Selective logging intensity and volume extraction mapping
Spaceborne lidar and high-resolution optical imagery can estimate timber volume removed per logging event, answering not just whether cutting occurred but how much. Combining GEDI waveform data with Planet SuperDove time series turns canopy geometry into extraction accounts.
Sensors
- GEDI (NASA Global Ecosystem Dynamics Investigation): Full-waveform lidar aboard the ISS. Footprint diameter approximately 25 m, with along-track shot spacing of 60 m and eight parallel tracks offset by roughly 600 m. Measures canopy height, canopy cover, vertical foliage profile and relative height metrics (RH50, RH98) that correlate directly with above-ground biomass. Repeat coverage at a given point is irregular and latitude-limited to ±51.6°, so dense temporal stacking requires months to years rather than days.
- ICESat-2 ATL08 (NASA): Photon-counting lidar with three pairs of 17 m ground tracks. The ATL08 land-vegetation product reports terrain and canopy height at 100 m segments. Finer along-track resolution is achievable by reprocessing raw photon clouds. Complements GEDI where waveform saturation is a concern in very dense canopy, though spatial coverage per pass is sparser.
- Planet SuperDove: Eight-band multispectral constellation at 3 m native resolution (resampled to 3 m product). Daily to near-daily revisit globally. The red-edge and NIR bands support canopy gap fraction retrieval and fractional cover change. Used here as the change-detection layer that localises where GEDI waveform differences should be interrogated.
- Sentinel-2 MSI (ESA): 13-band multispectral at 10 m (visible/NIR) and 20 m (red-edge, SWIR). Five-day revisit at the equator with two satellites. Free and openly archived since 2015. Provides the longer historical baseline for pre-harvest canopy state and supports fractional canopy cover indices such as NDVI and LAI proxies. Cloud remains the principal constraint in humid tropical zones.
What a floating roof gives away
A forest canopy is, in physical terms, a surface with a measurable height and a measurable gap fraction. Remove a tree and both change. The key insight behind volume-extraction mapping is that the relationship between canopy gap size, gap depth and the stem volume that caused it is not arbitrary. Published allometric work in tropical forests shows that a felled tree of a given diameter at breast height produces a predictable gap area in the canopy layer, plus a secondary zone of collateral damage from the crown fall. Integrating gap area and the change in canopy height across a logging event therefore gives an estimate of extracted volume, not merely a record that disturbance occurred.
This is the distinction that separates extraction accounting from deforestation alerting. Alerting asks: did the canopy change? Extraction accounting asks: by how much did the canopy height distribution shift, and what volume of timber does that shift imply? The two questions require different sensors and different maths.
How lidar turns height loss into cubic metres
GEDI's full-waveform returns encode the vertical distribution of canopy material within each 25 m footprint. The relative height metric RH98, the height below which 98 percent of the return energy falls, is a close proxy for top-of-canopy height. Pre- and post-harvest differences in RH98 across footprints that intersect a logged area give a canopy height model difference. That difference, combined with the gap fraction change estimated from Planet SuperDove fractional cover, feeds into published volume-estimation frameworks that relate canopy structural change to extracted stem basal area.
ICESat-2 ATL08 adds a second independent height observation. Where the two lidar datasets agree on the magnitude of height loss, confidence in the extraction estimate rises. Where they diverge, the discrepancy often flags partial crown damage rather than full stem removal, which carries a different volume signature. Neither sensor alone is sufficient: GEDI has the waveform richness but irregular revisit; ICESat-2 has denser along-track sampling but coarser segment resolution in the standard product.
The honest limit here is footprint density. In a concession of, say, 50,000 hectares, GEDI may accumulate enough footprints for statistically meaningful height-change estimates over a three-to-six-month window, but it cannot resolve individual tree removals. The method produces concession-level or coupe-level extraction estimates, not stem-by-stem tallies.
Planet SuperDove as the spatial backbone
Lidar footprints are sparse in space. Planet SuperDove at 3 m resolution provides the continuous spatial fabric that tells you where to look. A time series of SuperDove images, differenced before and after a logging event, maps canopy gap openings at a scale where individual crown gaps of 50 to 200 square metres are detectable. The red-edge band (band 6 on SuperDove, centred near 705 nm) is particularly sensitive to fractional canopy cover loss because chlorophyll absorption drops sharply at that wavelength, making partial gap openings visible that a simple NDVI difference would miss.
The operational workflow pairs the SuperDove gap map with the GEDI height-change layer. SuperDove defines the spatial extent and gap-size distribution of each logging event. GEDI provides the height-loss magnitude at footprints within that extent. Together they parameterise the gap-fraction-to-volume model. Sentinel-2 extends the pre-harvest baseline back to 2015 at no cost, which is often necessary when a concession has a long and poorly documented extraction history.
Cloud cover is the unavoidable constraint in tropical regions. In the Congo Basin or Indonesian Borneo, persistent cloud can obscure a logging event for weeks. Planet's daily revisit improves the probability of obtaining a cloud-free observation within a given window, but it does not eliminate the problem. Analysts should expect some events to be partially obscured and should report the cloud-affected fraction of any concession assessment honestly.
From gap geometry to a volume estimate: the published physics
The physical basis draws on two well-established relationships. First, canopy gap fraction and gap size distribution are related to the leaf area index and canopy cover through Beer-Lambert radiative transfer, a relationship formalised in numerous published studies using airborne lidar and field validation. Second, tropical timber species follow published allometric equations that relate crown projected area to stem diameter and, from diameter, to merchantable volume. The FAO and IPCC both publish regional allometric tables that form the reference for this conversion.
In practice, the conversion carries uncertainty. Allometric equations have species-specific and site-specific variance. Selective logging in tropical forests typically targets a small number of high-value species whose crown geometry may differ from the regional average. The method therefore produces a volume estimate with an honest confidence interval, typically reported as a range rather than a point value, and the interval widens when species composition is unknown. Buyers should treat the output as an independent cross-check on concession harvest declarations, not as a substitute for ground-based forest inventory.
What the method cannot see
Selective logging in closed-canopy forest is designed, partly, to be inconspicuous. Stems felled under a closed upper canopy may not open a gap large enough to register in a 3 m optical image if the crown was suppressed. Very-low-intensity harvesting, fewer than one or two stems per hectare, may fall below the detection threshold of the combined method. The minimum detectable gap in SuperDove imagery is roughly 30 to 50 square metres under good atmospheric conditions, which corresponds to a crown of moderate size. Smaller gaps, or gaps obscured by adjacent crowns closing laterally, will be missed.
Skid trail damage, the collateral canopy loss from dragging logs to extraction points, is often larger in area than the felling gaps themselves. The method captures skid trail gaps as part of the total disturbed area, which is correct for total extraction impact but means the volume estimate conflates the felled-tree signal with collateral damage. Separating the two requires additional contextual analysis, such as linear feature detection to identify trail geometry.
Satellize applies this combined GEDI and Planet workflow within its analytics service, where concession-level extraction reports can be scheduled quarterly or triggered by a SuperDove change-detection alert. The Tonga crop-estimation programme demonstrated the team's ability to translate canopy structural signals into quantitative agricultural outputs; the same radiometric rigour applies here in a forestry context.
Delivering a number a regulator can use
The end product of this analysis is a per-coupe extraction estimate expressed in cubic metres of merchantable timber, with a stated uncertainty range, a map of detected canopy gaps attributed to the logging event, and a comparison against any available harvest declaration. That comparison is the commercially and legally relevant output. Governments issuing timber concessions, certification bodies auditing FSC or PEFC claims, and commodity banks financing forestry operations all need an independent, satellite-derived volume figure they can set against the declared harvest.
Delivery formats are GIS layers (GeoTIFF canopy height difference rasters, vector gap polygons) plus a structured report that documents the sensor dates used, the cloud-affected fraction, the allometric assumptions applied, and the resulting volume range. Latency from tasking to delivery depends primarily on the GEDI accumulation window required to achieve adequate footprint density over the concession, which is typically two to four months for a first assessment and shorter for subsequent monitoring cycles once the baseline is established.
Typical figures
| Optical spatial resolution | 3 m (Planet SuperDove); 10 m visible/NIR, 20 m red-edge/SWIR (Sentinel-2) |
| Lidar footprint diameter | ~25 m (GEDI); ~17 m track width, 100 m segment (ICESat-2 ATL08 standard product) |
| Optical revisit | Daily to near-daily (Planet SuperDove); 5 days at equator (Sentinel-2, two-satellite) |
| Lidar revisit at a point | Irregular; GEDI accumulates adequate footprint density over a given area in approximately 2–6 months depending on latitude and concession size |
| Canopy height detection limit | RH98 change of ~1–2 m detectable at GEDI footprint level; absolute accuracy ±2–3 m in closed tropical canopy per published GEDI validation |
| Minimum detectable gap (optical) | ~30–50 m² under good atmospheric conditions in Planet SuperDove imagery |
| Spectral bands used | Red-edge (~705 nm), NIR, SWIR (SuperDove and Sentinel-2); full-waveform 1064 nm (GEDI); 532 nm photon-counting (ICESat-2) |
| Archive depth | Sentinel-2: from 2015; Planet SuperDove: from approximately 2021 at full 8-band; GEDI: from April 2019; ICESat-2: from October 2018 |
| GEDI latitude coverage | ±51.6° (ISS orbital inclination limit; excludes high-latitude boreal forests) |
| Delivery formats | GeoTIFF canopy height difference raster, vector gap polygons (GeoPackage/Shapefile), structured PDF/Excel extraction report with uncertainty ranges |
Analytics Satellize can run
| Pre/post-harvest canopy height difference raster | GEDI RH98 differencing across footprints intersecting the logging event, interpolated using SuperDove gap extent as a spatial mask | GeoTIFF layer showing height loss in metres per 25 m cell, with footprint density map |
| Canopy gap map with size-class distribution | SuperDove red-edge fractional cover change detection; gap polygons classified by area into size bins (<100 m², 100–500 m², >500 m²) | Vector polygon layer (GeoPackage) with gap area, date of first detection and cloud-flag attribute |
| Per-coupe timber volume extraction estimate | Gap-fraction-to-basal-area conversion using published Beer-Lambert canopy radiative transfer, combined with regional FAO/IPCC allometric equations for merchantable volume | Tabular report per coupe: estimated volume range (m³), stated uncertainty interval, allometric assumptions used |
| Harvest declaration cross-check | Comparison of satellite-derived volume estimate against concession holder's declared harvest tonnage or volume, with discrepancy flag and confidence rating | Structured PDF report suitable for regulatory or certification body submission |
| Skid trail and extraction infrastructure footprint | Linear feature detection on SuperDove imagery using directional morphological filtering to separate trail geometry from felling gaps | Vector line layer of detected skid trails with estimated disturbed-area buffer |
| Quarterly extraction monitoring time series | Repeat SuperDove change detection triggering GEDI/ICESat-2 extraction query on a scheduled basis; cumulative volume tracked across reporting periods | Dashboard-ready GIS feed updated quarterly, with alert triggered when new gap area exceeds a client-defined threshold |
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.