Warming and ocean acidification may decrease estuarine dissolved organic carbon export to the ocean

Estuaries make a disproportionately large contribution of dissolved organic carbon (DOC) to the global carbon cycle, but it is unknown how this will change under a future climate. As such, the response of DOC fluxes from microbially dominated unvegetated sediments to individual and combined future climate stressors of warming (from Δ-3 °C to Δ+5 °C on ambient mean temperatures) and ocean acidification (OA, ~2 times the current partial pressure of CO2, pCO2) was investigated 10 ex situ. Warming alone increased sediment heterotrophy, resulting in a proportional increase in sediment DOC uptake, with sediments becoming net sinks of DOC (3.5 to 8.8 mmol-C m d) at warmer temperatures (Δ+3 °C and Δ+5 °C, respectively). This temperature response changed under OA conditions, with sediments becoming more autotrophic and a greater sink of DOC (1 to 4 times greater than under current-pCO2). This response was attributed to the stimulation of heterotrophic bacteria with the autochthonous production of labile organic matter by microphytobenthos. Extrapolating these results to the global 15 area of unvegetated subtidal estuarine sediments, the future climate of warming (Δ+3 °C) and OA may decrease the estuarine export of DOC by ~80 % (~150 Tg-C yr) and have a disproportionately large impact on the global DOC budget.


Introduction
The aquatic dissolved organic carbon (DOC) pool is one of the largest pools of organic carbon on earth (Hedges, 1987), roughly equivalent in size to the atmospheric CO 2 reservoir (Siegenthaler and Sarmiento, 1993). The role of DOC in long-term carbon storage in the ocean has been a focus of research for decades (Siegenthaler and Sarmiento, 1993;Hansell et al., 2009;Bauer and Bianchi, 2011;Wagner et al., 2020), with DOC reaching the ocean interior being effectively stored for millennia (Hansell et al., 2009). Although phytoplankton in the surface ocean are the main source of DOC globally, with an estimated production of around 50 Pg C yr −1 , only 0.3 % of the DOC they produce reaches the ocean interior (Hansell et al., 2009). Most of the DOC produced by phytoplankton is rapidly remineralized in the water column by heterotrophic bacteria (Azam, 1998). Only more recently has the coastal zone been considered a major source of DOC export to the open ocean and deep sea (Duarte et al., 2005;Maher and Eyre, 2010;Krause-Jensen and Duarte, 2016). The shallow coastal zone accounts for 1 to 10 % of global net primary production (NPP) (Duarte and Cebrián, 1996), with up to 33 % of the associated DOC exported offshore and stored in the ocean interior (Krause-Jensen and Duarte, 2016).
Although shallow estuaries and fringing wetlands make up only ∼ 22 % of the world's coastal area (Costanza et al., 1997) and 8.5 % of the total marine area (Costanza et al., 1997), they are quantitatively significant in terms of DOC processing and offshore transport (Smith and Hollibaugh, 1993). In 1998, Bauer and Druffel used radioisotopic carbon ( 14 C) to identify the source and age of DOC and particulate organic carbon inputs into the open ocean interior. They found that ocean margins accounted for greater organic carbon inputs into the ocean interior than the surface ocean by more than an order of magnitude. Assuming one-third of the DOC produced in the coastal zone (100-1900 Tg C yr −1 ; M. N. Simone et al.: Warming and OA may decrease DOC export rior (Krause-Jensen and Duarte, 2016), 30 to 630 Tg C yr −1 , or up to 3.5× more DOC, could reach the ocean interior from coastal areas than from the open ocean (180 Tg C yr −1 , Hansell et al., 2009). This is despite coastal areas having a DOC production rate only 0.2 to 3.9 % of that of the open ocean (Duarte, 2017). As such, small changes to the coastal production and export of DOC may have a disproportionate influence on the global DOC budget.
Euphotic estuarine sediments occupy the coastal boundary between terrestrial and marine ecosystems. Microalgal communities (microphytobenthos, or MPBs) are ubiquitous in these sediments, occupying ∼ 40 to 48 % of the coastal surface area (Gattuso et al., 2020) and generating up to 50 % of total estuarine primary productivity (Heip et al., 1995;Mac-Intyre et al., 1996;Underwood and Kromkamp, 1999). MPBs exude some of the carbon they fix as extracellular substances, including carbohydrates (Oakes et al., 2010), and can therefore be a source of relatively labile DOC in net-autotrophic sediments (Cook et al., 2004;Oakes and Eyre, 2014;Maher and Eyre, 2010). The dominant sink of DOC in estuarine sediments, however, is uptake by heterotrophic bacteria (Azam, 1998). These heterotrophic bacteria not only consume autochthonous DOC from upstream (Boto et al., 1989), but their biomass is influenced by the lability of sediment organic matter (OM) (Hardison et al., 2013), which can be altered by MPB production (Hardison et al., 2013;Cook et al., 2007). Estuarine sediments are therefore a potentially important sink for DOC.
Unvegetated estuarine sediments can affect the quantity and quality of DOC input to the ocean by (1) acting as a source of autochthonous DOC, through MPB production (Duarte, 2017;Krause-Jensen and Duarte, 2016;Maher and Eyre, 2010), or (2) modifying allochthonous and terrigenous DOC inputs (Fichot and Benner, 2014). Through efficient mineralization of DOC (Opsahl and Benner, 1997), estuaries can act as a sink for DOC and a source of CO 2 to the ocean (Frankignoulle et al., 1998;Fichot and Benner, 2014;Sandberg et al., 2004). Given the disproportionate contribution of estuaries to the export of DOC to offshore marine ecosystems, relative to their surface area, it is important to understand how this balance of DOC sources and sinks within estuaries may change with future shifts in climate, particularly expected increases in temperature and ocean acidification (OA) associated with elevated atmospheric CO 2 concentrations.
Climate projection models assuming a high-emission scenario suggest that atmospheric CO 2 concentrations could more than double by the end of the century, increasing the partial pressure of CO 2 (pCO 2 ) in surface waters to 1000 µatm and decreasing pH by 0.3 units, together termed OA (RCP8.5; IPCC, 2019). There is also expected to be an increase in mean surface ocean temperature by 2-4 • C (RCP8.5; IPCC, 2019) and increased frequency of unseasonably warm days (Morak et al., 2013;Fischer and Knutti, 2015).
Primary producers fix dissolved inorganic carbon (DIC) during photosynthesis and release DOC directly through exudation and/or indirectly when they are grazed upon. Photosynthetically produced DOC is the main source of DOC in the ocean (Hansell et al., 2009) and fuels local microbial mineralization (Azam, 1998). Heterotrophic bacteria within estuarine sediments respire the carbon from DOC as CO 2 , which can then be recaptured by photoautotrophs (Riekenberg et al., 2018), closing the microbial loop (Azam, 1998). DOC or DIC that is not captured is ultimately effluxed to the overlying water column and may be transported from estuaries to the coastal ocean. Individually, increased temperature and CO 2 have been reported to enhance primary productivity and DOC production in Arctic (Czerny et al., 2013) and temperate phytoplankton communities (Wohlers et al., 2009;Engel et al., 2011;Liu et al., 2017;Novak et al., 2018;Taucher et al., 2012), and in temperate stream sediments (Duan and Kaushal, 2013). However, one study in a temperate fjord reported no enhancement of DOC production despite CO 2 -enhanced phytoplankton productivity . This uncertainty of the response to individual climate stressors is exacerbated when considering how the combination of OA and warming will affect DOC processing. To date, only one study has considered this combined stressor effect on DOC (Sett et al., 2018), observing no difference in DOC production by temperate phytoplankton relative to current conditions under the combined stressors (Sett et al., 2018).
To understand the potential effect of future climate on DOC fluxes, it is essential that both individual and combined effects of OA and warming are considered. Here we focus on changes in DOC fluxes in unvegetated estuarine sediments, as these systems have the potential for significant uptake of DOC that is currently exported to the coastal ocean. In this study, benthic DOC responses in unvegetated estuarine sediments were investigated over an 8 • C temperature range under both current and projected future high-pCO 2 conditions in an ex situ laboratory incubation.
We expected that warming would promote a stronger heterotrophic than autotrophic microbial response in shallow euphotic sediments (Patching and Rose, 1970;Vázquez-Domínguez et al., 2012;Yang et al., 2016) and, as such, there would likely be more DOC remineralization (Lønborg et al., 2018) than "new" DOC production by photoautotrophs (Wohlers et al., 2009;Engel et al., 2011;Novak et al., 2018). Moreover, despite the potential stimulation of primary productivity in unvegetated muddy sediments by OA (Vopel et al., 2018), or more likely high pCO 2 , and potential enhancement of DOC production Liu et al., 2017), this increase in labile DOC may promote bacterial productivity and DOC mineralization (Hardison et al., 2013). In addition, increased DOC availability alone may increase heterotrophic bacterial biomass production and activity . We therefore predicted that increases in DOC production from OA alone or in combination with warming may be counteracted by increased consumer activity, potentially diminishing the available DOC pool under future climate conditions.

Core collection
Sediment (∼ 20 cm depth) was collected and capped in acrylic cores (9 cm diameter × 47 cm length) allowing for ∼ 1.8 L of overlying water on 9 and 16 January 2018 (15 and 12 cores, respectively). Thalassinidean shrimp, Trypaea australiensis, burrows were avoided and therefore excluded from the collected cores as their occasional inclusion would result in considerable variability in sediment processes (Webb and Eyre, 2004) that would mask potential treatment effects. To ensure sediments were subtidal, cores were collected during low tide. Immediately after core collection, ∼ 700 L of site water was also collected to fill a laboratory incubation setup.

Incubation setup
Within 6 h of core collection, all cores were in the laboratory, submerged and uncapped in site water. The cores were placed in one of four temperature tanks -control (23 • C), − 3 • C (20 • C), + 3 • C (25 • C), and + 5 • C (28 • C) -filled with ∼ 80 L of site water, with temperatures maintained and monitored via thermo-regulating aquarium pumps. Each tank had three cores (n = 3).
The ex situ study design allowed control of temperature, pCO 2 , and light that would be difficult to achieve in situ. Due to limited space, this investigation was conducted over 2 weeks with two complementary incubations repeated back to back. The incubation in the first week (9-12 January 2018) had cores in the four temperature tanks subjected to high pCO 2 (∼ 1000 µatm), achieved with a CO 2 -enriched airstream (initially adjusted and set when attached to a LICOR LI-7000) bubbled into tank water via airstones and air pumps. The incubation in the second week (16-20 January 2018) maintained current pCO 2 conditions (∼ 450 µatm) by circulating ambient laboratory air through the tank water via airstones and air pumps. An additional tank was included in week 1 alongside the high-pCO 2 incubation. This tank was a control tank equivalent to the control tank present week 2, allowing for comparison of the two separate incubations (see Table 1 for details). The temperature and pCO 2 manipulations were within 12 and 4 %, respectively, of their in situ ranges (see Sect. 2.1) to reduce any potential shock effect for the sediment community.
Water columns within cores were stirred at ∼ 60 rpm throughout the incubations via magnetic stir bars (∼ 5 cm above sediment surface) interacting with an external rotating magnet, ensuring water columns were well mixed whilst avoiding sediment disturbance (Ferguson et al., 2003(Ferguson et al., , 2004. High-pressure sodium lamps (400 W; Philips Son-T Argo 400) were used to simulate mean daytime field conditions, providing ∼ 270-280 µmol quanta m −2 s −1 of photosynthetically active radiation at the water surface of the tanks. Lamps were turned on in the mornings in line with natural diel light cycling, following a similar in situ ∼ 12 h dark-12 h light cycle. Cores were pre-incubated at treatment conditions for 36-48 h, before solute-flux incubations. This pre-incubation period would be sufficient for three to six generations of the dominant microbial members of unvegetated estuarine sediments, diatoms, and cyanobacteria (Mori et al., 1996;Greene et al., 1992), allowing time for the microbial community to acclimatize to the new treatment conditions.

Solute-flux incubation
Immediately after pre-incubation, cores were capped for a 20 h (10 h dark, 10 h light) solute-flux incubation to measure rates of O 2 , DIC, and DOC production and consumption over a diel cycle. To adhere to natural diel cycling, cores were capped at dusk to start the incubation on a dark cycle. Samples were collected from three cores per tank at each of three time points in the diel cycle (dark start (dusk), dark end and light start (dawn), and light end (dusk)). Water was collected and syringe-filtered to determine concentrations of DIC (0.45 µm Minisart filter, 100 mL serum bottle; without headspace, poisoned with 50 µL of saturated HgCl 2 , stored at room temperature) and DOC (GF/F filter, 40 mL glass vial with silicon septum; without headspace, poisoned with 20 µL of HgCl 2 , injected with 200 µL of 85 % H 3 PO 4 , stored at room temperature). As water was removed for sampling, it was replaced with gravity-fed water maintained in a collapsible bag under the same atmospheric conditions and temperature. After all cores were sampled, dissolved oxygen (DO)  Table 1. Start conditions for incubations at current and high pCO 2 levels ( * ) showing mean (±SD) of various carbonate parameters. CON * is the overlapping control core present in the high-pCO 2 incubation week.

Scenario
Sal concentrations, temperature, and pH were measured using a high-precision Hach HQ40d multiprobe meter with an LDO probe and pH probe, calibrated to 3-point NIST buffer scale (R 2 = 0.99). Probes were inserted into a resealable port fitted in each lid, ensuring no incubation water exchanged with tank water. After the dawn sampling, lamps were switched on.
DIC concentrations were determined with an AIRICA system (MARIANDA, Kiel, Germany) via infrared absorption using a LICOR LI-7000 and corrected for accuracy against certified reference material, batch no. 171 (Dickson, 2010). Measurements on four analytical replicates of 1.5 mL sample volume were used to calculate DIC concentration as the mean of the last three out of four measurements (typical overall uncertainty, < 1.5 µmol kg −1 ). DIC and pH measurements were then used to calculate the remaining carbonate chemistry parameters (Table 1) using CO 2 SYS (Pierrot et al., 2006). Total borate concentrations (Uppström, 1974) and boric acid (Dickson, 1990) and stoichiometric equilibrium constants for carbonic acid from Mehrbach et al. (1973), as refit by Dickson and Millero (1987), were used. Comparison of pH (free scale) measured with a Hach HQ40d multiprobe meter and pH calculated from measured total alkalinity and DIC (Table S1) indicated an uncertainty for potentiometric pH measurements of ±0.05 pH units. Propagating the uncertainty in pH measurements with the uncertainty of DIC measurements translates to a pCO 2 uncertainty of ± ∼ 110 and ∼ 56 µatm under high-pCO 2 and current conditions, respectively. This uncertainty is well within the treatment variability measured among cores (Table 1) and is therefore considered unlikely to have contributed substantially to differences in treatment response. DOC concentrations were measured via continuous-flow wet oxidation using an Aurora 1030 W total organic carbon analyser (Oakes et al., 2011) (uncertainty of ∼ 3 %).

Data analysis
The dissolved oxygen and DIC measurements were used to estimate benthic microalgal production inside the cores. Net primary production and respiration (NPP and R, µmol O 2 m −2 h −1 ) were defined as the light and dark cycle oxygen evolution, respectively, where DIC and DOC light and dark fluxes (µmol C m −2 h −1 ) were defined using the evolution of DIC and DOC concentrations, respectively. Fluxes (NPP, R, DIC, or DOC) were calculated as where end and start are the dissolved oxygen, inorganic carbon, or organic carbon concentrations (µmol O 2 L −1 or µmol C L −1 ) at the end and start of the light or dark cycle, V is the water column volume (L), T is hours of incubation, and A is surface area of the core.
Gross primary productivity (GPP, µmol O 2 m −2 h −1 ) was calculated using NPP and R, as follows: (2) The production-to-respiration ratio (P / R) was calculated using GPP and R scaled for a 12 h light-12 h dark diel cycle (Eyre et al., 2011).
Finally, net fluxes for DIC and DOC were calculated from the dark and light fluxes from Eq.
(1) and presented as mmol C m −2 d −1 for a 12 h light-12 h dark diel cycle.
Temperature sensitivity coefficients (Q 10 values) were used to evaluate the temperature dependence of metabolic rates on temperature increases of 10 • C. This was expressed simply as an exponential function: where R 1 and R 2 are the R, NPP, or GPP rates measured at temperatures of 20 • C (T 1 ) and optimal temperatures (T opt ), where rates are highest, respectively.

Scaling rates
Rates in the overlapping control cores each week were checked to ensure comparability between incubations. If means (±SD) were significantly different (did not overlap), rates from individual treatment cores were scaled to the overall mean control rate of both weeks (n = 6). This was done by calculating the relative proportion of treatment rates (tProp., µmol N m −2 h −1 ) to the control rates present in its week (Eq. 6), where tRate is the individual core rate, and control is the mean rate of control cores present during the incubation (n = 3). This proportional rate was then multiplied by the overall control mean rates (averaged across both weeks, n = 6) to scale individual core rates and calculate comparable treatment means (n = 3) across incubations (see Sect. 3.1 for details on scaled rates).

Statistical analysis
Homogeneity of variances (Levene's test) was tested before analysis to minimize the potential for type I error. All tests were run in MATLAB (Mathworks, 2011) with significance defined at a maximum alpha of < 0.05. Where Levene's test returned a significant result, either datasets were log-transformed or else, if negative values were present, an alpha of 0.01 was used for the subsequent analyses of variance (ANOVAs).

Net variability with temperature and CO 2
Net fluxes were compared among treatments to identify the individual and combined effects of temperature and pCO 2 on O 2 , DIC, and DOC fluxes. To investigate the effect of increased pCO 2 alone, data from control temperature cores at both current and high pCO 2 levels (n = 2) were compared using a one-way ANOVA. A two-way ANOVA on each dataset identified whether there were interacting effects on O 2 , DIC, and DOC fluxes of temperature (n = 4) and pCO 2 (n = 2). Finally, one-way ANOVAs were also run for each pCO 2 level to investigate differences in sediment responses across temperatures (n = 4). Post hoc Tukey's tests were then used to determine which temperatures had similar or different responses.

Diel variability with temperature for DIC and DOC fluxes
Differences between dark and light cycles were compared to further investigate changes observed in DIC and DOC net variability. Similar analyses to those described above were applied. To examine differences among temperatures (n = 4), light condition (n = 2), and whether light condition significantly interacted with temperature response, two-way ANOVAs were applied to current-and high-pCO 2 cores, separately. Following this, each light condition was further investigated to consider the individual temperature responses in the light and dark separately using one-way ANOVAs and post hoc Tukey's tests.

Overlapping control scaling
Mean rates calculated from overlapping control cores present in each week were compared to establish whether the two sets of incubations were directly comparable, and whether changes attributed to high pCO 2 levels were truly due to that treatment, not just a temporal shift in how the sediments were behaving. The P / R ratios were similar for incubations (0.84 ± 0.01 and 0.83 ± 0.04, Table S2); however, the magnitude of the R and NPP fluxes was ∼ 23 % greater for control cores in the high-pCO 2 week (Table S2; discussed in Sect. 4.0). As such, R and NPP rates of cores were scaled to mean control rates (n = 6) using the proportional rate difference calculated between the treatments and the individual controls present in the respective weeks (n = 3) (Eq. 6). Scaled rates were within ±13 % of actual rates. There were no significant differences between controls for light or dark production of DIC or DOC.

Productivity and respiration responses to temperature and OA
Temperature had a strong effect on R, NPP, GPP, and P / R, whereas only light cycle NPP and, in turn, GPP and P / R were affected by OA. The response of R to temperature was similar at both current and high pCO 2 levels (no two-way interaction: F 3,16 = 0.77, p = 0.53; Fig. 2a), and it was not affected by pCO 2 (CO 2 effect, two-way interaction: F 1,16 = 0.99, p = 0.34; Fig. 2a). Accordingly, Q 10 values for R were similar for current (1.66) and high-pCO 2 (1.69) conditions (Table 2). R changed significantly across the 8 • C temperature range, increasing by ∼ 11 and ∼ 29 % in higher-temperature cores ( + 3 and +5 • C, respectively) and decreasing by ∼ 16 % in −3 • C cores (temperature effect, two-way interaction: F 3,16 = 36.93, p < 0.0001; Fig. 2a).
The differences in P / R among treatments further highlighted significant interaction of temperature and pCO 2 (two-way interaction: F 3,16 = 5.86, p = 0.007; Fig. 2d), suggesting the effect of pCO 2 on primary productivity was strong enough to alter the overall productivity of the sediments. Under current pCO 2 conditions, P / R reflected GPP with a clear separation between control and −3 • C sediments having a higher P / R (0.84 ± 0.01 and 0.89 ± 0.07, respectively) than the significantly lower ratios calculated in increased-temperature cores (0.42 ± 0.11 and 0.33 ± 0.05 for +3 and +5 • C, respectively) (one-way interaction: F 3,8 = 49.41, p < 0.0001; Fig. 2d). Similarly, under high-pCO 2 conditions, the effect of GPP on P / R was clear. The positive effect of high-pCO 2 conditions on GPP pushed the P / R ratio of −3 • C and control temperature cores to ∼ 1 (1.09 ± 0.16 and 1.03 ± 0.03, respectively), suggesting the ecosystem shifted toward net autotrophy under those conditions. The positive effect of high-pCO 2 conditions was also highlighted at +3 • C, with P / R (0.77 ± 0.13) remaining close to the current ecosystem ratio (0.84 ± 0.01) instead of significantly dropping like those calculated under current pCO 2 conditions or in +5 • C cores (0.25 ± 0.04, one-way interaction: F 3,8 = 38.58, p<0.0001; Fig. 2d).

DOC fluxes
At current pCO 2 levels, increasing temperature resulted in a significant shift in net DOC fluxes, going from effluxes at the two lower temperatures ( −3 • C and control) to uptakes at the two higher temperatures (one-way interaction: F 3,8 = 6.96, p = 0.013; Fig. 5a). The relative light-and-dark cycle contributions of these net trends at current pCO 2 levels were also affected by temperature (two-way interaction: F 3,16 = 13.18, p = 0.0001; Fig. 5b). Significant changes in DOC fluxes in the dark shifted from an efflux at −3 • C to an uptake at control temperature, with higher uptake rates at +5 • C (26 % higher than control rates; one-way dark: F 3,8 = 8.64, p = 0.007; Fig. 5b). In contrast, the highest DOC effluxes in the light were at control temperatures, significantly decreasing with both increasing and decreasing Middle horizontal line in each box is the exclusive median, with the start of the upper and lower quartiles represented by the top and bottom edges of the box, respectively. Letters identify significantly different means across temperatures under current pCO 2 conditions, and numerals identify significantly different means across temperatures under high-pCO 2 conditions, 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. Solid and dashed horizontal lines identify significant effects of OA and temperature, respectively, where double solid and dashed lines identify significant interaction of temperature and OA (two-way ANOVA). Levels of significance are denoted with " * " for 0.05, " * * " for 0.01, and " * * * " for 0.001. Data in Table S5. temperatures to DOC fluxes around zero (one-way interaction: F 3,8 = 16.76, p = 0.001; Fig. 5b).
High-pCO 2 conditions alone (at ambient mean temperatures, 23 • C) had a significant effect on net DOC, shifting from a slight efflux at current pCO 2 levels (∼ 0.5 mmol C m −2 d −1 ) to a significant uptake at high pCO 2 levels (∼ 10.9 mmol C m −2 d −1 ; one-way interaction: F 3,4 = 25.1, p = 0.007; Fig. 5a). The trend in temperature response was similar for current and high-pCO 2 conditions (two-way interaction: F 3,16 = 0.88, p = 0.47; Fig. 5a), but there was a significant shift from small efflux at lower temperatures to considerable uptakes at all temperatures under high-pCO 2 conditions (two-way CO 2 effect: F 1,16 = 61.46, p < 0.0001; Fig. 5a). Differences between dark and light DOC fluxes under high-pCO 2 conditions were independent of temperature (two-way interaction: F 3,16 = 1.94, p = 0.16; Fig. 5c), with the overall magnitude of influxes in the dark being significantly greater than those in the light (two-way light condition: F 1,16 = 15.83, p = 0.001; Fig. 5c). Loss of statistically different temperature responses for high-pCO 2 light and dark responses (temperature effect, two-way interaction: F 3,16 = 1.05, p = 0.40; Fig. 5c) was in large part due to within-treatment variability in the high-pCO 2 cores.

Discussion
An important component of this study was testing the interaction and individual effects of warming and OA on DOC processing. This was necessarily achieved through a comparison of core incubations occurring in different weeks. As such, it is important to consider the limitations of this approach. Control treatments in different weeks would ideally and high-pCO 2 conditions (hatched boxes). Light (grey boxes) and dark fluxes (spotted boxes) of DOC (µmol C m −2 h −1 ) for (b) current pCO 2 and (c) high-pCO 2 conditions. In panel (a) letters identify significantly different means across temperatures under current pCO 2 conditions, and numerals identify significantly different means across temperatures under high-pCO 2 conditions; in panels (b) and (c) letters identify significantly different means across temperatures in dark cycles, and numerals identify significantly different means across temperatures in light cycles. Letters or numerals that are the same indicate no significant difference, as determined by a one-way ANOVA and post hoc Tukey's test. Solid and dashed horizontal lines identify significant effects of pCO 2 or light and temperature, respectively, where double solid and dashed lines identify significant interaction of temperature and light (two-way ANOVA). Levels of significance are denoted with " * " for 0.05, " * * " for 0.01, and " * * * " for 0.001. Data in Table S6. be the same in all respects, but there were some differences. For instance, NPP and R were higher in the incubation week for current pCO 2 conditions (Table S2), likely due to small changes in environmental conditions, such as salinity differences (24 versus 17.7 for current and high-pCO 2 conditions, respectively; Table 1). However, these differences did not significantly affect DOC fluxes, nor the heterotrophy of the sediments (P / R = 0.84 ± 0.01 and 0.83 ± 0.04; Table S2). Moreover, sediments in separate weeks maintained the same OM content (∼ 3.5 %) and molar C : N ratio (∼ 16), suggesting that differences in processing have very little short-term impact on the overall OM pool in the sediment due to the OM pool size being about 3 orders of magnitude higher than any diel flux (organic carbon pool ∼ 12 000 mM). Thus, because all conditions in the laboratory setup were the same for each incubation (with the exception of pCO 2 in treatment tanks, which was intentionally manipulated to be different), the difference in fluxes between controls was attributed to differences in when the sediments and overlying waters were collected. Therefore, the scaling of NPP and R (Table S3) was done for the sake of treatment comparison, resulting in scaled rates within 13 % of actual measured values, which had a negligible effect on P / R (< 1 % across all treatments). The final NPP and R rates in comparisons across treatments should thus be considered relative to control rates and be interpreted as approximate values (±13 %).
Understanding current ecosystem functioning is of primary interest when trying to determine how disturbances in the environment may change metabolic rates and pathways of OM mineralization (Jørgensen, 1996;D'Avanzo et al., 1996;Malone and Conley, 1996). Based on unadjusted R rates, the nearly 1 : 1 ratio of DIC production to O 2 consumption in the dark (respiratory quotient of ∼ 1.13 ± 0.05; Fig. 4) suggests that aerobic respiration dominated the sediments (Eyre and Ferguson, 2002). Similarly, unadjusted NPP rates suggest that aerobic processes dominated benthic production in the light, with a 1 : 1 ratio of O 2 and DIC fluxes (Fig. 4;Eyre and Ferguson, 2002). Sediments in the current study were net heterotrophic, with a P / R in control cores of ∼ 0.84 ± 0.01 and ∼ 0.83 ± 0.04 during current-and high-pCO 2 incubation weeks, respectively. Despite the undeniable range of P / R ratios that unvegetated estuarine sediments may experience (1.2 (Oakes et al., 2012) to 0.01 (Ferguson and Eyre, 2013)), the ratios in the current study were similar to mean global estimates for unvegetated estuarine sediments (∼ 0.82, calculated from values in Duarte et al., 2005) and calculated from P and R values of 22 estuaries globally (∼ 0.87, compiled by Smith and Hollibaugh, 1993), suggesting that the metabolic function of sediments in the current study are representative of estuarine sediments globally and the impacts observed in this study should be broadly applicable.

DOC fuels benthic respiration
DOC appeared to be a significant driver of benthic respiration (Fig. 5b). At control temperatures (23 • C) net DOC fluxes were near zero (0.47 ± 0.93 mmol C m −2 d −1 ), indicating that the diel production and uptake of DOC across the sediment-water interface were balanced (Fig. 5a). The control rates in the present study were close to benthic DOC flux rates reported for subtropical estuarine sediments in most seasons, ∼ 1.5 mmol C m −2 d −1 , except summer (Maher and Eyre, 2010). Relative to our control (summer) rates, Maher and Eyre (2010) reported higher net DOC flux rates (∼ 10 mmol C m −2 d −1 ) as a result of DOC effluxes in both the light and dark (Maher and Eyre, 2010). We observed similar light DOC effluxes (610 µmol C m −2 h −1 ) to summer sediments in Maher and Eyre (2010) (∼ 647 µmol C m −2 h −1 ), whereas dark DOC uptake was observed in the current study (−571 µmol C m −2 h −1 ), and Maher and Eyre (2010) reported dark DOC effluxes (254 µmol C m −2 h −1 ). This release of DOC in the dark was attributed to enhanced microbial coupling in the sediments in warmer temperatures (Maher and Eyre, 2010). In the current study, and in previous reports, DOC uptake suggests that bacteria not only intercepted DOC produced from within the pore waters (potentially satisfying up to 60 % of total mean bacterial production; Boto et al., 1989) but also took up available DOC from the water column to satisfy its metabolic requirements (Boto et al., 1989;Brailsford et al., 2019), effectively acting as a DOC sink. Under conditions of reduced light availability and/or intensity, sediments are expected to have an amplified heterotrophic response in addition to a reduction in microalgal production of DOC.

OA increases DOC assimilation
Positive responses in primary production were associated with OA. The ∼ 72 % increase in NPP rates at ambient temperatures was consistent with general stimulation of primary production in finer sediments with increased DIC availability (Vopel et al., 2018;Oakes and Eyre, 2014). Sediments may become DIC-limited when algal demand is relatively high compared to porewater supply of CO 2 (Cook and Røy, 2006), and MPBs therefore may benefit from an increase in CO 2 availability. MPBs in fine sediments are restricted to dissolved substrates (i.e. nutrients and DIC) accessed via diffusion from deeper and adjacent sediments, and the overlying water column (Boudreau and Jørgensen, 2001). This makes them more likely to deplete accessible DIC than MPBs in permeable sediments. Primary producers in permeable sediments, like those in reef ecosystems, therefore do not often experience the same increase in primary production with increased CO 2 (Trnovsky et al., 2016;Cyronak and Eyre, 2016;Eyre et al., 2018;Cook and Røy, 2006;Vopel et al., 2018). As well as differences in diffusive versus advective modes of solute transfer between the sediment types (Cook and Røy, 2006), variable response may also be attributable to sandier sediments being limited by other factors such as nutrient and OM availability, given that coarser sediments are generally more oligotrophic (Admiraal, 1984;Heip et al., 1995). In comparison, nutrients were non-limiting in the less permeable sediments used in the current study, based on nutrient concentrations that increased during all incubations (see the Supplement and Table S7). MPB growth rates in sediments with low permeability are more likely limited by DIC availability. Primary productivity responses to pCO 2 would likely differ in permeable sediments where general access to CO 2 is greater.
Given that MPBs exude carbon (Maher and Eyre, 2010), we would expect increased GPP to correspond with increased DOC production and flux. However, although OA stimulated primary production (Fig. 2), we instead saw increased DOC uptake in the dark (Fig. 5). A likely explanation is that bacterial uptake of DOC was stimulated through the provision of labile carbon from MPBs (Morán et al., 2011;Hardison et al., 2013). As such, DOC appeared to fuel much of the dark cycle respiration, as DOC uptake in the dark reflected dark DIC production (respiration), except for sediments at −3 • C under current pCO 2 conditions. Under current pCO 2 conditions, uptake of DOC in the dark accounted for only ∼ 50 % of the DIC respired in the dark. The portion of DIC accounted for by dark DOC uptake increased from 50 to 100 % under the high-pCO 2 conditions. In part, this may have been due to increased availability of labile organic carbon (Moran and Hodson, 1990) arising from the increase in NPP under high-pCO 2 conditions across all temperatures (Fig. 2b), which would reduce the need for bacteria to synthesize ectoenzymes (Chróst, 1991(Chróst, , 1992, resulting in more rapid turnover of carbon to the water column.

Warming drives increased heterotrophy and DOC assimilation
Sediments in this study, like other manipulative studies in both permeable sands (Lantz et al., 2017;Trnovsky et al., 2016) and cohesive sediments (Apple et al., 2006), demonstrated increased heterotrophy with increased temperature. This shift to heterotrophy is often attributed to the imbalance in the thermal sensitivity of heterotrophic over autotrophic metabolism (Yang et al., 2016;Allen et al., 2005). More specifically, differences in activation energy dictated by differences in physiology and biochemical processes (Patching and Rose, 1970;Apple et al., 2006) result in increases in heterotrophic activity with increasing temperature than exceed increases in autotrophic activity (Yang et al., 2016). However, in this study, under current pCO 2 conditions, the increases in R and GPP from −3 • C to control temperatures were similar (∼ 16 and ∼ 11 %, respectively), whereas at higher temperatures GPP decreases far exceeded increases in R (7× and 3× for 26 and 28 • C, respectively). There-fore, unlike previous studies, decreases in MPB productivity at higher temperatures appeared to be a greater driver towards heterotrophy than increases in respiration rates. In other words, temperature increases not only increased the rate of DOC uptake but also likely decreased the rate of DOC production.

Warming reduces GPP and DOC production under current pCO 2 conditions
Primary production is the main source of DOC in marine ecosystems (Wagner et al., 2020). Decreasing trends in GPP with warming under current pCO 2 conditions seen here have been described previously where photosynthetic growth and production decline at higher temperatures (Thomas et al., 2012). Photosynthetic productivity is often linked to seasonal temperature (Apple et al., 2006), which is also associated with differences in environmental factors such as light, nutrient concentrations, and dissolved organic matter (DOM) quality and availability (Geider, 1987;Herrig and Falkowski, 1989). Although the relative availability of light and nutrients does influence productivity rates (Kana et al., 1997) and would be expected to influence in situ seasonal production, the current study controlled light and initial nutrient concentrations in the water column to isolate the effect of temperature. Thus, decreasing GPP was driven by warming, suggesting that MPBs in these subtropical sediments likely had a temperature optimum around current mean summer temperatures of ∼ 23 • C (GPP: 1515 ± 37 µmol O 2 m −2 h −1 ; Fig. 2c). Longer-term warming could allow for possible migration of more tolerant species to settle from lower latitudes (Hallett et al., 2018), shifting the composition of the benthic community. The introduction of more tolerant species could reduce the increase in heterotrophy and net DOC removal from the water column seen here. However, the species diversity of the estuarine sediments will ultimately decrease as they are pushed to temperature extremes (Thomas et al., 2012), reducing the functional redundancy of the microbial community. This decreased functional redundancy has the potential to make unvegetated estuarine sediments less resilient to environmental perturbations under future climate conditions.

Warming increases respiration and DOC assimilation
Unlike photosynthetic productivity, heterotrophic respiration often has a linear rate increase with temperature to the thermal optimum due to heterotrophs not being constrained by the same abiotic variables (e.g. nutrient and light availability) as primary producers (Apple et al., 2006(Apple et al., , 2008Geider, 1987;Yap et al., 1994). In this study, respiration rates under both current and high pCO 2 conditions increased from the lowest rates measured at −3 • C to maximum rates (> 50 % greater) at + 5 • C (Fig. 2a). Consistent with overall lower respiration rates relative to other subtropical unvegetated sediments (∼ 900 to ∼ 1500 µmol O 2 m −2 h −1 ; Ferguson and Eyre, 2013), the temperature dependence of respiration under both current and high-pCO 2 conditions (Q 10 = 1.66 and 1.69, respectively) was slightly lower than is typical for biological systems (Q 10 = 2; Valiela, 1995) but similar to temperature dependence described in other estuarine systems (Q 10 = 1.5-1.9; Morán et al., 2011), with values towards the lower end of this range possibly being a result of resource limitation (López-Urrutia and Morán, 2007). A potential limiting resource for bacteria in estuarine sediments is DOM (Church, 2008), ultimately controlling the flow of carbon through the microbial loop (Kirchman and Rich, 1997). However, in the presence of sufficient DOM, warming has been associated with increased bacterial DOM incorporation (Kirchman and Rich, 1997). In line with this, increased uptake of DOC at higher temperatures and efflux at lower temperatures was observed. Although DOC is mainly produced by photoautotrophs, DOC can be produced in the dark through, for example, chemodegradation of detrital organic carbon and cell lysis by viruses and during grazing (Carlson, 2002). As such, the efflux of DOC in the dark at −3 • C suggests that heterotrophic bacterial productivity, and therefore DOC uptake, was reduced by lowered temperatures (Raymond and Bauer, 2000), resulting in a failure to intercept all DOC produced in the pore waters. This failure to intercept DOC may be compounded if nutrient supply is limited (Brailsford et al., 2019), as it is common for heterotrophic bacteria to rely on refractory DOC when labile sources are not readily available (Chróst, 1991), which can occur under conditions of nutrient-limited biological productivity (Allen, 1978).

Global estuarine loss of DOC from unvegetated sediments in the future
Up to 3.5× more DOC reaches the ocean interior from coastal areas than the open ocean (Duarte, 2017;Krause-Jensen and Duarte, 2016;Hansell et al., 2009). As such, small changes to the coastal export of DOC may have a disproportionately large influence on the global DOC budget. Our findings suggest a reduced export of DOC to the ocean from the coastal zone under high-pCO 2 conditions, across the full 8 • C temperature range due to changes in carbon processing within unvegetated sediments. Despite the lack of seasonality in the study, the inclusion of an 8 • C temperature range, including temperatures below current mean temperatures, suggests that seasonal temperature variation is unlikely to have a significant effect on the relative change in DOC in the future (Fig. 5). Although any upscaling of a single controlled experiment to a global scale is highly speculative, we believe it is valuable to demonstrate the potential for a high-pCO 2 climate to cause globally significant change in DOC export from coastal zones. Furthermore, putting our findings in a global context provides a guideline value for potential change. The following estimates should be considered in this context, and it should be expected that different hydrodynamic settings, sediment and/or sediment community composition, and sources of organic matter could affect the outcome. For example, the response to warming and pCO 2 may be different for pelagic communities and/or in deeper waters that are subject to stratification (Li et al., 2020), where access to nutrients and CO 2 may become limiting (Rost et al., 2008). We have applied our results to global coastal DOC exports (Maher and Eyre, 2010;Duarte, 2017) as an initial step in estimating responses of unvegetated sediment habitats to future high-pCO 2 climate. We do not assume that the responses of unvegetated sediments to the future climate found here are applicable to other ecosystems dominated by macrophytes, and we thus did not apply our findings to vegetated coastal habitats. To estimate total DOC export from coastal zone in a future high-pCO 2 climate of +3 • C and OA, the sediment uptake rate of 19 ± 4 mmol C m −2 d −1 was scaled to the global surface area of unvegetated estuarine sediments (1.8 × 10 12 m 2 ; Costanza et al., 1997). On this basis, an estimated 150 Tg C would be removed from the coastal zone by unvegetated estuarine sediments annually under OA conditions with an accompanying 3 • C temperature increase. To then calculate the potential impact of this uptake, we applied our estimates to existing future global coastal DOC export estimates (Maher and Duarte, 2017). Mean benthic DOC export from estuaries, including intertidal and vegetated habitats, has been estimated at 168 Tg C yr −1 (90-247 Tg C yr −1 ) (Maher and Eyre, 2010). Under this scenario, the switch to DOC uptake by sediments under future climate conditions (Fig. 5a) would result in ∼ 90 % reduction in total mean estuarine DOC export (Maher and Eyre, 2010), decreasing the load from ∼ 168 to ∼ 18 Tg C yr −1 . Other global estimates of DOC exported from coastal vegetated ecosystems range from 114 up to 1853 Tg C yr −1 (Duarte, 2017), with scaled estimates suggesting unvegetated estuarine sediments may consume 8 to 132 % of this DOC under a future high-pCO 2 climate. As such, this basic upscaling suggests that, by impacting DOC fluxes in unvegetated sediments, future climate conditions have the potential to significantly impact global DOC export from coastal systems to the open ocean. This has implications for global marine productivity and carbon transfer to the ocean interior (Krause-Jensen and Duarte, 2016). However, to get a more accurate insight into global carbon cycling, the response of DOC export from estuarine vegetated habitats to future climate also needs to be studied.
Author contributions. All listed authors have contributed substantially to preparation and drafting of this paper and have approved the final submitted manuscript. Specifically, MNS conceived the project, collected data, ran data analysis and interpretation, and led the writing of the manuscript. KGS, JMO, and BDE helped conceive the project, contributed to interpretation, and helped draft the manuscript.
Competing interests. The authors declare that they have no conflict of interest.