Open Access
Add to BibSonomy
Abstract
Knowing variations of phytoplankton community characteristics is of great significance to many marine ecological and biogeochemical processes in oceanography and related fields research. Satellite remote sensing provides the only viable path for continuously detecting phytoplankton community characteristics in the large-scale spatial areas. However, remote sensing approaches are currently hindered by limited understanding on reflectance responses to variations from phytoplankton community compositions and further do not achieve a true application by satellite observations. Here we analyze in situ observation data sets from three cruises in a dynamic marine environment covering those coastal water areas in the marginal seas of the Pacific Northwest (Bohai Sea, Yellow Sea, and East China Sea). The size/species-specific phytoplankton assemblages can be quantitatively defined by the high-performance liquid chromatography (HPLC)-derived phytoplankton pigments and customized diagnostic pigment analysis, as well as a matrix factorization “CHEMTAX” program. Therein, note that a suit of updated weight values for diagnostic pigments are proposed with better performance than others. The above-mentioned size/species-specific phytoplankton assemblages include three size classes, i.e., micro-, nano-, and picoplankton, and eight species typically existing in the investigated water areas. Relationship analysis illustrates us that relatively close and robust models can be established to associate three size-specific and four dominant species-specific phytoplankton biomasses with the total chlorophyll a. Those models are then applied to the Geostationary Ocean Color Imager (GOCI) images for the whole 2015 year, which generated annual mean distributions of size/species-specific phytoplankton biomasses. The current study represents a meaningful attempt to achieve the satellite remote-sensing retrievals on the phytoplankton community composition, especially the species-specific phytoplankton biomass in the study region.
© 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
As the bottom of the food chain, marine phytoplankton assemblages plays very important roles in the aquatic ecosystem [1–3], and also makes significant influence on the biogeochemical processes, such as carbon cycling [2,4,5], nitrogen fixing [6–8], and silica dynamics [9,10]. Therefore, exploring and grasping variations of phytoplankton assemblages is of great significance to many aspects in the field of oceanography research.
Quantifying variations of phytoplankton assemblages is a complex and difficult task, since the phytoplankton assemblages are a huge “biosystem” with hundreds and even thousands of phytoplankton species from the perspective of marine biology [11–13]. Indeed, a measurement approach by in situ collected water samples and subsequent laboratory analysis can provide accurate and detailed information on the compositions and characteristics of phytoplankton assemblages [14–16]. However, the use of the in situ observation approach would show distinct limitations on spatial and temporal sampling, as well as labor intensive and time consuming with low efficiency [17]. Remote sensing technique will exert its incomparable advantages in large-scale, periodic, and conventional monitoring to marine environment changes [18,19].
Remote sensing monitoring of variations of phytoplankton assemblages has been usually carried out by detecting the chlorophyll a concentration in marine waters [20,21]. Although it can explain well the total phytoplankton biomass, note that the only chlorophyll a indicator is not sufficient to quantify comprehensive and fine characteristics of phytoplankton assemblages, such as phytoplankton species compositions and changes. Fortunately, this has received increasing concerns in recent years, and more and more studies on the remote sensing of fine characteristics of phytoplankton assemblages were being done [22].
Considering that the cell sizes are key characteristics of phytoplankton community, and are closely related to many marine biogeochemical cycles [23–25], the phytoplankton assemblages have been conventionally divided into three size classes, namely so-called phytoplankton size classes (PSCs), microplankton (> 20 μm), nanoplankton (2-20 μm), and picoplankton (< 2 μm) [25]. Detecting the PSCs can be thus regarded as an important route to investigate fine characteristics of phytoplankton assemblages in marine waters. Some remote sensing algorithms have been developed to estimate the PSCs, especially in the past decade, such as abundance-based approaches [22,26–31], and spectra-based approaches [32–36]. Those proposed algorithms functioned well, especially in open ocean waters, and yet note that very limited studies aim at optically complex turbid coastal waters. Anyhow, the remote sensing estimation of the PSCs is essentially to detect three sub-concentrations of chlorophyll a for micro-, nano-, and picoplankton in marine waters, which will refine the spatiotemporal distribution characteristics of the phytoplankton biomass to some extent.
On the other hand, quantifying concentrations of phytoplankton diagnostic pigments or species in marine waters is another key approach to explore and document fine variations of phytoplankton assemblages. Several up-to-date studies showed pioneering efforts on remote sensing estimations to phytoplankton diagnostic pigments or species concentrations in marine waters. For instance, Chase et al. (2017) [37] made use of in situ hyperspectral remote-sensing reflectance measurements to develop an inversion algorithm for detecting phytoplankton pigments, including chlorophylls a, b, c1 + c2, and the photoprotective carotenoids, which defined phytoplankton pigment absorption as a sum of Gaussian functions. Moisan et al. (2011) [38] utilized phytoplankton absorption spectra and HPLC pigment observations to extract pigment-specific absorption in high correlations, and further used the derived pigment-specific absorption in a linear inverse calculation to estimate the various phytoplankton pigments, with good correlations (R2 > 0.5) for 7 pigments in the 18 pigment fields. In the northern South China Sea, Pan et al. (2013) [39] developed empirical algorithms for estimating species specific cell abundance, including the cell abundances of prochlorococcus, synechococcus, and pico-eukaryotes, which were related to the total chlorophyll a and zeaxanthin pigment concentrations. Kramer et al. (2018) [40] evaluated and refined a bio-optical algorithm to discriminate diatom dominance from other phytoplankton in the surface ocean, by using the relationship between ratios of remote sensing reflectance and chlorophyll a concentration.
Clearly, the investigations on the estimation of phytoplankton diagnostic pigments or species concentrations have gradually received increasing concerns, and yet limitations are distinct based on our knowledge on the current research works. Owing to a lack of those potentially “fine” spectral absorption characteristics on which the inversion algorithms were established, there still exists a certain gap that cannot achieve a true application by satellite observations. Meanwhile, the phytoplankton species or pigments that can be identified by those existing algorithms are still limited, which inadequately supports us to know the phytoplankton community structures well. Note that optical complexity in those coastal turbid waters may present potential challenges and limitations in extracting phytoplankton species information. For instance, the non-algal particles may preclude the ability to retrieve information about the phytoplankton groups, when they are a large contribution to total absorption. On the other hand, phytoplankton community structures in turbid coastal waters may show distinct differences from those in open ocean waters because of terrigenous discharge and interference from nonplankton constituents [41–43].
In the current study, we collect an in situ observed data set from five cruises in the marginal seas of northwest Pacific Ocean, covering those optically complex turbid coastal waters (Bohai Sea, Yellow Sea, and East China Sea). Data of phytoplankton pigments, including the total chlorophyll a and other nineteen diagnostic pigments, are measured for analyzing the characteristics of phytoplankton community structures in the region. Our main motive is to explore and potentially document the relationships between those size/species-specific phytoplankton and the total biomass, and to form remote sensing models of estimating the PSCs and typically dominant phytoplankton species contents in the study regions. Note that those proposed models should be truly applied into satellite image for addressing the spatiotemporal patterns of phytoplankton community structures. At last, necessary discussions on some key points of our work are appended.
2. Materials and methods
2.1. Our study regions
Our study regions are those marginal seas of the Pacific Northwest, including Bohai Sea (BS), Yellow Sea (YS), East China Sea (ECS). BS covers a total area of roughly 77,000 km2, and is the only semi-closed inner sea surrounded by Liaodong peninsula, North China, and Shandong peninsula. BS has an average water depth of about 18 m and a maximum depth up to 70 m [44]. Connecting to BS, YS locates between the Chinese mainland and the Korean peninsula, and covers an area of about 380,000 km2. BS can be divided into two parts, namely North Yellow Sea and South Yellow Sea (Fig. 1). Compared with BS and YS, ECS is a larger sea region that covers an approximate 770,000 km2 area, from the Yangtze River mouth to the South China Sea [45]. These water regions are typically optically complex turbid coastal waters, with a chlorophyll a range of about 0.01 - 15 mg m−3, and have become highly productive and polluted with green macroalgae bloom in recent years [46,47]. During a two-year period of 2016 to 2017, we carried out three cruise surveys covering many stations in BS, YS, and ECS. At each station, surface water samples were collected with Niskin bottles mounted on a CTD/rosette system, and then used for High Performance Liquid Chromatography (HPLC) analysis. All obtained data were quality controlled by the data statistical rules of Aiken et al. (2009) [48].
2.2. Determination of size/species-specific phytoplankton assemblages
Phytoplankton diagnostic pigments and the total chlorophyll a are important inputs to determine size/species-specific phytoplankton biomass. To determine those pigment concentrations, water samples were first filtered onto 47 mm Whatman GF/F glass fiber filters, and then immediately put into liquid nitrogen for storage for subsequent analysis. Concentrations of the total chlorophyll a and nineteen diagnostic pigments were measured by reversed-phase HPLC with a Zorbax Eclipse XDB-C8 column [49,50].
As well as the total chlorophyll a, seven diagnostic pigments, including fucoxanthin, peridinin, 19′-hexanoyloxyfucoxanthin, 19′-butanoyloxyfucoxanthin, alloxanthin, chlorophyll b and divinyl chlorophyll b, and zeaxanthin, were used to analyze the size-specific phytoplankton biomass, namely phytoplankton size classes (PSCs). Following Vidussi et al. (2001) [51] and Uitz et al. (2006) [31], this study used diagnostic pigments analysis (DPA) to estimate the fractions of chlorophyll a in different size classes (i.e., micro-, nano-, and picoplankton). The HPLC-derived total chlorophyll a concentration (C) can be simulated by the weighted sum of seven diagnostic pigments (Cw) as follows:
where wi is the weight, corresponding to the i-th diagnostic pigment Pi (in mg m−3). The accurate determination about those weight values will be shown in the Section 3.Vidussi et al. (2001) [51] and Uitz et al. (2006) [31] attributed fucoxanthin and peridinin to microplankton. However, Devred et al. (2011) [29] apportioned part of the fucoxanthin pigment to nanoplankton, such as prymnesiophytes and chrysophytes [52]. Accordingly, the fraction of microplankton (Fm) can be calculated as follows:
where P1, n can be estimated using an equation ofwhere the q1 and q2 values are 0.356 and 1.190 [22], respectively.Based on the approach of Vidussi et al. (2001) [51] and Uitz et al. (2006) [31] that attributed 19′-hexanoyloxyfucoxanthin, 19′-butanoyloxyfucoxanthin, and alloxanthin to the pool of nanoplankton, Brewin et al. (2010) [26] proposed and apportioned part of 19′-hexanoyloxyfucoxanthin and 19′-butanoyloxyfucoxanthin to the picoplankton pool at low chlorophyll a concentration (< 0.08 mg m−3) level. The fraction of nanoplankton (Fn) can be derived as below:
And the updated calculation on the fraction of picoplankton (Fp) is like below:
At last, the concentrations of three size-specific phytoplankton assemblages can be derived as follows:
where the subscripts for C, namely m, n, and p, correspond to micro-, nano-, and picoplankton, respectively.Another work is to determine the species-specific phytoplankton biomass. Based on previous studies [53–55], there existed eight relatively dominant species in the study regions, including diatoms, dinoflagellates, chlorophytes, cryptophytes, chrysophytes, cyanobacteria, rasinophytes, and prynesiophytes. The concentrations of these species could be derived by using a matrix factorization “CHEMTAX” program. Eleven diagnostic pigments, including fucoxanthin (Fuco), peridinin (Per), 19′-butanoyloxyfucoxanthin (19′-but), 19′-hexanoyloxyfucoxanthin (19′-hex), neoxanthin (Neo), prasinoxanthin (Pra), violaxanthin (Vio), alloxanthin (Allo), lutein (Lut), zeaxanthin (Zea), and chlorophyll b and divinyl chlorophyll b (Chlb), were used as inputs, as well as the total chlorophyll a (Chl-a). Following Mackey et al. (1996) [56], Furuya et al. (2003) [57], Kong (2012) [54] studies, an initial pigment ratio matrix, as shown in Table 1, was also used as inputs in the “CHEMTAX” program. The final abundance estimates were averaged from the best 10% outputs (the smallest residual) of totally 60 times’ run.
Table 1. Pigment ratios to the total chlorophyll a for eight taxonomic groups in the study regions based on CHEMTAX analysis.
2.3. GOCI image data
The current study made use of GOCI image data to detect the spatiotemporal variations of phytoplankton community compositions, which could record eight image data per day from 00:15 to 07:45 GMT with a nominal spatial resolution of 500 m. From the Korea Ocean Satellite Center (http://kosc.kiost.ac.kr/eng/p10/kosc_p11.html), GOCI Level-1B data (calibrated top-of-atmosphere radiance) has been collected with a total of 2898 images during the whole 2015 year. These raw data were cropped to BS, YS, and ECS, and then processed to Level-2 data products by using the GOCI Data Processing System (GDPS, version 1.4) [58], and its atmospheric correction and default parameterization. In this study, the chlorophyll a concentration product was processed and used, which is the inputs of those developed models (see below) for deriving size/species-specific phytoplankton assemblages.
2.4. Mathematical analysis method
In this study, the Leave-One-Out (LOO) method is mainly used for diagnostic pigment analysis (DPA), and also for analyzing fitting relationships between the size/species-specific and the total chlorophyll a concentrations, and serves for the model development and validation. Two processes are needed in the LOO method: 1) using randomly one sample from the total of samples (n) for validation, and carrying out the function fitting with those leaving n-1 samples for establishing a relationship model, and 2) iterating the above process at n times until each sample has been used as validation once, and finally forming a mathematical model with n-times averaged parameters. The performances of mathematical function models in this study can be assessed with comparisons between estimated data values and those in situ measured ones. Coefficient of determination (R2) is used to show how well the estimated values agree with the in situ measured ones. Meanwhile, other three indicators, including mean absolute percentage error (MAPE), root mean squared error (RMSE), and Mean ratio, are quantitively used for models’ performance assessment. They can be calculated as follows [59]:
where xi and yi represent those in situ observed and estimated values, respectively.3. Results
3.1. Diagnostic pigment analysis
By means of diagnostic pigment analysis on the in situ collected data set, this study calculates the fractions of the chlorophyll a in the three phytoplankton size classes (i.e., Fm, Fn and Fp), where the chlorophyll a is conventionally used as a delegate for the phytoplankton biomass. The total phytoplankton biomass (total chlorophyll a concentration, C) is reconstructed by the product of the seven typical diagnostic pigments mentioned above and their corresponding weights. Importantly, how to accurately determine the weight values for these diagnostic pigments becomes a purpose of the DPA in the present study. According to Eq. (1), a multiple linear regression analysis is carried out on our HPLC data set by the LOO method. As such, the DPA produces a robust performance with a determination coefficient (R2) of 0.970 (p<0.001) and a MAPE of 12.2% (see Fig. 2). The obtained weight values are 2.20, 0.95, 1.33, 4.22, 3.57, 0.74, and 1.13 for fucoxanthin, peridinin, 19′-hexanoyloxyfucoxanthin, 19′-butanoyloxyfucoxanthin, alloxanthin, chlorophyll b and divinyl chlorophyll b, and zeaxanthin, respectively.
Fig. 2 Comparison between the in situ measured C and DPA-modeled Cw by our collected data set during five cruise surveys in the study areas (both in log space). Note that the parameter Cw means the weighted sum of seven diagnostic pigments.
3.2. Statistical distribution of phytoplankton community structure
Statistical distributions on phytoplankton community structure are presented here by the PSCs and species-specific phytoplankton concentrations. Figure 3(a) shows the frequency distribution of fractions of micro-, nano- and picoplankton in this study. Apparently, the picoplankton contributes the least to the total phytoplankton biomass, with an average contribution of ~23% (mean concentration, 0.233 mg m−3) and about 20% contribution in nearly half of the total samples, compared with the micro- and nanoplankton. Note that the micro- and nanoplankton show similar contributions to the total phytoplankton biomass with about 37% (mean concentration, 0.362 mg m−3) and 40% (mean concentration, 0.404 mg m−3), respectively. Eight typical phytoplankton species, including diatoms, dinoflagellates, chlorophytes, cyanobacteria, cryptophytes, prasinophytes, chrysophytes, and prynesiophytes, mostly exist in the investigated water bodies [53–55]. In general, diatoms contributes the most (mean contribution ratio, approximately 32%, Fig. 3(b)) to the total phytoplankton biomass, than other species. Cryptophytes shows the second most contribution with an average ratio of about 21%. Chlorophytes, cyanobacteria, prynesiophytes, and prasinophytes shows similar contributions to the total biomass between each other, all with a mean contribution of approximately 10%. Note that dinoflagellates and chrysophytes roughly presents the least contribution with an aggregated mean ratio below 10%.
Fig. 3 Histograms on the frequency distributions of percentages of micro-, nano- and picoplankton (A) and species-specific phytoplankton concentrations (B).
3.3. Relationships between the size/species-specific and the total phytoplankton biomass
By our collected data sets, the current study analyzes the relationships between the size-specific and the total chlorophyll a concentrations, where the size-specific chlorophyll a concentrations, namely Cm, Cn and Cp, can be obtained according to Eqs. (2)–(8). According to the LOO method, the collected samples in the data set are all utilized in the process of the relationship calibration and validation. Many mathematical functions are attempted to establish their relationships, including linear, quadratic, power, exponential, logarithmic, etc., with a finding that the power function performs best than others after comparisons (not show here for simplicity). The detailed function form is as follows:
where Cpsc represents the chlorophyll a concentrations of the micro-, nano-, and picoplankton size classes; k1 and k2 are the region-specific empirical parameter and power-law exponent, respectively. The LOO method is utilized to train the power law model and test its performance. As shown in Table 2, relatively high goodness of fit can be obtained, with R2 of 0.797, 0.754, and 0.678 (p<0.001) for those relationships between micro-, nano-, and picoplankton, with the total phytoplankton biomass. The derived MAPE values during the process of model training are approximately 40-50%. Note that the finally formed model (including its parameters and accuracy indicator) is very representative catering to our used observation data, because the k1, k2, R2, MAPE, and RMSE, shown in Table 2, are those averaged values after many iterations by the LOO method. Fortunately, the validation errors are roughly comparable to that in the model training (Fig. 4). This indicates the robustness of the model and potential applicability in the study areas.Table 2. Model parameters and calibration accuracy for the relationships between size-specific and the total phytoplankton biomass (i.e., the chlorophyll a of three size classes vs. the total chlorophyll a). Note that the corresponding levels of significance are all p<0.001 for the three models.
Fig. 4 Scatter plots of the estimated size-specific phytoplankton chlorophyll a concentrations versus those in situ measurements, which is used for model validation by the LOO method.
Similar to that of phytoplankton size classes, this study also attempts to establish the relationships between the species-specific and the total phytoplankton biomass. After testing many mathematical functions (not shown here), we select the simple linear model, namely,
where Calgae represents the concentrations of each algae species, and k1 and k2 are the region-specific parameters, since it performs best than others. Table 3 shows the derived model parameters and calibration accuracies. Note that the shown relationships are just for four algae species, namely diatoms, dinoflagellates, chlorophytes, and chrysophytes. Unfortunately, other algae species do not show very close relationships with the total chlorophyll a concentration (not shown here). The calibrated R2 are relatively high for dinoflagellates (0.947, p<0.001), chlorophytes (0.606, p<0.001), and diatoms (0.808, p<0.001), while a slightly low R2 is obtained for chrysophytes (0.291), yet still with a very high level of significance (p<0.001) and a relatively low and acceptable MAPE (47.7%) by the model calibration with the LOO method (Table 3). Those predictive errors for the four formed models are shown in Fig. 5, which are derived through the validation samples by using the LOO method. About 40%-60% are obtained for the MAPE of these models, and corresponding RMSE are all 0.04 mg m−3 for dinoflagellates, chlorophytes, and chrysophytes, along with 0.17 mg m−3 for diatoms. Honestly, the predictive accuracies derived by the proposed models may be not very high, yet currently acceptable to some extent, considering that the existing remote sensing estimations are still limited specific to the detailed algae concentrations.Table 3. Model parameters and calibration accuracy for the relationships between species-specific and the total phytoplankton biomass (i.e., algae species concentrations vs. the total chlorophyll a). Note that the corresponding levels of significance are all p<0.001 for those models.
Fig. 5 Scatter plots of the estimated species-specific phytoplankton chlorophyll a concentrations versus those in situ measurements, which is used for model validation by the LOO method.
3.4. Spatial distribution of the size/species-specific phytoplankton biomass
This study applies the above established models to our collected 2898 GOCI satellite images during the whole 2015 year. By taking the BS and YS regions for the demonstration applications, the PSCs (micro-, nano-, and picoplankton) and four typical algae concentrations (dinoflagellates, chlorophytes, diatoms, and chrysophytes) are derived to produce the one-year averaged spatial distributions (Figs. 6 and 7). In general, the microplankton shows the highest content than that of nanoplankton and picoplankton, especially in the nearshore regions. For instance, the Cm in most areas of the BS and NYS can reach above 1.5 mg m−3, also in the nearshore zones along the coastlines. The nanoplankton shows the middle level in the three PSCs, nearly throughout the whole regions. The Cn in the central regions of the YS, approximately 0.8 mg m−3, is relatively lower than that in other regions (mainly near 1 mg m−3). With the lowest content than the micro- and nanoplankton, the picoplankton roughly distributes in the range of 0.3 - 0.5 mg m−3 for the central regions of the YS, when about 0.8 mg m−3 for the BS, NYS, and those nearshore waters.
Fig. 6 Annual mean distributions of the concentrations for the PSCs, derived from GOCI satellite measurements in 2015.
Fig. 7 Annual mean distributions of species-specific phytoplankton biomass derived from GOCI satellite measurements in 2015.
The distribution of different phytoplankton species is generally consistent with that of PSC, because their concentrations have a significant relationship with chlorophyll a. We can see it clearly that Diatoms is dominant algae of our study area and the concentration of Chrysophytes and Prymnesiophytes are very low (< 0.4 mg m−3). However, different algae models have different precision, which we discussed before (section 3.2). Compared to other models, models for Cyanobacteria and Prymnesiophytes may have relatively low accuracies (so their images are not shown here), which may needs further research work for obtaining accurate outputs on Cyanobacteria and Prymnesiophytes in future.
3.5. Comparison with previously published diagnostic pigment weights
The DPA is the important basis to quantify the phytoplankton community structure. Four suits of DPA-derived weights for diagnostic pigments, as shown in Table 4, have been currently reported and utilized in the previous studies [22,28,31,60]. Based on our in situ observed HPLC data, we show the performances by using those previously reported diagnostic pigment weights (Fig. 8). Although these previous schemes present relatively high goodness of fit, with R2 values larger than 0.95 (p<0.001), there all exists a distinct underestimation to a greater and lesser extent. Quantitatively speaking, their derived MAPE are relatively high with more than 30% for [31,60] schemes, and yet a certain improvement for [22,28] schemes, still with about 20% for MAPE) (Figs. 8(a)–8(d)). By contrast, high degree of fitting (R2 = 0.97, p < 0.001) and low predictive deviation (MAPE = 12%) (Fig. 8(e)) assure that our proposed scheme on phytoplankton pigment weights are more suitable for the study regions.
Table 4. Diagnostic pigment weights reported in the previous studies and our DPA-derived weights in this study.
Fig. 8 Comparisons between the previously reported diagnostic pigment weights and that in this study, based on our in situ observed HPLC data in the study areas.
The above comparison has demonstrated that our proposed suit of weights for diagnostic pigments perform better than that in other previous studies for the study water regions. Even so, we still need to test whether different mathematical fitting methods play an important role in the determination of those weights or not. We compare the weight parameters obtained by three common approaches, namely Random-sample Fitting (RF), Order-sample Fitting (OF), and Leave-one-out (LOO) methods. RF approach: 70% data samples are randomly selected for calibration and the remaining samples are used as validation data set. OF: HPLC data samples are sorted by the chlorophyll a, and then three samples are divided into a subgroup, of which two are classified into calibration data set, and the rest is used for validation data set. As shown in Table 5, different methods produce very similar weights for diagnostic pigments, with the consistent R2 of 0.97 (p<0.001) and similar validation errors. Therefore, the above results respond to our concern that different fitting methods can provide us approximately consistent weights.
Table 5. Diagnostic pigment weights obtained by three fitting methods (RF, OF, and LOO) and corresponding accuracy assessment (MAPE with a unit of %; RMSE, unit mg m−3; Mean ratio, MR). Note that all shown R2 are with a very high significance of level (p <0.001).
4. Discussion
4.1. Rationality and limitation of satellite-derived products on size/species-specific phytoplankton biomass
This study shows the satellite-derived products on phytoplankton assemblages, and particularly focuses on the detection of their community structure. With an important finding that the chlorophyll a can be used as an index of not only the biomass but also the community structure of phytoplankton in the study region, the least-square fitting relationships can be established to estimate the PSCs and major phytoplankton species (diatoms, dinoflagellates, chlorophytes, and chrysophytes) concentrations from the chlorophyll a. This successful quantification on the PSCs and major phytoplankton species concentrations in this study is essentially based on an existing fact that they can be separately expressed as a part of the total chlorophyll a, also more importantly, have some regular dynamic changes in the total phytoplankton biomass, like reported in previous studies [53,55,61]. These regularities can be just described well by our proposed relationships between the size/species-specific with the total phytoplankton biomass in the study region. Note that the chlorophyll a concentration is the only input to those refined relationships, which has been accurately produced by GOCI satellite data (with high spatial and temporal resolution) and credible operational retrieval algorithm. Also, relatively low and acceptable validation errors (Figs. 4 and 5) demonstrate the reliability and rationality of those developed relationships and derived satellite products.
In fact, the natural condition existing in our study areas, i.e., the size/species-specific phytoplankton assemblages distribute with a rough regularity in the total phytoplankton biomass in a statistical quantitative term, is not unique. Several previous studies have also reported similar cases in those investigated water regions. At a large-scale global scope, Hirata et al. (2011) [30] have defined the relationships between chlorophyll a and a total of ten pigment groups, including three phytoplankton size classes and seven phytoplankton functional types, based on a large quantity of in situ observed data set, and demonstrated that the phytoplankton community shift could be inferred by the variation in the total phytoplankton biomass. In the northeast coast of the United States, Pan et al. (2011) [62] established regression relationships between concentrations of phytoplankton functional groups and the total cell counting abundance, and found that some major phytoplankton functional groups, such as diatoms, cryptophytes, dinoflagellates, and cyanobacteria, regularly increased along with increase in the cell counts. Brewin et al. (2010, 2014) [26,27] showed and quantified the regular variations of the raw size-specific chlorophyll a values along with the total chlorophyll a in the Atlantic Ocean. Brotas et al. (2013) [63] depicted scatter plots to present close relationships between the PSCs and the total chlorophyll a in the eastern Atlantic Ocean. Similar reports could be also found in some other literatures [57,64–66].
Nevertheless, the current work cannot derive more phytoplankton species information in the water bodies, except those dominant species, including diatoms, dinoflagellates, chlorophytes, and chrysophytes. Large predictive errors would be produced for other phytoplankton populations, such as approximately 140% for prasinophytes, 130% for cryptophytes, and even beyond 300% for cyanobacteria in the predictive MAPE values (not shown in figure here), possibly indicative of potential out-of-order changes for these species in the total phytoplankton biomass. Thus, more work in future should be focused on the extraction to those non-dominant phytoplankton assemblages for comprehensive grasp to the changes of phytoplankton community structures.
4.2. Implications for marine environmental monitoring and future work
Detecting phytoplankton assemblages’ variations is very important aspects of marine environment monitoring, since phytoplankton community significantly makes an effect on marine environment changes [20,42,44,63,67]. Currently, using a chlorophyll a indicator is still the main means to characterize phytoplankton assemblages’ variations [25,31,51]. Although it has been widely and conventionally applied for a delegate of the total phytoplankton biomass [68,69], the chlorophyll a indicator hardly addresses more on phytoplankton assemblages. While existing approaches of estimating the phytoplankton assemblage’s information have been demonstrated successful in the global ocean, they have not been previously demonstrated in optically complex coastal regions. We have shown here that it is possible to retrieve size and taxonomic group information in turbid waters. Specifically, this study shows a remote sensing approach for detailed monitoring on distribution characteristics of phytoplankton assemblages, and focuses on those sub-phytoplankton biomasses on size/species-specific phytoplankton populations. Instead of the total chlorophyll a concentration, our study provides the satellite-derived spatiotemporal distributions of micro-, nano-, and picoplankton contents, and four typical algae concentrations, including diatoms, dinoflagellates, chlorophytes, and chrysophytes. These satellite-derived products enrich and improve the detection of phytoplankton assemblages, and greatly serve for monitoring marine environment changes. Thus, we recommend that future ocean monitoring plans include those satellite-derived fine sub-phytoplankton biomass products to aid in the interpretation of spatial and temporal patterns of marine environment changes.
5. Conclusion
From two perspectives of the sizes and species of phytoplankton, this study observes the characteristics of phytoplankton assemblage compositions in a dynamic coastal environment (BS, YS, and ECS), by using the reversed-phase HPLC technique. With self-contained in situ observation data sets, we improve key weights for phytoplankton diagnostic pigments that support an accurate in situ measurement-based quantification on the size-specific phytoplankton assemblages in the study region, also namely the PSCs that include micro-, nano-, and picoplankton types. Meanwhile, eight species of phytoplankton, typically existing in our investigated water areas, can be produced by means of a “CHEMTAX” program, with eleven diagnostic pigments as inputs, together with the total chlorophyll a. Of note, our study has demonstrated that unique close relationships exist between size-specific phytoplankton assemblages and four dominant phytoplankton species, including diatoms, dinoflagellates, chlorophytes, and chrysophytes, with the total phytoplankton biomass in the study region. Those obtained findings of this study allow us to truly achieve the satellite remote sensing retrievals to document the spatiotemporal dynamics on the phytoplankton community composition. Overall, the present study illustrates the potential for retrieving fine characteristics of phytoplankton assemblages from satellite ocean color observations, other than the total chlorophyll a, and provides in-depth knowledge on phytoplankton assemblages.
Funding
National Natural Science Foundation of China (41876203, 41576172); National Key Research and Development Program of China (2016YFC1400901, 2016YFC1400904); Jiangsu Provincial Programs for Marine Science and Technology Innovation (HY2017-5); Jiangsu Six Talent Summit Project (JY-084); Qing Lan Project, College Students Practice Innovation Training Program of Nuist (201710300027Z); NSFC Open Research Cruise (NORC2018-01), funded by Shiptime 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 also thank the two reviewers and the editor for their suggestions and comments that help improve the quality of the manuscript.
References
1. S. Dutkiewicz, A. E. Hickman, O. Jahn, S. Henson, C. Beaulieu, and E. Monier, “Ocean colour signature of climate change,” Nat. Commun. 10(1), 578 (2019). [CrossRef] [PubMed]
2. P. G. Falkowski, M. E. Katz, A. H. Knoll, A. Quigg, J. A. Raven, O. Schofield, and F. J. R. Taylor, “The Evolution of Modern Eukaryotic Phytoplankton,” Science 305(5682), 354–360 (2004). [CrossRef] [PubMed]
3. 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]
4. J. I. House, I. C. Prentice, and C. L. Quéré, “Maximum impacts of future reforestation or deforestation on atmospheric CO2,” Glob. Change Biol. 8(11), 1047–1052 (2002). [CrossRef]
5. C. L. Sabine, R. M. Key, R. A. Feely, and D. Greeley, “Inorganic carbon in the Indian Ocean: Distribution and dissolution processes,” Global Biogeochemical Cycles 16, 15–11–15–18 (2002). [CrossRef]
6. N. S. Agawin, S. Rabouille, M. J. Veldhuis, L. Servatius, S. Hol, H. M. van Overzee, and J. Huisman, “Competition and facilitation between unicellular nitrogen‐fixing cyanobacteria and non—nitrogen‐fixing phytoplankton species,” Limnol. Oceanogr. 52(5), 2233–2248 (2007). [CrossRef]
7. T. H. Mague, F. C. Mague, and O. Holmhansen, “Physiology and chemical composition of nitrogen-fixing phytoplankton in the central North Pacific Ocean,” Mar. Biol. 41(3), 213–227 (1977). [CrossRef]
8. M. M. Mills and K. R. Arrigo, “Magnitude of oceanic nitrogen fixation influenced by the nutrient uptake ratio of phytoplankton,” Nat. Geosci. 3(6), 412–416 (2010). [CrossRef]
9. S. Tavernini, E. Pierobon, and P. Viaroli, “Physical factors and dissolved reactive silica affect phytoplankton community structure and dynamics in a lowland eutrophic river (Po river, Italy),” Hydrobiologia 669(1), 213–225 (2011). [CrossRef]
10. T. Wakatsuki and T. Masunaga, “Dissolved silica dynamics and phytoplankton population in Citarum watershed, Indonesia,” J. Food Agric. Environ. 7, 132 (2009).
11. M. Hoppenrath, M. Elbrächter, and G. Drebes, “Marine Phytoplankton. Selected microphytoplankton species from the North Sea around Helgoland and Sylt,” J. Phycol. 46(3), 622–625 (2010). [CrossRef]
12. E. Litchman, D. T. P. Paula, C. A. Klausmeier, M. K. Thomas, and K. Yoshiyama, “Linking traits to species diversity and community structure in phytoplankton,” Hydrobiologia 653(1), 15–28 (2010). [CrossRef]
13. C. S. Reynolds, J. Padisák, and U. Sommer, “Intermediate disturbance in the ecology of phytoplankton and the maintenance of species diversity: a synthesis,” Hydrobiologia 249(1-3), 183–188 (1993). [CrossRef]
14. B. C. Booth, “Estimating cell concentration and biomass of autotrophic plankton using microscopy,” in Handbook of Methods in Aquatic Microbial Ecology (Lewis, 1993), pp. 199–205.
15. E. MacIsaac and J. G. Stockner, “Enumeration of phototrophic picoplankton by autofluorescence microscopy,” in Handbook of Methods in Aquatic Microbial Ecology (Lewis, 1993), pp. 187–197.
16. C. M. Yentsch and P. K. Horan, “Cytometry in the aquatic sciences,” Cytometry 10(5), 497–499 (1989). [CrossRef] [PubMed]
17. H. Sosik, S. Sathyendranath, J. Uitz, H. Bouman, and A. Nair, “In situ methods of measuring phytoplankton functional types,” in Phytoplankton Functional Types from Space (International Ocean Colour Coordinating Group, 2014).
18. T. Platt, C. Fuentes-Yaco, and K. T. Frank, “Marine ecology: Spring algal bloom and larval fish survival,” Nature 423(6938), 398–399 (2003). [CrossRef] [PubMed]
19. R. P. Wilson, D. Grémillet, J. Syder, M. A. Kierspel, S. Garthe, H. Weimerskirch, C. Schäfer-Neth, J. A. Scolaro, C.-A. Bost, J. Plötz, and D. Nel, “Remote-sensing systems and seabirds: their use, abuse and potential for measuring marine environmental variables,” Mar. Ecol. Prog. Ser. 228, 241–261 (2002). [CrossRef]
20. R. D. Moral, “Vegetation ordination of subalpine meadows using adaptive strategies,” Can. J. Bot. 61(12), 3117–3127 (1983). [CrossRef]
21. J. E. O’Reilly, M. Stéphane, M. B. Greg, D. A. Siegel, K. L. Carder, S. A. Garver, K. Mati, and M. C. Charles, “Ocean color chlorophyll algorithms for SeaWiFs,” J. Geophys. Res. 103, 24937–24953 (1998).
22. 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]
23. S. Chisholm, Phytoplankton size, primary productivity and biogeochemical cycles in the sea (Plenum Press, 1992).
24. C. Le Quéré and S. Pesant, “Plankton functional types in a new generation of biogeochemical models: integration of plankton abundance data for the evaluation of marine biogeochemical models; Cambridge, United Kingdom, 1–3 October 2008,” Eos Trans, AGU 90(4), 30–31 (2009). [CrossRef]
25. J. M. Sieburth, V. Smetacek, and J. Lenz, “Pelagic ecosystem structure: Heterotrophic compartments of the plankton and their relationship to plankton size fractions 1,” Limnol. Oceanogr. 23(6), 1256–1263 (1978). [CrossRef]
26. 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]
27. R. J. Brewin, S. Sathyendranath, P. K. Lange, and G. Tilstone, “Comparison of two methods to derive the size-structure of natural populations of phytoplankton,” Deep Sea Res. Part I Oceanogr. Res. Pap. 85, 72–79 (2014). [CrossRef]
28. R. J. Brewin, S. Sathyendranath, G. Tilstone, P. K. Lange, and T. Platt, “A multicomponent model of phytoplankton size structure,” J. Geophys. Res. Oceans 119(6), 3478–3496 (2014). [CrossRef]
29. E. Devred, S. Sathyendranath, V. Stuart, and T. Platt, “A three component classification of phytoplankton absorption spectra: Application to ocean-color data,” Remote Sens. Environ. 115(9), 2255–2266 (2011). [CrossRef]
30. 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]
31. J. Uitz, H. Claustre, A. Morel, and S. B. Hooker, “Vertical distribution of phytoplankton communities in open ocean: An assessment based on surface chlorophyll,” J. Geophys. Res. Oceans 111, C08005 (2006).
32. T. Kostadinov, D. Siegel, and S. Maritorena, “Retrieval of the particle size distribution from satellite ocean color observations,” J. Geophys. Res. Oceans 114, C09015 (2009).
33. A. Fujiwara, T. Hirawake, K. Suzuki, and S.-I. Saitoh, “Remote sensing of size structure of phytoplankton communities using optical properties of the Chukchi and Bering Sea shelf region,” Biogeosciences 8(12), 3567–3580 (2011). [CrossRef]
34. Z. Li, L. Li, K. Song, and N. Cassar, “Estimation of phytoplankton size fractions based on spectral features of remote sensing ocean color data,” J. Geophys. Res. Oceans 118(3), 1445–1458 (2013). [CrossRef]
35. E. Organelli, A. Bricaud, D. Antoine, and J. Uitz, “Multivariate approach for the retrieval of phytoplankton size structure from measured light absorption spectra in the Mediterranean Sea (BOUSSOLE site),” Appl. Opt. 52(11), 2257–2273 (2013). [CrossRef] [PubMed]
36. S. Roy, S. Sathyendranath, H. Bouman, and T. Platt, “The global distribution of phytoplankton size spectrum and size classes from their light-absorption spectra derived from satellite data,” Remote Sens. Environ. 139, 185–197 (2013). [CrossRef]
37. A. P. Chase, E. Boss, I. Cetinić, and W. Slade, “Estimation of Phytoplankton Accessory Pigments From Hyperspectral Reflectance Spectra: Toward a Global Algorithm,” J. Geophys. Res. 122, 9725– 9743 (2017).
38. 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. 116, C09018 (2011).
39. X. Pan, G. T. Wong, T.-Y. Ho, F.-K. Shiah, and H. Liu, “Remote sensing of picophytoplankton distribution in the northern South China Sea,” Remote Sens. Environ. 128, 162–175 (2013). [CrossRef]
40. S. J. Kramer, C. S. Roesler, and H. M. Sosik, “Bio-optical discrimination of diatoms from other phytoplankton in the surface ocean: Evaluation and refinement of a model for the Northwest Atlantic,” Remote Sens. Environ. 217, 126–143 (2018). [CrossRef]
41. B. Wang, “Cultural eutrophication in the Changjiang (Yangtze River) plume: History and perspective,” Estuar. Coast. Shelf Sci. 69(3-4), 471–477 (2006). [CrossRef]
42. M. Zhou, Z. Shen, and R. Yu, “Responses of a coastal phytoplankton community to increased nutrient input from the Changjiang (Yangtze) River,” Cont. Shelf Res. 28(12), 1483–1489 (2008). [CrossRef]
43. 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]
44. 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]
45. 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]
46. C. Hu, “A novel ocean color index to detect floating algae in the global oceans,” Remote Sens. Environ. 113(10), 2118–2129 (2009). [CrossRef]
47. L. Qi, C. Hu, M. Wang, S. Shang, and C. Wilson, “Floating algae blooms in the East China Sea,” Geophys. Res. Lett. 44, 11501–11509 (2017).
48. J. Aiken, Y. Pradhan, R. Barlow, S. Lavender, A. Poulton, P. Holligan, and N. Hardman-Mountford, “Phytoplankton pigments and functional types in the Atlantic Ocean: a decadal assessment, 1995–2005,” Deep Sea Res. Part II Top. Stud. Oceanogr. 56(15), 899–917 (2009). [CrossRef]
49. L. Van Heukelem and C. S. Thomas, “Computer-assisted high-performance liquid chromatography method development with applications to the isolation and analysis of phytoplankton pigments,” J. Chromatogr. A 910(1), 31–49 (2001). [CrossRef] [PubMed]
50. M. Zapata, F. Rodríguez, and J. L. Garrido, “Separation of chlorophylls and carotenoids from marine phytoplankton: a new HPLC method using a reversed phase C8 column and pyridine-containing mobile phases,” Mar. Ecol. Prog. Ser. 195, 29–45 (2000). [CrossRef]
51. F. Vidussi, H. Claustre, B. B. Manca, A. Luchetta, and J. C. Marty, “Phytoplankton pigment distribution in relation to upper thermocline circulation in the eastern Mediterranean Sea during winter,” J. Geophys. Res. Oceans 106(C9), 19939–19956 (2001). [CrossRef]
52. S. Roy, C. A. Llewellyn, E. S. Egeland, and G. Johnsen, Phytoplankton Pigments: Characterization, Chemotaxonomy and Applications in Oceanography (Cambridge University Press, 2011).
53. J. Sun, D. Liu, S. Yang, J. Guo, and S. Qian, “The preliminary study on phytoplankton community structure in the central Bohai Sea and the Bohai Strait and its adjacent area,” Oceanol. Limnol. Sin. 33, 461–471 (2002).
54. 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]
55. 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]
56. 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]
57. 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]
58. J. H. Ryu, H. J. Han, S. Cho, Y. J. Park, and Y. H. Ahn, “Overview of geostationary ocean color imager (GOCI) and GOCI data processing system (GDPS),” Ocean Sci. J. 47(3), 223–233 (2012). [CrossRef]
59. D. Antoine, F. d’Ortenzio, S. B. Hooker, G. Bécu, B. Gentili, D. Tailliez, and A. J. Scott, “Assessment of uncertainty in the ocean reflectance determined by three satellite ocean color sensors (MERIS, SeaWiFS and MODIS‐A) at an offshore site in the Mediterranean Sea (BOUSSOLE project),” J. Geophys. Res. 113, C07013 (2008).
60. J. Uitz, H. Claustre, F. B. Griffiths, J. Ras, N. Garcia, and V. Sandroni, “A phytoplankton class-specific primary production model applied to the Kerguelen Islands region (Southern Ocean),” Deep Sea Res. Part I Oceanogr. Res. Pap. 56(4), 541–560 (2009). [CrossRef]
61. J. Sun, D. Liu, J. Bai, H. Gao, and X. Han, “Phytoplankton Community of the Bohai Sea in winter 2001,” Period. Ocean Univ. China 34, 413–422 (2004).
62. X. Pan, A. Mannino, H. G. Marshall, K. C. Filippino, and M. R. Mulholland, “Remote sensing of phytoplankton community composition along the northeast coast of the United States,” Remote Sens. Environ. 115(12), 3731–3747 (2011). [CrossRef]
63. 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]
64. 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,” Limnol. Oceanogr. 47(2), 404–417 (2002). [CrossRef]
65. J. Lin, W. Cao, G. Wang, and S. Hu, “Satellite-observed variability of phytoplankton size classes associated with a cold eddy in the South China Sea,” Mar. Pollut. Bull. 83(1), 190–197 (2014). [CrossRef] [PubMed]
66. G. Wang, W. Cao, G. Wang, and W. Zhou, “Phytoplankton size class derived from phytoplankton absorption and chlorophyll-a concentrations in the northern South China Sea,” Chin. J. Oceanology Limnol. 31(4), 750–761 (2013). [CrossRef]
67. W. Huang, X. Zhu, J. Zeng, Q. Chen, X. Xu, Y. Liao, J. Liu, and P. Du, “Microcosm experiments on the influence of different N/P ratios on phytoplankton community structure in the coast of Zhejiang Province,” Acta Oceanol. Sin. 34, 128–138 (2012).
68. J. E. O’Reilly, S. Maritorena, D. A. Siegel, M. C. O’Brien, D. Toole, B. G. Mitchell, M. Kahru, F. P. Chavez, P. Strutton, and G. F. Cota, “Ocean color chlorophyll a algorithms for SeaWiFS, OC2, and OC4: version 4,” SeaWiFS Post. Calib. Valid. Anal. 3, 9–23 (2000).
69. A. Morel and D. Antoine, ATBD 2.9--Pigment Index Retrieval in Case 1 Waters (2000).
