Introduction

Changes in the global nitrogen cycle have pushed the Earth’s system outside the relatively stable state that had seen human civilisations arise1. Anthropogenic rates of N2 fixation, mostly due to industrialised forms of agriculture, now exceed natural biological fixation rates2. Estuaries and coastal waters receive much of this anthropogenic nitrogen, where it stimulates the production of excess organic matter (eutrophication3). Nitrogen is permanently removed from the coastal zone via denitrification and anammox, which convert fixed nitrogen to di-nitrogen gas (N2) that is ultimately lost to the atmosphere. Denitrification typically accounts for most of the N2 produced in coastal sediments, with a widely reported ratio of 71:29 for denitrification:anammox4. This, in turn, acts as a control on ecosystem primary productivity5. However, denitrification also produces nitrous oxide (N2O)6, a greenhouse gas with a global warming potential 296 times more than CO27.

In addition to excess nitrogen loading, increasing sea surface temperatures (SST) and ocean acidification (OA) are also impacting coastal ecosystems, with the cumulative effects increasing the risk of hypoxia potentially exacerbating the acidification of subsurface waters. The IPCC “high emissions” scenario projects a doubling of atmospheric CO2 concentrations, reaching upwards of 1000 ppmv by the end of the century (RCP8.58), resulting in increased global mean SST of ~2–4 °C9. The partial pressure of CO2 (pCO2) in surface water will increase in line with atmospheric concentrations, leading to ocean acidification (OA), with increased concentrations of dissolved inorganic carbon (DIC), and a ~0.3 unit pH decrease10. Understanding how warming and OA will affect denitrification, and the associated production of N2O is essential to assess the nitrogen removal capacity of the coastal zone, and the positive warming feedback loop from N2O production in a high-CO2 climate.

Warming increases potential dark denitrification up to an optimal temperature (Topt)11,12,13. Potential dark denitrification rates may have a Topt greater than anammox, suggesting that denitrification may become even more dominant as temperatures increase which, in turn, may increase the potential incomplete reduction and production of N2O11. However, warming was unlikely to directly affect the relative reduction of N by anammox or denitrification based on similar Topts for the two pathways in temperate estuarine sediments14. In contrast, OA effects on denitrification have not been widely investigated. In one study with high CO2 concentrations and pH buffered to 7.2 there was an inverse relationship between CO2 concentration and the rate of total N removal or denitrification efficiency, through the inhibition of denitrifying microbes’ intracellular electron transport and consumption system15. Yet, despite both warming and OA having the potential to affect denitrification and N2O production rates, both individually and together, this combined effect has never been studied. Denitrifying and nitrifying gene abundance (16S rRNA genes, amoA and nirS) under individual and combined stressors of elevated temperatures (Δ+4 °C) and CO2 (750 ppm) have been investigated, with lower gene abundance in combined stressor treatments than at elevated temperatures under current CO216. A lower gene abundance may indicate that despite warming, OA could lead to a decrease in N2 and N2O production in a high-CO2 climate, but no study has directly measured these N2 and N2O flux rates under these conditions. Here we evaluate how denitrification and N2O production will be affected by a combination of increased temperature and pCO2.

Through a series of whole core incubation studies, we were able to assess the combined impact of warming and OA on estuarine sediment denitrification rates and N2O production. We found that OA may offset the warming increase in estuarine sediment denitrification and N2O production by up to 64% and reduce a potential positive climate feedback loop by inhibiting denitrification pathways.

Results

Denitrification

Denitrification (D14) rates were not significantly different between overlapping incubation controls in light and dark cycles (Supplementary Table 1) and thus were determined to be comparable and rates were left unadjusted. Supporting this, in situ, sediment characteristics were also unchanged across the two separate incubation times, with surface sediments C:N of 15.88 ± 1.34, and δ13C and δ 15N values of ~22‰ and 4.5‰, respectively (error <0.2‰ for C:N and ~1% for δ13C and δ 15N measured values).

There was a significant interacting effect of pCO2 and temperature on denitrification rates (F3,15 = 4.14, p = 0.025), however this difference was only measurable at 28 °C (Fig. 1a). For temperatures below 28 °C, increasing temperature resulted in significant increases in D14 (F3,8 = 22.03, p = 0.0003; F3,7 = 53.53, p < 0.0001; for current and high pCO2, respectively, Fig. 1a). Both pCO2 conditions resulted in a decrease in net D14 at 28 °C, but this decrease was significantly greater under high-pCO2 (F1,3 = 20.34, p = 0.02). The ~70% decrease in net D14 under high-pCO2 was a result of a decrease in light cycle net production that was 40% greater than the decrease under current-pCO2 rates.

Fig. 1: Effect of temperature on total denitrification (D14), and denitrification of NO3- from the water column (Dw), and from nitrification (Dn) fluxes under current and high-pCO2.
figure 1

Net diel (a) D14, (d) Dw, and (g) Dn fluxes (µmol-N m−2 h−1), under current-pCO2 (open boxes) and high-pCO2 (hatched boxes) conditions at three temperature offsets from control, 23 °C. Letters identify significantly different means across temperatures under current-pCO2 and numerals identify significantly different means across temperatures under high-pCO2 conditions, where letters or numbers that are the same indicate no significant difference, as determined by a one-way ANOVA and post hoc Tukey’s test. Light and dark fluxes for b/c D14, e/f Dw, and h/i Dn fluxes (µmol-N m−2 h−1) under current/high-pCO2 conditions, respectively. Here, letters identify significantly different means across temperatures in light and numerals identify significantly different means across temperatures in dark cycles, where letters or numerals that are the same indicate no significant difference, as determined by a one-way ANOVA and post hoc Tukey’s test. Statistical details are in Supplementary Tables 4 and 5. All boxes show mean values ‘×’, with the middle horizontal line indicating the exclusive median and the top and bottom edges indicating the start of the upper and lower quartiles, respectively.

Full size image

Greater insight into these differences was gained by the dissection of D14 into Dw and Dn. The D14 response was dominated by Dn across all treatments (>50% of total D14, Fig. 1), except at 28 °C under current-pCO2, where Dn only accounted for 44% of D14. However, there were no significant interactive effects of pCO2 and temperature (F3,16 = 2.68, p = 0.08) and/or effect of pCO2 on Dn production of N2 (F1,16 = 0.71, p = 0.41). Differences in D14 production between pCO2 conditions were therefore driven by an interactive effect of pCO2 and temperature on Dw response (F3,15 = 16.43, p < 0.0001).

Under current-pCO2, net Dw rates increased significantly with increased temperature (F3,8 = 41.58, p < 0.0001; Fig. 1d), with control rates (38 µmol-N m−2 d−1) doubling in Δ+3 °C and Δ+5 °C cores (74 and 81 µmol-N m−2 d−1, respectively). In contrast, under high-pCO2 Dw rates only increased at 26 °C (to rates ~27% less than those at current-pCO2). Dw rates were significantly lower at 28 °C (F3,8 = 46.18, p < 0.0001; Fig. 1d), around ~67% below rates measured at current-pCO2. This greater drop at 28 °C compared to 26 °C was a result of significant drops in light and dark production, versus drops just in dark production (Fig. 1e, f).

The Topt and Q10 denitrification rates changed with OA, with responses between dark and diel cycles varying between the total (D14) and its components (Dw and Dn). In general, the Q10 of diel rates was greater than for dark rates, and Dw Q10 responded differently to OA than Dn and D14. OA increased dark D14 and Dn Q10 from 3.14 and 3.28 under current-pCO2 to 5.92 and 13.37, respectively, without affecting the Topt of 26 °C (Table 1). In contrast, dark Dw Q10 decreased with OA, from a current-pCO2 Q10 similar to that of D14 and Dn (3.33 to 1.61), also reducing the apparent Topt by 5 °C, from 28 to 23 °C. These same Q10 and Topt trends with D14, Dn, and Dw were calculated from diel rates, with OA increasing diel D14 and Dn Q10 from 5.91 and 5.61 to 9.06 and 12.71, respectively, with Topt at the same 26 °C, and Dw decreasing with OA from 5.20 to 3.19, with Topt decreasing from 28 °C to 26 °C. Including light in diel N2 production increased Q10 of Dw under both current and high-pCO2 by 88% and 98%, respectively, D14 by 88% and 53%, respectively, and Dn under current-pCO2 only, by 71%, with a slight decrease of 5% under high-pCO2.

Table 1 Dark and Diel Q10 and (Topt) values for D14, Dw, and Dn under current and high-pCO2
Full size table

N2O fluxes

There was a significant interactive effect of warming and OA on sediment net diel N2O responses (F1,12 = 24.02, p = 0.0003; Fig. 2a). Significant effluxes due to warming (162 ± 34 nmol-N m−2 h−1; one-way: F1,6 = 31.19, p = 0.0014; Fig. 2a) diminished under combined warming and OA conditions (by ~64% to 59 ± 8 nmol-N m−2 h−1), approaching OA-only efflux rates (63 ± 37 nmol-N m−2 h−1). Similar interactive effects of OA and warming were measured under light conditions (F1,12 = 18.72, p = 0.001; Fig. 2b), where control sediments shifted from an influx of N2O (99 ± 92 nmol-N m−2 h−1) to an efflux under warming and/or OA treatment conditions. Warming resulted in the largest effluxes (190 ± 34 nmol-N m−2 h−1) (F1,12 = 17.73, p = 0.001; Fig. 2b). In contrast, dark N2O fluxes did not significantly vary with treatment (p < 0.05; Fig. 2c). N2O contribution to gaseous N loss is <2% with its contribution increasing in warmer temperature but decreasing when warming and OA were considered together (Table 2). Note, negative net N2O contributions to gaseous N loss were calculated under control conditions (Table 2) because of the net N2O uptake in the light (Fig. 2b) coupled with a net loss of N2 (Fig. 1a).

Fig. 2: N2O fluxes across treatments.
figure 2

a Net diel, b light and c dark cycle rates (nmol m−2 h−1) under current-pCO2 at control and warming (T ↑ = Δ+3 °C) temperatures and under high-pCO2 conditions alone (CO2) and with warming (T ↑ CO2). All boxes show mean values ‘×’, with the middle horizontal line indicating the exclusive median and the top and bottom edges indicating the start of the upper and lower quartiles, respectively. Letters identify significantly different means between temperatures under current-pCO2 and numerals identify significantly different means across temperatures under high-pCO2 conditions, where letters or numbers that are the same indicate no significant difference, as determined by a one-way ANOVA and post hoc Tukey’s test. Statistical details are in Supplementary Tables 4 and 5.

Full size image
Table 2 Light, dark, and net contribution (%) of N2O to gaseous N loss
Full size table

Discussion

The influence of temperature on denitrification has been measured in dark anaerobic slurry incubations11,12,13 and in modelling studies17. However, by destroying the sediment vertical structure and removing the benthic-pelagic coupling and metabolic redox structure near the anoxic interface, anaerobic slurries omit the potential for substrate limitation to be captured in total denitrification rates. Reflecting this removal of substrate limitation in slurried sediments, Riekenberg et al. measured 3-36× higher N-uptake rates in slurries than measured in intact sediments. The whole core incubations used in the current study simulate actual benthic-pelagic coupling, capturing the potential limitations of coupled nitrification-denitrification and diffusive NO3- supply from the water column for denitrifiers in the sediment. As such, the whole core N2 flux measurements in the current study allowed us to understand temperature effects on denitrification based on NO3- from the water column (Dw) or from coupled nitrification/denitrification in the sediment (Dn).

Warming-related increases in N2 and N2O production under current-pCO2

The temperature dependence of net Dw rates under current-pCO2 was strong, with significant increases associated with warming. Similar to N2, N2O fluxes increased with increasing temperature (Fig. 2c), which is consistent with an increase in denitrification rates leading to greater N2O production11. It is important to note that in this experiment we did not trace nitrate reduction to N2O, but instead measured the net change in N2O in the overlying water columns. As such, this net N2O production may be a result of both denitrification and nitrification18. At projected warming (26 °C), net D14 and N2O rate increases in the light (Fig. 1b, e, h). The increased production of both N2 and N2O in the light is likely indicative of coupled nitrification-denitrification19 as seen in other net heterotrophic sediments20,21.

Under strong warming (28 °C), the decrease in D14 and Dn suggests a decrease in coupled nitrification-denitrification, despite a continued increase in Dw and N2O formation (Figs. 1 and 2, respectively). However, other studies have observed an increase in nitrification activity and gene abundance with increasing temperatures16, suggesting that there is a decoupling of denitrification and nitrification instead of a decrease in nitrification rates. This decoupling might have resulted from an increase in competition for NO3- produced in the sediments through nitrification. Dissimilatory nitrate reduction to ammonium (DNRA) can outcompete denitrification for NO3- in increasingly reduced sediments under sustained anaerobic conditions22,23,24 and may be favourable under warming conditions17, which in turn would decouple nitrification-denitrification. The DIC to NH4+ efflux ratio measured from the sediments (with a C:N of 16:1) in a parallel study under strong warming alone was 3.825 suggesting increased production of NH4+ (Supplementary Table 2). DNRA is consistent with excess production of NH4+ effluxed per mole of DIC compared to what is expected from the complete mineralisation of organic matter with a Redfield C:N ratio of 106:1626 or the in situ sediment organic matter, which had a C:N ratio of 16:125.

Reduced denitrification and N2O production under warming and ocean acidification

A decrease in dark and diel Dw rates and N2O fluxes at warmer temperatures (>23 °C) elevated-pCO2 (Figs. 1f, e and  2), may indicate unfavourable conditions for denitrifiers. A mesocosm study of temperate, Scottish estuarine sediments found that 16 S rRNA (nirS and amoA) gene abundances responsible for nitrogen cycling were lower in combined stressor treatments than at elevated temperatures under current-pCO216, indicating a reduction in the respective activities of denitrifiers and nitrifiers. However, the short duration of the incubations here would better reflect suppression of activity when exposed to unfavourable conditions, which may lead to a reduction of denitrifiers/nitrifiers in the long term.

Under strong warming (28 °C) and high-pCO2 there was a greater decrease in D14 than measured under current-pCO2, despite an equal decrease in Dn (Fig. 1a, g). This gap reflects the large decrease in Dw at increased temperatures under high-pCO2 conditions only (Fig. 1d). The decrease in Dw and nitrification rates (included in D14 and Dn) that resulted from OA with strong warming may also reflect changes in gene abundance16. Unlike sediment responses to strong warming under current-pCO2, sediment responses under high-pCO2 did not suggest an increased role of DNRA in NO3- reduction25. This difference may be a result of reduced nitrification16 potentially limiting the coupling with DNRA. For example, pushing sediments to DNRA might not only reduce N2O production, but it might increase N2O consumption as organisms carrying out DNRA have been found to have atypical NosZ genes27. Alternatively, OA may have a similar inhibitory effect on DNRA as has been observed for denitrification and nitrification16.

Q 10 values for denitrification affected by ocean acidification

Under current-pCO2, dark denitrification rates appeared to be strongly dependent on temperature changes (Table 3). Q10 values for denitrification previously reported for subtropical unvegetated coastal sediments11,13 are similar to the values in the current study. Dark denitrification Q10 values calculated here under current-pCO2 (3.14–3.33; Table 3) are closer to those reported by Canion et al.13 (3.8–5.0) than by Tan et al.11 (1.9–2.7). Instead, Tan et al.11 denitrification Q10 values appear to reflect Q10 values of temperate and polar regions13,28,29. Both Tan et al.11 and Canion et al.13 Q10 ranges were calculated from potential denitrification rates, compared to the actual denitrification rates measured here. Regardless, this strong positive Q10 is consistent with N2O production increasing with warming, as was also measured in this study (Fig. 2), supporting previous findings that increases in temperature and denitrification activity will increase the occurrence of denitrification and the production of N2O11.

Table 3 Mean (±SD) temperature (°C) and pCO2 (µatm) conditions for each incubation
Full size table

The Topt of Dw rates in this study was 28 °C, the experimental maximum temperature (the actual Topt may have been higher). This is comparable to previously reported potential optimum temperatures of at least 30 °C11,12,13. However, D14 and Dn rates in the current study had a Topt of 26 °C (Table 3). The inclusion of nitrification in these measurements suggests that nitrification may have a lower Topt than denitrification30. Although, sediment nitrification rates at a similar latitude in the North Ariake Sea in Japan, had an estimated Topt of 28 °C31, the reason for the rate decrease after only 26 °C remains speculative.

In the current study, for the first time, the effect of OA on the temperature sensitivity of denitrification rates has been calculated. In the dark, OA reduced temperature stimulation of Dw rates (Fig. 1e, f), reducing Q10 from 3.33 to 1.61. N2O flux rates reflected this reduced temperature sensitivity in denitrification rates, with lower dark and net N2O effluxes under combined warming and OA conditions (Fig. 2b, c).

N2O fluxes identify potential limitations in sediment denitrification

Sediments under ambient conditions in the current study had the highest denitrification rates in the light period yet were a sink of N2O, and had reduced denitrification rates in the dark, despite becoming a source of N2O. Sediment uptake of N2O is not uncommon32,33,34 and relates to its use by denitrifiers as a terminal electron acceptor for OM degradation in the absence of NOx34, or its use in the oxidation of reduced sulfur compounds to produce N235. NOx availability in the oligotrophic Clarence Estuary waters was low (<2 µM), forcing denitrifiers to compete for NOx with the photoautotrophic microorganisms and bacteria living within the photic surface sediments. As a result, denitrifiers were likely forced to use N2O despite the preference for NOx as an electron acceptor36,37. This competition with photoautotrophs is absent in the dark, allowing for the uninhibited utilisation of NOx by denitrifiers. As such, the preference of NOx as an electron acceptor coupled with the excess of NOx availability would promote the incomplete reduction of NOx to N2O, and not N236,37. Alternatively, nitrification may have been responsible for driving N2O fluxes and this may have been alleviated under light conditions when oxygen concentrations were higher38.

N2O production rates can be accelerated when concentrations of NH4+ and NOx are elevated6, or when organic matter is limited39 and/or refractory40. N2O flux was greatest in both the light and dark under warming conditions relative to control or OA-amended treatments (Fig. 2). Elevated temperatures are expected to increase NH4+ and NOx availability25 due to thermal stimulation of organic matter remineralisation41,42 and nitrification31. This, coupled with the increased denitrification measured in this study (Fig. 1), accounts for the significant increase in net N2O effluxes under conditions of warming alone.

Under acidified conditions, warming did not affect sediment net and light/dark N2O flux (Fig. 2) despite the interactive effect that was found for Dw (Fig. 1d). This effect of OA on N2O flux was two-fold. OA reduced denitrification rates at elevated temperatures, reducing the potential for light and dark production of N2O. However, in the light, sediments were still a source of N2O relative to control conditions. Previous findings show that OA increases the availability of NOx to the sediments25 with a likely stimulation of nitrification from increased O2 production by photoautotrophs19, possibly reducing the need for denitrifiers to utilise N2O in their reduction of OM. Alternatively, a reduced microbial electron transfer efficiency during denitrification under OA conditions may have encouraged N2O production, versus its reduction43.

Previous studies of the effect of temperature on denitrification have been limited to potential rates measured in the dark11,13. This study has shown that light and OA alter the diel Q10, Topt, and overall response of denitrification activity. This is chiefly due to the potential for OA to stimulate photoautotrophic activity in the light44. Because photoautotrophic production could either slow the rates of denitrification due to increased competition for nitrate or increased dissolved oxygen in the experiment45,46, or stimulate denitrification with a coupled increase in nitrification rates19,21, the resulting effects are polarised. Further complications emerge when you consider the possibility that OA may make conditions unfavourable for denitrifiers16. However, in the present study, denitrification was generally greater in the light than in the dark (Fig. 1), suggesting that nitrification likely increased substrate availability and alleviated competition between denitrifiers and photoautotrophs. In contrast, OA has the potential to reduce the stimulating effect of warming on Dw (Fig. 1d) and N2O production (Fig. 2a) by up to 64%. These contrasting stimulating and reducing effects on denitrification and N2O highlight the need to (1) include a full diel cycle to determine the temperature sensitivity of sediment processes, and (2) consider the interactions of future climate conditions. Failure to do so may hinder our ability to make accurate future projections of climate-related impacts. This study was only undertaken at one site. Further, work should be undertaken at sites with different environmental conditions to see if the responses of denitrification and N2O sediment-water fluxes to warming and OA are similar. Moreover, our warming treatments were offset from the ten-year average for the Clarence estuary. Additional studies should be undertaken in winter and summer with warming treatments offset from average seasonal temperatures to see if the responses of denitrification and N2O sediment-water fluxes to warming and OA are similar under different seasonal conditions and strive to include longer-term (months) experiments to reflect the chronic processes of warming and OA.

Methods

Site description

Cores were collected on three separate sampling excursions within the same 4 m \(\times\) 12 m in the Clarence River, Yamba, NSW, Australia (29°24.21’S, 153°19.44’E; Supplementary Fig. 1). Two excursions used subtidal sediments (Summer—January 2018) with ~0.3–1.0 m of overlying water at low-tide, to measure denitrification rates. The third excursion used intertidal sediments that were collected as part of a separate study (Spring—September 2019). Intertidal sediments at this site are only exposed for up to 3 h during very low tides and are subtidal the remainder of the time. The average sea level at high tide is approximately 2 m above the sediment surface. Subtidal sediments (0–2 cm) had a higher porosity (0.43) and organic matter (OM) content (~3.5% of dry weight), determined from mass loss after combustion (490 °C for 4 h) of dried sediment (60 °C), relative to intertidal sediments (0–2 cm), which had a porosity of ~0.23 and OM content of ~1.34%. The top 10 cm of sediments were characterised via dry sieving analysis47,48, with both subtidal and intertidal sediments having a well-sorted sandy-mud grain size distribution (Supplementary Table 3). Dissolved inorganic nitrogen (DIN) concentrations at the time of this study (~1.19–2.40 µM, Table 4) were relatively low and consistent with previous studies in the Clarence River (<2 µM16), as a result of low nutrient loading in the catchment area49. Temperature (±0.3 °C) and salinity (± <1%) were measured over 24 h using a submerged Hydrolab (HL7) at the site. The mean daily water temperature at the site ranged from 21.3 ± 1 °C (max 24 °C) in September to 23.5 ± 1 °C (max 26 °C) in January. Salinity varied from 10–35 over the tidal cycle at both times of year.

Table 4 Mean (±SD) DIN concentration (µM) at the start of dark (Start D), transition to light (D/L) and end of light (End L) incubations
Full size table

Core collection

Acrylic cores (9 cm diameter × 47 cm) were used to collect sediments (around 20 cm depth allowing for ~1.8 L of overlying site water). To measure denitrification rates, 33 sediment cores were collected on January 09th, 2018 and 27 cores were collected on January 16th, 2018. Cores were collected from inundated sediment at low tide to ensure all sediments were subtidal. To measure N2O fluxes, 16 cores were collected on September 16th, 2019. Large burrow openings were avoided during all collections to reduce the variability among cores that would otherwise result from the occasional inclusion of large macrofauna50. In addition to sediment cores, 700 L of site water was collected on each excursion to fill a laboratory incubation set-up, in which all cores were submerged within 6 h of collection.

Climate scenarios

This investigation considered the temperature dependence of denitrification rates under both current and high-pCO2 conditions (~480 and ~1000 µatm, respectively). As such, sediments were subjected to both cooling and warming, covering an 8 °C range from 20 to 28 °C. Control temperatures were defined as the ten-year average annual sea surface temperature in Yamba, ~23 °C (17-27 °C)51. Three discrete temperature offsets from “control” were selected, (1) the projected end-of-century SST, Δ+3 °C (RCP8.59), (2) a greater warming condition, Δ+5 °C, and (3) a lower temperature condition, Δ−3 °C. Although sediments occasionally experience these offset temperatures throughout the year, the objective of this study was not to stress the sediments, but instead to quantify the effect that incremental temperature increases have on benthic denitrification rates. Moreover, the generation time of the dominant microbial members of unvegetated estuarine sediments was from 6–15 h52,53, thus sediments were given sufficient time for multiple generations during a 48–72 h acclimation period to treatment conditions. Temperatures were maintained at 23 ± 0.5 °C (control), 26 ± 0.5 °C (RCP8.5 projected warming), 28 ± 0.5 °C (stronger warming), and 20 °C (cooling) throughout the experiment. OA conditions were simply defined as a doubling of current mean pCO2 (417 µatm54), under the high emission scenario (RCP8.58). To complement the denitrification incubation, N2O incubations used two temperatures, current mean ambient temperature (23 °C) and projected warming (26 °C), in a full factorial design with current and high-pCO2 levels (~450 and 900 µatm, respectively), to focus on projected warming and OA effects.

Incubation setup

In the laboratory, depending on the incubation, 4 to 5 tanks were fitted with a thermo-regulating aquarium pump to maintain temperature (± 0.1 °C). Current-pCO2 tanks were aerated with ambient laboratory air (~450 µatm) via an air pump and airstones, whereas high-pCO2 tanks were aerated with an enriched-CO2 airstream (~1000 µatm) to simulate the desired future atmospheric CO2 conditions (~900 µatm; Table 3) and to ensure water columns were saturated with O2 prior to the start of incubations (O2 concentrations were ~230 µmol L−1 at the start of incubations and remained above 152 µmol L−1 for all treatments, see44 for further detail). DIN concentrations were low at the start of incubations with details of start and end concentrations in Table 4). DIC concentrations were determined using an AIRICA system (MARIANDA, Kiel) via infrared absorption using a LI-COR LI-7000, and corrected for accuracy against certified reference material, batch #17155 for both the denitrification and N2O incubations. DIC was coupled with total alkalinity measurements using potentiometric titration on a Metrohm 869 Compact Sample Changer and an 848 Titrino plus dosing unit, according to the open-cell method described in Dickson56 for N2O incubations, and pH measurements made with a Hach HQ40d Multiprobe metre and probe calibrated to 3-point standard buffer scale (National Institute for Standards and Technology—NIST, R2 = 0.99) for denitrification incubations to calculate pCO2 offsets using CO2SYS following methods described in Simone et al.44,57.

Water columns within cores were stirred throughout the incubations via the interaction of magnetic stir bars with an external rotating magnet (60 rpm). Stir bars were positioned ~5 cm above the sediment surface to ensure sediments were not disturbed58,59. High-pressure sodium lamps (400 W; PHILIPS Son-T Argo 400) were positioned above the tanks so that each core would receive ~270–280 µmol quanta m−2 s−1 of photosynthetically active radiation (PAR) at the water surface, equivalent to mean daytime field conditions measured at 1 m depth with a HOBO light logger. These lamps were turned on in the mornings to simulate the day in line with natural diel light cycling and followed a ~ 12:12 h light:dark cycle, similar to that in situ.

For the denitrification incubations four temperature treatment tanks (20 °C, 23 °C (control), 26 °C, and 28 °C) were each filled with ~80 L of site water. Due to limited space, two separate denitrification incubations were done using the same laboratory setup to investigate N2 production responses to temperatures under current (~450 µatm; week of January 16th, 2018) and high-pCO2 conditions (~900 µatm; week of January 09th, 2018). Sediment cores were randomly allocated to the tanks. Two sets of 3 cores (n = 6) were placed uncapped in each tank, except for the control tank, which had an additional 3 cores submerged (n = 9). The additional 3 cores in the control tank were used for analysis of background isotope ratios in sediment and water compartments. For the high-pCO2 incubation, an additional tank was included to provide an overlapping control (control temperature, current-pCO2) and allow for comparison among treatments across the two incubations. The overlapping control cores were used to verify that sediments in the two separate incubations were behaving the same way in the absence of any treatment and to control for any differences between incubations resulting from the collection of sediments and overlying water at different times.

For the N2O incubations, four tanks were used to provide a fully factorial design of two temperatures (23 °C and 26 °C) at both current and high-pCO2 (~460 µatm and ~900 µatm), with each tank receiving four cores.

Sampling design and analysis

Denitrification

Cores were left open to acclimate in tank water with warming and OA conditions imposed for 48–72 h, before being capped to begin diel incubations, which encompassed an entire dark/light cycle. At the beginning of both the light and dark cycles, the starting concentration of NOx was determined for each core; 20 mL of water was syringe sampled from a resealable port in the lid of each core, filtered (0.45 µm cellulose acetate) into duplicate 12 mL polyethylene vials (2 mL headspace), and stored frozen until analysis using a LachatTM flow-injection system60. Three cores per treatment were then labelled at the beginning of the dark period with a 15 mmol−15N L−1 (99% 15N-KNO3-) stock solution to reach a final DIN concentration in cores of ~16.9 µmol L−1, enriching 95% of the NO3- pool and 89% of the DIN pool with the added 15N, to determine the changes to microbial reduction pathways through isotope pairing of N2. The three remaining cores in each treatment were labelled to the same enrichment at the beginning of the light period. Label additions were done separately at the beginning of dark and then at the beginning of light to isolate the isotope accumulation and transfers for each part of the diel cycle.

After ~10 h in the dark or light, labelled cores were sacrificed. Nitrogen stable isotope ratios (δ15N) of N2 were determined from sample water collected from sealed cores via syringe, without bubbles, in 12 mL glass exetainers preserved with 20 µL of saturated HgCl2 and stored with no headspace at room temperature. The additional three cores in the control tank without label addition were sacrificed in the same way and used as background isotope references. Sediments from background cores were used to determine the carbon and nitrogen content (%, δ13C and δ15N) of organic matter. Homogenised surface sediments were dried (60 °C) and acidified (carbon) or rinsed with KCl (2 M, nitrogen) to remove inorganic content settled or adsorbed on the sediment. Dry samples were then analysed on a Thermo Finnigan Flash EA 112 coupled via a Thermo Conflo III to a Thermo Delta V Plus IRMS to verify that the C:N of surface sediments (0-5 cm) and δ13C and δ15N values were comparable across the two incubations. A 2 mL He headspace was later added to the N2 samples ~24 h prior to analysis. N2 concentrations and isotope ratios were determined from the He headspace via isotope ratio mass spectrometry (IRMS), using a Thermo Trace Ultra gas chromatograph (GC) equipped with an autosampler61 coupled to a Delta V Plus IRMS via a Thermo Conflo III interface. All values are reported with respect to N2 in the air. Immediately before and after samples were run, N2 standards with known % 29N2 and 30N2 were used to correct for interference in the ion source caused by the generation of NO. The isotope pairing technique (IPT) of Nielsen (1992) was used to calculate actual N2 loss (D14) from the production of 29N2 and 30N2. This method also assumes no anammox is present, which was confirmed by a separate incubation (Supplementary Method 1), in which anammox was absent from all sediment depths (0 to 90 mm) (Supplementary Fig. 2). The absence of anammox meant that the original IPT62 could be used to calculate denitrification in the intact core incubations.

Denitrification (D14) was then separated into two components, denitrification from (1) NO3- diffusing into the sediments from the water column (Dw) and (2) NO3- produced within the sediments via nitrification (Dn)62, using the frequency of 14N and 15N substrate available. All final rates were presented as µmol-N m−2 d−1, assuming a 12:12 h light:dark cycle.

N2O fluxes

Cores were left open to acclimate in tank water under temperature and OA-amended conditions (with no substrate amendment) for ~48 h before being capped to begin incubations. Cores were capped prior to the simulation of daylight, with water samples collected immediately before lights were switched on, by gently syringe-sampling water (20 mL) from resealable ports fitted in each core lid. Water samples (6 mL) were then injected into duplicate 12 mL He flushed and sealed exetainers preserved with 20 µL HgCl2 without letting air escape, overturned, and allowed to equilibrate under pressure at room temperature (21 °C) for >48 h. Cores were left in the light for ~3 h before end-of-cycle sampling using the same method outlined for the start of the incubation. Cores were then uncapped and water columns were allowed to recirculate. In line with natural light cycling, the lights were turned off, and cores were resealed for dark sampling. Water collection for N2O samples in the dark was identical to that in the light. The time between the start and end of N2O incubations was ~3 h.

Dissolved N2O concentrations were determined from the equilibrated gas in the pressurised exetainers, using gas chromatography (GC, Schimadzu GC-9A) equipped with an electron capture detector63. A five-point standard curve was used to calibrate the GC with a precision typically >99%. The headspace concentrations (y in Eq. 1, ppm) were used to calculate the dissolved concentration (mol L−1) at equilibration temperature (21 °C) and subsequently adjusted to calculate the dissolved concentrations for the respective treatment temperatures (nmol L−1).

$${{{{{{\rm{N}}}}}}}_{2}{{{{{\rm{O}}}}}}=((({{{{{{\rm{y}}}}}}}-{{{{{\rm{Bg}}}}}})\times {10}^{-6})\times (({{{{{{\rm{P}}}}}}}\times {{{{{{\rm{\beta }}}}}}}_{{{{{{\rm{RT}}}}}}}\times {{{{{{{\rm{V}}}}}}}}_{{{{{{\rm{w}}}}}}})+({{{{{{{\rm{V}}}}}}}}_{{{{{{\rm{h}}}}}}})))/{{{{{{{\rm{V}}}}}}}}_{{{{{{\rm{w}}}}}}}\times {10}^{9}\times {{{{{{\rm{P}}}}}}}/({{{{{{\rm{R}}}}}}}\times {{{{{{\rm{T}}}}}}})$$
(1)

where βRT is the dimensionless Bunsen solubility coefficient (mL gas mL H2O−1 atm−1) calculated at room temperature for N2O64, P is the atmospheric pressure of the headspace in the exetainer, Vw, and Vh is the volume of liquid sample and headspace volumes, respectively. R is the constant for an ideal gas (0.082057 L atm K−1 mol−1), T is the treatment temperature (°K), and Bg is the background N2O concentration (mol L−1) measured in He flushed ‘blank’ exetainers.

Data analyses

Temperature sensitivity coefficients (Q10 values) were used to evaluate the response of dark and diel rates of Dw, Dn, and D14 to temperature increases of 10 °C. This was expressed simply as an exponential function:

$${{{Q}}}_{10}={\left(\frac{{{{R}}}_{2}}{{{{R}}}_{1}}\right)}^{{10}^{\circ}{{{{{\rm{C}}}}}}/\left({{{T}}}_{{{{{{\rm{opt}}}}}}}-{{{T}}}_{1}\right)}$$
(2)

where R1 and R2 are the denitrification rates (Dw, Dn, or D14) measured at temperatures 20 °C (T1) and optimal temperatures (Topt), respectively. Topt was defined as the incubation temperature for which denitrification rates were highest.

Statistical analyses

Statistical tests were conducted on denitrification and N2O flux rates using MATLAB65. Statistical significance was generally defined as alpha <0.05. All datasets were checked for homogeneity of variances (Levene’s test) before further analysis. If Levene’s test returned a significant result, data was log-transformed. If data could not be log-transformed, an alpha of 0.01 was used to limit type I errors.

Net variability with temperature and CO2

Net fluxes among treatments were compared to identify the effect of temperature and pCO2 condition. A one-way analysis of variance (ANOVA) was used to investigate the effect of high-pCO2 at each temperature. A two-way ANOVA then examined differences among temperatures (n = 4 or 2, for denitrification and N2O, respectively) at each level of pCO2 (n = 2) and identified any interacting effects. Finally, a separate one-way ANOVA was run for each level of pCO2 to investigate sediment responses across temperatures and to further investigate significant interactions identified in the two-way ANOVA. Post hoc Tukey’s tests were used to determine which scenarios had similar or different responses.

Diel variability of denitrification rates with temperature

To further investigate changes observed in net fluxes, differences between dark and light cycles were compared. First, a two-way analysis of variance (ANOVA) was used to examine differences among temperatures (n = 4), light conditions (n = 2), and whether light conditions significantly affected temperature response for cores in the current and high-pCO2 climates separately. Following this, a one-way ANOVA was run for each light condition to investigate any significant interaction found in the two-way ANOVA and to consider the individual temperature responses. If a significant effect was found in the one-way ANOVAs, post hoc Tukey’s tests were used to determine which treatments were significantly different from one another.

Treatment effects on light and dark N2O fluxes

Two-way ANOVAs were used to investigate potential differences in sediment N2O flux responses to warming (n = 2) and pCO2 (n = 2) in the light and dark, where separate one-way ANOVAs were run for each level of pCO2 to further investigate significant interactions identified in the two-way ANOVA.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.