Spaceborne lidar canopy height and forest structure
ICESat-2 and GEDI fire laser pulses through forest canopies and time the returning photons to map vertical structure at scales no passive sensor can reach. Each system has distinct geometry, sensitivity and coverage gaps that buyers must understand before committing to a monitoring design.
Sensors
- ICESat-2 ATLAS: Photon-counting lidar at 532 nm. Six beams arranged in three pairs, each pair separated by 3.3 km across-track, beams within a pair separated by 90 m. Along-track shot spacing approximately 0.7 m; effective footprint diameter roughly 17 m. 91-day exact-repeat orbit, polar-inclusive coverage to ±88° latitude. Vertical precision better than 10 cm over flat ice; forest retrievals noisier due to multiple scattering and canopy clutter.
- GEDI (Global Ecosystem Dynamics Investigation): Full-waveform lidar at 1064 nm, mounted on the International Space Station. Footprint diameter approximately 25 m, along-track shot spacing approximately 60 m, eight beam tracks with across-track spacing that varies with ISS attitude. Coverage limited to ±51.6° latitude by the ISS orbital inclination, excluding high-latitude boreal forests. Waveform decomposition yields canopy height, vertical foliage profile and ground return directly.
- Sentinel-2 MSI (supporting layer): 10 m multispectral imagery used to extrapolate lidar-derived height samples across continuous wall-to-wall maps via machine-learning fusion. Covered on the multispectral vegetation index page; included here only as the spatial bridge between sparse lidar tracks.
- Landsat 8/9 OLI (supporting layer): 30 m optical archive back to 1972 used to provide temporal context and spectral covariates for biomass modelling when fused with GEDI or ICESat-2 samples. Same extrapolation role as Sentinel-2; chosen when longer time series matter more than spatial resolution.
What a waveform actually tells you
Passive optical sensors see the top of the canopy. Lidar sees through it. When a laser pulse hits a forest, photons scatter back from every surface they encounter: the uppermost leaves, intermediate branches, understorey shrubs and eventually the ground. The time delay between those returns encodes height directly, because light travels at a fixed speed. From that vertical profile you can read canopy top height, the height at which a given fraction of the canopy bulk sits (the so-called relative height metrics, RH50, RH75, RH98 and so on), and the ratio of plant material to open sky at each elevation layer.
GEDI records the full analogue waveform of returning energy. A Gaussian decomposition of that waveform separates discrete return peaks, each corresponding to a canopy layer or the ground. ICESat-2 ATLAS works differently: it counts individual photons with timing precise enough to reconstruct a point cloud along the track, then applies statistical noise filtering to isolate signal photons from solar background. Both approaches yield canopy height; they differ in what else they reveal. GEDI's waveform is richer for understorey structure in moderate-density forests. ICESat-2's photon cloud is denser along-track and works at night with lower solar noise, which matters in high-reflectance tropical conditions.
The geometry problem neither system has solved
Both instruments sample in narrow along-track swaths. GEDI's eight beams cover a swath of roughly 4.2 km but with gaps between individual footprints of about 60 m along track and irregular spacing across track depending on ISS yaw steering. ICESat-2's six beams span 6.6 km across-track but the beams themselves are point samples, not a continuous strip. The result is the same for both: you get precise vertical measurements at sampled locations, and nothing between them.
Turning sparse samples into a wall-to-wall canopy height map requires a fusion step. The standard published approach pairs lidar height metrics with coincident spectral and topographic covariates from Sentinel-2 or Landsat, then trains a regression or random-forest model to predict height everywhere the lidar did not look. The Global Land Analysis and Discovery group at the University of Maryland has published global canopy height maps using exactly this method with GEDI data. The accuracy of the extrapolated surface depends heavily on how well the spectral covariates correlate with height in the target forest type, and it degrades in structurally diverse or fragmented landscapes where a 10 m or 30 m pixel mixes many canopy conditions.
Humid tropics: where the data thins out
Cloud and aerosol extinction are the dominant operational constraint on GEDI in the humid tropics. The 1064 nm wavelength is less affected by molecular scattering than visible wavelengths, but optically thick cloud still blocks the pulse entirely, and dense smoke from agricultural burning attenuates the return enough to corrupt waveform decomposition. Published analyses of GEDI data availability over the Amazon basin show that cloud cover reduces usable shot density by 40 to 70 per cent in the wet season, with some grid cells losing nearly all valid observations for months at a time.
ICESat-2 at 532 nm is more sensitive to aerosol extinction than GEDI, making the problem worse in smoke-affected regions. Both systems accumulate valid shots over time, so a multi-year composite can fill most gaps, but that means the height estimate you are working with may be two or three years old in persistently cloudy areas. For deforestation monitoring in near-real time, lidar alone is insufficient; it needs to be paired with SAR-based canopy-loss detection, which operates regardless of cloud cover.
GEDI's ISS orbit also imposes a hard latitudinal ceiling at ±51.6°. Boreal forests in Canada, Russia and Scandinavia, which hold enormous carbon stocks, are entirely outside GEDI's reach. ICESat-2 covers those regions, but its photon-counting approach is less straightforward for understorey structure retrieval in dense conifer stands where ground returns are frequently obscured.
From height to biomass: the conversion is not free
Canopy height is a proxy for aboveground biomass, not a direct measurement of it. The relationship between height and biomass varies by forest type, stand age, species composition and disturbance history. Allometric equations that convert height and canopy cover to tonnes of carbon per hectare are calibrated from destructive field harvests, and those calibrations carry their own uncertainty, typically ±20 to 40 per cent at the plot level in tropical forests according to published assessments of pantropical biomass maps.
That uncertainty is worth stating plainly because it affects what lidar-derived biomass estimates can and cannot support. For national greenhouse gas inventory reporting under UNFCCC frameworks, the uncertainty must be quantified and documented. Lidar-derived estimates are generally more defensible than those from optical indices alone, because height is a more direct structural measurement than spectral reflectance, but they are not a substitute for ground-truth plots. A credible monitoring programme combines lidar sampling, spectral wall-to-wall mapping and a statistically designed field campaign. Satellize structures its forest analytics engagements around that three-layer logic, drawing on the same open data pipelines it uses for programmes such as the Tonga crop-estimation work.
What the archive already holds
GEDI data collection began in April 2019 following installation on the ISS. The mission has experienced interruptions due to ISS scheduling conflicts, but as of the publicly available record it has accumulated several years of global shots between ±51.6°. ICESat-2 launched in September 2018 and has operated continuously since, building a multi-year archive with 91-day exact repeats that allow change detection at the track level.
Both datasets are freely available. GEDI products (L1B through L4B, including the gridded aboveground biomass density product at 1 km) are distributed through NASA Earthdata. ICESat-2 ATL08 is the land and vegetation product, providing canopy and terrain height at 100 m segments along each beam. The archive depth means that for many forested regions, a baseline canopy height estimate from 2019 to 2020 already exists, and subsequent acquisitions can be differenced to detect height loss or gain consistent with logging, regrowth or disturbance.
Typical figures
| ICESat-2 ATLAS footprint diameter | ~17 m |
| GEDI footprint diameter | ~25 m |
| ICESat-2 along-track shot spacing | ~0.7 m |
| GEDI along-track shot spacing | ~60 m |
| ICESat-2 orbital repeat | 91 days (exact repeat); polar-inclusive to ±88° |
| GEDI latitudinal coverage | ±51.6° (ISS inclination limit; boreal forests excluded) |
| Vertical precision (ICESat-2, flat surface) | <10 cm; forest retrievals typically ±1–3 m depending on canopy density |
| GEDI canopy height accuracy (published) | RMSE ~2–5 m depending on forest type and waveform quality |
| Archive depth | ICESat-2 from September 2018; GEDI from April 2019 |
| Data access | Both freely available via NASA Earthdata (no licence fee) |
Analytics Satellize can run
| Wall-to-wall canopy height map | Lidar height metric extraction (ATL08 or GEDI L2A) fused with Sentinel-2 spectral covariates via random-forest regression; published GLAD methodology | GeoTIFF raster at 10–30 m resolution, delivered as GIS layer with uncertainty band per pixel |
| Aboveground biomass density estimate | GEDI L4A shot-level biomass predictions aggregated and gap-filled using GEDI L4B gridded product methodology; allometric uncertainty propagated | Gridded biomass map (tonnes C per hectare) with documented uncertainty range, formatted for UNFCCC inventory reporting |
| Canopy height change detection | Differencing of multi-year ICESat-2 ATL08 or GEDI L2A height metrics on repeat tracks to identify statistically significant height loss or gain | Change polygon layer flagging areas of height loss >2 m, with acquisition dates and confidence scores, updated per new overpass |
| Forest vertical structure profile | GEDI L2B relative height and plant area index metrics decomposed by canopy layer; used to classify forest structural types (closed canopy, open woodland, degraded) | Classified structural map with per-class area statistics, delivered as GIS layer and tabular summary report |
| Deforestation alert integration layer | Lidar-derived baseline height combined with SAR-based canopy-loss detection to confirm and quantify volume of loss events; reduces false positives from optical-only alerts | Alert feed with estimated volume loss per event, refreshed on SAR acquisition cadence (independent of cloud cover) |
| Carbon stock baseline for jurisdictional REDD+ | Plot-calibrated allometric model applied to fused lidar-spectral height surface; stratified random sampling design to quantify uncertainty per IPCC Tier 3 guidance | Jurisdiction-level carbon stock report with stratified uncertainty bounds and sampling design documentation |
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.