Wetland inundation dynamics as a methane flux proxy
SAR backscatter and coherence change can map flooded vegetation beneath forest canopy at sub-monthly revisit, giving carbon registries the inundation time-series they need to bound natural wetland methane flux and separate it from project-attributed reductions.
Sensors
- Sentinel-1 C-band SAR (ESA): 5.6 cm wavelength; Interferometric Wide Swath mode gives 10 m ground range resolution across a 250 km swath. Repeat cycle 6 days at the equator with both satellites active. C-band penetrates sparse canopy but is attenuated by dense closed-canopy forest, so flooded-forest detection works best in open or seasonally flooded woodland.
- ALOS-2 PALSAR-2 L-band SAR (JAXA): 23.6 cm wavelength penetrates dense tropical forest canopy and double-bounce returns from water-trunk interfaces are a reliable indicator of inundation beneath closed canopy. Spatial resolution 10 m in Fine Beam Single mode; repeat cycle approximately 14 days. Archive extends to 2014 and predecessor ALOS PALSAR to 2006, giving a useful historical baseline.
- Landsat 8/9 OLI (USGS/NASA): 30 m multispectral imagery used as an optical cross-check during cloud-free windows. NDWI and MNDWI indices reliably delineate open water; the 16-day repeat is too coarse to capture rapid inundation pulses alone, but the long archive back to 1972 (Landsat 1 through 9) anchors historical inundation frequency baselines.
- SMAP L-band radiometer (NASA): Provides daily global soil-moisture estimates at 36 km passive resolution (9 km enhanced). Not useful for local inundation mapping, but soil saturation context helps interpret SAR returns in transitional zones where shallow subsurface water precedes surface flooding. Useful for regional-scale flux modelling inputs.
Why inundation extent matters more than water area
Methane production in wetlands is almost entirely an anaerobic process. Methanogenic archaea in waterlogged soils generate CH4 that travels to the atmosphere through three pathways: diffusion through the water column, ebullition (bubble release), and plant-mediated transport through aerenchyma tissue. The third pathway dominates in vegetated wetlands and is controlled by the presence of standing water around plant stems, not simply by open water area.
This distinction matters enormously for carbon accounting. A wetland with 40 percent open water and 60 percent flooded forest may emit two to three times more methane per unit area than one with the same open-water fraction but a dry forest floor, because the aerenchyma pathway bypasses the oxidation that occurs in the water column. Mapping open water alone, which optical sensors do well, systematically underestimates the inundation extent that drives flux. SAR is the practical solution because it detects the double-bounce return from water beneath vegetation that optical imagery cannot see.
The physics of double-bounce and coherence loss
When a SAR pulse encounters a flooded forest, it reflects off the water surface, then off a vertical trunk, and returns to the sensor along nearly the same path it arrived. This double-bounce mechanism produces a characteristic bright return in HH polarisation that is absent over dry forest, where volume scattering from the canopy dominates and the signal is more diffuse. L-band exploits this effect through closed-canopy forest; C-band works in open woodland and marsh vegetation but loses penetration in dense multi-layered tropical forest.
Coherence change detection adds a second independent signal. Repeat-pass SAR interferometry measures phase coherence between two acquisitions. Inundation changes the dielectric properties of the surface and moves the scattering phase centre, causing coherence to drop sharply. A pixel that was coherent in a dry-season pass and incoherent in a wet-season pass is a strong candidate for flooding, even where backscatter intensity alone is ambiguous. Combining backscatter and coherence reduces false positives from wind-induced canopy motion, which also decorrelates but does not produce the characteristic HH brightness increase.
From inundation maps to flux bounds: what the published record supports
The relationship between inundation duration and methane flux is non-linear and site-dependent, but the published literature provides useful bounding values. Studies of Amazonian floodplains have reported mean fluxes from flooded forest in the range of 20 to 80 mg CH4 m⁻² day⁻¹ during inundated periods, with ebullition events adding episodic spikes that are difficult to predict remotely. Temperate freshwater marshes typically fall in a lower range, around 5 to 30 mg CH4 m⁻² day⁻¹, with strong temperature dependence.
A satellite-derived inundation time-series can be used to calculate inundation duration (days per year or per season) for each pixel, then multiplied by published flux rates for the relevant vegetation and climate class to produce an emission estimate with explicit uncertainty bounds. This is not a direct flux measurement. It is a spatially consistent proxy that can be updated at SAR revisit frequency, something eddy covariance towers and chamber measurements cannot match at landscape scale. The honest caveat is that flux-per-unit-area values carry large uncertainties across sites; the satellite contribution is the spatial and temporal coverage, not the flux physics.
For carbon project MRV, the relevant question is usually not the absolute flux but the change in flux attributable to a conservation or restoration intervention. A before-and-after inundation time-series, compared against a reference area with no intervention, can detect shifts in inundation regime that would be expected to alter flux. Registries increasingly require this kind of spatially explicit evidence to validate or reject project claims.
Practical limits that buyers should understand before commissioning work
Dense tropical forest canopy attenuates C-band SAR significantly. In closed-canopy Amazonian or Congo Basin forest, Sentinel-1 may miss flooded understorey entirely; PALSAR-2 L-band is the better choice but its 14-day repeat is coarser and commercial access to tasked data adds cost. Neither sensor resolves individual trees or narrow channels below roughly 10 m width.
Rainfall on the canopy and wet leaves produce a surface scattering return that can mimic the signature of shallow inundation, particularly in C-band. Temporal filtering across multiple passes reduces this confusion but introduces latency. SAR also cannot distinguish freshwater from saline inundation from backscatter alone, which matters in coastal wetlands where tidal and riverine flooding mix. Optical cross-checks with Landsat or Sentinel-2 during cloud-free windows help resolve ambiguous cases, but cloud cover over tropical wetlands during wet season can suppress optical observations for weeks at a time. The SAR time-series is the backbone precisely because of this, but it is not infallible.
Structuring the evidence for a carbon registry
A credible MRV package for a wetland carbon project needs four components: a historical inundation baseline (typically ten or more years of SAR and Landsat composites), a project-period inundation time-series at sub-monthly resolution, a reference area time-series to separate climate-driven variability from project effects, and explicit uncertainty quantification on both the inundation mapping and the flux conversion.
The archive depth available from Sentinel-1 (from 2014), PALSAR-2 (from 2014, with PALSAR from 2006), and Landsat (from 1972) is sufficient to build baselines that predate most active carbon projects. Satellize structures these evidence packages as versioned GIS layers with full provenance metadata, which is the format that third-party auditors and registry technical reviewers can actually work with. The Tonga crop-estimation programme used a similar time-series architecture, though for an entirely different application, demonstrating that systematic multi-sensor compositing at national scale is operationally tractable on open constellations.
Buyers should be aware that no current satellite system provides direct atmospheric methane attribution to a specific wetland patch at the spatial scale of a carbon project. That is the domain of a sibling page on methane emission verification. What inundation mapping provides is the hydrological state variable that drives flux models, delivered at a scale and frequency that ground surveys cannot match.
Typical figures
| Spatial resolution (SAR, primary) | 10 m (Sentinel-1 IW mode; PALSAR-2 Fine Beam Single mode) |
| Revisit period | 6 days at equator (Sentinel-1 A+B); 14 days (PALSAR-2); 16 days (Landsat 8/9) |
| SAR frequency / wavelength | C-band 5.405 GHz / 5.6 cm (Sentinel-1); L-band 1.2575 GHz / 23.6 cm (PALSAR-2) |
| Minimum detectable inundated patch | Approximately 0.01 ha at 10 m resolution; narrow channels below 10 m width are sub-pixel |
| Canopy penetration | C-band: open and sparse woodland; L-band: closed-canopy tropical forest (double-bounce detectable) |
| Archive depth | Sentinel-1 from 2014; PALSAR-2 from 2014 (PALSAR from 2006); Landsat from 1972 |
| Soil moisture context | SMAP daily global, 36 km passive / 9 km enhanced resolution |
| Flux proxy uncertainty | Published tropical flooded-forest flux ranges: 20–80 mg CH4 m⁻² day⁻¹; site-specific calibration required |
| Delivery formats | GeoTIFF inundation extent layers, NetCDF time-series stacks, versioned GIS packages with provenance metadata |
Analytics Satellize can run
| Inundation extent map (per SAR acquisition) | Thresholded HH backscatter change detection combined with reference dry-season composite; standard published approach for SAR flood mapping | GeoTIFF binary inundation mask per acquisition date, with confidence layer |
| Flooded-vegetation fraction map | Double-bounce decomposition in HH/HV polarisation space to separate open water, flooded forest and dry land; L-band preferred for closed canopy | Per-pixel fractional cover layer (open water / flooded vegetation / dry) as GeoTIFF |
| Inundation duration raster (annual or seasonal) | Pixel-wise summation of inundated days across the SAR time-series, gap-filled using temporal interpolation where acquisition gaps exist | Annual inundation duration GeoTIFF (days per year) for baseline and project periods |
| Coherence-change inundation detection | Repeat-pass InSAR coherence differencing between dry and wet season pairs; low coherence in wet season flagged as probable inundation beneath canopy | Coherence-change layer as supplementary evidence GeoTIFF, merged with backscatter product |
| Methane flux proxy estimate with uncertainty bounds | Inundation duration multiplied by published flux-rate ranges for vegetation and climate class; uncertainty propagated from both mapping error and flux-rate variability | Tabular flux estimate (low / central / high) per project polygon, with methodology annex suitable for registry submission |
| Reference-area comparison time-series | Parallel inundation time-series extraction for a matched reference wetland; statistical comparison to isolate project-attributable inundation change from climate variability | Time-series chart and summary statistics report for third-party auditor review |
| Multi-sensor optical cross-check composite | NDWI / MNDWI open-water mapping from cloud-free Landsat 8/9 scenes, co-registered and compared against SAR inundation layers to flag systematic discrepancies | Seasonal optical-SAR agreement layer and discrepancy report |
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.