Permeable sediments were collected at the edge of the shoreline in a city park during ebbing tide in June 2016. Sediments were collected below the mean water level, and the surface layer (0−2 cm) samples were discarded due to the accumulation of algal debris. The gathered sediment was rapidly transferred to the laboratory (4 h after collection). The salinity of the overlying water was 34 at high tide and decreased to 31.5 at low tide in June, indicating a salinity variation caused by terrestrial water injection (river water or terrestrial groundwater) on the sampled sediments. Bay water temperature near the site was approximate 18°C. The grain size of the collected sediment was determined using a Beckman Coulter laser particle size analyzer. Average grain size was 0.32 mm and the proportion of clay and silt (<0.063 mm) in the sediment was low (ca. 5% in total sediment particles). Sediment porosity was approximately 0.32, determined according to the approach described in Ibánhez and Rocha (2016). Sediment organic carbon content by mass was 0.14%, quantified with the acid fumigation method of Harris et al. (2001). Total nitrogen in the sediment was 0.008% determined on a Vario EL^® Cube elemental analyzer. Inorganic P in the sediment was 0.01% determined by an acid extraction method (1 mol/L HCl solution for 24 h; Aspila et al., 1976).

The permeable sediment used in the FTR experiments was thoroughly mixed and placed into reactor cells (10 cm length, 6.6 cm diameter). Compared with the intact sediment columns, repacked sediments are homogeneous, decreasing the build-up of preferential flow during the experiments and eliminating differences in the test sediment among treatment groups (Santos et al., 2012). In addition, the FTR cells used two collimator caps and GF/F membrane filters on both sides of the sediment to remove the possibility of particle interference and facilitate the dispersion of the circulating solution in the cells (Fig. 1). After assemblage, the reactor cells were wrapped with aluminum foil to avoid light exposure and transferred to a temperature-controlled incubator (20°C) to eliminate the influence of temperature variations on observed N reactions (Rocha, 1998, 2000) and simulate the benthic N turnover in summer in Ireland (Rocha et al., 2015).

The circulating solutions used in the FTR experiments were made with mixtures of oligotrophic marine water (salinity, 36) and river water (salinity, 0), collected from the Coliemore Harbor (53.274 8°N, 6.093 9°W) and the Phoenix Park in Dublin (53.355 8°N, 6.318 7°W), respectively. Before the experiments, the mixed solutions were aerated for 2 h to saturate dissolved oxygen (DO). To explore the effects of salinity and input ${\rm {NO}}_3^- $ concentration on consumption/production rates in the sediments, three treatments were employed in the experiments. The first treatment corresponds to the mixture of natural river water and marine water, and this was identified as the control. The ${\rm {NO}}_3^- $ concentrations in both river water and marine water were low (approximately 12 μmol/L and 2.3 μmol/L). After mixing, the final salinities of the reaction solutions were 0, 10, 15, 20, 30 and 36 (6 runs in the treatment). In the second treatment, river water was spiked with ${\rm {NO}}_3^- $ to a final concentration of approximately 100 μmol/L, while the seawater remained unchanged. This intends to simulate the dispersion of ${\rm {NO}}_3^- $ enriched freshwater along a salinity gradient in coastal zones. For the third treatment, both river water and seawater were spiked with ${\rm {NO}}_3^- $ to a final concentration of 100 μmol/L, eliminating the effect of mixing of fresh and marine waters over ${\rm {NO}}_3^- $ concentration, and therefore testing for the effect of changing salinity alone. The mixing proportions between river water and seawater in the second and third treatments were identical to the first treatment. In total, three treatments with 18 runs were included in the present study. Each run was performed in triplicate. Prior to mixing, both “endmember” water samples were filtered through GF/F filters (Whatman^®; average pore size, 0.7 μm) to remove suspended particles. The flow rate of porewater in the reactor cells was 1.5 mL/min, similar to the advection rate in permeable beaches (Rocha et al., 2009; Ibánhez et al., 2011), which guaranteed the long reaction time for solutes with sediment particle surface (Jiang et al., 2018a). The input solution was pumped from the bottom to the top to simulate the 1-D advection transport in benthic environments (Fig. 1). Prior to each experimental run, the input solution was pumped through the sediment overnight (ca. 12 h) at the working porewater velocity for acclimatization. Subsequently, the run started and water samples were collected at both the input solution and the output stream (Fig. 1) every 60 min (similar to the porewater residence time in the reactor cells) for 3 h (3 collections in total). In parallel, DO content at both input and output solution was determined with an OxiCal^® probe under the salinity calibration mode. Water samples for the determination of dissolved organic carbon (DOC) were stored in amber glass vials at 4°C after acidification to pH<2 using 2 mol/L HCl. Samples for fluorescent dissolved organic matter (FDOM) were stored in glass vials at −20°C. Water samples for the concentration of dissolved inorganic nitrogen (DIN) species, including ${\rm {NO}}_3^- $, nitrite (${\rm {NO}}_2^-$), and ammonium (${\rm {NH}}_4^+ $), as well as soluble reactive phosphorus (SRP) were stored in vacutainers at 4°C with a preservation period of less than one week (Jiang et al., 2017b). Samples for δ¹⁵N-${\rm {NO}}_3^- $ were stored in vacutainers at −20°C. The initial concentrations of DO, DOC, ${\rm {NO}}_3^- $, ${\rm {NH}}_4^+ $, ${\rm {NO}}_2^-$ and SRP at each run are outlined in Supplementary Table S1.

DOC was determined on a Vario TOC Cube elemental analyzer after purging with carbon-free gas (CO₂ and VOC free air; 10 min per sample). The detection limit for DOC was approximately 14 μmol/L based on the determination blanks. The method precision was ca. 6%. FDOM was measured on a Cary Varian Eclipse^® spectrofluorometer via 3-D excitation-emission-matrix spectral scanning (Ibánhez and Rocha, 2014) and subsequently processed with the DOM Fluor Toolbox in MATLAB (Stedmon and Bro, 2008). Water from a Millipore purification gradient system was used as blank. The signal from the daily blanks was subtracted from the sample spectra to remove Raman scatter peaks (Stedmon and Bro, 2008). Fluorescence intensities for each sample were divided by the integrated area of Raman peaks from the blank (Lawaetz and Stedmon, 2009). Consequently, results were expressed in Raman units (R.U.). Parallel Factor Analysis (PARAFAC) was used to decompose the fluorescence signal and to obtain the main fluorophores present in our dataset (Stedmon and Bro, 2008). The results of the PARAFAC modelling were validated by split-half analysis provided in the tool box.

The concentration of ${\rm {NO}}_3^- $, ${\rm {NO}}_2^- $, ${\rm {NH}}_4^+ $ and SRP in all water samples were determined using a Lachat Quickchem-8500 Flow Injection Analysis system according to standard colorimetric methods (Hansen and Koroleff, 1999). The equipment detection limit for all these solutes was below 0.3 μmol/L and the method precision was higher than 4%. δ¹⁵N-${\rm {NO}}_3^- $ was determined using the bacterial reduction method (Pseudomonas aureofaciens; Sigman et al., 2001) after removal of ${\rm {NO}}_2^- $ (Weigand et al., 2016). N₂O produced from the ${\rm {NO}}_3^- $ reduction vials were concentrated in a Thermo-Fisher Precon system (Thermo Fisher, USA) via a liquid nitrogen loop with subsequent flow-through Finngan chromatographic loops and finally into a Thermo-Fisher Delta V Advantage isotope system (Jin et al., 2020). The method precision was ca. 0.2‰ based on the long-run. Calibration was made using commercial standards (USGS 32, 34, 35; IAEA-${\rm {NO}}_3^- $).

Sediment microbiota of the mixed sediment preserved at −20°C was quantified as the 16S rRNA gene fragments to determine the microbial community structure. Briefly, DNA from approximately 0.3 g thawed sediment was extracted and V3-V4 regions of 16S rRNA genes were then amplified after 35 PCR cycles. After library construction using KAPA Hyper Prep Kit (Roche), the obtained library was paired-end sequenced on an Illumina Miseq system (Jiang et al., 2020). The potential functional profiles of sediment microbiota were predicted in R environment using an open-source package that identifies the functional capabilities of microbial communities based on 16S rRNA marker gene profiles (AßHauer et al., 2015; Wang et al., 2020).

The mean concentration of solute in each treatment was the average from three reactors. The reaction rate for each solute was calculated with the input-output concentration difference from the individual reactor according to Ibánhez and Rocha (2014) and the mean rate was the average of the three reactors. Accordingly, the statistical errors from the concentration variation between reactors could not propagate to the reaction rate calculation and reduce the standard deviation. Positive reaction rates indicate the production of a specific solute and negative reaction rates represent a net consumption. To identify the effects of different salinity and input ${\rm {NO}}_3^- $ concentration on reaction rates, the data were analyzed in Minitab 18 by one-way ANOVA (tested factor, such as salinity or ${\rm {NO}}_3^- $ input concentration) after checking for normality and homogeneity of variance at a 95% significance level.

The operational taxonomic units in the sediment were 1 705. These operational taxonomic units were identified as more than 10 phyla with the dominant phylum as Protobacteria (Table 1). Important opportunity phyla included Actinobacteria, Bacteroidetes and Firmicutes. Among these phyla, the majority was identified as heterotrophic and actively involved in the nitrogen, oxidative and sulfur metabolism (Supplementary Fig. S1). At the class-level, the dominant species was Alpha-proteobacteria (Table 1). For the microbiota involved in the ${\rm {NO}}_3^- $ biogeochemical processing, the dominant nitrifier in the sediment was Nitrosomonas and Nitrospira, while potential denitrifiers were mainly identified as Bacillus, Lactobacillus and Paracoccus (Table 1).

In all the experiments, DO concentrations at the input endmember were near saturation. After passing through the reactor cells, DO concentrations markedly decreased (Figs 2a-c). In the control (i.e., with waters not amended with ${\rm {NO}}_3^- $), DO concentrations in the outflow varied between 257 and 194 μmol/L. Accordingly, DO consumption rates ranged between 9.2 nmol/cm³ and 15.6 nmol/cm³ sediment per hour with the highest value recorded at intermediate salinity (20). In the remaining treatments, the addition of ${\rm {NO}}_3^- $ did not significantly influence the DO concentration in the outflow. Accordingly, DO consumption rates among the treatments showed similar distribution along the used salinity range (p>0.05).

In the control, DOC concentration in the input solutions from 360 μmol/L (river water) to 225 μmol/L (seawater). After flowing through the reactor columns, DOC concentrations increased in all salinity treatments (Fig. 2d), indicating that permeable sediments acted as a source of DOC. The highest production rate was 3.0 nmol/(cm³·h) at a salinity of 20. When additional ${\rm {NO}}_3^- $ was introduced into the system, DOC concentrations at the output stream were also higher than the values obtained at the input solution (Figs 2e and f). Compared with the control group, the DOC production rates from the two ${\rm {NO}}_3^- $ spiked treatments were significantly higher (p<0.05).

At the output of the control group, ${\rm {NO}}_3^- $ concentrations increased, leading to an active production trend (Fig. 3a). When river water was spiked with ${\rm {NO}}_3^- $, the sediments (salinity 0 to 30) shifted to a net sink for ${\rm {NO}}_3^- $ with the most active ${\rm {NO}}_3^- $ removal occurring at a salinity of 10 (Fig. 3b). In addition, the active ${\rm {NO}}_3^- $ removal further enhanced the sediment DOC production capability (Fig. 4a). When maintaining high and similar ${\rm {NO}}_3^- $ concentration levels at all the salinities used (3rd treatment), ${\rm {NO}}_3^- $ reduction rates at high salinity markedly increased (Fig. 3c), producing a linear correlation between ${\rm {NO}}_3^- $ removal rates and salinity (Fig. 4b). Due to the identical ${\rm {NO}}_3^- $ input concentrations in the 3rd treatment, no clear pattern was found between ${\rm {NO}}_3^- $ reaction rates and ${\rm {NO}}_3^- $ concentrations (Fig. 4c), while two parabolic lines were outlined in the 2nd treatment, suggesting the impacts from ${\rm {NO}}_3^- $ availability on its benthic removal. Coupled with ${\rm {NO}}_3^- $ reaction dynamics, δ¹⁵N-${\rm {NO}}_3^- $ co-varied, ${\rm {NO}}_3^- $ production in the control group triggered decreases in δ¹⁵N-${\rm {NO}}_3^- $ (Fig. 3d), while the occurrence of ${\rm {NO}}_3^- $ consumption led to an increase in δ¹⁵N-${\rm {NO}}_3^- $ in all cases (Figs 3e and f). ${\rm {NO}}_3^- $ reaction rates, including production and consumption, showed a linear correlation with the Δδ¹⁵N-${\rm {NO}}_3^- $ (difference in δ¹⁵N-${\rm {NO}}_3^- $ between output and input) as outlined in Fig. 4d.

In the control, after passing through the sediments, ${\rm {NH}}_4^+ $ concentrations significantly increased (p<0.05) with higher production rates in the 10−30 salinity range (Fig. 5a). When exogenous ${\rm {NO}}_3^- $ was added, the sediments continued to act as a source of ${\rm {NH}}_4^+ $. However, reaction rates dropped (p<0.05), especially in the intermediate salinity runs (Figs 5b and c). For ${\rm {NO}}_2^- $ in the control, the sediment was an active producer and concentrations peaked at salinities of 10 and 20 (Fig. 5d). In contrast, after ${\rm {NO}}_3^- $ addition, ${\rm {NO}}_2^- $ cycling in the sediments changed. Both ${\rm {NO}}_2^- $ production and removal were observed, randomly distributed along the salinity gradient. Moreover, the magnitude of reaction rates shrank (Figs 5e and f).

SRP concentrations in the input endmembers were lower than 1.2 μmol/L in the control group (Fig. 5g). SRP concentrations at the outflow water dropped, resulting in a net SRP consumption. This SRP consumption tended to remain similar among the different salinity treatments. Moreover, the addition of ${\rm {NO}}_3^- $ did not significantly influence SRP removal rates (Figs 5h and i).

Four FDOM components were identified via PARAFAC modeling of our FDOM dataset. The characteristic excitation and emission wavelength of each fluorophore are outlined in Table 2. The first FDOM fluorophore is similar to the spectra of pure tyrosine. It corresponds to the so-called Peak B in the Coble nomenclature (Coble, 1996) while FDOM Component 2 is similar to the spectra of pure Tryptophan and corresponds to Peak T (Coble, 1996). They both are associated to labile, low molecular weight, protein-like FDOM compounds (Stubbins et al., 2014). Their maximum fluorescence intensities (F_max) varied between approximately 0.08 to 0.16 R.U. in the input solutions used in all treatments. The ratios between input and output F_max were frequently higher than 1 (i.e., net production) along the salinity gradient employed for both protein-like FDOM components, particularly pronounced in the case of the intermediate salinity runs (Figs 6a and b). In the experiments amended with ${\rm {NO}}_3^- $, the input-output balance was also frequently higher than 1, especially in the salinity range of 10 to 30.

The other two FDOM components identified through PARAFAC modelling of our dataset correspond to two types of recalcitrant FDOM, terrestrial humic (Peak C) and microbial humic-like FDOM fluorophores (Peak M; Coble, 1996). These corresponds to high molecular weight, complex humic and fulvic substances of low degradability (Stubbins et al., 2014). In our experiment, both humic-like components showed that the sediment acted always as a source of these compounds to the circulating porewater (Figs 6c and d). Under low ${\rm {NO}}_3^- $ concentrations in the circulating porewater (control treatment), maximum production of these FDOM components was verified at intermediate salinities, similar to the two protein-like FDOM components. Nevertheless, when high with ${\rm {NO}}_3^- $ concentrations were presented (i.e., at low salinities in the 2nd treatment and in the entire 3rd treatment), an enhancement in the production of the two humic-like FDOM components is observed (Figs 6c and d).

FTRs are powerful tools to investigate biogeochemical reactions in permeable sediments under the 1-D advection condition. Still, porewater transport processes in FTRs are not directly comparable with fluid circulation in “real world” conditions. Specifically, in natural environments, solute transport between permeable sediments and overlying water is largely dependent on advection resulting from the pressure gradient caused by waves, current interaction with topography, tide or piezometric head among other process (Rocha, 2008). The advection rate thus is variable in time but also spatially, as the interfacial flow direction at any point is controlled by the local permeability tensor (Rocha et al., 2005). Due to the tight coupling among reaction rates and advective transport in permeable sediments (Rocha, 2008; Ibánhez et al., 2011; Huettel et al., 2014), direct extrapolation of reaction kinetics and relationships between biogeochemical transformations and environmental variables obtained with FTR experiments requires a “double-check” via a data comparison.

DO, ${\rm {NO}}_3^- $ and SRP reaction rates obtained in this study are comparable to others previously published FTR studies or simulation results. In the current research, DO consumption rates ranged from 92 nmol/(cm²·h) to 156 nmol/(cm²·h) for the entire sediment column (10 cm sediment column). These values are in the range of those obtained in different marine permeable sediments such as those found by Ibánhez and Rocha (2016) in the Ria Formosa Lagoon using the FTR technique (56 nmol/(cm²·h) to 168 nmol/(cm²·h), according to O₂) and Rao et al. (2007) in sediments from the South Atlantic Bight using a series of 15.8 cm length of FTRs (86.9 nmol/(cm²·h) to 240 nmol/(cm²·h), according to O₂). These are also comparable to the benthic chamber incubations in the German Bight using (120 nmol/(cm²·h) to 160 nmol/(cm²·h), according to O₂; Janssen et al., 2005).

Present results for ${\rm {NO}}_3^- $ production rates (0.4 nmol/(cm²·h) to 21 nmol/(cm²·h) for the 10 cm sediment column) are also comparable to a series of incubation experiments targeting permeable sediments, such as nitrification rate estimates in the Ria Formosa Lagoon (7.0 nmol/(cm²·h) to 10.4 nmol/(cm²·h); Ibánhez and Rocha, 2017) and the German Bight (2.2 nmol/(cm²·h) to 9.4 nmol/(cm²·h); Marchant et al., 2016). In terms of the ${\rm {NO}}_3^- $ reduction rates, our results (9.8 nmol/(cm²·h) to 48 nmol/(cm²·h)) are comparable to those obtained by Jiang et al. (2018a) (7.0 nmol/(cm²·h) to 68 nmol/(cm²·h)) who applied similar input ${\rm {NO}}_3^- $ concentrations in FTR experiments run with Ria Formosa sediments. The rapid removal of SRP in the test sediments also agrees well with several publications on SRP sorption dynamics in sandy coastal sediments (e.g., Leote et al., 2013; Yang et al., 2018).

${\rm {NO}}_3^- $ production and reduction showed significantly distinct rates among the three treatments tested in our experiments (control, river water amended with ${\rm {NO}}_3^- $ and all amended with ${\rm {NO}}_3^- $). The net ${\rm {NO}}_3^- $ production observed in the control treatment suggests the dominance of nitrification over ${\rm {NO}}_3^- $ reduction, likely due to the low levels of ${\rm {NO}}_3^- $ and DO saturation imposed in the input solutions. This is generally observed in pristine (${\rm {NO}}_3^- $ unpolluted) estuaries in remote regions (Seitzinger, 1987; Jiang et al., 2019). The intensity of net ${\rm {NO}}_3^- $ production changed along the salinity gradient imposed, with a clear accumulation of both ${\rm {NO}}_3^- $ and ${\rm {NO}}_3^- $ at intermediate salinities (~20, Figs 3a and 5d). Coupled with enhanced ${\rm {NO}}_3^- $ production at intermediate salinities, δ¹⁵N-${\rm {NO}}_3^- $ decreased, indicating the occurrence of enhanced nitrification (Marchant et al., 2016). Similar enhanced ${\rm {NO}}_3^- $ production at intermediate salinities were found by Zhou et al. (2017) and Magalhães et al. (2005) in benthic habitats.

In the control, ${\rm {NH}}_4^+ $, i.e., the substrate for nitrification, did not show consumption rates equivalent to the net ${\rm {NO}}_3^- $ production observed. Hence, the studied permeable sediments acted as a source of ${\rm {NH}}_4^+ $. In permeable sediments, several sources might be contributors to the porewater ${\rm {NH}}_4^+ $ pool. N fixation, mediated by benthic microbiota, including a wide range of Alpha-proteobacteria (Dekas et al., 2018), should be considered as an important contributor (outlined in Fig. 7). Notably, carbon supply has been frequently found to be a pivotal factor for fixation activity (Changjiang River Estuary, Hou et al., 2018; Hydrate Ridge, Dekas et al., 2018). In the present study, the sediment organic matter was limited due to the nature of the “sands” and the input solution was oligotrophic, indicating that N fixation via Alpha-proteobacteria was depressed. ${\rm {NH}}_4^+ $ desorption from the sediments due to changes in the partition coefficient induced by varying salinity (Ibánhez and Rocha, 2017) also impacts porewater ${\rm {NH}}_4^+ $ levels, while this effect might be minimized in the incubation period, due to 12 h FTR stabilization prior to the sample collection. Accordingly, the main source of ${\rm {NH}}_4^+ $ feeding nitrification should be that produced through mineralization and ammonification. Considering that DOC and labile FDOM were not consumed in the different treatments, sedimentary organic matter (SOM) seemed to be the main carbon source to support the intensive oxidative metabolism (Supplementary Fig. S1), especially aerobic respiration. This agreed with the observation of Jiang et al. (2018a) and Ibánhez and Rocha (2014), who found that, despite low SOM levels in permeable sediments of Ria Formosa, these acted as net sources of DOC. During SOM decomposition via heterotrophic processes, organic N is mineralized into inorganic fractions and these products can be nitrified (outlined in Fig. 7). Such reaction chain leads to the coupling of SOM aerobic mineralization and nitrification in benthic environments, as outlined by the DOC production and nitrification rate in Fig. 4a. The decrease of δ¹⁵N-${\rm {NO}}_3^- $ reinforces the occurrence of active nitrification in the 1st treatment. According to the intensive requirement (first-order kinetics) on the substrate availability in benthic nitrifiers (Ibánhez and Rocha, 2017), it is not surprising to find the pattern that high reaction rate of nitrification at intermediate salinities in the control treatment is obtained. This reaction peak is in line with results from experiments run by Chambers et al. (2013), who also show nitrification is stimulated at intermediate salinities. Moreover, salinity increase leads to a reduction of ${\rm {NH}}_4^+ $ adsorption potentials of estuarine sediments (Weston et al., 2010), indicating that ${\rm {NH}}_4^+ $ produced from mineralization or ${\rm {NH}}_4^+ $ adsorbed on sand particles could be more rapidly available for nitrification.

The variability of nitrification rates under different salinities may also result from the composition of the nitrifier microbial community. Two types of nitrifiers, ammonia-oxidizing Archaea (AOA) and ammonia-oxidizing Bacteria (AOB) have been identified in coastal sediments (Bernhard and Bollmann, 2010). In the collected sediment, AOB was the dominant nitrifier in sediments with high-abundance Nitrosomonas and Nitrospira (Table 1). Salinity preferences have widely been identified in both nitrifiers. For instance, Nitrospira showed a great abundance at salinity of 3.4 along Chinese tidal flat wetlands (Sun et al., 2020). Jones and Hood (1980) isolated Nitrosomonas strains from the salt marsh at Lac des Allemands, USA that showed an activity peak at salinity 6 to 10. Wang and Gu (2014) revealed that the growth of Nitrosomonas strains in mangrove sediments (Mai Po Nature Reserve) was stimulated by saltwater (salinity, 20). Although there exist differences in the microbial community composition between sites, these observed salinity-preference patterns suggest that nitrification rates in coastal sediments are likely to occur at increasing salinity in the mixing gradient between fresh and seawater and peak at the medium salinity (brackish environment). During the short-term incubation, the variation of microbial community in the sediment was assumed to be limited (Mariotti et al., 1981), supported by the linear correlation between δ¹⁵N-${\rm {NO}}_3^- $ and reaction rates (Fig. 4d). Alternatively, the salinity induced microbial activity variation is deemed to be responsible for the reaction intensity change.

When external ${\rm {NO}}_3^- $ was introduced into the system, the balance between ${\rm {NO}}_3^- $ production and removal in the studied permeable sediment markedly changed, shifting to a net sink for ${\rm {NO}}_3^- $. Furthermore, compared to the control treatment, ${\rm {NO}}_2^- $ in the porewater sustained at low-level, indicating a rapid transformation for nitrogen oxides in the sediment. However, the production/reduction of ${\rm {NO}}_3^- $ in the 2nd and 3rd treatments along the salinity gradient was markedly different, suggesting overlapped influences from both salinity and ${\rm {NO}}_3^- $ concentration changes. With an identical ${\rm {NO}}_3^- $ input throughout the salinities used (3rd treatment), ${\rm {NO}}_3^- $ removal rates peaked and δ¹⁵N-${\rm {NO}}_3^- $ enrichment in the highest salinity scenario (Figs 3c and f). In the benthic environment, though the presence of Anammox, denitrification is assumed to be the major contributor of ${\rm {NO}}_3^- $ removal, especially in the aerobic condition (Burgin and Hamilton, 2007). In the present study, a high abundance of potential denitrifers in the sediment was observed, while microbial carriers for Anammox (e.g., Pirellula and Planctomyces) were not obtained (Table 1). Additionally, the combination of increased δ¹⁵N-${\rm {NO}}_3^- $ and high ${\rm {NO}}_3^- $ removal rates (Fig. 4d) reinforces the presence of dominant removal reaction, i.e., denitrification here, for the ${\rm {NO}}_3^- $ concentration decreases (Granger et al., 2008).

In previous research, positive relationships between salinity and benthic denitrification or ${\rm {NO}}_3^- $ reduction were observed (Laverman et al., 2007; Marks et al., 2016; Zhou et al., 2017). The denitrification process can be conducted by a wide range of bacteria. These bacteria have distinct salinity preferences (Santoro et al., 2006). The bacteria frequently observed in the marine environment, e.g. Pseudomonas sp. and Marinobacter sp. (Yoshie et al., 2004), are still functional in the freshwater environment (Yoshie et al., 2004; Miao et al., 2015), while the inverse trend might not be as likely because of the high osmotic pressure in the saline environment. In wastewater treatment reactors, Bacillus and Paracoccus were frequently observed and showed higher activity with increasing salinity (Silva et al., 2018). Moreover, Bacillus was frequently isolated in marine environments, such as shrimp culture ponds (Song et al., 2011). Additionally, at the time of sediment collection during the present study, the salinity of the overlying water in the sampling site was 34 (high tide) and 31.5 (low tide). Thus, it would not be surprising to observe the highest ${\rm {NO}}_3^- $ consumption rate in the most saline treatment when amended with ${\rm {NO}}_3^- $.

In freshwater ecosystems, such as rivers and groundwater, widespread anthropogenic enrichment of ${\rm {NO}}_3^- $ due to the intensive application of chemical fertilizers, sewage discharge, and manure disposal is observed (Galloway et al., 2014). The mixing between freshwater and seawater in the coastal zone can produce large spatial and temporal gradients in both salinity and ${\rm {NO}}_3^- $ concentration. Under such circumstance, the peak of denitrification occurred in the intermediate salinity (Fig. 4b), outlined by the 2nd treatment. Given the strong reaction potential in the saline environment, the occurrence of this peak indicates the balance between ${\rm {NO}}_3^- $ supply and salinity. In low salinity waters, because the ${\rm {NO}}_3^- $ supply was sufficient (>60 μmol/L here; five times higher than the riverine ${\rm {NO}}_3^- $ concentration, Fig. 3b), salinity mediated bacterial activity was the dominant factor controlling ${\rm {NO}}_3^- $ removal rates, producing a negative correlation between ${\rm {NO}}_3^- $ concentration and denitrification rate (Fig. 4c). Under high salinity conditions, the dilution from a large portion of oligotrophic seawater decreases ${\rm {NO}}_3^- $ concentration. As a reaction strictly driven by the substrate availability, the significant requirement of ${\rm {NO}}_3^- $ from denitrifiers and the constraint of ${\rm {NO}}_3^- $ on denitrification rate has been determined via in situ observation (seepage face; Ibánhez et al., 2011), simulation experiments (FTRs; Jiang et al., 2018a), as well as denitrifier-inoculated reactors (Jafari et al., 2015). Moreover, in laboratory culture, nitrate reductase in Bacillus (potential denitrifier at our site) was also proved to be activated by high-level ${\rm {NO}}_3^- $ in culture solution (Schulp and Stouthamer, 1970). Accordingly, it can be deduced that sediment denitrification processes at high salinity from the 2nd treatment was constrained by ${\rm {NO}}_3^- $ supplies due to seawater dilution, especially compared to the highest ${\rm {NO}}_3^- $ consumption rate measurement when seawater was spiked with ${\rm {NO}}_3^- $ (3rd treatment).

In coastal zones, this trend can be a key factor in understanding the modulation of land-derived ${\rm {NO}}_3^- $ conducted through permeable sediments (transformation pathways outlined in Fig. 7). In oligotrophic environments, permeable sediments continuously provide ${\rm {NO}}_3^- $ into the overlying water via organic matter mineralization and nitrification, which can be used by primary producers. Enhanced ${\rm {NO}}_3^- $ loading from human activities stimulates sediment denitrification and initiates the role of “${\rm {NO}}_3^- $ cleaner”. This balance between these contrasting roles in benthic sediments may help the coastal system maintain primary productivity and ecological stability, acting as a “nutrient modulator” (Huettel et al., 2014). Moreover, compared to the rapid shift in the role played in the ${\rm {NO}}_3^- $ processing, the preferential salinity of sediment microbial community was stable in a short temporal scale (3rd treatment). It suggests that environmental stresses in coastal systems result from the coupling of intensive ${\rm {NO}}_3^- $ loading and fresh water discharge (triggered by weather extremes, such as floods in the upper stream driven by storms) and might exceed the ${\rm {NO}}_3^- $ removal capability deduced from saline environments. Consequently, a significant fraction of redundant ${\rm {NO}}_3^- $ might temporally escape from the benthic sediment and transfer into the food web, triggering unexpected ecological issues. This potential input-feedback loop deserves more attention from a coastal management perspective, especially in the handling of weather extremes.

Notably, in the present study, ${\rm {NO}}_3^- $ removal occurred in overall oxic conditions, indicating that denitrifiers rely on microniches on the surface of the sediment particles (Jiang et al., 2018a). Coupled with ${\rm {NO}}_3^- $ reduction, SOM was continuously decomposed, releasing additional DOC fractions (linear regression in Fig. 4a) such as humic-FDOM solutes into the porewater when compared with the control treatment (Figs 6c and d). This observation agrees well with FTR experiments conducted with sediments from the Ria Formosa Lagoon (Ibánhez and Rocha, 2014), where the addition of ${\rm {NO}}_3^- $ to the circulating porewater enhanced DOC and humic-like FDOM production. In their experiments, SOM was characterized by a relatively high C:N concentration ratio, similar to that found in our sampling site (C:N=15), which together with low standing stocks of SOM explains the observed C limitation of the microbial community of Ria Formosa permeable sediments (Ibánhez and Rocha, 2014). Nevertheless, enhanced humic-like FDOM production, linked to active organic matter degradation, after ${\rm {NO}}_3^- $ addition to the porewater suggest that the microbial community was also N limited. Thus, as in our case, ${\rm {NO}}_3^- $ addition increases the capacity of the microbial community to process SOM thus favoring heterotrophic processes.

Coupled with the active metabolism functions (Supplementary Fig. S1), especially benthic mineralization and denitrification, that continuously consume SOM, release of SRP into the porewater should be observed in all treatments. In addition, increasing salinity potentially benefits the SRP desorption process because of enhanced ion exchanges on the particle surface. However, low SRP concentrations and the related net consumption are found in the outflow, regardless of salinity and DO consumption rates. Considering that primary production in the sediments was eliminated by imposing dark conditions and that oxic conditions were imposed, SRP adsorption on the sandy sediment particles due to the presence of metal oxidants, such as iron and manganese, was deemed to be responsible for the observed SRP concentration decline (Yang et al., 2018). In fact, inorganic P content in the tested sediment was higher than the concentration of sedimentary N, which was contrary to the Redfield ratio (N:P=16:1). This reinforces the strong SRP adsorption capability of coastal sediments on overlying water or porewater (Jiang et al., 2018b). In marine environments, the stoichiometric balance between N and P is a key for the phytoplankton community structure (Lin et al., 2020). Our results show that under oxic conditions, permeable sediments from the Dublin Bay may increase the N:P ratio in the outflow relative to the Redfield ratio, potentially driving the system to P-limitation. This change might induce phytoplankton community shift from the diatom-dominance to dinoflagellate-dominance due to the weak N competition capability of diatoms (Rocha et al., 2002), particularly together with the variation in water temperature (Li et al., 2011). Generally, diatoms are the most important primary supplier in the food web and directly contribute to filter feeding fish and shellfish populations. In comparison, dinoflagellates are an unfavored food source for high-level consumers due to toxin defensiveness (Verma et al., 2019). Therefore, long-term P adsorption in permeable sediments during the porewater exchange might lead to ecological pressures. In addition, when redox conditions change (aerobic to anaerobic), adsorbed P in the permeable sediments would be released into the overlying water (Jiang et al., 2018b), potentially driving the system to N limitation. Such shift between N and P limitation caused by permeable sediments could be of extreme importance for the biological structure of shallow coastal waters.