Coastal lagoon and estuarine salinity mapping from orbit
L-band microwave radiometry detects ionic concentration in surface water, but its 40 km footprint misses most lagoons. Fusing SMOS/SMAP with Sentinel-2 optical indices extends usable salinity estimates into water bodies too small for the radiometer alone.
Sensors
- SMOS MIRAS (ESA): L-band (1.4 GHz) synthetic aperture radiometer. Spatial resolution approximately 43 km at nadir, degrading toward swath edges. Global revisit every 2 to 3 days. Measures brightness temperature from which sea-surface salinity is retrieved to roughly 0.2 to 0.5 PSU accuracy over open ocean; accuracy degrades significantly in shallow water and near land due to radio-frequency interference and land contamination within the large footprint.
- SMAP Radiometer (NASA): L-band (1.41 GHz) conically scanning radiometer. Fixed 40 km spatial resolution, 2 to 3 day global revisit. Originally designed for soil moisture; its ocean salinity product (released from 2015) achieves comparable open-ocean accuracy to SMOS but shares the same resolution floor. Near-coast land contamination affects retrievals within roughly 100 km of shore, directly limiting utility in enclosed lagoons.
- Sentinel-2 MSI (ESA): Multispectral imager, 10 m resolution in visible and near-infrared bands, 20 m in red-edge and shortwave infrared. 5-day revisit at the equator with both satellites combined. Does not measure salinity directly, but band ratios (particularly blue-to-green, turbidity-sensitive indices and coloured dissolved organic matter proxies) correlate with salinity gradients in optically shallow, sediment-laden estuaries. Cloud cover is the primary operational limit.
- Landsat 8/9 OLI (USGS/NASA): 30 m multispectral resolution, 16-day revisit per satellite (8-day combined). Coastal aerosol band (Band 1, 443 nm) is specifically useful for shallow water and suspended sediment discrimination. Longer archive than Sentinel-2 (Landsat record extends to 1972 in various forms) supports trend analysis of salinity-proxy indices over decades, though 16-day revisit limits tidal-cycle resolution.
Why L-band and not something more convenient
Salinity in water changes its dielectric constant. L-band microwave energy at 1.4 GHz is sensitive to that change in a way that shorter wavelengths are not. The physics is straightforward: the emissivity of saline water at L-band differs measurably from fresh water, and the long wavelength is largely indifferent to surface roughness and foam that would confound higher-frequency approaches. This is why SMOS and SMAP, both L-band radiometers, are the only satellite instruments that retrieve salinity directly.
The inconvenient corollary is resolution. A passive microwave radiometer at 1.4 GHz requires an impractically large antenna to achieve fine spatial resolution from orbit. SMOS uses an interferometric Y-shaped array 8 metres across and still achieves only 43 km at nadir. SMAP's 6-metre reflector dish yields 40 km. A lagoon the size of the Étang de Berre in southern France (approximately 15 km across) falls entirely inside a single SMOS pixel. Most aquaculture ponds, tidal creeks and river mouths are far smaller still. The instrument works; the footprint does not match the problem.
What the optical bands can and cannot substitute
Sentinel-2 and Landsat OLI resolve individual aquaculture ponds at 10 to 30 m. They cannot measure salinity directly, but in many estuarine environments salinity correlates with other optically active constituents. Freshwater inputs typically carry higher loads of coloured dissolved organic matter (CDOM) and suspended sediment, both of which shift reflectance in the blue and green bands. In turbid, sediment-laden systems such as river deltas, a blue-to-green ratio or a CDOM index can serve as a reasonable salinity proxy when calibrated against in-situ measurements. The relationship is site-specific and breaks down in clear oligotrophic lagoons where freshwater and saline water are optically indistinguishable.
Cloud cover is the operational constraint that never disappears. Coastal regions with active freshwater inputs, particularly in monsoon-affected or temperate maritime climates, can sustain cloud cover for weeks at a time. A 5-day Sentinel-2 revisit becomes a 30-day gap in practice. This is not a solvable problem with current optical assets; it is a design constraint that any honest salinity monitoring plan must account for.
Fusion: making the coarse radiometer useful at lagoon scale
The productive approach is statistical downscaling, not replacement. SMOS or SMAP brightness temperatures carry genuine salinity signal at their native 40 km scale. Sentinel-2 optical indices carry spatial structure at 10 m but only a proxy salinity signal. Regression-based or machine-learning fusion methods trained on concurrent in-situ salinity measurements can distribute the radiometric salinity estimate across the optical spatial field, guided by the optical gradients. Published studies using this class of method in coastal China, the Gulf of Mexico and the Baltic have reported downscaled salinity products at 1 to 3 km resolution, though the accuracy of the downscaled product depends heavily on how well the training data captures the range of conditions in the target system.
The honest limit: downscaling does not create information that was absent from the radiometer. It redistributes radiometric signal using optical structure as a guide. Where the optical proxy is weak (clear water, low CDOM) the downscaled product degrades toward the coarse radiometric estimate. Validation with in-situ conductivity sensors remains essential, particularly during the calibration phase and after major hydrological events that alter the optical-salinity relationship.
Tidal mixing and freshwater plume dynamics as operational questions
Tidal forcing complicates salinity retrieval in two ways. First, tidal cycles operate at 6 to 12 hour periods, far shorter than the 2 to 3 day SMOS/SMAP revisit. A single overpass captures one moment in the tidal cycle; comparing images acquired at different tidal phases introduces apparent salinity changes that are real but not the signal of interest to a manager tracking seasonal freshwater input. Second, tidal pumping drives the spatial structure of salinity gradients, so a map produced at high tide looks different from one produced at low tide even if the underlying freshwater budget has not changed.
Freshwater plumes from rivers and storm drains are more tractable. Plumes persist for days to weeks, their extent visible in both L-band brightness temperature anomalies and optical turbidity indices. Sentinel-2 can resolve the leading edge of a plume at 10 m; SMOS can confirm the salinity depression at the plume core even when optical retrievals are compromised by suspended sediment. The combination is more informative than either sensor alone.
Aquaculture siting: what satellite data actually answers
Salinity tolerance windows govern species selection in coastal aquaculture. Penaeus monodon (black tiger prawn) tolerates 5 to 35 PSU; Pacific oysters are sensitive to prolonged exposure below roughly 10 PSU; seabass and seabream require broadly marine conditions. A buyer siting a new grow-out facility needs to know the seasonal salinity range at candidate locations, the frequency and duration of freshwater flood events, and the proximity to sources of agricultural runoff.
Satellite data addresses the first two questions reasonably well over sites larger than roughly 1 km, provided in-situ calibration data exists. It does not replace water chemistry testing for nutrients, pathogens or dissolved oxygen, and it cannot resolve salinity within individual ponds at aquaculture-farm scale using current public sensors. Satellize runs fusion analytics on open Sentinel and Landsat archives combined with SMOS and SMAP brightness temperature data; the Tonga crop-estimation programme demonstrated the same multi-source integration logic in a Pacific island context where in-situ data is sparse. For a prospective aquaculture operator, the practical first step is a historical salinity-proxy time series at candidate sites, built before any capital is committed.
Archive depth and what a decade of data reveals
SMOS has operated since November 2009; SMAP since April 2015. Landsat 8 data begins in 2013, with the broader Landsat archive extending to the early 1970s in lower-resolution form. This gives a decade-plus of L-band salinity data and a much longer optical proxy record. Multi-year analysis can reveal whether a lagoon's salinity regime is shifting, whether upstream land-use change or altered river regulation is pushing freshwater intrusion further into historically saline areas, and whether seasonal patterns are becoming more or less predictable.
The caveat is consistency. SMOS brightness temperature products have undergone multiple reprocessing campaigns; comparisons across the full archive require careful version control. Landsat band calibration is well-documented by USGS but atmospheric correction over water remains a source of inter-scene variability. Any trend analysis should account for these instrumental artefacts before attributing change to the environment.
Typical figures
| L-band radiometer spatial resolution | 40 km (SMAP) to 43 km (SMOS at nadir); degrades to ~60 km at swath edges |
| L-band revisit | 2 to 3 days globally (SMOS and SMAP independently) |
| Open-ocean salinity retrieval accuracy | 0.2 to 0.5 PSU (published mission targets); coastal accuracy substantially lower due to land contamination and RFI |
| Sentinel-2 MSI spatial resolution | 10 m (visible/NIR), 20 m (red-edge/SWIR); 5-day revisit (Sentinel-2A+B combined) |
| Landsat OLI spatial resolution | 30 m multispectral; 8-day combined revisit (Landsat 8 + 9) |
| Minimum lagoon size tractable for L-band | Approximately 100 km² for direct radiometric retrieval; smaller bodies require fusion downscaling with optical data |
| Downscaled salinity product resolution (fusion methods) | 1 to 3 km (published literature range); accuracy is site- and calibration-dependent |
| Archive depth | SMOS from 2009; SMAP from 2015; Landsat optical proxy from 2013 (OLI), earlier with TM/ETM+ |
| Primary operational limit | Cloud cover for optical; land contamination and RFI for L-band radiometry near shore |
| Delivery formats | GeoTIFF salinity-proxy grids, NetCDF time-series stacks, tabular site statistics, PDF seasonal reports |
Analytics Satellize can run
| Historical salinity-proxy time series at candidate aquaculture sites | Optical index regression (Sentinel-2 blue/green ratio, CDOM index) calibrated against published in-situ datasets or client-supplied gauge records | Site-specific CSV time series and PDF seasonal summary, covering available Landsat and Sentinel-2 archive |
| L-band brightness temperature anomaly maps for freshwater plume events | SMOS and SMAP Level-2 brightness temperature differencing against climatological baseline | GeoTIFF anomaly maps per overpass, with plume extent polygon in GeoJSON |
| Downscaled salinity field at 1 to 3 km for target lagoon | Statistical fusion of SMOS/SMAP brightness temperature with Sentinel-2 optical indices using regression or random-forest downscaling, trained on client-supplied or published in-situ salinity data | Monthly GeoTIFF salinity grids with uncertainty band, delivered via GIS-compatible feed |
| Tidal-phase-stratified salinity composites | Overpass timestamps cross-referenced against published tidal prediction models (e.g. FES2014) to bin retrievals by tidal phase before compositing | Separate high-tide and low-tide salinity-proxy composites per season, GeoTIFF format |
| Multi-year salinity regime trend assessment | Mann-Kendall trend test on annual salinity-proxy index time series derived from Landsat OLI and Sentinel-2 archive | Trend significance map and written interpretation report flagging sites with statistically detectable change |
| Aquaculture siting suitability layer | Composite scoring of salinity seasonality, freshwater flood frequency (from optical inundation indicators) and proximity to runoff sources, weighted by species-specific tolerance thresholds from published aquaculture literature | Ranked candidate-site GeoJSON layer with per-site data table, suitable for import into GIS or planning software |
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.