OverviewThis 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 2011-2020 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 2020 (Version 1.8). 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 2011-2020, including better detection of boreal loss due to fire, smallholder rotation agriculture in tropical forests, selective logging, 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 1.8 should be performed with caution. Read more about the Version 1.8 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.Frequency of updates: AnnualDate of content: 2001-2020Resolution: 30x30m

SpatioTemporal Asset Catalog (STAC) Item - hansen-gfc-2023-v1.11-80N-140E in glad-global-forest-change-1.11

Results from time-series analysis of Landsat images in characterizing global forest extent and change. The 'first' and 'last' bands are reference multispectral imagery from the first and last available years for Landsat spectral bands corresponding to red, NIR, SWIR1, and SWIR2. Reference composite imagery represents median observations from a set of quality-assessed growing-season observations for each of these bands. Please see the User Notes for this update, as well as the associated journal article: Hansen, Potapov, Moore, Hancher et al. "High-resolution global maps of 21st-century forest cover change." Science 342.6160 (2013): 850-853.

The amount of tree cover around forest remnants was calculated using 30-m-resolution global maps of forest extent and change from 2000 through 2023, version 1.11 (Hansen et al. 2013). Each pixel value represents the percentage of canopy closure for all vegetation taller than 5 m in height for the year 2000, which was classified as either forest (≥ 50% tree cover) or non-forest (Hasui et al. 2024). Therefore, such binarization is used loosely, as tree cover can comprise both natural forests and tree plantations (e.g., monocultures of oil palm, rubber, or eucalypt) (Tropek et al. 2014). Despite this, tree cover mostly represents natural forests, and tree plantations are a high-quality matrix type for forest species, given the structural similarity between the habitat (forest) and the matrix (tree plantation) (Prevedello & Vieira 2010). When bird surveys were conducted before 2002, we used the tree cover map for the year 2000. For surveys conducted from 2002 to 2017, we updated the tree cover map by subtracting the cumulative forest loss from 2001 up to the year prior to each survey’s first year (Moulatlet et al. 2021).

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.

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.

Results from time-series analysis of Landsat images in characterizing global forest extent and change from 2000 through 2015. For additional information about these results, please see the associated journal article (Hansen et al., Science 2013).

OverviewThis emissions layer is part of the forest carbon flux model described in Harris et al. (2021). This paper introduces a geospatial monitoring framework for estimating global forest carbon fluxes which can assist a variety of actors and organizations with tracking greenhouse gas fluxes from forests and in decreasing emissions or increasing removals by forests. Forest carbon emissions represent the greenhouse gas emissions arising from stand-replacing forest disturbances that occurred in each modeled year (megagrams CO2e emissions/ha, between 2001 and 2024). Emissions include all relevant ecosystem carbon pools (aboveground biomass, belowground biomass, dead wood, litter, soil organic carbon) and greenhouse gases (CO2, CH4, N2O). Emissions estimates for each pixel are calculated following IPCC Guidelines for national greenhouse gas inventories where stand-replacing disturbance occurred, as mapped in the Global Forest Change annual tree cover loss data of Hansen et al. (2013). The carbon emitted from each pixel is based on carbon densities in 2000, with adjustment for carbon accumulated between 2000 and the year of disturbance.Emissions reflect a gross estimate, i.e., carbon removals from subsequent regrowth are not included. Instead, gross carbon removals resulting from subsequent regrowth after clearing are accounted for in the companion forest carbon removals layer. The fraction of carbon emitted from each pixel upon disturbance (emission factor) is affected by several factors, including the direct driver of disturbance, whether fire was observed in the year of or preceding the observed disturbance event, whether the disturbance occurred on peat, and more. All emissions are assumed to occur in the year of disturbance. Emissions can be assigned to a specific year using the Hansen tree cover loss data; separate rasters for emissions for each year are not available from GFW. All input layers were resampled to a common resolution of 0.00025 × 0.00025 degrees each to match Hansen et al. (2013).We have made several updates to the model since its original release. For documentation through the current version, please refer to this blog. For a more detailed description of the changes included through the 2023 tree cover loss launch (released spring 2024) and a comparison of the model's fluxes with those from the Global Carbon Budget and national greenhouse gas inventories, please refer to this article.Three variations of emissions rasters are available for download:megagrams CO2e emissions/ha in pixels with >30% tree cover density (TCD) in 2000 or tree cover gain: Used for visualizing (mapping) emissions according to the default GFW TCD threshold because it represents the density of emissions per hectare. You would use this if you want to only include emissions in pixels that are more conservatively defined as forest.megagrams CO2e emissions/pixel in pixels with >30% TCD in 2000 or tree cover gain: Used for calculating the emissions in an area of interest (AOI) according to the default GFW TCD threshold because the values of the pixels in the AOI can be summed to obtain the total emissions for that area. You would use this if you want to only include emissions in pixels that are more conservatively defined as forest.megagrams CO2e emissions/pixel in pixels with any amount of tree cover in 2000 or tree cover gain: Used for calculating the emissions in an area of interest (AOI) without any TCD threshold because the values of the pixels in the AOI can be summed to obtain the total emissions for that area. This would represent the total emissions from tree cover loss in the AOI without applying a TCD threshold. You would use this if you want to include emissions in pixels that have low (<30%) TCD in 2000.The values in the megagrams CO2e/pixel layers were calculated by adjusting the emissions per hectare by the size of each pixel, which varies by latitude. Tree cover density in 2000 is according to Hansen et al. (2013) and tree cover gain between 2000 to 2020 is according to Potapov et al. (2022)Related Open Data Portal layers: Forest Carbon Removals, Net Forest Carbon FluxGoogle Earth Engine: asset (megagrams CO2e emissions/ha in pixels with >30% TCD) and visualization scriptResolution: 30 x 30mGeographic Coverage: GlobalFrequency of Updates: AnnualDate of Content: 2001-2024CautionsData are the product of modeling and thus have an inherent degree of error and uncertainty. Users are strongly encouraged to read and fully comprehend the metadata and other available documentation prior to data use.Values are applicable to forest areas only (canopy cover >30 percent and >5 m height or areas with tree cover gain). See Harris et al. (2021) for further information on the forest definition used in the analysis.Although emissions in each pixel are associated with a specific year of disturbance, emissions over an area of interest reflect the total over the model period of 2001-2024. Thus, values must be divided by 24 to calculate average annual emissions.Emissions reflect stand-replacing disturbances as observed in Landsat satellite imagery and do not include emissions from unobserved forest degradation.Emissions reflect a gross estimate, i.e., carbon removals from any regrowth that occurs after disturbance are not included. Instead, gross carbon removals are accounted for in the companion forest carbon removals layer.Emissions data contain temporal inconsistencies. Improvements in the detection of tree cover loss due to the incorporation of new satellite data and methodology changes between 2011 and 2015 may result in higher estimates of emissions in recent years compared to earlier years. Refer here for additional information.Forest carbon emissions do not reflect carbon transfers from ecosystem carbon pools to the harvested wood products (HWP) pool.This dataset has been updated since its original publication. See Overview for more information.

This dataset provides a comparison of forest extent agreement from seven remote sensing-based products across Mexico. These satellite-derived products include European Space Agency 2020 Land Cover Map for Mexico (ESA), Globeland30 2020 (Globeland30), Commission for Environmental Cooperation 2015 Land Cover Map (CEC), Impact Observatory 2020 Land Cover Map (IO), NAIP Trained Mean Percent Cover Map (NEX-TC), Global Land Analysis and Discovery Global 2010 Tree Cover (Hansen-TC), and Global Forest Cover Change Tree Cover 30 m Global (GFCC-TC). All products included data at 10-30 m resolution and represented the state of forest or tree cover from 2010 to 2020. These seven products were chosen based on: a) feedback from end-users in Mexico; b) availability and FAIR (findable, accessible, interoperable, and replicable) data principles; and c) products representing different methodological approaches from global to regional scales. The combined agreement map documents forest cover for each satellite-derived product at 30-m resolution across Mexico. The data are in cloud optimized GeoTIFF format and cover the period 2010-2020. A shapefile is included that outlines Mexico mainland areas.

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

The Intact Forest Landscapes (IFL) data set identifies unbroken expanses of natural ecosystems within the zone of forest extent that show no signs of significant human activity and are large enough that all native biodiversity, including viable populations of wide-ranging species, could be maintained. To map IFL areas, a set of criteria was developed and designed to be globally applicable and easily replicable, the latter to allow for repeated assessments over time as well as verification. IFL areas were defined as unfragmented landscapes, at least 50,000 hectares in size, and with a minimum width of 10 kilometers. These were then mapped from Landsat satellite imagery for the year 2000.Changes in the extent of IFLs were identified from 2000-2013 and from 2013-2016 within the original year 2000 IFL boundary using the global wall-to-wall Landsat image composite for years 2013, 2016, and the global forest cover loss dataset (Hansen et al., 2013). Areas identified as “reduction in extent” met the IFL criteria in 2000, but no longer met the criteria in 2016. The main causes of change were clearing for agriculture and tree plantations, industrial activity such as logging and mining, fragmentation due to infrastructure and new roads, and fires assumed to be caused by humans.This data can be used to assess forest intactness, alteration, and degradation at global and regional scales.

Background

Human-induced land use change (LUC), driven by activities such as forestry, logging, and the production of agricultural commodities (e.g. fruits, nuts, and meat) significantly impacts the Global Commons, encompassing the climate system, ice sheets, land biosphere, oceans, and the ozone layer. The convertion of natural forests into areas dedicated to these activities lead to disrupted ecosystems (Foley et al. 2005), severely degraded biodiversity (Newbold et al. 2015), and the release of substantial amounts of greenhouse gases (GHGs) into the atmosphere (Hong et al. 2021), further exacerbating climate change and ocean acidification (Doney et al. 2009). The expansion of the agricultural frontier is identified as the predominant direct cause of deforestation globally, with other industries like timber and mining also playing significant roles (Curtis et al. 2018). To achieve global climate targets, forestry, and other land use GHG emissions must decrease along a nonlinear trajectory and reach carbon neutrality by 2050 (Rockström et al. 2017). However, to successfully address this road map, improving our understanding of deforestation drivers is urgently needed.

Summary

This dataset is the result of data processing performed to estimate the extent to which commodities and other agricultural products have replaced forests, while mapping the CO2 emission impact making use of the best available spatially explicit data. Results are reported globally for 52 products at national level, as well as agroecological and thermal zones (FAO & IIASA) and a 50km cell vector grid.

In order to detect spatially-explicit deforestation drivers, the current extent of commodities and agricultural products was overlapped with global annual tree cover loss in the 10-year period from 2014 to 2023. Carbon stocks in the deforested areas were then assumed to have been emmited into the atmosphere. Recent, detailed crop and pasture maps for relevant commodities were used whenever available, and coarser resolution datasets were used as supplements when needed. Operations were performed in Google Earth Engine.

Datasets used

Forest and biomass carbon distribution

The Global Forest Change dataset (Hansen et al., 2013) is used to estimate deforestation between 2014 and 2023. This tree cover loss dataset measures the first instance of complete removal of tree cover canopy at a 30-meter resolution for all woody vegetation over 5 meters in height.

The WCMC Above and Below Ground Biomass Carbon Density (Soto-Navarro et al., 2020), for reference year 2010 at 300m pixel, is overlapped with resulting deforested areas pixels to dermine the biomass carbon present in the areas before deforestation.

Generalized deforestation drivers

Tree cover loss by dominant driver (Curtis et al., 2022) in 2023 is used to determine wide categories of deforestation drivers (commodities, shifting agriculture, forestry, wildfire and urbanization). Pixels indicating deforestation in the Global Forest Change dataset (Hansen et al., 2013) that overlap the commodities and shifting agriculture pixels from this dataset (Curtis et al., 2022) have their drivers further detailed with the data sources listed in the below.

EarthStat pasture areas layer (Ramankutty et al., 2008) is used to identify areas for which specific livestock categories are to be defined. The project provides pasture areas for reference year 2000 at ~10km resolution.

Detailed deforestation drivers

The Group on Earth Observations Global Agricultural Monitoring (GEOGLAM) commodity distibution layer (Becker-Reshef et al., 2023) is used to identify specific commodities (winter wheat, spring wheat, maize, rice and soybean) to deforestation pixels pertaining to the "commodities" class. The ressource provides commodity distribution mapping at 5km pixel resolution. Values are provided as percentage of pixel area occupied by given crop.

The Spatial Production Allocation Model (SPAM) physical area layer (You et al., 2014) for reference year 2020 is used to detail drivers pertaining to the "shifting agriculture" class. The dataset covers 46 crops and crop groups at ~9km pixel resolution. Values are provided as percentage of pixel area occupied by given crop or crop group.

The Gridded Livestock of the World (GLW3) (Gilbert et al., 2022) is used to determine which species (cattle, goat, sheep or horse) of livestock is raised in areas identified as pasture in the EarthStat layer and pertaining to the "commodities" class. The project provides livestock distribution for reference year 2015 at ~9km resolution. Values are provided as number of individuals located within the pixel. Values were converted into percentage of pixel area covered by grazing field for given species based on species density thresholds.

Data processing

Most of data processing takes place in Google Earth Engine, with scripts redacted in javascript. In summary, two strategies were implemented:

Proportional driver distribution strategy: When deforestation pixels (Hansen et al., 2013) overlapped with pixels from at least one of the detailed deforestation drivers data sources, the driver describe in the latter were associated with that deforested area. Whenever more than one of these data sources had non-null pixels overlapping the area, a proportional distribution was assumed (i.e. if SPAM indicated 100% of the area to be covered by cowpea crops, GEOGLAM 100% by maize, and GLW3 100% by cattle grazing fields, the pixel is assumed to have 33.3% of its deforested area associated with each of these drivers).

Main driver strategy: When deforestation pixels did not overlap with any non-null pixels from any of the detailed drivers sources, the pixel is assumed to have the entirety of its deforested area associated with one single main driver resulting from a crop-livestock mosaic. The mosaic is created by taking the highest value from each of the crop or livestock distribution rasters, and then assigning the raster category to be the new pixel value, ultimately creating a category raster layer containing the main crop, crop group or livestock species occupying that pixel area. Null or zero values in this mosaic are filled-in by nearest neighbour analysis, to a limit of 20 pixels expansion. This was enough to ensure that all deforestation pixels had at least one detailed driver with which it could be associated. The logic behind this operation resides in the fact that the deforestation layer (Hansen et al., 2013) has a larger temporal coverage (with the more recent data point being the reference year 2023), while the detailed driver layers can be as old as reference year 2015. This means we're assuming the main deforestation drivers continued to expand their limits to neighbouring areas during the years for which no data is available.

Resulting rasters from both strategies are put together and a zonal statistics operation is performed in order to populate the vector grid cells.

Files

This repository contains the following files:

deforested_area_by_LUC_driver_2014_2023.CSV contains the deforested area (hectares) and the corresponding driver in each grid cell (idenfied by the id field) in each year, in CSV text format.

carbon_emissions_by_LUC_driver_2014_2023.CSV contains the carbon emitted (Mg CO2 eq.) and the corresponding driver in each grid cell (idenfied by the id field) in each year, in CSV text format.

spatial_grid.gpkg contains the raw 50km cell grid, with identification of country (iso3 and name fields), region, and FAO agroecological zone (zone field) and thermal zone (thermal field), in Geopackage format. In order to visualize the data in a map, the user will need to join one of the csv files to this geopackage file by basing the join on the 'id' field.

summary_showcase.png is an image showcasing maps created using the database, as well as a diagram showing the datasets used to create the final dataset.

How to cite

Iablonovski, G.; Berthet, E. C.; Roberts, S. (2024). Yearly CO2 emissions from anthropogenic land use change by main driver (2014-2023) [Data set]. Zenodo. https://zenodo.org/doi/10.5281/zenodo.13308514

Authors and contact

Authors: Guilherme Iablonovski*, Etienne Charles Berthet, Sophie Roberts

*Corresponding author: Guilherme Iablonovski (guilherme.iablonovski@unsdsn.org)

The Cordillera Administrative Region (CAR) in the Philippines is among the last forest frontiers in the country and is also home to 13 major watersheds in Northern Luzon that supply irrigation and hydroelectricity to other regions. However, it is faced with the deterioration of the quality of its watersheds due to forest loss driven mainly by agricultural expansion and illegal logging. Thus, this study was conducted to analyze the spatial and temporal patterns of forest loss that could serve as a basis for policy decisions. Also, this paper determined the strength of relationships using Pearson’s correlation coefficient (r) between forest loss and seven independent variables, which includes forest cover, agricultural areas, built-up, road network, and socio-economic data. This study utilized the Hansen Global Forest Change (HGFC), a Landsat-derived dataset from 2001 to 2019. Results revealed that 70,925 hectares (ha) of forest loss were detected with an annual deforestation rate of 3,74...

The loaded dataset in Zenodo includes six parts:

(1) Shp file of the Congo Basin;

(2) Shp file of the validation points for evergreen forest cover map in 2000 (reference value '1' means forest, and '0' means non-forest);

(3) Shp file of the validation points for the forest loss and non-loss during 2001-2020, and it also includes the forest loss year (reference value '1' means forest loss, and '0' means non-loss; loss_year values of '1,2, ..., 20' mean the year of '2001, 2002, ..., 2020');

(4) GeoTIFF image of the evergreen forest cover map in the Congo Basin (file name 'Congo_EvergreenForestCoverMap_2000');

(5) GeoTIFF image of the annual forest loss map generated by Hansen's Global Forest Change product [1] (file name 'Congo_Hansen_GFC_annualForestLoss_2001to2020');

(6) GeoTIFF image of the annual forest loss map produced by the proposed method in the Science Advances paper (file name 'Congo_Proposed_annualForestLoss_2001to2020').

[1] M. C. Hansen et al., "High-Resolution Global Maps of 21st-Century Forest Cover Change," Science, vol. 342, no. 6160, pp. 850-853, 2013/11/15 2013.

Abstract: NetCDF files of the forest cover fraction, vegetation cover fraction and fraction of non-vegetated area on 250m grid resolution sinusoidal grid generated at ICDC from the original HDF files (see https://www.cen.uni-hamburg.de/en/icdc/data/land/modis-vcf-forest.html and https://doi.org/10.25592/uhhfdm.12921) obtained from https://lpdaac.usgs.gov/mod44bv061/ are used to compute globally gridded maps of these parameters at 0.5 degree grid resolution on an equi-rectangular climate modeling grid (CMG). The global maps contain the grid-cell mean fractions of the three mentioned parameters, their variance within the grid cells, and - for the forest cover fraction - the grid-cell mean standard deviation. In addition, the data set includes maps of the number of valid forest cover fraction values at 250 m resolution per 0.5 degree grid cell, a grid cell mean quality flag and fractions of the two most abundant quality flags (primary and secondary). Generally all valid data are used; the user is advised to check the quality flags to eventually discard data of low quality.

TableOfContents: grid cell mean forest cover fraction; grid cell mean forest cover fraction standard deviation; forest cover fraction variance within grid cell; grid cell mean vegetation cover fraction; vegetation cover fraction variance within grid cell; non-vegetated area cover fraction; non-vegetated area cover fraction variance within grid cell; number of useful vegetation cover fraction values per grid cell; grid cell mean quality flag; primary quality flag fraction; secondary quality flag fraction

Technical Info: dimension: 720 columns x 360 rows x unlimited; temporalExtent_startDate: 2022-03-06; temporalExtent_endDate: 2023-03-05; temporalResolution: yearly; spatialResolution: 0.5; spatialResolutionUnit: degrees; horizontalResolutionXdirection: 0.5; horizontalResolutionXdirectionUnit: degrees; horizontalResolutionYdirection: 0.5; horizontalResolutionYdirectionUnit: degrees; verticalResolution: none; verticalResolutionUnit: none; verticalStart: none; verticalEnd: none; instrumentName: MODerate Resolution Spectroradiometer (MODIS); instrumentType: visible_to_infrared_spectroradiometer; instrumentLocation: Earth Observation Satellite (EOS) Terra; instrumentProvider: NOAA/NASA

Methods: [1] https://lpdaac.usgs.gov/products/mod44bv061/; [2] Townshend, J., et al., User Guide for the MODIS Vegetation Continuous Fields product Collection 6.1, verison 1, https://lpdaac.usgs.gov/documents/1494/MOD44B_User_Guide_V61pdf; [3] Algorithm Theoretical Basis Document (ATBD), https://lpdaac.usgs.gov/documents/113/MOD44B_ATBD.pdf; [4] Carroll, M., et al., 2011. Vegetative Cover Conversion and Vegetation Continuous Fields. In: Ramachandran, B., C. O. Justice, and M. Abrams (eds.), Land Remote Sensing and Global Environment Change: NASA's Earth Observing System and the Science of ASTER and MODIS. Springer Verlag.; [5] Hansen, M., et al., 2005. Estimation of tree cover using MODIS data at global, continental and regional/local scales. Int. J. Rem. Sens., 26(19), 4359-4380.

Units: Units for all variables (see TableOfContents): percent; percent; 1; percent; 1; percent; 1; 1; 1; 1; 1

geoLocations: westBoundLongitude: -180.0 degrees East; eastBoundLongitude: 180.0 degrees East; southBoundLatitude: -90.0 degrees North; northBoundLatitude: 90.0 degrees North; geoLocationPlace: global on land

Size: 1 file, ~5.2 Mb

Format: netCDF

DataSources:

Original data on sinusoidal grid tiles in hdf-format: https://doi.org/10.5067/MODIS/MOD44B.061 (last accessed 2023-06-23), see also https://lpdaac.usgs.gov/products/mod44bv061/ (last accessed: 2023-06-23)

Reprocessed data on sinusoidal grid tiles in netCDF format: https://doi.org/10.25592/uhhfdm.12921 (last access 2023-07-14), see also https://www.cen.uni-hamburg.de/en/icdc/data/land/modis-vcf-forest.html (last accessed 2023-07-14)

Contact: stefan.kern (at) uni-hamburg.de

Web page: https://www.cen.uni-hamburg.de/en/icdc/data/land/modis-vcf-forest.html

We used the Global Forest Change dataset (Hansen et al. 2003) to group loss pixels by year of loss, and restricted possible timber clearings to those ≥3.0 ac in size. Resulting polygons are assigned a year of harvest between 2001 and 2018.

Forest carbon emissions represent the greenhouse gas emissions arising from stand-replacing forest disturbances that occurred in each modeled year (megagrams CO2 emissions/ha, between 2001 and 2020). Emissions include all relevant ecosystem carbon pools (aboveground biomass, belowground biomass, dead wood, litter, soil) and greenhouse gases (CO2, CH4, N2O). Emissions estimates for each pixel are calculated following IPCC Guidelines for national greenhouse gas inventories where stand-replacing disturbance occurred, as mapped in the Global Forest Change annual tree cover loss data of Hansen et al. (2013). The carbon emitted from each pixel is based on carbon densities in 2000, with adjustment for carbon accumulated between 2000 and the year of disturbance. Emissions reflect a gross estimate, i.e., carbon removals from subsequent regrowth are not included. Instead, gross carbon removals resulting from subsequent regrowth after clearing are accounted for in the companion forest carbon removals layer. The fraction of carbon emitted from each pixel upon disturbance (emission factor) is affected by several factors, including the direct driver of disturbance, whether fire was observed in the year of or preceding the observed disturbance event, whether the disturbance occurred on peat, and more. All emissions are assumed to occur in the year of disturbance. Emissions can be assigned to a specific year using the Hansen tree cover loss data.This emissions layer is part of the forest carbon flux model described in Harris et al. (2021), which introduces a geospatial monitoring framework for estimating global forest carbon fluxes which can assist governments and non-government actors with tracking greenhouse gas fluxes from forests and decreasing emissions or increasing removals by forests. All input layers were resampled to a common resolution of 0.00025 x 0.00025 degrees each to match Hansen et al. (2013).Emissions are available for download in two different area units: 1) megagrams of CO2 emissions/ha, and 2) megagrams of CO2 emissions/pixel. The first is appropriate for visualizing (mapping) emissions because it represent the density of emissions per hectare. The second is appropriate for calculating the emissions in an area of interest (AOI) because the values of the pixels in the AOI can be summed to obtain the total emissions for that area. The values in the latter were calculated by adjusting the emissions per hectare by the size of each pixel, which varies by latitude. Both datasets only include pixels within forests, as defined in the methods of Harris et al. (2021).Related Open Data Portal layers: Forest Carbon Removals, Net Forest Carbon FluxGoogle Earth Engine asset and visualization scriptResolution: 30 x 30mGeographic Coverage: GlobalFrequency of Updates: AnnualDate of Content: 2001-2020Cautions:Data are the product of modeling and thus have an inherent degree of error and uncertainty. Users are strongly encouraged to read and fully comprehend the metadata and other available documentation prior to data use. Values are applicable to forest areas only (canopy cover >30 percent and >5 m height). See Harris et al. (2021) for further information on the forest definition used in the analysis.Emissions reflect stand-replacing disturbances as observed in Landsat satellite imagery and do not include emissions from unobserved forest degradation.Emissions reflect a gross estimate, i.e., carbon removals from any regrowth that occurs after disturbance are not included. Instead, gross carbon removals are accounted for in the companion forest carbon removals layer.Emissions data contain temporal inconsistencies because tree cover loss from 2001-2010 and 2011-2020 were produced from different algorithms, with later years of loss likely to be more sensitive to smaller-scale forest disturbances.

This repository provides a set of essential variables to support research on forest loss driven by mining. All variables have been resampled to 30 arcsec spatial resolution (approximately 1 by 1 km at the equator) and are encoded in Geographic Tagged Image File Format (GeoTIFF). The grid extends from the longitude −180 to 180 degrees and from the latitude −90 to 90 degrees in the geographical reference system WGS84. Cells over water have no-data values. Below we describe the list of variables, sources, and processing steps.area_of_mines_circa_2018.tif: mining area in square metres. This layer was derived from a global-scale data set of mining polygons [Maus et al., 202a,b0] available from [doi:10.1594/PANGAEA.910894] under CC BY-SA 4.0 license. The mining area for each 30 arcsec grid was calculated intersecting cells and mining polygons.distance_to_mine_circa_2018.tif: distance to the nearest mine in metres. This layer was derived by calculating the Euclidean distance between each grid cell's centroid to the centroid of the closest grid cell with mine presence, i.e. cells where area_of_mines_circa_2018.tif > 0.area_of_forest_cover_circa_2000.tif: area of forest cover in square metres. This layer was derived from the Global Forest Change (GFC) dataset [Hansen et al., 2013] version 1.7 available from [https://earthenginepartners.appspot.com/science-2013-global-forest/download_v1.7.html] under CC BY 4.0 license. We aggregated the GFC data from 1 arcsec to our 30 arcsec grid cells by summing the area of forest cover pixels weighted by their surface intersection with the 30 arcsec cells.area_of_forest_cover_within_mines_circa_2000.tif: area of forest cover in square metres. This layer was derived using the same methods as area_of_forest_cover_circa_2000.tif; however, it only includes forest area intersecting mining polygons, i.e. the on-site forest cover circa 2000.area_of_forest_cover_loss_yearly_from_2001_to_2019.tif: area of forest cover loss in square metres. This GeoTIFF file has 19 layers (one layer per year) starting from 2000. We aggregated the GFC data from 1 arcsec to our 30 arcsec grid cells by summing the area of forest loss pixels weighted by their surface intersection with the 30 arcsec cells.ecoregions2017_code.tif: an integer with the ecoregions code (ECO_ID) rasterized from the Ecoregion 2017 polygons [Dinerstein et al., 2017; Resolve, 2017], which is available from [https://ecoregions2017.appspot.com/] under CC BY 4.0 license. The polygons were rasterized to a 30 arcsec grid by the major class present. The ecoregion class names corresponding to the GeoTIFF file values are available in the auxiliary file ecoregions_2017_concordance_tbl.csv, which contains the following variables ECO_ID, ECO_NAME, BIOME_NUM, BIOME_NAME, where ECO_ID is a unique identifier.The layers available from this repo can be stacked together with other variables essential for land-use modelling. Some of these variables are openly available at the same spatial extent and resolution, for example, grided population [NASA, 2018], elevation and slope [Amatulli et al., 2018a,b].

Datasets, code, and metadata associated with, Hansen, WD, Fitzsimmons, R, Olnes, J, Williams, AP. An alternate vegetation type proves resilient and persists for decades following forest conversion in the North American boreal biome. J Ecol. 2020; 109: 85– 98. https://doi.org/10.1111/1365-2745.13446. Abstract: 1. Changing climate and natural disturbances are increasingly causing forests to transition to alternate vegetation types (e.g., new tree species assemblages, grasslands). Determining whether and how these new vegetation types will persist is essential for forecasting earth system function during this century. However, long-term studies of past disturbance-induced forest conversion are rare, and future climates may not have a present day analog. Both limit the utility of empirical approaches for evaluating the fate of alternate vegetation types. 2. We conducted individual-based simulations to test how changing climate, disturbance, and biotic interactions shape the resilience of deciduous broadleaf forest, which has begun to replace spruce after severe wildfires in interior Alaska, USA. 3. Deciduous forest persisted in 86% of simulated stands and was especially resilient with fire return intervals of 50 years or shorter. However, when transitions to another vegetation type did occur, mixed forest was most common, particularly when fire return intervals were longer than 50 years and when seed source was distant. Recovery to spruce forest almost never occurred. Moose browsing and postfire drought also influenced outcomes, but effects were contingent on fire-regime characteristics. When fire return intervals were long and postfire seed sources were 500 m away or farther, moose browsing reduced deciduous sapling growth and survival, helping spruce better compete. Late 21st-century drought following short-interval fire was sufficient to occasionally cause conversion to nonforest. 4. Synthesis. Our analyses indicate that emerging postfire deciduous forest will almost certainly prove resilient for decades to centuries, which will shape biophysical and biogeochemical feedbacks to climate and alter subsequent disturbance. This paper offers a framework for quantifying the long-term resilience of alternate vegetation types following forest conversion and lends critical insights into the biotic and abiotic agents that are likely to underpin similar vegetation transitions across the North American boreal biome. File list: model_archive.zip ccsm4_climate_complete.csv ipsl_climate_complete.csv mri_climate_complete.csv output_processed_ccsm4.csv output_processed_ipsl.csv output_processed_mri.csv analysis-stability.Rmd : analysis of the condensed simulation output for results in the paper analysis-climate.Rmd: analysis of the climate data for results in the paper data_column_names_processed tables_Hansen.png data_column_names_complete_tables_Hansen.png table_descriptions_stability_project.csv: descriptions of each of the data tables in this project.

Forest Fragmentation indices for South East Asia: A spatial raster representation of forest fragmentation indices. The entire study area is divided into rectangular grid cells with an edge length of 5 arc minutes (approximately 9 × 9 km). The forest cover data is obtained from the Global Forest Change data base (Hansen et al., 2013). Combining the initial forest cover raster maps and the yearly forest loss raster maps serve to derive yearly measures of fragmentation for each grid cell with the landscapemetrics package (Hesselbarth et al., 2019) in R 3.6 (R Core Team, 2020), which implements the calculation of landscape metrics from the widely employed FRAGSTATS software (McGarigal et al., 2002). The fragmentation metrics calculated are the perimeter-area fractal dimension (PAFRAC), the mean of the core area index (CAI_MN), the area-weighted mean of the core area index (CAI_AM), the coefficient of variation of the core area index (CAI_CV), clumpiness index (CLUMPY), and the normalized landscape shape index (nLSI).