Ocean colour and coastal water quality retrieval

Sentinel-3 OLCI, MODIS and VIIRS measure water-leaving radiance across visible wavelengths to derive chlorophyll-a, suspended matter and dissolved organics. This page explains the physics, the algorithms, and the hard limits that apply near coastlines.

Sensors

Sentinel-3 OLCI: 21 spectral bands from 400 to 1020 nm, 300 m spatial resolution, global revisit roughly every two days with both Sentinel-3A and 3B combined. Designed specifically for ocean colour; bands are placed to isolate chlorophyll fluorescence at 681 nm and separate pigment absorption from sediment scattering. The 300 m floor is a genuine constraint for estuaries narrower than a few hundred metres.
MODIS Aqua: 36 bands; ocean colour bands at 250 m to 1 km depending on channel. Daily global coverage at 1 km for most ocean colour products. Operational since 2002, giving more than two decades of archive for trend analysis. Lower spatial resolution than OLCI limits coastal work in confined water bodies.
VIIRS SNPP / NOAA-20: 750 m resolution ocean colour bands, daily global coverage. Designed as the MODIS successor with improved radiometric stability. The M-band suite covers 412 to 862 nm. Useful for open-ocean and large coastal systems; shares the resolution constraint of MODIS in narrow inlets.
Landsat 8/9 OLI: 30 m resolution across six reflective bands including a dedicated coastal/aerosol band at 443 nm. Revisit is 8 days per satellite, 16 days for a single satellite. Not purpose-built for ocean colour, so atmospheric correction over water is more demanding and signal-to-noise ratios are lower than OLCI. Useful for small estuaries and lagoons where 300 m resolution is insufficient.

What water-leaving radiance actually tells you

When sunlight enters the ocean, some fraction scatters back upward through the surface. The spectral composition of that upwelling light, once you strip away the atmosphere's contribution, is the water-leaving radiance. It carries the fingerprint of whatever is in the water column: phytoplankton pigments absorb strongly in the blue and red, shifting the apparent colour toward green; suspended sediment scatters broadly, brightening the signal across the visible; coloured dissolved organic matter (CDOM) absorbs in the blue, pushing the spectrum toward longer wavelengths.

The three quantities routinely derived from this signal are chlorophyll-a concentration (a proxy for phytoplankton biomass and primary productivity), total suspended matter (TSM, which tracks sediment load and turbidity), and CDOM absorption. Each has a distinct spectral signature, but they overlap enough that separating them requires careful algorithm design, particularly in coastal waters where all three are often elevated simultaneously.

Atmospheric correction: the step that makes or breaks everything

Roughly 90 percent of the signal reaching a satellite sensor over the ocean originates in the atmosphere, not the water. The water-leaving radiance is a small residual. Atmospheric correction must account for Rayleigh scattering from air molecules, aerosol scattering and absorption, and sun glint from the surface. Over the open ocean, algorithms exploit the near-infrared bands where water absorbs almost completely, using those bands to characterise aerosol properties and then extrapolating the correction into the visible. This approach, developed for MODIS and adapted for OLCI, works well in clear water.

Coastal and inland waters break the assumption. Turbid water reflects significantly in the near-infrared, so the aerosol-characterisation step picks up water signal and overcorrects. Several correction schemes address this: the POLYMER algorithm uses spectral matching across the full visible-NIR range; the C2RCC (Case-2 Regional CoastColour) neural network, which is the standard processor for OLCI coastal products in the Copernicus system, was trained on a large dataset of in-situ measurements specifically to handle optically complex water. Neither approach is perfect. In very turbid estuaries or under absorbing aerosol plumes such as dust or smoke, retrievals can still fail or carry large uncertainty.

The adjacency effect and why small estuaries are hard

Near coastlines, bright land pixels scatter light into the field of view of adjacent water pixels. This adjacency effect inflates the apparent reflectance of water near shore, mimicking or masking the water-leaving signal. Correction methods exist, but they require accurate knowledge of the point spread function of the sensor and the reflectance of the surrounding land, neither of which is trivially known. The practical consequence is that a buffer zone of at least one to three pixels adjacent to land is often unreliable, which at OLCI's 300 m resolution means roughly 300 to 900 m of coastal water may be compromised.

For narrow estuaries, tidal creeks and lagoons, Landsat OLI at 30 m is the more practical sensor despite its weaker radiometry. The trade-off is explicit: better spatial detail, noisier water-leaving signal, less frequent revisit. Sentinel-2 MSI is sometimes used for coastal water quality despite having no dedicated ocean colour bands, but its atmospheric correction over water is less mature and its signal-to-noise ratio in the blue bands is lower than OLCI. Buyers should be clear about which constraint matters most for their water body before choosing a sensor.

Bio-optical algorithms: from radiance to concentration

The OC4 algorithm, developed by NASA for SeaWiFS and adapted for MODIS and VIIRS, estimates chlorophyll-a from the ratio of reflectance in the blue bands to reflectance in the green. It is an empirical relationship fitted to global in-situ data. It performs well in open ocean Case-1 waters where phytoplankton dominate the optical signal. In coastal Case-2 waters, where sediment and CDOM decorrelate the blue-green ratio from chlorophyll, OC4 can be substantially wrong, sometimes by a factor of two to five.

Neural-network approaches such as C2RCC invert the full spectral shape simultaneously, estimating chlorophyll-a, TSM and CDOM absorption together. They are better suited to optically complex water but are sensitive to conditions outside their training data. Hybrid schemes that switch between empirical and semi-analytical algorithms depending on water type are increasingly common in operational systems. The Copernicus Marine Service distributes OLCI-based Level-3 products at 300 m globally, with daily and monthly composites that fill cloud gaps through temporal averaging. Cloud cover remains the dominant source of data loss in any optical retrieval: a persistent overcast can leave a coastal zone unobserved for days or weeks.

Operational uses and honest expectations

Routine ocean colour monitoring supports harmful algal bloom (HAB) detection, water intake quality management for desalination plants, fisheries productivity assessment, coral reef stress monitoring, and compliance reporting for coastal discharge permits. The Copernicus Marine Service provides operational chlorophyll and TSM products for European seas and globally, with latency typically of one to two days for near-real-time products and weekly for reprocessed datasets.

What satellite retrieval cannot do is replace in-situ measurement for regulatory compliance at specific discharge points, confirm species identity in an algal bloom (satellite colour indicates biomass, not taxonomy), or see through cloud. A bloom that develops and dissipates within a cloudy period may be missed entirely. For applications where those gaps matter, satellite data works best as a wide-area screening layer that directs field sampling rather than a standalone measurement.

Satellize runs ocean colour analytics on open Sentinel-3 and VIIRS data streams, applying C2RCC and OC4 retrievals with client-specific validation against in-situ reference points where available. For coastal zones too narrow for OLCI, Landsat OLI can be substituted with appropriate caveats on retrieval uncertainty.

Resolution, revisit and what to ask before commissioning a programme

The central question for any coastal water quality programme is spatial scale versus temporal frequency. OLCI gives daily coverage at 300 m. Landsat gives 30 m coverage every 8 days. No current open-access sensor gives both. Commercial very-high-resolution optical satellites can image water at sub-5 m resolution, but their ocean colour atmospheric correction pipelines are less mature and their revisit at any given point is irregular unless tasked.

Before specifying a sensor, it is worth asking: how large is the water body of interest, how fast do the relevant parameters change, and what cloud climatology applies to the region? A lagoon 500 m wide in a persistently cloudy tropical climate is a very different problem from a 50 km coastal embayment in the Mediterranean. The answer shapes not just sensor choice but compositing strategy, gap-filling methods and the confidence interval that can honestly be placed on any delivered product.

Typical figures

Spatial resolution (OLCI)	300 m
Spatial resolution (MODIS / VIIRS ocean colour bands)	1 km (MODIS), 750 m (VIIRS)
Spatial resolution (Landsat OLI coastal/aerosol band)	30 m
Revisit (Sentinel-3A + 3B combined)	Approximately every 1–2 days globally
Revisit (Landsat 8 + 9 combined)	8 days at any given point
Key spectral bands for ocean colour	400–750 nm visible (OLCI: 21 bands); coastal aerosol ~443 nm, chlorophyll fluorescence ~681 nm
Typical product latency (near-real-time)	1–2 days (Copernicus Marine Service operational products)
Archive depth	MODIS Aqua from 2002; Sentinel-3A OLCI from 2016; Landsat 8 from 2013
Minimum detectable chlorophyll-a (open ocean)	Approximately 0.01–0.05 mg m⁻³ (OLCI, clear water); uncertainty rises sharply in turbid coastal water
Cloud impact	Optical retrieval fails under cloud; persistent overcast can cause gaps of days to weeks in any given coastal area

Analytics Satellize can run

Chlorophyll-a concentration maps	OC4 empirical band-ratio algorithm (open ocean) or C2RCC neural-network inversion (coastal); applied to OLCI or MODIS L1 radiance	Georeferenced GeoTIFF or NetCDF layer, daily or composited to weekly; uncertainty flag layer included
Total suspended matter (turbidity) maps	C2RCC simultaneous inversion of TSM and CDOM alongside chlorophyll-a from OLCI spectral shape	GIS layer with per-pixel quality flags; time-series CSV for defined monitoring zones
Harmful algal bloom screening alerts	Threshold exceedance on chlorophyll-a and fluorescence line height (FLH) derived from OLCI band 10 (681 nm) relative to baseline	Near-real-time alert (email or API push) when bloom index exceeds client-defined threshold in a defined coastal polygon
CDOM absorption coefficient maps	Semi-analytical inversion (C2RCC or QAA-class algorithm) separating CDOM absorption from phytoplankton and sediment contributions	Monthly GeoTIFF stack; trend report showing seasonal and interannual change
Long-term water quality trend analysis	Time-series regression on MODIS Aqua and Sentinel-3 OLCI archive; seasonal decomposition to separate trend from natural variability	PDF analytical report with station-level trend charts and anomaly maps; underlying data in CSV
Estuary and small water body retrieval (Landsat OLI)	Atmospheric correction using ACOLITE or LaSRC over water; empirical or semi-analytical retrieval of chlorophyll-a and TSM at 30 m	30 m resolution GeoTIFF per Landsat overpass; explicit retrieval uncertainty note attached to each scene

Sources

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.