The Nakdong River is situated in the southeastern portion of the Korean Peninsula and is the second longest river in the country ( Figure 1 ), with a main channel spanning 526 km and a 23,817 km² total catchment area. River discharge was significantly regulated by a massive 2.4 km long-barrage system constructed at the river mouth 1983–1987 (An, 2021). The barrage system comprises six main gates (width = 47.5 m, height = 9.2 m), and four regulating gates (width = 47.5 m, height = 8.3 m). The main gates were kept closed to prevent seawater intrusion. In contrast, the regulating gates remain open during low tide and are closed when the water level difference is < 0.2 m (Yoon et al., 2016; An, 2021). Freshwater discharge follows three phases based on precipitation: a dry period with less discharge from September to March (approximately 100 m³ s⁻¹), a normal period with moderate discharge from April to June (approximately 500 m³ s⁻¹), and a wet period with high discharge from July–September (exceeding 1200 m³ s⁻¹). The annual precipitation is approximately 1500 mm and is concentrated in the summer. The temperature ranges from 4 to 25°C, with distinct seasonal variation. The surface sediment is dominated by muddy sand (~10% mud), with a 2.6–3.9 φ mean grain size (Choy et al., 2008). The estuary experiences a dominant semidiurnal tide, with a tidal range varying from 0.4 m during low tide to 2.0 m maximum during high tide.

Short-term dam opening experiments were conducted in 2019, while long-term dam opening experiments were carried out in 2020 and 2021 (Brevé et al., 2019; Baird et al., 2020; An, 2021; van den Tweel et al., 2021). In 2019, two short-term dam opening experiments took place on June 6 and September 17, involving approximately 0.64 and 1.01 million tons of seawater respectively, flowing upstream of the Nakdong River for 38 minutes and over 1 hour (An, 2021). Following that, long-term dam opening experiments were conducted in both 2020 and 2021. Since 2020, the dam has consistently remained open for one lunar tidal cycle, around 1 month. This period encompasses two tidal cycles, encompassing both high and low tides, with seawater intrusion occurring exclusively during high tides (An, 2021). In this study, we define the long-term seawater intrusion cycle as seawater intrusion during the first high tide of week 1 (SIW 1), followed by seawater intrusion during the first low tide (SIW 1-1), and then repeating similarly during the second week (SIW 2, and SIW 2-2). The first long-term dam opening experiment (third overall, including short-term trials), occurred in June 2020 (June 4–8 = SIW 1; June 19–25 = SIW 2), resulting in a total of 9.3 million tons of seawater flowing into the river (An, 2021). The long-term dam opening method was explored in April, June, August, and October 2021, coinciding with approximately 1.79, 2.07, 2.24, and 2 million tons of seawater intrusion into the river in each respective lunar tidal cycle. As a result of the experimental dam openings in 2019, 2020, and 2021, the National Water Committee decided to permanently open the floodgates (An, 2021). The long-term dam opening method was implemented for the monthly dam opening operations throughout 2022, aligning with the lunar tidal cycle.

The field monitoring and sampling as well as subsequent laboratory experiments were conducted in September 2019 (second dam opening, short-term), June 2020 (third dam opening, long-term), and October 2021 (seventh dam opening, long-term), based on the specific dam opening experiments and corresponding seawater intrusion phases ( Table 1 ). The seawater intrusion phases can be divided into before the dam opening (BDO), seawater intrusion during the first high tide of week 1 (SIW 1), seawater intrusion during the first low tide of week 1 (SIW 1-1), and after the dam opening (ADO). S6.5, located 6.5 km upstream from the dam (the farthest point where seawater can reach in the first dam opening experiment), represents the station affected by each seawater intrusion and was evaluated from 2019 to 2021 ( Figure 1 ). S11, situated 11 km upstream from the dam, was included in the assessment in 2021, indicating a station where seawater intrusion had negligible effects. Details of the sites visited during each seawater intrusion event are listed in Table 1 .

Depth profiles or bottom-water temperature, salinity, Chl-a, and DO were acquired using an AAQ Series H-11 multiprobe system (JFE Advantech Co., Ltd.). Intact depth profiles of water quality data were only available in 2020, and 2021. Bottom-water samples were collected in a 5 L Niskin bottle for subsequent incubation experiments and nutrient analysis. Water samples for nutrient analysis were filtered immediately upon collection through glass fiber GF/F filters (Whatman International, Maidstone, Kent, UK) and transferred into 50-mL conical sterile polypropylene centrifugal tubes (Thermo Fisher Scientific, USA).

Sediment samples were collected from the sediment cores in 2019 and grabs in 2020, and 2021. Intact sediment cores and overlying water were collected using acrylic cores (small: internal diameter = 4 cm, height = 23 cm; large: internal diameter = 8 cm, height = 33 cm) by scuba divers with minimal disturbance to the sediment–water interface (An and Joye, 2001). The cores contained approximately 8–10 cm of sediments. Upon collection, the sediment cores and water samples were immediately sealed with polyvinyl chloride caps and transported to the laboratory (within 2–3 h) for subsequent experiments. Surface sediment (0–10 cm) samples were collected using a grab sampler, sealed in sterile plastic bags, and stored in a 4°C ice box before transportation while the sediment samples for metagenomic analysis were stored in 50 mL sterile tubes and frozen with dry ice or liquid nitrogen at the sampling sites. An intact core incubation was conducted in 2019 only to measure the actual benthic N transformation rates and SOD during seawater intrusion. Anaerobic slurry incubation was conducted during each cruise from 2019–2021 to measure the potential denitrification, anammox, and DNRA rates. Aerobic slurry incubation was conducted in 2020 to assess the nitrification potential.

The filtered bottom-water samples were stored in a −80°C freezer until nutrient analysis. Dissolved inorganic NH₄ ⁺, NOx (NO₃ ^- plus NO₂ ^-), and orthophosphate (PO₄ ^3-) concentrations were determined using standard methods with a spectrophotometer (Shimadzu, UV-1650PC, Japan) (Strickland and Parsons, 1972). Sediment characteristics, including water content (%), grain size proportion > 63 µm (sand content %), and sediment organic matter content (LOI %), were measured using a large core sampler in 2019 and a grab sampler after 2020. Sediment cores were sliced from 0–10 cm at 1 cm intervals and then mixed. Sediment characteristics were measured as previously described (Huang and An, 2022).

Intact sediment cores with their upper caps removed and in situ bottom-water (approximately 20 L) were pre-incubated overnight in the dark at in situ water temperature and O₂ saturation in an incubation tank. The tank was amended with Na¹⁵NO₃ ^- tracer (98–99% atom %, Sigma-Aldrich) at the beginning of the pre-incubation to reach a final concentration of 100 μmol L^-1. Immediately after the pre-incubation, the intact cores were incubated for approximately 24 h. Duplicate cores were subsequently used to determine the oxygen concentrations at 0, 1, 2, and 24 h using a membrane inlet mass spectrometer (MIMS) (Kana et al., 1994; Deng et al., 2015; Huang and An, 2022). As described above, the water samples were filtered and preserved for DNRA analysis at each time point.

Slurry incubations were conducted according to Thamdrup and Dalsgaard (2002) and Shan et al. (2016) to determine potential denitrification, anammox, and DNRA rates, and their relative contributions to total NO₃ ^- reduction. First, homogenized subsamples (0–10 cm) from the fresh sediment were sieved through a 1.0 mm copper sieve to remove visible shells or roots (Wang et al., 2020), and 3 g of sediment and He-purged bottom-water were transferred to 40 mL gas-tight glass vials (Exetainer, Labco, UK), which were sealed immediately to prevent bubbles or headspace. After being pre-incubated overnight to consume the residual O₂ and NOx, the slurries received one of three different tracer combinations (¹⁵NH₄Cl, ¹⁵NH₄Cl plus K¹⁴NO₃, or Na¹⁵NO₃, 98–99% atom %, Sigma-Aldrich) to a final concentration of approximately 100 μmol L^{-1  15}N in each vial. Triplicate vials were sacrificed at 0, 24, 48, and 72 h after 18–24 h, when the ¹⁵N tracer reached equilibrium. N₂ gas production was immediately analyzed by MIMS system, and ¹⁵NO₃ ^–amended slurry samples were used for ¹⁵NH₄ ⁺ analysis.

Using the OX/MIMS method (¹⁵NH₄ ⁺ oxidation and MIMS analysis technique) described by Yin et al. (2014), DNRA activity (potential rates from slurry incubation both in 2019 and 2020 or actual rates from core incubation) was determined. The ¹⁵NH₄ ⁺ samples were purged with helium gas to eliminate any ²⁹N₂ and ³⁰N₂ produced via denitrification and/or anammox in the intact core/slurry incubations and were transferred to 12 mL gas-tight glass vials. Subsequently, a hypobromite iodine solution (0.2 mL) was added to oxidize the ¹⁵N gas (²⁹N₂ and ³⁰N₂), which was analyzed using MIMS to calculate the DNRA rate. In 2021, DNRA activity was detected through KCl extractions and alkaline acid trap diffusion, followed by analysis using an Elemental Analyzer Isotope Ratio Mass Spectrometer (EA-IRMS) system at the University of Connecticut Marine Science (Holmes et al., 1999).

Potential nitrification rates were measured in oxic slurries enriched with ¹⁵NH₄ ⁺. Briefly, 3 g of sieved sediment and 40 mL in situ bottom-water were transferred into a 50-mL centrifugal tube and amended with ¹⁵NH₄ ⁺ to a final concentration of 100 μmol L^-1 15N in each tube. The slurries were collected and filtered immediately for initial ¹⁵NO₃ ^- analysis. Triplicate vials were incubated for 6 h on a shaker table at the in situ temperature and collected for the final ¹⁵NO₃ ^- analyses. ¹⁵NO₃ ^- samples were analyzed at the University of Connecticut Marine Science using an IRMS system (Casciotti et al., 2002).

In the anoxic slurry incubations, ¹⁵NH₄ ⁺-amended and ¹⁵NH₄ ⁺ + ¹⁴NO₃ ^- treatments were used to test whether all ambient ¹⁴NO₃ ^- was consumed during pre-incubation and to assess the presence of anammox activity in the sediment (based on ²⁹N₂ production). The ¹⁵NO₃ ^- treatment was used to determine potential denitrification and anammox rates, denitrification contribution to total N₂ production, and DNRA contribution to N reduction according to the equations reported by Thamdrup and Dalsgaard (2002).

Both denitrification and anammox contributed to total ²⁹N₂ production based on the ¹⁵NO₃ ^- tracer experiment. Relative denitrification and anammox contributions to total ²⁹N₂ production were quantified using the following Equation 1:

where P ₂₉, A ₂₉ and D ₂₉ (nmol N g^-1 d^-1) denote the total ²⁹N₂ production rate during slurry incubation, anammox contribution to the ²⁹N₂ production rate, and denitrification, respectively, to the ²⁹N₂ production rate. Denitrifiers can use ¹⁴NO₃ ^- and ¹⁵NO₃ ^- randomly to produce N₂ gas; denitrification occurs after ¹⁴N and ¹⁵N random pairing, according to. Therefore, D ₂₉ was estimated using the following Equation 2:

where P ₃₀ represents the total ³⁰N₂ production rate during slurry incubation and F _N represents the ¹⁵N mole fraction in ¹⁴NO₃ ^-, which can be obtained by measuring the ¹⁴NO₃ ^- concentrations before and after ¹⁴NO₃ ^- tracer addition.

Finally, the potential anammox and denitrification rates were calculated using the following Equations 3, 4:

where D _t and A ₂₉ (nmol N g^-1 day^-1) represent the potential denitrification and anammox rates, respectively. The potential DNRA (nmol N g^-1 day^-1) was calculated from the increase in ¹⁵NH₄ ⁺, which was converted from ¹⁵N gas (²⁹N₂ and ³⁰N₂). Potential nitrification (nmol N g^-1 day^-1) was calculated from the increase in ¹⁵NO₃ ^- during oxic slurry incubation.

In the intact-core incubations, the measured O₂, ²⁹N_2, and ²⁹N₂ concentration linear regression slopes against time were used to calculate the rate. The units for sediment oxygen uptake (µmol O₂ l^-1 h^-1) and benthic N reduction (denitrification, anammox, and DNRA, µmol N₂ l^-1 h^-1) in the core incubation were converted to mmol O₂ m^-2 day^-1 and mmol N m^-2 day^-1, respectively, by dividing them by the intact incubation sediment surface area.

Anammox activity was confirmed using a ¹⁵NH₄ ⁺ tracer, which supported the anoxic slurry incubation hypothesis (Thamdrup and Dalsgaard, 2002). Therefore, N₂ production was derived from both denitrification and anammox in intact core incubations, from which the actual denitrification and anammox activity was calculated using the revised isotope pairing technique (Risgaard-Petersen et al., 2003; Huang and An, 2022). The actual total N₂ production (P ₁₄) and anammox and denitrification rates (aao ₁₄ and den ₁₄) in the intact core were calculated according to Huang and An (2022).

The frozen sediment samples (0.5 g) were used for DNA extraction using the DNeasy PowerLyzer PowerSoil kit (Qiagen) as described in the manufacturer’s protocol at the Virginia Institute of Marine Science. The extracted genomic DNA was sequenced by Novogene Corporation (CA, USA) using an Illumina NovaSeq 600 platform (2 x 150 bp). The raw metagenome sequences were uploaded to the KBase pipeline. Functional gene profiles of each metagenome were examined using the FAMA program, particularly N cycling genes in nitrification (amoA), denitrification (nirS and nirK), anammox (hzo), and DNRA (nrfA). The gene abundances were aggregated for each process and normalized using the reference gene effective fragment count per kg (efpkg). The gene abundance variability of bottom-water microbial communities was measured in 2020.

One-way analysis of variance (ANOVA) was used to determine whether seawater intrusion had a significant impact on various hydrological parameters, nutrient concentrations, SOD, and N transformation rates (denitrification, anammox, and DNRA) at different sites and cruises. Homogeneity of variance and normal distribution tests were conducted prior to ANOVA. If any data did not meet the assumptions of ANOVA, non-parametric tests (such as the Kruskal–Wallis test) were utilized. Statistical significance at P<0.05 was considered significant. Pearson’s correlation analyses were used to examine the relationship between N transformation rates and environmental parameters. All statistical analyses were performed using Statistical Package for the Social Sciences (SPSS) version 25.

Overall, bottom-water temperatures and DO ranged from 18.2–26.3°C and 1.6–382.5 μmoL L^-1, respectively ( Table 1 ). Significant inter-annual differences (ANOVA, P = 0.042) and spatial differences (ANOVA, P = 0.04) were observed in the bottom-water DO. DO was high at S11 (346.6 μ moL L^-1) and low at S6.5 (172.9 μmoL L^-1). Salinity ranged from 0.1 to 9.2 PSU, with no significant spatial or inter-annual differences observed, despite a recorded salinity of 2.3 PSU at S6.5 and 0.2 PSU at S11. However, significant seawater intrusion impact was observed, evidenced by salinity measuring 0.2 PSU before the dam opening, 3.7 PSU during seawater intrusion, and returning to 0.2 PSU after the dam opening (Kruskal–Wallis test, P = 0.043). Bottom-water Chl-a values ranged from 0.8–21.5 µg L^-1, and spatial variations were found between S6.5 (7.2 ± 3.7 µg L^-1) and S11 (16.8 ± 6.6 µg L^-1) (ANOVA, P=0.013).

Bottom-water NOx, NH₄ ⁺, and PO₄ ^3- concentrations ranged from 17–166.6, 0.9–30.3, and 0.1–2.9 μmoL L^-1, respectively. Significant inter-annual variation in bottom-water PO₄ ^3- levels (ANOVA, P< 0.01) was observed across 2019 (2.7 ± 0.3 μmoL L^-1), 2020 (0.9 ± 0.6 μmoL L^-1), and 2021 (0.3 ± 0.3 μmoL L^-1), but not in bottom-water NOx, and NH₄ ⁺. Bottom-water NOx was low for station S6.5 (88.3 ± 48.7 μmoL L^-1), and high for station S11 (117.8 ± 14.9 μmoL L^-1), bottom-water NH₄ ⁺, and PO₄ ^3- was high for station S6.5 (12.4 ± 10, and 1.3 ± 1.1 μmoL L^-1), and low for station S11 (1.3± 0.5, and 0.1 ± 0.1 μmoL L^-1), but only weak spatial variations were observed in bottom-water NH₄ ⁺ levels (Kruskal–Wallis test, P = 0.053). After seawater intrusion, bottom-water NOx decreased from 108.4 to 68.3 μmoL L^-1, while bottom-water NH₄ ⁺ increased from 5.9 to 15.9 μmoL L^-1 for station S6.5, although the seawater intrusion impact was not significant (ANOVA, P>0.05).

The surface sediment (0–10 cm) water content ranged from 21–41% (data not shown). The organic matter content was the highest in 2020 (3.1%), followed by 2019 (1.2%), and 2021 (0.9%) (Kruskal–Wallis test, P = 0.056). A weak spatial variation was observed in sediment organic matter (Kruskal–Wallis test, P = 0.053). The S6.5 site was rich in organic matter (1.9%), whereas the S11 site had low organic matter content (0.8%). After seawater intrusion in 2019 and 2020, organic matter content increased from 0.8% to 1.9% and from 1.1% to 3.4%, respectively, compared to levels before the dam opening ( Table 1 ). The sediment was dominated by sand (grain size > 63 μmoL L^-1), which accounted for 80% of the samples.

After seawater intrusion, notable changes were observed in the bottom-water of the S6.5 station regarding salinity and DO, whereas changes were not evident in the surface and middle waters ( Figure 2 ). Before the dam opening, bottom-water salinity remained below 0.4 PSU. However, during seawater intrusion in 2019, 2020, and 2021, salinity substantially increased, reaching levels as high as 4.5, 9.2, and 1.4 PSU, respectively ( Table 2 and Figure 2 ).

Bottom-water oxygen levels displayed normoxic conditions before the dam opening in 2019 (265.3 μmoL L^-1), 2020 (218.1 μmoL L^-1), and 2021 (264.4 μmoL L^-1). However, during seawater intrusion, a significant decrease was observed, with oxygen levels dropping to 206.6 μmoL L^-1 in 2019, 6.9 and 1.6 μmoL L^-1 in 2020, and ranging between 252.2 and 137.2 μmoL L^-1 in 2021, all during SIW 1 or SIW 1-1 ( Table 2 and Figure 2 ). After seawater intrusion, bottom-water oxygen conditions varied among normoxia, hypoxia, and weak hypoxia in 2019, 2020, and 2021 ( Table 2 and Figure 2 ).

In the intact core incubation conducted in 2019, a slight increase in SOD was observed consistently before, during, and after the dam opening, with average values of 4.4 ± 0.4, 4.6 ± 1.0, and 6.9 ± 0.1 mmol O₂ m^-2 day^-1, respectively ( Figure 3A ). Initially, the actual N₂ production rates (P ₁₄) sharply decreased from 0.3 ± 0.1 mmol N m^-2 day^-1 before the dam opening to 0.02 ± 0.02 mmol N m^-2 day^-1 during seawater intrusion; however, they subsequently rebounded to 0.35 ± 0.19 mmol N m^-2 day^-1 after the dam opening ( Figure 3B ). Denitrification predominantly contributed to the actual nitrogen production rates, regardless of seawater intrusion. Both actual denitrification (den ₁₄) and anammox (aao ₁₄) rates displayed similar trends before, during, and after seawater intrusion, with average values of 0.22 ± 0.09, 0.01 ± 0.01, and 0.28 ± 0.15 mmol N m^-2 day^-1, respectively, for den ₁₄, and 0.07 ± 0.03, 0.01 ± 0.01, and 0.07 ± 0.04 mmol N m^-2 day^-1, respectively, for aao ₁₄ ( Figure 3B ). DNRA was not detected in freshwater at the S6.5 station before the dam opening. However, it significantly increased to 0.08 ± 0.00 mmol N m^-2 day^-1 during seawater intrusion and became negligible again after the dam opening ( Figure 3B ). The nitrate reduction process was dominated by denitrification both before and after the dam opening. Notably, DNRA emerged as the dominant nitrate reduction process during seawater intrusion ( Figure 3B ).

After seawater intrusion, potential nitrification in the sediment decreased, ranging from 46.8 ± 20.3 nmol N g^-1 day^-1 before the dam opening to 15.5 ± 1.4 nmol N g^-1 day^-1 during seawater intrusion. Subsequently, the potential nitrification did not recover from seawater intrusion after the dam opening, remaining at 8.3 ± 4.2 nmol g^-1 day^-1 ( Figure 3C ).

The potential denitrification rate in sediment varied considerably, ranging from 10.6 ± 0.8 nmol N g^-1 day^-1 at S6.5 in 2019 after the dam opening to 925.2 ± 80.7 nmol N g^-1 day^-1 at S11 in 2021 before the dam opening ( Figure 4A ). Significantly different denitrification rates were observed across the years 2019–2021 (ANOVA, p = 0.027). Specifically, the denitrification rates showed a pattern with the highest rate in 2021 (501 ± 277.5 nmol N g^-1 day^-1), followed by a lower rate in 2020 (220.3 ± 69.6 nmol N g^-1 day^-1), and the lowest rate recorded in 2019 (57.2 ± 54.3 nmol N g^-1 day^-1). Weak spatial variations were also noted, ranging from 233 ± 181.6 nmol N g^-1 day^-1 at S6.5 to 613.9 ± 440.3 nmol N g^-1 day^-1 at S11 (ANOVA, p = 0.051). No significant impact of seawater intrusion on potential denitrification was observed, although the rates decreased marginally from 387.6 ± 365.9 nmol N g^-1 day^-1 before the dam opening to 303.8 ± 200.3 nmol N g^-1 day^-1 after seawater intrusion. Overall, denitrification accounted for 30–100% of total NO₃ ^- reduction, representing 76% at S6.5, 100% at S11, and 82%, 78%, and 81% before, during, and after the dam opening, respectively ( Figure 5 ).

In 2019 and 2020, denitrification decreased when bottom-water oxygen condition was normoxia and hypoxia, while it increased in 2021 when bottom-water oxygen was weak hypoxia ( Figure 4A ). The potential denitrification rates were 116.8 ± 5.7, 44.2 ± 0.4, and 10.6 ± 0.8 nmol N g^-1 day^-1 in 2019, and 297 ± 15.8, 150.1 ± 34.5, and 174 ± 16.9 nmol N g^-1 day^-1 in 2020, before, during, and after the dam opening, respectively ( Figure 4A ). The potential denitrification rates consistently responded to seawater intrusion in 2019 and 2020, during which time seawater intrusion significantly suppressed denitrification. The potential denitrification rate was enhanced after seawater intrusion in 2021 (211.2 ± 41, 569.8 ± 93.0, and 496 ± 81 nmol N g^-1 day^-1 before the dam opening, during SIW 1, and SIW 1-1, respectively; Figure 4A ). Seawater intrusion reduced denitrification rates and its contribution to the total NO₃ ^- reduction process. Denitrification dominated the NO₃ ^- reduction process from 2019–2021, regardless of seawater intrusion, except in 2020 after seawater intrusion ( Figure 5 ).

The potential for DNRA in the sediment was nearly undetectable in 2021. In 2020 peak rates occurred at 310.3 ± 82.1 nmol N g^-1 day^-1 after seawater intrusion, with an overall mean of 43.1 ± 96.3 nmol N g^-1 day^-1 (all measurements conducted at S6.5) ( Figure 4B ). Overall, after seawater intrusion, potential DNRA increased from 22 ± 37.4 nmol N g^-1 day^-1 to 69.4 ± 135.6 nmol g^-1 day^-1, subsequently declining to 9.1 ± 11.7 nmol N g^-1 day^-1 after the dam opening. Our overall study findings indicated that neither the spatial variation nor the seawater intrusion effect on DNRA was statistically significant. However, a significant inter-annual variation was observed in the potential DNRA activity from 2019–2021 (Kruskal–Wallis test, p = 0.018). Potential DNRA was the highest in 2020 (107 ± 137 nmol N g^-1 day^-1), low in 2019 (0.9 ± 0.2 nmol N g^-1 day^-1), and 2021 (0.1 ± 0.0 nmol N g^-1 day^-1). DNRA accounted for 0–63% of total NO₃ ^- reduction, accounting for 6%, 15%, and 7% before, during, and after seawater intrusion, respectively ( Figure 5 ).

Potential DNRA rates increased in 2019 and 2020 under bottom-water normoxia and hypoxia in response to seawater intrusion ( Figure 4B ). However, DNRA exhibited a more pronounced increase in 2020 (65.2 ± 13.8 nmol N g^-1 day^-1 before the dam opening and 310.3 ± 82.1 nmol N g^-1 day^-1 after seawater intrusion) compared to the change from 0.7 ± 0.1 nmol N g^-1 day^-1 before the dam opening to 1.1 ± 0.5 nmol N g ^-1 day^-1 after seawater intrusion in 2019 ( Figure 4B ). In 2021, DNRA was almost undetected (maintaining approximately 0.1 nmol N g^-1 day^-1 before, during, and after the dam opening), and little change was observed ( Figure 4B ). Seawater intrusion significantly enhanced the DNRA rates in 2020 and improved the contribution of DNRA to the total NO₃ ^- reduction process. DNRA did not dominate the NO₃ ^- reduction process, except in 2020, after seawater intrusion (63% of total NO₃ ^- reduction, Figure 5 ).

Potential anammox activity in the sediment samples was low relative to denitrification and DNRA, it ranged from 0 in 2021 both at S6.5 and S11 to 75.7 ± 7.3 nmol N g^-1 day^-1 at S6.5 in 2021 ( Figure 4C ). Station S6.5 exhibited higher potential anammox (27.7 ± 25.7 nmol N g^-1 day^-1) compared to station S11 (0.8 ± 1.2 nmol N g^-1 day^-1). The highest potential anammox was recorded in 2021 (28.8 ± 38.8 nmol N g^-1 day^-1), 2019 (20.0 ± 16.1 nmol N g^-1 day^-1), and 2020 (18.7 ± 10.9 nmol N g^-1 day^-1). However, no significant spatial or inter-annual variation was observed in potential anammox rates (Kruskal–Wallis test, p > 0.05). Overall, potential anammox activity decreased from 32.9 ± 31.6 nmol N g^-1 day^-1 before the dam opening to 22.5 ± 25.3 nmol N g^-1 day^-1 after seawater intrusion, and it continued to decrease after the dam opening (6.3 ± 5.2 nmol N g^-1 day^-1); however, no significant seawater impact was observed. Generally, after seawater intrusion, potential anammox decreased in both 2019 bottom-water normoxia and 2021 bottom-water weak hypoxia, however, it increased in 2020 under severe bottom-water hypoxia ( Figure 4C ). Anammox accounted for 13% at S6.5, 0.1% at S11, and 14%, 9%, and 12% before, during, and after seawater intrusion, respectively, of total NO₃ ^- reduction ( Figure 5 ). Overall, anammox did not contribute significantly to total NO₃ ^- reduction, whereas seawater intrusion reduced the contribution of anammox to NO₃ ^- reduction ( Figure 5 ).

Overall, the nitrification gene abundance ranged from 0.04–0.18 effective fragment count per kg (efpkg). An obvious inter-annual variation was observed in nitrification gene abundance (ANOVA, p = 0.005), with an average value of 0.16 ± 0.3, 0.12 ± 0.02, and 0.06 ± 0.02 efpkg in 2021, 2020, and 2019, respectively ( Figure 6A ). The nitrification gene (amoA) abundance increased numerically in 2019 under bottom-water normoxia, and little changed in 2020 under severe bottom-water hypoxia after seawater intrusion. The nitrification differed among before, during, and after the dam opening, but this change in the gene abundance was not statistically significant.

The nitrogen reduction process was dominated by denitrification, as confirmed by the abundance of nitrogen cycling genes. The total abundance of denitrification (nirS and nirK), DNRA (nrfA), and anammox (hzo) gene ranged from 1.8–3.3, 0.9–2.1, 0–0.01 efpkg, respectively ( Figures 6B-D ). Significant inter-annual variations were found in DNRA gene abundance (ANOVA, p = 0.004), but not in denitrification nor anammox gene abundance (ANOVA, p > 0.1). The total gene abundance for denitrification was highest in 2019 (2.8 ± 0.4 efpkg), high in 2020 (2.1 ± 0.3 efpkg), and low in 2021 (1.9 ± 0.2 efpkg). The denitrification gene abundance both decreased in 2019 and 2020 and increased in 2021. Overall, this change in denitrification gene abundance is similar to the change in potential denitrification rates in the sediment slurry incubation. The DNRA gene abundance was highest in 2020 with an average value of 2.0 ± 0.2 efpkg similar to the highest potential DNRA rate in 2020 observed in sediment slurry incubation ( Figures 4B , 6D ). The total gene abundance for DNRA both increased in 2019 and 2020 after seawater intrusion, similar to the potential DNRA response found in sediment slurry incubation. The DNRA gene abundance did not differ much numerically in 2021 after seawater intrusion, at the same time, little change was observed in potential DNRA. The total anammox gene abundance was almost negligible when compared to denitrification and DNRA gene abundance. Generally, it increased both in 2019 and 2021 after seawater intrusion, but decreased in 2020, which differed from the potential anammox response observed in sediment slurry incubation.

Pearson’s correlation analysis revealed that the bottom-water DO was negatively correlated with bottom-water salinity ( Table 3 ). A positive relationship was observed between bottom-water DO and Chl-a levels. Bottom-water NO_X was positively correlated with bottom-water DO content, but negatively correlated with bottom-water salinity. Bottom-water NH₄ ⁺ concentrations were positively correlated to bottom-water salinity; however, they were negatively correlated to bottom-water DO, and bottom-water NO_X. No significant correlations were observed between bottom-water salinity and potential denitrification, anammox, or DNRA rates. However, potential sediment denitrification rates were negatively correlated with bottom-water PO₄ ^3- concentrations and positively correlated with bottom-water Chl-a levels. Potential anammox activity in the sediment was not correlated with any bottom-water environmental parameters. Potential DNRA in the sediment was negatively correlated with bottom-water DO concentrations. Nitrification gene abundance in the sediment was negatively correlated with bottom-water PO₄ ^3- concentrations, and positively related to denitrification. Denitrification gene abundance in the sediment was negatively correlated with nitrification gene abundance, and positively correlated with bottom-water PO₄ ^3- concentrations. In contrast, DNRA gene abundance in the sediment was negatively correlated with bottom-water DO, and denitrification gene abundance. The abundance of the Anammox gene did not significantly correlate with any bottom-water environmental parameters or with the abundance of other nitrogen cycling genes in the sediment.

After the dam opening and seawater intrusion, there was an immediate and rapid increase in bottom-water salinity ( Figure 2 ). At the S6.5 station, located upstream from the dam and deeper than S11, seawater accumulated and became trapped due to density disparities between seawater and freshwater (An, 2021). Consequently, the trapped bottom seawater restricted the exchange with the overlying freshwater, resulting in a decline in dissolved oxygen (DO) levels. Pearson correlation analyses conducted ( Table 3 ) revealed a significantly negative correlation between bottom-water salinity and oxygen concentration (P=0.003). This correlation indicates that as salinity increased post-seawater intrusion, the DO concentration gradually decreased. For instance, in 2020, bottom-water DO levels dropped to 6.9 μmoL L^-1 and 1.6 μmoL L^-1 during SIW 1 and SIW 1-1, respectively, indicating substantial hypoxia in contrast to the levels before the dam opening (218.1 μmoL L^-1). Jia et al. (2021) extensively discussed the effects of seawater intrusion on water density stratification in the Modaomen Estuary, China, highlighting frequent occurrences of bottom-water hypoxia linked to substantial water-column stratification due to the interplay of high freshwater discharge and saltwater intrusion. Additionally, Poirrier (2013) reported instances of bottom-water hypoxia and anoxia in Lake Pontchartrain, attributed to salinity stratification resulting from saltwater intrusion via the Mississippi River Gulf Outlet through the Inner Harbor Navigation Canal.

In addition to water stratification, an increase in sediment oxygen demand (SOD) for organic matter decomposition may also contribute to the oxygen decline following seawater intrusion. This was confirmed by our intact core incubation result conducted in 2019, SOD increased from 4.4 ± 0.4 before the dam opening to 4.6 ± 1.0 mmol O₂ m^-2 day^-1 after seawater intrusion ( Figure 3A ). The increased SOD did not result in bottom-water hypoxia in 2019, we hypothesize this is because seawater trapping was negligible as the dam only keep open for 1 hour ( Table 2 ). The degree of seawater trapping was dependent on the duration of the dam opening and subsequent seawater intrusion (An, 2021). After these intrusion events, bottom-water oxygen levels showed different states, ranging from normoxic to anoxic or weakly anoxic. Seawater trapping was severe in 2020, weak in 2021, and negligible in 2019. Consequently, bottom-water oxygen levels were anoxic or hypoxic in 2020, weakly hypoxic in 2021, and normal in 2019 ( Table 2 ). These varying bottom-water oxygen conditions might impact benthic nitrogen cycling.