The last two decades have been characterized by an increased interest in inverse methods dedicated to the estimation of inherent optical properties (IOPs) of natural surface waters from ocean colour remote sensing observations. The ability to retrieve IOPs from space offers particular benefits of improving the investigations of biogeochemical status of the surface oceanic layer at large basin and global scales over long periods of time. Observation of IOPs from space also allows monitoring of water quality status of coastal and inland waters at unprecedented spatio-temporal scales. In the frame of an EUMETSAT Copernicus study we developed IOP products for Sentinel-3 OLCI over open ocean, coastal, and inland waters. For that purpose, two kinds of semi-analytical approaches (one-step and two step algorithms) with their own assumptions have been tested. After different validation exercises based on synthetic data (error free), in situ data sets collected in open-ocean, coastal and inland waters, and match-up data sets, the best approach has been selected, and implemented to provide the full set of IOPs (with their own uncertainties) from OLCI on Sentinel-3. Level-2 and Level-3 examples of these IOPs products are provided and compared to the current IOPs NASA products.

The Ocean Colour Thematic Center (OCTAC) operates the Ocean Colour component of the Copernicus Marine Environment Monitoring Service (CMEMS), providing worldwide global, pan-European and regional (Atlantic, Arctic, Baltic, Mediterranean and Black Sea) high-quality satellite ocean colour products based on Earth Observation Ocean Colour missions. OC TAC relies on current and legacy Ocean Colour satellite sensors: MERIS, from ESA, Sentinel 3 OLCI from EUMETSAT, SeaWiFS and MODIS from NASA, and VIIRS from NOAA.
Global and regional satellite products are higher level observational combined products, thus providing an added value to standard products delivered by the space agencies. To this aim, observations from multiple missions are processed together to ensure homogenized and inter-calibrated datasets providing higher coverage as compared with those of the single sensors. Regional satellite products provide higher accuracy higher than standard Ocean Colour data available from space ground segments thanks to the regionalization of processing chains that takes into account the bio-optical characteristics of each regional seas for production and data validation.
For the global and each of the regional sea, OCTAC selected the state-of-the-art product algorithm for chlorophyll retrieval on the basis of optical characteristics of the basin and round robin procedure. For the regional seas, daily chlorophyll fields are produced by applying two different algorithms for open ocean and coastal waters. The data are then merged into a single chlorophyll field providing a regional product with an improved accuracy of estimates in coastal waters.
This talk will provide an overview of the CMEMS-OCTAC catalogue, the accuracy of the multi-sensor reflectance and chlorophyll products, and the plans for new products including phytoplankton functional types at regional and global scale, primary production at global scale, assessment of long term chlorophyll trends.

We use MERIS and OLCI data to track the distribution of floating vegetation and plankton blooms in the open ocean using a global detection scheme implemented by ESA’s Grid Processing on Demand project. MCI measures the excess radiance at 709 nm above a baseline value linearly interpolated from bands at 681 and 754 nm. Since the red edge radiance increase of floating marine vegetation has its peak at close to 709 nm, and some intense surface plankton blooms cause a narrow-band radiance peak centred near 709 nm, the MCI index proves a sensitive indicator of both types of event. The scheme uses the global data set at the reduced resolution of 1200 m to give monthly composite global images at 5 km spatial resolution. In the equatorial Atlantic, Caribbean, Gulf of Mexico area, the images clearly show the changing distribution of Sargassum. In other areas of the world, MCI indicates presence of other floating vegetation and surface-bloom-forming phytoplankton, chiefly Trichodesmium. Areas of high MCI are also observed in Antarctic ice. Coral reefs give a strong signal, but these are at known, fixed locations. In the area of the South Atlantic Anomaly, high cosmic ray density at satellite altitudes causes many single point high values of MCI.

Using data from the start to the end of MERIS (June 2002 to April 2012) and from the start of OLCI in October 2016, we show how the distributions of Sargassum and plankton blooms have changed over the years 2002 to 2018. Sargassum changed significantly in 2011, from being predominantly in the Gulf of Mexico/Sargasso Sea (North Atlantic tropical gyre) to covering an area extending through the Caribbean and across the Atlantic to Africa.

Chlorophyll fluorescence is light re-emitted by chlorophyll molecules during return from excited to non-excited states. It is greatly beneficial to studies of marine phytoplankton biomass, physiology, and composition and is required for user applications and services, including by CMEMS.
Quantification of solar-induced phytoplankton Fluorescence has two main advantages in marine bio-geochemistry studies. This is, on one hand, the improvement of the chlorophyll retrieval, which is customarily based on the detection of the chlorophyll absorption signal. On the other hand, additional information on phytoplankton physiological state, biomass and maximum layer depth can be gained through the ratio of the chlorophyll Fluorescence to absorption signal.
Customarily phytoplankton chlorophyll fluorescence is determined from space through a fluorescence line-height algorithm using a band around 680nm as the signal band. These approaches are affected by water and suspended particle scattering and water, chlorophyll and CDOM absorption. That is because the peak around 681nm is not only caused by fluorescence, but it is also an effect of being in a spectral minimum of combined water, CDOM and chlorophyll absorption, while the increase in this band is, despite of fluorescence, caused by scattering at water molecules and suspended matter.
We have developed an algorithm which retrieves a, what we call ‚water leaving fluorescence‘ from fitting a gaussian function upon other terms to top-of-atmosphere reflectance bands between 650 and 750nm and which is specifically designed for OLCI bands 8-12.
In detail the function that is fitted to the measurements is composed of two Gaussian functions, an exponential decay and an offset. There is flexibility about the number of channels that can be included (which is the number of sampling points where the formula is evaluated) together with the number of parameters to be retrieved (possible parameters are chlorophyll fluorescence, chlorophyll absorption, position of the fluorescence peak). The combination of those parameters and the chlorophyll absorption signal from blue-green allows insight into the phytoplankton physiological state, biomass and maximum layer depth.
We will present an application and a validation of the algorithm.

Estimating the uncertainties of water color data products is one of the key issues for the user of such data and, likewise, the algorithms provider. This is particular the case for optically complex waters such as present in coastal zones, estuaries, rivers and lakes. These waters contain a large number of different organisms and substances with different and varying optical properties. Depending on its concentration a constituent can dominate the reflectance spectrum but, at low concentrations, can also be masked by other constituents.

Our approach to determine the inherent optical properties (IOPs) and the concentrations of the water constituents is based on machine learning using artificial neural networks (NNs) and is implemented in the C2RCC processor.

For both missions, S2 and S3, we have developed NN based procedures for the correction of the influence of the atmosphere and the retrieval of IOPs and concentrations. This procedure is in line with the QA4EO / GUM guidelines ("Evaluation of measurement data — Supplement 1 to the Guide to the expression of uncertainty in measurement”).

Since we use radiative transfer simulations for training of the NNs, the simulated data can also be used to determine the uncertainties and train NNs to determine the uncertainties. However, the application of this procedure requires that the measured reflectance spectrum has to be analyzed if it is within the scope of the simulated data. For this first test we use an autoassociative neural network.

In this paper we will demonstrate using S2 data of inland waters and S3 data of coastal waters:
- The impact of the bio-optical and atmosphere model on uncertainties
- Ambiguities and masking effects
- The impact of different optimization procedures and NN architectures using TensorFlow deep learning procedures.

Another effect on uncertainties have the spectral and radiometric properties of the sensor, and, in the case of small scale patchiness of the distribution of water constituents, also the spatial resolution. These impacts will be analysed by comparing S2 and S3 OLCI data.

Finally we will present examples of products and their associated uncertainties of S2 and S3 data for different water types.

The Global Climate Observation System (GCOS) includes primary ocean color (OC) radiometric data among the Essential Climate Variables (ECVs) required to support the work of the United Nations Framework Convention of Climate Change (UNFCCC). Still, their actual application implies uncertainties lower than 5% for data collected in the blue-green spectral bands. These requirements are particularly challenging in costal waters due to the simultaneous presence of non-covarying in-water optically active components (i.e., pigments, colored dissolved organic matter and suspended sediments) and potential contributions from sea-bottom and nearby land.
This work presents recent theoretical quantifications of adjacency effects (AE) in Earth Observation of coastal waters. AE, often neglected by the operational atmospheric correction procedures, may become a major source of uncertainty in satellite data and derived products.
Specific attention is given to AE affecting marine observations by two EO-dedicated satellite sensors of the Copernicus Space component: i) the Ocean and Land Colour Instrument (OLCI) on board Sentinel-3 delivering OC observations of the sea; and ii) the MultiSpectral Imagery (MSI) on board Sentinel-2 delivering high-resolution optical land imagery, but also marine observations up to 20 km offshore.
AE are quantified and analyzed for a comprehensive set of typical mid-latitude coastal environments and for two specific sites. These are: the Aqua Alta Oceanographic Tower (AAOT) validation site located in the Northern Adriatic Sea, included in the Ocean Color component of the Aerosol Robotic Network (AERONET-OC); and the marine region surrounding the Lampedusa Island located in the Southern Mediterranean Sea and considered for hosting a long-term system vicarious calibration (SVC) infrastructure.
The study analyzes the relevance of AE at the satellite sensor with respect to standardized signal-to-noise ratios (SNR). Considerations are also drawn on the impact of AE in satellite radiometric products.
As an example, results indicate that AE are above the noise level in OLCI reduced resolution data up to ~36 km offshore for typical land covers (green and dry vegetation, white sand, bare soil, concrete, snow), except in the case of green vegetation at the red wavelengths. Conversely, OLCI full resolution data are considerably less sensitive to AE because of lower SNR.
Results further indicate a remarkable seasonality on AE. Additionally, even though AE at the satellite sensor are markedly function of the land albedo, AE in derived products may be masked by compensations resulting from the atmospheric correction process.
The analysis at the AAOT site further indicates the need to account for adjacency contributions to increase the accuracy of matchup analysis, while specific investigations for the positioning of an SVC site around the Lampedusa Island indicate significant AE up to ~15 km offshore.