A Unique Diel Pattern in Carbonate Chemistry in the Seagrass Meadows of Dongsha Island: The Enhancement of Metabolic Carbonate Dissolution in a Semienclosed Lagoon

In contrast to other seagrass meadows where seawater carbonate chemistry generally shows strong diel variations with higher pH but lower partial pressure of CO2 (pCO2) during the daytime and lower pH but higher pCO2 during nighttime due to the alternation in photosynthesis and respiration, the seagrass meadows of the inner lagoon (IL) on Dongsha Island had a unique diel pattern with extremely high pH and low pCO2 across a diel cycle. We suggest that this distinct diel pattern in pH and pCO2 could be associated with the enhancement of total alkalinity (TA) production coupled to carbonate sediment dissolution in a semienclosed lagoon. The confinement of the IL may hamper water exchange and seagrass detritus export to the adjacent open ocean, which may result in higher organic matter loading to the sediments, and longer residence time of the water in the IL, accompanied by microbial respiration (both aerobic and anaerobic) that may reduce carbonate saturation level to drive carbonate dissolution and thus TA elevation, thereby forming such a unique diel pattern in carbonate chemistry. This finding further highlights the importance of considering TA production through metabolic carbonate dissolution when evaluating the potential of coastal blue carbon ecosystems to buffer ocean acidification and to absorb atmospheric CO2, in particular in a semienclosed setting.

In contrast to other seagrass meadows where seawater carbonate chemistry generally shows strong diel variations with higher pH but lower partial pressure of CO 2 (pCO 2 ) during the daytime and lower pH but higher pCO 2 during nighttime due to the alternation in photosynthesis and respiration, the seagrass meadows of the inner lagoon (IL) on Dongsha Island had a unique diel pattern with extremely high pH and low pCO 2 across a diel cycle. We suggest that this distinct diel pattern in pH and pCO 2 could be associated with the enhancement of total alkalinity (TA) production coupled to carbonate sediment dissolution in a semienclosed lagoon. The confinement of the IL may hamper water exchange and seagrass detritus export to the adjacent open ocean, which may result in higher organic matter loading to the sediments, and longer residence time of the water in the IL, accompanied by microbial respiration (both aerobic and anaerobic) that may reduce carbonate saturation level to drive carbonate dissolution and thus TA elevation, thereby forming such a unique diel pattern in carbonate chemistry. This finding further highlights the importance of considering TA production through metabolic carbonate dissolution when evaluating the potential of coastal blue carbon ecosystems to buffer ocean acidification and to absorb atmospheric CO 2 , in particular in a semienclosed setting.

INTRODUCTION
As one of the most productive ecosystems on Earth, seagrass meadows have been recognized for their important role in "blue carbon" storage (Duarte et al., 2010;Fourqurean et al., 2012). In addition to the significant carbon sequestration potential of seagrass meadows, previous studies have shown that the high level of seagrass primary productivity may alter seawater carbonate chemistry by taking up a large amount of carbon dioxide (Frankignoulle and Distèche, 1984;Gattuso et al., 1998), and some recent studies have further proposed that seagrass meadows have the potential to mitigate ocean acidification (OA) induced by rising atmospheric CO 2 (Manzello et al., 2012;Unsworth et al., 2012;Pacella et al., 2018).
Carbonate chemistry dynamics in the water column of seagrass meadows is well known to be driven by a variety of metabolic activities, including plant photosynthesis/respiration (Gattuso et al., 1998;Semesi et al., 2009), carbonate formation/dissolution (Barrón et al., 2006;Burdige et al., 2008;Howard et al., 2018;Saderne et al., 2019b), benthic metabolism (Berg et al., 2019), and hydrodynamic processes (Ruesink et al., 2015;Baldry et al., 2020). As a result, the carbonate system in the overlying water column of seagrass meadows generally shows large variabilities at diel and tidal time scales as well as seasonal variations (Frankignoulle and Distèche, 1984;Waldbusser and Salisbury, 2014;Cyronak et al., 2018;Saderne et al., 2019a). In fact, several in situ investigations have revealed conflicting evidence as to whether seagrass meadows can buffer against OA. For instance, Challener et al. (2016) found significant diel and seasonal variability in seawater pH and the saturation state of aragonite ( a ) in a Florida seagrass meadow; the high pH and a values, which would alleviate OA, were observed during the daytime/growing season due to dissolved inorganic carbon (DIC) uptake, while the low pH and a values, which would exacerbate OA, occurred during the nighttime/decay season due to DIC release. Furthermore, Hendriks et al. (2014) also reported diel pH changes in Mediterranean seagrass meadows, where the pH during 47% of the observation time was lower than the source seawater pH, suggesting a certain period of time when the seagrass meadows could not mitigate OA but exacerbated OA. These surveys clearly demonstrate that the OA buffering potential of seagrass meadows may have considerable temporal variability.
Our previous study documented the divergent behaviors in carbonate chemistry in two hydrodynamically contrasting seagrass meadows at Dongsha Island in the northern South China Sea (NSCS) during a 6-day survey in August 2015 (Figure 1), in which higher pH and total alkalinity (TA) but lower partial pressure of CO 2 (pCO 2 ) were observed in the semi-enclosed inner lagoon (IL) than on the open north shore, and a unexpected diel pattern with an extremely high pH and pCO 2 across a diurnal cycle was found in the sheltered IL (Chou et al., 2018). In this study, we revisited the same sites to examine whether the divergent behaviors in carbonate chemistry between two sites and the distinct diel pattern in the sheltered IL would occur repeatedly in all seasons. More importantly, recent studies have suggested that sedimentary TA production and export may represent an important but overlooked sink in blue carbon ecosystems (Sippo et al., 2016;Maher et al., 2018;Saderne et al., 2020;Reithmaier et al., 2021). Thus, we also collected sediment cores and porewater samples to examine the potential role of TA production coupled to carbonate dissolution in regulating the carbonate dynamics in the overlying waters. We found that the confinement of the semienclosed IL may provide an ideal scenario for metabolic carbonate dissolution in sediments and thus TA production, including high organic matter (OM) content, low sediment permeability and a long porewater residence time, thus distinguishing the seagrass meadows in the IL on Dongsha Island from those in the open environment.

Study Sites
The study sites were the same as those in our previous work and have been comprehensively described in Chou et al. (2018). Briefly, the Dongsha Atoll is a circular coral reef located in the NSCS, and Dongsha Island is situated on the western margin of the atoll. A semienclosed IL occupies the central part of Dongsha Island (Figure 1). Two hydrodynamically contrasting seagrass meadows were chosen so that their characteristics in seawater carbonate chemistry could be compared. One meadow is located on the northern shore (NS), where water can freely exchange with the adjacent open ocean; the other is situated in the IL, where water exchange is largely hampered due to confinement by a sand barrier. Both sites are multispecies seagrass meadows with the same dominant species (Thalassia hemprichii and Cymodocea rotundata). Previous study have revealed that the average total seagrass biomass, seagrass cover, canopy height, and shoot density were 949.4 ± 62.1 and 827.4 ± 74.8 (g DW m −2 ), 81.91 ± 2.13 and 85.14 ± 6.33 (%), 22.36 ± 1.22 and 20.62 ± 1.52 (cm), and 2677 ± 485 and 2920 ± 639 (shoots m −2) on the NS and in the IL, respectively, based on four surveys conducted in April, August, October 2010, and February 2011 .

Sampling and Carbonate Chemistry Analysis
This study was conducted in four different seasons between 2016 and 2019. During each investigation, a multiparameter sonde (Ocean Seven 316 Plus CTD; Idronaut S.r.I, Brugherio, Italy) attached to a frame was deployed within the seagrass canopy 0.1 m above the sediment at each site to record temperature, salinity, water depth, and dissolved oxygen (DO). The CTD mooring were from January 2 to 7 and from January 7 to 12, 2016 in winter, and from November 16 to 19 and from November 11 to 19 with an interruption between November 14 and 16, 2016 in autumn at the NS and IL sites, respectively. In spring, the CTD mooring was from April 6 to 18, 2018 at the both sites, and with an interruption between April 7 and 10 at the NS site. Furthermore, probably due to the low water level with periods of emersions, a malfunction of the CTD caused a missing of temperature and salinity data at the IL site during the spring deployment. And so, the temperature, salinity and DO data were disregarded for the NS site when water level less than 15 cm. In summer, the CTD mooring was from June 28 to July 3, 2019 at the two sites.
Discrete seawater samplings were conducted in January 2016 (1/1-12; winter), November 2016 (11/11-13; autumn), April 2018 (4/12-18; spring), and June-July 2019 (6/28-7/3; summer). The water depth at the IL and NS sampling sites was approximately 0.5 to 1 m. During these sampling periods, discrete surface seawater samples for carbonate chemistry analysis were taken at approximately 6:00 a.m., 12:00 a.m., and 18:00 p.m. every day, using a 9-L Nalgene TM HDPE carboy. The water samples were first passed through a 100-µM nylon mesh, and then transferred into 350-mL pre-cleaned borosilicate bottles thoroughly flushed with sample with extensive overflow (>200 mL). These samples were subsequently injected with 100 µL of saturated HgCl 2 solution and stored at room temperature in darkness until they were analyzed in a shorebased laboratory. Measurements of TA, DIC, and pH followed the standard operating procedures described in Dickson et al. (2007), and the procedures were consistent with those used in our previous studies (Chou et al., 2018(Chou et al., , 2020. Briefly, DIC and TA were determined using the non-dispersive infrared method on a DIC analyzer (AS-C3, Apollo SciTech) and Gran titration on an automatic TA titrator (AS-ALK2, Apollo SciTech), respectively, and both measurements had accuracies and precisions of 0.2% or better. The pH was spectrophotometrically measured at 25 • C with a precision of 0.005 (Chou et al., 2016). The pCO 2 and a were calculated from the measured DIC and TA data using the Excel macro CO2SYS version 2.1 (Pelletier et al., 2011). Furthermore, in order to examine the effect of temperature on pH and pCO 2 variations, the measured pH at 25 • C (pH 25 ) was corrected to in situ temperature (pH in situ ), and the pCO 2 was calculated at both in situ temperature (pCO 2in situ ) and the average temperature of 26.2 • C (TpCO 2 ). Similarly, TA and DIC were normalized to the average salinity of 35.6 (NTA (NDIC) = TA (DIC) × 35.6/salinity) to remove the effect of evaporation/precipitation on their variations (Chou et al., 2018).

Sediment Cores and Porewater
Samplings and Grain Size, Total Carbon and Nitrogen, and Calcium Ion Analyses Sediment cores and porewater samples were collected in the summer of 2019 at the IL, NS, and another unvegetated site located on the southern shore (SS; Figure 1), which served as a reference site. Because of the difficulty inherent in collecting a core of sandy sediments (Drupp et al., 2016), only one 25-cm push core from the IL, one 16.5-cm push core from the NS and one 18-cm push core from the SS were collected. The IL, NS, and SS cores were sectioned into 5, 3, and 3 layers ∼5 to 6 cm thick, respectively, which were stored in a −20 • C freezer until subsequent analysis. Grain size distribution was determined via wet-sieving through the Wentworth series of screens with mesh openings from 1.0 mm to 63 µm (Folk, 1966). Median grain size was calculated using the GRADISTAT software (Blott and Pye, 2001). Each core section was measured for total carbon (TC) and total nitrogen (TN) contents using an elemental analyzer (Elementar, Vario EL-III, Germany), according to Hung and Gong (2010).
Porewater samples were collected using porewater wells and modified from Falter and Sansone (2000). A total volume of 25 ml porewater was extracted from each well at 2, 4, 6, 8, 12, 16, and 20 cm sediment depths using a Luer-Lok syringe. The sampling procedure and sample preservation of porewater followed the methods in Kindeberg et al. (2020). In addition to carbon chemistry parameters (TA, DIC, and pH), the calcium ion concentration of the porewater was determined using an ICP-MS system (Agilent 7700×, Agilent Technologies) based on that in Su and Ho (2019).

Statistical Analysis
Due to the variance changes across seasons and sites, a Wilcox's robust ANOVA (WR-ANOVA) was chosen to account for heteroscedasticity among habitat groups. Tests between medians were chosen rather than between means, as seagrass habitats displayed skewed carbonate chemistry observations (Baldry et al., 2020). Differences in the medians of pH 25 , pH in situ , pCO 2in situ , TpCO 2 , DIC, normalized DIC (NDIC), TA, and normalized TA (NTA) between the IL and NS sites in each season were assessed using a WR-ANOVA that was implemented by the function "med1way" of the R package "WRS2" (Mair and Wilcox, 2020). All statistical tests were performed using R software v4.1.1 (R Core Team, 2021) with a 95% confidence level.

Temporal Variations in Temperature, Salinity, Water Depth, and Dissolved Oxygen
The mean (±SD) temperature during the sampling period was 23.6 ± 1.8, 27.1 ± 0.8, 26.4 ± 2.1, and 31.1 ± 1.1 • C at the NS site (red symbols in Figures 2A-D), and 21.7 ± 1.4, 26.3 ± 1.1, not determined (n.d.), and 31.4 ± 1.7 • C at the IL site (blue symbols in Figures 2A-D) in winter, autumn, spring and summer, respectively. Not surprisingly, the highest average temperature was in summer, and the lowest was in winter, and the average temperature in spring and autumn was in the intermediate range.
The mean (±SD) salinity was 34.6 ± 0.3, 33.8 ± 0.5, 33.8 ± 0.4, and 33.8 ± 0.5 at the NS site (red symbols in Figures 2E-H), and 36.8 ± 0.3, 34.7 ± 0.3, n.d., and 34.9 ± 0.6 at the IL site (blue symbols in Figures 2E-H) in winter, autumn, spring, and summer, respectively. The highest average salinity was in winter, but the average salinity did not display remarkable difference among the other seasons at the both sites. Generally, the average salinity at the IL site was higher than the NS site in all seasons. The mean (±SD) water depth was 0.71 ± 0.17, 0.62 ± 0.37, 0.37 ± 0.18, and 1.07 ± 0.29 at the NS site (red symbols in Figures 2I-L), and 0.48 ± 0.11, 0.70 ± 0.20, 0.24 ± 0.10, and 0.50 ± 0.20 at the IL site (blue symbols in Figures 2I-L) in winter, autumn, spring and summer, respectively. The lowest average water depth was in spring at the both sites, and the highest average water depth was in autumn and summer at the IL and NS sites, respectively. Overall, the water depth showed a larger daily variation range at the NS site than the IL site in all seasons. The mean (±SD) DO saturation during the complete recording period was 93 ± 33, 112 ± 37, 83 ± 42, and 66 ± 33 at the NS site (red symbols in Figures 2M-P), and 77 ± 12, 82 ± 13, 69 ± 21, and 59 ± 19 at the IL site (blue symbols in Figures 2M-P) in winter, autumn, spring, and summer, respectively. The highest and lowest average DO saturation was in autumn and summer, respectively, at the both sites. Similar to the water depth, DO saturation generally showed a larger daily variation range at the NS site than the IL site in all seasons.

Temporal Variations in Carbonate Chemistry
The seasonal daily variations in pH 25 , pH in situ , pCO 2in situ , TpCO 2 , TA, NTA, DIC, and NDIC at the NS and IL sites are shown in Figure 3, and the means (±SD) and the statistical analysis results on the difference in the medians of these parameters between the two sites are given in Table 1. Despite the notable daily variations in temperature and salinity, the temperature corrected pH (pH 25 ) and pCO 2 (TpCO 2 ) generally displayed the similar daily variation pattern as those in in situ condition. Likewise, the salinity normalized NTA and NDIC also showed the parallel daily variation pattern as those in in situ condition.
The horizontal bands superimposed on Figures 3A-D,I-P represent the typical seasonal variation ranges of pH, DIC, and TA, respectively, in the adjacent NSCS, which were collected from the SouthEast Asian Time-series Study (SEATS) station at 18 • N and 116 • E (Figure 1) during 19 cruises between 1999 and 2003 (Chou et al., 2005;Tseng et al., 2007). The horizontal dashed lines shown in Figures E-H denote the atmospheric pCO 2 (400 µatm). As shown, at the NS site, the pH 25 was generally higher than the typical ranges in the adjacent NSCS from midday to midnight; however, the pH 25 was lower from midnight to midday in all seasons, and the opposite pattern was found for NDIC. Similarly, TpCO 2 was lower than the atmospheric pCO 2 from midday to midnight, while it was higher from midnight to midday in all seasons, except in spring, when TpCO 2 across a daily cycle was almost always lower than the atmospheric pCO 2 . NTA across a daily cycle was generally higher than the typical range in the adjacent NSCS, except in summer, when NTA varied within the typical range. In contrast, at the IL site, the pH 25 across a daily cycle was higher than the typical range in the adjacent NSCS, while the TpCO 2 across a daily cycle was lower than the atmospheric pCO 2 in all seasons, suggesting that OA buffering and atmospheric CO 2 absorbing capacities for the seagrass meadows in the IL on Dongsha Island existed in all seasons. Furthermore, NDIC across a daily cycle was generally lower than the typical range in the adjacent NSCS, except in fall when NDIC was higher. In contrast, NTA across a daily cycle was almost always higher than the typical TA range in the adjacent NSCS in all seasons.

Porewater Carbonate Chemistry and Calcium Ion Concentration
The vertical porewater profiles of the carbonate parameters and calcium ion concentrations at the SS, NS, and IL sites are shown in Figures 4A-F. Overall, porewater DIC, TA, and pCO 2 profiles showed an increasing trend with sediment depth, and values were generally greater in the porewater than in the overlying water column at all sites. In contrast, pH and a revealed a decreasing trend, and values were lower in the porewater than in the overlying water column. Nevertheless, the vertical gradients of the carbonate parameters differed strikingly among the different sites. The sharpest vertical gradients of porewater carbonate parameters were found at the IL site, where the average depth-integrated changes in DIC, TA, pCO 2 , pH, and a relative to the sediment-water interface (SWI) were +3502 ± 1218 µM, +1791 ± 865 µM, +27457 ± 15775 µatm, −2.31 ± 0.18 pH units, and −12.29 ± 0.73, respectively, and the smallest vertical gradients of carbonate parameters were observed at the unvegetated SS site, where the average depth-integrated changes in DIC, TA, pCO 2 , pH, and a were +77 ± 23 µM, +27 ± 28 µM, +151 ± 75 µatm, −0.18 ± 0.15 pH units, and −0.51 ± 0.31, respectively. The vertical profiles of the carbonate parameters at the NS site were closer in magnitude to those at the unvegetated SS site than those at the IL site, where the average depth-integrated changes in DIC, TA, pCO 2 , pH, and a were +386 ± 114 µM, +91 ± 67 µM, +1456 ± 627 µatm, −0.44 ± 0.09 pH units, and −2.90 ± 0.57, respectively. Furthermore, it is worth noting that porewater below 2 cm at the IL site was generally undersaturated with respect to aragonite ( a < 1; except at 12 cm), while porewater remained supersaturated ( a > 1) throughout the entire profiles at the NS and SS sites ( Figure 4E).
The vertical profiles of calcium ion concentration in porewaters also differed markedly among the different sites. The calcium ion concentration remained fairly constant at 9.5-10 mM throughout the entire profile at the SS and NS sites, while it gradually increased from the SWI to a maximum of 13.1 mM at 12 cm and then decreased to 20 cm at the IL site. Similar to those of the carbonate parameters, the average depth-integrated change in calcium ion concentration at the IL site (+0.76 ± 0.84 mM) was noticeably higher than those at the NS (+0.23 ± 0.06 mM) and SS (+0.17 ± 0.17 mM) sites.

Median Grain Size, Total Carbon, and Total Nitrogen in Sediments
The vertical distributions of MGS, TC, and TN in the sediments at the SS, NS, and IL sites are shown in Figures 4G-I. Similar to the results of the porewater, the MGS, TC, and TN in the sediments also revealed considerable difference among the different sites. The coarsest grain size was found at the unvegetated SS site (MGS ranging from 800 to 1000 µm), followed by that at the NS site (MGS ranging from 400 to 600 µm), and the finest grain size was observed in the top 12.5 cm layer at the IL site (MGS ranging from 200 to 400 µm). TC and TN showed the highest contents at the IL site, with a mean of 17.1 and 0.82 wt%, respectively. At the unvegetated SS site, TC and TN remained relatively constant throughout the core and had the lowest contents, with a mean of 12.8 and 0.01%, respectively. The TC and TN contents at the NS site were closer in magnitude to those at the SS site than those at the IL site, with a mean of 14.1 and 0.04 wt%, respectively.

DISCUSSION
The Differences in pH and Partial Pressure of CO 2 Between the Northern Shore and Inner Lagoon Sites In terms of carbonate chemistry, two striking differences between the NS and IL sites can be clearly seen in Figures 3A-H. First, pH was generally higher, while pCO 2 was lower at the IL sites than those at the NS site (noting that due to the smaller sampling numbers and larger variations, the medians of pH in autumn and the medians of pCO 2 in winter and autumn did not show significant difference between the two sites according to a WR-ANOVA analysis, Table 1). Second, the amplitude of daily variations in pH and pCO 2 at the IL site were smaller than the NS site. Because the temperature-corrected pH and pCO 2 (pH 25 and TpCO 2 ) displayed the similar patterns (Figures 3A-H) and  the NS and IL sites exhibited similar temperatures in both autumn and summer (Figures 2B,D), temperature effect could not explain the observed different behaviors in pH and pCO 2 between the two sites.
Previous studies have indicated that hydrodynamics is of major importance in seagrass biology, ecology, and ecophysiology (Koch, 1994;Koch et al., 2006). Unfortunately, with the present dataset we cannot clearly quantify the influence Differences in the medians of these parameters between the IL and NS were assessed using a Wilcox's robust ANOVA. A significance level of 0.05 was used to determine significant statistical differences.
of physical mixing processes on the dynamics of pH and pCO 2 . Nonetheless, the water depth revealed smaller variation ranges at the IL site than the NS site in all seasons (Figures 2I-L), suggesting a weaker water exchange process and thus a potential higher control of metabolic activities in carbonate chemistry at the sheltered IL site. Additionally, as shown in Figure 5, the variations of DO saturation (green symbols) were generally in phase with temperature (red symbols) but out of phase with water depth (blue symbols) at the both sites, implying that factor(s) other than water exchange could also play an important role in regulating the variations of biogeochemical characteristics in the study area. From these lines of evidence, we suggest that the physical mixing processes and the diverse biogeochemical processes deriving from the different hydrodynamic regimes between the two sites could altogether cause the observed different behaviors in pH and pCO 2 between the NS and IL sites.

The Enhancement of Metabolic Carbonate Dissolution in the Inner Lagoon
As revealed in the results, the CO 2 dynamics at the NS site showed strong diel variation in all seasons (red squares in Figure 3), which was similar to the common pattern found in other seagrass meadows (Frankignoulle and Distèche, 1984;Gattuso et al., 1998;Bouillon et al., 2007;Turk et al., 2015;Ganguly et al., 2017;Cyronak et al., 2018;Berg et al., 2019;McCutcheon et al., 2019). In contrast, the unexpected high pH and low pCO 2 values across a diel cycle were found repeatedly at the IL site (blue circles in Figure 3) in four different seasons. We suggest that the semienclosed setting of the IL may provide an ideal circumstance for sedimentary TA production through enhanced carbonate dissolution that could represent an important driver to form the observed unique diel pattern in carbonate chemistry at the IL site.
Previous studies have shown that seagrasses may induce carbonate sediment dissolution by the combined effect of OM being supplied and oxygen pumping via their roots and rhizomes, which may fuel OM respiration and thus CO 2 release and a lower carbonate saturation state ( ), consequently leading to carbonate mineral dissolution (Burdige and Zimmerman, 2002;Burdige et al., 2008). This process is often referred to as "metabolic carbonate dissolution" (Kindeberg et al., 2020), as represented in the following equations: (1) where CH 2 O represents a simplified formula for OM undergoing remineralization. The net reaction can therefore be described stoichiometrically as: which increases both the TA and DIC of the porewater with a ratio of 1:1.
We suggest that the "metabolic carbonate dissolution" can be stimulated in the seagrass meadows of the IL, due to several auspicious conditions. First, the confinement of the semienclosed IL may hamper seagrass detritus export to the adjacent open ocean and thus cause more OM to accumulate in the reef sediments at the IL site, as evidenced by the elevated TC and TN contents shown in Figures 4H,I. More OMs may fuel stronger aerobic and/or anaerobic respiration that can decrease carbonate saturation level, and thus drive carbonate dissolution. This explanation is further supported by the undersaturated values of a (<1) and elevated levels of Ca 2+ in the porewater at the IL site (Figures 4E,F). Second, the relatively calm hydrodynamic environment at the IL site could also be favorable for fine-grained sediment accumulation as shown in Figure 4G, which may facilitate DIC and TA production in various ways. For instance, fine-grained sediments can provide greater available reactive surface area for all of the metabolic processes. Moreover, the overall finer-grained sediments may result in a lower permeability that may reduce oxygen penetration and thus favor the occurrence of anaerobic OM remineralization, which could also contribute to DIC and TA production (Hu and Cai, 2011;Reithmaier et al., 2021). Finally, the less-energetic hydrodynamic and low sediment permeability may collectively result in a longer porewater residence time and thus allow for the buildup of DIC and TA in the porewater at the IL site as shown in Figures 4A,B. Besides, previous studies have indicated the occurrence of carbonate mineral reprecipitation along with dissolution in sediments at dense seagrass site (Hover et al., 2001;Hu and Burdige, 2007). Though we cannot be sure whether the coupled carbonate dissolution and reprecipitation occurred at our study sites due to the limitation of data, we think that carbonate mineral reprecipitation could be more likely to take place in the porewater at the NS site compared to the Il site, because the former was supersaturated but the latter was undersaturated with respect to aragonite (Figure 4E), and this could also partially explain why TA and DIC in the porewater at the NS site were lower than the IL site. Nevertheless, we suggest that the role of coupled carbonate dissolution and reprecipitation in regulating CO 2 system in porewater of seagrass meadows remains to be further examined.
Although the "metabolic carbonate dissolution" can simultaneously release DIC and TA in the porewater (reaction 3), which can then be exported to the overlying water column via the advection induced by tide, current, and wave actions and/or the diffusion driven by the chemical gradient between the SWI. The exported DIC can be taken up again through the high productivity of the seagrasses, whereas photosynthesis cannot consume TA. Consequently, the TA exported from porewater can remain in the overlying water column. The confinement of the IL site may also hinder water exchange with the adjacent open ocean, thus providing another favorable circumstance for the accumulation of the sedimentary generated TA in the overlying water column. Thermodynamically, TA increases with a constant DIC may not only drive pH increases and pCO 2 decreases in seawater but also enhance their buffer capacities, which could at least partially explain why a weak diel pattern with an extremely high pH and low pCO 2 across a diurnal cycle was seasonally observed within the seagrass meadows at the IL site.
It is worth noting that the studied seagrass meadows are not totally isolated from the Dongsha Atoll, where the living coral reef flat encircles the lagoon (Figure 1). Depending on the direction and strength of water exchange, the CO 2 system in individual ecosystem can affect each other (Sippo et al., 2016;Saderne et al., 2019a). In fact, a recent study has shown that TA in the Dongsha Atoll lagoon is depleted relative to the offshore water due to calcification therein (Fan et al., 2021). Thus, if the produced TA through the aforementioned processes in seagrass meadows can be exported to the adjacent Dongsha Atoll lagoon, it could at least partially offset the TA depletion from calcification, and thus could be beneficial for the associated coral reefs facing future OA. However, quantifying the export of TA by benthic ecosystems requires hydrological and bathymetrical data, and models (Saderne et al., 2019a), which are not available for us now. Nevertheless, we highlight the importance of quantifying the hydrodynamics-induced TA and DIC exchanges between coastal ecosystems, which could be crucial to better understand how costal blue carbon ecosystems could contribute to the growth of downstream reefs (Sippo et al., 2016;Saderne et al., 2019a).

Relationship Between Normalized Total Alkalinity and Normalized Dissolved Inorganic Carbon
Besides organic and CaCO 3 metabolisms associated to the plant themselves, and the related flora and fauna, aerobic and anaerobic processes in sediments may also affect seawater carbonate chemistry in seagrass meadows (Chou et al., 2018;Saderne et al., 2019a). Generally, the stoichiometry of the relationship between NTA and NDIC is used to interpret the dominant processes driving the carbonate system because the relative variation in TA and DIC follows a well-established stoichiometric ratio that is specific to the respective processes, where the ratios for organic metabolism (photosynthesis/respiration), carbonate metabolism (calcification/carbonate dissolution), and net "metabolic carbonate dissolution" (reaction 3) are 0/1, 2/1, and 1/1, respectively (Hu and Cai, 2011;Rassmann et al., 2020). In addition to organic and CaCO 3 metabolisms, a series of redox reactions in sediments, including denitrification, manganese reduction, iron reduction, and sulfate reduction (SR), may also play an important role in regulating the carbonate system. As sulfate is far more abundant than other oxidant in seawater, SR is often the dominant redox reaction in sediment with a relative variation ratio in TA and DIC of 1/1 (Burdige, 2011), as denoted in the following equation: (4) Figure 6 shows the covariation in NTA and NDIC for the water samples collected at the IL ( Figure 6A) and NS (Figure 6B) sites. As shown, the slopes of NTA vs. NDIC of the regression lines for the IL and NS are 0.73 and 0.35, respectively. Neither slope follows the specific ratio of any single process, implying that organic and CaCO 3 metabolisms, and SR may collectively control TA and DIC variations at the IL and NS sites (Drupp et al., 2016). Though partitioning between the causes of the observed CO 2 system variation is difficult, the higher slope suggests that the "metabolic carbonate dissolution" and/or SR could play a more FIGURE 7 | A tentative conceptual model explaining the enhancement of metabolic carbonate dissolution in the seagrass meadows of the inner lagoon on Dongsha Island. First, the elevated OM loading may fuel aerobic respiration (reaction 1) that can produce DIC and reduce carbonate saturation level leading to carbonate dissolution (reaction 2) and thus TA and DIC production, i.e., the "metabolic carbonate dissolution." Secondly, presuming that sulfate reduction (reaction 3) would take place in the deeper anoxic layer and the resulting sulfide subsequently re-oxide to sulfate (reaction 4) in the shallow oxic layer, the whole process can serve as a proton shuttle that can also facilitate carbonate dissolution. Finally, the sedimentary exported DIC can be taken up again by the high productivity of the seagrasses, whereas photosynthesis (reaction 5) cannot consume TA. Consequently, the sedimentary exported TA can remain in the overlying seawater, and thereby forming a unique diel pattern with extremely high pH and low pCO 2 values across a diel cycle. important role in regulating carbonate chemistry at the IL site compared to the NS site.
Besides, it is worth noting that although we did not measure sulfate and sulfide in this study, the presence of SR in the sediments at the IL site is to be expected, providing that the favorable conditions of elevated OM loading and finer MGS were observed. If SR takes place, the resulting H 2 S in ironpoor carbonate sediments could easily re-oxidize to sulfate within the shallow oxic layer. This process can serve as a proton shuttle to facilitate carbonate dissolution (Ku et al., 1999;Burdige et al., 2008;Drupp et al., 2016), as represented in the following equations: Therefore, SR coupled to sulfide re-oxidation may represent another potential pathway to stimulate carbonate dissolution at the IL site. Nonetheless, the net impact of reactions 4-6 on carbonate system are indistinguishable from "metabolic carbonate dissolution" denoted by reaction 3. Thus, more studies of detailed porewater chemistry in sulfate and sulfide as well as pyrite formation in sediments will be needed to further clarify the potential contribution of SR to TA production (Reithmaier et al., 2021). In fact, the partitioning between the sources of TA from CD and SR remains a scientific and analytical challenge to date in blue carbon studies .
In summary, based on these discussions, we tentatively proposed a conceptual model summarizing the potential processes controlling TA and DIC dynamics in the IL in Figure 7, which may collectively enhance sedimentary CaCO 3 dissolution and thus TA production.

The Potential Role of Organic Alkalinity
In seawater dissolved organic acids produced by organic metabolism dissociate into conjugate bases, which may react with protons during seawater titration and thereby contribute to the titration alkalinity (i.e., organic alkalinity, TA org ; Ko et al., 2016). A number of studies have shown that TA org can be significant in productive coastal environments (Cai et al., 1998;Kulinski et al., 2014). Therefore, TA org could also play a non-ignorable role in regulating carbonate chemistry in seagrass meadows. TA org has been measured indirectly as a difference between TA measured by direct titration (TA meas ) and TA calculated (TA cal ) from the measured DIC and pH (i.e., TA = TA meas -TA cal ; Yang et al., 2015). Since the CO 2 system was overdetermined in the present study, we can determine TA org by calculating TA. Figure 8 shows the comparison of the TA meas and TA cal at the NS and IL site, in which the difference between the data points and the 1:1 line ( TA) denotes TA org . As shown, the TA meas and TA cal generally follow the 1:1 line in winter, autumn and summer at the NS site, and in winter and autumn at the IL site, suggesting no significant TA org production during these periods. In contrast, the TA meas are apparently higher than TA cal in spring at the NS site, and in spring and summer at the IL site, indicating substantial TA org production during these periods. The higher TA org production in spring and summer at the IL FIGURE 8 | Comparison of the measured TA (TA meas ) by direct titration and the TA calculated (TA cal ) from the measured DIC and pH in the inner lagoon (IL; circles), and on the northern shore (NS; squares). The difference between TA meas and TA meas ( TA) denotes organic TA (TA org ; see text for definition).
site could be associated with the more active organic metabolism in seagrass meadows during the warm seasons . Moreover, it is worth noting that spring and summer are also the only two seasons when the median of TpCO 2 is significant lower at the IL site compared to the NS site according to a W-R ANOVA analysis (Table 1). Also, at the NS site, when TA org production was higher in spring, the amplitudes of daily variations in pH and pCO 2 appear to be smaller than the other seasons (Figures 3A-H). These results imply that in addition to the aforementioned processes in Sections "The Enhancement of Metabolic Carbonate Dissolution in the Inner Lagoon" and "Relationship Between Normalized Total Alkalinity and Normalized Dissolved Inorganic Carbon, " TA org production could also play a non-ignorable role in regulating carbonate chemistry in seagrass meadows, especially during the warm seasons. Therefore, we suggest that TA org production could be a potentially interesting and relevant research topic in studying CO 2 system in costal blue carbon ecosystems, which needs more research.

CONCLUDING REMARKS
To the best of our knowledge, the seasonal recurrence of high pH and low pCO 2 values across a diel cycle within seagrass meadows, such as those at the IL site on Dongsha Island, has never been reported in other marine ecosystems. We suggest that this unique diel pattern in carbonate chemistry could be associated with the semienclosed nature of the IL site, including, higher OM loading, finer-grained sediment, smaller sediment permeability and longer residence time of the porewater, that may stimulate both aerobic and anaerobic respiration leading to CaCO 3 dissolution and thus TA production. Overall, our results demonstrate that CaCO 3 dissolution in vegetated reef sediments can be enhanced in a semienclosed environmental setting. Most importantly, the present study provides the first observational evidence showing that the intense mechanistic coupling between metabolic processes and carbonate dissolution in seagrass meadows in reef sediment with a confinement setting may create a localized buffering effect against OA and a strong CO 2 sink, and thus may provide a valuable theoretical proposition for conserving and restoring seagrass meadows as a promising strategy for climate change mitigation. We also highlight the importance in quantifying the hydrodynamicsinduced TA and DIC exchanges between coastal ecosystems, which could be crucial to better understand how costal blue carbon ecosystems could contribute to the growth of downstream reefs. Furthermore, we find that TA org could play a non-ignorable role in regulating carbonate chemistry in seagrass meadows, and thus may be a potentially interesting and relevant research topic in studying CO 2 system in costal blue carbon ecosystems. Finally, since the present work was not conducted over a single year, the observed seasonal variation could be also affected by inter-annual variability, more studies with better temporal coverage are still needed to further elucidate the potential impact of inter-annual variability in our findings.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: The data used for this study are available at PANGEA (https://doi.org/10.1594/PANGEA.928748).

AUTHOR CONTRIBUTIONS
W-CC conceived the study, performed the field work, and wrote the manuscript. C-CY, Y-HC, and H-CT performed the field work and contributed to data collection and analysis. L-FF, C-CH, W-JH, Y-YS, and KS conceived the study and contributed to data interpretation. G-CG, H-YC, and C-KS contributed to the data collection and interpretation. All authors read and approved the final manuscript.