Earth Observation-based Quantification of Woody Biomass

Several terrestrial biosphere Carbon Dioxide Removal (CDR) approaches rely on the capture and storage of carbon in living woody biomass, including both aboveground biomass (AGB) and belowground biomass (BGB). The quantification of BGB is often dependent upon AGB¹, which is more directly observable. Thus, for these pathways, accurately and conservatively quantifying the gross storage of carbon in in both AGB and BGB is crucial for demonstrating net CO₂ removal.

Aboveground biomass (AGB) encompasses all living vegetation above the soil surface, including stems, branches, foliage, and bark. Woody biomass refers to plants whose structure includes lignified stems, such as bamboo, shrubs, and trees. Belowground woody biomass (BGB) encompasses the subsurface woody biomass in root systems.

This Module provides a transparent, scientifically grounded process by which Isometric quantifies project-scale AGB and BGB using Earth Observation based third-party or internally-developed biomass datasets. These datasets are typically derived from satellite-based remote sensing, ground-based calibration data, and predictive statistical models. This Module outlines the procedures by which Isometric selects, validates, applies, and transparently documents Earth Observation based AGB data products, to ensure consistency with the Isometric Standard and governing Protocol.

Earth Observation aboveground biomass (AGB) maps represent the culmination of decades of remote sensing innovation, ecological research, and data integration methods aimed at quantifying vegetative carbon stocks across large spatial extents. These maps typically integrate satellite observations, airborne remote sensing, and in situ forest inventory data to generate spatially continuous predictions of woody aboveground biomass. The goal is to achieve reliable, wall-to-wall AGB estimates across landscapes where field-based measurement alone is economically or technically infeasible.

AGB maps differ from simpler land cover maps or Normalized Difference Vegetation Index (NDVI) in that they attempt to estimate actual tonnes of dry biomass per hectare — and often include uncertainty quantification layers for each pixel. This integration of empirical data, remote sensing, and modeling has enabled a new generation of biomass maps that underpin conservation policy, MRV systems, and nature-based carbon credit issuances.

Common sources of remotely sensed data that underpin Earth Observation based AGB mapping include:

• LiDAR (Light Detection and Ranging): provides high-resolution 3D structural data on canopy height and vegetation density -- includes missions such as GEDI.
• Synthetic Aperture Radar (SAR): sensitive to forest structure and moisture and can penetrate cloud cover, making them useful in the often cloud-covered, humid tropics. Common instruments include the new ESA BIOMASS satellite.
• Optical sensors: offer spectral reflectance data used to derive vegetation indices correlated with canopy vigor and density. Most well studied and utilized, and include well-known missions such as Landsat, MODIS, and Sentinel-2.

These data are calibrated and validated against field measurements using statistical models ranging from linear regressions to complex machine learning algorithms like Random Forests, Support Vector Regression, or ensemble approaches. Model outputs are then spatialized across the landscape using remote sensing inputs, yielding gridded AGB predictions -- such as biomass in tonnes per hectare. Some products also include associated error metrics per pixel.

While these datasets have vastly improved our capacity to monitor biomass, limitations persist. These include, but are not limited to:

• Signal saturation: Optical and radar sensors tend to saturate in high-biomass regions (>200 t/ha), leading to underestimation.
• Coarse resolution: Lower spatial or temporal resolution of Earth Observation products may not capture small-scale heterogeneity or edge effects on forest biomass.
• Geographic bias: Most training data is skewed toward accessible or well-studied sites, reducing generalizability or applicability to more remote sites.
• Uncertainty: Errors in field data, sensor noise, or misregistration may compound and be improperly propagated.

These uncertainties make it essential that Earth Observation based AGB maps used for crediting are transparently benchmarked, adjusted if necessary, and conservatively interpreted.

This Module was developed based on the current state of the art, publicly available science regarding quantification of woody biomass and long-term monitoring of terrestrial ecosystems. This Module aims to be scientifically stringent and robust. We recognize that some requirements may exceed the status quo in the market and that there are will be opportunities to improve the rigor of this Module.

Additionally, this Module will be reviewed when there is an update to published scientific literature, government policies, or legal requirements which would affect net CO₂e removal quantification or the monitoring guidelines outlined in this Module, or at a minimum of every 2 years.

This Module quantifies the total CO₂ equivalent stored as aboveground woody biomass across the project area ($CO_2e_{AGB,t}$ in tonnes) at a given time point, t, through the use of third-party biomass datasets that meet specified eligibility requirements. Where included as a pool, it also quantifies the total CO₂ equivalent stored as belowground woody biomass across the project area ($CO_2e_{BGB,t}$) at a given time point, t, which is derived from assessed aboveground biomass estimates. This Module applies under the following conditions:

• Reliable field measurement data can be consistently collected, reported, and verified; and

• R-Y90T-0The Project lies within a biome, ecoregion, or forest type for which one or more AGB data products are available that meet the eligibility guidelines as determined by Isometric.

Throughout this Module, the use of “must” indicates a requirement, whereas “should” indicates a recommendation.

Eligibility criteria are adapted from those corresponding to a validation stage of 2 or higher in the hierarchy set by the CEOS validation working group using best practices described in Duncanson et al., (2021)². These include:

• The methodology for the production of the data product must be based on peer-reviewed scientific literature, with supporting citations provided by the data provider.
• Product accuracy must be assessed against reference data covering at least 30 locations or time points across a range of biomass conditions, with the validation details documented in the supporting documentation and metadata.
• The product name, model version, and date must be provided for each dataset. 

For application to a specific project area, the following must be demonstrated:

• The data product must provide full spatial coverage of the project area. 
• The product’s validation dataset should include the project’s biome, forest type, and expected biomass range. If not or only partially, justification for the product’s applicability must be provided. 
• The temporal coverage of the data product is anticipated to continue at a frequency aligned with the project’s Reporting Period interval (e.g., there is no imminent planned retirement).

Third-party AGB map data products must be developed by an Isometric-approved provider whose processes and products have indicated alignment with the best available science and reliable performance of the data products. At each use, the data provider must provide an attestation indicating alignment with the above criteria for the provided data, as well as relevant metadata (e.g., model version) to ensure continuity. Other data products may be permitted with Isometric approval where alignment with the requirements and reliable data performance can be demonstrated.

It is anticipated that the models underlying the map generation may be trained on broader data sets not limited to the project area or conditions. Suitability for application within the project area is further assessed against field data collected within the project area as part of the benchmarking process described in Section 6, and further local calibration to the project area may also be supported with these data.

To calculate the total aboveground biomass in the project area at a given time point ($M_{AGB,t}$ in tonnes), the map must first be masked to remove any pixels which don’t fully reside within the bounds of the project area to ensure there is not contamination from vegetation dynamics outside of the project. The total biomass is then calculated as:

$$M_{AGB,t} = \bar{m}_t \times A_{project}$$

(Equation 1)

Where: 

• $\bar{m}_t$ is the mean aboveground biomass density at time $t$ within the masked map in tonnes per ha 
• $A_{project}$ is the total project area in hectares

The total aboveground biomass carbon stock in tonnes CO₂ equivalent at time, $t$, is further calculated as:

$$CO_2e_{AGB}(t) = \frac{44}{12} \times CF \times M_{AGB}(t)$$

(Equation 2)

Where:

• $\frac{44}{12}$ is the ration of mass of CO₂ to mass C, used to convert to tonnes of CO₂
• $CF$ is the carbon fraction for the tree species in the project area, in tonne dry biomass per tonne C

The carbon fraction, $CF$, must be chosen from the following hierarchy:

1. A regional and species-specific factor that is justified based on scientific literature (e.g., Doraisami et al., 2022³). This is the preferred approach to have the most accurate estimate and avoid overestimation;
2. If the above is not available, then a genus-specific or national average factor that is justified based on scientific literature can be used;
3. As a last resort, if it is proven that the above two factors are not available, then a default factor of 47%, which is a mean across species, can be used^4,5.

Where belowground biomass (BGB) is an included as a carbon pool, the belowground biomass is calculated as:

$$CO_2e_{BGB}(t) = CO_2e_{AGB}(t) \times RS$$

(Equation 3)

Where:

• $RS$ is the dimensionless project-specific root-to-shoot ratio

Appropriate root-to-shoot ratios should be selected by regional and species-specific factors that are justified based on scientific literature (e.g., USFS's Component Ratio Method or similar national-level species-specific ratios). This is the preferred approach to have the most accurate estimate and avoid overestimation. If sufficient evidence is provided to demonstrate that no suitable project-specific factor can be obtained, matching to the ecological zone and continent of the project area, based on the IPCC 2019 Table 4.4⁴, must be used. In this case, sufficient evidence documenting the unsuccessful search for project specific factors must also be supplied. Acceptable evidence must show (1) a list of search terms used within a research database (e.g., Web of Science, Google Scholar) that encapsulate the region and species relevant to the Project, and (2) the relevant species are not included in the list of species for which root-to-shoot ratio data are available on the TRY Plant Trait Database⁶.

Where the root:shoot ratios are species or genus specific, the project-area root:shoot ratio must be calculated as a weighted value to represent the species mix within the project area.

The uncertainty in selected $RS$ factors must be reported from the same source dataset. For example, the IPCC 2019 Chapter on Forest Land⁴ provides an uncertainty in the root-to-shoot ratio.

• G-0EDS-0Plot size should be similar to pixel-area of the Earth Observation based AGB map being validated.
  • It is recommended that plots be a minimum of 0.1 ha to help ensure stable quantification of biomass. Where pixels are smaller than this size, it is recommended to keep field plots larger and to then cluster neighboring pixels for benchmarking against the field plots.
• There is not a strict requirement for the number of field plots that must be measured. Where the benchmarking data are being used to support further calibration, a minimum of 30% of the data will be withheld from from model calibration/training to be used for data product validation and uncertainty assessment, unless the requirements for a cross-validation procedure are met (see Section 6.1). As such, it can be expected that the larger the number of data points collected, the more data that will be available for calibration and the better constrained the uncertainty characterization will be. As rough guidance, we note the recommendation in the field plot module that the number of plots should be sufficient to achieve good precision in the mean AGB estimate, with a recommended target being +/- 10% at 90% confidence
• G-M39E-0Plot measurements should also be temporally aligned, to the greatest possible extent, with map products.
• Because the purpose of benchmarking is to characterize the map's performance across the range of conditions to which it is applied, rather than to estimate mean stock directly, there may be differences in how the sampling design is structured. The sampling design must be structured to capture the range of AGB values present within the project area. Projects are encouraged to develop a stratification approach informed by differences in planting design, management, and/or environmental factors.
  • From the second benchmarking event onward, Projects should evaluate the sampling design and describe any updates made to the plan to better capture the range of AGB values in the Project Area (e.g., through structurally guided sampling⁷). This is recommended where plots at the previous benchmarking event did not span the 10th to 90th percentile of AGB values in the map for that event.
• Within each plot, the species mix should be representative of the broader project/strata species mix to ensure a representative spectral signal.

The field data are then used to benchmark the data product prior to its use for quantification and to characterize the uncertainty of its estimates relative to the field-measured AGB. Where the benchmarking reveals a bias in the map’s AGB estimates relative to the withheld validation data, a correction may be applied to the map estimates prior to the assessment of uncertainty.

Cross-validation procedures can allow the full benchmarking dataset to be used for calibration/training, while still maintaining a robust assessment of model skill and uncertainty. However, this requires additional criteria be met as part of the procedure to ensure transparency and replicability. Projects using cross-validation procedures wherein the full benchmark data are also used to support training must meet the following requirements:

• Model code and data must be supplied to enable reproducibility of the model evaluation results;
• Plots measured at more than one benchmarking event must be held out together;
• If cross-validation is used to select from multiple candidate models or feature sets, Projects must use either nested cross-validation or report validation statistics for the competing candidate models.

Models and measurements of aboveground biomass inherently include uncertainty from various data sources which contribute to the quantification. While it is not expected that all uncertainties are exhaustively quantified, Project Proponents must evaluate, report, and conservatively account for identifiable and significant sources following the in the approved approaches outlined in the Isometric Standard. Potential sources of uncertainty to consider include, but are not limited to:

• Pixel-level uncertainty estimates of the AGB map associated with map data sources and development process (e.g., remote sensing instrument errors, model errors)
• Uncertainty associated with any local corrections applied to the AGB product as a result of benchmarking
• Errors associated with field measurements and estimation procedures used for benchmarking (e.g., DBH measurement errors, height estimation bias, wood density variability, plot representativeness)

Project Proponents must follow the requirements within the Area-based Quantification of Woody Biomass for reporting the methodological details and data for the field plots collected for the benchmarking process. 

Data providers are responsible for providing the attestation and relevant metadata for the data products to demonstrate their eligibility in alignment with the requirements of Section 4.0. 

Isometric would like to thank the following external contributors to this Module:

• Sujit Ghosh PhD
• Renoster

1. Pan, Y., Birdsey, R. A., Fang, J., Houghton, R., Kauppi, P. E., Kurz, W. A., ... & Hayes, D. (2011). A large and persistent carbon sink in the world’s forests. Science, 333(6045), 988-993. https://doi.org/10.1126/science.1201609 ↩

2. Duncanson, L., Armston, J., Disney, M., Avitabile, V., Barbier, N., Calders, K., Carter, S., Chave, J., Herold, M., MacBean, N., McRoberts, R., Minor, D., Paul, K., Réjou-Méchain, M., Roxburgh, S., Williams, M., Albinet, C., Baker, T., Bartholomeus, H., Bastin, J.F., Coomes, D., Crowther, T., Davies, S., de Bruin, S., De Kauwe, M., Domke, G., Dubayah, R., Falkowski, M., Fatoyinbo, L., Goetz, S., Jantz, P., Jonckheere, I., Jucker, T., Kay, H., Kellner, J., Labriere, N., Lucas, R., Mitchard, E., Morsdorf, F., Næsset, E., Park, T., Phillips, O.L., Ploton, P., Puliti, S., Quegan, S., Saatchi, S., Schaaf, C., Schepaschenko, D., Scipal, K., Stovall, A., Thiel, C., Wulder, M.A., Camacho, F., Nickeson, J., Román, M., Margolis, H. (2021). Aboveground Woody Biomass Product Validation Good Practices Protocol. Version 1.0. In L. Duncanson, M. Disney, J. Armston, J. Nickeson, D. Minor, and F. Camacho (Eds.), Good Practices for Satellite- Derived and Product Validation, (p. 236): Land Product Validation Subgroup (WGCV/CEOS), doi:10.5067/doc/ceoswgcv/lpv/agb.001 ↩

3. Doraisami, M., Kish, R., Paroshy, N. J., Domke, G. M., Thomas, S. C., & Martin, A. R. (2022). A global database of woody tissue carbon concentrations. Scientific Data, 9(1), 284. https://doi.org/10.1038/s41597-022-01396-1 ↩

4. Aalde, Harald. vol. 4, Institute for Global Environmental Strategies (IGES), 2006. IPCC Guidelines for National Greenhouse Gas Inventories: Agriculture, Forestry and Other Land Use. https://www.osti.gov/etdeweb/biblio/20880391 ↩ ↩² ↩³

5. McGroddy, M. E., Daufresne, T., & Hedin, L. O. (2004). Scaling of C: N: P stoichiometry in forests worldwide: Implications of terrestrial redfield‐type ratios. Ecology, 85(9), 2390-2401. https://doi.org/10.1890/03-0351 ↩

6. Kattge, J., Bönisch, G., Díaz, S., Lavorel, S., Prentice, I. C., Leadley, P., ... & Cuntz, M. (2020). TRY plant trait database–enhanced coverage and open access. Global change biology, 26(1), 119-188. ↩

7. Goodbody, T. R., Coops, N. C., Queinnec, M., White, J. C., Tompalski, P., Hudak, A. T., ... & Woods, M. E. (2023). sgsR: a structurally guided sampling toolbox for LiDAR-based forest inventories. Forestry, 96(4), 411-424. https://doi.org/10.1093/forestry/cpac055 ↩