Tidal wetland inundation frequency mapping for blue-carbon stock validation
Inundation frequency controls which plants grow where and how fast carbon accumulates below ground. Sentinel-1 SAR time-series, tied to tide-gauge records, maps that frequency at sub-field scale and tests whether blue-carbon project boundaries match reality.
Sensors
- Sentinel-1 SAR IW mode (C-band, ESA): 5 x 20 m ground range resolution in Interferometric Wide swath mode; 6-day repeat at the equator with two satellites combined. C-band backscatter distinguishes open water from flooded vegetation, but saturates under closed mangrove canopy above roughly 100 t/ha above-ground biomass, a hard limit for dense stands.
- Sentinel-2 MSI (ESA): 10 m visible and near-infrared bands; NDWI (bands 3 and 8) detects surface water and vegetation moisture. Cloud cover is the dominant constraint in tropical wetland settings; 5-day revisit with both satellites but effective clear-sky acquisitions in monsoonal regions can drop to a handful per season.
- ICESat-2 ATL03/ATL08 (NASA): Photon-counting lidar delivers coastal surface elevations to roughly 10 cm vertical accuracy along narrow ground tracks spaced approximately 3 km apart at mid-latitudes. Used to anchor a digital elevation model of the intertidal zone, without which inundation frequency cannot be modelled from tide-gauge records alone.
- NOAA/UHSLC tide-gauge records: Hourly water-level data from the nearest verified gauge, used to assign a tidal stage to each SAR acquisition. Accuracy depends on gauge proximity; sites more than 30 km from the project area or with complex tidal prisms may require harmonic correction or a local pressure sensor.
Why inundation frequency is the argument, not just the context
Blue-carbon accounting assigns different carbon accumulation rates to different vegetation zones: low-marsh, high-marsh, mangrove fringe, transition scrub. Those rates are not interchangeable. Studies published in the peer-reviewed literature show that below-ground carbon stocks in tidal wetlands correlate strongly with hydroperiod, the fraction of time a given point sits under water. A project that mis-classifies a high-inundation zone as a low-inundation zone overstates its carbon density. That is not a rounding error; it can shift a stock estimate by tens of percent.
Project documents typically draw zone boundaries from a single field survey or a static vegetation map. Neither captures inter-annual tidal variability, sea-level trend, or the effect of storm-driven berm breaches. An independent inundation frequency map derived from a multi-year radar archive gives auditors something the project proponent cannot easily contest: a pixel-by-pixel record of how often each part of the site was wet.
What the radar backscatter time-series actually measures
Sentinel-1 C-band radar responds to surface roughness and dielectric properties. Open water returns very low backscatter (typically below -15 dB in VV polarisation) because the smooth surface reflects energy away from the sensor. Flooded vegetation beneath a canopy produces a double-bounce return that is markedly brighter, often 3 to 8 dB above the dry-soil baseline. Bare mudflat sits between those extremes and shifts with surface moisture. Classifying each pixel in each acquisition as flooded, moist or dry, then counting the fraction of acquisitions in the flooded state, produces an inundation frequency map.
The critical step is tying each acquisition to a tidal stage from the nearest gauge. A pixel that appears flooded in 40 percent of images acquired near mean high water tells a very different hydrological story from one that appears flooded in 40 percent of images acquired across all tidal stages. Without that linkage the frequency map is descriptive but not physically interpretable. With it, you can estimate the elevation of the flooding threshold for each vegetation zone and cross-check that against the ICESat-2 elevation model.
The saturation problem under dense mangrove is real and must be stated plainly. Where above-ground biomass exceeds roughly 100 tonnes per hectare, C-band backscatter no longer responds reliably to inundation beneath the canopy. The double-bounce signal is attenuated by the canopy itself. For those pixels, the SAR record gives you canopy-surface information, not ground inundation. L-band SAR (ALOS-2 PALSAR-2, or the forthcoming NISAR) penetrates better, but Sentinel-1 is the only freely available, high-revisit option at present.
Calibrating the map against tide gauges and elevation
The workflow has three interlocking components. First, ICESat-2 ground tracks are used to build or refine a bare-earth elevation model of the intertidal zone. Published ICESat-2 coastal studies report vertical accuracy of 8 to 12 cm over flat mudflat; accuracy degrades under canopy where photon returns are sparse. Second, the local tidal datum is established from the nearest NOAA or UHSLC gauge, or from a harmonic tidal model such as TPXO where no gauge exists within a useful distance. Third, each Sentinel-1 acquisition is tagged with the predicted water level at acquisition time, and a logistic relationship between water level and flood probability is fitted per pixel.
The output is not a binary flooded-or-dry map. It is a continuous inundation frequency surface, typically expressed as the percentage of tidal cycles during which each pixel is submerged, referenced to a defined tidal datum. That surface can then be overlaid on the project's claimed carbon-stock zone boundaries. Where the frequency contours diverge substantially from the claimed boundaries, the auditor has a documented basis for querying the stock estimate.
Honest limits: cloud, canopy, gauge distance and resolution floor
Sentinel-1 is a radar system and therefore cloud-transparent, which is the main reason it is used here rather than optical sensors. But the 5 x 20 m IW mode resolution means that narrow tidal creeks, typically 2 to 4 m wide in mature marsh systems, are sub-pixel features. The map reflects the average inundation state of a 100 square-metre cell, not the creek itself. Vegetation growing on creek banks may be systematically wetter than the pixel average suggests.
Sentinel-2 NDWI is useful for cross-checking open-water extent during cloud-free windows, but it adds little information in the dense-canopy zones where the SAR is already struggling. Its 10 m resolution does not resolve the creek-width problem either.
Gauge distance matters more than is often acknowledged. Tidal prisms in complex estuaries can produce water-level differences of 20 to 40 cm between a gauge and a site 15 km away. That error propagates directly into the inundation frequency estimate. Where no gauge is within a defensible distance, the uncertainty should be quantified and reported, not silently absorbed.
Archive depth is a genuine advantage. The Sentinel-1 archive runs from 2014, giving a decade of acquisitions for trend analysis. That is long enough to detect multi-year shifts in inundation frequency associated with sea-level rise or land-use change upstream, both of which affect carbon stock trajectories.
From frequency map to carbon-zone audit
The practical output for a carbon auditor is a comparison layer: the project's claimed zone boundaries on one side, the satellite-derived inundation frequency contours on the other. Disagreements fall into two categories. Boundary errors, where the project has drawn a line in the wrong place, are the simpler case; they call for a revised area calculation. Systematic errors, where an entire zone class has been assigned the wrong hydroperiod, are more consequential because they imply the wrong carbon accumulation coefficient has been applied throughout.
Satellize produces this comparison layer as a GIS deliverable, with the inundation frequency surface, the uncertainty bounds from gauge-distance and canopy-saturation effects, and a tabular summary of zone-area discrepancies. The methodology is the same class of approach used in published peer-reviewed inundation mapping studies; the deliverable is formatted for submission to a carbon registry audit process rather than a journal.
The Tonga crop-estimation programme demonstrated the same underlying principle in a different context: that a time-series of open-constellation imagery, calibrated against ground truth, can produce a quantitative estimate that is defensible to an external reviewer. The same discipline applies here.
What a project developer should bring to this analysis
The satellite archive exists and is free. The analytical bottleneck is almost always the ground data: a georeferenced project boundary in a standard format, the identity and distance of the nearest tide gauge, any existing vegetation survey or soil-core data that can anchor the carbon-stock coefficients, and clarity on which carbon accounting standard the project is registered under (since different standards specify different zone-classification rules).
If the project site is in a region with poor gauge coverage, a short-duration pressure-sensor deployment during the analysis period can reduce the tidal-calibration uncertainty substantially. That is a field decision, not a satellite decision, but it is the single intervention most likely to improve the defensibility of the final map.
Typical figures
| SAR spatial resolution (IW mode) | 5 m range x 20 m azimuth (Sentinel-1); ground range detected product typically resampled to 10 m |
| SAR revisit (two-satellite constellation) | 6 days at equator; shorter at higher latitudes |
| Optical cross-check resolution (Sentinel-2 NDWI) | 10 m (bands 3 and 8) |
| Elevation accuracy (ICESat-2 over flat mudflat) | 8 to 12 cm vertical; degrades under canopy |
| C-band backscatter saturation threshold | Approximately 100 t/ha above-ground biomass; dense mangrove canopy unreliable above this |
| Tidal-stage calibration source | NOAA/UHSLC hourly gauge records or TPXO harmonic model where gauge distance exceeds ~30 km |
| Minimum resolvable creek width | Greater than ~10 m; narrower creeks are sub-pixel in Sentinel-1 IW |
| Archive depth (Sentinel-1) | 2014 to present; approximately 10 years of repeat acquisitions available |
| Inundation frequency output format | Continuous 0–100 % surface, GeoTIFF or GeoPackage, referenced to local tidal datum |
| Latency for new acquisitions | Sentinel-1 data typically available via Copernicus Dataspace within 1 to 3 hours of acquisition |
Analytics Satellize can run
| Inundation frequency surface | Per-pixel logistic flood classification across Sentinel-1 VV/VH time-series, tagged by tidal stage from nearest gauge; frequency computed as fraction of acquisitions in flooded state | GeoTIFF raster, 10 m resolution, with per-pixel uncertainty band; delivered with processing log |
| Carbon-zone boundary comparison layer | Overlay of project-claimed zone polygons against inundation frequency contours; discrepancy area calculated per zone class | GeoPackage with attributed polygons and tabular area-discrepancy summary for registry submission |
| Tidal elevation threshold map | Logistic regression of inundation frequency against ICESat-2 bare-earth elevations to derive flooding-threshold elevation per vegetation zone | Raster and point-data layer showing elevation of zone boundaries; cross-referenced to local tidal datum |
| Canopy-saturation mask | Sentinel-2 NDVI thresholding combined with backscatter texture analysis to flag pixels where C-band saturation makes inundation classification unreliable | Binary mask layer included with frequency surface; narrative section in audit report quantifying affected area |
| Multi-year inundation trend | Annual inundation frequency computed for each year of the Sentinel-1 archive; Mann-Kendall trend test applied per pixel to detect statistically significant change | Trend-significance raster and time-series plots for user-defined zones; relevant to permanence and baseline assessments |
| Gauge-calibration uncertainty report | Sensitivity analysis varying assumed water-level offset across plausible gauge-distance error range; propagated into frequency estimate confidence intervals | PDF technical annex quantifying uncertainty contribution from tidal calibration, formatted for carbon registry audit appendix |
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.