In 2023, the deforested area in the Legal Amazon in Brazil amounted to approximately 802,300 hectares. Just a year earlier, the Amazon deforested area surpassed 1.2 million hectares. What is behind the growing Amazon deforestation in Brazil? Illegal logging, expansion of agricultural areas for soybean cultivation, and an increase in wildfire outbreaks are all among the leading causes of deforestation in the Brazilian Amazon. Politics, however, has also played an important role. For example, the authorized budget for Brazil’s Ministry of the Environment has been on a mostly downward trend since 2013, when it reached a decade-long peak of nearly seven billion Brazilian reals. How big is the Brazilian deforestation issue? In 2023, Brazil registered by far the largest area of primary forest loss in the world, amounting to more than one million hectares. This was roughly the same area as the remaining top nine countries combined. As the country with the second-largest forest area worldwide, these developments are cause for concern amidst the conversation on climate change mitigation. With the global tree cover loss annually increasing, and the emission of greenhouse gases from forest areas along with it, reaching net-zero emissions targets by 2050 grows ever more challenging.

Deforestation led to an annual loss of 4.4 million hectares of forest in Africa between 2015 and 2020. The conversion of forest to other land uses affected mostly the eastern and southern areas of the continent, at a deforestation rate of 2.2 million hectares per year. In Western and Central Africa, around 1.9 million hectares of forest were lost per year in the same period. Despite a small reduction observed in the period 2015-2020, the continental deforestation rate has overall increased since 1990. From that year until 2020, Africa has seen the greatest loss in forest area more than any region of the world.

Approximately *** million hectares of forest area were lost due to deforestation between 1990 and 2020. However, the deforestation rate has slowed over the last three decades, down from the total **** million hectares per year in 1990-2000 to **** million hectares per year in 2015-2020. The climatic domain that has shown the highest pace of deforestation is the tropical, which saw *** million hectares per year lost in 2015-2020. On the other hand, the boreal domain has the lowest deforestation rate, with a figure of **** million hectares per year in 2015-2020.

WWF developed a global analysis of the world's most important deforestation areas or deforestation fronts in 2015. This assessment was revised in 2020 as part of the WWF Deforestation Fronts Report.Emerging Hotspots analysisThe goal of this analysis was to assess the presence of deforestation fronts: areas where deforestation is significantly increasing and is threatening remaining forests. We selected the emerging hotspots analysis to assess spatio-temporal trends of deforestation in the pan-tropics.Spatial UnitWe selected hexagons as the spatial unit for the hotspots analysis for several reasons. They have a low perimeter-to-area ratio, straightforward neighbor relationships, and reduced distortion due to curvature of the earth. For the hexagon size we decided on a unit of 1,000 ha, based on the resolution of the deforestation data (250m) meant that we could aggregate several deforestation events inside units over time. Hexagons that are closer to or equal to the size of a deforestation event means there could only be one event before the forest is gone and limit statistical analysis.We processed over 13 million hexagons for this analysis and limited the emerging hotspots analysis to only hexagons with at least 15% forest cover remaining (from the all-evidence forest map). This prevented including hotspots in agricultural areas or where all forest has been converted.OutputsThis analysis uses the Getis-Ord and Mann-Kendall statistics to identify spatial clusters of deforestation which have a non-parametric significant trend across a time series. The spatial clusters are defined by the spatial unit and a temporal neighborhood parameter. We use a neighborhood parameter of 5km to include spatial neighbors in the hotspots assessment and time slices for each country described below. Deforestation events are summarized by a spatial unit (hexagons described below) and the results comprise a trends assessment which defines increasing or decreasing deforestation in the units determined at 3 different confidence intervals (90%, 95% and 99%) and the spatio-temporal analysis classifying areas into 8 hot unique or cold spot categories. Our analysis identified 7 hotspot categories:Hotspot TypeDefinitionNewA location with a statistically significant increasing hotspots only in the final time stepConsecutiveAn uninterrupted run of statistically significant hotspot in the final time-steps IntensifyingA statistically significant hotspot for >90% of the bins, including the final time stepPersistentA statistically significant hotspot for >90% of the bins with no upward or downward trend in clustering intensityDiminishingA statistically significant hotspot for >90% of the time steps, with where the clustering is decreasing, or the most recent time step is not hot.SporadicA on-again then off-again hotspot where <90% of the time-step intervals have been statistically significant hot spots and none have been statistically significant cold spots.HistoricalAt least ninety percent of the time-step intervals have been statistically significant hot spots, with the exception of the final time steps..For the evaluation of spatio-temporal trends of tropical deforestation we selected the Terra-i deforestation dataset to define the temporal deforestation patterns. Terra-i is a freely available monitoring system derived from the analysis of MODIS (NVDI) and TRMM (rainfall) data which are used to assess forest cover changes due to anthropic interventions at a 250 m resolution [ref]. It was first developed for Latin American countries in 2012, and then expanded to pan-tropical countries around the world. Terra-i has generated maps of vegetation loss every 16 days, since January 2004. This relatively high temporal resolution of twice monthly observations allows for a more detailed emerging hotspots analysis, increasing the number of time steps or bins available for assessing spatio-temporal patterns relative to annual datasets. Next, the spatial resolution of 250m is more relevant for detecting forest loss than changes in individual tree cover or canopies and is better adapted to process trends on large scales. Finally, the added value of the Terra-i algorithm is that it employs an additional neural network machine learning to identify vegetation loss that is due to anthropic causes as opposed to natural events or other causes. Our dataset comprised all Terra-i deforestation events observed between 2004 and 2017. Temporal unitThe temporal unit or time slice was selected for each country according to the distribution of data. The deforestation data comprised 16-day periods between 2004 and 2017 for a total of 312 potential observation time periods. These were aggregated to time bins to overcome any seasonality in the detection of deforestation events (due to clouds). The temporal unit is combined with the spatial parameter (i.e. 5km) to create the space-time bins for hotspot analysis. For dense time series or countries with a lot of deforestation events (i.e. Brazil) a smaller time slice was used (i.e. 3 months, n=54) with a neighborhood interval of 8 months, meaning that the previous year and next year together were combined to assess statistical trends relative to the global variables together. The rule we employed was that the time slice x neighborhood interval was equal to 24 months, or 2 years, in order to look at general trends over the entire time period and prevent the hotspots analysis from being biased to short time intervals of a few months.Deforestation FrontsFinally, using trends and hotpots we identify 24 major deforestation fronts, areas of significantly increasing deforestation and the focus of WWF's call for action to slow deforestation.

Annual forest loss in the Brazilian Amazon had in 2012 declined to less than 5,000 sqkm, from over 27,000 in 2004. Mounting empirical evidence suggests that changes in Brazilian law enforcement strategy and the related governance system may account for a large share of the overall success in curbing deforestation rates. At the same time, Brazil is experimenting with alternative approaches to compensate farmers for conservation actions through economic incentives, such as payments for environmental services, at various administrative levels. We develop a spatially explicit simulation model for deforestation decisions in response to policy incentives and disincentives. The model builds on elements of optimal enforcement theory and introduces the notion of imperfect payment contract enforcement in the context of avoided deforestation. We implement the simulations using official deforestation statistics and data collected from field-based forest law enforcement operations in the Amazon region. We show that a large-scale integration of payments with the existing regulatory enforcement strategy involves a tradeoff between the cost-effectiveness of forest conservation and landholder incomes. Introducing payments as a complementary policy measure increases policy implementation cost, reduces income losses for those hit hardest by law enforcement, and can provide additional income to some land users. The magnitude of the tradeoff varies in space, depending on deforestation patterns, conservation opportunity and enforcement costs. Enforcement effectiveness becomes a key determinant of efficiency in the overall policy mix.

1. Tropical forests are subject to diverse deforestation pressures while their conservation is essential to achieve global climate goals. Predicting the location of deforestation is challenging due to the complexity of the natural and human systems involved but accurate and timely forecasts could enable effective planning and on-the-ground enforcement practices to curb deforestation rates. New computer vision technologies based on deep learning can be applied to the increasing volume of Earth observation data to generate novel insights and make predictions with unprecedented accuracy.

2. Here, we demonstrate the ability of deep convolutional neural networks (CNNs) to learn spatiotemporal patterns of deforestation from a limited set of freely available global data layers, including multispectral satellite imagery, the Hansen maps of annual forest change (2001-2020) and the ALOS PALSAR digital surface model, to forecast deforestation (2021). We designed four model architectures, based on 2D CNNs, 3D CNNs, and Convolutional Long Short-Term Memory (ConvLSTM) Recurrent Neural Networks (RNNs), to produce spatial maps that indicate the risk to each forested pixel (~30 m) in the landscape of becoming deforested within the next year. They were trained and tested on data from two ~80,000 km² tropical forest regions in the Southern Peruvian Amazon.

3. The networks could predict the location of future forest loss to a high degree of accuracy (F₁ = 0.58-0.71). Our best performing model (3D CNN) had the highest pixel-wise accuracy (F₁ = 0.71) when validated on 2020 forest loss (2014-2019 training). Visual interpretation of the mapped forecasts indicated that the network could automatically discern the drivers of forest loss from the input data. For example, pixels around new access routes (e.g. roads) were assigned high risk whereas this was not the case for recent, concentrated natural loss events (e.g. remote landslides).

4. CNNs can harness limited time-series data to predict near-future deforestation patterns, an important step in harnessing the growing volume of satellite remote sensing data to curb global deforestation. The modelling framework can be readily applied to any tropical forest location and used by governments and conservation organisations to prevent deforestation and plan protected areas.

Amazon Annual Deforestation Rate: Roraima data was reported at 436.000 sq km in 2024. This records an increase from the previous number of 284.000 sq km for 2023. Amazon Annual Deforestation Rate: Roraima data is updated yearly, averaging 240.000 sq km from Dec 1988 (Median) to 2024, with 37 observations. The data reached an all-time high of 630.000 sq km in 1989 and a record low of 84.000 sq km in 2002. Amazon Annual Deforestation Rate: Roraima data remains active status in CEIC and is reported by National Institute for Space Research. The data is categorized under Brazil Premium Database’s Environmental, Social and Governance Sector – Table BR.EVC001: Amazon Annual Deforestation Rate. 2024 is preliminary data.

The zip file contains the map used for the statistics calculated in the publication 'A Continental Assessment of the Drivers of Tropical Deforestation With a Focus on Protected Areas', available at https://doi.org/10.3389/fcosc.2022.830248.

This is a first version of the drivers of forest loss map, It is an allocation of crowdsourced data on predominant drivers of forest loss (data available here https://doi.org/10.1038/s41597-022-01227-3) to the Hansen et al (2013) dataset of global tree loss (v 1.7), set to a resolution of 100x100 m.

Furthermore, the classes 'Other subsistence agriculture' and 'Shifting cultivation' are parts of the 'Subsistence agriculture' class on the original crowdsourced data, separated by extracting shifting agriculture according to Heinimann et al (2017).

The map is in Goode Homolosine projection.

Primary forest cover and forest cover loss in Wallacea for the years 2000-2018 to train a deforestation model and produce maps of projected probability of deforestation until 2053. Full details about this dataset can be found at https://doi.org/10.5285/c7148c20-c6b3-43e1-9f99-b6e38e4dfdaf

How to download this data:Click on "View Map" button at the top (the Download button allows you to download the footprint of tiles but not the actual alerts)Click on the tile where your area of interest is locatedCopy the whole URL from the pop-up and paste it into your internet browser. Download will begin automaticallyAdditional DetailsThis data set, assembled by Global Forest Watch, aggregates deforestation alerts from three alert systems (GLAD-L, GLAD-S2, RADD) into a single, integrated deforestation alert layer. This integration allows users to detect deforestation events faster than any single system alone, as the integrated layer is updated when any of the source alert systems are updated. The source alert systems are derived from satellites of varying spectral and spatial resolutions. 30m GLAD Landsat-based alerts are up-sampled to match the 10m spatial resolution of Sentinel-based alerts (GLAD-S2, RADD). This avoids the double counting of overlapping alerts, which are instead classified at a higher confidence level, indicated by darker pixels. Alerts are classified as high confidence when detected twice by a single alert system. Alerts detected by multiple alert systems are classified as highest confidence. With multiple sensors picking up change in the same location, we can be more confident that it was not a false positive and may not need to wait for additional satellite imagery to increase confidence in detected loss. *This data product utilizes a special encoding*Each pixel (alert) encodes the date of disturbance and confidence level in one integer value. The leading integer of the decimal representation is 2 for a low-confidence alert, 3 for a high-confidence alert, and 4 for an alert detected by multiple alert systems, followed by the number of days since December 31, 2014. 0 is the no-data value. For example:20001 is a low confidence alert on January 1st, 201530055 is a high confidence alert on February 24, 201521847 is a low confidence alert on January 21, 202041847 is a highest confidence alert (detected by multiple alert systems) on January 21, 2020. Alert date represents the earliest detection0 represents no alertResolution: 10 x 10mGeographic Coverage: 30°N to 30°SFrequency of Updates: DailyDate of Content: January 1st, 2015 – presentCautionsConfidence level may change retroactively as source data is updated GLAD-L: Available for entire tropics (30°N to 30°S) from January 1, 2018 to the present, and from 2015 to the present for select countries in the Amazon, Congo Basin, and insular Southeast Asia GLAD-S2: Available for the primary humid tropical forest areas of South America from January 2019 to the present RADD: Available for the primary humid tropical forest areas of South America, sub-Saharan Africa and insular Southeast Asia at a 10m spatial resolution, with coverage from January 2019 to the present for Africa and January 2020 to the present for South America and Southeast Asia In order to integrate the three alerting systems on a common grid, GLAD-L is resampled from 30m resolution to 10m resolution to match GLAD-S2 and RADD. As a result, pixels in the integrated layer may not exactly align with pixels in the individual GLAD-L layer. Each pixel in the integrated layer preserves the earliest date of detection from any alerting system, even if multiple systems have reported an alert in that pixel. In some situations, this may lead to inconsistent visualizations when switching from the integrated layer to individual alerting system layers. It is advisable to use in the integrated layer when you are interested in the earliest date of detection by any alerting system. However, it is better to use the individual alerting system layers if you are interested in a specific alert type. Although called ‘deforestation alerts’ these alerts detect forest or tree cover disturbances. This product does not distinguish between human-caused and other disturbance types. Where alerts are detected within plantation forests (more likely to happen in the GLAD-L system), alerts may indicate timber harvesting operations, without a conversion to a non-forest land use. The term deforestation is used because these are potential deforestation events, and alerts could be further investigated to determine this. LicenseCC by 4.0SourcesGLAD Alerts:Hansen, M.C., A. Krylov, A. Tyukavina, P.V. Potapov, S. Turubanova, B. Zutta, S. Ifo, B. Margono, F. Stolle, and R. Moore. 2016. Humid tropical forest disturbance alerts using Landsat data. Environmental Research Letters, 11 (3). GLAD-S2 Alerts:Pickens, A.H., Hansen, M.C., Adusei, B., and Potapov P. 2020. Sentinel-2 Forest Loss Alert. Global Land Analysis and Discovery (GLAD), University of Maryland. RADD Alerts:Reiche, J., Mullissa, A., Slagter, B., Gou, Y., Tsendbazar, N.E., Braun, C., Vollrath, A., Weisse, M.J., Stolle, F., Pickens, A., Donchyts, G., Clinton, N., Gorelick, N., Herold, M. 2021. Forest disturbance alerts for the Congo Basin using Sentinel-1. Environmental Research Letters. https://doi.org/10.1088/1748-9326/abd0a8

WWF developed a global analysis of the world's most important deforestation areas or deforestation fronts in 2015. This assessment was revised in 2020 as part of the WWF Deforestation Fronts Report.An Emerging Hotspot Analysis was used to derive deforestation fronts using the ArcGIS Emerging Hot Spot Analysis tool. This analysis was undertaken in 10km² hexagons, within country boundaries, based on the remote sensing data series from Terra-i data for Latin America, Africa, Asia and Oceania for the period from 2004 to 2017 for which validated data was available. Locations with the highest incidence of deforestation were selected in the tropical and subtropical biomes in each country. Following the hotspot analysis, a visual interpretation of the spatial clustering of deforestation hotspots in the selected biomes by country was conducted in order to delineate the boundaries of deforestation fronts, which comprise all countries in which deforestation hotspots were detected. Deforestation fronts are places that contain an important area of remaining forests where there is a relatively larger spatial concentration or clustering of deforestation hotspots (measured in 10km2 hexagons). As a result of this exercise, based on the Terra-i data, 30 countries were retained in the analysis.

The data in this repository is available under the Open Database License: http://opendatacommons.org/licenses/odbl/1.0/. Any rights in individual contents of the database are licensed under the Database Contents License: http://opendatacommons.org/licenses/dbcl/1.0/

This repository includes two datasets. The first is a collection of polygons covering mines globally and the associated forest cover loss from 2000 to 2019. The polygons were derived by merging the "global-scale mining polygons version 2" (Maus et al., 2022) and mining and quarry polygon features extracted from the OpenStreetMap database (OpenStreetMap contributors, 2017). To remove double counting of areas the overlaps between the datasets were resolved by uniting intersecting features into single polygon features, i.e. keeping only the external borders of intersecting features. A random visual check was conducted, and a few small manual editing of polygons was performed where errors were identified.

The resulting dataset is encoded as a Geopackage in the file 'global_mining_polygons.gpkg'. The GeoPackage includes a single layer with 192,584 entries called 'mining_polygons' with the following attributes:

• id unique feature identifier
• isoa3 ISO 3166-1 alpha-3 country codes
• country country names
• area area of the polygon in squared kilometres
• geom the geometry of the features in geographical coordinates WGS84

The second dataset provides annual time series of global tree cover loss within mines from 2000 to 2019, covering all polygons in the above dataset. The area of tree cover loss for each polygon was calculated from the Global Forest Change database (Hansen et al., 2013). Each polygon also has additional string attributes with biomes derived from Ecoregions 2017 ^{© Resolve} (Dinerstein et al., 2017) and the level of protection derived from The World Database on Protected Areas (UNEP-WCMC and IUCN, 2022).

This dataset is encoded in CSV format in the file 'global_mining_forest_loss.csv', which includes 416,412 entries and 53 variables, such that:

• id unique feature identifier
• id_hcluster unique feature identifier
• list_of_commodities a comma-separated list of commodities
• isoa3 ISO 3166-1 alpha-3 country codes
• country country names
• ecoregion ecoregion name
• biome biome name
• year the year
• area_forest_loss_XXX_YYY

The values of tree cover loss are disaggregated per initial percentage of tree cover (XXX) and per protection level (YYY).

• XXX can take one of:
  • 000: total tree cover loss independently from the initial tree cover
  • 025: tree cover loss on pixels with initial tree cover between 0 and 25%
  • 050: tree cover loss on pixels with initial tree cover between 25 and 50%
  • 075: tree cover loss on pixels with initial tree cover between 50 and 75%
  • 100: tree cover loss on pixels with initial tree cover between 75 and 100%
• YYY can take one of:
  • la: tree cover loss within strict nature reserve
  • Ib: tree cover loss within wilderness area
  • II: tree cover loss within national park
  • III: tree cover loss within natural monument or feature
  • IV: tree cover loss within habitat/species management area
  • V: tree cover loss within protected landscape/seascape
  • VI: tree cover loss within PA with sustainable use of natural resources
  • p: tree cover loss within any type of protection, including not applicable, not assigned, or not reported
  • none: when YYY is omitted, total tree cover loss within the polygon

For details about the protection levels definition see the UNEP-WCMC and IUCN (2022). The id can be used to link polygons to forest loss data.

Overview

This dataset includes the images (visible bands for Landsat-8 or NICFI PlanetScope), auxiliary data (infrared, NCEP, forest gain, OpenStreetMap, SRTM, GFW), and data about forest loss (Global Forest Change) used to train, validate and test a model to classify direct deforestation drivers in Cameroon.

Description of the files

• 'my_examples_landsat_final_detailed.zip': Landsat-8 images, auxiliary data and forest loss data used to train, validate and test a model for a detailed classification of deforestation drivers in Cameroon (15 classes: ‘Oil palm plantation’, ‘Timber plantation’, ‘Fruit plantation (e.g. banana)’, ‘Rubber plantation’, ‘Other large-scale plantation (e.g. tea, sugarcane)’, ‘Grassland/Shrubland’, ‘Small-scale oil palm plantation’, ‘Small-scale maize plantation’, ‘Other small-scale agriculture’, ‘Mining’, ‘Selective logging’, ‘Infrastructure’, ‘Wildfire’, ‘Hunting’, ‘Other’)
• 'my_examples_planet_final_detailed.zip': NICFI PlanetScope images, auxiliary data and forest loss data used to train, validate and test a model for a detailed classification of deforestation drivers in Cameroon (15 classes)
• 'my_examples_landsat_final.zip': Landsat-8 images, auxiliary data and forest loss data used to train, validate and test a model for a classification of deforestation drivers by groups in Cameroon (4 classes: 'Plantation', 'Grassland/Shrubland', 'Smallholder agriculture', 'Other')
• 'my_examples_planet_final.zip': NICFI PlanetScope images, auxiliary data and forest loss data used to train, validate and test a model for a classification of deforestation drivers by groups in Cameroon (4 classes)
• 'my_examples_landsat_detailed_timeseries.zip': Landsat-8 images, auxiliary data and forest loss data used to test a model for a detailed classification of deforestation drivers in Cameroon (15 classes) using multiple images and a time series analysis
• 'my_examples_planet_detailed_timeseries.zip': NICFI PlanetScope images, auxiliary data and forest loss data used to test a model for a detailed classification of deforestation drivers in Cameroon (15 classes) using multiple images and a time series analysis

• ‘labels.zip’: in csv files, the labels for each image in each folder described above (image identified by folder and coordinates or ‘path’) and matches the format of the csv files used as inputs to train, validate and test our classification model

  For ‘labels.zip’, we have subfolders for Landsat and PlanetScope. Then, for each type of imagery, we have subfolders for ‘detailed’, ‘groups’ and ‘time series’ which correspond to the different ‘my_examples’ folders listed above.

  For each folder, subfolders named with the coordinates of the centre of the images contain each:
  • A folder ‘images’, with a sub-folder ‘visible’ containing the PNG RGB image; and a sub-folder ‘infrared’ containing the infrared bands in a NPY file.
  • A folder ‘auxiliary’ with topographic and forest gain information in a NPY format, OpenStreetMap and peat data in a JSON format, and a sub-folder ‘ncep’ containing all data from NCEP in a NPY format.
  • The forest loss pickle file delimiting the area of forest loss.

Details about the images

• For Landsat-8 data (courtesy of the U.S. Geological Survey), this dataset contains 332x 332 pixels RGB calibrated top-of-atmosphere (TOA) reflectance images pan-sharpened to a 15 m resolution (less than 20% cloud cover)

• For NICFI PlanetScope data (catalog owner: Planet), this dataset contains 332x 332 pixels monthly RGB composite with a 4.77 m resolution

Details about the auxiliary data

• Forest gain from GFC: 30-m resolution, yearly data for 2000-2021, downloaded via Google Earth Engine
• Near infrared, shortwave infrared 1 and 2 bands from Landsat-8 TOA: 30-m resolution, data every 16 days for 2013-2023, downloaded via Google Earth Engine and selected using the same process as for Landsat-8 RGB images
• From NCEP Climate Forecast System Version 2 (CFSv2) 6-hourly Products: surface level albedo and volumetric soil moisture content (depths: 0.1 m, 0.4 m, 1.0 m, 2.0m) in 0.01%; radiative fluxes (clear-sky longwave flux downward and upward, clear-sky solar flux downward and upward, direct evaporation from bare soil, longwave and shortwave radiation flux downward and upward, latent, ground and sensible heat net flux), potential evaporation rate, and sublimation in W/m²; humidity (specific, maximum specific, minimum specific) in 10-4 kg/kg; ground level precipitation in 0.1 mm; air pressure at surface level in 10 Pa; wind level (u and v component) in 0.01 m/s, water runoff at surface level in 232.01 kg/ m²; temperature in K: 22264-m resolution, available four times a day for 2011-2023, downloaded directly from the NOAA website and selected the mean of the monthly mean over 5 years before the forest loss event, the monthly maximum over 5 years before the forest loss event, and the monthly minimum over 5 years before the forest loss event for each parameter
• Closest street and closest city from OpenStreetMap in km: directly downloaded with the Nominatim API
• Altitude in m, slope and aspect in 0.01° from Shuttle Radar Topography Mission (SRTM): 30-m resolution, measured for 2000, downloaded via Google Earth Engine
• Presence of peat from GFW: 232-m resolution, measured for 2017, directly downloaded on the GFW website

Details about Global Forest Change

For each image, there is a corresponding 'forest_loss_region' .pkl file delimiting a forest loss region polygon from Global Forest Change (GFC). GFC consists of annual maps of forest cover loss with a 30-m resolution.

License

The NICFI PlanetScope images fall under the same license as the NICFI data program license agreement (data in 'my_examples_planet_final.zip', 'my_examples_planet_final_detailed.zip', 'my_examples_planet_detailed_timeseries.zip': subfolders '[coordinates]'>'images'>'visible').

OpenStreetMap® is open data, licensed under the Open Data Commons Open Database License (ODbL) by the OpenStreetMap Foundation (OSMF) (data in all 'my_examples' folders: subfolders '[coordinates]'>'auxiliary'>'closest_city.json'/'closest_street.json'). The documentation is licensed under the Creative Commons Attribution-ShareAlike 2.0 license (CC BY-SA 2.0).

The rest of the data is under a Creative Commons Attribution 4.0 International License. The data has been transformed following the code that can be found via this link: https://github.com/aedebus/Cam-ForestNet (in 'prepare_files').

Amazon Annual Deforestation Rate: Tocantins data was reported at 23.000 sq km in 2024. This records a decrease from the previous number of 32.000 sq km for 2023. Amazon Annual Deforestation Rate: Tocantins data is updated yearly, averaging 124.000 sq km from Dec 1988 (Median) to 2024, with 37 observations. The data reached an all-time high of 1,650.000 sq km in 1988 and a record low of 23.000 sq km in 2024. Amazon Annual Deforestation Rate: Tocantins data remains active status in CEIC and is reported by National Institute for Space Research. The data is categorized under Brazil Premium Database’s Environmental, Social and Governance Sector – Table BR.EVC001: Amazon Annual Deforestation Rate. 2024 is preliminary data.

This repository includes the raster datasets which mask plantations—mainly oil palm and rubber—in the Hansen's data of forest loss during the period 2001–2015. The pixel values representing plantations are encoded as 1.The plantation mask covers the main production countries of oil palm (Malaysia, Nigeria, and Indonesia) and rubber (Indonesia, Thailand, Malaysia, and Vietnam) in 2000.

Probability of forest for 1900, 70% tree coverThis raster contains the median modelled probability of forest for the year 1900 and computed using a tree cover threshold of 70%.Proba_for_1900_70TC.tifProbability of forest for 2000, 70% tree coverThis raster contains the median modelled probability of forest for the year 2000 and computed using a tree cover threshold of 70%.Proba_for_2000_70TC.tifForest and savanna distribution for 1900, 70% tree coverThis raster contains the modelled distribution of forest (coded 1) and savanna (coded 0) for the year 1900, and using a probability threshold of 0.5 and for 70% tree cover threshold.Forest1900_thres05_70TC.tifForest and savanna distribution for 2000, 70% tree coverThis raster contains the modelled distribution of forest (coded 1) and savanna (coded 0) for the year 2000, and using a probability threshold of 0.5 and for 70% tree cover threshold.Forest2000_probthres05_70TC_UNC.tifForest changes between 1900 and 2000, 70% tree coverThis raster c...

From 1990 and up until 2010, South America was the region in the world with the highest rate of forest loss, with an estimated 5.2 million hectares of net forest lost per year in the first decade of this century. Since then, the destruction of South American forests has slowed down to an average of 2.6 million hectares per year, the second largest forest loss rate in the world after Africa. The figures suggest that, despite reforestation efforts, forest areas in South America continue to be endangered by massive deforestation and wildfires.

OverviewThis 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.CautionsThis analysis is limited by various data and attribution issues and methodological assumptions, including the following:Commodity data sets have limited coverage and quality. Only oil palm has recent, detailed maps of extent at a global level. The analysis also uses detailed data on South American soy. Outside of these regions and commodities, the analysis relies on global 10-kilometer resolution data on crop and pasture extent. These data are from 2010 (2000 for pasture), so the amount of forest replaced by a specific commodity is assumed to be proportional to its area during that year and may be misrepresented if significant expansion or contraction of that commodity has occurred since then. While Goldman et al. (2020) presents results using detailed pasture data for Brazil, this data set includes pasture results for the coarse method only.The data cannot capture complex land-use change transitions. The analysis does not consider other possible land uses between the deforestation event and the establishment of the commodity. The analysis also does not consider any forms of indirect land-use change (e.g., the target commodity displacing other activities that may, in turn, expand into forested areas).The data measure tree cover loss rather than deforestation directly. All tree cover loss in an area later used for one of the target commodities is assumed to be deforestation because forest replaced with a crop or pasture represents a permanent land-use change. Historical data from Indonesia and Malaysia were used to filter out older oil palm plantations from the analysis to avoid counting old, unproductive oil palm trees being felled as tree cover loss.The data may miss some forms of tree cover loss. The Hansen et al. (2013) tree cover loss data may not detect all changes related to commodity production. Much of the production of cocoa and coffee occurs on very small farms (less than one hectare) that may not be captured by the tree cover loss data. The analysis may also underestimate the conversion of dry forest and woody savanna areas, which are not well represented in the tree cover loss data. For the detailed soy analysis, we define tree cover as any woody vegetation with a minimum of 10 percent canopy cover (analyses for other commodities use 30 percent) to minimize underestimations in South American biomes such as the Cerrado and the Chaco.Further discussion about the methods, assumptions, and limitations of this analysis is available in Goldman et al. (2020).CitationGoldman, E., M.J. Weisse, N. Harris, and M. Schneider. 2020. “Estimating the Role of Seven Commodities in Agriculture-Linked Deforestation: Oil Palm, Soy, Cattle, Wood Fiber, Cocoa, Coffee, and Rubber.” Technical Note. Washington, DC: World Resources Institute. Available online at: wri.org/publication/estimating-the-role-of-sevencommodities-in- agriculture-linked-deforestationLicenseCreative Commons Attribution 4.0 International License (CC-BY 4.0)

This data set, a collaboration between the GLAD (Global Land Analysis & Discovery) lab at the University of Maryland, 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. 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,” or the complete removal of tree cover canopy at the Landsat pixel scale. 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 2022 (Version 1.10). 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 2011-2022, including better detection of boreal loss due to fire, smallholder rotation agriculture in tropical forests, selective losing, and short cycle plantations. Eventually, a future “Version 2.0” will include reprocessing for 2000-2010 data, but in the meantime integrated use of the original data and Version 1.7 should be performed with caution. Read more about the Version 1.7 update 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.

This is the replication package (data and code) that accompanies the article "DETERring Deforestation in the Amazon: Environmental Monitoring and Law Enforcement". In the article, we study Brazil's recent use of satellite technology to overcome law enforcement shortcomings resulting from weak institutional environments. DETER is a system that processes satellite imagery and issues near-real-time deforestation alerts to target environmental enforcement in the Amazon. We propose a novel instrumental variable approach for estimating enforcement's impact on deforestation. Clouds limiting DETER's capacity to detect clearings serve as a source of exogenous variation for the presence of environmental authorities. Findings indicate that monitoring and enforcement effectively curbed deforestation. Results are not driven by the displacement of illegal activity intoneighboring areas, and hold across several robustness checks.