Post-fire regeneration trajectory monitoring for carbon permanence assessment
After fire crosses a carbon project boundary, spectral recovery and biomass recovery diverge for years. Sentinel-2 NBR time-series and GEDI biomass products can distinguish genuine carbon permanence from a green veneer over a structurally depleted stand.
Sensors
- Sentinel-2 MSI: 10 m resolution in visible and near-infrared bands; 20 m in the shortwave-infrared bands (B11, B12) used for NBR calculation. Five-day revisit at the equator with both satellites. Cloud cover is the principal constraint in tropical project areas, where consecutive clear acquisitions can be separated by weeks rather than days.
- Landsat Collection 2 (Landsat 8 and 9 OLI): 30 m resolution, 16-day single-satellite revisit. The archive extends to 1972 for Landsat 1 MSS and to 1984 for the TM-era dNBR record, making it the only source for pre-fire baseline dNBR values in projects registered before the Sentinel era. Collection 2 surface-reflectance products include uncertainty layers.
- GEDI L4B gridded biomass product: Mean above-ground biomass density at 1 km grid cells, derived from waveform lidar footprints of approximately 25 m diameter. GEDI operates from the International Space Station at latitudes between roughly 51.6° N and S. The L4B product carries published prediction interval estimates; single-footprint uncertainty can exceed 50% in structurally complex or steep terrain.
- VIIRS 375 m active fire product: Near-real-time fire detection at 375 m pixel resolution with twice-daily overpass from Suomi-NPP and NOAA-20. Used here not for fire-risk assessment but to establish the precise burn date and spatial extent of the ignition event inside the project boundary, anchoring the start of the recovery time-series.
Why permanence fails quietly, not dramatically
Carbon registries require that sequestered carbon remain stored for decades. A fire that burns through a project area is an obvious threat, but the more insidious problem is what happens next. If the stand regenerates to its pre-disturbance structure and biomass density within the crediting period, permanence may survive. If it transitions to a lower-biomass stable state, perhaps shrubland or early-successional pioneer forest, the carbon credits already issued become a liability. Registries differ on how they handle this: some require buffer pool contributions, others mandate reversal reporting. All of them require evidence.
That evidence is increasingly expected to be satellite-derived and time-stamped. The question is whether the available sensors can actually answer the question being asked, which is not 'has the canopy recovered?' but 'has the carbon stock recovered?'. Those are different questions, and conflating them is the central risk in post-fire monitoring.
What NBR time-series can and cannot tell you
The normalised burn ratio, calculated from near-infrared and shortwave-infrared reflectance, is sensitive to both char and live vegetation moisture. In the months after a fire, dNBR (the difference between pre-fire and post-fire NBR) tracks burn severity reliably. The composite burn index, validated against field plots in multiple published studies, correlates well with canopy mortality at the stand scale. Sentinel-2, with its 20 m SWIR bands and five-day revisit, produces a dense enough time-series to fit a recovery trajectory curve against published spectral recovery benchmarks for each forest type.
The honest limit is that spectral greenness recovers faster than biomass. Pioneer species and understorey shrubs can push NBR back toward pre-fire values within two to four years in moist tropical forests, while above-ground carbon density may still be 40 to 70 percent below baseline. A project monitor who reads NBR recovery as carbon recovery will overstate permanence. The gap between spectral and structural recovery is widest in forests with high pre-fire canopy height, precisely the high-carbon stands that attract the largest credit volumes.
Where GEDI closes the gap, and where it does not
GEDI waveform lidar measures canopy height and vertical structure directly, not as a proxy. The L4B gridded biomass product translates those measurements into above-ground biomass density estimates with published uncertainty bounds. In a post-fire context, comparing L4B values at the fire-affected area before and after the disturbance, controlling for the L4B prediction interval, gives a structurally grounded biomass trajectory that NBR alone cannot provide.
The constraints are real. GEDI's 1 km gridded product smooths over within-pixel heterogeneity that matters enormously in a patchy burn. The ISS orbital precession means that individual footprint locations shift over time, complicating direct pixel-to-pixel comparison. Coverage is limited to roughly 51.6° N and S, which excludes boreal project areas. And GEDI acquisition paused between 2023 and reactivation periods, creating temporal gaps. For small project areas, fewer than a few thousand hectares, the number of GEDI footprints falling inside the burn perimeter may be insufficient for statistically meaningful biomass change estimates.
Building a defensible recovery trajectory
A credible post-fire permanence assessment rests on three things anchored in time: the burn event itself, the pre-fire baseline, and the post-fire trajectory. VIIRS 375 m active fire detections establish the ignition date and approximate perimeter. Landsat Collection 2 dNBR from the archive establishes pre-fire spectral condition and, for older projects, a multi-year baseline against which severity can be calibrated. Sentinel-2 NBR composites, cloud-masked and gap-filled using methods such as the Whittaker smoother or harmonic regression, then track the recovery curve at 10 to 20 m resolution through subsequent growing seasons.
Published spectral recovery curves for tropical, temperate, and boreal forest types, derived from studies using Landsat time-series, provide the reference envelope. If the project's NBR trajectory falls within the expected recovery range for its forest type and climate zone, and GEDI biomass estimates are consistent with structural recovery, the permanence claim is defensible. If the trajectory plateaus below the reference envelope for two or more consecutive growing seasons, that is evidence of state transition requiring registry notification and potential buffer pool drawdown.
Satellize structures this workflow as a repeating annual delivery: burn perimeter confirmed against VIIRS, dNBR severity map from Sentinel-2 and Landsat, annual NBR recovery index against forest-type reference curves, and GEDI biomass delta where footprint density permits. The Tonga crop-estimation programme demonstrated the same principle of time-series consistency checking applied to agricultural recovery; the forest carbon context adds the permanence liability dimension.
The ambiguity you must report, not hide
No remote-sensing workflow currently resolves the spectral-versus-structural recovery ambiguity with certainty at the project scale. That is not a failure of method; it is a physical constraint. Optical sensors measure reflected light from the top of the canopy. A closed canopy of fast-growing pioneer species reflects light in ways that are nearly indistinguishable from a recovering late-successional stand at the same leaf-area index. Only lidar, field plots, or airborne hyperspectral data can reliably separate them, and none of those are available at the revisit frequency of Sentinel-2.
The honest position for a carbon auditor is that satellite-derived NBR recovery trajectories are necessary but not sufficient for permanence certification. They are strong evidence of state transition when recovery is clearly absent. They are weaker, though still useful, evidence when recovery appears to be on track, because the structural question remains open until canopy height and biomass density are confirmed by lidar or field measurement. Reporting should distinguish between 'spectral recovery consistent with permanence' and 'biomass recovery confirmed', and registries should require that distinction in MRV submissions.
Typical figures
| Optical spatial resolution (NBR) | 10 m (NIR, Sentinel-2 B8) / 20 m (SWIR, Sentinel-2 B11, B12); 30 m Landsat OLI |
| Revisit frequency | 5 days (Sentinel-2A+B combined at equator); 16 days (single Landsat satellite) |
| GEDI biomass product resolution | 1 km gridded (L4B); individual waveform footprints approximately 25 m diameter |
| GEDI latitude coverage | Approximately 51.6° N to 51.6° S; boreal projects outside this range are not covered |
| Active fire detection resolution | 375 m (VIIRS, Suomi-NPP and NOAA-20); twice-daily overpass |
| Landsat dNBR archive depth | TM-era surface reflectance from 1984; Collection 2 with uncertainty layers from 2013 (Landsat 8) |
| Minimum detectable burn severity | dNBR threshold for low-severity burns approximately 100 to 180 (composite burn index calibrated); high cloud frequency in tropics can delay detection by weeks |
| Latency (operational delivery) | Sentinel-2 imagery typically available within 3 hours of acquisition via Copernicus Data Space; analysis products within 24 to 48 hours depending on cloud masking requirements |
| Delivery formats | GeoTIFF (dNBR, NBR index, biomass delta), GeoJSON (burn perimeter), PDF annual permanence trajectory report, GIS-ready time-series stack |
Analytics Satellize can run
| Burn perimeter and severity map | VIIRS active fire detection for event anchoring; Sentinel-2 and Landsat dNBR calculation; composite burn index classification | GeoTIFF severity map and GeoJSON perimeter with area statistics, delivered within 48 hours of confirmed clear acquisition post-fire |
| Annual NBR recovery index | Harmonic regression or Whittaker-smoothed Sentinel-2 NBR time-series; comparison against published spectral recovery envelopes for the relevant forest biome | Annual GIS layer showing per-pixel recovery status (on-track, lagging, plateau) relative to forest-type reference curve |
| Biomass delta estimate | GEDI L4B pre- and post-fire gridded biomass comparison with published prediction interval propagation; footprint density assessment for statistical adequacy | Tabular biomass change report with confidence intervals and explicit flag where footprint density is insufficient for project-scale inference |
| State-transition risk flag | Two-consecutive-season plateau detection in NBR trajectory below forest-type recovery envelope; cross-checked against GEDI biomass trend where available | Registry-ready alert report distinguishing spectral-recovery-consistent-with-permanence from confirmed-biomass-recovery, with methodology disclosure |
| Multi-decadal fire history and baseline dNBR | Landsat Collection 2 archive dNBR time-series from 1984 to present; pre-fire baseline condition characterisation for permanence baseline setting | GeoTIFF archive stack and summary PDF documenting historical fire frequency and severity within the project boundary |
| Annual permanence trajectory summary for MRV submission | Integration of NBR recovery index, GEDI biomass delta, and fire event record into a structured permanence evidence package following registry MRV reporting conventions | Structured PDF and machine-readable JSON evidence package suitable for submission to carbon registry auditors |
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.