Taylor etal 2013

Taylor etal 2013
JOURNAL OF GEOPHYSICAL RESEARCH: OCEANS, VOL. 118, 1–13, doi:10.1002/jgrc.20201, 2013
Estimation of relative phycoerythrin concentrations from
hyperspectral underwater radiance measurements––A statistical
Bettina B. Taylor,1 Marc H. Taylor,1 Tilman Dinter,2 and Astrid Bracher1,2
Received 22 August 2013; revised 11 April 2013; accepted 15 April 2013.
[1] Phycobiliproteins are a family of water-soluble pigment proteins that play an important
role as accessory or antenna pigments and absorb in the green part of the light spectrum
poorly used by chlorophyll a. The phycoerythrins (PEs) are one of four types of
phycobiliproteins that are generally distinguished based on their absorption properties. As
PEs are water soluble, they are generally not captured with conventional pigment analysis.
Here we present a statistical model based on in situ measurements of three transatlantic
cruises which allows us to derive relative PE concentration from standardized hyperspectral
underwater radiance measurements (Lu). The model relies on Empirical Orthogonal
Function (EOF) analysis of Lu spectra and, subsequently, a Generalized Linear Model with
measured PE concentrations as the response variable and EOF loadings as predictor
variables. The method is used to predict relative PE concentrations throughout the water
column and to calculate integrated PE estimates based on those profiles.
Citation: Taylor, B. B., M. H. Taylor, T. Dinter, and A. Bracher (2013), Estimation of relative phycoerythrin concentrations from
hyperspectral underwater radiance measurements––A statistical approach, J. Geophys. Res. Oceans, 118, doi:10.1002/jgrc.20201.
1994; Six et al., 2007; Zhao et al., 2011]. In addition to the
covalently bound chromophores, several noncovalent interactions between the apoprotein and the chromophores are
essential for the light-harvesting function and influence the
spectral properties of the proteins. Generally, four types of
phycobiliproteins are distinguished on the basis of their
absorption maxima (Absmax): phycoerythrocyanins (PEC,
Absmax 575 nm), phycoerythrins (PE, Absmax 495–575
nm), phycocyanins (PC, Absmax 615–640 nm) and allophycocyanins (APC, Absmax 650–655 nm). Cryptophyte
biliproteins are often classified as a separate group. In this
paper, we will concentrate on the PEs.
[4] PEs carry only one or two chromophore types, phycoerythrobilin (PEB, Absmax 545 nm) and phycourobilin
(PUB, Absmax 495 nm). They exhibit very diverse spectral characteristics depending on their molecular structure
which varies among phytoplankton species and with environmental conditions such as light and nutrient availability
[Haverkamp et al., 2009; Hoge et al., 1998; Lantoine and
Neveux, 1999; Palenik, 2001; Sidler, 1994]. The spectral
characteristics of PEs are strongly influenced by the number and proportions of PUB and PEB, the chemical structure and geometry of the molecule, and their direct local
environment, which can lead to varying absorption maxima
[MacColl, 1998]. PEB is found in all PEs, whereas PUB is
present in varying concentrations in some forms of PE
[Glazer, 1999; Six et al., 2007, Wood et al., 1998].
Generally, species with high-PUB content are more abundant in clear oligotrophic waters of the open ocean,
whereas low-PUB or no-PUB PEs characterize species that
are native in more eutrophic coastal waters [Chekalyuk and
Hafez, 2008; Lantoine and Neveux, 1999; Wood et al.,
1998]. PEs generally have a single fluorescence emission
Description of Phycoerythrins
[2] The phycobiliproteins, a group of water-soluble,
brightly colored proteins, are the major light-harvesting
pigments of cyanobacteria, red algae and cryptophytes
[Sidler, 1994; Zhao et al., 2011]. They consist of openchain tetrapyrroles known as phycobilins or bilins, that are
covalently bound to the apoproteins via thioether bonds
and they efficiently absorb light in the green part of the
light spectrum poorly used by chlorophyll a (chl a) [French
and Young, 1952; Glazer et al., 1982; Kronick, 1986]. The
colors of the phycobiliproteins originate mainly from the
tetrapyrrole chromophores [Sidler, 1994]. In intact cells of
cyanobacteria and chloroplasts of red algae, the phycobiliproteins are generally assembled in structures called phycobilisomes which are attached in regular arrays to the
external surface of the thylakoid membranes [Gantt and
Conti, 1966; Glazer, 1988; Sidler, 1994].
[3] Phycobiliproteins are usually classified according to
their spectral properties and phylogenetic occurrence,
although the classifications are constantly revised and discussed in the literature due to great spectral diversity and
variation [e.g., Glazer, 1999; Hill and Rowan, 1989; Sidler,
Alfred Wegener Institute for Polar and Marine Research, Bremerhaven,
Institute of Environmental Physics, University of Bremen, Bremen,
Corresponding author: B. B. Taylor, Department of Climate Dynamics,
Alfred Wegener Institute for Polar and Marine Research, Bussestrasse 24,
Bremerhaven D-27570, Germany. ([email protected])
©2013. American Geophysical Union. All Rights Reserved.
obtained from methods such as laser-induced emission
measurements from airborne platforms [Hoge et al., 1998]
or as shipboard-device [Chekalyuk and Hafez, 2008; Chekalyuk et al., 2012]. However, these measurements require
very sophisticated instrumentation that is not easily
acquired and handled and have been confined to underway
surface measurements or water samples from discrete
depths. Automated submersible flow cytometers have been
developed [Dubelaar et al., 1999; Olson et al., 2003] and
used as moored devices measuring time series, but, as far
as we are aware, not for profiling. Some submersible and
profiling fluorometers exploiting the fingerprints of the specific excitation spectra of PE-containing phytoplankton
have been designed [Beutler et al., 2002, 2004; Cowles et
al., 1993; Desiderio et al., 1997; Horiuchi and Wolk,
2008] and some of these PE fluorometers are commercially
available. These profiling fluorometers have been mainly
deployed in freshwater environments [Beutler et al., 2002;
Leboulanger et al., 2002; Proctor and Roesler, 2010].
Thus, even though PEs and PE-containing phytoplankton
have been subject to many studies, there is still a lack of information about the depth distribution in marine ecosystems and the deep maxima are often overlooked. Here we
demonstrate a method which deduces profiles of total PE
down to the depth of measurable light availability.
[7] The spectral diversity of the PEs opens many possibilities to distinguish between PE-containing groups as has
been shown with many of the methods mentioned above,
but also carries many difficulties, when only specific wavelengths are used for the excitation and detection of the different groups. Ample research has shown that PE content
and spectral properties can vary with light availability (thus
also with depth) and nutrient status [Lantoine and Neveux,
1999; Wood et al., 1998; Wyman et al., 1985]. For example, in some species PE is not only used in photosynthesis,
but also as nitrogen storage [Wyman et al., 1985]. All the
methods mentioned above use some kind of fluorescence
excitation to detect PEs; some methods use only one excitation and emission wavelength (for example many flow
cytometers), others use several excitation and/or emission
wavelengths to distinguish between spectral varieties of
PE. As input parameter into our model, we use PE values,
determined by a method that takes the whole diversity of
PEs into account and thus, on the expense of losing the information about specific spectral types of PE, measures
total PE.
[8] Building on the fact that PE absorption and fluorescence have an impact on the underwater light field, we propose a statistical method that allows for the prediction of
relative PE concentrations from underwater upwelling radiance (Lu) measurements. Several studies have shown that
PE-containing phytoplankton influence the submarine light
field due to absorption, scattering and fluorescence processes [Hoge and Swift, 1990; Kirk, 1986; Morel, 1997].
Taking advantage of this fact, we apply Empirical Orthogonal Function (EOF) analysis to decompose a matrix of corresponding Lu spectra, as this method has been shown to be
particularly useful for assessing variance structure in spectral measurements such as remote sensing reflectance
[Craig et al., 2012; Lubac and Loisel, 2007; Mueller,
1976; Toole and Siegel, 2001]. So far, the idea to retrieve
information about pigments from radiometric data has been
maximum regardless of the presence or absence of PUB,
indicating that only PEB chromophores fluoresce, although
the presence of PUB can influence the emission wavelength
of PEB. PEB fluorescence peaks at 580 nm, but can be
shifted to shorter wavelengths for PUB-containing PEs
[Chekalyuk and Hafez, 2008; Falkowski and Raven, 1997;
Hoge et al., 1998; Lantoine and Neveux, 1997; Ong and
Glazer, 1991].
1.1. Phycoerythrins in the Marine Environment
[5] Our understanding of the functioning of marine ecosystems starts with the phytoplankton because of their fundamental role in the marine food web and carbon cycle.
PE-containing species of the phytoplankton are a globally
important group of photosynthetic organisms and are distributed ubiquitously throughout oceanic regions, ranging
from polar through temperate to tropical waters in coastal
and open ocean regions [Hoef-Emden, 2008; Partensky et
al., 1999; Scanlan and West, 2002; Wood et al., 1998].
Within the picophytoplankton (cells < 2 mm), the PE-rich
cyanobacteria are a widely recognized and studied phytoplankton group in the marine environment [Carr and
Mann, 1994; Everroad and Wood, 2012; Partensky et al.,
1999; Scanlan and West, 2002]. They have been shown to
be important primary producers ; species of the genus Synechococcus, for example, have been estimated to account for
64% of total photosynthesis in the North Pacific [Iturriaga
and Mitchell, 1986]. Other picocyanobacterial strains such
as a PE-containing Cyanobium, a Cyanobium-like lineage
and species of the genus Synechocystis can occur alongside
Synechococcus in open ocean areas [Everroad and Wood,
2006; Waterbury and Rippka, 1989]. Another important
PE-containing genus is Trichodesmium, a group of filamentous cyanobacteria that forms extensive colonies in surface
waters of oligotrophic, tropical and subtropical oceans
[Capone et al., 1997; Staal et al., 2007; Subramaniam et
al., 1999]. The two eukaryotic PE-containing groups (cryptophytes and red algae) are more prevalent in coastal,
brackish, and freshwater environments; red algae occur
mainly as macroalgae [Clay et al., 1999; Cole and Sheath,
1990; Gabrielson et al., 1989; Gillot, 1989; Hoef-Emden,
1.2. Challenges of Phycoerythrin Measurement
[6] PEs seem to be ideal marker pigments for PE-containing cyanobacteria. However, contrary to all other algal
pigments which are soluble in organic solvents, measurements of phycobiliproteins are still scarce. A number of
methods have been proposed to measure PEs in water samples [e.g., Algarra et al., 1988; Downes and Hall, 1998;
Kim et al., 2011; Lantoine and Neveux, 1997; Ong et al.,
1984] or in vivo [Beutler et al., 2002; Chekalyuk and
Hafez, 2008; Chekalyuk et al., 2012; Cowles et al., 1993;
Hoge et al., 1998], but none are routinely used in oceanography, thus leaving PEs often undetected in pigment analysis. The most commonly used method, which detects PEs
through its orange fluorescence, is flow cytometry. Nevertheless, these methods need discrete water samples and
involve a certain amount of, often complex and time consuming, sample preparation. Very detailed information
about spatial and spectral variability and distribution of
PEs without the need of discrete water samples can be
concentration), shock-frozen in liquid nitrogen and stored at
80 C. A total of 131 water samples were taken (26, 62, and
43 on C1, C2, and C3, respectively) and 59 radiance profiles
were measured (13, 22, and 24 on C1, C2, and C3, respectively) at the same time as the water samples. At each station,
we took at least two water samples at the surface and the chl
a maximum; if possible a third sample was taken at 100 m.
2.2. Phycoerythrin Measurements
[11] PE measurements were based on the in vivo method
by Wyman [1992] and the spectrofluorometric assay by
Lantoine and Neveux [1997] and Neveux et al. [2006]. In
detail, the polycarbonate filters were placed into 3 ml of
50/50 mixture of glycerol and phosphate buffer (0.1 mol
L1 NaH2PO4 (pH ¼ 6.5)) and the cells on the filters were
resuspended by vigorous shaking on a lab bench vortex
mixer. The choice of filters and buffer was made following
extensive tests with different methods and personal communication with J. Neveux. Samples were kept on ice and
in the dark for 1h and vortexed again before fluorescence
was measured with a Fluorolog FL3–22 spectrofluorometer
(Horiba). We performed an excitation scan from 450 to 560
nm (emission: 575 nm). Excitation slits were set to 5 nm.
As we were not able to purify PEs for calibration purposes,
our results remain relative values and could not be converted into absolute concentrations. The relative PE concentration per L seawater was calculated from the
integrated area below the blank-subtracted excitation spectra between 450 and 560 nm. Spectra were normalized to
the Raman scatter [e.g., Coble et al., 1993; Sepp€al€a et al.,
2005] and the fluorometer was equipped with a reference
detector to monitor and compensate for variations in the xenon lamp output; thus even though the concentrations are
relative, they are comparable between samples and cruises.
Figure 1. Map of the Atlantic Ocean with tracks of the
three cruises. Points show all stations where concurrent
measurements of PE and Lu were conducted.
mainly applied to the main photosynthetic pigment chl a
and has been successfully extended to retrieve global chl a
concentrations from satellite data [e.g., Gordon et al.,
1980; McClain, 2009].
[9] The method presented here is used to predict PE profiles and values of integrated PE based on upwelling radiance spectra. The aim was to develop a method that can be
easily applied to radiometric measurements and thus is applicable to measurements at all depths as long as light is detectable. With only a few discrete water samples and PE
measurements, information about whole profiles of total PE
can be extracted.
2.3. Pigment Analysis
[12] The composition of pigments which are soluble in
organic solvents was analyzed by High Performance Liquid
Chromatography (HPLC) following a method described by
Hoffmann et al. [2006] adjusted to our instruments as
detailed by Taylor et al. [2011].
2.1. Sample Collection
[10] Samples were collected during three cruises: the
ANT-XXIV/4 expedition of the RV Polarstern in April/May
2008 along a South-to-North transect through the Atlantic
Ocean from Punta Arenas (Chile) to Bremerhaven (Germany), the ANT-XXV/1 expedition of the RV Polarstern in
November 2008 along a North-to-South transect through the
eastern Atlantic Ocean from Bremerhaven (Germany) to
Cape Town (South Africa) and the ANT-XXVI/4 expedition
of the RV Polarstern in April/May 2010 along a South-toNorth transect through the Atlantic Ocean from Punta Arenas (Chile) to Bremerhaven (Germany). For convenience the
cruises will be called C1 (ANTXXIV/4), C2 (ANTXXV/1)
and C3 (ANTXXVI/4) (Figure 1). Sampling stations generally coincided once a day at noon local time and involved
CTD casts with water samplers as well as above- and belowwater radiance and irradiance measurements. Water samples
were filtered on 0.4 mm polycarbonate filters for PE analysis
and on GF/F filters for analysis of other pigments, shock-frozen in liquid nitrogen and stored at 80 C. Samples for flow
cytometry were preserved with 0.1% glutaraldehyde (final
2.4. Flow Cytometry
[13] Phytoplankton cells were enumerated from preserved and frozen, unstained samples by using their specific
chl a and PE autofluorescence as described by Marie et al.
[2005]. Flow cytometry was performed on a FACScalibur
with an excitation beam of 488 nm, two light scatter detectors at 180 (forward scatter) and at 90 (side scatter) and
several photomultipliers detecting at 530 6 15 nm (green
fluorescence), 585 6 21 nm (orange fluorescence) and
>670 nm (red fluorescence). Phytoplankton groups were
separated according to their red and orange fluorescence
and scattering characteristics. Yellow-green FluoresbriteV
Microspheres with a diameter of 1 mm (Polysciences) were
used as an internal standard. The data were analyzed with
the instrument software ‘‘CellQuest.’’
2.5. Radiometric Measurements
[14] Underwater optical light fields were measured with
hyperspectral radiometers (RAMSES, TriOS GmbH, Germany) measuring radiance profiles. The instrument covers
a wavelength range of 320–950 nm with an optical
resolution of 3.3 nm and a spectral accuracy of 0.3 nm. All
measurements were collected with sensor-specific automatically adjusted integration times (between 4 ms and 8
s). Radiometric profiles measuring upwelling radiance (Lu)
[W m2 nm1 sr1] were collected at the same time as the
CTD profiles on a second winch down to a maximum depth
of 190 m. Irradiance at the surface (Edþ) [W m2] was
measured as a reference with a second sensor placed
above-water and was utilized to normalize the measured
under-water data to a maximum value of Edþ as described
by Smith and Baker [1984]. The radiance sensor had a field
of view of 7 , while the irradiance sensor had a cosine collector fixed in front of the instrument. The in-water sensor
was equipped with an inclination and a pressure sensor. To
avoid ship shadow, the ship was oriented such that the sun
was illuminating the side where the measurements were
taking place. The pitch and roll data measured by the ship
did not exceed values larger than 5 . For the in-water data,
the inclination in either dimension was smaller than 14
[Matsuoka et al., 2007].
Cjl ¼ X:J X:l :
e eT ;
k¼1 k k k
[18] with EOFs equaling the Eigenvectors E, and K
being a diagonal matrix containing the eigenvalues, by
which the explained variance of each EOF can be calculated. E is an NN matrix containing loadings for each
sample by mode. EOF expansion coefficients Z (i.e., ‘‘principal components’’) are then calculated as the projection of
the data X onto E:
Z¼XE Zik ¼
Xij ejk ;
where Z is an MN matrix carrying the loadings for each
radiance wavelength (nm) by mode.
2.6.2. Generalized Linear Model
[19] The number n of significant EOF modes was determined according to North’s Rule-of-Thumb [North et al.,
1982]. The significant EOF modes from E were then used
as covariates in the prediction of measured PE concentrations using a multiple Generalized Linear Model (GLM):
2.6. Statistical Methods
[15] We first processed the data of all cruises together,
followed by the analysis of each cruise separately. As
cruise C3 yielded different results than the two other
cruises (possible reasons will be discussed below), we also
analyzed a data set consisting of the data of cruises
C1 þ C2 only.
2.6.1. Empirical Orthogonal Function Analysis
[16] All analyses were conducted using the statistical
computing software ‘‘R’’ [R Development Core Team,
2011]. Spectral data were subjected to an Empirical Orthogonal Function (EOF) analysis (sometimes referred to
as Principal Component Analysis) in order to reduce the
high dimensionality of the data and derive the dominant
signals (‘‘modes’’) that best describe variance within the
data set. We averaged the Lu spectra within one cast that
were measured within 61 m depth of the respective discrete PE measurement. The sampling rate of the radiometer
depended on integration time, which is affected by light
availability. As a result, the number of sampled spectra
decreased with depth and the number of spectra used for
the average ranged from 3 to 10 samples. The averaged Lu
spectra were used to create a data matrix X consisting of M
rows of Lu radiances, from 350–800 nm in 1 nm increments, by N sample columns. The resulting data matrix
consisted of M¼451 rows (nm), while the number of N columns (samples) varied between models. Prior to analysis,
the Lu spectra were standardized by first subtracting the
mean (centering) and then dividing by the standard deviation (scaling), in order that each spectral sample (columns)
had a mean of zero and standard deviation of one (i.e.,
dimensionless). This standardization step allowed us to
focus on spectral shape rather than magnitude. All results
shown in this paper are from standardized spectra. Subsequently, a covariance matrix was calculated:
loge ðEðPEÞÞ ¼ þ 1 e1 þ 2 e2 þ þ n en ;
where e1;2;:::n are the leading n significant EOF modes from
E, is the intercept, and 1;2;:::n are the regression coefficients. The GLM assumed a Gaussian error distribution and
used a log-normal link function for the expectation of PE,
E(PE). The model assumes that error dispersion is constant
and independent of PE, which is consistent with the error
of the method used to determine its concentration [Lantoine
and Neveux, 1997]. The log-link function GLM provided a
better fit over that of a simple linear model (in terms of
sum of squared differences) and also had the advantage of
preventing the prediction of negative values.
[20] A stepwise routine was used to search for smaller
models based on fewer terms, through minimization of the
Akaike information criterion (AIC). Once the best model
was determined, the significance of included terms was
defined by the change in AIC (AIC) following each
term’s removal. This is an appropriate test for the comparison of nested models because it includes a penalty for the
number of parameters in the model.
2.6.3. PE Prediction
[21] The model was applied to spectra from depths that
did not have corresponding PE measurements in order to
create profiles of estimated PE. This new set of radiance
spectral data Y was first projected onto the EOF domain
using the EOF coefficients Z:
Z1 Y
¼ F;
where F is an NN matrix giving the loadings for each
sample, as with E. Predicted PE is then calculated using the
best fitted GLM and the new EOF loadings F:
log e E PE pred ¼ a þ b 1 f1 þ 2 f2 þ ::: þ b n fn;
where PEpred equals the predicted relative PE concentrations and f1;2;:::n are the EOF loadings F that correspond to
the significant terms of the multiple regression model
[17] The covariance matrix C was then subjected to an
Eigen decomposition :
Table 1. Percent of Total Variance Explained by the Significant EOFs Derived from Radiance Spectra from Various Combinations of
Cruise Data
values in the wavelength region of 400–500 nm with no
other conspicuous effects in the region above 550 nm. EOF
coefficient 2 also showed main deviations (positive and
negative) in the region of 400–500 nm, while EOF mode 3
showed a negative peak between 450 and 500 nm and positive values in the wavelength region of 500–600 nm.
(equation (4)). Confidence intervals were calculated for
each PEpred based on the standard error of the GLM
2.6.4. Integrated PE
[22] Using the above procedure, PE concentration was
predicted from radiance spectra throughout each station’s
sampled profile. A spline function was fitted to the predicted values by depth and used for the calculation of integrated values for each profile (in relative PE concentration
m2). In order to include a majority of stations in the comparison, integrated values were calculated down to a depth
of 95 m, since most station profiles were sampled to at least
this depth.
3.2. Generalized Linear Models
[24] Figure 3 shows the predicted versus the measured
PE values for each analysis. The models show better fits for
data sets of individual cruises than when all cruises were
grouped together. In particular, adding C3 to the model
using only data from C1 and C2 lowered R2 substantially,
although the correlation was still highly significant at
p<0.0001. All cruises showed a large range of PE values;
however C3 had very low-PE values compared to the two
other cruises. Reasons for these variations will be discussed
[25] In order to estimate the importance of each EOF in
the best model, we used the AIC as a measure of each
smaller model (without the respective EOF) relative to the
best model. The bigger the AIC, the greater the importance of the removed EOF (Table 2). In all models EOF 3
was the most significant predictor of PE concentrations.
3.1. EOF Analysis of Lu Spectra
[23] The decomposition of the standardized spectra by
the EOF analysis returned between 5 and 9 significant
EOFs explaining most of the variance of the data matrix
(Table 1). Figure 2 illustrates the 3 dominant EOF coefficients for each of the cruises (for visual comparison, we
have removed the eigenvalue units in order to present all
coefficients on a similar scale). The shapes of these 3 EOF
coefficients were similar in all analyses and explained most
of the variance (> 98%). EOF coefficients four to nine generally added less than 1% and thus are not discussed in
detail in this paper. The gray area in the plots indicates the
area where the main signal of PEs is to be expected due to
their absorption and fluorescence features in the spectral
range of 450–600 nm. EOF coefficient 1 shows positive
3.3. PE Prediction
[26] Using the regression models we predicted PE profiles and integrated PE values from the radiance measurements. Figure 4 shows the integrated PE (relative
concentration m2) for all stations of C1, C2, and C3 of
which we had Lu profiles down to 95 m. The profiles were
calculated with the models of the respective cruises and the
Figure 2. Leading three EOF coefficients (scaled) as calculated from the standardized Lu spectra for
the single cruise models. Each cruise is designated by a color. The gray area indicates the wavelength
range where the main influence of PEs is to be expected (450–600 nm).
Figure 3. Observed versus predicted PE values for models based on various combinations of cruise
data. All models are highly significant at p < 0.0001. The 1:1 line is shown in gray for reference.
flectance data to derive information about chl a, particulate
backscattering, nutrients or other biochemical or optical parameters. We have applied this method to upwelling underwater radiance in the hope to be able to use underwater
light measurements for the estimation of pigment profiles
which cannot be measured easily in the same resolution
with common laboratory methods.
[28] PEs and PE-containing cyanobacteria have been
studied extensively. Especially the genus Synechococcus, a
group of small unicellular cyanobacteria with a wide geographical distribution and considerable impact on the
global carbon cycle, has been the interest of many research
projects since its discovery in the late 1970s [Waterbury et
al., 1979]. Most data on the depth distribution are still
based on discrete water samples and laborious measurement techniques involving a great amount of sample preparation and instrumentation, thus often relying on a small
number of samples only. More recently, some specific PE
sensors have been developed, which estimate PE fluorescence using one or several excitation and emission wavelengths (see section 1.3.).
gray areas around the profiles show the 95% confidence
intervals of the predictions. The cumulative error of the
multiple regression EOF terms increases the confidence
interval for the higher range of predicted PE concentrations. Integrated PE values are highest in the Bay of Biscay,
off the West coast of Portugal, in the Mauritanian upwelling and off the coast of Namibia. Seven example PE profiles from stations of different biogeographical regions of
the Atlantic Ocean are depicted in the figure. The profiles
show that PE distribution by depth can be very different,
ranging from the main concentration residing in surface
waters to deep maxima at >50 m. See Figure A1 in the Appendix for all predicted profiles.
[27] EOF analysis is a statistical method that has proved
useful in extracting information about water constituents
from spectral data since the 1970s [Craig et al., 2012;
Lubac and Loisel, 2007; Mueller, 1976; Toole and Siegel,
2001]. Most authors used this method on ocean color re-
Table 2. Change in the Akaike Information Criterion (AIC) Following Individual Term Removals from the Best Multiple Regression
Figure 4. Map of the Atlantic Ocean with all stations of which a value of integrated PE could be calculated from the Lu profiles. The integrated values were calculated with the single cruise models. Each
cruise is designated by a color. Arranged around the map are seven examples of predicted PE profiles,
including the values of measured PE (open circles) and the depth of the chl a maximum (hatched band).
The gray areas around the profiles show the 95% confidence intervals of the predictions. Confidence
intervals are narrow for low concentrations and cannot always be visualized.
al. [2012] also used a type of spectral scaling in order to
focus their analysis on shape rather than magnitude. They
suggest that the use of normalized spectra emphasizes the
color of the water carrying most of the information about
chl a and phytoplankton absorption.
[30] The same analysis was conducted with flow cytometry data (i.e., cell counts of orange fluorescing cells and
total orange fluorescence as measured by the flow cytometer) instead of PE data, but the relationships were not as
strong as with the PE data. We suggest two possible reasons for the better performance with PE data:
[31] (1) Sample volume: The flow cytometry data are
derived from measurements of very small sample volumes
[29] The presented results show a clear relationship
between PE concentrations and the spectral shape of Lu,
especially for the data sets of the single cruises, thus corroborating the fact that PE influences the surrounding
underwater light field, by absorption and scattering processes, and also by fluorescence emission. In order to focus
on changes in spectral shape of the Lu spectra rather than
magnitude, the spectra were standardized to minimize the
component of variability due to spectral amplitude thus
accounting for the inclusion of samples from different
depths with different light availability. We tried the same
analysis with nonstandardized data and got relatively poor
model fits, compared to the standardized data set. Craig et
here. The spectral variability of PE in terms of the two
chromophores PUB and PEB which absorb at 495 nm
and 545 nm respectively will extend the influence of PE
toward shorter wavelengths when high PUB-species are
present, whereas no-PUB species would not influence the
light spectrum below 500 nm.
[35] As indicated by AIC, the predictor that explained
the variance in the data best was EOF coefficient 3, which
explains the variance in the spectral region of 450–600 nm
where we expect the main influence of PEs on the underwater light field, corroborating the validity of our model.
The fact that we cannot explain the influence of PE on the
Lu spectra by EOF mode 3 only, but also need to include
other modes for a better correlation, is probably a function
of the spectral variability of PEs and of the other pigments
present in the cells. High-PUB PEs will have an impact on
the underwater light field in the spectral region below 500
nm and the variance in this region is also described by EOF
coefficients 1 and 2. The variability in the PUB/PEB ratio
as well as the composition of accessory pigments could
also be the reason that the number of EOF coefficients
included in the models varies depending on the cruises
which were included. It emphasizes the fact that the effect
of PE on the underwater light field is apparent within a
broad wavelength range and not only at the peaks of
absorption or fluorescence.
[36] Using the GLMs based on EOFs, we could predict
PE profiles based on radiance spectra for each station. Integrated values and seven example profiles are shown in Figure 4, with measured PE and the depth of the chl a
maximum provided as reference. In open ocean regions,
the PE signal can be linked to Synechococcus distribution
patterns as published in the literature: Synechococcus is
known to be more abundant in areas which are seasonally
or permanently enriched with nutrients by strong upwelling
or coastal inputs [Olson et al., 1990; Partensky et al.,
1996, 1999; Zubkov et al., 1998]. Highest integrated values
are found in association with the Mauretanian upwelling
and off the coasts of Western Europe and South West
Africa. At stations near the coast the detection of the
marker pigment alloxanthin (pigment data not shown here)
suggests the presence of cryptophytes which are probably
responsible for the PE signal in these environments less
typical for cyanobacteria.
[37] The predicted PE profiles generally reproduce the
measured PE values well and the bands of 95% confidence
intervals show the predictions to be robust. The GLM that
we apply assumes a constant error across the range of measured PE concentrations. Translated to error as a percent of
the concentration, the model is less sensitive in identifying
variation at the lower range of PE concentrations and is
more appropriate for identifying profile features at higher
concentrations. The profiles show varying vertical distributions, which is not surprising as vertical profiles are influenced by numerous physico-chemical factors such as
stability of the water column, light, temperature or nutrients,
hence making it difficult to predict which profiles should be
found in which region at a certain point in time. We often
find that the PE-maximum is closely linked to the chl a maximum (see Figure A1 of the Appendix), a relationship that
has been reported before [Partensky et al., 1996]. However,
exceptions exist where the PE maximum is located slightly
(<0.0005 L) whereas for the PE measurements the sample
volume was generally between 1 and 4 L. This large difference in sample volume could account for the fact that the
Lu estimates provide a better fit to the PE data than to the
cell counts.
[32] (2) Excitation wavelengths : From our own analysis
of PE excitation spectra and other publications we know
that the excitation of 488 nm used in the flow cytometric
analysis is not optimal for all samples [e.g., Olson et al.,
1988]. Samples dominated by low-PUB species have their
excitation maximum at 550 nm, for no-PUB species this
peak can even be shifted to 560 nm, whereas high-PUB
species generally have two maxima at 495 nm and at
550 nm [Hoge et al., 1998; Lantoine and Neveux, 1999;
Olson et al., 1988]. It has been demonstrated before [Olson
et al., 1988] that high-PUB cells show higher fluorescence
when excited by a 488 nm laser than low-PUB cells. The
low-fluorescence response of low- or no-PUB cells to the
488 nm laser of the flow cytometer might explain why
some cells could be missed in this method, whereas the relative PE concentrations measured with the fluorometric
technique are based on excitation scans ranging from 450
to 560 nm.
[33] Another fact that will add to the differences between
flow cytometry and PE measurements, but also between PE
and Lu measurements, is the species composition. Noteworthy species such as Trichodesmium, which form filaments
and colonies, will probably not be detected in flow cytometry samples due to the size of the filaments and the small
sample volume. There is a greater chance to encounter Trichodesmium filaments in a larger water sample, as was
taken for the PE measurements; however, even the large
Niskin bottles might miss the naturally buoyant colonies.
Although we never detected any visible filaments on the
sample filters or colonies in the surface waters, we cannot
be sure that they were not present. The radiometer, however, would detect the spectral impact of a Trichodesmium
[34] For all data sets, EOF modes 1–3 explained between
98 and 99% of the total variance of standardized spectra.
EOF mode 1 showed one main feature in the region of
400–500 nm probably due to chl a and other accessory pigments, while EOF mode 2 showed positive and negative
spectral features in the same wavelength region, possibly
due to a mixed effect of absorption and backscattering on
the underwater light field. EOF mode 3 explained variation
in the spectral region of 450–600 nm. As described in detail
in the introduction, absorption spectra of PEs cover the
spectral region of 475–580 nm (with absorption maxima
at 495 nm and 545 nm for the PUB and PEB chromophores, respectively), and fluoresce at 580 nm, depending
on various factors such as taxonomy, pigment composition,
light availability and nutrients [Falkowski and Raven,
1997; Glazer and Stryer, 1984; Glazer et al., 1982; Hoge
et al., 1998; Lantoine and Neveux, 1999]. As a result of
these spectral properties, we would expect that the presence
or absence of PE would coincide with deviations in the Lu
spectra in the spectral range of approximately 450–600 nm.
This spectral region is generally known as the ‘‘green gap’’
as all other known photosynthetic pigments do not absorb
or absorb only weakly in this region. Thus, we can assume
that the interference from other pigments would be smallest
Figure 5. Concentrations of total pigments [mg/m3] measured by HPLC, Synechococcus [cells/ml]
measured by flow cytometry and PE [relative concentration/L] measured by a spectrofluorometric
method by latitude. Each cruise is designated by color. Values are given for surface (left plots) and at the
depth of the chl a maximum (right plots).
ments. However, the range of relative PE concentrations of
C3 was restricted to a much smaller range than in the two
other cruises (0.001–0.25 relative units/L for C3, as
opposed to 0.001–0.89 relative units/L and 0.001–0.72 relative units/L for C1 and C2, respectively). The cruises C1
and C3 were conducted on nearly the same cruise track and
in the same season, with 2 years between them. To investigate possible reasons for these low-PE values, we looked at
the pigment analysis and flow cytometry data of all three
cruises (Figure 5). The amount of total pigment gives a
general idea of the phytoplankton biomass present at each
station and the flow cytometry data shown here gives the
number of cells per milliliters of Synechococcus species,
the most abundant phytoplankton with PE-containing phycobilisomes in open ocean waters. For comparison, the
third plot in Figure 5 shows the measured PE concentration
(relative units/L). Both C1 and C2 encountered major phytoplankton blooms with a high percentage of PE-containing
phytoplankton in the Mauritanian upwelling, whereas C3
does not show any elevated pigment or PE levels at that latitude. C3 only encountered higher pigment concentration
north of Spain and the percentage of PE-containing phytoplankton was very low in that area. These differences in
pigment composition might explain why we obtain different models from the different cruises. They emphasize how
important the input data are for the model output and how
below the chl a maximum (for example profiles A and B in
Figure 4). In some cases (as in profile C in Figure 4) there
are two PE maxima and only one of them matches the chl a
maximum. Profile F is an example where the PE maximum
would have been missed entirely if samples had been taken
only in the chl a maximum.
[38] The fact that we use relative PE concentrations limits
our ability to compare with other PE quantifications. However, the aim of this study is not to show a distribution of
absolute values of PE, but to demonstrate a method, which
allows us to deduce PE values from radiometric measurements. The PE determination chosen as input parameter for
the model dictates the output. Thus, should we or other laboratories have the possibility to quantify PE in absolute values, the output of the model would also be in absolute
concentrations. However, purifying phycobiliproteins is not
a straight forward procedure and, more importantly for the
quantification of PEs, purified proteins often have different
spectral properties than the proteins in vivo, as their spectral
properties are strongly influenced by interactions in the complex assemblages of the phycobilisomes [Glazer, 1988].
[39] In our GLMs, we found the strongest relationship
for the single cruise models and for the model with pooled
data of C1þC2. However, the third cruise (C3), did not
improve the correlation although within the cruise itself
there was a strong relationship between Lu and PE measure9
Figure A1. Predicted PE profiles for all stations, calculated with the single cruise models, including
the values of measured PE (open circles) and the depth of the chl a maximum (hatched band). The gray
areas around the profiles show the 95% confidence intervals of the predictions. Confidence intervals are
narrow for low concentrations and cannot always be visualized.
depth profiles than could be acquired through measurements of discrete water samples. We have shown that the
influence that PE-containing phytoplankton has on the
underwater light field can be exploited to calculate PE concentrations from Lu measurements. For any cruise or site
where Lu and concurrent PE measurements are available,
we can derive a model for PE estimation. There is still a
need for discrete water samples for PE measurements as
input and validation data for the model, but with our model,
the resolution of the output data is much higher than it
would ever be with discrete water samples and laboratory
crucial it is to have a broad set of data encompassing a
wide range of pigment concentrations. Large interannual
changes in (pico-) phytoplankton composition and dynamics have been reported previously [Dandonneau et al.,
2004; Head and Pepin, 2010a, 2010b; Partensky et al.,
1996] and should be taken into account.
Conclusions and Further Research
[40] We have developed a statistical approach to derive
PE concentrations from underwater radiance measurements, which enables us to obtain a broader PE data set of
Figure A1.
AWI as well as Oliver Zielinski and his group at the ICBM in Oldenburg
for the use of their equipment. We are grateful to Erika Allhusen, Mirko
Lunau, Anja Bernhardt and Sonja Wiegmann for help with the laboratory
work and the crew, principal investigators and other scientists on board the
RV Polarstern for support on board and fruitful discussions afterward.
[41] From our results, we conclude that, for the best
approximation to reality, each cruise would need its own
model to accommodate seasonal and interannual changes.
Another possibility, which could be tested with more data,
would be a separation of the data sets according to biogeographical regions or provinces instead of cruises and compare the model outcomes. We would also welcome a
comparison of our model output with the output data set of
specific PE sensors that are commercially available.
Algarra, P., M. Estrada, and F. X. Niell (1988), Phycobiliprotein distribution across the Western Mediterranean divergence, Deep Sea Res. Part
A, 35(8), 1425–1430.
Beutler, M., K. H. Wiltshire, B. Meyer, C. Moldaenke, C. Luring, M.
Meyerhofer, U. P. Hansen, and H. Dau (2002), A fluorometric method
for the differentiation of algal populations in vivo and in situ, Photosynth. Res., 72(1), 39–53.
Beutler, M., K. H. Wiltshire, C. Reineke, and U. P. Hansen (2004), Algorithms and practical fluorescence models of the photosynthetic apparatus
of red cyanobacteria and Cryptophyta designed for the fluorescence
detection of red cyanobacteria and cryptophytes, Aquat. Microbial Ecol.,
35(2), 115–129.
Capone, D. G., J. P. Zehr, H. W. Paerl, B. Bergman, and E. J. Carpenter (1997), Trichodesmium, a globally significant marine cyanobacterium,
doi :10.1126/
Carr, N. G., and N. H. Mann (1994), The oceanic cyanobacterial picoplankton, in The Molecular Biology of Cyanobacteria, edited by D. A. Bryant,
pp. 27–48, Kluwer Acad., Dordrecht.
Appendix A
[42] Figure A1 shows calculated profiles of phycoerythrin
(PE) distribution for all stations of the three cruises. Values
of measured PE and the depth of the chlorophyll a (chl a)
maximum are included for reference. We often find that the
PE-maximum is closely linked to the chl a maximum.
[43] Acknowledgments. We would like to thank four anonymous
reviewers for very helpful comments. We thank AWI, Helmholtz Impulse
Fond (HGF Young Investigators Group Phytooptics) and OCEANET for
funding. M.H.T. was supported by the German Research Foundation, project BiPhyCoSi (ID: LO-1143/6). We also thank the remote sensing group
at the Helmholtz-Zentrum Geesthacht, Allan Cembella and his group at the
Chekalyuk, A. M., and M. Hafez (2008), Advanced laser fluorometry of
natural aquatic environments, Limnol. Oceanogr., 6, 591–609.
Chekalyuk, A. M., M. R. Landry, R. Goericke, A. G. Taylor, and M. A.
Hafez (2012), Laser fluorescence analysis of phytoplankton across a
frontal zone in the California current ecosystem, J. Plankton Res., 34(9),
761–777, doi:10.1093/plankt/fbs034.
Clay, B. L., P. Kugrens, and R. E. Lee (1999), A revised classification of
Cryptophyta, Bot. J. Linnean Soc., 131(2), 131–151, doi:10.1111/
Coble, P. G., C. A. Schultz, and K. Mopper (1993), Fluorescence contouring analysis of DOC intercalibration experiment samples––A comparison of techniques, Mar. Chem., 41(1–3), 173–178, doi:10.1016/0304–
Cole, K. M., and R. G. Sheath (Eds.) (1990), Biology of the Red Algae,
Cambridge Univ. Press, Cambridge, U.K.
Cowles, T. J., R. A. Desiderio, and S. Neuer (1993), Insitu characterization
of phytoplankton from vertical profiles of fluorescence emission-spectra,
Mar. Biol., 115(2), 217–222, doi:10.1007/Bf00346338.
Craig, S. E., C. T. Jones, W. K. W. Li, G. Lazin, E. Horne, C. Caverhill,
and J. J. Cullen (2012), Deriving optical metrics of coastal phytoplankton biomass from ocean colour, Remote Sens. Environ., 119, 72–83,
Dandonneau, Y., P. Y. Deschamps, J. M. Nicolas, H. Loisel, J. Blanchot,
Y. Montel, F. Thieuleux, and G. Becu (2004), Seasonal and interannual variability of ocean color and composition of phytoplankton
communities in the North Atlantic, equatorial Pacific and South Pacific, Deep Sea Res. Part II, 51(1–3), 303–318, doi :10.1016/
Desiderio, R. A., C. Moore, C. Lantz, and T. J. Cowles (1997), Multiple excitation fluorometer for in situ oceanographic applications, Appl. Optics,
36(6), 1289–1296, doi:10.1364/Ao.36.001289.
Downes, M. T., and J. A. Hall (1998), A sensitive fluorometric technique
for the measurement of phycobilin pigments and its application to the
study of marine and freshwater picophytoplankton in oligotrophic environments, J. Appl. Phycol., 10(4), 357–363.
Dubelaar, G. B. J., P. L. Gerritzen, A. E. R. Beeker, R. R. Jonker, and K.
Tangen (1999), Design and first results of CytoBuoy: A wireless flow
cytometer for in situ analysis of marine and fresh waters, Cytometry,
37(4), 247–254.
Everroad, R. C., and A. M. Wood (2006), Comparative molecular evolution
of newly discovered picocyanobacterial strains reveals a phylogenetically informative variable region of beta-phycoerythrin, J. Phycol.,
42(6), 1300–1311, doi:10.1111/j.1529–8817.2006.00282.x.
Everroad, R. C., and A. M. Wood (2012), Phycoerythrin evolution and
diversification of spectral phenotype in marine Synechococcus and
related picocyanobacteria, Mol. Phylogenet. Evol., 64(3), 381–392,
Falkowski, P. G., and J. A. Raven (1997), Aquatic Photosynthesis, Blackwell Sci, Malden, Massachusetts.
French, C. S., and V. K. Young (1952), The fluorescence spectra of red
algae and the transfer of energy from phycoerythrin to phycocyanin to
chlorophyll, J. Gen. Physiol., 35(6), 873–890.
Gabrielson, P. W., D. J. Garbary, M. R. Sommerfeld, R. A. Townsend, and
P. L. Tyler (1989), Phylum Rhodophyta, in Handbook of Protoctista:
The Structure, Cultivation, Habitats and Life Histories of the Eukaryotic
Microorganisms and Their Descendants Exclusive of Animals, Plants
and Fungi, edited by L. Margulis, J. O. Corliss, M. Melkonian and D. J.
Chapman, pp. 102–118, Jones and Barlett, Boston.
Gantt, E., and S. F. Conti (1966), Phycobiliprotein Localization in Algae,
Brookhaven Symp. Biol., 19, 393–405.
Gillot, M. (1989), Phylum Cryptophytoa (Cryptomonads), in Handbook of
Protoctista: The Structure, Cultivation, Habitats and Life Histories of
the Eukaryotic Microorganisms and Their Descendants Exclusive of Animals, Plants and Fungi, edited by L. Margulis, J. O. Corliss, M. Melkonian, and D. J. Chapman, pp. 139–151, Jones and Barlett, Boston.
Glazer, A. N. (1988), Phycobilisomes, Methods Enzymol., 167, 304–312.
Glazer, A. N. (1999), Cyanobacterial photosynthetic apparatus: An overview, in Marine Cyanobacteria, special issue 19, edited by L. Charpy
and A. W. D. Larkum, pp. 419–430, Bull. de l’Inst. oceanographique,
Glazer, A. N., and L. Stryer (1984), Phycofluor Probes, Trends Biochem.
Sci., 9(10), 423–427, doi:10.1016/0968-0004(84)90146-4.
Glazer, A. N., J. A. West, and C. Chan (1982), Phycoerythrins as chemotaxonomic markers in red algae––A survey, Biochem. Syst. Ecol., 10(3),
Gordon, H. R., D. K. Clark, J. L. Mueller, and W. A. Hovis (1980), Phytoplankton pigments from the Nimbus-7 coastal zone color scanner––
Comparisons with surface measurements, Science, 210(4465), 63–66,
Haverkamp, T. H. A., D. Schouten, M. Doeleman, U. Wollenzien, J. Huisman, and L. J. Stal (2009), Colorful microdiversity of Synechococcus
strains (picocyanobacteria) isolated from the Baltic Sea, ISME J., 3(4),
397–408, doi:10.1038/ismej.2008.118.
Head, E. J. H., and P. Pepin (2010a), Monitoring changes in phytoplankton
abundance and composition in the Northwest Atlantic: A comparison of
results obtained by continuous plankton recorder sampling and colour
satellite imagery, J. Plankton Res., 32(12), 1649–1660, doi:10.1093/
Head, E. J. H., and P. Pepin (2010b), Spatial and inter-decadal variability in
plankton abundance and composition in the Northwest Atlantic (1958–
2006), J. Plankton Res., 32(12), 1633–1648, doi:10.1093/plankt/fbq090.
Hill, D. R. A., and K. S. Rowan (1989), The biliproteins of the Cryptophyceae, Phycologia, 28(4), 455–463, doi:10.2216/i0031–8884-28-4-455.1.
Hoef-Emden, K. (2008), Molecular phylogeny of phycocyanin-containing
cryptophytes: Evolution of biliproteins and geographical distribution, J.
Phycol., 44(4), 985–993, doi:10.1111/j.1529–8817.2008.00530.x.
Hoffmann, L. J., I. Peeken, K. Lochte, P. Assmy, and M. Veldhuis (2006),
Different reactions of Southern Ocean phytoplankton size classes to iron
fertilization, Limnol. Oceanogr., 51(3), 1217–1229.
Hoge, F. E., and R. N. Swift (1990), Photosynthetic Accessory Pigments Evidence for the Influence of Phycoerythrin on the Submarine LightField, Remote Sens. Environ., 34(1), 19–35.
Hoge, F. E., C. W. Wright, T. M. Kana, R. N. Swift, and J. K. Yungel
(1998), Spatial variability of oceanic phycoerythrin spectral types
derived from airborne laser-induced fluorescence emissions, Appl.
Optics, 37(21), 4744–4749, doi:10.1364/Ao.37.004744.
Horiuchi, T., and F. Wolk (2008), Multispectral measurements of fluorescence and phytoplankton, Sea Technol., 49(9), 15–19.
Iturriaga, R., and B. G. Mitchell (1986), Chroococcoid cyanobacteria - a
significant component in the food web dynamics of the open ocean, Mar.
Ecol. Prog. Ser., 28(3), 291–297, doi:10.3354/meps028291.
Kim, J. J., Y. M. Jeon, J. H. Noh, and M. Y. Lee (2011), Isolation and characterization of a new phycoerythrin from the cyanobacterium Synechococcus sp ECS-18, J. Appl. Phycol., 23(1), 137–142, doi:10.1007/
Kirk, J. T. O. (1986), Optical properties of picoplankton suspensions, in
Photosynthetic Picoplankton, edited by T. Platt and W. K. W. Li, vol.
214, pp. 501–520, Can. Bull. of Fish. and Aquat. Sci., Ottawa.
Kronick, M. N. (1986), The use of phycobiliproteins as fluorescent labels in
immunoassay, J. Immunol. Methods, 92(1), 1–13.
Lantoine, F., and J. Neveux (1997), Spatial and seasonal variations in abundance and spectral characteristics of phycoerythrins in the tropical northeastern Atlantic Ocean, Deep Sea Res. Part I, 44(2), 223–246.
Lantoine, F., and J. Neveux (1999), Phycoerythrins in the sea: Abundance
and spectral diversity, in Marine Cyanobacteria, special issue 19, edited
by L. Charpy and A. W. D. Larkum, pp. 443–450, Bull. de l’Inst.
oceanographique, Monaco.
Leboulanger, C., U. Dorigo, S. Jacquet, B. Le Berre, G. Paolini, and J. F.
Humbert (2002), Application of a submersible spectrofluorometer for
rapid monitoring of freshwater cyanobacterial blooms: A case study,
Aquat. Microb. Ecol., 30(1), 83–89, doi:10.3354/Ame030083.
Lubac, B., and H. Loisel (2007), Variability and classification of remote
sensing reflectance spectra in the eastern English Channel and southern
North Sea, Remote Sens. Environ., 110, 45–48.
MacColl, R. (1998), Cyanobacterial phycobilisomes, J. Struct. Biol.,
124(2–3), 311–334, doi:10.1006/jsbi.1998.4062.
Marie, D., N. Simon, and D. Vaulot (2005), Phytoplankton cell counting by
flow cytometry, in Algal Culturing Techniques, edited by R. A. Andersen, pp. 253–268, Elsevier Acad. Press, Amsterdam.
Matsuoka, A., Y. Hout, K. Shimada, S. Saitoh, and M. Babin (2007), Biooptical characteristics of the western Arctic Ocean: Implications for
ocean color algorithms, Can. J. Remote Sens., 33(6), 503–518.
McClain, C. R. (2009), A decade of satellite ocean color observations,
Morel, A. (1997), Consequences of a Synechococcus bloom upon the optical properties of oceanic (case 1) waters, Limnol. Oceanogr., 42(8),
Mueller, J. L. (1976), Ocean color spectra measured off oregon coast––
Characteristic vectors, Appl. Optics, 15(2), 394–402.
Neveux, J., M. M. B. Tenorio, C. Dupouy, and T. A. Villareal (2006), Spectral diversity of phycoerythrins and diazotroph abundance in tropical
waters, Limnol. Oceanogr., 51(4), 1689–1698.
North, G. R., T. L. Bell, R. F. Cahalan, and F. J. Moeng (1982), Sampling
Errors in the Estimation of Empirical Orthogonal Functions, Mon.
Weather Rev., 110(7), 699–706.
Olson, R. J., S. W. Chisholm, E. R. Zettler, and E. V. Armbrust (1988),
Analysis of Synechococcus pigment types in the sea using single and
dual beam flow cytometry, Deep Sea Res. Part A, 35(3), 425–440,
Olson, R. J., S. W. Chisholm, E. R. Zettler, and E. V. Armbrust (1990), Pigments, size, and distribution of Synechococcus in the North Atlantic and
Pacific Oceans, Limnol. Oceanogr., 35(1), 45–58.
Olson, R. J., A. Shalapyonok, and H. M. Sosik (2003), An automated submersible flow cytometer for analyzing pico- and nanophytoplankton:
FlowCytobot, Deep Sea Res. Part I, 50(2), 301–315.
Ong, L. J., and A. N. Glazer (1991), Phycoerythrins of marine unicellular
cyanobacteria: 1. Bilin types and locations and energy-transfer pathways
in Synechococcus spp phycoerythrins, J. Biol. Chem., 266(15), 9515–
Ong, L. J., A. N. Glazer, and J. B. Waterbury (1984), An unusual phycoerythrin from a marine Cyanobacterium, Science, 224(4644), 80–83.
Palenik, B. (2001), Chromatic adaptation in marine Synechococcus strains,
Appl. Environ. Microbiol., 67(2), 991–994, doi:10.1128/aem.67.2.991–
Partensky, F., J. Blanchot, F. Lantoine, J. Neveux, and D. Marie (1996),
Vertical structure of picophytoplankton at different trophic sites of the
tropical northeastern Atlantic Ocean, Deep Sea Res. Part I, 43(8), 1191–
Partensky, F., J. Blanchot, and D. Vaulot (1999), Differential distribution
and ecology of Prochlorococcus and Synechococcus in oceanic waters:
A review, in Marine Cyanobacteria, edited by L. Charpy and A. W. D.
Larkum, vol. 19, pp. 457–475, Bull. de l’Inst. oceanographique,
Proctor, C. W., and C. S. Roesler (2010), New insights on obtaining phytoplankton concentration and composition from in situ multispectral Chlorophyll fluorescence, Limnol. Oceanogr., 8, 695–708, doi:10.4319/
R Development Core Team (2011), R: A Language and Environment for
Statistical Computing. R Found. for Stat. Comput., Vienna, Austria,
ISBN 3–900051-07-0. [Available at http://www.R-project.org, accessed
9 May 2013.].
Scanlan, D. J., and N. J. West (2002), Molecular ecology of the marine cyanobacterial genera Prochlorococcus and Synechococcus,
FEMS Microbiol. Ecol., 40(1), 1–12, doi :10.1016/s0168–
Sepp€al€a, J., P. Ylostalo, and H. Kuosa (2005), Spectral absorption and fluorescence characteristics of phytoplankton in different size fractions
across a salinity gradient in the Baltic Sea, Int. J. Remote Sens., 26(2),
387–414, doi:10.1080/01431160410001723682.
Sidler, W. A. (1994), Phycobilisome and Phycobiliprotein structures, in
The Molecular Biology of Cyanobacteria, edited by D. A. Bryant,
Kluwer Acad., Dordrecht.
Six, C., J. C. Thomas, L. Garczarek, M. Ostrowski, A. Dufresne, N. Blot,
D. J. Scanlan, and F. Partensky (2007), Diversity and evolution of phycobilisomes in marine Synechococcus spp.: A comparative genomics
study, Genome Biol., 8(12), doi:10.1186/Gb-2007-8-12-R259.
Smith, R. C., and K. S. Baker (1984), The analysis of ocean optical data,
Proc. Soc. Photo Opt. Instrum. Eng., 489, 119–126.
Staal, M., S. T. Hekkert, G. J. Brummer, M. Veldhuis, C. Sikkens, S. Persijn, and L. J. Stal (2007), Nitrogen fixation along a north-south transect
in the eastern Atlantic Ocean, Limnol. Oceanogr., 52(4), 1305–1316,
Subramaniam, A., E. J. Carpenter, D. Karentz, and P. G. Falkowski (1999),
Bio-optical properties of the marine diazotrophic cyanobacteria Trichodesmium spp. I. Absorption and photosynthetic action spectra, Limnol.
Oceanogr., 44(3), 608–617.
Taylor, B. B., E. Torrecilla, A. Bernhardt, M. H. Taylor, I. Peeken, R.
Röttgers, J. Piera, and A. Bracher (2011), Bio-optical provinces in the
eastern Atlantic Ocean and their biogeographical relevance, Biogeosciences, 8(12), 3609–3629, doi:10.5194/bg-8–3609-2011.
Toole, D. A., and D. A. Siegel (2001), Modes and mechanisms of ocean
color variability in the Santa Barbara Channel, J. Geophys. Res.,
106(C11), 26,985–27,000.
Waterbury, J. B., and R. Rippka (1989), Subsection I. Order Chroococcales, in Bergey’s Manual of Systematic Bacteriology, edited by J. T.
Staley, M. P. Bryant, N. Pfennig and J. G. Holt, pp. 1728–1746, Williams
& Wilkins, Baltimore, MD.
Waterbury, J. B., S. W. Watson, R. R. L. Guillard, and L. E. Brand (1979),
Widespread occurence of a unicellular, marine, planktonic cyanobacterium, Nature, 277(5694), 293–294, doi:10.1038/277293a0.
Wood, A. M., D. A. Phinney, and C. S. Yentsch (1998), Water column
transparency and the distribution of spectrally distinct forms of phycoerythrin-containing organisms, Mar. Ecol. Prog. Ser., 162, 25–31,
Wyman, M. (1992), An in vivo method for the estimation of phycoerythrin
concentrations in marine cyanobacteria (Synechococcus spp.), Limnol.
Oceanogr., 37(6), 1300–1306.
Wyman, M., R. P. F. Gregory, and N. G. Carr (1985), Novel role for phycoerythrin in a marine Cyanobacterium, Synechococcus strain DC2, Science, 230(4727), 818–820, doi:10.1126/science.230.4727.818.
Zhao, K.-H., R. J. Porra, and H. Scheer (2011), Phycobiliproteins, in Phytoplankton Pigments––Characterization, Chemotaxonomy and Applications in Oceanography, edited by S. Roy, C. A. Llewellyn, E. S. Egeland
and G. Johnson, pp. 375–411 Cambridge Univ. Press, Cambridge.
Zubkov, M. V., M. A. Sleigh, G. A. Tarran, P. H. Burkill, and R. J. G. Leakey
(1998), Picoplanktonic community structure on an Atlantic transect from
50 degrees N to 50 degrees S, Deep Sea Res. Part I, 45(8), 1339–1355.
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF