Sensors
- Sentinel-2 MSI: 10 m resolution in visible and NIR bands, 5-day revisit at mid-latitudes. Blue (Band 2, 490 nm) and green (Band 3, 560 nm) penetrate clear shallow water to roughly 15–25 m depending on turbidity; SWIR bands (1610 nm and 2190 nm) resolve grain-size and moisture contrasts in emerged sediment. Free archive from 2015.
- WorldView-2 multispectral: 2 m multispectral resolution including a dedicated coastal-blue band at 427 nm, which penetrates water more deeply than standard blue channels and sharpens the bathymetric signal in very shallow, clear water. Used in published Caesarea Maritima work to map submerged harbour structures at sub-metre precision.
- Landsat 8/9 OLI: 30 m multispectral resolution with a dedicated coastal-aerosol band (Band 1, 443 nm) suited to shallow-water depth retrieval. 16-day revisit per satellite; combined Landsat 8 and 9 gives roughly 8-day coverage. Archive extends to 1972 (earlier missions), giving decades of coastline-change context.
- PlanetScope (contextual): 3 m daily revisit useful for monitoring turbidity windows: bathymetric retrieval requires low suspended-sediment conditions, and daily imagery lets analysts select the clearest acquisition rather than waiting for a fixed revisit.
What differential attenuation actually tells you
Water absorbs light selectively. Blue wavelengths (roughly 440–490 nm) penetrate furthest; green (530–570 nm) attenuates faster; red is gone within a few metres. By comparing the ratio of blue to green reflectance across a scene, analysts can infer relative water depth, a technique formalised by Stumpf et al. (2003) and since applied operationally across dozens of coastal archaeology projects. The method requires calibration against at least a handful of in-situ soundings or known-depth reference points, but once calibrated it produces continuous depth maps across the entire shallow-water extent of a scene.
The practical depth ceiling is roughly 15–25 m in clear oceanic water, dropping sharply as turbidity rises. In the silted, sediment-laden shallows typical of ancient harbour remnants, useful penetration is often only 5–10 m. That is still enough to map the residual basin geometry, trace submerged breakwater alignments and identify anomalous depth signatures that correspond to infilled dock channels. At Caesarea Maritima, WorldView-2 coastal-blue data resolved harbour structures at depths consistent with the known Roman-period construction levels, corroborating and extending results from acoustic survey.
The emerged half: reading sediment texture from SWIR reflectance
Once sediment has prograded above the waterline, the bathymetric signal disappears. The SWIR bands take over. Sentinel-2 bands at 1610 nm and 2190 nm are sensitive to soil moisture content and, to a lesser degree, grain-size distribution. Coarser, better-drained sediments reflect differently from fine silts and clays, and the boundary between ancient harbour fill and surrounding agricultural alluvium often produces a measurable spectral contrast, particularly after rain when moisture differences are amplified.
In the Nile Delta, where successive harbour sites from the Pharaonic through Byzantine periods lie buried under metres of alluvium, SWIR-based soil-moisture mapping has been used alongside historical map overlays and geophysical survey to constrain the likely extent of infilled basins. The satellite data cannot resolve individual buried features at this scale, but it can delineate sediment units across tens of square kilometres in a single acquisition, guiding where ground-based geophysical effort is best deployed. That triage function alone can save a field project weeks of unproductive survey.
Turbidity is the enemy: when the method fails and what to do
Satellite-derived bathymetry breaks down in turbid water. Suspended sediment scatters blue and green light before it reaches the bottom, producing a shallow apparent depth regardless of actual depth. Many ancient harbour sites sit in estuarine or river-influenced coastal zones where turbidity is chronically high. The honest answer is that bathymetric retrieval is often impossible in those conditions from a single image.
The practical workaround is multitemporal selection. Turbidity in coastal zones fluctuates with wind, tide and river discharge. A dense image archive, whether from Sentinel-2's five-day revisit or PlanetScope's daily coverage, allows analysts to identify low-turbidity windows, sometimes only a few days per year in silty estuaries, and use those acquisitions for depth retrieval. Seasonal patterns matter: post-storm turbidity spikes are predictable, and late-summer low-flow conditions in Mediterranean rivers often produce the clearest water. Planning acquisition around those windows is as important as the retrieval algorithm itself.
Coastline reconstruction: stitching the wet and dry evidence together
Ancient harbours are rarely either fully submerged or fully emerged. The interesting archaeology typically spans the intertidal and shallow subtidal zone, with some structures above water and others below. A complete reconstruction requires combining the bathymetric depth map with the emerged sediment classification, georeferencing both against a common datum, and integrating the result with any available historical cartography, LiDAR or geophysical survey.
Sentinel-2's consistent radiometric calibration across its archive from 2015 onwards makes it practical to stack multiple acquisitions, select the clearest per-pixel depth estimate and produce a composite bathymetric surface with reduced noise. The 10 m pixel size limits structural resolution: individual column bases or mooring blocks will not appear, but basin outlines, channel axes and breakwater alignments tens of metres across are within reach. WorldView-2 at 2 m closes some of that gap in the submerged zone, at higher tasking cost and with a narrower swath.
Landsat's archive depth is the underappreciated asset here. Comparing a 1980s Landsat scene with a current Sentinel-2 acquisition shows how much coastline has changed in the intervening decades, whether through natural progradation, land reclamation or erosion. For sites where the archaeological horizon is still being actively buried or exposed, that temporal baseline is direct evidence of preservation risk.
What the method cannot do, stated plainly
Satellite bathymetry maps water depth, not what lies beneath the sediment surface. A submerged harbour basin that has been completely infilled to the surrounding seabed level will produce no bathymetric anomaly at all. The method works only where some residual topographic expression remains, either as a depression in the water column or as a sediment-type contrast on the emerged surface. Completely buried, flat-topped infill is invisible to this approach and requires acoustic sub-bottom profiling or ground-penetrating radar.
Resolution is a genuine constraint. At Sentinel-2's 10 m, features narrower than roughly 20–30 m are unreliable. Harbour entrance channels, individual quay walls and small dock structures fall below this threshold. WorldView-2 improves matters substantially in the submerged zone but is not freely available and requires tasking or archive purchase. Cloud cover is the perennial obstacle in temperate and tropical coastal zones; analysts should budget for it and not assume a single acquisition will suffice.
Satellize runs bathymetric and sediment-mapping analytics on open Sentinel-2 and Landsat archives, with commercial WorldView-2 tasking available on client licence. The workflow is similar in structure to the sediment-classification component of the Tonga crop-estimation programme, adapted for coastal rather than agricultural contexts.
Integrating satellite output into a field programme
The most productive use of satellite-derived bathymetry in harbour archaeology is not as a standalone survey but as a pre-field prioritisation tool. A satellite-derived depth map and sediment classification, produced before a field season, tells a diving or geophysical team where the residual basin geometry is most legible, where sediment boundaries suggest ancient infill margins and where turbidity or depth will defeat acoustic methods. That spatial intelligence compresses the time needed to establish a survey grid.
Published projects at Caesarea Maritima and in the Nile Delta have followed exactly this sequence: satellite analysis first, targeted geophysical and diving survey second, excavation or coring third. The satellite layer does not replace the field work. It makes the field work cheaper and better aimed.
Typical figures
| Spatial resolution (bathymetric) | 10 m (Sentinel-2), 2 m (WorldView-2 multispectral), 30 m (Landsat 8/9 OLI) |
| Spatial resolution (SWIR sediment mapping) | 20 m (Sentinel-2 SWIR bands resampled to 20 m native), 30 m (Landsat OLI) |
| Revisit frequency | 5 days (Sentinel-2, mid-latitudes); 8 days combined (Landsat 8+9); 1–2 days (PlanetScope, contextual) |
| Maximum bathymetric depth | 15–25 m in clear oceanic water; typically 5–10 m in turbid estuarine or silted coastal settings |
| Key spectral bands | Coastal blue ~443–490 nm (depth penetration); green ~530–570 nm (depth ratio); SWIR 1610 nm and 2190 nm (sediment moisture and texture) |
| Minimum detectable basin feature | Approximately 20–30 m across at Sentinel-2 resolution; ~5–10 m at WorldView-2 resolution |
| Archive depth | Sentinel-2 from 2015; Landsat from 1972 (Landsat 1 MSS); WorldView-2 from 2009 (commercial archive) |
| Calibration requirement | In-situ depth soundings or known-depth reference points required for absolute bathymetric calibration; relative depth maps possible without |
| Cloud and turbidity constraint | Both defeat the method; multitemporal acquisition selection is the primary mitigation strategy |
| Deliverable formats | GeoTIFF depth raster, classified sediment polygon layer (GeoPackage/Shapefile), composite coastline reconstruction overlay, PDF interpretation report |
Analytics Satellize can run
| Satellite-derived bathymetric depth map | Band-ratio depth retrieval (Stumpf et al. log-linear method) calibrated against in-situ soundings or published chart data; applied to Sentinel-2 or WorldView-2 blue and green bands | GeoTIFF depth raster with uncertainty layer; PDF map with annotated basin geometry |
| Turbidity-window selection and composite clear-water mosaic | Multitemporal NDWI and turbidity index screening across Sentinel-2 archive to identify low-suspended-sediment acquisitions; per-pixel minimum turbidity compositing | Ranked acquisition list with turbidity scores; composite clear-water GeoTIFF for bathymetric input |
| SWIR sediment texture and moisture classification | Supervised classification of Sentinel-2 SWIR band combinations (Bands 11 and 12) to discriminate coarse sandy fill, fine silt, clay-rich alluvium and waterlogged zones on emerged surfaces | Classified sediment polygon layer (GeoPackage); spectral signature report by unit |
| Integrated palaeocoastline reconstruction overlay | Co-registration and fusion of bathymetric depth map, SWIR sediment classification and historical cartographic reference; manual interpretation by analyst against published site literature | Georeferenced reconstruction overlay (GeoTIFF and vector); annotated PDF interpretation report |
| Multidecadal coastline change analysis | Waterline extraction from Landsat archive (1972 to present) using NDWI thresholding; change vector analysis between epochs to quantify progradation, erosion and infill rates | Time-series waterline GeoPackage; progradation rate table by coastal segment; PDF summary |
| Field-survey prioritisation map | Spatial overlay of bathymetric anomalies, sediment unit boundaries and depth uncertainty to rank zones by archaeological potential and survey accessibility | Priority-zone GIS layer with ranked polygons; field-briefing PDF with recommended transect locations |
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.