Photosynthetic Production Determines Bottom Water Oxygen Variations in the Upwelling Coastal South China Sea Over Recent Decades

Dissolved oxygen (DO) in seawater is fundamental to marine ecosystem health. How DO in coastal upwelling areas responds to upwelling intensity under climate change is of particular interest and vital importance, because these productive regions account for a large fraction of global fishery production and marine biodiversity. The Yuedong upwelling (YDU) in the coastal northern South China Sea can be served as a study case to explore long-term responses of DO to upwelling and climate due to minor influence of riverine input. Here, bottom water DO conditions were recovered by sedimentary C28Δ22/Δ5,22 ratios of steroids in three short cores, with lower ratio value indicating higher DO concentration. The ratio records showed oscillations in varying degrees and exhibited no clear trends before ∼1980s, after which, however, there occurred a persistent decreasing trend or basically remained at lower values. Thus, inferred DO variations by the C28Δ22/Δ5,22 ratio records are not compatible with regional YDU-involved physical processes under climate change, such as southwesterly wind-induced onshore advection of reduced-oxygenated source waters from outer shelf and oceanic warming that would rather lead to less oxygenation in bottom waters in recent decades. Intriguingly, the alcohol records of n-C20:1/C28Δ5,22 and br-C15/C28Δ5,22 ratios, indicative of the relative strengths between biogeochemical oxygen consumption (i.e., by zooplankton and microbes) and photosynthetic oxygen production (i.e., by phytoplankton), changed almost in parallel with the C28Δ22/Δ5,22 records in three cores. Accordingly, we propose that net photosynthetic oxygen production outweighs source water– and warming-induced increasing deoxygenation in the study area. This study may suggest an important biogeochemical mechanism in determining bottom water DO dynamics in shallow coastal upwelling regions with minor contribution of riverine input.


INTRODUCTION
The fate of dissolved oxygen (DO) in seawater has attracted increasing attentions from multi-disciplines, owing to its vital role in ecological, biological, and geochemical dynamics in marine environments (Diaz and Rosenberg, 2008;Breitburg et al., 2009;Breitburg et al., 2018;Levin and Breitburg, 2015;Watson, 2016;Schmidtko et al., 2017;Fennel and Testa, 2019). The inventory of oceanic DO is determined by the balance between its primary sources (i.e., photosynthetic production, air-sea gas exchange, and physical oxygen supply) and sinks (i.e., aerobic respiration and consumption, oxidation of reduced chemical species, and physical oxygen export) (Breitburg et al., 2018;Fennel and Testa, 2019). Over recent decades, various investigations based on timeseries observations, numerical models, and geological records have been carried out to understand the physical-biogeochemical processes controlling the spatiotemporal distributions of lowoxygen (i.e., hypoxic and anoxic) conditions within the global ocean (e.g., Breitburg et al., 2018;Fennel and Testa, 2019, and references therein).
Generally, the occurrence of oxygen-deficiency in estuarine and river-dominated shelf regions is mainly attributed to the significant feedback of freshwater discharge and anthropogenic input, whereas the coastal upwelling areas without direct riverine and anthropogenic impacts experience low-oxygen conditions mostly due to supplies of oxygen-poor and nutrient-rich source waters at times of upwelling (Breitburg et al., 2018;Fennel and Testa, 2019). Long-term links between DO condition and upwelling intensity have been evidenced by sedimentary records of nitrogen isotope from the northeastern tropical Pacific margin, showing enhanced oxygenation in response to weakened wind-induced upwelling since 1850 (Deutsch et al., 2014). However, different relationships between DO and upwelling have been documented in other upwelling regions; e.g., overall enhanced oxygenation (Cardich et al., 2019) is observed in response to strengthened upwelling off Peru since 1860 (Gutierrez et al., 2011). These occurrences likely indicate that the proposed negative impact of upwelling activity on DO condition may be too simplistic, especially in terms of long-term (i.e., decadal and centurial scales) relationships. Besides, close links between upwelling-favorable wind and climate change proposed previously (Bakun, 1990) have been recently evidenced by observations, models, and records in many upwelling regions (Gutierrez et al., 2011;Sydeman et al., 2014;Wang et al., 2015), showing strengthened wind-induced upwelling in response to climatic warming. This scenario may further complicate long-term responses of DO condition to upwelling variation and climate change; however, this issue has not been investigated in upwelling regions in the coastal northern South China Sea (SCS) (Hu and Wang, 2016), including the Yuedong upwelling (YDU).
Recently, sedimentary records of long-chain diols in the YDU area revealed increasing trends in the upwelling intensity and annual mean sea surface temperature (SST) over recent decades (Zhu et al., 2018), supporting previous studies showing close links between upwelling intensification and climatic warming (Bakun, 1990;Gutierrez et al., 2011;Sydeman et al., 2014;Wang et al., 2015). Long-chain diol records also demonstrated insignificant or indirect input from the Pearl River (Zhu et al., 2018), making the YDU as an ideal spot to explore long-term responses of DO variability to upwelling and climate. However, as available DO records are lacking in the YDU region, how DO responds to enhanced upwelling and increased SST remains unclear. This issue is relevant to fishery production, marine biodiversity, and ecosystem health over time so that proper proxies are essential to reconstruct long-term DO variations in the YDU area to fill in this gap.
The preferential degradation of Δ 5 -sterols relative to their 5αstanol counterparts may also confound the 5α-stanol/Δ 5 -sterol ratios to reflect anaerobic conversion processes (e.g., Arzayus and Canuel, 2004;Bogus et al., 2012). The higher rate of degradation than hydrogenation of Δ 5 -sterols has been proposed to explain higher 5α-stanol/Δ 5 -sterol ratios in sediments from the York River estuary (Arzayus and Canuel, 2004). This is followed by a subsequent study on surface sediments in a cross-shelf transect offshore the Pakistan continental margin, suggesting that the increasing trend in the C 27 Δ 0 /Δ 5 ratio is attributable to the faster degradation of C 27 Δ 5 compared with that of C 27 Δ 0 (Bogus et al., 2012).

Study Area
The YDU, located in the inshore area from Hong Kong to the Nanri Islands (Figure 1), is a common phenomenon with a large spatial extent occurring in summer, leading to colder SST and higher salinity and nutrients than surrounding waters (Jing et al., 2011;Hu and Wang, 2016). The local southwesterly wind stress is one of the most important dynamical factors to induce the coastal YDU with apparent inter-annual variability (Jing et al., 2011;Hu and Wang, 2016). Recently, sedimentary records of long-chain diol index (LDI) and diol index 2 (DI-2), which are respective indicators for annual mean SST and southwesterly wind-induced upwelling intensity in the YDU area, revealed increasing warming and enhancing upwelling over recent decades (Zhu et al., 2018).
To the west of YDU is the Pearl River estuary (PRE; Figure 1), receiving discharges from the subtropical Pearl River. The river is the second largest river in China in terms of discharge, 80% of which occurs during the wet season (April to September). After being poured out of the estuary, the Pearl River plume swings seasonally (Dong et al., 2004;Su, 2004); it generally turns toward the west during the dry winter due to northeasterly winds and Coriolis effect and extends offshore toward south and southeast during wet summer under southwesterly winds when the Pearl River discharge reaches its maximum.

Sample Collection
In this work, three short box cores ( Figure 1; Table 1) at the margin of the YDU were studied. They were collected in 2009 during the China Ocean Carbon (CHOICE-C) Cruise I onboard the Dongfanghong II. The core surfaces were well preserved upon collection, as demonstrated by the fairly clear water above the sediment surface in the box corers. After the overlying water being siphoned out, core barrels were pushed into each box to collect sub-cores. Sediment in the sub-cores was then (usually within an hour) extruded onboard using a hydraulic jack, and sectioned at 2-cm intervals. The sectioned samples were sealed in plastic jars, then frozen until they were freeze-dried at −50°C in the laboratory, and grounded with an agate mortar and pestle for further analyses and tests. The chronology of the cores has been determined by the 210 Pb method and reported by Zhu et al. (2018). Table 1 shows the dating results and other information of the three cores.

Lipid Extraction, Separation and Measurement
The core sediments have been reported for long-chain diols and the detailed experiments for lipid extraction and separation can be found in the work of Zhu et al. (2018). Briefly, freeze-dried and powdered sediments were extracted ultrasonically with MeOH, dichloromethane (DCM)/MeOH (1:1, v/v), and DCM, and all Frontiers in Earth Science | www.frontiersin.org December 2021 | Volume 9 | Article 759317 extracts were combined after centrifugation. Following saponification with KOH/MeOH and extraction into hexane, the neutral lipids were purified using silica gel chromatography by elution with DCM/n-hexane (4:1, v/v) and DCM/MeOH (2:1, v/v), respectively. The latter fraction containing alcohols was converted to trimethylsilyl derivatives with bis(trimethylsilyl)trifluoroacetamide (BSTFA) at 60°C for 2 h before gas chromatography-mass spectrometry (GC-MS) analyses. GC-MS analysis was performed at the State Key Laboratory of Organic Geochemistry, Guangzhou Institute of Geochemistry, Chinese Academy of Sciences, with a Thermo Scientific Trace gas chromatograph coupled to a Thermo Scientific DSQ II mass spectrometer. Separation was achieved with a 60 m × 0.32 mm i.d., fused silica column (J and W DB-5) coated with a 0.25-μm film thickness. The oven temperature was programmed from 80°C (held 2 min) to 220°C at 6°C/min, then to 290°C (held 5 min) at 8°C/min, and at last to 315°C (held 25 min) at 2°C/min. Identification and quantification of alcohol compounds, including n-alcohols, br-alcohols, sterols, and stanols, were based on their characteristic mass fragments, i.e., m/z 103 for n-and br-alcohols, 255 for Δ 5 -and Δ 5,22 -sterols, 257 for Δ 22sterols, and 215 for 5α-stanols (Huang et al., 2013;Yang et al., 2014;Nakakuni et al., 2017;Nakakuni et al., 2018).
Further, an approach proposed by Nakakuni et al. (2017); Nakakuni et al. (2018) was applied to determine the accuracy of the analytical 5α-stanol/Δ 5 -sterol ratio results, especially the C 27 Δ 22 /Δ 5,22 , C 27 Δ 0 /Δ 5 , and C 28 Δ 22 /Δ 5,22 ratio pairs used in this study (see below). Four samples from cores S101 (2-4 cm and 22-24 cm) and S201 (0-2 cm and 22-24 cm) were measured replicated for quality control, because both cores with relatively low or even unquantifiable contents of Δ 5 -sterols (this scenario can be seen roughly in mass spectrogram) were particularly susceptible to large deviations caused by pronounced background noise. Estimated values of relative standard deviation of the three ratio pairs in the four samples (2-4 cm and 22-24 cm in S101, and 0-2 cm and 22-24 cm in S201) ranged from 4% to 10%, similar to those (<10%) reported by Nakakuni et al. (2017); Nakakuni et al. (2018), implying high precision of the analytical results in the present study.

Degradation-Corrected Downcore Sterol Concentration
Using an approach similar to (Middelburg, 1989) power model, log-log plots of degradation rate constant (k) vs. time since deposition (t) were proposed by Canuel and Martens (1996) to evaluate the influence of post-diagenesis on downcore sterol concentration according to the following equation: log k −1.12 × log t − 0.065, or k 0.86 × t −1. 12 (1) Further, given a known apparent initial age (a 0 ), the sterols corr (the initial sterols at the sediment surface before  post-depositional loss) for any downcore sample at depositional time t (sterols t ) can be solved using the following equations: The sterols t was initially termed as the sum of several sterols (Canuel and Martens, 1996); however, given almost similar degradation behaviors between individual sterols under anaerobic conditions (Sun and Wakeham, 1994;Canuel and Martens, 1996;Harvey and Macko, 1997;Sun and Wakeham, 1998;Grossi et al., 2001), the sterols t can be also utilized to calculate sterols corr for each individual sterol. The value of a 0 has been found to be several decades in coastal and shelf areas (Middelburg, 1989;Jia et al., 2013). For simplicity, an empirical 59-year value for a 0 calculated by Jia et al. (2013) for site A9 outside the PRE (see Figure 1 for location) was used in this study.
TABLE 2 | Sterol and stanol compounds identified in core S401 (see Figure 2 for TIC). Detailed information (i.e., structure, nomenclature, and source) of individual compounds can be found in, e.g., the works of Rampen et al. (2010) and Nakakuni et al. (2017Nakakuni et al. ( , 2018.

Contents of sterols (and individual sterols as demonstrated in
Values of the C 28 Δ 22 /Δ 5,22 ratio (and other two ratio pairs as demonstrated in Supplementary Table S3) oscillated in large amplitudes in cores S101 and S201 (ranged 0.7-5.7 and 1.2-13.3, FIGURE 5 | Linear correlation of sterols with n-C 14-18 (red points), n-C 26-30 (blue points), n-C 20:1 (green points) and br-C 15 (orange points) alcohols in the three cores.
Frontiers in Earth Science | www.frontiersin.org December 2021 | Volume 9 | Article 759317 7 respectively) but swung less variably in core S401 (ranged 0.4-1.1) (Figure 4). Distributions of the C 28 Δ 22 /Δ 5,22 ratio varied similarly in shape between cores S101 and S201 but changed diversely in terms of multi-year variations as compared to core S401 (Figure 4) likely due to the lowresolution data of the core (Table 1). Temporal variations of the C 28 Δ 22 /Δ 5,22 ratio exhibited no clear trends before ∼1980s; after that, however, there occurred a persistent decreasing trend in cores S101 (from 3.2 to 0.7) and S401 (from 1.1 to 0.4) or basically remained at relatively lower values in core S201 (mostly <3 with two exceptions) (Figure 4).
Moreover, contents of n-C 14-18 and n-C 20:1 alcohols exhibited strong positive correlations with sterols (and individual sterols as demonstrated in Supplementary Table S2) in the three cores ( Figure 5). The abundances of br-C 15 alcohols resembled to the n-alcohol ratios of C 15-17 /C 16-20 down the cores (Supplementary Figure S1) but correlated insignificantly with sterols (and individual sterols as demonstrated in Supplementary Table  S2) especially in cores S201 and S401 ( Figure 5).

Origin of Sterols and Their Application as Primary Production (PP)
The biological sources of the C 27 Δ 5,22 , C 27 Δ 5 , C 28 Δ 5,22 , and C 30 Δ 22 sterols in the study area are likely predominantly marine organisms. This is supported by the better correlations of sterols (and individual sterols as demonstrated in Supplementary Table S2) with marine-sourced n-C 14-18 alcohols than with terrigenous-originated n-C 26-30 alcohols in the three cores ( Figure 5). Presently, it is difficult to constrain the species-specific sources of these biological sterols, owing to their ubiquity in a wide range of marine organisms, including phytoplankton and zooplankton (Volkman, 1986;Volkman et al., 1998;Volkman, 2003;Rampen et al., 2010), which are abundant in the study area (Wang et al., 2011;Duan et al., 2014;Ren et al., 2020). However, the strong correlations among these sterols in the three cores (Supplementary Table S2) suggest common factors in modulating production of various phytoplankton and subsequent consumption by zooplankton in food chains. This is further supported by the strong correlations between sterols (and individual sterols as demonstrated in Supplementary Table S2) and n-C 20:1 alcohol ( Figure 5), a diagnostic biomarker for copepods (Kattner and Krause, 1989), which dominate zooplankton composition in the YDU area (Ren et al., 2020). Therefore, here, sterols, the sum of sterols produced by various phytoplankton (i.e., diatoms and dinoflagellates) and/or phytoplankton-dependent zooplankton, rather than individual sterol, are more suitable to roughly indicate PP in the study area. Nevertheless, the sterols loss due to degradation processes in the water column and/or sediments (e.g., Sun and Wakeham, 1994;Canuel and Martens, 1996;Prahl et al., 2000;Sinninghe Damsté et al., 2002;Wakeham et al., 2002;Hernández-Sánchez et al., 2014) may confound sterols burial fluxes to reflect initial export PP.
Present-day investigations in the YDU area reveal that total phytoplankton abundances are not highly variable at various depths of the upper 30-m water column (i.e., 78.9 × 10 2 unit in surface water vs. 50.7 × 10 2 unit at 20-to 30-m water depth; Wang et al., 2011). This occurrence should be due to that the photic zone (i.e., 0-30 m) can extend readily to the seafloor of the study sites (water depths <35 m; Table 1). Phytoplankton-derived lipid biomarkers could undergo least degradation in such shallow water depths, as even in deep-sea water columns, including the Arabian Sea (0-3,380 m; Prahl et al., 2000;Wakeham et al., 2002) and the South East Atlantic Ocean (0-100 m; Hernández-Sánchez et al., 2014), the loss of biomarkers (including sterols) has been found insignificant. Accordingly, the application of buried sterols to infer paleo-PP in the YDU area is feasible as long as degradation in sediments is properly considered.
Thus, we used a model proposed by Canuel and Martens (1996) to estimate the degradation loss of sterols in sediments and further to assess its influence on the downcore changes of sterols, similar with the approach for organic carbondegradation correction applied in the PRE  using the (Middelburg, 1989) model. It should be noted that the (Canuel and Martens, 1996) model is established for the anoxic environments of Cape Lookout Bight but is also likely applicable for coastal areas where aerobic conditions are usually confined to uppermost few-millimeter sediments (e.g., Hansen and Blackburn, 1991;Sun and Wakeham, 1994;Sun and Wakeham, 1998;Arndt et al., 2013). As illustrated in Figure 6, the initial sterols ( sterols corr ) changed almost similarly with the burial sterols, indicating that downcore sterols can be used to infer PP variations in the study area, despite the degradation loss could be as high as ∼24% in sediment depths.
According to Figure 4, the sterols with largely similar variation features in most parts of the three cores suggest common factors controlling PP in the YDU area over the past few decades. The change in abundances of algal sterols, which is associated with PP variability, has been also documented offshore the PRE (i.e., at site A9; Figure 1) over the past century , demonstrating, again, the reliability of burial sterols to reflect export PP in the coastal northern SCS. Moreover, the variability of PP offshore the PRE is considered as a result from Frontiers in Earth Science | www.frontiersin.org December 2021 | Volume 9 | Article 759317 the fluvial nutrient influx from the Pearl River . However, riverine nutrient inputs could be largely ruled out as a cause for PP variations in the study area due to minor influence of the Pearl River plume, as have been described in detail in our previous study (Zhu et al., 2018). This is further supported by the relatively lower sterol abundances in core S101 than in cores S201 and S401 ( Figure 3A; Supplementary Table S1), because core S101 was located closest to, whereas the other two cores were distributed further away from, the PRE (Figure 1). Our previous study on long chain diols derived DI-2 records in three cores revealed that the increasing trend in the upwelling intensity with inter-annual variability is a common phenomenon covering the YDU area over recent decades (Zhu et al., 2018). Thus, we deem that the upwelling-induced nutrients may have exerted an important effect on the variations of PP and sterol abundances at the sites.

Origin of Stanols
The diatom community is almost the exclusively dominant algae in the study area, accounting for ∼88% of the total phytoplankton composition with Proboscia alata being the major species, followed by T. nitzschioides, Pseudo-nitzschia pungens, and S. costatum (Wang et al., 2011). The S. costatum is predominantly abundant in the PRE (Qiu et al., 2010;Shen et al., 2011); however, C 28 Δ 24(28) that can be produced in vivo by S. costatum (Barrett et al., 1995;Rampen et al., 2010) has not been identified (or reported) in this region (Hu et al., 2009;Strong et al., 2012;Strong et al., 2013;Jia et al., 2013;Guo et al., 2019). The C 28 Δ 24(28) was not detected either in core S101 (closest to the PRE), suggesting that C 28 Δ 24(28) would be little produced by algae (i.e., S. costatum) in the coastal northern SCS. However, C 28 Δ 24(28) was identified, despite with minimal amounts (averaged 7 ± 4 ng/g; Supplementary Table S1) in core S401 (farthest to the PRE), thus indicating that anaerobic conversion of C 28 Δ 5,24(28) is more likely responsible for the presence of C 28 Δ 24(28) at the site. The contribution of T. nitzschioide to C 27 Δ 22 (Barrett et al., 1995;Rampen et al., 2010) in the three cores, however, could not be completely neglected based on the present study. Dinoflagellates, despite account for a minute fraction of the total phytoplankton community, are the second most abundant algae in the YDU area (Wang et al., 2011;Duan et al., 2014). The dominance of Scrippsiella sp. in dinoflagellate floras in the study area (Duan et al., 2014) likely suggests this species as a considerable contributor to C 27 Δ 0 . This is further supported by the much high values of the C 27 Δ 0 /Δ 5 ratio; i.e., an average of 32.7 in core S201 ( Figure 3C), as such a similarly high C 27 Δ 0 / Δ 5 ratio has been also documented in Scrippsiella sp. (Mansour et al., 1999). The possibility of dinoflagellate G. sanguineum as a biological source for C 28 Δ 22 (Mansour et al., 1999), however, could be largely ruled out, owing to the absence of this species in the YDU area (Wang et al., 2011;Duan et al., 2014). Similarly, some other marine organisms that have been reported to contain considerable fractional abundances of C 28 Δ 22 (Volkman et al., 1990;Nelson et al., 2000;Nelson et al., 2001) have also not been observed (or reported) in the study area (Wang et al., 2011;Duan et al., 2014;Ren et al., 2020).
Implication of the C 28 Δ 22 /Δ 5,22 Ratio for Bottom Water DO Condition As explained above, the C 27 Δ 22 /Δ 5,22 and C 27 Δ 0 /Δ 5 ratios are not suitable as reliable indicators for anaerobic transformation processes at the study sites, as T. nitzschioide and Scrippsiella sp. may contribute to C 27 Δ 22 and C 27 Δ 0 , respectively. In contrast, the insignificant contribution of living organisms to other 5αstanols (i.e., C 28 Δ 24(28) and C 28 Δ 22 ) suggests they are more likely FIGURE 6 | Comparison between sterols (red curves) and sterols corr (blue curves), and between C 28 Δ 22 /Δ 5,22 (orange curves) and C 28 Δ 22 /Δ 5,22 corr (green curves) in the three cores. The C 28 Δ 5,22 corr profiles were calculated using the method by Canuel and Marten (1996), as the similar approach for sterols corr .
Frontiers in Earth Science | www.frontiersin.org December 2021 | Volume 9 | Article 759317 derived from the anaerobic conversion of their Δ 5 -sterol counterparts. The C 28 Δ 22 /Δ 5,22 ratio is selected in this study because the three cores all contained detailed data of this ratio. However, another confounding factor should be ruled out before the application of the ratio as a redox indicator, i.e., the C 28 Δ 22 /Δ 5,22 ratio variations down the cores could be caused by differential degradation, given that 5α-stanols are more refractory than their respective Δ 5 -sterol counterparts (e.g., Nishimura and Koyama, 1977;Wakeham, 1989;Arzayus and Canuel, 2004;Bogus et al., 2012). Lines of evidence have revealed that biomarker (including sterols) degradation is insignificant even in deep-sea water columns as compared to in sediments (e.g., Prahl et al., 2000;Sinninghe Damsté et al., 2002;Wakeham et al., 2002;Hernández-Sánchez et al., 2014). Therefore, the potential loss of C 28 Δ 5,22 in the studied shallow water column (<35 m; Table 1) is not considered, but instead, its loss in sediments were evaluated. Here, we used the model by Canuel and Martens (1996) to estimate the initial C 28 Δ 5,22 (C 28 Δ 5,22 corr , as the case for sterols corr ) and then assessed the potential influence of degradation process on the downcore C 28 Δ 22 /Δ 5,22 variations. As illustrated in Figure 6, profiles of the corrected C 28 Δ 22 /Δ 5,22 (C 28 Δ 22 /C 28 Δ 5,22 corr ) followed those of buried C 28 Δ 22 /Δ 5,22 in three cores, indicating that degradation loss of C 28 Δ 5,22 in sediment depths is not responsible for the downcore changes in the C 28 Δ 22 / Δ 5,22 ratios. Therefore, by elimination, we deem the anaerobic conversion of C 28 Δ 5,22 is a more reasonable process to interpret past C 28 Δ 22 / Δ 5,22 variations. Previous studies have revealed that the 5α-stanol/ Δ 5 -sterol ratios increase sharply in the oxic-anoxic transition zone in waters (Wakeham, 1989;Wakeham et al., 2007;Berndmeyer et al., 2014;Wakeham, 2020) and sediments (Nishimura and Koyama, 1977). Conditions in such a zone are well suited to the development of a large and most active microbial population (Karl, 1978;Larock et al., 1979;Lin et al., 2008;Rodriguez-Mora et al., 2013;Berndmeyer et al., 2014). In the YDU area, the significant reduction of C 28 Δ 5,22 in the water column could be largely ruled out due to the absence of anaerobic waters (and thus oxic-anoxic transition zone), as revealed by monitoring records (Zhang et al., 2018; Figure 7) and satellite observations (the World Ocean Atlas 2018; https://www.nodc.noaa.gov/cgi-bin/OC5/SELECT/ woaselect.pl?parameter 3).
Thus, we believe the anaerobic conversion of C 28 Δ 5,22 should have occurred principally in sediments. The hydrogenation reduction of Δ 5 -steros has been found to be rapid in the oxicanoxic transition zone in microbiologically active sediments in the early sedimentation process, below which the hydrogenation rate is greatly attenuated (Nishimura and Koyama, 1977). This scenario may thus suggest that, in the coastal YDU, the greater part of C 28 Δ 22 originating from the hydrogenation of C 28 Δ 5,22 may have been produced mainly in the microbiologically active oxic-anoxic transition zone, which would lie in surface sediments. This is because 1) aerobic conditions are usually confined to the uppermost few-millimeter layer in coastal areas (e.g., Hansen and Blackburn, 1991;Sun and Wakeham, 1994;Sun and Wakeham, 1998;Arndt et al., 2013) and 2) microbial population is highest near the sediment surface and drops off steeply with depths in coastal marine sediments (e.g., Jørgensen and Revsbech, 1989;Sahm et al., 1999;Arndt et al., 2013). Moreover, comparison of surface sediments (≤2 cm) from different lakes with diverse bottom water redox potentials demonstrates an increasing trend in the 5α-stanol/Δ 5 -sterol ratios from relatively oxidizing (i.e., 0.36 at Lake Kizaki) to reducing (i.e., 0.66 at Lake Suigetsu) environments (Nishimura and Koyama, 1977). Similarly, the 5αstanol/Δ 5 -sterol ratios are found higher in surface sediments with comparatively lower bottom water DO levels in the Yangtze River estuary (Zhu et al., 2012) and the PRE (Guo, 2015), implying that the ratio records in downcore sediments may reflect bottom water (surface sediment) redox conditions in the past. This is followed by a recent study on short cores around Penguin Island, demonstrating that sedimentary variations of the 5α-stanol/Δ 5sterol ratios may be attributed to changes in bottom water conditions at the study sites at the time of deposition (Ceschim et al., 2016).
FIGURE 7 | Comparison of the C 28 Δ 22 /Δ 5,22 ratios (orange curve) in core S101 with bottom water DO concentrations (4-year running average; green curve) at proximate station MM13 (Figure 1) during 1991-2007. The annual mean DO concentration is estimated from monthly data collected by the Environmental Protection Department, Hong Kong (https://cd.epic.epd.gov.hk/EPICRIVER/marine). Note that monitoring DO data are available since the early 1990s, so given time span (before 2006) and resolution (averaged 3.6 years per 2-cm interval) in core S101 (Table 1), this figure shows a rough comparison between records of the C 28 Δ 22 /Δ 5,22 ratio and monitoring DO content (4-year running average) during 1991-2007.
Frontiers in Earth Science | www.frontiersin.org December 2021 | Volume 9 | Article 759317 Accordingly, in this study, sedimentary C 28 Δ 22 /Δ 5,22 ratios are likely also predominantly determined by bottom water DO conditions that modulate the extent of anaerobic conversion of C 28 Δ 5,22 at its initial deposition as a component of surface sediments. Here, DO condition is not an either-or concept between anaerobic and aerobic but a description of likely continuous change in DO concentration, although a common quantitative relation between the C 28 Δ 22 /Δ 5,22 ratio and DO concentration cannot be made at present and could be site specific. This is because the so-called anaerobic conversion has been found to take place under various redox conditions not strictly devoid of oxygen with a reported wide range of DO concentrations from 0 to ∼10.6 mg/L (Wakeham, 1989;Wakeham, 2020;Wakeham et al., 2007;Berndmeyer et al., 2014;Guo, 2015). This implies that in the YDU area the anaerobic conversion of C 28 Δ 5,22 may occur persistently and with varying degrees in surface sediments in response to variable bottom water DO concentrations. Given that the mean temporal resolutions of the three cores were ∼2-7 years per 2-cm sediment interval (Table 1), the C 28 Δ 22 /Δ 5,22 ratio in each sediment interval is effectively a result of multi-year accumulated anaerobic conversion processes. Such an integrated continuous accumulation record would dampen or smooth out greater variations on shorter time scales (e.g., monthly and seasonal) and thereby should tend to reflect longer-term (i.e., multi-year) mean state of DO conditions. This point is supported by previous studies showing that the 5α-stanol/Δ 5 -sterol ratios are not reflective of seasonal-biased DO signals in surface sediments, where bottom water experience dramatically seasonal DO variability (Nishimura and Koyama, 1977;Zhu et al., 2012;Guo, 2015), implying that seasonal DO dynamics is hard to be recorded in sedimentary C 28 Δ 22 /Δ 5,22 ratios. Therefore, we believe multi-year mean state of bottom water DO conditions should be responsible for sedimentary C 28 Δ 22 /Δ 5,22 records in the study area. This is consistent with a rough comparison between the C 28 Δ 22 /Δ 5,22 ratios in core S101 (3.6 years per 2-cm sedimentary interval; Table 1) and 4-year averaged bottom water DO variations at proximal monitoring station MM13 (Figure 1) during 1991-2007, showing higher C 28 Δ 22 /Δ 5,22 ratio values corresponding with lower DO concentrations (Figure 7). According to the ratio records down the three cores (Figure 4), DO concentrations varied at each site but similarly in shape between sites S101 and S201 with no clear temporal trend before ∼1980s; after that, DO concentrations increased gradually (i.e., at sites S101 and S401) or shifted to a relatively abundant state (i.e., at site S201).

Potential Factors on Bottom Water DO Variations
Our novel sterol data ( Figure 3A; Supplementary Table S1), together with already published diol data (Zhu et al., 2018), reinforce that the Pearl River plume contributes insignificantly to the study area, consistent with a recent study suggesting the influence of runoff from the Pearl River only within a certain range (Xu et al., 2020). Therefore, the input from the Pearl River is not considered as the most important driver on bottom water DO dynamics, but instead, the climate-induced upwellinginvolved physical-biogeochemical processes (i.e., source water, wind stress, oceanic warming, photosynthetic production, and aerobic consumption) should be important to influence past changes in bottom water DO concentrations in the YDU area. These upwelling-involved processes have also been proposed as potentially important drivers on bottom water DO dynamics in many other upwelling regions without direct riverine influence (e.g., Fennel and Testa, 2019 and references therein).
The substantial impact of upwelling intensity on DO condition has been observed in upwelling regions off Peru, showing overall enhanced oxygenation in response to gradually intensified upwelling, which is attributed to a subtle oxygenation trend in the upwelled source waters (Cardich et al., 2019). The advection of upwelled source waters has been also proposed as a dominant factor controlling DO variability in many other upwelling regions (e.g., Grantham et al., 2004;Monteiro et al., 2006;Monteiro et al., 2008;Chan et al., 2008;Mohrholz et al., 2008). However, a 29year observation across the northern SCS shelf area during 1976-2004 reveals a significant decrease in DO concentrations at various depths of the upper 200-m water column (Ning et al., 2009), thus likely transporting reduced-oxygenated source waters from outer shelf to inner coastal area at times of upwelling over recent decades. This scenario would rather decrease bottom water oxygenation and thus promote the anaerobic conversion of C 28 Δ 5,22 in more reducing surface sediments, leading to elevated C 28 Δ 22 /Δ 5,22 ratios in recent decades that is in contrast with our present records. Therefore, the physical supply of source waters from outer shelf due to upwelling activity is not likely the major cause for bottom water DO variability in the study area.
Increases in southwesterly wind stress have been simulated to decrease bottom water hypoxia by eroding vertical stratification and aerating deep waters in the PRE (Wei et al., 2016;Lu et al., 2018). However, a time-series (2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015) observation in the nearby Mirs Bay reveals that larger hypoxic areas in bottom waters are often observed during years with longer-lasting southwesterly wind as a result of enhanced stratification, extended residence time of bottom waters, and onshore transport of low-oxygenated source waters induced by stable upwelling (Zhang et al., 2018). This inconsistency between the two proximal regions just to the west of YDU suggests that the actual impact of local southwesterly wind on DO condition may be highly site specific in the coastal northern SCS. Specific to our study sites, the wind stress appears to exert a minor effect on bottom water DO condition, as the C 28 Δ 22 /Δ 5,22 ratio correlated insignificantly with DI-2 inferred local wind stress in the three cores (Supplementary Figure S2).
Oceanic warming has a high potential to enhance stratification, decrease the DO solubility, and increase the rate of organic matter mineralization, thus eventually increasing oxygen deficiency and causing low-oxygen condition in bottom waters (Ning et al., 2009;Keeling et al., 2010;Schmidtko et al., 2017;Breitburg et al., 2018;Fennel and Testa, 2019). However, the important role of oceanic warming in regulating bottom water DO condition could be largely ruled out in the study area, because the C 28 Δ 22 /Δ 5,22 ratio changed Frontiers in Earth Science | www.frontiersin.org December 2021 | Volume 9 | Article 759317 uncorrelated with LDI-derived annual mean SST in three cores (Supplementary Figure S2). Similar findings suggesting less importance of oceanic warming on DO condition in the coastal area than in the open ocean have been also made in other studies (e.g., Breitburg et al., 2018;Fennel and Testa, 2019 and references therein). The above difficulties in the interpretation on the C 28 Δ 22 /Δ 5,22 ratio-inferred bottom water DO variability prompt us to consider the possible role of biogeochemical processes, because we notice that the C 28 Δ 22 /Δ 5,22 ratios correlate substantially reversed with the sterols abundances in three cores (Figure 4). This occurrence suggests that bottom water DO concentrations change positively with total PP in the YDU area, which is compatible with recent studies focused on a small area of the western YDU, showing that low-oxygen areas do not coincide with the regions of high PP (Wei et al., 2016;Zhang et al., 2018). This phenomenon may be related with the competitive results of the twofold impacts induced by elevated PP in shallow, clear FIGURE 8 | Comparison of n-C 20:1 /C 28 Δ 5,22 (blue curves) and br-C 15 /C 28 Δ 5,22 (red curves) with C 28 Δ 22 /Δ 5,22 (green curves) in the three cores.
FIGURE 9 | Schematic diagram of the biogeochemical mechanism of photosynthetic production in determining bottom water DO variations in surface sediments in the shallow, clear YDU area, showing (A) decreased bottom water oxygenation (−) and more reducing surface sediment (+) indicated by higher C 28 Δ 22 /Δ 5,22 ratio (+) in response to lower PP indicated by less abundant sterols (−), and (B) increased bottom water oxygenation (+) and less reducing surface sediment (−) indicated by lower C 28 Δ 22 /Δ 5,22 ratio (−) in response to higher PP indicated by more abundant sterols (+). Here, anaerobic conversion of C 28 Δ 5,22 is setting to occur principally in the oxic-anoxic transition zone in microbiologically active surface sediment, and thus, the C 28 Δ 22 /Δ 5,22 ratio at each sedimentary interval is mainly governed by bottom water DO level at the time of deposition. For detail in, e.g., figure (B), higher abundances of phytoplankton can fuel more abundant zooplankton ( Figure 5) in the total photic water column; however, higher amounts of phytoplankton (and phytoplankton-dependent zooplankton) derived relatively labile organic matter may not always lead to increased microbial biomass (i.e., at sites S201 and S401; Figure 5). Regardless of certain degree of DO consumption by zooplankton respiration and microbial remineralization, the dominant role of DO production by phytoplankton photosynthesis in the photic zone ( Figure 8) may still lead to higher extent of net bottom water DO penetrating into surface sediment, resulting in less reducing condition and thus lower sedimentary C 28 Δ 22 /Δ 5,22 ratio. waters (Fennel and Testa, 2019); i.e., exacerbated oxygen depletion due to high rate of respiration and mineralization, and sufficient oxygen supply through photosynthetic production.
To give a rough assessment of the competitive results between the biogeochemical oxygen sink and source, we compared the records of zooplankton and microbes (consumers) with records of phytoplankton (producers). As illustrated in Figure 8, distributions of the n-C 20:1 /C 28 Δ 5,22 and br-C 15 /C 28 Δ 5,22 ratios, indicative of the relative abundances of zooplankton and microbes compared with phytoplankton, respectively, changed almost in parallel with the C 28 Δ 22 /Δ 5,22 ratios in the three cores. These occurrences suggest that, in our studied shallow-water area with less turbidity due to minor influence of the Pearl River plume (Zhu et al., 2018; this study), the oxygen production via photosynthesis may have exceeded biological oxygen consumption by zooplankton respiration and microbial remineralization, leading to a strong positive relationship between water column PP and bottom water (surface sediment) DO ( Figure 4). Figure 9 illustrates general scenarios of the dominant role of photosynthetic production in the photic water column (i.e., 0-30 m; Wang et al., 2011) just above the seafloor of the study sites (water depths <35 m; Table 1) in determining bottom water redox conditions in surface sediments. We suggest that net photosynthetic oxygen production outweighs upwelling-involved source water-and oceanic warming-induced deoxygenation in bottom waters in the YDU area. We believe this phenomenon may also occur elsewhere with similar coastal upwelling conditions, such as shallow, clear waters with minor influence of riverine terrigenous input. However, the dominance of biogeochemical forcing on bottom water DO condition may be greatly attenuated in large river-impacted upwelling regions. For example, in the Mirs Bay (Figure 1) that is the western boundary of YDU affected by the Pearl River plume, the physical processes (i.e., freshwater discharge), rather than biogeochemical dynamics, are proposed as the most important drivers for inter-annual bottom water DO variability (Zhang et al., 2018).

CONCLUSION
A series of sterols and stanols were identified, and the 5αstanol/Δ 5 -sterol ratios were examined for their applicability for redox reconstruction in sediment cores in the YDU area in the coastal northern SCS. The sterols abundances with roughly similar variation features were observed in most parts of the three cores, suggesting common factors controlling PP, which may be related to upwelling-induced nutrients over the past few decades. The C 27 Δ 22 /Δ 5,22 and C 27 Δ 0 /Δ 5 ratios were not so reliable to reflect redox conditions because marine organism may produce these two stanols in vivo. However, some other ratio pairs such as the C 28 Δ 22 /Δ 5,22 ratio can be accepted as a feasible redox indicator to infer changes of redox conditions in surface sediments, which are largely dependent on bottom water DO levels. The C 28 Δ 22 /Δ 5,22 ratio records in three cores showed oscillations in varying degrees and exhibited no clear trends in bottom water DO concentrations before ∼1980s but indicated a persistent increasing trend in oxygenation (at sites S101 and S401) or basically shifted to relatively higher DO level (at site S201) since then. This occurrence could be caused by the changes in net ecosystem production, as indicated by the n-C 20:1 /C 28 Δ 5,22 and br-C 15 /C 28 Δ 5,22 records that changed similarly with the C 28 Δ 22 /Δ 5,22 records in the three cores. This study may provide a perspective on the responses of bottom water DO variation to upwelling and climate change in the future in shallow, clear coastal upwelling regions.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material; further inquiries can be directed to the corresponding author.