Sansha Bay (26°30′–26°58′E, 119°26′–120°10′N) is located on the northeast coast of Fujian Province, China ( Figure 1 ). The surface current outside Sansha Bay is influenced by the CCC in winter and the South China Sea Warm Current (SCSWC) in summer. In late autumn and winter, seawater from the East China Sea or from farther north of China is transported along the coast under the influence of winter winds. In addition, observations based on radium isotopes (²²⁶Ra and ²²⁸Ra) have shown that groundwater input in Sansha Bay also affects the nutrient flux (Wang et al., 2018). Seasonal activities such as seeding, growing, and harvesting on the densely clustered floating mariculture mats in the bay also have a substantial effect on the spatial–temporal distribution of its hydrochemical parameters (Wang et al., 2018). In addition, driven by the Asian monsoon, the hydrographic characteristics of Sansha Bay were significantly influenced by the plume water of the Jiao River, which showed the highest freshwater discharge rate in warm months (from April to September) and the lowest in cold months (from October to April the following year). Han et al. (2021) observed a water mass with relatively lower salinity than CCC meandering in Sansha Bay and influenced by intensive mariculture activities, nutrient concentrations in this Bay were much higher than other water masses in this coastal area.

A research cruise was carried out in Sansha Bay on 11–12 December, 2021 ( Figure 1 ). Water samples were collected at 25 stations using a 5L Niskin bottle guided by a Conductivity–Temperature–Depth (CTD; Seabird^® WQM 2019) recorder, which simultaneously measured sea surface temperature (SST) and salinity (SSS). Salinity in the water column samples was determined using a portable salinometer (Portasal 8410A, Guildline Co., Canada) and used to calibrate data from the CTD recorder. Water samples were obtained from two or three depth layers at each sampling site, depending on the overall water depth. “Surface waters” refer to waters 1 m below the surface, and “bottom waters” denote waters 1 m above the sediment bed.

Nutrient samples were filtered using cellulose acetate membranes with a pore size of 0.45 μm that had been acid-cleaned. Two hundred and fifty mL of filtered seawater was frozen and stored at −20°C and used for routine spectrophotometric analysis of NO₃ ⁻, NO₂ ⁻, PO₄ ³⁻, and Si(OH)₄ concentrations using the Technicon AA3 automatic analyzer (Bran-Lube, GmbH; Han et al., 2021). Concentrations of NO₃ ⁻ + NO₂ ⁻ were determined using a Cu–Cd column reduction method, and NO₂ ⁻ contents were determined using spectrophotometry with standard pink azo dye (Dai et al., 2008). Concentrations of PO₄ ³⁻ and Si(OH)₄ were measured using standard spectrophotometric methods (Knap et al., 1996). The detection limits for NO₃ ⁻ + NO₂ ⁻, PO₄ ³⁻, and Si(OH)₄ were 0.1, 0.08, and 0.08 µM, respectively. The relative standard deviations (RSD) of repeat measurements of selected samples were<5%.

Isotope analyses of NO₃ ⁻ were carried out according to the method of Sigman et al. (2001) and Casciotti et al. (2002) (Equations 1 and 2) using an isotope-ratio mass spectrometer (IRMS). The international nitrate reference materials NITS USGS34 (δ¹⁵N = −1.8‰; δ¹⁸O = −27.9‰) and NITS USGS35 (δ¹⁵N = 2.7‰; δ¹⁸O = 57.5‰) were used to correct for drift, oxygen isotopic exchange, and blanks. The average standard deviation was typically 0.2‰ for δ¹⁵N and 0.5‰ for δ¹⁸O, which is applicable to samples with NO₃ ⁻ concentrations ≥ 1 μM. Isotope ratios are reported in delta (δ) notation in units of per mil (‰):

where ( 15 N / 14 N ) r e f e r e n c e denotes N₂ in air and ( 18 N / 16 N ) r e f e r e n c e denotes Vienna Standard Mean Ocean Water (VSMOW).

Phytoplankton absorbs nutrients from seawater according to the Redfield ratio (Redfield, 1963), leaving relative excess of nitrogen or phosphorus. Excess nitrogen or phosphorus do not have direct contribution to eutrophication but could be considered as a potential factor, known as potential eutrophication (Sun et al., 2006). In order to highlight the limiting characteristics of nutrient salts, dissolved inorganic nitrogen (DIN, including NO₃ ^-, NO₂ ^- and NH₄ ⁺) and active phosphate (PO₄ ^3-), which play a bottleneck role in phytoplankton growth (Zhuang et al., 2021b), were selected as evaluation parameters. The potential eutrophication evaluation model ( Table 1 , Guo et al., 1998; Sun et al., 2016; Yang et al., 2020) was adopted for the evaluation of the nutrient condition in the study area.

To evaluate the nitrate deficit, we used the parameter N* proposed by Gruber and Sarmiento (1997), where N* = ([NO₃ ⁻] − 16*[PO₄ ³⁻] + 2.9) × 0.87. In the equation, 16 is the Redfield ratio of 16:1, and 0.87 is a calculated factor based on the specific stoichiometric ratios in denitrification and remineralization processes. The constant (2.9) was introduced to set the mean N* to zero and is based on global measurements (Gruber and Sarmiento, 1997). The N* index reflects the net effect of N₂ fixation and denitrification, and negative N* values imply a nitrate deficit in the ocean (Deutsch and Weber, 2012).

SIAR (stable isotope analysis in R) is a software package that uses a Bayesian stable isotope mixing model, which is used to calculate the relative proportion of various nitrate sources. In the mixing model, the Bayesian framework is utilized to calculate the probability distribution amongst the different nitrate sources. The model framework is as follows:

where X i j denotes the isotope values (j = 2, δ¹⁵N-NO₃ ⁻, and δ¹⁸O-NO₃ ⁻) of the sample i (i = 1, 2, 3, … N); S j k is the isotope value j of the source k (k = 1, 2, 3, … K) and is normally distributed with an average μ j k and standard deviation ω j k ; P_k is the proportion of source k, as calculated using the SIAR model; C j k is the fractionation factor for j on source k and is normally distributed with an average λ j k and standard deviation τ j k ; ϵ_jk is the residual error of the additional unquantified variations between individual samples and is normally distributed with an average 0 and standard deviation σ_j. The model uses CSV Microsoft Excel of δ¹⁵N-NO₃ ⁻ and δ¹⁸O-NO₃ ⁻, S j k , and C j k as inputs. It then outputs numerical and graphical depictions of the relative contributions of the potential sources (Zhang et al., 2018) (Equation 3). More detailed information of the Bayesian stable isotope mixing model has been provided by Moore and Semmens (2008); Xue et al. (2009), and Zhang et al. (2018).

The winter waters in Sansha Bay, like most coastal harbors affected by the CCC in southeastern China, appear to be characterized by eutrophication (e.g., Cai et al., 2013; Yang et al., 2018). Approximately 81% of the data points (50/62) in this study exceeded the thresholds for eutrophication in harbor waters proposed by Guo et al. (Guo et al., 1998; NO₃ ⁻ >21.4 µM, PO₄ ³⁻>1.45 µM), indicating that most of the winter water in Sansha Bay was eutrophic ( Figure 4 ). The small amount of winter run off from rivers into Sansha Bay and another area of low surface nutrient concentrations was found at the mouth of the Sai River, suggesting that winter river input does not greatly influence nutrient levels in the bay. The lowest nutrient concentrations in the surface and bottom waters were recorded at the mouth of Sansha Bay., indicating that offshore water input may not be the main source of nutrients into the bay. Areas of high nutrient concentrations were observed mainly in mariculture areas, indicating that mariculture activities contribute to nutrient levels in the bay. The influence of the winter CCC extends from 26°N to 35°N along the southeastern coast (Wang et al., 2003). As shown in Figure 5 , the NO₃ ⁻/PO₄ ³⁻ ratios in CCC-influenced harbors along the southeastern coast of China generally decrease from north to south. Owing to the input of nutrients from the Yangtze River, NO₃ ⁻/PO₄ ³⁻ ratios can exceed 80 in the winter waters of the Yangtze River estuary (Liu et al., 2009). The CCC carries signals from the land sources of the Yangtze River and a large volume of nutrients to the southeastern coastal harbors during winter, which is one of the main reasons for the high N/P ratios in these harbors (Yang et al., 2018). The NO₃ ⁻/PO₄ ³⁻ ratios in Hangzhou Bay, Xiangshan Bay, and Sanmen Bay are relatively similar (Cai et al., 2013; Wu et al., 2020) and lower than those of the Yangtze River estuary, but they further decrease to 14.3 ± 2.2 in Sansha Bay ( Figure 5 ; Supplementary Table 1 ).

The winter waters of Sansha Bay have lower NO₃ ⁻/PO₄ ³⁻ and NO₃ ⁻/DSi ratios than many other nearshore harbors influenced by the CCC ( Supplementary Table 1 ). Correlation analysis showed significant positive relationships between concentrations of NO₃ ⁻ and both PO₄ ³⁻ and DSi ( Figure 4 ). The relationship between NO₃ ⁻ and PO₄ ³⁻ concentrations was examined by linear regression and indicates that NO₃ ⁻ concentrations increased with increasing PO₄ ³⁻; however, the slope was only 11.5, lower than the Redfield ratio ( Figure 4A ). Similarly, NO₃ ⁻ concentrations increased with increasing DSi, with a slope of 0.9 ( Figure 4A ). These results indicate that the winter waters of Sansha Bay are eutrophic but have relatively low N/P ratios.

The distribution of the N* index reflects the excess (positive values) or deficiency (negative values) of nitrate relative to phosphate (Deutsch and Weber, 2012). The N* values decrease gradually along the southeastern coast, with values approaching 0 in Sansha Bay, indicating that strong denitrification processes occur as the excess nitrate flows southward and enters coastal harbors. This may be related to intense nitrogen removal via sedimentary denitrification (Li et al., 2021). Indeed, observations show that denitrification at the mouth of the Yangtze River is the main pathway for the removal of excess nitrogen, with nitrogen removal rates reaching up to 28.49 ng N g⁻¹·h⁻¹ (Li et al., 2021).

Based on the dual isotope method for nitrate source identification (Kendall et al., 2007; Xue et al., 2009), the δ¹⁵N-NO₃ ⁻ (8.8‰–11.9‰) and δ¹⁸O-NO₃ ⁻ (2.2‰–6.0‰) values in Sansha Bay waters ( Figure 6 ) suggest that the main sources of nitrate are manure and sewage, as evidenced by the more positive δ¹⁵N-NO₃ ⁻ values. The δ¹⁵N-NO₃ ⁻ values are lower at the mouths of the main rivers entering the bay ( Figure 2 ) when the river input is low ( Supplementary Figure 3 ), suggesting that the more positive anthropogenic signal may be introduced by mariculture activities within the bay. Previous studies have suggested that the main sources of nutrients in Sansha Bay during winter are the eutrophic coastal currents and mariculture inputs (Han et al., 2021); however, the extent of the latter depends heavily on the type of mariculture. Jiang et al. (2020) found that seaweed aquaculture in Xiangshan Bay absorbed nutrients and alleviated eutrophication. The type of mariculture may also affect the distribution of dual nitrate isotopes. In Xiangshan Bay, where seaweed aquaculture is dominant, δ¹⁵N-NO₃ ⁻ and δ¹⁸O-NO₃ ⁻ values range from 5.7‰ to 8.8‰ and from 1.8‰ to 6.8‰, respectively (Yang et al., 2018). In Laizhou Bay, where sea cucumber aquaculture is dominant, δ¹⁵N-NO₃ ⁻ and δ¹⁸O-NO₃ ⁻ values range from 2.0‰ to 11.8‰ and from −7.8‰ to 12.6‰, respectively (Kang and Xu, 2016).

The application of Bayesian mixing models reveals that the surface water in Sansha Bay has a mixture of sources ( Supplementary Figure 4 ), which poses a challenge for nitrate source analysis. According to previous studies, the main nitrate sources in water are likely manure and sewage, reduced nitrogen fertilizer, nitrate derived from soil nitrogen, and atmospheric deposition (Dai et al., 2008; Ye et al., 2016). The relative contributions of these four potential nitrate sources have been calculated in many studies across a broad range of study areas (e.g., Moore and Semmens, 2008; Xue et al., 2012; Zhang et al., 2018; Lao et al., 2019). Therefore, we calculated the potential sources of nitrate using δ¹⁵N-NO₃ ⁻ and δ¹⁸O-NO₃ ⁻ values in a Bayesian isotope mixing model (Chen et al., 2022b). The uncertainty associated with these calculations is difficult to constrain with the present data because (a) the range of isotopic values from potential nitrate sources is relatively large ( Supplementary Table 2 ), and (b) the end-members of the potential nitrate sources originate from the reference literature and not from the measurement of samples in Sansha Bay. Nevertheless, such an estimate can provide an insight into nitrate sources in Sansha Bay. The end-member values for the potential nitrate sources are presented in Supplementary Table 2 , and the results are shown in Figure 6 . The results suggest that manure and sewage were the predominant sources of nitrate in the surface waters of Sansha Bay (51%–75%, average 63%), followed by soil organic nitrogen (3%–45%, average 26%) and reduced nitrogen fertilizer (0%–24%, average 11%).

Mariculture activities are commonplace in the harbors along the southeastern coast of China and include kelp farming in Xiangshan Bay (Yang et al., 2018), mixed crab and seaweed farming in Sanmen Bay (Cai et al., 2013), and mixed shellfish and seaweed farming in Sansha Bay (Han et al., 2021). Most mariculture activities increase nutrient loading in water owing to the excretion of farmed organisms and the addition of feed (Bouwman et al., 2013). In this study, the higher nutrient concentrations in Sansha Bay compared with the bay mouth suggest nutrient loading from mariculture. However, the low NO₃ ⁻/PO₄ ³⁻ ratios in Sansha Bay compared with the bay mouth indicate the possibility of denitrification in the bay, which may be caused by the structural modification of nutrients, potentially owing to seaweed farming. Zhang et al. (2022) suggested that seaweed farming may increase N/P ratios, whereas bivalve farming may decrease N/P ratios. Overall, the N/P ratio of nutrients removed by mariculture is close to 18. In Sansha Bay, where mixed shellfish and seaweed farming occurs, excess nitrate (relative to phosphate) may be absorbed during mariculture activities under conditions of sufficient nitrogen, and excess nitrogen may be removed from the water during the harvest season (Han et al., 2021).

Previous studies have shown that when algae have high rates of nitrate assimilation, the δ¹⁵N of the remaining nitrate will increase significantly owing to the preferential utilization of ¹⁴N by the organism (Ahad et al., 2006; Chen et al., 2013). Average δ¹⁵N-NO₃ ⁻ values in Sansha Bay during winter were as high as 9.8‰ ± 0.6‰, higher than the spring ratios measured in the Yangtze River Estuary (6.9‰, Chen et al., 2013) and the Pearl River Estuary (2.2‰–4.4‰, Chen et al., 2022b). These values are also higher than those measured in the upper bay area of Xiangshan Bay, where kelp farming is predominant (8.1‰ ± 0.5‰), and the lower bay area, which is significantly influenced by dilution via input from the Yangtze River (6.2‰ ± 0.3‰) (Yang et al., 2018). These results indicate that Sansha Bay experiences strong biological absorption and transformation of nitrate.

The intense biological production and deposition of organic debris in the surface layer during seaweed cultivation provides abundant organic matter to the sediment (Zhuang et al., 2022). The re-mineralization of this organic matter may enhance sedimentary denitrification (Wu et al., 2021), thus removing more nitrogen (while phosphates remain unaffected) and reducing N/P ratios. Denitrification enriches the residual nitrate, with δ¹⁵N and δ¹⁸O having a ratio close to 2:1 (Kendall, 1998; Mengis et al., 1999). The higher δ¹⁵N-NO₃ ⁻ and δ¹⁸O-NO₃ ⁻ values in the bottom waters of the mariculture area in Sansha Bay compared with the surface waters support the idea that denitrification processes are occurring in the sediment ( Figure 3 ). These findings suggest that in areas where shellfish and seaweed are cultivated, the strong absorption of nitrate by organisms and enhanced sedimentary denitrification stimulated by the downward settling of organic matter result in the reduction of N/P ratios in the water.