This study presents a method for estimating secondary phytoplankton pigments from satellite ocean color observations. We first compiled a large training data set composed of 12,000 samples; each sample is composed of 10 in situ phytoplankton high‐performance liquid chromatography (HPLC)‐measured pigment concentrations, GlobColour products of chlorophyll‐a concentration, and remote sensing reflectance (Rrs(λ)) data at different wavelengths, in addition to advanced very high resolution radiometer sea surface temperature measurements. The resulting data set regroups a large variety of encountered situations between 1997 and 2014. The nonlinear relationship between the in situ and satellite components was identified using a self‐organizing map, which is a neural network classifier. As a major result, the self‐organizing map enabled reliable estimations of the concentration of chlorophyll‐a and of nine different pigments from satellite observations. A cross‐validation procedure showed that the estimations were robust for all pigments (R² > 0.75 and an average root‐mean‐square error = 0.016 mg/m³). A consistent association of several phytoplankton pigments indicating phytoplankton group specific dynamic was shown at a global scale. We also showed the uncertainties for the estimation of each pigment.

We revisited the partitioning of the Mediterranean Sea into bioregions by processing satellite Sea Surface Temperature (SST) and Chlorophyll‐a concentration (Chla) from ocean color observations combined with Argo mixed‐layer depth for a period ranging from 2003 up to 2020. This regionalization was performed using an innovative classification based on self‐organizing maps, the so‐called 2S‐SOM. We clustered the Mediterranean Sea waters into seven bioregions governed by specific physical and biogeochemical characteristics. We studied the interannual variability of these bioregions over the 18 years. We showed that the temperature is increasing with a mean rate of 0.4°C per decade. The chlorophyll‐a concentration is quasi constant, but we evidenced a noticeable change in satellite‐derived phytoplankton communities: Diatoms concentration is decreasing while cyanobacteria concentration is increasing. The bioregion trends could be related to global warming. The whole Mediterranean Sea is shifting to an eastern Mediterranean state.

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 a (Chl a) 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 a) combined with an omics-based biomarker developed using Tara 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 a 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.

Satellite remote sensing is a powerful tool to monitor the global dynamics of marine plankton. Previous research has focused on developing models to predict the size or taxonomic groups of phytoplankton. Here, we present an approach to identify community types from a global plankton network that includes phytoplankton and heterotrophic protists and to predict their biogeography using global satellite observations. Six plankton community types were identified from a co-occurrence network inferred using a novel rDNA 18 S V4 planetary-scale eukaryotic metabarcoding dataset. Machine learning techniques were then applied to construct a model that predicted these community types from satellite data. The model showed an overall 67% accuracy in the prediction of the community types. The prediction using 17 satellite-derived parameters showed better performance than that using only temperature and/or the concentration of chlorophyll a. The constructed model predicted the global spatiotemporal distribution of community types over 19 years. The predicted distributions exhibited strong seasonal changes in community types in the subarctic–subtropical boundary regions, which were consistent with previous field observations. The model also identified the long-term trends in the distribution of community types, which suggested responses to ocean warming.