Mangrove recruitment and propagule-dispersal zone mapping from tidal hydrodynamics
Natural mangrove recovery depends on where tidal currents carry propagules and whether pioneer patches can establish. Sentinel-1 SAR and Sentinel-2 spectral indices, combined with tidal hydrodynamic modelling, map the zones most likely to recruit, and track whether they actually do.
Sensors
- Sentinel-1 SAR (C-band, ESA): 10 m ground-range resolution in IW mode, 6-day revisit at the equator with both satellites active. C-band backscatter penetrates cloud and darkness, enabling consistent inundation mapping across tidal cycles. Flooded soil beneath sparse canopy returns a double-bounce signal that distinguishes wet bare mudflat from dry sediment, though dense closed canopy attenuates the signal and limits sub-canopy inundation detection.
- Sentinel-2 MSI (ESA): 10 m resolution in visible and near-infrared bands (B2, B3, B4, B8), 20 m in red-edge and SWIR bands (B5, B6, B7, B8A, B11, B12), 5-day revisit with both satellites. Red-edge bands (705 nm, 740 nm) are particularly sensitive to chlorophyll content in sparse, low-biomass pioneer stands that would be invisible to a simple NDVI threshold on Landsat. Cloud contamination in humid tropical settings routinely reduces usable acquisitions to fewer than one per month in the wet season.
- ALOS-2 PALSAR-2 (JAXA, L-band): L-band (1.27 GHz) penetrates mangrove canopy far more effectively than C-band, enabling detection of standing water beneath closed-canopy stands. Spatial resolution in standard mode is 10 m (fine beam) or 6 m (ultra-fine). Revisit is 14 days, and the sensor is not free-access, so archive depth and acquisition frequency depend on tasking agreements. Particularly useful for distinguishing waterlogged soils under recovering canopy from dry ground.
- Landsat 8/9 OLI (USGS/NASA): 30 m resolution, 16-day revisit per satellite, 8-day combined. The archive extends to 1972 (Landsat 1 MSS), making it the only freely available source for multi-decade baselines of mangrove extent and disturbance history. At 30 m, pioneer patches smaller than roughly 0.09 ha are sub-pixel; the sensor is best used for establishing pre-disturbance canopy extent and long-term trajectory analysis rather than detecting early recruitment.
Why tidal plumbing matters more than seed supply
Mangrove propagules are not passively scattered. Species such as Rhizophora produce buoyant propagules that remain viable for months, but viability alone does not determine where they strand. Tidal current velocity, creek geometry, and inundation duration control stranding location. A mudflat that floods for two to four hours per tidal cycle sits in the recruitment sweet spot for most Indo-Pacific Rhizophora species; flood longer and anaerobic stress kills seedlings before they root, flood less and desiccation does the same job. This means that mapping propagule-dispersal zones is, in large part, a problem of mapping tidal hydrodynamics at fine spatial scale.
Satellite data cannot directly observe water velocity, but it can observe the outcome: inundation frequency. A dense Sentinel-1 time series, typically 30 or more acquisitions spanning several tidal cycles, allows each pixel to be assigned an inundation frequency value. Pixels that flood on roughly 20 to 50 percent of acquisitions correspond well to the intermediate inundation regime that field studies associate with successful Rhizophora establishment. That correspondence is not a law of physics; it depends on tidal range, local topography, and species assemblage. The satellite result is a probability surface, not a guarantee.
Reading the SAR time series: what the backscatter is actually telling you
On bare or sparsely vegetated mudflat, C-band backscatter drops sharply when the surface floods. The water surface is specularly smooth at low incidence angles, scattering energy away from the sensor. This creates a reliable binary: flooded pixels go dark, dry pixels stay bright. Stacking 30 to 50 IW-mode Sentinel-1 scenes and computing per-pixel inundation frequency produces a tidal zonation map that resolves creek networks down to roughly 10 m width. Creeks narrower than one pixel (10 m) are invisible; their hydrological influence on adjacent mudflat is real but cannot be directly observed.
The double-bounce effect complicates interpretation under even sparse mangrove cover. Vertical pneumatophores and prop roots act as corner reflectors when standing in water, producing anomalously high backscatter that can be misclassified as dry ground. Separating this from genuine dry-soil returns requires either multi-temporal coherence analysis or cross-referencing with Sentinel-2 to confirm vegetation presence. Neither approach eliminates the ambiguity entirely; it reduces it.
Finding pioneer patches at 10 metres: what the spectral indices can and cannot see
Young mangrove recruits, typically one to three years post-establishment, have lower canopy closure, lower leaf-area index, and different canopy architecture than mature stands. Sentinel-2's red-edge bands (B5 at 705 nm and B6 at 740 nm) are sensitive to these differences. The red-edge chlorophyll index (CIre = B7/B5 minus 1) separates low-biomass pioneer stands from both bare sediment and dense mature canopy more reliably than NDVI alone, which saturates at moderate canopy densities. Published studies using Sentinel-2 red-edge indices in Southeast Asian mangroves have reported detection of pioneer patches down to roughly 0.01 to 0.05 ha, though this lower bound is highly site-dependent.
Cloud is the practical constraint. In the humid tropics, a single cloud-free Sentinel-2 scene may require compositing over two to three months of acquisitions. A pioneer patch that establishes and then dies within that window is invisible. The operational approach is to build seasonal composites, typically dry-season, and compare successive years. Canopy-closure trajectories over three to five years are a more reliable indicator of genuine recruitment than any single-date spectral index. Individual seedlings, which are centimetres tall and sub-pixel by orders of magnitude, cannot be detected from orbit at any currently operational resolution.
Combining the layers: from inundation map to dispersal-zone probability surface
The analytic workflow has three stages. First, a Sentinel-1 inundation-frequency raster identifies the tidal zonation across the study area. Second, the location of seed-producing adult stands is extracted from a Sentinel-2 or Landsat canopy map; these are the propagule sources. Third, a simple dispersal model, typically a distance-weighted or hydrodynamically-informed kernel applied along tidal creek networks, identifies which inundation-suitable zones are within plausible dispersal range of source stands. The result is a recruitment-probability surface: areas that are hydrologically suitable and within propagule supply range score highest.
This is not a substitute for field validation. The surface identifies where recovery is physically possible, not where it is occurring. Comparing the recruitment-probability surface against the Sentinel-2 pioneer-patch detection layer from successive years tests whether the model is capturing real dynamics. Where high-probability zones show no spectral evidence of recruitment after several years, the explanation may be substrate unsuitability, grazing pressure, water-quality stress, or simply that propagule supply was interrupted by an upstream disturbance that the model did not capture.
Honest limits and the role of field calibration
The 10 m pixel of Sentinel-1 and Sentinel-2 sets a hard floor on what can be detected. Patches smaller than roughly 0.01 ha are sub-pixel. In fragmented post-disturbance landscapes, where early recruitment is exactly the kind of fine-grained, spatially scattered process that matters most for conservation planning, this is a genuine constraint. L-band ALOS-2 PALSAR-2 can improve inundation detection under canopy but does not improve the spatial resolution of pioneer-patch detection, which is a spectral rather than a SAR problem.
Tidal range varies enormously between sites. A method calibrated on a macro-tidal estuary in northern Australia will not transfer without recalibration to a micro-tidal lagoon in the Caribbean. The inundation-frequency thresholds that correspond to the recruitment zone are site-specific and ideally derived from co-located tide-gauge records or validated hydrodynamic models rather than assumed from the literature. Satellize's analytics for this use case incorporate site-specific tidal datum correction where gauge data are available, following the same approach used in the Tonga crop-estimation programme for localising remote-sensing outputs to ground-truth conditions.
What a monitoring programme actually looks like in practice
A credible satellite-based recruitment monitoring programme requires a minimum archive depth of three to five years to distinguish genuine canopy-closure trajectories from inter-annual spectral noise. The Sentinel archive, which begins in 2014 for Sentinel-2 and 2014 for Sentinel-1, is sufficient for most post-2016 disturbance events. For earlier disturbances, Landsat provides the baseline extent map, and Sentinel provides the recruitment tracking from the point at which it becomes available.
Deliverables for conservation managers typically include an annual pioneer-patch map (GIS polygon layer with area and spectral confidence attributes), a recruitment-probability surface updated when new tidal-creek network data are available, and a canopy-closure trajectory chart per defined monitoring zone. Alert products, flagging zones where the recruitment-probability surface predicts recovery but successive composites show no spectral change, are operationally useful for prioritising field inspection effort. The satellite cannot tell a ranger why recruitment has failed; it can tell them precisely where to look.
Typical figures
| Spatial resolution (inundation mapping) | 10 m (Sentinel-1 IW mode) |
| Spatial resolution (pioneer-patch spectral detection) | 10 m visible/NIR, 20 m red-edge/SWIR (Sentinel-2) |
| Minimum detectable pioneer patch | Approximately 0.01 to 0.05 ha (site-dependent; sub-pixel below this floor) |
| Revisit (Sentinel-1, both satellites) | 6 days at equator; reduced at higher latitudes |
| Revisit (Sentinel-2, both satellites) | 5 days at equator; cloud-effective revisit in humid tropics often 30 to 90 days |
| SAR frequency (Sentinel-1) | C-band, 5.405 GHz |
| SAR frequency (ALOS-2 PALSAR-2) | L-band, 1.2575 GHz; 10 m (fine beam) or 6 m (ultra-fine) resolution |
| Archive depth | Sentinel-1/2 from 2014; Landsat from 1972 (30 m, for baseline extent only) |
| Latency (standard processing) | Sentinel-1/2 Level-1 products available within 3 hours of acquisition via Copernicus Data Space |
| Delivery formats | GeoTIFF raster, GeoPackage or Shapefile polygon layers, PDF monitoring report |
Analytics Satellize can run
| Tidal inundation-frequency raster | Multi-temporal Sentinel-1 SAR backscatter thresholding and per-pixel inundation-frequency calculation across a minimum 30-scene stack, with site-specific tidal datum correction | GeoTIFF raster, continuous 0 to 1 inundation-frequency values, annual update |
| Tidal creek network extraction | SAR-derived water-surface mask skeletonisation and morphological filtering to identify channel centrelines at 10 m resolution | GeoPackage polyline layer with estimated channel width attributes |
| Pioneer recruitment patch map | Sentinel-2 red-edge chlorophyll index (CIre) and NDVI seasonal compositing with change detection against prior-year baseline; thresholding to isolate low-biomass mangrove spectral class | Annual GeoPackage polygon layer with patch area, spectral confidence score, and year-of-first-detection attribute |
| Propagule-dispersal probability surface | Distance-weighted dispersal kernel applied along tidal creek network from mapped adult-stand source polygons, intersected with inundation-frequency suitability mask | GeoTIFF raster, 10 m resolution, continuous probability values; updated when source-stand or creek-network layers are revised |
| Canopy-closure trajectory analysis | Annual Sentinel-2 composite time series stacking; per-pixel linear regression of red-edge index over three to five years to classify positive, neutral, or negative canopy-development trajectories | Trajectory classification raster and summary chart per monitoring zone, delivered as PDF report and GeoTIFF |
| Failed-recruitment alert layer | Logical intersection of high-probability dispersal zones with zones showing no statistically significant spectral improvement over two consecutive annual composites | GeoPackage polygon alert layer with zone ID and area, for field inspection prioritisation; issued annually after dry-season composite processing |
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.