Ecosystem-services carbon-stock accounting from satellite biomass layers
Estimating forest carbon stocks from space is tractable in low-to-moderate biomass ecosystems but runs into a hard physics wall in dense tropical forests. This page explains which sensors work where, why, and what the incoming ESA BIOMASS mission changes.
Sensors
- GEDI (Global Ecosystem Dynamics Investigation): Full-waveform LiDAR mounted on the ISS. Samples canopy height and vertical structure at 25 m footprints along orbital tracks spaced roughly 600 m apart at the equator. Provides relative height (RH) metrics that correlate strongly with above-ground biomass; does not give wall-to-wall coverage but is used to calibrate and spatially extrapolate SAR-derived biomass maps. Coverage limited to latitudes between approximately 51.6° N and S.
- ESA BIOMASS (P-band SAR): Launched April 2025. Operates at 435 MHz (P-band, ~69 cm wavelength), which penetrates the full forest canopy and interacts with large woody stems and trunks rather than leaves. Designed to estimate above-ground biomass in high-density tropical forests where C-band saturates, with a target accuracy of ±20% or ±10 Mg/ha (whichever is larger) at 200 m resolution and 1-hectare aggregation. Global tropical forest coverage planned in three-year repeat cycles.
- ALOS-2 PALSAR-2 (L-band SAR): Operates at 1.27 GHz (L-band, ~23 cm wavelength). Penetrates deeper into canopy than C-band, with biomass saturation typically occurring at 100–150 Mg/ha in tropical forests, compared to roughly 50 Mg/ha for C-band. Provides 10 m single-polarisation and 25 m dual-polarisation imagery with 14-day repeat. JAXA's global forest/non-forest maps and the ALOS Forest/Non-Forest product are publicly archived.
- Sentinel-1 (C-band SAR): Operates at 5.405 GHz (C-band, ~5.5 cm wavelength). Free, open archive from 2014. Six-day repeat at mid-latitudes with two satellites. Highly effective for forest-cover change detection and biomass estimation in low-to-moderate biomass ecosystems (below roughly 50 Mg/ha), including temperate forests, degraded tropical forests and early-regrowth stands. Saturates rapidly in intact tropical forest; useful there primarily as a change-detection layer rather than a biomass estimator.
- ICESat-2 (ATLAS photon-counting LiDAR): Provides canopy height profiles along ground tracks at 100 m along-track resolution. Complements GEDI in higher-latitude forests (GEDI does not cover beyond ~52°) and in boreal zones. Less suited to dense tropical canopy than GEDI's full-waveform approach, but the two datasets are increasingly fused in published biomass mapping workflows.
Why biomass estimation is a physics problem before it is a data problem
Carbon stock accounting starts with above-ground biomass (AGB), typically expressed in tonnes of dry matter per hectare, then multiplied by a carbon fraction of roughly 0.47 for most forest types. The challenge is that the most carbon-dense forests are also the ones that defeat the most common remote-sensing approaches. Optical sensors see the top of the canopy; they cannot infer what is beneath. Cloud cover, which is near-permanent over many tropical forests, compounds this. SAR gets through clouds but hits a different wall: signal saturation.
Saturation is not a calibration error. It is a consequence of microwave physics. At C-band frequencies, the radar signal interacts primarily with small branches and leaves. Once the canopy is dense enough, the signal is scattered back before it reaches the large woody stems that hold most of the carbon. Above roughly 50 Mg/ha of AGB, additional biomass produces almost no additional backscatter. The relationship between signal and carbon stock becomes flat. Longer wavelengths penetrate further: L-band pushes the saturation ceiling to around 100–150 Mg/ha in tropical conditions. P-band, with its ~69 cm wavelength, interacts with trunks and large branches and is expected to remain sensitive well beyond 300 Mg/ha, which is why ESA's BIOMASS mission is built around it.
What GEDI changed, and what it still cannot do
GEDI's full-waveform LiDAR records the vertical distribution of canopy material from ground to top-of-canopy, giving metrics such as RH95 (the height below which 95% of returned energy falls) that correlate well with AGB across biomes. Published studies using GEDI data have produced global AGB maps with substantially lower uncertainty in high-biomass tropical forests than SAR-only approaches achieved previously. GEDI data are freely available through NASA's Earthdata portal.
The limitation is spatial sampling. GEDI lays down tracks, not a continuous grid. At the equator, adjacent tracks are roughly 600 m apart, leaving large gaps between observations. Producing a wall-to-wall biomass map requires using GEDI-derived heights as calibration points and then spatially predicting across the gaps using SAR backscatter, canopy texture from optical imagery, or terrain covariates. The statistical uncertainty of that extrapolation is the dominant error source in most national-scale AGB products. GEDI's ISS orbit also means coverage stops at about 51.6° latitude, excluding boreal forests, which hold very large carbon stocks in their soils and peatlands even where AGB is modest.
The BIOMASS mission: what P-band actually promises
ESA's BIOMASS satellite, launched in April 2025, is the first spaceborne P-band SAR. Its primary science objective is a global map of tropical forest AGB at 200 m resolution, updated across a three-year observation cycle. The mission design target is an AGB uncertainty of ±20% or ±10 Mg/ha (whichever is larger), which, if achieved, would be sufficient to support national greenhouse-gas inventory reporting under IPCC Tier 2 requirements for many countries that currently lack adequate ground-truth data.
P-band SAR also enables interferometric techniques (PolInSAR and TomoSAR) that can retrieve vertical forest structure rather than a single AGB estimate. This is scientifically significant: knowing how biomass is distributed vertically helps separate living crown from understory and improves allometric conversions. The mission carries a secondary objective of subsurface geology mapping in arid regions, but for carbon accounting purposes, the tropical forest biomass product is the headline deliverable. Honest caveat: BIOMASS data products are not yet publicly released at the time of writing; validation against field plots is ongoing.
Peatlands and wetlands: where biomass is the wrong metric
In peatlands, the majority of ecosystem carbon is not in above-ground biomass at all. It is in the peat column, which can be several metres deep and hold thousands of tonnes of carbon per hectare. Satellite biomass layers derived from SAR or LiDAR measure the trees and shrubs growing on the peat surface, not the peat itself. For carbon accounting in peatland ecosystems, AGB estimation is a secondary concern; the primary signals are water-table depth (inferred from SAR backscatter changes and InSAR subsidence, covered in the peatland drainage sibling page) and land-cover change that exposes bare peat.
Wetland forests, including flooded Amazonian várzea and igapó, present a different challenge. The periodic flooding changes SAR backscatter dramatically, and the interaction between water surface and vertical stems produces a double-bounce signal that can be misinterpreted as high biomass. Multi-temporal SAR analysis, distinguishing flooded from non-flooded states, is necessary before any biomass inversion is applied. These environments are among the most uncertain in current global AGB products.
From satellite layers to inventory-grade carbon numbers
National greenhouse-gas inventories under the UNFCCC require not just a point estimate of carbon stock but a quantified uncertainty, traceable to measurement methods. Satellite-derived AGB layers feed into this as activity data (how much forest area exists and how it is changing) combined with emission factors (how much carbon per unit area). The IPCC Good Practice Guidance accepts remote-sensing data for activity data at Tier 2 and Tier 3 levels, provided uncertainty is characterised.
Voluntary carbon markets impose their own methodological requirements, which vary by standard body (Verra VCS, Gold Standard and others). Most require a conservative discounting of estimated carbon stocks to account for uncertainty, which means a higher-uncertainty satellite product directly reduces the number of credits a project can issue. This creates a direct financial incentive to invest in better calibration data, including field plots and airborne LiDAR, rather than relying solely on open satellite products. Satellize structures AGB analytics workflows around this hierarchy: open SAR and GEDI data provide the spatial backbone, commercial or client-supplied field data tighten the calibration, and the delivered product includes a per-pixel uncertainty layer rather than a single-value map.
Practical limits buyers should understand before commissioning a study
No current satellite system produces a wall-to-wall tropical forest AGB map accurate to better than roughly ±20–30% at 1-hectare scale, and that figure worsens in the wettest, most structurally complex forests. BIOMASS may improve this, but its products are not yet validated. Field plots remain indispensable for calibration; a satellite-only approach that skips ground-truth will produce numbers that look precise and are not.
Temporal consistency matters as much as accuracy. Carbon markets require multi-year monitoring, which means the AGB estimation method must be consistent across years even as satellite constellations change. Sentinel-1's free archive back to 2014 is valuable precisely for this reason: it provides a consistent C-band record against which change can be measured, even if absolute AGB retrieval saturates in dense forest. Fusion workflows that combine Sentinel-1 change detection with GEDI-calibrated AGB baselines and, in time, BIOMASS-derived structural layers represent the current state of practice for serious inventory work.
Typical figures
| Spatial resolution (GEDI) | 25 m circular footprint; track spacing ~600 m at equator (not wall-to-wall) |
| Spatial resolution (BIOMASS) | 200 m (target for AGB product); finer for raw SAR imagery |
| Spatial resolution (Sentinel-1) | 10 m (IW mode GRD); 20–40 m effective for backscatter analysis |
| Spatial resolution (ALOS-2 PALSAR-2) | 10 m (single-pol), 25 m (dual-pol FBD mode) |
| SAR frequency / wavelength | C-band ~5.5 cm (Sentinel-1); L-band ~23 cm (PALSAR-2); P-band ~69 cm (BIOMASS) |
| AGB saturation threshold (C-band) | ~50 Mg/ha above-ground dry biomass |
| AGB saturation threshold (L-band) | ~100–150 Mg/ha (tropical conditions) |
| Revisit / temporal cadence | Sentinel-1: 6 days (two-satellite); PALSAR-2: 14 days; GEDI: non-repeating track sampling; BIOMASS: global tropical coverage over 3-year cycle |
| GEDI latitude coverage | ~51.6° N to 51.6° S (ISS orbital inclination limit) |
| Archive depth (Sentinel-1) | From April 2014 (Sentinel-1A launch); open and free via Copernicus Data Space |
Analytics Satellize can run
| Above-ground biomass map with per-pixel uncertainty | SAR backscatter inversion calibrated against GEDI RH metrics; Random Forest or Gaussian process regression; uncertainty propagated from calibration sample size and model variance | GeoTIFF raster stack (AGB estimate + uncertainty band) at client-specified resolution, with methodology note suitable for IPCC Tier 2 reporting |
| Carbon stock change time series | Multi-temporal Sentinel-1 backscatter differencing to detect disturbance and regrowth; change events attributed to AGB loss or gain using pre-disturbance biomass baseline | Annual carbon flux table (Mg CO2-equivalent per stratum) with confidence intervals; compatible with VCS and national inventory formats |
| Forest degradation detection layer | Sentinel-1 coherence and backscatter time series; selective logging and edge degradation detectable as partial backscatter reduction below full deforestation threshold | Quarterly GIS polygon layer flagging degradation events, with severity classification (low / moderate / severe) and date of first detection |
| GEDI-SAR fusion biomass baseline | Spatial prediction of GEDI-derived AGB point estimates across SAR coverage gaps using gradient boosting with SAR backscatter, texture and terrain covariates; cross-validated against withheld GEDI footprints | Wall-to-wall AGB raster at 100 m resolution with cross-validation RMSE reported by forest stratum |
| Peatland above-ground carbon layer | Land-cover classification separating peat-swamp forest, degraded peat and bare peat; AGB estimated for vegetated classes via SAR; bare-peat extent flagged separately as high-priority field-verification zone | Stratified carbon stock summary table and classified raster; bare-peat alert polygons delivered as shapefile |
| Voluntary carbon market monitoring report | Annual AGB change quantification within project boundary; conservative discounting applied per Verra VCS methodology; leakage belt monitoring using same SAR stack | Annual PDF monitoring report with supporting GIS layers and raw data archive, formatted for third-party verification submission |
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.