Evaluating the importance of phytoplankton community structure to the optical properties of the Santa Barbara Channel, California
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Limnol. Oceanogr., 59(3), 2014, 927–946
E 2014, by the Association for the Sciences of Limnology and Oceanography, Inc.
doi:10.4319/lo.2014.59.3.0927
Evaluating the importance of phytoplankton community structure to the optical
properties of the Santa Barbara Channel, California
Rebecca K. Barrón,1,* David A. Siegel,1,2 and Nathalie Guillocheau 1
1 Earth Research Institute, University of California, Santa Barbara, California
2 Department of Geography, University of California Santa Barbara, California
Abstract
Observations from the Santa Barbara Channel, California, were used to evaluate relationships among optical
properties and phytoplankton community structure. Phytoplankton community structure was determined by
statistically analyzing 10 diagnostic phytoplankton pigment concentrations using empirical orthogonal function
(EOF) analysis. The first four EOF modes explained 82% of phytoplankton community structure variability and
were interpreted as a mixed community mode composed mostly of nanoplankton, a mode dominated by
microplankton (diatoms and dinoflagellates), a mode describing alternating diatom and dinoflagellate
dominance, and a mode reflecting picoplankton presence. Variations in colored dissolved organic matter
(CDOM) and phytoplankton absorption spectra were related to changes of the mixed microplankton modal
amplitudes. Characteristics of the CDOM spectrum were further dependent on whether diatoms or dinoflagellates
were dominant. The particle backscattering coefficient was significantly correlated with EOF modes describing
the mixed microplankton and the picoplankton communities. The influence of phytoplankton community
structure was also seen in the performance of standard ocean color algorithms using the in situ data set. The
present results demonstrate that many optical characteristics vary significantly with changes in phytoplankton
community structure and suggest that improvements in remote-sensing algorithms will require model coefficients
to vary accordingly. Further, changes in phytoplankton community composition affect both dissolved and
particle absorption and scattering properties, not simply the phytoplankton-specific properties, creating
challenges for the development of algorithms aimed at assessing phytoplankton community structure from
satellite observations.
Ocean color remote sensing has revolutionized our where ocean color is dominated by phytoplankton proper-
understanding of the global ocean by providing informa- ties and other constituents are assumed to roughly covary
tion about phytoplankton distributions, rates of net with changes in Chl a (International Ocean Colour
primary production, and particle characteristics of the Coordinating Group [IOCCG] 2000; Siegel et al. 2013).
surface ocean. Space-borne instrumentation quantifies the However, in complex ocean optical environments, the
reflectance spectrum of the sea surface, and the information contributions of other optical properties relative to
it contains, on spatial and temporal scales impossible to phytoplankton chlorophyll concentrations vary, making
achieve by any other means. Remote-sensing reflectance of this empirical approach fraught with difficulties (Dierssen
ocean waters, Rrs(l) or, equivalently, normalized water- 2010; Szeto et al. 2011). Semianalytical models break the
leaving radiance, LwN(l) can be modeled as a function of assumption of common proportions among IOPs that
the absorption and backscattering coefficients of seawater, plague empirical relationships (Lee et al. 2002; Maritorena
termed inherent optical properties (IOPs). IOPs are et al. 2002); however, they have yet to include the influences
additive functions of seawater itself as well as suspended that variations in phytoplankton community structure may
and dissolved constituents, such as phytoplankton, detri- create.
tus, mineral particles, colored dissolved organic matter Coastal areas are often optically complex due to high
(CDOM), bacteria, viruses, and air bubbles (Mobley et al. and highly variable levels of turbidity, phytoplankton
2002; Stramski et al. 2004). Thus, understanding the link productivity, and CDOM from upwelling and/or terrestrial
between IOPs and Rrs(l) is critical for studying and inputs, and the associated IOPs may vary independently
monitoring biological and biogeochemical change in the (IOCCG 2000; Toole and Siegel 2001). Such areas of
world oceans. optical complexity, termed Case II conditions, often require
Ocean color algorithms model biological and optical local calibration for empirical coefficients used in remote-
properties from remotely sensed observations of ocean sensing algorithms. Magnuson et al. (2004) found that local
reflectance spectra (O’Reilly et al. 1998; Lee et al. 2002; tuning of the Garver-Siegel-Maritorena (GSM) semianaly-
Maritorena et al. 2002). The Ocean Color algorithm (OC4) tical and the SeaWiFS operational (OC4) empirical models
quantifies chlorophyll a (Chl a) concentration from ocean for the Chesapeake Bay and the Mid-Atlantic Bight
reflectance using an empirical relationship between ratios resulted in better Chl a retrieval statistics than the
of Rrs(l) bands and in situ measurements of Chl a (O’Reilly respective global models. Kostadinov et al. (2007) locally
et al. 1998). This algorithm is adequate for the open ocean, optimized the GSM model for the Santa Barbara Channel
(SBC) but found no significant improvement in model
* Corresponding author: rebecca@eri.ucsb.edu performance. The lack of improvement was attributed to
927928 Barrón et al. the assumption that the IOP spectral shapes were constant phytoplankton species have been directly linked to in time. Changes in the seawater constituents can cause increases in CDOM absorption via the release of photo- changes in IOP spectral shape. protective pigments called mycosporine-like amino acids Changes in phytoplankton functional type affect IOPs (MAAs; Vernet and Whitehead 1996). Absorption and directly due to differences in size, shape, density, and scattering of detritus can be significant and correlate to cellular composition and indirectly due to excretions and phytoplankton assemblage, especially during or following a biological relationships, such as nutrient cycling, affecting phytoplankton bloom (Antoine et al. 2011). Associated the local environment. For example, Stramski et al. (2001) biogeochemical cycling of the detrital matter can also have showed that both the magnitude and the spectral shapes of an effect on CDOM absorption. absorption and scattering varied significantly for particu- For all the reasons listed above, phytoplankton com- late assemblages with the same total Chl a concentration. munity composition should play a significant role in Cell size, shape, and composition all affect the relative determining IOP characteristics and therefore should affect proportions of light scatter in the forward and backward ocean reflectance spectra (Mobley and Stramski 1997; directions (Morel and Ahn 1990; Stramski et al. 2001, Dierssen 2010). Over the last decade, several studies have 2004). Over a significant portion of the size range, Mie used ocean color reflectance determinations to assess theory predicts that smaller homogeneous spheres will phytoplankton community structure (Sathyendranath et al. scatter proportionally more light in the backward direc- 2001; Alvain et al. 2005; Torrecilla et al. 2011). Alvain et al. tion than larger ones. However, recent observational and (2006) explained the variability in Chl a concentrations laboratory studies have found that larger phytoplankton modeled with the OC4 band-ratio algorithm by the direct contribute significantly to particulate backscattering relationship of LwN(l) spectral shape with phytoplankton (Dall’Olmo et al. 2009; Whitmire et al. 2010; Westberry community composition. Although this novel approach et al. 2010). Differences in cellular composition and shape explained chlorophyll variability well, shortcomings with also affect light scatter. This can lead to differences in the this approach are manifested in the fact that the OC4 amount of light backscattered for organisms with the same algorithm does not consider the independent variability of assumed scattering cross sections. Vaillancourt et al. (2004) IOPs. More recently, Alvain et al. (2012) evaluated their found that dinoflagellates had the highest backscattering original model and found it to be sensitive to variations in efficiency of all of the phytoplankton species in their 12- particle backscattering as well as CDOM and phytoplank- culture study. Whitmire et al. (2010) showed that back- ton absorption. Therefore, constraining IOP variability scattering ratios for diatoms were largely a function of size, and, in particular, the roles of phytoplankton community whereas dinoflagellate backscattering ratios were high structure changes is needed to advance ocean color models. regardless of size. Here, we will evaluate the relationship of phytoplankton Different species of phytoplankton can have unique community structure to the optical properties in a complex absorption spectra due to the differences in the amount of coastal ocean. We identify phytoplankton communities by Chl a per cell as well as the presence of various accessory applying a multivariate statistical procedure to phyto- pigments unique to their functional type (Kirk 1994; plankton indicator pigment concentrations collected over Stramski et al. 2001; Dierssen 2010). Phytoplankton species a 4-yr period alongside apparent and inherent optical have the ability to increase the intracellular pigment property measurements. The goal of this work is to concentration, thereby increasing the Chl a content per ultimately assess relationships between IOPs and phyto- cell and effectively decreasing the absorption per pigmented plankton community structure that can help us identify particle with respect to the same concentration of pigment sources of error for remote-sensing algorithms for areas suspended in solution (Morel and Bricaud 1981). This is with high biological and biogeochemical diversity. The referred to as pigment packaging and is common among shifts in biological and biogeochemical properties observed diatom species (Nelson et al. 1993). Bricaud et al. (2004) span those of the global ocean, thereby making this study concluded that pigment packaging, attributed to differences relevant to a diverse array of oceanic systems. in size class, was an important source of variability of phytoplankton absorption in the global ocean. In addition Methods to Chl a, phytoplankton contain many accessory pigments that absorb light at various wavelengths to aid in Study site—The SBC is a dynamic coastal system with photosynthetic processes and/or provide protection from near-surface Chl a concentrations ranging from 0.3 to ultraviolet (UV) light (Roy et al. 2011). Accessory 28 mg m23, while stream and river inputs only episodically pigment content may be unique to various phytoplankton influence overall particle loads (Toole and Siegel 2001; groups and can be observed as differences in the Otero and Siegel 2004; Kostadinov et al. 2012). Phyto- absorbance spectra (Dierssen 2010). plankton community composition in the SBC varies Secondary effects of phytoplankton functional types on seasonally from a microphytoplankton community, indic- IOPs include that of CDOM and particulate detritus ative of a eutrophic coastal upwelling system, to a associated with a given phytoplankton community. CDOM community comprised of nano- and picophytoplankton accounts for the majority of UV and blue spectral light resembling a more oceanic, oligotrophic system (Anderson absorption in the ocean (Nelson and Siegel 2013). et al. 2008). Due to the low stream-water inputs during Phytoplankton community structure may also affect the most of the year, changes in phytoplankton abundances CDOM composition in the ocean. In particular, some and characteristics dominate the variability of all IOPs in
Santa Barbara Channel optical properties 929
Fig. 1. Map of the SBC. PnB stations are indicated by their station number.
the SBC (Toole and Siegel 2001; Kostadinov et al. 2007; frozen until analysis using an NaOH extraction procedure
Antoine et al. 2011). described in Shipe and Brzezinski (2001) and, more
Data for this article were collected as part of the Plumes recently, in Krause et al. (2013). POC samples were filtered
and Blooms (PnB) program, which conducts monthly onto GF/F filters shipboard and immediately stored in
cruises across the SBC at seven stations from Goleta Point liquid nitrogen until analysis on a CE440 Elemental
to Santa Rosa Island (Fig. 1). The surveys consist of Analyzer.
conductivity–temperature–depth measurements, optical pa- The phytoplankton pigments were determined via high-
rameters measured in situ with profiling instruments and in performance liquid chromatography (HPLC) analysis
the laboratory from discrete samples, and various chemical (described in more detail below) and included Chl a in
and biological determinations. A brief description of the pigment suite. Fluorometric Chl a analysis using a
measurement methodologies is provided below, and more standard acetone extraction method was also conducted,
detailed descriptions are available from the PnB website although only Chl a concentrations determined via HPLC
(http://www.icess.ucsb.edu/PnB) and in previous studies are presented here to provide consistency.
(Toole and Siegel 2001; Anderson et al. 2008; Kostadinov
et al. 2012). The data used for this paper span from IOPs—The coefficient of absorption for CDOM, ag(l),
November 2005 to August 2009 and make up a subset of was determined using surface water samples that were
the entire record that began in 1996 and continues collected in glass amber bottles preconditioned for carbon
presently. analysis. Samples were immediately stored in a shipboard
refrigerator at 4uC. Samples were filtered through a 0.2 mm
Discrete water sample analyses—Discrete samples were membrane filter in the laboratory and analyzed on a
taken using 5-liter Niskin bottles deployed on a rosette with Shimadzu 2401-PC spectrophotometer within 24 h of
a Seabird 9/11 conductivity–temperature–depth system. collection.
Samples that were taken at discrete depths were analyzed Several large peak-like features in the UV spectral region
for nutrients, biogenic and lithogenic silica (BSi and LSi), were found in the CDOM spectra from PnB (Fig. 2a). The
particulate organic carbon (POC), phytoplankton accesso- features are a departure from a typical CDOM spectrum, a
ry pigments, Chl a, and CDOM. Nutrient samples were spectrum decreasing exponentially with increasing wave-
collected directly from the Niskin bottles and stored in a length, and were similar in shape to MAA absorption
shipboard freezer, then transferred to the laboratory freezer signatures presented in other studies (Whitehead and
until analysis on a Lachat QuikChem 8000 Flow Injection Vernet 2000). The (presumably) MAA peaks in this study
Analyzer (http://www.msi.ucsb.edu/services/analytical-lab/ were quantified by modeling a ‘‘baseline’’ CDOM spectrum
instruments/flow-injection-analyzer). BSi and LSi samples by log-transforming the ag(l) data (see Fig. 2), then making
were filtered shipboard onto 0.4 mm membrane filters and a linear regression fit to the data points surrounding the930 Barrón et al.
Fig. 2. Example of MAA peak and MAA Index calculation: (a) CDOM absorption spectra for Sta. 5 on 24 January 2006 in the solid
line and molded ‘‘baseline’’ spectra (310–390 nm) in the dashed line and (b) residual between modeled ag(l) and real ag(l) data (310–
390 nm).
MAA absorption region (300–310 nm and 390–400 nm). coefficient, bp(l), was determined by bp(l) 5 b(l) 2 bw(l),
The procedure was similar to that of calculating the where bw(l) is the scattering coefficient of pure seawater
spectral slope, S (Nelson et al. 2007, 2010), but was used taken from Smith and Baker (1981).
here to estimate what the CDOM spectra would appear to Vertical profiles of volume scattering function at 140u,
be if the MAA-like signal were not present. The modeled b(140u,l), were measured at wavelengths l 5 442, 470, 510,
baseline spectra were then subtracted from the real ag(l) 589, and 671 nm using a HobiLabs Hydroscat-6 profiling
data. The term ‘‘MAA index’’ is defined here as the instrument. The data were filtered with a moving average
summation of the residual between ag(l) and the modeled and then binned to 1 m. The Hydroscat data were corrected
baseline between l 5 310 and 390 nm (Fig. 2b). for light attenuated in the measurement path of the
Samples for determining the particle absorption coeffi- instrument, called a s(l) correction, using data collected
cient, ap(l), were collected by filtering seawater onto a GF/F simultaneously by the AC-9 (Kostadinov et al. 2010, 2012).
filter and then stored immediately in liquid nitrogen until b(140u, l) was then converted to a particle backscattering
laboratory analysis. The ap(l) samples were analyzed on the coefficient, bbp(l), using
Shimadzu 2401-PC using the quantitative filtration tech-
nique (Mitchell 1990). The optical path-length amplification bbp (l)~2pxp ½b(1400 , l){bw (1400 , l) ð1Þ
factor was determined using natural phytoplankton samples The value of xp 5 1.14 was determined from the results of
collected from PnB cruises (Guillocheau 2003). The filters Dall’Olmo et al. (2009), and bw(140u, l) was determined
were then extracted in methanol overnight to remove from Morel (1974). The data were then averaged for the
extractable phytoplankton products and reanalyzed on the upper 15 m of the water column to estimate average surface
spectrophotometer for absorption of detrital material, ad(l). bbp(l) values.
This allowed for the quantification of phytoplankton-
specific absorption as aph(l) 5 ap(l) 2 ad(l). Remote-sensing reflectance spectra—Light reflectance
Vertical profiles of the beam attenuation coefficient, c(l), was determined using a free-falling Biospherical Instru-
and absorption coefficient, a(l), spectra were collected at ments Profiling Reflectance Radiometer, PRR-600. The
each station using a Wetlabs AC-9 profiling instrument at instrument captured vertical profiles of upwelling radiance,
wavelengths l 5 440, 488, 510, 555, 630, and 676 nm, which Lu(l), and downwelling irradiance, Ed(l), at wavelengths
were linear interpolated to match those captured by the l 5 412, 443, 490, 510, 555, and 656 nm. Values of the
Hydroscat (described below). Correction algorithms and remote-sensing reflectance spectrum, Rrs(l), were calculat-
data analysis procedures for this instrument are presented ed from the ratio of the upwelling radiance just beneath the
in Kostadinov et al. (2012). Surface values were obtained sea surface, Lu(02, l), to the corresponding downwelling
by averaging the upper 15 m of the vertical profiles. The irradiance spectrum, Ed(02, l), and then propagated across
total scattering coefficient, b(l), was calculated as the the sea surface using the relationship described in Lee et al.
difference between the beam attenuation and absorption (2002). Further data-processing details can be found in
coefficients, or b(l) 5 c(l) 2 a(l). The particle scattering Kostadinov et al. (2012). Estimates of Rrs(l) were then usedSanta Barbara Channel optical properties 931
Table 1. Diagnostic phytoplankton pigments for chemotaxonomy. Modified from Vidussi
et al. (2001).
Pigments Abbreviation Taxonomic significance Size class
Peridinin Per Dinoflagellates Micro (.20 mm)
199butanoyloxy-
fucoxanthin But Chromophytes nanoflagellates Nano (2–20 mm)
Fucoxanthin Fuco Diatoms Micro (.20 mm)
Violaxanthin Viol Photoprotection
199hexanoyloxy-
fucoxanthin Hex Chromophytes nanoflagellates Nano (2–20 mm)
Alloxanthin Allo Cryptophytes Nano (2–20 mm)
Zeaxanthin Zea Cyanobacteria prochlorophytes Pico (,2 mm)
Lutien Lut Photoprotection
Chlorophyll b Chl b Green flagellates prochlorophytes Nano (2–20 mm)
Chlorophyll a Chl a
to retrieve IOPs and Chl a concentrations. The GSM model for determining phytoplankton community compositions,
uses a semianalytical algorithm that relates Rrs(l) to the because the EOF method is well suited for understand-
absorption and scattering properties of seawater and ing covariability among the diagnostic pigments. The
retrieves bbp(443), ag(443), and Chl a concentration as CHEMTAX program requires a priori ratios of pigment
outputs (Maritorena et al. 2002). The OC4 model is an concentrations and does not allow those pigment ratios to
empirical algorithm that uses band ratios of Rrs(l) at blue vary in either time or space. Hence, the multivariate
to green wavelengths to derive Chl a concentrations statistical approaches are more flexible in that respect.
(O’Reilly et al. 1994). The globally optimized versions of Anderson et al. (2008) performed a statistical analysis
the GSM model (Maritorena et al. 2002) and OC4v6 model using PnB diagnostic pigment data from 1998 to 2003.
were used (http://oceancolor.gsfc.nasa.gov/REPROCESSING/ The pigment samples used in that study were analyzed by
R2009/ocv6). a team at the San Diego State University Center for
Hydro-Optics and Remote Sensing (CHORS). Quality
Phytoplankton community composition—Phytoplankton assurance discrepancies in the HPLC procedures from the
community composition was determined using a multivar- CHORS lab surfaced shortly after the Anderson et al.
iate statistical approach applied to phytoplankton pigment (2008) article was published (Hooker et al. 2009). The
concentrations collected at each PnB station. Pigment calibration issues in the CHORS HPLC phytoplankton
samples were collected from surface waters, immediately pigment data set resulted in overestimating pigment
concentrated by filtration onto GF/F filters, and stored in concentrations that were typically found in relatively
liquid nitrogen. Samples were shipped in liquid nitrogen to higher concentrations and underestimating pigments that
the Horn Point Laboratory for HPLC analysis (Hooker were found in lower concentrations. We chose to omit the
et al. 2009). The 10 pigments chosen for this analysis are CHORS data for this work and use here only data from
considered diagnostic pigments and represent the presence the Horn Point Laboratory. It should be noted that the
of different phytoplankton functional groups (Table 1; methodological discrepancies between the two data sets
following Vidussi et al. 2001). We added Chl a and lutein (a were unlikely to have affected the outcome of the analysis
photoprotective pigment for many species) to our pigment in Anderson et al. (2008) due to the nature of the
suite in an attempt to better characterize phytoplankton statistical procedure. That is, quantification issues would
community responses under bloom conditions and chang- not necessarily affect the patterns of covariability among
ing light conditions. pigments to first order, the EOF method would assess
Following Anderson et al. (2008), we performed an these patterns nearly independent of the issues in their
empirical orthogonal function (EOF) analysis using the individual quantification, and the interpretations made by
diagnostic pigment data set after removing the mean and Anderson et al. (2008) are fully supported by the present
standardizing to unit variance. An EOF analysis decom- analysis.
poses spatial and temporal variability of a data set
containing several variables into a set of independent Results
orthogonal functions, or modes (Emery and Thomson
1997). The modes of variability determined for the Oceanographic conditions—Sea surface temperature
diagnostic pigment data represent a phytoplankton ‘‘com- (SST) during this study ranged from 10uC to nearly 22uC
munity,’’ and the amplitude function associated with each (mean 14.7uC), where maximum temperatures occurred in
mode indicates the intensity of presence of the given the early fall of each year and the minimum temperatures
community (Anderson et al. 2008). We chose to use the occurred in the spring (Fig. 3a). This is driven by
EOF method rather than CHEMTAX method (Mackey upwelling-favorable winds in the spring causing the vertical
et al. 1997), a commonly used chemotaxonomic method transport of cool, nutrient-rich waters to the surface and932 Barrón et al. Fig. 3. Surface seawater properties at Sta. 4: (a) Temperature. (b) NO3 + NO2 concentration. (c) Chl a concentration determined via HPLC. (d) BSi concentration. more stratified conditions with a shallow, warm surface highest in the spring, also consistent with upwelling- layer in the fall (Toole and Siegel 2001; Brzezinski and favorable conditions (Fig. 3b). Surface Chl a concentra- Washburn 2011). Although seasonal patterns of upwelling tions determined via HPLC analysis are shown in Fig. 3c. and stratification occur, physical mixing processes in the The mean Chl a concentration was 2.96 mg L21, and the SBC are more so influenced by lateral advection driven by median was 2.06 mg L21. An extremely high concentration wind forcing and relaxation along the California coast of 28.3 mg L21 was observed in May 2008 and was (Harms and Winant 1998; Washburn et al. 2011). Synoptic coincident with high levels of biogenic silica (Fig. 3d), scale wind forcing and relaxations have a significant effect indicating the presence of diatom populations (Shipe and on changes in biogeochemistry, such as nutrient status, in Brzezinski 2001; Krause et al. 2013). Patterns in biogenic the SBC. Brzezinski and Washburn (2011) found that wind- silica tended to mimic patterns in Chl a through much of driven upwelling was responsible for the highest levels of the time series. Diatom populations are typically abundant phytoplankton productivity and nutrient concentrations in during or just following periods of strong upwelling the SBC and that cyclonic eddies enhance productivity either accompanied by higher rates of primary productivity and by entrainment of upwelled water, isopycnal uplift, or a Chl a concentrations, whereas increased dinoflagellate combination of both. Eddy-enhanced productivity was most abundance has been observed in response to more shallow prominent in the fall (Brzezinski and Washburn 2011). eddy-driven mixing processes together with low levels of Dissolved nitrate + nitrite (NO3 + NO2) concentrations dissolved silicate concentrations (Anderson et al. 2008; at the sea surface, referred to hereafter as nitrate, were Brzezinski and Washburn 2011).
Santa Barbara Channel optical properties 933 Fig. 4. Eigenvectors for the four dominant EOF modes. The percentile listed next to the titles are the percent of the phytoplankton diagnostic pigment covariability captured by the given mode. The numbers listed above the eigenvector bars are r2 3 100 of the linear regression between the respective pigment concentrations and the amplitude function for the given mode. Phytoplankton community structure—Phytoplankton com- strongest form in either the positive or the negative munity structure was quantified using an EOF analysis of direction. The closer the AF values are to zero, the less phytoplankton pigment concentrations following Ander- relevant that EOF mode is for that time and location. son et al. (2008). The first four EOF modes explained 82% Extreme AF values are most frequently observed at Sta. 5 of community structure variability. Figure 4 shows bar and 6; these stations are near the center of the cyclonic plots of the eigenvector loadings for the four most eddy frequently present in the SBC and thus located significant EOF modes. The phytoplankton community where upwelling is thought to persistently occur (Harms composition associated with each mode is interpreted by and Winant 1998; Washburn et al. 2011). the relative value of the eigenvectors (height of the bars) Mode 1 of the EOF analysis captured 37% of the and the relationship between the individual pigment covariability of the pigment data (Fig. 4). The eigenvector concentrations and the EOF amplitude functions for each loadings for Mode 1 were positive for all pigments, mode (i.e., the values of 100 3 r2 are the numbers above indicating that all pigment levels increase and decrease in each bar). The amplitude function (AF) for each mode is concert as the Mode 1 AF changes. The amplitude a value that indicates the intensity of the overall pattern functions are well correlated with indicator pigments for of community structure for every time and space point the functional types: green flagellates, prochlorophytes, analyzed (Fig. 5). The dashed lines in Fig. 5 indicate the chromophytes, nanoflagellates, and others in the nano- first and second standard deviations of the mean AF for plankton size class (Table 1). AF did not correlate well with each mode. Data outside these lines represent extreme AF Chl a (r2 5 0.05), indicating that large changes in Chl a values and are interpreted as community presence in its (i.e., blooms) are not typically associated with these
934 Barrón et al. Fig. 5. Amplitude functions for the four dominant EOF modes. Figures highlight the mean and median AF as well AF 6 1 standard deviation and 62 standard deviations from the mean. phytoplankton assemblages. These results are broadly A positive correlation was also seen with peridinin (r2 5 consistent with the results of Anderson et al. (2008), who 0.16), the marker pigment for dinoflagellates. Hence, this interpreted their Mode 1 as an early upwelling nanoplank- mode was interpreted as a ‘‘mixed’’ microplankton commu- ton community. Mode 1 AF did not correlate well with nity, as the fucoxanthin and peridinin eigenvectors have the nutrient concentrations, with the exception of a weak same sign, indicating the co-occurrence of diatoms and correlation with SiO4 (r 5 0.22), and only a weak dinoflagellates. The eigenvectors for many of the indicator correlation is found with SST (r 5 20.16). This shows pigments of the smaller functional types, such as prochlor- that the present Mode 1 community possesses some of the ophytes, cyanobacteria, nanoflagellates, chromophyte, and traits of those found in the previous study (Table 2). green flagellates, were negative, showing that the smaller However in this study, extreme AF values of Mode 1 did phytoplankton groups alternate in importance with micro- not seem to occur seasonally (e.g., in early spring, such as in plankton. Mode 2 AF showed a strong negative correlation Anderson et al. 2008) but were present at various times of with SST and strong positive correlations with B-Si, POC, the year (Fig. 5 Mode 1). Here, Mode 1 was interpreted as Chl a, and the MAA index (Table 2). High values of the a baseline community rather than a more characteristic MAA index indicate the optical presence of MAAs in the upwelling indicator. Differences between Anderson et al. dissolved phase of the seawater. Positive extreme values of (2008) and the present study may simply be due to lack of Mode 2 AF (e.g., indicating strong microplankton presence) overlap in the observational periods for the two studies. occurred mostly in the spring, further aiding our interpre- The EOF loadings for Modes 2 through 4 occurred in tation of Mode 2 as the spring bloom mode. Negative both the positive and the negative direction, indicating extreme AF values of Mode 2 indicated the lack of opposing relationships among some of the pigments. Mode microplankton in the community, the strongest of which 2 captured 22% of the covariability of the pigment data set. occurred in the late spring and early summer of 2009 (Fig. 5 Mode 2 AF values correlated strongly with Chl a (r2 5 Mode 2). 0.82) and fucoxanthin (r2 5 0.72) concentrations, suggest- Mode 3, representing 13% of the variance in the pigment ing that diatom blooms drive variability of this second mode. data set, was interpreted as a microplankton community
Santa Barbara Channel optical properties 935
Table 2. Linear regression results from EOF AF vs. seawater properties. Bold numbers
indicate significant results with p # 0.05.
Linear regression correlation coefficient (r)
Property Mode 1 Mode 2 Mode 3 Mode 4 n
Salinity 0.04 20.11 0.46 20.34 207
Temperature 20.16 20.50 20.15 0.49 210
NO3 + NO2 0.01 0.07 0.03 20.50 195
PO4 0.13 0.11 0.02 20.55 195
SiO2 0.22 0.13 20.10 20.41 195
L-Si 20.09 0.21 20.03 20.05 225
B-Si 20.16 0.40 0.77 20.15 225
POC 20.06 0.55 0.82 0.15 125
Chl a 0.14 0.54 0.70 20.04 225
MAA 0.05 0.64 20.32 0.24 225
with alternating dominance between diatom (positive) and IOPs—Mean component absorption spectra from PnB
dinoflagellate (negative) populations. Negative loadings for cruises (November 2005–August 2009) are shown in Fig. 6.
this mode indicated a phytoplankton community dominat- Variability about the mean is displayed here as one
ed by dinoflagellates (Mode 3 AF vs. peridinin, r2 5 0.52), standard deviation (shaded areas). The phytoplankton
and positive loadings indicated a community dominated by absorption coefficient (aph[l]) was more variable at shorter
diatoms (Mode 3 AF vs. fucoxanthin, r2 5 0.25). Mode 3 wavelengths and particularly variable in the UV portion
AF showed strong positive correlations with concentra- of the spectrum. The absorption peaks at aph(440) and
tions of B-Si, POC, Chl a, and salinity (Table 2), consistent aph(675) nm are due to Chl a absorption. The shoulders
with conditions during diatom blooms in the SBC seen in the mean aph(l) spectrum surrounding the Chl a
(Anderson et al. 2008; Brzezinski and Washburn 2011). blue peak are due to accessory pigments, mostly caroten-
Mode 3 AF correlated well in the negative direction with oids. Variability about the mean is a result of the variability
MAA index, thus indicating a positive relationship with the in phytoplankton community composition and abundance
dinoflagellate-dominated community and MAAs (see and can be attributed to the presence of different accessory
discussion to follow). Negative extreme values of Mode 3 pigments contained by various species (e.g., diagnostic
co-occurred with positive extreme values of Mode 2 in the pigments) as well as size differences between phytoplankton
spring of 2006 and again in the winter of 2006–2007, groups (Nelson et al. 1993; Ciotti et al. 2002; Bricaud et al.
indicating dinoflagellate-dominated blooms. Positive ex- 2004). The above previous studies have shown that lower
treme AF of Mode 3 co-occurred with Mode 2 positive AF absorption coefficient values as well as flatter spectra have
in the spring of 2007 and, more so, in 2008, indicating the been observed by microplankton, whereas smaller phyto-
strong dominance of diatoms during these blooms. plankton groups have shown higher overall absorption
Mode 4 captured 10% of covariability in the data set. coefficient values with sharper absorption peaks. This
Values of the Mode 4 AF were well correlated with SST reduction in the magnitude and flattening of phytoplank-
and inversely correlated with nutrient concentrations ton absorption spectra as intracellular pigment concentra-
(Table 2). Mode 4 showed positive extremes in the tions or cell size increase is referred to as the package effect
amplitude function during stratified conditions and nega- (Morel and Bricaud 1981; Nelson et al. 1993). The effects of
tive extremes when turbulent mixing was occurring. pigment packaging can be better assessed by normalizing
Indicator pigments most highly correlated with AF of the spectra to the Chl a concentration (aph*[l] 5 aph[l]/[Chl
Mode 4 were zeaxanthin (r2 5 0.38), an indicator for a]; Fig. 6b). Here, the standard deviations fit more tightly
picoplankton, and violaxanthin (r2 5 0.32), a photopro- around the mean spectrum, particularly at higher wave-
tective pigment. Mode 4 was interpreted as a picoplankton- lengths. The change in mean spectral shape and tightening
dominated or stratified mode, and the most strongly of the standard deviation for aph(l)* reflect the strong
positive AF occurred in the summer and negative AF in presence of microplankton in the SBC due to the high Chl a
the winter. The eigenvector for peridinin was also on the concentrations associated with microplankton groups.
positive side, while the eigenvectors for all of the other The average detrital absorption coefficient spectrum,
phytoplankton functional types were opposite. It makes ad(l), increases toward shorter wavelengths (Fig. 6c).
sense that the indicator pigments for dinoflagellates covary Overall, ad(l) accounts for only a small portion of the
positively with those for other stratified functional types, as total particle absorption. Increasing standard deviations
they are also found in the summer–fall in the SBC toward decreasing wavelengths are indicative of changes in
(Brzezinski and Washburn 2011). Mode 4 AF correlated the spectral decay slope within the data set. The mean
well in the positive direction with temperature and MAA CDOM absorption coefficient, ag(l), also decays with
index and negatively with dissolved nutrients, further increasing wavelength (Fig. 6d). Mean values of ag(400) are
confirming the relationship with warm, stratified condi- much greater than the other component absorption
tions (Table 2). coefficients following global patterns (Nelson and Siegel936 Barrón et al. Fig. 6. Average absorption properties of the SBC for the study period: (a) Phytoplankton absorption (aph(l)). (b) Phytoplankton- specific absorption (aph*(l)). (c) Detrital absorption (ad (l)). (d) CDOM absorption (ag(l)). 2013). The standard deviations in ag(l) are for the most phytoplankton species (also observed through laboratory part relatively small, with the exception of a bulge found studies; Whitmire et al. 2010). between l 5 300 and 400 nm. This is due to peaklike Remote-sensing reflectance spectrum, Rrs(l), is a func- features near l 5 335 nm in several samples from the PnB tion of the absorption and backscattering spectra presented cruises (see Fig. 2 for an example). above. The values of Rrs(l), obtained in situ using the PRR Mean particulate total and backward scattering spectra instrument, were indicative of a productive, coastal ocean, are relatively flat, yet are highly variable in magnitude where the reflectance is highest in the green portion of (Fig. 7a, b). Average values of particle scattering coeffi- the spectrum (Fig. 7d). The data presented here are a cient, bp(l), of the surface ocean resulted in a fairly flat subset of the PnB IOP data set used in Kostadinov et al. spectrum with values ranging spectrally between 0.45 (2012) as only Horn Point Laboratory–analyzed HPLC and 0.5 m21 (Fig. 7a). Average particulate backscattering data are used. coefficient spectra, bbp(l), ranged from 0.0039 to 0.0059 m21. The spectral shape for bbp(l) observed in this Relationship of IOPs to phytoplankton community study is typical for coastal, eutrophic regions (Kostadinov structure—The data series of each IOP were linearly et al. 2012). Mean values of the particulate backscattering regressed with the AFs for each of the EOF modes where ratio, bbp : bp, often referred to as b̃bp(l), were 1–1.2% with a the AF values are used as a proxy for different slight decreasing trend in the blue spectral region (Fig. 7c). phytoplankton communities. Linear regression analyses Values of bbp : bp is governed by the index of refraction and were done for every wavelength of the IOPs in effort to the slope of particle size distribution, where low values of characterize spectral relationships between the optical bbp : bp are indicative of large particles. The values presented parameters and phytoplankton community structure. here (Fig. 7c) and in other studies for the SBC (Kostadinov Relationships were examined using slope diagrams, where et al. 2010, 2012; Antoine et al. 2011) are typical for coastal the y-axis in Figs. 8 and 9 display the slope, m, for the systems where the particle load is dominated by larger linear regression, y 5 mx + n, of IOP vs. AF data series for
Santa Barbara Channel optical properties 937 Fig. 7. Average particle scattering properties and average ocean reflectance in the SBC for the study period: (a) Total particle scatter (bp(l)). (b) Particle backscatter (bbp(l)). (c) Particle backscatter normalized by total particle scatter (bbp(l) : bp(l)). (d) Remote-sensing reflectance measured in situ with the PRR (Rrs(l)). each available wavelength (following Kostadinov et al. fraction of the absorption spectrum when the phytoplank- 2007). Figure 8 show the spectral slope values for the ton community was comprised of mostly nanoplankton. regression between aph(l), ad(l), aph*(l), and ag(l) and the Linear regressions with Mode 2 AF were significant top four EOF AF values (maph[l], mad[l], maph*[l], and for all absorption properties over large portions of the mag[l]; in the following, we will denote the slope spectra spectrum. The slope diagrams for maph and mad display without the explicit spectral notation). Similar regressions positive relationships for the entire spectra, indicating that were performed between bbp(l), bp(l), and Rrs(l) observa- values of absorption due to phytoplankton (e.g., ‘‘living’’ tions and the pigment EOF AF values (Fig. 9). The n-value particles), as well as absorption due to nonliving particles, shown in the figures indicates the number of independent were higher overall when microplankton groups dominated observations used in each linear regression. The linear the phytoplankton community. The spectral shape of the relationships represented by the slope diagrams were maph and mad slope diagrams mimic the shape of the interpreted to be significant when the 95% confidence average absorption spectra, reinforcing the notion that intervals surrounding the regression slope values are microplankton groups are strongly influencing these divergent from zero. properties. Values of aph*(l) were negatively related to Slope spectra for the Mode 1 AF (the nanoplankton Mode 2 AF at l 5 663 nm and l 5 , 380–500 nm, community) showed significant positive relationships with capturing the Chl a absorption peaks as well as some of the aph(l) and aph*(l) at approximately l 5 400–500 nm and l surrounding accessory pigments. Negative slopes at the Chl 5 660–685 nm, capturing the regions of maximum Chl a a absorption peaks are indicative of the package effect; that absorption. Mode 1 mag and mad were not significantly is, chlorophyll-specific phytoplankton absorption decreases different from zero at any wavelength, indicating that there because of the package effect. The package effect is a were no significant relationships between the nonliving common physiological strategy for large phytoplankton
938 Barrón et al. Fig. 8. Slope diagrams for linear regressions of various absorption properties vs. EOF AF for each mode, where mIOP refers to the term m in the linear equation y 5 mx + n. species, such as diatoms (Kirk 1994). Values slopes were generally interpreted as positive linear rela- of maph* were positive for wavelengths approximately less tionships of IOPs vs. diatoms, and negative linear than l 5 380 nm. This is likely due to the presence of regression slopes were interpreted as positive relationships MAAs associated with one or more of the microplankton of IOPs vs. dinoflagellates. However, some logical excep- groups. tions were made below, as the EOF leaves room for Positive linear relationships were observed for Mode 2 objectivity. Mode 3 maph were significantly positive for mag starting l 5 400 nm and continuing into the UV wavelengths l 5 400–700 nm with the characteristic portion of the spectrum. The slope diagram does not reflect phytoplankton absorption spectrum. This portion of the that of the average CDOM absorption spectra (Fig. 6d) but slope diagram had a similar shape to that for Mode 2, does show ‘‘peaks’’ in this relationship around l 5 335 nm. indicating that diatoms largely influence the relationship The increased absorption in the UV wavelengths, in both for this wavelength range. At wavelengths less than 400 nm, the particulate and the dissolved phase, may have been due the linear regression slope became negative and showed a to MAAs. The linear relationships indicated a positive significant relationship at about l 5 360 nm. UV peaks correlation of the MAA index with microplankton groups. (e.g. MAAs) were observed in several CDOM spectra, and The EOF loadings for Mode 3 indicated an alternation the relationship here indicates that increased UV absorp- in community dominance between diatom (positive) and tion was associated with the dominance of dinoflagellate dinoflagellate (negative) groups. Alternating positive and groups. negative AF was interpreted as the progressive alternation Values of Mode 3 maph* were significantly negative for of these groups in time and space; thus, the interpretation the entire spectra and from 400 to 700 nm look similar to of the slope diagrams was so. Positive linear regression that for Mode 2. For wavelengths less than 400 nm, the
Santa Barbara Channel optical properties 939 Fig. 9. Slope diagrams for linear regressions of various scattering properties and Rrs(l) vs. EOF AF for each mode, where mIOP refers to the term m in the linear equation y 5 mx + n. slope diagram shows a strengthening negative relationship Significant positive linear regression slope spectra for the into the UV portion of the spectrum, indicating a positive absorption properties vs. Mode 4 AFs were observed for correlation with dinoflagellates and UV absorption. Based aph(l), aph*(l), and ag(l) only in the UV portion of the on the shape of the slope diagram for Mode 3 maph, the spectrum. The slope diagrams for particulate absorption aph(l) relationship is likely reflecting the pigment packaging properties make sense for this mode, which is indicative of effect due to diatoms from l 5 400 to 700 nm rather than high-light-adapted, small phytoplankton groups. Many increased phytoplankton-specific absorption due to dino- high-light-adapted species contain MAAs in their cyto- flagellate presence. Negative maph and maph* in the UV plasm (Roy et al. 2011). The shape of the slope diagram for portion of the spectrum, however, are more likely indicative mag looks like an inverted version of the slope diagram for of the strong UV-absorbing capability of some dinoflagel- Mode 3; for example, the linear relationship strengthens late species and therefore are interpreted as positive linear further into the UV portion of the spectrum, indicating a relationships between aph(l) and aph*(l) and dinoflagellate strong relationship with CDOM absorption coefficient. functional type. The linear regressions for ag(l) vs. AF also Linear regression spectral slopes for particle scattering resulted in significantly negative mag in the UV portion of properties vs. AF and Rrs(l) vs. AF are shown in Fig. 9. the spectrum. Slight peaklike features around l 5 335 and Mode 1 mbbp was not significant at any wavelength; 360 nm, like those in Mode 2, are observed also in the however, mbp were significantly positive, and mbbp : bp were negative direction for this mode. The shape of the slope significantly negative across all wavelengths. This shows diagram slightly resembles the shape of the average CDOM that the nanoplankton community is significantly corre- absorption spectra in that the relationship increases fur- lated to higher total scattering and that backscattering ther into the UV rather than decreasing as for Mode 2. efficiency is lower when these groups are present. mbbp was This shows that both the baseline CDOM concentrations positive across the entire spectra for Mode 2, indicating and the UV peaks are correlated with dinoflagellate that larger backscattering coefficient values are observed presence. when the phytoplankton community is dominated by
940 Barrón et al.
microplankton and Chl a concentrations are high. Values the optical environment in different ways. Alvain et al.
of Mode 2 mbp were significantly positive at longer (2005) statistically examined accessory pigment concentra-
wavelengths. Mode 2 AF also showed a significant positive tions in relation to in situ LwN(l) and Rrs(l) from satellite
relationship with bbp : bp at all wavelengths. Mode 3 mbbp measurements, so their method is also not limited by
values were significantly positive for wavelengths 442, 589, predetermined ratios (Alvain et al. 2005, 2006). However,
and 671 nm. Mode 3 spectra for mbp were significantly peridinin concentrations are below detection in the open
positive for the entire spectral range, while mbbp : bp was ocean (Case 1), and the effects of dinoflagellates on light
weakly significantly negative for only 510 nm. All of the reflectance and satellite-derived chlorophyll were omitted
linear regressions for scattering properties with Mode 4 AF in those studies.
were insignificant. The linear regression results for Mode 1 AF vs. the
Linear regressions for Rrs(l) vs. AF resulted in signifi- various IOP parameters showed that changes in the Chl a
cantly negative relationships in the blue and green part of concentrations were only slightly represented by this
the spectrum for Modes 1, 2, and 3. mRrs for Modes 2 and 3 community, a nanoplankton-dominated community, by
were both significantly positive at l 5 656 nm. Rrs(l) weak positive correlations with aph(l) and aph*(l) from
to first order is related to the ratio of backscattering approximately l 5 400–500 nm. However, detrital and
coefficient divided by the absorption coefficient. The first CDOM absorption properties were not well constrained by
three modes are consistent with the results of changes the dynamics of this mode at all. This was not surprising
because absorption is the dominant process. AF for Mode considering that Mode 1 did not include any bloom events,
4 looks more like a backscattering signal with its positive which would be expected to have a higher effect on changes
slope values throughout the spectrum (Toole and Siegel in detritus and carbon biogeochemistry in the surface
2001). High AF values for Mode 4 indicate a higher water. Similarly, bbp(l) was also not constrained by Mode 1
picoplankton presence but also a clearer ocean. The AF. Antoine et al. (2011) used data from the SBC as well
relationship with Rrs(l) could simply be due to lower as from the Mediterranean Sea and found that bbp(l)
absorption. Backscattering is also driving variability in was a good indicator for Chl a and thus phytoplankton
Rrs(l) for the longest wavelengths in the slope spectra with abundance only when strong changes in abundance
Mode 2 and 3 AFs. occurred, such as a bloom or a dilution event. Their results
relate well to Mode 1 results, considering that it is rep-
Discussion resenting background community assemblage.
IOPs were most strongly correlated with the importance
Comparison with previous studies—The results above of the microplankton groups (Mode 2) and the differences
show that changes in phytoplankton community structure between diatom and dinoflagellate functional types (Mode
will affect IOPs. An EOF analysis of a suite of phyto- 3). Characteristics of the diatoms captured in the IOP
plankton pigment concentrations was performed, showing properties were consistent with the package effect, as
that the first four modes accounted for 82% of the shown by the negative linear relationships with aph*(l), yet
variability in the pigment data. The four modes of positive correlations were observed with the scattering
variability presented encompassed all of the phytoplankton coefficients bbp(l) and bp(l). Positive correlations with
community regimes in the SBC, consistent with previous scattering properties were somewhat counterintuitive. Mie
studies (Toole and Siegel 2001; Anderson et al. 2008; theory shows that larger homogeneous spheres will scatter
Brzezinski and Washburn 2010). Several recent studies proportionally more light in the forward direction than
have related IOPs and remote-sensing reflectance to smaller ones. In general, larger phytoplankton may have
chemotaxonomic phytoplankton functional types and/or lower indices of refraction than smaller ones (e.g., Stramski
size classes determined via HPLC pigment concentrations. 1999). However, Mode 2 was most highly correlated with
Bricaud et al. (2004) found that deviations from average Chl a concentrations, and the higher backscatter associated
aph*(440) in oceanic waters were driven by phytoplankton with this mode may have been a reflection of the sheer
size classes, determined using the methods outlined in increase in phytoplankton abundances from baseline
Vidussi et al. (2001) and Uitz et al. (2006). Kostadinov et al. conditions (Antoine et al. 2011). Several recent studies
(2012) used size classifications also determined from have found that larger phytoplankton size classes contrib-
phytoplankton pigment concentrations to explain the ute much more to bulk backscattering than previously
variability of particle scattering in the SBC. The novelty thought (Dall’Olmo et al. 2009; Westberry et al. 2010;
of the approach of this study is that it objectively analyzed Whitmire et al. 2010). Mode 2 was also correlated
phytoplankton community variability using the covariance positively with ad(l) and ag(l), indicating an increase in
structure contained within the pigment data rather than detrital particulates during blooms. The positive correla-
preset ratios. This approach splits the microplankton size tions of bbp(l) with Mode 2 likely were caused by a
class into two separate regimes (e.g., EOF Mode 3), combination of phytoplankton-sized particles, detritus, and
whereas in the phytoplankton pigment concentration size any other increased particle abundance (e.g., bacteria as
class approaches, the diatom and dinoflagellate functional consumers in response to increased detritus) that may have
groups are often combined into a single class. Diatom and covaried with bloom conditions.
dinoflagellate functional types may play very different roles The influence of dinoflagellates on IOP properties is
in aquatic environments with regard to biogeochemical manifested in numerous ways, as shown through positive
cycling (Nair et al. 2008) and, as shown in this work, affect correlations, presented as negative correlations due to theSanta Barbara Channel optical properties 941
sign of the amplitude function, with the Mode 3 AF vs. et al. 2010). Therefore, the peak-like features are referred
aph(l) and aph*(l) in the UV region, positive correlation to as MAAs in this article.
with ag(l) in what appears to be a UV-absorbing substance Residual CDOM spectra, a step in quantifying the MAA
as well as background CDOM absorption, and positive index (see Methods section above; Fig. 2b), surprisingly
correlations to bbp(l) at select wavelengths. Whitmire et al. revealed characteristic peaks at very low absorption in
(2010) found that dinoflagellate species had higher back- many spectra, totaling 55% of the PnB data used for this
scattering ratios than diatom species, which is consistent study. Values of the MAA index were significantly
with the results found in this article considering the slight correlated with Modes 2–4 (Table 2). The strongest
negative correlation of Mode 3 AF with bbp : bp (510). The relationship occurred with Mode 2 (r 5 0.64), the mixed
authors attribute their observations to the complex cellular microplankton mode; however, the negative correlation
composition and dense DNA content of dinoflagellates. with Mode 3 (r 5 20.32) indicates that dinoflagellates
Vaillancourt et al. (2004) found that dinoflagellates had the drive the relationship. The MAA index is also correlated
highest bbp : bp(l) values of 20 species of phytoplankton well with peridinin concentrations, r2 5 0.70 (Fig. 10),
examined, consistent with the analysis on the EOF modes supporting the speculation that the MAAs are related to
in this study. Overall, this study was able to constrain much dinoflagellates. It was also not surprising that a significant
of the variability in the IOPs based on phytoplankton correlation was observed with Mode 4 (r 5 0.24), as this
community structure. The main microplankton functional mode indicates stratified, high light conditions—conditions
types in the SBC, diatoms and dinoflagellates, explained that would be prime for species that are able to produce
notable differences in all IOPs. MAAs as a UV-shading mechanism.
Many phytoplankton species contain MAAs for pre-
Phytoplankton community and CDOM absorption sumably photoprotective purposes, including the pico-
spectra—Absorption properties in both the particulate plankton Prochlorococcus and the harmful algal diatom
and the dissolved phase were strongly influenced by Pseudo-nitzschia (Roy et al. 2011). However, dinoflagellates
phytoplankton community structure in the UV portion are the only functional group documented to date that have
of the spectrum, as shown in Fig. 8. This feature was co-occurred with dissolved MAAs in the water column
particularly noteworthy for CDOM absorption coeffi- (Vernet and Whitehead 1996; Tilstone et al. 2010). In the
cients, with stronger correlations peaking between l 5 300 EOF analysis, peridinin is found on the positive side of our
and 400 nm. The characteristic was due to the presence of eigenvectors for Mode 4—the mode that signals the
peak-like features in the CDOM spectra that were presence of a stratified water column. Although peridinin
observed as deviations from the baseline CDOM curve is weakly correlated with the Mode 4 AF (r2 5 0.09), it is
and peak around l 5 335 nm, and often (but not always), influential only for Mode 3 (r2 5 0.52), indicating how
a peak is found at approximately l 5 360 nm as well (see prevalent this functional group is in the SBC.
example in Fig. 2). This spectral region is where the largest
variability among the ag(l) spectra was found (Fig. 6d). Roles of phytoplankton community structure on remote-
Large peaks (e.g., those visible to the eye in individual sensing retrievals—Coastal areas can be optically complex
spectral plots) were found in 33% of the CDOM spectra in due to a number of factors, including terrestrial runoff,
the PnB data set used for this analysis. The peaks may be phytoplankton blooms, and, as the SBC data set has
indicative of a UV-absorbing substance related to, and shown, high variability in the phytoplankton community
potentially derived from, phytoplankton, as they co- structure. Model performance for ocean color remote
occurred with strong UV absorption in the aph(l) spectra sensing in coastal areas can sometimes be improved by
as well. Vernet and Whitehead (1996) observed increased local calibration. That is, site-specific characterization of
UV absorbance in the particulate and dissolved phases IOPs in coastal areas can be incorporated into the models,
during a red-tide bloom of the dinoflagellate Lingulodinum such as the GSM model that requires IOP spectral slopes as
polyedra off the southern California coast in conjunction constants. In an attempt to improve model performance,
with increased MAA concentrations. Absorbance peaks Kostadinov et al. (2007) executed local calibration of the
were observed at ap(360) as well as ag(360) for samples GSM and OC4 models for the SBC. However, the locally
collected in situ during the bloom, and a shoulder in the tuned models did not perform substantially better than the
absorbance spectra for filtrate of growth media was globally tuned models. Here we examine the roles that
observed around l 5 310 nm when the species was grown phytoplankton community structure may play on bio-
in isolation. Recently, Tilstone et al. (2010) observed optical model performance. The globally tuned GSM
increased UV absorption for in situ CDOM on the Iberian model retrieved acdm(443), bbp(443), and Chl a concentra-
Peninsula with similar spectral shoulders as observed in tion using measured Rrs(l) spectra. Measured reflectance
this study. The UV peaks observed in their study spectra were also used to retrieve Chl a concentrations
correlated well with increased MAA concentrations, using the empirical OC4v6 algorithm. Model data residuals
peridinin concentrations, and dinoflagellate presence. were calculated by comparing the measured IOPs or Chl a
There are no measurements of MAAs for the PnB data concentrations with the modeled parameters. Figure 11
set, but it is likely that the peaks in UV absorption shows the data model residuals, where negative residuals
observed in this study were due to the presence of MAAs indicate an overestimation from the model and a positive
based on the similarity to previous studies (Vernet and residual indicates model underestimation of the given
Whitehead 1996; Whitehead and Vernet 2000; Tilstone constituent.You can also read
