Tidal current and flushing capacity assessment for aquaculture site licensing
Satellite altimetry, assimilated ocean models and Sentinel-2 turbidity time series can characterise surface tidal flushing at candidate cage sites, giving applicants a defensible evidence base before committing to costly in-situ ADCP deployments.
Sensors
- Sentinel-6 Michael Freilich (radar altimeter): Provides sea-surface height anomalies along ground tracks spaced roughly 315 km apart at the equator, with a 10-day exact-repeat cycle. Assimilated into CMEMS ocean models to constrain geostrophic surface current fields at mesoscale resolution of approximately 25 km. Not useful for resolving tidal currents in fjords or narrow straits directly; its value is in anchoring the broader circulation model.
- Copernicus Marine Service NEMO-based models (CMEMS): Assimilated operational ocean models, such as the Atlantic-Iberian Biscay Irish (IBI) and the Arctic Ocean Physics Analysis, deliver gridded surface current fields at 1/36-degree (~3 km) horizontal resolution with daily to sub-daily output. Tidal constituents are included in regional configurations. Latency for near-real-time products is typically 24 hours; reanalysis products extend back to 1993.
- CMEMS HF radar surface currents: Shore-based HF radar networks, aggregated and distributed through CMEMS, measure surface current vectors at 1–3 km spatial resolution and hourly temporal resolution within roughly 200 km of the coast. Coverage is patchy: Norway's coast and parts of the Scottish Minch have operational networks; many Southeast Asian candidate sites do not. Where available, these are the most direct satellite-era surface current observation for tidal characterisation.
- Sentinel-2 MSI (multispectral imager): 10 m resolution in visible bands, 5-day revisit at mid-latitudes with two satellites. Turbidity derived from the red and near-infrared bands (B4 at 665 nm, B5 at 705 nm) using published algorithms such as the Nechad et al. SPM approach. Cloud cover is the primary constraint: Scottish and Norwegian sites average 60–80% cloud cover, so building a tidal-phase-resolved time series can require 12–18 months of acquisition. Provides surface information only.
- Jason-3 (radar altimeter): Predecessor to Sentinel-6 on the same reference ground track, 10-day repeat, contributing to the multi-mission along-track sea-surface height record used in model assimilation. Extends the altimetric record back to 2016 (Jason-3) and, through the merged TOPEX/Poseidon lineage, to 1992. Useful for validating the low-frequency circulation context of a site.
What the Norwegian standard actually demands, and where satellites fit
NS 9410, the Norwegian standard for environmental monitoring of marine fish farms, requires applicants to demonstrate that current speeds at the proposed site are sufficient to disperse waste and prevent organic enrichment of the seabed beyond defined thresholds. The standard references a minimum current speed of 2 cm/s at cage depth as a rough lower bound for adequate flushing, though the formal dispersion modelling it mandates uses full vertical current profiles, not surface values alone. Scotland's CAR licensing regime and several Southeast Asian frameworks impose analogous requirements, differing mainly in the specific benthic threshold metrics.
Satellites cannot deliver a vertical current profile. That remains the domain of moored acoustic Doppler current profilers (ADCPs). What satellite-derived products can do is characterise the surface tidal regime, identify the dominant circulation patterns, and flag sites where flushing is likely to be marginal before an applicant spends weeks deploying instruments in the wrong location. That pre-screening function has real economic value: a 30-day ADCP deployment in a Norwegian fjord costs on the order of tens of thousands of euros, and choosing the wrong candidate site is an expensive mistake.
Reading the tide from orbit: altimetry, model assimilation and HF radar
Sentinel-6 Michael Freilich and its predecessor Jason-3 measure sea-surface height to centimetre-level accuracy along fixed ground tracks. Individual passes over a fjord entrance or a strait are rare, but the long time series, dating back through the Jason lineage to 1992, feeds the assimilation schemes of operational ocean models distributed through the Copernicus Marine Service. The NEMO-based IBI model, for instance, produces tidal-constituent-resolved current fields at roughly 3 km horizontal resolution. That is adequate for characterising the tidal regime of a broad sea loch or an open coastal bay, but it cannot resolve the complex bathymetric steering that governs currents inside a narrow fjord arm.
Where HF radar networks exist, the picture sharpens considerably. CMEMS aggregates observations from shore-based HF radar systems across European waters, delivering hourly surface current vectors at 1–3 km resolution. A site in the Minch or along the Norwegian Helgeland coast can be characterised through a harmonic tidal analysis of this record, resolving the principal M2 and S2 constituents and their phase relationships. The honest caveat is coverage: much of coastal Southeast Asia and many remote Scottish sea lochs fall outside any existing HF radar footprint, leaving the analyst dependent on the coarser model output alone.
Turbidity as a flushing proxy: what Sentinel-2 can and cannot tell you
Water renewal at a cage site leaves a visible signature in suspended particulate matter concentrations. A site that flushes well on each tidal cycle will show turbidity values that oscillate with tidal phase and respond rapidly to wind events. A poorly flushed embayment will accumulate particulates and show persistently elevated turbidity relative to adjacent open water. Sentinel-2's 10 m resolution in the red and red-edge bands allows semi-analytical retrieval of surface suspended matter concentrations using published algorithms, with detection limits in the range of 1–3 g/m³ in clear coastal water.
Building a tidal-phase-resolved turbidity time series requires patience. Cloud cover at Scottish and Norwegian latitudes means that, even with the five-day revisit of the two-satellite Sentinel-2 constellation, acquiring 20 or more usable images at known tidal states can take 12 to 18 months. In clearer tropical waters, the same archive depth is achievable in three to four months. The method is also strictly a surface measurement: it says nothing about bottom boundary layer dynamics, which is precisely where benthic organic enrichment occurs. Turbidity time series are therefore a screening indicator, not a substitute for sub-surface monitoring.
Combining the data streams into a site characterisation
A defensible pre-licensing assessment typically proceeds in three stages. First, the CMEMS tidal current climatology is extracted for the candidate site and its neighbours, identifying the dominant tidal constituents, peak spring current speeds at the surface, and the residual circulation direction. Sites where the modelled surface current falls below 2–3 cm/s through most of the tidal cycle are flagged as high-risk for poor flushing without further investigation.
Second, the Sentinel-2 turbidity archive is processed to build a time series keyed to tidal phase using co-registered tidal predictions. Persistent turbidity anomalies relative to adjacent open-water reference polygons indicate reduced exchange. Third, where HF radar data exist, a harmonic tidal analysis of the observed surface current record provides direct validation of the model output and a quantitative estimate of tidal excursion, the distance a water parcel travels over a half tidal cycle, which is a useful first-order flushing metric. Sites passing all three screens are prioritised for ADCP deployment; sites failing any one are either deprioritised or redesigned.
Honest limits: what this approach cannot resolve
The vertical structure of currents in stratified fjords is not captured by any of these methods. Norwegian fjords in summer develop strong haloclines that decouple surface and bottom flows; a site with vigorous surface tidal currents may have near-stagnant water at cage depth. Sentinel-2 sees only the top metre or two of the water column. CMEMS models represent stratification in a bulk sense but are not validated at the spatial scales of individual farm sites. HF radar is by definition a surface measurement.
Cloud cover imposes hard limits on optical time series in high-latitude and monsoon-affected regions. SAR-derived surface roughness can infer current boundaries in some conditions, but the relationship between radar backscatter and current speed is ambiguous and not suitable for quantitative flushing assessment without extensive ground-truthing. Satellite-derived products are best understood as a way to reduce the uncertainty before in-situ work begins, not to replace it. Formal dispersion modelling under NS 9410 or equivalent frameworks will still require ADCP data. Satellize's role in this workflow is to deliver the satellite screening layer and the associated uncertainty characterisation, structured to align with what the regulatory submission actually needs.
Typical figures
| CMEMS model surface current resolution | ~3 km (IBI regional configuration); 1/12-degree global |
| CMEMS model temporal resolution | Hourly to daily; tidal constituents resolved in regional models |
| HF radar surface current resolution | 1–3 km spatial, hourly temporal (where network coverage exists) |
| Sentinel-6 / Jason-3 altimetry repeat cycle | 10 days; ground track spacing ~315 km at equator |
| Sentinel-2 MSI spatial resolution (turbidity bands) | 10 m (B4), 20 m (B5, B6); 5-day revisit (two-satellite) |
| Turbidity detection limit (Sentinel-2) | ~1–3 g/m³ SPM in clear coastal water; degrades in highly turbid scenes |
| CMEMS reanalysis archive depth | 1993 to present (altimetry-assimilated); Jason lineage from 1992 |
| Near-real-time model latency | ~24 hours for CMEMS physics analysis products |
| Vertical measurement depth | Surface only for all satellite and HF radar inputs; no sub-surface profiling |
| Delivery formats | NetCDF (current fields, model output), GeoTIFF (turbidity maps), CSV (tidal harmonic analysis), PDF site report |
Analytics Satellize can run
| Tidal current climatology for candidate site | Harmonic tidal analysis of CMEMS model output and HF radar observations; M2, S2, K1, O1 constituent extraction | GIS layer of peak spring and neap surface current speeds; tabular tidal constituent summary for regulatory submission |
| Tidal excursion estimate | Integration of modelled or observed surface current velocity over a half tidal cycle to derive water parcel displacement distance | Site-specific tidal excursion map (metres per half-cycle) with confidence range based on model vs. HF radar agreement |
| Sentinel-2 turbidity time series (tidal-phase resolved) | Semi-analytical SPM retrieval (Nechad et al. approach) applied to atmospherically corrected Sentinel-2 imagery; images keyed to co-registered tidal predictions | Monthly turbidity anomaly rasters and time-series plots for candidate site polygon vs. open-water reference; cloud-fraction metadata per scene |
| Flushing risk classification | Multi-criteria scoring combining modelled current speed percentiles, tidal excursion, and turbidity anomaly persistence against published threshold ranges | Traffic-light site classification (adequate, marginal, poor) with narrative uncertainty statement; structured for inclusion in regulatory pre-application report |
| ADCP deployment prioritisation brief | Spatial optimisation of mooring locations based on current field gradients and bathymetric steering identified in model output | Recommended ADCP mooring coordinates with rationale; identifies where satellite screening reduces or increases uncertainty |
| Historical circulation context report | Extraction of CMEMS reanalysis current statistics (1993 to present) for inter-annual and seasonal variability characterisation | PDF report with seasonal current roses, inter-annual anomaly time series, and comparison to regulatory minimum thresholds |
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.