Sensors
- ESA SMOS (Soil Moisture and Ocean Salinity): L-band (1.4 GHz) passive microwave radiometer providing sea-surface salinity at approximately 40–50 km spatial resolution with a 3-day global revisit. Accuracy in open ocean is around 0.2–0.5 PSU; degrades substantially within 100–200 km of coastlines due to land-sea radio-frequency contamination.
- NASA/CONAE Aquarius SAC-D: Three-beam L-band radiometer/scatterometer combination that operated 2011–2015, delivering salinity at roughly 150 km resolution and weekly global coverage. Provides the best-documented open-ocean salinity climatology for calibration baselines; coastal accuracy limitations mirror SMOS.
- Copernicus Marine Service GLORYS12 reanalysis: Global ocean reanalysis at 1/12° (approximately 8–9 km) horizontal resolution, assimilating altimetry, SST and in-situ profiles. Provides daily salinity fields back to 1993, filling the coastal gap left by L-band radiometry with physically constrained model estimates.
- Sentinel-3 OLCI: 21-band ocean colour radiometer at 300 m resolution with a 2-day revisit (two satellites combined). Does not measure salinity directly, but turbidity and coloured dissolved organic matter retrievals serve as optical proxies for fresh-water influence, helping locate the visible expression of salinity fronts at sub-kilometre scales.
Why salinity fronts matter more than the shrimp themselves
Penaeus monodon and Litopenaeus vannamei are euryhaline but not indifferent. Both species exhibit strong aggregation behaviour at salinity gradients during pre-spawning migration and post-larval ingress, preferring the 10–25 PSU transition zone where estuarine outflow meets coastal water. That preference is not incidental. The gradient concentrates suspended particulate matter, zooplankton and detritus, making it simultaneously a feeding opportunity and a navigational cue. Fishers who have worked the same estuary for decades know this empirically. Satellite salinity mapping makes it quantitative and predictive rather than anecdotal.
The commercial implication is straightforward. A front that shifts 20 km offshore following a heavy rainfall event can move an entire catchable aggregation out of range of artisanal gear overnight. Conversely, a front that tightens during the dry season compresses the target zone and raises catch-per-unit-effort sharply. Knowing where the front will be three to five days ahead, rather than where it was yesterday, is the operational difference between a productive season and a wasted one.
What L-band radiometry actually measures, and where it stops working
L-band microwave emission from the ocean surface is sensitive to the dielectric constant of seawater, which varies with salinity and temperature. SMOS measures brightness temperature at 1.4 GHz across a 2D interferometric aperture, then inverts it to salinity using a dielectric model. In open ocean conditions with sea-surface temperatures above 15°C, retrieval accuracy is approximately 0.2–0.5 PSU per observation, improving to around 0.1 PSU with monthly averaging. That is sufficient to map large-scale frontal systems in the Bay of Bengal, the Gulf of Mexico or the South China Sea.
Coastal zones are a different matter. Land masses emit strongly at L-band, and the SMOS synthetic aperture produces sidelobes that contaminate retrievals within roughly 100–200 km of the coast depending on land geometry and antenna orientation. In shallow enclosed seas and narrow gulfs, this contamination can be pervasive. Additionally, the 40–50 km footprint cannot resolve the sharp, kilometre-scale salinity gradients that form at the mouths of individual estuaries. Any workflow that presents SMOS data alone as a coastal shrimp-migration tool is overstating the sensor. The honest application is to use SMOS and Aquarius climatologies to establish the large-scale salinity regime and its seasonal envelope, then blend with GLORYS reanalysis and OLCI-derived turbidity to reconstruct the nearshore gradient at operationally useful resolution.
Building a usable front product from imperfect inputs
The practical workflow has three layers. First, SMOS Level-3 daily composites (or the Aquarius climatological baseline for historical seasons) define the offshore salinity field and its anomaly relative to the long-term mean. A positive freshwater anomaly pushing seaward signals enhanced riverine discharge and a likely offshore migration of the estuarine front. Second, GLORYS12 daily salinity fields at 1/12° are bias-corrected against any available in-situ CTD or Argo float data and used to resolve the nearshore gradient structure that SMOS cannot see. Third, Sentinel-3 OLCI turbidity and CDOM retrievals at 300 m are used to locate the visible expression of the front, validating the model field and providing the sub-kilometre positional fix that drives actual fishing-ground advisories.
Front detection itself uses standard gradient magnitude operators applied to the blended salinity field, with frontal probability estimated over a rolling 5-day window to smooth noise. The output is a probability-weighted front position with an uncertainty envelope, not a single line. That distinction matters operationally: a front with a 30 km positional uncertainty is still useful for planning a multi-day trip; it is not useful for directing a single boat to a precise location on a given morning.
Seasonal patterns and the rainfall complication
In the Indo-Pacific shrimp belt, the dominant driver of coastal salinity fronts is the monsoon cycle. During the southwest monsoon (June to September across much of South and Southeast Asia), river discharge peaks, freshwater plumes extend far offshore, and the 10–25 PSU zone can sit 50–100 km from the estuary mouth. Post-monsoon, as discharge falls, the front retreats shoreward and compresses. This is when nearshore aggregations of P. monodon tend to peak in many fisheries, coinciding with the main commercial season.
Rainfall complicates prediction in two ways. Intense localised events can shift a front by tens of kilometres within 48 hours, faster than GLORYS reanalysis can assimilate. And in years with anomalous monsoon strength, the entire seasonal envelope shifts, making climatological baselines misleading. The most defensible approach is to run the salinity front product as a real-time anomaly relative to the GLORYS climatology rather than as an absolute position, flagging when current conditions depart significantly from the seasonal norm. That framing also makes the product honest about what it cannot do: it cannot predict a front shift caused by a convective rainfall event that has not yet been assimilated into the model.
Operational limits a buyer should understand before commissioning
Spatial resolution of the blended product is bounded by GLORYS at roughly 8–9 km in the nearshore zone, with OLCI providing a qualitative refinement to approximately 300 m where turbidity contrast is sufficient. This means the product is suited to fleet-level planning across a fishing ground, not to directing individual vessels to metre-scale targets. In very shallow water (less than 10 m depth), both model physics and optical retrievals degrade: GLORYS does not resolve sub-grid bathymetric effects, and OLCI turbidity retrievals are affected by bottom reflectance.
Cloud cover interrupts OLCI observations. During active monsoon periods, optical data may be unavailable for days or weeks at a time, leaving the product dependent entirely on the model field. SMOS and Aquarius are unaffected by cloud but face the coastal contamination problem already described. L-band salinity retrieval also degrades at sea-surface temperatures below approximately 5°C, which is not a constraint in tropical shrimp fisheries but is worth noting for temperate applications. Finally, the salinity front is a necessary but not sufficient predictor of shrimp aggregation: water temperature, dissolved oxygen and prey availability all co-determine where animals concentrate. This product is most valuable as one layer in a multi-variable environmental advisory, not as a standalone forecast.
From data layer to fishing-ground advisory
The end product for a fisheries authority or aquaculture operator is a weekly front-position map with a 5-day probabilistic outlook, delivered as a GIS layer or georeferenced PDF, annotated with the salinity anomaly magnitude and the confidence rating of the OLCI validation. Where in-situ salinity logger networks exist near estuary mouths, those observations can be ingested to ground-truth and bias-correct the model field in near real time.
Satellize runs this kind of multi-source environmental analytics on open constellations, the same approach it uses in its Tonga crop-estimation programme. For a coastal fisheries authority, the practical first step is a single-season pilot covering one estuarine system, producing a hindcast comparison of front position against catch-log data to establish whether the salinity signal has predictive skill in that specific geography before committing to an operational service.
Typical figures
| Salinity retrieval spatial resolution (SMOS) | 40–50 km per observation; degrades further within 100–200 km of coastlines |
| Salinity retrieval spatial resolution (GLORYS12 reanalysis) | Approximately 8–9 km (1/12° grid), daily fields |
| Optical front proxy resolution (Sentinel-3 OLCI) | 300 m; cloud-limited, especially during monsoon periods |
| SMOS revisit | 3-day global; Level-3 composites typically 7-day or monthly for improved accuracy |
| Open-ocean salinity accuracy (SMOS) | 0.2–0.5 PSU per observation; approximately 0.1 PSU with monthly averaging |
| Frequency (L-band radiometry) | 1.4 GHz (L-band protected band) |
| GLORYS12 archive depth | 1993 to near-present (updated with approximately 1-week latency) |
| Aquarius SAC-D archive | August 2011 to June 2015; used for climatological baseline |
| Minimum detectable salinity gradient | Approximately 0.5 PSU over 40–50 km (SMOS open ocean); finer gradients require model or in-situ augmentation |
| Delivery formats | NetCDF, GeoTIFF, GIS vector (front polyline with uncertainty band), georeferenced PDF advisory |
Analytics Satellize can run
| Weekly blended salinity front map | Fusion of SMOS Level-3, GLORYS12 daily fields and OLCI turbidity proxy; gradient magnitude detection with frontal probability scoring over a 5-day rolling window | GeoTIFF and GIS vector layer showing front position, uncertainty envelope and salinity anomaly magnitude |
| 5-day probabilistic front-position outlook | GLORYS12 forecast fields (where available) combined with climatological anomaly regression; uncertainty quantified as positional range in kilometres | Georeferenced PDF advisory with confidence rating, suitable for fisheries authority distribution to fleet operators |
| Seasonal salinity anomaly report | Comparison of current-season SMOS and GLORYS fields against Aquarius-era and GLORYS climatological baseline; anomaly expressed in PSU departure and front-position offset | PDF seasonal bulletin with time-series charts and maps, issued monthly during the commercial shrimp season |
| OLCI turbidity front validation layer | Sentinel-3 OLCI total suspended matter and CDOM retrieval; cloud-masked compositing over 5-day windows; overlay against model salinity gradient to assess positional agreement | GeoTIFF composite with agreement score; flags where optical and model fronts diverge by more than a defined threshold |
| Hindcast skill assessment | Retrospective comparison of modelled front position against available catch-per-unit-effort records or in-situ salinity logs; Pearson correlation and spatial offset statistics | Technical report quantifying predictive skill for a named estuarine system; basis for go/no-go decision on operational service |
| In-situ bias-correction pipeline | Kalman-filter assimilation of estuary-mouth salinity logger data into the GLORYS field to reduce nearshore model error | Corrected daily salinity grid in NetCDF; logger data ingestion specification document |
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.