WRI Google Drivers of Tree Cover Loss (1km)

This data set shows the dominant driver of tree cover loss from 2001-2023 using the following five categories: * **Commodity-driven deforestation:** Large-scale deforestation linked primarily to commercial agricultural expansion. * **Shifting agriculture:** Temporary loss or permanent deforestation due to small- and medium-scale agriculture. * **Forestry:** Temporary loss from plantation and natural forest harvesting, with some deforestation of primary forests. * **Wildfire:** Temporary loss, does not include fire clearing for agriculture. * **Urbanization:** Deforestation for expansion of urban centers. The commodity-driven deforestation and urbanization categories represent permanent deforestation, while tree cover usually regrows in the other categories. The data were generated using decision tree models to separate each 10 km grid cell into one of the five categories. The decision trees were created using 4,699 sample grid cells, and use metrics derived from the Hansen tree cover, tree cover gain, and tree cover loss, NASA fires, global land cover, and population count. Separate decision trees were created for each driver and each region (North America, South America, Europe, Africa, Eurasia, Southeast Asia, Oceania), for a total of 35 decision trees. The final outputs were combined into a global map, which is then overlaid with tree cover loss data to indicate the intensity of loss associated with each driver around the world. All model code, reference samples, decision trees, and the final model are available in the Supplementary Materials of the paper. Shows the dominant driver of tree cover loss within each 10 km grid cell and the intensity of that loss for the time period 2001-2023.

- This data set, a collaboration between the Global Land Analysis & Discovery (GLAD) () lab at the University of Maryland (UMD), Google, USGS, and NASA, measures areas of tree cover loss across all global land (except - Antarctica and other Arctic islands) at approximately 30 × 30 meter resolution. The data were generated using multispectral satellite imagery from the Landsat 5 thematic mapper (TM), the Landsat 7 thematic mapper plus (ETM+), and the Landsat 8 Operational Land Imager (OLI) sensors. Over 1 million satellite images were processed and analyzed, including over 600,000 Landsat 7 images for the 2000-2012 interval, and more than 400,000 Landsat 5, 7, and 8 images for updates for the 2011-2022 interval, and additional images used for 2023, 2024 and 2025. The clear land surface observations in the satellite images were assembled and a supervised learning algorithm was applied to identify per pixel tree cover loss.  - In this data set, “tree cover” is defined as all vegetation greater than 5 meters in height, and may take the form of natural forests or plantations across a range of canopy densities. Tree cover loss is defined as “stand replacement disturbance” which is considered to be clearing of at least half of tree cover within a 30-meter pixel. The exact threshold is variable both through space and time, and is biome-dependent. Tree cover loss may be the result of human activities, including forestry practices such as timber harvesting or deforestation (the conversion of natural forest to other land uses), as well as natural causes such as disease or storm damage. Fire is another widespread cause of tree cover loss, and can be either natural or human-induced.  - This data set has been updated five times since its creation, and now includes loss up to 2025 (Version 1.13). The analysis method has been modified in numerous ways, including new data for the target year, re-processed data for previous years (2011 and 2012 for the Version 1.1 update, 2012 and 2013 for the Version 1.2 update, and 2014 for the Version 1.3 update), and improved modelling and calibration. These modifications improve change detection for 2011-2025, including better detection of boreal loss due to fire, smallholder rotation agriculture in tropical forests, selective losing, and short cycle plantations. Since the entire historical timeseries was not reprocessed with the updated methodology, time-series assessments should be performed with caution. Read more about the Version 1.13 update here:  and access on GEE here: .  - When zoomed out (< zoom level 13), pixels of loss are shaded according to the density of loss at the 30 x 30 meter scale. Pixels with darker shading represent areas with a higher concentration of tree cover loss, whereas pixels with lighter shading indicate a lower concentration of tree cover loss. There is no variation in pixel shading when the data is at full resolution (≥ zoom level 13).  - The tree cover canopy density of the displayed data varies according to the selection - use the legend on the map to change the minimum tree cover canopy density threshold.  Identifies areas of gross tree cover loss

This dataset enhances pantropical near-real-time forest disturbance monitoring systems by using a deep learning model to classify direct drivers of forest disturbances on a monthly basis. Produced by Wageningen University and Research (WUR), Laboratory of Geo-information Science and Remote Sensing, with support from Global Nature Watch (GNW), the dataset is built upon GNW's integrated disturbance alerts, which contains WUR's RADD (Radar for Detecting Deforestation) alerts and University of Maryland's GLAD (Global Land Analysis & Discovery) alerts, harnessing post-disturbance imagery from Sentinel-1 and Sentinel-2 satellites. Only high and highest confidence alerts are classified. This approach enables the identification of these direct causes driving deforestation: small-scale agriculture, large-scale agriculture, road development, selective logging, mining, wildfire, flooding, and other natural disturbances. The alert drivers offer more targeted enforcement of laws and regulations, improved estimations of ecological impacts, and better understanding of carbon emissions related to forest disturbances. Read the paper here: Rapid satellite monitoring reveals complexity of tropical tree cover loss drivers at fine scale | Research Square **Alert Driver class definitions** **Non-natural drivers (human-caused)** **Small-scale agriculture**: Clearings smaller than 2 ha, commonly related to shifting cultivation (temporary clearing for cultivation with later regrowth) or smallholder farming. In smallholder landscapes with mixed disturbances, this may include artisanal logging and fuelwood collection (typically burned for cooking).  **Small-scale agriculture with fire**: Clearings for small-scale agriculture (see above) where fire was likely used for the clearing, observed by a coinciding VIIRS fire alert and low post-disturbance Sentinel-2 Normalized Burn Ratio **Large-scale agriculture**: Clearings larger than 2 ha for the establishment of crops or pastures, commonly related to industrial agriculture (e.g. production of soy, palm oil, beef, etc.), clearcuts, and large-scale clearings for land speculation. **Large-scale agriculture with fire**: Clearings for large-scale agriculture (see above) where fire was likely used for the clearing, observed by a coinciding VIIRS fire alert and low post-disturbance Sentinel-2 Normalized Burn Ratio **Road development**: Clearings for the establishment of roads, commonly related to facilitating industrial timber harvests, but can include roads for any purpose.  **Selective logging**: Small-scale disturbances caused by selective tree felling and skidding (paths where felled logs are dragged or transported), commonly related to industrial timber harvests.  **Mining**: Forest clearing to facilitate artisanal and industrial mineral extraction. **Natural drivers** **Flooding**: Disturbances or clearings caused by floodings and meandering rivers. This includes both natural and human-induced flooding. **Other natural disturbances**: Disturbances or clearings without visible human-induced cause. This includes windthrows, droughts, landslides and naturally dying trees. **Other drivers** **Wildfire**: Large-scale disturbances due to fire, without immediate land clearing for agricultural activity. This includes both human-induced and naturally induced fires. This class excludes controlled fires used for agricultural clearing, but includes escaped wildfires caused by controlled fires. **Unlabeled**: Pixels where a confidence threshold for a prediction is not reached. The dataset covers the primary humid tropical forest (Turubanova et al. 2018) of the Amazon (Bolivia, Brazil, Colombia, Ecuador, French Guiana, Guyana, Peru, Suriname and Venezuela, restricted to the biogeographic outlines of the Amazon biome as defined by RAISG), Congo Basin (Cameroon, Central African Republic, Democratic Republic of the Congo, Equatorial Guinea, Gabon, Republic of the Congo) and Southeast Asia (Brunei, East- Timor, Indonesia, Malaysia and Papua New Guinea). The model was trained on a dataset of labeled disturbance alerts, interpreted with high-resolution Planet imagery across the Amazon, Congo Basin, and Southeast Asia. The model integrates the complementary strengths of radar (Sentinel-1) and optical (Sentinel-2) imagery, leveraging convolutional neural networks (CNNs) to classify disturbances. Even though the dataset initially displayed historical alert data (2022-2025), the classifications were applied in a simulated near-real-time monitoring scenario to classify monthly disturbance alerts into one of the specified classes, achieving an overall accuracy of 0.89 Macro-F1. Accuracy varies by region, with the Congo Basin performing best. After the addition of the DIST-ALERT product, there are more alerts and therefore, areas of misclassification may be larger. Most misclassifications stem from confusion between agriculture and mining, which often co-occur, and between flooding and wildfires. Historical drivers (2022-2025) were classified up to three times within the three months following each alert, with the most confident result retained. These 2022-2025 classifications were postprocessed by selecting per disturbance patch the most occurring driver class, leading to some initial classifications being overwritten by another class in postprocessing. The current method, running from January 2026 onwards, selects the most occurring driver class only for disturbance patches within the three-month windows, As a result, an alert can be updated up to 3 times, and the latest classification is retained. Classifications may therefore change for up to 4 months after the initial alert and do not change after that. . A class ‘unconfident’ is built from pixels that do not exceed a 0.5 class probability from the classification output. When the most occurring class for a disturbance patch is ‘unconfident’, this alerts remains unlabeled. Post-processing steps identified where fire coincided with the clearing of agricultural land, based on the presence of VIIRS fire alerts (within a 500 m buffer around the alert and a 6-month window) and a low post-disturbance normalized-burn ratio (lower than 0.0) in the following month’s Sentinen-2 composite. Smoothing was applied to all classes except road development and selective logging, which occur at finer scales. For patches smaller than 5 km, the dominant driver was assigned. For larger patches in the 2022-2025 data, pixel-level classifications were retained but smoothed with a 5 × 5-pixel majority filter. This step was applied to the whole time series; therefore, the historical product has an advantage in differentiation between small- and large-scale agriculture compared to the operational product. In a near-real-time setting, large clearings may initially be labeled as small-scale until they exceed the 2 ha threshold. In the future, monthly driver attributions for disturbance alerts will expand to new regions. The driver attributions will be continuously mapped and openly distributed via Global Nature Watch, as an extension to the integrated disturbance alerts. We also provide an annual version of this dataset, which incorporates some post processing – specifically the differentiation between large-scale and small-scale agriculture is made at the end of the year, meaning that agricultural clearings which are identified incrementally are identified based on their size at the end of the year. See the integrated disturbance alert layer metadata for additional details about the disturbance alerts. The alert drivers are available on Google Earth Engine with asset IDs: **Latest data** (ee.ImageCollection containing data from January 1, 2026 – end of last month): - projects/wurnrt-drivers/assets/operational_postprocessed_sa_collection - projects/wurnrt-drivers/assets/operational_postprocessed_afr_collection - projects/wurnrt-drivers/assets/operational_postprocessed_sea_collection *Note: this collection updates monthly, so new images will be added over time, but the collection path itself will stay the same.* **Historical archive** (ee.Image containing data from 2022–2025): - projects/wurnrt-drivers/assets/distribution/driverclassification_afr_202201_202512 - projects/wurnrt-drivers/assets/distribution/driverclassification_sa_202201_202512 - projects/wurnrt-drivers/assets/distribution/driverclassification_sea_202201_202512 Assigns probable drivers to the integrated disturbance alerts using machine learning and Sentinel imagery

This data is produced by the Global Land Analysis & Discovery (GLAD) lab at the University of Maryland (UMD) and measures areas of tree cover loss due to fire compared to all other drivers across all global land (except Antarctica and other Arctic islands) at approximately 30 × 30-meter resolution. The data were generated using global Landsat-based annual change detection metrics for 2001-2025 as input data to a set of regionally calibrated classification tree ensemble models. The result of the mapping process can be viewed as a set of binary maps (tree cover loss due to fire vs. tree cover loss due to all other drivers).  In this data set, “tree cover” is defined as all vegetation greater than 5 meters in height and may take the form of natural forests or plantations across a range of canopy densities. Tree cover loss is defined as “stand replacement disturbance” which is clearing of at least half of tree cover within a 30-meter pixel. The exact threshold is variable both through space and time and is biome-dependent. Tree cover loss due to fire may be caused by natural or human-induced fire activity.  The analysis method for the base tree cover loss map on GNW that is used as input for this dataset has been modified in numerous ways to improve detection of boreal loss due to fire, smallholder rotation agriculture in tropical forests, selective logging, and short cycle plantations for data covering the 2011-2025 period. Due to these changes, comparing trends across the 2000-2010 and 2011-2025 periods should be performed with caution. You can read more about updates to the tree cover loss modeling process here.  When zoomed out (< zoom level 13), pixels of loss are shaded according to the density of loss at the 30 x 30-meter scale. Pixels with darker shading represent areas with a higher concentration of tree cover loss, whereas pixels with lighter shading indicate a lower concentration of tree cover loss. There is no variation in pixel shading when the data is at full resolution (≥ zoom level 13).  The tree cover density of the displayed data varies according to the selection - use the legend on the map to change the minimum tree cover canopy density threshold.  This data is available for download from UMD here and also accessible through Google Earth Engine using the following image IDs (view the download page from UMD for more information):  - Fire certainty: users/sashatyu/2001-2025\_fire\_forest\_loss  - Date of loss: users/sashatyu/2001-2025\_fire\_forest\_loss\_annual  Identifies areas of tree cover loss due to fires compared to all other drivers of tree cover loss

Due to the increasing size and complexity of global forest monitoring data sources, analysis and interpretation tools for this data are ever more important for intervention efforts, allowing for the quick identification and interpretation of significant forest loss. The emerging hot spots data set identifies the most significant clusters of primary forest loss between 2002-2025 at a country level basis, on a tropical scale. The term ‘hot spot’ is defined as an area that exhibits statistically significant clustering in the spatial patterns of loss. In this analysis, observed patterns of primary forest loss are likely to be attributable to underlying, as opposed to random, spatial processes. The different categories of hot spots are described below:  New: A location that is a statistically significant hot spot only for the year 2025 and has never been a hot spot before.  Sporadic: A location that is an on-again then off-again hot spot. Less than 22 of the 24 years have been statistically significant hot spots.  Intensifying: A location that has been a statistically significant hot spot for more than 21 of the 24 years (>90%), including the most recent year (2025). In addition, the intensity of clustering of high counts in each year is increasing.  Persistent: A location that has been a statistically significant hot spot for more than 21 of the 24 years (>90%), with no discernible trend indicating an increase or decrease in the intensity of clustering over time.  Diminishing: A location that has been a statistically significant hot spot for more than 21 of the 24 years (>90%). In addition, the intensity of clustering of high counts in each year is decreasing, or the most recent year (2025) is not hot.  The emerging hot spots analysis uses the annual Hansen et al 2013 tree cover loss data set between the years 2002 – 2025, the Turubanova et al. 2018 primary forest extent data set for the year 2001, and the ESRI ArcGIS Emerging Hot Spot Analysis geoprocessing tool. In this analysis, primary forest is defined as mature natural humid tropical forest cover that has not been completely cleared and regrown in recent history. Forest loss is defined as ‘stand replacement disturbance,’ or the complete removal of tree cover canopy at the Landsat pixel scale. The emerging hot spots analysis tool uses a combination two statistical measures, the Getis-Ord Gi\* statistic to identify the location and degree of spatial clustering of forest loss, and the Mann-Kendall trend test to evaluate the temporal trend over time.  The forest loss data used in this analysis has a user’s accuracy of 87% and a producer’s accuracy of 83.1% across the tropical biome. Additionally, because this analysis was run for individual countries, results are relative to the patterns and amount of loss in each country. Results should not be directly compared between countries - please use caution when viewing layer at a global scale.  To identify statistically significant clusters of primary forest loss on a country level basis

This data set estimates agriculture-linked deforestation for oil palm, soy, cattle, cocoa and coffee annually for the years 2001-2015. While agriculture is generally recognized to be a major driver of deforestation, few studies have attempted to estimate the role that particular commodities play in global deforestation, and even fewer have been spatially explicit. In this analysis, we estimate the extent to which these commodities are replacing forests and map their impacts using the best available spatially explicit data. We report results globally at the second administrative level (e.g., county, municipality, or other administrative subdivision, depending on the country). To identify the specific commodities that have replaced forested land, we analyzed the overlap of current commodity extent with global annual tree cover loss from 2001 to 2018. We used recent, detailed crop maps for global oil palm and South American soy and supplemented with coarser resolution global data where needed for the other commodities and regions. Provides annual agriculture-linked deforestation estimates for oil palm, soy, cattle, cocoa and coffee for the years 2001-2015 by administrative boundary