Coastal erosion monitoring for fishing community and infrastructure exposure
Multi-decadal satellite imagery quantifies shoreline retreat rates at fishing ports and aquaculture sites, turning decades of Landsat and Sentinel observations into exposure assessments that inform where to invest and what to abandon.
Sensors
- Landsat 5/7/8/9 (USGS/NASA): 30 m multispectral resolution, 16-day revisit per satellite. The archive runs from 1972 (Landsat 1 MSS at 60–80 m) through to the present, giving the longest continuous shoreline record available from any public satellite programme. NDWI derived from Green and NIR bands isolates the land-water boundary; tidal stage at acquisition is the dominant source of positional error.
- Sentinel-2 MSI (ESA Copernicus): 10 m resolution in visible and NIR bands, 5-day revisit at mid-latitudes with both Sentinel-2A and 2B operating. The 10 m Green and NIR bands support NDWI-based shoreline extraction with sub-pixel accuracy achievable through sub-pixel edge detection, though tidal ambiguity remains. Archive from 2015.
- Sentinel-1 SAR (ESA Copernicus): C-band synthetic aperture radar at 10 m (IW mode GRD), 6-day revisit. Cloud-independent, which matters considerably during monsoon and tropical storm seasons when optical sensors are blind for weeks. The land-water contrast in backscatter is strong enough to extract shoreline position, though intertidal vegetation and wave roughness introduce their own ambiguities.
- Planet SuperDove: 3–4 m resolution, near-daily revisit over most coastal areas. Useful for resolving fine-scale features such as narrow beach berms, revetment toe positions and small aquaculture pond embankments that fall below Landsat and Sentinel detection limits. Archive from approximately 2017; commercial licence required.
Why a fishing port needs a fifty-year shoreline record
Coastal erosion is not an event. It is a rate. A jetty built in 1985 on a coastline retreating at 0.8 metres per year sits 30 metres further into the surf zone today than when it was designed. The engineering load cases used for that structure almost certainly did not account for that shift. Satellite archives make the rate visible and, crucially, auditable.
The Landsat programme provides imagery from 1972 onward, giving analysts up to five decades of shoreline position data at 30 m resolution. That temporal depth is not matched by any ground survey programme for most developing-world fishing communities. Sentinel-2 extends the record at higher resolution from 2015. Together they allow change rates to be computed per coastal segment, with seasonal variability separated from the long-term trend.
What a water index actually measures, and where it goes wrong
The Normalised Difference Water Index (NDWI), computed as (Green minus NIR) divided by (Green plus NIR), produces high positive values over open water and negative values over dry land. The zero-crossing approximates the instantaneous waterline. That approximation carries a systematic error: the waterline is not the shoreline. The true shoreline, defined as the mean high water mark or a comparable datum, lies somewhere landward of the waterline at the moment of image acquisition, depending on tidal stage.
For a macrotidal coast with a 6 m tidal range and a gently sloping intertidal zone, a 1 m error in tidal stage translates to tens of metres of apparent shoreline position error. This is not a failure of the method; it is a physical constraint that must be corrected using co-registered tidal model outputs (such as FES2014 or TPXO) or filtered by restricting analysis to images acquired near mean sea level. Published studies using the CoastSat toolkit have demonstrated sub-pixel shoreline extraction accuracy of roughly 10 m or better on sandy beaches after tidal correction, but that figure degrades on rocky, mangrove-fringed or heavily developed coasts where the land-water boundary is ambiguous.
SAR avoids cloud contamination but introduces its own complications. Wave run-up, wind-roughened intertidal pools and emergent aquatic vegetation all affect C-band backscatter in ways that shift the apparent waterline. Neither method is unconditionally superior; the strongest analyses use both, with optical providing the long archive and SAR filling cloud-season gaps.
Quantifying exposure at landing sites and aquaculture infrastructure
Once a shoreline change rate is established per coastal segment, exposure analysis is straightforward in principle. A fish landing site 15 m from the current waterline on a coast retreating at 1.2 m per year has a physical exposure horizon of roughly 12 years before the waterline reaches the structure toe, absent any intervention. That number is not a prediction of failure; it is a planning input.
Aquaculture ponds built on low-lying coastal flats present a different geometry. The risk is not direct wave erosion of the pond berm but inundation from the seaward side as the protective beach or mangrove buffer narrows. Satellite time series can track both the seaward shoreline and the pond berm position independently, flagging sites where the buffer width is shrinking toward a threshold that operators or insurers have defined.
Resolution matters here. A 30 m Landsat pixel cannot resolve a 5 m-wide berm. Planet SuperDove at 3–4 m can, but its archive depth is limited to roughly 2017 onward and commercial access is required. The honest answer for most clients is a hybrid: Landsat for the long-term trend, Sentinel-2 for the recent decade at 10 m, and commercial imagery for site-specific engineering-grade assessment where budgets allow.
Storm seasons, cloud cover and the SAR argument
Tropical cyclone seasons coincide with the periods of highest coastal change. A coastline that loses 3 m of beach during a single storm event will show that loss in the next clear optical image, which may arrive weeks or months later. By then, partial recovery may have occurred, and the peak erosion is invisible in the archive.
Sentinel-1 acquires through cloud at 6-day intervals. A SAR-derived shoreline extracted two days after a storm landfall captures the maximum erosion extent before sand transport begins to rebuild the beach. That observation is operationally important for damage assessment and for separating storm-driven episodic erosion from chronic background retreat. The limitation is that SAR waterline extraction on complex coasts requires more manual quality control than optical NDWI, and the Sentinel-1 archive only extends to 2014.
From change rate to investment decision
A shoreline change rate map is not itself a decision. It becomes one when overlaid with asset registries: jetty locations, cold-store facilities, hatchery buildings, road access points. The overlap produces a ranked exposure list. Sites retreating fastest with the highest asset value and shortest replacement lead time sit at the top. That ranking is what a port authority, a development bank or a national fisheries ministry actually needs.
Satellize applies this workflow to open constellations as a standard analytics product, with commercial tasking added where resolution demands it. The approach is directly analogous to the spatial analysis methods used in the Kingdom of Tonga crop-estimation programme, adapted from agricultural field boundaries to coastal infrastructure polygons. Outputs are delivered as GIS layers with associated change-rate tables, suitable for direct ingestion into infrastructure asset management systems.
One honest caveat: satellite-derived shoreline change rates describe the past. They are a reasonable basis for near-term projection on coasts with stable sediment budgets and no planned interventions. On coasts where river damming has recently cut sediment supply, or where a new groyne field has been installed, the historical rate may be a poor guide to the next decade. Ground-truthing at the time of commission is not optional; it is the step that converts a satellite product into an engineering input.
Typical figures
| Spatial resolution (optical) | 30 m (Landsat 5/7/8/9); 10 m (Sentinel-2 MSI); 3–4 m (Planet SuperDove, commercial) |
| Spatial resolution (SAR) | 10 m ground range detected (Sentinel-1 IW mode) |
| Revisit interval | 16 days per Landsat satellite; 5 days (Sentinel-2A+B combined); 6 days (Sentinel-1A+B combined); near-daily (Planet SuperDove) |
| Archive depth | 1972–present (Landsat MSS/TM/ETM+/OLI); 2015–present (Sentinel-2); 2014–present (Sentinel-1); ~2017–present (Planet SuperDove) |
| Shoreline extraction accuracy | Approximately 10 m or better on sandy beaches after tidal correction (optical NDWI, published CoastSat benchmark); degrades on rocky or vegetated coasts |
| Minimum detectable change rate | Approximately 0.5–1 m/year over a multi-decadal Landsat time series; finer rates require longer baselines or higher-resolution imagery |
| Primary spectral bands used | Green and NIR (NDWI); SWIR for improved water discrimination in turbid or shallow conditions; C-band (5.4 GHz) for SAR waterline |
| Tidal correction requirement | Required for quantitative change rates; global tidal models (FES2014, TPXO) used; residual error depends on local tidal range and beach slope |
| Delivery formats | GeoTIFF shoreline position rasters; GeoJSON/Shapefile shoreline polylines; CSV change-rate tables per coastal segment; PDF exposure summary reports |
| Cloud limitation | Optical sensors unusable under persistent cloud; SAR fills gaps but requires additional QC on complex coasts |
Analytics Satellize can run
| Multi-decadal shoreline change rate map | NDWI-based waterline extraction from Landsat and Sentinel-2 time series, tidal correction via global tidal model, linear regression per coastal segment (CoastSat-class method) | GIS layer (GeoJSON/Shapefile) with change rate (m/year) and uncertainty per 500 m coastal segment, plus PDF summary |
| Storm-event erosion extent | Pre/post-storm SAR waterline differencing (Sentinel-1 IW GRD backscatter threshold), cloud-independent acquisition within days of landfall | GeoTIFF difference raster and GeoJSON erosion polygon with estimated volume loss where beach slope data are available |
| Infrastructure exposure ranking | Spatial overlay of shoreline retreat projections with client-supplied asset registry; ranked by years-to-impact at current change rate | Ranked asset table (CSV/Excel) with site-level exposure scores, delivered as GIS layer and PDF report |
| Aquaculture buffer-width time series | Dual shoreline extraction tracking seaward beach position and pond berm position independently; buffer width computed per site per image date | Time-series chart and GIS layer per site; alert flag when buffer narrows below client-defined threshold |
| Seasonal vs. chronic change decomposition | Harmonic regression on annual shoreline position time series to separate seasonal oscillation from long-term trend; applied to Landsat/Sentinel-2 combined archive | Per-segment trend and seasonal amplitude table; GIS layer distinguishing chronic erosion zones from seasonally dynamic zones |
| High-resolution site assessment (commercial tasking) | Planet SuperDove 3–4 m imagery for narrow intertidal features below Sentinel-2 detection limit; NDWI and manual digitisation | Site-specific shoreline position map at 3–4 m resolution, with comparison to Sentinel-2 baseline |
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.