<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE article PUBLIC "-//NLM//DTD Journal Publishing with OASIS Tables v3.0 20080202//EN" "https://jats.nlm.nih.gov/nlm-dtd/publishing/3.0/journalpub-oasis3.dtd">
<article xmlns:xlink="http://www.w3.org/1999/xlink" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:oasis="http://docs.oasis-open.org/ns/oasis-exchange/table" xml:lang="en" dtd-version="3.0" article-type="research-article"><?xmltex \bartext{Research article}?>
  <front>
    <journal-meta><journal-id journal-id-type="publisher">OS</journal-id><journal-title-group>
    <journal-title>Ocean Science</journal-title>
    <abbrev-journal-title abbrev-type="publisher">OS</abbrev-journal-title><abbrev-journal-title abbrev-type="nlm-ta">Ocean Sci.</abbrev-journal-title>
  </journal-title-group><issn pub-type="epub">1812-0792</issn><publisher>
    <publisher-name>Copernicus Publications</publisher-name>
    <publisher-loc>Göttingen, Germany</publisher-loc>
  </publisher></journal-meta>
    <article-meta>
      <article-id pub-id-type="doi">10.5194/os-20-217-2024</article-id><title-group><article-title>Linking satellites to genes with machine learning to estimate phytoplankton community structure from space</article-title><alt-title>Linking satellites to genes with machine learning</alt-title>
      </title-group><?xmltex \runningtitle{Linking satellites to genes with machine learning}?><?xmltex \runningauthor{R.~El~Hourany et al.}?>
      <contrib-group>
        <contrib contrib-type="author" corresp="yes" rid="aff1">
          <name><surname>El Hourany</surname><given-names>Roy</given-names></name>
          <email>roy.elhourany@univ-littoral.fr</email>
        <ext-link>https://orcid.org/0000-0002-6454-1645</ext-link></contrib>
        <contrib contrib-type="author" corresp="no" rid="aff2 aff3 aff4">
          <name><surname>Pierella Karlusich</surname><given-names>Juan</given-names></name>
          
        <ext-link>https://orcid.org/0000-0003-1739-4424</ext-link></contrib>
        <contrib contrib-type="author" corresp="no" rid="aff2 aff4 aff5">
          <name><surname>Zinger</surname><given-names>Lucie</given-names></name>
          
        <ext-link>https://orcid.org/0000-0002-3400-5825</ext-link></contrib>
        <contrib contrib-type="author" corresp="no" rid="aff1">
          <name><surname>Loisel</surname><given-names>Hubert</given-names></name>
          
        </contrib>
        <contrib contrib-type="author" corresp="yes" rid="aff6">
          <name><surname>Levy</surname><given-names>Marina</given-names></name>
          <email>marina.levy@locean.ipsl.fr</email>
        <ext-link>https://orcid.org/0000-0003-2961-608X</ext-link></contrib>
        <contrib contrib-type="author" corresp="yes" rid="aff2 aff4">
          <name><surname>Bowler</surname><given-names>Chris</given-names></name>
          <email>cbowler@biologie.ens.fr</email>
        </contrib>
        <aff id="aff1"><label>1</label><institution>Univ. Littoral Côte d’Opale, Univ. Lille, CNRS, IRD, UMR 8187, LOG, Laboratoire d’Océanologie et de Géosciences,  62930 Wimereux, France</institution>
        </aff>
        <aff id="aff2"><label>2</label><institution>Institut de Biologie de l'Ecole Normale Supérieure (IBENS), Ecole Normale Supérieure, CNRS, INSERM, <?xmltex \hack{\break}?>Université PSL, 75005 Paris, France</institution>
        </aff>
        <aff id="aff3"><label>3</label><institution>FAS Division of Science, Harvard University, Cambridge, MA, USA</institution>
        </aff>
        <aff id="aff4"><label>4</label><institution>Research Federation for the study of Global Ocean Systems Ecology and Evolution,  <?xmltex \hack{\break}?>FR2022/Tara GOSEE, 75016 Paris, France</institution>
        </aff>
        <aff id="aff5"><label>5</label><institution>Naturalis Biodiversity Center, 2300 RA Leiden, the Netherlands</institution>
        </aff>
        <aff id="aff6"><label>6</label><institution>Sorbonne Université, LOCEAN-IPSL, Laboratoire d'Océanographie et du Climat; Expérimentations et Approches Numériques, CNRS, IRD, MNHN, 75005 Paris, France</institution>
        </aff>
      </contrib-group>
      <author-notes><corresp id="corr1">Roy El Hourany (roy.elhourany@univ-littoral.fr), Marina Levy (marina.levy@locean.ipsl.fr), and Chris Bowler (cbowler@biologie.ens.fr)</corresp></author-notes><pub-date><day>21</day><month>February</month><year>2024</year></pub-date>
      
      <volume>20</volume>
      <issue>1</issue>
      <fpage>217</fpage><lpage>239</lpage>
      <history>
        <date date-type="received"><day>7</day><month>December</month><year>2022</year></date>
           <date date-type="rev-request"><day>14</day><month>December</month><year>2022</year></date>
           <date date-type="rev-recd"><day>9</day><month>November</month><year>2023</year></date>
           <date date-type="accepted"><day>5</day><month>December</month><year>2023</year></date>
      </history>
      <permissions>
        <copyright-statement>Copyright: © 2024 </copyright-statement>
        <copyright-year>2024</copyright-year>
      <license license-type="open-access"><license-p>This work is licensed under the Creative Commons Attribution 4.0 International License. To view a copy of this licence, visit <ext-link ext-link-type="uri" xlink:href="https://creativecommons.org/licenses/by/4.0/">https://creativecommons.org/licenses/by/4.0/</ext-link></license-p></license></permissions><self-uri xlink:href="https://os.copernicus.org/articles/.html">This article is available from https://os.copernicus.org/articles/.html</self-uri><self-uri xlink:href="https://os.copernicus.org/articles/.pdf">The full text article is available as a PDF file from https://os.copernicus.org/articles/.pdf</self-uri>
      <abstract><title>Abstract</title>

      <p id="d1e169">Ocean color remote sensing has been used for more than 2 decades to estimate primary productivity. Approaches have also been developed to disentangle phytoplankton community structure based on spectral data from space, in particular when combined with in situ measurements of photosynthetic pigments. Here, we propose a new ocean color algorithm to derive the relative cell abundance of seven phytoplankton groups, as well as their contribution to total chlorophyll <inline-formula><mml:math id="M1" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> (Chl <inline-formula><mml:math id="M2" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula>) at the global scale. Our algorithm is based on machine learning and has been trained using remotely sensed parameters (reflectance, backscattering, and attenuation coefficients at different wavelengths, plus temperature and Chl <inline-formula><mml:math id="M3" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula>) combined with an omics-based biomarker developed using <italic>Tara</italic> Oceans data representing a single-copy gene encoding a component of the photosynthetic machinery that is present across all phytoplankton, including both prokaryotes and eukaryotes. It differs from previous methods which rely on diagnostic pigments to derive phytoplankton groups. Our methodology provides robust estimates of the phytoplankton community structure in terms of relative cell abundance and contribution to total Chl <inline-formula><mml:math id="M4" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> concentration. The newly generated datasets yield complementary information about different aspects of phytoplankton that are valuable for assessing the contributions of different phytoplankton groups to primary productivity and inferring community assembly processes. This makes remote sensing observations excellent tools to collect essential biodiversity variables (EBVs) and provide a foundation for developing marine biodiversity forecasts.</p>
  </abstract>
    
<funding-group>
<award-group id="gs1">
<funding-source>Centre National d’Etudes Spatiales</funding-source>
<award-id>Postdoctoral fellowship 2019-2021</award-id>
<award-id>TOSCA BIODIV: Phytoplankton Diversity from Space</award-id>
</award-group>
<award-group id="gs2">
<funding-source>Sorbonne Université</funding-source>
<award-id>ML4BioChange Emergence program 2021-2023</award-id>
</award-group>
<award-group id="gs3">
<funding-source>European Research Council</funding-source>
<award-id>Diatomic (grant agreement No. 835067)</award-id>
</award-group>
<award-group id="gs4">
<funding-source>Agence Nationale de la Recherche</funding-source>
<award-id>OCEANOMICS (ANR-11-BTBR-0008)</award-id>
<award-id>FRANCE GENOMIQUE (ANR-10-INBS-09-08)</award-id>
<award-id>MEMO LIFE (ANR-10-LABX-54)</award-id>
<award-id>PSL Research University (ANR-11-IDEX-0001-02)</award-id>
</award-group>
<award-group id="gs5">
<funding-source>Fonds Français pour l'Environnement Mondial</funding-source>
<award-id>Postdoctoral fellowship</award-id>
</award-group>
</funding-group>
</article-meta>
  </front>
<body>
      

<sec id="Ch1.S1" sec-type="intro">
  <label>1</label><title>Introduction</title>
      <p id="d1e212">The production of organic matter (i.e., productivity) in marine ecosystems relies largely on phytoplankton. These unicellular photosynthetic microorganisms are evolutionarily diverse and exhibit a wide range of cell morphologies, sizes, photosynthetic accessory pigments, elemental requirements, and biogeochemical and trophic functions <xref ref-type="bibr" rid="bib1.bibx56" id="paren.1"/>. They play a key role in regulating ocean biogeochemistry <xref ref-type="bibr" rid="bib1.bibx34" id="paren.2"/> and global climate, partly through the absorption of atmospheric CO<inline-formula><mml:math id="M5" display="inline"><mml:msub><mml:mi/><mml:mn mathvariant="normal">2</mml:mn></mml:msub></mml:math></inline-formula> and export of carbon to the deep ocean <xref ref-type="bibr" rid="bib1.bibx36 bib1.bibx75 bib1.bibx73" id="paren.3"/>.</p>
      <?pagebreak page218?><p id="d1e233"><?xmltex \hack{\newpage}?>In order to investigate the potential impacts of  environmental changes on marine ecosystem functioning <xref ref-type="bibr" rid="bib1.bibx42 bib1.bibx37" id="paren.4"/>, high-resolution, real-time, and global-scale data on phytoplankton community structure are required <xref ref-type="bibr" rid="bib1.bibx54" id="paren.5"/>. However, existing knowledge about the global distribution of phytoplankton communities from in situ observations is highly fragmented, spatially disparate, and temporally punctuated. It is furthermore limited by both the challenges of in situ data collection and the associated costs of measurement techniques, which range from microorganism imaging and flow cytometry to DNA sequencing <xref ref-type="bibr" rid="bib1.bibx38 bib1.bibx45 bib1.bibx70 bib1.bibx64 bib1.bibx58 bib1.bibx63 bib1.bibx25 bib1.bibx56" id="paren.6"/>.</p>
      <p id="d1e246">Ocean color remote sensing offers an interesting alternative to map the global distribution of phytoplankton communities at the sea surface at a high spatiotemporal resolution. Since 1978, ocean color satellites have been used to observe the concentration of the main phytoplankton pigment, chlorophyll <inline-formula><mml:math id="M6" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> (Chl <inline-formula><mml:math id="M7" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula>), considered  a proxy for phytoplankton biomass <xref ref-type="bibr" rid="bib1.bibx51 bib1.bibx68" id="paren.7"/>. Recently, ocean color data have also been used to gain information about phytoplankton communities, such as their size structure and their taxonomic or functional composition. This interest has facilitated the integration of the concept of phytoplankton functional types (PFTs) and taxonomic phytoplankton groups (PGs) into studies exploring various ecological and biogeochemical aspects <xref ref-type="bibr" rid="bib1.bibx48 bib1.bibx41" id="paren.8"/>. PFTs refer to distinct categories associated with biogeochemical processes (e.g., silicifiers, calcifiers) and physiological adaptations to environmental factors (e.g., light, nutrients, turbulence) or to more practical categories identified through specific analytical techniques (e.g., pigment types) (IOCCG report no. 14). On the other hand, phytoplankton groups correspond to taxonomic classes (e.g., diatoms, haptophytes, cyanobacteria). It is important to note that phytoplankton from different taxonomic groups can perform the same ecosystem function; e.g., both diatoms and silicoflagellates can biosilicify but represent different taxonomic groups. Specialized algorithms applied to ocean color data have consequently been developed to detect specific taxa with distinctive optical characteristics (e.g., <xref ref-type="bibr" rid="bib1.bibx15" id="altparen.9"/>;  <xref ref-type="bibr" rid="bib1.bibx43" id="altparen.10"/>), the dominance of PFTs (e.g., <xref ref-type="bibr" rid="bib1.bibx3" id="altparen.11"/>), the relative abundance of PGs and size classes in terms of their contribution to Chl <inline-formula><mml:math id="M8" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> (e.g., <xref ref-type="bibr" rid="bib1.bibx40" id="altparen.12"/>;  <xref ref-type="bibr" rid="bib1.bibx82 bib1.bibx83" id="altparen.13"/>), and recently plankton assemblages and communities (e.g., <xref ref-type="bibr" rid="bib1.bibx47" id="altparen.14"/>).</p>
      <p id="d1e295">The diagnostic pigment analysis method (DPA; <xref ref-type="bibr" rid="bib1.bibx78" id="altparen.15"/>) relies on the association of secondary phytoplankton pigments with different broad taxonomic phytoplankton groups. DPA classification was later refined by <xref ref-type="bibr" rid="bib1.bibx77" id="text.16"/>, who gave different weightings to the diagnostic pigments to retrieve three phytoplankton size classes (PSCs) from total Chl <inline-formula><mml:math id="M9" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula>. The advantage of this method is that phytoplankton pigments can be measured cost-effectively through high-performance liquid chromatography (HPLC). Today, large in situ HPLC datasets are available with broad spatial and temporal coverage. These  HPLC datasets have enabled the development of several DPA-based ocean color algorithms, which has made it possible to evaluate the abundance of different phytoplankton groups and size classes from ocean color satellite data (e.g., <xref ref-type="bibr" rid="bib1.bibx77 bib1.bibx39 bib1.bibx40 bib1.bibx72 bib1.bibx24 bib1.bibx52 bib1.bibx27 bib1.bibx28 bib1.bibx13 bib1.bibx83" id="altparen.17"/>). However, the limitation of the DPA approach is that it is associated with large uncertainties in the classification of phytoplankton due to the presence of some pigments in different phytoplankton taxa and cell size classes, which also vary with acclimation to light, temperature, and nutrient availability <xref ref-type="bibr" rid="bib1.bibx11 bib1.bibx18" id="paren.18"/>.</p>
      <p id="d1e318">In this work, we propose an alternate approach to develop an ocean color algorithm for phytoplankton group detection from in situ metagenomic observations. The approach is ground-truthed on data collected by <italic>Tara</italic> Oceans, which constitutes the most comprehensive and harmonized molecular dataset available on phytoplankton taxonomic community structure on a global scale. More specifically, we used metagenomics reads to extract the global-scale distribution and abundance of the single-copy gene <italic>psbO</italic>, which is present across all phytoplankton groups and  provides an unbiased picture of phytoplankton cell abundances <xref ref-type="bibr" rid="bib1.bibx57" id="paren.19"/>. We used these data, together with satellite-derived optical, physical, and biogeochemical parameters, to train an unsupervised machine learning algorithm able to discern the nonlinear relationship between phytoplankton taxonomic community structure and data derived from satellites. This new algorithm allowed us to derive the spatiotemporal variability of seven PGs between 1997 and 2021. We then compared the performance of this new algorithm with that of two previous DPA-based algorithms  <xref ref-type="bibr" rid="bib1.bibx27 bib1.bibx83" id="paren.20"/>.</p>
</sec>
<sec id="Ch1.S2">
  <label>2</label><title>Materials</title>
      <p id="d1e341">In this section, we present the datasets that were used for training the algorithm and for evaluating the outputs. The input dataset includes the in situ distribution and abundance of phytoplankton groups inferred from metagenomics data from <italic>Tara</italic> Oceans and their associated satellite matchups.  The outputs of the new algorithm are compared to a global dataset of in situ HPLC diagnostic pigments, as well as with estimates from two DPA-based remote sensing algorithms.</p>

      <?xmltex \floatpos{t}?><fig id="Ch1.F1" specific-use="star"><?xmltex \currentcnt{1}?><?xmltex \def\figurename{Figure}?><label>Figure 1</label><caption><p id="d1e349">Global biogeographical patterns of marine phytoplankton relative cell abundance and Chl <inline-formula><mml:math id="M10" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> fraction per group based on <italic>psbO</italic> reads obtained from metagenomes from seawater samples collected during the  <italic>Tara</italic> Oceans expeditions. Two sub-datasets are represented in this figure; the first, <inline-formula><mml:math id="M11" display="inline"><mml:mrow><mml:msub><mml:mi>D</mml:mi><mml:mtext>RCA</mml:mtext></mml:msub></mml:mrow></mml:math></inline-formula>, is composed of <italic>psbO</italic>-derived relative cell abundance, and the second, <inline-formula><mml:math id="M12" display="inline"><mml:mrow><mml:msub><mml:mi>D</mml:mi><mml:mtext>ChlF</mml:mtext></mml:msub></mml:mrow></mml:math></inline-formula>, <italic>psbO</italic>-derived Chl <inline-formula><mml:math id="M13" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> fraction per phytoplankton group.</p></caption>
        <?xmltex \igopts{width=455.244094pt}?><graphic xlink:href="https://os.copernicus.org/articles/20/217/2024/os-20-217-2024-f01.png"/>

      </fig>

<?xmltex \hack{\newpage}?>
<?pagebreak page219?><sec id="Ch1.S2.SS1">
  <label>2.1</label><title>Input dataset</title>
<sec id="Ch1.S2.SS1.SSS1">
  <label>2.1.1</label><?xmltex \opttitle{Metagenomic read abundance of the  {\textit{psbO}} gene}?><title>Metagenomic read abundance of the  <italic>psbO</italic> gene</title>
      <p id="d1e428">The <italic>psbO</italic> gene encodes the manganese-stabilizing protein of around 270 amino acids, which constitutes a core subunit of photosystem II (PSII) and is unique to organisms carrying out oxygenic photosynthesis. The <italic>psbO</italic> gene is a single-copy gene in the vast majority of eukaryotes and prokaryotes. We used <italic>psbO</italic> reads from the metagenomes generated by the <italic>Tara</italic> Oceans expedition as a proxy for phytoplankton relative cell abundance <xref ref-type="bibr" rid="bib1.bibx57" id="paren.21"/>. Among the 210 <italic>Tara</italic> Oceans stations,  145 stations sampled <italic>psbO</italic> reads in different ocean regimes from oligotrophic to eutrophic waters (Chl <inline-formula><mml:math id="M14" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> from 0.01 to 10 mg m<inline-formula><mml:math id="M15" display="inline"><mml:msup><mml:mi/><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">3</mml:mn></mml:mrow></mml:msup></mml:math></inline-formula>, median at 0.3 mg m<inline-formula><mml:math id="M16" display="inline"><mml:msup><mml:mi/><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">3</mml:mn></mml:mrow></mml:msup></mml:math></inline-formula>) from 2009 to 2013. Seawater samples were filtered in order to differentiate five planktonic size fractions (0.22–3, 0.8–5, 5–20, 20–180, 180–2000 <inline-formula><mml:math id="M17" display="inline"><mml:mrow class="unit"><mml:mi mathvariant="normal">µ</mml:mi></mml:mrow></mml:math></inline-formula>m). For the purpose of this study, we pooled the five size fractions into a single aggregated sample.</p>
      <p id="d1e492">The <italic>psbO</italic> data enabled us to taxonomically differentiate seven phytoplankton groups: diatoms, dinoflagellates, green algae, haptophytes, pelagophytes, cryptophytes, and prokaryotes (cyanobacteria) (Fig. <xref ref-type="fig" rid="Ch1.F1"/>). The <italic>psbO</italic> read abundances of these seven groups are expressed as relative phytoplankton cell abundance (%). Phytoplankton that were not assigned to any of these seven groups (unclassified) represented less than 5 % of the total relative cell abundance among all size classes.</p>
      <p id="d1e503">The <italic>psbO</italic> measurements are proxies for relative cell abundance since this protein-encoding gene is generally present as a single copy and is found in all phytoplankton groups. For example, if we take a micro-sized diatom compared to a pico-sized <italic>Synechococcus</italic>, both have one <italic>psbO</italic> gene and are therefore counted as one within the <italic>psbO</italic> quantification. However, we know that a diatom's Chl <inline-formula><mml:math id="M18" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> content is significantly greater than that of <italic>Synechococcus</italic> <xref ref-type="bibr" rid="bib1.bibx2 bib1.bibx35 bib1.bibx21 bib1.bibx6" id="paren.22"/>. This is where the conversion via size-dependent weights is essential in the case of Chl <inline-formula><mml:math id="M19" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> content estimation.</p>
      <p id="d1e540">We should note, however, that filters may retain cells smaller than the nominal pore size because of net clogging, being trapped in fecal pellets, and being present as symbioses and colonies. This has been observed with prokaryotic pico-sized cells such as <italic>Synechococcus</italic> and <italic>Prochloroccocus</italic> being overrepresented in the 180–2000 <inline-formula><mml:math id="M20" display="inline"><mml:mrow class="unit"><mml:mi mathvariant="normal">µ</mml:mi></mml:mrow></mml:math></inline-formula>m size fraction (Fig. <xref ref-type="fig" rid="Ch1.F2"/>). To minimize this impact, we based our size weighting on four size fractions, while excluding the 180–2000 <inline-formula><mml:math id="M21" display="inline"><mml:mrow class="unit"><mml:mi mathvariant="normal">µ</mml:mi></mml:mrow></mml:math></inline-formula>m size range. Chl <inline-formula><mml:math id="M22" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> fraction per group is expressed as follows:
              <disp-formula id="Ch1.E1" content-type="numbered"><label>1</label><mml:math id="M23" display="block"><mml:mrow><mml:mtable class="split" rowspacing="0.2ex" displaystyle="true" columnalign="right left"><mml:mtr><mml:mtd/><mml:mtd><mml:mrow><mml:mtext>Chl </mml:mtext><mml:mi>a</mml:mi><mml:mspace linebreak="nobreak" width="0.125em"/><mml:msub><mml:mtext>fraction</mml:mtext><mml:mtext>PG</mml:mtext></mml:msub></mml:mrow></mml:mtd></mml:mtr><mml:mtr><mml:mtd/><mml:mtd><mml:mrow><mml:mspace width="0.25em" linebreak="nobreak"/><mml:mspace linebreak="nobreak" width="0.25em"/><mml:mspace width="0.25em" linebreak="nobreak"/><mml:mo>=</mml:mo><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mrow><mml:mtext>Chl </mml:mtext><mml:mi>a</mml:mi><mml:mspace width="0.125em" linebreak="nobreak"/><mml:msub><mml:mtext>fraction</mml:mtext><mml:mtext>in situ</mml:mtext></mml:msub><mml:mo>⋅</mml:mo><mml:mfenced close=")" open="("><mml:mrow><mml:msubsup><mml:mo>∑</mml:mo><mml:mrow><mml:mi>s</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow><mml:mn mathvariant="normal">4</mml:mn></mml:msubsup><mml:mo>(</mml:mo><mml:mi>p</mml:mi><mml:mi>s</mml:mi><mml:mi>b</mml:mi><mml:msubsup><mml:mi>O</mml:mi><mml:mtext>s</mml:mtext><mml:mtext>PG</mml:mtext></mml:msubsup><mml:mo>⋅</mml:mo><mml:msub><mml:mtext>size</mml:mtext><mml:mtext>s</mml:mtext></mml:msub><mml:mo>)</mml:mo></mml:mrow></mml:mfenced></mml:mrow><mml:mrow><mml:msubsup><mml:mo>∑</mml:mo><mml:mrow><mml:mi>s</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow><mml:mn mathvariant="normal">4</mml:mn></mml:msubsup><mml:msubsup><mml:mo>∑</mml:mo><mml:mrow><mml:mtext>PG</mml:mtext><mml:mo>=</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow><mml:mn mathvariant="normal">7</mml:mn></mml:msubsup><mml:mo>(</mml:mo><mml:mi>p</mml:mi><mml:mi>s</mml:mi><mml:mi>b</mml:mi><mml:msubsup><mml:mi>O</mml:mi><mml:mtext>s</mml:mtext><mml:mtext>PG</mml:mtext></mml:msubsup><mml:mo>⋅</mml:mo><mml:msub><mml:mtext>size</mml:mtext><mml:mtext>s</mml:mtext></mml:msub><mml:mo>)</mml:mo></mml:mrow></mml:mfrac></mml:mstyle></mml:mrow></mml:mtd></mml:mtr></mml:mtable><mml:mo>,</mml:mo></mml:mrow></mml:math></disp-formula>
            where <inline-formula><mml:math id="M24" display="inline"><mml:mrow><mml:mi>p</mml:mi><mml:mi>s</mml:mi><mml:mi>b</mml:mi><mml:msubsup><mml:mi>O</mml:mi><mml:mtext>s</mml:mtext><mml:mtext>PG</mml:mtext></mml:msubsup></mml:mrow></mml:math></inline-formula> is the <italic>psbO</italic> read abundance for a specific<?pagebreak page220?> phytoplankton group (PG) and for one of the four size fractions (<inline-formula><mml:math id="M25" display="inline"><mml:mi>s</mml:mi></mml:math></inline-formula>), and size corresponds to the mid-value of the corresponding size range, following the protocol in <xref ref-type="bibr" rid="bib1.bibx71" id="text.23"/>, i.e., <inline-formula><mml:math id="M26" display="inline"><mml:mrow><mml:mo>×</mml:mo><mml:mn mathvariant="normal">0.9</mml:mn></mml:mrow></mml:math></inline-formula> for the [0.6–1.2] size class, <inline-formula><mml:math id="M27" display="inline"><mml:mrow><mml:mo>×</mml:mo><mml:mn mathvariant="normal">2.9</mml:mn></mml:mrow></mml:math></inline-formula> for the [0.8–5] size class, <inline-formula><mml:math id="M28" display="inline"><mml:mrow><mml:mo>×</mml:mo><mml:mn mathvariant="normal">12.5</mml:mn></mml:mrow></mml:math></inline-formula> for the [5–20] size class, and <inline-formula><mml:math id="M29" display="inline"><mml:mrow><mml:mo>×</mml:mo><mml:mn mathvariant="normal">100</mml:mn></mml:mrow></mml:math></inline-formula> for the [20–180] size class. Applying Eq. (<xref ref-type="disp-formula" rid="Ch1.E1"/>) pools all size fractions per group while considering the <italic>psbO</italic> read values and the size factors mentioned above.</p>
      <p id="d1e787">There are hence two levels of information derived from the molecular dataset: relative abundance of <italic>psbO</italic> reads as a proxy for relative cell abundance and the fraction of Chl <inline-formula><mml:math id="M30" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> that each group represents. Both types of information have different implications. Chl <inline-formula><mml:math id="M31" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> is often used as a proxy for biomass, which is a relevant parameter for energy and matter fluxes (e.g., food webs, biogeochemical cycles), while cell abundance corresponds to species abundance for unicellular organisms, which is an important measure for inferring community assembly processes.</p>

      <?xmltex \floatpos{t}?><fig id="Ch1.F2" specific-use="star"><?xmltex \currentcnt{2}?><?xmltex \def\figurename{Figure}?><label>Figure 2</label><caption><p id="d1e809">Relative abundance of <italic>psbO</italic> reads as a proxy for phytoplankton group cell abundance observed in each size fraction. The box plots represent the distribution of each group, and each panel shows the different size fractions. The equivalent plots for the <italic>psbO</italic> read values normalized by sequencing depth are displayed in the Supplement in Fig. S1.</p></caption>
            <?xmltex \igopts{width=455.244094pt}?><graphic xlink:href="https://os.copernicus.org/articles/20/217/2024/os-20-217-2024-f02.png"/>

          </fig>

</sec>
<sec id="Ch1.S2.SS1.SSS2">
  <label>2.1.2</label><title>Satellite datasets</title>
      <p id="d1e832">We used ocean color products from the GlobColour project (R2019, full archive reprocessed, 2020) from 1997 to the present day, downloaded from the GlobColour portal. These products were constructed by merging data from various satellite sensors: Sea-viewing Wide Field-of-view Sensor (SeaWiFS), Moderate Resolution Imaging Spectroradiometer (MODIS), Visible Infrared Imaging Radiometer Suite (VIIRS), Medium Resolution Imaging Spectrometer (MERIS), and Ocean and Land Colour Instrument (OLCI). We used 16 GlobColour products: chlorophyll <inline-formula><mml:math id="M32" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> concentration (Chl <inline-formula><mml:math id="M33" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula>, product name: CHL1-AVW), remote sensing reflectances (<inline-formula><mml:math id="M34" display="inline"><mml:mrow><mml:msub><mml:mi>R</mml:mi><mml:mtext>rs</mml:mtext></mml:msub></mml:mrow></mml:math></inline-formula>) at 11 wavelengths (412, 443, 469, 490, 510, 531, 547, 555, 620, 645, and 670 nm), light attenuation coefficient at 490 nm (Kd490), photosynthetically available radiation (PAR), normalized fluorescence light height (NFLH), and particulate backscattering at 443 nm (bbp). These products have daily and 4 km spatiotemporal resolution. In addition, we used the Climate Change Initiative (CCI) sea surface temperature (SST) product at 4 km resolution and daily frequency distributed by the Copernicus Marine Services (CMEMS) portal.</p>
</sec>
</sec>
<sec id="Ch1.S2.SS2">
  <label>2.2</label><title>HPLC datasets</title>
      <p id="d1e869">To compare <italic>psbO</italic>-derived phytoplankton group distributions with more conventional DPA-based products, we compiled a global HPLC dataset regrouping 12 000 HPLC observations from several HPLC datasets between 1997 and 2014 (Fig. <xref ref-type="fig" rid="Ch1.F3"/>, Table S1): MAREDAT, NOMAD, SeaBASS, AESOP-CSIRO, and other oceanographic campaigns including GeP&amp;Co, <italic>Polarstern</italic>, BROKE-West, and SAZ-Sense Voyage <xref ref-type="bibr" rid="bib1.bibx80 bib1.bibx79 bib1.bibx22 bib1.bibx8 bib1.bibx9 bib1.bibx10 bib1.bibx53 bib1.bibx81 bib1.bibx23" id="paren.24"/>. This HPLC dataset was collocated with satellite GlobColour and the CCI SST product matchups. It depicts the abundance of the pigments most widely used to identify major phytoplankton groups: fucoxanthin (Fuco), peridinin (Perid), alloxanthin (Allo), zeaxanthin (Zea), chlorophyll <inline-formula><mml:math id="M35" display="inline"><mml:mi>b</mml:mi></mml:math></inline-formula> (Chl <inline-formula><mml:math id="M36" display="inline"><mml:mi>b</mml:mi></mml:math></inline-formula>), 19<inline-formula><mml:math id="M37" display="inline"><mml:msup><mml:mi/><mml:mo>′</mml:mo></mml:msup></mml:math></inline-formula>-hexanoyloxyfucoxanthin (19HF), and 19<inline-formula><mml:math id="M38" display="inline"><mml:msup><mml:mi/><mml:mo>′</mml:mo></mml:msup></mml:math></inline-formula>-butanoyloxyfucoxanthin (19BF) (Table 1). To estimate the Chl <inline-formula><mml:math id="M39" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> fraction for each phytoplankton group, namely diatoms, dinoflagellates, haptophytes, green algae, cryptophytes, pelgophytes, and prokaryotes, diagnostic pigments were used. The Chl <inline-formula><mml:math id="M40" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> fraction per group is expressed by
            <disp-formula id="Ch1.E2" content-type="numbered"><label>2</label><mml:math id="M41" display="block"><mml:mrow><mml:mtext>Chl </mml:mtext><mml:msub><mml:mi>a</mml:mi><mml:mtext>PG</mml:mtext></mml:msub><mml:mo>=</mml:mo><mml:mtext>Chl </mml:mtext><mml:msub><mml:mi>a</mml:mi><mml:mtext>in situ</mml:mtext></mml:msub><mml:mo>⋅</mml:mo><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mrow><mml:mtext>DP</mml:mtext><mml:mo>⋅</mml:mo><mml:mi mathvariant="italic">α</mml:mi></mml:mrow><mml:mrow><mml:mo>∑</mml:mo><mml:mtext>DP</mml:mtext><mml:mo>⋅</mml:mo><mml:mi mathvariant="italic">α</mml:mi></mml:mrow></mml:mfrac></mml:mstyle><mml:mo>,</mml:mo></mml:mrow></mml:math></disp-formula>
          where <inline-formula><mml:math id="M42" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> is a coefficient associated with a diagnostic pigment (DP) for a specific PG.</p>
      <p id="d1e982">Four sets of coefficients <inline-formula><mml:math id="M43" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> are proposed for a global ocean application and are presented in Table 1 <xref ref-type="bibr" rid="bib1.bibx77 bib1.bibx72 bib1.bibx11 bib1.bibx49" id="paren.25"/>. An examination of the values assigned to the coefficients by these four studies reveals disparities that do not consistently align across all pigments. Notably, while the coefficients for diatoms exhibit similarity across the four sets, differences arise. For instance, in the case of dinoflagellates, only <xref ref-type="bibr" rid="bib1.bibx77" id="text.26"/> and <xref ref-type="bibr" rid="bib1.bibx11" id="text.27"/> show close coefficients associated with Perid, while in the case of haptophytes, <xref ref-type="bibr" rid="bib1.bibx11" id="text.28"/>, <xref ref-type="bibr" rid="bib1.bibx72" id="text.29"/>, and <xref ref-type="bibr" rid="bib1.bibx49" id="text.30"/> estimate close coefficients attributed to 19HF. The discrepancies can be attributed to variations in the datasets utilized for coefficient estimation and differences in the methodologies employed. We averaged the output of the four sets of coefficients to increase the robustness of the results while considering the different outputs of the utilization of these coefficients.</p>
      <p id="d1e1011">Simultaneously, <italic>Tara</italic> Oceans HPLC measurements <xref ref-type="bibr" rid="bib1.bibx55" id="paren.31"/>, which are available for the same stations and sampling time as for <italic>psbO</italic>, were considered to evaluate the correspondence between pigments and <italic>psbO</italic>-derived phytoplankton groups.</p>

<?xmltex \floatpos{t}?><table-wrap id="Ch1.T1" specific-use="star"><?xmltex \currentcnt{1}?><label>Table 1</label><caption><p id="d1e1030">Phytoplankton groups and size classes associated with their diagnostic pigments and coefficients <inline-formula><mml:math id="M44" display="inline"><mml:mi mathvariant="italic">α</mml:mi></mml:math></inline-formula>. Major phytoplankton groups, emphasized in bold, serve to label the outputs resulting from the application of coefficient <inline-formula><mml:math id="M45" display="inline"><mml:mi mathvariant="italic">α</mml:mi></mml:math></inline-formula> to the respective diagnostic pigment.</p></caption><oasis:table frame="topbot"><oasis:tgroup cols="7">
     <oasis:colspec colnum="1" colname="col1" align="left"/>
     <oasis:colspec colnum="2" colname="col2" align="justify" colwidth="2.2cm"/>
     <oasis:colspec colnum="3" colname="col3" align="justify" colwidth="4cm"/>
     <oasis:colspec colnum="4" colname="col4" align="right"/>
     <oasis:colspec colnum="5" colname="col5" align="right"/>
     <oasis:colspec colnum="6" colname="col6" align="right"/>
     <oasis:colspec colnum="7" colname="col7" align="right"/>
     <oasis:thead>
       <oasis:row>
         <oasis:entry colname="col1">Phytoplankton</oasis:entry>
         <oasis:entry colname="col2">Phytoplankton</oasis:entry>
         <oasis:entry colname="col3">Diagnostic pigment (DP)</oasis:entry>
         <oasis:entry colname="col4">Uitz et al.</oasis:entry>
         <oasis:entry colname="col5">Soppa et al.</oasis:entry>
         <oasis:entry colname="col6">Brewin et al.</oasis:entry>
         <oasis:entry colname="col7">Losa et al.</oasis:entry>
       </oasis:row>
       <oasis:row rowsep="1">
         <oasis:entry colname="col1">size class</oasis:entry>
         <oasis:entry colname="col2">group</oasis:entry>
         <oasis:entry colname="col3"/>
         <oasis:entry colname="col4">(2006)</oasis:entry>
         <oasis:entry colname="col5">(2014)</oasis:entry>
         <oasis:entry colname="col6">(2015)</oasis:entry>
         <oasis:entry colname="col7">(2017)</oasis:entry>
       </oasis:row>
     </oasis:thead>
     <oasis:tbody>
       <oasis:row>
         <oasis:entry colname="col1">Micro</oasis:entry>
         <oasis:entry rowsep="1" colname="col2"><bold>Diatoms</bold>, <?xmltex \hack{\hfill\break}?>haptophytes, <?xmltex \hack{\hfill\break}?>chrysophytes, <?xmltex \hack{\hfill\break}?>dinoflagellates</oasis:entry>
         <oasis:entry rowsep="1" colname="col3">Fucoxanthin (Fuco) <?xmltex \hack{\hfill\break}?>(Jeffrey, 1980)</oasis:entry>
         <oasis:entry rowsep="1" colname="col4">1.41</oasis:entry>
         <oasis:entry rowsep="1" colname="col5">1.55</oasis:entry>
         <oasis:entry rowsep="1" colname="col6">1.51</oasis:entry>
         <oasis:entry rowsep="1" colname="col7">1.27</oasis:entry>
       </oasis:row>
       <oasis:row rowsep="1">
         <oasis:entry colname="col1"/>
         <oasis:entry colname="col2"><bold>Dinoflagellates</bold></oasis:entry>
         <oasis:entry colname="col3">Peridinin (Perid) (Jeffrey, 1980; Jeffrey and Hallegraeff, 1987)</oasis:entry>
         <oasis:entry colname="col4">1.41</oasis:entry>
         <oasis:entry colname="col5">0.41</oasis:entry>
         <oasis:entry colname="col6">1.35</oasis:entry>
         <oasis:entry colname="col7">2.43</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">Nano</oasis:entry>
         <oasis:entry rowsep="1" colname="col2"><bold>Haptophytes</bold>, <?xmltex \hack{\hfill\break}?>chrysophytes,<?xmltex \hack{\hfill\break}?>dinoflagellates</oasis:entry>
         <oasis:entry rowsep="1" colname="col3">19<inline-formula><mml:math id="M48" display="inline"><mml:msup><mml:mi/><mml:mo>′</mml:mo></mml:msup></mml:math></inline-formula>-Hexanoyloxyfucoxanthin (19HF) (Wright and Jeffrey, <?xmltex \hack{\hfill\break}?>1987)</oasis:entry>
         <oasis:entry rowsep="1" colname="col4">1.27</oasis:entry>
         <oasis:entry rowsep="1" colname="col5">0.86</oasis:entry>
         <oasis:entry rowsep="1" colname="col6">0.95</oasis:entry>
         <oasis:entry rowsep="1" colname="col7">1.07</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1"/>
         <oasis:entry rowsep="1" colname="col2"><bold>Green algae</bold>, <?xmltex \hack{\hfill\break}?>prasinophytes</oasis:entry>
         <oasis:entry rowsep="1" colname="col3">Chlorophyll <inline-formula><mml:math id="M49" display="inline"><mml:mi>b</mml:mi></mml:math></inline-formula> (Chl <inline-formula><mml:math id="M50" display="inline"><mml:mi>b</mml:mi></mml:math></inline-formula>)<?xmltex \hack{\hfill\break}?>(Vidussi et al., 2001)</oasis:entry>
         <oasis:entry rowsep="1" colname="col4">1.01</oasis:entry>
         <oasis:entry rowsep="1" colname="col5">1.17</oasis:entry>
         <oasis:entry rowsep="1" colname="col6">0.85</oasis:entry>
         <oasis:entry rowsep="1" colname="col7">1.30</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1"/>
         <oasis:entry rowsep="1" colname="col2"><bold>Cryptophytes</bold></oasis:entry>
         <oasis:entry rowsep="1" colname="col3">Alloxanthin (Allo) <?xmltex \hack{\hfill\break}?>(Gieskes and Kraay, 1983)</oasis:entry>
         <oasis:entry rowsep="1" colname="col4">0.6</oasis:entry>
         <oasis:entry rowsep="1" colname="col5">2.39</oasis:entry>
         <oasis:entry rowsep="1" colname="col6">2.71</oasis:entry>
         <oasis:entry rowsep="1" colname="col7">2.06</oasis:entry>
       </oasis:row>
       <oasis:row rowsep="1">
         <oasis:entry colname="col1"/>
         <oasis:entry colname="col2"><bold>Pelagophytes</bold>, <?xmltex \hack{\hfill\break}?>haptophytes</oasis:entry>
         <oasis:entry colname="col3">19<inline-formula><mml:math id="M51" display="inline"><mml:msup><mml:mi/><mml:mo>′</mml:mo></mml:msup></mml:math></inline-formula>-Butanoyloxyfucoxanthin (19BF) (Wright and Jeffrey, <?xmltex \hack{\hfill\break}?>1987)</oasis:entry>
         <oasis:entry colname="col4">0.35</oasis:entry>
         <oasis:entry colname="col5">1.06</oasis:entry>
         <oasis:entry colname="col6">1.27</oasis:entry>
         <oasis:entry colname="col7">–</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">Pico</oasis:entry>
         <oasis:entry colname="col2"><bold>Prokaryotes</bold><?xmltex \hack{\hfill\break}?>(<bold>cyanobacteria</bold>), <?xmltex \hack{\hfill\break}?>green algae,<?xmltex \hack{\hfill\break}?>prasinophytes,<?xmltex \hack{\hfill\break}?>chrysophytes, <?xmltex \hack{\hfill\break}?>euglenophytes</oasis:entry>
         <oasis:entry colname="col3">Zeaxanthin (Dandonneau et al., 2004; Guillard et al., 1985)</oasis:entry>
         <oasis:entry colname="col4">0.86</oasis:entry>
         <oasis:entry colname="col5">2.04</oasis:entry>
         <oasis:entry colname="col6">0.93</oasis:entry>
         <oasis:entry colname="col7">2.36</oasis:entry>
       </oasis:row>
     </oasis:tbody>
   </oasis:tgroup></oasis:table><table-wrap-foot><p id="d1e1047">Coefficients are based on the global HPLC dataset corresponding to the sum of the weighted diagnostic pigments to the total Chl <inline-formula><mml:math id="M46" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula>; Chl <inline-formula><mml:math id="M47" display="inline"><mml:mrow><mml:mi>a</mml:mi><mml:mo>=</mml:mo><mml:mo>∑</mml:mo><mml:mi mathvariant="italic">α</mml:mi></mml:mrow></mml:math></inline-formula> DP.</p></table-wrap-foot><?xmltex \gdef\@currentlabel{1}?></table-wrap>

      <?xmltex \floatpos{t}?><fig id="Ch1.F3" specific-use="star"><?xmltex \currentcnt{3}?><?xmltex \def\figurename{Figure}?><label>Figure 3</label><caption><p id="d1e1400"><bold>(a)</bold> Geographical location of the global HPLC dataset stations regrouping observations from 1997 and 2014. Panel <bold>(b)</bold> represents a comparison between in situ HPLC Chl <inline-formula><mml:math id="M52" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> measurement and its matchup using the GlobColour Chl <inline-formula><mml:math id="M53" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> product.</p></caption>
          <?xmltex \igopts{width=398.338583pt}?><graphic xlink:href="https://os.copernicus.org/articles/20/217/2024/os-20-217-2024-f03.png"/>

        </fig>

</sec>
<sec id="Ch1.S2.SS3">
  <label>2.3</label><title>Phytoplankton group satellite products</title>
      <p id="d1e1436">In order to compare the outputs of our method to those of existing DPA-based remote sensing algorithms, we used two previously published algorithms.</p><?xmltex \hack{\newpage}?>
<?pagebreak page221?><sec id="Ch1.S2.SS3.SSS1">
  <label>2.3.1</label><?xmltex \opttitle{CMEMS  phytoplankton Chl $a$ fraction}?><title>CMEMS  phytoplankton Chl <inline-formula><mml:math id="M54" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> fraction</title>
      <p id="d1e1455">This daily GlobColour product contains the concentration of each phytoplankton group (expressed in terms of Chl <inline-formula><mml:math id="M55" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> concentration fraction) based on the <xref ref-type="bibr" rid="bib1.bibx83" id="text.32"/> algorithm from 2002 to the present at the global scale and with a 4 km resolution. This algorithm estimates the Chl <inline-formula><mml:math id="M56" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> concentration of diatoms, dinoflagellates, haptophytes, green algae, and prokaryotes. The algorithm was implemented with HPLC-based phytoplankton groups using the DPA approach (Soppa et al., 2014) merged to ocean color (OC) <inline-formula><mml:math id="M57" display="inline"><mml:mrow><mml:msub><mml:mi>R</mml:mi><mml:mtext>rs</mml:mtext></mml:msub></mml:mrow></mml:math></inline-formula> products (412, 443, 490, 510, 531, 547, 555, 670, and 678 nm) and accounting for the influence of SST<?pagebreak page222?> on the derived PG quantities (product number: OCEANCOLOUR_GLO_BGC_L3_MY_009_103).</p>
</sec>
<sec id="Ch1.S2.SS3.SSS2">
  <label>2.3.2</label><title>SOM phytoplankton pigments</title>
      <p id="d1e1494">SOM-Pigments <xref ref-type="bibr" rid="bib1.bibx27" id="paren.33"/> is a machine-learning-based algorithm that allows the estimation of phytoplankton pigment concentrations in oceanic waters from satellite ocean color data (chlc<inline-formula><mml:math id="M58" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula>, <inline-formula><mml:math id="M59" display="inline"><mml:mrow><mml:msub><mml:mi>R</mml:mi><mml:mtext>rs</mml:mtext></mml:msub></mml:mrow></mml:math></inline-formula> at four wavelengths: 412, 443, 490, and 555 nm) and SST. This algorithm is based on the use of self-organizing maps (SOMs) and an unsupervised neural network, and it was calibrated using the HPLC dataset described above.</p>
      <p id="d1e1518">The SOM-Pigments algorithm applied to GlobColour products allows  estimating the daily concentration of 10 phytoplankton pigments (chlorophyll <inline-formula><mml:math id="M60" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> – Chl <inline-formula><mml:math id="M61" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula>, divinyl-chlorophyll <inline-formula><mml:math id="M62" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> – DVChl <inline-formula><mml:math id="M63" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula>, chlorophyll <inline-formula><mml:math id="M64" display="inline"><mml:mi>b</mml:mi></mml:math></inline-formula> – Chl <inline-formula><mml:math id="M65" display="inline"><mml:mi>b</mml:mi></mml:math></inline-formula>, divinyl-chlorophyll <inline-formula><mml:math id="M66" display="inline"><mml:mi>b</mml:mi></mml:math></inline-formula> – DVChl <inline-formula><mml:math id="M67" display="inline"><mml:mi>b</mml:mi></mml:math></inline-formula>, 19<inline-formula><mml:math id="M68" display="inline"><mml:msup><mml:mi/><mml:mo>′</mml:mo></mml:msup></mml:math></inline-formula>exfucoxanthin – 19HF, 19<inline-formula><mml:math id="M69" display="inline"><mml:msup><mml:mi/><mml:mo>′</mml:mo></mml:msup></mml:math></inline-formula>butfucoxanthin – 19BF, fucoxanthin – Fuco, peridinin – Perid, alloxanthin – Allo, zeaxanthin – Zea) at the global scale from 1997 to the present with a resolution of 4 km. We then used the coefficients in Table 1 of <xref ref-type="bibr" rid="bib1.bibx77" id="text.34"/>, <xref ref-type="bibr" rid="bib1.bibx72" id="text.35"/>, and <xref ref-type="bibr" rid="bib1.bibx11" id="text.36"/> to convert pigments into the Chl <inline-formula><mml:math id="M70" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> concentration of five phytoplankton groups, namely diatoms, dinoflagellates, haptophytes, green algae, and prokaryotes.</p>
</sec>
</sec>
</sec>
<sec id="Ch1.S3">
  <label>3</label><title>Methods</title>
      <p id="d1e1623">The algorithm  to estimate phytoplankton groups from satellite data was built using SOM (Kohonen, 2013) and topology-constrained organization. This allowed us to confirm the nonlinear relationships between phytoplankton group composition and satellite data through topology conservation. Next, we used the ascending hierarchical clustering algorithm to identify the large-scale patterns generated by SOM. This allowed us to emphasize the predominant data structure learned by SOM and to characterize phytoplankton biomes. The steps of the training and operational phase of the SOM methodology are illustrated in flowcharts found in the Supplement (Figs. S2–S4). Finally, to characterize the differences between the DPA- and <italic>psbO</italic>-based approaches, we used  random forest models to highlight the cumulative importance of a pigment composition to estimate a phytoplankton group abundance. In the following section, each methodology and algorithm is explained in detail.</p>
<sec id="Ch1.S3.SS1">
  <label>3.1</label><title>Structure of the training and test databases</title>
      <p id="d1e1636">The initial dataset (<inline-formula><mml:math id="M71" display="inline"><mml:mi>D</mml:mi></mml:math></inline-formula>) consists of the 145 <italic>Tara</italic> Oceans observations of <italic>psbO</italic> relative abundance of the seven defined phytoplankton groups, the Chl <inline-formula><mml:math id="M72" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> fraction per group, and the associated matchups of 21 satellite-derived parameters (Chl <inline-formula><mml:math id="M73" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula>, SST, <inline-formula><mml:math id="M74" display="inline"><mml:mrow><mml:msub><mml:mi>R</mml:mi><mml:mtext>rs</mml:mtext></mml:msub></mml:mrow></mml:math></inline-formula> at 15 wavelengths from 412 to 709 nm, NFLH, Kd at 490 m, PAR, and bbp at 443 nm). The unclassified phytoplankton fraction was also considered, despite negligible values, to ensure coherence of the total phytoplankton pool. To extract the matchup for a given observation, a <inline-formula><mml:math id="M75" display="inline"><mml:mrow><mml:mn mathvariant="normal">3</mml:mn><mml:mo>×</mml:mo><mml:mn mathvariant="normal">3</mml:mn></mml:mrow></mml:math></inline-formula> pixel box was employed, centered around the observation's coordinates on the same day. The average of the non-outlier pixels was computed. If this approach was unproductive due to fewer than four valid pixels within the <inline-formula><mml:math id="M76" display="inline"><mml:mrow><mml:mn mathvariant="normal">3</mml:mn><mml:mo>×</mml:mo><mml:mn mathvariant="normal">3</mml:mn></mml:mrow></mml:math></inline-formula> box or the absence of any pixel, a <inline-formula><mml:math id="M77" display="inline"><mml:mrow><mml:mn mathvariant="normal">3</mml:mn><mml:mo>×</mml:mo><mml:mn mathvariant="normal">3</mml:mn></mml:mrow></mml:math></inline-formula> pixel extraction was performed for the adjacent days (<inline-formula><mml:math id="M78" display="inline"><mml:mrow><mml:mo>+</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:math></inline-formula> and <inline-formula><mml:math id="M79" display="inline"><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:math></inline-formula>) <xref ref-type="bibr" rid="bib1.bibx27 bib1.bibx28" id="paren.37"/>. Following these matchup exercises, we performed a baseline comparison between in situ Chl <inline-formula><mml:math id="M80" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> and satellite-derived Chl <inline-formula><mml:math id="M81" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula>. This comparison is deemed satisfactory, with an average error rate of 33 %.</p>
      <p id="d1e1752">We built two sub-datasets, the first (<inline-formula><mml:math id="M82" display="inline"><mml:mrow><mml:msub><mml:mi>D</mml:mi><mml:mtext>RCA</mml:mtext></mml:msub></mml:mrow></mml:math></inline-formula>) relating <italic>psbO</italic>-derived relative cell abundance of the seven defined phytoplankton groups to the 17 satellite-derived parameters and the second (<inline-formula><mml:math id="M83" display="inline"><mml:mrow><mml:msub><mml:mi>D</mml:mi><mml:mtext>ChlF</mml:mtext></mml:msub></mml:mrow></mml:math></inline-formula>) joining <italic>psbO</italic>-derived Chl <inline-formula><mml:math id="M84" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> fraction per phytoplankton group and the same 17 satellite-derived parameters. We then constructed two algorithms using either <inline-formula><mml:math id="M85" display="inline"><mml:mrow><mml:msub><mml:mi>D</mml:mi><mml:mtext>RCA</mml:mtext></mml:msub></mml:mrow></mml:math></inline-formula> or <inline-formula><mml:math id="M86" display="inline"><mml:mrow><mml:msub><mml:mi>D</mml:mi><mml:mtext>ChlF</mml:mtext></mml:msub></mml:mrow></mml:math></inline-formula>, both based on the same SOM methodology described below. Following the positioning of <italic>Tara<?pagebreak page223?></italic> Oceans stations and the distribution of Chl <inline-formula><mml:math id="M87" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> values within both datasets (Figs. S5 and S6), both algorithms are suitable for case 1 water applications (i.e., open ocean).</p>
      <p id="d1e1823">The rationale behind this is that the phytoplankton community should be treated as a whole; consequently, the variability of each phytoplankton group is dependent on each other in a relative way. <inline-formula><mml:math id="M88" display="inline"><mml:mrow><mml:msub><mml:mi>D</mml:mi><mml:mtext>RCA</mml:mtext></mml:msub></mml:mrow></mml:math></inline-formula> and <inline-formula><mml:math id="M89" display="inline"><mml:mrow><mml:msub><mml:mi>D</mml:mi><mml:mtext>ChlF</mml:mtext></mml:msub></mml:mrow></mml:math></inline-formula> both present missing values (Table 2), most likely due to cloud coverage, coastal proximity, and/or ice presence. In situ <italic>psbO</italic>-based observations also contained missing values due to an absence of certain measurements at a given station. Since the in situ dataset contains a low number of observations (145 stations), every observation is valuable. In order to overcome the several limitations faced with this training dataset, we used the SOM algorithm that can deal with missing values and allow a robust generalization in the case of limited observations <xref ref-type="bibr" rid="bib1.bibx46" id="paren.38"/>. Before applying the SOM, we ensured that all variables, phytoplankton observations, and satellite parameters were weighted alike while normalizing their values by their variance.</p>

<?xmltex \floatpos{t}?><table-wrap id="Ch1.T2" specific-use="star"><?xmltex \currentcnt{2}?><label>Table 2</label><caption><p id="d1e1858">Percentage of missing values within the initial database.</p></caption><oasis:table frame="topbot"><?xmltex \begin{scaleboxenv}{.85}[.85]?><oasis:tgroup cols="10">
     <oasis:colspec colnum="1" colname="col1" align="left"/>
     <oasis:colspec colnum="2" colname="col2" align="left"/>
     <oasis:colspec colnum="3" colname="col3" align="right" colsep="1"/>
     <oasis:colspec colnum="4" colname="col4" align="right"/>
     <oasis:colspec colnum="5" colname="col5" align="right"/>
     <oasis:colspec colnum="6" colname="col6" align="right"/>
     <oasis:colspec colnum="7" colname="col7" align="right"/>
     <oasis:colspec colnum="8" colname="col8" align="right"/>
     <oasis:colspec colnum="9" colname="col9" align="right"/>
     <oasis:colspec colnum="10" colname="col10" align="right"/>
     <oasis:thead>
       <oasis:row rowsep="1">
         <oasis:entry colname="col1"><italic>Tara</italic> Oceans</oasis:entry>
         <oasis:entry namest="col2" nameend="col3" align="center" colsep="1"><italic>psbO</italic></oasis:entry>
         <oasis:entry namest="col4" nameend="col10" align="center">Sat </oasis:entry>
       </oasis:row>
     </oasis:thead>
     <oasis:tbody>
       <oasis:row rowsep="1">
         <oasis:entry colname="col1"><inline-formula><mml:math id="M90" display="inline"><mml:mi>D</mml:mi></mml:math></inline-formula> (145 stations)</oasis:entry>
         <oasis:entry colname="col2">Relative cell abundance</oasis:entry>
         <oasis:entry colname="col3">Chl <inline-formula><mml:math id="M91" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> fraction per group</oasis:entry>
         <oasis:entry colname="col4">Chl <inline-formula><mml:math id="M92" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula></oasis:entry>
         <oasis:entry colname="col5"><inline-formula><mml:math id="M93" display="inline"><mml:mrow><mml:msub><mml:mi>R</mml:mi><mml:mtext>rs</mml:mtext></mml:msub></mml:mrow></mml:math></inline-formula> 412–709 nm</oasis:entry>
         <oasis:entry colname="col6">SST</oasis:entry>
         <oasis:entry colname="col7">bbp443</oasis:entry>
         <oasis:entry colname="col8">Kd490</oasis:entry>
         <oasis:entry colname="col9">NFLH</oasis:entry>
         <oasis:entry colname="col10">PAR</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">Percentage of missing values</oasis:entry>
         <oasis:entry colname="col2">–</oasis:entry>
         <oasis:entry colname="col3">7 %</oasis:entry>
         <oasis:entry colname="col4">18 %</oasis:entry>
         <oasis:entry colname="col5">43 %–53 %</oasis:entry>
         <oasis:entry colname="col6">30 %</oasis:entry>
         <oasis:entry colname="col7">55 %</oasis:entry>
         <oasis:entry colname="col8">53 %</oasis:entry>
         <oasis:entry colname="col9">37 %</oasis:entry>
         <oasis:entry colname="col10">14 %</oasis:entry>
       </oasis:row>
     </oasis:tbody>
   </oasis:tgroup><?xmltex \end{scaleboxenv}?></oasis:table><?xmltex \gdef\@currentlabel{2}?></table-wrap>

</sec>
<sec id="Ch1.S3.SS2">
  <label>3.2</label><?xmltex \opttitle{Self-organizing map applied to \textit{Tara} Oceans \textit{psbO} data}?><title>Self-organizing map applied to <italic>Tara</italic> Oceans <italic>psbO</italic> data</title>
<sec id="Ch1.S3.SS2.SSS1">
  <label>3.2.1</label><title>General concept of SOM</title>
      <p id="d1e2028">The SOM algorithm is utilized for clustering multidimensional databases by assigning them to classes represented by a fixed network of neurons known as the self-organizing map (SOM). The SOM  consists of a rectangular grid of <inline-formula><mml:math id="M94" display="inline"><mml:mrow><mml:mi>p</mml:mi><mml:mo>×</mml:mo><mml:mi>q</mml:mi></mml:mrow></mml:math></inline-formula> neurons and defines a discrete distance between neurons, enabling the partitioning of the dataset. Each cluster is associated with a neuron and represented by a prototype vector. Observations in the dataset are assigned to the nearest neuron based on the Euclidean norm. A key feature of SOM is its ability to provide topological ordering, where close neurons on the map correspond to similar observations in the data space. The estimation of a neuron's vector and the topological order are determined through a minimization process of a cost function that depends on the distance between the neuron and its assigned observation. SOMs have been widely employed to complete missing data, utilizing the truncated distance <xref ref-type="bibr" rid="bib1.bibx33 bib1.bibx17 bib1.bibx66 bib1.bibx60" id="paren.39"/>. The truncated distance is defined as a modification of the standard Euclidean distance between two observations that accounts only for the existing components of the vectors. This modification of the distance measure allows for the comparison of observations with incomplete information by considering only the existing components and effectively handling missing data. The SOM algorithm can then use this truncated distance measure in its learning process to complete missing data and integrate incomplete information, enabling more robust analysis and visualization of the data.</p><?xmltex \hack{\newpage}?>
</sec>
<sec id="Ch1.S3.SS2.SSS2">
  <label>3.2.2</label><title>Training phase</title>
      <p id="d1e2055">Briefly, we first split the <italic>Tara</italic> Oceans <italic>psbO</italic> datasets so as to obtain 80 % of the data to train the SOM and 20 % of the data as a test set, the latter consisting of 30 observations with complete <italic>psbO</italic> information. We did this separately for <inline-formula><mml:math id="M95" display="inline"><mml:mrow><mml:msub><mml:mi>D</mml:mi><mml:mtext>RCA</mml:mtext></mml:msub></mml:mrow></mml:math></inline-formula> and <inline-formula><mml:math id="M96" display="inline"><mml:mrow><mml:msub><mml:mi>D</mml:mi><mml:mtext>ChlF</mml:mtext></mml:msub></mml:mrow></mml:math></inline-formula> sub-datasets so as to generate SOMRCA, which stands for the algorithm specialized in relative cell abundance estimation, and SOMChlF for the algorithm specialized in Chl <inline-formula><mml:math id="M97" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> fraction per phytoplankton group.</p>
      <p id="d1e2097">During the SOM training, different combinations of satellite variables were used to determine the best set of variables to estimate the seven phytoplankton groups in terms of relative cell abundance and Chl <inline-formula><mml:math id="M98" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> fraction. For each combination of variables, we increased the number of neurons from 10 to 1000 neurons, with an interval of 10 neurons, to determine the optimal size of the SOM. For each SOM obtained, we quantified quantization and topographic errors. The quantization error represents the difference between an observation and its closest neuron. This error is monitored during the training procedure until it reaches stability at a minimum value with increasing training epochs. This is where the training should stop to prevent overfitting. The quantization error is expressed as follows:
              <disp-formula id="Ch1.E3" content-type="numbered"><label>3</label><mml:math id="M99" display="block"><mml:mrow><mml:mi>q</mml:mi><mml:mi>e</mml:mi><mml:mo>=</mml:mo><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mn mathvariant="normal">1</mml:mn><mml:mi>n</mml:mi></mml:mfrac></mml:mstyle><mml:munderover><mml:mo movablelimits="false">∑</mml:mo><mml:mrow><mml:mi>i</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow><mml:mi>n</mml:mi></mml:munderover><mml:mfenced open="∥" close="∥"><mml:mrow><mml:msub><mml:mi>x</mml:mi><mml:mi>i</mml:mi></mml:msub><mml:mo>-</mml:mo><mml:msub><mml:mi>w</mml:mi><mml:mrow><mml:mi>c</mml:mi><mml:mi>i</mml:mi></mml:mrow></mml:msub></mml:mrow></mml:mfenced><mml:mo>,</mml:mo></mml:mrow></mml:math></disp-formula>
            where <inline-formula><mml:math id="M100" display="inline"><mml:mrow><mml:msub><mml:mi>x</mml:mi><mml:mi>i</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> is the vector of an input observation <inline-formula><mml:math id="M101" display="inline"><mml:mi>i</mml:mi></mml:math></inline-formula>, <inline-formula><mml:math id="M102" display="inline"><mml:mrow><mml:msub><mml:mi>w</mml:mi><mml:mrow><mml:mi>c</mml:mi><mml:mi>i</mml:mi></mml:mrow></mml:msub></mml:mrow></mml:math></inline-formula> is the vector of the closest neuron <inline-formula><mml:math id="M103" display="inline"><mml:mi>c</mml:mi></mml:math></inline-formula> of a sample <inline-formula><mml:math id="M104" display="inline"><mml:mrow><mml:msub><mml:mi>x</mml:mi><mml:mi>i</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula>, and <inline-formula><mml:math id="M105" display="inline"><mml:mi>n</mml:mi></mml:math></inline-formula> is the number of observations.</p>
      <p id="d1e2214">However, the topographic error is a representation of having, for each observation of the database, distant first and second best-matching neurons and is expressed as follows:
              <disp-formula id="Ch1.E4" content-type="numbered"><label>4</label><mml:math id="M106" display="block"><mml:mrow><mml:mi>t</mml:mi><mml:mi>e</mml:mi><mml:mo>=</mml:mo><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mn mathvariant="normal">1</mml:mn><mml:mi>n</mml:mi></mml:mfrac></mml:mstyle><mml:munderover><mml:mo movablelimits="false">∑</mml:mo><mml:mrow><mml:mi>i</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow><mml:mi>n</mml:mi></mml:munderover><mml:mi>d</mml:mi><mml:mo>(</mml:mo><mml:msub><mml:mi>x</mml:mi><mml:mi>i</mml:mi></mml:msub><mml:mo>)</mml:mo><mml:mo>,</mml:mo></mml:mrow></mml:math></disp-formula>
            where <inline-formula><mml:math id="M107" display="inline"><mml:mrow><mml:mi>d</mml:mi><mml:mo>(</mml:mo><mml:msub><mml:mi>x</mml:mi><mml:mi>i</mml:mi></mml:msub><mml:mo>)</mml:mo><mml:mo>=</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:math></inline-formula> if the first and second closest neurons to <inline-formula><mml:math id="M108" display="inline"><mml:mrow><mml:msub><mml:mi>x</mml:mi><mml:mi>i</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> are not adjacent; otherwise, <inline-formula><mml:math id="M109" display="inline"><mml:mrow><mml:mi>d</mml:mi><mml:mo>(</mml:mo><mml:msub><mml:mi>x</mml:mi><mml:mi>i</mml:mi></mml:msub><mml:mo>)</mml:mo><mml:mo>=</mml:mo><mml:mn mathvariant="normal">0</mml:mn></mml:mrow></mml:math></inline-formula>.</p>
      <p id="d1e2313">Minimizing this quantity is important to ensure the preservation of the topological order within the SOM with an increasing number of interpolated neurons. A leave-one-out cross-validation procedure was performed to assign three performance metrics to help choose the best combination of SOM size and satellite variables: regression coefficient (<inline-formula><mml:math id="M110" display="inline"><mml:mrow><mml:msup><mml:mi>R</mml:mi><mml:mn mathvariant="normal">2</mml:mn></mml:msup></mml:mrow></mml:math></inline-formula>) and root mean squared error (RMSE). One should note that, for SOMChlF, the <inline-formula><mml:math id="M111" display="inline"><mml:mrow><mml:msup><mml:mi>R</mml:mi><mml:mn mathvariant="normal">2</mml:mn></mml:msup></mml:mrow></mml:math></inline-formula> was calculated using log-transformed Chl <inline-formula><mml:math id="M112" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> values, and RMSE was calculated using non-transformed Chl <inline-formula><mml:math id="M113" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> values. At each iteration of the cross-validation procedure, we randomly chose one observation as a test, whereas the other observations served to train the SOM with the given grid size. We calculated the closest neuron to the test observation based on its satellite variables only and associated these latter with the neuron's vector for the seven<?pagebreak page224?> phytoplankton groups. When all the observations were used as a test, we calculated a mean <inline-formula><mml:math id="M114" display="inline"><mml:mrow><mml:msup><mml:mi>R</mml:mi><mml:mn mathvariant="normal">2</mml:mn></mml:msup></mml:mrow></mml:math></inline-formula> and an RMSE associated with the given size map, while comparing the estimated and observed phytoplankton group values. The best SOM configuration and variable combination are based on an optimum where <inline-formula><mml:math id="M115" display="inline"><mml:mrow><mml:mi>t</mml:mi><mml:mi>e</mml:mi></mml:mrow></mml:math></inline-formula>, <inline-formula><mml:math id="M116" display="inline"><mml:mrow><mml:mi>q</mml:mi><mml:mi>e</mml:mi></mml:mrow></mml:math></inline-formula>, and the RMSE are in low ranges while avoiding overfitting. The chosen SOM was tested using the 20 % test set, providing independent performance metrics to evaluate the generalization of the chosen SOM. As a result, we present in the paper the performance metrics of the chosen SOM configuration based on the cross-validation procedure and the test set.</p>
      <p id="d1e2385">The optimal combination of satellite parameters for the SOMRCA and SOMChlF algorithms was determined to be Chl <inline-formula><mml:math id="M117" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula>, SST, <inline-formula><mml:math id="M118" display="inline"><mml:mrow><mml:msub><mml:mi>R</mml:mi><mml:mtext>rs</mml:mtext></mml:msub></mml:mrow></mml:math></inline-formula> at four wavelengths (412, 443, 490, and 555 nm), bbp, and Kd490. The grid size for SOMRCA was set at 242 neurons, while SOMChlF had a grid size of 222 neurons. This selection was based on several factors, including a high regression coefficient between estimated and observed phytoplankton values, low error values of quantization and topographic error, and a low global RMSE encompassing all phytoplankton groups. The choice of <inline-formula><mml:math id="M119" display="inline"><mml:mrow><mml:msub><mml:mi>R</mml:mi><mml:mtext>rs</mml:mtext></mml:msub></mml:mrow></mml:math></inline-formula> bands aligns with previous work conducted on the PHYSAT method by <xref ref-type="bibr" rid="bib1.bibx3" id="text.40"/> and <xref ref-type="bibr" rid="bib1.bibx5" id="text.41"/>. The PHYSAT method utilizes water-leaving radiance anomalies in the same four selected bands to identify dominant phytoplankton functional types. In the clear open ocean, the information contained in the remote sensing reflectance (<inline-formula><mml:math id="M120" display="inline"><mml:mrow><mml:msub><mml:mi>R</mml:mi><mml:mtext>rs</mml:mtext></mml:msub></mml:mrow></mml:math></inline-formula>) bands beyond 555 nm is limited due to the strong absorption by water <xref ref-type="bibr" rid="bib1.bibx76 bib1.bibx74" id="paren.42"/>. It should be noted that the <inline-formula><mml:math id="M121" display="inline"><mml:mrow><mml:msub><mml:mi>R</mml:mi><mml:mtext>rs</mml:mtext></mml:msub></mml:mrow></mml:math></inline-formula> bands selected are commonly measured by all sensors used to build the <inline-formula><mml:math id="M122" display="inline"><mml:mrow><mml:msub><mml:mi>R</mml:mi><mml:mtext>rs</mml:mtext></mml:msub></mml:mrow></mml:math></inline-formula> product of GlobColour. This overlap of different sensors enhances data availability and coverage, thus increasing the importance of these <inline-formula><mml:math id="M123" display="inline"><mml:mrow><mml:msub><mml:mi>R</mml:mi><mml:mtext>rs</mml:mtext></mml:msub></mml:mrow></mml:math></inline-formula> bands within the initial dataset.</p>
      <p id="d1e2471">Through the iterative training process described above, the results show a significant increase in the general performance of the method when the number of neurons increases to a certain extent (Fig. <xref ref-type="fig" rid="Ch1.F4"/>). Using a number of neurons larger than the training dataset still allows a refined discretization. In this case, some neurons will capture a sample of the database, which permits defining a referent vector for these neurons. When the neuron did not capture any data observation, the discrete distance between the neighboring neurons was used to determine the referent vector <inline-formula><mml:math id="M124" display="inline"><mml:mi>w</mml:mi></mml:math></inline-formula> of each neuron that did not capture any data <xref ref-type="bibr" rid="bib1.bibx67 bib1.bibx27" id="paren.43"/>. This leads to preserving the topological order provided by new interpolated neurons. However, the quantization error’s lowest values above 350 neurons might indicate overfitting.</p>

      <?xmltex \floatpos{t}?><fig id="Ch1.F4"><?xmltex \currentcnt{4}?><?xmltex \def\figurename{Figure}?><label>Figure 4</label><caption><p id="d1e2488">Quality assessment based on the quantization and topographic error related to the training of  SOMChlF and SOMRCA as a function of increasing SOM size (number of neurons) using Chl <inline-formula><mml:math id="M125" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula>, SST, <inline-formula><mml:math id="M126" display="inline"><mml:mrow><mml:msub><mml:mi>R</mml:mi><mml:mtext>rs</mml:mtext></mml:msub></mml:mrow></mml:math></inline-formula> at four wavelengths (412, 443, 490, and 555 nm), bbp, and Kd490. In parallel, the average regression coefficient and the root mean squared error as a function of increasing SOM size were calculated through a “leave-one-out” cross-validation procedure. The dashed black and red lines respectively correspond to the <inline-formula><mml:math id="M127" display="inline"><mml:mrow><mml:msup><mml:mi>R</mml:mi><mml:mn mathvariant="normal">2</mml:mn></mml:msup></mml:mrow></mml:math></inline-formula> and the RMSE using the <inline-formula><mml:math id="M128" display="inline"><mml:mi>K</mml:mi></mml:math></inline-formula>-nearest neighbor algorithm. Finally, the dotted lines correspond to the chosen SOM size for SOMChlF <inline-formula><mml:math id="M129" display="inline"><mml:mo>=</mml:mo></mml:math></inline-formula> 242 neurons and SOMRCA <inline-formula><mml:math id="M130" display="inline"><mml:mo>=</mml:mo></mml:math></inline-formula> 222 neurons.</p></caption>
            <?xmltex \igopts{width=241.848425pt}?><graphic xlink:href="https://os.copernicus.org/articles/20/217/2024/os-20-217-2024-f04.png"/>

          </fig>

</sec>
<sec id="Ch1.S3.SS2.SSS3">
  <label>3.2.3</label><title>Operational phase</title>
      <p id="d1e2556">During the operational phase, we estimated the phytoplankton group variability using the best combination of satellite parameters. The set of parameters of a pixel was projected onto the SOM. In doing so, the parameters at each pixel were normalized by the variance of that same parameter within the initial training dataset to maintain an equal weight among the parameters and were assigned with the closest best-matching neuron using the truncated distance. At the end of the assignment phase, each pixel was associated with a referent vector corresponding to the best-matching neuron, which includes the seven phytoplankton groups as a function of relative cell abundance in the case of SOMRCA or Chl <inline-formula><mml:math id="M131" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> fraction in the case of SOMChlF. Since the training was undertaken for the<?pagebreak page225?> whole phytoplankton community at once, alongside the total Chl <inline-formula><mml:math id="M132" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> information, the SOM allows the inherent structure of the data to be preserved.</p>
      <p id="d1e2573">For this phase, level-3 mapped 4 km daily images were used to estimate the phytoplankton groups at the same spatiotemporal resolution.</p>
</sec>
<sec id="Ch1.S3.SS2.SSS4">
  <label>3.2.4</label><title>Masking and uncertainty evaluation</title>
      <p id="d1e2585">Given that our initial dataset is of limited size, it is possible that it does not contain certain naturally occurring cases. In order to prevent abnormal predictions for cases not observed in the initial dataset, we conducted a quality evaluation of the method's output. This evaluation involved quantifying a reliability index by comparing the set of satellite parameter values at a particular pixel with the values of the same parameters in the initial dataset. If a satellite variable's value fell outside the range defined within the initial dataset by the mean value of the same variable’s distribution plus or minus 2 standard deviations, it was considered distant. This evaluation was performed for all satellite variables per pixel, and the reliability index was determined by dividing the number of accepted variables by the total number of existing variables. A higher reliability index indicates greater reliability of the method, while regions with lower reliability index values require additional attention.</p>
      <p id="d1e2588">In the context of the global ocean, numerous uncertainties are associated with in situ measurements,  model parameterization, satellite parameters, and regions. The SOM algorithm is known to effectively reduce noise and mitigate the impact of uncertainties within the dataset <xref ref-type="bibr" rid="bib1.bibx20" id="paren.44"/>. However, the main source of uncertainty in the estimation process stems from selecting the best-matching neuron. This involves finding and associating the closest neuron in the SOM with a new or unfamiliar observation, such as a satellite pixel. Due to the topology conservation, a pixel could be assigned to several close neurons, forming a neighborhood along a distance gradient. Consequently, a single satellite observation can represent various probabilities of phytoplankton group combinations influenced, to a certain extent, by the uncertainties of the satellite parameter.</p>
      <p id="d1e2594">To account for uncertainties in the estimations, we opted to associate each pixel and phytoplankton group (based on relative cell abundance or Chl <inline-formula><mml:math id="M133" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> fraction) with a weighted standard deviation derived from the values of the 10 closest neurons. The weights were determined by the distances between the first 10 matching neurons and the pixel. This approach allowed us to incorporate uncertainties into the assignment process and provide a confidence measure for each pixel's assignment. By considering both the reliability index and the weighted standard deviation, we could assess the influence of uncertainties in the satellite variables.</p>
      <p id="d1e2604">However, we should acknowledge the importance of addressing the uncertainties in the <italic>psbO</italic> measurements and their potential impacts on the algorithm's outputs, which are not taken into account in this study. This exclusion is primarily due to the absence of a comprehensive framework that accounts for all the associated steps in the quantification of <italic>psbO</italic>, including aspects such as filtration, extraction, and the accuracy of <italic>psbO</italic> analysis. <xref ref-type="bibr" rid="bib1.bibx57" id="text.45"/> conducted a thorough comparative study, evaluating <italic>psbO</italic> quantities against data obtained from confocal and optical microscopy, as well as cytometry, revealing  agreement of 70 % (Spearman's rho <inline-formula><mml:math id="M134" display="inline"><mml:mo>=</mml:mo></mml:math></inline-formula> 0.64–0.71, <inline-formula><mml:math id="M135" display="inline"><mml:mi>p</mml:mi></mml:math></inline-formula> value <inline-formula><mml:math id="M136" display="inline"><mml:mo>&lt;</mml:mo></mml:math></inline-formula> 0.001). However, it is essential to recognize that like <italic>psbO</italic>, every quantification method is subject to uncertainties stemming from the various steps of the quantification process, emphasizing the necessity of comprehensive assessments within every in situ measurement protocol.</p>

      <?xmltex \floatpos{t}?><fig id="Ch1.F5" specific-use="star"><?xmltex \currentcnt{5}?><?xmltex \def\figurename{Figure}?><label>Figure 5</label><caption><p id="d1e2650">Results of the two-step cross-validation (blue) and test (red) procedures for SOMChlF <bold>(a)</bold> and SOMRCA <bold>(b)</bold> with the chosen best combination of satellite parameters and a SOM grid, respectively, of 242 and 222 neurons. From the initial dataset consisting of 145 observations, two sets were split: 80 % to be used in a one-leave-out cross-validation procedure and 20 % as an independent test. For the cross-validation, each observation among the 115 observations was used iteratively as a training set and as a test set until all observations served as tests (blue dots).  This procedure was used to identify the best satellite combination and SOM grid size. Finally, the remaining 30 observations were used as a test to evaluate the generalization capacity of the SOM with the chosen configuration (red dots). One should note that, for SOMChlF, the <inline-formula><mml:math id="M137" display="inline"><mml:mrow><mml:msup><mml:mi>R</mml:mi><mml:mn mathvariant="normal">2</mml:mn></mml:msup></mml:mrow></mml:math></inline-formula> was calculated using log-transformed Chl <inline-formula><mml:math id="M138" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> values. For complete evaluation metrics refer to Table 3.</p></caption>
            <?xmltex \igopts{width=455.244094pt}?><graphic xlink:href="https://os.copernicus.org/articles/20/217/2024/os-20-217-2024-f05.png"/>

          </fig>

</sec>
</sec>
<sec id="Ch1.S3.SS3">
  <label>3.3</label><title>Characterization of phytoplankton biomes</title>
      <p id="d1e2692">To emphasize the predominant data structure learned by SOMChlF, the ascending hierarchical clustering algorithm (AHC) was used to characterize phytoplankton biomes on the basis of their Chl <inline-formula><mml:math id="M139" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> fractions (a proxy for a phytoplankton group's biomass) and optical signature.</p>
      <p id="d1e2702">The AHC is a bottom-up clustering algorithm. The AHC starts with individuals and combines them according to their similarity (with respect to the chosen distance) to obtain new clusters. The exact number of biomes is not known a priori, but at the end of the SOM <inline-formula><mml:math id="M140" display="inline"><mml:mo>+</mml:mo></mml:math></inline-formula> AHC procedure, several possibilities of a number of clusters to be taken into account were revealed. A compromise was made between the number of clusters we could explain from a physical point of view and the number of clusters for which we needed to include the maximum  information embedded in the dataset. This procedure has been used with success in several studies <xref ref-type="bibr" rid="bib1.bibx61 bib1.bibx62 bib1.bibx65 bib1.bibx69 bib1.bibx29" id="paren.46"/>. At the end of the AHC clustering phase, each neuron of the SOMChlF was  associated with a cluster. The association of several neurons in a cluster allows us to identify common phytoplankton community structures and therefore characterize phytoplankton biomes. Upon applying SOMChlF as described in the operational phase section, each pixel of a satellite image could be associated with a cluster.</p>
</sec>
<sec id="Ch1.S3.SS4">
  <label>3.4</label><title>Evaluation of pigments to estimate phytoplankton groups</title>
      <p id="d1e2724">Each phytoplankton group's <italic>psbO</italic> abundance was associated with its corresponding HPLC pigment measurements performed for the same <italic>Tara</italic> Oceans station. The ability of pigments to predict a specific phytoplankton group was evaluated using a bagged random forest algorithm (number of learners set to 200), following the permutation-based importance method.</p>
      <p id="d1e2733">Using this method, a pigment composition of the seven major phytoplankton pigments cited in Table 1 was tested to predict the abundance of each of the seven <italic>psbO</italic>-derived<?pagebreak page226?> phytoplankton groups and therefore estimate their importance relative to each group. The concentration of each pigment was converted in terms of pigment ratios, which is a ratio relative to the sum of all pigment concentrations, and in parallel, the <italic>psbO</italic>-derived relative abundance was used.</p>
      <p id="d1e2742">The bagged random forest algorithm is a set of decision trees, each composed of internal nodes and leaves. Within the internal nodes, the algorithm uses pigment data as the predictor variable to partition the dataset into subsets based on pigment characteristics. These subsets are then utilized to predict the abundance of specific phytoplankton groups, enabling effective analysis of the importance of pigments to describe the variability of a phytoplankton group. Since this algorithm is used in a case of regression, the training is done while minimizing the error between the <italic>psbO</italic>-derived phytoplankton group abundance and the predicted one. The permutation-based importance method will randomly shuffle each pigment and compute the change in the model’s performance to predict the abundance of a phytoplankton group.</p>
</sec>
</sec>
<sec id="Ch1.S4">
  <label>4</label><title>Results and discussion</title>
<sec id="Ch1.S4.SS1">
  <label>4.1</label><title>Performances, uncertainties, and spatial limitation of the SOMRCA and SOMChlF algorithms</title>
      <p id="d1e2764">To assess the integrity of inter-variable relationships within the input data represented by the self-organizing map (SOM), a comparison of correlation coefficients and distributions of phytoplankton group values was conducted between SOMRCA and SOMChlF with their respective measures, <inline-formula><mml:math id="M141" display="inline"><mml:mrow><mml:msub><mml:mi>D</mml:mi><mml:mtext>RCA</mml:mtext></mml:msub></mml:mrow></mml:math></inline-formula> and <inline-formula><mml:math id="M142" display="inline"><mml:mrow><mml:msub><mml:mi>D</mml:mi><mml:mtext>ChlF</mml:mtext></mml:msub></mml:mrow></mml:math></inline-formula>. This analysis indicated that the correlation coefficients and value distributions remained unaffected within both SOMRCA and SOMChlF compared to the initial dataset, illustrating the capacity of SOM to retain the characteristics of the original dataset after training (Figs. S5 and S6).</p>
      <p id="d1e2789">The cross-validation and test exercises demonstrated an average <inline-formula><mml:math id="M143" display="inline"><mml:mrow><mml:msup><mml:mi>R</mml:mi><mml:mn mathvariant="normal">2</mml:mn></mml:msup></mml:mrow></mml:math></inline-formula> of 0.68 for SOMRCA and 0.74 for SOMChlF across all phytoplankton groups (Fig. <xref ref-type="fig" rid="Ch1.F5"/>, Table 3). Aggregating all Chl <inline-formula><mml:math id="M144" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> fractions showed satisfactory agreement between estimated total Chl <inline-formula><mml:math id="M145" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> and in situ values<?pagebreak page227?> (<inline-formula><mml:math id="M146" display="inline"><mml:mrow><mml:msup><mml:mi>R</mml:mi><mml:mn mathvariant="normal">2</mml:mn></mml:msup><mml:mo>=</mml:mo><mml:mn mathvariant="normal">0.83</mml:mn></mml:mrow></mml:math></inline-formula>), indicating the preservation of the initial phytoplankton quantity expressed in total Chl <inline-formula><mml:math id="M147" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula>. For SOMRCA, the RMSE ranged between 2 % and 24 % in the test set and between 2 % and 19 % in cross-validation. The highest errors were observed for prokaryotes, reaching 24 % due to their high relative cell abundance in the initial dataset. In the case of SOMChlF, the RMSE ranged between 0.02 and 0.24 mg m<inline-formula><mml:math id="M148" display="inline"><mml:msup><mml:mi/><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">3</mml:mn></mml:mrow></mml:msup></mml:math></inline-formula>  in cross-validation and 0.02 and 0.31 in the test set, with the highest error associated with the estimation of Chl <inline-formula><mml:math id="M149" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula>, stemming from the cumulative Chl <inline-formula><mml:math id="M150" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> fractions of phytoplankton groups. Notably, the largest RMSE  among phytoplankton groups was observed for the diatom Chl <inline-formula><mml:math id="M151" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> fraction, attributed to their substantial Chl <inline-formula><mml:math id="M152" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> content and its exponential relationship with total Chl <inline-formula><mml:math id="M153" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula>. The MRD highlighted a distinct contrast between SOMRCA and SOMChlF performance. Notably, SOMRCA exhibited a significantly higher median relative deviation approximately 3 times that of SOMChlF's MRD. The MRD for SOMRCA fluctuated between 0.36 and 0.81 for cross-validation and between 0.28 and 0.92 for the test set, with dinoflagellates exhibiting the highest MRD. In contrast, SOMChlF's MRD per group ranged between 0.13 and 0.24 for phytoplankton Chl <inline-formula><mml:math id="M154" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> fraction and 0.33 for Chl <inline-formula><mml:math id="M155" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> in the test set. This discrepancy emphasizes the complexity of determining the phytoplankton community structure in terms of relative cell abundance, indicating the likelihood of diverse community structures responding to the same satellite-derived environmental context.</p>
      <p id="d1e2904">Given the limited size of the initial dataset, applying SOMRCA and SOMChlF to the global satellite data must be done with caution.  For each pixel and at each time step between 1997 and 2021, we performed the quality control described in Sect. 3.2.3 to provide a measure of the applicability of this method (Fig. <xref ref-type="fig" rid="Ch1.F6"/>).  Regions of low confidence can be identified where the value of the reliability index does not exceed 60 % throughout the time series. This threshold was arbitrarily chosen while evaluating the frequency histogram of this index's values in Fig. <xref ref-type="fig" rid="Ch1.F6"/>. A value of 60 % roughly translates to the exclusion of three out of eight satellite parameter values considered outliers at a certain pixel. These regions are mainly found in coastal and turbid waters, as well as the South Pacific Ocean gyre, and are characterized either by very high or very low Chl <inline-formula><mml:math id="M156" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> values. This result is expected because the SOM algorithm is mainly adapted for case 1 waters and cannot extrapolate beyond the distribution of values in the initial dataset. Furthermore, moderate-confidence regions in which around 20 % of the pixels fall out of the accepted bounds are highlighted by a reliability index under 80 %. These regions are mainly found at high latitudes, especially in the Southern Ocean, mainly due to the limited number of available samples in the area and the particular optical characteristics of that region <xref ref-type="bibr" rid="bib1.bibx50" id="paren.47"/>.</p>
      <p id="d1e2921">Uncertainty values reached 30 % relative cell abundance for SOMRCA and 0.15 mg m<inline-formula><mml:math id="M157" display="inline"><mml:msup><mml:mi/><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">3</mml:mn></mml:mrow></mml:msup></mml:math></inline-formula> of Chl <inline-formula><mml:math id="M158" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> for SOMChlF, revealing distinct regional patterns in both cases. Notably, the observed uncertainties generally aligned with the concentration gradient in Chl <inline-formula><mml:math id="M159" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> fraction and cell abundance per group. The uncertainty associated with SOMRCA's outputs corresponded to the high relative deviation noted in the test and cross-validation, suggesting the potential acceptance of multiple community structures represented by the neurons of SOMRCA for a single satellite pixel, thus contributing to increased uncertainty levels. Regions at high latitudes exhibited the highest uncertainties for diatoms, green algae, and haptophyte relative cell abundances, while the Southern Ocean displayed heightened uncertainties specifically for prokaryotic cell abundance.</p>
      <p id="d1e2951">The increased uncertainty within the Southern Ocean, particularly for prokaryotes, could be attributed to the limited sampling conducted in this geographical region. This limitation resulted in a notable dissimilarity between satellite data collected in this area and the data sampled in the initial dataset, aligning with the findings of the reliability index. This finding is consistent with the documented very low abundance of cyanobacteria in the Southern Ocean <xref ref-type="bibr" rid="bib1.bibx32" id="paren.48"/>, which may contribute to heightened model uncertainty for this particular region.</p>

<?xmltex \floatpos{t}?><table-wrap id="Ch1.T3" specific-use="star"><?xmltex \currentcnt{3}?><label>Table 3</label><caption><p id="d1e2960">Results of the cross-validation and test exercises of SOMRCA and SOMChlF based on the regression coefficient (<inline-formula><mml:math id="M160" display="inline"><mml:mrow><mml:msup><mml:mi>R</mml:mi><mml:mn mathvariant="normal">2</mml:mn></mml:msup></mml:mrow></mml:math></inline-formula>), the root mean squared error (RMSE), and the median relative deviation (MRD).  One should note that, for SOMChlF, the <inline-formula><mml:math id="M161" display="inline"><mml:mrow><mml:msup><mml:mi>R</mml:mi><mml:mn mathvariant="normal">2</mml:mn></mml:msup></mml:mrow></mml:math></inline-formula> and MRD were calculated using log-transformed Chl <inline-formula><mml:math id="M162" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> values and RMSE was calculated using real Chl <inline-formula><mml:math id="M163" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> values.</p></caption><oasis:table frame="topbot"><?xmltex \begin{scaleboxenv}{.89}[.89]?><oasis:tgroup cols="13">
     <oasis:colspec colnum="1" colname="col1" align="left"/>
     <oasis:colspec colnum="2" colname="col2" align="right"/>
     <oasis:colspec colnum="3" colname="col3" align="right"/>
     <oasis:colspec colnum="4" colname="col4" align="right" colsep="1"/>
     <oasis:colspec colnum="5" colname="col5" align="right"/>
     <oasis:colspec colnum="6" colname="col6" align="right"/>
     <oasis:colspec colnum="7" colname="col7" align="right" colsep="1"/>
     <oasis:colspec colnum="8" colname="col8" align="right"/>
     <oasis:colspec colnum="9" colname="col9" align="right"/>
     <oasis:colspec colnum="10" colname="col10" align="right" colsep="1"/>
     <oasis:colspec colnum="11" colname="col11" align="right"/>
     <oasis:colspec colnum="12" colname="col12" align="right"/>
     <oasis:colspec colnum="13" colname="col13" align="right"/>
     <oasis:thead>
       <oasis:row>
         <oasis:entry colname="col1"/>
         <oasis:entry rowsep="1" namest="col2" nameend="col7" align="center" colsep="1">SOMRCA relative cell abundance (%) </oasis:entry>
         <oasis:entry rowsep="1" namest="col8" nameend="col13" align="center">SOMChlF phytoplankton chlorophyll <inline-formula><mml:math id="M164" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> fraction (mg m<inline-formula><mml:math id="M165" display="inline"><mml:msup><mml:mi/><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">3</mml:mn></mml:mrow></mml:msup></mml:math></inline-formula>) </oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1"/>
         <oasis:entry namest="col2" nameend="col4" align="center" colsep="1">Cross-validation </oasis:entry>
         <oasis:entry namest="col5" nameend="col7" align="center" colsep="1">Test </oasis:entry>
         <oasis:entry namest="col8" nameend="col10" align="center" colsep="1">Cross-validation </oasis:entry>
         <oasis:entry namest="col11" nameend="col13" align="center">Test </oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1"/>
         <oasis:entry rowsep="1" namest="col2" nameend="col4" align="center" colsep="1"><inline-formula><mml:math id="M166" display="inline"><mml:mrow><mml:mi>n</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">115</mml:mn></mml:mrow></mml:math></inline-formula></oasis:entry>
         <oasis:entry rowsep="1" namest="col5" nameend="col7" align="center" colsep="1"><inline-formula><mml:math id="M167" display="inline"><mml:mrow><mml:mi>n</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">30</mml:mn></mml:mrow></mml:math></inline-formula></oasis:entry>
         <oasis:entry rowsep="1" namest="col8" nameend="col10" align="center" colsep="1"><inline-formula><mml:math id="M168" display="inline"><mml:mrow><mml:mi>n</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">115</mml:mn></mml:mrow></mml:math></inline-formula></oasis:entry>
         <oasis:entry rowsep="1" namest="col11" nameend="col13" align="center"><inline-formula><mml:math id="M169" display="inline"><mml:mrow><mml:mi>n</mml:mi><mml:mo>=</mml:mo><mml:mn mathvariant="normal">30</mml:mn></mml:mrow></mml:math></inline-formula></oasis:entry>
       </oasis:row>
       <oasis:row rowsep="1">
         <oasis:entry colname="col1"/>
         <oasis:entry colname="col2"><inline-formula><mml:math id="M170" display="inline"><mml:mrow><mml:msup><mml:mi>R</mml:mi><mml:mn mathvariant="normal">2</mml:mn></mml:msup></mml:mrow></mml:math></inline-formula></oasis:entry>
         <oasis:entry colname="col3">RMSE (%)</oasis:entry>
         <oasis:entry colname="col4">MRD</oasis:entry>
         <oasis:entry colname="col5"><inline-formula><mml:math id="M171" display="inline"><mml:mrow><mml:msup><mml:mi>R</mml:mi><mml:mn mathvariant="normal">2</mml:mn></mml:msup></mml:mrow></mml:math></inline-formula></oasis:entry>
         <oasis:entry colname="col6">RMSE (%)</oasis:entry>
         <oasis:entry colname="col7">MRD</oasis:entry>
         <oasis:entry colname="col8"><inline-formula><mml:math id="M172" display="inline"><mml:mrow><mml:msup><mml:mi>R</mml:mi><mml:mn mathvariant="normal">2</mml:mn></mml:msup></mml:mrow></mml:math></inline-formula></oasis:entry>
         <oasis:entry colname="col9">RMSE (mg m<inline-formula><mml:math id="M173" display="inline"><mml:msup><mml:mi/><mml:mrow><mml:mi mathvariant="normal">−</mml:mi><mml:mn mathvariant="normal">3</mml:mn></mml:mrow></mml:msup></mml:math></inline-formula>)</oasis:entry>
         <oasis:entry colname="col10">MRD</oasis:entry>
         <oasis:entry colname="col11"><inline-formula><mml:math id="M174" display="inline"><mml:mrow><mml:msup><mml:mi>R</mml:mi><mml:mn mathvariant="normal">2</mml:mn></mml:msup></mml:mrow></mml:math></inline-formula></oasis:entry>
         <oasis:entry colname="col12">RMSE (mg m<inline-formula><mml:math id="M175" display="inline"><mml:msup><mml:mi/><mml:mrow><mml:mi mathvariant="normal">−</mml:mi><mml:mn mathvariant="normal">3</mml:mn></mml:mrow></mml:msup></mml:math></inline-formula>)</oasis:entry>
         <oasis:entry colname="col13">MRD</oasis:entry>
       </oasis:row>
     </oasis:thead>
     <oasis:tbody>
       <oasis:row>
         <oasis:entry colname="col1">Diatoms</oasis:entry>
         <oasis:entry colname="col2">0.65</oasis:entry>
         <oasis:entry colname="col3">2.7</oasis:entry>
         <oasis:entry colname="col4">0.74</oasis:entry>
         <oasis:entry colname="col5">0.72</oasis:entry>
         <oasis:entry colname="col6">2</oasis:entry>
         <oasis:entry colname="col7">0.58</oasis:entry>
         <oasis:entry colname="col8">0.66</oasis:entry>
         <oasis:entry colname="col9">0.24</oasis:entry>
         <oasis:entry colname="col10">0.18</oasis:entry>
         <oasis:entry colname="col11">0.86</oasis:entry>
         <oasis:entry colname="col12">0.12</oasis:entry>
         <oasis:entry colname="col13">0.20</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">Dinoflagellates</oasis:entry>
         <oasis:entry colname="col2">0.79</oasis:entry>
         <oasis:entry colname="col3">5.45</oasis:entry>
         <oasis:entry colname="col4">0.81</oasis:entry>
         <oasis:entry colname="col5">0.83</oasis:entry>
         <oasis:entry colname="col6">6.15</oasis:entry>
         <oasis:entry colname="col7">0.92</oasis:entry>
         <oasis:entry colname="col8">0.65</oasis:entry>
         <oasis:entry colname="col9">0.06</oasis:entry>
         <oasis:entry colname="col10">0.26</oasis:entry>
         <oasis:entry colname="col11">0.61</oasis:entry>
         <oasis:entry colname="col12">0.07</oasis:entry>
         <oasis:entry colname="col13">0.16</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">Green algae</oasis:entry>
         <oasis:entry colname="col2">0.61</oasis:entry>
         <oasis:entry colname="col3">5.64</oasis:entry>
         <oasis:entry colname="col4">0.61</oasis:entry>
         <oasis:entry colname="col5">0.67</oasis:entry>
         <oasis:entry colname="col6">4.32</oasis:entry>
         <oasis:entry colname="col7">0.47</oasis:entry>
         <oasis:entry colname="col8">0.71</oasis:entry>
         <oasis:entry colname="col9">0.08</oasis:entry>
         <oasis:entry colname="col10">0.20</oasis:entry>
         <oasis:entry colname="col11">0.62</oasis:entry>
         <oasis:entry colname="col12">0.19</oasis:entry>
         <oasis:entry colname="col13">0.23</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">Haptophytes</oasis:entry>
         <oasis:entry colname="col2">0.66</oasis:entry>
         <oasis:entry colname="col3">3.42</oasis:entry>
         <oasis:entry colname="col4">0.36</oasis:entry>
         <oasis:entry colname="col5">0.33</oasis:entry>
         <oasis:entry colname="col6">3.04</oasis:entry>
         <oasis:entry colname="col7">0.28</oasis:entry>
         <oasis:entry colname="col8">0.76</oasis:entry>
         <oasis:entry colname="col9">0.07</oasis:entry>
         <oasis:entry colname="col10">0.16</oasis:entry>
         <oasis:entry colname="col11">0.68</oasis:entry>
         <oasis:entry colname="col12">0.05</oasis:entry>
         <oasis:entry colname="col13">0.13</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">Prokaryotes</oasis:entry>
         <oasis:entry colname="col2">0.6</oasis:entry>
         <oasis:entry colname="col3">19.27</oasis:entry>
         <oasis:entry colname="col4">0.59</oasis:entry>
         <oasis:entry colname="col5">0.67</oasis:entry>
         <oasis:entry colname="col6">20.53</oasis:entry>
         <oasis:entry colname="col7">0.76</oasis:entry>
         <oasis:entry colname="col8">0.57</oasis:entry>
         <oasis:entry colname="col9">0.05</oasis:entry>
         <oasis:entry colname="col10">0.25</oasis:entry>
         <oasis:entry colname="col11">0.76</oasis:entry>
         <oasis:entry colname="col12">0.10</oasis:entry>
         <oasis:entry colname="col13">0.23</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">Cryptophytes</oasis:entry>
         <oasis:entry colname="col2">0.62</oasis:entry>
         <oasis:entry colname="col3">1.98</oasis:entry>
         <oasis:entry colname="col4">0.45</oasis:entry>
         <oasis:entry colname="col5">0.78</oasis:entry>
         <oasis:entry colname="col6">1.98</oasis:entry>
         <oasis:entry colname="col7">0.61</oasis:entry>
         <oasis:entry colname="col8">0.7</oasis:entry>
         <oasis:entry colname="col9">0.05</oasis:entry>
         <oasis:entry colname="col10">0.16</oasis:entry>
         <oasis:entry colname="col11">0.77</oasis:entry>
         <oasis:entry colname="col12">0.04</oasis:entry>
         <oasis:entry colname="col13">0.10</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">Pelagophytes</oasis:entry>
         <oasis:entry colname="col2">0.64</oasis:entry>
         <oasis:entry colname="col3">2.6</oasis:entry>
         <oasis:entry colname="col4">0.78</oasis:entry>
         <oasis:entry colname="col5">0.36</oasis:entry>
         <oasis:entry colname="col6">1.96</oasis:entry>
         <oasis:entry colname="col7">0.65</oasis:entry>
         <oasis:entry colname="col8">0.68</oasis:entry>
         <oasis:entry colname="col9">0.02</oasis:entry>
         <oasis:entry colname="col10">0.21</oasis:entry>
         <oasis:entry colname="col11">0.74</oasis:entry>
         <oasis:entry colname="col12">0.02</oasis:entry>
         <oasis:entry colname="col13">0.24</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">Chlorophyll <inline-formula><mml:math id="M176" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula></oasis:entry>
         <oasis:entry colname="col2"/>
         <oasis:entry colname="col3"/>
         <oasis:entry colname="col4"/>
         <oasis:entry colname="col5"/>
         <oasis:entry colname="col6"/>
         <oasis:entry colname="col7"/>
         <oasis:entry colname="col8">0.83</oasis:entry>
         <oasis:entry colname="col9">0.23</oasis:entry>
         <oasis:entry colname="col10">0.24</oasis:entry>
         <oasis:entry colname="col11">0.72</oasis:entry>
         <oasis:entry colname="col12">0.31</oasis:entry>
         <oasis:entry colname="col13">0.33</oasis:entry>
       </oasis:row>
     </oasis:tbody>
   </oasis:tgroup><?xmltex \end{scaleboxenv}?></oasis:table><?xmltex \gdef\@currentlabel{3}?></table-wrap>

      <?xmltex \floatpos{t}?><fig id="Ch1.F6"><?xmltex \currentcnt{6}?><?xmltex \def\figurename{Figure}?><label>Figure 6</label><caption><p id="d1e3600">Applicability of the satellite <italic>psbO</italic>-based method. The geographical <bold>(a)</bold> and value distributions <bold>(b)</bold> of the reliability index were calculated between 1997 and 2021 by testing the set of satellite parameters at a given pixel against the values in the original dataset (<inline-formula><mml:math id="M177" display="inline"><mml:mi>D</mml:mi></mml:math></inline-formula>).</p></caption>
          <?xmltex \igopts{width=241.848425pt}?><graphic xlink:href="https://os.copernicus.org/articles/20/217/2024/os-20-217-2024-f06.png"/>

        </fig>

</sec>
<sec id="Ch1.S4.SS2">
  <label>4.2</label><title>Comparison with global HPLC pigment dataset</title>
      <?pagebreak page228?><p id="d1e3633">The global in situ HPLC dataset was then used to estimate Chl <inline-formula><mml:math id="M178" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> fractions for each phytoplankton group using the diagnostic pigment approach (DPA). This dataset was compared to the Chl <inline-formula><mml:math id="M179" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> fraction matching each phytoplankton group that was estimated by SOMChlF (Fig. <xref ref-type="fig" rid="Ch1.F7"/>). A total of 2671 matchups were found following the same procedure described in Sect. 3.1. Evaluating the sum of Chl <inline-formula><mml:math id="M180" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> fractions and comparing it with in situ Chl <inline-formula><mml:math id="M181" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> can be considered a baseline evaluation of this method. This comparison showed a satisfying correspondence score of <inline-formula><mml:math id="M182" display="inline"><mml:mrow><mml:msup><mml:mi>R</mml:mi><mml:mn mathvariant="normal">2</mml:mn></mml:msup><mml:mo>=</mml:mo><mml:mn mathvariant="normal">0.72</mml:mn></mml:mrow></mml:math></inline-formula>. Relatively good correspondence is noted for diatoms and haptophytes, showing an <inline-formula><mml:math id="M183" display="inline"><mml:mrow><mml:msup><mml:mi>R</mml:mi><mml:mn mathvariant="normal">2</mml:mn></mml:msup><mml:mo>=</mml:mo><mml:mn mathvariant="normal">0.64</mml:mn></mml:mrow></mml:math></inline-formula> between in situ and SOMChlF for diatoms and 0.65 for haptophytes. Moderate correspondence was found for green algae, cryptophytes, and pelagophytes, with an <inline-formula><mml:math id="M184" display="inline"><mml:mrow><mml:msup><mml:mi>R</mml:mi><mml:mn mathvariant="normal">2</mml:mn></mml:msup></mml:mrow></mml:math></inline-formula> ranging between 0.43 and 0.39. Prokaryotes and dinoflagellates had the lowest correspondence between both outputs. The comparison between DPA-based phytoplankton groups and SOMChlF estimates is highly uncertain. It compares two types of information indicating the same phytoplankton group, with different underlying assumptions about how to define and describe a certain group. For some of the groups, these results are coherent. For example, the diatom Chl <inline-formula><mml:math id="M185" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> fraction is well captured by the latter, and the values agree with those estimated using HPLC observations; however, we noted a major overestimation within the HPLC DPA method. For prokaryotes, this comparison leads us to say that using zeaxanthin as an indicator of the cyanobacterial contribution to Chl <inline-formula><mml:math id="M186" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> may not entirely represent this group.</p>
      <p id="d1e3722">The permutation-based importance analysis using random forest, performed on the in situ <italic>Tara</italic> Oceans <italic>psbO</italic> and HPLC measurements, emphasizes the necessity of a multivariate approach for predicting phytoplankton community structure based on pigments (see Fig. <xref ref-type="fig" rid="Ch1.F8"/>). Notably, the diagnostic pigments mentioned in Table 1 exhibited dominant importance in determining the relative abundance of their respective assigned phytoplankton groups. For instance, peridinin represented dinoflagellates, chlorophyll <inline-formula><mml:math id="M187" display="inline"><mml:mi>b</mml:mi></mml:math></inline-formula> characterized green algae, and zeaxanthin indicated prokaryotes (Table 1). These pigments demonstrated the highest importance for their respective groups, as illustrated in Fig. <xref ref-type="fig" rid="Ch1.F8"/>, accompanied by a positive Spearman correlation. However, individually, these pigments accounted for less than 25 % of the variance in their respective groups. Conversely, in the case of cryptophytes, diatoms, and haptophytes, no pigment stood out in terms of importance, and the observed correlations were related to covariation between pigments (e.g., Chl <inline-formula><mml:math id="M188" display="inline"><mml:mi>b</mml:mi></mml:math></inline-formula> and Fuco in diatoms), possibly influenced by Chl <inline-formula><mml:math id="M189" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> variability. Therefore, the variability within each group is best explained not by a single diagnostic pigment, but rather by the overall pigment composition. It is crucial to consider how natural variability can influence the interpretation of pigment composition in relation to phytoplankton community structure. Pigment ratios not only vary with phytoplankton composition but also reflect the diverse strategies employed by different phytoplankton types to acclimate to environmental factors such as light, temperature, nutrients, and other variables.</p>

      <?xmltex \floatpos{t}?><fig id="Ch1.F7" specific-use="star"><?xmltex \currentcnt{7}?><?xmltex \def\figurename{Figure}?><label>Figure 7</label><caption><p id="d1e3759">Comparison between the outputs of SOMChlF and the DPA approach applied to an in situ global HPLC dataset.  2672 matchups were found between the outputs of SOMChlF and the in situ dataset and analyzed in this figure. The <inline-formula><mml:math id="M190" display="inline"><mml:mrow><mml:msup><mml:mi>R</mml:mi><mml:mn mathvariant="normal">2</mml:mn></mml:msup></mml:mrow></mml:math></inline-formula> and MRD result from calculations based on log-transformed data, and RMSE is based on non-log-transformed data.</p></caption>
          <?xmltex \igopts{width=327.206693pt}?><graphic xlink:href="https://os.copernicus.org/articles/20/217/2024/os-20-217-2024-f07.png"/>

        </fig>

</sec>
<sec id="Ch1.S4.SS3">
  <label>4.3</label><title>Global patterns of satellite-derived phytoplankton groups</title>
      <p id="d1e3787">We then applied our method to GlobColour satellite data to generate a daily database spanning from 1997 to 2021, capturing the relative cell abundance and Chl <inline-formula><mml:math id="M191" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> fraction of seven phytoplankton groups of interest. Figure <xref ref-type="fig" rid="Ch1.F9"/> presents the annual patterns of relative cell abundance and Chl <inline-formula><mml:math id="M192" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> fraction for each phytoplankton group, derived from this satellite dataset.</p>
      <p id="d1e3806">Regarding relative cell abundance, the prokaryotes stand out as a dominant group. This group largely dominated tropical regions, with a relative abundance of up to 80 % in subtropical gyres. Haptophytes, green algae, and diatoms exhibited higher abundance in middle and high latitudes as well as the equatorial region, showing a maximum relative abundance of 30 %. The remaining three phytoplankton groups displayed relative abundances that barely exceeded 10 % of the total phytoplankton community. Pelagophytes and dinoflagellates were primarily observed in middle and subtropical latitudes, while cryptophytes were found in coastal areas and high latitudes.</p>
      <p id="d1e3809">Examination of how each phytoplankton group contributed to total Chl <inline-formula><mml:math id="M193" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> revealed that diatoms had a significant contribution at high latitudes and equatorial regions. Prokaryotes, on the other hand, had  an overall low to moderate contribution to total Chl <inline-formula><mml:math id="M194" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula>.</p>

      <?xmltex \floatpos{t}?><fig id="Ch1.F8"><?xmltex \currentcnt{8}?><?xmltex \def\figurename{Figure}?><label>Figure 8</label><caption><p id="d1e3829">Evaluation of secondary pigment weighting for the estimation of different phytoplankton groups. Panel <bold>(a)</bold> represents the percentage of variance of each phytoplankton group explained by a set of frequently used phytoplankton secondary pigments. This analysis has been done using a random forest algorithm applied to the in situ <italic>Tara</italic> Oceans <italic>psbO</italic> and HPLC datasets. A Spearman correlation coefficient has been calculated between each pigment and the phytoplankton groups <bold>(b)</bold>.</p></caption>
          <?xmltex \igopts{width=241.848425pt}?><graphic xlink:href="https://os.copernicus.org/articles/20/217/2024/os-20-217-2024-f08.png"/>

        </fig>

      <?pagebreak page229?><p id="d1e3850">Qualitatively, the information captured by SOMChlF was clustered into five groups, each characterized by a distinct remote sensing reflectance spectrum that corresponded to the phytoplankton community structure (Fig. <xref ref-type="fig" rid="Ch1.F10"/>). To illustrate the link between each group's contribution to total Chl <inline-formula><mml:math id="M195" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> concentration and relative cell abundance, we depicted the latter while evaluating the pixel's assigned relative abundance values for each of the five clusters. This approach revealed that three out of the five clusters are dominated by prokaryotes in terms of cell abundance (C1, C2, and C3). However, based on their relative contribution to Chl <inline-formula><mml:math id="M196" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula>, C1 was found to be dominated by prokaryotes and dinoflagellates, C2 exhibited a mixed composition, and C3 and C4 represented diatoms and other eukaryotes, whereas C5 was predominantly composed of diatoms. The shift from relative cell abundance to size-integrated relative Chl <inline-formula><mml:math id="M197" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> fraction illustrates how cell size influences Chl <inline-formula><mml:math id="M198" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> contribution and variability.</p>

      <?xmltex \floatpos{t}?><fig id="Ch1.F9" specific-use="star"><?xmltex \currentcnt{9}?><?xmltex \def\figurename{Figure}?><label>Figure 9</label><caption><p id="d1e3885">Annual composites of the relative abundances and Chl fractions of the seven <italic>psbO</italic>-derived phytoplankton groups based on satellite data (compiled using data from 1997–2021). The uncertainties related to each group and each method are because of their different possible combinations through the weighted standard deviations, as described in Sect. 3.2.3. We note that the scales for uncertainty are smaller than those in the abundance and Chl <inline-formula><mml:math id="M199" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> columns.</p></caption>
          <?xmltex \igopts{width=398.338583pt}?><graphic xlink:href="https://os.copernicus.org/articles/20/217/2024/os-20-217-2024-f09.png"/>

        </fig>

      <p id="d1e3904">Each cluster is characterized by a specific optical signature in terms of <inline-formula><mml:math id="M200" display="inline"><mml:mrow><mml:msub><mml:mi>R</mml:mi><mml:mtext>rs</mml:mtext></mml:msub></mml:mrow></mml:math></inline-formula> spectra. The <inline-formula><mml:math id="M201" display="inline"><mml:mrow><mml:msub><mml:mi>R</mml:mi><mml:mtext>rs</mml:mtext></mml:msub></mml:mrow></mml:math></inline-formula> values per wavelength were normalized based on their corresponding variance, enabling intercomparison regardless of magnitude. For instance, C1, which exhibits higher reflectance in the blue wavelength, represents clear, oligotrophic waters. In such environments with low nutrients and high surface stratification, picophytoplankton groups like cyanobacteria thrive due to their high surface-to-size ratio <xref ref-type="bibr" rid="bib1.bibx59 bib1.bibx19" id="paren.49"/>. C2 represents normalized <inline-formula><mml:math id="M202" display="inline"><mml:mrow><mml:msub><mml:mi>R</mml:mi><mml:mtext>rs</mml:mtext></mml:msub></mml:mrow></mml:math></inline-formula> spectra with insignificant differences between normalized bands, suggesting an average state where the phytoplankton community appears mixed. In C3 and C4, we observed an increase in normalized <inline-formula><mml:math id="M203" display="inline"><mml:mrow><mml:msub><mml:mi>R</mml:mi><mml:mtext>rs</mml:mtext></mml:msub></mml:mrow></mml:math></inline-formula> values in the green compared to the blue wavebands, indicating higher Chl <inline-formula><mml:math id="M204" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> in these environments. Given that C3 and C4 are located in high-latitude regions with ample nutrient resources and exceptional seasonal variability of light<?pagebreak page230?> intensity, larger-cell-sized phytoplankton groups, including diatoms, are favored, leading to increased biomass and Chl <inline-formula><mml:math id="M205" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> contribution <xref ref-type="bibr" rid="bib1.bibx16" id="paren.50"/>. C5, with the greatest difference between <inline-formula><mml:math id="M206" display="inline"><mml:mrow><mml:msub><mml:mi>R</mml:mi><mml:mtext>rs</mml:mtext></mml:msub></mml:mrow></mml:math></inline-formula> in the blue and green, represents eutrophic waters, known for their high productivity and diatom-dominated blooms <xref ref-type="bibr" rid="bib1.bibx16" id="paren.51"/>.</p>

      <?xmltex \floatpos{t}?><fig id="Ch1.F10" specific-use="star"><?xmltex \currentcnt{10}?><?xmltex \def\figurename{Figure}?><label>Figure 10</label><caption><p id="d1e3988">Satellite-derived biomes of phytoplankton communities, obtained by unsupervised clustering (hierarchical clustering) of SOMChlF neurons. Relative cell abundances per phytoplankton group as well as normalized and denormalized <inline-formula><mml:math id="M207" display="inline"><mml:mrow><mml:msub><mml:mi>R</mml:mi><mml:mtext>rs</mml:mtext></mml:msub></mml:mrow></mml:math></inline-formula> spectra were also derived. The global map shows the most frequent community structure recorded during the 1997–2021 period. A spatiotemporal analysis was conducted to highlight latitudinal patterns.</p></caption>
          <?xmltex \igopts{width=497.923228pt}?><graphic xlink:href="https://os.copernicus.org/articles/20/217/2024/os-20-217-2024-f10.png"/>

        </fig>

      <p id="d1e4009">Based on the global distributions of these clusters, several biomes can be defined. C1 is centered in subtropical gyres, C2 is found in transitional zones such as midlatitude regions and the equatorial region, C3 is observed in the Southern Ocean, C4 corresponds to high-latitude regions, and C5 is prevalent in coastal and eutrophic waters.</p>
      <p id="d1e4012">Different temporal variability is evident for each cluster across different latitudinal bands. In northern high latitudes, an increase in C5 indicates maximal productivity occurring in that region around May. At midlatitudes, the winter maximum is marked by an increase in C5 and C4 clusters. A secondary, less pronounced peak can be observed in autumn, attributed to the break in the thermocline and remineralization processes. During summer, C1 dominates the midlatitude regions. In tropical regions, C1 is predominant, with a cyclic increase in C2 suggesting coastal influences, likely due to the proximity of C2 to nutrient-rich zones like upwelling systems. In contrast to northern high latitudes, the Southern Ocean exhibits a different temporal variability. The presence of prokaryotes is signified by C1 in this region, whereas C3 dominates during the bloom season in January. This analysis confirms the Antarctic nature of C3 in contrast to C4, highlighting differences in water types between the two regions based on phytoplankton community structure and satellite data.</p>
      <p id="d1e4015">This division into parallel and transitional biomes underscores the significant influence of latitudinal physical gradients, including light availability and temperature, on the structuring of the phytoplankton community in terms of types and size. These findings align with previous global phytoplankton studies conducted in situ <xref ref-type="bibr" rid="bib1.bibx42 bib1.bibx71" id="paren.52"/> as well as satellite estimates <xref ref-type="bibr" rid="bib1.bibx4 bib1.bibx40 bib1.bibx5 bib1.bibx27 bib1.bibx82 bib1.bibx83" id="paren.53"/>.</p>

      <?xmltex \floatpos{t}?><fig id="Ch1.F11" specific-use="star"><?xmltex \currentcnt{11}?><?xmltex \def\figurename{Figure}?><label>Figure 11</label><caption><p id="d1e4026">Intercomparison of five satellite-derived phytoplankton group Chl <inline-formula><mml:math id="M208" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> fractions based on the SOMChlF, SOM-Pigments <xref ref-type="bibr" rid="bib1.bibx27" id="paren.54"/>, and <xref ref-type="bibr" rid="bib1.bibx83" id="text.55"/> algorithms for the year 2020. The annual average and the average per latitude of each Chl <inline-formula><mml:math id="M209" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> fraction are calculated to reveal global and latitudinal patterns.</p></caption>
          <?xmltex \igopts{width=426.791339pt}?><graphic xlink:href="https://os.copernicus.org/articles/20/217/2024/os-20-217-2024-f11.png"/>

        </fig>

</sec>
<sec id="Ch1.S4.SS4">
  <label>4.4</label><title>Intercomparison of satellite-derived phytoplankton group products</title>
      <p id="d1e4063">A comparison was performed between SOMChlF output and two operational products based on the Xi et al. (2021) and SOM-Pigments <xref ref-type="bibr" rid="bib1.bibx27" id="paren.56"/> algorithms. We based this on the five phytoplankton groups common to all three algorithms: diatoms, dinoflagellates, green algae, haptophytes, and prokaryotes for the year 2020. The annual patterns show substantial agreement between all three satellite-derived phytoplankton estimates (Fig. <xref ref-type="fig" rid="Ch1.F11"/>). However, some differences between the estimated quantities of Chl <inline-formula><mml:math id="M210" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> phytoplankton groups can be noted. For diatoms, the outputs based on <xref ref-type="bibr" rid="bib1.bibx27" id="text.57"/> and SOMChlF exhibit higher Chl <inline-formula><mml:math id="M211" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> values, while those based on <xref ref-type="bibr" rid="bib1.bibx83" id="text.58"/> show low values near the equatorial latitudes. For green algae and haptophytes, the three products show matching latitudinal variability, with only minor discrepancies in values at high and subtropical latitudes. For prokaryotes, the outputs of <xref ref-type="bibr" rid="bib1.bibx83" id="text.59"/> show higher estimates, particularly near the Arctic and equatorial regions. Lastly, for dinoflagellates, the SOM-Pigments method yielded lower Chl <inline-formula><mml:math id="M212" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> values, especially in subtropical gyres, whereas SOMChlF showed the highest Chl <inline-formula><mml:math id="M213" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> estimates for this taxonomic group.</p>
      <p id="d1e4109">Upon comparing the uncertainty patterns with those observed in Xi et al. (2021), similar trends were identified for the Chl <inline-formula><mml:math id="M214" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> fraction of eukaryotic phytoplankton, displaying consistency in following the Chl <inline-formula><mml:math id="M215" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> concentration gradient as seen in our study. Notably, regions such as the gyres exhibited lower uncertainties, whereas higher uncertainties were evident in high-latitude regions and marginal seas. Conversely, when examining the uncertainty in the retrieval of prokaryote Chl <inline-formula><mml:math id="M216" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> by <xref ref-type="bibr" rid="bib1.bibx83" id="text.60"/>, lower uncertainties were noted in polar regions, contrasting with higher uncertainties observed in low-latitude regions. Similarly, in <xref ref-type="bibr" rid="bib1.bibx12" id="text.61"/>, the uncertainty maps for diatoms and dinoflagellates depicted distribution patterns akin to our uncertainty estimations in the North Atlantic Ocean.</p>
      <p id="d1e4139">This coherence in uncertainty patterns between HPLC-based products and our <italic>psbO</italic>-based product can be attributed to the direct relationship between DPA pigment concentration and total Chl <inline-formula><mml:math id="M217" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula>, as well as between <italic>psbO</italic>-derived Chl <inline-formula><mml:math id="M218" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> fractions and total Chl <inline-formula><mml:math id="M219" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula>. Consequently, similar patterns in predictions, as well as in the uncertainties, emerge.</p>
      <?pagebreak page231?><p id="d1e4169">However, addressing the similarities and differences between the outputs of the above-cited methods referring to the same phytoplankton group is not a straightforward task. These methods are based on distinct assumptions and resolutions of phytoplankton groups; the estimation of phytoplankton groups using pigments is inherently imperfect and relies on assumptions that introduce considerable variability and bias in determining the contribution of specific pigments to the assessment of phytoplankton groups. For instance, several studies showed that the DPA approach tends to overestimate diatoms <xref ref-type="bibr" rid="bib1.bibx14 bib1.bibx18" id="paren.62"/>. This approach may compromise the relevance of satellite images when used. However, the added value of such an approach resides in the availability of the large HPLC dataset, which allows the development of robust algorithms. On the other hand, the method described in this paper and the generated outputs are based for the first time on a complete and harmonized database of phytoplankton taxonomic community structure on a global scale, which is an approach that provides an unbiased picture of phytoplankton cell abundances. However, the major limitation of this approach at this time is the low number of observations from which the metric has been derived.</p>
</sec>
</sec>
<?pagebreak page232?><sec id="Ch1.S5" sec-type="conclusions">
  <label>5</label><title>Conclusions</title>
      <p id="d1e4185">By employing an alternative approach utilizing in situ metagenomic observations, a reliable ocean color algorithm for detecting phytoplankton groups was developed in this work. This achievement is noteworthy considering the limited availability of omics data used in our analysis. The successful implementation was made possible by leveraging machine learning techniques and preserving the data structure using self-organizing maps. The methodology demonstrated satisfactory performance in producing robust estimates for the seven major phytoplankton groups, albeit with some limitations in terms of global generalization due to the limited availability of data. For instance, it is important to exercise caution when interpreting estimates for regions such as the subtropical gyres. As DNA sequencing costs continue to decrease and new expeditions generate molecular data from undersampled ocean regions, we expect the training datasets to increase rapidly in future years, which should further increase the accuracy of our method. Furthermore, this study presents a new global dataset of the relative cell abundances of the seven phytoplankton groups and their contributions to total Chl <inline-formula><mml:math id="M220" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula>. These two types of information carry different implications. Chl <inline-formula><mml:math id="M221" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> serves as a biomass proxy, which is crucial for energy and matter fluxes in various ecological and biogeochemical processes. On the other hand, cell abundance represents species abundance for unicellular organisms, providing insights into community assembly processes.</p>
      <?pagebreak page233?><p id="d1e4202">This dataset opens up possibilities for intercomparisons with existing approaches, such as DPA-based methods using in situ and satellite data. The results provide coherent yet distinct information about phytoplankton communities, contributing to a better understanding of their composition. While our focus was on seven broad phytoplankton groups, it is worth mentioning that the deep taxonomic resolution achievable through molecular methods allows for species-level monitoring, which can be an interesting avenue for future implementation.</p>
      <p id="d1e4205">The methodology presented in this work provides a unique opportunity to observe the state of the major phytoplankton groups at the global scale in real time and at high resolution. This makes remote sensing observations excellent tools to collect EBVs,  play the role of broker between monitoring initiatives and decision-makers, and provide a foundation for developing marine biodiversity forecasts under different policy and management scenarios. To reach this objective, remote sensing data inherently need to be validated with in situ observations as well. Of further interest is the launch of NASA's Plankton, Aerosol, Cloud, ocean Ecosystem (PACE) mission, a strategic climate continuity mission that will make global hyperspectral ocean color measurements possible. This will allow extended data records on ocean ecology and global biogeochemistry, revolutionizing the detection of phytoplankton communities from space. From the perspective of the PACE mission, this study is a step towards further understanding the effect of environmental changes on phytoplankton community structure and diversity.</p><?xmltex \hack{\clearpage}?>
</sec>

      
      </body>
    <back><app-group>

<?pagebreak page234?><app id="App1.Ch1.S1">
  <?xmltex \currentcnt{A}?><label>Appendix A</label><title>List of acronyms</title>
      <?pagebreak page235?><p id="d1e4220"><table-wrap id="Taba" position="anchor"><oasis:table><oasis:tgroup cols="2">
     <oasis:colspec colnum="1" colname="col1" align="left"/>
     <oasis:colspec colnum="2" colname="col2" align="left"/>
     <oasis:tbody>
       <oasis:row>
         <oasis:entry colname="col1"><bold>Acronym</bold></oasis:entry>
         <oasis:entry colname="col2"><bold>Definition</bold></oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1"/>
         <oasis:entry colname="col2"/>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">PFT</oasis:entry>
         <oasis:entry colname="col2">Phytoplankton functional type</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">PG</oasis:entry>
         <oasis:entry colname="col2">Phytoplankton taxonomic group</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">PSC</oasis:entry>
         <oasis:entry colname="col2">Phytoplankton size class</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">HPLC</oasis:entry>
         <oasis:entry colname="col2">High-performance liquid chromatography</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">DPA</oasis:entry>
         <oasis:entry colname="col2">Diagnostic pigment analysis</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">DP</oasis:entry>
         <oasis:entry colname="col2">Diagnostic pigment</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">Chl <inline-formula><mml:math id="M222" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula></oasis:entry>
         <oasis:entry colname="col2">Chlorophyll <inline-formula><mml:math id="M223" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula></oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">Fuco</oasis:entry>
         <oasis:entry colname="col2">Fucoxanthin</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">Perid</oasis:entry>
         <oasis:entry colname="col2">Peridinin</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">Allo</oasis:entry>
         <oasis:entry colname="col2">Alloxanthin</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">Zea</oasis:entry>
         <oasis:entry colname="col2">Zeaxanthin</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">Chl<inline-formula><mml:math id="M224" display="inline"><mml:mi>b</mml:mi></mml:math></inline-formula></oasis:entry>
         <oasis:entry colname="col2">Chlorophyll <inline-formula><mml:math id="M225" display="inline"><mml:mi>b</mml:mi></mml:math></inline-formula></oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">19HF</oasis:entry>
         <oasis:entry colname="col2">19<inline-formula><mml:math id="M226" display="inline"><mml:msup><mml:mi/><mml:mo>′</mml:mo></mml:msup></mml:math></inline-formula>-Hexanoyloxyfucoxanthin</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">19BF</oasis:entry>
         <oasis:entry colname="col2">19<inline-formula><mml:math id="M227" display="inline"><mml:msup><mml:mi/><mml:mo>′</mml:mo></mml:msup></mml:math></inline-formula>-Butanoyloxyfucoxanthin</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">DVChl <inline-formula><mml:math id="M228" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula></oasis:entry>
         <oasis:entry colname="col2">Divinyl-chlorophyll <inline-formula><mml:math id="M229" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula></oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">DVChl <inline-formula><mml:math id="M230" display="inline"><mml:mi>b</mml:mi></mml:math></inline-formula></oasis:entry>
         <oasis:entry colname="col2">Divinyl-chlorophyll <inline-formula><mml:math id="M231" display="inline"><mml:mi>b</mml:mi></mml:math></inline-formula></oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">MAREDAT</oasis:entry>
         <oasis:entry colname="col2">MARine Ecosystem DATa</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">PSII</oasis:entry>
         <oasis:entry colname="col2">Photosystem II</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">OC</oasis:entry>
         <oasis:entry colname="col2">Ocean color</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">SeaWiFS</oasis:entry>
         <oasis:entry colname="col2">Sea-viewing Wide Field-of-view Sensor</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">MODIS</oasis:entry>
         <oasis:entry colname="col2">Moderate Resolution Imaging Spectroradiometer</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">VIIRS</oasis:entry>
         <oasis:entry colname="col2">Visible Infrared Imaging Radiometer Suite</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">MERIS</oasis:entry>
         <oasis:entry colname="col2">Medium Resolution Imaging Spectrometer</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">OLCI</oasis:entry>
         <oasis:entry colname="col2">Ocean and Land Colour Instrument</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">PACE</oasis:entry>
         <oasis:entry colname="col2">Plankton, Aerosol, Cloud, ocean Ecosystem mission</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">IOCCG</oasis:entry>
         <oasis:entry colname="col2">International   Ocean Colour Coordinating Group</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">CCI</oasis:entry>
         <oasis:entry colname="col2">Climate Change Initiative</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">CMEMS</oasis:entry>
         <oasis:entry colname="col2">Copernicus   Marine Environment Monitoring Service</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1"><inline-formula><mml:math id="M232" display="inline"><mml:mrow><mml:msub><mml:mi>R</mml:mi><mml:mtext>rs</mml:mtext></mml:msub></mml:mrow></mml:math></inline-formula></oasis:entry>
         <oasis:entry colname="col2">Remote sensing reflectance</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">Kd490</oasis:entry>
         <oasis:entry colname="col2">Attenuation coefficient at 490 nm</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">PAR</oasis:entry>
         <oasis:entry colname="col2">Photosynthetically available radiation</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">NFLH</oasis:entry>
         <oasis:entry colname="col2">Normalized fluorescence line height</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">bbp</oasis:entry>
         <oasis:entry colname="col2">Particulate backscattering coefficient</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">SST</oasis:entry>
         <oasis:entry colname="col2">Sea surface temperature</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">SOM</oasis:entry>
         <oasis:entry colname="col2">Self-organizing map</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">AHC</oasis:entry>
         <oasis:entry colname="col2">Ascending hierarchical clustering</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">TD</oasis:entry>
         <oasis:entry colname="col2">Truncated distance</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1"><inline-formula><mml:math id="M233" display="inline"><mml:mi>D</mml:mi></mml:math></inline-formula></oasis:entry>
         <oasis:entry colname="col2">Initial dataset</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1"><inline-formula><mml:math id="M234" display="inline"><mml:mrow><mml:msub><mml:mi>D</mml:mi><mml:mtext>RCA</mml:mtext></mml:msub></mml:mrow></mml:math></inline-formula></oasis:entry>
         <oasis:entry colname="col2">Phytoplankton groups' relative cell abundance sub-dataset to train SOMRCA</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1"><inline-formula><mml:math id="M235" display="inline"><mml:mrow><mml:msub><mml:mi>D</mml:mi><mml:mtext>ChlF</mml:mtext></mml:msub></mml:mrow></mml:math></inline-formula></oasis:entry>
         <oasis:entry colname="col2">Phytoplankton groups' chlorophyll <inline-formula><mml:math id="M236" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> fraction per group sub-dataset to train SOMChlF</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">SOMRCA</oasis:entry>
         <oasis:entry colname="col2">SOM algorithm dedicated to estimating phytoplankton groups' relative cell   abundance</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">SOMChlF</oasis:entry>
         <oasis:entry colname="col2">SOM algorithm dedicated to estimating phytoplankton groups' Chl <inline-formula><mml:math id="M237" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> fraction</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">RMSE</oasis:entry>
         <oasis:entry colname="col2">Root mean squared error</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1"><inline-formula><mml:math id="M238" display="inline"><mml:mrow><mml:msup><mml:mi>R</mml:mi><mml:mn mathvariant="normal">2</mml:mn></mml:msup></mml:mrow></mml:math></inline-formula></oasis:entry>
         <oasis:entry colname="col2">Regression coefficient</oasis:entry>
       </oasis:row>
       <oasis:row>
         <oasis:entry colname="col1">MRD</oasis:entry>
         <oasis:entry colname="col2">Median relative deviation</oasis:entry>
       </oasis:row>
     </oasis:tbody>
   </oasis:tgroup></oasis:table></table-wrap>
        <?xmltex \hack{\clearpage}?></p>
</app>
  </app-group><notes notes-type="codedataavailability"><title>Code and data availability</title>

      <p id="d1e4790">The <italic>psbO</italic> dataset can be found at <uri>https://www.ebi.ac.uk/biostudies/studies/S-BSST761</uri> (<xref ref-type="bibr" rid="bib1.bibx7" id="altparen.63"/>). The GlobColour dataset can be found at  <uri>https://hermes.acri.fr/index.php</uri> (<xref ref-type="bibr" rid="bib1.bibx1" id="altparen.64"/>). The SST CCI dataset can be found at <ext-link xlink:href="https://doi.org/10.48670/moi-00169" ext-link-type="DOI">10.48670/moi-00169</ext-link> (<xref ref-type="bibr" rid="bib1.bibx31" id="altparen.65"/>). The global HPLC pigment compiled dataset sources are MAREDAT (Peloquin et al., 2013), <italic>Polarstern</italic> data (Bracher, 2015a, b, c), <italic>Tara</italic> Oceans Expedition (Pesant et al., 2015), GeP&amp;Co (Dandonneau et al., 2004), the NOMAD NASA Bio-Optical Marine Algorithm Dataset (Werdell and Bailey, 2005), and numerous campaigns found on the NASA SeaBASS portal (Werdell et al., 2003; Wright et al., 2010; de Salas et al., 2011). The MATLAB version of both operational SOM <italic>psbO</italic> algorithms, the training <italic>psbO</italic>, and satellite matchup dataset, along with the compiled HPLC data and their corresponding SOMChlF output matchups, can be accessed on Zenodo at the following links: <ext-link xlink:href="https://doi.org/10.5281/zenodo.10361485" ext-link-type="DOI">10.5281/zenodo.10361485</ext-link> (<xref ref-type="bibr" rid="bib1.bibx30" id="altparen.66"/>) and <ext-link xlink:href="https://doi.org/10.5281/zenodo.10571578" ext-link-type="DOI">10.5281/zenodo.10571578</ext-link> (<xref ref-type="bibr" rid="bib1.bibx26" id="altparen.67"/>). The MATLAB software library SOM Toolbox 2.1 was used, implementing the self-organizing map and hierarchical ascending classification algorithm © 1999 (by Esa Alhoniemi, Johan Himberg, Jukka Parviainen, and Juha Vesanto; accessible at <uri>https://github.com/ilarinieminen/SOM-Toolbox/tree/master</uri>, <xref ref-type="bibr" rid="bib1.bibx44" id="altparen.68"/>).  The MATLAB function for the random forest algorithm was used to run the algorithm (MATLAB version R2020b).</p>
  </notes><app-group>
        <supplementary-material position="anchor"><p id="d1e4846">The supplement related to this article is available online at: <inline-supplementary-material xlink:href="https://doi.org/10.5194/os-20-217-2024-supplement" xlink:title="pdf">https://doi.org/10.5194/os-20-217-2024-supplement</inline-supplementary-material>.</p></supplementary-material>
        </app-group><notes notes-type="authorcontribution"><title>Author contributions</title>

      <p id="d1e4855">Conceptualization: REH, ML, CB. Methodology: REH. Validation: REH, JPK. Formal analysis: REH, JPK, ML, CB. Investigation: REH, JPK, LZ, HL, ML, CB. Resources: ML, CB. Data curation: JPK, REH. Writing (original draft preparation): REH. Writing (review and editing): REH, JPK, LZ, HL, ML, CB. Visualization: REH. Supervision: ML, CB. Project administration: ML, CB. Funding acquisition: REH, ML, CB.</p>
  </notes><notes notes-type="competinginterests"><title>Competing interests</title>

      <p id="d1e4861">The contact author has declared that none of the authors has any competing interests.</p>
  </notes><notes notes-type="disclaimer"><title>Disclaimer</title>

      <p id="d1e4867">Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors.</p>
  </notes><ack><title>Acknowledgements</title><p id="d1e4873">The authors acknowledge the recommendations and guidance of Emmanuel Boss (University of Maine) and Sylvie Thiria (Sorbonne University), as well as the constructive comments and suggestions provided by the editor and the four reviewers, which contributed to the enhancement of the paper. Roy El Hourany acknowledges CNES postdoc fellowship 2019–2021, ANR junior professor (ANR-22-CPJ1-0003-01), CNES TOSCA 2020–2021, Sorbonne University Emergence program 2021–2023, ML4BioChange, and IFSEA that benefits from grant ANR-21-EXES-0011.  Juan Pierella Karlusich acknowledges postdoctoral funding from the Fonds Français pour l'Environnement Mondial. Chris Bowler acknowledges the ERC Advanced Award Diatomic  (grant agreement no. 835067); the Horizon Europe projects AtlantECO (grant agreement no. 862923), Marco-Bolo (grant agreement no. 101082021), and BlueRemediomics (grant agreement no. 101082304); the French government “Investissements d'Avenir” programs OCEANOMICS (ANR-11-BTBR-0008), FRANCE GENOMIQUE (ANR-10-INBS-09-08), MEMO LIFE (ANR-10-LABX-54), and PSL Research University (ANR-11-IDEX-0001-02); and the BNP Paribas Foundation's Climate and Biodiversity Initiative. This article is contribution number 148 of <italic>Tara</italic> Oceans.</p></ack><notes notes-type="financialsupport"><title>Financial support</title>

      <p id="d1e4881">This research has been supported by the Centre National d'Etudes Spatiales (postdoctoral fellowship 2019–2021 and TOSCA BIODIV: Phytoplankton Diversity from Space), Sorbonne Université (ML4BioChange Emergence program 2021–2023), the European Research Council H2020 (diatomic project, grant agreement no. 835067), the Agence Nationale de la Recherche (CPJ – ANR-22-CPJ1-0003-01, Graduate school IFSEA – ANR-21-EXES-0011, OCEANOMICS  – ANR-11-BTBR-0008, FRANCE GENOMIQUE – ANR-10-INBS-09-08, MEMO LIFE – ANR-10-LABX-54, and PSL Research University – ANR-11-IDEX-0001-02), the Fonds Français pour l'Environnement Mondial (postdoctoral fellowship), the ERC Advanced Award Diatomic (grant agreement no. 835067), AtlantECO (grant agreement no. 862923), the Horizon Europe projects Marco-Bolo (grant agreement no. 101082021) and BlueRemediomics (grant agreement no. 101082304), and the BNP Paribas Foundation's Climate and Biodiversity Initiative.</p>
  </notes><notes notes-type="reviewstatement"><title>Review statement</title>

      <p id="d1e4887">This paper was edited by Jochen Wollschlaeger and reviewed by Alison Chase and three anonymous referees.</p>
  </notes><ref-list>
    <title>References</title>

      <ref id="bib1.bibx1"><?xmltex \def\ref@label{{ACRI-ST(2019)}}?><label>ACRI-ST(2019)</label><?label ACRI-ST?><mixed-citation>ACRI-ST: The European Service for Ocean Colour – GlobColour, [data set], <uri>https://hermes.acri.fr/index.php</uri> (last access: 29 January 2024), 2019.</mixed-citation></ref>
      <ref id="bib1.bibx2"><?xmltex \def\ref@label{{Agustí(1991)}}?><label>Agustí(1991)</label><?label Agusti1991?><mixed-citation>Agustí, S.: Allometric Scaling of Light Absorption and Scattering by Phytoplankton Cells, Can. J. Fish. Aquat. Sci., 48, 763–767, <ext-link xlink:href="https://doi.org/10.1139/f91-091" ext-link-type="DOI">10.1139/f91-091</ext-link>, 1991.</mixed-citation></ref>
      <ref id="bib1.bibx3"><?xmltex \def\ref@label{{Alvain et~al.(2005)Alvain, Moulin, Dandonneau, and
Br{\'{e}}on}}?><label>Alvain et al.(2005)Alvain, Moulin, Dandonneau, and Bréon</label><?label Alvain2005?><mixed-citation>Alvain, S., Moulin, C., Dandonneau, Y., and Bréon, F.: Remote sensing of phytoplankton groups in case 1 waters from global SeaWiFS imagery, Deep-Sea Res. Pt. I, 52, 1989–2004, <ext-link xlink:href="https://doi.org/10.1016/j.dsr.2005.06.015" ext-link-type="DOI">10.1016/j.dsr.2005.06.015</ext-link>, 2005.</mixed-citation></ref>
      <ref id="bib1.bibx4"><?xmltex \def\ref@label{{Alvain et~al.(2006)Alvain, Moulin, Dandonneau, Loisel, and
Br{\'{e}}on}}?><label>Alvain et al.(2006)Alvain, Moulin, Dandonneau, Loisel, and Bréon</label><?label Alvain2006?><mixed-citation>Alvain, S., Moulin, C., Dandonneau, Y., Loisel, H., and Bréon, F. M.: A species-dependent bio-optical model of case I waters for global ocean color processing, Deep-Sea Res. Pt. I, 53, 917–925, <ext-link xlink:href="https://doi.org/10.1016/j.dsr.2006.01.011" ext-link-type="DOI">10.1016/j.dsr.2006.01.011</ext-link>, 2006.</mixed-citation></ref>
      <ref id="bib1.bibx5"><?xmltex \def\ref@label{{{Ben Mustapha} et~al.(2013){Ben Mustapha}, Alvain, Jamet, Loisel, and
Dessailly}}?><label>Ben Mustapha et al.(2013)Ben Mustapha, Alvain, Jamet, Loisel, and Dessailly</label><?label BenMustapha2013?><mixed-citation>Ben Mustapha, Z., Al<?pagebreak page236?>vain, S., Jamet, C., Loisel, H., and Dessailly, D.: Automatic classification of water-leaving radiance anomalies from global SeaWiFS imagery: Application to the detection of phytoplankton groups in open ocean waters, Remote Sens. Environ., 146, 97–112, <ext-link xlink:href="https://doi.org/10.1016/j.rse.2013.08.046" ext-link-type="DOI">10.1016/j.rse.2013.08.046</ext-link>, 2013.</mixed-citation></ref>
      <ref id="bib1.bibx6"><?xmltex \def\ref@label{{Bock et~al.(2022)Bock, Subramaniam, Juhl, Montoya, and
Duhamel}}?><label>Bock et al.(2022)Bock, Subramaniam, Juhl, Montoya, and Duhamel</label><?label bock2022quantifying?><mixed-citation>Bock, N., Subramaniam, A., Juhl, A. R., Montoya, J., and Duhamel, S.: Quantifying per-cell chlorophyll <inline-formula><mml:math id="M239" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> in natural picophytoplankton populations using fluorescence-activated cell sorting, Front. Mar. Sci., 9, 850646, <ext-link xlink:href="https://doi.org/10.3389/fmars.2022.850646" ext-link-type="DOI">10.3389/fmars.2022.850646</ext-link>, 2022.</mixed-citation></ref>
      <ref id="bib1.bibx7"><?xmltex \def\ref@label{{Bowler and Pierella Karlusich(2022)}}?><label>Bowler and Pierella Karlusich(2022)</label><?label Bowler?><mixed-citation>Bowler, C. and Pierella Karlusich, J. J.: A robust approach to estimate relative phytoplankton cell abundances from metagenomes,  BioStudies, S-BSST761,  [data set], <uri>https://www.ebi.ac.uk/biostudies/studies/S-BSST761</uri> (last access: 29 January 2024), 2022.</mixed-citation></ref>
      <ref id="bib1.bibx8"><?xmltex \def\ref@label{{{Bracher}(2015{\natexlab{a}})}}?><label>Bracher(2015a)</label><?label bracher2015ppcd1?><mixed-citation>Bracher, A., Taylor, M. H., Taylor, B., Dinter, T., Röttgers, R., and Steinmetz, F.: Using empirical orthogonal functions derived from remote-sensing reflectance for the prediction of phytoplankton pigment concentrations, Ocean Sci., 11, 139–158, <ext-link xlink:href="https://doi.org/10.5194/os-11-139-2015" ext-link-type="DOI">10.5194/os-11-139-2015</ext-link>, 2015a.</mixed-citation></ref>
      <ref id="bib1.bibx9"><?xmltex \def\ref@label{{{Bracher}(2015{\natexlab{b}})}}?><label>Bracher(2015b)</label><?label bracher2015ppcd2?><mixed-citation>Bracher, A., Taylor, M. H., Taylor, B., Dinter, T., Röttgers, R., and Steinmetz, F.: Using empirical orthogonal functions derived from remote-sensing reflectance for the prediction of phytoplankton pigment concentrations, Ocean Sci., 11, 139–158, <ext-link xlink:href="https://doi.org/10.5194/os-11-139-2015" ext-link-type="DOI">10.5194/os-11-139-2015</ext-link>, 2015b.</mixed-citation></ref>
      <ref id="bib1.bibx10"><?xmltex \def\ref@label{{{Bracher}(2015{\natexlab{c}})}}?><label>Bracher(2015c)</label><?label bracher2015ppcd3?><mixed-citation>Bracher, A., Taylor, M. H., Taylor, B., Dinter, T., Röttgers, R., and Steinmetz, F.: Using empirical orthogonal functions derived from remote-sensing reflectance for the prediction of phytoplankton pigment concentrations, Ocean Sci., 11, 139–158, <ext-link xlink:href="https://doi.org/10.5194/os-11-139-2015" ext-link-type="DOI">10.5194/os-11-139-2015</ext-link>, 2015c.</mixed-citation></ref>
      <ref id="bib1.bibx11"><?xmltex \def\ref@label{{Brewin et~al.(2015)Brewin, Sathyendranath, Jackson, Barlow, Brotas,
Airs, and Lamont}}?><label>Brewin et al.(2015)Brewin, Sathyendranath, Jackson, Barlow, Brotas, Airs, and Lamont</label><?label Brewin2015?><mixed-citation>Brewin, R. J., Sathyendranath, S., Jackson, T., Barlow, R., Brotas, V., Airs, R., and Lamont, T.: Influence of light in the mixed-layer on the parameters of a three-component model of phytoplankton size class, Remote Sens.   Environ., 168, 437–450, <ext-link xlink:href="https://doi.org/10.1016/J.RSE.2015.07.004" ext-link-type="DOI">10.1016/J.RSE.2015.07.004</ext-link>, 2015.</mixed-citation></ref>
      <ref id="bib1.bibx12"><?xmltex \def\ref@label{{Brewin et~al.(2017)Brewin, Ciavatta, Sathyendranath, Jackson,
Tilstone, Curran, Airs, Cummings, Brotas, Organelli, Dall'Olmo, and
Raitsos}}?><label>Brewin et al.(2017)Brewin, Ciavatta, Sathyendranath, Jackson, Tilstone, Curran, Airs, Cummings, Brotas, Organelli, Dall'Olmo, and Raitsos</label><?label Brewin2017?><mixed-citation>Brewin, R. J., Ciavatta, S., Sathyendranath, S., Jackson, T., Tilstone, G., Curran, K., Airs, R. L., Cummings, D., Brotas, V., Organelli, E., Dall'Olmo, G., and Raitsos, D. E.: Uncertainty in ocean-color estimates of chlorophyll for phytoplankton groups, Front. Mar. Sci., 4, 104, <ext-link xlink:href="https://doi.org/10.3389/FMARS.2017.00104/BIBTEX" ext-link-type="DOI">10.3389/FMARS.2017.00104/BIBTEX</ext-link>, 2017.</mixed-citation></ref>
      <ref id="bib1.bibx13"><?xmltex \def\ref@label{{Brewin et~al.(2010)Brewin, Sathyendranath, Hirata, Lavender,
Barciela, and Hardman-Mountford}}?><label>Brewin et al.(2010)Brewin, Sathyendranath, Hirata, Lavender, Barciela, and Hardman-Mountford</label><?label Brewin2010?><mixed-citation>Brewin, R. J. W., Sathyendranath, S., Hirata, T., Lavender, S. J., Barciela, R. M., and Hardman-Mountford, N. J.: A three-component model of phytoplankton size class for the Atlantic Ocean, Ecol. Model., 221, 1472–1483, <ext-link xlink:href="https://doi.org/10.1016/j.ecolmodel.2010.02.014" ext-link-type="DOI">10.1016/j.ecolmodel.2010.02.014</ext-link>, 2010.</mixed-citation></ref>
      <ref id="bib1.bibx14"><?xmltex \def\ref@label{{Brewin et~al.(2014)Brewin, Sathyendranath, Tilstone, Lange, and
Platt}}?><label>Brewin et al.(2014)Brewin, Sathyendranath, Tilstone, Lange, and Platt</label><?label Brewin2014?><mixed-citation>Brewin, R. J. W., Sathyendranath, S., Tilstone, G., Lange, P. K., and Platt, T.: A multicomponent model of phytoplankton size structure, J. Geophys. Res.-Oceans, 119, 3478–3496, <ext-link xlink:href="https://doi.org/10.1002/2014JC009859" ext-link-type="DOI">10.1002/2014JC009859</ext-link>, 2014.</mixed-citation></ref>
      <ref id="bib1.bibx15"><?xmltex \def\ref@label{{Brown(1995)}}?><label>Brown(1995)</label><?label Brown1995?><mixed-citation>Brown, C.: Global Distribution of Coccolithophore Blooms, Oceanography, 8, 59–60, <ext-link xlink:href="https://doi.org/10.5670/oceanog.1995.21" ext-link-type="DOI">10.5670/oceanog.1995.21</ext-link>, 1995.</mixed-citation></ref>
      <ref id="bib1.bibx16"><?xmltex \def\ref@label{{Brun et~al.(2015)Brun, Vogt, Payne, Gruber, O'brien, Buitenhuis,
Le~Qu{\'{e}}r{\'{e}}, Leblanc, and Luo}}?><label>Brun et al.(2015)Brun, Vogt, Payne, Gruber, O'brien, Buitenhuis, Le Quéré, Leblanc, and Luo</label><?label brun2015ecological?><mixed-citation> Brun, P., Vogt, M., Payne, M. R., Gruber, N., O'Brien, C. J., Buitenhuis, E. T., Le Quéré, C., Leblanc, K., and Luo, Y.-W.: Ecological niches of open ocean phytoplankton taxa, Limnol. Oceanogr., 60, 1020–1038, 2015.</mixed-citation></ref>
      <ref id="bib1.bibx17"><?xmltex \def\ref@label{{Charantonis et~al.(2015)Charantonis, Testor, Mortier, D’Ortenzio,
and Thiria}}?><label>Charantonis et al.(2015)Charantonis, Testor, Mortier, D’Ortenzio, and Thiria</label><?label charantonis2015completion?><mixed-citation> Charantonis, A. A., Testor, P., Mortier, L., D'Ortenzio, F., and Thiria, S.: Completion of a sparse GLIDER database using multi-iterative Self-Organizing Maps (ITCOMP SOM), Procedia Comput. Sci., 51, 2198–2206, 2015.</mixed-citation></ref>
      <ref id="bib1.bibx18"><?xmltex \def\ref@label{{Chase et~al.(2020)Chase, Kramer, Ha{\"{e}}ntjens, Boss, Karp-Boss,
Edmondson, and Graff}}?><label>Chase et al.(2020)Chase, Kramer, Haëntjens, Boss, Karp-Boss, Edmondson, and Graff</label><?label Chase2020?><mixed-citation>Chase, A. P., Kramer, S. J., Haëntjens, N., Boss, E. S., Karp-Boss, L., Edmondson, M., and Graff, J. R.: Evaluation of diagnostic pigments to estimate phytoplankton size classes, Limnol. Oceanogr.-Meth., 18, 570–584, <ext-link xlink:href="https://doi.org/10.1002/LOM3.10385" ext-link-type="DOI">10.1002/LOM3.10385</ext-link>, 2020.</mixed-citation></ref>
      <ref id="bib1.bibx19"><?xmltex \def\ref@label{{Chisholm(1992)}}?><label>Chisholm(1992)</label><?label chisholm1992phytoplankton?><mixed-citation>Chisholm, S. W.: Phytoplankton Size, in: Primary Productivity and Biogeochemical Cycles in the Sea, edited by: Falkowski, P. G., Woodhead, A. D., Vivirito, K., Environmental Science Research, vol 43. Springer, Boston, MA, <ext-link xlink:href="https://doi.org/10.1007/978-1-4899-0762-2_12" ext-link-type="DOI">10.1007/978-1-4899-0762-2_12</ext-link>, 1962.</mixed-citation></ref>
      <ref id="bib1.bibx20"><?xmltex \def\ref@label{{da~Silva and Costa(2013)}}?><label>da Silva and Costa(2013)</label><?label da2013clustering?><mixed-citation> da Silva, L. E. B. and Costa, J. A. F.: Clustering, noise reduction and visualization using features extracted from the self-organizing map, in: Intelligent Data Engineering and Automated Learning–IDEAL 2013: 14th International Conference, IDEAL 2013, Hefei, China, 20–23  October 2013, Proceedings 14, Springer, 242–251, 2013.</mixed-citation></ref>
      <ref id="bib1.bibx21"><?xmltex \def\ref@label{{Dairiki et~al.(2020)Dairiki, Motokawa, Murata, and
Taguchi}}?><label>Dairiki et al.(2020)Dairiki, Motokawa, Murata, and Taguchi</label><?label dairiki2020does?><mixed-citation> Dairiki, C., Motokawa, S., Murata, A., and Taguchi, S.: How does cell volume influence the total light absorption efficiency of a mixed population of dinoflagellates with similar cell shapes and pigment compositions?, Plankton Benthos Res., 15, 250–258, 2020.</mixed-citation></ref>
      <ref id="bib1.bibx22"><?xmltex \def\ref@label{{Dandonneau et~al.(2004)Dandonneau, Deschamps, Nicolas, Loisel,
Blanchot, Montel, Thieuleux, and B{\'{e}}cu}}?><label>Dandonneau et al.(2004)Dandonneau, Deschamps, Nicolas, Loisel, Blanchot, Montel, Thieuleux, and Bécu</label><?label Dandonneau2004?><mixed-citation>Dandonneau, Y., Deschamps, P.-Y., Nicolas, J.-M., Loisel, H., Blanchot, J., Montel, Y., Thieuleux, F., and Bécu, G.: Seasonal and interannual variability of ocean color and composition of phytoplankton communities in the North Atlantic, equatorial Pacific and South Pacific, Deep-Sea Res. Pt. I, 51, 303–318, <ext-link xlink:href="https://doi.org/10.1016/j.dsr2.2003.07.018" ext-link-type="DOI">10.1016/j.dsr2.2003.07.018</ext-link>, 2004.</mixed-citation></ref>
      <ref id="bib1.bibx23"><?xmltex \def\ref@label{{de~Salas et~al.(2011)de~Salas, Eriksen, Davidson, and
Wright}}?><label>de Salas et al.(2011)de Salas, Eriksen, Davidson, and Wright</label><?label de2011protistan?><mixed-citation> de Salas, M. F., Eriksen, R., Davidson, A. T., and Wright, S. W.: Protistan communities in the Australian sector of the Sub-Antarctic Zone during SAZ-Sense, Deep-Sea Res. Pt. II, 58, 2135–2149, 2011.</mixed-citation></ref>
      <ref id="bib1.bibx24"><?xmltex \def\ref@label{{{Di Cicco} et~al.(2017){Di Cicco}, Sammartino, Marullo, and
Santoleri}}?><label>Di Cicco et al.(2017)Di Cicco, Sammartino, Marullo, and Santoleri</label><?label DiCicco2017?><mixed-citation>Di Cicco, A., Sammartino, M., Marullo, S., and Santoleri, R.: Regional Empirical Algorithms for an Improved Identification of Phytoplankton Functional Types and Size Classes in the Mediterranean Sea Using Satellite Data, Front. Mar. Sci., 4, 126, <ext-link xlink:href="https://doi.org/10.3389/fmars.2017.00126" ext-link-type="DOI">10.3389/fmars.2017.00126</ext-link>, 2017.</mixed-citation></ref>
      <ref id="bib1.bibx25"><?xmltex \def\ref@label{{Dutkiewicz et~al.(2020)Dutkiewicz, Cermeno, Jahn, Follows, Hickman,
Taniguchi, and Ward}}?><label>Dutkiewicz et al.(2020)Dutkiewicz, Cermeno, Jahn, Follows, Hickman, Taniguchi, and Ward</label><?label Dutkiewicz2020?><mixed-citation>Dutkiewicz, S., Cermeno, P., Jahn, O., Follows, M. J., Hickman, A. E., Taniguchi, D. A. A., and Ward, B. A.: Dimensions of marine phytoplankton diversity, Biogeosciences, 17, 609–634, <ext-link xlink:href="https://doi.org/10.5194/bg-17-609-2020" ext-link-type="DOI">10.5194/bg-17-609-2020</ext-link>, 2020.</mixed-citation></ref>
      <ref id="bib1.bibx26"><?xmltex \def\ref@label{El Hourany(2024)}?><label>El Hourany(2024)</label><?label ElHourany2024d?><mixed-citation>El Hourany, R.: Satellite-derived phytoplankton community structure from space using psbO and machine learning,  Zenodo [code], <ext-link xlink:href="https://doi.org/10.5281/zenodo.10571578" ext-link-type="DOI">10.5281/zenodo.10571578</ext-link>, 2024.</mixed-citation></ref>
      <ref id="bib1.bibx27"><?xmltex \def\ref@label{{{El Hourany} et~al.(2019{\natexlab{a}}){El Hourany}, {Abboud-Abi
Saab}, Faour, Aumont, Cr{\'{e}}pon, and Thiria}}?><label>El Hourany et al.(2019a)El Hourany, Abboud-Abi Saab, Faour, Aumont, Crépon, and Thiria</label><?label ElHourany2019a?><mixed-citation>El Hourany, R., Abboud-Abi Saab, M., Faour, G., Aumont, O., Crépon, M., and Thiria, S.: Estimation of secondary phytoplankton pigments from satellite observations using self-organizing maps (SOM), J. Geophys. Res.-Oceans, 124, 1357–1378, <ext-link xlink:href="https://doi.org/10.1029/2018JC014450" ext-link-type="DOI">10.1029/2018JC014450</ext-link>, 2019a.</mixed-citation></ref>
      <ref id="bib1.bibx28"><?xmltex \def\ref@label{{{El Hourany} et~al.(2019{\natexlab{b}}){El Hourany}, {Abboud-Abi
Saab}, Faour, Mejia, Cr{\'{e}}pon, and Thiria}}?><label>El Hourany et al.(2019b)El Hourany, Abboud-Abi Saab, Faour, Mejia, Crépon, and Thiria</label><?label ElHourany2019?><mixed-citation>El Hourany, R., Abboud-Abi Saab, M., Faour, G., Mejia, C., Crépon, M., and Thiria, S.: Phytoplankton Diversity in the Mediterranean Sea From Satellite Data Using Self-Organizing Maps, J. Geophys. Res.-Oceans, 124, 5827–5843, <ext-link xlink:href="https://doi.org/10.1029/2019JC015131" ext-link-type="DOI">10.1029/2019JC015131</ext-link>, 2019b.</mixed-citation></ref>
      <ref id="bib1.bibx29"><?xmltex \def\ref@label{{{El Hourany} et~al.(2021){El Hourany}, Mejia, Faour, Cr{\'{e}}pon,
and Thiria}}?><label>El Hourany et al.(2021)El Hourany, Mejia, Faour, Crépon, and Thiria</label><?label ElHourany2021?><mixed-citation>El Hourany, R., Mejia, C., Faour, G., Crépon, M., and Thiria, S<?pagebreak page237?>.: Evidencing the Impact of Climate Change on the Phytoplankton Community of the Mediterranean Sea Through a Bioregionalization Approach, J. Geophys. Res.-Oceans, 126, e2020JC016808, <ext-link xlink:href="https://doi.org/10.1029/2020JC016808" ext-link-type="DOI">10.1029/2020JC016808</ext-link>, 2021.</mixed-citation></ref>
      <ref id="bib1.bibx30"><?xmltex \def\ref@label{El Hourany et~al.(2024)}?><label>El Hourany et al.(2024)</label><?label ElHourany2024?><mixed-citation>El Hourany, R., Pierella Karlusich, J. J., Zinger, L., Loisel, H., Levy, M., and Bowler, C.:  Linking satellites to genes with machine learning to estimate phytoplankton community structure from space,  Zenodo [data set], <ext-link xlink:href="https://doi.org/10.5281/zenodo.10361485" ext-link-type="DOI">10.5281/zenodo.10361485</ext-link>, 2024.</mixed-citation></ref>
      <ref id="bib1.bibx31"><?xmltex \def\ref@label{ESA SST CCI and C3S reprocessed sea surface temperature analyses(2019)}?><label>ESA SST CCI and C3S reprocessed sea surface temperature analyses(2019)</label><?label ESA?><mixed-citation>ESA SST CCI and C3S reprocessed sea surface temperature analyses: E.U. Copernicus Marine Service Information (CMEMS), Marine Data Store (MDS), [data set], <ext-link xlink:href="https://doi.org/10.48670/moi-00169" ext-link-type="DOI">10.48670/moi-00169</ext-link>, 2019.</mixed-citation></ref>
      <ref id="bib1.bibx32"><?xmltex \def\ref@label{{Flombaum et~al.(2013)Flombaum, Gallegos, Gordillo, Rincon, Zabala,
Jiao, Karl, Li, Lomas, Veneziano, Vera, Vrugt, and Martiny}}?><label>Flombaum et al.(2013)Flombaum, Gallegos, Gordillo, Rincon, Zabala, Jiao, Karl, Li, Lomas, Veneziano, Vera, Vrugt, and Martiny</label><?label Flombaum2013?><mixed-citation>Flombaum, P., Gallegos, J. L., Gordillo, R. A., Rincon, J., Zabala, L. L., Jiao, N., Karl, D. M., Li, W. K. W., Lomas, M. W., Veneziano, D., Vera, C. S., Vrugt, J. A., and Martiny, A. C.: Present and future global distributions of the marine Cyanobacteria Prochlorococcus and Synechococcus, P. Natl. Acad. Sci. USA, 110, 9824–9829, <ext-link xlink:href="https://doi.org/10.1073/pnas.1307701110" ext-link-type="DOI">10.1073/pnas.1307701110</ext-link>, 2013.</mixed-citation></ref>
      <ref id="bib1.bibx33"><?xmltex \def\ref@label{{Folguera et~al.(2015)Folguera, Zupan, Cicerone, and
Magallanes}}?><label>Folguera et al.(2015)Folguera, Zupan, Cicerone, and Magallanes</label><?label folguera2015self?><mixed-citation> Folguera, L., Zupan, J., Cicerone, D., and Magallanes, J. F.: Self-organizing maps for imputation of missing data in incomplete data matrices, Chemometr. Intell. Lab., 143, 146–151, 2015.</mixed-citation></ref>
      <ref id="bib1.bibx34"><?xmltex \def\ref@label{{Fuhrman(2009)}}?><label>Fuhrman(2009)</label><?label Fuhrman2009?><mixed-citation>Fuhrman, J. A.: Microbial community structure and its functional implications, 459, 193–199, <ext-link xlink:href="https://doi.org/10.1038/nature08058" ext-link-type="DOI">10.1038/nature08058</ext-link>, 2009.</mixed-citation></ref>
      <ref id="bib1.bibx35"><?xmltex \def\ref@label{{Fujiki and Taguchi(2002)}}?><label>Fujiki and Taguchi(2002)</label><?label fujiki2002variability?><mixed-citation> Fujiki, T. and Taguchi, S.: Variability in chlorophyll a specific absorption coefficient in marine phytoplankton as a function of cell size and irradiance, J. Plankton Res., 24, 859–874, 2002.</mixed-citation></ref>
      <ref id="bib1.bibx36"><?xmltex \def\ref@label{{Guidi et~al.(2009)Guidi, Stemmann, Jackson, Ibanez, Claustre,
Legendre, Picheral, and Gorsky}}?><label>Guidi et al.(2009)Guidi, Stemmann, Jackson, Ibanez, Claustre, Legendre, Picheral, and Gorsky</label><?label Guidi2009?><mixed-citation>Guidi, L., Stemmann, L., Jackson, G. A., Ibanez, F., Claustre, H., Legendre, L., Picheral, M., and Gorsky, G.: Effects of phytoplankton community on production, size, and export of large aggregates: A world-ocean analysis, Limnol. Oceanogr., 54, 1951–1963, <ext-link xlink:href="https://doi.org/10.4319/LO.2009.54.6.1951" ext-link-type="DOI">10.4319/LO.2009.54.6.1951</ext-link>, 2009.</mixed-citation></ref>
      <ref id="bib1.bibx37"><?xmltex \def\ref@label{{Henson et~al.(2021)Henson, Cael, Allen, and Dutkiewicz}}?><label>Henson et al.(2021)Henson, Cael, Allen, and Dutkiewicz</label><?label Henson2021?><mixed-citation>Henson, S. A., Cael, B. B., Allen, S. R., and Dutkiewicz, S.: Future phytoplankton diversity in a changing climate, Nat. Commun., 12, 1–8, <ext-link xlink:href="https://doi.org/10.1038/s41467-021-25699-w" ext-link-type="DOI">10.1038/s41467-021-25699-w</ext-link>, 2021.</mixed-citation></ref>
      <ref id="bib1.bibx38"><?xmltex \def\ref@label{{Hillebrand and Azovsky(2001)}}?><label>Hillebrand and Azovsky(2001)</label><?label Hillebrand2001?><mixed-citation>Hillebrand, H. and Azovsky, A. I.: Body size determines the strength of the latitudinal diversity gradient, Ecography, 24, 251–256, <ext-link xlink:href="https://doi.org/10.1034/J.1600-0587.2001.240302.X" ext-link-type="DOI">10.1034/J.1600-0587.2001.240302.X</ext-link>, 2001.</mixed-citation></ref>
      <ref id="bib1.bibx39"><?xmltex \def\ref@label{{Hirata et~al.(2008)Hirata, Aiken, Hardman-Mountford, Smyth, and
Barlow}}?><label>Hirata et al.(2008)Hirata, Aiken, Hardman-Mountford, Smyth, and Barlow</label><?label Hirata2008?><mixed-citation>Hirata, T., Aiken, J., Hardman-Mountford, N., Smyth, T., and Barlow, R.: An absorption model to determine phytoplankton size classes from satellite ocean colour, Remote Sens. Environ., 112, 3153–3159, <ext-link xlink:href="https://doi.org/10.1016/J.RSE.2008.03.011" ext-link-type="DOI">10.1016/J.RSE.2008.03.011</ext-link>, 2008.</mixed-citation></ref>
      <ref id="bib1.bibx40"><?xmltex \def\ref@label{{Hirata et~al.(2011)Hirata, Hardman-Mountford, Brewin, Aiken, Barlow,
Suzuki, Isada, Howell, Hashioka, Noguchi-Aita, and Yamanaka}}?><label>Hirata et al.(2011)Hirata, Hardman-Mountford, Brewin, Aiken, Barlow, Suzuki, Isada, Howell, Hashioka, Noguchi-Aita, and Yamanaka</label><?label Hirata2011b?><mixed-citation>Hirata, T., Hardman-Mountford, N. J., Brewin, R. J. W., Aiken, J., Barlow, R., Suzuki, K., Isada, T., Howell, E., Hashioka, T., Noguchi-Aita, M., and Yamanaka, Y.: Synoptic relationships between surface Chlorophyll-<inline-formula><mml:math id="M240" display="inline"><mml:mi>a</mml:mi></mml:math></inline-formula> and diagnostic pigments specific to phytoplankton functional types, Biogeosciences, 8, 311–327, <ext-link xlink:href="https://doi.org/10.5194/bg-8-311-2011" ext-link-type="DOI">10.5194/bg-8-311-2011</ext-link>, 2011.</mixed-citation></ref>
      <ref id="bib1.bibx41"><?xmltex \def\ref@label{{Hood et~al.(2006)Hood, Laws, Armstrong, Bates, Brown, Carlson, Chai,
Doney, Falkowski, Feely, Friedrichs, Landry, {Keith Moore}, Nelson,
Richardson, Salihoglu, Schartau, Toole, and Wiggert}}?><label>Hood et al.(2006)Hood, Laws, Armstrong, Bates, Brown, Carlson, Chai, Doney, Falkowski, Feely, Friedrichs, Landry, Keith Moore, Nelson, Richardson, Salihoglu, Schartau, Toole, and Wiggert</label><?label Hood2006?><mixed-citation>Hood, R. R., Laws, E. A., Armstrong, R. A., Bates, N. R., Brown, C. W., Carlson, C. A., Chai, F., Doney, S. C., Falkowski, P. G., Feely, R. A., Friedrichs, M. A., Landry, M. R., Keith Moore, J., Nelson, D. M., Richardson, T. L., Salihoglu, B., Schartau, M., Toole, D. A., and Wiggert, J. D.: Pelagic functional group modeling: Progress, challenges and prospects, Deep-Sea Res. Pt. II, 53, 459–512, <ext-link xlink:href="https://doi.org/10.1016/J.DSR2.2006.01.025" ext-link-type="DOI">10.1016/J.DSR2.2006.01.025</ext-link>, 2006.</mixed-citation></ref>
      <ref id="bib1.bibx42"><?xmltex \def\ref@label{{Ibarbalz et~al.(2019)Ibarbalz, Henry, Brand{\~{a}}o, Martini,
Busseni, Byrne, Coelho, Endo, Gasol, Gregory, Mah{\'{e}}, Rigonato,
Royo-Llonch, Salazar, Sanz-S{\'{a}}ez, Scalco, Soviadan, Zayed, Zingone,
Labadie, Ferland, Marec, Kandels, Picheral, Dimier, Poulain, Pisarev,
Carmichael, Pesant, Acinas, Babin, Bork, Boss, Bowler, Cochrane, de~Vargas,
Follows, Gorsky, Grimsley, Guidi, Hingamp, Iudicone, Jaillon, Karp-Boss,
Karsenti, Not, Ogata, Poulton, Raes, Sardet, Speich, Stemmann, Sullivan,
Sunagawa, Wincker, Pelletier, Bopp, Lombard, and Zinger}}?><label>Ibarbalz et al.(2019)Ibarbalz, Henry, Brandão, Martini, Busseni, Byrne, Coelho, Endo, Gasol, Gregory, Mahé, Rigonato, Royo-Llonch, Salazar, Sanz-Sáez, Scalco, Soviadan, Zayed, Zingone, Labadie, Ferland, Marec, Kandels, Picheral, Dimier, Poulain, Pisarev, Carmichael, Pesant, Acinas, Babin, Bork, Boss, Bowler, Cochrane, de Vargas, Follows, Gorsky, Grimsley, Guidi, Hingamp, Iudicone, Jaillon, Karp-Boss, Karsenti, Not, Ogata, Poulton, Raes, Sardet, Speich, Stemmann, Sullivan, Sunagawa, Wincker, Pelletier, Bopp, Lombard, and Zinger</label><?label Ibarbalz2019?><mixed-citation>Ibarbalz, F. M., Henry, N., Brandão, M. C., Martini, S., Busseni, G., Byrne, H., Coelho, L. P., Endo, H., Gasol, J. M., Gregory, A. C., Mahé, F., Rigonato, J., Royo-Llonch, M., Salazar, G., Sanz-Sáez, I., Scalco, E., Soviadan, D., Zayed, A. A., Zingone, A., Labadie, K., Ferland, J., Marec, C., Kandels, S., Picheral, M., Dimier, C., Poulain, J., Pisarev, S., Carmichael, M., Pesant, S., Acinas, S. G., Babin, M., Bork, P., Boss, E., Bowler, C., Cochrane, G., de Vargas, C., Follows, M., Gorsky, G., Grimsley, N., Guidi, L., Hingamp, P., Iudicone, D., Jaillon, O., Karp-Boss, L., Karsenti, E., Not, F., Ogata, H., Poulton, N., Raes, J., Sardet, C., Speich, S., Stemmann, L., Sullivan, M. B., Sunagawa, S., Wincker, P., Pelletier, E., Bopp, L., Lombard, F., and Zinger, L.: Global Trends in Marine Plankton Diversity across Kingdoms of Life, Cell, 179, 1084–1097, <ext-link xlink:href="https://doi.org/10.1016/J.CELL.2019.10.008" ext-link-type="DOI">10.1016/J.CELL.2019.10.008</ext-link>, 2019.</mixed-citation></ref>
      <ref id="bib1.bibx43"><?xmltex \def\ref@label{{Iglesias-Rodr{\'{i}}guez et~al.(2002)Iglesias-Rodr{\'{i}}guez, Brown,
Doney, Kleypas, Kolber, Kolber, Hayes, and
Falkowski}}?><label>Iglesias-Rodríguez et al.(2002)Iglesias-Rodríguez, Brown, Doney, Kleypas, Kolber, Kolber, Hayes, and Falkowski</label><?label Iglesias-Rodriguez2002?><mixed-citation>Iglesias-Rodríguez, M. D., Brown, C. W., Doney, S. C., Kleypas, J., Kolber, D., Kolber, Z., Hayes, P. K., and Falkowski, P. G.: Representing key phytoplankton functional groups in ocean carbon cycle models: Coccolithophorids, Global Biogeochem. Cy., 16, 1–20, <ext-link xlink:href="https://doi.org/10.1029/2001GB001454" ext-link-type="DOI">10.1029/2001GB001454</ext-link>, 2002.</mixed-citation></ref>
      <ref id="bib1.bibx44"><?xmltex \def\ref@label{{ilarinieminen(2012)}}?><label>ilarinieminen(2012)</label><?label ilarinieminen?><mixed-citation>ilarinieminen: SOM Toolbox 2.1, [code], <uri>https://github.com/ilarinieminen/SOM-Toolbox/tree/master</uri> (last access: 29 January 2024), 2012.</mixed-citation></ref>
      <ref id="bib1.bibx45"><?xmltex \def\ref@label{{Irigoien et~al.(2004)Irigoien, Hulsman, and Harris}}?><label>Irigoien et al.(2004)Irigoien, Hulsman, and Harris</label><?label Irigoien2004?><mixed-citation>Irigoien, X., Hulsman, J., and Harris, R. P.: Global biodiversity patterns of marine phytoplankton and zooplankton, Nature, 429, 863–867, <ext-link xlink:href="https://doi.org/10.1038/nature02593" ext-link-type="DOI">10.1038/nature02593</ext-link>, 2004.</mixed-citation></ref>
      <ref id="bib1.bibx46"><?xmltex \def\ref@label{{Jouini et~al.(2013)Jouini, L{\'{e}}vy, Cr{\'{e}}pon, and
Thiria}}?><label>Jouini et al.(2013)Jouini, Lévy, Crépon, and Thiria</label><?label Jouini2013?><mixed-citation>Jouini, M., Lévy, M., Crépon, M., and Thiria, S.: Reconstruction of satellite chlorophyll images under heavy cloud coverage using a neural classification method, Remote Sens. Environ., 131, 232–246, <ext-link xlink:href="https://doi.org/10.1016/j.rse.2012.11.025" ext-link-type="DOI">10.1016/j.rse.2012.11.025</ext-link>, 2013.</mixed-citation></ref>
      <ref id="bib1.bibx47"><?xmltex \def\ref@label{{Kaneko et~al.(2023)Kaneko, Endo, Henry, Berney, Mah{\'{e}}, Poulain,
Labadie, Beluche, El~Hourany et~al.}}?><label>Kaneko et al.(2023)Kaneko, Endo, Henry, Berney, Mahé, Poulain, Labadie, Beluche, El Hourany et al.</label><?label kaneko2023predicting?><mixed-citation>Kaneko, H., Endo, H., Henry, N., Berney, C., Mahé, F., Poulain, J., Labadie, K., Beluche, O., El Hourany, R., et al.: Predicting global distributions of eukaryotic plankton communities from satellite data, ISME Commun., 3, 101, <ext-link xlink:href="https://doi.org/10.1038/s43705-023-00308-7" ext-link-type="DOI">10.1038/s43705-023-00308-7</ext-link>, 2023.</mixed-citation></ref>
      <ref id="bib1.bibx48"><?xmltex \def\ref@label{{{Le Qu{\'{e}}r{\'{e}}} et~al.(2005){Le Qu{\'{e}}r{\'{e}}}, Harrison,
{Colin Prentice}, Buitenhuis, Aumont, Bopp, Claustre, {Cotrim Da Cunha},
Geider, Giraud, Klaas, Kohfeld, Legendre, Manizza, Platt, Rivkin,
Sathyendranath, Uitz, Watson, and Wolf-Gladrow}}?><label>Le Quéré et al.(2005)Le Quéré, Harrison, Colin Prentice, Buitenhuis, Aumont, Bopp, Claustre, Cotrim Da Cunha, Geider, Giraud, Klaas, Kohfeld, Legendre, Manizza, Platt, Rivkin, Sathyendranath, Uitz, Watson, and Wolf-Gladrow</label><?label LeQuere2005?><mixed-citation>Le Quéré, C., Harrison, S. P., Colin Prentice, I., Buitenhuis, E. T., Aumont, O., Bopp, L., Claustre, H., Cotrim Da Cunha, L., Geider, R., Giraud, X., Klaas, C., Kohfeld, K. E., Legendre, L., Manizza, M., Platt, T., Rivkin, R. B., Sathyendranath, S., Uitz, J., Watson, A. J., and Wolf-Gladrow, D.: Ecosystem dynamics based on plankton functional types for global ocean biogeochemistry models, Global Change Biol., 11, 2016–2040, <ext-link xlink:href="https://doi.org/10.1111/j.1365-2486.2005.1004.x" ext-link-type="DOI">10.1111/j.1365-2486.2005.1004.x</ext-link>, 2005.</mixed-citation></ref>
      <ref id="bib1.bibx49"><?xmltex \def\ref@label{{Losa et~al.(2017)Losa, Soppa, Dinter, Wolanin, Brewin, Bricaud,
Oelker, Peeken, Gentili, Rozanov et~al.}}?><label>Losa et al.(2017)Losa, Soppa, Dinter, Wolanin, Brewin, Bricaud, Oelker, Peeken, Gentili, Rozanov et al.</label><?label losa2017synergistic?><mixed-citation>Losa, S. N., Soppa, M. A., Dinter, T., Wolanin, A., Brewin, R. J., Bricaud, A., Oelker, J., Peeken, I., Gentili, B., Rozanov, V., et al.: Synergistic exploitation of hyper-and multi-spectral precursor sentinel measurements to determine phytoplankton functional types (SynSenPFT), Front. Mar. Sci., 4, 203, <ext-link xlink:href="https://doi.org/10.3389/fmars.2017.00203" ext-link-type="DOI">10.3389/fmars.2017.00203</ext-link>, 2017.</mixed-citation></ref>
      <ref id="bib1.bibx50"><?xmltex \def\ref@label{{Mitchell et~al.(1991)Mitchell, Brody, Holm-Hansen, McClain, and
Bishop}}?><label>Mitchell et al.(1991)Mitchell, Brody, Holm-Hansen, McClain, and Bishop</label><?label Mitchell1991?><mixed-citation>Mitchell, B. G., Brody, E. A., Holm-Hansen, O., McClain, C., and Bishop, J.: Light limitation of phytoplankton biomass and macronutrient utilization in the Southern Ocean, Limnol. Oceanogr., 36, 1662–1677, <ext-link xlink:href="https://doi.org/10.4319/lo.1991.36.8.1662" ext-link-type="DOI">10.4319/lo.1991.36.8.1662</ext-link>, 1991.</mixed-citation></ref>
      <ref id="bib1.bibx51"><?xmltex \def\ref@label{{O'Reilly et~al.(1998)O'Reilly, Maritorena, Mitchell, Siegel, Carder,
Garver, Kahru, and McClain}}?><label>O'Reilly et al.(1998)O'Reilly, Maritorena, Mitchell, Siegel, Carder, Garver, Kahru, and McClain</label><?label OReilly1998?><mixed-citation>O'Reilly, J. E., Maritorena, S., Mitchell, B. G., Siegel, D. a., Carder, K. L., Garver, S. a., Kahru, M., and McClain, C.: Ocean color chlorophyll algorithms for SeaWiFS, J. Geophys. Res., 103, 24937, <ext-link xlink:href="https://doi.org/10.1029/98JC02160" ext-link-type="DOI">10.1029/98JC02160</ext-link>, 1998.</mixed-citation></ref>
      <ref id="bib1.bibx52"><?xmltex \def\ref@label{{Organelli et~al.(2013)Organelli, Bricaud, Antoine, and
Uitz}}?><label>Organelli et al.(2013)Organelli, Bricaud, Antoine, and Uitz</label><?label Organelli2013?><mixed-citation>Organelli, E., Bricaud, A., Antoine, D., and Uitz, J.: Multivariate approach for the retrieval of phytoplankton size structure from measured light absorption spectra in the Mediterranean Sea (BOUSSOLE site), Appl. Opt., 52, 2257, <ext-link xlink:href="https://doi.org/10.1364/AO.52.002257" ext-link-type="DOI">10.1364/AO.52.002257</ext-link>, 2013.</mixed-citation></ref>
      <?pagebreak page238?><ref id="bib1.bibx53"><?xmltex \def\ref@label{{Peloquin et~al.(2013)Peloquin, Swan, Gruber, Vogt, Claustre, Ras,
Uitz, Barlow, Behrenfeld, Bidigare, Dierssen, Ditullio, Fernandez, Gallienne,
Gibb, Goericke, Harding, Head, Holligan, Hooker, Karl, Landry, Letelier,
Llewellyn, Lomas, Lucas, Mannino, Marty, Mitchell, Muller-Karger, Nelson,
Prezelin, Repeta, {Smith Jr}, Smythe-Wright, Stumpf, Subramaniam, Suzuki,
Trees, Vernet, Wasmund, and Wright}}?><label>Peloquin et al.(2013)Peloquin, Swan, Gruber, Vogt, Claustre, Ras, Uitz, Barlow, Behrenfeld, Bidigare, Dierssen, Ditullio, Fernandez, Gallienne, Gibb, Goericke, Harding, Head, Holligan, Hooker, Karl, Landry, Letelier, Llewellyn, Lomas, Lucas, Mannino, Marty, Mitchell, Muller-Karger, Nelson, Prezelin, Repeta, Smith Jr, Smythe-Wright, Stumpf, Subramaniam, Suzuki, Trees, Vernet, Wasmund, and Wright</label><?label Peloquin2013?><mixed-citation>Peloquin, J., Swan, C., Gruber, N., Vogt, M., Claustre, H., Ras, J., Uitz, J., Barlow, R., Behrenfeld, M., Bidigare, R., Dierssen, H., Ditullio, G., Fernandez, E., Gallienne, C., Gibb, S., Goericke, R., Harding, L., Head, E., Holligan, P., Hooker, S., Karl, D., Landry, M., Letelier, R., Llewellyn, C. A., Lomas, M., Lucas, M., Mannino, A., Marty, J.-C., Mitchell, B. G., Muller-Karger, F., Nelson, N., O'Brien, C., Prezelin, B., Repeta, D., Jr. Smith, W. O., Smythe-Wright, D., Stumpf, R., Subramaniam, A., Suzuki, K., Trees, C., Vernet, M., Wasmund, N., and Wright, S.: The MAREDAT global database of high performance liquid chromatography marine pigment measurements, Earth Syst. Sci. Data, 5, 109–123, <ext-link xlink:href="https://doi.org/10.5194/essd-5-109-2013" ext-link-type="DOI">10.5194/essd-5-109-2013</ext-link>, 2013.</mixed-citation></ref>
      <ref id="bib1.bibx54"><?xmltex \def\ref@label{{Pereira et~al.(2013)Pereira, Ferrier, Walters, Geller, Jongman,
Scholes, Bruford, Brummitt, Butchart, Cardoso, Coops, Dulloo, Faith, Freyhof,
Gregory, Heip, H{\"{o}}ft, Hurtt, Jetz, Karp, McGeoch, Obura, Onoda,
Pettorelli, Reyers, Sayre, Scharlemann, Stuart, Turak, Walpole, and
Wegmann}}?><label>Pereira et al.(2013)Pereira, Ferrier, Walters, Geller, Jongman, Scholes, Bruford, Brummitt, Butchart, Cardoso, Coops, Dulloo, Faith, Freyhof, Gregory, Heip, Höft, Hurtt, Jetz, Karp, McGeoch, Obura, Onoda, Pettorelli, Reyers, Sayre, Scharlemann, Stuart, Turak, Walpole, and Wegmann</label><?label Pereira2013?><mixed-citation> Pereira, H. M., Ferrier, S., Walters, M., Geller, G. N., Jongman, R. H., Scholes, R. J., Bruford, M. W., Brummitt, N., Butchart, S. H., Cardoso, A. C., Coops, N. C., Dulloo, E., Faith, D. P., Freyhof, J., Gregory, R. D., Heip, C., Höft, R., Hurtt, G., Jetz, W., Karp, D. S., McGeoch, M. A., Obura, D., Onoda, Y., Pettorelli, N., Reyers, B., Sayre, R., Scharlemann, J. P., Stuart, S. N., Turak, E., Walpole, M., and Wegmann, M.: Essential biodiversity variables, Science, 339, 277–278, 2013.</mixed-citation></ref>
      <ref id="bib1.bibx55"><?xmltex \def\ref@label{{Pesant et~al.(2015)Pesant, Not, Picheral, Kandels-Lewis, {Le Bescot},
Gorsky, Iudicone, Karsenti, Speich, Trouble, Dimier, and
Searson}}?><label>Pesant et al.(2015)Pesant, Not, Picheral, Kandels-Lewis, Le Bescot, Gorsky, Iudicone, Karsenti, Speich, Trouble, Dimier, and Searson</label><?label Pesant2015?><mixed-citation>Pesant, S., Not, F., Picheral, M., Kandels-Lewis, S., Le Bescot, N., Gorsky, G., Iudicone, D., Karsenti, E., Speich, S., Trouble, R., Dimier, C., and Searson, S.: Open science resources for the discovery and analysis of Tara Oceans data, Sci. Data, 2, 1–16, <ext-link xlink:href="https://doi.org/10.1038/SDATA.2015.23" ext-link-type="DOI">10.1038/SDATA.2015.23</ext-link>, 2015.</mixed-citation></ref>
      <ref id="bib1.bibx56"><?xmltex \def\ref@label{{{Pierella Karlusich} et~al.(2020){Pierella Karlusich}, Ibarbalz, and
Bowler}}?><label>Pierella Karlusich et al.(2020)Pierella Karlusich, Ibarbalz, and Bowler</label><?label PierellaKarlusich2020?><mixed-citation>Pierella Karlusich, J. J., Ibarbalz, F. M., and Bowler, C.: Phytoplankton in the Tara Ocean,  Annu. Rev. Mar. Sci., 12, 233–265, <ext-link xlink:href="https://doi.org/10.1146/ANNUREV-MARINE-010419-010706" ext-link-type="DOI">10.1146/ANNUREV-MARINE-010419-010706</ext-link>, 2020.</mixed-citation></ref>
      <ref id="bib1.bibx57"><?xmltex \def\ref@label{{{Pierella Karlusich} et~al.(2022){Pierella Karlusich}, Pelletier,
Zinger, Lombard, Zingone, Colin, Gasol, Dorrell, Henry, Scalco, Acinas,
Wincker, de~Vargas, and Bowler}}?><label>Pierella Karlusich et al.(2022)Pierella Karlusich, Pelletier, Zinger, Lombard, Zingone, Colin, Gasol, Dorrell, Henry, Scalco, Acinas, Wincker, de Vargas, and Bowler</label><?label PierellaKarlusich2022?><mixed-citation>Pierella Karlusich, J. J., Pelletier, E., Zinger, L., Lombard, F., Zingone, A., Colin, S., Gasol, J. M., Dorrell, R. G., Henry, N., Scalco, E., Acinas, S. G., Wincker, P., de Vargas, C., and Bowler, C.: A robust approach to estimate relative phytoplankton cell abundances from metagenomes, Molec. Ecol. Resour.,  23, 16–40, <ext-link xlink:href="https://doi.org/10.1111/1755-0998.13592" ext-link-type="DOI">10.1111/1755-0998.13592</ext-link>, 2022.</mixed-citation></ref>
      <ref id="bib1.bibx58"><?xmltex \def\ref@label{{Powell and Glazier(2017)}}?><label>Powell and Glazier(2017)</label><?label Powell2017?><mixed-citation>Powell, M. G. and Glazier, D. S.: Asymmetric geographic range expansion explains the latitudinal diversity gradients of four major taxa of marine plankton, Paleobiology, 43, 196–208, <ext-link xlink:href="https://doi.org/10.1017/PAB.2016.38" ext-link-type="DOI">10.1017/PAB.2016.38</ext-link>, 2017.</mixed-citation></ref>
      <ref id="bib1.bibx59"><?xmltex \def\ref@label{{Raven(1998)}}?><label>Raven(1998)</label><?label raven1998twelfth?><mixed-citation> Raven, J.: The twelfth Tansley Lecture. Small is beautiful: the picophytoplankton, Funct. Ecol., 12, 503–513, 1998.</mixed-citation></ref>
      <ref id="bib1.bibx60"><?xmltex \def\ref@label{{Rejeb et~al.(2022)Rejeb, Duveau, and Rebafka}}?><label>Rejeb et al.(2022)Rejeb, Duveau, and Rebafka</label><?label rejeb2022self?><mixed-citation>Rejeb, S., Duveau, C., and Rebafka, T.: Self-organizing maps for exploration of partially observed data and imputation of missing values, Chemometr.   Intell. Lab., 231, 104653, <ext-link xlink:href="https://doi.org/10.1016/j.chemolab.2022.104653" ext-link-type="DOI">10.1016/j.chemolab.2022.104653</ext-link>, 2022.</mixed-citation></ref>
      <ref id="bib1.bibx61"><?xmltex \def\ref@label{{Reygondeau et~al.(2014)Reygondeau, Irisson, Ayata, Gasparini,
Benedetti, Albouy, Hattab, Guieu, and Koubbi.}}?><label>Reygondeau et al.(2014)Reygondeau, Irisson, Ayata, Gasparini, Benedetti, Albouy, Hattab, Guieu, and Koubbi.</label><?label Reygondeau2014?><mixed-citation>Reygondeau, G., Irisson, J.-O., Ayata, S. D., Gasparini, S., Benedetti, F., Albouy, C., Hattab, T., Guieu, C., and Koubbi, P.: Definition of the Mediterranean Eco-regions and Maps of Potential Pressures in These Eco-regions, Tech. Rep., Perseus Deliverable 1, <uri>http://www.perseus-net.eu/assets/media/PDF/deliverables/3336.6_Final.pdf</uri> (last access: 26 Janaury 2024), 2014.</mixed-citation></ref>
      <ref id="bib1.bibx62"><?xmltex \def\ref@label{{Richardson et~al.(2003)Richardson, Risien, and
Shillington}}?><label>Richardson et al.(2003)Richardson, Risien, and Shillington</label><?label Richardson2003?><mixed-citation>Richardson, A. J., Risien, C., and Shillington, F. A.: Using self-organizing maps to identify patterns in satellite imagery, Prog. Oceanogr., 59, 223–239, <ext-link xlink:href="https://doi.org/10.1016/j.pocean.2003.07.006" ext-link-type="DOI">10.1016/j.pocean.2003.07.006</ext-link>, 2003. </mixed-citation></ref><?xmltex \hack{\newpage}?>
      <ref id="bib1.bibx63"><?xmltex \def\ref@label{{Righetti et~al.(2019)Righetti, Vogt, Gruber, Psomas, and
Zimmermann}}?><label>Righetti et al.(2019)Righetti, Vogt, Gruber, Psomas, and Zimmermann</label><?label Righetti2019?><mixed-citation> Righetti, D., Vogt, M., Gruber, N., Psomas, A., and Zimmermann, N. E.: Global pattern of phytoplankton diversity driven by temperature and environmental variability, Sci. Adv., 5, 6253–6268, 2019.</mixed-citation></ref>
      <ref id="bib1.bibx64"><?xmltex \def\ref@label{{Rodr{\'{i}}guez-Ramos et~al.(2015)Rodr{\'{i}}guez-Ramos,
Mara{\~{n}}{\'{o}}n, and Cerme{\~{n}}o}}?><label>Rodríguez-Ramos et al.(2015)Rodríguez-Ramos, Marañón, and Cermeño</label><?label Rodriguez-Ramos2015?><mixed-citation>Rodríguez-Ramos, T., Marañón, E., and Cermeño, P.: Marine nano- and microphytoplankton diversity: redrawing global patterns from sampling-standardized data, Global Ecol. Biogeogr., 24, 527–538, <ext-link xlink:href="https://doi.org/10.1111/GEB.12274" ext-link-type="DOI">10.1111/GEB.12274</ext-link>, 2015.</mixed-citation></ref>
      <ref id="bib1.bibx65"><?xmltex \def\ref@label{{Rossi et~al.(2014)Rossi, Ser-Giacomi, L{\~{o}}pez, and
Hern{\'{a}}ndez-Garc{\'{i}}a}}?><label>Rossi et al.(2014)Rossi, Ser-Giacomi, Lõpez, and Hernández-García</label><?label Rossi2014?><mixed-citation>Rossi, V., Ser-Giacomi, E., Lõpez, C., and Hernández-García, E.: Hydrodynamic provinces and oceanic connectivity from a transport network help designing marine reserves, Geophys. Res. Lett., 41, 2883–2891, <ext-link xlink:href="https://doi.org/10.1002/2014GL059540" ext-link-type="DOI">10.1002/2014GL059540</ext-link>, 2014.</mixed-citation></ref>
      <ref id="bib1.bibx66"><?xmltex \def\ref@label{{Saitoh(2016)}}?><label>Saitoh(2016)</label><?label 7805741?><mixed-citation>Saitoh, F.: An ensemble model of self-organizing maps for imputation of missing values, in: 2016 IEEE 9th International Workshop on Computational Intelligence and Applications (IWCIA), 9–14, <ext-link xlink:href="https://doi.org/10.1109/IWCIA.2016.7805741" ext-link-type="DOI">10.1109/IWCIA.2016.7805741</ext-link>, 2016.</mixed-citation></ref>
      <ref id="bib1.bibx67"><?xmltex \def\ref@label{{Sarzeaud and Stephan(2000)}}?><label>Sarzeaud and Stephan(2000)</label><?label Sarzeaud2000?><mixed-citation> Sarzeaud, O. and Stephan, Y.: Data interpolation using Kohonen networks, Proceedings of the International Joint Conference on Neural Networks, 6, 197–202, 2000.</mixed-citation></ref>
      <ref id="bib1.bibx68"><?xmltex \def\ref@label{{Sathyendranath et~al.(2014)Sathyendranath, Aiken, Alvain, Barlow,
Bouman, Bracher, Brewin, Bricaud, Brown, Ciotti, Clementson, Craig, Devred,
Hardman-Mountford, Hirata, Hu, Kostadinov, Lavender, Loisel, Moore, Morales,
Mouw, Nair, Raitsos, Roesler, Shutler, Sosik, Soto, Stuart, Subramaniam, and
Uitz}}?><label>Sathyendranath et al.(2014)Sathyendranath, Aiken, Alvain, Barlow, Bouman, Bracher, Brewin, Bricaud, Brown, Ciotti, Clementson, Craig, Devred, Hardman-Mountford, Hirata, Hu, Kostadinov, Lavender, Loisel, Moore, Morales, Mouw, Nair, Raitsos, Roesler, Shutler, Sosik, Soto, Stuart, Subramaniam, and Uitz</label><?label Sathyendranath2014?><mixed-citation>Sathyendranath, S., Aiken, J., Alvain, S., Barlow, R., Bouman, H., Bracher, A., Brewin, R., Bricaud, A., Brown, C. W., Ciotti, A. M., Clementson, L. A., Craig, S. E., Devred, E., Hardman-Mountford, N., Hirata, T., Hu, C., Kostadinov, T. S., Lavender, S., Loisel, H., Moore, T. S., Morales, J., Mouw, C. B., Nair, A., Raitsos, D., Roesler, C., Shutler, J. D., Sosik, H. M., Soto, I., Stuart, V., Subramaniam, A., and Uitz, J.: Phytoplankton functional types from Space, International Ocean-Colour Coordinating Group, Dartmouth, Nova Scotia, B2Y 4A2, Canada, ioccg; 15 Edn., <uri>https://epic.awi.de/id/eprint/36000/</uri> (last access: 26 Janaury 2024), 2014.</mixed-citation></ref>
      <ref id="bib1.bibx69"><?xmltex \def\ref@label{{Sawadogo et~al.(2009)Sawadogo, Brajard, Niang, Lathuiliere, Crepon,
and Thiria}}?><label>Sawadogo et al.(2009)Sawadogo, Brajard, Niang, Lathuiliere, Crepon, and Thiria</label><?label Sawadogo2009?><mixed-citation>Sawadogo, S., Brajard, J., Niang, A., Lathuiliere, C., Crepon, M., and Thiria, S.: Analysis of the Senegalo-Mauritanian upwelling by processing satellite remote sensing observations with topological maps, 2009 International Joint Conference on Neural Networks, Atlanta, GA, USA, 28260–2832, <ext-link xlink:href="https://doi.org/10.1109/IJCNN.2009.5178623" ext-link-type="DOI">10.1109/IJCNN.2009.5178623</ext-link>, 2009.</mixed-citation></ref>
      <ref id="bib1.bibx70"><?xmltex \def\ref@label{{Smith(2007)}}?><label>Smith(2007)</label><?label Smith2007?><mixed-citation>Smith, V. H.: Microbial diversity–productivity relationships in aquatic ecosystems, FEMS Microb. Ecol., 62, 181–186, <ext-link xlink:href="https://doi.org/10.1111/J.1574-6941.2007.00381.X" ext-link-type="DOI">10.1111/J.1574-6941.2007.00381.X</ext-link>, 2007.</mixed-citation></ref>
      <ref id="bib1.bibx71"><?xmltex \def\ref@label{{Sommeria-Klein et~al.(2021)Sommeria-Klein, Watteaux, Ibarbalz,
Karlusich, Iudicone, Bowler, and Morlon}}?><label>Sommeria-Klein et al.(2021)Sommeria-Klein, Watteaux, Ibarbalz, Karlusich, Iudicone, Bowler, and Morlon</label><?label Sommeria-Klein2021?><mixed-citation> Sommeria-Klein, G., Watteaux, R., Ibarbalz, F. M., Karlusich, J. J. P., Iudicone, D., Bowler, C., and Morlon, H.: Global drivers of eukaryotic plankton biogeography in the sunlit ocean, Science, 374, 594–599, 2021.</mixed-citation></ref>
      <ref id="bib1.bibx72"><?xmltex \def\ref@label{{Soppa et~al.(2014)Soppa, Hirata, Silva, Dinter, Peeken, Wiegmann, and
Bracher}}?><label>Soppa et al.(2014)Soppa, Hirata, Silva, Dinter, Peeken, Wiegmann, and Bracher</label><?label Soppa2014?><mixed-citation>Soppa, M. A., Hirata, T., Silva, B., Dinter, T., Peeken, I., Wiegmann, S., and Bracher, A.: Global retrieval of diatom abundance based on phytoplankton pigments and satellite data, Remote Sens., 6, 10089–10106, <ext-link xlink:href="https://doi.org/10.3390/rs61010089" ext-link-type="DOI">10.3390/rs61010089</ext-link>, 2014.</mixed-citation></ref>
      <ref id="bib1.bibx73"><?xmltex \def\ref@label{{Tara Ocean~Foundation(2022)}}?><label>Tara Ocean Foundation(2022)</label><?label tara2022priorities?><mixed-citation>Tara Ocean Foundation: Tara Oceans, European Molecular Biology Laboratory (EMBL), Priorities for ocean microbiome research, Nat. Microbiol., 7, 937–947, <ext-link xlink:href="https://doi.org/10.1038/s41564-022-01145-5" ext-link-type="DOI">10.1038/s41564-022-01145-5</ext-link>, 2022.</mixed-citation></ref>
      <ref id="bib1.bibx74"><?xmltex \def\ref@label{{Taylor et~al.(2011)Taylor, Torrecilla, Bernhardt, Taylor, Peeken,
R{\"{o}}ttgers, Piera, and Bracher}}?><label>Taylor et al.(2011)Taylor, Torrecilla, Bernhardt, Taylor, Peeken, Röttgers, Piera, and Bracher</label><?label taylor2011bio?><mixed-citation>Taylor, B. B., Torrecilla, E., Bernhardt, A., Taylor, M. H., Peeken, I., Röttgers, R., Piera, J., and Bracher, A.: Bio-optical provinces in the eastern Atlantic Ocean and their biogeographical relevance, Biogeosciences, 8, 3609–3629, <ext-link xlink:href="https://doi.org/10.5194/bg-8-3609-2011" ext-link-type="DOI">10.5194/bg-8-3609-2011</ext-link>, 2011.</mixed-citation></ref>
      <?pagebreak page239?><ref id="bib1.bibx75"><?xmltex \def\ref@label{{Tilman et~al.(2014)Tilman, Isbell, and Cowles}}?><label>Tilman et al.(2014)Tilman, Isbell, and Cowles</label><?label Tilman2014?><mixed-citation>Tilman, D., Isbell, F., and Cowles, J. M.: Biodiversity and Ecosystem Functioning, Annu. Rev. Ecol. Evol. Syst, 45, 471–493, <ext-link xlink:href="https://doi.org/10.1146/annurev-ecolsys-120213-091917" ext-link-type="DOI">10.1146/annurev-ecolsys-120213-091917</ext-link>, 2014.</mixed-citation></ref>
      <ref id="bib1.bibx76"><?xmltex \def\ref@label{{Torrecilla et~al.(2011)Torrecilla, Stramski, Reynolds,
Mill{\'{a}}n-N{\'{u}}{\~{n}}ez, and Piera}}?><label>Torrecilla et al.(2011)Torrecilla, Stramski, Reynolds, Millán-Núñez, and Piera</label><?label torrecilla2011cluster?><mixed-citation> Torrecilla, E., Stramski, D., Reynolds, R. A., Millán-Núñez, E., and Piera, J.: Cluster analysis of hyperspectral optical data for discriminating phytoplankton pigment assemblages in the open ocean, Remote Sens. Environ., 115, 2578–2593, 2011.</mixed-citation></ref>
      <ref id="bib1.bibx77"><?xmltex \def\ref@label{{Uitz et~al.(2006)Uitz, Claustre, Morel, and Hooker}}?><label>Uitz et al.(2006)Uitz, Claustre, Morel, and Hooker</label><?label Uitz2006?><mixed-citation>Uitz, J., Claustre, H., Morel, A., and Hooker, S. B.: Vertical distribution of phytoplankton communities in open ocean: An assessment based on surface chlorophyll, J. Geophys. Res., 111, C08005, <ext-link xlink:href="https://doi.org/10.1029/2005jc003207" ext-link-type="DOI">10.1029/2005jc003207</ext-link>, 2006.</mixed-citation></ref>
      <ref id="bib1.bibx78"><?xmltex \def\ref@label{{Vidussi et~al.(2001)Vidussi, Claustre, Manca, Luchetta, and
Marty}}?><label>Vidussi et al.(2001)Vidussi, Claustre, Manca, Luchetta, and Marty</label><?label Vidussi2001?><mixed-citation>Vidussi, F., Claustre, H., Manca, B. B., Luchetta, A., and Marty, J.-C.: Phytoplankton pigment distribution in relation to upper thermocline circulation in the eastern Mediterranean Sea during winter, J.   Geophys. Res.-Oceans, 106, 19939–19956, <ext-link xlink:href="https://doi.org/10.1029/1999JC000308" ext-link-type="DOI">10.1029/1999JC000308</ext-link>, 2001.</mixed-citation></ref>
      <ref id="bib1.bibx79"><?xmltex \def\ref@label{{Werdell and Bailey(2005)}}?><label>Werdell and Bailey(2005)</label><?label Werdell2005?><mixed-citation>Werdell, P. J. and Bailey, S. W.: An improved in-situ bio-optical data set for ocean color algorithm development and satellite data product validation, Remote Sens. Environ., 98, 122–140, <ext-link xlink:href="https://doi.org/10.1016/j.rse.2005.07.001" ext-link-type="DOI">10.1016/j.rse.2005.07.001</ext-link>, 2005. </mixed-citation></ref><?xmltex \hack{\newpage}?>
      <ref id="bib1.bibx80"><?xmltex \def\ref@label{{Werdell et~al.(2003)Werdell, Bailey, Fargion, Pietras, Knobelspiesse,
Feldman, and McClain}}?><label>Werdell et al.(2003)Werdell, Bailey, Fargion, Pietras, Knobelspiesse, Feldman, and McClain</label><?label werdell2003unique?><mixed-citation> Werdell, P. J., Bailey, S., Fargion, G., Pietras, C., Knobelspiesse, K., Feldman, G., and McClain, C.: Unique data repository facilitates ocean color satellite validation, Eos, 84, 377–387, 2003.</mixed-citation></ref>
      <ref id="bib1.bibx81"><?xmltex \def\ref@label{{Wright et~al.(2010)Wright, van~den Enden, Pearce, Davidson, Scott,
and Westwood}}?><label>Wright et al.(2010)Wright, van den Enden, Pearce, Davidson, Scott, and Westwood</label><?label wright2010phytoplankton?><mixed-citation>Wright, S. W., van den Enden, R. L., Pearce, I., Davidson, A. T., Scott, F. J., and Westwood, K. J.: Phytoplankton community structure and stocks in the Southern Ocean (30–80<inline-formula><mml:math id="M241" display="inline"><mml:msup><mml:mi/><mml:mo>∘</mml:mo></mml:msup></mml:math></inline-formula> E) determined by CHEMTAX analysis of HPLC pigment signatures, Deep-Sea Res. Pt. II, 57, 758–778, 2010.</mixed-citation></ref>
      <ref id="bib1.bibx82"><?xmltex \def\ref@label{{Xi et~al.(2020)Xi, Losa, Mangin, Soppa, Garnesson, Demaria, Liu,
D'Andon, and Bracher}}?><label>Xi et al.(2020)Xi, Losa, Mangin, Soppa, Garnesson, Demaria, Liu, D'Andon, and Bracher</label><?label Xi2020?><mixed-citation>Xi, H., Losa, S. N., Mangin, A., Soppa, M. A., Garnesson, P., Demaria, J., Liu, Y., D'Andon, O. H. F., and Bracher, A.: Global retrieval of phytoplankton functional types based on empirical orthogonal functions using CMEMS GlobColour merged products and further extension to OLCI data, Remote Sens. Environ., 240, 111704, <ext-link xlink:href="https://doi.org/10.1016/J.RSE.2020.111704" ext-link-type="DOI">10.1016/J.RSE.2020.111704</ext-link>, 2020.</mixed-citation></ref>
      <ref id="bib1.bibx83"><?xmltex \def\ref@label{{Xi et~al.(2021)Xi, Losa, Mangin, Garnesson, Bretagnon, Demaria,
Soppa, Hembise Fanton~d'Andon, and Bracher}}?><label>Xi et al.(2021)Xi, Losa, Mangin, Garnesson, Bretagnon, Demaria, Soppa, Hembise Fanton d'Andon, and Bracher</label><?label xi2021global?><mixed-citation>Xi, H., Losa, S. N., Mangin, A., Garnesson, P., Bretagnon, M., Demaria, J., Soppa, M. A., Hembise Fanton d'Andon, O., and Bracher, A.: Global chlorophyll a concentrations of phytoplankton functional types with detailed uncertainty assessment using multisensor ocean color and sea surface temperature satellite products, J. Geophys. Res.-Oceans, 126, e2020JC017127, <ext-link xlink:href="https://doi.org/10.1029/2020JC017127" ext-link-type="DOI">10.1029/2020JC017127</ext-link>, 2021.</mixed-citation></ref>

  </ref-list></back>
    <!--<article-title-html>Linking satellites to genes with machine learning to estimate phytoplankton community structure from space</article-title-html>
<abstract-html/>
<ref-html id="bib1.bib1"><label>ACRI-ST(2019)</label><mixed-citation>
      
ACRI-ST: The European Service for Ocean Colour – GlobColour, [data set], <a href="https://hermes.acri.fr/index.php" target="_blank"/> (last access: 29 January 2024), 2019.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib2"><label>Agustí(1991)</label><mixed-citation>
      
Agustí, S.: Allometric Scaling of Light Absorption and Scattering by
Phytoplankton Cells, Can. J. Fish. Aquat. Sci., 48,
763–767, <a href="https://doi.org/10.1139/f91-091" target="_blank">https://doi.org/10.1139/f91-091</a>, 1991.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib3"><label>Alvain et al.(2005)Alvain, Moulin, Dandonneau, and
Bréon</label><mixed-citation>
      
Alvain, S., Moulin, C., Dandonneau, Y., and Bréon, F.: Remote sensing of
phytoplankton groups in case 1 waters from global SeaWiFS imagery, Deep-Sea Res. Pt. I, 52, 1989–2004,
<a href="https://doi.org/10.1016/j.dsr.2005.06.015" target="_blank">https://doi.org/10.1016/j.dsr.2005.06.015</a>, 2005.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib4"><label>Alvain et al.(2006)Alvain, Moulin, Dandonneau, Loisel, and
Bréon</label><mixed-citation>
      
Alvain, S., Moulin, C., Dandonneau, Y., Loisel, H., and Bréon, F. M.: A
species-dependent bio-optical model of case I waters for global ocean color
processing, Deep-Sea Res. Pt. I, 53,
917–925, <a href="https://doi.org/10.1016/j.dsr.2006.01.011" target="_blank">https://doi.org/10.1016/j.dsr.2006.01.011</a>, 2006.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib5"><label>Ben Mustapha et al.(2013)Ben Mustapha, Alvain, Jamet, Loisel, and
Dessailly</label><mixed-citation>
      
Ben Mustapha, Z., Alvain, S., Jamet, C., Loisel, H., and Dessailly, D.:
Automatic classification of water-leaving radiance anomalies from global
SeaWiFS imagery: Application to the detection of phytoplankton groups in open
ocean waters, Remote Sens. Environ., 146, 97–112,
<a href="https://doi.org/10.1016/j.rse.2013.08.046" target="_blank">https://doi.org/10.1016/j.rse.2013.08.046</a>, 2013.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib6"><label>Bock et al.(2022)Bock, Subramaniam, Juhl, Montoya, and
Duhamel</label><mixed-citation>
      
Bock, N., Subramaniam, A., Juhl, A. R., Montoya, J., and Duhamel, S.:
Quantifying per-cell chlorophyll <i>a</i> in natural picophytoplankton populations
using fluorescence-activated cell sorting, Front. Mar. Sci., 9,
850646, <a href="https://doi.org/10.3389/fmars.2022.850646" target="_blank">https://doi.org/10.3389/fmars.2022.850646</a>, 2022.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib7"><label>Bowler and Pierella Karlusich(2022)</label><mixed-citation>
      
Bowler, C. and Pierella Karlusich, J. J.: A robust approach to estimate relative phytoplankton cell abundances from metagenomes,  BioStudies, S-BSST761,  [data set], <a href="https://www.ebi.ac.uk/biostudies/studies/S-BSST761" target="_blank"/> (last access: 29 January 2024), 2022.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib8"><label>Bracher(2015a)</label><mixed-citation>
      
Bracher, A., Taylor, M. H., Taylor, B., Dinter, T., Röttgers, R., and Steinmetz, F.: Using empirical orthogonal functions derived from remote-sensing reflectance for the prediction of phytoplankton pigment concentrations, Ocean Sci., 11, 139–158, <a href="https://doi.org/10.5194/os-11-139-2015" target="_blank">https://doi.org/10.5194/os-11-139-2015</a>, 2015a.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib9"><label>Bracher(2015b)</label><mixed-citation>
      
Bracher, A., Taylor, M. H., Taylor, B., Dinter, T., Röttgers, R., and Steinmetz, F.: Using empirical orthogonal functions derived from remote-sensing reflectance for the prediction of phytoplankton pigment concentrations, Ocean Sci., 11, 139–158, <a href="https://doi.org/10.5194/os-11-139-2015" target="_blank">https://doi.org/10.5194/os-11-139-2015</a>, 2015b.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib10"><label>Bracher(2015c)</label><mixed-citation>
      
Bracher, A., Taylor, M. H., Taylor, B., Dinter, T., Röttgers, R., and Steinmetz, F.: Using empirical orthogonal functions derived from remote-sensing reflectance for the prediction of phytoplankton pigment concentrations, Ocean Sci., 11, 139–158, <a href="https://doi.org/10.5194/os-11-139-2015" target="_blank">https://doi.org/10.5194/os-11-139-2015</a>, 2015c.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib11"><label>Brewin et al.(2015)Brewin, Sathyendranath, Jackson, Barlow, Brotas,
Airs, and Lamont</label><mixed-citation>
      
Brewin, R. J., Sathyendranath, S., Jackson, T., Barlow, R., Brotas, V., Airs,
R., and Lamont, T.: Influence of light in the mixed-layer on the parameters
of a three-component model of phytoplankton size class, Remote Sens.   Environ., 168, 437–450, <a href="https://doi.org/10.1016/J.RSE.2015.07.004" target="_blank">https://doi.org/10.1016/J.RSE.2015.07.004</a>, 2015.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib12"><label>Brewin et al.(2017)Brewin, Ciavatta, Sathyendranath, Jackson,
Tilstone, Curran, Airs, Cummings, Brotas, Organelli, Dall'Olmo, and
Raitsos</label><mixed-citation>
      
Brewin, R. J., Ciavatta, S., Sathyendranath, S., Jackson, T., Tilstone, G.,
Curran, K., Airs, R. L., Cummings, D., Brotas, V., Organelli, E., Dall'Olmo,
G., and Raitsos, D. E.: Uncertainty in ocean-color estimates of chlorophyll
for phytoplankton groups, Front. Mar. Sci., 4, 104,
<a href="https://doi.org/10.3389/FMARS.2017.00104/BIBTEX" target="_blank">https://doi.org/10.3389/FMARS.2017.00104/BIBTEX</a>, 2017.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib13"><label>Brewin et al.(2010)Brewin, Sathyendranath, Hirata, Lavender,
Barciela, and Hardman-Mountford</label><mixed-citation>
      
Brewin, R. J. W., Sathyendranath, S., Hirata, T., Lavender, S. J., Barciela,
R. M., and Hardman-Mountford, N. J.: A three-component model of
phytoplankton size class for the Atlantic Ocean, Ecol. Model., 221,
1472–1483, <a href="https://doi.org/10.1016/j.ecolmodel.2010.02.014" target="_blank">https://doi.org/10.1016/j.ecolmodel.2010.02.014</a>, 2010.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib14"><label>Brewin et al.(2014)Brewin, Sathyendranath, Tilstone, Lange, and
Platt</label><mixed-citation>
      
Brewin, R. J. W., Sathyendranath, S., Tilstone, G., Lange, P. K., and Platt,
T.: A multicomponent model of phytoplankton size structure, J.
Geophys. Res.-Oceans, 119, 3478–3496, <a href="https://doi.org/10.1002/2014JC009859" target="_blank">https://doi.org/10.1002/2014JC009859</a>,
2014.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib15"><label>Brown(1995)</label><mixed-citation>
      
Brown, C.: Global Distribution of Coccolithophore Blooms, Oceanography, 8,
59–60, <a href="https://doi.org/10.5670/oceanog.1995.21" target="_blank">https://doi.org/10.5670/oceanog.1995.21</a>, 1995.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib16"><label>Brun et al.(2015)Brun, Vogt, Payne, Gruber, O'brien, Buitenhuis,
Le Quéré, Leblanc, and Luo</label><mixed-citation>
      
Brun, P., Vogt, M., Payne, M. R., Gruber, N., O'Brien, C. J., Buitenhuis,
E. T., Le Quéré, C., Leblanc, K., and Luo, Y.-W.: Ecological niches
of open ocean phytoplankton taxa, Limnol. Oceanogr., 60, 1020–1038,
2015.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib17"><label>Charantonis et al.(2015)Charantonis, Testor, Mortier, D’Ortenzio,
and Thiria</label><mixed-citation>
      
Charantonis, A. A., Testor, P., Mortier, L., D'Ortenzio, F., and Thiria, S.:
Completion of a sparse GLIDER database using multi-iterative Self-Organizing
Maps (ITCOMP SOM), Procedia Comput. Sci., 51, 2198–2206, 2015.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib18"><label>Chase et al.(2020)Chase, Kramer, Haëntjens, Boss, Karp-Boss,
Edmondson, and Graff</label><mixed-citation>
      
Chase, A. P., Kramer, S. J., Haëntjens, N., Boss, E. S., Karp-Boss, L.,
Edmondson, M., and Graff, J. R.: Evaluation of diagnostic pigments to
estimate phytoplankton size classes, Limnol. Oceanogr.-Meth.,
18, 570–584, <a href="https://doi.org/10.1002/LOM3.10385" target="_blank">https://doi.org/10.1002/LOM3.10385</a>, 2020.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib19"><label>Chisholm(1992)</label><mixed-citation>
      
Chisholm, S. W.: Phytoplankton Size, in: Primary Productivity and Biogeochemical Cycles in the Sea,
edited by: Falkowski, P. G., Woodhead, A. D., Vivirito, K.,
Environmental Science Research, vol 43.
Springer, Boston, MA, <a href="https://doi.org/10.1007/978-1-4899-0762-2_12" target="_blank">https://doi.org/10.1007/978-1-4899-0762-2_12</a>, 1962.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib20"><label>da Silva and Costa(2013)</label><mixed-citation>
      
da Silva, L. E. B. and Costa, J. A. F.: Clustering, noise reduction and
visualization using features extracted from the self-organizing map, in:
Intelligent Data Engineering and Automated Learning–IDEAL 2013: 14th
International Conference, IDEAL 2013, Hefei, China, 20–23  October 2013,
Proceedings 14, Springer, 242–251, 2013.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib21"><label>Dairiki et al.(2020)Dairiki, Motokawa, Murata, and
Taguchi</label><mixed-citation>
      
Dairiki, C., Motokawa, S., Murata, A., and Taguchi, S.: How does cell volume
influence the total light absorption efficiency of a mixed population of
dinoflagellates with similar cell shapes and pigment compositions?, Plankton
Benthos Res., 15, 250–258, 2020.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib22"><label>Dandonneau et al.(2004)Dandonneau, Deschamps, Nicolas, Loisel,
Blanchot, Montel, Thieuleux, and Bécu</label><mixed-citation>
      
Dandonneau, Y., Deschamps, P.-Y., Nicolas, J.-M., Loisel, H., Blanchot, J.,
Montel, Y., Thieuleux, F., and Bécu, G.: Seasonal and interannual
variability of ocean color and composition of phytoplankton communities in
the North Atlantic, equatorial Pacific and South Pacific, Deep-Sea Res. Pt. I, 51, 303–318,
<a href="https://doi.org/10.1016/j.dsr2.2003.07.018" target="_blank">https://doi.org/10.1016/j.dsr2.2003.07.018</a>, 2004.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib23"><label>de Salas et al.(2011)de Salas, Eriksen, Davidson, and
Wright</label><mixed-citation>
      
de Salas, M. F., Eriksen, R., Davidson, A. T., and Wright, S. W.: Protistan
communities in the Australian sector of the Sub-Antarctic Zone during
SAZ-Sense, Deep-Sea Res. Pt. II, 58,
2135–2149, 2011.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib24"><label>Di Cicco et al.(2017)Di Cicco, Sammartino, Marullo, and
Santoleri</label><mixed-citation>
      
Di Cicco, A., Sammartino, M., Marullo, S., and Santoleri, R.: Regional
Empirical Algorithms for an Improved Identification of Phytoplankton
Functional Types and Size Classes in the Mediterranean Sea Using Satellite
Data, Front. Mar. Sci., 4, 126, <a href="https://doi.org/10.3389/fmars.2017.00126" target="_blank">https://doi.org/10.3389/fmars.2017.00126</a>,
2017.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib25"><label>Dutkiewicz et al.(2020)Dutkiewicz, Cermeno, Jahn, Follows, Hickman,
Taniguchi, and Ward</label><mixed-citation>
      
Dutkiewicz, S., Cermeno, P., Jahn, O., Follows, M. J., Hickman, A. E., Taniguchi, D. A. A., and Ward, B. A.: Dimensions of marine phytoplankton diversity, Biogeosciences, 17, 609–634, <a href="https://doi.org/10.5194/bg-17-609-2020" target="_blank">https://doi.org/10.5194/bg-17-609-2020</a>, 2020.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib26"><label>El Hourany(2024)</label><mixed-citation>
      
El Hourany, R.: Satellite-derived phytoplankton community structure from space using psbO and machine learning,  Zenodo [code], <a href="https://doi.org/10.5281/zenodo.10571578" target="_blank">https://doi.org/10.5281/zenodo.10571578</a>, 2024.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib27"><label>El Hourany et al.(2019a)El Hourany, Abboud-Abi
Saab, Faour, Aumont, Crépon, and Thiria</label><mixed-citation>
      
El Hourany, R., Abboud-Abi Saab, M., Faour, G., Aumont, O., Crépon,
M., and Thiria, S.: Estimation of secondary phytoplankton pigments from
satellite observations using self-organizing maps (SOM), J. Geophys. Res.-Oceans, 124, 1357–1378, <a href="https://doi.org/10.1029/2018JC014450" target="_blank">https://doi.org/10.1029/2018JC014450</a>, 2019a.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib28"><label>El Hourany et al.(2019b)El Hourany, Abboud-Abi
Saab, Faour, Mejia, Crépon, and Thiria</label><mixed-citation>
      
El Hourany, R., Abboud-Abi Saab, M., Faour, G., Mejia, C., Crépon,
M., and Thiria, S.: Phytoplankton Diversity in the Mediterranean Sea From
Satellite Data Using Self-Organizing Maps, J. Geophys. Res.-Oceans, 124, 5827–5843, <a href="https://doi.org/10.1029/2019JC015131" target="_blank">https://doi.org/10.1029/2019JC015131</a>, 2019b.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib29"><label>El Hourany et al.(2021)El Hourany, Mejia, Faour, Crépon,
and Thiria</label><mixed-citation>
      
El Hourany, R., Mejia, C., Faour, G., Crépon, M., and Thiria, S.:
Evidencing the Impact of Climate Change on the Phytoplankton Community of
the Mediterranean Sea Through a Bioregionalization Approach, J. Geophys. Res.-Oceans, 126, e2020JC016808,
<a href="https://doi.org/10.1029/2020JC016808" target="_blank">https://doi.org/10.1029/2020JC016808</a>, 2021.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib30"><label>El Hourany et al.(2024)</label><mixed-citation>
      
El Hourany, R., Pierella Karlusich, J. J., Zinger, L., Loisel, H., Levy, M., and Bowler, C.:  Linking satellites to genes with machine learning to estimate phytoplankton community structure from space,  Zenodo [data set], <a href="https://doi.org/10.5281/zenodo.10361485" target="_blank">https://doi.org/10.5281/zenodo.10361485</a>, 2024.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib31"><label>ESA SST CCI and C3S reprocessed sea surface temperature analyses(2019)</label><mixed-citation>
      
ESA SST CCI and C3S reprocessed sea surface temperature analyses:
E.U. Copernicus Marine Service Information (CMEMS), Marine Data Store (MDS),
[data set], <a href="https://doi.org/10.48670/moi-00169" target="_blank">https://doi.org/10.48670/moi-00169</a>, 2019.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib32"><label>Flombaum et al.(2013)Flombaum, Gallegos, Gordillo, Rincon, Zabala,
Jiao, Karl, Li, Lomas, Veneziano, Vera, Vrugt, and Martiny</label><mixed-citation>
      
Flombaum, P., Gallegos, J. L., Gordillo, R. A., Rincon, J., Zabala, L. L.,
Jiao, N., Karl, D. M., Li, W. K. W., Lomas, M. W., Veneziano, D., Vera,
C. S., Vrugt, J. A., and Martiny, A. C.: Present and future global
distributions of the marine Cyanobacteria Prochlorococcus and Synechococcus,
P. Natl. Acad. Sci. USA, 110, 9824–9829,
<a href="https://doi.org/10.1073/pnas.1307701110" target="_blank">https://doi.org/10.1073/pnas.1307701110</a>, 2013.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib33"><label>Folguera et al.(2015)Folguera, Zupan, Cicerone, and
Magallanes</label><mixed-citation>
      
Folguera, L., Zupan, J., Cicerone, D., and Magallanes, J. F.: Self-organizing
maps for imputation of missing data in incomplete data matrices, Chemometr. Intell. Lab., 143, 146–151, 2015.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib34"><label>Fuhrman(2009)</label><mixed-citation>
      
Fuhrman, J. A.: Microbial community structure and its functional
implications, 459, 193–199, <a href="https://doi.org/10.1038/nature08058" target="_blank">https://doi.org/10.1038/nature08058</a>, 2009.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib35"><label>Fujiki and Taguchi(2002)</label><mixed-citation>
      
Fujiki, T. and Taguchi, S.: Variability in chlorophyll a specific absorption
coefficient in marine phytoplankton as a function of cell size and
irradiance, J. Plankton Res., 24, 859–874, 2002.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib36"><label>Guidi et al.(2009)Guidi, Stemmann, Jackson, Ibanez, Claustre,
Legendre, Picheral, and Gorsky</label><mixed-citation>
      
Guidi, L., Stemmann, L., Jackson, G. A., Ibanez, F., Claustre, H., Legendre,
L., Picheral, M., and Gorsky, G.: Effects of phytoplankton community on
production, size, and export of large aggregates: A world-ocean analysis,
Limnol. Oceanogr., 54, 1951–1963, <a href="https://doi.org/10.4319/LO.2009.54.6.1951" target="_blank">https://doi.org/10.4319/LO.2009.54.6.1951</a>,
2009.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib37"><label>Henson et al.(2021)Henson, Cael, Allen, and Dutkiewicz</label><mixed-citation>
      
Henson, S. A., Cael, B. B., Allen, S. R., and Dutkiewicz, S.: Future
phytoplankton diversity in a changing climate, Nat. Commun.,
12, 1–8, <a href="https://doi.org/10.1038/s41467-021-25699-w" target="_blank">https://doi.org/10.1038/s41467-021-25699-w</a>, 2021.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib38"><label>Hillebrand and Azovsky(2001)</label><mixed-citation>
      
Hillebrand, H. and Azovsky, A. I.: Body size determines the strength of the
latitudinal diversity gradient, Ecography, 24, 251–256,
<a href="https://doi.org/10.1034/J.1600-0587.2001.240302.X" target="_blank">https://doi.org/10.1034/J.1600-0587.2001.240302.X</a>, 2001.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib39"><label>Hirata et al.(2008)Hirata, Aiken, Hardman-Mountford, Smyth, and
Barlow</label><mixed-citation>
      
Hirata, T., Aiken, J., Hardman-Mountford, N., Smyth, T., and Barlow, R.: An
absorption model to determine phytoplankton size classes from satellite ocean
colour, Remote Sens. Environ., 112, 3153–3159,
<a href="https://doi.org/10.1016/J.RSE.2008.03.011" target="_blank">https://doi.org/10.1016/J.RSE.2008.03.011</a>, 2008.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib40"><label>Hirata et al.(2011)Hirata, Hardman-Mountford, Brewin, Aiken, Barlow,
Suzuki, Isada, Howell, Hashioka, Noguchi-Aita, and Yamanaka</label><mixed-citation>
      
Hirata, T., Hardman-Mountford, N. J., Brewin, R. J. W., Aiken, J., Barlow, R., Suzuki, K., Isada, T., Howell, E., Hashioka, T., Noguchi-Aita, M., and Yamanaka, Y.: Synoptic relationships between surface Chlorophyll-<i>a</i> and diagnostic pigments specific to phytoplankton functional types, Biogeosciences, 8, 311–327, <a href="https://doi.org/10.5194/bg-8-311-2011" target="_blank">https://doi.org/10.5194/bg-8-311-2011</a>, 2011.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib41"><label>Hood et al.(2006)Hood, Laws, Armstrong, Bates, Brown, Carlson, Chai,
Doney, Falkowski, Feely, Friedrichs, Landry, Keith Moore, Nelson,
Richardson, Salihoglu, Schartau, Toole, and Wiggert</label><mixed-citation>
      
Hood, R. R., Laws, E. A., Armstrong, R. A., Bates, N. R., Brown, C. W.,
Carlson, C. A., Chai, F., Doney, S. C., Falkowski, P. G., Feely, R. A.,
Friedrichs, M. A., Landry, M. R., Keith Moore, J., Nelson, D. M.,
Richardson, T. L., Salihoglu, B., Schartau, M., Toole, D. A., and Wiggert,
J. D.: Pelagic functional group modeling: Progress, challenges and
prospects, Deep-Sea Res. Pt. II, 53,
459–512, <a href="https://doi.org/10.1016/J.DSR2.2006.01.025" target="_blank">https://doi.org/10.1016/J.DSR2.2006.01.025</a>, 2006.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib42"><label>Ibarbalz et al.(2019)Ibarbalz, Henry, Brandão, Martini,
Busseni, Byrne, Coelho, Endo, Gasol, Gregory, Mahé, Rigonato,
Royo-Llonch, Salazar, Sanz-Sáez, Scalco, Soviadan, Zayed, Zingone,
Labadie, Ferland, Marec, Kandels, Picheral, Dimier, Poulain, Pisarev,
Carmichael, Pesant, Acinas, Babin, Bork, Boss, Bowler, Cochrane, de Vargas,
Follows, Gorsky, Grimsley, Guidi, Hingamp, Iudicone, Jaillon, Karp-Boss,
Karsenti, Not, Ogata, Poulton, Raes, Sardet, Speich, Stemmann, Sullivan,
Sunagawa, Wincker, Pelletier, Bopp, Lombard, and Zinger</label><mixed-citation>
      
Ibarbalz, F. M., Henry, N., Brandão, M. C., Martini, S., Busseni, G.,
Byrne, H., Coelho, L. P., Endo, H., Gasol, J. M., Gregory, A. C., Mahé,
F., Rigonato, J., Royo-Llonch, M., Salazar, G., Sanz-Sáez, I., Scalco,
E., Soviadan, D., Zayed, A. A., Zingone, A., Labadie, K., Ferland, J., Marec,
C., Kandels, S., Picheral, M., Dimier, C., Poulain, J., Pisarev, S.,
Carmichael, M., Pesant, S., Acinas, S. G., Babin, M., Bork, P., Boss, E.,
Bowler, C., Cochrane, G., de Vargas, C., Follows, M., Gorsky, G., Grimsley,
N., Guidi, L., Hingamp, P., Iudicone, D., Jaillon, O., Karp-Boss, L.,
Karsenti, E., Not, F., Ogata, H., Poulton, N., Raes, J., Sardet, C., Speich,
S., Stemmann, L., Sullivan, M. B., Sunagawa, S., Wincker, P., Pelletier, E.,
Bopp, L., Lombard, F., and Zinger, L.: Global Trends in Marine Plankton
Diversity across Kingdoms of Life, Cell, 179, 1084–1097,
<a href="https://doi.org/10.1016/J.CELL.2019.10.008" target="_blank">https://doi.org/10.1016/J.CELL.2019.10.008</a>, 2019.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib43"><label>Iglesias-Rodríguez et al.(2002)Iglesias-Rodríguez, Brown,
Doney, Kleypas, Kolber, Kolber, Hayes, and
Falkowski</label><mixed-citation>
      
Iglesias-Rodríguez, M. D., Brown, C. W., Doney, S. C., Kleypas, J.,
Kolber, D., Kolber, Z., Hayes, P. K., and Falkowski, P. G.: Representing key
phytoplankton functional groups in ocean carbon cycle models:
Coccolithophorids, Global Biogeochem. Cy., 16, 1–20,
<a href="https://doi.org/10.1029/2001GB001454" target="_blank">https://doi.org/10.1029/2001GB001454</a>, 2002.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib44"><label>ilarinieminen(2012)</label><mixed-citation>
      
ilarinieminen: SOM Toolbox 2.1, [code], <a href="https://github.com/ilarinieminen/SOM-Toolbox/tree/master" target="_blank"/> (last access: 29 January 2024), 2012.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib45"><label>Irigoien et al.(2004)Irigoien, Hulsman, and Harris</label><mixed-citation>
      
Irigoien, X., Hulsman, J., and Harris, R. P.: Global biodiversity patterns of
marine phytoplankton and zooplankton, Nature, 429, 863–867,
<a href="https://doi.org/10.1038/nature02593" target="_blank">https://doi.org/10.1038/nature02593</a>, 2004.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib46"><label>Jouini et al.(2013)Jouini, Lévy, Crépon, and
Thiria</label><mixed-citation>
      
Jouini, M., Lévy, M., Crépon, M., and Thiria, S.: Reconstruction
of satellite chlorophyll images under heavy cloud coverage using a neural
classification method, Remote Sens. Environ., 131, 232–246,
<a href="https://doi.org/10.1016/j.rse.2012.11.025" target="_blank">https://doi.org/10.1016/j.rse.2012.11.025</a>, 2013.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib47"><label>Kaneko et al.(2023)Kaneko, Endo, Henry, Berney, Mahé, Poulain,
Labadie, Beluche, El Hourany et al.</label><mixed-citation>
      
Kaneko, H., Endo, H., Henry, N., Berney, C., Mahé, F., Poulain, J.,
Labadie, K., Beluche, O., El Hourany, R., et al.: Predicting global
distributions of eukaryotic plankton communities from satellite data, ISME
Commun., 3, 101, <a href="https://doi.org/10.1038/s43705-023-00308-7" target="_blank">https://doi.org/10.1038/s43705-023-00308-7</a>, 2023.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib48"><label>Le Quéré et al.(2005)Le Quéré, Harrison,
Colin Prentice, Buitenhuis, Aumont, Bopp, Claustre, Cotrim Da Cunha,
Geider, Giraud, Klaas, Kohfeld, Legendre, Manizza, Platt, Rivkin,
Sathyendranath, Uitz, Watson, and Wolf-Gladrow</label><mixed-citation>
      
Le Quéré, C., Harrison, S. P., Colin Prentice, I., Buitenhuis,
E. T., Aumont, O., Bopp, L., Claustre, H., Cotrim Da Cunha, L., Geider, R.,
Giraud, X., Klaas, C., Kohfeld, K. E., Legendre, L., Manizza, M., Platt, T.,
Rivkin, R. B., Sathyendranath, S., Uitz, J., Watson, A. J., and Wolf-Gladrow,
D.: Ecosystem dynamics based on plankton functional types for global ocean
biogeochemistry models, Global Change Biol., 11, 2016–2040,
<a href="https://doi.org/10.1111/j.1365-2486.2005.1004.x" target="_blank">https://doi.org/10.1111/j.1365-2486.2005.1004.x</a>, 2005.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib49"><label>Losa et al.(2017)Losa, Soppa, Dinter, Wolanin, Brewin, Bricaud,
Oelker, Peeken, Gentili, Rozanov et al.</label><mixed-citation>
      
Losa, S. N., Soppa, M. A., Dinter, T., Wolanin, A., Brewin, R. J., Bricaud, A.,
Oelker, J., Peeken, I., Gentili, B., Rozanov, V., et al.: Synergistic
exploitation of hyper-and multi-spectral precursor sentinel measurements to
determine phytoplankton functional types (SynSenPFT), Front. Mar.
Sci., 4, 203,
<a href="https://doi.org/10.3389/fmars.2017.00203" target="_blank">https://doi.org/10.3389/fmars.2017.00203</a>, 2017.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib50"><label>Mitchell et al.(1991)Mitchell, Brody, Holm-Hansen, McClain, and
Bishop</label><mixed-citation>
      
Mitchell, B. G., Brody, E. A., Holm-Hansen, O., McClain, C., and Bishop, J.:
Light limitation of phytoplankton biomass and macronutrient utilization in
the Southern Ocean, Limnol. Oceanogr., 36, 1662–1677,
<a href="https://doi.org/10.4319/lo.1991.36.8.1662" target="_blank">https://doi.org/10.4319/lo.1991.36.8.1662</a>, 1991.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib51"><label>O'Reilly et al.(1998)O'Reilly, Maritorena, Mitchell, Siegel, Carder,
Garver, Kahru, and McClain</label><mixed-citation>
      
O'Reilly, J. E., Maritorena, S., Mitchell, B. G., Siegel, D. a., Carder, K. L.,
Garver, S. a., Kahru, M., and McClain, C.: Ocean color chlorophyll
algorithms for SeaWiFS, J. Geophys. Res., 103, 24937,
<a href="https://doi.org/10.1029/98JC02160" target="_blank">https://doi.org/10.1029/98JC02160</a>, 1998.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib52"><label>Organelli et al.(2013)Organelli, Bricaud, Antoine, and
Uitz</label><mixed-citation>
      
Organelli, E., Bricaud, A., Antoine, D., and Uitz, J.: Multivariate approach
for the retrieval of phytoplankton size structure from measured light
absorption spectra in the Mediterranean Sea (BOUSSOLE site), Appl. Opt.,
52, 2257, <a href="https://doi.org/10.1364/AO.52.002257" target="_blank">https://doi.org/10.1364/AO.52.002257</a>, 2013.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib53"><label>Peloquin et al.(2013)Peloquin, Swan, Gruber, Vogt, Claustre, Ras,
Uitz, Barlow, Behrenfeld, Bidigare, Dierssen, Ditullio, Fernandez, Gallienne,
Gibb, Goericke, Harding, Head, Holligan, Hooker, Karl, Landry, Letelier,
Llewellyn, Lomas, Lucas, Mannino, Marty, Mitchell, Muller-Karger, Nelson,
Prezelin, Repeta, Smith Jr, Smythe-Wright, Stumpf, Subramaniam, Suzuki,
Trees, Vernet, Wasmund, and Wright</label><mixed-citation>
      
Peloquin, J., Swan, C., Gruber, N., Vogt, M., Claustre, H., Ras, J., Uitz, J., Barlow, R., Behrenfeld, M., Bidigare, R., Dierssen, H., Ditullio, G., Fernandez, E., Gallienne, C., Gibb, S., Goericke, R., Harding, L., Head, E., Holligan, P., Hooker, S., Karl, D., Landry, M., Letelier, R., Llewellyn, C. A., Lomas, M., Lucas, M., Mannino, A., Marty, J.-C., Mitchell, B. G., Muller-Karger, F., Nelson, N., O'Brien, C., Prezelin, B., Repeta, D., Jr. Smith, W. O., Smythe-Wright, D., Stumpf, R., Subramaniam, A., Suzuki, K., Trees, C., Vernet, M., Wasmund, N., and Wright, S.: The MAREDAT global database of high performance liquid chromatography marine pigment measurements, Earth Syst. Sci. Data, 5, 109–123, <a href="https://doi.org/10.5194/essd-5-109-2013" target="_blank">https://doi.org/10.5194/essd-5-109-2013</a>, 2013.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib54"><label>Pereira et al.(2013)Pereira, Ferrier, Walters, Geller, Jongman,
Scholes, Bruford, Brummitt, Butchart, Cardoso, Coops, Dulloo, Faith, Freyhof,
Gregory, Heip, Höft, Hurtt, Jetz, Karp, McGeoch, Obura, Onoda,
Pettorelli, Reyers, Sayre, Scharlemann, Stuart, Turak, Walpole, and
Wegmann</label><mixed-citation>
      
Pereira, H. M., Ferrier, S., Walters, M., Geller, G. N., Jongman, R. H.,
Scholes, R. J., Bruford, M. W., Brummitt, N., Butchart, S. H., Cardoso,
A. C., Coops, N. C., Dulloo, E., Faith, D. P., Freyhof, J., Gregory, R. D.,
Heip, C., Höft, R., Hurtt, G., Jetz, W., Karp, D. S., McGeoch, M. A.,
Obura, D., Onoda, Y., Pettorelli, N., Reyers, B., Sayre, R., Scharlemann,
J. P., Stuart, S. N., Turak, E., Walpole, M., and Wegmann, M.: Essential
biodiversity variables, Science, 339, 277–278, 2013.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib55"><label>Pesant et al.(2015)Pesant, Not, Picheral, Kandels-Lewis, Le Bescot,
Gorsky, Iudicone, Karsenti, Speich, Trouble, Dimier, and
Searson</label><mixed-citation>
      
Pesant, S., Not, F., Picheral, M., Kandels-Lewis, S., Le Bescot, N., Gorsky,
G., Iudicone, D., Karsenti, E., Speich, S., Trouble, R., Dimier, C., and
Searson, S.: Open science resources for the discovery and analysis of Tara
Oceans data, Sci. Data, 2, 1–16, <a href="https://doi.org/10.1038/SDATA.2015.23" target="_blank">https://doi.org/10.1038/SDATA.2015.23</a>, 2015.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib56"><label>Pierella Karlusich et al.(2020)Pierella Karlusich, Ibarbalz, and
Bowler</label><mixed-citation>
      
Pierella Karlusich, J. J., Ibarbalz, F. M., and Bowler, C.: Phytoplankton in
the Tara Ocean,  Annu. Rev. Mar. Sci., 12, 233–265, <a href="https://doi.org/10.1146/ANNUREV-MARINE-010419-010706" target="_blank">https://doi.org/10.1146/ANNUREV-MARINE-010419-010706</a>, 2020.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib57"><label>Pierella Karlusich et al.(2022)Pierella Karlusich, Pelletier,
Zinger, Lombard, Zingone, Colin, Gasol, Dorrell, Henry, Scalco, Acinas,
Wincker, de Vargas, and Bowler</label><mixed-citation>
      
Pierella Karlusich, J. J., Pelletier, E., Zinger, L., Lombard, F., Zingone,
A., Colin, S., Gasol, J. M., Dorrell, R. G., Henry, N., Scalco, E., Acinas,
S. G., Wincker, P., de Vargas, C., and Bowler, C.: A robust approach to
estimate relative phytoplankton cell abundances from metagenomes, Molec.
Ecol. Resour.,  23, 16–40, <a href="https://doi.org/10.1111/1755-0998.13592" target="_blank">https://doi.org/10.1111/1755-0998.13592</a>, 2022.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib58"><label>Powell and Glazier(2017)</label><mixed-citation>
      
Powell, M. G. and Glazier, D. S.: Asymmetric geographic range expansion
explains the latitudinal diversity gradients of four major taxa of marine
plankton, Paleobiology, 43, 196–208, <a href="https://doi.org/10.1017/PAB.2016.38" target="_blank">https://doi.org/10.1017/PAB.2016.38</a>, 2017.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib59"><label>Raven(1998)</label><mixed-citation>
      
Raven, J.: The twelfth Tansley Lecture. Small is beautiful: the
picophytoplankton, Funct. Ecol., 12, 503–513, 1998.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib60"><label>Rejeb et al.(2022)Rejeb, Duveau, and Rebafka</label><mixed-citation>
      
Rejeb, S., Duveau, C., and Rebafka, T.: Self-organizing maps for exploration of
partially observed data and imputation of missing values, Chemometr.   Intell. Lab., 231, 104653, <a href="https://doi.org/10.1016/j.chemolab.2022.104653" target="_blank">https://doi.org/10.1016/j.chemolab.2022.104653</a>, 2022.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib61"><label>Reygondeau et al.(2014)Reygondeau, Irisson, Ayata, Gasparini,
Benedetti, Albouy, Hattab, Guieu, and Koubbi.</label><mixed-citation>
      
Reygondeau, G., Irisson, J.-O., Ayata, S. D., Gasparini, S., Benedetti, F.,
Albouy, C., Hattab, T., Guieu, C., and Koubbi, P.: Definition of the
Mediterranean Eco-regions and Maps of Potential Pressures in These
Eco-regions, Tech. Rep., Perseus Deliverable 1,
<a href="http://www.perseus-net.eu/assets/media/PDF/deliverables/3336.6_Final.pdf" target="_blank"/> (last access: 26 Janaury 2024),
2014.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib62"><label>Richardson et al.(2003)Richardson, Risien, and
Shillington</label><mixed-citation>
      
Richardson, A. J., Risien, C., and Shillington, F. A.: Using self-organizing
maps to identify patterns in satellite imagery, Prog. Oceanogr.,
59, 223–239, <a href="https://doi.org/10.1016/j.pocean.2003.07.006" target="_blank">https://doi.org/10.1016/j.pocean.2003.07.006</a>, 2003.


    </mixed-citation></ref-html>
<ref-html id="bib1.bib63"><label>Righetti et al.(2019)Righetti, Vogt, Gruber, Psomas, and
Zimmermann</label><mixed-citation>
      
Righetti, D., Vogt, M., Gruber, N., Psomas, A., and Zimmermann, N. E.: Global
pattern of phytoplankton diversity driven by temperature and environmental
variability, Sci. Adv., 5, 6253–6268, 2019.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib64"><label>Rodríguez-Ramos et al.(2015)Rodríguez-Ramos,
Marañón, and Cermeño</label><mixed-citation>
      
Rodríguez-Ramos, T., Marañón, E., and Cermeño, P.:
Marine nano- and microphytoplankton diversity: redrawing global patterns
from sampling-standardized data, Global Ecol. Biogeogr., 24,
527–538, <a href="https://doi.org/10.1111/GEB.12274" target="_blank">https://doi.org/10.1111/GEB.12274</a>, 2015.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib65"><label>Rossi et al.(2014)Rossi, Ser-Giacomi, Lõpez, and
Hernández-García</label><mixed-citation>
      
Rossi, V., Ser-Giacomi, E., Lõpez, C., and Hernández-García,
E.: Hydrodynamic provinces and oceanic connectivity from a transport network
help designing marine reserves, Geophys. Res. Lett., 41,
2883–2891, <a href="https://doi.org/10.1002/2014GL059540" target="_blank">https://doi.org/10.1002/2014GL059540</a>, 2014.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib66"><label>Saitoh(2016)</label><mixed-citation>
      
Saitoh, F.: An ensemble model of self-organizing maps for imputation of missing
values, in: 2016 IEEE 9th International Workshop on Computational
Intelligence and Applications (IWCIA), 9–14,
<a href="https://doi.org/10.1109/IWCIA.2016.7805741" target="_blank">https://doi.org/10.1109/IWCIA.2016.7805741</a>, 2016.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib67"><label>Sarzeaud and Stephan(2000)</label><mixed-citation>
      
Sarzeaud, O. and Stephan, Y.: Data interpolation using Kohonen networks,
Proceedings of the International Joint Conference on Neural Networks, 6,
197–202, 2000.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib68"><label>Sathyendranath et al.(2014)Sathyendranath, Aiken, Alvain, Barlow,
Bouman, Bracher, Brewin, Bricaud, Brown, Ciotti, Clementson, Craig, Devred,
Hardman-Mountford, Hirata, Hu, Kostadinov, Lavender, Loisel, Moore, Morales,
Mouw, Nair, Raitsos, Roesler, Shutler, Sosik, Soto, Stuart, Subramaniam, and
Uitz</label><mixed-citation>
      
Sathyendranath, S., Aiken, J., Alvain, S., Barlow, R., Bouman, H., Bracher, A.,
Brewin, R., Bricaud, A., Brown, C. W., Ciotti, A. M., Clementson, L. A.,
Craig, S. E., Devred, E., Hardman-Mountford, N., Hirata, T., Hu, C.,
Kostadinov, T. S., Lavender, S., Loisel, H., Moore, T. S., Morales, J., Mouw,
C. B., Nair, A., Raitsos, D., Roesler, C., Shutler, J. D., Sosik, H. M.,
Soto, I., Stuart, V., Subramaniam, A., and Uitz, J.: Phytoplankton
functional types from Space, International Ocean-Colour Coordinating Group,
Dartmouth, Nova Scotia, B2Y 4A2, Canada, ioccg; 15 Edn.,
<a href="https://epic.awi.de/id/eprint/36000/" target="_blank"/> (last access: 26 Janaury 2024), 2014.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib69"><label>Sawadogo et al.(2009)Sawadogo, Brajard, Niang, Lathuiliere, Crepon,
and Thiria</label><mixed-citation>
      
Sawadogo, S., Brajard, J., Niang, A., Lathuiliere, C., Crepon, M., and Thiria,
S.: Analysis of the Senegalo-Mauritanian upwelling by processing satellite remote sensing observations
with topological maps, 2009 International Joint Conference on Neural Networks, Atlanta, GA, USA,
28260–2832, <a href="https://doi.org/10.1109/IJCNN.2009.5178623" target="_blank">https://doi.org/10.1109/IJCNN.2009.5178623</a>, 2009.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib70"><label>Smith(2007)</label><mixed-citation>
      
Smith, V. H.: Microbial diversity–productivity relationships in aquatic
ecosystems, FEMS Microb. Ecol., 62, 181–186,
<a href="https://doi.org/10.1111/J.1574-6941.2007.00381.X" target="_blank">https://doi.org/10.1111/J.1574-6941.2007.00381.X</a>, 2007.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib71"><label>Sommeria-Klein et al.(2021)Sommeria-Klein, Watteaux, Ibarbalz,
Karlusich, Iudicone, Bowler, and Morlon</label><mixed-citation>
      
Sommeria-Klein, G., Watteaux, R., Ibarbalz, F. M., Karlusich, J. J. P.,
Iudicone, D., Bowler, C., and Morlon, H.: Global drivers of eukaryotic
plankton biogeography in the sunlit ocean, Science, 374, 594–599,
2021.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib72"><label>Soppa et al.(2014)Soppa, Hirata, Silva, Dinter, Peeken, Wiegmann, and
Bracher</label><mixed-citation>
      
Soppa, M. A., Hirata, T., Silva, B., Dinter, T., Peeken, I., Wiegmann, S., and
Bracher, A.: Global retrieval of diatom abundance based on phytoplankton
pigments and satellite data, Remote Sens., 6, 10089–10106,
<a href="https://doi.org/10.3390/rs61010089" target="_blank">https://doi.org/10.3390/rs61010089</a>, 2014.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib73"><label>Tara Ocean Foundation(2022)</label><mixed-citation>
      
Tara Ocean Foundation: Tara Oceans, European Molecular Biology Laboratory (EMBL), Priorities for ocean microbiome research, Nat. Microbiol., 7, 937–947, <a href="https://doi.org/10.1038/s41564-022-01145-5" target="_blank">https://doi.org/10.1038/s41564-022-01145-5</a>, 2022.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib74"><label>Taylor et al.(2011)Taylor, Torrecilla, Bernhardt, Taylor, Peeken,
Röttgers, Piera, and Bracher</label><mixed-citation>
      
Taylor, B. B., Torrecilla, E., Bernhardt, A., Taylor, M. H., Peeken, I., Röttgers, R., Piera, J., and Bracher, A.: Bio-optical provinces in the eastern Atlantic Ocean and their biogeographical relevance, Biogeosciences, 8, 3609–3629, <a href="https://doi.org/10.5194/bg-8-3609-2011" target="_blank">https://doi.org/10.5194/bg-8-3609-2011</a>, 2011.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib75"><label>Tilman et al.(2014)Tilman, Isbell, and Cowles</label><mixed-citation>
      
Tilman, D., Isbell, F., and Cowles, J. M.: Biodiversity and Ecosystem
Functioning, Annu. Rev. Ecol. Evol. Syst, 45, 471–493,
<a href="https://doi.org/10.1146/annurev-ecolsys-120213-091917" target="_blank">https://doi.org/10.1146/annurev-ecolsys-120213-091917</a>, 2014.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib76"><label>Torrecilla et al.(2011)Torrecilla, Stramski, Reynolds,
Millán-Núñez, and Piera</label><mixed-citation>
      
Torrecilla, E., Stramski, D., Reynolds, R. A., Millán-Núñez, E.,
and Piera, J.: Cluster analysis of hyperspectral optical data for
discriminating phytoplankton pigment assemblages in the open ocean, Remote
Sens. Environ., 115, 2578–2593, 2011.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib77"><label>Uitz et al.(2006)Uitz, Claustre, Morel, and Hooker</label><mixed-citation>
      
Uitz, J., Claustre, H., Morel, A., and Hooker, S. B.: Vertical distribution of
phytoplankton communities in open ocean: An assessment based on surface
chlorophyll, J. Geophys. Res., 111, C08005, <a href="https://doi.org/10.1029/2005jc003207" target="_blank">https://doi.org/10.1029/2005jc003207</a>,
2006.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib78"><label>Vidussi et al.(2001)Vidussi, Claustre, Manca, Luchetta, and
Marty</label><mixed-citation>
      
Vidussi, F., Claustre, H., Manca, B. B., Luchetta, A., and Marty, J.-C.:
Phytoplankton pigment distribution in relation to upper thermocline
circulation in the eastern Mediterranean Sea during winter, J.   Geophys. Res.-Oceans, 106, 19939–19956,
<a href="https://doi.org/10.1029/1999JC000308" target="_blank">https://doi.org/10.1029/1999JC000308</a>, 2001.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib79"><label>Werdell and Bailey(2005)</label><mixed-citation>
      
Werdell, P. J. and Bailey, S. W.: An improved in-situ bio-optical data set for
ocean color algorithm development and satellite data product validation,
Remote Sens. Environ., 98, 122–140, <a href="https://doi.org/10.1016/j.rse.2005.07.001" target="_blank">https://doi.org/10.1016/j.rse.2005.07.001</a>,
2005.


    </mixed-citation></ref-html>
<ref-html id="bib1.bib80"><label>Werdell et al.(2003)Werdell, Bailey, Fargion, Pietras, Knobelspiesse,
Feldman, and McClain</label><mixed-citation>
      
Werdell, P. J., Bailey, S., Fargion, G., Pietras, C., Knobelspiesse, K.,
Feldman, G., and McClain, C.: Unique data repository facilitates ocean color
satellite validation, Eos, 84,
377–387, 2003.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib81"><label>Wright et al.(2010)Wright, van den Enden, Pearce, Davidson, Scott,
and Westwood</label><mixed-citation>
      
Wright, S. W., van den Enden, R. L., Pearce, I., Davidson, A. T., Scott, F. J.,
and Westwood, K. J.: Phytoplankton community structure and stocks in the
Southern Ocean (30–80°&thinsp;E) determined by CHEMTAX analysis of HPLC pigment
signatures, Deep-Sea Res. Pt. II, 57,
758–778, 2010.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib82"><label>Xi et al.(2020)Xi, Losa, Mangin, Soppa, Garnesson, Demaria, Liu,
D'Andon, and Bracher</label><mixed-citation>
      
Xi, H., Losa, S. N., Mangin, A., Soppa, M. A., Garnesson, P., Demaria, J., Liu,
Y., D'Andon, O. H. F., and Bracher, A.: Global retrieval of phytoplankton
functional types based on empirical orthogonal functions using CMEMS
GlobColour merged products and further extension to OLCI data, Remote
Sens. Environ., 240, 111704, <a href="https://doi.org/10.1016/J.RSE.2020.111704" target="_blank">https://doi.org/10.1016/J.RSE.2020.111704</a>, 2020.

    </mixed-citation></ref-html>
<ref-html id="bib1.bib83"><label>Xi et al.(2021)Xi, Losa, Mangin, Garnesson, Bretagnon, Demaria,
Soppa, Hembise Fanton d'Andon, and Bracher</label><mixed-citation>
      
Xi, H., Losa, S. N., Mangin, A., Garnesson, P., Bretagnon, M., Demaria, J.,
Soppa, M. A., Hembise Fanton d'Andon, O., and Bracher, A.: Global chlorophyll
a concentrations of phytoplankton functional types with detailed uncertainty
assessment using multisensor ocean color and sea surface temperature
satellite products, J. Geophys. Res.-Oceans, 126,
e2020JC017127, <a href="https://doi.org/10.1029/2020JC017127" target="_blank">https://doi.org/10.1029/2020JC017127</a>, 2021.

    </mixed-citation></ref-html>--></article>
