Optical distinguishability of phytoplankton species and its implications for hyperspectral remote sensing discrimination potential

Full text

Optical distinguishability of phytoplankton species and its implications for
hyperspectral remote sensing discrimination potential
Yuan Zhang
a
, Fang Shen
a,*
, Haiyang Zhao
a
, Xuerong Sun
b
, Qing Zhu
a
, Mengyu Li
a
a
State Key Laboratory of Estuarine and Coastal Research, East China Normal University, Shanghai, China
b
Centre for Geography and Environmental Science, Department of Earth and Environmental Sciences, Faculty of Environment, Science and Economy, University of
Exeter, Cornwall, United Kingdom
ARTICLE INFO
Keywords:
Phytoplankton
Optical property
Phytoplankton types discrimination
Hyperspectral remote sensing；
ABSTRACT
Different phytoplankton types play distinct roles in marine ecosystems, biogeochemical processes, and responses
to climate change. Traditionally, phytoplankton classication has heavily relied on chemical analysis methods
based on phytoplankton pigments, such as High-Performance Liquid Chromatography (HPLC) analysis. This
approach limits the classication resolution to the phylum level of phytoplankton, making it difcult to rene
classication to the genus or species level. With the observation of the hyperspectral ocean satellite PACE
(Plankton, Aerosol, Cloud, ocean Ecosystem mission) louched by NASA in February 2024, there is potential to
achieve ner classication of phytoplankton based on differences in spectral characteristics. This study cultivates
various phytoplankton species in the laboratory to observe their light absorption properties (e.g., specic ab-
sorption coefcients spectra under unit concentration), investigating the spectral differences between different
phyla and among species within the Dinoagellates and Diatoms. Based on the observed absorption and scat-
tering properties of each phytoplankton species, we simulated the remote sensing reectance of different species
under various ocean color components, examining the potential of hyperspectral remote sensing discrimination
of phytoplankton types, and analyzing the impact of Chlorophyll a (Chla), colored dissolved organic matter
(CDOM), and non-algal particles (NAP) concentrations on the remote sensing discrimination. The results show
signicant differences in absorption spectra between different groups of phytoplankton (i.e., Diatoms, Di-
noagellates, Xanthophytes, Coccolithophores, Chlorophytes, Cyanobacteria, Cryptophytes). Among species
within the Dinoagellate group, there are also signicant spectral differences, while species within the Diatom
group exhibit relatively small variations in their spectral shapes. As Chla concentration increases, the potential
for remote sensing discrimination of phytoplankton species also increases; conversely, lower Chla concentrations
pose greater challenges for remote sensing disscrimiantion. Other ocean color components, such as increased
CDOM or NAP concentrations, interfere with the spectral characteristics of phytoplankton in the blue-green
spectral domain. Using hierarchical clustering for phytoplankton classication, the results indicate that Cya-
nobacteria and Chlorophytes can be well distinguished from other group at lower NAP concentrations, while
Diatoms, Cryptophytes, and Xanthophytes are not easily distinguishable from each other. Differentiating be-
tween species within the same group using remote sensing data presents signicant challenges. This study
provides a comprehensive investigation into the optical characteristics of different phytoplankton types, laying a
foundation for their remote sensing classication and deepening the understanding of the potential of hyper-
spectral remote sensing for detailed phytoplankton classication.
1. Introduction
Although the biomass of phytoplankton is only 1 % of that of
terrestrial plants, it contributes approximately 50 % of global primary
productivity (Field et al., 1998; Behrenfeld, 2014). Owing to differences
in morphological and physiological characteristics, different phyto-
plankton play different roles in biogeochemical processes and marine
ecosystems (Le Qu´
er´
e et al., 2005). For example, different diatom spe-
cies have different cell sizes and cell wall silicon‑carbon ratios, resulting
in different abilities to transport carbon to the deep sea (Tr´
eguer et al.,
* Corresponding author.
E-mail address: fshen@sklec.ecnu.edu.cn (F. Shen).
Contents lists available at ScienceDirect
Journal of Sea Research
journal homepage: www.elsevier.com/locate/seares
https://doi.org/10.1016/j.seares.2024.102540
Received 29 June 2024; Accepted 3 September 2024
Journal of Sea Research 202 (2024) 102540
Available online 6 September 2024
1385-1101/© 2024 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license ( http://creativecommons.org/licenses/by-
nc-nd/4.0/ ).

2018). Approximately 75–80 % of the species of toxic algal bloom spe-
cies belong to the dinoagellate group, whereas certain dinoagellate
species do not produce toxins (Janouˇ
skovec et al., 2017). Therefore,
taxonomic studies of phytoplankton diversity are required to enhance
our understanding of phytoplankton function in global biogeochemical
cycling processes.
Owing to differences in size, shape, internal and external structure,
and pigment composition, different phytoplankton can alter the optical
radiation signal (Ciotti et al., 2002; Brewin et al., 2010); therefore, op-
tical radiation-based phytoplankton taxonomy is an important method
for current phytoplankton diversity observation studies (Mouw et al.,
2017). Optical remote sensing techniques can provide large-scale and
long-term measurements of phytoplankton biomass, spatial distribution,
taxonomy, and population composition accompanied with the current
rapid development of sensor technology. Hyperspectral sensors can
provide abundant spectral data for research on biological species of
phytoplankton, thereby enhancing their ability to differentiate phyto-
plankton (Dierssen et al., 2021). Launches of successful hyperspectral
satellites, such as GaoFen-5 (Liu et al., 2019), PRISMA (Cogliati et al.,
2021), EnMAP (Guanter et al., 2015), and the upcoming PACE mission
dedicated to monitoring and investigating phytoplankton species
(Werdell et al., 2019), can improve the amount of data available for the
remote sensing of phytoplankton species (Dierssen et al., 2020).
Prospective efforts have been made to determine the optical remote
sensing potential of various phytoplankton types. For example, Torre-
cilla et al. (2011) found that remote-sensing reectance can better
identify phytoplankton, and 425–435 and 495–540 nm are the most
useful spectral ranges for phytoplankton group research. Xi et al. (2015)
used the phytoplankton absorption coefcient, simulated R
rs
(λ), and
simulated R
rs
(λ)-derived absorption of water compounds to identify the
phytoplankton groups. Their results showed that the direct use of the
phytoplankton absorption coefcient resulted in excellent phyto-
plankton, whereas the absorption coefcients derived from the simu-
lated R
rs
(λ) introduced errors, leading to the worst results. Wolanin et al.
(2016) applied the absorption coefcients of diatoms, coccolithophores,
and cyanobacteria to simulate the R
rs
(λ) under different ocean color
conditions. Consequently, through derivative analysis, they explored the
band setting of hyperspectral sensors suitable for distinguishing phyto-
plankton groups. In general, most previous studies are based on in situ or
model-simulated absorption or remotely sensed reectance spectral
data, combined with signal processing tools, to analyze the optical
properties of phytoplankton, which usually rely on empirical
relationships.
To date, studies have rarely focused on the optical properties of
phytoplankton species and their hyperspectral characteristics of remote
sensing, both in laboratory phytoplankton monoculture settings and in
eld surveys in natural aqutic environments. Moreover, the use of these
hyperspectral characteristics in distinguishing between phytoplankton
species, genera, and phyla remains unclear. Recently, a study investi-
gated the absorption properties of phytoplankton groups, and various
data analyses (i.e., matrix inversion, derivative analyses, and cluster
analysis) were successfully applied to discriminate or quantify the in-
formation of eight phytoplankton groups to a certain extent (Sun et al.,
2019). However, knowledge and understanding of the optical properties
of specic phytoplankton species, particularly their hyperspectral
characteristics of remote sensing, is still lacking, which limits subse-
quent remote sensing applications. Specically: (i) The absorption
characteristics of phytoplankton species remain unclear. More labora-
tory cultures and measurements are required to examine the optical
properties of different phytoplankton species to increase the funda-
mental understanding of the optical variability of distinct species. (ii)
The ability and limitation of using remote sensing reectance in
discriminating phytoplankton species has not been properly investi-
gated, particularly in coastal waters (e.g., those heavily inuenced by
Colored Dissolved Organic Matter (CDOM) and Non-Algal Particles
(NAP)). (iii) At the remote sensing application level, the design of the
sensor bandwidth and position setting is reliant on and guided by the
optical properties of phytoplankton species. In summary, further
research on the optical properties of phytoplankton species is required to
expand the understanding of phytoplanktons.
The objectives of this study are as follows: (i) to investigate the ab-
sorption properties characteristics through laboratory measurements,
analyzing both inter-group and intra-group differences; (ii) to explore
the remote sensing identication capability based on phytoplankton
species and groups using remote sensing reectance; (iii) to analyze the
impact of CDOM and NAP concentrations on the identication of
phytoplankton species; and (iv) to discuss the potential for hyperspectral
discrimination phytoplankton types. The results of this study provide a
foundation for understanding the feasibility of discriminating phyto-
plankton using hyperspectral remote sensing.
2. Data and methods
2.1. Laboratory measurements
2.1.1. Phytoplankton species cultures and HPLC pigments
Table 1 lists 14 uni-algal species from seven phytoplankton taxo-
nomic groups cultured in the laboratory. It includes six species in di-
atoms, three species in dinoagellates, and one each in xanthophytes,
coccolithophores, chlorophytes, cyanobacteria, and cryptophytes
groups. Pseudo-nitzschia pungens, Karenia mikimotoi, Prorocentrum dong-
haiense, Heterosigma akashiwo, Emiliania huxleyi, Platymonas sub-
cordiformis and Synechococcus sp. were obtained from Shanghai
Guangyu Biological Technology Co, Ltd., and other species from Xiamen
University. The algae species were cultured in a light incubator at a
temperature of 20 ◦C under a light dark cycle ratio of 12:12 h and an
optical density of 2500 lx.
For each uni-algal species, three different volumes of pure algal
liquid were diluted to 1 L with pure seawater, resulting in three samples
at different concentrations. Subsequently, each diluted uni-algal sample
was simultaneously ltered onto two Whatman GF/F Glass Microber
lters (0.7
μ
m, 25 mm) under low vacuum pressure. The lters were
stored at −40 ◦C for further analysis. For each species and concentration,
one lter was used to measure the phytoplankton pigment concentration
and the other to measure the phytoplankton absorption coefcient
(aph(λ)).
Phytoplankton pigment concentrations derived from high-
performance liquid chromatography (HPLC) were analyzed by the
Instrumental Analysis Center of Shanghai Jiao Tong University, using
the method proposed by Bidigare et al. (2005). Fourteen phytoplankton
pigments were measured, including Fucoxanthin (Fuco)，19
′
-but-
Fucoxanthin (ButFuco)，19
′
-hex-Fucoxanthin (HexFuco)，Diadinox-
anthin (Diadino)，Lutein (Lut)，Peridinin (Perid)，Neoxanthin
(Neo)，Violaxanthin (Viola)，Zeaxanthin (Zea)，Chl a，Chlorophyll b
(Chlb)，Chlorophyll c
1
(Chl c
1
)，Chlorophyll c
2
(Chl c
2
)，Chlorophyll
Table 1
Information of algae species in the laboratory.
Species Index Species Group
A.1 Skeletonema costatum
Diatoms
A.2 Thalassiosira weissogii
A.3 Chaetoceros debilis
A.4 Chaetoceros curvisetus
A.5 Phaeodactylum tricornutum
A.6 Pseudo-nitzschia pungens
B.1 Karenia mikimotoi
DinoagellatesB.2 Zooxanthella
B.3 Prorocentrum donghaiense
CHeterosigma akashiwo Xanthophytes
DEmiliania huxleyi Coccolithophores
EPlatymonas subcordiformis Chlorophytes
FSynechococcus sp. Cyanobacteria
GCryptomonas sp. Cryptophytes
Y. Zhang et al.
Journal of Sea Research 202 (2024) 102540
2

c
3
(Chl c
3
). The average value of each pigment from three samples of
each algal species was used in this study.
2.1.2. Phytoplankton absorption and backscattering coefcients
The absorption coefcients of total and nonalgal particles, (ap(λ))
and (aNAP(λ)), respectively. Were measured using a PerkinElmer Lambda
1050 UV/VIS spectrophotometer equipped with a 15-cm integrating
sphere in the range of 380–720 nm at 2-nm resolution and 1-nm inter-
polation. Adhering to the NASA and IOCCG ocean optics protocols
(Mitchell et al., 2003; Roesler et al., 2018), the ap(λ)and aNAP(λ)values
were obtained before and after pigment extraction in methanol,
respectively. Further, the phytoplankton absorption coefcient (aph(λ))
was determined as the difference ap(λ) − aNAP. The chlorophyll-specic
absorption coefcient (a*
ph(λ)) was calculated as aph(λ)/Chla. To reduce
this error, the average value of a*
ph(λ)from three samples of each algal
species was used in this study.
The attenuation and absorption coefcients of the unialgal species
were measured using the spectral absorption and attenuation meter (ac-
s, WETLabs Inc.) by pouring diluted samples into the tube. The ac-s
meter has 84 bands in the spectrum range of 400–730 nm. Based on
the measurement protocol of the ac-s instrument (reference or weblink),
the temperature and salinity of the diluted samples were measured for
further corrections. The difference between the attenuation and ab-
sorption coefcients is the scattering coefcient.
The backscattering coefcients of the uni-algal cultures b
b_ph
(λ),
were measured using the back scattering meter (ECO bb9, WETLabs
Inc.). During the measurements, all the lenses of bb9 were completely
submerged in the uni-algal water samples in the pure black bucket under
dark light conditions, and the volume scattering functions at nine
wavebands (i.e., 412, 440, 488, 510, 532, 595, 650, 676, and 715 nm) in
the 124◦direction were collected. In addition, b
b_ph
(λ) was further cor-
rected using the corresponding attenuation and absorption coefcients
from the ac-s meter coupled with the measured temperature and salinity
(Mueller et al. (2002)).
Because of the large volume of uni-algal samples required for
measuring the backscattering coefcients, only b
b_ph
(λ) of S. costatum,
T. weissogii, K. mikimotoi, E. huxleyi were collected through bb9. In
contrast, those of the remaining 10 species were calculated using the
scattering coefcients from the ac-s multiplied by the empirical value of
the phytoplankton backscattering scale coefcient of 0.005 (Shen et al.,
2019).
2.2. Remote sensing reectance dataset
In contrast to oceanic waters, which are dominated by phyto-
plankton, CDOM and NAP also affect R
rs
(λ) in coastal waters. Therefore,
using the measured a*
ph(λ)and b
b_ph
(λ) coefcients of the 14 phyto-
plankton species, the R
rs
(λ) values of different phytoplankton species
with different water component compositions were simulated to study
the spectral characteristics of R
rs
(λ) in optically complex environments.
According to Gordon et al. (1988) and Lee et al. (2002), R
rs
(λ) can be
simulated using the total absorption coefcient (a(λ)) and total back-
scattering coefcient(b
b
(λ)). The specic calculation process is as
follows:
Rrs(λ) = 0.52 ×rrs(λ)
1−1.7×rrs(λ)，#(1)
rrs(λ) = 0.0895 ×u(λ) + 0.1247 ×u(λ)2，#(2)
u(λ) = bb(λ)
a(λ) + bb(λ).#(3)
The total absorption and backscattering coefcients a(λ) and b
b
(λ)
can be expressed as the linear sum of the contributions of all ocean color
components (Sathyendranath et al., 2001), as follows:
a(λ) = aw(λ) + ag(λ) + aNAP(λ) + aph (λ)，#(4)
bb(λ) = bb w(λ) + bb NAP (λ) + bb ph(λ)，#(5)
where aw(λ)is the absorption coefcient of pure seawater, ag(λ)is the
absorption coefcient of the CDOM，bb w(λ)is the backscattering co-
efcient of pure seawater, and bb NAP(λ)is the backscattering coefcient
of the NAP. The aw(λ)and bb w(λ)values were obtained from Pope and
Fry (1997) and Morel (1974), respectively. Table 2 lists the calculation
formulae for ag(λ), aNAP(λ), aph(λ), bb NAP (λ), and bb ph(λ).
In the formula for bb ph(λ)presented in Table 2, λ0 is the reference
wavelength, b*
bph(λ0)is the chlorophyll-specic bb ph(λ)at the reference
wavelength, bb ph (λ0)is bb ph (λ)at reference wavelength, and
η
is the
spectral slope. The backscattering parameters (b*
bph(λ0)，
η
) were ob-
tained by tting the bb ph(λ)formula to the measured data.
Because of the strong absorption of water, phytoplankton pigments,
and Chl a uorescence, when tting the backscattering data measured
by bb9, the 532 nm wavelength was set as the reference wavelength, and
the data at 488, 510, 532, 595, and 650 nm were used for calculation
(Whitmire et al., 2010). To t the backscattering coefcients calculated
using the ac-s meter, a reference wavelength of 555 nm was selected.
Finally, the mean value of the backscatter parameters for each algal
species under the three concentrations was used in this study. Theoret-
ically,
η
should be positive; that is, the backscattering coefcients should
decrease with an increase in wavelength. However, certain values of
η
derived from the ac-s meter were negative. Under these circumstances,
η
was set to an empirical value of 1 (Lee et al., 2002). Table 3 lists the
retrieved backscattering parameters for each of the algal species.
To eliminate the inuence of the value of R
rs
(λ) and only focus on the
spectral shape, R
rs
(λ) was normalized as per the process outlined in a
previous study (Xi et al., 2015),
A=λmax
λmin Rrs(λ)dλ
λmax −λmin
，#(7)
Rrs(λ) = Rrs (λ)
A，#(8)
where Rrs (λ)is the normalized Rrs (λ), and λmin and λmax represent the
minimum and maximum values of the spectral range, respectively. In
this study, λmin and λmax were 380 and 720 nm, respectively.
Table 2
Formulae of absorption and backscattering coefcients of water components.
Description Math Formula References
Absorption coefcient
of CDOM(ag(λ))
ag(λ) = ag(440)exp−Sg× (λ−
440)Sg=0.015
(Bricaud et al.,
1981; Yu,
2013)
Absorption coefcient
of NAP (aNAP(λ))
aNAP(λ) = aNAP(440)exp( − 0.009 ×
(λ−440) )
aNAP(440) = 0.01 ×CNAP
(Shen et al.,
2012)
Absorption coefcient
of phytoplankton
(aph(λ))
aph(λ) = a*
ph(λ) × CChl a
(Devred et al.,
2006)
Backscattering
coefcient of NAP
(bb NAP(λ))
bb NAP(λ) = bb NAP(532) × 532
λn
n=0.4114 ×bb NAP(532)−0.3
bb NAP(532) = 0.0183 ×bNAP(532)
bNAP(532) = 0.2×CNAP
(Liu, 2013)
Backscattering
coefcient of
phytoplankton
(bb ph(λ))
bb ph(λ) = bb ph(λ0) × λ0
λ
η
bb ph(λ0) = b*
bph(λ0) × CChl a
(Werdell et al.,
2014)
Y. Zhang et al.
Journal of Sea Research 202 (2024) 102540
3

2.3. Divergence index
Derivative analysis can effectively highlight spectral characteristics
and has been widely used in spectral studies. The rst (Sʹ(λ)) and second
derivatives (Sʹʹ(λ)) of the spectra were calculated using the following
equations:
Sʹ(λ) = S(λ+Δλ) − S(λ−Δλ)
2Δλ，#(9)
Sʹʹ(λ) = S(λ+Δλ) − 2S(λ) + S(λ−Δλ)
(Δλ)2，#(10)
where Δλ represents the spectral interval between two adjacent bands
and S(λ)represents the spectra. Because derivative analysis can be easily
affected by noise, the measured spectra were smoothed using the
Savitzky-Golay lter with a polynomial order and frame length of 4 and
21, respectively.
An Sʹ(λ)value of 0 indicates the extremum point (maximum or
minimum) of the spectra. Whereas, Sʹʹ(λ)value of 0 indicates the in-
ection point of the spectra. Thus, setting the sensor band at a higher
frequency where the Sʹ(λ)and Sʹʹ(λ)are equal to 0 can facilitate the
capture of the spectral characteristics (Lee et al., 2007).
To quantify differences between spectra of different phytoplankton
groups and species, the divergence index (DI(λ)) was presented in a
previous study (Shen et al., 2019),
DI(λ) = 1
π
×arccosS1•S2
|S1| × |S2|，#(11)
where S1 and S2 represent the two different spectra. The range of DI(λ)
varied from 0 to 1. The closer the DI index is to 0, the greater the sim-
ilarity between the two spectra. In contrast, the closer it is to 1, the
greater the difference between the two spectra.
3. Results and discussion
3.1. Difference of phytoplankton absorption property
In this section, the difference of the absorption spectra among
various groups is investigated. When analyzing the absorption charac-
teristics and differences among severn phytoplankton groups,
S. costatum and K. mikimotoi were chosen to represent diatoms and di-
noagellates, respectively.
3.1.1. Various phytoplankton groups
To focus on the shape of absorption spectra, the a*
ph(λ)was further
normalized at 440 nm (i.e., the a*
ph(λ)was divided by its a*
ph(440)). The
normalized a*
ph(λ)values of the uni-alagl cultures used in this study are
shown in Fig. 1.
Because all the algae species used in this study contained Chl a,
obvious absorption peaks were observed at approximately 440 and 675
nm in all seven phytoplankton groups (Fig. 1). However, the shapes of
the absorption spectra of the seven groups in the other bands were
different. The absorption spectra of chlorophytes and cyanobacteria
were signicantly different from those of other phytoplankton groups.
For example, the absorption spectrum of chlorophytes exhibited ab-
sorption peak and valley at 480 and 456 nm, respectively. In addition,
chlorophytes exhibited a shoulder peak at approximately 650 nm,
whereas the remaining phytoplankton groups exhibited an absorption
valley. The absorption coefcient of chlorophytes increased with
increasing wavelength in the range of 550–675 nm, whereas no such
phenomenon was observed in the other phytoplankton groups. The ab-
sorption spectra of cyanobacteria exhibited an absorption peak and
valley at 490 and 478 nm, respectively, and its absorption spectrum did
not change signicantly with an increase in wavelength in the range of
570–610 nm. Both the dinoagellates and coccolithophores exhibited
absorption peaks at approximately 465 nm; however, the peak of di-
noagellates was weaker than that of coccolithophores. In addition, it is
evident in Fig. 1 that the absorption spectra of diatoms, xanthophytes,
and cryptophytes were slightly different.
Table 4 presents the ratio of pigment to Chl a concentrations in the
seven phytoplankton groups. Chlorophytes were found to be the only
group containing the pigments lutein and Chl b. Based on the
chlorophyll-specic absorption of Chl b shown in Fig. 2(Clementson and
Wojtasiewicz, 2019), it can be inferred that the absorption peaks of
chlorophytes at 480 and 650 nm were mainly associated with Chl b.
Further, the differences in the absorption spectra of different phyto-
plankton groups were owing to the different types and contents of
diagnostic pigments of each group (Catlett and Siegel, 2018). In cya-
nobacteria, only zeaxanthin and diadinoxanthin were detected, and it
was the only phytoplankton group that contained zeaxanthin. This
indicated that the absorption peak of cyanobacteria at 490 nm was most
probably inuenced by zeaxanthin, which is consistent with the spectra
of zeaxanthin (Fig. 2). Except for 19
′
-hex-fucoxanthin, dinoagellates
and coccolithophores shared the same type of pigment (Table 4). For
example, only two groups contained Chl c
3
. According to the spectra of
Chl c
3
shown in Fig. 2, the absorption peak of dinoagellates and coc-
colithophores at approximately 465 nm was most probably owing to the
absorption by Chl c
3
. Moreover, because the ratio of Chl c
3
to Chl a of
coccolithophores Was higher than that of dinoagellates, the absorption
peak of coccolithophores was more obvious at approximately 465 nm.
Xanthophytes and cryptophytes contained the same type of pigment,
which may explain why the absorption spectra of xanthophytes and
cryptophytes were similar. Except for violaxanthin, diatoms, xantho-
phytes, and cryptophytes have the same type of pigment.
To quantitatively analyze the differences between the absorption
spectra of different phytoplankton groups, the DI(λ) values of the ab-
sorption spectra of the seven phytoplankton groups were calculated, as
shown in Fig. 3. It was found that the DI(λ) between the absorption
spectra of any two phytoplankton groups was less than 0.1 in the ranges
of 400–430 and 680–720 nm. Whereas, large differences were mainly
observed in the range of 430–680 nm. The DI(λ) of the absorption
spectra of any two groups of diatoms, xanthophytes, and cryptophytes
were less than 0.2 for the entire wavelengths, except for that in the range
of 570–585 nm. In contrast, the DI(λ) between diatoms and the
remaining phytoplankton groups (except xanthophytes and crypto-
phytes) was greater than 0.5, in the range of 400–500 nm. In particular,
the DI(λ) between diatoms and coccolithophores, chlorophytes, and
cyanobacteria were approximately 1, indicating that the existence of
large differences in the spectra between diatoms and the three groups.
Except in the ranges of 455–458 and 566–589 nm, the DI(λ) values
between diatoms and dinoagellates were less than 0.4.
Table 3
Backscattering parameters of 14 algae species.
Species Index Species Reference band (nm) b*
bph(λ0)/
(m
2
•mg
−1
)
η
A.1 S. costatum 532 0.00084 1.35
A.2 T. weissogii 532 0.00053 1.60
A.3 C. debilis 555 0.00038 0.51
A.4 C. curvisetus 555 0.00047 0.13
A.5 P. tricornutum 555 0.00227 0.13
A.6 P. pungens 555 0.00147 0.64
B.1 K. mikimotoi 532 0.00041 0.50
B.2 Zooxanthella 555 0.00046 1.00
B.3 P. donghaiense 555 0.00068 1.00
C Xanthophytes 555 0.00009 1.00
D Coccolithophores 532 0.00104 1.59
E Chlorophytes 555 0.00018 0.45
F Cyanobacteria 555 0.00065 3.22
G Cryptophytes 555 0.00014 1.00
Y. Zhang et al.
Journal of Sea Research 202 (2024) 102540
4

When comparing the dinoagellates and other groups, the DI(λ)
values between the absorption spectra dinoagellates and that of xan-
thophytes (Fig. 3f), coccolithophores and cryptophytes (Fig. 3m) were
less than 0.65 in the entire spectral range. Consider the coccolithophores
as an example. Except that the DI(λ) between dinoagellate and coc-
colithophores (Fig. 3g), coccolithophores and chlorophytes (Fig. 3n)
were approximately 0.6 at 455 nm, that between the coccolithophores
and the other groups were approximately 1, indicating the existence of
Fig. 1. Normalized a
*
ph(λ)of seven phytoplankton groups. The red, gold, green, and blue shadows in (a) represent the spectral ranges where the absorption spectrum
of dinoagellates, chrysophytes, chlorophytes, and cyanobacteria exhibit considerable differences compared with other phytoplankton groups, respectively. (For
interpretation of the references to color in this gure legend, the reader is referred to the web version of this article.)
Table 4
Ratio of pigment concentration to Chl a concentration in seven phytoplankton groups.
Group Chl c
3
Chl c
2
Chl c
1
ButFuco Fuco Neo Viola HexFuco Diadino Zea Lut Chl b
Dinoagellates 0.01 0.09 –0.02 0.14 0.01 –0.06 0.07 – – –
Coccolithophores 0.20 0.24 – – 0.28 0.32 –0.22 0.15 – – –
Diatoms –0.34 0.02 –0.70 – – – 0.11 – – –
Cyanobacteria – – – – – – – 0.01 0.19 – –
Xanthophytes –0.06 0.03 –0.31 –0.06 –0.03 – – –
Chlorophytes – – – – 0.03 0.07 0.06 – – – 0.06 0.62
Cryptophytes –0.07 0.03 –0.37 –0.06 –0.04 – – –
Fig. 2. Chlorophyll-specic absorption spectra of phytoplankton pigments from (Clementson and Wojtasiewicz, 2019).
Y. Zhang et al.
Journal of Sea Research 202 (2024) 102540
5

signicant differences. Similarly, large differences in DI(λ) values were
observed between cyanobacteria and the other groups. For example, the
value of DI(λ) between the cyanobacteria and chlorophytes was
approximately 0.7 at 480 nm, whereas those between cyanobacteria and
the other phytoplankton groups reached 1. In the ranges of 465–476 and
635–640 nm, the DI(λ) between the absorption spectra of chlorophytes
and the other phytoplankton groups was greater than 0.5. Further, that
between the absorption spectra of xanthophytes and cryptophytes was
less than 0.2 in the entire spectral range. In addition, the DI(λ) values
between the absorption spectra of xanthophytes and the other
Fig. 3. Divergence index of absorption spectra among seven phytoplankton groups.
Y. Zhang et al.
Journal of Sea Research 202 (2024) 102540
6

phytoplankton groups were the same as those between the cryptophytes
and other phytoplankton groups.
3.1.2. Various diatoms species
The a*
ph(λ)values for the six diatom species are shown in Fig. 4. The
shapes of the absorption spectra of the six diatoms were similar, with
only slight differences. For example, in the range of 475–500 nm,
S. costatum, T. weissogii, and P. tricornutum exhibited shoulder peaks.
Whereas, in the range of 575–600 nm, with the exception of S. costatum,
the other ve diatom species exhibited small absorption peaks.
Fig. 5 shows that the absorption spectra of S. costatum and other
diatoms had DI(λ) values of approximately 1 at 577 nm, indicating the
high potential of this alga for optical differentiation. Nonetheless, with
the exception of S. costatum, the DI(λ) values between the absorption
spectra of the two diatoms were less than 0.5 across the entire spectral
range. Particularly, the DI(λ) values between the absorption spectra of
C. debilis, C. curvisetus, and P. tricornutum were less than 0.25 across the
entire spectral range, thereby indicating the poor capacity of these algae
in differentiating among one another.
Moreover, the difference in absorption spectra between xantho-
phytes and certain diatom species was less than that between the two
diatoms species. For example, the DI(λ) value between the absorption
spectra of T. weissogii and xanthophytes was smaller than that between
T. weissogii and P. pungens. Furthermore, the absorption spectra of
T. weissogii and xanthophytes were more similar to each other than
those of T. weissogii and P. pungens.
As shown in Table 5, the six diatom species contain the same four
pigment types; however, the ratio between the concentration of the
phytoplankton pigments and Chl a concentration was different.
3.1.3. Various dinoagellate species
Fig. 6 depicts the a*
ph(λ)spectra of the three methanogenic species.
The absorption spectrum of K. mikimotoi differed signicantly from
those of Zooxanthella and P. donghaiense. All three species contained an
extra absorption peak near 465 nm, with K. mikimotoi exhibiting the
smallest absorption peak. K. mikimotoi displayed a pronounced absorp-
tion valley at approximately 650 nm, whereas Zooxanthella and
P. donghaiense lacked this characteristic.
Within the three dinoagellate species, differences were smaller
between Zooxanthella and P. donghaiense, with DI(λ) values lower than
0.45 in the full spectral range (Fig. 7). This similarity results in similar DI
values between Zooxanthella, P. donghaiense, and other phytoplankton
groups. Large differences were observed between K. mikimotoi and the
other two dinoagellate species, where the DI(λ) value reached 1 at 640
nm. In the range of 440–480 nm, except for the small DI(λ) between the
absorption spectra of Zooxanthella (or P. donghaiense) and coccolitho-
phores, the DI(λ) between the absorption spectra of Zooxanthella and
other phytoplankton groups was close to 1. In the 620–650 nm spectral
range, with the exception of chlorophytes, the DI value between the
absorption spectra of Zooxanthella and other phytoplankton groups was
close to 1, whereas that between the absorption spectra of K. mikimotoi
and chlorophytes was close to 1 in this spectral range. In addition, the DI
(λ) between the absorption spectra of K. mikimotoi and other phyto-
plankton groups was less than 0.5.
Different types of pigments in the three dinoagellate species may be
the primary reason for differences in their absorption spectra (Table 6).
In general, peridinin is a diagnostic pigment for dinoagellates and
fucoxanthin is a diagnostic pigment for diatoms (Vidussi et al., 2001).
However, as a dinoagellate, K. mikimotoi does not contain peridinin
and rather contains fucoxanthin. In addition, compared with Zooxan-
thella and P. donghaiense, K. mikimotoi also contains 19
′
-hex-fucoxanthin,
Chl c
3
, and neoxanthin, which are found in other groups such as diatoms.
According to Table 6 and Fig. 7, the additional absorption peaks of
Zooxanthella and P. donghaiense at approximately 465 nm were mainly
caused by peridinin. Although K. mikimotoi does not contain peridinin,
the presence of Chl c
3
may be the reason for the absorption peak at
approximately 465 nm. In addition, the absorption spectra of the three
dinoagellate species at 550–650 nm were signicantly different, which
may be due to the different types of Chl c.
3.2. Differnece of phytoplankton R
rs
We further analyzed the changes in optical properties under different
biomass conditions, as shown in Fig. 8. To highlight the spectral char-
acteristics, the rst derivative of normalized R
rs
(λ) (Rʹ
rs(λ)) was calcu-
lated. Fig. 8 shows that different phytoplankton groups at NAP
concentrations are 0 g/m
3
, a
g
(440) is 0 m
−1
, Rʹ
rs(λ)spectrum changes
with Chl a concentration. It is evident that when Chl a concentration is
0.1 mg/m
3
, the Rʹ
rs(λ)of different phytoplankton groups the perfor-
mance were consistent, which is greater than 0 before 410 nm, less than
0 after 410 nm, and close to 0 after 520 nm. Thus, the original R
rs
(λ) rst
increases with the increase in wavelength, then decreased with the in-
crease in wavelength after 410 nm, and nally became constant. With an
increase in Chl a concentration, the Rʹ
rs(λ)of different phytoplankton
groups showed differences. However, above 610 nm, the Rʹ
rs(λ)
remained close to 0. This was mainly owing to the strong absorption of
water in the red band, resulting in the spectral characteristics of the
phytoplankton contribution being covered. (See Figs. 9 and 10.)
It was found that Rʹ
rs(λ)of cyanobacteria was quite different from
that of other phytoplankton groups. The Rʹ
rs(λ)of cyanobacteria was
greater than 0 in the range–440-460 nm. At approximately 440 and 460
nm, the original R
rs
(λ) of cyanobacteria had an extreme point. In the
range of 450–500 nm, the Rʹ
rs(λ)of chlorophytes is different from that of
other phytoplankton groups. It was found from the absorption spectrum
of chlorophytes that the reason for this phenomenon may be the ab-
sorption peak at approximately 480 nm.
Rʹ
rs(λ)of diatoms under the four conditions of Chl a concentration in
the 580–675 nm spectral range is the lowest among the seven phyto-
plankton groups, and a valley was formed at approximately 655 nm. In
Fig. 8, with an increase in Chl a concentration, the Rʹ
rs(λ)of diatoms
formed a peak at approximately 685 nm and passed through 0 at 675
and 695 nm, indicating that the original R
rs
(λ) exhibited an extreme
point. With the continuous increase in Chl a concentration, the Rʹ
rs(λ)of
other phytoplankton groups also formed a peak at approximately 685
nm, with extreme points at 675 and 695 nm. Combined with the ab-
sorption spectrum analysis, it was found that the characteristics of R
rs
(λ)
in the range of 650–700 nm were mainly caused by Chl a. Because the
a*
ph(λ)of diatoms was higher, the remote sensing characteristics in this
spectral range were observed when the Chl a concentration in diatoms
was low.
Fig. 4. Normalized a
*
ph(λ)of six diatom species.
Y. Zhang et al.
Journal of Sea Research 202 (2024) 102540
7

3.3. Impact of CDOM and NAP
In coastal waters, R
rs
is inuenced by various water color compo-
nents and their concentrations. Therefore, we investigated the effects of
different concentrations of CDOM and NAP on the optical properties of
phytoplankton. We explored Rrs(λ)under two scenarios, specically: (1)
CDOM cannot be ignored, whereas NAP was ignored. The specic values
were: Chl a concentration is 2.5 mg/m
3
，a
g
(440) is 0.075–0.5 m
−1
, NAP
concentration 0.1 g/m
3
; (2) Neither CDOM nor NAP can be ignored. The
specic values were: Chl a concentration is 2.5 mg/m
3
, a
g
(440) is 0.1
m
−1
, NAP concentration is 1–50 g/m
3
.
Fig. 9 shows the different phytoplankton groups at NAP concentra-
tion was 0 g/m
3
and Chl a concentration was 2.5 mg/m
3
, the Rʹ
rs(λ)
spectrum changed with different a
g
(440). With increase in the a
g
(440),
the characteristics of Rʹ
rs(λ)before 550 nm gradually disappeared. When
a
g
(440) was 0.5 m
−1
, the Rʹ
rs(λ)of different phytoplankton groups was
greater and lesser than 0 within the range of lesser and greater than 570
nm. Thus, the original R
rs
(λ) increased and decreased with the increase
in wavelength when less and greater than 570 nm, respectively. Further,
it was found that when a
g
(440) was 0.5 m
−1
, the Rʹ
rs(λ)of cyanobacteria
was still different from the other phytoplankton groups at approximately
450 and 500 nm.
The Rʹ
rs(λ)of different phytoplankton groups in the range of
550–610 nm exhibited large uctuations (rst decreased rapidly and
then increased rapidly). This can be primarily attributed to the ab-
sorption coefcient of water signicantly changes within the spectral
Fig. 5. Divergence index of absorption spectra amongsix diatom species.
Table 5
Phytoplankton pigments and the ratio of pigment concentration to Chl a con-
centration in six diatom species.
Species Name Chl c
2
Chl c
1
Fuco Diadino
S. costatum 0.32 0.02 0.70 0.11
P. pungens 0.13 0.04 0.60 0.09
C. curvisetus 0.05 0.09 0.40 0.11
P. tricornutum 0.11 0.25 0.72 0.25
T. weissogii 0.04 0.03 0.42 0.15
C. debilis 0.06 0.06 0.44 0.09
Fig. 6. Normalized a
*
ph(λ)of three dinoagellate species. The orange and green
shadows in (b) represent the spectral ranges where the absorption spectrum of
Zooxanthella and K. mikimotoi have great differences compared with other
dinoagellate species, respectively. (For interpretation of the references to color
in this gure legend, the reader is referred to the web version of this article.)
Y. Zhang et al.
Journal of Sea Research 202 (2024) 102540
8

range.
Fig. 10 shows that the Rʹ
rs(λ)changes with NAP concentrations with
Chl a concentration of 2.5 mg/m
3
and a
g
(440) of 0.1 m
−1
. Similar to the
effect of CDOM, with the increase in NAP concentration, the Rʹ
rs(λ)
characteristics of different phytoplankton groups gradually disappeared
before 550 nm. When the NAP concentration was 50 g/m
3
, the Rʹ
rs(λ)of
Fig. 7. Divergence index of absorption spectra between three dinoagellate species.
Table 6
Phytoplankton pigments and the ratio of pigment concentration to Chl a concentration in three dinoagellate species.
Species Name Chl c
3
Chl c
2
Chl c
1
Perid ButFuco Fuco Neo HexFuco Diadino
K. mikimotoi 0.10 0.09 – – 0.02 0.14 0.10 0.06 0.07
P. donghaiense –0.21 –0.99 0.09 – – – 0.11
Zooxanthella sp. –0.41 0.02 0.87 0.06 – – – 0.20
Fig. 8. When the NAP concentration is 0 g/m
3
and ag(440)is 0 m−1, the Rʹ
rs(λ)of different phytoplankton groups changes with the Chl a concentration, where (a) -
(d) represent that the Chl a concentration is 0.10 mg/m3, 1 mg/m
3
, 2.5 mg/m
3
and 5 mg/m
3
respectively.
Y. Zhang et al.
Journal of Sea Research 202 (2024) 102540
9

cyanobacteria could still be distinguished from other phytoplankton
groups at approximately 450 and 500 nm.
In general, with an increase in CDOM or NAP concentration, the
R
rs
(λ) spectral characteristics contributed by phytoplankton in the short-
wavelength position gradually disappeared. This was primarily because
the absorption of CDOM and NAP decreased with increasing wave-
length, thus exhibiting strong absorption under blue light. Although the
backscattering of NAP also decreased with increasing wavelength, the
absorption of NAP was much greater than the backscattering in the blue
light range.
3.4. Potential of hyperspectral remote sensing
PACE (NASA’s Plankton, Aerosol, Cloud, ocean Ecosystem mission)
is the rst global hyperspectral mission for ocean and atmospheric
observation, equipped with a hyperspectral imager called OCI. OCI
features continuous hyperspectral capabilities from ultraviolet to near-
infrared, with a nominal spectral step size of 2.5 nm and an average
bandwidth of approximately 5 nm, covering a spectral range of 340–890
nm. The spatial resolution of OCI imagery is about 1 km
2
at its lowest
point, with a swath width of 2663 km, enabling global coverage every
two days. To explore its potential for monitoring phytoplankton di-
versity, we used the HCA method to further analyze the possibility of
hyperspectral resolution of phytoplankton under three different
scenarios, as shown in Fig. 11.
Fig. 11 (a) shows scenario A, wherein the phytoplankton was
dominant, CDOM and NAP were ignored. The cyanobacteria and
chlorophytes could be well distinguished from other group, mainly
because the absorption spectra of cyanobacteria and chlorophytes were
quite different from other phytoplankton groups. In addition, coccoli-
thophores could also be well distinguished from the other phyto-
plankton groups. It was difcult to distinguish between T. weissogii,
C. debilis, xanthophytes, and cryptophytes mainly because the absorp-
tion spectra of diatoms, xanthophytes, and cryptophytes are similar.
Dinoagellates can also be distinguished from other phytoplankton
groups; however, distinguishing between the three dinoagellate spe-
cies was challenging.
Fig. 11(b) shows scenario B, wherein CDOM was not ignored while
NAP was. Cyanobacteria could still be effectively distinguished from the
other phytoplankton groups, whereas chlorophytes could be distin-
guished from the other phytoplankton groups under most conditions.
However, a misclassication with other phytoplankton groups occurred.
Dinoagellates could be distinguished from the other phytoplankton
groups under most conditions; however, they were misclassied as di-
atoms and coccolithophores.
Fig. 11 (c) shows the results of HCA based on R
rs
(λ) in scenario C
wherein both CDOM and NAP were not ignored. With the increase in
NAP concentration, except for cyanobacteria, which could be
Fig. 9. When the NAP concentration is 0 g/m
3
and the Chl a concentration is 2.5 mg/m
3
, the Rʹ
rs(λ)of different phytoplankton groups changes with the ag(440),
where (a) - (c) represent ag(440)is 0.05m−1, 0.15m−1, 0.5m−1 (respectively).
Fig. 10. When the Chl a concentration is 2.5 mg/m
3
and ag(440)is 0.1m−1, the Rʹ
rs(λ)of different phytoplankton groups changes with the NAP concentration, where
(a) - (c) represent the NAP concentration is 1g/m3, 10 g/m3 and 50 g/m3 (respectively).
Y. Zhang et al.
Journal of Sea Research 202 (2024) 102540
10

distinguished from the other phytoplankton groups under certain con-
ditions, the distinguishability between different phytoplankton groups
decrease. The HCA was not suitable for the remote sensing dis-
tinguishing of phytoplankton species in this scenario, and a new method
must be developed.
In general, when phytoplankton biomass is high or the NAP con-
centration is low, phytoplankton can be distinguished based on R
rs
(λ).
When the NAP concentration is high, that is, in turbid coastal waters, the
effect of identifying phytoplankton groups based on R
rs
(λ) is reduced.
The mean absolute percentage error of Zhu et al. (2019) established that
the phytoplankton species composition model increases by 17 % with
increase in the NAP concentration from 0 to 200 g/m
3
. Cyanobacteria
and chlorophytes are easy to identify compared with other phyto-
plankton groups, and can be well distinguished from other phyto-
plankton group under the condition of high NAP concentration, Xi et al.
(2017) reached the same conclusion. From Fig. 11, it was found that the
linkage distance between diatoms, xanthophyta, and cryptophytes was
small, and remote sensing discrimination encounters considerable
challenges.
4. Conclusion
By capturing continuous and detailed spectral information, hyper-
spectral technology is gradually becoming a forefront hotspot in
phytoplankton remote sensing research. This study investigates the
optical properties of different phytoplankton types and their potential
for discrimination using hyperspectral remote sensing through labora-
tory measurements and remote sensing reectance simulations. The
main conclusions are as follows:
(1) There were signicant differences in the absorption spectra of
different groups, particularly in case of cyanobacteria and
Fig. 11. The results of HCA based on three scenarios.
Y. Zhang et al.
Journal of Sea Research 202 (2024) 102540
11

chlorophytes. This is mainly because of the different types of
phytoplankton pigments contained in different phytoplankton
groups. Among the three dinoagellates, the absorption spectra
of K. mikimotoi were quite different from those of other dinoa-
gellate species, mainly because of the large differences in the
types of phytoplankton pigments contained in different di-
noagellates. The absorption coefcient spectra of the six diatom
species were similar because they contained the same phyto-
plankton pigments. Moreover, the difference in the absorption
spectra between xanthophyta and some diatoms was less than
that between diatoms.
(2) With an increase in Chl a concentration, the difference in R
rs
(λ) of
different phytoplankton groups increased. When Chl a concen-
tration was low, distinguishing phytoplankton groups based on
R
rs
(λ) was challenging.
(3) As the concentrations of colored dissolved organic matter or non-
algal particles increase, the remote sensing spectral features
contributed by phytoplankton in the short-wavelength region
become progressively obscured. This presents a challenge for
distinguishing phytoplankton species using remote sensing
reectance under such conditions.
(4) The HCA results based on R
rs
(λ) showed that when phytoplankton
were dominant, with the exception of diatoms, cryptoalgae, and
xanthophyta, different groups could be effectively distinguished
by R
rs
(λ). Cyanobacteria and chlorophytes were the easiest algae
species to be identied based on R
rs
(λ). However, remote-sensing
discrimination between different species in the same phyto-
plankton group remains a challenging task.
This study explores the optical properties of phytoplankton and an-
alyzes the potential of using hyperspectral technology to distinguish
different phytoplankton types. It provides an important basis for the
development of hyperspectral remote sensing methods for phyto-
plankton diversity inversion.
CRediT authorship contribution statement
Yuan Zhang: Formal analysis, Investigation, Writing – review &
editing. Fang Shen: Conceptualization, Funding acquisition. Haiyang
Zhao: Methodology, Software, Visualization, Writing – original draft.
Xuerong Sun: Writing – review & editing. Qing Zhu: Data curation,
Writing – review & editing. Mengyu Li: Data curation, Methodology,
Writing – review & editing.
Declaration of competing interest
The authors declare that they have no known competing nancial
interests or personal relationships that could have appeared to inuence
the work reported in this paper.
Data availability
Data will be made available on request.
Acknowledgments
This study was funded by the National Natural Science Foundation of
China [Grant No. 42076187 and No.42271348].
References
Behrenfeld, M.J., 2014. Climate-mediated dance of the plankton. Nat. Clim. Chang. 4,
880–887.
Bidigare, R.R., Van Heukelem, L., Trees, C.C., 2005. In: Andersen, R.A. (Ed.), Analysis of
Algal Pigments by High- Performance Liquid Chromatography[M]. Algal Culturing
Techniques. Elsevier Academic Press, pp. 327–345.
Brewin, R.J.W., Lavender, S.J., Hardman-Mountford, N.J., Hirata, T., 2010. A spectral
response approach for detecting dominant phytoplankton size class from satellite
remote sensing. Acta Oceanol. Sin. 29, 14–32.
Bricaud, A., Morel, A., Prieur, L., 1981. Absorption by dissolved organic matter of the sea
(yellow substance) in the UV and visible domains. Limnol. Oceanogr. 26, 43–53.
Catlett, D., Siegel, D.A., 2018. Phytoplankton pigment communities can be modeled
using unique relationships with spectral absorption signatures in a dynamic coastal
environment. J. Geophys. Res. Oceans 123, 246–264.
Ciotti, ´
A.M., Lewis, M.R., Cullen, J.J., 2002. Assessment of the relationships between
dominant cell size in natural phytoplankton communities and the spectral shape of
the absorption coefcient. Limnol. Oceanogr. 47, 404–417.
Clementson, L.A., Wojtasiewicz, B., 2019. Dataset on the absorption characteristics of
extracted phytoplankton pigments. Data Brief 24, 103875.
Cogliati, S., Sarti, F., Chiarantini, L., Cosi, M., Lorusso, R., Lopinto, E., Miglietta, F.,
Genesio, L., Guanter, L., Damm, A., P´
erez-L´
opez, S., Schefer, D., Tagliabue, G.,
Panigada, C., Rascher, U., Dowling, T.P.F., Giardino, C., Colombo, R., 2021. The
PRISMA imaging spectroscopy mission: overview and rst performance analysis.
Remote Sens. Environ. 262, 112499.
Devred, E., Sathyendranath, S., Stuart, V., Maass, H., Ulloa, O., Platt, T., 2006. A two-
component model of phytoplankton absorption in the open ocean: theory and
applications. J. Geophys. Res. Oceans 111.
Dierssen, H., Bracher, A., Brando, V., Loisel, H., Ruddick, K., 2020. Data needs for
hyperspectral detection of algal diversity across the globe. Oceanography 33, 74–79.
Dierssen, H.M., Ackleson, S.G., Joyce, K.E., Hestir, E.L., Castagna, A., Lavender, S.,
McManus, M.A., 2021. Living up to the hype of hyperspectral aquatic remote
sensing: science, resources and outlook. Front. Environ. Sci. 9.
Field, C.B., Behrenfeld, M.J., Randerson, J.T., Falkowski, P., 1998. Primary production of
the biosphere: integrating terrestrial and oceanic components. Science 281,
237–240.
Gordon, H.R., Brown, O.B., Evans, R.H., Brown, J.W., Smith, R.C., 1988. A semianalytic
radiance model of ocean color. J. Geophys. Res. Atmos. 93, 10909–10924.
Guanter, L., Kaufmann, H., Segl, K., Foerster, S., Rogass, C., Chabrillat, S., Kuester, T.,
Hollstein, A., Rossner, G., Chlebek, C., Straif, C., Fischer, S., Schrader, S., Storch, T.,
Heiden, U., Mueller, A., Bachmann, M., Mühle, H., Müller, R., Habermeyer, M.,
Ohndorf, A., Hill, J., Buddenbaum, H., Hostert, P., van der Linden, S., Leit˜
ao, P.,
Rabe, A., Doerffer, R., Krasemann, H., Xi, H., Mauser, W., Hank, T., Locherer, M.,
Rast, M., Staenz, K., Sang, B., 2015. The EnMAP spaceborne imaging spectroscopy
mission for earth observation. Remote Sens. 7, 8830–8857.
Janouˇ
skovec, J., Gavelis, G.S., Burki, F., Dinh, D., Bachvaroff, T.R., Gornik, S.G.,
Bright, K.J., Imanian, B., Strom, S.L., Delwiche, C.F., Waller, R.F., Fensome, R.A.,
Leander, B.S., Rohwer, F.L., Saldarriaga, J.F., 2017. Major transitions in
dinoagellate evolution unveiled by phylotranscriptomics. Proc. Natl. Acad. Sci.
114, E171–E180.
Le Qu´
er´
e, C., Harrison, S.P., Prentice, I.C., Buitenhuis, E.T., Aumont, O., Bopp, L.,
Claustre, H., Cunha, L.C.D., Geider, R., Giraud, X., Klaas, C., Kohfeld, K.E.,
Legendre, L., Manizza, M., Platt, T., Rivkin, R.B., Sathyendranath, S., Uitz, J.,
Watson, A.J., Wolf-Gladrow, D., 2005. Ecosystem dynamics based on plankton
functional types for global ocean biogeochemistry models. Glob. Chang. Biol. 11,
2016–2040.
Lee, Z., Carder, K.L., Arnone, R.A., 2002. Deriving inherent optical properties from water
color: a multiband quasi-analytical algorithm for optically deep waters. Appl. Opt.
41, 5755–5772.
Lee, Z., Carder, K., Arnone, R., He, M., 2007. Determination of primary spectral bands for
remote sensing of aquatic environments. Sensors 7, 3428–3441.
Liu, M., 2013. Scattering Properties of Suspended Particles in High Turbid Waters and
Remote Sensing Application. Master Thesis. East China Normal University,
Shanghai, China.
Liu, Y., Zhang, J., Zhang, Y., Sun, W., Jiao, L., Sun, D., Hu, X., Ye, X., Li, Y., Liu, S.,
Cao, K., Chai, M., Zhou, W., 2019. The advanced hyperspectral imager: aboard
China’s GaoFen-5 satellite. IEEE Geosci. Remote Sens. Mag. 7, 23–32.
Mitchell, B.G., Kahru, M., Wieland, J., Stramska, M., 2003. Determination of spectral
absorption coefcients of parti- cles, dissolved material and phytoplankton for
discrete water samples. In: Mueller, J.L., Fargion, G.S., McClain, C.R. (Eds.), Ocean
Optics Protocols for Satellite Ocean Color Sensor Validation, Revision 4, Volume Iv:
Inherent Optical Properties: Instruments, Characterizations, Field Measurements and
Data Analysis Protocols. National Aeronautical and Space Administration.
Morel, A., 1974. Optical properties of pure water and pure sea water. In: Jerlov, N.G.,
Nielsen, E.S. (Eds.), Optical Aspects of Oceanography, New York. Academic Press.
Mouw, C.B., Hardman-Mountford, N.J., Alvain, S., Bracher, A., Brewin, R.J.W.,
Bricaud, A., Ciotti, A.M., Devred, E., Fujiwara, A., Hirata, T., Hirawake, T.,
Kostadinov, T.S., Roy, S., Uitz, J., 2017. A consumer’s guide to satellite remote
sensing of multiple phytoplankton groups in the global ocean. Front. Mar. Sci. 4, 41.
Mueller, J.L., Fargion, G.S., McClain, C.R., 2002. Ocean Optics Protocols for Satellite
Ocean Color Sensor Validation, Revision 4, Volume IV: Inherent Optical Properties:
Instruments, Characterizations, Field Measurements and Data Analysis Protocols.
Goddard Space Flight Space Center.
Pope, R.M., Fry, E.S., 1997. Absorption spectrum (380-700 nm) of pure water. II.
Integrating cavity measurements. Appl. Opt. 36, 8710–8723.
Roesler, C., Stramski, D., Sa, E.J.D., R¨
ottgers, R., Reynolds, R.A., 2018. Chapter 5:
Spectrophotometric measure- ments of particulate absorption using lter pads. In:
Neeley, A.R., Mannino, A. (Eds.), Ocean Optics & Biogeochemistry Protocols for
Satellite Ocean Colour Sensor Validation, Volume 1: Inherent Optical Property
Measurements and Protocols: Absorption Coefcient (v1.0). International Ocean
Colour Coordinating Group (IOCCG) in Collaboration With National Aeronautics and
Space Administration (NASA).
Y. Zhang et al.
Journal of Sea Research 202 (2024) 102540
12

Sathyendranath, S., Cota, G., Stuart, V., Maass, H., Platt, T., 2001. Remote sensing of
phytoplankton pigments: a comparison of empirical and theoretical approaches. Int.
J. Remote Sens. 22, 249–273.
Shen, F., Zhou, Y., Hong, G., 2012. Absorption property of non-algal particles and
contribution to total light absorption in optically complex waters, a case study in
Yangtze and adjacent coast. In: International Conference on Environment Science,
pp. 61–66. Melbourne, Australia.
Shen, F., Tang, R., Sun, X., Liu, D., 2019. Simple methods for satellite identication of
algal blooms and species using 10-year time series data from the East China Sea.
Remote Sens. Environ. 235, 111484.
Sun, X.R., Shen, F., Brewin, R.J.W., Liu, D.Y., Tang, R.G., 2019. Twenty-year variations
in satellite-derived chlorophyll-a and phytoplankton size in the Bohai Sea and
Yellow Sea. J. Geophys. Res. Oceans 124, 8887–8912.
Torrecilla, E., Stramski, D., Reynolds, R.A., Mill´
an-Nú˜
nez, E., Piera, J., 2011. Cluster
analysis of hyperspectral optical data for discriminating phytoplankton pigment
assemblages in the open ocean. Remote Sens. Environ. 115, 2578–2593.
Tr´
eguer, P., Bowler, C., Moriceau, B., Dutkiewicz, S., Gehlen, M., Aumont, O., Bittner, L.,
Dugdale, R., Finkel, Z., Iudicone, D., Jahn, O., Guidi, L., Lasbleiz, M., Leblanc, K.,
Levy, M., Pondaven, P., 2018. Inuence of diatom diversity on the ocean biological
carbon pump. Nat. Geosci. 11, 27–37.
Vidussi, F., Claustre, H., Manca, B.B., Luchetta, A., Marty, J.C., 2001. Phytoplankton
pigment distribution in relation to upper thermocline circulation in the eastern
Mediterranean Sea during winter. J. Geophys. Res. Oceans 106, 19939–19956.
Werdell, P.J., Roesler, C.S., Goes, J.I., 2014. Discrimination of phytoplankton functional
groups using an ocean reectance inversion model. Appl. Opt. 53, 4833–4849.
Werdell, P.J., Behrenfeld, M.J., Bontempi, P.S., Boss, E., Cairns, B., Davis, G.T., Franz, B.
A., Gliese, U.B., Gorman, E.T., Hasekamp, O., Knobelspiesse, K.D., Mannino, A.,
Martins, J.V., McClain, C.R., Meister, G., Remer, L.A., 2019. The plankton, aerosol,
cloud, ocean ecosystem mission: status, Science, advances. Bull. Am. Meteorol. Soc.
100, 1775–1794.
Whitmire, A.L., Pegau, W.S., Karp-Boss, L., Boss, E., Cowles, T.J., 2010. Spectral
backscattering properties of marine phytoplankton cultures. Opt. Express 18,
15073–15093.
Wolanin, A., Soppa, M., Bracher, A., 2016. Investigation of spectral band requirements
for improving retrievals of phytoplankton functional types. Remote Sens. 8, 871.
Xi, H., Hieronymi, M., R¨
ottgers, R., Krasemann, H., Qiu, Z., 2015. Hyperspectral
differentiation of phytoplankton taxonomic groups: a comparison between using
remote sensing reectance and absorption spectra. Remote Sens. 7, 14781–14805.
Xi, H., Hieronymi, M., Krasemann, H., R¨
ottgers, R., 2017. Phytoplankton group
identication using simulated and in situ hyperspectral remote sensing reectance.
Front. Mar. Sci. 4, 272.
Yu, X., 2013. Measurements of Pigment Absorption Coefcients and Retrieval Models of
Pigment Concentration in Turbid Coastal Waters. Master Thesis. East China Normal
University, Shanghai, China.
Zhu, Q., Shen, F., Shang, P., Pan, Y., Li, M., 2019. Hyperspectral remote sensing of
phytoplankton species composition based on transfer learning. Remote Sens. 11,
2001.
Y. Zhang et al.
Journal of Sea Research 202 (2024) 102540
13