Ocean acidification (OA) is expected to drive the transition of coral reef ecosystems from net calcium carbonate (CaCO3) precipitating to net dissolving within the next century. Although permeable sediments represent the largest reservoir of CaCO3 in coral reefs, the dissolution of shallow CaCO3 sands under future pCO2 levels has not been measured under natural conditions. In situ, advective chamber incubations under elevated pCO2 (~800 µatm) shifted the sediments from net precipitating to net dissolving. Pore water advection more than doubled dissolution rates (1.10 g CaCO3 m−2 d−1) when compared to diffusive conditions (0.42 g CaCO3 m−2 d−1). Sediment dissolution could reduce net ecosystem calcification rates of the Heron Island lagoon by 8% within the next century, which is equivalent to a 25% reduction in the global average calcification rate of coral lagoons. The dissolution of CaCO3 sediments needs to be taken into account in order to address how OA will impact the net accretion of coral reefs under future predicted increases in CO2.
 The atmospheric CO2 concentration is expected to stabilize between 660 and 790 µatm by the year 2100, an increase of ~400 µatm from the present day concentration [Pachauri and Reisinger, 2007]. This increase in CO2 is predicted to decrease oceanic pH by ~0.3 units by the end of the century, a phenomenon termed ocean acidification (OA) [Feely et al., 2004]. Multiple studies have demonstrated a reduction in the net ecosystem calcification (NEC) rates of coral reefs due to increasing oceanic pCO2 [Andersson et al., 2009; Shamberger et al., 2011]. However, most studies to date have not separated the effects of OA on calcium carbonate (CaCO3) dissolution and coral calcification, despite dissolution being potentially more sensitive to OA than calcification [Andersson et al., 2009].
 Calcium carbonate found within coral reefs exists in two main pools; reef framework and CaCO3 sediments. Sediments often exceed the areal coverage of living coral and framework by up to one order of magnitude [Gattuso et al., 1998]. While reef framework is the main site of CaCO3 production due to the presence of living coral and other calcifying organisms, sediments represent the accumulation of reef generated CaCO3 over thousands of years [Smith et al., 2009]. Pore water advection refers to the bulk exchange of sediment pore water with the overlying water column and is largely driven by tides, wave action, sediment topography, sediment permeability, and currents [Precht and Huettel, 2003]. In contrast, exchange by diffusion results from a concentration gradient between the water column and pore water. Advection can enhance biological and geochemical processes occurring within the pore waters and fluxes of solutes into the water column [Wild et al., 2004; Eyre et al., 2008; Cyronak et al., 2013a].
 To date, most studies have examined the dissolution kinetics of shallow CaCO3 sediments in the laboratory or under diffusive conditions [Tynan and Opdyke, 2011; Yamamoto et al., 2012], making it difficult to predict dissolution kinetics in the environment. The few field investigations done that incorporate advective conditions and diel cycles have demonstrated that advection influences benthic alkalinity fluxes (i.e., CaCO3 precipitation and dissolution) [Rao et al., 2012; Cyronak et al., 2013a; Cyronak et al., 2013b]. However, none of the studies manipulated pCO2 to simulate predicted future OA conditions. In order to understand how OA will impact the accumulation of CaCO3 in coral reefs, it is necessary to determine the rates at which CaCO3 sediments will dissolve in situ under pore water advection, diel cycles, and predicted increases in pCO2. Herein, we performed in situ diffusive and advective chamber incubations on Heron Island under current and increased pCO2 scenarios.
 Heron Island is a coral cay in the Great Barrier Reef surrounded by a large (26.4 km2) and shallow (1.7m) coral lagoon covered mostly (~75%–85%) by CaCO3 sands [Wild et al., 2004; Eyre et al., 2008]. The high permeability and porosity of the sands allow seawater to easily flow in and out of the sediments. It is estimated that ~15% of the Heron Island lagoon seawater volume is filtered through permeable sands every day [Wild et al., 2004]. Our incubations were conducted in an area dominated by CaCO3 sands free from macrophytes and macrofauna burrows. Sediments from this site are comprised mostly of aragonite (65%) and high Mg2+ calcite (32%) with a Mg2+ content of 15.2% [Cyronak et al., 2013a].
 A total of 23, 24 h long incubations were conducted on three separate days (28 and 30 April 2013 and 2 May 2013). On each day, duplicate incubations for each treatment were run, with one advective high pCO2 incubation lost on 28 April 2013. The incubations on 28 April 2013 started at sunset (~18:00), while the other two incubations started at sunrise (~06:00). Benthic chambers as detailed in Eyre et al.  were used to measure benthic fluxes under diffusive and advective conditions (see supporting information Figure S1). The chambers were placed in the sediments without lids and left open to the overlying water for at least 1 h. Advective chambers were run at 80 RPM during the first incubation and 40 RPM thereafter. Previous studies have shown that these stirring rates induce pore water advection rates similar to those measured in situ at Heron Island [Wild et al., 2004; Glud et al., 2008]. The six replicates of each treatment were averaged for further analysis. Prior to the first sample being taken, chambers were closed to the overlying water and CO2 additions began. The pCO2 of the chambers was raised by pressurizing a closed loop of silicone tubing (Tygone 3350, Cole-Palmer, Vernon Hills, IL, USA, 0.48 cm i.d.) at 2 bar with 99.9% pure CO2 gas (Figure S1). The pH was monitored until the desired offset (~0.2) was reached, and the chambers were allowed to equilibrate for 30 min before the first samples were taken.
 Chambers were sampled (120 mL) every 12 h via syringe, and water was brought into the laboratory within 15 minutes for analysis. Dissolved oxygen (DO) (± 1%) was measured immediately in unfiltered samples using a Hach LDO probe (Hach Co., Loveland, CO, USA). Samples for pH (± 0.008) and total alkalinity (TA) (± 0.1%) were 0.45 µm filtered and kept in an airtight container with no bubbles until analysis within 4 h using a Metrohm Titrando automated titrator with a Metrohm Electrode Plus pH probe (Metrohm, Herisau, Switzerland). The pH probe was calibrated to pH NBS buffers (4, 7, 10), and TA was corrected against Dickson reference material (Batch 122). The pH was also monitored every 0.5 h in selected chambers using a SAMI2-pH (Sunburst Sensors, Missoula, MT, USA) unit, which measures pH in the total scale using metacresol purple as an indicator dye. The pH in the total scale is ~0.13 lower than the NBS scale; however, the exact magnitude is dependent on temperature, salinity, and individual pH electrode, therefore pH between the two scales is presented as measured. The pH and photosynthetically active radiation (PAR) were monitored in the water column every 15 min at the study site using a Hydrolab DS5X (Hach Co., Loveland, CO, USA). The pH from the Hydrolab was corrected to standards and seawater samples measured using the Metrohmn electrode (Metrohm, Herisau, Switzerland) (NBS scale). pCO2 and ΩAr were calculated using CO2SYS with pH, TA, and laboratory temperature measured during the TA titrations and a constant salinity of 35 as the input parameters. CO2SYS was set to the parameters as described in McMahon et al. .
 CaCO3 dissolution rates were calculated from TA fluxes assuming that half of the molar TA flux is equivalent to the molar CaCO3 flux. Other benthic fluxes that could contribute to TA fluxes within the chambers (i.e., nutrients) were shown to be minimal at the same study site, implying that TA fluxes represent CaCO3 precipitation and dissolution [Rao et al., 2012; Cyronak et al., 2013a]. The molar concentration of CaCO3 (100.1 g mol−1) was used to convert fluxes to g m−2. Student's t tests were performed to compare means between treatments while linear regression analysis was performed on any regressions.
 During the incubation on 2 May 2013, all carbonate parameters (pH, TA, pCO2, and ΩAr) varied over a diel cycle (Figure 1). The average pCO2 (n = 6) of the control incubations was 418 ± 48 and 427 ± 20 µatm in the diffusive and advective chambers, respectively. The high pCO2 treatments had significantly higher (p < 0.05, n = 6) averages of pCO2 (803 ± 98 and 839 ± 90 µatm) (Figure 1 and Table S2). This resulted in a ΔpCO2 between the treatments and controls of 385 and 412 µatm in the diffusive and advective chambers, respectively, close to predicted changes for the end of the century [Pachauri and Reisinger, 2007]. Average pH was 0.24 units lower (p < 0.05, n = 6) between the control and high pCO2 treatments under both diffusive and advective conditions, while ΩAr was reduced by an average of 1 unit (p < 0.05, n = 6) (Figure 1 and Table S2). The offset of pH between low and high pCO2 treatments was relatively stable throughout the day long incubations (Figure 1). Average pH in the water column of the Heron Island lagoon where the incubations took place was 8.12 with a minimum of 7.90 and maximum of 8.41. Therefore, the variability in the chamber incubations was similar to the natural variability of the lagoon water column. This variability was maintained at an offset in the high pCO2 treatments (Figure 1), in contrast to most previous CO2 manipulations in which pCO2 was kept constant over a diel cycle.
 Both dissolved oxygen (DO) and TA fluxes followed trends that were indicative of biological drivers (Figure 2a). DO fluxes were positive during the day (net benthic production; NPP) and negative during the night (respiration), while alkalinity fluxes revealed the sediments to be net precipitating in the day and net dissolving at night (Table S2). Net precipitation rates in the day were similar between all treatments except the advective low pH treatment, which was 30% lower than the control. Since NPP and TA fluxes are correlated (Figure S3), a 15% reduction in daytime production rates may partially explain the 30% decrease in daytime precipitation rates in the advective chambers. At night, elevated pCO2 resulted in a 35% and 68% increase in CaCO3 dissolution rates (p < 0.05, n = 6) in the diffusive and advective treatments, respectively.
 The amount of photosynthetically active radiation (PAR) varied between the days that incubations were performed, with daytime averages of 455, 272, and 521 µmol quanta m−2 s−1 during the first, second, and third day, respectively. Despite this variability in PAR, all of the chambers were net productive on a daily basis, while they varied between net CaCO3 precipitating and dissolving dependent on the treatment (Figure 2a). Both the diffusive and advective high pCO2 treatments were net dissolving while the controls were net precipitating (p < 0.05, n = 6). Advection interacted with high pCO2 to stimulate the night, day, and net dissolution rates of CaCO3 sediments above diffusive rates by 12%, 26%, and 150% (p < 0.05, n = 6), respectively (Figure 2a and Table S2). The net difference between control and high pCO2 conditions was 0.42 ± 0.22 g CaCO3 m−2 d−1 under diffusive conditions and 1.10 ± 0.22 g CaCO3 m−2 d−1 under advective conditions. This equates to a 162% increase in dissolution rates between diffusive and advective chambers (Figure 2b). This is consistent with increased flow of low pH surface waters into the interstitial pore waters under advective conditions. A significant positive linear trend was observed between net dissolution rates and the average chamber pCO2 under advective conditions (r2 = 0.660, p < 0.005, n = 11), while under diffusive conditions the regression was not significant (r2 = 0.333, p = 0.05, n = 12) (Figure 2c).
 The goal of this study was to assess in situ CaCO3 dissolution rates in permeable sediments over a natural diel cycle under OA conditions, and their influence on coral reef NEC rates. Previous observations in the Heron Island lagoon estimated an average community NEC rate of 6.15 g CaCO3 m−2 d−1 (2,246 g m−2 yr−1) [McMahon et al., 2013]. Using this estimate of NEC, CaCO3 sands currently contribute from 1.0% to 3.7% of community CaCO3 precipitation under diffusive and advective conditions, respectively. In the elevated pCO2 treatments, CaCO3 sands dissolved at rates equivalent to 5.9% of the daily NEC rate under diffusive conditions and 14.6% under advective conditions. The net differences between control and high pCO2 treatments equate to 6.8% and 17.9% of the NEC rate (Figure 2b). Because solute transport in permeable sands is often dominated by advective exchange [Santos et al., 2012] and sands make up at least 80% [Wild et al., 2004; Eyre et al., 2008] of the benthos of Heron Island lagoon, a 400 µatm increase in the average pCO2 could result in a reduction of current Heron Island NEC rates by ~14%.
 The relationship between TA and dissolved inorganic carbon (DIC) offers insights into changes in the carbonate system [Andersson and Gledhill, 2013]. Under OA conditions, elevated pCO2 increased DIC concentrations while TA remained constant (Figure 3). Advection in both the low and high pCO2 treatments shifted the slope of the TA versus DIC relationship by ~13% toward a more biological dominated system (i.e., more influence from photosynthesis/respiration than CaCO3 precipitation/dissolution) (Figure 3 and Table S2). This increase in the influence of organic processes under advective flow is consistent with other permeable sand studies [Rao et al., 2012; Cyronak et al., 2013a]. More biological control on the carbonate system would result in a larger range of ΩAr values in the overlying water column over a diel cycle [Andersson and Gledhill, 2013]. This also indicates that more CO2 per unit of TA would be fluxed out of the sediments under advective flow, partially inhibiting any buffering effect [see Andersson and Mackenzie, 2012] that sediment-derived TA may have in the water column. This is also supported by the lower net production rates measured in the high CO2 and advective incubations, indicating less of a net uptake of CO2 from the water column under future conditions (Figure 2a).
 Considering advection dominates solute exchange in coral reef permeable sands [Wild et al., 2004] and flushing rates in our advective chambers are comparable to in situ rates [Glud et al., 2008], we used the advective regression in order to model how the dissolution of CaCO3 sands could affect community NEC rates as a function of average water column pCO2. Assuming 80% coverage of CaCO3 sands, an increase in average pCO2 to 800 µatm in the overlying water (predicted by the year 2100) would result in an 8% decrease in the NEC rate of the Heron Island lagoon (Figure 4a). If our results are compared to the global average NEC rate of coral lagoons [Milliman, 1993] (800 g CaCO3 m−2 yr−1), the dissolution of CaCO3 sediments alone could reduce the annual NEC of coral lagoons 25% by 2100 (Figure 4a). Other models predict a 42% decrease in the CaCO3 production of the global coastal ocean by 2100 due to reductions in calcification rates alone [Andersson et al., 2005; Andersson et al., 2006]. Therefore, increases in sediment dissolution could decrease CaCO3 production an additional 25% above calcification-based model predictions. Also, the same models predicted an increase in CaCO3 sediment dissolution of only 20% by the year 2100 [Andersson et al., 2005; Andersson et al., 2006], while our model indicates that the dissolution of shallow CaCO3 sediments could increase 380% by the year 2100. Pore water advection at rates similar to those induced in our study have been calculated to occur at depths of up to ~50m and may occur deeper, depending on the physical characteristics of the sediments and surface gravity waves [Precht and Huettel, 2003]. Therefore, our results may be applicable to large areas of the coastal ocean and demonstrate the necessity of further elucidating other drivers of sediment dissolution in order to adequately predict changes to the global CaCO3 budget under future CO2 levels.
 Applying our model to the global estimate of coral NEC rates assumes that CaCO3 dissolution will be similar across coral reef ecosystems. However, there are multiple variables that could affect the dissolution rates of CaCO3 sediments. For instance, the % Mg content, structural disorder, presence of impurities, porosity of sediments, and biological activity can all influence dissolution [Morse et al., 2006; Cyronak et al., 2013a]. The exact influence of % Mg on dissolution rates of biogenic CaCO3 has yet to be fully elucidated and has been further complicated by the recent discovery of dolomite producing coralline algae [Nash et al., 2013]. However, the mineralogical makeup of Heron Island sediments (15 mol% Mg2+) puts it within the greatest frequency of occurrences for middepth bank CaCO3 sands [Morse et al., 2006]. This implies that the dissolution rates measured in the Heron Island sediments may be a reasonable first-order estimate of future dissolution rates throughout different ecosystems with similar sediment mineralogy and pore water advection rates.
 Correlations between benthic NPP and average water column depth in Heron Island (r2 = 0.77, p < 0.05, n = 26) [Eyre et al., 2013] offer insights into how net dissolution rates may vary with depth (Figure S4). Because NPP is significantly correlated to fluxes of alkalinity (CaCO3 dissolution) (Figure S3), any reduction in NPP would also result in an increase in net daily CaCO3 dissolution (Figure 4b). By the year 2100, sediment at an average depth of 3 m could undergo dissolution rates equal to the global average NEC rate of coral lagoons. If there is no photosynthesis in the sands, sediment dissolution rates approach the global average NEC rate under current CO2 levels. This suggests that the average depth of coral lagoons is critical in determining how shallow water systems will respond to future pCO2 levels. While our calculations illustrate the potential impact of depth on sediment dissolution in coral lagoons, additional studies may be needed to better understand permeable CaCO3 sediment dissolution in deeper environments such as continental shelves.
 In summary, considering that up to 90% of CaCO3 in coral reefs is contained within the sediments, measuring their dissolution rates in situ provides insights into how OA will affect the net accretion of coral reefs. Our results demonstrate that elevated pCO2 (~400 µatm above current) and advection act in synergy to increase dissolution ~5 times above the current precipitation rates of shallow CaCO3 sediments. Also, pore water advection may more than double any rates previously estimated under diffusive conditions. Other studies have estimated CaCO3 sedimentation rates of ~0.5 kg m−2 yr−1 in sandy reef environments [Ryan et al., 2001; Harney and Fletcher, 2003]. Comparing this sedimentation rate to the dissolution rates measured in this study demonstrates that by the year 2100 dissolution of CaCO3 sediments could reduce sedimentation rates by up to 80% of current values. Assuming our short term observations persist in the long term, this has drastic implications to the formation of valuable reef habitat, especially under rising sea levels.
 We would like to thank the staff of Heron Island Research Station for their help with field work. The research was supported by an ARC grant (DP110103638).
 The Editor thanks Merinda Nash and an anonymous reviewer for their assistance in evaluating this paper.