Acta Oceanologica Sinica 2010, Vol. 29, No. 2, P. 14-32 DOI: 10.1007/s13131-010-0018-y http://www.hyxb.org.cn E-mail: hyxbe@263.net A spectral response approach for detecting dominant phytoplankton size class from satellite remote sensing Robert J W Brewin 1∗ , Samantha J Lavender 1,2 , Nick J Hardman-Mountford 3,4 , Takafumi Hirata 3,4† 1 School of Marine Science and Engineering, University of Plymouth, UK 2 ARGANS Ltd, Plymouth, UK 3 Plymouth Marine Laboratory (PML), Plymouth, UK 4 National Centre for Earth Observation, PML, Plymouth, UK Received 11 December 2008; accepted 7 April 2009 c The Chinese Society of Oceanography and Springer-Verlag Berlin Heidelberg 2010 Abstract An important goal in ocean colour remote sensing is to accurately detect different phytoplank- ton groups with the potential uses including the validation of multi-phytoplankton carbon cycle models; synoptically monitoring the health of our oceans, and improving our understanding of the bio-geochemical interactions between phytoplankton and their environment. In this paper a new algorithm is developed for detecting three dominant phytoplankton size classes based on distinct differences in their optical signatures. The technique is validated against an independent cou- pled satellite reflectance and in situ pigment dataset and run on the 10-year NASA Sea viewing Wide Field of view Sensor (SeaWiFS) data series. Results indicate that on average 3.6% of the global oceanic surface layer is dominated by microplankton, 18.0% by nanoplankton and 78.4% by picoplankton. Results, however, are seen to vary depending on season and ocean basin. Key words: phytoplankton size, remote sensing, absorption, ocean colour, SeaWiFS 1 Introduction The oceanic carbon cycle is partly controlled by photosynthetic, microscopic, single celled algae termed phytoplankton. Phytoplankton use inorganic carbon to photosynthesise organic matter, which in turn is recycled in the water column or exported towards the sediments. Carbon cycle models have previously used phytoplankton abundance and distributions in con- junction with ocean physical models to understand the role the ocean has in cycling atmospheric carbon diox- ide. This has resulted in calculations of the oceanic im- pact on the carbon cycle, such as House et al. (2002), who estimated that 26% of carbon dioxide released into the atmosphere is absorbed by the ocean. As marine carbon cycling is specifically linked to the activity of particular phytoplankton groups, modellers have attempted to improve such estima- tions by explicitly incorporating phytoplankton com- munity structure or incorporating specific phytoplank- ton functional types (PFTs) into biogeochemical mod- els. This has led to a variety of multi-phytoplankton coupled ocean-ecosystem models (Taylor et al., 1993; Vanden Berg et al., 1996; Baretta-Bekker et al., 1997; Moore et al., 2002; Gregg et al., 2003; Blackford et al., 2004; Le Qu´ er´ e et al., 2005). Some authors have challenged the predictive ca- pability of such multi-phytoplankton models (Ander- son, 2005; Flynn, 2006). In order to validate and im- prove these models more information is required on the synoptic distribution of different phytoplankton groups as well as on their response to different envi- ronmental factors. Measuring phytoplankton biomass synoptically is very difficult and expensive to conduct in the ocean, and currently the only practical method is through satellite remote sensing. Ocean scientists have strived toward an ultimate goal of utilising remote sensing data in order to pro- duce synoptic global and regional maps of individual phytoplankton species. As remote sensing instruments Foundation item: This work is funded by the National Environmental Research Council, UK, through a PhD studentship at the Centre for observation of Air-Sea Interactions & fluXes (CASIX), the National Centre for Earth Observation and NERC Oceans 2025 programme (Themes 6 and 10). ∗ Corresponding author, E-mail: robert.brewin@plymouth.ac.uk † Present address: Graduate School of Environmental Earth Science, Hokkaido University NIOVV5, kita–ku, Sapporo 060–0810, Japan