Ocean colour and marine primary productivity estimation
Ocean colour radiometry retrieves chlorophyll-a concentration from the spectral ratio of water-leaving radiance, turning subtle blue-to-green shifts into estimates of phytoplankton biomass and marine net primary productivity. Atmospheric correction dominates the challenge: more than 90 % of the signal at the sensor is atmosphere, not ocean.
Sensors
- Sentinel-3 OLCI: 21 spectral bands from 400 to 1020 nm at 300 m spatial resolution, global coverage every roughly 1.4 days with two satellites (3A and 3B). The dedicated water-colour bands at 412, 443, 490, 510, 560 and 665 nm are designed for chlorophyll-a retrieval in both open-ocean and coastal waters. Free and open data.
- MODIS-Aqua: Nine ocean-colour bands at 1 km resolution, daily global coverage. The archive runs from 2002 to present, making it the backbone of long-term productivity climatologies. Instrument sensitivity has degraded modestly with age; cross-calibration against Sentinel-3 is standard practice.
- NASA PACE OCI: Launched February 2024. Hyperspectral ocean colour instrument covering 340 to 890 nm continuously at roughly 5 nm spectral resolution and 1 km spatial resolution. Designed to distinguish phytoplankton functional types beyond what broadband sensors can resolve. Still in early validation phase.
- SeaWiFS (archive): Operated 1997 to 2010; eight bands optimised for open-ocean chlorophyll. Its archive underpins the multi-decadal ocean-colour climate data records maintained by NASA and ESA. Not operationally active, but indispensable for trend analysis spanning more than two decades.
- VIIRS (Suomi NPP / NOAA-20 / NOAA-21): Ocean colour bands at 750 m resolution with daily global coverage. Used operationally by NOAA for near-real-time chlorophyll products. Provides continuity as MODIS-Aqua ages out.
What the colour of the sea is actually measuring
Phytoplankton contain chlorophyll-a, a pigment that absorbs strongly in the blue (around 443 nm) and red (around 665 nm) and reflects in the green (around 550 nm). Clean ocean water absorbs red and transmits blue. As phytoplankton concentration rises, blue absorption increases and the water-leaving radiance ratio of blue to green falls. That ratio, typically expressed as Rrs(443)/Rrs(555) or similar band combinations, is the core observable behind every global chlorophyll-a product.
The relationship between that ratio and actual chlorophyll-a concentration is empirical. The widely used OC3M and OC4 algorithms, developed from in-situ match-up datasets, fit a fourth-order polynomial to the log of the band ratio. In clear, open-ocean waters (Case 1 waters, where phytoplankton and their co-varying substances dominate optical properties) these algorithms perform well, with uncertainties typically cited at 30 to 35 % in the global open ocean. That is not a rounding error; it reflects genuine natural variability in the pigment-to-biomass relationship.
Case 2 waters: where the simple ratio breaks down
Coastal and inland waters are optically complex. Suspended sediment, coloured dissolved organic matter (CDOM) from river runoff, and bottom reflectance in shallow areas all alter water-leaving radiance independently of phytoplankton. These are Case 2 waters, and the blue-green ratio approach can produce errors of a factor of two or worse in them.
Retrieving chlorophyll-a in Case 2 waters requires either semi-analytical algorithms (such as the Quasi-Analytical Algorithm, QAA, or GIOP) that decompose the total absorption and backscattering signal into constituent parts, or machine-learning approaches trained on regional in-situ datasets. Sentinel-3 OLCI's 300 m resolution and its red-edge band at 709 nm give it a meaningful advantage in coastal zones over kilometre-scale sensors. Even so, results in highly turbid estuaries carry substantial uncertainty and should always be validated against in-situ measurements before operational use.
The atmosphere owns 90 % of the signal. Getting it back matters.
At a sun-synchronous orbit altitude, roughly 90 to 95 % of the total radiance reaching an ocean-colour sensor comes from the atmosphere, principally Rayleigh scattering and aerosol scattering. The water-leaving radiance that carries the biological information is the residual. Atmospheric correction is therefore not a preprocessing step; it is the central technical problem of ocean colour remote sensing.
Standard approaches estimate aerosol optical properties from near-infrared bands where water-leaving radiance is assumed negligible (the 'black pixel' assumption) and then extrapolate the aerosol correction into the visible. The assumption fails in turbid coastal waters, where NIR reflectance from suspended sediment is non-zero, requiring iterative or alternative correction schemes. Cloud contamination is a harder limit: a cloudy pixel is simply lost. In persistently overcast regions such as the Southern Ocean or the Bay of Bengal during monsoon season, cloud cover can eliminate usable observations for weeks at a time.
From chlorophyll to net primary productivity
Chlorophyll-a concentration is a proxy for phytoplankton biomass, but biomass alone does not tell you how much carbon the ocean is fixing. Net primary productivity (NPP) models add two further inputs: photosynthetically active radiation (PAR, the 400 to 700 nm solar flux reaching the surface) and sea-surface temperature (SST), which modulates phytoplankton growth rates.
The most widely used NPP model classes are the Vertically Generalised Production Model (VGPM, Behrenfeld and Falkowski 1997) and its descendants, including the Carbon-based Productivity Model (CbPM) and CAFE. Each differs in how it represents the depth structure of the phytoplankton community and the light-saturation response. Published global NPP estimates from these models range from roughly 40 to 60 petagrams of carbon per year, with the spread reflecting both model structure and uncertainty in the satellite inputs. SST comes from MODIS or VIIRS thermal bands; PAR is derived from the same ocean-colour sensors or from geostationary meteorological satellites.
The practical output for a fisheries or carbon-accounting client is a gridded NPP map, typically at monthly or 8-day composites to reduce cloud gaps, expressed in milligrams of carbon per square metre per day.
What the data honestly cannot do
Ocean colour radiometry measures the surface optical layer, typically the top optical depth, which in clear water can be 20 to 30 metres but collapses to a few metres in turbid conditions. Subsurface chlorophyll maxima, common in stratified tropical waters, are invisible to passive sensors. Satellite-derived chlorophyll is a column-integrated proxy, not a depth profile.
Spatial resolution is a genuine constraint for coastal applications. MODIS at 1 km and even OLCI at 300 m miss the fine-scale fronts and upwelling filaments that drive local productivity hotspots. High-resolution multispectral satellites such as Sentinel-2 MSI (10 to 20 m) have the spatial detail but lack the radiometric calibration and the dedicated atmospheric correction heritage of purpose-built ocean-colour sensors; their chlorophyll retrievals in Case 2 waters are active research, not mature operational products.
Satellize runs chlorophyll and NPP analytics on Sentinel-3 OLCI and MODIS-Aqua data streams, applying QAA-family semi-analytical algorithms for coastal zones and standard band-ratio products for open-ocean clients. The Tonga crop-estimation programme gave us a working pattern for combining multi-source satellite analytics with in-country ground truth, a pattern that transfers directly to fisheries productivity assessments in island-nation exclusive economic zones.
Putting it to work: fisheries, carbon and policy
The most direct operational use of chlorophyll-a and NPP maps is fisheries stock assessment support. Pelagic fish such as skipjack tuna aggregate at productivity fronts; NPP anomaly maps relative to a climatological baseline can identify years when recruitment conditions are unusually poor or strong. This is not a fish-finding service; it is an environmental context layer that improves the interpretation of catch and survey data.
For blue carbon accounting and marine spatial planning, time-series NPP products allow governments to quantify interannual variability in their EEZ's biological carbon pump, relevant to national ocean carbon inventories under UNFCCC reporting frameworks. The archive depth matters here: MODIS-Aqua back to 2002, combined with SeaWiFS back to 1997, gives a 27-year record against which current conditions can be assessed. That length of record is long enough to detect multi-decadal trends in some ocean basins, though attribution of those trends to climate forcing versus natural variability requires careful statistical treatment.
Typical figures
| Typical spatial resolution (open ocean) | 1 km (MODIS-Aqua, VIIRS); 300 m (Sentinel-3 OLCI); 1 km (PACE OCI) |
| Revisit / temporal resolution | Daily global coverage (MODIS, VIIRS, PACE); ~1.4 days with Sentinel-3A+3B combined |
| Standard compositing period (cloud mitigation) | 8-day or monthly composites; daily products available but cloud-gap-filled with caveats |
| Key spectral bands for chlorophyll retrieval | 412, 443, 490, 510, 560, 665 nm (plus 709 nm red-edge on OLCI for Case 2 waters) |
| Chlorophyll-a detection range | ~0.01 to >50 mg m⁻³; algorithm accuracy degrades at extremes and in Case 2 waters |
| Typical chlorophyll-a retrieval uncertainty (Case 1 open ocean) | 30 to 35 % (empirical algorithms); higher in Case 2 coastal and turbid waters |
| NPP output units | mg C m⁻² day⁻¹, gridded; global totals typically 40 to 60 Pg C yr⁻¹ across model variants |
| Archive depth | SeaWiFS 1997–2010; MODIS-Aqua 2002–present; Sentinel-3 OLCI 2016–present |
| Near-real-time latency | MODIS and VIIRS Level-2 products available within ~3 hours of overpass via NASA OBPG |
| Delivery formats | NetCDF-4 (standard); GeoTIFF on request; WMS/WCS via Copernicus Marine Service (CMEMS) |
Analytics Satellize can run
| Chlorophyll-a concentration map (Case 1 waters) | OC3M / OC4 empirical band-ratio algorithm applied to atmospherically corrected Level-2 Rrs | GeoTIFF or NetCDF layer, 8-day or monthly composite, with per-pixel quality flags |
| Chlorophyll-a concentration map (coastal / Case 2 waters) | Quasi-Analytical Algorithm (QAA) or GIOP semi-analytical inversion of Sentinel-3 OLCI Rrs | GeoTIFF layer with uncertainty band; flagged pixels where sediment or CDOM confound retrieval |
| Net primary productivity estimate | VGPM or CbPM model combining satellite chlorophyll, SST (MODIS/VIIRS thermal) and PAR | Gridded NPP map (mg C m⁻² day⁻¹), monthly time series, EEZ or basin aggregate table |
| Chlorophyll anomaly and trend report | Z-score anomaly relative to MODIS-Aqua or Sentinel-3 climatological baseline (2003–present) | PDF report with time-series plots and spatial anomaly maps; annual or quarterly cadence |
| Productivity front detection | Gradient magnitude of chlorophyll-a field; front persistence scoring across compositing periods | Vector shapefile of active frontal zones; optional overlay with SST front layer |
| EEZ productivity baseline assessment | Multi-year NPP climatology from SeaWiFS/MODIS/Sentinel-3 merged record; interannual variability quantified | Structured data report for fisheries management or national ocean carbon inventory; GIS-ready outputs |
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.