In-situ sampling was done in the ECS and the Tsushima Strait (TS) during cruises on board the T/V Nagasaki Maru in July of 2011–2014 (Figure 1). At each station, a Conductivity Temperature Depth profiler with Niskin bottles was used to conduct hydrographic measurements and water sampling. Water was sampled from 2 to 4 layers, which included the surface, subsurface chlorophyll a maximum (SCM), above the SCM and below the SCM.

Surface water samples for CHL, phytoplankton pigments, and nutrients (including DIN = NO₃⁻ + NO₂⁻, DIP: PO43-) were collected using an acid-washed bucket. Nutrient samples were immediately frozen in polyethylene tubes and transferred to the shore laboratory under frozen conditions for analysis using an Auto-Analyzer (AACS-IV, BLTEC). DIN and DIP concentrations were determined according to absorption spectrometry (Murphy and Riley, 1962; Armstrong et al., 1967; Koroleff, 1983). The observation area in 2013 was overlapped with 2011, therefore three transects (ECS-2011, TW-2012, and KW-2014) in the featured region were selected to show the differences in regional water mass properties and phytoplankton communities of the euphotic water column (Figure 1).

Sea surface temperature (SST) and MODIS/Aqua CHL were obtained from the NASA Ocean Color website (http://oceancolor.gsfc.nasa.gov/). Monthly composites for our study area (122–132°N, 28–36°E), for July 2011, 2012, and 2014, were extracted from global coverage datasets.

Water samples for HPLC pigment analysis (1–2 L) were filtered onto 25 mm Whatman GF/F glass fiber filters under a vacuum pressure (<0.01 MPa) and dim light and immediately frozen in liquid nitrogen for measurement in the laboratory. Concentrations of 19 pigments were measured using KANSO CO. LTD. according to the method suggested by Van Heukelem and Thomas (Van Heukelem and Thomas, 2001) using reverse-phase HPLC with a Zorbax Eclipse XDB-C8 column (150 × 4.6 mm, 3.5 μm; Agilent Technologies, USA FluoroProbe (bbe moldaenke, Germany)).

CHEMTAX was used to interpret phytoplankton group composition based on pigment data. Matrices of pigment data and initial pigments/CHL ratio for each phytoplankton group were used as input of the CHEMTAX program (Mackey et al., 1996). Thirteen marker pigments were selected and classified into nine phytoplankton groups according to published papers in the ECS and adjacent areas (Furuya et al., 2003; Zhu et al., 2009; Liu et al., 2016). Samples were separated into different subsets according to their similarity of pigment compositions by clustering analysis to ensure that the pigment ratios were constant for CHEMTAX analysis. For each dataset, the CHEMTAX program was run 10 times using the output ratio matrix of the last run as the next ratio matrix input, and the optimized phytoplankton compositions were selected from the most stable final ratios among the 10 outputs (Latasa, 2007). More details on deriving the appropriate proportions of phytoplankton groups from CHL, using the CHEMTAX program, are described in detail in our earlier work (Gomes et al., 2018).

After vertical profiles of phytoplankton excitation spectra were collected using the Multi-Exciter, the model suggested by Wang et al. (2016) was applied to estimate phytoplankton groups from the measured fluorescence spectra. In brief, this model includes three steps: (1) Continuum removal and band ratio were processed to obtain spectral features from phytoplankton fluorescence excitation spectra F(λ), which are associated with the composition of phytoplankton groups. (2) Principle component analysis (PCA) was used to compress the redundancy of spectral features and to obtain PC scores, which were used to derive the CHL fractions of the phytoplankton groups. (3) A sigmoid function was used to estimate phytoplankton groups from spectral data as:

where f denotes CHL fractions of brown algae, cyanobacteria, or green algae; Si is the i-th PC score; c₀ and c_i are regression coefficients between phytoplankton group fractions and PC scores (Si); k is the number of PCs, which was set to 8 because 99% of the variance of the original dataset can be explained by the first 8 PC scores.

In this study, four groups were determined according to the spectral characteristics of each group (brown algae, cyanobacteria, green algae, and cryptophytes). The upper and lower limits of f were set as 1 and 0, respectively, to avoid unrealistic estimates. Cryptophytes were calculated as 1–f_{brown algae}-f_{cyanobacteria}-f_{green algae} by considering that the sum of four phytoplankton groups is 1. More details of the model can be found in Wang et al. (2016). Datasets from 2011 to 2013 are obtained from the study by Wang et al. (2016), which, after calibration, showed statistically high relationships with CHEMTAX-derived phytoplankton groups (R² = 0.64, 0.68, 0.36, and 0.49 and RMSE = 0.117, 0.078, 0.072, and 0.060 for brown algae, cyanobacteria, green algae, and cryptophytes, respectively). Data from 2014 were recalibrated with HPLC samples collected in the same year to improve the accuracy of the model. Correlation coefficients (R²) varied from 0.72, 0.87, 0.61, and 0.19 and RMSE were <0.2 for brown algae, cyanobacteria, green algae, and cryptophytes, respectively.

This study focused on the vertical structure and variability of water masses and their influence on phytoplankton communities in the ECS and TS. Surface ocean color CHL images for the four summers that comprised our study revealed high CHL near Changjiang estuary, gradually decreasing offshore (Figure 2). These spatial distribution patterns are highly consistent with previous studies (Yamaguchi et al., 2012, 2013). However, our study reveals that besides surface variations, there are large changes in temperature and CHL with depth that needed to be examined, to understand how the movement of water masses in deeper layers can affect phytoplankton composition (Furuya et al., 2003; Liu et al., 2012).

The eastern ECS generally maintains its characteristics of oceanic waters, the KW flows into the ECS forming a mixture of water masses with CDW and water from the Yellow Sea, flowing further into the TS as TW (Isobe et al., 2002). In the ECS, high temperature, low salinity, and high N:P ratios were observed above 20 m indicating the influence of CDW. SCMs were located in the upper waters <30 m (Figures 8a–g). The phytoplankton groups varied annually: brown algae dominated in 2013, while mixed phytoplankton communities were recorded in 2011 (Figures 8h–k). CDW flowing into the mid-shelf of the ECS caused DIP-limitation and strongly inhibited the growth of large phytoplankton such as diatoms and dinoflagellate (denoted as brown algae in this study) (Wang et al., 2014, Wang et al., 2016; Gomes et al., 2018; Xu et al., 2019; Ishizaka, 2021). Similar patterns were reported for surface CDW dominated waters suggested by Xu et al. (2019) who suggested dominance by brown algae in 2013 was likely caused by the offshore movement of low N:P-ratio CDW to the mid-shelf ECS. At locations where vertical mixing was strong enough to erode the halocline, high salinity and high nutrient water that shoaled into the euphotic zone supported phytoplankton growth to form SCMs near the nitracline (Lee et al., 2017).

In TS (2012) and Kuroshio region (2014), surface CHL was low, and SCMs were recorded below 30 m (Figure 8c). Brown algae contributed to the high CHL in the entire euphotic column especially in SCM (Figure 8h). A previous study in TS showed that 70% of the freshwater at upper depths (<50 m) flowed from the ECS to TS, however, DIN and DIP had been depleted by the phytoplankton in the ECS (Isobe et al., 2002; Morimoto et al., 2009). Therefore, the dominance of brown algae represents a typical feature of phytoplankton composition in coastal regions of global oceans (Simpson and Sharples, 2012; Wang et al., 2014). In the Kuroshio region (2014), brown algae were replaced by green algae in the SCMs at deep depth. The paucity of nutrients in the surface waters of the Kuroshio region greatly limited the growth of large phytoplankton, but conversely facilitated the growth of small phytoplankton because of their high photosynthetic efficiency under stable light conditions (Ishizaka et al., 1994; Wang et al., 2014).

Phytoplankton groups derived using multiple excitation fluorometer and calibrated with similar phytoplankton community distribution derived using in-situ HPLC pigment biomarkers offer two major advantages. First, this submersible profiling instrument was able to provide accurate high-resolution vertical profiles of phytoplankton in ECS and TS far greater than that possible by discrete depth HPLC-pigment samples (Figure 9, left). At almost all stations HPLC pigment analysis could not adequately capture the phytoplankton variability associated with different water masses. For example, temperature, nutrient, and CHL concentrations, all showed clear vertical and spatial changes (Figures 5c, 6c,f,i,l, 7c), but it was difficult the associated fine-scale changes in phytoplankton composition with HPLC derived phytoplankton groups (Figure 9, left). Conversely, the fluorometer provided phytoplankton abundance every 1 m (Figure 9, right) with general results that were consistent with HPLC-derived phytoplankton groups.

The interaction between water mass changes and phytoplankton variabilities was observed on a larger scale due to the large amount of data generated. Three transects were chosen to show the relative abundances of phytoplankton groups in KW (KW-2014), TS (TW-2012), and ECS (ECS-2011). Even in KW-2014, which was mostly affected by KW, temperature and nutrient gradients resulted in three layers of different phytoplankton groups. Surface and deep water were gradually mixed at 20–50 m, with temperatures of 22–24.5°C, and nutrient concentrations were low. Consequently, CHL in the surface waters was low, and cyanobacteria and green were the major contributors to CHL. Below 50 m, at the western and eastern stations, brown and green algae contributed to the higher CHL. Compared with KW-2014, section TW-2012 was mostly dominated by brown algae, but the shift to dominance by green algae (probably prochlorophytes, as discussed in the following paragraph) at the eastern station is probably related to the flow of low-nutrient, high-temperature KW (Liu et al., 2016; Ishizaka, 2021). This was well-captured by the fluorometer making possible large-scale observations, especially for retracing imperceptible variations, difficult to observe by sampling at discrete depths.

The objective of the Multi-Exciter calibration in our study is to extract spectral characteristics of different phytoplankton groups from in-situ HPLC CHEMTAX data. Compared with the initial calibration, the in-situ calibration in this study takes advantage to obtain spectral differences of each phytoplankton group without defining “norm spectra.” The “norm spectra” for the Multi-Exciter was initially calibrated by the manufacturing company with freshwater species cultured in the laboratory (Yoshida et al., 2011), similar to another multiple excitation fluorometer, the FluoroProbe. Several researchers who used the FluoroProbe pointed out that the accuracy of the fluorometer is largely dependent on the calibration (See et al., 2005; Richardson et al., 2010), and the determination of the norm spectra of each group is the fundamental component for evaluation (Beutler et al., 2002). It has been known that the norm spectra between species are not always the same under different environmental conditions and varied phytoplankton composition (Loftus and Seliger, 1975), which leads to under/overestimation by the fluorometer (Beutler et al., 2002; Alexander et al., 2012). Therefore, applying the present model facilitates using the Multi-Exciter without measuring cultured algae spectra from the specific study areas.

There are, however, some limitations to the use of multiple excitation fluorometers in the field. First, the fluorescent spectra data in shallow layers (<5 m in this study) were influenced by non-photochemical quenching under high light conditions, so that it cannot correctly reflect phytoplankton physiological properties and total abundance and species composition may be influenced (Koblížek et al., 2001; Morrison, 2003).

Second, our approach cannot resolve as many phytoplankton groups as HPLC-derived pigments in association with CHEMTAX. For example, in this study, prochlorophytes were not included for calibration, whereas this group was frequently observed in the Kuroshio region of the ECS (Furuya et al., 2003; Xu et al., 2019). Prochlorophytes, which are part of a cyanobacterial group, are about 0.6–0.8 μm in size and contain the dominant and characteristic pigment divinylchlorophyll a. The spectral shape of divinyl-chlorophyll a is similar to CHL, showing fluorescence emission in the red band, which overlaps with that of CHL (Zapata et al., 2000; Bricaud et al., 2004). The complementary pigments of prochlorophytes include chlorophyll b₂, zeaxanthin, and α-carotene, which is a similar pigment composition to prasinoxanthin containing prasinophytes (Goericke and Repeta, 1992). The Multi-Exciter will detect this group as green algae rather than cyanobacteria. Our HPLC analysis of pigments shows that divinyl-chlorophyll a was present in the Kuroshio region (0.01–0.35 mg m⁻³), with maximum values in the SCM of KW-2014. The Multi-Exciter detected green algae in association with the high KW in mixed layers where temperature was around 22–24.5°C (Figures 5c, 7i). Therefore, the green algae detected using the Multi-Exciter in the KW region of 2014 should be revised as prochlorophytes, as previous studies have revealed that prochlorophytes was temperature-sensitive and restricted in KW in the eastern ECS (Lee et al., 2014; Ishizaka, 2021). This example serves to stress the necessity and importance of using a combination of Multi-Exciter and HPLC pigment analysis at certain depths to ascertain the whole structure of phytoplankton groups.