Coastal erosion monitoring at maritime and littoral heritage sites

Satellite time-series can quantify shoreline retreat at coastal forts, middens and shipwreck sites years before a site is lost, giving heritage managers defensible rates and ranked priorities for emergency recording.

Sensors

Sentinel-2 MSI: 10 m multispectral resolution in visible and near-infrared bands; 5-day revisit at mid-latitudes with two satellites. The near-infrared and shortwave-infrared bands define the water-land boundary precisely enough to extract satellite-derived waterlines for DSAS transect analysis. Free archive from 2015.
Landsat 5/7/8/9 (USGS archive): 30 m multispectral resolution; archive extends to 1984, giving a 40-year baseline unavailable from any other open source. Coarser than Sentinel-2 but indispensable for detecting decadal retreat trends at sites with no earlier survey data.
PlanetScope: 3 m resolution, near-daily revisit. Resolves storm-event shoreline displacements that a 5-day sensor misses entirely. Commercially tasked; coverage is on-demand rather than guaranteed global. Useful for high-priority sites in the days following a named storm.
Sentinel-1 SAR (C-band): Approximately 10 m resolution in Interferometric Wide Swath mode; 6-day repeat. Cloud and darkness do not interrupt acquisitions, which matters for winter storm seasons when optical data gaps are worst. Coherence change between repeat passes flags surface disturbance, including newly exposed or collapsed structural fabric.

What a retreating shoreline actually destroys

Coastal heritage is not a single asset class. A Napoleonic sea fort, a Neolithic shell midden, a post-medieval fish-trap and a submerged Bronze Age landscape all sit in the same erosion envelope, but they fail differently. The fort loses masonry in discrete collapse events. The midden loses stratigraphy continuously, centimetre by centimetre, as wave undercutting and freeze-thaw work through the cliff face. The submerged landscape is exposed or reburied by sediment transport that can reverse seasonally. Satellite monitoring has to be designed around the failure mode, not just the rate.

Published erosion rates at vulnerable coastal heritage sites in northern Europe range from under 0.5 m per year at armoured or naturally resistant shores to more than 5 m per year at soft-sediment cliffs. At 2 m per year, a site with 20 m of lateral extent has roughly a decade before it is gone. That sounds generous until you account for recording backlogs, funding cycles and the reality that a single storm can remove several years' worth of average retreat in 48 hours.

How satellite-derived waterlines become retreat rates

The standard published workflow applies the Digital Shoreline Analysis System, developed by the USGS, to a stack of satellite-derived waterlines. Each image contributes one waterline, extracted at the water-land boundary using a spectral index, typically the Modified Normalised Difference Water Index calculated from green and shortwave-infrared bands. DSAS then casts shore-normal transects at regular intervals, typically 10 to 50 m apart, and fits linear or weighted linear regression to the waterline positions through time. The output is an end-point rate and a net shoreline movement distance per transect, with uncertainty bounds.

Sentinel-2 at 10 m resolution can resolve waterline positions to roughly 5 to 10 m in favourable conditions, meaning retreat rates below about 0.5 m per year are not reliably detectable over a five-year window. The Landsat archive extends the temporal baseline enough to detect slower trends, at the cost of spatial precision. Tidal stage at acquisition time is the largest confounding variable: a waterline extracted at low tide and another at high tide will produce a spurious retreat signal. Correcting for tidal offset using a published tidal model is not optional; it is the difference between a usable result and noise.

What SAR coherence adds that optical cannot

Optical waterline extraction tells you where the shoreline is. SAR coherence tells you whether the surface between two acquisitions has changed, regardless of whether it is wet, dark or cloud-covered. In a Sentinel-1 repeat-pass pair, coherence drops sharply where the surface has been disturbed, whether by wave overwash, cliff collapse, or the redistribution of beach sediment over structural remains.

For coastal forts and masonry structures, a coherence drop localised to a specific wall face or bastion is a reliable indicator of structural change worth ground-truthing. For midden sites, coherence loss across the cliff-top edge after a storm event can be mapped and compared with pre-storm optical imagery to estimate the volume of material removed. Neither method is a substitute for field survey, but together they give a heritage manager a ranked list of sites that changed since the last visit, not a uniform inspection schedule driven by administrative convenience.

One honest limit: SAR coherence at C-band is sensitive to vegetation change as well as structural change. Sites with dense cliff-top vegetation will show coherence loss after wind events that have nothing to do with erosion. Separating the two requires either cross-checking with optical imagery or, where available, higher-frequency SAR bands that are less sensitive to canopy motion.

The archive problem and the storm problem

Most coastal heritage sites have no georeferenced baseline survey. The Landsat archive from 1984 onward is the closest thing to a universal starting point, and even that requires careful co-registration and tidal correction before it yields defensible rates. For sites where earlier aerial photography or Ordnance Survey mapping exists in digitised form, those can be added as additional waterline epochs, extending the baseline to the mid-twentieth century.

Storm events are where the analysis gets uncomfortable. A single major storm can move a shoreline by tens of metres in hours, and the next cloud-free optical acquisition may not arrive for weeks. By then, the beach has partially recovered, the cliff face has been re-colonised by vegetation, and the satellite record shows a net change that understates the peak displacement. PlanetScope's near-daily revisit reduces this gap substantially, but it does not eliminate it. Pairing satellite monitoring with a simple wave-height threshold alert, using published hindcast or forecast products, allows a team to task high-resolution commercial imagery specifically around storm events rather than on a fixed schedule.

Turning rates into a triage list

The point of quantified retreat rates is not the number itself. It is the ranked priority list that follows. A site losing 3 m per year with 15 m of deposit remaining has roughly five years of accessible stratigraphy left. A site losing 0.3 m per year with 50 m of deposit remaining can wait. Combining DSAS-derived rates with a simple deposit-depth estimate, drawn from existing site records or a published regional geology layer, produces a time-to-loss figure per site that is directly actionable for emergency recording programmes.

Satellize has applied this kind of multitemporal analytics logic to agricultural monitoring in the Kingdom of Tonga, where the same principles of change detection and priority ranking apply across a dispersed island geography. The coastal heritage context is methodologically analogous: many sites, limited field access, and a need to direct scarce resources to the places changing fastest. The DSAS workflow and SAR coherence change detection are both well-documented in the peer-reviewed literature and can be applied to any coastline with an adequate Sentinel or Landsat archive.

What satellite monitoring cannot do, and what comes next

No satellite currently operational can see below the waterline with enough resolution and reliability to monitor intertidal or subtidal structural remains directly. Satellite-derived bathymetry using Sentinel-2 band ratios works in clear, shallow water, but turbidity, which is highest precisely during the storm events that cause the most damage, makes it unreliable at the moments that matter most. Submerged heritage monitoring from orbit remains a research frontier rather than an operational capability.

On land, the 10 m resolution floor of Sentinel-2 means that individual features smaller than roughly 20 to 30 m across are not reliably detected as separate objects. A narrow midden exposure, a single collapsed wall section or a small shipwreck scatter on a beach may fall below the detection threshold. This is where commercial tasking at 30 to 50 cm resolution, from operators such as Maxar or Airbus, provides the detail that open constellations cannot. Satellite monitoring at Sentinel scale is best understood as a screening tool that identifies which sites warrant that more expensive, targeted acquisition.

Typical figures

Optical spatial resolution	10 m (Sentinel-2), 30 m (Landsat), 3 m (PlanetScope)
SAR spatial resolution	~10 m (Sentinel-1 IW mode, single-look complex)
Revisit frequency	5 days (Sentinel-2, two satellites); 6 days (Sentinel-1); near-daily (PlanetScope, commercial); 16 days (Landsat 8/9)
Archive depth	Landsat: 1984 to present (USGS). Sentinel-2: 2015 to present. Sentinel-1: 2014 to present.
Minimum detectable retreat rate	~0.5 m/yr over 5-year window at 10 m resolution with tidal correction; ~1–2 m/yr at 30 m (Landsat)
Tidal correction requirement	Required for all waterline extraction; published tidal models (e.g. FES2014, TPXO) used to normalise acquisitions to mean sea level
Cloud cover limitation	Optical methods fail under persistent cloud; SAR unaffected. Winter storm seasons in mid-to-high latitudes may produce optical data gaps of 2–6 weeks.
Delivery formats	GeoTIFF shoreline rasters, GeoPackage/Shapefile DSAS transect outputs, PDF priority-ranking reports, GIS-ready coherence change layers
DSAS transect spacing	Typically 10–50 m along-shore, set to match site feature scale
Cost of open-data inputs	Sentinel and Landsat imagery: free. PlanetScope and commercial high-resolution: licensed per km² or per scene.

Analytics Satellize can run

Multitemporal shoreline position stack	MNDWI-based waterline extraction from Sentinel-2 and Landsat time-series, tidal-stage corrected using published tidal model	GeoPackage of dated waterlines per site, ready for DSAS ingestion
DSAS retreat-rate transects	Digital Shoreline Analysis System linear regression on waterline stack; end-point rate and weighted linear regression rate computed per transect	Shapefile with EPR, WLR and 95% confidence intervals per transect; summary statistics table
Storm-event shoreline displacement map	Pre- and post-storm PlanetScope or Sentinel-2 differencing; displacement magnitude mapped per 10 m cell	GeoTIFF displacement raster and PDF event report within 5 days of cloud-clear post-storm acquisition
SAR coherence change detection layer	Sentinel-1 repeat-pass coherence differencing between consecutive 6-day pairs; coherence loss thresholded to flag surface disturbance	GeoTIFF coherence-change layer per acquisition pair; alert flag for sites exceeding disturbance threshold
Time-to-loss priority ranking	DSAS retreat rate combined with deposit-extent estimate from site records or published geology; simple time-to-loss calculation per site	Ranked site table with estimated years of accessible deposit remaining; PDF briefing for heritage managers
Annual monitoring report	Aggregated DSAS rates, coherence change summary and storm-event log for a defined site portfolio over a 12-month window	PDF report with maps, transect plots and updated priority ranking; GIS data package

Sources

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.