Upload
ecnu
View
0
Download
0
Embed Size (px)
Citation preview
Estimation of phytoplankton pigment concentration in the Gulf ofAqaba (Eilat) by in situ and remote sensing single-wavelengthalgorithms
L. SOKOLETSKY{, Z. DUBINSKY{, M. SHOSHANY{ and
N. STAMBLER{{Faculty of Life Sciences, Bar-Ilan University, Ramat-Gan 52900, Israel;e-mail: [email protected]{Department of Geography, Bar-Ilan University, Ramat-Gan 52900, Israel
(Received 19 April 2001; in final form 25 November 2002 )
Abstract. Bio-optical relationships between inherent and apparent opticalproperties, and between optical properties and phytoplankton pigmentconcentration (C) averaged in a discrete layer, were developed. Theserelationships were derived from analysis of data collected during theperiod 1996–1998 in the Gulf of Aqaba (Eilat), a ‘Case 1’ type waterbody. Parameterization of these relationships was accomplished by com-bining Gershun’s equation, radiative transfer theory for average cosine ofunderwater light field, and a set of different bio-optical models. An analysisof the asymptotic light field was carried out. Semi-analytical single-wavelength(at l~443 nm) algorithms for in situ and remote sensing (RS) estimation ofmean pigment concentration were developed, and evaluated by sensitivity anderror analysis. The advantages of RS single-wavelength algorithms incomparison with current two- and multi-wavelengths RS algorithms arediscussed.
1. Introduction
Estimation of phytoplankton in marine and freshwater bodies is an important
goal in ecosystem research and monitoring, and environmental quality control. In
recent years, due to the availability of new airborne video cameras and satellite
sensors such as SeaWiFS or MODIS, new remote sensing (RS) algorithms were
developed, allowing estimation of phytoplankton pigment concentration (C) (e.g.
Gordon et al. 1988, Morel 1988, Hoge 1994, Morel and Gentili 1996, Gitelson et al.
1996, Fraser et al. 1997, O’Reilly et al. 1998, Avard et al. 2000, Gross et al. 2000).
For this purpose C is considered as the sum of chlorophyll a (Chl a) and degraded
algal pigments, pheophytins (pheo). These algorithms are typically based on four
radiation characterizations: (a) spectral water-leaving radiance Lw (l); (b)
normalised spectral water-leaving radiance: Lwn (l)~Lw(l)/[(cos h0)ta(l)], where
h0 is solar zenith angle, and ta(l) is spectral transmittance of the atmosphere; (c)
International Journal of Remote SensingISSN 0143-1161 print/ISSN 1366-5901 online # 2003 Taylor & Francis Ltd
http://www.tandf.co.uk/journalsDOI: 10.1080/0143116031000069807
INT. J. REMOTE SENSING, 20 DECEMBER, 2003,
VOL. 24, NO. 24, 5049–5073
above-water remote-sensed reflectance Rrs (l)~Lw(l)/Ed(l, 0z), where Ed(l, 0z)
is downwelling irradiance just above sea surface; or (d) underwater remote-sensed
reflectance Rrsw (l)~Lu(l, 02)/Ed(l, 02), where Lu(l, 02) and Ed(l, 02) are
upwelling radiance and downwelling irradiance just below sea surface, respectively.
Most of the existing algorithms have power or polynomial form, utilizing these
characteristics at different wavelength bands as independent variables. Empirical
data for the verification of these algorithms were collected from basin-specific
seasonal measurements or from generalization of regional data sets. In both cases a
delicate approach to utilization of existing algorithms for other water basins and
seasons is necessary.Generalisation of in situ and RS models deriving phytoplankton concen-
tration for water bodies under a wide range of geographical conditions is
of primary importance. The aim of this study was to simplify these models by
utilising a single wavelength band, which could potentially contribute to their
geographical generalisation. The spectral band we selected was the 443 nm
band, which was found to be the most suitable for this purpose by numerous
investigators (e.g. Gordon et al. 1988, Morel 1988, Tilzer et al. 1994, Bricaud et al.
1995, Waters 1995, Garver and Siegel 1997, Stramska and Dickey 1998, Berwald
et al. 1998, Antoine and Morel 1999). Use of the blue range of the spectrum, in
addition to other optical ranges for in situ estimation of pigment biomass, has been
discussed by Gordon and Morel (1983), Smith et al. (1991) and Bartlett et al.
(1998). Garver and Siegel (1997), using only one wavelength (441 nm), obtained
excellent correspondence between Chl a measured in situ and Chl a retrieved
from Rrsw.
The main goal of the present work is to establish new bio-optical relationships
based on state-of-the-art underwater investigations, and to parameterise such
relationships for the Gulf of Aqaba (Eilat). From a practical point of view, the
relationships developed have been used for the solution of the inverse bio-optical
problem, namely, estimation of layer-averaged pigment concentration.
2. Study area
According to most ‘chlorophyllous definitions’ of the trophic state of water
basins (e.g. Shifrin 1988, Morel and Berthon 1989, Dera 1995, Antoine et al. 1996,
Vinogradov et al. 1997), the waters of the Gulf of Aqaba exhibit meso-oligotrophic
rather than oligotrophic-type characteristics (Levanon-Spanier et al. 1979). The
primary productivity rates measured in the Gulf of Aqaba (Levanon-Spanier et al.
1979, Lindell and Post 1995, Iluz 1998) also indicate the seasonal meso-
oligotrophicity of the Gulf.
In the past only a limited number of phytoplankton concentration studies have
been conducted in the Gulf of Aqaba, and these have been limited to single dates
and locations (Levanon-Spanier et al. 1979, Dubinsky et al. 1990, Stambler 1992,
Lindell and Post 1995, Iluz 1998, Badran and Foster 1998). This does not allow
development of statistically reliable bio-optical algorithms for the region under
consideration. The present study is based on an extensive oceanographic survey
consisting of about 70 cruises that were conducted in the Gulf between January
1996 and December 1998.
The chosen site for the measurement of optical, hydro-physical and other
5050 L. Sokoletsky et al.
characteristics was station A1 (29‡ 28’N, 34‡ 56’E), which is situated at the northern
tip of the Gulf about 5 km off the coast. Bottom depth at this station is
approximately 700 m. This station was determined as a good representative of the
offshore waters in the Gulf.
Trophic conditions in the region under consideration were estimated according
to pigment concentration averaged in the penetration and euphotic layers (DZp and
DZe, respectively). These layers are defined as layers from the just below sea surface
(Z~0–) to the depth at which photosynthetically active radiation, EPAR (e.g.
downwelling irradiance averaged within the spectral range from 400 to 700 nm) is
reduced to 1/e (Z~Zp) or 1% (Z~Ze) of its sub-surface value, respectively. Thus,
trophic conditions were classified as oligotrophic during the May–October stratified
period (when observed penetration-layer averaged concentrations vCpw were
<0.12–0.25 mg m{3 and euphotic-layer averaged concentrations vCew were
<0.30–0.46 mg m{3), and mesotrophic during the November–April mixing period
(vCpw <0.33–0.72 mg m{3 and vCew <0.37–0.68 mg m{3). Concomitantly, Zp
and Ze varied from 13.6 and 73.7 m, respectively on 4 March 1996 up to 28.9 and
114.2 m, respectively on 15 June 1998. These observations and classification are in
agreement with the early observations of Levanon-Spanier et al. (1979).
3. Data acquisition and pre-processingData of Chl a and pheo concentrations were derived from in situ measurements
conducted at depths Z from 0– to y600 m, and radiometry data — from 0– to
y100–120 m, in accordance with the depth of the euphotic layer. The number of
vertical profiles of C was 52. Samples were taken usually every 20 metres and after
that measurements were repeated several times.
The fresh samples were analysed within a few hours after collection.
Determination of C concentrations in the water column was performed by the
standard fluorometric method (Holm-Hansen et al. 1965, Schanz et al. 1997).
Statistical analysis of repeated concentration measurements give a mean coefficient
of variance CV~15.8% for Chl a for initial dataset (N~45) and CV~9.8% for a
dataset remaining after rejecting obviously erroneous measurements (N~24). With
the aim of diminishing of stochastic (space and temporal) error and estimation of
intermediate values of concentrations, quadratic or cubic smoothing of raw data
were applied.
The spectra of the downwelling irradiance Ed(l, Z), and upwelling radiance
Lu(l, Z) were acquired with a submersible Profiling Reflectance Radiometer (Model
PRR-600 from Biospherical Instruments Ltd) at seven spectral bands (412, 443,
490, 510, 555, 665, and 694 nm) with 10-nm bandwidths. All spectra were measured
every several centimetres and in two directions (down and up). The least significant
bit that the spectroradiometer could resolve corresponded to a spectral irradiance
value of <0.01 mW m{2 cm{1. Initial number of measured optical profiles
(including repeated measurements) was 56. However, after rejection of erroneous
measurements and incomplete optical or concentration data, 24 pairs (concen-
trationszoptics) of vertical profiles and 30 pairs of bio-optical data just below air-
sea interface (Z~0–), were saved for further data processing.For more exact assessment of the downwelling irradiance just below sea surface
Ed(l, 0–), additional meteorological data of incoming global solar irradiance at
Estimation of phytoplankton pigment concentration 5051
Eilat during 1990–1998 and estimates obtained from the regionally-tuned algorithm
at l~443 nm (Sokoletsky et al. 2000), were used. In order to reduce the effects of
noise in the data, the third order polynomials were used (henceforth, the symbol l
will be omitted):
In Ed Zð Þ½ �~{ a0za1Zza2Z2za3Z
3� �
, ð1Þwhere coefficients a0, a1, a2, and a3 were determined by comparison of mea-
sured values of ln[Ed(Z)] to their model values (simultaneously within all range of
depths) by means of non-linear least-squares method (NLSM). Equation 1 was
found to be significant in most cases (the coefficient of determination R2 was
greater than 0.99). A similar approach was used for smoothing of Lu(l, Z) data.
The vertical attenuation coefficient Kd (Z) for downwelling irradiance, which is
determined by
Kd Zð Þ~{Lln Ed Zð Þ½ �
LZð2Þ
was then derived from equations 1 and 2 as follows:
Kd Zð Þ~a1z2a2Zz3a3Z2: ð3Þ
4. Average cosine of the underwater light field: RTT implementation
The average cosine of underwater light field (henceforth, ‘average cosine’) �mm is
an important geometrical characteristic of the underwater light regime, and serves
as a factor relating an inherent optical property (IOP), a, with the apparent optical
property (AOP), Kd (Plass et al. 1981, Stavn 1988, Kirk 1994, McCormic 1995,
Pelevin and Rostovtseva 2001) as follows:
a~�mmKd: ð4Þ
Equation 4 can be considered as a form of Gershun’s equation (Gershun 1939),
and was tested by Plass et al. (1981), who showed that the mean accuracy of this
equation is in the range of 0.5%, and the maximum deviation for clear oceanic
waters is 1.8%. Later processing of numerous experimental ocean data by Pelevin
and Rostovtseva (2001) slightly increased the boundary of accuracy to 4%.
In turn, �mm (at given vertical depth Z) can be represented in the approximate
form (Berwald et al. 1995):
�mm~�mm?z �mm0{�mm?ð Þ exp {PzZð Þ, ð5Þwhere �mm0 and �mm? are average cosines at Z~0– and Z~‘ (i.e. at infinitely-thick
oceanic layer), respectively; Pz is defined from equation 5 as the rate (at any vertical
depth Z) of exponential decrease of the quantity �mm{�mm?ð Þ= �mm0{�mm?ð Þ. Taking into
account the independence of �mm0 and �mm? of the depth, Pz has a physical meaning as
the rate of decrease of the average cosine �mm.
It is reasonable to assume that if the sun is in the zenith, then �mm0 and �mm? are
IOPs, i.e. optical properties independent of the ambient light field. Indeed, Berwald
et al. (1995), by means of HYDROLIGHT radiative transfer numerical model and
following least-squares fitting, were able to get equations relating these properties to
5052 L. Sokoletsky et al.
another IOP, single-scattering albedo v0:
�mm0~mw 0:000421v0
1{v0
� �2
{0:0274v0
1{v0
� �z1
" #, ð6Þ
�mm?~mw {1:59 v0ð Þ4z1:71 v0ð Þ3
{0:467 v0ð Þ2{0:347v0z1
h i, ð7Þ
v0~b= azbð Þ, ð8Þwhere mw is cosine of solar zenith angle at Z~0–; b is a beam scattering coefficient.
Pz is often taken as either a constant (McCormick 1985, Zaneveld 1989) or a
weak function of v0 (Berwald et al. 1995), however, authors of the latter paper
noted a significant deviation of real behaviour Pz within the first few optical depths
from such representations. In the recent work of Kirk (1999), by straightforward
application of radiative transfer theory, expression for the rate of exponential
decreasing (Pd) of �mm with the average distance (d), traversed by photons, has been
derived:
Pd~b 1{�mmsð Þ, ð9Þwhere �mms is an average cosine of single scattering (asymmetry parameter). For
conversion from Pd to Pz, we substitute d by Z=�mmd (e.g. Kirk 1991), where �mmd is the
downwelling average cosine of underwater light field, obtaining the expression for
Pz:
Pz~b 1{�mmsð Þ
�mmd
: ð10Þ
The relation between �mmd and �mm is described by the equation (Aas 1987, Morel and
Gentili 1991):
�mmd~�mm 1z2Rð Þ
1{R, ð11Þ
where R is reflectance, defined as the ratio of upwelling irradiance Eu to
downwelling irradiance Ed. Returning to equation 5 and taking into account
equations 10 and 11, we get a implicit expression for �mm:
�mm~�mm?z �mm0{�mm?ð Þ exp {bZ 1{�mmsð Þ 1{Rð Þ
�mm 1z2Rð Þ
� �: ð12Þ
Solution of this equation can be obtained in explicit form by sequential substitution
of �mm into equation 12; by doing so we get solution in infinite form:
�mm~�mm?z �mm0{�mm?ð Þ exp {q
�mm?z �mm0{�mm?ð Þ exp { qP
� �" #
, ð13Þ
where q is certain generalised parameter, equalling
q~bZ 1{�mmsð Þ 1{Rð Þ
1z2R: ð14Þ
Now we can derive some important relationships among the optical properties
of the underwater light field �mm Zð Þ, Pz(Z) and Kd(Z) themselves. Indeed, as far as the
optical properties a, b, R and �mms are concerned, these are only C-dependent and
Z-independent. Therefore, from equations 4, 10 and 11, for any two depths Z1 and
Estimation of phytoplankton pigment concentration 5053
Z2, follows the similarity relationship:
Pz Z1ð ÞPz Z2ð Þ~
�mm Z2ð Þ�mm Z1ð Þ~
Kd Z1ð ÞKd Z2ð Þ , ð15aÞ
or in its differential form:
LlnPz Zð ÞLZ
��������~ L�mm Zð Þ
LZ
��������~ LlnKd Zð Þ
LZ
��������: ð15bÞ
Figure 1 illustrates the computation [based on equations (6), (7) and (13)] of the
normalised average cosine �mm=mw as a function of q and v0.
The form of the relationship between �mm=mw and q (at v0~const) is similar to the
forms of relationships �mm=mw vs. Z or vs. optical depth t~Zc, represented in several
earlier papers (e.g. Preisendorfer 1959, Zaneveld 1989, Sathyendranath and Platt
1991, Gordon et al. 1993, Berwald et al. 1995, 1998, McCormick 1995). At first
�mm=mw strongly decreases as the parameter q increases; however, with further increase
of q, it reaches an almost constant value, i.e. the underwater light field becomes
asymptotic. A more detailed study of the asymptotic light field will be undertaken
below, in a separate section.
Note, that recent experimental data (IOPs, AOPs and vertical profiles of C)
gathered by Sosik et al. (1998) in northwest Atlantic coastal waters, revealed a clear
vertical increase of Kd(443) even in the case of oligotrophic (C v0.5 mg m{3) and
homogeneous waters.
Objective analysis of above equations and general background (see e.g. Kirk
1994) show that the underwater light regime is governed by various components
Figure 1. Vertical profiles (along axis q) of the average cosine of underwater light field atzenith sun �mm computed for the set of single-scattering albedo v0 (thin curves) andpercentage criterion for asymptotic property x (thick curves). v0 accepts values of0.2, 0.3, …, 0.9, 0.95 (on the right to the left); x increases from 5% on the bottom to50% on the top by 5% steps.
5054 L. Sokoletsky et al.
of the complex atmosphere–water column system (including air/water interface),
primarily by:
1. sun position, affecting mw and R;2. atmospheric conditions (cloudiness, pollution, winds, etc) affecting mw;
3. vertical depth Z, affecting Pz and �mm;
4. concentration of solid and dissolved matter in the water, affecting all IOPs
and AOPs, including �mm0, �mm?, �mms and �mm.
5. Bio-optical modelling
The following initial bio-optical relationships (all taken at wavelength of
443 nm) were used in the present work:
. Total absorption a (m{1) vs. C (e.g. Prieur and Sathyendranath 1981, Morel
1991, Bricaud et al. 1995, Ciotti et al. 1999):
a~aw, diszaaCba, ð16Þ
where aw;dis is absorption coefficient for water and dissolved matter, aa and ba
are positive coefficients;. Total backscattering bb (m{1) vs. C (e.g. Gordon et al. 1988, Sathyendranath
and Platt 1988):
bb~bbwzabbCbbb, ð17Þ
where bbw is the water backscattering coefficient, given as 0.00239 m{1 by
Morel (1974), abb and bbb are positive coefficients;
. The ratio of particle backscattering (bbp~bb2bbw) to particle scattering
(bp~b2bw) (Ulloa et al. 1994, Sathyendranath et al. 2001) is:
bbp
bp~0:0078{0:0042 log10 C, ð18Þ
where bw is the water scattering coefficient, given as 0.00478 m{1 by Morel
(1974);
. Reflectance R vs. bb/a and cosine (ma) of solar zenith angle in air (Morel and
Gentili 1991) is:
R~½0:6279{0:2227gb{0:0513 gbð Þ2{
0:3119{0:2465gbð Þma�bb
a
� �,
ð19Þ
where gb is the contribution of molecular backscattering into total
backscattering, i.e.:
gb~bbw=bb; ð20Þ
. Average single-scattering cosine �mms vs. C, derived by approximation of the
data from table 1 of Kirk (1991) (see also Walker 1994, p. 58, Gordon and
Table 1. Values of IOP parameters, estimated at l~443 nm.
aw;dis (m{1) aa ba abb bbb
0.0186 0.0297 0.788 0.00175 0.253
Estimation of phytoplankton pigment concentration 5055
Boynton 1997, Boynton and Gordon 2000) is:
�mms~0:975 exp {2:594bb=bð Þ; ð21Þ. Underwater remote-sensed reflectance Rrsw (sr{1) vs. bb/a (Kirk 1994):
Rrsw~0:083bb=a: ð22Þ
It should be noted that equations 16–22 illustrate the assumption about the
depth-independency of a, b, bb, R, Rrsw and �mms.
6. Parameterization of bio-optical equations
Attempts to estimate the parameters of the equations listed above by solving
them for a large range of the depths lead to absurd results. Such a situation is
characteristic for the solution of large non-linear systems of equations. For this
reason, we used a special computer algorithm, which can be divided into three
steps:
1. Values a and bb at l~443 nm and Z~0–) were computed from measured
values of Kd, Rrsw, C and mw (‘pseudodata’). The last parameter was
computed by Snell’s law. The computation was executed by exact solution of
the system of equations 4, 6, 8, 18 and 22 for each sample (N~30).
2. Pseudodata obtained from (1) were approached by means of three-
parametric power bio-optical models (equations 16 and 17) by NLSM.3. The remaining inherent optical properties (b, c, gb, v0, �mms, �mm0, and �mm?), the
quasi-inherent optical property (Rrsw) and the apparent optical properties (R,
�mm? Pz, �mm, �mmd, and Kd) were computed by equations listed in Sections 4 and 5.
The input parameters for modelling (C and Z) were taken from the following
ranges, near to real measurement conditions: (0ƒCƒ1 mg m{3 and
0ƒZƒ100 m).
As can be seen from table 1 and figures 2–4, the bio-optical models (at
l~443 nm) found are close to ones encountered in literature, describing Case 1
waters. For instance, an a vs. C model close to ours can be found in Morel (1988),
Cleveland (1995), Lee et al. (1998); b vs. C in Gordon et al. (1993) and Haltrin
(1999); bb vs. C in Sathyendranath and Platt (1988) and Morel and Maritorena
(2001); v0 vs. C in Berwald et al. (1995), Ciotti et al. (1999) and Morel and
Maritorena (2001); Kd(0–) vs. C in Platt et al. (1994). It should be noted, however,
that our models at Cv0.1 mg m{3 are expected to be less accurate than at
0.1ƒCƒ1 mg m{3 since they are based mainly on data within the second, higher
range of pigment concentrations.
The total accuracy of models found was estimated from comparison of
computed optical properties with directly-observed optical properties [Kd(0–) and
Rrsw], or with optical pseudodata (a and bb); it was represented (table 2) by values
of normalised (to average values of the ‘true’ optical property) root-mean-squared
error (NRMSE), coefficient of determination (R2) and significance level (p).
Developed and parameterised above, bio-optical models define characteristics of
the underwater light field R, q, Pz, �mm=mw and mwKd as functions of ma, C and Z.
However, in contrast to C and Z, impact of the sun position on these characteristics
is not as large. We verified this impact by doing all computations for the two
5056 L. Sokoletsky et al.
extreme values of the cosine (ma) of solar zenith angle, namely, ma~0.6 (this is the
minimal value for the entire period of observations) and for ma~1 (its maximal
theoretical value) for the chosen range of input parameters. The normalised
(to the average values of selected optical property, computed for both values of ma)
root-mean-squared differences (NRMSD) estimated for �mm=mw and mwKd are
less than 0.3% (table 3) and, therefore, these important optical characteristics may
be assumed to be independent of the sun’s position over a broad range of its
variations.
The generalised parameter q is a quantity proportional to the depth Z
(equation 14), with a slope increasing with pigment concentration (figure 5). This is
Figure 2. Relationships (the main model) between different optical properties and totalpigment concentration. Note that Kd(0–) and R(0–) are computed for zenith sun.
Estimation of phytoplankton pigment concentration 5057
Figure 3. Relationships (observed data, the main and simplified models) between Kd(0–) atsun in zenith and sub-surface pigment concentration C0.
Figure 4. As figure 3, but for water-leaving radiance Rrsw.
Table 2. Accuracy of bio-optical models at l~443 nm and Z~0– (N~30).
Optical property NRMSE (%) R2 p
a 11.7 0.793 4.41610{11
bb 12.2 0.243 5.65610{3
Kd 11.4 0.813 1.05610{11
Rrsw 16.8 0.443 5.98610{5
5058 L. Sokoletsky et al.
explained due to the larger positive contribution of beam scattering b than that of
the negative contribution of the factor (1{�mms) to parameter q.
Vertical profiles of Pz, �mm and Kd (at zenith sun) plotted for the C values in the
chosen range, are shown in figures 6–8, respectively.
7. An asymptotic light field
Below sufficient depth in a homogeneous ocean, the angular shape of the
radiance distribution and the rates of decay of the magnitude of radiances with
depth are constant and depend only on the IOP of the water. This is referred to as
the asymptotic light field (e.g. Preisendorfer 1959, Berwald et al. 1998). Taking into
account that Pz, �mm and Kd vary with depth in the same manner (see the similarity
relationship in equation 15), we can define an asymptotic light field more generally
as that range of depths at which the magnitudes of Pz, �mm and Kd become constants,
depending only on the IOP of the water.The criterion for asymptotic property (x) giving the upper boundary of the
asymptotic light field (Zx) was chosen analogously to the criterion applied by
Table 3. Normalized root-mean-squared difference (NRMSD, in %) for characteristics ofthe underwater light field computed between ma~0.6 and ma~1.
R q Pz �mm=mw mwKd
16.7 2.3 0.25 0.22 0.25
Figure 5. Vertical profiles (along axis of depths) of generalized parameter q (thin curves)computed for C~0, 0.01, 0.05, 0.1, 0.2, 0.3, 0.4, 0.5, 0.7 and 1.0 mg m{3 on the leftto the right. x (thick curves) increases from 5% on the bottom to 40% on the top by5% steps.
Estimation of phytoplankton pigment concentration 5059
Gordon et al. (1993), namely,
Pz ?ð Þ{Pz Zxð ÞPz ?ð Þ ~
�mm Zxð Þ{�mm?�mm?
~Kd ?ð Þ{Kd Zxð Þ
Kd ?ð Þ vx
for all Z > Zx:
ð23Þ
Similar expressions can be written for qxwq(Zx).
Figure 7. Vertical profiles (along axis of depths) of the average cosine �mm of underwater lightfield (at zenith sun), computed for C~0, 0.01, 0.05, 0.1, 0.2, 0.3, 0.4, 0.5, 0.7 and1.0 mg m{3 on the right to the left (thin curves). x (thick curves) varies as in figure 5.
Figure 6. Vertical profiles (along axis of depths) of the rate of exponential decrease Pz ofaverage cosine of underwater light field (thin curves), computed for C~0, 0.01, 0.05,0.1, 0.2, …,1.0 mg m{3 on the left to the right. x (thick curves) varies as in figure 5.
5060 L. Sokoletsky et al.
The plots illustrating asymptotic light ‘fields’ corresponding to different values
of x (in %) are presented in figures 1 and 5–7. These plots clearly demonstrate the
significant variability of AOPs (Pz, �mm and Kd) along the vertical axes q (figure 1) or
Z (figures 5–7). For example, a general model, based only on RTT (figure 1), yields
more than 50% variability of these AOPs as the single-scattering albedo v0 varied
over the range 0.63 to 0.83, and the generalised parameter q varied from 0 to ‘. In
more realistic cases when v0 ranges from 0.79 to 0.83 (this corresponds to the
pigment concentration C ranging from 0.1 to 1 mg m{3) and vertical depths vary
from 0 to 100 m, the above-mentioned AOPs vary from 24 to 36%, depending on C
(figures 5–7). The onset of the asymptotic light field Zx (in m) is a function of the
percentage criterion x, and pigment concentration C. However, for the following
ranges of parameters: 5%ƒxƒ35% and 0.1ƒCƒ1 mg m{3, the uncertainty due to
C is less than 14%, and Zx can be estimated from the approximated equation 24,
derived from a general in situ model (with R2~0.999):
Zx~156:5{10:1xz0:295x2{0:00355x3: ð24Þ
8. Simplified bio-optical models
Although above-presented in situ and remote sensing bio-optical models allow
the possibility of solving the direct problem (estimation of Kd and Rrsw from mw, C
and Z), the solution of the inverse problem (estimation of C from Kd (or Rrsw), mw
and Z) may, in some cases, lead to absurd results. With the aim of removing
such situations and simplifying the solution of the inverse problem, the
following simplified models were developed and used in further investigations (at
l~443 nm, 0ƒCƒ1 mg m{3 and 0ƒZƒ100 m for Kd and Z~0– for Rrsw):
Figure 8. Vertical profiles of Kd computed by two models for the following set of pigmentconcentrations C (from the left to the right): 0, 0.01, 0.05, 0.1, 0.2, …, 1.0 mg m{3.The main model is shown by the thick curves, and the simplified model by the thincurves.
Estimation of phytoplankton pigment concentration 5061
Kd Zð Þ~Kw, dis Zð ÞzaK Zð Þ C Zð Þ½ �bK , ð25Þwhere Kw;dis(Z) is downwelling attenuation coefficient for water and dissolved
matter:
Kw, dis Zð Þ~ aw, dis
mw�mmw0
� �1zk1Zzk2Z
2� �
, ð26Þ
aK(Z) is a positive coefficient:
aK Zð Þ~ aa
mw�mmw0
� �1za1Zza2Z
2� �
, ð27Þ
Rrsw~0:015{0:0241C0z0:0290 C0ð Þ2{0:0125 C0ð Þ3
, ð28Þwhere bK was assumed equal to ba~0.788; C0 is sub-surface pigment concentration;
�mmw0 is sub-surface average cosine �mm0 at mw~1 (i.e. at zenith sun), computed by
equations 6, 8 and 16–18 and averaged for the overall range of C. The remaining
values of parameters for our simplified in situ (Kd vs. C) model were estimated from
comparison of this model with the main in situ model by the least-squares method
(table 4).A simplified RS model (equation 28) was derived, based on observed values of
C0 and Rrsw (N~30). The plots for which both types of models (main and
simplified) were used simultaneously (figures 3 and 8 for Kd and figure 4 for Rrsw),
confirm the high accuracy of our simplified models.
9. Sensitivity and error analysis
Sensitivity analysis for the single-wavelength Kd(0–) and Rrsw vs. C models,
derived above, was conducted by using an error amplification factor H, deter-
mined as a ratio of relative error of output value (C for our application) to
relative error of input value (F) at very small values of the F error. It is easy to
show that H can be expressed by the following equation (see, for instance, Shifrin
1988, p. 212):
H C vs:F Cð Þ½ �~ F Cð ÞC LF Cð Þ=LC½ � , ð29Þ
where F(C) is presented by Kd(0–) or Rrsw.
From equations 25, 28 and 29 it follows:
H C vs:Kdð Þ~ Kd
bK Kd{Kw, disð Þ , ð30Þ
H C0 vs:Rrswð Þ~ Rrsw
C0 {0:0241z0:0579C0{0:0374 C0ð Þ2h i : ð31Þ
The plot of error amplification factor for C vs. Kd at two depths, 0– and 100 m
(figure 9) demonstrate a sharp increase of error H as C approaches 0, and slow
Table 4. Parameters of equations 24–26.
�mmw0 bk k1 (m{1) k2 (m{1) a1 (m{1) a2 (m{1)
0.888 0.788 21.51610{4 1.51610{5 1.04610{2 24.85610{5
5062 L. Sokoletsky et al.
asymptotic behavior as C tends to infinity (limit values of H is 1/bK~1.270). In
order to obtain magnitudes of errors, resulting from the seasonal variations in
pigment concentrations, dependences H as functions of mean values of observed
C(Z), averaged separately for the mixing period (November–April) and for
the stratified period (May–October), were also plotted in figure 9. As could be
expected, during a mixing period (when the pigment concentrations were
C0~0.52¡0.21 mg m{3 and C100~0.45¡0.12 mg m{3) the magnitude of H was
usually lower than during a stratified period (when C0~0.15¡0.05 mg m{3 and
C100~0.60¡0.30 mg m{3).
Values of the error amplification factor H for C0 vs. Rrsw (figure 10) are negative
for all values of C0, in the range under consideration, but their absolute values are
close to values for H(C0 vs. Kd). The relationship between H(C0 vs. Rrsw) and C0
has a more complex character than that between H(C0 vs. Kd) and C0, and is much
lower related to season. Indeed, values of H(C0 vs. Rrsw) computed for the mean
sub-surface pigment concentrations (¡ standard deviations) measured during both
periods, are {4:05z0:7{2:4 and {4:9z0:9
{1:85 for the mixing and stratified period,
respectively. Note that absolute values of H(C0 vs. Rrsw) are even somewhat lower
than those (about 6), obtained by Vasilkov et al. (2001) for Case 1 and Case 2
waters, using bio-optical simulations for the 5-wavelength RS algorithm.
If we accept that the error of retrieving Rrsw at 443 nm is within ¡5% (see e.g.
Figure 9. Error amplification factor (H) computed for the estimation of C(Z) from Kd(Z).The thick curve was obtained for Z~0– and thin curve for Z~100 m. The symbolsshow H computed for the averaged pigment concentration values measured indifferent seasonal periods; open triangles correspond to mixing period, while closecircles correspond to stratified period. The symbols are shown for the depths from0– to 100 m in 10 m steps.
Estimation of phytoplankton pigment concentration 5063
Vasilkov et al. 2001, Gordon and Wang 1994, Wang and Gordon 1994), then the
error of remote-sensed retrieving of C0 can be estimated as +20% and +24:5% for
mixing and stratified period, respectively. Therefore, it seems obvious that if the
first SeaWiFS accuracy goal of 5% in water-leaving radiances (Mueller and Austin
1992) is achievable, then the second goal of 35% accuracy of Chl a retrieval is also a
realistic minimum for the waters under consideration.
10. Relationships between pigment concentrations averaged in select layers
One of the challenging objectives in biological oceanography is estimation of
underwater layer-averaged biomass or total pigment concentration. Remote-sensing
imagery gives estimations of C or Chl a only from the near-surface layer rather
than from deeper layers. For example, as was shown by Gordon and McCluney
(1975), approximately 90% of the information contained in water-leaving radiance
originates from the near-surface (‘penetration’) layer. However, in Case 1 waters,
main biomass (expressed by chlorophyll concentration) is usually situated below, in
the layer approximately next to the ‘euphotic’ depth, Ze. Thus preliminary in situ
determination of relationships between pigment concentrations averaged in
different layers has paramount value. Such relationships were observed for
different water basins by various investigators (e.g. Morel and Berthon 1989,
Arrigo et al. 1998, Ignatiades 1998).
We assumed a simple linear regression between C0 and layer-averaged concen-
trations, vCpw (for Z varying from 0– to Zp), vC50w (for Z varying from 0– to
50 m) and vCew (for Z varying from 0– to Ze), which were parameterized based
Figure 10. Error amplification factor H for C0 vs. Rrsw. The solid curve is computed forcontinuous values of C0 from 0 to 1 mg m{3, while the symbols show H computed forthe mean values of C0 (¡ standard deviations) observed during two differentseasonal periods.
5064 L. Sokoletsky et al.
on in situ measurements as follows (figure 11):
vCp > ~0:895C0z0:061 R2~0:972� �
, ð32Þ
vC50 > ~0:687C0z0:172 R2~0:856� �
, ð33Þ
vCe > ~0:474C0z0:290 R2~0:725� �
: ð34ÞThus, equations 32–34 along with equations of simplified models (Section 8) can
be utilized for in situ or remote sensing single-wavelength estimation of mean
phytoplankton pigment concentration in select layers.
11. In situ/RS estimation of layer-averaged pigment concentrations
The in situ and RS versions of bio-optical relationships represented above allow
development of an algorithm for estimation of layer-averaged pigment concentra-
tions from the measured quantities of mwKd(443) or Rrsw(443). It would be
reasonable to begin from estimation of sub-surface pigment concentration C0, and
then use equations 32–34 to estimate vCDZw.
An analytical solution for C0 estimated in situ was found by inversion of
equation 25 taking into account equations 26–27 at Z~0– as follows:
C0~mw�mmw0Kd 0{ð Þ{aw, dis
aa
� �b{1K
%
0, if mwKd 0{ð Þ¡0:0209 m{1
74:8 mwKd 0{ð Þ{0:0209½ �1:270, otherwise
( ð35Þ
Analogously, an analytical solution for C0 estimated by the RS algorithm was
Figure 11. Relationships between sub-surface pigment concentration C0 and layer-averagedpigment concentrations vCDZw. The observed data are shown by symbols: solid,dotted and dot-and-dash curves are approaches to vCpw, vC50w and vCew vs.C0 relationships, respectively.
Estimation of phytoplankton pigment concentration 5065
developed by inversion of equation 26:
C0~4:440{653:1Rrswz24060 Rrswð Þ2: ð36ÞFigures 12 and 13 show modelled dependences of mwKd(443, 0–) and Rrsw(443),
respectively, as a function of C0, vCpw, vC50w and vCew. Comparison
between modelled (by both algorithms) and measured euphotic layer-averaged
pigment concentrations (figure 14 and table 5) demonstrate sufficiently high and
approximately the same accuracy of in situ and RS algorithms.
12. Discussion and conclusion
New single-wavelength (at l~443 nm) algorithms for in situ and remote sensing
estimation of layer-averaged pigment concentration are developed and parameter-
ized for the Gulf of Aqaba (Eilat). The novelty of these algorithms is joint
consideration of depth-dependent average cosine of underwater light field and semi-
analytical bio-optical relationships. The error analysis outlined in this paper
indicates that a single wavelength band centred on 443 nm may be utilised for C
estimation even in a strongly stratified water column with accuracy, which meets
current goals of the marine biological community. Due to such simplification,
developed bio-optical models could potentially contribute to geographical general-
isation of new local algorithms. Such algorithms can be used for real-time optical
measurements conducted from in situ platforms (e.g. ships, optical moorings,
drifters and profiling floats) as well as from remote platforms (e.g. aircrafts or
satellites). It is also clear that single-wavelength algorithms could be used in further
applications of bio-optical methods for water quality assessment and monitoring.
Examples of such applications include estimation of penetrating in-depth solar
irradiance, phytoplankton primary productivity or modelling of time-spatial
Figure 12. Modelling of relationship between mw Kd (0–) and the pigment concentrationsjust below the surface (C0) as well as averaged over selected layers.
5066 L. Sokoletsky et al.
Figure 14. Comparison between modelled (by means of in situ and remote sensing lgorithms)and measured values of pigment concentrations averaged over euphotic layer.
Table 5. Results of statistical verification of in situ and remote sensing algorithms foraveraged (in euphotic layer) pigment concentration; linear regression of estimatedconcentrations vs. measured concentrations is assumed.
Algorithm Slope Intercept NRMSE (%) R2 p
In situ 0.731 0.116 20.0 0.555 2.36610{6
RS 0.989 0.017 24.9 0.559 2.05610{6
Figure 13. As figure 12, but for Rrsw.
Estimation of phytoplankton pigment concentration 5067
structure of plankton community (e.g. Olesen et al. 1999, Kamenir et al. 2000,
Reynolds et al. 2001, Sathyendranath et al. 2001, Sokoletsky et al. 2001).
In spite of its own merits, several aspects of the proposed algorithms remain
questionable and require further investigation. One such question is to what degree
do the equations for average cosine, used in the present work, reflect the real
underwater light situation? An especially important point here is the impact of
absorption on the geometry of underwater light fluxes. For instance, if to use the
value of single-scattering albedo from our bio-optical modelling [v0~0.82 (¡0.01),
see figure 2(a)], then, according to the findings of Berwald et al. (1995), the
contribution of absorption to Pt~Pz/c (and, hence, to Pz) should be about 13%.
The second important question, not discussed in the present work, is the choice
and number of wavelength bands for RS estimation of chlorophyll (for Case 1
waters). The number of bands usually used in the current RS applications, from
two to four (e.g. O’Reilly et al. 1998) is not necessarily optimal. Moreover, there are
examples in which good results were obtained from a single wavelength (441 nm) in
the blue range (Garver and Siegel 1997, O’Reilly et al. 1998). The following are
additional arguments supporting the possibility of using only one wavelength:
1. Spectral remote-sensed reflectances [Rrs(l) and Rrsw(l)], with high accuracy,
proportional to spectral normalised water-leaving radiance Lwn(l) (e.g.
Gordon 1990, Clark 1997, Fraser et al. 1997, Gordon and Voss 1999, Mobley
1999, Loisel et al. 2001), the main feature of which is maximal removal of the
effects of the atmosphere and the solar zenith angle from upwelling signal.
Therefore, use of Rrs(l) and Rrsw(l) also promotes removal of these effects.
2. Remote sensing observations (McClain et al. 1998, Gordon and Voss 1999)
of Lwn(l) accompanied by ground-truth data, demonstrate the minimal error
of Lwn(l) retrieval for lv555 nm. A possible explanation of this may be the
large variability of the Raman scattering contribution to Lw(l) at l~550 nm
(from 18% for pure water to 3% for 1.0 mg Chl m{3) in comparison to
l~440 nm (from 6% for pure water to 2% for 1.0 mg Chl m{3) (Waters
1995).
3. The uncertainties of retrieving Lwn(l) and, hence, Rrs(l) or Rrsw(l) may have
different signs and there is no correlation between the sensor noise at
different wavelengths (Gordon 1990, Clark 2001). Moreover, it can be
assumed that the absolute values of spectral Rrs(l), Rrsw(l) or Lwn(l) errors
are just the same, positive and negative signs of errors equiprobable, and
spectral errors independent from one another. Then, from sensitivity analysis
of model equations 16, 17 and 22 it follows that the mean error of C
estimation from Rrs(443), Rrsw(443) or Lwn(443) would not be greater than
that from any spectral differences or ratios of these parameters. For example,
from the spectral radiance data of Wang and Gordon (1994), it follows that
for modelled range of C: 0.1ƒCƒ1 mg m{3, the mean-square error of C vs.
Lwn(443) algorithm on 3.9% and 5.5% lower than the mean-square error of C
vs. Lwn(443) – Lwn(550) and Lwn(443)/Lwn(550) algorithms, respectively.
These findings may be expanded easily to a higher number of wavelengths by
the propagation errors method. Nevertheless, conclusions about the possibility and
even preference of single-wavelength use have preliminary value only. It seems that
further experimental (particularly in different sites of the Gulf) and theoretical
5068 L. Sokoletsky et al.
investigations are necessary for verification of this conclusion, as well as for
refinement of presented one-wavelength algorithms. Undoubtedly, significant
improvements in the bio-optical models may be obtained from additional
microstructure information and particle optics.
Acknowledgments
This study was part of the Ph.D. study of the first author and was conducted
within the framework of the ‘Red-Sea Program’, a joint German, Egyptian,
Palestinian and Israeli program funded by the German Ministry of Science,
Technology and Education (BMBF). The authors cordially thank Professors
J. T. O. Kirk (Kirk Marine Optics), N. J. McCormick (University of Washington),
A. Morel (Universite Pierre et Marie Curie), R. H. Stavn (University of North
Carolina), and K. J. Voss (University of Miami) for valuable suggestions and
comments offered in the course of paper preparation. We would like to thank
anonymous reviewers who provided helpful comments on the manuscript. We also
gratefully acknowledge Ms Sharon Victor for English assistance.
ReferencesAAS, E., 1987, Two stream irradiance model for deep waters. Applied Optics, 26, 2096–2101.ANTOINE, D., and MOREL, A., 1999, A multiple scattering algorithm for atmospheric
correction of remotely sensed ocean colour (MERIS instrument): principle andimplementation for atmospheres carrying various aerosols including absorbing ones.Journal of Geophysical Research, 20, 1875–1916.
ANTOINE, D., ANDRE, J.-M., and MOREL, A., 1996, Oceanic primary production 2.Estimation at global scale from satellite (coastal zone color scanner) chlorophyll.Global Biogeochemical Cycles, 10, 57–69.
ARRIGO, K. R., ROBINSON, D. H., WORTHEN, D. L., SCHIEBER, B., and LIZOTTE, M. P.,1998, Bio-optical properties of the southwestern Ross Sea. Journal of GeophysicalResearch, 103, 21683–21695.
AVARD, M. M., SCHIEBE, F. R., and EVERITT, J. H., 2000, A potential tool for estimatingchlorophyll concentration in lakes and reservoirs. Journal of Freshwater Ecology, 15,125–133.
BADRAN, M. I., and FOSTER, P., 1998, Environmental quality of the Jordanian coastal watersof the Gulf of Aqaba, Red Sea. Aquatic Ecosystem Health and Management, 1, 75–89.
BARTLETT, J. S., ABBOTT, M. R., LETELIER, R. M., and RICHMAN, J. G., 1998, Chlorophyllconcentration estimated from irradiance measurement at fluctuating depths. In:Proceedings of Ocean Optics XIV, Kailua-Kona, November 1998. http://picasso.oce.orst.edu/ORSOO/pubs/oopaper.pdf.
BERWALD, J., STRAMSKI, D., MOBLEY, C. D., and KIEFER, D. A., 1995, Influences ofabsorption and scattering on vertical changes in the average cosine of the underwaterlight field. Limnology and Oceanography, 40, 1347–1357.
BERWALD, J., STRAMSKI, D., MOBLEY, C. D., and KIEFER, D. A., 1998, Effect of Ramanscattering on the average cosine and diffuse attenuation coefficient of irradiance inthe ocean. Limnology and Oceanography, 43, 564–576.
BOYNTON, G. C., and GORDON, H. R., 2000, Irradiance inversion algorithm for estimatingthe absorption and backscattering coefficients of natural waters: Raman-scatteringeffects. Applied Optics, 39, 3012–3022.
BRICAUD, A., BABIN, M., MOREL, A., and CLAUSTRE, H., 1995, Variability in thechlorophyll-specific absorption coefficients of natural phytoplankton: analysis andparameterization. Journal of Geophysical Research, 100, 13321–13332.
CIOTTI, A. M., CULLEN, J. J., and LEWIS, M. R., 1999, A semi-analytical model of theinfluence of phytoplankton community structure on the relationship between lightattenuation and ocean color. Journal of Geophysical Research, 104, 1559–1578.
CLARK, D. K., 1997, ATBD-MOD-18. MODIS: Bio-optical Algorithms: Case 1
Estimation of phytoplankton pigment concentration 5069
Waters. http://eospso.gsfc.nasa.gov/ftp_ATBD/REVIEW/MODIS/ATBD-MOD-18/atbd-mod-18.pdf.
CLARK, D. K., 2001, MODIS Terra-Phytoplankton Pigments. Data Quality Summary.http://modis-ocean.gsfc.nasa.gov/qa/dataqualsum/;chlorMODIS_qualsum.pdf.
CLEVELAND, J. S., 1995, Regional models for phytoplankton absorption of chlorophyll aconcentration. Journal of Geophysical Research, 100, 13333–13344.
DERA, J., 1995, Underwater irradiance as a factor affecting primary production. Dissertationand monographs. Institute of Oceanology of the Polish Academy of Sciences, Sopot,112.
DUBINSKY, Z., STAMBLER, N., BEN-ZION, M., MCCLOSKEY, L., MUSCATINE, L., andFALKOWSKI, P. G., 1990, The effect of external nutrient resources on the opticalproperties and photosynthetic efficiency of Stylophora pistillata. Proceedings of theRoyal Society of London, B239, 231–246.
FRASER, R. S., MATTOO, S., YEH, E.-N., and MCCLAIN, C. R., 1997, Algorithm foratmospheric and glint corrections of satellite measurements of ocean pigment.Journal of Geophysical Research, 102, 17107–17118.
GARVER, S. A., and SIEGEL, D. A., 1997, Inherent optical property inversion of ocean colorspectra and its biogeochemical interpretation. 1. Time series from the Sargasso Sea.Journal of Geophysical Research, 102, 18607–18625.
GERSHUN, A. A., 1939, The light field. Journal of Mathematical Physics, 18, 51–151.GITELSON, A., KARNIELI, A., GOLDMAN, N., YACOBI, Y. Z., and MAYO, M., 1996,
Chlorophyll estimation in the Southeastern Mediterranean using CZCSimages: adaptation of an algorithm and its validation. Journal of Marine Systems,9, 283–290.
GORDON, H. R., 1990, Radiometric considerations for ocean color remote sensors. AppliedOptics, 29, 3228–3236.
GORDON, H. R., and BOYNTON, G. C., 1997, A radiance-irradiance inversion algorithm forestimating the absorption and backscattering coefficients of natural waters:homogeneous waters. Applied Optics, 36, 2636–2641.
GORDON, H. R., and MCCLUNEY, W. R., 1975, Estimation of the depth of sunlightpenetration in the sea for remote sensing. Applied Optics, 14, 413–416.
GORDON, H. R., and MOREL, A., 1983, Remote assessment of ocean color for interpretationof satellite visible imagery. In Lecture Notes on Coastal and Estuarine Studies, Vol. 4,edited by R. T. Barber, N. K. Mooers, M. J. Bowman, and B. Zeitzschel (New York:Springer-Verlag), 114 pp.
GORDON, H. R., and VOSS, K. J., 1999, MODIS normalised Water-Leaving Radiance, MOD18. http://modarch.gsfc.nasa.gov/MODIS/ATBD/atbd_mod17.pdf.
GORDON, H. R., and WANG, M., 1994, Retrieval of water-leaving radiance and aerosoloptical thickness over the oceans with SeaWiFS: A preliminary algorithm. AppliedOptics, 33, 443–452.
GORDON, H. R., BROWN, O. B., EVANS, R. H., BROWN, J. W., SMITH, R. C., BAKER, K. S.,and CLARK, D. K., 1988, A semi-analytic model of ocean color. Journal ofGeophysical Research, 93, 10909–10924.
GORDON, H. R., DING, K., and GONG, W., 1993, Radiative transfer in the ocean:computations relating to the asymptotic and near-asymptotic daylight field. AppliedOptics, 32, 1606–1619.
GROSS, L., THIRIA, S., FROUIN, R., and MITCHELL, B. G., 2000, Artificial neural networksfor modeling the transfer function between marine reflectance and phytoplanktonpigment concentration. Journal of Geophysical Research, 105, 3483–3495.
HALTRIN, V. I., 1999, Chlorophyll-based model of seawater optical properties. AppliedOptics, 38, 6826–6832.
HOGE, F. E., 1994, Asymmetrical spectral curvature algorithms: oceanic-constituentssensitivities. Applied Optics, 33, 7764–7769.
HOLM-HANSEN, O., LORENZEN, C. J., HOLMES, R. W., and STRICKLAND, J. D. H., 1965,Fluorometric determination of chlorophyll. Journal du Conseil Permanent Inter-national Pour L’Exploration de la Mer, 30, 3–15.
IGNATIADES, L., 1998, The productive and optical status of the oligotrophic waters of the
5070 L. Sokoletsky et al.
Aegean Sea (Cretan Sea), Eastern Mediterranean. Journal of Plankton Research, 20,985–995.
ILUZ, D., 1998, The light field, phytoplankton pigmentation and productivity in the Gulf ofElat (in Hebrew). Ph.D. Thesis. Bar-Ilan University, Ramat-Gan, Israel.
KAMENIR, Y., BRENNER, S., DUBINSKY, Z., HAESE, C., ILUZ, D., LAZAR, B., AL-QUTOB,M., SOKOLETSKY, L., and STAMBLER, N., 2000, Time-space structure of a microbialfood web: Oligotrophic Gulf Eilat (Red Sea) simulation model. In Book of abstractsof The 7th European Marine Microbiology Symposium (EMMS). (Noordwijkerhout:Netherlands Institute of Ecology – Centre of Estuarine and Coastal Ecologyand Netherlands Institute of Sea Research), p. 38. http://www.nioz.nl/emms/abstracts.pdf.
KIRK, J. T. O., 1991, Volume scattering function, average cosines, and the underwater lightfield. Limnology and. Oceanography, 36, 455–467.
KIRK, J. T. O., 1994, Light and Photosynthesis in Aquatic Ecosystems, 2nd ed., CambridgeUniversity Press, Cambridge, 520 pp.
KIRK, J. T. O., 1999, Multiply scattering of a photon flux: implications for theintegral average cosine of the underwater light field. Applied Optics, 38,3134–3140.
LEE, Z. P., CARDER, K. L., STEWARD, R. G., PEACOCK, T. G., DAVIS, C. O., and PATCH, J.S., 1998, An empirical algorithm for light absorption by ocean water based on color.Journal of Geophysical Research, 103, 27967–27978.
LEVANON-SPANIER, I., PADAN, E., and REISS, Z., 1979, Primary production in a desert-enclosed sea – the Gulf of Elat (Aqaba), Red Sea. Deep-Sea Research, 26, 673–685.
LINDELL, D., and POST, A. F., 1995, Ultraphytoplankton succession is triggered by deepwinter mixing in the Gulf of Aqaba (Eilat), Red Sea. Limnology and Oceanography,40, 130–141.
LOISEL, H., STRAMSKI, D., MITCHELL, B. G., FELL, F., FOURNIER-SICRE, V., LEMASLE, B.,and BABIN, M., 2001, Comparison of the ocean inherent optical properties obtainedfrom measurements and inverse modeling. Applied Optics, 40, 2384–2397.
MCCLAIN, C. R., CLEAVE, M. L., FELDMAN, G. C., GREGG, W. W., HOOKER, S. B., andKURING, N., 1998, Science quality SeaWiFS data for global biosphere research. SeaTechnology, 39, 10–16.
MCCORMIC, N. J., 1995, Mathematical models for the mean cosine of irradiance and thediffuse attenuation coefficient. Limnology and Oceanography, 40, 1013–1018.
MOBLEY, C. D., 1999, Estimation of the remote-sensing reflectance from above-surfacemeasurements. Applied Optics, 38, 7442–7455.
MOREL, A., 1974, Optical properties of pure seawater. In Optical Aspects of Oceanography,edited by N. G. Jerlov and E. Steemann Nielsen, (New York: Academic Press)pp. 1–24.
MOREL, A., 1988, Optical modeling of the upper ocean in relation to its biogenous mattercontent (Case I waters). Journal of Geophysical Research, 91, 10749–10768.
MOREL, A., 1991, Light and marine photosynthesis: a spectral model with geochemical andclimatological implications. Progress in Oceanography, 26, 263–306.
MOREL, A., and BERTHON, J. F., 1989, Surface pigments, algal biomass profiles, andpotential production of the euphotic layer: relationships reinvestigated in view ofremote-sensing applications. Limnology and Oceanography, 34, 1545–1562.
MOREL, A., and GENTILI, B., 1991, Diffuse reflectance of oceanic waters: its dependence onsun angle as influenced by the molecular scattering contribution. Applied Optics, 30,4427–4438.
MOREL, A., and GENTILI, B., 1996, Diffuse reflectance of oceanic waters. III. Implication ofbidirectionality for the remote-sensing problem. Applied Optics, 35, 4850–4862.
MOREL, A., and MARITORENA, S., 2001, Bio-optical properties of oceanic waters: Areappraisal. Journal of Geophysical Research, 106, 7163–7180.
MUELLER, J. L., and AUSTIN, R. W., 1992, Ocean Optics Protocols. NASA Tech. Memo.104566, Vol. 5, S.B. Hooker and E.R. Firestone, Eds., NASA Goddard Space FlightCenter, Greenbelt, Maryland, 43 pp.
OLESEN, M., LUNDSGAARD, C., and ANDRUSHAITIS, A., 1999, Influence of nutrients and
Estimation of phytoplankton pigment concentration 5071
mixing on the primary production and community respiration in the Gulf of Riga.Journal of Marine Systems, 23, 143–127.
O’REILLY, J. E., MARITORENA, S., MITCHELL, B. G., SIEGEL, D. A., CARDER, K. L.,GARVER, S. A., KAHRU, M., and MCCLAIN, C., 1998, Ocean color chlorophyllalgorithms for SeaWiFS. Journal of Geophysical Research, 103, 24937–24953.
PELEVIN, V. N., and ROSTOVTSEVA, V. V., 2001, Sea water scattering and absorptionsmodels development using the classification of ocean waters on base contact mea-surement data. Proceedings of the International Conference ‘‘Current Problemsin Optics of Natural Waters’’. ONW-2001, edited by I. Levin and G. Gilbert.St. Petersburg, 25–28 September 2001, pp. 377–382.
PLASS, G. N., HUMPHREYS, T. J., and KATTAWAR, G. W., 1981, Ocean-atmosphericinterface: its influence on radiation. Applied Optics, 20, 917–931.
PLATT, T., SATHYENDRANATH, S., WHITE III, G. N., and RAVINDRAN, P., 1994, Attenuationof visible light by phytoplankton in a vertically structured ocean: solutions andapplications. Journal of Plankton Research, 16, 1461–1487.
PREISENDORFER, R. W., 1959, On the existence of characteristic diffuse light in naturalwaters sea. Journal of Marine Research, 18, 1–9.
PRIEUR, L., and SATHYENDRANATH, S., 1981, An optical classification of coastal and oceanicwaters based on the specific spectral absorption curves of phytoplankton pigments,dissolved organic matter, and other particulate materials. Limnology and Oceano-graphy, 26, 671–689.
REYNOLDS, C. S., IRISH, A. E., and ELLIOTT, J. A., 2001, The ecological basis for simulatingphytoplankton responses to environmental change (PROTECH). Ecological Model-ling, 140, 271–291.
SATHYENDRANATH, S., and PLATT, T., 1988, The spectral irradiance field at the surface andin the interior of the ocean: a model for applications in oceanography and remotesensing. Journal of Geophysical Research, 93, 9270–9280.
SATHYENDRANATH, S., and PLATT, T., 1991, Angular distribution of the submarine lightfield: Modification by multiple scattering. Proceedings of the Royal Society of LondonSeries A-Mathematical Physical and Engineering Sciences, 433, 287–297.
SATHYENDRANATH, S., COTA, G., STUART, V., MAASS, H., and PLATT, T., 2001, Remotesensing of phytoplankton pigments: a comparison of empirical and theoreticalapproaches. International Journal of Remote Sensing, 22, 249–273.
SCHANZ, F., SENN, P., and DUBINSKY, Z., 1997, Light absorption by phytoplankton and thevertical light attenuation: ecological and physiological significance. Oceanography andMarine Biology: An Annual Review, 35, 71–95.
SHIFRIN, K. S., 1988, Physical Optics of Ocean Water. (New York: AIP Translation Series.American Institute of Physics).
SMITH, R. C., WATERS, K. J., and BAKER, K. S., 1991, Optical variability and pigmentbiomass in the Sargasso Sea as determined using deep-sea optical mooring data.Journal of Geophysical Research, 96, 8665–8686.
SOKOLETSKY, L., DUBINSKY, Z., SHOSHANY, M., and STAMBLER, N., 2000, Non-meteorological predictive models of solar flux and atmospheric transmittance atweakly-varied climatic conditions. Extended abstract (paper No. 1003 on CD-ROM)in Ocean Optics XV Conference, Monaco, 16–20 October 2000.
SOKOLETSKY, L., DUBINSKY, Z., SHOSHANY, M., and STAMBLER, N., 2001, Radiativetransfer equations, bio-optical modelling and phytoplankton pigment estimation inthe Gulf of Aqaba (Eilat). Proceedings of the International Conference ‘‘CurrentProblems in Optics of Natural Waters’’, edited by I. Levin and G. Gilbert.St. Petersburg, 25–28 September 2001, pp. 290–296.
SOSIK, H. M., GREEN, R. E., and OLSON, R. J., 1998, Optical variability in coastal waters ofthe northwest Atlantic. In Proceedings of SPIE: Ocean Optics XIV, edited by S. G.Ackleson and R. Frouin, 2963, 14. http://www.whoi.edu/science/B/sosiklab/ooxiv.htm.
STAMBLER, N., 1992, Harvesting and utilization of light by hermatypic corals (in Hebrew).Ph.D. Thesis. Bar-Ilan University, Ramat-Gan, Israel.
STAVN, R. H., 1988, Lambert-Beer law in ocean waters: optical properties of water and
5072 L. Sokoletsky et al.
of dissolved/suspended material, optical energy budgets. Applied Optics, 27,222–231.
STRAMSKA, M., and DICKEY, T. D., 1998, Short-term variability of the underwater light fieldin the oligotrophic ocean in response to surface waves and clouds. Deep-SeaResearch, 45, 1393–1410.
TILZER, M. M., GIESKES, W. W., HEUSEL, R., and FENTON, N., 1994, The impact ofphytoplankton on spectral water transparency in the Southern Ocean: implicationsfor primary productivity. Polar Biology, 14, 127–136.
ULLOA, O., SATHYENDRANATH, S., and PLATT, T., 1994, Effect of the particle-sizedistribution on the backscattering ratio in seawater. Applied Optics, 33, 7070–7077.
VASILKOV, A. P., HOMMEL, D., and CARDER, C., 2001, Mass loading (turbidity) visible/infrared imager/radiometer suite algorithm theoretical basis document. Version 4: May2001. Raytheon Systems Company. http://npoesslib.ipo.noaa.gov/atbd/viirs/Y2410-MassLoading-ATBD-v4.pdf.
VINOGRADOV, M. E., SHUSHKINA, E. A., VEDERNIKOV, V. E., NEZLIN, N. P., and GAGARIN,V. I., 1997, Primary production and plankton stocks in the Pacific Ocean and theirseasonal variation according to remote sensing and field observations. Deep-SeaResearch, 44, 1979–2001.
WALKER, R. E., 1994, Marine light field statistics. (New York: Wiley).WANG, M., and GORDON, H. R., 1994, A simple, moderately accurate, atmospheric
correction algorithm for SeaWiFS. Remote Sensing of Environment, 50, 231–239.WATERS, K. J., 1995, Effects of Raman scattering on the water-leaving radiance. Journal of
Geophysical Research, 100, 13151–13161.ZANEVELD, J. R. V., 1989, An asymptotic closure theory for irradiance in the sea and its
inversion to obtain the inherent optical properties. Limnology and Oceanography, 34,1442–1452.
Estimation of phytoplankton pigment concentration 5073