Three main estuaries in Fujian Province were selected as the study area: From north to south, the Min River, Mulan River, and Jiulong River. The three estuaries all have a subtropical monsoon climate, with an average annual rainfall of more than 1300 mm and an average annual temperature of 19.6 to 21.0°C (Zhang et al., 2011; Tong et al., 2012; Luo et al., 2019). In August 2017, eight sampling points were selected for shrimp culture ponds near the three estuaries, and surface sediments were collected underwater using a Plexiglas tube ( Figure 1 ). The shrimp ponds at the sampling points were all reclaimed from estuary swamp wetlands, with similar land-use transformation years (7–9 years), and all were muddy pond slopes and pond bottoms. The shrimp ponds contained white shrimp (Litopenaeus vannamei). Shrimp seedlings are put in at the end of May every year, and farming ends after all the shrimps are harvested in mid- to late October. Shrimp farming ponds use local river water as a source of farming water. Each shrimp farming pond is equipped with two impeller aerators with a power of 1.5 kW, and the startup time is about 20:00 to 2:00 the next morning. The switch of the aerator during the day depends on the weather, the growth of fish and shrimp, and the water quality. The feed was provided twice a day at 7:00 and 17:00, and the feeding amount was in accordance with the conventional requirements. The basic features of shrimp ponds are listed in Table 1 .

Sediment density was measured using the syringe method (Percival and Lindsay, 1997), and the moisture content was dried using the weight loss method at 80°C to a constant weight. After removing sedimentary carbonate using 0.1 M HCl, the concentraction of total organic carbon (TOC) and total nitrogen (TN) in the sediments were determined using a carbon-hydrogen-nitrogen elementary analyser (VVarioELIII, Elementary, Germany) and C N elemental analyser (Elementar Vario MAX CN, Germany), respectively (Lin et al., 2017; Hu et al., 2022). The sediment concentrations of NH₄ ⁺ and NO₃ ^- were extracted with 2 M KCl solution and then determined using a continuous flow analyser (SAN Plus, Skalar Analytical B.V., The Netherlands) (Sun et al., 2022).

The potential DNF, ANA, and DNRA rates were measured using the nitrogen isotope tracing method (Yin et al., 2014; Deng et al., 2015; Gao et al., 2017). In brief, the slurry was prepared by mixing sediment into helium-purged water with a sediment/water volume ratio of 1:7 and then transferring to a helium-purged 12-mL vial (Labco Exetainers) (Hou et al., 2015). The vials were then pre-incubated for 36 h to remove surplus NO₃ ^-, NO₂ ^-, and O₂ at in situ temperatures. After pre-incubation, sterile anoxic solutions of ¹⁵NO₃ ^- (¹⁵N at 99%) was added to all vials via the septa, with the final content of ¹⁵N being approximately 100 μM (Hou et al., 2013). Then, 200 μL of ZnCl₂ solution (50%) was added to half of the replicates (as initial samples). Then, half of the slurries were incubated for 8 h, and 200 μL of ZnCl₂ solution (50%) was added at the end of incubation to terminate the reaction (Lin et al., 2017). Both of the contents of ²⁹N₂ and ³⁰N₂ were measured by membrane inlet mass spectrometry (MIMS) during incubation, to calculate the DNF and ANA rates based on the difference in ²⁹N₂ and ³⁰N₂ produced among the final and initial results (Gao et al., 2019).

The DNF and ANA rates were approximated based on the accumulations of ²⁹N₂ and ³⁰N₂, respectively (Deng et al., 2015). The contributions of DNF and ANA to ²⁹N₂ production were calculated using Equation (1).

Here, P₂₉ , D₂₉ , and A₂₉ (nmol N g⁻¹ h⁻¹) denote the total ²⁹N₂ production rate and the production rate of ²⁹N₂ from DNF and ANA during the slurry experiments, respectively. The ratio of ¹⁴N and ¹⁵N produced by ¹⁴NO₃ ^- or ¹⁵NO₃ ^- with random isotope pairing and D₂₉ was calculated using equation (2) (Risgaard-Petersen et al., 2003).

Here, P₃₀ (nmol N g⁻¹ h⁻¹) represents the production rate of total ³⁰N₂, and F _N (%) denotes the proportion of ¹⁵N in NO₃ ^-, which was estimated from the measured content of added ¹⁵NO₃ ^- and surplus NO₃ ^- (Shan et al., 2016). Finally, the DNF potential rates were calculated using Equation (3), and the ANA was calculated using Equation (4).

Here, D_total and A₂₉ (nmol N g⁻¹ h⁻¹) denote DNF and ANA rates, respectively.

The DNRA rate was calculated using the ¹⁵NH₄ ⁺ oxidation and MIMS analysis (OX/MIMS) method (Yin et al., 2014). First, the sediment slurry was pre-incubated. After the preincubation, 100 μL of ¹⁵NO₃ (¹⁵N at 99.6% and a final content of approximately 100 μM ¹⁵N) was added to all the vials. Immediately, half of the slurry was saved (as initial sample) with 200 μL of ZnCl₂ solution (50%). The rest of the vials (as final samples) were further incubated 8 h before adding 200 μL of ZnCl₂ solution (50%). The potential rates of DNRA were calculated using Equation (5).

Here, R_DNRA (nmol N g^-1 h^-1) is the total potential rate of DNRA; (¹⁵NH₄ ⁺)_Final and (¹⁵NH₄ ⁺)_Initial (nmol N L^-1) are the content of ¹⁵NH₄ ⁺ in the final and initial sample, respectively; and V (L), W (g), and T (h) denote the volume of the vial, dry weight of the sediment, and time, respectively (Gao et al., 2019).

The differences between the NO₃ ^- reduction rates among all points were analysed by one-way analysis of variance (ANOVA) (homogeneity of variance was tested by the LSD test, and Dunnett’s T3 was used to test for heterogeneity of variance). The relationships between the environmental variables and DNF, ANA, and DNRA rates were revealed by Pearson correlation analyses. SPSS 22.0 was used for one-way ANOVA and pearson analysis. While Redundancy analysis (RDA) was used to evaluate variations in NO₃ ^- reduction rates with respect to environmental variables using software Canoco 4.5 (Sun et al., 2020). The significance level was set at 0.05.

The physicochemical characteristics of each site are presented in Table 2 . Sediment bulk density varied from 1.27 to 1.78 g·cm^-3 in the study area, and the water content in the sediment varied from 31 to 59%. The salinities of the shrimp ponds ranged from 0.06 to 1.93 ppt. The TOC, TN, and TOC/TN ratios in the sediment varied between 0.58 to 1.97%, 0.07 to 0.16%, and 4.51 to 13.75, respectively. The concentration of NH₄ ⁺ and NO₃ ^- in sediments varied from 2.42 to 8.41 and 0.02 to 0.14 μmol·g^-1, respectively. In addition to the physicochemical characteristics of the sediments, the area and average depth of each shrimp pond were also different, and the largest area (1.113–1.255 ha) and average depth (2.1–2.8 m) of shrimp ponds were found in the Mulan estuary. The unit water output varied between 9.77–10.6 kg ha⁻¹, and the food coefficient ranged from 1.38–1.86 in the shrimp pond ( Table 1 ).

The sediment potential rates of DNF ranged from 2.87 to 18.11 nmol N g^-1 h^-1, and there were significant differences between the all sites (n = 24, p < 0.05). The highest DNF rate was found at site MA, while the lowest DNF rate was at site PH ( Figure 2 ). DNF contributed 64.2–91.5% of the total NO₃ ^- reduction rate ( Figure 3 ). The rates of ANA varied from 0.10 to 1.92 nmol N g^-1 h^-1 in the study area, and a distinct spatial difference in the ANA rates was observed among sites (n = 24, p < 0.05). The highest ANA rate occurred at site PH, whereas the lowest ANA rate occurred at site JF. Compared to DNF, ANA had less effect on NO₃ ^- reduction and contributed 1.2–31.4% to total nitrogen loss. Potential DNRA rates varied from 0.21 to 1.25 nmol N g^-1 h^-1. Significant differences were observed among the sites in the study area (n = 24, p < 0.05). DNRA and DNF showed the same distribution, with the highest value appearing at the MA site and the lowest value appearing at the PH site. The contribution of DNRA accounted for 4.38–12.35% of the total NO₃ ^- reduction.

The RDA was implemented to evaluate variations in the NO₃ ^- reduction rate with respect to environmental variables. The first two RDA dimensions were found to account for 70.56% of the cumulative variance, and the first and second axes contributed 55.12 and 15.44%, respectively ( Figure 4 ). The RDA results showed that salinity and TOC were significantly associated with various NO₃ ^- reduction process. The DNF rates were significantly and negatively correlated with bulk density (r = - 0.559, p < 0.05, n = 24) and positively correlated with water content (r = 0.464, p < 0.05, n =24), TOC (r = 0.590, p < 0.01, n = 24), TOC/TN (r = 0.495, p < 0.05, n = 24), and NO₃ ^- (r = 0.525, p < 0.01, n = 24) ( Table 3 ). Of the detected environmental factors, salinity and the ANA rate showed a significant negative correlation (r = - 0.656, p < 0.01, n = 24), and there was a positive correlation with TN (r = 0.498, p < 0.05, n =24), TOC (r =0.506, p < 0.05, n = 24) and NH₄ ⁺ (r = 0.577, p < 0.05, n = 24). The sediment potential rate of DNRA was negatively associated with bulk density (r = 0.441, p < 0.05, n = 24) and salinity (r = - 0.637, p < 0.01, n = 24) and positively related to water content (r = 0.472, p < 0.05, n = 24), TOC (r = 768, p < 0.01, n = 24), TN (r = 0.546, p < 0.01, n = 24), TOC/TN (r = 0.608, p < 0.01, n = 24), NH₄ ⁺ (r = 0.604, p < 0.05, n = 24), and NO₃ ^- (r = 0.508, p < 0.05, n = 24) ( Table 3 ). There are positive correlations of most of the dissimilatory NO₃ ^- reduction processes with TOC, TOC/TN, NH₄ ⁺, and NO₃ ^- were found ( Table 3 ).

DNF rate were measured by ¹⁵N tracer techniques according to the assumption of N₂, because of the ratio of N₂O to N₂ from DNF in aquatic ecosystems is very low (Dong et al., 2002). Salinity affected the DNF rate to some extent, although no statistically significant correlation was observed, and a decrease in denitrification activity associated with higher salinity was observed. RDA showed a negative correlation between salinity and DNF rate. Salinity may result in physiological stress on DNF, which in turn affects the DNF rate, which generally decreases with an increase in salinity (Rysgaard et al., 1999; Seo et al., 2008). In general, the DNF rate was positively related to the TOC content in previous studies, such as in lakes, rice fields, estuaries, and coastal environments (Vymazal, 2007; Deng et al., 2015; Yang et al., 2022). Our study showed that the DNF rate showed a positive relationship with the TOC content in the shrimp ponds, which is similar to previous research (Gao et al., 2019). Because shrimp ponds have been in a high organic carbon environment for a long time, they are beneficial for the growth of denitrifying-associated bacteria and promote the DNF rate (Canion et al., 2014; Plummer et al., 2015; Yang et al., 2017b). A remarkable relationship between the DNF rate and NO₃ ^- was observed in our study ( Table 3 ), and it was confirmed that NO₃ ^- is a major driver of the rate of NO₃ ^- reduction in shrimp ponds. DNF microorganisms use NO₃ ^- as an electron acceptor and substrate, which is strongly dependent on the NO₃ ^- concentration (Giles et al., 2012; Shan et al., 2016; Palacin-Lizarbe et al., 2020).

Spatial variations in ANA rates have also been found, and numerous studies have reported that salinity is a crucial environmental factor affecting ANA (Wang et al., 2012; Hou et al., 2015). Salinity was significantly correlated with the ANA rate in this study, suggesting that high salinity might inhibit the rate of ANA activity. It has been reported that ANA activity is not directly caused by an energy source, but is influenced by C and N substrates in sediments (Hou et al., 2015). A strong correlation was found between ANA and the C and N substrates in our study; this C and N matrix mainly comes from the decomposition of a large amount of feed and manure residues in the culture period (Wu et al., 2014). Studies have shown that TOC and NH₄ ⁺ concentrations at suitable concentrations can promote the ANA rate (Cheng et al., 2016; Damashek and Francis, 2018), as well as a significant positive correlation between TOC, NH₄ ⁺, and ANA. However, it should be noted that high TOC concentrations may inhibit ANA activity (Bettazzi et al., 2010).

DNRA is often considered to be closely related to salinity, and previous studies have shown that increasing salinity accelerates the DNRA rate (Gardner et al., 2006; Giblin et al., 2010). However, some studies have found that the abundance of the nrfA gene associated with DNRA rates is not significantly affected by salinity (Yin et al., 2017). Interestingly, our results showed that DNRA and ANA rates declined significantly with increasing salinity, as both showed similar results ( Table 3 ). RDA also showed a negative correlation between salinity and ANA, DNRA rates. TOC and NH₄ ⁺ could be important factors influencing DNRA in sediments (Bu et al., 2017), which was also proven by the significant positive correlation between the rate of DNRA and TOC in our study. High TOC concentrations may provide adequate organic substrates for the growth of DNRA-associated microorganisms (Deng et al., 2015). DNRA can more efficiently utilise NO₃ ^- as an electron acceptor in a carbon-rich environment, thereby increasing the DNRA rate (Kraft et al., 2014; Cheng et al., 2016).

The DNF rate was compared with the ANA and DNRA rates to identify potential connections among them ( Figure 5 ). These correlation analyses show that DNF is closely related to ANA and DNRA in the study area. During the interaction between ANA and DNF, DNF is a major source of NO₂ ^- for ANA in coastal wetland sediments (Meyer et al., 2005), and there was a significant relationship found between DNF and ANA rates in the shrimp pond (R² = 0.34, p < 0.05). Studies have also found a coupling related process between ANA and DNF (Hou et al., 2015). The ANA rates were significantly associated with DNRA rates (R² = 0.66, p < 0.01), indicating that DNRA can produce NO₂ ^- as an intermediate product, which is an alternative substrate for ANA (Lam et al., 2009). DNF and DNRA compete for NO₃ ^- under hypoxic or anaerobic conditions. Our results showed that the DNF rates were significantly related with the DNRA rates in the sediment (R² = 0.61, p < 0.05) ( Figure 5 ), indicating that the growth of NO₃ ^- substrate promoted both of them.

Competition among DNF, ANA, and DNRA can determine the fate of NO₃ ^- owing to their diverse roles in dissimilatory NO₃ ^- reduction. DNF was the dominant route contributing 64.2–91.6% (81.14%) of the total the dissimilatory NO₃ ^- reduction processes in the shrimp pond, while the contributions of ANA and DNRA were 1.2–31.4% (10.48%) and 4.34–12.4% (8.38%), respectively ( Figure 6 ). According to several studies, DNF plays a major role in removing nitrogen from various aquatic systems, while the proportions of ANA and DNRA differ between ecosystems (Deng et al., 2015; Shan et al., 2016; Yang et al., 2022). Different aquatic environments contribute different amounts of dissimilatory NO₃ ^- reduction, as shown in Table 4 . Studies have shown that the DNF, ANA, and DNRA rates in urban rivers and seas are higher than those in shrimp ponds (Song et al., 2013; Cheng et al., 2016). However, the rate of DNF in a paddy field was between 2.37 and 8.30 nmol N g^-1 h^-1, and ANA and DNRA were also relatively low (Shan et al., 2016). The DNF rate in estuarine and lake regions ranges from 0.09 to 11.47 nmol N g^-1 h^-1, which is lower than that in shrimp ponds (Deng et al., 2015; Yang et al., 2022). Compared with the estuary wetland, the dissimilation NO₃ ^- reduction rate increased after reclamation for shrimp ponds, showing an increasing trend with increasing reclamation years (Gao et al., 2019). It was found that the biggest difference between the shrimp ponds system and other systems may be more dependent on the difference caused by the high concentration of C and N matrix input. We found that the rate of DNF ranged from 3.24 to 16.64 nmol N g^-1 h^-1 in shrimp ponds, while the rate of ANA was slightly higher than that of DNRA ( Figure 2 ). Further analysis showed that DNRA was higher than ANA in aquaculture ponds in the Min and Jiulong River estuaries, indicating that the contribution of DNRA to the total NO₃ ^- reduction was still higher than that of natural wetlands (Gao et al., 2019). ANA was found to be much higher than DNRA in the Mulan River estuary, especially the PH site at 31.4% ( Figure 3 ). This may be related to the large size and depth of shrimp ponds in the estuary of the Mulan River, however, specific causes need to be identified.

If the average DNF and ANA rates were extrapolated to the study area, the N losses were approximately 32.43–131.64 g N m^-2 yr^-1 for DNF and 2.38–15.85 g N m^-2 yr^-1 for ANA. According to the area of aquaculture ponds in China’s coastal wetland (2.6 ×10⁶ ha) (Chen et al., 2016a; Chen et al., 2016b), approximately 2.18 × 10⁶ t N can be removed from shrimp ponds by this process annually, indicating that coastal reclamation is a considerable way to remove nitrogen. Recent results suggest that if natural wetlands are gradually converted into shrimp ponds, reactive nitrogen may be retained (Murphy et al., 2016). Shrimp ponds may be more important than nitrogen loss (Gao et al., 2019). Consequently, sediment N loads increase, resulting in decreased water quality and shrimp disease outbreaks (Castillo Soriano et al., 2013).