Old-growth forest structural complexity mapping from spaceborne lidar
Canopy height alone cannot distinguish old-growth forest from a mature plantation. GEDI waveform lidar reveals vertical structure, gap fractions and understory layers that betray ecological age, even from 400 km up.
Sensors
- GEDI (Global Ecosystem Dynamics Investigation): Full-waveform lidar on the ISS. Footprints of approximately 25 m diameter, spaced 60 m along-track and 600 m across-track. Delivers canopy height (RH95, RH98), plant area index profiles, foliage height diversity and ground-return separation. Operational since April 2019; coverage between roughly 51.6° N and S.
- ICESat-2 ATL08: Photon-counting lidar. Ground-track spacing of approximately 3.3 km, 100 m along-track segments for vegetation products. Provides canopy height and relative height percentiles. Complements GEDI in higher latitudes and offers a second independent height estimate for cross-validation.
- Sentinel-1 SAR (C-band): C-band synthetic aperture radar at 10 m resolution, 6-day repeat at the equator for dual-satellite operations. Backscatter and coherence sensitive to canopy volume and moisture. Used to spatially extrapolate GEDI point samples across continuous wall-to-wall coverage, filling the gaps between non-contiguous footprints.
- Sentinel-2 MSI: 13-band multispectral imager at 10 to 20 m resolution, 5-day revisit. Provides spectral proxies for canopy structure such as NDVI, red-edge chlorophyll indices and texture metrics. Useful for time-series phenology to separate evergreen old-growth from seasonally deciduous secondary growth of similar height.
What a waveform knows that a canopy-height map does not
A 30-year-old eucalyptus plantation and a 300-year-old temperate rainforest can share almost identical top-of-canopy height. The difference lives below that ceiling. Old-growth stands accumulate structural complexity over decades: multiple distinct canopy layers, a dense shrub and fern understory, standing deadwood, and frequent gap openings where fallen giants let light reach the floor. A single-return or photogrammetric canopy model collapses all of that into one number.
GEDI does not. Its full waveform records the time-resolved return of a laser pulse as it passes through successive canopy layers and reaches the ground. From that waveform, analysts extract relative height percentiles (RH25 through RH98), the ratio of mid-canopy to top-canopy energy, plant area index at multiple height bins, and the foliage height diversity index, a Shannon-entropy measure of how evenly foliage is distributed vertically. Old-growth stands consistently show higher foliage height diversity and a stronger mid-canopy return relative to the peak, compared with even-aged plantations or early secondary regrowth of similar RH95.
The sampling problem: brilliant physics, awkward geometry
GEDI's limits deserve honesty. The 25 m footprint is generous by lidar standards but coarse relative to individual tree crowns in many forest types. More consequentially, the ISS orbit produces a non-contiguous, quasi-random sampling pattern. Across-track spacing of roughly 600 m means that at any single overpass, most of a forest remains unsampled. Over months of accumulation the coverage thickens, but it never becomes wall-to-wall. For a protected area of a few thousand hectares, the number of usable footprints after quality filtering (cloud, slope, urban adjacency flags) can be surprisingly thin.
The standard response is statistical extrapolation: use the GEDI samples as training labels and spatially continuous predictors, typically Sentinel-1 backscatter and coherence, Sentinel-2 spectral and texture bands, and terrain derivatives, as covariates in a random-forest or gradient-boosted regression. The result is a 10 m or 20 m wall-to-wall map of predicted structural metrics. The prediction carries uncertainty, and that uncertainty should be reported per pixel rather than suppressed. In areas of steep terrain, GEDI waveforms are distorted by slope, and the ground-return separation degrades; slopes above roughly 30 degrees warrant caution.
ICESat-2 ATL08 adds a second lidar source with different orbital geometry and photon-counting physics. It does not replace GEDI's waveform richness, but its independent height estimates allow cross-validation and can extend coverage into boreal and sub-Arctic latitudes where GEDI's 51.6-degree inclination limit leaves gaps.
Separating old-growth from tall secondary forest: the structural fingerprint
The core analytical challenge is that secondary forests can reach 30 to 40 m canopy height within 40 to 60 years in humid tropics, overlapping the height range of genuine old-growth. The discriminating features are structural, not altitudinal. Published work using GEDI data has shown that foliage height diversity and the ratio of RH50 to RH98 are among the strongest discriminators. Old-growth stands tend to have RH50/RH98 ratios above 0.55 to 0.65, indicating that half the canopy energy comes from well below the top, a signature of dense mid-story development. Secondary forests, even tall ones, tend to be more top-heavy.
Sentinel-1 C-band coherence adds a complementary signal. Coherence over dense, structurally complex canopies decorrelates rapidly between passes because the many small scatterers move independently. Over structurally simpler, even-aged stands, coherence is marginally higher. The effect is subtle at C-band and should not be over-interpreted, but as one variable among many in a fusion model it contributes discriminating power. L-band SAR, where available from commercial sources, penetrates deeper and is more sensitive to large woody structure, though that is outside the open-constellation scope of this page.
Cloud cover, latency and what the archive can and cannot answer
GEDI data are publicly archived and searchable through NASA Earthdata. The Level 2A and 2B products provide footprint-level canopy height and plant area index profiles. Processing latency from acquisition to public release has typically been weeks to a few months, making GEDI unsuitable for near-real-time monitoring but well suited to annual or biennial structural assessments.
Cloud cover does not affect lidar the way it affects optical sensors. GEDI's laser is blocked by optically thick cloud, and affected footprints are flagged in the quality field. In persistently cloudy tropical regions, the effective sampling density after quality filtering can drop by 30 to 50 percent seasonally. Analysts should compute cloud-filtered footprint density before committing to a mapping extent, rather than discovering thin coverage after the model is built.
The Sentinel-1 fusion layer is cloud-independent and available globally at 10 m, with a consistent archive back to 2014 for many regions. This makes it possible to examine whether structural-complexity predictions have shifted over a decade, a useful check for monitoring secondary-forest maturation or detecting selective disturbance that does not trigger a complete canopy-height change.
From map to conservation decision
A structural-complexity map is not the same as an old-growth classification, and conflating the two causes policy errors. The map identifies stands with structural attributes consistent with old-growth condition. Ground verification, species-composition data and disturbance history are needed to confirm ecological status. What the map does well is prioritise. In a large forested landscape, it can direct limited field teams to the highest-value candidates, and it can flag structurally simple stands within a nominally protected area that may have been selectively logged without triggering a detectable canopy-height change.
Satellize runs structural-complexity analytics on open GEDI and Sentinel archives, producing per-footprint metric extractions, wall-to-wall fusion layers and change summaries as GIS deliverables. The workflow is the same class of approach used in the Tonga crop-estimation programme: open-constellation data, rigorous uncertainty quantification, and outputs calibrated to a specific decision rather than a generic map product. Conservation agencies and forest-tenure bodies wanting to define high-value zones for protection or payment-for-ecosystem-services schemes are the natural clients for this analysis.
Typical figures
| GEDI footprint diameter | ~25 m |
| GEDI along-track footprint spacing | ~60 m |
| GEDI across-track beam spacing | ~600 m (non-contiguous) |
| GEDI latitudinal coverage | ~51.6° N to 51.6° S |
| Sentinel-1 SAR spatial resolution | 10 m (IW mode, GRD) |
| Sentinel-1 revisit (dual satellite) | 6 days at equator |
| Sentinel-2 spatial resolution | 10 m (VIS/NIR), 20 m (red-edge, SWIR) |
| Sentinel-2 revisit | 5 days (dual satellite) |
| GEDI archive depth | April 2019 to present |
| Minimum detectable structural unit | Stand-level (~1 ha aggregation recommended for reliable extrapolation) |
Analytics Satellize can run
| Per-footprint GEDI structural metrics | GEDI L2A/L2B waveform product extraction: RH25, RH50, RH75, RH95, RH98, foliage height diversity, plant area index profiles | Point-vector GIS layer (GeoPackage or Shapefile) with per-footprint metric table and quality flags |
| Wall-to-wall structural-complexity map | Random-forest or gradient-boosted regression using GEDI footprints as labels and Sentinel-1 backscatter/coherence plus Sentinel-2 spectral/texture bands as continuous covariates | 10 m or 20 m raster (GeoTIFF) of predicted foliage height diversity and RH50/RH98 ratio, with per-pixel uncertainty band |
| Old-growth structural-condition classification | Threshold and cluster analysis on fusion-model outputs, informed by published RH50/RH98 and foliage height diversity breakpoints from peer-reviewed GEDI studies | Classified raster (high / moderate / low structural complexity) with confidence layer, plus area statistics by protection zone |
| Decadal structural-change summary | Multi-year GEDI footprint comparison combined with Sentinel-1 coherence time series to detect gradual canopy structural degradation not visible in height-only metrics | Change-detection report with flagged polygons and before/after metric tables |
| Sampling-adequacy assessment | Cloud-filtered footprint density computation and spatial autocorrelation analysis to determine whether GEDI coverage is sufficient for reliable extrapolation across the target area | Coverage diagnostic map and written assessment, delivered before full mapping engagement is scoped |
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.