MERCATOR OCEAN JOURNA: 
SEPTEMBER 2021 
Figure 2 and Table 2 summarize the evolution of the OCTAC 
zatalogue to 2021 focused on: 
- development and improvements of the NRT and REP 
mMulti-sensor processing chains, 
inclusion of the Copernicus Sentinel-2 and Sentinel-3 
in the single-sensor and multi-sensor datasets, 
improvements in the algorithm for Chlorophyll 
vetrieval based on optical characteristics of the basin 
and round-robin procedures, 
- development of new datasets on Phytoplankton 
Functional Groups and community structure and on 
Primary Production, 
increase of the number of L4 "gap free” datasets. 
1.2.1 Multi-sensor products 
From 2015 to date, OCTAC members put a great effort to 
develop and further improve NRT and REP multi-sensor 
products (Figure 2, Table 2). For Global NRT and REP 
products, the Copernicus-GlobColour processor used data 
from different sensors including: SeaWiFS, MODIS Aqua, 
MODIS Terra, MERIS, VIIRS NPP VIIRS-JPSS1 OLCI-S3A and 
S3B (Garnesson et al., 2019). 
Several OCTAC products are generated taking the advantage of 
the ESA OC-CCI initiative targeting climate quality consistency 
with a minimal inter-sensor bias to produce consistent long 
term multi-sensors (SeaWIFS, MODIS, MERIS, VIIRS and OLCI 
FE 
L2 SeaWiFS (REP), MERIS 
(REP), MODIS, VIIRS NPP 
and JPSS1, OLCI S3A and 
S3B 
| L1 SeaWiFS, MODIS, ME- 
Global REP RIS VIIRS OLCI 
GlobColour 
Garnesson et al 2019) 
OC-CCI vb 
{OC-CCI, 2020) 
Arctic 
NRT+REP 
L1 SeaWiFS, MODIS, ME- 
RIS, VIIRS, OLCI 
OC-CCI v5 upgraded to 
1 km full resolution 
Atlantic 
NRT+REP 
L1 SeaWiFS, MODIS, ME- 
RIS., VIIRS, OLCI 
NC-CCI vBupgraded to 
1] km full resolution 
Mediter- 
ranean 
NRT+REP 
| L2 SeaWiFS (REP), MERIS 
{REP), MODIS, VIIRS NPP ac 
JPSS1, OLCI S3A and S3B 
L2 SeaWiFS (REP), MERIS 
{REP), MODIS, VIIRS NPP ac 
JPSS1,0LCI S3A and S3B 
L1 OLCI S3A and S3B 
CNR MED+BS Proces- 
sor (Volpe et al 2019) 
Black Sea 
NRT+REP 
CNR MED+BS Proces- 
sor (Volpe et al 2019) 
Baltic NRT 
CNR HZG processor 
. SeaWiFS, MODIS, MERIS, * OC-CCI v4.2 upgraded 
Baltic NRT VIIRS, OLCI L3 merged Rrs | to 1 km full resolution 
Jlobal ocean colour time-series (Sathyendranath et al., 2017; 
Sathyendranath et al., 2019). The last version of OC-CCI time- 
series is ingested by OCTAC and converted into CMEMS OC 
format to generate the global OC REP at 4 km resolution. In the 
REP processing chains, the OC-CCI processor is used to 
generate consistent timeseries of reflectances of the Arctic, 
Atlantic and Baltic Seas at 1 km. The same processor is also 
used in NRT for the Arctic, and Atlantic ocean from 2018. 
Since 2020, in Mediterranean and Black Seas, the REP 
processing chain became identical with the CNR NRT 
multi-sensor processor (Volpe et al., 2019) and thus is no 
ıonger based on the OC-CCI L3. Both the NRT and REP 
processing chains involve the pre-processing of L2 data 
from space sensors with different wavelengths that are 
merged over a common set of wavelengths corresponding 
to the SeaWIFS bandset (Volpe et al., 2019). 
The use of multiple sensors permitted to significantly 
increase the spatial coverage of the daily observations. For 
instance,Figure 3 shows the effect of the merging of two 
sensors (VIIRS and MODIS Aqua) and the successive 
introduction of OLCI in 2019 and then of NOAA VIIIRS 20 in 
2020. The number of clear-sky pixels for the Multi product 
is larger by 20 - 40 % than products from a single-sensor. 
Alas, the incremental effect was of -10% for the thirc 
sensor and 4% when the fourth sensor was added. 
DT 
ET] 
Zlended (Garnesson et al., 2019) 0C5 
‘Gohin et al., 2002) and CI (Hu et al., 
20171 
Monthly average 
Advanced Optimal 
Interpolation (variant of 
Saulquin et al., 2010) 
Monthly average 
DINEOF 
0C3, 0C4, 0C5 and Cl, depending on 
pixel water type (0C-CCI 2014) 
OC5CI developed by PML: 
Case 1: CI (Hu et al., 2012) 
Case 2: 0C5 (Gohin et al., 2002) 
OC5CI developed by PML: 
Case 1: CI (Hu et al., 2012) 
Case 2: 0C5 (Gohin et al., 2002) 
Blend of Case1 (MedOC: Volpe et al., 
2007, 2019) and Case 2 (Ad4: Berthor. 
&. Zibordi, 2004). 
Monthlv average 
Monthlvy average 
Variant of DINEOF (Volpe 
et al., 2018) & Monthly 
means 
— => LT 
Merging of BS_0C2 and MLP (Kajiyama HE EI E MEEE 
et al., 2018) ea Y 
Only Monthly means 
" OLCI Neural Network Swarm (Hierony- 
mi et al., 2015) 
MLP ensemble (Brando et al., 2021) 
Only Monthlv means 
[able 2: Overview of processing chains for alobal ocean and regional seas in 202°