Intact forest landscape boundary verification for high-conservation-value claims
Intact Forest Landscape designations underpin billions in carbon credits and supply-chain claims, yet their boundaries are static polygons in a landscape that changes daily. Independent satellite verification exposes the gap between what is claimed and what the canopy actually shows.
Sensors
- Maxar WorldView-3: 30 cm panchromatic, 1.24 m multispectral (8 bands including SWIR). Detects tracks as narrow as 2 m wide when contrast with surrounding canopy is sufficient. On-demand tasking; revisit roughly 1 day at mid-latitudes but cloud cover in humid tropics routinely delays acquisition by days to weeks.
- Airbus Pléiades Neo: 30 cm panchromatic, 1.2 m multispectral (6 bands). Comparable resolution to WorldView-3; stereo acquisition in a single pass enables canopy height estimation as a secondary check on structural intactness.
- Sentinel-1 SAR (C-band, IW mode): 10 m ground range resolution, 6-day repeat at the equator (12-day single-satellite). Cloud-penetrating; detects road surfaces and cleared edges through backscatter contrast even under persistent overcast. Freely available from Copernicus. Cannot resolve features below roughly 5 m.
- Planet Basemaps (PlanetScope): 3 m multispectral, near-daily revisit globally. Useful for temporal change detection across large IFL extents and for flagging candidate disturbance zones for subsequent VHR tasking. Spatial resolution is insufficient to confirm track width or settlement structure independently.
What an Intact Forest Landscape boundary actually represents
The IFL concept, developed by Potapov and colleagues at the Global Forest Watch programme and periodically updated, defines areas of at least 500 km² that show no remotely detectable human influence at the Landsat scale (roughly 30 m). The 2020 update mapped approximately 1.11 billion hectares globally. That 30 m detection threshold is the first thing a verification analyst should write down, because it is also the threshold below which a great deal of economically motivated disturbance is deliberately kept.
Carbon project developers and corporate buyers cite IFL status as evidence of high carbon stocks, high biodiversity value and, critically, additionality. The claim is not unreasonable in principle. IFL areas do tend to carry higher above-ground biomass than degraded forests. The problem is that the boundary is a snapshot, the snapshot ages, and the economic incentive to keep disturbance just below the detection limit of the mapping methodology is substantial.
The 30-metre gap and how it is exploited
A Landsat pixel is 900 m². A logging track of 4 m width occupies less than 0.5 percent of a single pixel. At that sub-pixel scale, spectral mixing keeps the pixel green. Selective extraction roads, skid trails and even small clearings for equipment staging can persist for years inside a nominally intact polygon without triggering any update to the IFL dataset. The IFL map is not wrong; it is simply operating at its stated resolution.
This is not a theoretical concern. Published studies using VHR imagery inside IFL boundaries in the Congo Basin and Amazon have found road networks and logging camps that are invisible at Landsat resolution. The disturbance is real, the carbon impact is real, and the boundary verification claim rests on a dataset that was never designed to detect it.
Verification therefore requires a two-stage approach. First, Sentinel-1 SAR time-series analysis across the full claimed boundary at 10 m resolution to flag backscatter anomalies consistent with cleared surfaces or compacted tracks. Second, targeted VHR tasking (WorldView-3 or Pléiades Neo at 30 cm) over flagged zones to confirm or dismiss the anomaly. SAR alone produces false positives from flooding and soil moisture variation; VHR alone is too expensive to cover large extents systematically. The combination is the point.
Temporal drift: the boundary that does not move when the forest does
IFL boundaries are updated periodically, not continuously. The most recent global update available as of 2024 uses data through 2020. A project claiming IFL status in 2024 on the basis of a 2020 polygon is making a four-year temporal assumption. In high-pressure landscapes, four years is long enough for a logging concession to open, operate and partially close.
Proper verification constructs a temporal stack: the baseline IFL polygon, annual Planet Basemap composites from the claim date backwards to the last IFL update, and Sentinel-1 change detection across the same period. Any disturbance event that post-dates the IFL snapshot but pre-dates the carbon credit issuance is material to the claim. Auditors who skip this step are not auditing the project; they are auditing the map.
Settlement edges, road networks and the politics of contested boundaries
Some of the most commercially significant IFL areas overlap with contested land tenure. Indigenous territories, logging concessions and agricultural frontier zones can all share the same polygon. Where tenure is disputed, the IFL boundary itself becomes a political object: governments, concession holders and conservation organisations may each assert a different version of what is inside and what is outside.
Satellite evidence is not politically neutral in this context, but it is at least falsifiable. A 30 cm image either shows a road or it does not. A Sentinel-1 time-series either shows a persistent backscatter change consistent with clearing or it does not. The analyst's job is to document what the sensors show, state the resolution and detection limits clearly, and leave the legal interpretation to the appropriate jurisdiction.
Settlement detection deserves specific attention. Smallholder encroachment at forest edges typically produces a distinctive spectral signature: small rectangular clearings, often aligned with access tracks, with bare soil or early-succession vegetation replacing closed canopy. At 30 cm resolution, individual structures are visible. At 3 m (Planet), the pattern is identifiable even if individual buildings are not. Either way, the edge of human occupation can be mapped with a precision that the IFL methodology was not designed to provide.
What verification can and cannot prove
A well-constructed satellite verification package can confirm or refute the spatial claim: is the area inside the asserted IFL boundary visually and spectrally consistent with closed-canopy, undisturbed forest at the time of credit issuance? It can date disturbance events to within the revisit cycle of the archive used. It can extract road network geometry for comparison against any roads declared in the project design document.
What it cannot do is independently verify above-ground biomass without lidar or field calibration data (those are covered in sibling pages on canopy structure and spaceborne lidar). It cannot resolve the legal status of any disturbance it detects. And it cannot reliably distinguish between a road that was present before the IFL boundary was drawn and one that appeared afterwards, unless a sufficient archive of VHR imagery exists for the specific location, which outside the major commercial corridors it often does not.
Satellize structures these verification packages as auditor-ready evidence sets: georeferenced image chips, change-detection layers, road network extractions and a plain-language findings report, formatted to the standards expected by carbon registries and corporate sustainability auditors. The Tonga crop-estimation programme demonstrated the same evidence-chain discipline in a different domain; the methodology transfers directly to forest boundary work.
Practical thresholds for a credible verification programme
Any organisation commissioning IFL boundary verification should expect to specify three things at the outset: the minimum detectable disturbance width (2 m is achievable with WorldView-3 under good contrast conditions, 5 m is a safer planning assumption), the temporal window to be covered, and the acceptable cloud-gap tolerance (in persistently cloudy regions, SAR must carry the primary detection burden and VHR confirmation may require multiple tasking attempts across a season).
Cost scales with area and cloud probability. A compact project boundary of 50,000 hectares in a seasonally dry forest can be verified with one or two VHR acquisitions and a Sentinel-1 stack. A multi-million-hectare IFL claim in the Congo Basin or the Indonesian archipelago requires a systematic tiling strategy, SAR-first screening, and a realistic expectation that some tiles will remain cloud-obscured for months. Stating that limit honestly is part of the service.
Typical figures
| VHR spatial resolution (panchromatic) | 30 cm (WorldView-3, Pléiades Neo); practical minimum detectable track width ~2 m under high contrast |
| VHR multispectral resolution | 1.2–1.24 m (WorldView-3 and Pléiades Neo); 8-band SWIR available on WorldView-3 |
| SAR resolution (Sentinel-1 IW mode) | 10 m ground range; minimum detectable cleared surface approximately 5 m width |
| SAR revisit (Sentinel-1, equatorial) | 6 days (two-satellite constellation); 12 days single-satellite |
| Planet Basemap resolution | 3 m multispectral; near-daily raw revisit, monthly and quarterly composites available |
| Cloud penetration | Sentinel-1 SAR: full cloud penetration. VHR optical: zero cloud penetration; tropical acquisition delays of days to weeks are routine |
| Archive depth | Sentinel-1: 2014 to present (Copernicus). Landsat: 1972 to present (USGS). Planet: ~2016 to present. Commercial VHR: variable by location, typically 2008 to present in high-interest areas |
| IFL baseline dataset resolution | Landsat 30 m; sub-pixel disturbance (tracks <5 m wide) not detectable in IFL mapping methodology |
| Delivery formats | GeoTIFF change layers, GeoPackage road-network extractions, georeferenced image chips, PDF findings report |
| Temporal dating precision | Disturbance events datable to within the revisit interval of the archive used; typically ±6 days for SAR, ±1–3 months for optical composites |
Analytics Satellize can run
| IFL boundary conformance map | VHR visual interpretation and automated canopy-gap detection against claimed polygon; comparison with IFL 2020 dataset | GeoTIFF conformance layer with flagged discrepancy zones and a tabular area summary |
| Road and track network extraction | Sentinel-1 SAR backscatter change detection for screening; VHR-based semi-automated line extraction (morphological filtering) for confirmed features | GeoPackage road-network layer with width estimates, date-of-first-detection and confidence rating per segment |
| Temporal disturbance chronology | Annual Planet Basemap NDVI differencing and Sentinel-1 coherence change detection across the verification window | Time-stamped change-event table and animated GIF stack for auditor review |
| Settlement edge mapping | VHR object-based image analysis (OBIA) for structure and clearing detection; spectral unmixing at Planet resolution for edge delineation | Polygon layer of identified settlement features with area, date and distance-to-IFL-boundary attributes |
| Cloud-gap risk assessment | Historical cloud-frequency analysis from Sentinel-2 scene metadata for the project area | Per-month cloud-probability raster and recommended acquisition window calendar |
| Auditor-ready evidence package | Compilation of georeferenced image chips, change layers and road extractions into a structured evidence set aligned with carbon registry documentation standards | Zipped archive with a plain-language findings report and a methodology annex stating detection limits |
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.