Sea-surface temperature risk mapping for salmon lice proliferation in fjords
Satellite thermal infrared imagery can map the water temperatures that govern sea-lice development rates across salmon-farming fjords, but persistent cloud cover in Norway and Scotland makes multi-sensor data fusion essential for operational use.
Sensors
- Sentinel-3 SLSTR: Dual-view thermal infrared at 1 km nadir resolution, with a roughly one-day revisit over European waters at the latitude of Norway and Scotland. Designed for SST retrieval to better than 0.3 K bias against drifting buoys under clear sky. The 1 km pixel is too coarse to resolve individual farm pens but captures fjord-scale thermal gradients and cold freshwater intrusions.
- MODIS Aqua/Terra: 11 µm and 12 µm thermal bands at 1 km resolution, twice-daily combined revisit. The long archive (Aqua since 2002, Terra since 1999) supports multi-year climatological baselines of fjord SST. Cloud masking removes substantial fractions of the record in Atlantic-facing fjords; gap-filling requires temporal compositing or multi-sensor blending.
- VIIRS SNPP and NOAA-20: 375 m moderate-resolution and 750 m imagery day-night band, with SST products derived from the M-band thermal channels at 750 m. Slightly finer spatial detail than MODIS. The two-satellite constellation improves daily coverage. NOAA's STAR group publishes near-real-time SST products operationally.
- Landsat 8/9 TIRS: Thermal infrared at 100 m native resolution (resampled to 30 m in distributed products), 16-day revisit per satellite. The finer pixel footprint can resolve temperature differences between the mouth and head of a sea loch, and between sheltered bays where lice populations build up and the main fjord channel. The long revisit is a significant constraint for operational risk monitoring.
Why temperature is the governing variable for lice risk
Sea lice do not behave uniformly across a fjord. The copepodid larval stage, which is the infectious phase for salmon, develops faster in warmer water and survives longer in the water column when temperatures are above roughly 10 °C. The peer-reviewed development-rate models for Lepeophtheirus salmonis, including the widely cited Stien et al. (2005) formulation used in Norwegian regulatory frameworks, express development time as a temperature-dependent function: at 6 °C a full moult cycle takes roughly twice as long as at 12 °C. That relationship means a fjord with a warm surface layer in late spring carries a materially different infestation risk from one where cold freshwater runoff suppresses surface temperatures.
Spatial heterogeneity within a single fjord can be substantial. Stratification driven by meltwater, river discharge or tidal flushing creates pockets of anomalously warm or cold water that persist for days. A farm sited in a sheltered bay may experience surface temperatures several degrees above the main channel. Satellite SST mapping makes these gradients visible at the fjord scale, which farm-level sensors and sparse buoy networks cannot.
What the sensors actually see, and what they miss
Thermal infrared sensors measure the skin temperature of the top few micrometres of the water surface. That skin layer is a reasonable proxy for the upper metre or two relevant to copepodid survival, but it is not identical to the temperature at cage depth. Wind mixing and haloclines can decouple the skin from the bulk mixed layer. Any SST-based risk index should be understood as a surface signal, not a full water-column measurement.
Cloud is the dominant practical constraint. The fjordic environments of western Norway and the Scottish west coast sit under persistent Atlantic cloud systems. In winter months, cloud cover can exceed 80 % of observations on any given day. No thermal infrared sensor penetrates cloud. Operational systems therefore rely on multi-day compositing, typically five to ten day windows, to assemble spatially complete SST fields. That latency limits the value of the product for real-time lice alert purposes; it is more useful for trend monitoring and seasonal risk forecasting than for detecting a sudden thermal event overnight.
Landsat TIRS at 100 m offers the finest spatial detail of the freely available sensors, but its 16-day single-satellite revisit means cloud-free coverage over any given Scottish sea loch may occur only a handful of times per season. Sentinel-3 SLSTR and VIIRS trade spatial resolution for much better temporal sampling. In practice, an operational programme fuses both: coarser daily imagery for temporal continuity, finer imagery when cloud gaps allow for spatial calibration.
Turning a temperature map into a lice-risk index
The analytical step from SST to risk index applies published temperature-development-rate relationships to each pixel. A common approach computes accumulated thermal units, sometimes called degree-days above a threshold temperature, over a rolling window. Where the accumulated thermal exposure crosses a threshold consistent with one or more completed lice moult cycles, the pixel is flagged as elevated risk. The Norwegian Mattilsynet (Food Safety Authority) and the Scottish Sea Farms regulatory framework both use temperature as a primary input to lice management trigger thresholds, which gives the satellite-derived index a direct regulatory context.
The index can be disaggregated spatially to show which farm sites within a fjord are in the highest-risk thermal zone at any point in the season. It can also be expressed as a seasonal anomaly: how does this spring's thermal accumulation compare to the ten-year median derived from MODIS or VIIRS archives? An anomalously warm spring is a forward indicator of elevated lice pressure weeks before farm-level counts confirm it.
Limits that an honest risk product must declare
Three ambiguities deserve explicit acknowledgement in any delivered product. First, SST predicts development rate, not infestation load. Lice abundance also depends on larval dispersal from neighbouring farms, treatment history, host fish density and currents. Temperature is one governing variable, not the whole picture. Second, the skin-layer measurement introduces a bias in strongly stratified conditions; a halocline at one or two metres depth can mean the bulk mixed layer is warmer or cooler than the satellite skin by more than one degree. Third, the spatial resolution of operationally available daily SST products (750 m to 1 km) cannot resolve the thermal environment of individual net pens, which are typically 30 to 50 m in diameter. The product is a site-scale and fjord-scale risk indicator, not a pen-level measurement.
Users should also be aware that cloud-composited SST products smooth out short-duration thermal events. A warm pulse lasting 24 to 48 hours may be missed entirely in a five-day composite. For sites where rapid thermal fluctuation is known to occur, supplementary in-situ sensors remain necessary.
Operational design for a fjord monitoring programme
A practical programme for a regulatory body or farming group typically combines three data streams. A daily VIIRS or MODIS SST layer provides the temporal backbone, cloud-composited over five to ten days with the most recent clear-sky observation preserved where available. Landsat TIRS acquisitions, ingested opportunistically when cloud-free, anchor the spatial calibration and provide high-resolution snapshots during key seasonal windows. In-situ temperature loggers at a subset of farm sites provide ground truth for the satellite retrievals and flag skin-to-bulk discrepancies.
Satellize runs this kind of multi-sensor SST fusion on open constellations for aquaculture clients, drawing on the same analytical approach applied in its Tonga crop-estimation programme: open-satellite time series processed against published biological models to produce operationally useful indices rather than raw imagery. The deliverable for a lice-risk programme is typically a weekly GIS layer with pixel-level risk scores and a site-ranked alert table for the regulator or farm manager.
Archive depth is a genuine asset here. MODIS Terra data runs back to 2000, giving more than two decades of fjord SST from which to build climatological baselines and identify years with anomalously early or prolonged warm-water conditions. That historical context is as valuable to a regulator designing treatment-trigger thresholds as the near-real-time risk map is to a farm manager.
Typical figures
| Spatial resolution (daily SST, VIIRS) | 750 m (M-band thermal channels) |
| Spatial resolution (high-res SST, Landsat TIRS) | 100 m native, 30 m resampled product |
| Revisit (MODIS Aqua + Terra combined) | Twice daily, cloud permitting |
| Revisit (Landsat 8 + 9 combined) | 8 days (16 days per satellite) |
| SST retrieval accuracy (Sentinel-3 SLSTR) | < 0.3 K bias vs. drifting buoys under clear sky (ESA specification) |
| Spectral bands used | Thermal infrared: 10.8 µm and 12 µm (split-window retrieval) |
| Cloud obstruction in target environments | Typically 60–80 % of daily passes in Atlantic fjordic climates; compositing required |
| Archive depth | MODIS Terra from 2000; Landsat from 1984 (TM); VIIRS from 2012 |
| Latency (near-real-time VIIRS SST products) | 3–6 hours from overpass (NOAA STAR operational products) |
| Delivery formats | GeoTIFF, NetCDF, GIS-ready vector risk layers, tabular site-alert CSV |
Analytics Satellize can run
| Fjord-scale SST composite map | Multi-sensor split-window SST retrieval with cloud masking and temporal gap-filling by most-recent-clear-sky compositing | Weekly GeoTIFF or NetCDF layer covering named farm zones |
| Thermal degree-day accumulation index | Published temperature-development-rate functions (e.g. Stien et al. formulation) applied pixel-by-pixel to rolling 30-day SST time series | Gridded risk-score layer with farm-site point extracts, updated weekly |
| Seasonal SST anomaly relative to climatological baseline | Pixel-level departure from MODIS multi-year median for the same calendar period, expressed in degrees Celsius | Anomaly map and ranked-site table issued at start of each calendar month |
| Lice-risk alert table for named farm sites | Threshold exceedance detection on degree-day index; site coordinates intersected with risk grid | Automated alert feed (email or API) when site crosses defined risk tier |
| Thermal stratification proxy from skin-to-composite temperature variance | Standard deviation of SST across a five-day window as a proxy for mixing instability; flagged where variance exceeds 1 °C | Supplementary data quality flag layer accompanying the main SST product |
| Multi-year seasonal risk climatology report | Statistical summary of degree-day accumulation distributions across the MODIS archive for each farm zone, identifying high-risk and low-risk years | Annual PDF report with time-series charts and site-level statistics |
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.