Forest carbon stock estimation from canopy structure
Above-ground biomass baselines for Verra VCS and Gold Standard projects depend on canopy structure data that ground plots alone cannot supply at scale. Multi-sensor fusion of L-band SAR, spaceborne lidar and InSAR extends the measurable range well beyond C-band's saturation ceiling.
Sensors
- ALOS-2 PALSAR-2 (L-band SAR, JAXA): L-band (1.27 GHz) penetrates the canopy and scatters from woody trunks and large branches, maintaining sensitivity to above-ground biomass up to roughly 150–200 t/ha in dense tropical forest before saturation. Single-polarisation resolution is 10 m (Fine mode); revisit is 14 days. Global mosaic products are published annually by JAXA.
- Sentinel-1 (C-band SAR, ESA/Copernicus): C-band (5.405 GHz) backscatter saturates at approximately 80–100 t/ha in closed-canopy forest, making it unreliable as a standalone biomass estimator in high-biomass stands. Useful for forest/non-forest masking, change detection and as a fusion input. IW mode: 10 m resolution, 6-day revisit at mid-latitudes with both satellites operating.
- GEDI (Global Ecosystem Dynamics Investigation, NASA): Full-waveform spaceborne lidar aboard the International Space Station. Delivers canopy height, canopy cover and vertical structure metrics at 25 m footprint diameter, sampled in parallel tracks spaced roughly 600 m apart. Directly measures relative height (RH) metrics used in published allometric models. Coverage is limited to approximately 51.6° N/S by the ISS orbit.
- TanDEM-X (X-band InSAR, DLR): The TanDEM-X pair generates interferometric height products. The global DEM has 12 m posting and roughly 1 m absolute height accuracy over flat terrain. Forest canopy height can be extracted by differencing the X-band phase centre height from a bare-earth reference, with accuracy degrading in dense, tall canopies where X-band penetrates only partially.
- Landsat 8/9 OLI (USGS/NASA): 30 m multispectral imagery used to derive canopy cover fractions and NDVI-based stratification layers that guide allometric model selection across forest types. 16-day revisit per satellite; combined archive runs from 1972 (Landsat 1 MSS) to present, enabling long baseline stratification.
Why the registry asks for tonnes but satellites measure geometry
Verra VCS and Gold Standard methodologies require an above-ground biomass (AGB) baseline expressed in tonnes of dry matter per hectare, then converted to carbon at a ratio of approximately 0.47. No satellite measures biomass directly. What sensors measure is either the intensity of radar backscatter scattered from woody material, or the height of the canopy surface above bare ground. The conversion from those geometric and dielectric signals to tonnes involves allometric equations, and allometric equations carry uncertainty that must be propagated honestly through any MRV submission.
That said, the alternative is worse. Ground-plot networks in tropical forests are expensive, spatially sparse and subject to access constraints. A credible baseline combines plot data with wall-to-wall satellite-derived structure maps: plots anchor the allometry, satellites supply the spatial extrapolation. The question is which sensors to use where, and what their limits actually are.
What a floating roof gives away: the saturation problem in SAR
C-band radar, used by Sentinel-1, is sensitive to small branches and leaf surfaces. In open woodland and early-succession forest, where biomass is below roughly 80–100 t/ha, C-band backscatter rises usefully with AGB. Above that threshold the canopy becomes opaque to C-band and backscatter plateaus. Estimating biomass in a mature Amazonian stand from Sentinel-1 alone is like estimating the depth of a lake by measuring how wet the surface looks.
L-band (ALOS-2 PALSAR-2) penetrates deeper, reaching trunks and major branches. Sensitivity extends to 150–200 t/ha in many tropical forest types before saturation sets in, though the exact ceiling varies with forest structure and moisture conditions. P-band, used by ESA's forthcoming BIOMASS mission, penetrates further still, but that mission is not yet delivering operational data. For the present, L-band plus lidar is the practical answer to high-biomass stands.
GEDI's contribution: height as a proxy for stem volume
GEDI fires laser pulses at 242 Hz from the ISS and records the full return waveform. From the waveform, analysts extract relative height metrics such as RH95 (the height below which 95% of returned energy falls), which correlates strongly with Lorey's height and, through allometric relationships, with AGB. Published studies using GEDI data have reported AGB estimation errors in the range of 30–50% at the individual footprint level, falling to 10–20% when aggregated over project-scale areas of several thousand hectares.
The honest caveat: GEDI's sampling is not wall-to-wall. The 600 m track spacing means that in a 1,000 ha project area, direct GEDI observations may cover only a small fraction of the ground. The standard workflow uses GEDI observations to calibrate a continuous AGB map derived from SAR or optical predictors, not to map AGB directly. Cloud cover does not affect lidar the way it affects optical sensors, but dense smoke from fire can attenuate the pulse.
Extending the dynamic range: multi-sensor fusion
No single sensor covers the full biomass range from degraded scrub to intact old-growth without saturation or coverage gaps. The practical solution is to fuse predictors from multiple sensors, each contributing where it is most informative. A typical pipeline stratifies the landscape using optical canopy cover (Landsat or Sentinel-2), applies L-band backscatter as the primary AGB predictor in medium-biomass strata, and uses GEDI-derived canopy height to calibrate and extend estimates in high-biomass strata where L-band begins to saturate.
TanDEM-X interferometric canopy height adds a spatially continuous height layer at 12 m posting, useful in areas with limited GEDI coverage, though accuracy in tall closed-canopy forest is lower than GEDI's direct waveform measurement. The fusion approach requires careful cross-calibration and propagation of uncertainty from each input. Allometric model choice is often the largest single source of error, not sensor noise, particularly when a pantropical equation is applied to a locally distinctive forest type.
What the uncertainty budget looks like in practice
A credible AGB map for a carbon project should report uncertainty at multiple levels: sensor-level (radiometric calibration, speckle in SAR), model-level (allometric equation selection, parameter uncertainty), and spatial-aggregation-level (how plot sampling error propagates to the mapped area). For a well-designed project in a structurally uniform tropical forest, total AGB uncertainty at the project scale is typically 15–30% at the 90% confidence level, though this varies considerably with forest type, available ground-truth data and the biomass range present.
Registries increasingly expect uncertainty to be quantified and disclosed, not suppressed. A baseline that reports a suspiciously tight confidence interval is a red flag to auditors. Satellize's analytical outputs for carbon MRV include explicit uncertainty layers alongside point-estimate maps, which is what a methodologically literate verifier will ask for first.
From map to registry submission
A forest carbon baseline submission under VCS VM0015 or similar methodologies requires the AGB map to be accompanied by documentation of the sensor data sources, processing chain, allometric models used, validation statistics against independent plot data and a spatial uncertainty map. The map itself is not the deliverable; the auditable evidence package is.
The JAXA global PALSAR-2 mosaic, GEDI Level 4A AGB product and the Copernicus open data archive are all publicly accessible and citable, which matters for third-party verification. Satellize builds these pipelines on open constellations, adding commercial SAR tasking where archive data is insufficient or where a project requires a specific acquisition geometry. The Tonga crop-estimation programme demonstrated the same principle at agricultural scale: open-data analytics, independently verifiable, with uncertainty quantified at the output stage.
Typical figures
| SAR spatial resolution (ALOS-2 PALSAR-2 Fine mode) | 10 m single-look; typically multi-looked to 25–50 m for AGB mapping to reduce speckle |
| SAR spatial resolution (Sentinel-1 IW mode) | 10 m range × 10 m azimuth ground range detected |
| GEDI footprint diameter | 25 m; track spacing approximately 600 m; latitude coverage 51.6° N/S |
| TanDEM-X DEM posting | 12 m global; absolute height accuracy approximately 1 m over flat terrain, degraded in tall closed canopy |
| L-band AGB sensitivity range | Useful up to approximately 150–200 t/ha; saturation onset varies with forest moisture and structure |
| C-band AGB sensitivity range | Useful up to approximately 80–100 t/ha; not recommended as primary estimator in high-biomass closed-canopy forest |
| PALSAR-2 global mosaic revisit | Annual (JAXA publishes yearly 25 m global mosaics) |
| Sentinel-1 revisit | 6 days at mid-latitudes (both satellites); 12 days with one satellite |
| Typical project-scale AGB uncertainty | 15–30% at 90% confidence, depending on forest type, biomass range and ground-truth density |
| Archive depth | PALSAR/PALSAR-2 from 2007; Sentinel-1 from 2014; Landsat from 1972; GEDI from April 2019 |
Analytics Satellize can run
| Wall-to-wall AGB density map | Random forest or gradient-boosted regression of PALSAR-2 backscatter and GEDI RH metrics against field plots, following published JAXA/NASA GEDI AGB mapping protocols | GeoTIFF raster at 25–50 m resolution, with per-pixel uncertainty band, clipped to project boundary |
| Carbon stock baseline report | AGB-to-carbon conversion at 0.47 factor (IPCC Tier 1) applied to AGB map, with allometric model documentation and uncertainty propagation table | PDF report and supporting GIS layers formatted for VCS VM0015 or Gold Standard methodology submission |
| Biomass stratification layer | Optical canopy cover fraction (Landsat/Sentinel-2 NDVI) used to stratify forest into low/medium/high biomass classes, guiding sensor weighting in fusion model | Polygon shapefile of biomass strata with class statistics table |
| Canopy height model (CHM) | GEDI RH95 observations interpolated and gap-filled using TanDEM-X interferometric height, calibrated against GEDI waveform data | GeoTIFF CHM at 25 m posting with coverage confidence layer |
| Allometric uncertainty sensitivity analysis | Monte Carlo propagation of allometric parameter uncertainty across three published pantropical and regional equations, following methods in peer-reviewed remote-sensing literature | Tabular report showing AGB estimate and 90% confidence interval under each equation; recommendation for preferred model with justification |
| Multi-date AGB change map for baseline period | Annual PALSAR-2 mosaic differencing combined with Sentinel-1 change detection to attribute AGB change to disturbance events vs. growth, over the project's historical baseline window | Time-series GeoTIFF stack and summary statistics table by year |
Who does the work
We can get this done for you. Satellize runs its own analyst desk and a strong science team. You do not buy a data feed and work out what it means; our people source the imagery, run the analysis described on this page, and hand you the answer with its confidence limits stated. Discuss this requirement.