Coastal wetland inundation and methane emission proxy mapping with SAR
SAR backscatter and coherence reveal flooded ground beneath tidal marsh and mangrove canopies, providing inundation extent as a spatial proxy for anaerobic methane production. The method is powerful but indirect: emission estimates carry large uncertainty and depend on ancillary temperature and vegetation data.
Sensors
- Sentinel-1 C-band SAR (ESA): 10 m ground range resolution in IW mode, 6-day repeat at mid-latitudes with two satellites, free and open archive from 2014. C-band (5.4 GHz) penetrates light herbaceous marsh canopies but is attenuated by dense mangrove; strong double-bounce signal over flooded low-stature vegetation.
- ALOS-2 PALSAR-2 L-band SAR (JAXA): 3–10 m resolution depending on mode, 14-day repeat. L-band (1.27 GHz) penetrates dense mangrove and forested wetland canopies far more effectively than C-band, producing a stronger double-bounce return from flooded trunks and roots. Archive extends from 2014; commercial access required.
- NISAR (NASA/ISRO, planned launch 2025): Dual-frequency L- and S-band SAR on a single platform, 12-day global repeat. L-band will extend PALSAR-2 style canopy penetration; S-band adds a complementary intermediate frequency. Specifically designed with wetland carbon monitoring as a science objective.
- Capella Space X-band SAR (commercial): Sub-metre resolution in spotlight mode, revisit on demand via tasking. X-band does not penetrate canopies and is primarily useful for open-water inundation extent and temporal gap-filling between L-band passes, or for validating flood edges in unvegetated mudflats.
What a flooded forest trunk tells a radar
When standing water sits beneath a forest or tall marsh, microwave energy from a SAR illuminates the scene twice: once off the water surface and once off the vertical trunk or stem, then back to the sensor along the same path. This double-bounce mechanism produces a backscatter return that is typically 5–15 dB stronger than the same scene without inundation, depending on frequency and incidence angle. It is the physical signature that makes SAR uniquely suited to this problem. No optical sensor can see through a closed mangrove canopy to the water table below.
Frequency matters considerably. C-band (Sentinel-1, ~5.4 GHz) interacts with leaves and small branches, so dense mangrove canopies partially absorb and scatter the signal before it reaches the trunk-water interface. L-band (PALSAR-2, ~1.27 GHz) has a longer wavelength that passes through the same canopy with far less attenuation, producing a cleaner double-bounce. For open herbaceous marsh, C-band works well. For closed-canopy mangrove, L-band is the more reliable tool, and the forthcoming NISAR mission was partly scoped with this distinction in mind.
Coherence loss as a flood detector
Backscatter intensity alone can be ambiguous: a dry rough surface and a flooded surface can produce similar returns in some geometries. Interferometric coherence adds a second discriminant. When two SAR passes over the same area are compared, stable dry surfaces maintain high phase coherence; standing water breaks coherence almost completely because the water surface moves between passes. A coherence time-series from Sentinel-1's 6-day repeat cycle can therefore map inundation dynamics at sub-weekly temporal resolution, distinguishing tidal flooding pulses from longer storm-surge events.
The method has limits. Temporal decorrelation from vegetation growth, wind, or rainfall can mimic inundation-driven coherence loss, particularly in rapidly growing herbaceous marsh. Analysts typically combine coherence change with intensity change and apply ancillary tidal-stage data to reduce false positives. Spatial resolution in Sentinel-1 IW mode is 10 m in range and roughly 14 m in azimuth before multi-looking, which sets the minimum mappable inundation patch. Narrow tidal channels below that width are missed entirely.
From inundation map to methane flux estimate
Coastal wetlands, particularly mangroves and tidal marshes, are among the most methane-productive ecosystems per unit area when anaerobic conditions prevail. The link to SAR is indirect but defensible: flooded area extent is the primary spatial control on where anaerobic methanogenesis can occur. Multiplying inundation extent by published emission factors, which for tidal marshes range from roughly 0.5 to over 100 mg CH₄ m⁻² day⁻¹ depending on salinity, temperature and vegetation type, produces a regional flux estimate.
The uncertainty in that estimate is large, and it is important to say so plainly. Emission factors from the literature span two orders of magnitude across site types. Soil temperature modulates methanogenesis exponentially; without in-situ or satellite thermal data at compatible resolution, the temperature correction is a major source of error. Salinity suppresses methane production in favour of sulphate reduction, so a brackish marsh and a freshwater marsh with identical inundation extents can have very different fluxes. SAR-derived inundation maps are a necessary input to regional wetland methane budgets, but they are not sufficient on their own.
Building a time-series: practical workflow
A typical operational workflow ingests Sentinel-1 IW Ground Range Detected scenes, applies terrain correction using a digital elevation model (SRTM or Copernicus DEM at 30 m), and converts to gamma-nought backscatter. A change-detection classifier, often a threshold applied to the difference between a dry-season baseline and each subsequent acquisition, flags candidate flooded pixels. Coherence images from co-registered Single Look Complex pairs add the second discriminant. The output is a binary or probabilistic inundation mask per acquisition date.
For mangrove-dominated sites, the same workflow is repeated with ALOS-2 PALSAR-2 data where available, and the two frequency results are compared. Disagreement between C-band and L-band inundation maps is itself informative: pixels flooded in L-band but not C-band are likely beneath a dense canopy that C-band cannot penetrate. That discrepancy layer can be used to flag areas where C-band alone would underestimate flooded extent. NISAR's dual-frequency design will eventually make this cross-sensor reconciliation routine rather than a manual step.
Honest limits of the approach
SAR cannot measure methane concentrations. The emission estimate is a model output, not a measurement, and should be presented with explicit uncertainty bounds rather than a single number. Inundation maps themselves carry errors: wind-roughened open water can suppress the specular return and be misclassified as dry land; dense floating vegetation (water hyacinth, for example) can mask open water from above and produce low backscatter that resembles dry ground.
Revisit frequency is a real operational constraint. Sentinel-1's 6-day repeat is adequate for tracking spring-neap tidal cycles but misses individual storm-surge events that last only 12 to 24 hours. The 14-day PALSAR-2 repeat is worse for tidal dynamics. Capella Space and other commercial X-band operators can task revisits within hours, but X-band does not penetrate canopies, so the gap-fill is only useful at flood edges. Cloud cover is irrelevant for SAR, which is one of its genuine advantages over optical sensors in persistently overcast coastal tropics.
Satellize runs Sentinel-1 and PALSAR-2 inundation time-series as part of its satellite-data analytics offering; its Tonga crop-estimation programme illustrates the same principle of combining open-constellation data with ancillary environmental layers to produce actionable regional estimates.
What a government or carbon-market buyer should ask for
Any credible deliverable for this use case should specify the SAR frequency used, the baseline period for dry-condition reference, the DEM source and its vertical accuracy, the emission-factor database applied (IPCC Wetlands Supplement tables are the most widely cited), and the assumed temperature and salinity corrections. A map without those metadata is not auditable.
For carbon-market applications, inundation-derived flux estimates are typically used as one input to a broader Tier 2 or Tier 3 national inventory calculation, not as a standalone MRV product. The IPCC 2013 Wetlands Supplement provides the methodological framework most national inventory compilers follow. Buyers should expect the SAR-derived inundation layer to reduce spatial uncertainty in the inventory while the flux model and its assumptions remain the dominant source of overall error.
Typical figures
| Spatial resolution (Sentinel-1 IW) | 10 m range × ~14 m azimuth (single look); typically 20–40 m after multi-looking for flood mapping |
| Spatial resolution (ALOS-2 PALSAR-2 SM mode) | 3–10 m depending on polarisation and look angle |
| Revisit (Sentinel-1, two satellites) | 6 days at mid-latitudes; 12 days with one satellite |
| Revisit (ALOS-2 PALSAR-2) | 14 days global |
| Radar frequency / band | C-band 5.4 GHz (Sentinel-1); L-band 1.27 GHz (PALSAR-2, NISAR); X-band ~9.6 GHz (Capella) |
| Canopy penetration | L-band: good through closed mangrove; C-band: moderate for herbaceous marsh, poor for dense mangrove; X-band: negligible |
| Minimum mappable inundation patch | ~0.04 ha at 20 m resolution; sub-pixel tidal channels not resolved |
| Archive depth | Sentinel-1: 2014–present (free); ALOS-2: 2014–present (commercial licence) |
| Emission factor uncertainty | 0.5–100+ mg CH₄ m⁻² day⁻¹ depending on site type; order-of-magnitude spread across literature |
| Cloud sensitivity | None: SAR is cloud-transparent at all three frequencies |
Analytics Satellize can run
| Inundation extent map (per acquisition) | Gamma-nought backscatter thresholding against dry-season baseline, terrain-corrected with Copernicus DEM 30 m | GeoTIFF raster, binary or probabilistic flood mask, per Sentinel-1 pass |
| Inundation dynamics time-series | Multi-temporal coherence change detection on co-registered SLC pairs; tidal-stage ancillary filtering | Stacked GeoTIFF or NetCDF time-series; inundation frequency raster over user-defined period |
| Canopy-penetration discrepancy layer | Pixel-wise comparison of C-band (Sentinel-1) and L-band (PALSAR-2) flood masks; disagreement flags dense-canopy zones | GIS polygon layer identifying areas where C-band alone underestimates flooded extent |
| Regional methane flux proxy estimate | Inundation extent × IPCC Wetlands Supplement Tier 2 emission factors, temperature-corrected with ERA5 reanalysis skin temperature | Tabular flux estimate with uncertainty bounds by vegetation zone; methodology report for inventory use |
| Tidal flood pulse characterisation | Spring-neap cycle extraction from 6-day Sentinel-1 time-series; comparison with tide-gauge or hydrodynamic model output | Flood frequency histogram and inundation duration map per pixel; PDF report |
| Mangrove structure and flood interaction assessment | L-band HV/HH polarisation ratio as canopy density proxy combined with inundation mask; cross-referenced with Global Mangrove Watch extent data | Stratified inundation map by mangrove canopy density class; GIS layer and summary statistics |
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.