Coastal hazard mapping for submarine cable landing station risk assessment

Satellite-derived elevation models, shoreline time series and surge inundation layers quantify erosion, flooding and sediment-mobility risk at submarine cable landing stations. Optical sensors and lidar altimetry each contribute something the other cannot.

Sensors

Sentinel-2 MSI: 10 m visible and near-infrared bands, 5-day revisit at the equator with both satellites. Multitemporal composites spanning the full archive from 2015 allow decadal shoreline position extraction using the NIR/SWIR land-water boundary. Cloud contamination in tropical coastal zones can reduce usable acquisitions to fewer than 10 per year without compositing.
Landsat 8/9 OLI: 30 m multispectral, 16-day revisit per satellite, archive extending to 1984 for Landsat 5 and beyond. The longer record is the main asset here: 40-year shoreline change trajectories are possible where cloud cover permits. Coastal blue band (Band 1, 430–450 nm) aids shallow water discrimination, though turbidity limits depth retrieval to roughly 5–10 m in clear water and far less in estuarine conditions.
TanDEM-X: Global DEM at 12 m posting (90 m public release; 12 m available commercially), vertical accuracy typically 2 m LE90 over flat coastal terrain. Covers only above mean sea level. Inundation modelling under storm-surge scenarios uses this surface, but the product cannot represent sub-tidal bathymetry, and low-lying sandy coasts with relief under 1 m push against the vertical accuracy limit.
ICESat-2 ATL03: Photon-counting lidar with along-track spacing of roughly 0.7 m and vertical precision of approximately 3 cm over calm water surfaces. In clear, shallow coastal water (typically less than 30 m depth, often less than 10 m in practice) the green 532 nm channel penetrates to retrieve sub-aqueous topography along narrow ground tracks. Track spacing of roughly 91 km at the equator means spatial coverage is sparse; it supplements, rather than replaces, bathymetric survey.

Why a landing station is not just a building on a beach

A submarine cable landing station concentrates enormous economic and strategic value into a site chosen decades ago for nautical convenience, not long-term coastal stability. The station itself may be 50 to 200 metres inland, but the cable conduit runs across the intertidal zone and into the surf, where it is exposed to wave scour, sediment mobility and storm surge. If the beach retreats past the conduit entry point, or if a surge event overtops the site, the consequences extend far beyond a wet building.

Satellite data cannot replace a geotechnical survey or a bathymetric cruise. What it can do is provide the decadal context that in-situ surveys rarely capture: how fast the shoreline has been moving, whether the site sits on a prograding or retreating coastal cell, and what inundation extent a modelled surge scenario would produce given the best available topographic surface. That context is precisely what risk underwriters and cable consortium planners need before committing to a site or pricing an insurance policy.

Reading retreat in the archive: Sentinel-2 and Landsat shoreline time series

Shoreline position can be extracted from multispectral imagery using the water index contrast at the land-water boundary, most commonly the Modified Normalised Difference Water Index (MNDWI), which uses green and shortwave-infrared reflectance. Applied to Sentinel-2 from 2015 and Landsat from the 1980s, the method produces a time series of shoreline positions that, after tidal correction referenced to a datum such as mean high water, reveals net retreat or advance rates. Published studies using the CoastSat open-source workflow have demonstrated sub-pixel accuracy of roughly 5–15 m for sandy beaches imaged under calm conditions.

The honest caveat is cloud cover. Tropical sites, which include a disproportionate share of cable landing points in the Pacific, Indian Ocean and West Africa, can have fewer than 20 cloud-free acquisitions per year from any single sensor. Multisensor compositing across Sentinel-2 and Landsat improves the situation, but seasonal gaps remain. Where the shoreline is rocky or backed by mangrove, the MNDWI boundary is ambiguous and manual validation is necessary. Retreat rates derived from satellite data should be reported with confidence intervals, not as single-point estimates.

What a floating roof gives away: TanDEM-X and surge inundation

Storm-surge inundation modelling requires a digital elevation model of the coastal surface. TanDEM-X at 12 m posting is the most consistent global source above mean sea level, with a vertical accuracy that the DLR characterises as better than 2 m LE90 for flat terrain. That is sufficient to distinguish a site at 4 m elevation from one at 1 m, but it is not sufficient to resolve the difference between 0.8 m and 1.4 m on a low-lying atoll. For sites where the entire compound sits within 2 m of mean sea level, the DEM uncertainty is comparable to the hazard gradient, and the inundation map should be labelled accordingly.

Surge scenarios are typically derived from hydrodynamic models such as ADCIRC or SLOSH, which produce water-surface elevation outputs that can be draped over the TanDEM-X surface to estimate inundation extent and depth. The satellite contribution is the terrain surface, not the hydrodynamics. Where local lidar surveys exist, they should replace the TanDEM-X layer; where they do not, TanDEM-X is the best globally consistent alternative. The ICESat-2 ATL03 product can provide spot checks on the TanDEM-X elevations along its ground tracks, which is useful for validating the DEM before running inundation scenarios.

The turbid-water problem: what satellites cannot tell you about cable burial

Nearshore bathymetry, meaning the water depth from the surf zone out to the cable's designed burial depth, is critical for assessing whether the cable conduit has adequate sediment cover and whether scour events could expose it. Satellite-derived bathymetry using the ratio of blue and green reflectance works in clear, shallow water, and ICESat-2's green lidar channel extends this to perhaps 10–30 m depth in oligotrophic conditions. In turbid, estuarine or sediment-laden coastal water, optical penetration collapses to less than 2 m and sometimes less than 0.5 m.

A large proportion of cable landing points are located near river mouths or in regions with high suspended sediment loads, precisely because sheltered, low-energy coasts were historically preferred for cable laying. At these sites, satellite data provides essentially no bathymetric information. This is not a limitation that better processing overcomes; it is physics. The practical implication is that nearshore bathymetric surveys using acoustic methods remain necessary for cable burial depth assessment, and satellite data should not be presented as a substitute.

Putting the layers together into a risk product

A useful risk assessment for a cable landing station combines at least three satellite-derived layers: a decadal shoreline change rate with uncertainty bounds, a topographic surface with vertical accuracy characterised for the site's relief class, and a sediment-mobility or beach-state indicator derived from seasonal NDWI variability. These feed into a site-level risk matrix that distinguishes between erosion risk to the conduit entry point, inundation risk to the station building and compound, and sediment-mobility risk to the cable in the nearshore zone.

Satellize runs this class of coastal analytics on open-constellation data, including the Sentinel-2 and Landsat archives, and can task commercial optical or SAR sensors for sites where the open archive is cloud-compromised. The deliverable is a GIS package with attributed risk polygons, a time-series chart of shoreline position, and a written interpretation that is explicit about what the satellite data cannot resolve. For clients planning new cable routes, the same methodology applies to candidate landing sites before a site visit is committed to, which is where the cost saving is most tangible.

The Kingdom of Tonga crop-estimation programme demonstrated that rigorous archive-based analytics can serve Pacific island governments with limited in-country survey capacity. The same principle applies to cable operators assessing landing sites across remote island chains where ground surveys are expensive and infrequent.

Typical figures

Shoreline position spatial resolution	10 m (Sentinel-2), 30 m (Landsat 8/9)
Shoreline change detection accuracy	Approximately 5–15 m per epoch on sandy beaches (CoastSat-class methods); higher uncertainty on rocky or vegetated coasts
Sentinel-2 revisit	5 days at equator (both satellites combined); usable cloud-free acquisitions vary from fewer than 10 to more than 50 per year by site
TanDEM-X DEM posting	12 m (commercial); 90 m (public release)
TanDEM-X vertical accuracy	Better than 2 m LE90 over flat coastal terrain; degrades on slopes and in vegetated areas
ICESat-2 ATL03 vertical precision	Approximately 3 cm over calm water; along-track spacing 0.7 m; ground-track repeat 91-day exact repeat, ~91 km cross-track spacing at equator
Satellite-derived bathymetry depth limit	Up to ~30 m in clear oligotrophic water (ICESat-2 green lidar); less than 2 m or effectively zero in turbid or estuarine conditions
Landsat archive depth	Landsat 5 from 1984; Landsat 8 from 2013; Landsat 9 from 2021
Delivery format	GeoTIFF DEM and inundation rasters, GeoPackage shoreline time-series vectors, PDF risk summary with uncertainty characterisation

Analytics Satellize can run

Decadal shoreline change rate map	MNDWI-based waterline extraction applied to Sentinel-2 and Landsat multitemporal stack, tidal correction referenced to published tidal datum, linear and non-linear trend fitting with confidence intervals	GIS vector layer with per-transect retreat/advance rates and 95% confidence bounds; time-series chart per site
Storm-surge inundation extent under defined return-period scenarios	TanDEM-X 12 m DEM bathtub fill modelling using hydrodynamic model water-surface elevations (e.g. ADCIRC outputs or published FEMA/GEBCO-derived surge envelopes) as input water levels	Inundation depth raster per scenario (1-in-10, 1-in-50, 1-in-100 year); attributed polygon of compound and conduit exposure; DEM uncertainty flag where site elevation is within 2 m of model uncertainty
Seasonal beach-state and sediment-mobility indicator	Intra-annual NDWI and MNDWI variability across Sentinel-2 archive used as proxy for beach width oscillation and sandbar migration; high seasonal variance flags sites with active sediment transport	Annual beach-state summary table; flag for sites where seasonal shoreline oscillation exceeds 20 m
ICESat-2 nearshore depth profile (clear-water sites only)	ATL03 photon-counting data processed using published refraction-corrected bathymetry extraction (e.g. Parrish et al. method class); valid only where Kd(490) indicates low turbidity	Along-track depth profile GeoJSON with turbidity validity flag; explicit notation where retrieval is not possible due to water clarity
Multi-decadal coastal cell classification	Landsat archive shoreline positions from 1984 to present used to classify each 500 m coastal cell as erosional, stable or accretional; contextualises site-level risk within the broader coastal sediment system	Coastal cell classification GeoPackage with trend statistics; PDF narrative explaining sediment system context for the landing station site
Site risk matrix and written interpretation	Integration of shoreline change, inundation and sediment-mobility layers into a qualitative-quantitative risk matrix; explicit characterisation of what satellite data cannot resolve at the site	PDF report with risk matrix, data-gap register and recommended follow-on survey scope (bathymetric, geotechnical or lidar)

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.