Skip to main content
CTrees Logo
  • Land Carbon

Global Aboveground Biomass

Estimate aboveground biomass annually for 2000-2025 in any hectare of land on Earth

Lead scientists

  • ,
  • ,

Standardized open data is available via Earthmover, Registry of Open Data on AWS, and Google Earth Engine Community Catalog. Reach out for customized data solutions.

Map of global aboveground biomass in 2025
Map of global aboveground biomass in 2025

Track biomass change at a hectare level worldwide

This innovative CTrees product estimates global above-ground biomass (AGB) from 2000 to present at a standard spatial resolution of 100 meters, and has a range of diverse use cases and applications.

Our methodology uses a sophisticated multiscale spatio-temporal machine learning framework that integrates satellite imagery with extensive airborne LiDAR remote sensing measurements. CTrees also leverages a wealth of ground-level AGB data collected by our team, collaborators, national forest inventory datasets, and the broader research community.

The process includes four key steps:

  1. AI-enhanced baseline metrics: Establishing advanced vegetation height metrics and AGB maps for 2020.
  2. Annual AGB time series: Utilizing innovative change detection modeling for continuous AGB estimates.
  3. Post-processing: Removing spikes from the time series.
  4. Validation and uncertainty assessment: Ensuring data accuracy and reliability at scale.

CTrees’ AGB products are founded on over 20 years of peer-reviewed expertise in open-source and publicly distributed national and global datasets supported by NASA programs in remote sensing technology, carbon cycle, and terrestrial ecology. Our data products are accompanied by robust uncertainty metrics and support a wide range of applications, including greenhouse gas (GHG) inventory and reporting at both national and global levels, local carbon accounting, nature-based climate solutions, and a variety of forest and vegetation carbon market projects.

Global aboveground biomass for 2000-2025
Map of global aboveground biomass for 2000-2025

Data at a Glance

Units

Mg ha-1

For conversion to MgC/ha, multiply by 0.5 or a forest-specific carbon fraction

For conversion to MgCO2e/ha, multiply MgC/ha by 3.667

Spatial resolution

Standard 100-meter product: Global

Temporal coverage

Annual, 2000-2025

Unit, scaling factor

Integer 16, Scale: 0.1

Map projection and file format

Projection: Geographic Projection (EPSG:4326 - WGS 84)

Format: Cloud-Optimized GeoTIFF (COG), Zarr and Icechunk Repository

Building an annual AGB time series

Overview

CTrees' global aboveground biomass maps combine ground inventory data with airborne and satellite remote sensing to make forest biomass change accessible and measurable at the hectare scale worldwide. With annual layers available at 100-meter resolution from 2000 to 2025, CTrees’ global AGB maps reflect a combination of coverage, resolution, and temporal depth that no existing open product provides.

The underlying methodology builds on peer-reviewed work, and the description of this product is a preprint currently under review. The method has four key components:

  1. NASA GEDI and ICESat-2 (ATL08) lidar, used to map vegetation structure (canopy height) with models fitted per ecoregion.
  2. The DenseNet model that maps aboveground biomass for the 2020 base year, trained on forest-inventory and airborne lidar estimates.
  3. A change-detection model that extends the 2020 map into an annual 2000–2025 series, with change/no-change masks so that real loss and regrowth survive cross-year calibration.
  4. A per-pixel uncertainty layer with every annual layer. Key publications are cited at the end.

The carbon maps meet accuracy standards for emission factor calculations and will be updated with new data from NASA-ISRO's Synthetic Aperture Radar (NISAR) and ESA’s Biomass in 2024 and 2025.

CTrees’ innovative approach to developing global AGB maps makes use of both optical and microwave satellite observations and benefits from a large number of airborne scanning LiDAR (ALS) data acquired globally by our team, our collaborators, and publicly funded experiments. These datasets include: HLS, Landsat, ALOS PALSAR, Copernicus DEM, GEDI height metrics, ICESat-2, LiDAR CHM, forest cover change (e.g., Tropical Moist Forests, Global Forest Change), forest inventory data, and airborne liDAR biomass.

INPUT DATA

  • Landsat and Harmonized Landsat and Sentinel-2 (HLS) images provide surface reflectance data at about 30 meters sensitive to vegetation cover and density.
  • ALOS PALSAR images with global coverage of radar backscatter measurement at different polarizations sensitive to forest structure.
  • Copernicus DEM provides surface elevation data at 30-meter resolution measured by X-band synthetic aperture radar sensors.
  • GEDI provides waveform LiDAR data sampling global forests between ~51.6° N and 51.6° S latitudes. GEDI L2A includes elevation and height metrics. GEDI L4A provides above-ground biomass density estimations.
  • ICESAT-2 Advanced Topographic Laser Altimeter System (ATLAS) point counting LiDAR samples across global vegetation provide vegetation structure and height in areas of low-density vegetation across boreal and woodland savanna ecoregions.

BIOMASS TRAINING DATA

CTrees machine learning models are trained on more than 500,000 ground plots from national forest inventories, including a variety of boreal, temperate, and tropical countries along with more than one million hectares of airborne and drone-based LiDAR based estimates. A portion of training data is set aside for validation and uncertainty assessment for different ecoregions.

ACTIVITY AND CHANGES LAYER

Global Forest Change (stand-replacement loss), JRC Tropical Moist Forests (deforestation and degradation), and GABAM (burned area), along with vegetation index changes, are used to generate the change and no change mask.

Spatial modeling

Spatial modeling of AGB will be performed using ecoregion-specific LightGBM models, which have been proven to outperform other models in terms of performance metrics relying on accuracy, spatial integrity, and computation time. The steps for applying the model to develop maps include:

  • Regional Stratification: Two K-mean clusters maps derived from available height metrics maps and location coordinates, 120 clusters for height metric mapping and 60 clusters for biomass mapping. The clustering strategy ensures that regional variability is accurately represented while minimizing boundary artifacts
  • Height metrics: Global 100-meters spatial maps of four relative height (RH50, RH75, RH90, and RH98) metrics for 2020. GEDI and ICSat -2 height retrievals are quality-filtered, cross-calibrated, and aggregated to 100 meters. These aggregates then train a per-ecoregion machine learning model that produces wall to wall height maps.
  • Base AGB map: Base AGB map for the year 2020 using RH metric maps and RS data trained by ground and LiDAR based plots across ecoregion clusters at 100-meter resolution.
  • AGB time series maps: Application of change detection machine learning model using annual RS data to develop annual AGB maps from 2000-2025.
  • Uncertainty estimates: Spatial uncertainty estimation at the pixel level by developing a bootstrapping approach for biomass estimation and error propagation from different sources of uncertainty.

Validation of AGB maps

By developing uncertainty maps at the pixel level, we employ various validation approaches to assess how our biomass maps align with plot-level data across the entire biomass spectrum. We analyze mean above-ground biomass (AGB) estimates at sub-national levels (state and county), using adequately sized plot data to identify bias and make corrections at a larger scale. Additionally, we compare mean and total carbon estimates at the national level with countries that routinely monitor and report GHG inventories in the forest sector.

When repeat inventory measurements are available, our annual biomass maps serve as a crucial tool to confirm changes in national carbon stocks. The results presented below encapsulate the overall validation and performance of our maps, showcasing their accuracy and reliability.

Plot-level validation of the global 1-hectare map reveals consistent estimates across the entire global biomass range with gradual systematic error increasing in high-biomass forests where existing satellite data and forest heights are less sensitive to biomass variations due to larger diameter trees and denser canopy forests.

Validation of the biomass map at the smallest jurisdictional unit that can simulate a project-level mean carbon stock shows consistent accuracy of the maps across multiple years (Yu et al., 2022). More recent year validation with the U.S. Forest Service Forest Inventory and Analysis (FIA) data will be performed after renewal of access to FIA plot locations. The uncertainty shown in the county-level comparison may be related to size of FIA plots (0.44 ha) compared to map pixel (1-ha), definition of forest area, and the mismatch between the plot data collection and remote sensing imagery within the year of inventory.

Comparison with the U.S. state-level FIA estimates captures the consistency of biomass map accuracy across a continental scale covering a variety of temperate and boreal ecosystems with a combination of needle-leaf and broad-leaf and deciduous forests (Yu et al. 2022).

At the global scale, CTrees carbon stocks can be compared to countries that have regular forest inventory measurements and report carbon stocks to the UN Framework Convention on Climate Change (UNFCCC), as part of GHG inventory reporting, and to the UN Food and Agricultural Organization (FAO). The results demonstrate that the overall variations of carbon stocks from CTrees maps agree with national data accumulated from inventory data collected over 5-10 years for GHG reporting.

Open carbon data and customized maps

Via Earthmover

Via Registry of Open Data on AWS

Via Google Earth Engine Community Catalog

For customized maps and data solutions, please reach out to our team.

Get in touch

References

Ferraz, A., Saatchi, S., Xu, L., Hagen, S., Chave, J., Yu, Y., ... & Ganguly, S. (2018). Carbon storage potential in degraded forests of Kalimantan, Indonesia. Environmental Research Letters, 13(9), 095001.

McRoberts, R. E., Næsset, E., Saatchi, S., & Quegan, S. (2022). Statistically rigorous, model-based inferences from maps. Remote Sensing of Environment, 279, 113028.

Saatchi, S. (2025). Mapping Global Live Woody Vegetation Biomass at Optimum Spatial Resolutions. Zenodo. https://doi.org/10.5281/zenodo.15858551 (opens in a new tab)

Saatchi, Sassan S., et al. "Benchmark map of forest carbon stocks in tropical regions across three continents." Proceedings of the national academy of sciences 108.24 (2011): 9899-9904.

Yang, Y. (2026). From Snapshot Maps to Continuous Monitoring of Global Forest Carbon at 100 m Resolution (2000–2025) https://doi.org/10.31223/X5KJ4Q (opens in a new tab)

Yu, Y., Saatchi, S., Domke, G. M., Walters, B., Woodall, C., Ganguly, S., ... & Melendy, L. (2022). Making the US national forest inventory spatially contiguous and temporally consistent. Environmental Research Letters, 17(6), 065002.

Xu, L., Saatchi, S. S., Yang, Y., Yu, Y., Pongratz, J., Bloom, A. A., ... & Schimel, D. (2021). Changes in global terrestrial live biomass over the 21st century. Science Advances, 7(27), eabe9829.

Xu, L., Saatchi, S. S., Shapiro, A., Meyer, V., Ferraz, A., Yang, Y., ... & Ebuta, D. (2017). Spatial distribution of carbon stored in forests of the Democratic Republic of Congo. Scientific Reports, 7(1), 15030.

Expertise in biomass modeling

Data for carbon monitoring and reporting

Conduct annual GHG monitoring

Track global carbon changes at 1-hectare resolution from 2000 to present, providing insights into emissions and removals over time.

Estimate carbon in any ecosystem

Access biomass carbon data for forests, urban areas, woodlands, grasslands, croplands, coastal wetlands, and mangroves.

Quantify emission factors in forest carbon projects

Use high-resolution carbon maps and land-use data to support robust baseline assessments and verify additionality in carbon offset initiatives.

Report on jurisdictional forest carbon trends

Evaluate the progress and impact of national and sub-national land sector policies and market-based solutions aimed at reducing emissions from deforestation and forest degradation.

Page last updated Nov 16, 2026

Related Datasets

Stay Connected

Subscribe to our monthly newsletter and be the first to know about new data releases, research, and events.

By submitting your email address, you agree to receive communications from CTrees. You can opt-out at any time.