Identifying four phytoplankton functional types from space: An ecological approach

Deriving maps of phytoplankton taxa based on remote sensing data using bio-optical properties of phytoplankton alone is challenging. A more holistic approach was developed using artificial neural networks, incorporating ecological and geographical knowledge together with ocean color, bio-optical characteristics, and remotely sensed physical parameters. Results show that the combined remote sensing approach could discriminate four major phytoplankton functional types (diatoms, dinoflagellates, coccolithophores, and silicoflagellates) with an accuracy of more than 70%. Models indicate that the most important information for phytoplankton functional type discrimination is spatio-temporal information and sea surface temperature. This approach can supply data for large-scale maps of predicted phytoplankton functional types, and an example is shown. As the foundation of the aquatic food chain, phytoplankton are an integral part of the ecosystem, affecting trophic dynamics, nutrient cycling, habitat condition, and fisheries resources (Irigoien et al. 2002). Phytoplankton are responsible for .45% of the total primary production of plants on Earth (Falkowski et al. 2004) and uptake of the greenhouse gas carbon dioxide (CO2), and they contribute to the biological pump. Although the role of marine phytoplankton is significant, knowledge of spatio-temporal distribution and abundance of functional types is limited, especially in the open oceans. Research has been restricted in both time and space because information is often obtained from relatively expensive ship-based in situ measurements. Deriving maps of phytoplankton functional types (PFTs) from remotely sensed data is a new and potentially important technological application which offers high spatio-temporal coverage. Anderson (2005) reported that ecology is poorly understood due to a lack of in situ data as well as functional-type information related to the chemical and physical regime, which in turn has hindered the development of a convincing PFT prediction model. Empirical relationships between in situ pigment measurements and remotely sensed ocean color data were determined by Alvain et al. (2005) who generated global maps of haptophytes, prochlorococcus, synechococcus-like cyanobacteria, and diatoms. Development of several bio-optical methods for the identification of different PFTs have been used to map coccolithophore bloom distributions (Brown Acknowledgments We thank present and past staff of SAHFOS who have contributed to the maintenance of the CPR time series. Special thanks to Abigail McQuatters-Gollop and Yaswant Pradhan for useful discussions and comments on the manuscript. D. E. Raitsos is supported by a scholarship from the University of Plymouth. This study was also supported by the U.K. Natural Environment Research Council through the Atlantic Meridional Transect consortium (NERC/O/S/2001/00680) and Center for Observation of Air-Sea Interactions and Fluxes (CASIX). This is contribution No. 159 of the AMT programme and No. 46 for CASIX. Limnol. Oceanogr., 53(2), 2008, 605–613 E 2008, by the American Society of Limnology and Oceanography, Inc.