Remote sensing of phytoplankton groups in case 1 waters from global SeaWiFS imagery

Ocean color sensors enable a quasi-permanent monitoring of the chlorophyll a concentration (Ch1 a) in surface waters. This ubiquitous photosynthetic pigment cannot, however, be used to distinguish between phytoplankton species. Distinguishing phytoplankton groups from space is nevertheless necessary to better study some biochemical processes such as carbon fixation at the global scale, and is thus one of the major challenges of ocean color research. In situ data have shown that the water-leaving radiances (nLw), measured by ocean color sensors at different wavelengths in the visible spectrum, vary significantly for a given Ch1 a. This natural variability is due partly to differences in optical properties of phytoplankton species. Here, we derive relationships between nLw and phytoplankton species by using a large set of quantitative inventories of phytoplankton pigments collected during nine cruises from Le Havre (France) to Noumea (New Caledonia) in the framework of the GeP&CO program. Coincident SeaWiFS nLw data between 412 and 555 nm are extracted and normalized to remove the effect of Ch1 a. These normalized spectra vary significantly with in situ pigment composition, so that four major phytoplankton groups, i.e., haptophytes, Prochlorococcus, Synechocoecus-like cyanobacteria and diatoms, can be distinguished. This classification (PHYSAT) is applied to the global SeaWiFS dataset for year 2001, and global maps of phytoplankton groups are presented. Haptophytes and diatoms are found mostly in high latitudes and in eutrophic regions. Diatoms show a strong seasonal cycle with large-scale blooms during spring and summer. These results, obtained with only five channels in the visible spectrum, demonstrate that ocean color measurements can be used to discriminate between dominant phytoplankton groups provided that sufficient data are available to establish the necessary empirical relationships. (c) 2005 Elsevier Ltd. All rights reserved.