**1. Introduction**

Interest in phytoplankton diversity has increased in recent years due to its possible role in regulating climate by production and consumption of greenhouse gases. For example, gases can diffuse across the air-sea interface, many of which are synthesized and emitted by certain phytoplankton species or groups. It has been suggested that these variations play an important role in moderating our climate through backscattering of solar radiation and within cloud formation. Climate will ultimately control fundamental environmental conditions that regulate algal growth, including water temperature, nutrients, and light and thus can be expected to result in changes in the species composition, trophic structure and function of marine ecosystems. In the past several decades, the scientific community has witnessed changes in phytoplankton distribution.

The marine phytoplankton community is diverse and includes on the order of tens of thousands of phytoplankton species (Jeffrey & Vesk 1997). On regional scales, phytoplankton biogeography is controlled by the physical, chemical, and meteorological characteristics that force ecosystem dynamics. There is a renewed impetus for new technologies to provide information about the phytoplankton community composition over global scales. Real-time, large-scale taxonomic information, if available, could open up new possibilities and approaches geared toward monitoring highly-dynamic oceanic processes and phenomena such as algal blooms (including harmful algal blooms), frontal structures, eddies, and episodic events (storms, river outflow, and wind mixing). Phytoplankton diversity information provides a valuable quantitative database for structuring sophisticated predictive models that includes taxonomic phytoplankton community information such as size spectra, probability distribution of taxa, and upper trophic level estimations including fisheries productivity (Cheson & Case 1986, De Angelis & Waterhouse 1987). There have been several reviews and books written on phytoplankton community structure, dynamics, and biogeochemistry as measured by ocean color (Mitchell 1994, Martin 2004, Mueller et al. 2004, Miller et al. 2005, Richardson and LeDrew 2006, Longhurst 2007, Robinson 2010).

Ocean Color Remote Sensing of Phytoplankton Functional Types 103

various oceanographic regions (Irigolen et al. 2004). Marine phytoplankton diversity, similar to terrestrial vegetation, is described as a uni-modal function of phytoplankton biomass with an occurrence of maximum diversity at intermediate levels of phytoplankton biomass (sub-bloom concentrations) and low diversity during massive blooms (Irigolen et al. 2004). Similar to terrestrial vegetation, phytoplankton diversity is a unimodal function of phytoplankton biomass, with maximal diversity at intermediate levels of phytoplankton

Fig. 2. Ratios of diatoxanthin and carotenoids (to chlorophyll-a) near the Delmarva Peninsula against the SeaWiFS 10 year average. Differences between the values and the SeaWiFS average emphasize that the phytoplankton community is structured differently

Moreover, the diversity–productivity relationship of oceanic plankton is similar to the classical relationship observed in early shipboard collections of phytoplankton. The results of PFT distribution return a significantly different pattern compared with the 10-year SeaWiFS average, underscoring that underlying biological processes produce a complex

from the bulk community (e.g. chlorophyll-a).

ecological food web matrix in the ocean (Figure 2).

biomass and minimum diversity during massive blooms.

Fig. 1. Distribution of *Nitzschia americana* (solid square), *Nitzschia bicapitata* (open circle) and *Nitzschia pseudonana* (open triangle) from shipboard observations. Modified from Hasle (1976). These data were compiled over several years in order to understand their global distribution.

Approaches have been proposed for remote sensing detection of phytoplankton functional types (PFTs) and may be classified as direct (defined as they exploit optical signatures of phytoplankton that may be detected by sensors on satellite platforms) or indirect (e.g. exploit relationships between chlorophyll-a concentration and functional types). Both approaches have their advantages and limitations. Direct methods are limited to few phytoplankton groups that possess optical signatures distinct from other constituents in seawater. Indirect methods exploit well-established algorithms for retrieval of chlorophyll-a concentrations from ocean color satellite products in combination with phytoplankton community structure, which is validated by *in situ* observations. Optical sensors have ushered in a new capability of remotely sensing the ocean synoptically compared with *in situ* data collected on board ships, which may take years to compile but is is capable of exploring ancillary properties inaccessible to remote sensing (Hasle 1976, Figure 1). Early shipboard studies have demonstrated the complexity of the marine ecosystem and how distribution, function, and physiology are linked closely to the biochemistry of carbon, nitrogen, sulfur, and other elements in the sea. This chapter discusses how information on PFTs may be derived from a combination of satellite data and *in situ* observations. It is anticipated that developing an understanding of the large-scale distributions and variability in PFTs will contribute to various ecological problems of the day of paramount economic importance, such as providing an ecological basis for fisheries fluctuations and occurrence of harmful algal blooms. Future satellite sensors with improved spectral resolution or temporal resolution holds promise for improving the accuracy of retrieval of PFTs and understanding their role in the ecology of the ocean.

Global patterns of diversity for the majority of marine organisms are virtually uncharacterized (Worm et al. 2005). For planktonic organisms, which form the base of the marine food web, global ocean geographic distribution of diversity has only been characterized for planktonic foraminifera (Rutherford et al. 1999). Marine biodiversity patterns show a worldwide consistency despite differences in environmental conditions of

Fig. 1. Distribution of *Nitzschia americana* (solid square), *Nitzschia bicapitata* (open circle) and *Nitzschia pseudonana* (open triangle) from shipboard observations. Modified from Hasle (1976). These data were compiled over several years in order to understand their global

Approaches have been proposed for remote sensing detection of phytoplankton functional types (PFTs) and may be classified as direct (defined as they exploit optical signatures of phytoplankton that may be detected by sensors on satellite platforms) or indirect (e.g. exploit relationships between chlorophyll-a concentration and functional types). Both approaches have their advantages and limitations. Direct methods are limited to few phytoplankton groups that possess optical signatures distinct from other constituents in seawater. Indirect methods exploit well-established algorithms for retrieval of chlorophyll-a concentrations from ocean color satellite products in combination with phytoplankton community structure, which is validated by *in situ* observations. Optical sensors have ushered in a new capability of remotely sensing the ocean synoptically compared with *in situ* data collected on board ships, which may take years to compile but is is capable of exploring ancillary properties inaccessible to remote sensing (Hasle 1976, Figure 1). Early shipboard studies have demonstrated the complexity of the marine ecosystem and how distribution, function, and physiology are linked closely to the biochemistry of carbon, nitrogen, sulfur, and other elements in the sea. This chapter discusses how information on PFTs may be derived from a combination of satellite data and *in situ* observations. It is anticipated that developing an understanding of the large-scale distributions and variability in PFTs will contribute to various ecological problems of the day of paramount economic importance, such as providing an ecological basis for fisheries fluctuations and occurrence of harmful algal blooms. Future satellite sensors with improved spectral resolution or temporal resolution holds promise for improving the accuracy of retrieval of PFTs and

Global patterns of diversity for the majority of marine organisms are virtually uncharacterized (Worm et al. 2005). For planktonic organisms, which form the base of the marine food web, global ocean geographic distribution of diversity has only been characterized for planktonic foraminifera (Rutherford et al. 1999). Marine biodiversity patterns show a worldwide consistency despite differences in environmental conditions of

understanding their role in the ecology of the ocean.

distribution.

various oceanographic regions (Irigolen et al. 2004). Marine phytoplankton diversity, similar to terrestrial vegetation, is described as a uni-modal function of phytoplankton biomass with an occurrence of maximum diversity at intermediate levels of phytoplankton biomass (sub-bloom concentrations) and low diversity during massive blooms (Irigolen et al. 2004). Similar to terrestrial vegetation, phytoplankton diversity is a unimodal function of phytoplankton biomass, with maximal diversity at intermediate levels of phytoplankton biomass and minimum diversity during massive blooms.

Fig. 2. Ratios of diatoxanthin and carotenoids (to chlorophyll-a) near the Delmarva Peninsula against the SeaWiFS 10 year average. Differences between the values and the SeaWiFS average emphasize that the phytoplankton community is structured differently from the bulk community (e.g. chlorophyll-a).

Moreover, the diversity–productivity relationship of oceanic plankton is similar to the classical relationship observed in early shipboard collections of phytoplankton. The results of PFT distribution return a significantly different pattern compared with the 10-year SeaWiFS average, underscoring that underlying biological processes produce a complex ecological food web matrix in the ocean (Figure 2).

Ocean Color Remote Sensing of Phytoplankton Functional Types 105

Our limited understanding of marine ecosystem responses to various physico-chemical climate drivers, which include genetic and phenotypic adaptation to the unprecedented pace of climate change call for novel methods to study the problem, and remote sensing is one of the potential avenues (Falkowski et al. 2004, Chavez et al. 1999). Understanding the cyclical relationship of phytoplankton groups and how the production of greenhouse gases relates to temperature and productivity provides a better understanding of their coupling to climate (Falkowski et al. 2004). Changes in phytoplankton community structure could potentially provide an early warning for climate-driven perturbations in marine ecosystems (Hallegraeff 2010). Establishing the biogeography of several morphologically and genetically distinct marine species is a complex target for which remote sensing may

Historically, satellite sensors have provided oceanographers with bulk phytoplankton pigment concentrations (e.g. chlorophyll-*a*) over global scales with synoptic resolution (McClain, 2009). Coastal Zone Color Scanner (CZCS) was the beginning of the era of satellite remote sensing of ocean color. It was a relatively simple sensor and was successful at the retrieval of chlorophyll-a concentrations in open-ocean waters. Models were then developed that utilized satellite-derived chlorophyll-a data along with information on photosynthetically available radiation and temperature at the sea surface to compute water column primary production by remote sensing. However, CZCS performed poorly in coastal waters. SeaWiFS and MODIS, the successor to the CZCS, had better radiometric precision and more wavebands and was designed to address the limitations of CZCS, in particular for applications in coastal waters. Satellites with even higher spectral resolution followed (e.g. MERIS). Though the newer generations of satellites were designed for improving performance in coastal waters, there has been increasing recognition of their value in the detection of phytoplankton functional types as well. These satellite missions have provided knowledge and understanding of a variety of events including frontal features and episodic blooms on a global basis. The scientific community has now moved onto predicting ecologically significant characteristics of the food web such as size structure as interpreted for phytoplankton size classes including picoplankton to microplankton (Aiken et al. 2008). Ocean Color has been recognized as an essential climate variable by the Global Climate Observation System (GCOS). As the global time series of ocean color data grows (CZCS, MODIS, & SeaWiFS), there is an increase in satellite products available to study decadal-scale variations in phytoplankton distribution and primary production. Understanding the spatial and temporal distribution of PFTs will allow us to improve our knowledge of biologically-mediated fluxes of elements between the upper ocean and the ocean interior (Falkowski & Raven, 1997). The performance of biogeochemical models in the ocean has improved substantially as a result of incorporating PFTs into ecosystem models. The spatial variability and concentration of various PFTs are critical to improving primary-productivity estimates, and understanding the feedbacks of climate change. The ability to observe PFTs on a global scale that relate to key biogeochemical processes such as nitrogen fixation, silicification, and calcification is valuable to studies of marine elemental cycles. Despite the paucity of data on functional groups, our understanding of ecosystem linkages is improving as we accrue larger amounts of data. Major divisions of phytoplankton taxonomic groups such as diatoms, coccolithophores, dinoflagellates, chlorophytes, and cyanobacteria are often separated into distinct functional groups, as these

provide at best, a partial solution.

taxonomic groups have unique biogeochemical signatures

Fig. 3. Monthly SeaWiFS data for 2004 interpreted for phytoplankton, size classes: pico (blue); nano (green); micro (red, Aiken et al. 2009).

Fig. 3. Monthly SeaWiFS data for 2004 interpreted for phytoplankton, size classes: pico

(blue); nano (green); micro (red, Aiken et al. 2009).

Our limited understanding of marine ecosystem responses to various physico-chemical climate drivers, which include genetic and phenotypic adaptation to the unprecedented pace of climate change call for novel methods to study the problem, and remote sensing is one of the potential avenues (Falkowski et al. 2004, Chavez et al. 1999). Understanding the cyclical relationship of phytoplankton groups and how the production of greenhouse gases relates to temperature and productivity provides a better understanding of their coupling to climate (Falkowski et al. 2004). Changes in phytoplankton community structure could potentially provide an early warning for climate-driven perturbations in marine ecosystems (Hallegraeff 2010). Establishing the biogeography of several morphologically and genetically distinct marine species is a complex target for which remote sensing may provide at best, a partial solution.

Historically, satellite sensors have provided oceanographers with bulk phytoplankton pigment concentrations (e.g. chlorophyll-*a*) over global scales with synoptic resolution (McClain, 2009). Coastal Zone Color Scanner (CZCS) was the beginning of the era of satellite remote sensing of ocean color. It was a relatively simple sensor and was successful at the retrieval of chlorophyll-a concentrations in open-ocean waters. Models were then developed that utilized satellite-derived chlorophyll-a data along with information on photosynthetically available radiation and temperature at the sea surface to compute water column primary production by remote sensing. However, CZCS performed poorly in coastal waters. SeaWiFS and MODIS, the successor to the CZCS, had better radiometric precision and more wavebands and was designed to address the limitations of CZCS, in particular for applications in coastal waters. Satellites with even higher spectral resolution followed (e.g. MERIS). Though the newer generations of satellites were designed for improving performance in coastal waters, there has been increasing recognition of their value in the detection of phytoplankton functional types as well. These satellite missions have provided knowledge and understanding of a variety of events including frontal features and episodic blooms on a global basis. The scientific community has now moved onto predicting ecologically significant characteristics of the food web such as size structure as interpreted for phytoplankton size classes including picoplankton to microplankton (Aiken et al. 2008). Ocean Color has been recognized as an essential climate variable by the Global Climate Observation System (GCOS). As the global time series of ocean color data grows (CZCS, MODIS, & SeaWiFS), there is an increase in satellite products available to study decadal-scale variations in phytoplankton distribution and primary production.

Understanding the spatial and temporal distribution of PFTs will allow us to improve our knowledge of biologically-mediated fluxes of elements between the upper ocean and the ocean interior (Falkowski & Raven, 1997). The performance of biogeochemical models in the ocean has improved substantially as a result of incorporating PFTs into ecosystem models. The spatial variability and concentration of various PFTs are critical to improving primary-productivity estimates, and understanding the feedbacks of climate change. The ability to observe PFTs on a global scale that relate to key biogeochemical processes such as nitrogen fixation, silicification, and calcification is valuable to studies of marine elemental cycles. Despite the paucity of data on functional groups, our understanding of ecosystem linkages is improving as we accrue larger amounts of data. Major divisions of phytoplankton taxonomic groups such as diatoms, coccolithophores, dinoflagellates, chlorophytes, and cyanobacteria are often separated into distinct functional groups, as these taxonomic groups have unique biogeochemical signatures

Ocean Color Remote Sensing of Phytoplankton Functional Types 107

The Ocean Color community has utilized a reductionist approach to describe phytoplankton taxonomic composition and refers to optical signatures of certain phytoplankton groups/species or PFTs. In this chapter, we will refer to Phytoplankton Functional Types (PFTs) in order to utilize current terminology. Please take into account that we may be defining groups within different levels of complexity that relate to phylogeny or bio-optically similarities etc. Taxonomic characterization has utilized different terminology over the years and we delve into the history of its classification

Present studies on diversity are not too far removed from the original ideas invoked by G.E. Hutchinson's "Paradox of the Plankton" (Hutchinson, 1961) and J. H. Connell's "Intermediate Disturbance Hypothesis" (Connell, 1978). Ocean variability, through eddies, storms, seasonal cycles, and El Niño, have been documented to play a role in controlling diversity. How this level of stochastic variability manifests itself spatially and temporally due to climate change is unknown. Falkowski & Oliver (2007) suggest that resource competition theory is adequate to describe global distribution of marine eukaryotic phytoplankton taxa during equator-to-pole and continent-to-land thermal gradients. These results are supportive in light of describing phytoplankton beyond a bulk approach of

A functional group is used as an ecological term to define a group of organisms that have similar biogeochemical processes e.g. nitrogen fixation, silicifiers, and calcifiers (Fauchald & Jumars 1979, Reynolds et al. 2002, Nair et al. 2008). Functional groups in phytoplankton are defined as groups of organisms related through common biogeochemical processes but are not necessarily phylogenetically related (Iglesias-Rodriguez et al. 2002). Homologously similar sets of organisms are referred to as "functional groups" or "biogeochemical guilds" following Totterdell et al. (1993). For marine phytoplankton, some attempts have been made to organize species within certain classes of phytoplankton according to their ecological and biogeochemical function or habitat distribution perspective (Iglesias-Rodríguez et al. 2002, Smayda & Reynolds, 2003,Vila & Masó, 2005). Thus, terminology has finally evolved to the usage of Phytoplankton Functional Type (PFT) or Phytoplankton Functional group but these classifications are not necessarily straightforward and include a wide range of different taxonomic classification approaches such as size, biogeochemistry,

**3. On the road to predicting remote sensing reflectance: Fundamental basis** 

Algorithm development for remote sensing is focused on estimating the quantities of various optically-active constituents in the ocean. Various pigments present in phytoplankton play a role in determining the total absorption coefficient of phytoplankton, which is a key determinant of the spectral variability in remote-sensing reflectance. We utilize a reconstruction model to demonstrate the spectral variability of photoprotective and photosynthetic pigments (Sathyendranath et al. 1987, Bidigare 1990). There are differences in the weight-specific absorption spectra of various *in vitro* (extracted) phytoplankton pigments (Figure 5). The comparison of maximal peaks of mass-specific absorption spectra

or traditional taxonomic hierarchies (see discussion by Nair et al. 2008).

**between pigments and in vivo absorption** 

**2. "History of terminology for phytoplankton functional types": Ecological,** 

**biogeochemical, and optical definitions** 

utilizing only chlorophyll a.

because it may define many different levels of complexity.

Fig. 4. Schematic diagram showing sources of variation in optical characteristics which influence ocean color. (Keller et al. 1989)

Aiken et al. (2009) shows how different size classes of phytoplankton are distributed globally, and size is often related to function (Nair et al. 2008). Nonetheless, there are limitations to discriminating phytoplankton taxonomic composition or even functional groups by remote sensing, due to our inability to discriminate various phytoplankton types optically. Remote-sensing reflectance (Rrs (λ)) is influenced by absorption and backscattering by seawater, phytoplankton, colored dissolved organic matter, detrital matter and other suspended material (Garver & Siegel 1997). We show the various optical pools that contribute to absorption, which contributes to the spectral variability of Rrs (λ, Figure 4). Remote sensing of particular types of phytoplankton is only possible if optical characteristics are identified for those types that can be used to distinguish them from all other types of material in the water.

New technological developments and improved scientific knowledge have allowed for the development of several approaches for detecting phytoplankton biomass and some functional groups of phytoplankton including coccolithophores (Balch et al. 1991, Balch et al. 1996) and *Trichodesmium* (Subramanian et al. 1994, 1999 a,b, Hu et al. 2010). More recently, algorithms have been developed to distinguish additional phytoplankton groups and size classes (Sathyendranath et al. 2004, Alvain et al. 2005, Moisan et al. 2011ab). High Performance Liquid Chromatography (HPLC) is the *in situ* method of choice to serve as ground truth for satellite products due to its accuracy and rapid processing. A relatively new analysis tool called CHEMical TAXonomy (CHEMTAX) has been developed that is designed to yield information on the phytoplankton composition using HPLC data (Mackey et al.1996), but it has to be recognized that HPLC models have their limitations (e.g. Latasa 2007).

Fig. 4. Schematic diagram showing sources of variation in optical characteristics which

Aiken et al. (2009) shows how different size classes of phytoplankton are distributed globally, and size is often related to function (Nair et al. 2008). Nonetheless, there are limitations to discriminating phytoplankton taxonomic composition or even functional groups by remote sensing, due to our inability to discriminate various phytoplankton types optically. Remote-sensing reflectance (Rrs (λ)) is influenced by absorption and backscattering by seawater, phytoplankton, colored dissolved organic matter, detrital matter and other suspended material (Garver & Siegel 1997). We show the various optical pools that contribute to absorption, which contributes to the spectral variability of Rrs (λ, Figure 4). Remote sensing of particular types of phytoplankton is only possible if optical characteristics are identified for those types that can be used to distinguish them from all

New technological developments and improved scientific knowledge have allowed for the development of several approaches for detecting phytoplankton biomass and some functional groups of phytoplankton including coccolithophores (Balch et al. 1991, Balch et al. 1996) and *Trichodesmium* (Subramanian et al. 1994, 1999 a,b, Hu et al. 2010). More recently, algorithms have been developed to distinguish additional phytoplankton groups and size classes (Sathyendranath et al. 2004, Alvain et al. 2005, Moisan et al. 2011ab). High Performance Liquid Chromatography (HPLC) is the *in situ* method of choice to serve as ground truth for satellite products due to its accuracy and rapid processing. A relatively new analysis tool called CHEMical TAXonomy (CHEMTAX) has been developed that is designed to yield information on the phytoplankton composition using HPLC data (Mackey et al.1996), but it has to be recognized that HPLC models have their limitations (e.g. Latasa

influence ocean color. (Keller et al. 1989)

other types of material in the water.

2007).
