Statistical analysis of absorption spectra of phytoplankton and of pigment concentrations observed during three POMME cruises using a neural network clustering method.

We present a neural network methodology for clustering large data sets into pertinent groups. We applied this methodology to analyze the phytoplankton absorption spectra data gathered by the Laboratoire d'Océanographie de Villefranche. We first partitioned the data into 100 classes by means of a self-organizing map (SOM) and then we clustered these classes into 6 significant groups. We focused our analysis on three POMME campaigns. We were able to interpret the absorption spectra of the samples taken in the first oceanic optical layer during these campaigns, in terms of seasonal variability. We showed that spectra from the PROSOPE Mediterranean campaign, which was conducted in a different region, were strongly similar to those of the POMME-3 campaign. This analysis led us to propose regional empirical relationships, linking phytoplankton absorption spectra to pigment concentrations, that perform better than the previously derived overall relation.

Retrieval of pigment concentrations and size structure of algal populations from their absorption spectra using multilayered perceptrons.

Spectral absorption coefficients of phytoplankton can now be derived, under some assumptions, from hyperspectral ocean color measurements and thus become accessible from space. In this study, multilayer perceptrons have been developed to retrieve information on the pigment composition and size structure of phytoplankton from these absorption spectra. The retrieved variables are the main pigment groups (chlorophylls a, b, c, and photosynthetic and nonphotosynthetic carotenoids) and the relative contributions of three algal size classes (pico-, nano-, and microphytoplankton) to total chlorophyll a. The networks have been trained, tested, and validated using more than 3,700 simultaneous absorption and pigment measurements collected in the world ocean. Among pigment groups, chlorophyll a is the most accurately retrieved (average relative errors of 17% and 16% for the test and validation data subsets, respectively), while the poorest performances are found for chlorophyll b (average relative errors of 51% and 40%). Relative contributions of algal size classes to total chlorophyll a are retrieved with average relative errors of 19% to 33% for the test subset and of 18% to 47% for the validation subset. The performances obtained for the validation data, showing no strong degradation with respect to test data, suggest that these neural networks might be operated with similar performances for a large variety of marine areas.

Statistical analysis of a database of absorption spectra of phytoplankton and pigment concentrations using self-organizing maps.

We present a statistical analysis of a large set of absorption spectra of phytoplankton, measured in natural samples collected from ocean water, in conjunction with detailed pigment concentrations. We processed the absorption spectra with a sophisticated neural network method suitable for classifying complex phenomena, the so-called self-organizing maps (SOM) proposed by Kohonen [Kohonen, Self Organizing Maps (Springer-Verlag, 1984)]. The aim was to compress the information embedded in the data set into a reduced number of classes characterizing the data set, which facilitates the analysis. By processing the absorption spectra, we were able to retrieve well-known relationships among pigment concentrations and to display them on maps to facilitate their interpretation. We then showed that the SOM enabled us to extract pertinent information about pigment concentrations normalized to chlorophyll a. We were able to propose new relationships between the fucoxanthin/Tchl-a ratio and the derivative of the absorption spectrum at 510 nm and between the Tchl-b/Tchl-a ratio and the derivative at 640 nm. Finally, we demonstrate the possibility of inverting the absorption spectrum to retrieve the pigment concentrations with better accuracy than a regression analysis using the Tchl-a concentration derived from the absorption at 440 nm. We also discuss the data coding used to build the self-organizing map. This methodology is very general and can be used to analyze a large class of complex data.