Sensors
- Sentinel-1 SAR-C (ESA): C-band (5.405 GHz) synthetic aperture radar; IW mode ground range resolution approximately 5 x 20 m, swath 250 km, revisit 6 days per satellite (12 days single-satellite). Backscatter roughness patterns reveal surface current convergence, slicks, and internal wave signatures associated with tidal flow.
- Jason-3 radar altimeter (CNES/EUMETSAT/NOAA/NASA): Ku- and C-band dual-frequency altimeter; sea-surface height accuracy approximately 3-4 cm RMS along 10-day repeat ground tracks spaced ~315 km at the equator. Assimilated into FES2014 and TPXO9 to derive tidal constituents at coastal nodes.
- FES2014 tidal model (LEGOS/CNES, altimeter-assimilated): Global finite-element tidal solution assimilating TOPEX/Poseidon, Jason-1/2/3 and other altimeter records. Provides amplitude and phase for 34 tidal constituents at 1/16-degree resolution (~6 km). Coastal accuracy degrades in shallow, complex straits; published RMS error for M2 amplitude is typically 1-3 cm open ocean, rising to 5-15 cm in energetic shelf seas.
- Copernicus DEM GLO-30 (ESA/Airbus): 30 m horizontal resolution digital elevation model used as a bathymetric proxy in the absence of chart data, and for coastline delineation to constrain tidal model boundaries. Nearshore depths inferred from GLO-30 carry significant uncertainty; it is not a substitute for hydrographic survey.
- TOPEX/Poseidon heritage altimetry (NASA/CNES, 1992-2006): The foundational dataset from which global tidal constituent databases (TPXO series, FES family) were constructed. Its 10-day repeat cycle and dual-frequency design set the standard for open-ocean tidal recovery. The archive underpins every modern assimilated tidal model used in resource screening.
What the altimeter actually measures, and how tidal models are built from it
A radar altimeter measures the round-trip travel time of a microwave pulse to the sea surface. Correct for the geoid, atmospheric delay, and sea state, and what remains is sea-surface height (SSH) varying in time. Tidal signals dominate that variation in most coastal locations. By stacking years of repeat-track SSH observations, analysts can separate the major tidal constituents, principally M2 (principal lunar semidiurnal, period 12.42 hours), S2, N2, K1, and O1, using harmonic analysis. TOPEX/Poseidon provided the first global dataset dense enough to do this reliably, and its successors (Jason-1 through Jason-3) have extended the record to more than three decades.
FES2014 and the OSU TPXO9 model assimilate these altimeter-derived constituents into hydrodynamic finite-element solutions that respect coastline geometry and bathymetry. The output is amplitude and phase for each constituent at every model node. From those, you can reconstruct the tidal elevation time series at any point and, by applying the shallow-water momentum equations, estimate depth-averaged tidal current speed and direction. In open water this works well. In a 500-metre-wide strait with 40-metre depth variations, the model resolution of roughly 6 km means the computed current is a heavily smoothed approximation of what a turbine would actually experience.
What a SAR image gives away about flow that a tidal model cannot
C-band SAR measures centimetre-scale surface roughness. Strong tidal currents modulate that roughness in detectable ways. Where current shear concentrates surface films (biological slicks, thin oil, or simply suppressed capillary waves), the SAR backscatter drops sharply, producing the dark streaks visible in Sentinel-1 imagery of the Pentland Firth, the Alderney Race, and similar energetic sites. Conversely, current-induced surface convergence can roughen the water, brightening the image. Internal waves generated by tidal flow over sills appear as characteristic parallel bright-dark couplets. All of these signatures locate the spatial structure of the current far more precisely than any 6-km model grid can.
The limitation is equally clear: SAR sees the surface skin. It cannot directly measure current speed from a single image. Analysts derive qualitative flow structure and, with some caution, relative speed gradients by comparing backscatter intensity across a scene acquired at a known tidal phase. Combining multiple Sentinel-1 acquisitions at different tidal phases builds a picture of how the slick geometry shifts, which maps onto the tidal ellipse. The result is a spatially detailed, temporally sparse characterisation of surface flow patterns, not a current speed time series.
Combining the two sources for site screening
The practical workflow starts with the tidal model. FES2014 or TPXO9 gives depth-averaged current speed estimates at candidate sites. Locations where the model predicts peak spring currents exceeding roughly 1.5-2 m/s are worth examining further, since tidal-stream power density scales with the cube of velocity and sites below that threshold rarely justify the capital cost of an array. The model also provides the tidal phase at each location, which is needed to interpret SAR acquisitions correctly.
Sentinel-1 imagery, selected to match high-water and low-water phases at the candidate site, then adds spatial resolution. A site that looks promising in the tidal model but shows no coherent slick or roughness signature in SAR may have its energy dissipated across a broad shallow area rather than concentrated in a navigable channel. Conversely, a site where SAR consistently shows a sharp, stable slick boundary at peak ebb and flood is demonstrating that the current is spatially organised, which matters for array layout. The GLO-30 DEM contributes coastline geometry and a crude bathymetric proxy, useful for constraining the hydrodynamic model in data-sparse regions, but nearshore depths from an optical-derived or radar DEM carry errors of several metres and should be treated accordingly.
This combined screening can narrow a regional search area from dozens of candidate headlands to three or four sites worth deploying an acoustic Doppler current profiler (ADCP). That is its proper role: reducing the cost and time of the field campaign, not replacing it.
Where satellite methods stop and the ADCP takes over
Tidal-stream turbines operate at a specific hub height, typically 15-25 metres below the surface for bottom-mounted monopile designs. The current speed at that depth can differ substantially from the depth-averaged value that tidal models compute and from the surface value that SAR infers. The vertical current profile in a tidal channel is shaped by bottom roughness, stratification, and turbulence intensity, none of which satellite data can resolve.
Turbulence intensity is particularly critical: a site with high mean current but strong turbulent fluctuations imposes fatigue loads that shorten turbine blade life. No current satellite sensor measures turbulence directly. Significant wave height from altimetry is a related but distinct quantity. For any site that passes the satellite screening stage, a moored ADCP campaign of at least one spring-neap cycle (approximately 14 days, ideally a full year) is the minimum needed to characterise the vertical profile, turbulence intensity, and seasonal variability at hub height. Satellite data accelerates the selection of where to deploy that instrument, not the instrument itself.
Honest limits of the archive and the physics
Sentinel-1's 6-day revisit (per satellite, over Europe; longer elsewhere) means that for a given tidal phase, only a handful of acquisitions per year fall within the narrow window when the current is near peak spring. Building a statistically meaningful multi-image composite for a remote site may require two to three years of archive data. The Sentinel-1 archive runs from April 2014, giving roughly a decade of usable imagery, which is sufficient for most screening purposes.
FES2014 performs well for the dominant M2 constituent in open shelf seas but degrades in shallow, geometrically complex channels where the model grid cannot resolve the bathymetry. Published validation studies in the Alderney Race show M2 amplitude errors of 10-20 cm in the most energetic parts of the channel, translating to current speed errors of 15-25% at peak spring. TPXO9 shows similar limitations. Neither model is a substitute for a local hydrodynamic model calibrated with in-situ data.
Satellize runs tidal constituent extraction and SAR slick-mapping workflows on open Sentinel-1 and FES2014 data as part of its renewable energy analytics services, the same analytical approach it applies to Pacific island agriculture in the Tonga crop-estimation programme, adapted for a very different physical problem. The output is a ranked site-screening report with uncertainty ranges stated explicitly, not a turbine design specification.
Typical figures
| SAR spatial resolution (Sentinel-1 IW mode) | ~5 m range x 20 m azimuth (ground range detected); multi-look processing typically applied to 10-20 m pixels for slick analysis |
| SAR revisit at a given site | 6 days (two-satellite constellation over Europe); 12 days single-satellite; longer at high latitudes outside ESA priority zones |
| Tidal model spatial resolution (FES2014 / TPXO9) | 1/16 degree (~6 km); finite-element versions locally finer but not uniformly so in coastal straits |
| Tidal constituent accuracy (M2, open shelf) | Amplitude RMS ~1-3 cm open ocean; 5-20 cm in energetic coastal channels (published validation range) |
| Altimeter ground-track spacing (Jason-3) | ~315 km at equator; 10-day repeat cycle; SSH accuracy ~3-4 cm RMS |
| DEM resolution (GLO-30) | 30 m horizontal; vertical accuracy ~4 m LE90 over land; nearshore bathymetric proxy only, errors of several metres |
| Sentinel-1 SAR archive depth | April 2014 to present (~10 years); accessible via Copernicus Data Space Ecosystem |
| Minimum detectable surface current signature (SAR slick) | No hard published threshold; slicks associated with currents >0.5 m/s in low-wind (<4 m/s) conditions are typically detectable; high wind masks signatures |
| Deliverable formats | GeoTIFF tidal current speed maps, GeoPackage/Shapefile slick polygons, PDF site-screening report with ranked candidates and uncertainty ranges |
Analytics Satellize can run
| Tidal constituent map (M2, S2, K1, O1 amplitude and phase) | Harmonic analysis of FES2014/TPXO9 model output interpolated to candidate site grid; cross-checked against Jason-3 along-track SSH where tracks intersect the area of interest | GeoTIFF raster stack of constituent amplitudes and phases; tabulated peak spring current speed estimates with stated model uncertainty |
| SAR surface slick and roughness anomaly map | Sentinel-1 IW GRDH backscatter normalisation, adaptive thresholding for dark-feature extraction, tidal-phase tagging of each acquisition using FES2014 prediction | GeoPackage of slick polygons per acquisition, colour-coded by tidal phase; multi-image composite showing persistent slick locations |
| Ranked site-screening report | Overlay of tidal model peak spring current speed, SAR slick persistence score, GLO-30 bathymetric proxy, and navigational constraint layer; sites ranked by estimated energy density with explicit uncertainty bands | PDF report with site-ranked table, supporting maps, and recommended ADCP deployment priorities |
| Tidal current speed time series at candidate points | FES2014 constituent superposition reconstructed at hourly resolution for a user-specified period; spring-neap cycle envelope computed | CSV time series; summary statistics (mean spring peak, mean neap peak, directional asymmetry ratio) |
| Internal wave and tidal front detection | Sentinel-1 image gradient analysis to identify parallel bright-dark couplets indicative of internal wave generation over sills; spatial correlation with bathymetric sill locations from GLO-30 | GeoTIFF overlay of detected internal wave crests; interpretive note on flow structure implications for turbine siting |
| Multi-year SAR tidal phase composite | Selection and co-registration of all available Sentinel-1 acquisitions within ±1 hour of model-predicted peak spring ebb and flood; median backscatter composite per phase | Two GeoTIFF composites (peak ebb, peak flood) showing stable surface flow structure; difference image highlighting asymmetric flow patterns |
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.