MERCATOR OCEAN JOURNA: 
SEPTEMBER 2021 
1.2.3 Chlorophyll Algorithm improvements 
Algorithms improvements for Chlorophyll retrieval were 
carried out based on optical characteristics of the basin and 
round-robin procedures. It produced blended Chl-a maps 
applying appropriate algorithms across the open ocean and 
coastal waters depending on the occurring water types. The 
same regional bio-optical algorithms used in NRT production 
are applied to generate REP Chlorophyll in all basins with 
the sole exception of the Baltic Sea (Table 2). 
Within the Copernicus-GlobColour processor, the 
Chlorophyll-a (CHL) multi-sensor daily product merged 
Chlorophyll-a values are recomputed using an equivalent 
scheme for each sensor (Garnesson et al., 2019). For 
oligotrophic waters, the product relied on the Cl algorithm 
‚Hu et al., 2012). For mesotrophic and coastal waters, it 
-elied on the 0C5 algorithm (Gohin et al., 2002) tuned for 
each sensor. The blended 0C5 and CI was obtained using 
the same approach as NASA with a transition between 
concentration from 0.15 to 0.2 mg/m* to ensure a smooth 
merging. 
ın the Arctic and Atlantic Seas the regional Chlorophyll 
algorithm is OC5CCI, a variation of 0C5 (Gohin et al., 2002), 
Jeveloped by IFREMER and PML. To this end, an OC5CCI 
look-up table was generated specifically for application 
over OC-CCI daily-merged remote sensing reflectances. 
The resulting OC5CCI algorithm was tested and selected 
after a calibration exercise and sensibility analysis of the 
existing algorithms (0C3, 0C4, OCI, OC5CI, 0C5, OC5CCI) that 
included a round robin quantitative performance 
assessment against in situ data. 
In the Mediterranean Sea, the blended Chlorophyll product 
are based on two algorithms: the MedOC4, an updated version 
of the regionally parameterized Maximum Band Ratio (Volpe 
at al., 2007, 2019) for Open Ocean waters (Case |) and the 
ADOC4 algorithm (D’Alimonte and Zibordi, 2003) for 
optically complex waters (Case Il domain). 
Since 2020, the determination of the water type accounts 
specifically for waters with high Chlorophyll concentration 
due to phytoplankton blooms (e.g., Gulf of Lions) or mixing 
(e.g., Alborän Sea) that can be erroneously identified as 
Case I| waters. 
In the Black Sea, the retrieval of the Chl concentration is 
based on two different regional algorithms: 
a band-ratio algorithm based on two wavelengths 
{490 and 555 nm) (Zibordi et al., 2015), 
a Multilayer Perceptron (MLP) neural net based on Rrs 
values at three wavelengths (490, 510 and 555 nm) 
that features interpolation capabilities helpful to fit 
data non-linearities. 
The merging scheme (Kajiyama et al., 2018) has been 
designed to use the band-ratio algorithm and the MLP 
neural net in waters exhibiting lower and higher optica 
complexity, respectively. 
ın the Baltic Sea, since 2015, NRT products were based 
only on MODIS-Aqua data, while the Sentinel-3A/OLCI data 
stream was introduced in 2017. Then, the MODIS-Aqua 
data-stream was retired in December 2019. The OLC 
Neural Network Swarm (ONNS) algorithm is used to 
retrieve Chlorophyll concentration. In ONNS, the C2RCC 
atmospherically corrected remote sensing reflectance Is 
classified into 13 optical water types (OWT), then different 
OWT-optimized neural networks (NNs) are deployed, and in 
the end, the results of individual NNs are blended according 
to their OWT fuzzy logic weights (Hieronymi et al., 2017). 
For REP timeseries, an ensemble MLP algorithm 
specifically developed for the Baltic Sea was introduced in 
2021. Similarly to the Black Sea merging scheme (Kajiyama 
et al., 2018), the CHL is retrieved by combining results from 
individual MLPs based on different Rrs spectral subsets, 
weighting their contribution through the corresponding 
novelty index (Brando et al., 2021). This ensemble approach 
substituted the regional recalibration of the 0C4v6 with ir 
situ data by Pitarch et al. (2016). 
1.2.4 Phytoplankton type variables and Primary 
Production 
New datasets on Phytoplankton type variables were 
introduced in the OCTAC catalogue from 2019 for the globa 
ocean and all regional seas. Also, in 2020 a Primary 
Production (PP) dataset was introduced for the global 
Ocean. Phytoplankton Size Classes (PSCs) and 
Phytoplankton Functional Types (PFTs) are expressed as 
Chlorophyll_a concentration (mg m-*). Phytoplankton Size 
Zlasses (PSCs) include Micro-phytoplankton (Micro), Nano- 
phytoplankton (Nano) and Pico-phytoplankton (Pico). 
The PP is retrieved based on Antoine and Morel (1996) 
algorithm, using: 
ocean colour products (merged Chlorophyll-a, 
PAR ([photosynthetically active radiation], diffuse 
attenuation coefficient [Kd]), 
Sea-Surface Temperature (SST) from OSTIA (SST. 
GLO_SST_L4_REP_OBSERVATIONS_010_011), 
mixed layer depth climatology, estimated accordinc 
to the definition from CMEMS (GLOBAL_ANALYSIS_ 
FORECAST PHY 001 024)