFW and FRA provide valuable information on the state of forests, and they both have important, and complementary, applications.

Global Forest Watch and Forest Resources Assessment | GFW Blog

https://secure.gravatar.com/avatar/bf3dd86ffe8c109dcc6ad6d415f8ad0b?s=96&d=mm&r=g

James Anderson

Laura Cué la Rosa

<p>Global Forest Watch’s (GFW) deforestation alerts provide information about loss of tree cover in near-real-time, making them a critical tool for those working to combat deforestation. Existing alerts on GFW have been used to <a href="https://www.globalforestwatch.org/blog/users-in-action/indigenous-forest-monitors-reduce-deforestation/">help Indigenous and local communities</a> protect their forests, <a href="https://www.wri.org/insights/indigenous-peoples-local-communities-use-satellite-data-deforestation?ap3c=IGbOJV0jomFif30GAGbOJV2IivnpZwia9EpGvhO1OTmkkCABCQ" target="_blank" rel="noreferrer noopener">halt and prosecute</a> environmental crimes and <a href="https://www.globalforestwatch.org/blog/users-in-action/data-protect-cameroon-mengame-gorilla-sanctuary/">protect</a> threatened <a href="https://www.globalforestwatch.org/blog/users-in-action/protecting-forest-habitats-mara-elephant-project/">wildlife</a>.</p>



<p>However, in the past, alerts only let users know there’s been a disturbance without identifying a cause, other than a few limited classes such as fires.</p>



<p>Now, <a href="http://gfw.global/43jOCp9">new drivers of deforestation alerts data</a> is available on GFW that shows the causes of alerts across the tropics. This innovative data set uses an AI method to assign drivers to GFW’s integrated deforestation alerts in the three major forested regions in the tropics, allowing users to see what’s driving loss in near-real-time in these regions for the first time.</p>



<p>Critically, this information helps users distinguish between tree cover loss from natural and human causes, enabling more targeted action from those managing forests and reducing the need for costly field visits. Further, the data also provides more rapid information to estimate ecological impacts and supports better understanding of carbon emissions related to forest disturbances.</p>



<figure class="wp-block-image size-full"><img decoding="async" loading="lazy" width="1310" height="788" src="https://content.globalforestwatch.org/wp-content/uploads/2025/10/drivers_app.png" alt="" class="wp-image-19750" srcset="https://content.globalforestwatch.org/wp-content/uploads/2025/10/drivers_app.png 1310w, https://content.globalforestwatch.org/wp-content/uploads/2025/10/drivers_app-1024x616.png 1024w, https://content.globalforestwatch.org/wp-content/uploads/2025/10/drivers_app-768x462.png 768w" sizes="(max-width: 1310px) 100vw, 1310px" /></figure>



<h2 class="wp-block-heading"><strong>What does the drivers of deforestation alerts data show?</strong></h2>



<h3 class="wp-block-heading"><strong><em>Driver classes</em></strong></h3>



<p>The data classifies GFW’s near-real-time integrated deforestation alerts into 11 natural and human-caused driver classes at 10-meter resolution:</p>



<div class="flourish-embed flourish-cards" data-src="visualisation/25392464?163031"><script src="https://public.flourish.studio/resources/embed.js"></script><noscript><img decoding="async" src="https://public.flourish.studio/visualisation/25392464/thumbnail" width="100%" alt="cards visualization" /></noscript></div>



<h3 class="wp-block-heading"><strong><em>Geographic coverage</em></strong></h3>



<p>The data covers the three major forest basins of the tropics in the Amazon, Congo and Indonesia, which cover around one third of the world’s land surface and include 80% of the world’s tropical forests. </p>



<figure class="wp-block-image size-full"><img decoding="async" loading="lazy" width="1501" height="482" src="https://content.globalforestwatch.org/wp-content/uploads/2025/10/alerts_coverage_gfw.png" alt="" class="wp-image-19751" srcset="https://content.globalforestwatch.org/wp-content/uploads/2025/10/alerts_coverage_gfw.png 1501w, https://content.globalforestwatch.org/wp-content/uploads/2025/10/alerts_coverage_gfw-1024x329.png 1024w, https://content.globalforestwatch.org/wp-content/uploads/2025/10/alerts_coverage_gfw-768x247.png 768w" sizes="(max-width: 1501px) 100vw, 1501px" /></figure>



<p>Tropical forests are some of the <a href="https://gfr.wri.org/many-benefits-forests">most important forest ecosystems</a>, critical for livelihoods, carbon storage, water provision, biodiversity and more, and are <a href="https://gfr.wri.org/latest-analysis-deforestation-trends">under threat from deforestation</a>. &nbsp;</p>



<h3 class="wp-block-heading"><strong><em>Temporal coverage and update frequency</em></strong></h3>



<p>Currently, data is available for alerts detected from 2022-2024, but the algorithm will soon be adapted to run on a monthly basis. In this near-real-time setup, each new alert is classified within 31 days of detection, and every month the model reprocesses alerts from the previous three months on a rolling basis. Classifications are only displayed if they are confidently classified by the algorithm, and labels may be updated if a more confident classification becomes available. As a result, each alert can be reclassified up to three times, with the layer showing the most confident classification. This process allows accuracy to improve over time as additional satellite imagery and context become available. Additionally, some drivers (e.g., selective logging and road development) are best identified soon after the canopy is disturbed, while other drivers (e.g., mining or agriculture) may take longer for their characteristic signals to appear.</p>



<h2 class="wp-block-heading"><strong>How was the data developed?</strong></h2>



<p>The <a href="https://www.globalforestwatch.org/blog/data-and-tools/integrated-deforestation-alerts/">integrated deforestation alerts</a> at 10-meter resolution are the basis for the driver classifications. Integrated alerts combine the power of GLAD-L, GLAD-S2 and RADD deforestation alerts to provide a faster, more confident view of forest disturbances than any one individual system.</p>



<p>To apply driver classifications to the integrated deforestation alerts, we used a deep learning model first developed for Suriname, the Republic of Congo and the Democratic Republic of the Congo. The model was trained on an extensive data set of forest disturbance events in the Amazon, Congo and Southeast Asian rainforests using visual interpretation of high-resolution Planet imagery to assign classification labels. The&nbsp;model&nbsp;uses post-disturbance Sentinel-1 and Sentinel-2&nbsp;satellite&nbsp;imagery for the&nbsp;driver&nbsp;classifications,&nbsp;integrating the&nbsp;complementary strengths of radar and optical satellite data. This means that the model can classify drivers&nbsp;rapidly,&nbsp;even&nbsp;in cloudy conditions.&nbsp;&nbsp;</p>



<p>To identify where fire coincided with clearing of agricultural land — a commonly used agricultural management technique in the tropics — we used VIIRS fire alerts and Normalized Burn Ratio (information that identifies burned areas) from Sentinel-2.</p>



<p>To reduce noise in the data,&nbsp;we&nbsp;carried out&nbsp;smoothing:&nbsp;&nbsp;</p>



<ul>
<li>First, alert pixels not exceeding 0.75 confidence were attributed&nbsp;as&nbsp;uncertain classifications and set to ‘unlabeled.’&nbsp;</li>
</ul>



<ul>
<li>Second,&nbsp;for&nbsp;each distinct alert patch (i.e.&nbsp;a spatially connected group of alert pixels),&nbsp;the dominant driver was selected and attributed to the entire patch, with a few exceptions&nbsp;for drivers&nbsp;occurring&nbsp;at a finer, pixel-level scale&nbsp;(road development and selective logging)&nbsp;and for&nbsp;extremely large patches.&nbsp;</li>
</ul>



<p>The accuracy of the classifications varies by region, with the Congo Basin having the highest overall performance. Most misclassifications are related to confusion between agriculture and mining,&nbsp;which often occur in the same area, and&nbsp;between&nbsp;flooding and wildfires&nbsp;in floodplains.&nbsp;</p>



<p><a href="https://www.sciencedirect.com/science/article/pii/S0034425723002067" target="_blank" rel="noreferrer noopener">Learn more here</a> about the methodology and limitations.</p>



<h2 class="wp-block-heading"><strong>How does the data compare to other data showing drivers of loss?</strong></h2>



<p>Other data sets from GFW and Land & Carbon Lab provide information on drivers of vegetation disturbances, but vary in terms of their detail, timeframe and update frequency, and therefore have different applications. For example, the <a href="https://www.globalforestwatch.org/blog/data-and-tools/new-drivers-data-forest-loss/">1-km annual drivers of tree cover loss data</a> available on GFW is more suited for historical analysis over many years, and the Land Disturbance Alert Classification System from Land & Carbon Lab provides information on vegetation disturbances both inside and outside of forests globally.</p>



<div class="flourish-embed flourish-table" data-src="visualisation/25391478?163031"><script src="https://public.flourish.studio/resources/embed.js"></script><noscript><img decoding="async" src="https://public.flourish.studio/visualisation/25391478/thumbnail" width="100%" alt="table visualization" /></noscript></div>



<p>Some attributes make the data sets more suitable for specific applications. For example, higher resolution allows more insight into co-occurrence of drivers in local areas, and sub-annual information allows insights into the seasonality of different drivers:</p>



<div class="flourish-embed flourish-table" data-src="visualisation/25549662?163031"><script src="https://public.flourish.studio/resources/embed.js"></script><noscript><img decoding="async" src="https://public.flourish.studio/visualisation/25549662/thumbnail" width="100%" alt="table visualization" /></noscript></div>



<h2 class="wp-block-heading"><strong>How can the data be used? A vital source for forest protectors</strong></h2>



<p>Deforestation alerts are <a href="https://www.globalforestwatch.org/blog/category-and-topics/?category=users-in-action&topic=deforestation-alerts">used around the world</a> by hundreds of Indigenous Peoples and local communities, as well as forest rangers and NGOs, to investigate potential unplanned or illegal disturbances in forests.</p>



<p>However, different causes for disturbances warrant different responses. For example, in the case of agricultural conversion, users may want to investigate the loss to assess the legality of the clearing. In the case of a natural disturbance such as a landslide, either no investigation or a different sort of investigation would be required. Users can reference the drivers of deforestation alerts data to determine what type of response is appropriate depending on the cause of the disturbance.</p>



<p>While field visits or the use of high-resolution satellite imagery to check classifications may still be required in some instances to plan appropriate action or to fully understand the context, the addition of driver information on GFW enables alert prioritization and more targeted action, saving resources and allowing investigations to be focused on more concerning losses.</p>



<p>The data can also be used to improve understanding of carbon emissions from forest disturbances. Some types of conversion clear all the vegetation from the land, whereas others leave trees behind. In addition, information on the cause of loss is required to estimate emissions from the subsequent land use after the forest has been cleared. It also indicates whether regrowth can be expected, which is likely after logging activities but not after conversion to permanent agriculture, for example.</p>



<h2 class="wp-block-heading"><strong>What’s next for the drivers of deforestation alerts data?</strong></h2>



<p><strong>Monthly updates: </strong>The 2025 data will be made available in early 2026, after which monthly updates will be available from 2026 onwards.</p>



<p><strong>Expansion to the whole of the tropics: </strong>We soon plan to expand the data across the full coverage of the integrated deforestation alerts (currently 30 degrees north to 30 degrees south). In the future, the model may be applied to disturbances in temperate and boreal forests as well. Further research could also allow disturbances of other vegetation types to be classified using a similar approach.</p>



<p><strong>Integration with other GFW tools</strong>: We plan to add the data to Forest Watcher to make it more accessible to forest protectors working in the field.</p>



<p>Stay tuned for more updates as we continue to improve drivers of deforestation alerts for greater coverage, access and application, and <a href="https://www.globalforestwatch.org/help/datasets/forest-change/">visit the Help Center</a> to learn more.  </p>



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

See What’s Driving Deforestation Alerts for the First Time on Global Forest Watch

<p>Global Forest Watch’s (GFW) deforestation alerts provide information about loss of tree cover in near-real-time, making them a critical tool for those working to combat deforestation. Existing alerts on GFW have been used to <a href="https://www.globalforestwatch.org/blog/users-in-action/indigenous-forest-monitors-reduce-deforestation/">help Indigenous and local communities</a> protect their forests, <a href="https://www.wri.org/insights/indigenous-peoples-local-communities-use-satellite-data-deforestation?ap3c=IGbOJV0jomFif30GAGbOJV2IivnpZwia9EpGvhO1OTmkkCABCQ" target="_blank" rel="noreferrer noopener">halt and prosecute</a> environmental crimes and <a href="https://www.globalforestwatch.org/blog/users-in-action/data-protect-cameroon-mengame-gorilla-sanctuary/">protect</a> threatened <a href="https://www.globalforestwatch.org/blog/users-in-action/protecting-forest-habitats-mara-elephant-project/">wildlife</a>.</p>



<p>However, in the past, alerts only let users know there’s been a disturbance without identifying a cause, other than a few limited classes such as fires.</p>



<p>Now, <a href="http://gfw.global/43jOCp9">new drivers of deforestation alerts data</a> is available on GFW that shows the causes of alerts across the tropics. This innovative data set uses an AI method to assign drivers to GFW’s integrated deforestation alerts in the three major forested regions in the tropics, allowing users to see what’s driving loss in near-real-time in these regions for the first time.</p>



<p>Critically, this information helps users distinguish between tree cover loss from natural and human causes, enabling more targeted action from those managing forests and reducing the need for costly field visits. Further, the data also provides more rapid information to estimate ecological impacts and supports better understanding of carbon emissions related to forest disturbances.</p>



<figure class="wp-block-image size-full"><img src="https://content.globalforestwatch.org/wp-content/uploads/2025/10/drivers_app.png" alt="" class="wp-image-19750"/></figure>



<h2 class="wp-block-heading"><strong>What does the drivers of deforestation alerts data show?</strong></h2>



<h3 class="wp-block-heading"><strong><em>Driver classes</em></strong></h3>



<p>The data classifies GFW’s near-real-time integrated deforestation alerts into 11 natural and human-caused driver classes at 10-meter resolution:</p>



<div class="flourish-embed flourish-cards" data-src="visualisation/25392464?163031"><script src="https://public.flourish.studio/resources/embed.js"></script><noscript><img src="https://public.flourish.studio/visualisation/25392464/thumbnail" width="100%" alt="cards visualization" /></noscript></div>



<h3 class="wp-block-heading"><strong><em>Geographic coverage</em></strong></h3>



<p>The data covers the three major forest basins of the tropics in the Amazon, Congo and Indonesia, which cover around one third of the world’s land surface and include 80% of the world's tropical forests. </p>



<figure class="wp-block-image size-full"><img src="https://content.globalforestwatch.org/wp-content/uploads/2025/10/alerts_coverage_gfw.png" alt="" class="wp-image-19751"/></figure>



<p>Tropical forests are some of the <a href="https://gfr.wri.org/many-benefits-forests">most important forest ecosystems</a>, critical for livelihoods, carbon storage, water provision, biodiversity and more, and are <a href="https://gfr.wri.org/latest-analysis-deforestation-trends">under threat from deforestation</a>. &nbsp;</p>



<h3 class="wp-block-heading"><strong><em>Temporal coverage and update frequency</em></strong></h3>



<p>Currently, data is available for alerts detected from 2022-2024, but the algorithm will soon be adapted to run on a monthly basis. In this near-real-time setup, each new alert is classified within 31 days of detection, and every month the model reprocesses alerts from the previous three months on a rolling basis. Classifications are only displayed if they are confidently classified by the algorithm, and labels may be updated if a more confident classification becomes available. As a result, each alert can be reclassified up to three times, with the layer showing the most confident classification. This process allows accuracy to improve over time as additional satellite imagery and context become available. Additionally, some drivers (e.g., selective logging and road development) are best identified soon after the canopy is disturbed, while other drivers (e.g., mining or agriculture) may take longer for their characteristic signals to appear.</p>



<h2 class="wp-block-heading"><strong>How was the data developed?</strong></h2>



<p>The <a href="https://www.globalforestwatch.org/blog/data-and-tools/integrated-deforestation-alerts/">integrated deforestation alerts</a> at 10-meter resolution are the basis for the driver classifications. Integrated alerts combine the power of GLAD-L, GLAD-S2 and RADD deforestation alerts to provide a faster, more confident view of forest disturbances than any one individual system.</p>



<p>To apply driver classifications to the integrated deforestation alerts, we used a deep learning model first developed for Suriname, the Republic of Congo and the Democratic Republic of the Congo. The model was trained on an extensive data set of forest disturbance events in the Amazon, Congo and Southeast Asian rainforests using visual interpretation of high-resolution Planet imagery to assign classification labels. The&nbsp;model&nbsp;uses post-disturbance Sentinel-1 and Sentinel-2&nbsp;satellite&nbsp;imagery for the&nbsp;driver&nbsp;classifications,&nbsp;integrating the&nbsp;complementary strengths of radar and optical satellite data. This means that the model can classify drivers&nbsp;rapidly,&nbsp;even&nbsp;in cloudy conditions.&nbsp;&nbsp;</p>



<p>To identify where fire coincided with clearing of agricultural land — a commonly used agricultural management technique in the tropics — we used VIIRS fire alerts and Normalized Burn Ratio (information that identifies burned areas) from Sentinel-2.</p>



<p>To reduce noise in the data,&nbsp;we&nbsp;carried out&nbsp;smoothing:&nbsp;&nbsp;</p>



<ul>
<li>First, alert pixels not exceeding 0.75 confidence were attributed&nbsp;as&nbsp;uncertain classifications and set to ‘unlabeled.’&nbsp;</li>
</ul>



<ul>
<li>Second,&nbsp;for&nbsp;each distinct alert patch (i.e.&nbsp;a spatially connected group of alert pixels),&nbsp;the dominant driver was selected and attributed to the entire patch, with a few exceptions&nbsp;for drivers&nbsp;occurring&nbsp;at a finer, pixel-level scale&nbsp;(road development and selective logging)&nbsp;and for&nbsp;extremely large patches.&nbsp;</li>
</ul>



<p>The accuracy of the classifications varies by region, with the Congo Basin having the highest overall performance. Most misclassifications are related to confusion between agriculture and mining,&nbsp;which often occur in the same area, and&nbsp;between&nbsp;flooding and wildfires&nbsp;in floodplains.&nbsp;</p>



<p><a href="https://www.sciencedirect.com/science/article/pii/S0034425723002067" target="_blank" rel="noreferrer noopener">Learn more here</a> about the methodology and limitations.</p>



<h2 class="wp-block-heading"><strong>How does the data compare to other data showing drivers of loss?</strong></h2>



<p>Other data sets from GFW and Land &amp; Carbon Lab provide information on drivers of vegetation disturbances, but vary in terms of their detail, timeframe and update frequency, and therefore have different applications. For example, the <a href="https://www.globalforestwatch.org/blog/data-and-tools/new-drivers-data-forest-loss/">1-km annual drivers of tree cover loss data</a> available on GFW is more suited for historical analysis over many years, and the Land Disturbance Alert Classification System from Land &amp; Carbon Lab provides information on vegetation disturbances both inside and outside of forests globally.</p>



<div class="flourish-embed flourish-table" data-src="visualisation/25391478?163031"><script src="https://public.flourish.studio/resources/embed.js"></script><noscript><img src="https://public.flourish.studio/visualisation/25391478/thumbnail" width="100%" alt="table visualization" /></noscript></div>



<p>Some attributes make the data sets more suitable for specific applications. For example, higher resolution allows more insight into co-occurrence of drivers in local areas, and sub-annual information allows insights into the seasonality of different drivers:</p>



<div class="flourish-embed flourish-table" data-src="visualisation/25549662?163031"><script src="https://public.flourish.studio/resources/embed.js"></script><noscript><img src="https://public.flourish.studio/visualisation/25549662/thumbnail" width="100%" alt="table visualization" /></noscript></div>



<h2 class="wp-block-heading"><strong>How can the data be used? A vital source for forest protectors</strong></h2>



<p>Deforestation alerts are <a href="https://www.globalforestwatch.org/blog/category-and-topics/?category=users-in-action&amp;topic=deforestation-alerts">used around the world</a> by hundreds of Indigenous Peoples and local communities, as well as forest rangers and NGOs, to investigate potential unplanned or illegal disturbances in forests.</p>



<p>However, different causes for disturbances warrant different responses. For example, in the case of agricultural conversion, users may want to investigate the loss to assess the legality of the clearing. In the case of a natural disturbance such as a landslide, either no investigation or a different sort of investigation would be required. Users can reference the drivers of deforestation alerts data to determine what type of response is appropriate depending on the cause of the disturbance.</p>



<p>While field visits or the use of high-resolution satellite imagery to check classifications may still be required in some instances to plan appropriate action or to fully understand the context, the addition of driver information on GFW enables alert prioritization and more targeted action, saving resources and allowing investigations to be focused on more concerning losses.</p>



<p>The data can also be used to improve understanding of carbon emissions from forest disturbances. Some types of conversion clear all the vegetation from the land, whereas others leave trees behind. In addition, information on the cause of loss is required to estimate emissions from the subsequent land use after the forest has been cleared. It also indicates whether regrowth can be expected, which is likely after logging activities but not after conversion to permanent agriculture, for example.</p>



<h2 class="wp-block-heading"><strong>What’s next for the drivers of deforestation alerts data?</strong></h2>



<p><strong>Monthly updates: </strong>The 2025 data will be made available in early 2026, after which monthly updates will be available from 2026 onwards.</p>



<p><strong>Expansion to the whole of the tropics: </strong>We soon plan to expand the data across the full coverage of the integrated deforestation alerts (currently 30 degrees north to 30 degrees south). In the future, the model may be applied to disturbances in temperate and boreal forests as well. Further research could also allow disturbances of other vegetation types to be classified using a similar approach.</p>



<p><strong>Integration with other GFW tools</strong>: We plan to add the data to Forest Watcher to make it more accessible to forest protectors working in the field.</p>



<p>Stay tuned for more updates as we continue to improve drivers of deforestation alerts for greater coverage, access and application, and <a href="https://www.globalforestwatch.org/help/datasets/forest-change/">visit the Help Center</a> to learn more.  </p>



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

<p>Les alertes de déforestation de Global Forest Watch (GFW) fournissent des informations sur la perte de couverture forestière en temps quasi réel, ce qui en fait un outil essentiel pour ceux qui luttent contre la déforestation. Les alertes existantes sur GFW ont été utilisées pour aider les communautés autochtones et locales à protéger leurs forêts, à mettre fin aux crimes environnementaux et à poursuivre leurs auteurs, ainsi qu'à protéger la faune sauvage menacée.</p>



<p>Cependant, par le passé, les alertes se contentaient d'informer les utilisateurs qu'un incident s'était produit, sans en identifier la cause, à l'exception de quelques cas particuliers tels que les incendies.</p>



<p>Désormais, de nouvelles données sur les facteurs de déforestation&nbsp;&nbsp; sont disponibles sur GFW et montrent les causes des alertes dans les régions tropicales. Cet ensemble de données innovant utilise une méthode d'intelligence artificielle pour attribuer des facteurs aux alertes de déforestation intégrées de GFW dans les trois principales régions forestières des tropiques, permettant ainsi aux utilisateurs de voir pour la première fois en temps quasi réel ce qui cause la perte de couverture forestière dans ces régions.&nbsp; Ces informations aident les utilisateurs à distinguer la perte de couverture forestière due à des causes naturelles de celle due à des causes humaines, ce qui permet aux gestionnaires forestiers de prendre des mesures plus ciblées et réduit le besoin de visites coûteuses sur le terrain. En outre, les données fournissent également des informations plus rapides pour estimer les impacts écologiques et permettent de mieux comprendre les émissions de carbone liées aux perturbations forestières.</p>



<figure class="wp-block-image size-full"><img src="https://content.globalforestwatch.org/wp-content/uploads/2025/11/fr_drivers_app.png" alt="" class="wp-image-19831"/></figure>



<h2 class="wp-block-heading"><strong>Que révèlent les données sur les facteurs déclencheurs des alertes de déforestation?</strong></h2>



<h3 class="wp-block-heading"><em>Classifications des alertes de déforestation</em></h3>



<p>Les données classent les alertes de déforestation intégrées en temps quasi réel de GFW en 11 catégories de facteurs naturels et anthropiques à une résolution de 10 mètres :</p>



<div class="flourish-embed flourish-cards" data-src="visualisation/26207415?163031"><script src="https://public.flourish.studio/resources/embed.js"></script><noscript><img src="https://public.flourish.studio/visualisation/26207415/thumbnail" width="100%" alt="cards visualization" /></noscript></div>



<h3 class="wp-block-heading"><em>Couverture géographique </em></h3>



<p>Les données couvrent les trois principaux bassins forestiers tropicaux de l'Amazonie, du Congo et de l'Indonésie, qui représentent environ un tiers de la superficie terrestre mondial et abritent 80 % des forêts tropicales de la planète. Les forêts tropicales comptent parmi les écosystèmes forestiers les plus importants. Elles jouent un rôle essentiel pour les moyens de subsistance, le stockage du carbone, l'approvisionnement en eau, la biodiversité et bien d'autres aspects, mais elles sont menacées par la déforestation. </p>



<figure class="wp-block-image size-full"><img src="https://content.globalforestwatch.org/wp-content/uploads/2025/10/alerts_coverage_gfw.png" alt="" class="wp-image-19751"/></figure>



<h3 class="wp-block-heading"><em><em>Couverture temporelle et fréquence de mise à jour</em></em></h3>



<p>Actuellement, les données sont disponibles pour les alertes détectées entre 2022 et 2024, mais l'algorithme sera bientôt adapté pour fonctionner sur une base mensuelle. Dans cette configuration entemps quasi réel, chaque nouvelle alerte est classée dans les 31jourssuivant sa détection et chaque mois, le modèle traite à nouveau les alertes des trois mois précédents sur une base continue. Les classifications ne sont affichées que si elles sont classées avec certitude par l'algorithme et les étiquettes peuvent être mises à jour si une classification plus fiable devient disponible. En conséquence, chaque alerte peut être reclassée jusqu'à trois fois, la couche affichant la classification la plus fiable. Ce processus permet d'améliorer la précision au fil du temps, à mesure que des images satellites et des informations contextuelles supplémentaires deviennent disponibles. En outre, certains facteurs (par exemple, l'exploitation forestière sélective et le développement routier) sont mieux identifiés peu après la perturbation de la canopée, tandis que d'autres (par exemple, l'exploitation minière ou l'agriculture) peuvent prendre plus de temps avant que leurs signaux caractéristiques n'apparaissent.</p>



<h2 class="wp-block-heading">Comment les données ont-elles été développées ?</h2>



<p>Les alertes de déforestation intégrées à une résolution de 10 mètres constituent la base de la classification des facteurs. Les alertes intégrées combinent la puissance des alertes de déforestation GLAD-L, GLAD-S2 et RADD pour fournir une vue plus rapide et plus fiable des perturbations forestières que n'importe quel système individuel.</p>



<p>Pour appliquer les classifications des facteurs à l'alerte intégrée de déforestation, nous avons utilisé un modèle d'apprentissage profond initialement développé pour le Suriname, la République du Congo et la République Démocratique du Congo. Le modèle a été entraîné à partir d'un vaste ensemble de données sur les perturbations forestières en Amazonie, au Congo et dans les forêts tropicales d'Asie du Sud-Est, en utilisant l'interprétation visuelle d'images Planet haute-résolution pour attribuer des étiquettes de classification. Le modèle utilise des images satellites Sentinel-1 et Sentinel-2 post-perturbation pour les classifications des facteurs, intégrant les atouts complémentaires des données radar et optiques des satellites. Cela signifie que le modèle peut classer rapidement les facteurs, même par temps nuageux.</p>



<p>Pour identifier les endroits où les incendies coïncidaient avec le défrichage de terres agricoles — une technique de gestion agricole couramment utilisée dans les régions tropicales —, nous avons utilisé les alertes incendie VIIRS et le ratio de combustion normalisée (informations qui identifient les zones brûlées) de Sentinel-2.</p>



<p>Afin de réduire le bruit dans les données, nous avons procédé à un lissage :</p>



<ul>
<li>Tout d'abord, les pixels d'alerte ne dépassant pas un niveau de confiance de 0,75 ont été attribués à des classifications incertaines et définis comme « non étiquetés ».</li>



<li>Ensuite, pour chaque zone d'alerte distincte (c'est-à-dire un groupe de pixels d'alerte connectés spatialement), le facteur dominant a été sélectionné et attribué à l'ensemble de la zone, à quelques exceptions près pour les facteurs se produisant à une échelle plus fine, au niveau des pixels (développement routier et exploitation forestière sélective) et pour les zones extrêmement vastes.</li>
</ul>



<p>La précision des classifications varie selon les régions, avec le bassin du Congo affichant les meilleurs résultats globaux. La plupart des erreurs de classification sont liées à la confusion entre l'agriculture et l'exploitation minière, qui se déroulent souvent dans la même zone et entre les inondations et les incendies de forêt dans les plaines inondables.</p>



<p>Pour en savoir plus sur la méthodologie et les limites, <a href="https://www.sciencedirect.com/science/article/pii/S0034425723002067" target="_blank" rel="noreferrer noopener">cliquez ici</a>.</p>



<h2 class="wp-block-heading">Comment ces données se comparent-elles à d'autres données indiquant les facteurs de perte ?</h2>



<p>D'autres ensembles de données provenant de GFW et de Land &amp; Carbon Lab fournissent des informations sur les facteurs de perturbation de la végétation, mais varient en termes de détail, de période couverte et de fréquence de mise à jour, et ont donc des applications différentes. Par exemple, les données annuelles sur les facteurs de perte de couvert forestier à 1 km disponibles sur GFW sont plus adaptées à une analyse historique sur plusieurs années, tandis que le système de classification des alertes de perturbation des terres de Land &amp; Carbon Lab fournit des informations sur les perturbations de la végétation à l'intérieur et à l'extérieur des forêts à l'échelle mondiale.</p>



<div class="flourish-embed flourish-table" data-src="visualisation/26207567?163031"><script src="https://public.flourish.studio/resources/embed.js"></script><noscript><img src="https://public.flourish.studio/visualisation/26207567/thumbnail" width="100%" alt="table visualization" /></noscript></div>



<p>Certains attributs rendent les ensembles de données plus adaptés à des applications spécifiques. Par exemple, une résolution plus élevée permet de mieux comprendre la co-occurrence des facteurs dans les zones locales, et les informations infra-annuelles permettent de mieux comprendre la saisonnalité des différents facteurs :</p>



<div class="flourish-embed flourish-table" data-src="visualisation/26207715?163031"><script src="https://public.flourish.studio/resources/embed.js"></script><noscript><img src="https://public.flourish.studio/visualisation/26207715/thumbnail" width="100%" alt="table visualization" /></noscript></div>



<h2 class="wp-block-heading">Comment ces données peuvent-elles être utilisées ? Une source vitale pour les protecteurs de la forêt</h2>



<p>Les alertes de déforestation sont utilisées dans le monde entier par des centaines de peuples autochtones et de communautés locales, ainsi que par des gardes forestiers et des ONG, pour enquêter sur d'éventuelles perturbations imprévues ou illégales dans les forêts.</p>



<p>Cependant, les différentes causes de perturbations nécessitent des réponses différentes. Par exemple, dans le cas d'une conversion agricole, les utilisateurs peuvent vouloir enquêter sur la perte afin d'évaluer la légalité du déboisement. Dans le cas d'une perturbation naturelle telle qu'un glissement de terrain, aucune enquête ou un autre type d'enquête serait nécessaire. Les utilisateurs peuvent se référer aux données relatives aux causes des alertes de déforestation afin de déterminer le type de réponse approprié en fonction de la cause de la perturbation.</p>



<p>Bien que des visites sur le terrain ou l'utilisation d'images satellite haute résolution pour vérifier les classifications puissent encore être nécessaires dans certains cas afin de planifier des mesures appropriées ou de bien comprendre le contexte, l'ajout d'informations sur les facteurs déterminants dans la base de données GFW permet de hiérarchiser les alertes et de mener des actions plus ciblées, ce qui permet d'économiser des ressources et de concentrer les enquêtes sur les pertes les plus préoccupantes.</p>



<p>Les données peuvent également être utilisées pour mieux comprendre les émissions de carbone résultant des perturbations forestières. Certains types de conversion éliminent toute la végétation du terrain, tandis que d'autres laissent des arbres sur place. En outre, des informations sur la cause de la perte sont nécessaires pour estimer les émissions résultant de l'utilisation ultérieure des terres après le déboisement. Elles indiquent également si une repousse est à prévoir, ce qui est probable après des activités d'exploitation forestière, mais pas après une conversion à l'agriculture permanente, par exemple.</p>



<h2 class="wp-block-heading">Quelle est la prochaine étape pour les données des facteurs déclencheurs des alertes de déforestation ?</h2>



<p><strong>Mises à jour mensuelles :</strong> les données de 2025 seront disponibles début 2026, après quoi des mises à jour mensuelles seront disponibles à partir de 2026.</p>



<p><strong>Extension à l'ensemble des régions tropicales :</strong> nous prévoyons prochainement d'étendre les données à l'ensemble de la couverture des alertes de déforestation intégrées (actuellement de 30 degrés nord à 30 degrés sud). À l'avenir, le modèle pourrait également être appliqué aux perturbations dans les forêts tempérées et boréales. Des recherches supplémentaires pourraient également permettre de classer les perturbations d'autres types de végétation à l'aide d'une approche similaire.</p>



<p><strong>Intégration avec d'autres outils GFW :</strong> nous prévoyons d'ajouter les données à Forest Watcher afin de les rendre plus accessibles aux protecteurs des forêts travaillant sur le terrain.</p>



<p>Restez à l'écoute pour plus d'informations, car nous continuons à améliorer les alertes de déforestation afin d'en élargir la couverture, l'accès et l'application. <a href="https://www.globalforestwatch.org/help/datasets/forest-change/">Consultez le Centre d'aide</a> pour plus d'informations sur l'utilisation des données. </p>



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

Découvrez pour la première fois ce qui déclenche les alertes de déforestation sur Global Forest Watch

New drivers of deforestation alerts data, now available on GFW, shows the causes of alerts across the tropics for the first time.

See What’s Driving Deforestation Alerts for the First Time on GFW

Manuka Khan

A New Satellite Data App Supports Better Monitoring of European Forests

A new European Forest App supports better monitoring of European forests by providing tailored information and insights in an easy-to-use platform.

A New Satellite Data App Supports Better Monitoring of European Forests | Data &amp; Tools | Global Forest Watch Blog

carbon

carbon | Global Forest Watch Blog

David Gibbs

Melissa Rose

What’s New With GFW’s Forest Carbon Monitoring