Open Access
Add to BibSonomy
Abstract
Phytoplankton community is an important organism indicator of monitoring water quality, and accurately estimating its composition and biomass is crucial for understanding marine ecosystems and biogeochemical processes. Identifying phytoplankton species remains a challenging task in the field of oceanography. Phytoplankton fluorescence is an important biological property of phytoplankton, whose fluorescence emissions are closely related to its community. However, the existing estimation approaches for phytoplankton communities by fluorescence are inaccurate and complex. In the present study, a new, simple method was developed for determining the Chlorophytes, Chrysophytes, Cryptophytes, Diatoms, Dinoflagellates, and Prymnesiophytes based on the fluorescence emission spectra measured from the HOBI Labs Hydroscat-6P (HS-6P) in the Bohai Sea, Yellow Sea, and East China Sea. This study used single bands, band ratios, and band combinations of the fluorescence signals to test their correlations with the six dominant algal species. The optimal band forms were confirmed, i.e., X1 (i.e., fl(700), which means the fluorescence emission signal at 700 nm band) for Chlorophytes, Cryptophytes, Dinoflagellates, and Prymnesiophytes (R = 0.947, 0.862, 0.911, and 0.918, respectively) and X7 (i.e., [fl(700) + fl(550)]/[fl(550)/fl(700)], where fl(550) denotes the fluorescence emission signal at 550 nm band) for Chrysophytes and Diatoms (R = 0.893 and 0.963, respectively). These established models here show good performances, yielding low estimation errors (i.e., root mean square errors of 0.16, 0.02, 0.06, 0.36, 0.18, and 0.03 for Chlorophytes, Chrysophytes, Cryptophytes, Diatoms, Dinoflagellates, and Prymnesiophytes, respectively) between in situ and modeled phytoplankton communities. Meanwhile, the spatial distributions of phytoplankton communities observed from both in situ and fluorescence-derived results agreed well. These excellent outputs indicate that the proposed method is to a large extent feasible and robust for estimating those dominant algal species in marine waters. In addition, we have applied this method to three vertical sections, and the retrieved vertical spatial distributions by this method can fill the gap of the common optical remote sensing approach, which usually only detects the sea surface information. Overall, our findings indicate that the proposed method by the fluorescence emission spectra is a potentially promising way to estimate phytoplankton communities, in particular enlarging the profiling information.
© 2018 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
Phytoplankton is not only the principal primary producer in the ocean, but also the main supplier of dissolved oxygen and initiator of the food network in aquatic ecosystems, and plays a vital role in the energy flow, material circulation and information transmission of aquatic ecosystems [1–4]. Biodiversity is a natural property of phytoplankton, and its community structures are highly variable at temporal and spatial scales [5–7]. Studying the composition of various phytoplankton communities is of great significance for marine water quality management and nutrient pathways [7,8]. It is inadequate to quantify and identify phytoplankton community structures, only by using an analysis of Chl a measurement [9]. Such a single proxy could neither reflect changes in the structure of phytoplankton community nor accurately estimate biomass [10,11]. Instead, an accurate estimation of phytoplankton species is essential for understanding marine ecosystems and biogeochemical processes.
At present, several traditional methods have been used to identify phytoplankton communities, such as high-performance liquid chromatography (HPLC) [12–16], microscopic approaches [17–19], and flow cytometry [20,21]. These measurement methods are based on accurate field water sampling and are time-consuming. The phytoplankton communities determined through these methods are discontinuous due to discrete points; thus, these methods cannot obtain real continuous spatial distributions of phytoplankton communities. In addition, experiments using these methods require experienced analysts with phytoplankton taxonomic information [22]. Although these available methods have some limitations, they are able to accurately acquire phytoplankton information. At the same time, these traditional methods can provide a foundation of data for the retrieval method of remote sensing and fluorescence. Recently, several satellite-based remote sensing methods have been developed to derive phytoplankton communities and size structures [23–25], and these methods can estimate a large dynamic scale of optical properties of natural waters. Nevertheless, remote sensing method, as a passive technique, can only obtain water surface information, and is also seriously subject to the environment conditions (i.e., cloudy and whitecaps). Moreover, remote sensing causes worse inversion results in the near shore turbid coastal waters due to inaccurate atmospheric corrections [26]. Instead, the fluorescence method can address these limitations.
Phytoplankton taxonomic groups can be differentiated by their compositions [27] since each phytoplankton community includes different pigment composition. At the same time, in vivo fluorescence of photosynthetic pigments in phytoplankton cells, which is based on different excitation or emission spectra for various phytoplankton intracellular pigments, provides a potential method for determining phytoplankton communities [28,29]. The fluorescence emission spectra of all phytoplankton are approximately 680 nm, and the feature of this fluorescence emission spectra is unique in aquatic ecosystem [30]. Hence, fluorescence emission spectra have a great potential to identify phytoplankton community information [31].
Indeed, some in situ commercial submersible spectrofluorometers, such as the FluoroProbe (bbe-Moldaenke, Kiel, Germany), Multi-Exciter (JFE Advantech Co., Ltd., Japan) and laser-induced fluorescence system (LIF) (Second Institute of Oceanography, State Oceanic Administration, China), were developed to detect phytoplankton communities by measuring the excitation spectra. However, most methods depend on a constant value assumption of the norm spectra determined by culturing pure algae in the laboratory, and these constant norm spectra are not always suitable for analyzing individual phytoplankton in some investigated water areas [32]. This may lead to the inaccurate estimations of phytoplankton communities [32–37] [Table 1]. For instance, the methods of Catherine et al. (2012) [35] and Harrison et al. (2016) [37] studies are based on the hypothesis that the norm spectra of different algae cultured in the laboratory are assumed constants. Nevertheless, such a hypothesis may not be always true, since the norm spectra of phytoplankton vary from changes in phytoplankton between different areas, as shown in previous studies [32,33]. The statistical approach, used in Wang et al. (2016) [34] study, can be only applied to site-specific areas over the ECS or Tsushima Strait, though it does not depend on the norm spectra. The bi-Gaussian mixture model, developed in Chen et al. (2015) study [36] by the LIF field-measured data, requires water sampling and is time-consuming. The capacity of existing fluorescence-based methods to distinguish phytoplankton community does not satisfy the current requirement. Overall, developing approaches based on phytoplankton fluorescence spectra for determining phytoplankton communities remains an essential and exigent task.
Table 1. Summary analysis of advantages and disadvantages of previous methods for estimating phytoplankton communities.
In the current study, a new method based on fluorescence emission spectra collected by the HOBI Labs Hydroscat-6P (HS-6P) was developed for the first time to derive phytoplankton community in the Bohai Sea, Yellow Sea, and East China Sea. The purpose of this study was to establish a way for deriving phytoplankton communities using field-fluorescence measurements. The capacity and sensitivity of this method were also evaluated and analyzed. In addition, we also have discussed the rationality and limitation of this model and further analyzed its potential application.
2. Materials and methods
2.1 Study area and data collection
The study areas of this study are the Bohai Sea (BS), Yellow Sea (YS) and East China Sea (ECS). The semi-enclosed BS, with a surface area of 7.70 × 104 km2, is the smallest and shallowest of the three seas with an average depth of 18 m (maximum depth of 83 m). The YS has a surface area of 3.8 × 105 km2 and an average depth of 44 m (maximum depth of 140 m) [25,38]. The ECS, at 7.70 × 105 km2, is the largest of the three seas, with an average depth 370 m (maximum depth of ~2,719 m) [39]. Historically, the BS, YS, and ECS have been highly productive and polluted due to severe terrestrial discharge [40–42]. In summer, the ECS receives a large amount of eutrophic freshwater from the Yangtze River and forms the Yangtze River dilution water [43–45], which promotes the growth of phytoplankton and regulates the community structures. The data sets used in this study were collected from three cruise surveys that were carried out in the BS and YS in January 2017, in the ECS in September 2016, and in Zhejiang coastal waters (ZJW) in May 2016 [Fig. 1].
Fig. 1 Locations of the sampling stations in the Bohai Sea (BS), Yellow Sea (YS) in Jan 2017, in the East China Sea (ECS) in Sep 2016, and in Zhejiang coastal waters in May 2016. The colors indicate bathymetry.
2.2 Determination of the phytoplankton community based on pigment concentrations
Pigment concentrations were determined through the quantitative filtration technology (QFT). The pigments were analyzed by filtering water samples onto 47 mm Whatman GF/F glass fiber. At the same time, the filter samples were promptly stored in liquid nitrogen and then were reserved in a deep freezer (−80°C) onshore for pursuant laboratory analysis. The concentrations of 19 phytoplankton pigments were confirmed by reversed-phase high-performance liquid chromatography (HPLC) with a Zorbax Eclipse XDB-C8 column (150 mm × 4.6 mm, 3.5µm; Agilent Technologies) using the method of Van Heukelem and Thomas (Van Heukelem et al., 2001) and were calibrated by commercial pigment standards (Sigma Aldrich, St. Louis and DHI, Hørsholm, Denmark). These phytoplankton communities were identified by the retention time and absorption spectrum of the photodiode array detectors.
The phytoplankton communities were deduced from the HPLC pigment concentrations through the CHEMTAX program (version 1.95) [14,46], which is used for identifying and quantifying phytoplankton with a matrix of field pigment data and the initial matrix of pigment [47–49]. In this study, the initial input matrix of pigment used in the CHEMTAX calculation was based on those in the BS, YS and ECS according to the previous studies [48,50,51], and six dominant phytoplankton species were considered including Diatoms, Dinoflagellates, Chlorophytes, Chrysophytes, Cryptophytes, and Prymnesiophytes. Other algae with weak content were not considered this time. Before running CHEMTAX, a randomly determined factor F, where F = 1 + S × (R - 0.5), S is the scale factor (normally 0.7), and R is the random number between 0 and 1 generated by the Microsoft Excel RAND function, is selected to multiply the initial ratio to obtain 60 different ratios. Each of the 60 ratio tables was used as the starting point for a CHEMTAX optimization. The solution with the smallest residual was used for estimating taxonomic abundance. The optimal 10% of the results (n = 6) were selected to calculate the average and the standard deviation of the abundance estimations [46].
The phytoplankton communities in our investigated water areas showed wide dynamic ranges, which were collected from the three cruises [Table 2 and Fig. 1]. The Diatoms varied from very low value (~0) to 11.097 mg m−3 with the largest mean of 0.641 ± 1.507 mg m−3. A high CV value (235%) indicated its large variation. The mean of the Chrysophytes was the lowest, at 0.0260 ± 0.049 mg m−3, with a large CV (187%). This result indicated that large differences in phytoplankton communities existed in the investigated water regions. In addition, the values of Chlorophytes, Cryptophytes, Dinoflagellates, and Prymnesiophytes were between the mean of Diatoms and Chrysophytes.
Table 2. Description of in situ observed data sets collected from three cruises surveys in the Bohai Sea, Yellow Sea and East China Sea.
2.3 Fluorescence emission spectra measurements
The recently improved HOBI Labs Hydroscat-6P by the HOBI instrument services is a self-contained instrument for measuring particulate backscattering, bbp(λ), at six channels (i.e., 442, 488, 550, 620, 700, and 852 nm), and an important improvement is that it adds measurements of fluorescence signals. The HS-6P has two pairs of fluorescence channels whose beams and fields of views cross each other (see details in Hydroscat-6P User’s Manual from www.hobiservices.com), and has a standard depth rating of 500 m, which had been calibrated by the manufacturer before the cruise surveys. The fluorescence measurement at two bands (i.e., 550 nm and 700 nm) is the same as the optics method used for its backscattering measurements, which unceasingly excite chlorophyll fluorescence and CDOM fluorescence in the violet band by two light-emitting diodes (LEDs). The HS-6P measures chlorophyll fluorescence emission at 700 nm excited by 442 nm, and measures CDOM fluorescence emission at 550 nm excited by 420 nm. At each survey station, the instrument was lowered to approximately 3 m underwater to preheat the instrument, and then was elevated to just beneath the water surface and lowered again (at a speed of about 0.2 m/s) to the seabed to measure the vertical profile of phytoplankton fluorescence.
In this study, the collection of the fluorescence emission spectra from the HS-6P instrument [Fig. 2] showed large variations. The statistical values ranged from 9.61E-05 to 7.71E-02 with a mean value of 6.27E-03 ± 10−2 and from 1.10E-04 to 1.24E-02 with a mean value of 9.51E-04 ± 10−2 for 550 nm and 700 nm bands, respectively. Correspondingly, the fluorescence emission spectra at two bands, i.e., 550 nm and 700 nm, showed large CV values (207.5% and 168%, respectively). The changes of magnitude were mainly related to variations in phytoplankton biomass and CDOM, respectively, which were verified by the substantial linear relationships between fluorescence signals at 700 nm and Chl a, and between fluorescence signals at 550 nm and CDOM, respectively, as shown in Fig. 8.
Fig. 2 Frequency distribution of fluorescence signals of (A) 550 nm and (B) 700 nm bands. The black line represents a normal distribution curve.
Phytoplankton fluorescence is a critical attribute that relates the biological characteristics of phytoplankton to their optical responses [31,52,53]. By means of phytoplankton fluorescence signals, different phytoplankton communities can be detected using the so-called fluorescence taxonomic approach, which is actually based on algal features that various taxa possess different types of antennas and accessory pigments [27,54,55]. Thus, available fluorescence data is important source signals to explore retrieval algorithms on phytoplankton species. However, there are currently very limited filed fluorescence measurements that are used for identifying different phytoplankton communities. This study obtained valuable synchronous measurements on fluorescence and phytoplankton community data. Two fluorescence data signals, i.e., red (700 nm) and green (550 nm) bands, had been collected in the current study. A potential theoretical basis for our proposed models is that the red-band fluorescence signals have been widely known as sources from phytoplankton diagnostic pigments [30], when the green-band fluorescence signals is mainly from CDOM effect that may influence the establishment of phytoplankton community [33,56] fluorescence models. Coupling the red- and green- band fluorescence data into the potential retrieval models of phytoplankton community would be promising. The above mentioned theoretical foundations provide a potential feasibility for estimating phytoplankton community structure by using the fluorescence emission spectra.
2.4 Model description
Considering that empirical algorithms are primitive, easy to establish and significantly direct and efficient, especially for local areas [57], this study utilized single bands, ratios, and band combinations to develop the six phytoplankton community models over the BS, YS, and ECS, which is a similar method to those used in previous studies [58,59]. In this study, we designed eight band forms (see Table 3 for the details), which included single bands, band ratios, and band combinations. Some previous studies [58,59] showed that these band combinations can perform well in model development. Specifically, the above eight band forms were trained by MATLAB; thereby, each form was tested using all possible band combinations of the fluorescence emission spectra to provide optimal results for each form (see Fig. 3 for detail). In addition, the optimal band combinations with the highest correlation coefficient R were determined. According to the above test, we determined X1 (i.e., fl(700)) for the Chlorophytes, Cryptophytes, Dinoflagellates, and Prymnesiophytes, i.e., fl(700), with the highest R (0.947, 0.862, 0.911, and 0.918, respectively), and X7 for the Chrysophytes and Diatoms, i.e., [fl(700) + fl(550)]/[fl(550)/fl(700)], with the highest R (0.893 and 0.963, respectively). To determine good model performances, four mathematical function models were used to fit the six phytoplankton community models, which were linear, power, exponential, and logarithmic functions, and an optimal function model was recommended by each accuracy evaluation to retrieve the those phytoplankton communities. The model parameters and the accuracy assessment indictors (i.e., R2, RMSE, MAE, and bias) are shown in Fig. 4. Moreover, the comparison of the results of each fitted model were evaluated (see Fig. 4 for detail), which indicated that the linear model performed best for the Chlorophyte and Chrysophytes estimations, and the other four phytoplankton communities performed better under the power function.
Table 3. The band combination forms used in this study, X1 to X8 represent the eight different forms. Note that i and j indicate the two fluorescence emission bands.
Fig. 3 Correlation coefficient between the six phytoplankton species and fluorescence emission spectra with eight band combination forms. They include single fl(700) (X1), log10 (fl(700)) (X2), fl(550)-fl(700) (X3), fl(550)/fl(700) (X4), log10 (fl(550))/log10 (fl(700)) (X5), [fl(550)-fl(700)]/[fl(550)/fl(700)] (X6), [fl(550) + fl(700)]/[fl(550)/fl(700)] (X7), and [fl(550)-fl(700)]/[fl(550) + fl(700)] (X8). Note that each form with specific bands has been tested to be optimal in comparison with those using other bands.
Fig. 4 Evaluation (R2, RMSE, MAE, and bias) of phytoplankton community estimation models for all samples.
2.5 Accuracy evaluation
Statistical analyses including minimum, maximum, mean, standard deviation (S.D.), and coefficient of variation (CV) values were performed for pigment concentrations and fluorescence emission spectra in this study. To evaluate the models’ performances, this study utilized four indicators of the root mean square error (RMSE), mean absolute (MAE), determination coefficient (R2), and relative bias (bias), which could be calculated as follows:
where the yi and yiˊ denote the in situ measured and modeled values, respectively, for the ith sample; N represents the total number of samples.To further assess the ability of the developed models, a leave-one-out cross validation evaluation (LOO-CV) [60–62] was conducted. Generally, we randomly selected one sample from all samples (sample number, N) as model validation, and the remaining subset of N-1 samples were used for model training. This step continued until each sample was taken as a validation sample. For each fold, the model calibration data were used to obtain the regression coefficients. The final model determination was averaging the regression coefficients of the N-time model, which effectively avoided the selection bias. Hence, a reliable and stable model was determined by the LOO-CV method.
3. Results
3.1 Model development and validation
The six phytoplankton community models developed in this study are for dominant phytoplankton species including Chlorophytes, Chrysophytes, Cryptophytes, Diatoms, Dinoflagellates, and Prymnesiophytes.
The obtained results (see Section 2.4 for details) showed that the linear model performed best for the Chlorophyte and Chrysophytes estimations, and the other four phytoplankton communities performed better under the power function. As shown in Fig. 5, strong correlations were observed for the six phytoplankton species with excellent performances (i.e., R2>0.75, and even up 0.93) and significantly low errors (i.e., RMSE of 0.16, 0.02, 0.06, 0.36, 0.18, and 0.03 for Chlorophytes, Chrysophytes, Cryptophytes, Diatoms, Dinoflagellates, and Prymnesiophytes, respectively). Note that such strong correlations with low errors show good confidence. These results of the model calibration indicated that the six phytoplankton community estimation models developed here had a great potential.
Fig. 5 Scatter plots of X1 (A, D, E, and F), and X7 (B and C) versus the six phytoplankton species. The solid red lines are the fitted function curves, and the dotted red lines are the 95% confidence bounds.
To further evaluate the performance of the six proposed models of this study, a LOO-CV method was conducted. The scatter plots comparing in situ and modeled phytoplankton community values are shown in Fig. 6. Most of the model-estimated values generally followed the 1:1 line well, with 90.4%, 97.3%, 94%, 93.7%, 89%, and 92.5% data points distributing within the ± 10% ranges, and 99%, 100%, 100%, 98.2%, 99%, and 99% data points distributed within the ± 20% ranges for Chlorophytes, Diatoms, Chrysophytes, Dinoflagellates, Cryptophytes, and Prymnesiophytes, respectively. Each model showed excellent results, with one-to-one correspondence between the modeled and in situ phytoplankton community values and significantly low errors and high coefficient of determination values. These results indicate that the phytoplankton community estimation models are robust in the BS, YS, and ECS, but further research is required to determine their use for other regions.
Fig. 6 Comparisons between in situ measured and modeled (A) Chlorophytes (B) Diatoms (C) Chrysophytes (D) Dinoflagellate (E) Cryptophytes, and (F) Prymnesiophytes for leave-one-out cross validation (the red solid line:1:1). The dotted cyan and green lines indicate the ± 10% and ± 20% ranges, respectively, relative to the 1:1 line.
Following the model calibrations, the developed models were validated by the LOO-CV method to evaluate the precision of each model (see Fig. 6 for detail). The validation analysis showed significantly low errors and high correlation coefficients between in situ measured values and those derived from fluorescence spectra model. Here, we further analyzed the comparison between their spatial distributions derived both in field measurements and fluorescence derivations during the three cruises [Fig. 7]. Overall, most distribution patterns are consistent with each other. The six phytoplankton communities in the Zhejiang coastal waters showed much higher values than those in other regions. Both in situ and fluorescence-derived distributions showed low values and relatively higher values near the Yangtze River estuary. In addition, the Chlorophytes, Diatoms, and Dinoflagellates had a larger value compared to that of the other three species and were generally the dominant group. Despite some degree of discrepancy that existed in part of the North Yellow Sea (NYS) and near the mouth of the Yangtze River, there was general agreement in the large-scale patterns between the field measurements and fluorescence-derived results. This result further suggests that the proposed models are reliable to estimate the phytoplankton community structures from in situ fluorescence measurements in the study region.
Fig. 7 Comparison of spatial distributions of the six phytoplankton species between in situ measurements and fluorescence derivations during the three cruises.
3.2 Accuracy verification of fluorescence emission signals by the HS-6P instrument
In this study, the fluorescence emission spectra of phytoplankton at the two bands (i.e., 550 nm and 700 nm) could be collected by the HS-6P instrument. We compared the in situ HS-6P measurements and other measurements by commercial instruments. As shown in Figs. 8(A) and 8(B), the correlation between HS-6P fluorescence emission spectra at the 700 nm band and Chl a measured by the Turner fluorometer (Trilogy, Turner Designs Inc.) and submersible ECO fluorometer (Wetlabs Inc., Halifax, NS, Canada), respectively. Strong correlations existed between the 700 nm band and Chl a (R2 = 0.91 and 0.93, p<0.001, respectively). Similarly, high correlation (R2 = 0.89, p<0.001) was found between HS-6P fluorescence emission spectra at the 550 nm band and CDOM measured by ECO [Fig. 8C]. This indicates that the fluorescence emission spectra of HS-6P at the two bands can function well like the commercial spectrofluorometers.
Fig. 8 Correlations between field HS-6P measurements at the 700 nm band and (A) Chl a measured by Turner (N = 136), (B) Chl a measured by ECO (N = 158) and the correlation between field HS-6P measurements at the 550 nm band and (C) CDOM measured by ECO (N = 99). The colors indicate different cruise data sets.
3.3 Model application
Compared with optical remote sensing methods, shipborne fluorescence methods can obtain information of phytoplankton communities from both water surface and different profiles information for each station, and also are not affected almost by the weather. Here, the proposed models are preliminarily applied to three section verticals (see red lines marked in Fig. 1) surveyed during cruises in BS and YS in Jan 2017 and ECS in Sep 2016. The vertical resolution of the fluorescence emission spectra of each station is approximately from 0.1 to 0.4 m, and a high-resolution vertical distribution of phytoplankton communities can be obtained by applying the developed estimation models. In this study, we resampled from 0.1 to 0.4 m into 1 m to avoid possible instrument measurement noise. The preliminary application of this model can be seen in Figs. 9, which is of great significance for understanding the biogeochemical processes and ecosystems of marine waters. Clear spatial and vertical variations in the six phytoplankton species were observed in the BS, YS, and ECS vertical sections.
Fig. 9 Vertical distributions in the BS section of the Chlorophytes (A), Diatoms (B), Chrysophytes (C), Dinoflagellate (D), Cryptophytes (E), and Prymnesiophytes (F); in the YS section of the Chlorophytes (G), Diatoms (H), Chrysophytes (I), Dinoflagellate (J), Cryptophytes (K), and Prymnesiophytes (L); in the ECS section of the Chlorophytes (M), Diatoms (N), Chrysophytes (O), Dinoflagellate (P), Cryptophytes (Q), and Prymnesiophytes (R) estimated from the fluorescence emission spectra.
In the vertical section of the BS, the concentrations of Chlorophytes, Cryptophytes, and Dinoflagellate were higher than the concentrations of other species in the nearly whole water column, while the concentrations of Chrysophytes, Diatoms, and Prymnesiophytes were relatively low [Figs. 9(A)-(F)]. At the middle stations, the Chlorophytes, Cryptophytes, and Dinoflagellates showed relatively higher concentrations at the maximum layer of Chl a, while at the eastern stations, the concentrations were similarly distributed in the vertical section. For the Chrysophytes, Diatoms, and Prymnesiophytes, similar distributions with relatively low concentrations were found. As shown in Figs. 9(G)-(L), the concentration distributions of the six phytoplankton species in the YS vertical section were similar to those in the BS vertical section. At the western stations, the Chlorophytes, Cryptophytes, Dinoflagellate, and Prymnesiophytes were abundant at the subsurface Chl a maximum, which are possibly impacted by the nutrient-rich water regions such as the Subei Shoal. However, relatively higher concentrations were detected at the surface of the eastern stations for these phytoplankton species. Compared with the vertical sections of BS and YS, the Diatoms were one of the dominant algae in the ECS vertical section [Figs. 9(M)-(R)]. For all phytoplankton species, high concentrations appeared at the water surfaces of the eastern stations, which were possibly related to the nutrient-rich diluted water of the Yangtze River. The middle layers of the eastern stations showed higher concentrations, potentially influenced by Kuroshio waters. Those spatial distributions of the phytoplankton communities, estimated from the developed models by using in situ fluorescence emission spectra in this study, agree well, with that in previous studies [34,48].
4. Discussion
This study proposed a new method to retrieve phytoplankton community structures from in situ fluorescence emission spectra over the BS, YS, and ECS, which is for the first time conducting the investigations based on the fluorescence spectral data from the HS-6P instrument. Various synchronous measurements on those used parameters in this study are advantageous, which supplies a stable data basis for the construction of the six models [Fig. 5]. Additionally, eight band combinations were tested to provide the optimal form as the models’ input [Table 3 and Fig. 3]. High correlations (i.e., X1 and X7) support to establish accurate phytoplankton community estimation models for this study area. Moreover, we verified the stability and potential of the six phytoplankton community models by using the LOO-CV method, as shown in Fig. 6. Such results indicate that the phytoplankton community estimation models proposed here have a good potential.
It should be noted that the phytoplankton community estimation models were established based on HPLC pigment data by the CHEMTAX program, which might produce a certain bias compared to the true phytoplankton composition [14,46]. These deviations might introduce some uncertainty in the estimation of the models, like stated in the previous studies [34,63]. However, the concept of our proposed method does not absolutely rely on the HPLC analysis. Once the more accurate phytoplankton communities are available, the same method can be easily adopted to obtain the phytoplankton community retrieval algorithms. In addition, the developed models are based on regional data investigated over the BS, YS, and ECS, and may have limitations when applied to other water areas, since phytoplankton has specific physiological states under different environments, which may transform pigment composition and its fluorescence characteristics [34,56,64]. Even so, the data sets used in this study are from three cruises, which cover large dynamic ranges of phytoplankton and water conditions, (such as spring and autumn season of algae bloom [65,66], summer stratification and winter strong mixing [39,67]), that to some extent possess applicability in those similar water environments. However, more validation for the models demonstrated here and data collection from more water types are needed in future.
Compared with several traditional approaches of determining phytoplankton communities (such as microscopy, flow cytometry, and HPLC technique), fluorescence spectrometry has advantages in the joint ability of in situ observation, high accuracy, and low costs [68,69]. At present, many approaches using commercial fluorometers have been used for taxonomic discrimination of phytoplankton. However, most methods depend on a constant value assumption of the norm spectra determined by culturing pure algae in the laboratory, and these constant norm spectra do not always take effect in some water areas [32]. Based on previous studies [35,37], the obtained low precisions are not satisfactory [Table 1]. In addition to the fluorescence method, the light absorption of phytoplankton (aph(λ)) is strongly related to its pigment information, shape, and size structure [70,71]. However, some algorithms have been established to estimate pigment compositions, size structures, and communities of phytoplankton from aph(λ) by commercial instrument measurements such as WET labs AC-S [72–76]. In fact, the aph(λ) is mainly impacted by nonalgal particles and chromophoric dissolved organic matter (CDOM), which leads to difficulty in its measurement to a large degree [30]. In contrast, fluorescence has the potential to discriminate phytoplankton communities.
In the present study, we developed a new method to discriminate six algal species with good outputs [Figs. 5-7, and 9]. However, note that these models have been established by using our sampled concentration dynamic ranges of algal species. Detailed ranges are 0-2.787, 0-0.369, 0-0.887, 0-11.097, 0-2.319, and 0-0.616 mg m−3, for Chlorophytes, Chrysophytes, Cryptophytes, Diatoms, Dinoflagellates, and Prymnesiophytes, respectively. Thus, the valid ranges for the use of the developed fluorescence models should be generally consistent with the above scopes. Whether the models are suitable for estimations beyond the concentration scopes is still required for further research. In addition, a high-frequency, high-resolution, and long-term period vertical distribution of phytoplankton communities can be obtained based on the developed models. Currently, it is still hard to accurately identify the phytoplankton community structures. This study proposed a great potential method to derive the six algae based on the in situ fluorescence emission spectral by the HS-6P instrument. For instance, the Diatoms model showed the highest inversion accuracy (i.e., the R2 and RMSE values of 0.93 (p<0.001) and 0.36 mg m−3, respectively). Although the Cryptophytes model had the lowest inversion accuracy (i.e., the R2 and RMSE values of still 0.77 (p<0.001) and 0.06 mg m−3). Additionally, the retrieval precision of these algae existed some deviations, probably due to different algae concentration levels and their fluorescence signal intensities. At the same time, it must be emphasized that this fluorescence inversion approach is based on the collection of six phytoplankton community data sets in this study area, once more other actual phytoplankton communities can be obtained, our approach will be commendably adopted to establish estimation models of the phytoplankton community. Further research is needed in future work. Overall, our method proposes a direct and effective foundation for estimating phytoplankton communities through the fluorescence emission spectra of HS-6P. The two fluorescence emission spectral channels at 550 nm and 700 nm from the HS-6P instrument are shown to distinguish the phytoplankton communities well, though several other band channels had been utilized (e.g., 375 nm, 400 nm, 420 nm, 435 nm, 470 nm, 505 nm, 525 nm, 570 nm and 590 nm, respectively, see Yoshida et al. (2011) study [30]) in previous studies. Further research should be required to explore the potentials of other fluorescence channels in our study water regions.
5. Conclusion
Knowledge of phytoplankton communities and biomass will provide important information for understanding biogeochemical process and monitoring water quality in the aquatic systems. However, accurately discriminating phytoplankton species is still a challenge in marine areas. In the present study, we first developed a neoteric method to estimate phytoplankton communities in the BS, YS, and ECS from in situ fluorescence emission spectra collected by the HS-6P instrument. Eight band combination forms of fluorescence emission spectra were trained to determine those optimal forms for phytoplankton species retrievals [Table 3 and Fig. 3]. The X1 and X7 were ultimately recommended as inputs for the fitting models (see section 2.4). The validation assessment of the models using the leave-one-out cross-validation method showed good performances with relatively low errors [Fig. 6]. The goodness of fit for those models are high with R2 results in the range of 0.77-0.93 for those phytoplankton species. Correspondingly, the obtained RMSE are 0.16, 0.02, 0.06, 0.36, 0.18, and 0.03 for Chlorophytes, Chrysophytes, Cryptophytes, Diatoms, Dinoflagellates, and Prymnesiophytes, respectively. The generally consistent spatial distribution between those fluorescence-derived and in situ observed values also indicates our proposed models’ good performances. As an important application, our proposed models can derive the profiling information on the phytoplankton communities, as examples shown in Figs. 9, which is indeed beneficial for the comprehensive monitoring of phytoplankton assemblages. In short, this novel method utilizes the fluorescence emission spectra collected from the HS-6P instrument, and shows a potential capability to simply and accurately retrieve phytoplankton community information in the study water areas. The extension of application to other regions is expectable only if carrying out further validation and necessary region-specific parameterization.
Funding
National Key Research and Development Program of China (2016YFC1400904, 2016YFC1400901); National Natural Science Foundation of China (NSFC)(41876203, 41576172, 41506200); Jiangsu Provincial Programs for Marine Science and Technology Innovation (HY2017-5); Open Fund of State Key Laboratory of Satellite Ocean Environment Dynamics, Second Institute of Oceanography (QNHX1812); Qing Lan Project; and NSFC Open Research Cruise (NORC 2018-01), Ship-time Sharing Project of NSFC.
Acknowledgments
We acknowledge captains, officers, and crews of R/V Dongfanghong 2 and Science 3 for providing excellent assistance during field sampling and measurements. We acknowledge hard work received from Xiaojing Shen, Hailong Zhang, Cong Xiao, Zhaoxin Li, Xiaoping Su, and Ying Mao in our field investigations. We also thank the three reviewers and the editor for their suggestions and comments that help improve the quality of the manuscript.
References
1. D. Antoine and A. Morel, “Oceanic primary production: 1. Adaptation of a spectral light‐photosynthesis model in view of application to satellite chlorophyll observations,” Global Biogeochem. Cycles 10(1), 43–55 (1996). [CrossRef]
2. C. B. Field, M. J. Behrenfeld, J. T. Randerson, and P. Falkowski, “Primary production of the biosphere: integrating terrestrial and oceanic components,” Science 281(5374), 237–240 (1998). [CrossRef] [PubMed]
3. T. R. Parsons and C. M. Lalli, “Jellyfish Population Explosions: Revisiting a Hypothesis of Possible Causes,” Mer (Paris) 40, 111–121 (2002).
4. M.-E. Carr, M. A. M. Friedrichs, M. Schmeltz, M. N. Aita, D. Antoine, K. R. Arrigo, I. Asanuma, O. Aumont, R. Barber, M. Behrenfeld, R. Bidigare, E. T. Buitenhuis, J. Campbell, A. Ciotti, H. Dierssen, M. Dowell, J. Dunne, W. Esaias, B. Gentili, W. Gregg, S. Groom, N. Hoepffner, J. Ishizaka, T. Kameda, C. Le Quéré, S. Lohrenz, J. Marra, F. Mélin, K. Moore, A. Morel, T. E. Reddy, J. Ryan, M. Scardi, T. Smyth, K. Turpie, G. Tilstone, K. Waters, and Y. Yamanaka, “A comparison of global estimates of marine primary production from ocean color,” Deep Sea Res. Part II Top. Stud. Oceanogr. 53(5-7), 741–770 (2006). [CrossRef]
5. R. R. Bidigare and M. E. Ondrusek, “Spatial and temporal variability of phytoplankton pigment distributions in the central equatorial Pacific Ocean,” Deep Sea Res. Part II Top. Stud. Oceanogr. 43(4-6), 809–833 (1996). [CrossRef]
6. J. Ras, H. Claustre, and J. Uitz, “Spatial variability of phytoplankton pigment distributions in the Subtropical South Pacific Ocean: comparison between in situ and predicted data,” Biogeosciences 4, 353–369 (2007).
7. L. S. Bruckman, T. L. Richardson, J. A. Swanstrom, K. A. Donaldson, M. Allora Jr., T. J. Shaw, and M. L. Myrick, “Linear discriminant analysis of single-cell fluorescence excitation spectra of five phytoplankton species,” Appl. Spectrosc. 66(1), 60–65 (2012). [CrossRef]
8. E. Litchman, Chapter 16 - Resource Competition and the Ecological Success of Phytoplankton, 351–375 (Elsevier Inc., 2007).
9. I. C. Prentice, C. Le Quéré, E. T. Buitenhuis, J. I. House, C. Klaas, and W. Knorr, “Biosphere dynamics: challenges for earth system models,” State of the Planet Frontiers & Challenges in Geophysics 150, 269–278 (2004). [CrossRef]
10. M. Kruskopf and K. J. Flynn, “Chlorophyll content and fluorescence responses cannot be used to gauge reliably phytoplankton biomass, nutrient status or growth rate,” New Phytol. 169(3), 525–536 (2006). [CrossRef] [PubMed]
11. T. Hirata, N. J. Hardmanmountford, R. J. W. Brewin, J. Aiken, R. Barlow, K. Suzuki, T. Isada, E. Howell, T. Hashioka, M. Noguchiaita, and Y. Yamanaka, “Synoptic relationships between surface Chlorophyll-a and diagnostic pigments specific to phytoplankton functional types,” Biogeosciences 8(2), 311–327 (2011). [CrossRef]
12. D. F. Millie, H. W. Paerl, and J. P. Hurley, “Microalgal pigment assessments using high-performance liquid chromatography: a synopsis of organismal and ecological applications,” Can. J. Fish. Aquat. Sci. 50(11), 2513–2527 (1993). [CrossRef]
13. C. K. Wong and C. K. Wong, “HPLC pigment analysis of marine phytoplankton during a red tide occurrence in Tolo Harbour, Hong Kong,” Chemosphere 52(9), 1633–1640 (2003). [CrossRef] [PubMed]
14. M. D. Mackey, D. J. Mackey, H. W. Higgins, and S. W. Wright, “CHEMTAX-a program for estimating class abundances from chemical markers: application to HPLC measurements of phytoplankton,” Mar. Ecol. Prog. Ser. 144, 265–283 (1996). [CrossRef]
15. R. J. W. Brewin, S. Sathyendranath, T. Hirata, S. J. Lavender, R. M. Barciela, and N. J. Hardman-Mountford, “A three-component model of phytoplankton size class for the Atlantic Ocean,” Ecol. Modell. 221(11), 1472–1483 (2010). [CrossRef]
16. R. J. W. Brewin, S. Sathyendranath, T. Jackson, R. Barlow, V. Brotas, R. Airs, and T. Lamont, “Influence of light in the mixed-layer on the parameters of a three-component model of phytoplankton size class,” Remote Sens. Environ. 168, 437–450 (2015). [CrossRef]
17. S. W. Wright, D. P. Thomas, H. J. Marchant, H. W. Higgins, and D. J. Mackey, “Analysis of phytoplankton of the Australian sector of the Southern Ocean: comparisons of microscopy and size frequency data with interpretations of pigment HPLC data using the ‘CHEMTAX’ matrix factorisation program,” Euphytica 144, 285–298 (1996).
18. S. Roy, J. P. Chanut, M. Gosselin, and T. Simengando, “Characterization of phytoplankton communities in the lower St. Lawrence Estuary using HPLC-detected pigments and cell microscopy,” Mar. Ecol. Prog. Ser. 142, 55–73 (1996). [CrossRef]
19. H. Havskum, L. Schluter, R. Scharek, E. Berdalet, and S. Jacquet, “Routine quantification of phytoplankton groups - microscopy or pigment analyses?” Mar. Ecol. Prog. Ser. 273, 31–42 (2004). [CrossRef]
20. S. Sun and J. Lu, “Applications of flow cytometer in ecological studies of Nano- and Pico-phytoplankton,” Shengtaixue Zazhi 122, 12385–12386 (2000).
21. R. J. Olson, A. Shalapyonok, and H. M. Sosik, “An automated submersible flow cytometer for analyzing pico- and nanophytoplankton: FlowCytobot,” Deep Sea Res. Part I Oceanogr. Res. Pap. 50(2), 301–315 (2003). [CrossRef]
22. J. Gregor, R. Geriš, B. Maršálek, J. Heteša, and P. Marvan, “In situ quantification of phytoplankton in reservoirs using a submersible spectrofluorometer,” Hydrobiologia 548(1), 141–151 (2005). [CrossRef]
23. V. Brotas, R. J. W. Brewin, C. Sá, A. C. Brito, A. Silva, C. R. Mendes, T. Diniz, M. Kaufmann, G. Tarran, S. B. Groom, T. Platt, and S. Sathyendranath, “Deriving phytoplankton size classes from satellite data: validation along a trophic gradient in the eastern Atlantic Ocean,” Remote Sens. Environ. 134, 66–77 (2013). [CrossRef]
24. I. Ioccg, “Phytoplankton Functional Types from Space,” in Reports of the International Ocean-Colour Coordinating Group (IOCCG) (2014), pp. 1-156.
25. D. Sun, Y. Huan, Z. Qiu, C. Hu, S. Wang, and Y. He, “Remote sensing estimation of phytoplankton size classes From goci satellite measurements in Bohai Sea and Yellow Sea,” J. Geophys. Res. Oceans 122(10), 8309–8325 (2017). [CrossRef]
26. S. Wang, Y. Huan, Z. Qiu, D. Sun, H. Zhang, L. Zheng, and C. Xiao, “Remote sensing of particle cross-sectional area in the Bohai Sea and Yellow Sea: algorithm development and application implications,” Remote Sens. 8(10), 841 (2016). [CrossRef]
27. S. W. Jeffrey, R. F. C. Mantoura, and S. W. Wright, “Phytoplankton pigments in oceanography: guidelines to modern methods,” Monogr. Oceanogr. Methodol. 10, 1110 (1997).
28. C. S. Yentsch and C. M. Yentsch, “Fluorescence spectral signatures: the characterization of phytoplankton populations by the use of excitation and emission spectra [Algae, diatoms],” Appl. Environ. Microbiol. 60, 2911–2915 (1979).
29. C. S. Yentsch and D. A. Phinney, “Spectral fluorescence: an ataxonomic tool for studying the structure of phytoplankton populations,” J. Plankton Res. 7(5), 617–632 (1985). [CrossRef]
30. M. Yoshida, T. Horiuchi, and Y. Nagasawa, “In situ multi-excitation chlorophyll fluorometer for phytoplankton measurements: Technologies and applications beyond conventional fluorometers,” in Oceans, 2011, 1–4.
31. P. G. Falkowski and J. A. Raven, “Aquatic Photosynthesis,” (Princeton University Press, 2013).
32. H. L. Macintyre, E. Lawrenz, and T. L. Richardson, “Taxonomic discrimination of phytoplankton by spectral fluorescence,” (2010), pp. 129–169.
33. G. Johnsen and E. Sakshaug, “Biooptical characteristics of PSII and PSI in 33 species (13 pigment groups) of marine phytoplankton, and the relevance for pulse-amplitude-modulated and fast-repetition-rate fluorometry,” J. Phycol. 43(6), 1236–1251 (2007). [CrossRef]
34. S. Wang, C. Xiao, J. Ishizaka, Z. Qiu, D. Sun, Q. Xu, Y. Zhu, Y. Huan, and Y. Watanabe, “Statistical approach for the retrieval of phytoplankton community structures from in situ fluorescence measurements,” Opt. Express 24(21), 23635–23653 (2016). [CrossRef] [PubMed]
35. A. Catherine, N. Escoffier, A. Belhocine, A. B. Nasri, S. Hamlaoui, C. Yéprémian, C. Bernard, and M. Troussellier, “On the use of the FluoroProbe®, a phytoplankton quantification method based on fluorescence excitation spectra for large-scale surveys of lakes and reservoirs,” Water Res. 46(6), 1771–1784 (2012). [CrossRef] [PubMed]
36. P. Chen, D. Pan, and Z. Mao, “Application of a laser fluorometer for discriminating phytoplankton species,” Opt. Laser Technol. 67, 50–56 (2015). [CrossRef]
37. J. W. Harrison, E. T. Howell, S. B. Watson, and R. E. H. Smith, “Improved estimates of phytoplankton community composition based on in situ spectral fluorescence: use of ordination and field-derived norm spectra for the bbe FluoroProbe,” Can. J. Fish. Aquat. Sci. 73(10), 1472 (2016). [CrossRef]
38. D. Sun, Z. Qiu, C. Hu, S. Wang, L. Wang, L. Zheng, T. Peng, and Y. He, “A hybrid method to estimate suspended particle sizes from satellite measurements over Bohai Sea and Yellow Sea,” J. Geophys. Res. Oceans 121(9), 6742–6761 (2016). [CrossRef]
39. C. T. A. Chen, “Chemical and physical fronts in the Bohai, Yellow and East China seas,” J. Mar. Syst. 78(3), 394–410 (2009). [CrossRef]
40. R. C. Beardsley, R. Limeburner, H. Yu, and G. A. Cannon, “Discharge of the Changjiang (Yangtze River) into the East China Sea,” Cont. Shelf Res. 4(1-2), 57–76 (1985). [CrossRef]
41. J. Pang, M. Jiang, and L. I. Fulin, “Changes and development trend of runoff, sediment discharge and coastline of the yellow river estuary,” Haiyang Huzhao Tongbao 04, 1–6 (2000).
42. H. Fan and H. Huang, “Response of coastal marine eco-environment to river fluxes into the sea: a case study of the Huanghe (Yellow) River mouth and adjacent waters,” Mar. Environ. Res. 65(5), 378–387 (2008). [CrossRef] [PubMed]
43. H. Ichikawa and M. Chaen, “Seasonal variation of heat and freshwater transports by the Kuroshio in the East China Sea,” J. Mar. Syst. 24(1-2), 119–129 (2000). [CrossRef]
44. Z. Chen, J. Li, H. Shen, and W. Zhanghua, “Yangtze River of China: historical analysis of discharge variability and sediment flux,” Geomorphology 41(2-3), 77–91 (2001). [CrossRef]
45. H. J. Lee, K. T. Jung, J. K. So, and J. Y. Chung, “A three-dimensional mixed finite-difference Galerkin function model for the oceanic circulation in the Yellow Sea and the East China Sea in the presence of M tide,” Cont. Shelf Res. 22(1), 67–91 (2002). [CrossRef]
46. M. D. Mackey, D. J. Mackey, H. W. Higgins, and S. W. Wright, “CHEMTAX - a program for estimating class abundances from chemical markers:application to HPLC measurements of phytoplankton,” Mar. Ecol. Prog. Ser. 144, 265–283 (1997). [CrossRef]
47. S. W. Wright and R. L. van den Enden, “Phytoplankton community structure and stocks in the East Antarctic marginal ice zone (BROKE survey, January–March 1996) determined by CHEMTAX analysis of HPLC pigment signatures,” Deep Sea Res. Part II Top. Stud. Oceanogr. 47(12-13), 2363–2400 (2000). [CrossRef]
48. K. Furuya, M. Hayashi, Y. Yabushita, and A. Ishikawa, “Phytoplankton dynamics in the East China Sea in spring and summer as revealed by HPLC-derived pigment signatures,” Deep Sea Res. Part II Top. Stud. Oceanogr. 50(2), 367–387 (2003). [CrossRef]
49. Z. Y. Zhu, W. M. Ng, S. M. Liu, J. Zhang, J. C. Chen, and Y. Wu, “Estuarine phytoplankton dynamics and shift of limiting factors: a study in the Changjiang (Yangtze River) Estuary and adjacent area,” Estuar. Coast. Shelf Sci. 84(3), 393–401 (2009). [CrossRef]
50. F. Kong, R. Yu, Q. Zhang, T. Yan, and M. Zhou, “Pigment characterization for the 2011 bloom in Qinhuangdao implicated “brown tide” events in China,” Chin. J. Oceanology Limnol. 30(3), 361–370 (2012). [CrossRef]
51. X. Liu, B. Huang, Q. Huang, L. Wang, X. Ni, Q. Tang, S. Sun, H. Wei, S. Liu, C. Li, and J. Sun, “Seasonal phytoplankton response to physical processes in the southern Yellow Sea,” J. Sea Res. 95, 45–55 (2015). [CrossRef]
52. J. T. O. Kirk, “Light and photosynthesis in aquatic ecosystems,” J. Ecol. 45, 220 (1994).
53. S. Eddy, “Aquatic photosynthesis. second edition,” Northeastern Naturalist504, (2007).
54. N. Hoepffner and S. Sathyendranath, “Effect of pigment composition on absorption properties of phytoplankton,” Mar. Ecol. Prog. Ser. 73, 11–23 (1991). [CrossRef]
55. K. S. Rowan, “Photosynthetic pigments of algae,” Q. Rev. Biol. 30, 235 (1991).
56. Y. Huot and M. Babin, “Overview of fluorescence protocols: theory, basic concepts, and practice,” Springer Netherlands 4, 31– 74 (2010).
57. X. He, Y. Bai, D. Pan, N. Huang, X. Dong, J. Chen, C. T. A. Chen, and Q. Cui, “Using geostationary satellite ocean color data to map the diurnal dynamics of suspended particulate matter in coastal waters,” Remote Sens. Environ. 133, 225–239 (2013). [CrossRef]
58. Z. Qiu, “A simple optical model to estimate suspended particulate matter in Yellow River Estuary,” Opt. Express 21(23), 27891–27904 (2013). [CrossRef] [PubMed]
59. Z. Qiu, L. Zheng, Y. Zhou, D. Sun, S. Wang, and W. Wu, “Innovative GOCI algorithm to derive turbidity in highly turbid waters: a case study in the Zhejiang coastal area,” Opt. Express 23(19), A1179–A1193 (2015). [CrossRef] [PubMed]
60. M. Kearns and D. Ron, “Algorithmic stability and sanity-check bounds for leave-one-out cross-validation,” Neural Comput. 11(6), 1427–1453 (1999). [CrossRef] [PubMed]
61. G. C. Cawley, “Leave-One-Out Cross-Validation based model selection criteria for weighted LS-SVMs,” International Joint Conference on Neural Networks, 4426:1661–1668 (2006).
62. A. Vehtari, A. Gelman, and J. Gabry, “Practical bayesian model evaluation using leave-one-out cross-validation and WAIC,” Stat. Comput. 27, 1–20 (2015).
63. S. W. Wright, R. L. van den Enden, I. Pearce, A. T. Davidson, F. J. Scott, and K. J. Westwood, “Phytoplankton community structure and stocks in the Southern Ocean (30–80°E) determined by CHEMTAX analysis of HPLC pigment signatures,” Deep Sea Res. Part II Top. Stud. Oceanogr. 57(9-10), 758–778 (2010). [CrossRef]
64. K. Roháček and M. Barták, “Technique of the modulated chlorophyll fluorescence: basic concepts, useful parameters, and some applications,” Photosynthetica 37(3), 339–363 (1999). [CrossRef]
65. K. R. Arrigo, D. K. Perovich, R. S. Pickart, Z. W. Brown, G. L. van Dijken, K. E. Lowry, M. M. Mills, M. A. Palmer, W. M. Balch, F. Bahr, N. R. Bates, C. Benitez-Nelson, B. Bowler, E. Brownlee, J. K. Ehn, K. E. Frey, R. Garley, S. R. Laney, L. Lubelczyk, J. Mathis, A. Matsuoka, B. G. Mitchell, G. W. Moore, E. Ortega-Retuerta, S. Pal, C. M. Polashenski, R. A. Reynolds, B. Schieber, H. M. Sosik, M. Stephens, and J. H. Swift, “Massive phytoplankton blooms under Arctic sea ice,” Science 336(6087), 1408 (2012). [CrossRef] [PubMed]
66. Y. Xu, J. Ishizaka, H. Yamaguchi, E. Siswanto, and S. Wang, “Relationships of interannual variability in SST and phytoplankton blooms with giantjellyfish (Nemopilema nomurai) outbreaks in the Yellow Sea and East China Sea,” J. Oceanogr. 69(5), 511–526 (2013). [CrossRef]
67. I. M. Belkin, P. C. Cornillon, and K. Sherman, “Fronts in large marine ecosystems,” Prog. Oceanogr. 81(1-4), 223–236 (2009). [CrossRef]
68. C. Belzile, C. S. Roesler, J. P. Christensen, N. Shakhova, and I. Semiletov, “Fluorescence measured using the WETStar DOM fluorometer as a proxy for dissolved matter absorption,” Estuar. Coast. Shelf Sci. 67(3), 441–449 (2006). [CrossRef]
69. C. A. Stedmon, S. Markager, and R. Bro, “Tracing dissolved organic matter in aquatic environments using a new approach to fluorescence spectroscopy,” Mar. Chem. 82(3-4), 239–254 (2003). [CrossRef]
70. A. Bricaud, H. Claustre, and J. Eacute, “Ras, and K. Oubelkheir, “Natural variability of phytoplanktonic absorption in oceanic waters: influence of the size structure of algal populations,” J. Geophys. Res. Oceans 109, C11010 (2004).
71. S. R. Roy, F. Blouin, A. Jacques, and J. C. Therriault, “Absorption properties of phytoplankton in the Lower Estuary and Gulf of St. Lawrence (Canada),” J. Canadien Sci. Halieut. Aquat. 65(8), 1721–1737 (2008). [CrossRef]
72. A. Bricaud, M. Babin, A. Morel, and H. Claustre, “Variability in the chlorophyll‐specific absorption coefficients of natural phytoplankton: Analysis and parameterization,” J. Geophys. Res. Oceans 100(C7), 13321–13332 (1995). [CrossRef]
73. A. Bricaud, C. Mejia, D. Blondeau-Patissier, H. Claustre, M. Crepon, and S. Thiria, “Retrieval of pigment concentrations and size structure of algal populations from their absorption spectra using multilayered perceptrons,” Appl. Opt. 46(8), 1251–1260 (2007). [CrossRef] [PubMed]
74. A. M. Ciotti, M. R. Lewis, and J. J. Cullen, “Assessment of the relationships between dominant cell size in natural phytoplankton communities and the spectral shape of the absorption coefficient,” Am. Soc. Limnol. Oceanog. 47(2), 404–417 (2002). [CrossRef]
75. I. Laurion, F. Blouin, and S. Roy, “The quantitative filter technique for measuring phytoplankton absorption: interference by MAAs in the UV waveband,” Limnol. Oceanogr. Methods 1, 1–9 (2011). [CrossRef]
76. J. R. Moisan, T. A. H. Moisan, and M. A. Linkswiler, “An inverse modeling approach to estimating phytoplankton pigment concentrations from phytoplankton absorption spectra,” J. Geophys. Res. Oceans 116, C09018 (2011).
