Sensors
- ESA BIOMASS P-band SAR: 435 MHz P-band, launched April 2024. Full-polarisation backscatter and tomographic modes. P-band wavelength (~70 cm) penetrates closed canopies to interact with trunks and large branches, raising the saturation threshold to roughly 200–300 Mg/ha in some forest types compared with ~100 Mg/ha for L-band. Global coverage planned in repeating 25-day cycles.
- NASA GEDI (Global Ecosystem Dynamics Investigation): Full-waveform lidar on the ISS at ~25 m footprint, ~60 m along-track spacing between footprints. Delivers relative height metrics (RH50, RH75, RH98) correlated with canopy height and, through allometric models, with biomass. Coverage limited to approximately 51.6° N/S by ISS inclination. Not spatially continuous: used as training and validation data.
- JAXA ALOS-2 PALSAR-2: L-band SAR (1.27 GHz, ~23 cm wavelength), 3–10 m resolution in spotlight/stripmap modes, 14-day repeat. Penetrates canopy better than C-band but saturates at lower biomass densities than P-band. Supports the JAXA Global Forest/Non-Forest map and is a proven workhorse for tropical-forest biomass modelling.
- ESA Sentinel-1 C-band SAR: 5.4 GHz C-band, 10 m resolution in IW mode, 6–12 day revisit depending on latitude. Useful for low-biomass stands and forest/non-forest masking but saturates quickly, typically below 50–80 Mg/ha. Valuable for change detection and as a covariate rather than a primary biomass retrieval sensor.
Why radar wavelength determines what you can measure
Microwave radar interacts with forest structure through two dominant mechanisms. Volume scattering occurs as the signal bounces repeatedly within the canopy. Double-bounce scattering occurs when the signal reflects off a trunk and then the ground, or vice versa, producing a strong return correlated with basal area. The longer the wavelength, the deeper the penetration before the signal is attenuated.
C-band (Sentinel-1, ~5 cm wavelength) interacts mainly with leaves and small branches. Backscatter saturates at modest biomass densities, often 50–80 Mg/ha, which covers a large fraction of the world's secondary and plantation forests but misses the bulk of old-growth tropical carbon. L-band (~23 cm) reaches trunks and pushes saturation to roughly 80–150 Mg/ha in many studies, though the figure varies with stand structure and moisture. P-band (~70 cm) penetrates to the forest floor in most closed-canopy conditions, and published modelling work ahead of the BIOMASS mission suggested saturation thresholds of 200 Mg/ha or higher in tropical forests. Those numbers are still being validated against BIOMASS's first acquisitions.
What GEDI actually contributes, and where it cannot go
GEDI fires 1,064 nm laser pulses from the International Space Station and records the full return waveform. The ratio of energy returned at different heights gives relative height metrics: RH98 approximates the 98th-percentile canopy height, RH75 the 75th percentile. These metrics correlate well with above-ground biomass through allometric relationships, and published work using GEDI data has produced canopy height estimates with root-mean-square errors in the range of 2–5 m over many forest types.
The critical limitation is spatial coverage. GEDI samples discrete footprints, not continuous swaths. The ISS orbit restricts coverage to latitudes between roughly 51.6° N and 51.6° S, excluding boreal forests north of that band. The practical role of GEDI in biomass mapping is therefore as a training and calibration layer: its height retrievals are used to build statistical or machine-learning models that predict biomass from spatially continuous SAR backscatter. The model then extrapolates. Where GEDI footprint density is low, extrapolation uncertainty rises.
Fusing the sensors: how the retrieval chain works
The standard fusion workflow runs in several steps. Field inventory plots provide measured stem volume and, through species-specific allometric equations, above-ground biomass in Mg/ha. GEDI height metrics are calibrated against those plots and used to extend height estimates across the landscape. SAR backscatter, in one or more polarisations and wavelengths, is then regressed against the GEDI-derived biomass surface using methods ranging from Water-Cloud Model inversions to random forests and gradient-boosted trees. The trained model produces a wall-to-wall biomass raster at the SAR pixel scale.
Multi-source stacking improves performance. A common approach combines P-band or L-band backscatter, C-band coherence (which responds to canopy moisture and structure independently of backscatter), and GEDI RH metrics as co-predictors. ESA's BIOMASS mission is designed to support tomographic processing in a later phase, which reconstructs a three-dimensional scattering profile through the canopy rather than integrating all returns into a single backscatter value. Tomography could substantially reduce saturation effects but requires multiple passes from consistent geometries and is computationally demanding.
Allometric uncertainty: the error that never disappears
Every satellite-derived biomass estimate inherits uncertainty from the field plots used to build allometric equations. Allometric models convert diameter at breast height, tree height, and wood density into biomass. Published pantropical allometric equations carry uncertainty of roughly 10–30% at the individual tree level, and that propagates into plot-level estimates, then into the satellite model training data, and finally into the map.
This is not a flaw that better sensors will eliminate. It is a structural feature of indirect remote sensing of carbon. Responsible reporting requires that biomass maps include spatially explicit uncertainty layers, typically expressed as 90% confidence intervals in Mg/ha per pixel, and that those layers reflect both model fitting error and allometric propagation. Buyers of biomass data products who receive a single-value raster without uncertainty bounds should treat the product with caution. The IPCC guidance for national greenhouse gas inventories requires uncertainty quantification for exactly this reason.
What the maps can and cannot support
A well-constructed SAR-lidar fusion product can support national-scale carbon stock baselines, identify broad spatial gradients in forest density, and detect large-magnitude biomass change between epochs. At the pixel level in dense tropical forest, absolute accuracy is typically reported at plus or minus 20–40 Mg/ha in peer-reviewed benchmarking, though performance varies substantially with forest type, terrain, and the density of field calibration plots.
The maps are less suited to detecting small or diffuse biomass losses, such as selective logging that removes a few trees per hectare, because the signal change falls within the retrieval uncertainty. That use case belongs to degradation-specific methods covered separately. Temporal change detection using repeat SAR acquisitions can track large disturbances, but interpreting a backscatter decrease as a specific biomass loss requires careful separation from moisture-driven backscatter variation, which can mimic structural change.
Satellize builds biomass retrieval pipelines on open constellations (Sentinel-1, ALOS-2 archive, GEDI) and is integrating BIOMASS data as it becomes available through ESA's distribution channels. The methodology is the same class used in the Tonga crop-estimation programme: field-calibrated statistical models run against spatially continuous satellite inputs, with explicit uncertainty propagation reported at delivery.
Practical expectations for a national programme
A country commissioning a national above-ground biomass map should expect to provide, or fund the collection of, a network of field inventory plots distributed across the major forest strata. Sparse plot networks in heterogeneous forest produce wide uncertainty intervals regardless of sensor quality. The satellite data is not a substitute for ground truth; it is an interpolation engine between ground truth points.
Delivery timelines depend on SAR data availability and processing. Sentinel-1 and ALOS-2 data are available with latencies of days to weeks. BIOMASS data distribution timelines are governed by ESA's data policy and are still being established for the operational phase. A national baseline map, including uncertainty layers and validation report, is realistically a six-to-twelve month project from plot data receipt to final product, assuming no major gaps in the input SAR archive.
Typical figures
| Primary SAR frequency (BIOMASS) | P-band, 435 MHz (~70 cm wavelength), full polarisation |
| Primary SAR frequency (ALOS-2 PALSAR-2) | L-band, 1.27 GHz (~23 cm wavelength) |
| SAR spatial resolution | BIOMASS: ~50 m (nominal biomass product); ALOS-2: 3–25 m depending on mode; Sentinel-1 IW: 10 m |
| GEDI footprint and spacing | ~25 m footprint diameter; ~60 m along-track between footprints; not continuous |
| Revisit cadence | BIOMASS: ~25-day global repeat; ALOS-2: 14 days; Sentinel-1: 6–12 days |
| Biomass saturation threshold (approximate) | Sentinel-1 C-band: ~50–80 Mg/ha; ALOS-2 L-band: ~80–150 Mg/ha; BIOMASS P-band: estimated 200–300 Mg/ha (validation ongoing) |
| GEDI latitudinal coverage | ~51.6° N to 51.6° S (ISS orbit constraint) |
| Typical retrieval uncertainty (dense tropical forest) | ±20–40 Mg/ha at pixel level; varies with plot network density and forest type |
| Archive depth | Sentinel-1: from 2014; ALOS-2: from 2014; GEDI: from 2019; BIOMASS: from 2024 |
| Standard deliverable format | GeoTIFF raster (biomass estimate + uncertainty layer), GPKG validation report, optional COG for web delivery |
Analytics Satellize can run
| National above-ground biomass baseline map | SAR backscatter inversion (Water-Cloud Model or machine-learning regression) calibrated against GEDI RH metrics and field plots | Wall-to-wall GeoTIFF at 25–50 m resolution with per-pixel 90% confidence interval layer and methodology report |
| Biomass change detection between two epochs | Differencing of SAR-derived biomass rasters with significance thresholding based on propagated uncertainty | Change magnitude raster (Mg/ha lost or gained) with statistical significance mask and summary statistics by administrative unit |
| GEDI-calibrated canopy height surface | Spatial interpolation and SAR-assisted extrapolation of GEDI RH98 footprint data | Continuous canopy height GeoTIFF used as input to biomass model or delivered as standalone product |
| Stratified forest carbon stock summary | Zonal statistics over biomass raster applied to forest-type or administrative strata | Tabular report (CSV/XLSX) of mean, total, and uncertainty-bounded carbon stock by stratum, suitable for national inventory submission |
| SAR saturation risk assessment | Backscatter-biomass curve fitting per forest stratum to identify pixels where retrieval is operating near or above the saturation threshold | Saturation-flag raster overlaid on biomass map, with recommendations for supplementary field sampling in flagged zones |
| Allometric uncertainty propagation report | Monte Carlo propagation of allometric equation error through model training and prediction chain | Written technical annex quantifying allometric, model, and sensor contributions to total uncertainty, formatted for IPCC Tier 2 or Tier 3 reporting |
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.