Limnol. Oceanogr., 41(8), 1996, 1684-1696
0 1996, by the American
Society of Limnology
and Oceanography,
Inc.
The 199 1 coccolithophore bloom in the central North Atlantic.
2. Relating optics to coccolith concentration
William M. Balch
Bigelow Laboratory for Ocean Sciences, M&own
Point, W. Boothbay Harbor, Maine 04575
Katherine A. Kilpatrick
Division of Meterology and Physical Oceanography, Roscnstiel School for Marine and Atmospheric Science, University
of Miami, 4600 Rickenbacker Causeway, Miami, Florida 33 149- 1098
Patrick Holligan
Dept. of Oceanography, University of Southampton Highfield, Southampton, SO17 lBJ, United Kingdom
Derek Harbour
Plymouth Marine Laboratory, West Hoe, Plymouth, United Kingdom
Emilio Fernandez
Depto. Recursos Naturais e Medio Ambiente, Facultad de Ciencias de1 M,sr, Campus Lagoas-Marcosende,
Universidade de Vigo, E-36200, Vigo, Spain
Abstract
This study summarizes the relationships between various biological and optical properties of a mesoscale
coccolithophore
bloom observed in the North Atlantic during June 1991. Backscattering and coccolith
concentration were positively correlated. Backscattering and concentration of suspended calcite were even
better correlated because atomic absorption analyses of calcite calcium were equally accurate whether calcite
was attached or detached from cells, whereas it was difficult to enumerate the numbers of coccoliths attached
to cells by means of microscopy. As the bloom aged, the ratio of detached coccoliths to plated cells increased.
Dilution experiments provided the most precise relationships between coccolith backscattering and coccolith
abundance. The calcite-specific scattering coefficient was estimated from measurements of beam attenuation,
absorption, and calcite concentration. The contribution
of coccolith backscattering to total scattering was
modeled as a function of coccolith concentration and chlorophyll concentration. Even outside the coccolithophore bloom, coccoliths were responsible for 5-30% of the total backscattering. Anomalous diffraction
theory was used to show that calcite-specific scattering is the highest for I-3-pm spheres, which correspond
to the diameters of Emiliania huxleyi coccoliths (this prediction was close to the observed values). The
calcite-specific scattering coefficients of larger calcite particles (e.g. plated coccolithophore cells, foraminifera,
pteropods) would be expected to be considerably lower. These data were used to test an approach for predicting
coccolith concentration from water-leaving radiance in the blue and green wavelengths.
The advent of remote sensing by satellites has highlighted some classes of algal blooms that significantly
modulate the water-leaving radiance of surface waters.
Blue-green algae such as Trichodesmium can cause highreflectance, beige surface “slicks” in the tropics (Borstad
et al. 1992; Lewis et al. 1988). Dinoflagellates can cause
blooms or red tides in shelf waters at frontal scales (Pingree
et al. 1975; Holligan et al. 1984). Intense green openAcknowledgments
ocean diatom blooms have also been observed by means
We thank the officers and crew of the RRS Charles Darwin
of satellites at fi-ontal scalesin central ocean basins (Yoder
for ship handling and the staff of the Research Vessel Services,
et al. 1994). Coccolithophore blooms, which were first
Barry, U.K., for logistical support. Beam attenuation data used
recognized in satellite images more than a decade ago
to calculate total scattering were provided by Charles Trees and
Roy Lowry. Howard Gordon and Ken Voss provided help with
(Holligan et al. 1983), have been observed in both coastal
data interpretation. The comments of two anonymous reviewers
waters and the open ocean, and these blooms make the
are greatly appreciated.
ocean appear a milky turquoise. Coccolithophores, which
This work was supported by the ONR Ocean Optics Program
are members of the algal class Prymnesiophyceae, pro(NO00 14-9 1-J- 1048) NASA Global Biogeochemistry (NAGW
duce CaCO, scales called coccoliths. These scales range
2426), NASA EOSMODIS
(NAS5-3 1363) and NSF Biological
from
1 to 10 b.rn in diameter and have various shapes,
Oceanography Program (OCE 89-00 I89 and OCE 90-22227) to
such as rhombohedral crystals, disks (with or without a
W.M.B.
central hole and radial elements), and “trumpet” shapes
Contribution
2 19 to the U.S. Global Ocean Flux Program
(Reid 1980). Coccoliths are first attached to the cells and
and contribution
96003 of the Bigelow Laboratory for Ocean
later detach into the seawater.
Sciences.
1684
1685
Coccolithophore bio-optics
The main optical impact of coccoliths is an increase in
light scattering. Here, we refer to particle scattering (as
opposed to Rayleigh scattering) to indicate when the
wavelength of light is less than or equal to the particle
size. Such light scattering is mainly in the forward direction and is size-dependent. That fraction of light scattered
in the backward direction per unit thickness is termed
backscattering (bJ. (A list of notation is provided.) Volume-scattering functions of homogeneous particles in the
visible wavelengths (X) are a function of the spectral value
of the real part of the particle’s refractive index (yt), the
spectral value of the imaginary part of the refractive index
(y2’,where n’ = aX/4n and a is the absorption coefficient),
and the particle size distribution. There are relatively few
values of n and n’ available in the literature for coccolithophores (but see Morel 1987 and Bricaud et al. 1992).
The high refractive index of calcite relative to water makes
calcite an extremely efficient scatterer of light. Because
coccoliths do not absorb light at visible wavelengths (Balch
et al. 199 I), the imaginary part of their refractive index
is zero at visible wavelengths.
Four features set coccolithophore blooms apart from
monospecific blooms of noncalcifying algae: the optical
signature of coccolithophore blooms is mostly controlled
by light scattering from detached inorganic coccoliths, not
by absorption and scattering by particulate organic matter; organic biomass of such blooms is often relatively
low; coccolithophore blooms are observed at one part of
the organism’s life cycle, when its plates arc detaching;
and these blooms affect significantly larger areas than do
most other types of monospecific blooms. Areal extents
of typical coastal coccolithophore blooms are 50,000100,000 km2 (Holligan et al. 1983; Balch et al. 199 1;
Brown and Yoder 1993), and open-ocean blooms have
been observed to have areas > 0.5 x 1O6km2 (Holligan et
al. 1993).
Coccolithophores affect not only optics but also seawater chemistry; when in high abundance, they strongly
modulate the alkalinity and ZC02 levels in surface waters
(Robinson et al. 1994). Such chemical changes frequently
parallel optical changes. For example, alkalinity in coccolithophore blooms has been shown to be inversely related to optical backscattering of detached coccoliths.
Moreover, beam attenuation has been positively related
to Pco2 (Holligan et al. 1993).
Coccolithophore blooms also affect other aspects of
biology in the euphotic zone. For example, because of
their ability to survive in moderately stratified, low-nutrient waters, coccolithophores can outcompete diatoms
for nutrients in such conditions. Thus, coccolithophores
represent transitional species between well-mixed, diatom-dominated waters and well-stratified, dinoflagellatedominated waters (Margalef 1979). Moreover, because of
their ability to increase the light scattering of surface waters by detaching their coccoliths, coccolithophores shoal
the euphotic zone, which diminishes the light available
for deeper algal species. In turn, this may allow the nitracline to shoal because nitrate uptake is often lightdependent (MacIsaac and Dugdale 1972).
Even though optical scattering by coccoliths can be
Notation
a
4
a*
b
bChl
b*X
b,
bb’
bb Chl
BC
6b
6b Chl
c cacoj
C
C cnl
c
Chl
d
Y
K&J
K&w)
x
LV
n
n,I
kcco
N cells
m
P
Absorption coefficient, m-l
Particulate absorption coefficient, m-l
Chl a-specific absorption, m2 (mg Chl)-’
Scattering coefficient, m-l
Scattering due to Chl a, m-l
Calcite-specific scattering coefficient, m2 (mg calcite
C)- l
Backscattering coefficient, m-l
Calcite-specific backscattering coefficient, m-l
Chl a-specific backscattering coefficient, m-l
Scattering coefficient when Chl a + pheopigment =
1 mg m-3
bb: b, dimensionless
bbChl: bChl,dimensionless
Concn of calcite as C, j.68liter-’
Beam attenuation coefficient, m- l
Density of calcite, 2.711 x lo6 g m-3
Concn of Chl + pheopigment, mg m-3
Concn of Chl a, mg m-3
Particle diameter, m
Ratio of calcite to Chl by mass, mg calcite C (mg
Chl a)-’
Vertical attenuation coefficient for downward irradiance, m - 1
Vertical attenuation coefficient for downward irradiance at midpoint of euphotic zone, m-l
Avg Kd in euphotic zone, m-l
Wavelength, nm
Water-leaving radiance, W m-2 sr-l
Index of refraction, = 1.58 for calcite
Index of refraction of water, = 1.33
Imaginary part of the refractive index
Concn of coccoliths, per ml-l
Concn of coccolithophore cells, per ml-l
Relative refractive index, n/n,
Dimensionless parameter relating the size of particles to the wavelength of light and to their relative refractive index, 2n(d/X)(m - 1)
significant, there are relatively few studies that define the
precise relationship between coccolith abundance and light
scattering. Earlier studies showed that the relationship
between coccolith concentration and light scattering was
highly significant but that this relationship varied with
growth phase of the cells or coccolith detachment rates
(Ackleson et al. 1994; Balch et al. 199 1). The exact nature
of the growth effects has remained unknown for field populations.
Our companion paper (Balch et al. 1996) examines the
spatial variability of optical properties in a mesoscale
coccolithophore bloom in the North Atlantic Ocean. Here,
we relate the optical properties (backscattering and total
light scattering) to coccolith abundance, suspended calcite
concentration, and particle size in this 0.5 x 106-km2
bloom. We have applied simple models to explain the
effects of chlorophyll and calcite on inherent optical properties of absorption and scattering in seawater. Dilution
experiments were used to definitively relate backscattering to coccolith concentration at a given location within
the bloom. The light scattering results are compared to
predictions based on calcite spheres of various sizes.
1686
Balch et al.
Table 1. Statistical summary of linear optical relationships found during the B0FS cruise, including the independent variable
(IV.) and its associated standard error of prediction, dependent variable (D.V.), number of points, least-squares fit slope, standard
error (SE) of the slope, least-squares fit intercept (Int.), SE of the intercept, coefficient of determination,
and F-statistic (regression
mean square/residual mean square). All P values (the probability that the slope is equal to 0) <O.OOl (type 1 error). Units of the
variables are Nce1,9(cells ml-l), N,,,, (coccoliths ml-l), CCaCo3(mg C liter-‘), bb’ (m-l), b (m-l), b : a (dimensionless), and y [mg
calcite C (mg Chl a)-l].
SE
Y
I.V.
loi3PLkl
K&o31
b’b 436
b’b
546
p;
b436
b546
b546
b546
: a546
: a546
t
D.V.
0.3 1
51.69
9.6 lE-3
8.35E-3
5.24E-3
4.33E-3
0.552
0.492
7.70
5.28
t For data where
b,,,
log Wcoccol
wccm1
PLcco1
PLmo1
Lco31
K&o31
K!aco,l
Laco31
Y
Y
: a546> 10 only.
n
172
102
61
61
118
118
140
143
135
83
Slope
0.91
l.O5E-3
1.84E-7
1.35E-7
1.76E-4
1.36E-4
9.37E-3
8.41E-3
6.69E-2
5.94E-2
Methods
Observations were made on cruise 60 of the RRS
Charles Darwin (CD60) from 13 June to 3 July 199 1. The
cruise track is given by Balch et al. (1996, figure 1). Two
transects were made along the 20 and 15”W meridian.
Detailed optical measurements of volume scattering and
absorption were performed at 111 and 2 10 stations, respectively. For comparative purposes, the ship visited a
site well outside the bloom (59”39’N, 20’59’W; site of an
optical mooring). Other cruise details are given by Holligan et al. (1993). Balch et al. (1996) describes the methodology for measuring total light scattcrjng (b), backscattering (b,), and particulate absorption (a,), all at 436and 546-nm wavelengths. Briefly, backscattering due to
calcite (bb’) was calculated as the difference between the
bbof raw seawater and the bl, of the same sample following
30 s of bubbling with CO2 (which reduced the pH to 5.5
and dissolved the CaCO,).
Dilution experiments- Water was sampled with a Niskin bottle and transferred to a 1-liter polyethylene bottle.
A loo-ml aliquot of seawater was serially diluted with
0.2~pm-filtered seawater to achieve a series of coccolith
concentrations. Then, volume scattering was measured
and backscattering calculated for each concentration.
Total particulate carbon and calcite analyses-Total
particulate carbon was measured according to Fernandez
et al. (1993). The technique involved passing 0.5 liter of
seawater through precombusted Whatman GF/F filters
and freezing at - 20°C. A Carlo-Erba 1500 (series 2) CHN
analyzer was used to measure the total carbon, with no
pretreatment to remove CaCO,. Calcite was measured by
collecting seawater on precombusted Whatman GF/F filters and then measuring the atomic absorption of calcium
with the technique described by Holligan et al. (1993).
SE
3.72E-2
6.93E-5
1.82E-8
1.58E-8
6.33E-6
5.23E-6
5.28E-4
4.64E-4
6.63E-3
5.20E-3
Int.
1.60
48.27
3.36E-3
3.06E-3
-4.67E-3
-3.61E-3
4.42E- 1
4.10E-1
7.62
13.09
SE
1.21E-1
7.41
1.82E-3
1.58E-3
7.17E-4
5.92E-4
7.23E-2
6.28E-2
1.03
0.943
r2
F
0.779
0.690
600
231
102
74
775
674
315
328
102
130
0.634
0.556
0.866
0.849
0.695
0.700
0.433
0.617
This technique assumes that all particulate calcium is in
the form of C&O,. Briefly, l-liter samples were passed
through precombusted Whatman GF/F 25-mm filters,
and the filters .ryererinsed with several washes of filtered
seawater before freezing at - 20°C. Samples were extracted by adding 2 ml of 50% trace-metal-clean HCl to tubes
that contained the filters. Samples were incubated overnight at 40°C in a water bath. Eight milliliters of 1%
lanthanum chloride was added to each tube (to prevent
phosphorous suppression of ionization). The supernatant
was then injected into a flame photometric atomic absorption spectrometer that measured absorption at 422.7
nm with a IO-cm air-acetylene flame. Calibration curves
were prepared ‘with commercial calcium standards. Blank
filters were pre:?ared near the end of the cruise by mounting identical precombusted filters in the filter tower apparatus, applying vacuum, adding GF/F-filtered seawater
to wet the filter, and then by following rinsing and preparation techniques as described above. Total organic carbon was estimated as the difference between total particulate carbon and inorganic calcite carbon.
Cell and coccolith enumeration -Cell and coccolith
enumeration was performed according to Holligan et al.
(1984). Counts were performed with an inverted microscope on samples fixed with buffered Formalin and Lugol’s iodine.
Results
Standing stocks of cells, coccoliths, and carbon- We
observed well-defined patterns in the cell count and particulate analyses during the mcsoscalc coccolithophore
bloom. Coccolith concentrations [IV,,,,,] were about an
order of magnitude higher than coccolithophore cell concentrations [N,,Z,l,].Note, however, that the ratio of coccoliths to cells ‘wasnot constant; a least-squares linear fit
1687
Coccolithophore bio-optics
B
l
0.05
0x1o” 1x105
2x105
3x105
4x105
E. huxleyi coccoliths ml”
Fig. 1. A. Relationship between detached coccolith concentration and numbers of Emiliania huxleyi coccoliths per liter.
Line is a least-squares fit to the data. B. Suspended calcite conccntration vs. coccolith concentration. Line is a least-squares fit
to the data. Equation given in text.
explained only 53% of the variance with the standard
error of the dependent variable [IV,,,,,] of 45,259 coccoliths ml- l. A least-squares power fit explained considerably more variance (78%) (relationship and error terms
are given in Table l), implying that the ratio of detached
coccoliths to cells decreased from 26 to 17 as cell abundance increased from - 100 to 10,000 cells ml-l (Fig.
1A). The standard error of the dependent variable (log
[IV,,,,,]) was kO.3 1 log units, or a factor of -2.
A comparison of total particulate carbon vs. calcite
carbon concentrations revealed that the techniques were
internally consistent; in all cases (except one), total par-
100
260
300
4 0
Suspended calcite C (pg liter-‘)
Fig. 2. Calcite-specific backscattering at 436 nm and 546
nm as a function of concentration of suspended calcite. Lines
are least-squares fits to the data. (Equations given in Table 1.)
ticulate carbon exceeded calcite carbon. Although there
was not a highly significant relationship between total and
calcite carbon, the typical ratio of total particulate carbon
to calcite carbon varied between 2 and 6, and values
approached 1 as the calcite concentration increased (data
not shown). Plotting calcite carbon concentration ([C,,,,]
in pg liter-l) vs. coccolith concentration (per ml-l) provided an estimate of the carbon per coccolith. The lcastsquares relationship is given in Table 1 (Fig. 1B).
Backscattering and scattering as afunction of suspended
calcite-The blue and green wavelength bb’ values represent only the backscattering due to CaCO, (with no
1688
Balch et al.
Table 2.
colithophore
Results of dilution
bloom.
experiments
Slope*
Depth
1991
22 Jun
24 Jun
25 Jun
27 Jun
29 Jun
29 Jun b
30 Jun
* x 10-10.
t x 10-3.
Location
61.5N,
60.9N,
60.9N,
61.2N,
61.1N,
61.ON,
62.ON,
22.6W
22.9W
23.9W
15.2W
15.ow
15.6W
15.2W
at various
b-0
n
12
2
2
2
2
2
2
7
6
7
5
7
5
6
5.34
2.58
4.66
1.35
1.58
2.65
1.96
particulate
organic C). Our results of comparisons
between bb’ and the numbers of coccoliths or concentration
of CaCO, are presented in Table 1. Although the data set
for comparison of bb’and calcite concentrations was larger
than that for comparison of bb’ and coccolith concentrations (owing to the time-consuming nature of microscopic
cell counts), it is not apparent that this difference biased
our results since calcite concentration and coccolith abundance were sampled throughout the bloom. Highest coefficients of determination and lowest standard errors resulted when bb’ was predicted from calcite concentration
(Table 1, Fig. 2) (see below).
The best relationships between bb’ and coccolith concentration were achieved during the dilution experiments.
In seven experiments with water from the top 12 m, the
average coefficient of determination was 0.96 for all 436and 546-nm measurements (Table 2, Fig. 3). Moreover,
the slopes of the bb’ vs. coccolith relationships varied by
-4 x at either wavelength, with the largest slopes having
been observed early in the event (22, 24, and 25 June)
and reduced values at the end of the bloom (27 and 29
June).
Beam attenuation (c) was measured along-track and in
vertical profiles during the bloom (see Balch et al. 1997).
Given that beam attenuation is the sum of scattering (b)
and absorption (a), it was possible to calculate total scattering as the difference between c and a, where total absorption was taken as that due to water and particles only
(Balch et al. 1996). Note that scattering from calcite only
(b’; m-l) could not be calculated with this technique because the treatment to dissolve coccoliths could not be
performed in situ in the volume necessary to be viewed
by the transmissometer. Nevertheless, there was a positive relationship between total scattering (b436or bsd6)and
calcite concentration that was well fit by the linear relationships given in Table 1 (Fig. 4). For relating calcite
concentration to light scattering, the green wavelengths
are preferred to the blue wavelengths as the impact of
chlorophyll absorption is minimized. For the b,,, vs.
[C Ca,--3]relationship in Table 1, the y-intercept (0.4 10
m-l) was significantly different from zero and had confidence limits of 0.29-0.53 m-l. The slope of this line,
8.413 x 1o-3 m2 (mg calcite C)-l, represented the calcitespecific scattering coefficient (b*550).The 95% error limits
locations within
1nt.t
r2
the mesoscale coc-
Slope*
436 nm
-
1.76
0.31
0.97
0.50
-0.98
0.33
1.19
0.95,
0.91
0.91
0.9s
0.95
0.9s
0.99
1nt.t
r2
546 nm
3.72
1.82
3.56
1.01
1.22
2.07
1.54
1.3
-3.01
0.59
1.5
-0.80
0.036
0.43
0.99
0.97
0.91
0.84
0.97
1.oo
0.99
“.“I
A
0.06
+
436 nm
/
5.OE+7
1 .OE+8
2.c ‘+8
1.5E+8
0.05
l-
B
0.04i
546nm
/+
I
r
”
5.OE+7
Detached
1,
1 .OE+8
coccoliths
7 ’
I
1.5E+8
a ’
”
2.0 J+8
liter-’
Fig. 3. Results of seven dilution experiments in which seawater (containing coccolithophores and detached coccoliths) was
serially diluted, and calcite-dcpcndent
backscattering was estimated at 436- and 546-nm wavelengths. The lcast-squares fit
is shown for each experiment. The slope and intercept values
for each experiment are given in Table 2. Note that ordinate
scales arc different. Symbols (199 1 sample dates): + - 22 June;
A-24 June; O-25 June; a-27 June; V-29 June; Cl-29 June
b; x -30 June.
1689
Coccolithophore bio-optics
Coccoliths ml-’
400
Fig. 5. Calcite-specific
scattering at 546 nm vs. coccolith
concentration showing trend toward low b* values at high detached coccolith concentrations. Solid line-the
slope in Fig.
4B; dashed line-the
theoretical maximum b* calculated with
anomalous diffraction theory (van dc Hulst 198 1). Details of
the calculation given in discussion.
of scattering to absorption (b : a) to the ratio of the concentrations of the principal scatterer to the principal absorber (mg calcite C : mg Chl, hereafter dcsignatcd y).
Note, however, that the correlation was lower than expected; it was strongest at 550 nm, but still accounted for
only 43% of the variance. For data where b : a > 10 (representative of the most turbid parts of the bloom), the
b : a ratio vs. y least-squares fit accounted for 62% of the
variance (Table 1). At 436 nm, the correlations between
b : a vs. y never cxceedcd r2 of 0.44 (data not shown).
Discussion
100
Suspended
200
300
calcite (mg C m”)
Fig. 4. Scattering at 436 and 546 nm vs. the concentration
of calcite carbon. Lines are least-squares fits to the data; fitted
slopes and intercepts given in Table 2.
on the slope were 7.49-9.33 x 10m3m2 (mg calcite C)-I,
and there was some bias in the error, with b*550values
highest at low coccolith concentrations (Fig. 5).
Scattering per unit chlorophyll was calculated for comparison with the historical data collected by Morel (1987);
such values were generally 10 x greater in this coccolithophore feature than in typical nonbloom populations, and
there was no correlation between chlorophyll and light
scattering. A different approach to account for the absorption and scattering values was to compare the ratio
From a remote-sensing perspective, the timescale for
the onset of turbid conditions was driven by three factors:
the calcification rate, the coccolith detachment rate, and
numbers of coccoliths per coccolithophore. Numbers of
coccoliths per cell have been previously estimated at - 15
for Emiliania huxleyi (Paasche 1962). Linschooten et al.
(199 1) also found that E. huxleyi in a 16 : 8 L/D cycle
and grown on full-strength IMR medium (Eppley et al.
1967) had 15-20 coccoliths per cell at the end of the light
period. They pointed out that this number of coccoliths
was sufficient to cover the cells in a single layer. Our
previous estimates (Balch et al. 1993) have shown that
numbers of coccoliths can reach as high as 80 per cell at
midlogarithmic growth, and thcsc numbers decrease as
cells reach stationary phase. By USCof coccolith and cell
size only, we previously calculated that the first layer of
plates consists of - 15 coccoliths. The relationship bctween detached coccolith abundance and cell abundance
from this North Atlantic bloom was similar to previous
laboratory studies (Balch et al. 1993); on average, each
cell detached a maximum of 26 coccoliths, or almost two
layers, during the early stages of growth and a minimum
of 17, or about one layer, toward the later stages of the
bloom (Fig. 1, Table 2). This should not be confused with
1690
Balch et al.
the results of Fernandez et al. (1993, table l), who showed
that the coccolith : cell ratio in the same bloom varied
about one order of magnitude at five stations, from 12
to 108. The ratios cited here are average values derived
from all of the cell counts shown in Fig. 1. For example,
use of 95% C.I. for the slope and intercept of the first
regression in Table 1 (log coccoliths vs. log coccolithophores) gives a range in coccoliths : cell of 1l-65 when
there are 100 cells ml-l and a coccolith : cell ratio of 56 1 when the cell concentration is 10,000 cells ml-l. These
ratios for the entire data set are a bit lower than those
cited by Fernandez et al. (1993).
That the ratio of detached coccoliths per cell varied
and that there were other phytoplankton species undoubtedly caused large variance in ratios of total particulate carbon to calcite carbon. In the less concentrated
parts of the bloom, calcite carbon represented - 16-50%
of the total carbon, whereas in the densest part of the
bloom, virtually all of the carbon was as calcite. Where
did the organic carbon of the coccolithophores go? Cell
lysis may have occurred, but there was little evidence for
unusual numbers of viruses in the late bloom stages (see
Bratbak et al. 1993). It seems equally probable that the
lack of organic carbon resulted from highly efficient removal by grazers or by sinking (Holligan et al. 1993).
The calcite concentration data combined with coccolith
count values (Fig. 1B) provided an opportunity to check
the calcite carbon per coccolith for field populations of
E. huxleyi. Previous laboratory and field studies have
shown that this value is variable. For example, Paasche
(1962, 1963) gave a value of 0.17 pg C per coccolith for
cultured cells, whereas Linschooten et al. (199 1) estimated this value as 0.065-0.078. Balch et al. (199 1) estimated 0.26 pg C per coccolith based on CHN measurements of field populations; however, Holligan et al.
(1983) found distinctly higher values of 0.5 and 0.6. The
least-squares fit to the data of Fig. 1B gave 1.05 pg C per
coccolith, but this high value resulted because coccoliths
attached to cells were not included in the coccolith count.
When the coccospheres were multiplied by 20 (assuming
20 coccoliths per cell) and added to the detached coccolith
concentration, the count reduced to 0.47 pg C per coccolith (Fernandez et al. 1993), which is more reasonable
but is still higher than results from previous laboratory
studies. This calculation also implied that, on average,
more than half of the coccoliths were detached in the
North Atlantic bloom of 199 1, which has important ramifications for interpreting the light-scattering data.
Rackscattering and coccolith concentration-Two
fundamentally different approaches were used to understand
the impact of coccolith density on calcite-dependent
backscattering (bb’). The first approach involved measuring bb’ of seawater samples and plotting it against coccolith concentration with no dilution (Fig. 4) (hereafter
referred to as the nondilution approach). Note that in
making these regressions, data arc pooled from various
locations within the bloom, regardless of growth stage.
Our experience has shown that calcite backscattering
does not vary simply as a function of coccolith density;
other factors may be involved, such as the number of
detached vs. attached coccoliths, variance in the abundance of other calcifying algal species, coccolith size, and
coccolith integrity. All of these factors may vary with the
growth stage 01’ the bloom.
The second approach involved serial dilutions of individual samples and the examination of bb’ against coccolith abundance (hereafter called the dilution approach).
The purpose of the dilution experiments was to examine
the relationship between calcite-specific backscattering and
coccolith concentration for specific stages of bloom development. Results from the dilution experiments allowed deductions about how much variance in the bb’ vs.
coccolith relationship could be attributed to growth stage.
The results allowed limited speculation on the cause of
the variations; for example, they did not discern between
coccolith integrity and coccolith size. Because this analysis focused only on bb’ and not on bb, potentially complicating factors related to other organic particles were
not invoked to nterpret the results. For example, changes
in the refractive index were not expected to cause changes
in bb’ (unless calcite particles had been replaced with aragonite particles, which was not evident).
In our previous studies in the Gulf of Maine with the
nondilution approach (Balch et al. 199 l), 80-85% of the
variance in bb’ ,;ould be accounted for by the abundance
of detached coccoliths. In the North Atlantic coccolithophore bloom, only 56-63% of the variance in bb’ could
be explained by’ means of the nondilution approach (Table 1). For the BOFS expedition, use of the dilution approach increased the explained variance another 30-40%
(to >95%) for 11 of 14 experiments (Table 2, Fig. 3).
Two of the exceptions that had lower explained variance
were experiments performed at the station at 60.9”N,
23.9”W on 25 J.une, for which results were notably curvilinear. Multiple scattering is an unlikely reason for the
curvilinear behavior because the detached coccolith concentration was not particularly high to begin with, certainly no higher- than in other experiments in which the
relationship was clearly linear. The other low r2 value was
associated with the 546-nm data on 27 June (station at
61.2”N, 15.2”W; r2 = 0.84). This result seems to be due
to an anomalous data point at 2.25 x 10’ coccoliths liter-l. Note that the 436-nm data at the same station were
highly linear (r:! = 0.99), which supports the possibility
that the 546-nm data had a bad data point. Overall, the
dilution experirnent results clearly showed differences in
the bb’ vs. coccolith relationship for different parts of the
bloom. The most likely cause of the high variance between dilution experiment “slopes” was the differences
in the ratio of cletached coccoliths to plated cells, which
varied with bloom age. Nevertheless, we cannot exclude
changes in coccolith integrity or the size distribution of
the coccoliths (the latter is admittedly one of the most
important factors that affects light scattering of marine
particulates).
We estimatec the age of this bloom from the earliest
satellite image, which showed no high-reflectance water
(10 June 199 1); by the next available image (15 June
199 l), the bloom was well under way. Our use of a 10
1691
Coccolithophore bio-optics
June start date may cause a slight overestimate in the age.
From the dilution experiment data, we observed that the
detached coccolith : cell ratio increased with bloom age
(Table 3). Moreover, as the detached coccolith : cell ratio
increased, the slopes from the dilution expcriments-essentially the backscattering per detached coccolith-decreased (Fig. 3).
Many of the above problems can be avoided by using
the mass of CaCO, instead of numbers of coccoliths to
quantify coccolith abundance. The nondilution approach,
but with bb’ regressed against calcite concentration, explaincd - 8 5% of the variance in bb’for all cruise samples
(Fig. 2). This tighter fit probably was because coccoliths
attached to cells were included in atomic absorption estimates but were not included in microscope enumeration.
Calcite concentration explained 70% of the variance in
total scattering (bSb6;Fig. 4) for two reasons: estimates of
b,,, were based on the difference between beam attenuation and absorption and measurement errors likely were
greater for b than for b,‘; and the data included scattering
from other types of noncalcite particles. Another factor
that may have affected the relationship between calcite
concentration and light scattering was particle size, which
is addressed later.
Calcite, chlorophyll, and b : a-Kirk (198 la, b) used a
Monte Carlo approach to demonstrate that b : a is proportional to K,(z,)la (vertical attenuation coefficient for
downward irradiance at the midpoint of the euphotic zone
divided by the absorption coefficient; hereafter, the euphotic zone is defined as the depths at which downward
irradiance > 1%). The relationship also included two other factors: po, the cosine of the zenith angle of refracted
solar photons just beneath the surface, and G(po), the
fraction of scattering to vertical attenuation (determined
from the scattering phase function). Kirk’s semi-empirical
relationship is given as
K,(z,,)la = l/po[ 1 + G(po)bla]lh .
(1)
Kirk (199 1) applied a similar Monte Carlo analysis to
seven different water types of widely variable volumescattering functions (Petzold 1972). Hc showed that Eq.
1 was generally applicable. Moreover, assuming that the
incident light was vertical (ho = 1) and replacing &(z,)
with &(avg) (the average Kd above the depth at which
downward irradiance equals 1% of the surface value),
Kirk showed that G( 1) was approximately constant with
a C.V. of 3.9%. The specific value of G( 1) from San Diego
Harbor based on Petzold (1972) was 0.23 1; this is highly
relevant to our work because San Diego Harbor has been
shown to have a volume-scattering function very similar
to a previously observed coccolithophore bloom (Balch
et al. 199 1). Eq. 1 can therefore be rewritten as
K,(avg)la = (1 + 0.23 lb/a)“.
(2)
Eq. 1 and 2 yield plots of K,la vs. b : a that are slightly
curvilinear (concave-down) up to b : a of 30. For the limit
value of b : a = 0 (only absorption, no scatter), K,(z,)la
equals 1.O (Kirk 199 1, figure 1). To illustrate the role of
coccoliths in increasing the effective pathlength of light
Table 3. Detached coccoliths per plated coccolithophore (D.
cocco. : P. cocco.) and coccolith-specific backscattering (bb*, m-l
per coccolith) as a function of bloom age (d) (note that we assume
that the bloom began on 10 June 199 1, the date of the last clear
satellite image; see Balch et al. 1996). Data are given only for
stations where bb’ was measured. This table shows how normalization of b,’ by the number of detached coccoliths, not
including those attached to cells, can increase coccolith-specific
backscattering at early stages of bloom development.
1991
22
24
25
27
29
29
30
Jun
Jun
Jun
Jun
Jun
Jun
Jun
Bloom age
(d)
12
14
15
17
19
19
20
D. cocco. :
P. cocco.
h*
[X lo-l3 m2
(det. cocc.)- l]
7.65
11.94
9.07
19.49
18.48
28.29
36.99
3.72
1.82
3.56
1.01
2.07
1.22
1.54
moving vertically through the water column, WCsubstituted our b : a values into Eq. 2. In the North Atlantic
coccolithophore bloom, 87% of the b : a values at 550 nm
exceeded 5 and 62% exceeded 10, with the peak b : a
values of 45. Therefore, we would expect K,(avg)la values
to fall between - 1.5 and 3.4.
The b : a in typical phytoplankton blooms is usually a
function of chlorophyll, but as already discussed, b : a at
546 nm in this coccolithophore bloom was mostly a function of y, especially in the densest parts of the bloom.
The slope of the relationship between y and b : a at 550
nm in the most turbid regions (b : a > 10) was -0.06 mg
Chl (mg calcite)- l (Table 1).
To check for internal consistency between the observed
inherent optical properties and the quantity of calcite and
chlorophyll, we wrote the following specifically for turbid
parts of this bloom in which b : a > 10:
b/a = (b* x CaCO,)l(a* x Chl) > 10.
(3)
Eq. 3 was then rearranged, substituting y, to give Eq. 4,
again applicable only to the turbid parts of the bloom:
lOa*:b* < y.
(4)
Based on previous calculations of a* and b*, the left side
of Eq. 4 equaled 15.33 mg calcite C (mg Chl)-’
[=(10x0.0129 m2(mgChla)-‘)/8.413x
10-3m2(mgcalcite C)-l]. As expected, y values in the coccolithophore
bloom fell above this value 97% of the time. Moreover,
the slope in the plot of y vs. bSd6: a546for all data for
which b : a > 10 was 0.059 mg Chl (mg calcite C)-l (Table
l), which represents a statistical average of all our data.
The reciprocal of this slope [ 16.9 mg calcite C (mg Chl)- ‘1
fell within 10% of lOa* : b*.
Modeling light scattering due to coccoliths-Our field
results provided the means to assess the magnitude of
calcite-dependent light scattering relative to the total
backscatter of seawater. The calculations were based on
either coccolith-specific backscattering or calcite-specific
1692
Balch et al.
1
D
1x104
Coccolith concentration
1x105
(ml-l)
1x106
1x10'
1x10'
Calcite
:oncentration
1x103
(pg C liter-l)
Fig. 6. Modeled ratio for bo’ : bbtotat 436 nm as a function of (A) coccolith concentration
and (B) calcite concentration.
Modeled ratio for bb’ : bb tot at 546 nm as a function of (C)
coccolith concentration and (D) calcite concentration. Each line assumes a Chl a concentration,
from top to bottom, of 0.01, 0.03, 0.1, 0.3, 1.0, 3.0, and 10.0 pg liter-‘. A-Data
points from
the mesoscale coccolithophore bloom.
backscattering. However, several assumptions were required: we used 1.84 x 10e7 and 1.35 x 1O-7 (m2 coccolith)-’ for the coccolith-specific backscattcring coefficients at 436 and 546 nm (slopes in rows 3 and 4 of Table
1); we used the slopes from rows 5 and 6 of Table 1 to
calculate the calcite-specific backscattering coefficients;
and we used values for backscattcring due to water of
2.27 X 10m3and 0.965 x 1O-3 m-l for 436 and 546 nm,
respectively (Gordon et al. 1980).
Backscattering due to chlorophyll involved two calculations. The first calculation was based on a bio-optical
model by Gordon (1987) that was designed for case 1
waters (i.e. waters in which optical characteristics are
dominated by algae and their associated detritus). Particulate scattering at 550 nm (b, 546) was empirically predicted with a relationship from Gordon and Morel (1983;
based on 550-nm light): bp 550= B,C”-62, where B, represents the scattering coefficient
when the concentration
of chlorophyll plus pheopigment (C) is 1 mg m-3. Typically, B, ranges from 0.12 to 0.45, with an average of
0.3. Moreover, Gordon (1992) incorporated three factors
into his model: that detritus scatters inversely with wavelength; that phytoplankton scattering is much less wavelength-dependent; and that the wavelength-dependent effects arc most pronounced at low C, whereas at high C,
b is essentially wavelength-independent. The relationship
derived to satisfy the above observations was (Gordon
1992, equation 14)
h,(X) = (Cl.3 Ch1°.62)))([0.5+ (0.25 log Chl)]
+ (0.5 - [0.25(log Chl)]})55O/h/I,
(5)
where X represents the wavelength in nanometers. Next,
bb : b, or &,, was empirically related to chlorophyll at the
two wavelengths by Gordon and Morel (1983, p. 62):
& &436)
= 1.005(Chl-0.404)0.0 1,
(6)
and
&, Ch1(560) = 1 .009(Chl-0.262)0.0 1.
(7)
Note that Eq. ;’ applies to 560 nm and not to 546 nm,
but the effect OFthis difference is minor. Also, both Eq.
6 and 7 include multipl&ation by 0.0 1 because the original equations provide bb values in percentages (Gordon
and Morel 1983). The-value bb ,&I) was taken as the
product of bChl(X)and bbChl(X):
bbml@)= CkdVl[& cdVI-
(8)
The results oY this modeling, as well as data from the
bloom, are shown in Fig. 6. As chlorophyll increased,
chlorophyll bacltscattering increased according to Eq. 6-
1693
Coccolithophore bio-optics
0.0061
Range in diameter of E.
huxleyi coccoliths
Range in diameter
of foraminifera tests
Range in diameter
of most coccoliths
-
Pteropod shells and oolites
-1-3 mm in diameter
Size range of coccospheres
Range in length of
aragonite needles
Fig. 7. Theoretical calcite-specific scattering coefftcient for
calcite spheres vs. sphere diameter calculated with the anomalous diffraction theory (van de Hulst 198 1). Details of calculation given in text. The size ranges of various CaCO, particles
found in the sea are shown for reference (see Berger 1976; Seibold and Berger 1982; Winter and Siesscr 1994).
8. Thus, the fraction of total backscattering attributed to
coccoliths would have decreased solely because of the
percentage of organisms that contained chlorophyll relative to those that contained calcite. We acknowledge that
backscattering can be strongly influenced by absorption
in phytoplankton (Morel and Bricaud 1980) and that the
two variables are not necessarily independent. Consequently, the above equations that relate chlorophyll to
backscattering are empirical and thercforc implicitly include effects of chlorophyll absorption. The relationship
between bb’ : bband coccolith concentration or calcite concentration was curvilinear on a log-log plot, with the lines
converging toward higher coccolith concentrations. At
low chlorophyll concentrations (e.g. 0.1 pg Chl a liter-l),
a concentration of 2,000 coccoliths ml-l is sufficient to
increase the calcite-dependent scattering to 10% of the
total scattering at 546 nm. At a concentration of 10,000
coccoliths ml-l (still considered nonbloom), coccoliths
represent -35% of the total bb at both wavelengths. Values of bb’lbbtot are 50% between coccolith concentrations
of 20,000-40,000 coccoliths ml-l, depending on the chlorophyll level.
The data and predictions for bh’lbbtot showed values of
50% between calcite concentrations of 40-80 pg calcite
C liter-l, depending on the chlorophyll level. The model
predictions agreed with data mostly at high coccolith den-
sities, but there was considerable error at the lowest coccolith densities. Cruise data (plotted on the same figure)
often fell outside the expected range for water containing
0.01-10 mg Chl a liter-l, especially at 436-nm wavelength. The variation in coccolith-specific backscatter (Fig.
3) and calcite-specific backscattcrance (Fig. 2) may account for some of the problem. Also, Eq. 6, 7, and 8 use
average relationships to predict backscatter from chlorophyll that also have large confidence limits. Failure of
the data to fall within the bounds set by the different
chlorophyll levels in Fig. 6 may also be because of the
reduced accuracy of the chlorophyll component of the
model. Regional relationships for these equations might
provide one solution to this problem. This model only
begins to describe the envelopes of variability of calcitedependent scattering as a function of the concentrations
of coccoliths, calcite, or Chl a. More refinements are clearly necessary.
As a cross-check of the scattering coefficients we observed, we calculated calcite-specific scattering for calcite
spheres based on anomalous diffraction theory for nonabsorbing spheres (van de Hulst 198 1, p. 176). Morel
( 1987) defined the attenuation efficiency factor (Q,) as a
function of the dimensionless parameter p, which is a
function of particle diameter (d), wavelength (X), and refractive indices of water (n, = 1.33; Jerlov 1976) and
calcite (n = 1.583; Aas 198 1). The refractive indices then
were used to calculate the relative refractive index (m =
n/n, = 1.19 for coccoliths):
p = 2T(dlX)(rn - 1).
(9)
Next, p was used to approximate the cffrciency of attenuation:
Q,(p) = 2 - (4/p)sin p + (4/p2)( 1 - cos p).
(10)
This function was derived for particles where 2r(d/X) >>
1 and the quantity m - 1 < 1. For coccoliths, the value
of 27r(d/X) at 550 nm is 1l-23. As described by van de
Hulst (198 l), Eq. 10 models light extinction not only
when m is close to 1 but even up to values of m = 2 (see
figure 32 of van de Hulst 198 1). Calcite is effectively
nonabsorbing, so its absorption cfhciency, Q,, is zero and
its scattering efficiency, Qb, dominates the attenuation
efficiency (that is, Q, = Q, + Qb, so with calcite particles,
Q, = Qb). The calcite-specific scattering coefficient was
calculated by applying Eq. 9 and 10 to the following equation (Morel 1987):
b* = (3r/c,J)nw(m
- l)Qb(p)lp.
(11)
ccal is the density of calcite (2.7 11 x lo6 g mm3). These
calculations were performed for calcite spheres from 0.1
to 1,000 mm. Note that Eq. 10 can be used to calculate
the efficiency of attenuation (or in this case the efficiency
of scattering) provided that rd/X > 1. For these calculations, this would occur at a diameter of 0.14 pm (the
abscissa of Fig. 7 therefore begins at 0.1 pm). The calcitespecific scattering coefficient (b*) peaked at diameters between 1 and 3 pm, which is precisely the size range of E.
huxleyi coccoliths (Fig. 7). From this approach, a 2-pm
1694
Balch et al.
0.3-
ith concentration
0.25E
$j
0.25?.
n
$ 0.152
0.1 -
““:I
0
0.05
0.1
0.15
Oh5
011
0.115
b,/a+b,
0.2
0.25
0.3
0.35
0.3
]
0.1
0.05
0
012
(436 nm)
o.i5
013
0.
Fig. 8. Values of b&a + bb) at 546 nm plotted vs. b,l(a +
bh) at 436 nm. Concentrations of coccoliths are contoured that
show high dependence of b,l(a + bb) at 546 nm on coccolith
abundance and relative independence of b,/(a + bb) at 436 nm
to coccolith abundance. The values b,l(a + bb) are expected to
act similar to reflectance
respective wavelengths.
and water-lcaving
irradiance
at the
calcite sphere should have a b* of 9.7 x 1OP3m2 (mg calcite
C)-l.
The average observed calcite-specific scattering coefficient for 2-pm coccoliths [b* = 8.4 x 10e3 m2 (mg calcite
C)-l; Fig. 41 was 13% lower than the theoretical value of
b* for a 2-pm calcite sphere (P < 0.01). This difference
may have been related to the effect of shape on calcite
scattering (disks vs. spheres). Size differences may also
have caused the error limits in the slopes of Fig. 4 (2 SE
about these slopes represented +8%; Table 1, rows 5 and
6). As already shown, theoretical results for calcite spheres
suggested that calcite-specific scattering should increase
with decreasing size, peaking at a diameter of 1 pm and
declining below this size (Fig. 7).
Also note that we observed slightly higher b* values at
low concentrations of calcite. These high values may have
resulted from differences in calcite particle size or shape
when the concentration of suspended calcite was low (see
also Aas 1984). However, there was a cluster of b*,,,
values at early bloom stages[2 x 1O-2 m2 (mg calcite C)- l]
that were above the theoretical value for calcite spheres.
We speculate r.hat such high b*550values may have been
due to calcite.containing species other than E. huxleyi
(e.g. Coccolith YSpelagicus).
.
The peak of’the theoretical b* vs. size relationship in
Fig. 7 was - I .25 pm [b* = 1.19 x lop2 m2 (mg calcite
C)- ‘1. This peak may indicate which calcite particles cause
the most scattering in the sea. Most coccoliths, including
those of E. huxleyi, are l-2 pm in diameter, which, provided that their scattering behavior is similar to that of
calcite spheres, would give them high calcite-specific scattering efficiency. Moreover, one would expect lower-values of b* for calcite particles < 1 pm and > 3 pm, which
implies that, per unit mass, coccoliths arc likely more
important modulators of ocean scattering than are plated
coccolithophore cells or foraminifera or pteropods (> 1OOpm diameter). For comparison, Fig. 7 shows the size ranges
of these other CaCO, particles found in the sea.
Heretofore, it has been difficult to ascribe any significance to the m.orphology of coccoliths, but we can speculate from our results that the production of turbid conditions may be selectively advantageous to coccolithophores. Natural selection may therefore have led to calcite
scales with peak light scattering per unit mass. We previously hypothesized (Balch et al. 199 1) that the shedding
of highly scattering plates into the water by coccolithophores increases the effective pathlength of light within
surface waters. and thereby increases the probability of
photon capture by the coccolithophores and decreasesthe
probability of photon capture for deeper species of the
chlorophyll maximum. Our results here further support
this hypothesis.
The results presented in this paper also aid attempts
to estimate the #concentrationof calcite coccoliths by means
of remote sensing. Water-leaving radiance (L,) and reflectance are slrongly related (Gordon and Morel 1983),
and reflectance is a function of bJ(a + bb) (Gordon et al.
1988). Thus, &/(a + bb) is a good proxy for L,. We
therefore plotted bJ(a + bb) at 546 nm vs. bJ(a + bb) at
436 nm and ccntoured isopleths of coccolith density and
Chl a (Fig. 8). The results showed that the isopleths of
coccolith concentration were almost horizontal, whereas
the chlorophyll isopleths ran more diagonally. In other
words, calcite principally drove the bb/(a + bb) at 546 nm
through scattering effects, and chlorophyll affected both
b,l(a + bb) at 4.36 nm and b,l(a + bb) at 546 nm through
absorption and scattering effects. H. R. Gordon (pcrs.
comm.) has modeled the impact of coccoliths and chlorophyll on L, L36and L, 546at varying coccolith concentrations and has shown similar behavior. This information is important for the correction of remotely sensed
images of phytoplankton pigments that are contaminated
by high calcite, abundance. Conceptually, L, 546can be
used to derive an estimate of coccolith abundance (accurate to - 25,000 coccoliths ml-l), whereas both L, 436
and Lw 546 arc used to calculate pigment concentration
Coccolithophore bio-optics
(accurate to + 30%). The advent of simultaneous retrieval
of surface pigments and suspended calcite with satellite
remote sensing will allow a new level of understanding
of the cycling of organic and inorganic carbon in the sea.
References
AAS, E. 198 1. The refractive
index of phytoplankton.
Univ.
Oslo. Inst. Rep. Ser. 46.
1984. Influence of shape and structure on light scat-.
tcring by marine particles. Univ. Oslo Inst. Rep. Ser. 53.
ACIUESON,~., W.M. BALCH,ANDP. M. HOLLIGAN. 1994. The
response of water-leaving radiance to particulate calcite and
pigment concentration: A model for Gulf of Maine coccolithophore blooms. J. Gcophys. Rcs. 99: 7483-7499.
BALCH, W. M., P. M. HOLLIGAN, S. G. ACKLESON,AND K. J.
Voss. 199 1. Biological and optical properties of mesoscale
coccolithophore
blooms in the Gulf of Maine. Limnol.
Oceanogr. 36: 629-643.
BALCH,W. M., K. A. KILPATRICK,AND P.M. HOLLIGAN. 1993.
Coccolith formation and dctachmcnt by Emiliania huxleyi
(Prymnesiophyceae).
J. Phycol. 29: 5 6 6-5 7 5.
-,
AND C. R. TREES. 1996. The 199 1 coccolithophore bloom in the central North Atlantic 1. Optical properties and factors affecting their distribution.
Limnol.
Oceanogr. 41: 1669-l 683.
BERGER, W. H. 1976. Biogcnous deep sea sediments: Production, preservation and interpretation,
p. 266-326. In J. P.
Riley and R. Chester [eds.], Chemical oceanography. V. 5.
Academic.
BORSTAD,G. A.,J.F.R.
GOWR,AND E.J. CARPENTER.1992.
Development of algorithms for remote sensing of marine
Trichodesmium blooms, p. 193-210. In E. J. Carpenter
Trichodesmium and
[ed.], Marine pelagic cyanobacteria:
other diazotrophs. Kluwer.
BRATBAK,G., J.K. EGGE,AND M. HELDAL. 1993. Viralmortality of the marine alga Emiliania huxleyi (Haptophyceae)
and termination of algal blooms. Mar. Ecol. Prog. Ser. 93:
39-48.
BRICAUD,A., J. R. V. ZANEVELD,AND J. C. KITCHEN. 1992.
Backscattering efficiency of coccolithophorids:
USC of a
three-layered sphere model, p. 27-33. In Ocean Optics 11.
SPIE 1750.
BROWN,C., AND J. YODER. 1993. Distribution pattern of coccolithophorid
blooms in the western North Atlantic. Cont.
Shelf Res. 14: 175-198.
EPPLEY,R.W.,R.W.
HOLMES,ANDJ.D.H. STRICKLAND. 1967.
Sinking rates of marinc phytoplankton
measured with a
fluorometer. J. Exp. Mar. Biol. Ecol. 1: 191-208.
FERNANDEZ,E.,P. BOYD, P.M. HOLLIGAN,ANDD.S. HARBOUR.
1993. Production of organic and inorganic carbon within
a large scale coccolithophore
bloom in the North Atlantic
Ocean. Mar. Ecol. Prog. Ser. 97: 271-285.
GORDON,H. R. 1987. A bio-optical model describing the distribution of irradiance at the sea surface resulting from a
point source embedded in the ocean. Appl Opt. 26: 41334148.
-.
1992. Diffuse rcflectancc of the ocean: Influence of
nonuniform phytoplankton
pigment profile. Appl. Opt. 31:
2116-2129.
AND OTHERS. 1988. A semianalytic radiance model of
ocean color. J. Geophys. Res. 93: 10,909-10,924.
-,
AND A. Y. MOREL. 1983. Rcmotc assessment of ocean
1695
color for interpretation
of satcllitc visible imagery-a
review. Springer.
-,
R. C. SMITH, AND J. R. V. ZANEVELD. 1980. Introduction to ocean optics, p. 14-5 5. In Ocean Optics 4. SPIE
208.
HOLLIGAN,P. M., ANDOTHERS. 1993. A biogeochemical study
k’miliania huxleyi, in the North
of the coccolithophore,
Atlantic. Global Biogeochem. Cycles 7: 879-900.
-,
AND OTHERS. 1984. Vertical distribution and partitioning of organic carbon in mixed, frontal and stratified
waters of the English Channel. Mar. Ecol. Prog. Ser. 14:
111-127.
-,
M. VIOLLIER,D.S. HARBOUR,P. CAMUS,ANDM.&AMPAGNE-PHILIPPE.1983. Satellite and ship studies of coccolithophore production along a continental shelf edge. Nature 304: 339-342.
JERLOV,N. G. 1976. Marinc optics. Elscvicr.
KIRK, J. T. 0. 198 la. Monte carlo study of the nature of the
underwater light field in, and the relationships between optical properties of, turbid yellow waters. Aust. J. Mar. Freshwater Res. 32: 5 17-532.
198 1b. Estimation of the scattering cocffrcient of nat-.
ural waters using underwater irradiance mcasuremcnts. Aust.
J. Mar. Freshwater Res. 32: 533-539.
199 1. Volume scattering function, average cosines,
p
and the underwater light field. Limnol. Oceanogr. 36: 455467.
LEWIS,M. R., 0. ULLOA, AND T. PLATT. 1988. Photosynthetic
action, absorption, and quantum yield spectra for a natural
population of Oscillatoria in the North Atlantic. Limnol.
Oceanogr. 33: 92-98.
LINSCHOOTEN,C., AND OTHERS. 1991. Role of the light-dark
cycle and medium composition on the production of coccoliths by Emiliania huxleyi (Haptophyceae). J. Phycol. 27:
82-86.
MACISAAC,J. J., AND R. C. DUGDALE. 1972. Interactions of
light and inorganic nitrogen in controlling nitrogen uptake
in the sea. Deep-Sea Res. 19: 209-232.
MARGALEF,k. 1979. Lift-forms of phytoplankton as survival
alternatives in an unstable environment. Oceanol. Acta 1:
493-509.
MOREL,A. 1987. Chlorophyll-specific
scattering coefficient of
phytoplankton, a simplified theoretical approach. Deep-Sea
Res. 34: 1093-l 105.
AND A. BRICAUD. 1980. Theoretical results concerning
the optics of phytoplankton,
with special reference to remote sensing applications, p. 3 13-32 1. In J. F. R. Gower
[ed.], Oceanography from space. Plenum.
PAASCHE,E. 1962. Coccolith formation. Nature 193: 10941095.
. 1963. The adaptation of the carbon- 14 method for
the measurement of coccolith production in Coccolithus
huxleyi. Physiol. Plant. 16: 186-200.
PETZOLD,T. J. 1972. Volume scattering functions for selected
ocean waters. Scripps Inst. Oceanogr. Ref. 72-78.
PINGREE,R.D.,P.R. PUGH,P.M. HOLLIGAN,ANDG.R. FORSTER.
1975. Summer phytoplankton
blooms and red tides along
tidal fronts in the approaches to the English Channel. Nature 258: 672-677.
of the North Pacific
REID, F. M. H. 1980. Coccolithophorids
Central Gyre with notes on their vertical and seasonal distribution. Micropaleontology
26: 15 1- 17 6.
ROBINSON,J. E., AND OTHERS. 1994. The impact of a coccolithophore bloom on oceanic carbon uptake in the N.E.
Atlantic during summer 199 1. Deep-Sea Res. 41: 297-3 14.
SEIBOLD,E., ANDW. H. BERGER. 1982. The sea floor. Springer.
1696
Balch et al.
VAN DE HULST,H. C. 198 1. Light scattering by small particles.
Dover.
WINTER, A., AND W. G. SIESSER. 1994. Coccolithophores.
Cambridge.
YODER,J. A.,S.G. ACKLESON,R.T. BARBER,P. FLAMENT,AND
W. M. BALCH. 1994. A line in the sea. Nature 371: 689692.
Submitted: 16 February 1995
Accepted: 16 February 1996
Amended: 19 July 1996