Effects of Mixed Layer Depth on Phytoplankton Biomass in a Tropical Marginal Ocean: A Multiple Timescale Analysis

In open oceans, changes in mixed layer depth (MLD) may affect phytoplankton growth and biomass variations via the regulation of nutrient supply from deep waters. Estimates of relationships between variability in phytoplankton dynamics and the MLD remain limited, especially at different time scales. We compiled and analyzed averaged euphotic‐depth‐integrated chlorophyll‐a (IChl‐a) and surface chlorophyll‐a (SChl‐a) concentrations collected from 27 cruises during the period of 1999–2019 in the tropical northern South China Sea (SCS). Seasonal differences existed in both averaged IChl‐a and SChl‐a concentrations, with significantly high concentrations in the cold season. Inconsistent relationships between the averaged IChl‐a and SChl‐a concentrations between seasons implied that the use of SChl‐a concentration as a common indicator of phytoplankton biomass dynamics should be performed with caution. Over the past decades in the northern SCS, the averaged IChl‐a, SChl‐a, and MLD decreased to a greater extent in the cold season than in the warm season, while sea surface temperature (SST) rose rapidly and dramatically in both seasons. The MLD was observed to have better correlations with the averaged IChl‐a and SChl‐a concentrations than the SST in the time‐series data. Our results highlight the importance of IChl‐a concentration, which is an overall measure of phytoplankton responses to euphotic zone conditions, and the MLD could be used as a good index for changes in phytoplankton biomass under climate change.

biomass and distribution of phytoplankton can successively impact higher trophic levels, fisheries, and whole ecosystems (Boyce et al., 2010). However, the phytoplankton trends reported by either satellite or field observations at various locations still exhibit high variability at long-term seasonal, interannual and/or decadal time scales (Boyce et al., 2010;Chavez et al., 2011;Schollaert Uz et al., 2017), showing that temporal and spatial heterogeneity may influence the assessment of thermal stratification dynamics and phytoplankton changes.
Since the 1970s, satellite ocean color sensors have provided large-scale and long-term observations and are currently widely applied to detect phytoplankton biomass and distribution (McClain, 2009;Siegel et al., 2013). It is commonly inferred from measurements of total chlorophyll-a (Chl-a) pigment concentration, thus representing variances in marine primary production and capturing the first-order changes in phytoplankton in marine ecosystems (Boyce et al., 2010;Falkowski & Raven, 2013). However, this method has limitations both in relation to the upper water layers and in relation to dynamic detection at an hourly time scale. Therefore, in situ field sampling is expected to more accurately capture phytoplankton dynamics throughout the whole water column, although this type of sampling is difficult and costly to conduct. In general, surface chlorophyll-a (SChl-a) concentrations based on satellite observations are widely used to infer phytoplankton biomass dynamics in most large-scale longterm studies, while euphotic-depth-integrated Chl-a (IChl-a) concentrations, which are instead based on in situ field sampling, can reflect entire vertical distributions of phytoplankton biomass dynamics in the euphotic zone (Frolov et al., 2012;Morel & Berthon, 1989;Uitz et al., 2006).
Recently, studies have highlighted not only importance of the MLD for determining temperature ranges in oceanic and coastal regions and as a source of heat that drives global variability, but also the MLD in relation to vertical stratification for available light and nutrient for marine organisms (Gardner et al., 1995;Jeronimo & Gomez-Valdes, 2010;Sirevaag et al., 2011). Moreover, warming temperatures could cause increases in stratification and a shallowing of the MLD, hence less light and nutrient available for phytoplankton in the mixed layer, which then lead to a decrease of phytoplankton growth (Boyce et al., 2010;Winder & Sommer, 2012). However, too many factors come into play in the oceans, inconsistencies between the MLD and sea surface temperature (SST) in terms of patterns and increasing/decreasing rates have been gradually observed (Somavilla et al., 2017), raising debates regarding how best to consider parameters as representative of contemporary ocean warming and to link with phytoplankton dynamics in oceans. As commonly assumed for simplicity, the mixed layer deepens due to the effects of wind and heat loss to atmosphere but does not shoal smoothly (Price et al., 1986). Instead, the upper ocean stratifies due to solar heating or other heat sources (e.g., freshwater flux), and eventually, a new mixed layer reforms from the surface, thereby isolating phytoplankton and other particles in the so-called remnant layer (C. Ho & Marra, 1994;Lacour et al., 2019). That is, simply, the variability in MLD in fact involves a subtle interplay between circulation and atmospheric forcing (Schofield et al., 2018). In addition, in a comparison at different time scales, the amplitude of MLD variation at a diurnal time scale is usually small (Woods & Onken, 1982), whereas at seasonal and annual scales, the MLD effect becomes a process of greater significance (Carlson et al., 1994;Dall'Olmo & Mork, 2014;Dall'Olmo et al., 2016). Few studies, however, have placed the MLD influence in a time scale-dependent context or have assessed relationships between the MLD and phytoplankton at multiple temporal scales within a marine ecosystem.
The South China Sea (SCS), one of the largest marginal seas in the world, has high marine primary and fisheries production, but strong seasonal monsoons and adjacent oceanic currents make it sensitive to climate variability (K.-K. Liu et al., 2002;Qiu et al., 2010;Tseng et al., 2005). Together with year-round SCS Warm Current and Kuroshio intrusion, the northeast monsoon in winter drives a cyclonic gyre in the SCS, while the southwest monsoon in summer causes a reversal of the gyre in the southern part of the basin (Figure 1; Shaw & Chao, 1994), thereby supporting the role of seasonal, even longer temporal, changes in the area variations in evaporation and precipitation, leading to significant thermal stratification dynamics . The South East Asia Time-series Study (SEATS) station located at 116°E and 18°N was established to monitor long-term biogeochemical changes in the northern SCS, and regular field observations have been collected since 1999. Previous studies have noted that the depth of thermocline may be a major factor in phytoplankton variability at the SEATS station at a monthly time scale; additionally, the reported MLD at the SEATS station was approximately 20 m deep in summer and 70 m deep in winter, which is relatively shallow compared to the global averaged MLD (Tai et al., 2017;Wong et al., 2015;Wu et al., 2003). Moreover, satellite observations showed Chl-a peaks in both winter and summer, especially in monsoon season Wong et al., 2015), resulting in a seasonal pattern distinct from those of other oceans at tropical latitudes (Tseng et al., 2005). Stratification induced by a Methodology: Li-Tzu Hou, Bo-Shian Wang, Yung-Yen Shih, Fuh-Kwo Shiah, Chia-Ying Ko Writing -original draft: Li-Tzu Hou, Bo-Shian Wang, Chao-Chen Lai, Fuh-Kwo Shiah, Chia-Ying Ko combination of temperature and wind also regulated negative correlations between monthly and annual averaged SST and MLD (Tai et al., 2017). Given that the euphotic zone at the SEATS station can be as deep as 100 m in the ocean and that the Chl-a maximum layer is approximately 60 m, which is not close to the surface (K.-K. Liu et al., 2013), the IChl-a concentrations might more accurately represent the dynamics of phytoplankton distribution and biomass in the water column than the surface values in the SCS.
The objectives of this study were (a) to compare differences between IChl-a and SChl-a concentrations and (b) to explore effects of the MLD on both IChl-a and SChl-a concentrations and its potential use as an indicator in comparison to the commonly used SST parameter, particularly at extended time scales, for example, seasonal and decadal scales. We compiled data from 27 cruises at the SEAT station in the northern SCS over two decades, from 1999 to 2019, to identify in situ patterns linking phytoplankton and oceanographic variables, including SST and MLD. In addition, due to long-term impacts of climate change, which is becoming a greater threat in the Anthropocene (IPCC, 2013;Pörtner et al., 2014), we further hypothesized that at these extended time scales, high correlations should exist between phytoplankton and MLD because of ocean mixing as a better parameterization for modeled oceanic stratification and resultant biological carbon storage (Oka & Niwa, 2013;Tatebe et al., 2018).

Field Sampling
Phytoplankton and oceanographic measurements collected at the SEATS station in the northern SCS were compiled from 27 cruises over the course of more than two decades (1999-2019; Table S1 in Supporting Information S1). Sea temperature and Chl-a fluorescence data were continuously recorded by a conductivity-temperature-depth (CTD) profiling recorder (Sea-Bird model SBE9/11, Sea-Bird Electronics, Inc.), corresponding to one measurement per meter of water depth, during every cast in each cruise. Laboratory calibrations of all CTD sensors were done before each cruise and two sets of the CTD recorders were set up onboard in each cruise for simultaneously monitoring and comparing real-time estimates. The surveyed casts conducted during each cruise with sampling frequencies ranged from 3 to 6 hr over a 13-hr to 4-day period.

Phytoplankton Chl-a Biomass Measurement
Water samples for in situ phytoplankton Chl-a biomass measurements were taken from depths of 5, 10, 20, 30, 50, 80, and 100 m. The samples were filtered onboard through 47-mm GF/F filters (Whatman) under a vacuum pressure of less than 100 mmHg and preserved at −20°C until further processing. In the laboratory, the filters were extracted by 90% acetone and measured either by high-performance liquid chromatography (HPLC; Shimazu LC-10A, method refers to T.-Y. Ho et al., 2015) or by a fluorometer (Turner 10-AU, method refers to Strickland & Parsons, 1972), with the Chl-a standard validated by a UV-vis spectrophotometer and fluorometer for calibration. Both HPLC and fluorometer methods are well developed for measuring marine Chl-a concentrations, and their measured results are comparable (Murray et al., 1986). To derive phytoplankton Chl-a biomass within the euphotic zone, readings from an onboard CTD fluorescence meter were calibrated with discrete in situ Chl-a measurements using least squares linear regression. The coefficients of determination in the 27 cruises ranged from 0.42 to 0.97 ( Figure S1 in Supporting Information S1). The calibrated Chl-a fluorescence values were finally used to represent phytoplankton biomass dynamics in the following analyses. To limit perturbations caused by wave height, the Chl-a concentrations at a depth of 10 m were then used as surface measurements in this study. The averaged IChl-a concentrations as the euphotic-depth measurements were calculated by a trapezoidal integration method taking average from depths of 10-100 m (Uitz et al., 2006).

Data Management and Statistical Analysis
All surveyed casts were first systematically analyzed for characterizing diurnal-scale changes. To eliminate biased diurnal sampling effort due to sea conditions, only daytime casts implemented between 08:00 and 16:00 and nighttime casts between 20:00 and 04:00 were selected to standardize variability. The MLD was defined as depth at which the sea temperature cooled by 0.8°C relative to that at 10 m depth, that is, ΔT = 0.8°C (Tai et al., 2017).
To validate in situ measurements to a larger spatial scale, satellite Chl-a data were compared using daily MODIS-Aqua 4-km level 3 observations and assimilated monthly Chl-a data from the three-dimensional NASA Ocean Biogeochemical Model (NOBM) coupled circulation/biogeochemical/radiative processes of global oceans (Gregg, 2008). Both satellite Chl-a concentrations were mapped within a spatial range of 115.5°−116.5°E, 17.5°-18.5°N and a temporal range of July 2002 to December 2019.
Differences in SChl-a and averaged IChl-a concentrations between seasons were quantified using the Wilcoxon signed-rank test. Relationships among SChl-a, averaged IChl-a, and satellite Chl-a concentrations and changes in both SChl-a and averaged IChl-a concentrations and oceanographic variables, that is, MLD and SST, over time were estimated using regressions. Pearson's correlations between seasonal and interannual Chl-a concentrations and oceanographic variables were also assessed. The p-value lower than 0.05 was considered to be statistically significant. All statistical analyses were carried out by R 3.6.2 (R Core Team, 2019).

Intra-Annual Dynamics
At the SEATS station, the monthly averaged IChl-a concentrations ranged from 0.146 mg m −3 in May to 0.347 mg m −3 in December, and the monthly SChl-a concentrations ranged from 0.078 mg m −3 in December to 0.395 mg m −3 in April (Figures 2a and 2b), while the monthly averaged SST had a bell-shaped distribution with a peak of 29.71°C in July and a record low of 23.91°C in January ( Figure 2c) and an opposite trend for the MLD (17.67 and 81.83 m in May and January, respectively; Figure 2d). The observed Chl-a pattern was similar to middleand high-latitude open ocean water dynamics, such as the Bermuda Atlantic Time-series Study and the Hawaii Ocean Time-series station ALOHA, where phytoplankton biomass increase when the deep MLD penetrates the nutricline, which allows nutrients to enter the upper mixed layer in winter, and a shallower MLD causes low Chl-a concentrations in nutrient-limited surface waters in summer (Chavez et al., 2011;Cullen, 2015;Karl & Church, 2014). The intra-annual Chl-a dynamics in the SCS are caused by the consistent year-round solar irradiance, winter convective overturn by surface cooling, the influences of monsoon-enhanced wind-induced mixing with the strongest upwelling in winter and weak upwelling in summer, and cyclonic eddies caused by the Kuroshio intrusions during winter, which trigger phytoplankton blooms (Chen et al., 2006;Pan et al., 2015;Tseng et al., 2005).
Two seasons, that is, cold and warm, were classified at the SEATS station based on vertical oceanographic profiles ( Figures S2 and S3 in Supporting Information S1). In the cold season (December and January), SSTs were lower than 27.5°C, water column was generally well mixed, and Chl-a fluorescence showed a uniform vertical profile with a relatively even distribution of high concentrations above the MLD in the upper 50 m of the water column. In the warm season (March to November), SSTs were higher than 27.5°C, waters showed stratification above a depth of 60 m, and the peak Chl-a fluorescence or subsurface Chl-a maximum layer existed at subsurface depths of 50-80 m, which were below the MLDs.
Significantly higher median values of the averaged IChl-a and SChl-a concentrations were observed in the cold season than in the warm season (both p < 0.05, Figures 2a and 2b). Larger differences between the two seasons existed in SChl-a concentrations (with a difference of approximately 3.1 times) than in averaged IChl-a concentrations (with a difference of approximately 1.4 times), indicating consistent active phytoplankton growth and activity in the upper euphotic water layers year round in the northern SCS. These variations between SChl-a and IChl-a concentrations could be due to differences in the strengths of monsoons, which are strong in the cold season and weak in the warm season, inducing mixing and winter convective overturning by surface cooling, leading to low SChl-a values and high Chl-a peaks at the subsurface nutricline Tseng et al., 2005).
The small differences in averaged IChl-a concentrations between seasons at the SEATS station and the similarity between values observed in the warm season, particularly in July, September, and November, and those in the cold season may be due to episodic events, including typhoons and mesoscale processes, that alter phytoplankton community responses and succession (Alford et al., 2015;Zhao et al., 2008). We found typhoon disturbances with monthly cumulative numbers ranging from 2 to 32 in recent decades in the SCS (Figure 2e). Lin et al. (2003) proved that after a modulated typhoon passed, oceanic processes, including water mixing, uplift of nutrients from below the thermocline, and phytoplankton growth and activity in the water layers, caused phytoplankton biomass in the warm season to be comparable to that in the cold season, although Chl-a blooms were usually short lived and small in extent and not induced if nutrients at the sea surface were limited. The post-typhoon periods play an important role in the timeline of phytoplankton increases, and seasonal variation has also been observed in different aquatic ecosystems (C.-Y. Ko et al., 2017). Moreover, unlike previous studies showing the importance of mesoscale processes such as thermal fronts and eddies in the waters surrounding the northern SCS (Owen, 1981;Xian et al., 2012;Ye et al., 2017), we found that these processes rarely caused detectable Chl-a concentration anomalies among the 27 cruises in this study (Table S1 in Supporting Information S1). Only the OR1-664 cruise in November 2002 showed impacts of ENSO events on diurnal phytoplankton fluctuations with dramatically high hourly averaged IChl-a and SChl-a concentrations ( Figure S4 in Supporting Information S1); these findings may challenge understandings of anti-cyclonic ocean circulation existing in the SCS outside the aforementioned monsoon seasons, similar to the rare and special physical oceanographic conditions such as a combination of cold SST pool and peak in upwelling that occurred in the SCS in spring 2003 and autumn 2010 (D. S. Ko et al., 2014;Lin et al., 2003). Future work is needed to disentangle multiple ecological mechanisms and environmental stressors at play, including estimations of time-lag effects, different atmospheric scales, and potential feedbacks, to provide insights into the process of phytoplankton biomass dynamics during the observed patterns.
Interestingly, when estimating the relationships between averaged IChl-a and SChl-a concentrations, seasonal differences were observed, with a higher R 2 value in the warm season than in the cold season (Figure 3a). Therefore, the SChl-a concentrations might not fully reveal vertical distributions of phytoplankton biomass, at least at the seasonal scale, and should be used with caution as an indicator of phytoplankton biomass dynamics. Moreover, the widely used satellite Chl-a concentrations, either assimilated monthly, that is, NOBM Chl-a, or daily, that is, MODIS Chl-a, showed limited predictive power, except for the SChl-a versus NOBM Chl-a concentrations in the warm season (R 2 = 0.44), and revealed considerable variability in the Chl-a concentrations in the surface waters (Figures 3b and 3c). These results suggest that in the northern SCS, phytoplankton biomass and/or primary production derived from satellite Chl-a data could be underestimated, because the satellite 10.1029/2020EF001842 6 of 12 measurements are optical depth-averaged values of the upper water column and usually without consideration for vertical variations (Gordon et al., 1980;Lee & Carder, 2004;Ma et al., 2014). We suggest that such derivation formulas using remote sensing and satellite technology provide an easily accessible but highly heterogeneous uncertainty that will require careful calibration of samples. Scientists have been trying to uncover the vertical distributions of phytoplankton and depth-integrated primary production (IPP) by surface Chl-a data (Behrenfeld & Falkowski, 1997;Morel & Berthon, 1989;Saba et al., 2010;Uitz et al., 2006). Recently, a surprisingly strong relationship between SChl-a and averaged IChl-a concentrations was observed to hold for profiles with high concentrations of Chl-a in persistent thin (<3 m) subsurface phytoplankton layers in California (Frolov et al., 2012). Besides, complexity of physical-oceanographic processes, variability of phytoplankton physiology, and biological interactions in the phytoplankton community can cause highly heterogeneous shifts in the water layers (Cullen, 2015;Frolov et al., 2012;Gittings et al., 2018;Vanharanta et al., 2020). Therefore, unraveling the relationships between changes in phytoplankton biomass in the surface and euphotic-depth water layers in the northern SCS still needs to be conservative and open to adjustment to account for other factors.

Interannual Trends and Decadal Variations
From 1999 to 2019, both averaged IChl-a and SChl-a concentrations decreased to a greater extent in the cold season than in the warm season, with the averaged IChl-a concentrations decreasing more than the SChl-a concentrations ( Our results support the conclusions from Boyce et al. (2010) and Tai et al. (2017), who reported that with observed and projected increases in SST (Boyce et al., 2010;Pörtner et al., 2014), the warming of the surface ocean waters with stratification in tropical and subtropical oceans could reduce vertical mixing of the upper water layer, causing a shallower MLD and reducing nutrient input into the surface waters, thereby resulting in a decrease in phytoplankton biomass and productivity in both the surface and euphotic-depth layers. However, the North Pacific subtropical gyres, the North Atlantic, and the mid-latitude Eastern North Atlantic did not show significant changes in the MLD and IChl-a concentrations with respect to the increased SST in recent decades, where changes in salinity, density, wind-driven divergence and buoyancy losses could have more dominant effects (Chavez et al., 2011;Somavilla et al., 2017). Increasing SST can cause plankton community compositions to shift toward small-sized phytoplankton and cyanobacteria (Rasconi et al., 2015) and alter phytoplankton stoichiometry and metabolism (Toseland et al., 2013;Winder & Sommer, 2012), both of which also contribute to a decrease in phytoplankton biomass. In addition, despite increasing fishing effort after the 1990s, the fisheries in the SCS stagnated, and the mean temperature of catch did not increase with SST, in contrast to the pattern observed in other subtropical large marine ecosystems (Pauly & Liang, 2020). As a consequence, the severe effects of such SST increases, accompanying decreases in phytoplankton biomass, contribute to associated food chain disturbance and a more vulnerable SCS marine ecosystem under climate change. Accordingly, as a semi-enclosed marginal sea with little water exchange potential and consistently great fishing pressure, the SCS may be especially prone to contemporary increases in SST and decreases in MLD. We caveat that the continuous decreases in phytoplankton biomass may become a severe biotic crisis in the SCS.
Investigation of changes in the interannual variabilities in Chl-a concentrations in response to SST and MLD forcing revealed that both averaged IChl-a and SChl-a concentrations were inversely correlated with SST but positively correlated with MLD, with varied correlations ranging in magnitude from −0.26 to 0.64, −0.14 to 0.38, and −0.64 to 0.67 in the warm season, cold season and all years, respectively ( Figure 5). These strong correlations between Chl-a and SST and MLD also prevailed at low latitudes, as observed in previous studies (Boyce et al., 2010;Schollaert Uz et al., 2017). Our findings demonstrate that the MLD is a more significant and stronger predictor of Chl-a concentrations than SST in time-series data, although this relationship exhibits regional variability in response to modern disturbance trends (Hadfield & Sharples, 1996;Jeronimo & Gomez-Valdes, 2010;Tai et al., 2017). Notably, with different magnitudes of correlation coefficients in SST and MLD, the higher annual correlations observed between the averaged IChl-a concentrations and SST and MLD than those observed based on SChl-a concentrations. However, differently, high seasonal correlations existed between the SChl-a concentrations and SST and MLD, again indicating that the use of SChl-a concentrations to represent ocean water conditions derived from satellite remote sensing may require more consideration. Considering that physical oceanographical processes such as SST and MLD are usually easier to monitor and predict changes than biological processes and that the MLD directly represents nutrient availability in the euphotic zone, especially in oligotrophic open oceans (Tai et al., 2017;Wu et al., 2003), our analysis thus emphasizes the use of MLD to represent water column phytoplankton biomass and/or primary production under climate change. Furthermore, the averaged IChl-a concentrations were more tightly coupled with MLD than with SST, robustly demonstrating that the use of MLD could better predict oceanic Chl-a vertical concentrations, although the aforementioned seasonal dynamics between these oceanographic variations and phytoplankton performance exist. Overall, these findings add an important time scale context to previous work on MLD or SST correlations with observed phytoplankton changes and highlight the complexity of interactions between time scales and consequences of phytoplankton biomass changes.
Apart from long-term temporal changes, another interesting observation in this study was large diurnal variations among cruises and between seasons at the SEATS station in the northern SCS ( Figure S4 in Supporting Information S1). Specifically, in the cold season, high Chl-a concentrations were shown in the late evening, and low concentrations occurred before noon. Similar trends were reported in the equatorial Pacific (Claustre et al., 1999;Neveux, 2003;Vaulot & Marie, 1999), the North Pacific Ocean (Siegel et al., 1989), and the North Atlantic (Gardner et al., 1995) by the combined influences of changes in phytoplankton biomass on growth production and losses due to grazing, sinking, and aggregation and phytoplankton photoacclimation and photoinhibition processes in response to diurnal light availability (Claustre et al., 1999;Letelier et al., 1993;Mignot et al., 2014;Neveux, 2003). On the other hand, the warm season cruises did not show a clear diurnal Chl-a pattern with respect to sunlight or time. The diurnal fluctuations in the cold season were more than two-fold higher than those in the warm season, where the diurnal variation could range from 0.09 to 0.64 mg m −3 in the SChl-a concentrations within a 17-hr period (Cruise OR1-821, January 2007). Notably, at the SEATS station, the individual samplings around dawn might underestimate the diurnal SChl-a and averaged IChl-a concentrations, while those in the evening might overestimate these Chl-a concentrations, especially in the cold season. Therefore, we suggest that when investigating the short-term variability in phytoplankton dynamics, it is important to not only consider the diurnal effects but also be aware that a single in situ sampling event in the field may not be suitable to adequately represent the diurnal averaged concentrations. Simultaneously, weak interactions between averaged IChl-a and SChl-a concentrations and SST (all r < −0.2; Figure S5 in Supporting Information S1) and inconsistent interactions between averaged IChl-a and SChl-a concentrations (r ranging from 0.01 to 0.55; Figure S5 in Supporting Information S1) and MLD were found at hourly and diurnal timescales, further indicating that using SST and MLD to reflect phytoplankton biomass may be inadequate at short temporal scales.
Besides, variability of phytoplankton biomass in the face of primary production has been posed a number of relevant questions and different correlations between them usually exist in different seasons, climatic events, and regions (C. Ho & Marra, 1994;C.-Y. Ko et al., 2017;K.-K. Liu et al., 2002;Lin et al., 2003;Marañón et al., 2003). Austria et al. (2018) found a significant relationship between IChl-a concentrations and euphoticdepth-IPP based on one summer (Jun 2010) and two winter (January and December 2011) cruises conducted at the shelf of the SCS and further indicated that approximately 34% of IPP variations could be explained by the IChl-a concentrations at the SEATS station. We, therefore, believe that the phytoplankton biomass proxy, that is, the IChl-concentrations, presented in this study could be the indexes for rate parameters such as phytoplankton growth and primary production in the northern SCS.

Conclusions
The identification of oceanographic links with phytoplankton, especially its biomass, accounting for differences between euphotic and surface waters, represents an important goal for scientists and oceanic environmental managers. The monthly and seasonal dynamics in this study indicated that high phytoplankton biomass in the cold season and typhoon disturbance could play an important role in increasing oceanic phytoplankton biomass and/or production, particularly in oligotrophic environments such as the SCS. Our interannual and decadal observations showed a progressive increase in SST and decrease in MLD, averaged IChl-a and SChl-a concentrations. The phytoplankton biomass/production decreases with the shallowing of MLD and can be expected to continuously decrease with the current speed of global warming and climate change, hence reflecting the importance of having a long-term time-series station to monitor long-term oceanic environmental changes. Most importantly, given that the surface estimates are likely to bias the understandings of real oceanic phytoplankton biomass and/ or productivity, measurements of phytoplankton responses in the euphotic zone may still not be simply replaced by satellite remote sensing development. To better address contemporary environmental disruption, putting effort into long-term in situ monitoring remains essential for clarifying and overcoming concerns about environmental damage and climatic changes, as both IChl-a and MLD measurements can serve as alternative effective indicators.

Data Availability Statement
The data for this research can be found at http://140.112.70.58:5000/sharing/EWWB1Rb6W. The bathymetric map of the SCS was generated using GMT (Generic Mapping Tools). The typhoon records were compiled from the National Institute of Informatics (agora.ex.nii.ac.jp) from 2009 to 2018. Ocean eddies and meandering fronts were downloaded from Ocean Data Bank funded by the Ministry of Science and Technology (MOST), Taiwan (http://www.odb.ntu.edu.tw/). ENSO events, represented by MEI and ONI indices were downloaded from NOAA/National Weather Service (http://www.cpc.ncep.noaa.gov/and http://www.esrl.noaa.gov/psd/enso/mei/). This study at the SEATS station is based on a long-term contribution and collaboration of many researchers and technicians from the Taiwanese oceanographic community. The authors appreciate Dr. Kon-Kee Liu's vision on the establishment of SEATS program and launching a long-term observation in the SCS. The authors are grateful to have data producing and processing by all chef scientists and their teams, including Chung-Chi Chen, Wen-Chen Chou, Tung-Yuan Ho, Chih-An Huh, Guo-Wei Hung, Jen-Hua Tai, Chun-Mao Tseng, Liang-Saw Wen, and Yih Yang (in alphabetical order). The authors also much thank captains and crew of ORI, ORIII, Legend, and FR1 for their technical support during cruises. Valuable comments by the editor and reviewers were much appreciated and substantially improved the manuscript. This publication is based upon work supported by the Ministry of Science and Technology (MOST), Taiwan under Grant 105-2119-M-002-059-MY2 and Grant 107-2611-M-002-005.