Port expansion and nearshore bathymetric change detection
Multispectral and SAR satellites track dredging progress, reclamation fill and shoreline advance at port development sites. Satellite-derived bathymetry from blue-green band physics can estimate water depth to roughly 5–10 metres in clear conditions, with hard limits where turbidity takes over.
Sensors
- Sentinel-2 MSI: 10 m resolution in visible and near-infrared bands; 5-day revisit at the equator with both satellites. Coastal aerosol band (Band 1, 443 nm) and blue band (Band 2, 490 nm) are the primary inputs for satellite-derived bathymetry. Free and open archive from 2015.
- Maxar WorldView-2: 0.46 m panchromatic, 1.84 m multispectral including dedicated coastal blue (450 nm) and yellow bands. Sub-daily revisit over a target with tasking. The finer pixel gives better delineation of reclamation edges and vessel positions, but comes at commercial cost and with no systematic archive.
- Landsat 9 OLI: 30 m multispectral with a coastal aerosol band (Band 1, 435–451 nm) suited to bathymetric ratio methods. 16-day revisit, but the long archive back to Landsat 5 (1984) allows decadal shoreline-change baselines that no commercial sensor can match.
- Planet SuperDove: 3 m resolution, up to daily revisit. Eight spectral bands including coastal blue and green. Useful for tracking rapid fill-progress changes day-to-day, though its bathymetric performance in the blue band is less characterised than Sentinel-2 or WorldView-2 in peer-reviewed literature.
- Sentinel-1 SAR (C-band): Does not penetrate water for bathymetry, but provides cloud-independent detection of dredging vessels, floating plant, and reclamation-edge geometry at 10–20 m resolution on a 6-day repeat. Essential when tropical cloud cover makes optical observation unreliable for weeks at a time.
What the water colour is actually telling you
Shallow water transmits sunlight selectively. Blue and green wavelengths penetrate further than red before being absorbed or scattered back to a sensor. The ratio of log-transformed blue to green radiance correlates with depth in a relationship first formalised by Stumpf et al. (2003) and since validated across coral-reef and port environments worldwide. Sentinel-2's Band 1 (443 nm) and Band 2 (490 nm), WorldView-2's coastal and blue bands, and Landsat 9 OLI's Band 1 all exploit this physics.
The practical depth limit under good conditions is roughly 5 to 10 metres. Below that, the water column absorbs enough signal that depth and bottom reflectance become indistinguishable to the sensor. In port dredging programmes, where target depths are often 14 to 20 metres for Panamax or post-Panamax vessels, satellite-derived bathymetry (SDB) covers the shallow margins and the approach channels in their final shoaling stages, not the full dredge prism. That is an honest constraint and planners should size their survey budgets accordingly.
Turbidity is the method's enemy, and port sites produce plenty of it
Active dredging suspends fine sediment that raises water-column turbidity by orders of magnitude. High turbidity shifts the effective penetration depth toward zero: the sensor sees the plume surface, not the seabed. This means SDB results taken during or immediately after dredging are unreliable and should be flagged or discarded. A practical workflow acquires bathymetric imagery during scheduled dredging pauses or on days when wind and current have dispersed the plume, then uses the same imagery to map plume extent as a separate product.
Turbidity itself carries information. Plume geometry, drift direction and persistence can be monitored from Sentinel-2 at 10 m to assess whether spoil disposal is remaining within licensed dump-site boundaries. This is a regulatory compliance product, not a depth product, and it works in exactly the conditions where SDB fails. The two analyses are complementary rather than competing.
Tracking reclamation fill: where SAR earns its place
Reclamation fill is a land-creation process, so its progress is measurable by shoreline advance rather than bathymetry. Multispectral water-land classification from Sentinel-2 or Planet can detect a new land edge to within roughly one to two pixels, meaning 10–20 m positional accuracy for Sentinel-2 and 3–6 m for Planet. That is adequate for monthly progress reporting against a design polygon, though not for engineering sign-off.
SAR adds two things optical cannot. First, it works through cloud, which in tropical port environments can blank optical sensors for weeks during monsoon seasons. Sentinel-1 C-band imagery at 10 m resolution in Interferometric Wide Swath mode provides a consistent 6-day cadence regardless of weather. Second, backscatter intensity distinguishes rough fill material from calm water with high contrast, making automated edge detection straightforward even in low-light acquisitions. A combined optical-plus-SAR time series therefore gives more complete temporal coverage than either sensor alone.
Dredge-volume estimation: what is possible and what is not
If pre-dredge bathymetry is known from a hydrographic survey and post-dredge SDB can be computed from clear-water satellite imagery, the difference gives an approximate excavated volume over the depth range where SDB is valid. Studies using WorldView-2 in clear tropical waters have reported depth RMSE values of 0.3 to 1.5 metres against sonar ground truth, depending on bottom type and water clarity. Applied across an area, that translates to volume uncertainty that is acceptable for progress monitoring but not for payment certification.
For channels deeper than 10 metres, satellite data cannot substitute for multibeam echo-sounder surveys. The honest use of SDB in a dredging programme is to reduce the frequency of expensive vessel surveys in the shallow margins, flag areas where unexpected shoaling may have occurred, and provide a spatial context layer that helps survey crews prioritise their time.
Decadal baselines and the Landsat archive
Port expansion decisions are often contested on environmental grounds, and regulators or lenders may require evidence of pre-project shoreline position and nearshore habitat extent. The Landsat archive, continuous from 1984 with consistent radiometric calibration across Landsats 5, 7, 8 and 9, is the only freely available source for this. At 30 m resolution it cannot resolve individual structures, but it can establish where the waterline stood decades before construction began and quantify historical rates of natural accretion or erosion that would otherwise be argued over from anecdote.
Sentinel-2 extends the high-resolution record from 2015. Together, the two archives give a 40-year baseline at 30 m and a near-decade of 10 m imagery, which is sufficient to separate project-induced change from background coastal dynamics in most environments.
Putting it together for a project-finance audience
Lenders and project-finance advisers need periodic, independently sourced evidence that physical progress matches drawdown schedules. Satellite monitoring can provide monthly or fortnightly progress reports with georeferenced fill-extent polygons, plume-compliance summaries and, where water clarity permits, SDB depth grids over the shallow margins. These are not a replacement for the engineer's site report but they are an independent cross-check that does not require site access.
Satellize runs this kind of analytics workflow on open constellations supplemented by commercial tasking where resolution demands it. The Tonga crop-estimation programme demonstrated the same principle in a different domain: regular, calibrated change detection from satellite data delivered as a structured analytic output rather than raw imagery. For port teams or their lenders wanting to understand what a monitoring specification would look like for a specific site, the right starting point is a feasibility review against the site's water-clarity history and cloud climatology.
Typical figures
| Best spatial resolution (optical) | 0.46 m pan / 1.84 m multispectral (WorldView-2); 3 m (Planet SuperDove); 10 m (Sentinel-2); 30 m (Landsat 9 OLI) |
| Revisit cadence | Daily (Planet); 5 days at equator (Sentinel-2, both satellites); 6 days (Sentinel-1 SAR); 16 days (Landsat 9); sub-daily with tasking (WorldView-2) |
| SDB depth limit (optical) | 5–10 m in clear water; effectively zero in active-dredge turbidity conditions |
| SDB depth RMSE (published range) | 0.3–1.5 m against sonar ground truth in clear tropical water (WorldView-2 and similar) |
| Spectral bands used for bathymetry | Coastal aerosol (~443 nm) and blue (~490 nm); log-ratio method (Stumpf et al. 2003) |
| SAR frequency (Sentinel-1) | C-band, 5.405 GHz; cloud-independent; 10–20 m resolution in IW mode |
| Shoreline-edge detection accuracy | Approximately 1–2 pixels: ~10–20 m (Sentinel-2), ~3–6 m (Planet SuperDove) |
| Archive depth | Landsat from 1984; Sentinel-2 from 2015; Sentinel-1 from 2014; Planet from ~2016 |
| Latency (open data) | Sentinel-2 and Sentinel-1 typically available within 3–6 hours of acquisition via Copernicus Data Space |
| Delivery formats | GeoTIFF depth grids, GeoJSON shoreline polygons, PDF/Excel progress reports, GIS-ready change layers |
Analytics Satellize can run
| Monthly reclamation-fill progress map | Automated water-land classification (NDWI or modified NDWI) on Sentinel-2 or Planet, differenced against design polygon | GeoJSON fill-extent polygon with area and advance-rate statistics, PDF progress summary |
| Satellite-derived bathymetry grid | Log-ratio band transform (Stumpf et al. 2003) calibrated against available in-situ depth points; valid only in clear-water acquisitions flagged by turbidity index | GeoTIFF depth grid (0–10 m range) with uncertainty band and cloud/turbidity mask |
| Dredge-plume extent and drift analysis | Suspended sediment index from red and near-infrared bands on Sentinel-2; plume polygon delineated per acquisition | Time-series GIS layer of plume polygons with area, centroid drift vector and licensed dump-site compliance flag |
| Vessel detection and dredge-activity log | SAR backscatter object detection (Sentinel-1 IW mode) combined with AIS cross-reference where available | Tabular activity log of detected vessels by date and position; alert if plant is absent during contracted working windows |
| Decadal shoreline-change baseline | Waterline extraction from Landsat 5/7/8/9 archive (1984–present) using NDWI; linear regression on annual positions | Historical shoreline stack (GeoJSON), rate-of-change map, written baseline report for environmental or lender use |
| Cloud-gap-filled optical-SAR composite | Fusion of Sentinel-2 optical and Sentinel-1 SAR acquisitions to produce gap-free monthly composites during monsoon periods | Monthly GeoTIFF composite with data-source provenance layer; suitable for lender reporting under persistent cloud cover |
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.