Organic carbon mineralization, stock, and sequestration in an unvegetated mud flat and adjacent Suaeda japonica marsh of Garolim Bay, Yellow Sea: insights from δ 13 C–DIC

1 Introduction

Coastal wetlands are among the most biologically productive ecosystems globally and play a pivotal role in mitigating climate change by sequestering “blue carbon” (Alongi, 2020; Duarte et al., 2005; Kirwan and Megonigal, 2013; Pendleton et al., 2012). Although these areas account for only a small fraction (<0.01%) of the global surface area, their organic carbon (OC) burial rates are approximately 50-fold higher than those of mature terrestrial forest ecosystems (McLeod et al., 2011). These carbon stocks reflect the net balance between autochthonous and allochthonous carbon inputs and losses through microbial OC mineralization in sediment (Macreadie et al., 2017; Spivak et al., 2019; Alongi, 2020). Although blue carbon assessments have traditionally focused solely on OC burial, recent studies emphasize that a comprehensive understanding of net ecosystem carbon dynamics—and, more broadly, of the biogeochemical linkages across the land–sea continuum (Duarte et al., 2013; Krauss et al., 2018)—requires quantifying mineralization losses as well (Macreadie et al., 2017; Spivak et al., 2019). In particular, the isotopic signatures of accumulated OC and the dissolved inorganic carbon (DIC) produced via anaerobic mineralization remain poorly constrained in temperate coastal wetlands (Middelburg et al., 1997; He et al., 2010; Xia et al., 2021; Eagle et al., 2025).

The carbon and nitrogen contents and stable carbon isotope composition (δ¹³C-TOC) of sedimentary OC, combined with C:N ratios, are widely used as robust tracers for quantifying the relative contributions of multiple organic matter sources in marine sediments, including allochthonous marine- and terrestrial-derived materials (Kubo and Kanda, 2017; Lamb et al., 2006; Bouillon et al., 2008; Thornton and McManus, 1994). However, because these source signatures can overlap substantially in coastal wetlands, bulk sedimentary isotopes constrain the origin of the preserved, largely refractory OC pool but reveal little about which fraction is actively being degraded. The DIC produced by microbial OC mineralization offers a complementary constraint on this active C cycling (Jorgensen, 1982; Kristensen et al., 2008; Eagle et al., 2025). Waterlogged, coastal wetland sediments restrict oxygen availability, inhibiting aerobic decomposition and instead favoring anaerobic microbial pathways (Froelich et al., 1979). Aerobic and anaerobic decomposition processes ultimately produce DIC (CO₂, HCO₃⁻, and CO₃²⁻), which can either be released to the overlying water as CO₂ or sequestered as authigenic carbonates in sediments (Bouillon et al., 2008; Maher et al., 2015). The stable carbon isotope composition of DIC (δ¹³C-DIC) reflects the sources of mineralized organic matter and can reveal the relative contributions of autochthonous versus allochthonous carbon to microbial respiration (Alling et al., 2012; Miyajima et al., 2009). Nevertheless, the application of δ¹³C–DIC as a diagnostic tool for resolving organic matter sources in vegetated tidal wetland sediments remains limited, particularly across contrasting seasonal conditions (Bouillon et al., 2012). Such combined isotopic approaches—tracing both the preserved OC pool and the actively mineralized carbon pool within the same system—have proven effective in river-influenced marginal sediments (e.g., Fu et al., 2022). However, they have rarely been used to directly compare a vegetated salt marsh with an adjacent unvegetated tidal flat.

Suaeda japonica is among the most widely distributed halophytes along the west and south coasts of Korea, forming dominant monospecific stands in muddy intertidal flats at ecologically significant sites such as Ganghwa Island, Garolim Bay, and Suncheon Bay (Hwang et al., 2025). Nationally, S. japonica communities cover approximately 239,873 m², accounting for 13.2% of the total salt marsh area—the second largest after Phragmites australis communities (Lee et al., 2019). Salt marshes dominated by S. japonica are recognized as significant blue carbon ecosystems, with primary productivity and efficient transfer of photosynthetically fixed carbon to belowground biomass, contributing to organic burial in intertidal sediments (Hwang et al., 2025). Unvegetated tidal flats occupy a far greater cumulative area along Korea’s west coast than vegetated salt marshes and directly adjoin S. japonica stands at sites such as Garolim Bay, offering a naturally paired landscape for comparison. Unlike salt marshes, mangroves, and seagrasses already recognized as blue carbon, unvegetated tidal flats gained international recognition only recently (Chen and Lee, 2022), when the IPCC’s 63rd Plenary Session approved the outline of a methodology report including tidal flats as a candidate blue carbon category (IPCC, 2025). This underscores the need to directly compare carbon burial and mineralization dynamics between S. japonica marshes and adjacent unvegetated flats. Accordingly, the primary objectives of this study were to (1) identify the origins of OC stored in tidal flat sediments, (2) characterize the isotopic composition of the DIC produced by mineralization and its relationship to bulk sedimentary organic carbon, and (3) compare carbon preservation and mineralization processes between an S. japonica salt marsh and adjacent unvegetated tidal flat, thereby establishing site-specific OC budgets for both habitats to inform blue carbon assessment in Garolim Bay.

2 Materials and methods

2.1 Study area

Garolim Bay, located on the west coast of Korea, is characterized by a macrotidal regime with tidal ranges varying from 2.86 m during neap tide to 6.39 m during spring tide. Unlike eutrophic estuarine systems that receive high nutrient and organic matter loads from riverine inputs, Garolim Bay is not connected to major rivers and exchanges materials primarily through a tidal inlet opening toward the Yellow Sea. The intertidal flats occupy approximately 77% of the bay’s total area, providing an ideal natural laboratory for investigating intrinsic carbon cycling processes (An et al., 2025a). Among the dominant halophytes, S. japonica (SJ) forms extensive monospecific stands across the upper intertidal zone. With its root system (up to 10 cm) and high annual productivity, SJ enhances OC burial through belowground biomass accumulation and creates strong biogeochemical gradients in sediment redox conditions (Hyun et al., 2009; An et al., 2024).

2.2 Sampling and handling

Samples were collected during the non-growing (April) and growing (August) seasons. Sediment samples for biogeochemical analysis and metabolic rate measurements were collected at two contrasting sites: (1) sediments inhabited by the S. japonica (SJ); and (2) unvegetated mud flat (UMF) sites (Figure 1). At each site, sampling was conducted randomly near plant shoots or burrows at overall sites. Triplicate sediment samples were collected from each site for geochemical analysis using acrylic cores (40 cm length, 6 cm internal diameter) that were immediately sealed with butyl rubber stoppers. The samples were preserved in icebox until processed for pre-treatment. Sediment pore-water samples were extracted using a Rhizon sampler (Rhizon CSS; Rhizosphere Research Products B.V., Wageningen, The Netherlands) at 1-cm depth intervals through predrilled and taped holes in the core (Seeberg-Elverfeldt et al., 2005). Samples from adjacent depths were subsequently combined to obtain composite samples representing 0–2-, 2–5-, 5–10-, 10–15-, 15–20-, and 20–30-cm depth intervals. Pore-water samples were collected after discarding dead volume and rinsing the syringe with pore-water. To measure NH₄⁺ and PO₄^3- concentrations, aliquots of 1–2 mL of pore-water were frozen until analysis. To determine dissolved Fe²⁺ and SO₄^2-, 2 mL of pore-water was preserved with 12 N HCl (final conc. 0.1 N) and then stored at 4 °C. The pore-water samples for dissolved sulfide (H₂S) were precipitated with a zinc acetate solution (20%) and kept frozen. To prevent microbial activity, the dissolved OC (DOC) samples were acidified to a pH< 2 with 6 M HCl and preserved at 4 °C until further analysis. For total DIC analysis, 1.8-mL aliquots were collected in glass vials without headspace and preserved with 18 μL of saturated HgCl₂ (125 mM). Sediment samples for measuring total OC (TOC), total nitrogen (TN), and particle size content were sectioned at 0–2-, 2–5-, 5–10-, 10–15-, 15–20-, and 20–30-cm intervals and stored in a deep freezer (−20 °C) until laboratory processing.

2.3 Pore-water and solid-phase analysis

Dissolved NH₄⁺ and PO₄^3- in pore-water were measured using a continuous-flow nutrient analyzer (QuAAtro; Seal Analytical GmbH, Norderstedt, Germany); the reproducibility was generally within ±7%, and MOOS-3 (National Research Council Canada, Ottawa, ON, Canada) was used as a certified reference material. Dissolved Fe²⁺ in pore-water was determined using colorimetry with a ferrozine solution (detection limit: 1 μM; Stookey, 1970). SO₄^2- concentration was determined using ion chromatography (detection limit: 0.03 mM; 819 conductivity detector using an A-Supp5 column; Metrohm, Herisau, Switzerland). H₂S in the pore-water was quantified using the colorimetric methylene blue method (detection limit: 3 μM; Cline, 1969). Dissolved Ca²⁺ concentration was analyzed by inductively coupled plasma atomic emission spectrometry (ICP-AES, Optima 3300DV, PerkinElmer Co.).

DIC concentration and stable carbon isotope composition (δ¹³C–DIC) were determined using a Thermo Fisher GasBench II system (Thermo Fisher Scientific, Waltham, MA, USA) coupled to a Delta V Plus isotope ratio mass spectrometer (IRMS; Thermo Fisher Scientific). Pore-water samples (1–2 mL) were transferred to pre-evacuated, He-flushed 12-mL Exetainer vials. Thereafter, five drops of anhydrous phosphoric acid (H₃PO₄, 99%) were added to convert all DIC species to CO₂, and the vials were equilibrated at room temperature (25 °C) for 24 h. The evolved CO₂ was collected using a double-needle system and separated on a PoraPLOT Q fused silica column (25 m × 0.32 mm; Agilent Technologies, Santa Clara, CA, USA) before entering the IRMS. δ¹³C–DIC values were calibrated against NBS-19 and an internal lithium carbonate standard, and reported relative to Vienna Pee Dee Belemnite (VPDB). Analytical precision was ±0.1‰ (1σ). DIC concentrations were determined by comparing total peak areas (m/z 44, 45, 46) to a calibration curve generated from NaHCO₃ solutions.

Pore-water DOC concentrations were determined using a TOC-V_CPH analyzer (Shimadzu, Kyoto, Japan) and quality controlled using carbon reference material (Hansell Lab, U. Miami, RSMAS). Calibration standards were prepared by diluting a certified 100 mg L^-1 TOC stock solution (Sigma-Aldrich, St. Louis, MO, USA) with ultra-pure water. To determine TOC and TN in sediment samples, 1 g of freeze-dried and homogenized sediment was placed in a 15-mL tube, followed by the addition of 1 M HCl to remove inorganic carbon. The stable isotope analysis of TOC and TN was conducted using an elemental analyzer (Vario ISOTOPE Cube; Elementar, Langenselbold, Germany) coupled with an IRMS (Isoprime VisION IRMS; Elementar, Manchester, UK). The measured stable isotope ratios were expressed as’δ’ values, calculated using the following formula (Equation 1):

(1) δ13C=[(Rsample/Rstandard)−1]×1,000,

where R = ¹³C/¹²C for carbon isotopes. VPDB was used as the standard material for δ¹³C stable isotope calculations. The IAEA reference standard CH-3 (δ¹³C,−24.72‰ ± 0.1‰) was analyzed after every five samples to ensure analytical accuracy. Analytical error was confirmed to be within ±0.1‰ for δ¹³C. Total carbon (TC) was determined on untreated subsamples using the same elemental analyzer, and the sediment carbonate content was calculated as (TC - TOC) × 8.33, the stoichiometric ratio of CaCO3 to carbon. Chlorophyll a (Chl-a) concentrations in surface sediment (0–2 cm) were determined by extracting samples with 90% acetone for 24 h and measuring the extracts spectrophotometrically according to Parsons et al. (1984).

2.4 Rates of anaerobic OC mineralization

To determine anaerobic OC mineralization rates (OC_min), sediment from a depth of 0 to 30 cm was collected using acrylic cores (40 cm length, 10 cm internal diameter) that were immediately sealed with butyl rubber stoppers. Sediment cores were transferred to an N₂-filled glove bag and sliced at 0–2-, 2–5-, 5–10-, 10–15-, 15–20-, and 20–30-cm intervals. Subsequently, the sediments were homogenized, loaded into 50-mL centrifuge tubes, and incubated at in-situ temperatures in the dark. At regular intervals, two tubes were sacrificed, and pore-water samples were extracted by centrifugation at 3,500 rpm for 10 min. After reintroduction to the N₂-filled glove bag, supernatant pore-water was sampled for various chemical analyses and filtered through a 0.2-μm cellulose acetate syringe filter (Advantec; Toyo Roshi Kaisha, Ltd., Tokyo, Japan). DIC concentrations and δ¹³C–DIC were analyzed using the method described in Section 2.3; values at each time point are means of the duplicate tubes. Anaerobic OC_min was calculated from the linear regression of DIC concentration against incubation time. Because carbonate dissolution and precipitation alter pore-water DIC independently of OC_min, dissolved Ca²⁺ was measured in the 0–2-cm interval at the same time points as DIC. The change in carbonate carbon was calculated as ΔCaCO₃ = Δ[Ca²⁺]sol × (1 + K_Ca), where K_Ca = 1.6 is the adsorption constant for Ca²⁺ (Li and Gregory, 1974), following Thamdrup et al. (2000) and Hyun et al. (2017). Those authors applied this expression to correct DIC accumulation upward for authigenic carbonate precipitation, where dissolved Ca²⁺ declines during incubation. In the present study Ca²⁺ increased, so the expression was instead used to quantify the carbonate-derived contribution to measured DIC production.

2.5 Sedimentation rate

Sediment cores for ²¹⁰Pb dating were collected separately at SJ and UMF and sectioned at 5-cm intervals down to a 50-cm depth. Total ²¹⁰Pb and ²²⁶Ra activities were determined using gamma spectroscopy at the Korea Basic Science Institute (Daejeon, Republic of Korea). Excess ²¹⁰Pb (²¹⁰Pb_ex) at each depth interval was calculated as the difference between total ²¹⁰Pb and 226Ra activity, with uncertainty propagated from both measurements. The cumulative ²¹⁰Pb_ex inventory from the sediment surface to the base of each interval, A(z), was calculated as the depth-integrated sum of ²¹⁰Pb_ex × dry bulk density × interval thickness, and the total inventory of the core, A_total, was taken as the cumulative inventory at the base of the deepest interval. The age of each interval, t(z), was then calculated using the constant rate of supply (CRS) model (Equation 2) (Appleby and Oldfield, 1983; Sanchez-Cabeza and Ruiz-Fernández, 2012):

(2) t(z)=(1/λ)×ln[Atotal/(Atotal−A(z))],

where λ is the decay constant of ²¹⁰Pb (0.03114 year^-1, half-life 22.3 years). Because CRS ages diverge as the residual inventory approaches zero, intervals with a residual inventory below 1% of Atotal were excluded; applied identically at both sites, this removed one interval at UMF and none at SJ. Sedimentation rates (SR, cm year^-1) were then estimated as the slope of a linear regression of interval depth on CRS age, with the standard error of the slope as the reported uncertainty. Both cores showed a subsurface ²¹⁰Pb_ex maximum, indicating a bioturbated mixed layer within which CRS ages should not be read as true depositional ages (Appleby, 2001; Arias-Ortiz et al., 2018); the sensitivity of the derived rates to excluding these intervals, and to three alternative exclusion criteria, was therefore evaluated. ¹³⁷Cs was not measured, so the chronologies rest on ²¹⁰Pb alone.

2.6 OC stock and sequestration rate

OC stock per unit area was calculated for each sediment core following the standard method of Howard et al. (2014), as previously applied to Korean tidal flats including Garolim Bay (Lee et al., 2021). For each depth interval, the amount of carbon in the core section (g C cm^-2) was calculated as follows (Equations 3, 4):

(3) Dry bulk density(DBD)(g cm−3)=Mass of dry sediment(g)Volume sampled sediment(cm−3)

(4) OCsto(Mg C ha−1)=DBD(g cm−3)×TOC(g C g−1)×sediment thickness interval(cm)×100

Carbon amounts for all depth intervals (0–2, 2–5, 5–10, 10–15, 15–20, 20–25, and 25–30 cm) were summed to yield the TOC stock per unit area for each core (g C cm^-2), which was then converted to Mg C ha^-1 using a conversion factor of 0.01 (1 Mg = 10⁶ g; 1 ha = 10⁸ cm²). Because core lengths in the present study (30 cm) did not reach the 1-m depth commonly used as a standard for regional and global comparison, sedimentary OC stocks were additionally normalized to a 1-m-equivalent value following the approach of Lee et al. (2021).

OC sequestration rate (OC_seq, g C m^-2 year^-1) was calculated for each site following the standard approach of Xiaonan et al. (2008) and Lee et al. (2021) as follows (Equation 5):

(5) OCseq(g C m−2 year−1)=SR(cm year−1)×DBD(g cm−3)×TOC(%)/100×104,

where SR is the ²¹⁰Pb-derived sediment accumulation rate, DBD is dry bulk density averaged over the surface sediment interval, and TOC is OC content, with 10⁴ the unit conversion factor from cm² to m².

2.7 Plant biomass

Above- and below-ground biomass (AGB and BGB, respectively) and shoot density of S. japonica were determined by harvesting all plant material within randomly selected 25 cm × 25 cm quadrats (n = 3 on each sampling date). Plant shoots for determining AGB were cut at the sediment surface, and the sediment for estimating BGB was transferred to the laboratory. To remove visible stones and residues, plant material and detritus were carefully washed with tap water and subsequently rinsed twice with distilled water. Below-ground tissues were gently washed with distilled water under gentle flow to avoid the loss of fine roots. Plant material was then separated into leaves, stems, and below-ground material, and dried at 60 °C for >48 h to determine dry weight.

2.8 Data analysis

Statistical analyses were performed in Python (scipy, statsmodels, pingouin, scikit-posthocs). Because the depth intervals sampled from a given core are not independent observations, sediment and pore-water constituents were analyzed with a split-plot design: season and site were treated as between-core factors and tested against the among-core error term, and depth interval and its interactions as within-core factors, with core identity as a random effect. Residual normality (Shapiro–Wilk) and homogeneity of variance (Levene’s test) were assessed for each variable, and pairwise comparisons among the four site-season groups used Tukey’s HSD applied to core-level means, so that the post-hoc tests share the error stratum of the corresponding F test. Between-core effects are reported in the pore-water and solid-phase constituents and within-core (depth) effects in Supplementary Table S1. Source signatures of the DIC added during incubation were estimated with Keeling plots, in which measured δ¹³C-DIC is regressed against the reciprocal of DIC concentration, the y-intercept giving the isotopic composition of the added carbon source. Regressions were fitted by ordinary least squares separately for each depth interval of each site-season combination (five time points per regression), with the regression standard error taken as the uncertainty of each intercept. Depth-specific intercepts were combined using a weighted mean, each interval weighted by the DIC produced over the incubation, so that the integrated signature reflects the layers dominating depth-integrated mineralization. Individual regressions are shown in section 3.5 and their statistics in Supplementary Table S2; one interval yielding a positive intercept incompatible with any plausible carbon source was excluded. The offset was defined as the difference between the intercept and the mean bulk sediment δ¹³C-TOC of the same site and season, positive values indicating DIC isotopically enriched relative to bulk sedimentary organic carbon. Significance was set at p < 0.05.

3 Results

3.1 Sediment properties and biomass

The physical and ecological properties examined in this study are summarized in Table 1 and Supplementary Figure S1. Sediment temperature was 15 °C in April and 27 °C in August. Seasonal biomass data revealed a pronounced summer development of S. japonica at the vegetated site. The AGB increased from 0.30 ± 0.15 g m^-2 in April to 44.2 ± 33.4 g m^-2 in August, representing an approximately 146-fold increase. BGB increased from 0.04 ± 0.04 to 4.38 ± 2.57 g m^-2, corresponding to an about 110 fold increase over the same period. In contrast, shoot density decreased sharply from 672 ± 385 m^-2 in April to 64.1 ± 16.1 m^-2 in August, indicating that seasonal biomass accumulation reflected growth and thickening of individual plants rather than an increase in shoot abundance. Depth-integrated Chl-a (0–2 cm) was more than twice as high at SJ (170 ± 0.96 mg m^-2) as at UMF (87.4 ± 1.15 mg m^-2) in August, whereas in April the two sites differed by less than 20% (110 ± 1.25 mg m^-2 at UMF and 131 ± 18.3 mg m^-2 at SJ), with the variability among subsamples at SJ exceeding the difference between sites.

Table 1. Physicochemical properties, benthic microalgal biomass, and plant biomass at the unvegetated mud flat (UMF) and the Suaeda japonica marsh (SJ) in April and August.

Temperature, porosity, dry bulk density, and sediment type refer to the 0–30-cm depth interval; chlorophyll a refers to the 0–2-cm surface layer; and above-ground biomass (AGB), below-ground biomass (BGB), and shoot density were determined from quadrat sampling at SJ. Values are means ± 1 SD.

Values are presented as mean ± 1 SD. AGB, above-ground biomass; BGB, belowground biomass.

In sediments, mean porosity and DBD ranged from 0.51 to 0.59 and from 1.31 to 1.41 g cm^-3, respectively. Porosity, DBD, and water content showed little variation with depth. Across all cores, both habitats were dominated by fine-grained sediments, with silt comprising the largest fraction throughout the profiles in both seasons. Silt content generally ranged from 71.3 to 79.7%, whereas clay accounted for 8.9–17.1%, sand for 6.6–16.2%, and gravel was negligible or absent in most layers (Supplementary Figure S1).

3.2 Pore-water constituents

Pore-water concentrations of NH₄⁺, PO₄^3-, Fe²⁺, SO₄^2-, H₂S, DOC, DIC, and δ¹³C-DIC, measured throughout the 0–30-cm depth profile, are summarized in Table 2, and their depth distributions are shown in Figures 2, 3. Vertically, NH₄⁺ concentrations remained low and relatively uniform in the upper layers at both sites in April, whereas in August they increased sharply below ~10 cm at SJ. NH₄⁺ concentrations were significantly higher in August (36.2 ± 35.9 μM at UMF; 32.9 ± 34.3 μM at SJ) than those in April (21.0 ± 26.3 μM at UMF, 4.85 ± 4.11 μM at SJ) (F_1,8 = 70.9, p< 0.001), and also differed between sites (F_1,8 = 14.3, p = 0.005), with a significant season × site interaction (F_1,8 = 6.25, p = 0.037); SJ exhibited distinctly lower values in April than those of the other three site–season combinations. PO₄^3- concentrations at both sites generally increased with depth in the upper layers before decreasing toward the base of the cores. PO₄^3- concentrations were significantly higher in April (2.56 ± 2.12 μM at UMF, 3.05 ± 1.27 μM at SJ) than in August (1.66 ± 1.81 μM at UMF, 1.58 ± 1.03 μM at SJ) (F_1,8 = 11.3, p = 0.010), with no significant site effect (F_1,8 = 0.34, p = 0.58). Fe²⁺ concentrations generally increased with depth and exhibited subsurface maxima in deeper layers at both sites in both seasons. The Fe²⁺ concentration peak for SJ was most pronounced in the 2–6-cm layer in August. Fe²⁺ concentrations were significantly higher at UMF than at SJ (F_1,8 = 6.30, p = 0.036) but did not differ between seasons (F_1,8 = 2.35, p = 0.16). SO₄^2- levels were significantly higher in August than in April at both sites (F_1,8 = 149, p< 0.001), whereas the site effect was not significant (F_1,8 = 0.005, p = 0.95). Vertically, SO₄^2- concentrations remained largely uniform with depth, with a modest elevation in the root zone (2–10 cm) at SJ in August. H₂S levels were below the detection limit throughout the study period and are therefore not reported in Table 2.

Table 2. Mean concentration of the pore-water and solid-phase constituents.

Values represent the mean ± 1 SD of all depth intervals and replicate cores. F statistics are from a split-plot analysis in which season and site were treated as between-core factors, tested against the among-core error term, and depth interval and its interactions were treated as within-core factors, with core identity as a random effect (Section 2.8). All variables were determined on three replicate cores per site and season, giving degrees of freedom of 1, 8 for each between-core effect. Different superscript lowercase letters in the same column indicate significant differences among the four site–season groups (Tukey’s HSD applied to core-level means, p< 0.05). DOC, dissolved organic carbon; DIC, dissolved inorganic carbon; TOC, total organic carbon; TN, total nitrogen. Asterisks indicate significance as follows: *p< 0.05 and **p< 0.01.

Pore-water DOC was significantly higher at SJ (450 ± 134 μM in April, 565 ± 105 μM in August) than at UMF (352 ± 89.7 μM in April, 478 ± 96.6 μM in August) (F_1,8 = 78.0, p< 0.001) and higher in August than in April (F_1,8 = 133, p< 0.001), and generally exhibited a slight increase with depth or remained stable within each core, without a pronounced subsurface maximum. DIC concentrations ranged from 2.10–3.24 mM (2.50 ± 0.34 mM) at UMF and 1.61–2.57 mM (2.19 ± 0.22 mM) at SJ in April, and from 2.79–4.72 mM (3.53 ± 0.57 mM) at UMF and 3.38–6.15 mM (4.73 ± 0.97 mM) at SJ in August, increasing with depth at both sites, particularly in August. Pore-water δ¹³C-DIC ranged from −6.84‰ to −2.65‰ (−3.82 ± 1.14‰) at UMF and from −5.86‰ to −1.51‰ (−3.80 ± 1.32‰) at SJ in April. In August, values ranged from −7.30‰ to −3.62‰ (−5.80 ± 0.845‰) at UMF and from −11.8‰ to −6.96‰ (−10.2 ± 1.17‰) at SJ, becoming progressively depleted with depth—most markedly at SJ in August, where values below −11‰ were recorded in the deepest layers. Both DIC and δ¹³C-DIC differed significantly by season, site, and their interaction (DIC: F_1,8 = 604, 37.8, and 109; δ¹³C-DIC: F_1,8 = 359, 98.8, and 100; all p< 0.001; Table 2). Because depth intervals were nested within cores, depth and its interactions were tested within the core stratum (Section 2.8). Depth was a significant source of variation for all pore-water constituents (F = 2.65-48.6, p< 0.05), and the three-way depth × season × site interaction was significant for NH₄⁺, PO₄^3-, Fe²⁺, SO₄^2-, DIC, and δ¹³C-DIC (F_6,40 = 3.39-14.4, all p< 0.01; Supplementary Table S1), indicating that the vertical structure of pore-water chemistry differed between sites in a season-dependent manner rather than being uniformly displaced.

3.3 Vertical distribution of TOC, TN, CaCO₃, C:N ratio, and δ¹³C-TOC

The vertical distributions of TOC, TN, C:N ratio, and δ¹³C-TOC were examined in sediments from the UMF and SJ sites in April and August (Figure 3; Table 2). Overall, TOC contents were generally higher at the SJ site than at the UMF site in both seasons (0.48 and 0.58% at SJ vs. 0.42 and 0.47% at UMF; site F_1,8 = 33.6, p< 0.001) and higher in August than in April (F_1,8 = 29.3, p< 0.001), with no significant interaction (F_1,8 = 2.18, p = 0.18); TOC also varied significantly with depth (F_6,48 = 10.6, p< 0.001), and its vertical distribution differed between sites in a season-dependent manner (depth × season × site, F_7,48 = 3.87, p = 0.002; Supplementary Table S1). The highest TOC value was observed in the 0–2-cm layer in August at the SJ site (0.73%). The vertical profiles of TOC and TN showed slight variations with depth, without a distinct down-core trend. The C:N ratio and δ¹³C-TOC were both higher in August than in April (F_1,8 = 9.46 and 9.14, p = 0.015 and 0.016), with no significant site effect or interaction (all p > 0.07; Table 2). The C:N ratio (9.25 and 9.75 at UMF; 9.33 and 11.2 at SJ) and δ¹³C-TOC (-21.9‰ and -21.6‰ at UMF; -21.7‰ and -21.6‰ at SJ) varied little with depth in absolute terms, although both showed significant depth × site interactions (F_6,48 = 9.34 and 4.58, both p< 0.001; Supplementary Table S1). TN did not differ significantly by site or season (both p > 0.08). Sediment CaCO₃ content was low throughout both cores and at all depths, ranging from 0.011 to 1.96%, with no sample exceeding 2% (Figure 3). CaCO₃ was higher in April (1.01 ± 0.37% at UMF and 1.02 ± 0.26% at SJ) than in August (0.24 ± 0.20% at UMF and 0.22 ± 0.18% at SJ; season F_1,8 = 436, p< 0.001), with no difference between sites (F_1,8 = 0.017, p = 0.90) and no interaction (F_1,8 = 0.21, p = 0.66), and its vertical distribution differed between sites in a season-dependent manner (depth × season × site, F_7,48 = 14.6, p< 0.001; Supplementary Table S1).

3.4 Sedimentation rates, OC stock, and sequestration rates

Sedimentation rates (SR), OC stocks (OC_sto), and OC sequestration rates (OC_seq) are summarized in Table 3. The ²¹⁰Pb-derived sedimentation rate was 0.37 ± 0.02 cm year^-1 at UMF and 0.47 ± 0.03 cm year^-1 at SJ. Both cores showed a subsurface ²¹⁰Pb_ex maximum at 7.5 cm (UMF) and 12.5 cm (SJ); restricting the regression to intervals below these maxima lowered the rates by 6% and 18% respectively without altering the rank order between sites, and rates derived under four alternative exclusion criteria spanned only 0.376–0.409 cm year^-1 at UMF and 0.461–0.472 cm year^-1 at SJ. At UMF, ²¹⁰Pb_ex declined to the supported level at the base of the core (0.4 ± 8.2 Bq kg^-1), whereas at SJ it remained detectable (30.4 ± 9.3 Bq kg^-1), indicating that part of the inventory lies below the sampled interval; the rate reported for SJ is therefore a lower bound (Section 4.5; Supplementary Figure S2). Normalized to a 1-m equivalent depth (Section 2.6), OC stocks were 53.8 ± 0.61 and 60.5 ± 0.63 Mg C ha^-1 at UMF and SJ in April, and 60.4 ± 0.98 and 82.9 ± 1.12 Mg C ha^-1 at UMF and SJ in August. Stocks were higher at SJ in both seasons, and the difference widened from 12% in April to 37% in August, giving seasonal means of 57.1 ± 4.67 Mg C ha^-1 at UMF and 71.7 ± 15.8 Mg C ha^-1 at SJ. OC_seq was 20.5 ± 2.29 and 29.9 ± 0.82 g C m^-2 year^-1 at UMF and SJ in April, and 22.6 ± 4.11 and 40.0 ± 5.71 g C m^-2 year^-1 at UMF and SJ in August. Seasonal means were 21.6 ± 1.48 g C m^-2 year^-1 at UMF and 35.0 ± 7.21 g C m^-2 year^-1 at SJ, the latter approximately 62% higher.

Table 3. Sedimentation rates (SR), organic carbon stock (OC_sto), sequestration rate (OC_seq), and mineralization rate (OC_min) at the unvegetated mud flat (UMF) and the Suaeda japonica marsh (SJ).

For individual seasons, values are given as mean ± 1 SD; values in the “Ave.” rows are the mean ± 1 SD of the two seasonal values.

*Organic carbon stock normalized to the amount at a 1-m depth.

3.5 OC mineralization rates and δ¹³C-DIC

At both sites and in all incubations, pore-water DIC concentrations increased linearly with incubation time, demonstrating continuous mineralization of sedimentary OC (Figures 4; Supplementary Figure S3; Supplementary Table S3). In August, OC_min at the SJ site was highest in the surface sediment (0–2 cm, 2,312 nmol C cm^-3 day^-1), and then decreased sharply with depth to 65 nmol C cm^-3 day^-1 at 20–30 cm. At the UMF site, OC_min in August peaked in the 2–5-cm layer (378 nmol C cm^-3 day^-1) rather than at the surface (298 nmol C cm^-3 day^-1), declined to a minimum at 5–10 cm (65.5 nmol C cm^-3 day^-1), and remained between 91 and 144 nmol C cm^-3 day^-1 below that depth (Figure 4). The surface-layer (0–2 cm) OC_min at SJ in August was approximately 7.8-fold higher than that at UMF (298 nmol C cm^-3 day^-1), indicating a pronounced summer hotspot of OC_min confined to the rooted surface sediment at SJ. In April, OC_min was similar between the two sites (Figure 4). Depth-integrated OC_min down to a 30-cm sediment depth was approximately 1.3–2.2-fold greater at SJ (21.8 and 94.7 mmol m^-2 day^-1 in April and August, respectively) than at UMF (16.4 and 43.4 mmol m^-2 day^-1) across both seasons (Supplementary Table S3). Dissolved Ca²⁺ in the 0–2-cm incubations increased with time at both sites in both seasons, so no correction for authigenic carbonate precipitation was required; none of the four regressions was significant (r² = 0.16-0.54, p = 0.16-0.50), reflecting increases that were small relative to the analytical uncertainty of the measurements (Supplementary Figure S4). Taking the fitted increases at face value, the carbonate-derived contribution to depth-integrated DIC production was 7.9% and 5.3% at UMF and SJ in April, and 0.8% and 1.0% at UMF and SJ in August, with 95% upper confidence limits of 20.5%, 10.9%, 3.3%, and 2.1%, respectively. Because Ca²⁺ was measured only in the surface interval, these depth-integrated figures assume that carbonate dissolution is confined to that layer. Applying the surface fraction uniformly with depth instead would raise the August contribution to 5.9% at UMF and 2.1% at SJ.

The associated δ¹³C–DIC data showed a systematic depletion of ¹³C in DIC with incubation time at both sites in both seasons (Supplementary Figure S3), with δ¹³C–DIC values ranging from approximately −2.0‰ to −16.0‰ across depths and seasons. Keeling plot regressions were performed separately for each of the 24 depth intervals (Figure 5). Sixteen were significant at p< 0.05 (R² = 0.71–0.99), whereas eight were not (R² = 0.09–0.69, p = 0.07–0.63), the latter concentrated in the deeper intervals where the DIC increase over the incubation was small (Supplementary Table S2); non-significant regressions occurred in three of six intervals at UMF in each season, compared with one of six at SJ. The 20–30-cm interval at SJ in April yielded a positive intercept (+2.13‰), incompatible with any plausible carbon source, and was excluded. Combined across depths using weights proportional to the DIC produced in each interval (Section 2.8), intercepts were −10.52 ± 1.66‰ at UMF and −14.21 ± 5.80‰ at SJ in April, and −12.16 ± 1.58‰ at UMF and −16.32 ± 2.24‰ at SJ in August, corresponding to offsets from bulk δ¹³C-TOC of +11.38‰, +7.49‰, +9.44‰, and +5.28‰, respectively (Table 4; Figure 5). Intercepts were significantly higher at UMF than at SJ in August (t = 5.05, df = 10, p< 0.001). The weighted intercepts were higher at UMF than at SJ in August by 4.16‰, whereas in April the two sites differed by 3.69‰ with a much wider spread among the depth-specific intercepts at SJ (± 5.80‰ against ± 1.66‰ at UMF). Because the incubations were conducted on a single core set per site and season, the depth-specific intercepts are not independent replicates of the site, and we therefore describe this contrast rather than test it inferentially.

Table 4. Keeling plot intercepts for the DIC added to pore water during incubation, by site and season.

Intercepts of the individual depth intervals were combined using a weighted mean, each interval weighted by the DIC produced over the incubation (Section 2.8); values are given as weighted mean ± 1 weighted SD. Regression statistics for all 24 depth-specific intervals are given in Figure 5; Supplementary Table S2. The offset is the difference between the intercept and the mean bulk sediment δ¹³C-TOC of the same site and season, and is interpreted as an aggregate index of divergence between newly produced DIC and bulk sedimentary organic carbon rather than as the signature of a single source (Section 4.3).

^aDIC-production-weighted mean ± weighted SD of Keeling plot intercepts determined separately for each of the six depth intervals (0–2, 2–5, 5–10, 10–15, 15–20, 20–30 cm; five incubation time points per interval; individual regressions shown in Figure 5; Supplementary Table S2).

^bMean ± SD of δ¹³C–TOC across all sampled depths at that station/season.

4 Discussion

4.1 Source of sedimentary organic matter inferred from C:N and δ¹³C–TOC

C:N ratios and δ¹³C–TOC are widely used to trace organic matter sources in coastal sediments, as autochthonous and allochthonous end-members, including marine and freshwater algae, as well as C₃ and C₄ vascular plants, occupy characteristically distinct regions of C:N–δ¹³C space (Deines, 1980; Peterson and Fry, 1987; Lamb et al., 2006). Plotted against representative end-members compiled by Lamb et al. (2006), most samples from both the UMF and SJ sites (56.0%) fell within the marine algae field, with none falling within the freshwater algae or freshwater POC fields; the marine algae and marine POC fields overlap substantially in this parameter space (Lamb et al., 2006), so samples assigned to the marine algal field cannot be distinguished from marine POC on this basis (Figure 6) This pattern is consistent with the absence of a major riverine connection in Garolim Bay and with prior interpretations that C:N and δ¹³C–TOC values of approximately −22‰ in Korean intertidal sediments primarily reflect marine rather than freshwater or terrestrial input (He et al., 2010; Lee et al., 2014, 2019).

Notably, 23.8% of samples fell within the C₃ terrestrial plant range (C:N ≥ 12; Deines, 1980; Lamb et al., 2006), while none approached the C₄ field despite the capacity of Suaeda species to utilize both C₃ and C₄ processes to fix carbon (Shomer-Ilan et al., 1975; Fisher et al., 1997; Hwang et al., 2025) (Supplementary Table S4). This proportion was lowest at SJ in April (9.5%), when standing S. japonica biomass had not yet developed (Section 3.1), rose sharply to 42.9% at SJ in August, and remained comparatively stable and moderate at UMF across both seasons (19.0% in April, 23.8% in August), arguing against a signal confined to the vegetated site alone. Two non-exclusive processes likely explain this: in-situ input of fresh S. japonica detritus and roots during its August biomass peak (Section 3.1), and tidal redistribution of senesced marsh litter onto UMF, a process documented in other macrotidal marshes (Haines, 1977; Wang et al., 2018). Because C:N ratios are highly sensitive to small changes in TN (Sampei and Matsumoto, 2001; Lamb et al., 2006), and TN did not differ significantly by site or season in the present study (both p > 0.08; Table 2), some of the variability around the C:N threshold of 12 may reflect instability in the indicator itself rather than a true source shift. The remaining 20.2% of samples fell outside all six fields, likely representing intermediate mixtures rather than a distinct source (Lamb et al., 2006). This lack of a clear C₄-like signal parallels Ganghwa, another Yellow Sea wetland, where native C₃ halophyte sediments, including S. japonica, exhibited C₃-like δ¹³C–TOC values (−25.0‰ to −23.2‰), well below the C₄-like signatures (−20.4‰ to −15.8‰) that emerged following invasion by the C₄ grass Spartina alterniflora (Lee et al., 2024).

TOC differed significantly by site and season (site F_1,8 = 33.6, p< 0.001; season F_1,8 = 29.3, p< 0.001; Table 2) and was highest at SJ in August. Given the stability of TN noted above, the seasonal rise in the mean C:N ratio (F_1,8 = 9.46, p = 0.015) is consistent with the TOC-driven shift toward the C₃-classified fraction at SJ in August noted above, rather than a nitrogen change. δ¹³C–TOC varied only modestly overall. It was slightly higher in August than in April (F_1,8 = 9.14, p = 0.016) but did not differ between sites (F_1,8 = 3.67, p = 0.092; Table 2); its vertical distribution also shifted between sampling periods (depth × season, F_7,48 = 10.2, p< 0.001; Supplementary Table S1). The seasonal difference of 0.2-0.3‰ is small relative to the range spanned by the source end-members. Critically, this higher TOC at SJ was not accompanied by a difference in δ¹³C–TOC between sites in either season, indicating that vegetation increased the quantity of stored OC without fundamentally altering its bulk isotopic composition (Lee et al., 2019). Together with the finer sediment texture reported beneath S. japonica (Min, 2005), these results suggest that SJ canopies enhance trapping of marine-derived particles and locally supplied plant detritus from a common, marine-dominated carbon pool, rather than contributing an isotopically distinct autochthonous carbon source.

4.2 Impacts of S. japonica on sediment OC mineralization

Sedimentary OC mineralization is controlled by the availability of electron donors and acceptors, which can be modified by vegetation. The root–rhizome systems of marsh plants translocate both organic substrates and oxidants into the sediment, creating dynamic redox and biogeochemical microenvironments (Kostka et al., 2002; Canfield et al., 2005; Hyun et al., 2007; An et al., 2024). As shown in Section 3.5, depth-integrated OC_min was consistently higher at SJ than at UMF in both seasons, and the seasonal enhancement from April to August was more pronounced at SJ (4.3-fold) than at UMF (2.6-fold). This asymmetry suggests that the summer stimulation of mineralization at SJ is not simply a temperature-driven response shared by both habitats, but is associated with the vegetated site—most plausibly the seasonal development of the S. japonica root system. Root biomass peaks in August (Section 3.1), providing fresh, labile organic substrates and root exudates directly to the sediment (Kostka et al., 2002; Hines et al., 1999). This pattern is broadly consistent with the depth-integrated OC_min reported for S. japonica and unvegetated sediments in the Ganghwa tidal wetland in the Yellow Sea (An et al., 2020, 2024), supporting the generality of vegetation-enhanced mineralization across Korean S. japonica marshes despite differing hydrological settings between the two bays.

Within this overall enhancement, the highest rates were concentrated in the 0–2-cm layer at SJ in August, where surface OC_min exceeded the corresponding UMF rate by approximately 7.8-fold and clearly surpassed rates in the underlying SJ layers. This pronounced surface enrichment is consistent with a root-exudate-driven mechanism, in which labile carbon inputs are supplied preferentially to the sediment surface and rhizosphere rather than being distributed evenly with depth (Kostka et al., 2002; Canfield et al., 2005). Pore-water NH₄⁺ and PO₄^3-, the inorganic products of organic matter remineralization, were both markedly elevated in August relative to April (F_1,8 = 70.9 and 11.3, p< 0.001 and p = 0.010; Table 2), independently corroborating the seasonal intensification of mineralization inferred from the DIC-based rates. This interpretation is supported directly by the depth-resolved statistics: the three-way depth × season × site interaction was significant for NH₄⁺, PO₄^3-, Fe²⁺, SO₄^2-, DIC, and δ¹³C-DIC (F_6,40 = 3.39-14.4, all p< 0.01; Supplementary Table S1), indicating that the vertical structure of pore-water chemistry at the two sites diverged in a season-dependent manner rather than being uniformly displaced. Vegetation therefore appears to restructure the vertical distribution of mineralization rather than to elevate it evenly throughout the profile. The strong vertical gradient in OC_min, with rates decreasing from the surface toward deeper sediment, parallels the distribution of TOC and dissolved solutes. TOC contents peaked in the surface layer at the SJ site in August (Figure 3), and pore-water DOC was significantly higher at SJ than at UMF (F_1,8 = 78.0, p< 0.001; Table 2), consistent with an enhanced supply of reactive organic matter and its rapid conversion to dissolved products in the vegetated zone. This enhanced mineralization was also reflected in the coupled Fe–S geochemistry. Although sulfate concentrations were comparable between sites within each season (Table 2), the average concentration of pore-water Fe²⁺ was significantly lower at SJ than at UMF (F_1,8 = 6.30, p = 0.036), in both April (5.79 vs. 1.66 μM) and August (5.77 vs. 2.72 μM), consistent with more intensive sulfide production and authigenic FeS precipitation beneath the vegetated canopy (Bu et al., 2015; An et al., 2020). Together with observations from Spartina-dominated marshes, where high above- and belowground biomass, as well as extensive root exudation, increase the pool of labile DOC in sediments (Hines, 1991; 1999; Gribsholt et al., 2003; Bu et al., 2015; Zhang et al., 2017; Spivak and Reeve, 2015), these results suggest that vegetation-associated inputs of labile organic matter and root-associated redox processes enhance OC_min at SJ relative to that at UMF, with implications for the net blue carbon balance at the site.

4.3 Insights from ¹³C–DIC into active carbon pools and mineralization

Keeling plots of δ¹³C–DIC against 1/DIC revealed clear two-end-member mixing behavior for several depth–season combinations, consistent with mixing between an initial pore-water DIC pool and an added DIC source generated by in-situ mineralization processes. However, variability in the background DIC pool and contributions from carbonate dissolution may also have influenced the fitted intercepts (Figure 5). The intercepts were nevertheless consistently enriched relative to bulk δ¹³C–TOC, by +5.28‰ to +11.38‰ (Table 4). We define this difference as the offset (Δ = δ¹³C intercept – δ¹³C–TOC), with positive values indicating DIC isotopically enriched relative to bulk OC. The offset exceeded the small (often 0–2‰) fractionation typically attributed to heterotrophic respiration in marine and wetland sediments, suggesting either an additional contribution of relatively ¹³C-enriched carbon or more pronounced isotopic discrimination than is usually assumed for bulk OC_min.

Three processes could produce such an offset, and δ¹³C–DIC alone cannot uniquely distinguish among them: (1) selective remineralization of labile organic fractions slightly ¹³C-enriched relative to bulk TOC (Fu et al., 2022; Eagle et al., 2025), (2) partial methanogenesis (Hellings et al., 2000), and (3) dissolution of sedimentary carbonate, such as biogenic shell fragments (Fu et al., 2022). Methanogenesis is unlikely to be the principal source, as no intercept approached the strongly ¹³C-enriched values (absolute δ¹³C–DIC > +10‰) reported for systems in which methanogenesis dominates DIC production (Hellings et al., 2000), though a partial contribution cannot be excluded. A carbonate contribution likewise cannot be dismissed. A two-end-member mass balance model, attributing the entire offset to carbonate dissolution (δ¹³C ≈ 0‰) mixing with organic-derived DIC (δ¹³C ≈ δ¹³C–TOC) yields upper bounds of 24.4-52.0% across site-season combinations; these are upper bounds rather than source apportionments, because neither end-member was measured at these sites. Furthermore, the UMF intercepts (-10.52 ± 1.66‰ and -12.16 ± 1.58‰) overlap pore-water DIC values reported for Yellow Sea and East China Sea sediments (−8.2 ± 1.9‰), where biogenic carbonate dissolution has been identified as the dominant DIC source (Fu et al., 2022). Carbonate dissolution in coastal sediments is driven primarily by the acidity generated during aerobic mineralization and is inhibited once oxygen is depleted, because anaerobic mineralization generates alkalinity instead (Goossens et al., 2026); our incubations were anoxic throughout, so dissolution would be suppressed by the experimental conditions. Consistent with this, sediment carbonate content was below 2% of dry weight at every depth in both cores and fell to 0.24% or less in August (Section 3.3), placing these sediments among the carbonate-poor shelf settings in which dissolution is limited by the size of the reactive carbonate pool (Goossens et al., 2026) and an order of magnitude below the carbonate-rich shelf sediments where dissolution has been identified as the dominant DIC source (Fu et al., 2022). Dissolved Ca²⁺ measured during the incubations confirms this expectation, constraining the carbonate contribution to 0.9% and 1.0% of depth-integrated DIC production at UMF and SJ in August (Section 3.5), an order of magnitude below the isotopic upper bounds. Carbonate dissolution therefore was not the principal cause of the observed offsets. Radiocarbon or total alkalinity constraints would be required to partition the remaining organic contribution by age or reactivity. Therefore, we interpret the offset not as a measure of any single pathway, but as an aggregate index reflecting the combined influences of selective remineralization and carbonate dissolution on newly produced DIC. Specifically, it indicates the extent to which this DIC is isotopically offset from bulk sedimentary OC, a distinction increasingly recognized as important when decomposition signals are used in blue carbon accounting (Kelleway et al., 2022).

Under this interpretation, the offset was significantly larger at UMF than at SJ in August (+9.44‰ vs. +5.28‰; p< 0.001; Section 3.5), whereas in April the difference was directionally consistent but not significant (+11.38‰ vs. +7.49‰; p = 0.58). The emergence of a station contrast only during the growing season is itself informative, and is consistent with a vegetation-associated signal. In April, when standing S. japonica biomass had not yet developed (Section 3.1), the two habitats exhibited similar isotopic characteristics. By August, however, the vegetated site had diverged. Two non-exclusive explanations are plausible. Bare tidal flats typically accumulate more biogenic carbonate debris than densely vegetated marsh surfaces, favoring a greater carbonate contribution at UMF. Alternatively, at SJ, where a greater contribution of recently supplied, relatively labile S. japonica–derived detritus (Section 4.1), the actively respired fraction may closely resemble bulk TOC. In contrast, at UMF it may represent a minor, more reactive subset of a bulk pool otherwise enriched in older or allochthonous material (Fu et al., 2022). Consistent with either, the four offsets form a monotonic gradient (SJ–August< SJ–April< UMF–August< UMF–April) that decreases as vegetation cover and growing-season activity—and thus the supply of fresh organic matter—increase. The combined evidence from δ¹³C–DIC, pore-water Fe²⁺ and SO₄^2- profiles, and the independent OC_min estimates (Section 4.2) points to heterotrophic respiration, likely dominated by sulfate and iron reduction (Hines et al., 1989; Canfield et al., 2005; Hyun et al., 2007), as the primary mineralization pathway at both sites.

These insights have implications for the carbon budget of Garolim Bay. Because methanogenesis does not appear to be a principal DIC source, the mineralized carbon is returned predominantly as CO₂-system DIC, either through atmospheric outgassing or lateral export to adjacent waters, rather than as CH₄. This distinction is significant because CH₄ emissions, owing to their higher global warming potential, can offset blue carbon burial by as much as 20% in some vegetated coastal systems (Macreadie et al., 2019). The offset gradient further suggests that mineralization at the two sites may draw on carbon pools of differing turnover time, with rapid respiration of recently produced organic matter at SJ and a greater contribution from older or compositionally distinct carbon at UMF. This interpretation complements the relationship between mineralization and sequestration examined in Section 4.5 and suggests that combining δ¹³C–DIC with alkalinity or radiocarbon constraints could resolve whether mineralization losses deplete recently sequestered carbon or draw down a legacy reservoir.

4.4 Regional and global comparison of OC stock and sequestration rate

Both the normalized OC_sto at UMF and SJ (57.1 and 71.7 Mg C ha^-1, respectively) fall below global mean stocks reported for unvegetated tidal flats (86.3 ± 7.6 Mg C ha^-1), salt marshes (317 ± 19.1 Mg C ha^-1), mangroves (293 ± 7.2 Mg C ha^-1), and seagrass habitats (194 ± 20.2 Mg C ha^-1) (Supplementary Figure S5). These results indicate that both habitats in Garolim Bay store considerably less OC than the global average for their respective habitat categories, including unvegetated tidal flats. However, the present estimates are comparable to the multiyear (2017–2020) mean previously reported for Garolim Bay (48.2 ± 20.1 Mg C ha^-1; Lee et al., 2021), suggesting that they are broadly representative of longer-term conditions at this site rather than reflecting an anomalous sampling period.

The OC_seq at both sites were markedly lower than global means for unvegetated tidal flats (130 ± 29.9 g C m^-2 year^-1), salt marshes (245 ± 26.0 g C m^-2 year^-1), mangroves (231 ± 26.0 g C m^-2 year^-1), and seagrass habitats (138 ± 38.0 g C m^-2 year^-1) (Supplementary Figure S5). However, they were similar to previously reported rates for other unvegetated and vegetated tidal flats along the Korean west and south coasts, including Garolim Bay (16 g C m^-2 year^-1), Sinan (16 g C m^-2 y^-1), and Suncheon Bay (30 g C m^-2 year^-1), and somewhat lower than the rate reported for Ganghwa tidal flats (54 g C m^-2 year^-1) (Lee et al., 2021). This regional consistency, despite the large gap relative to global habitat means, suggests that comparatively low sequestration rates may be a broader characteristic of Korean west-coast tidal flats rather than a site-specific feature of Garolim Bay.

4.5 Mineralization relative to sequestration: burial efficiency of the sedimentary carbon budget

The OC_sto and OC_seq presented in Section 4.4 describe only the net accumulation side of the sedimentary carbon budget. Coastal wetlands, however, function simultaneously as carbon sources through OC_min and as sinks through the burial of photosynthetically fixed organic matter, so a comprehensive carbon budget requires joint consideration of both fluxes (McLeod et al., 2011; Macreadie et al., 2013; Alongi, 2018, 2020). Following the sediment mass-budget framework in which the organic carbon delivered to the sediment is partitioned between oxidation and burial (An et al., 2025b), and because the input flux was not measured independently here, we take the sum of OC_min and OC_seq as a lower bound on the carbon delivered and express their relative magnitude as an apparent burial efficiency, OC_seq/(OC_min + OC_seq) (Table 3). Annual values were 14.1% at UMF and 12.1% at SJ, decreasing from 22.2% and 23.8% in April to 10.6% and 8.8% in August. Because the sedimentation rate is a single decadal-mean value at each site, this seasonal contrast tracks the summer intensification of depth-integrated OC_min (2.6-fold at UMF and 4.3-fold at SJ) rather than a seasonal change in accumulation, and we do not present it as an independent result.

These efficiencies can be placed alongside direct sediment carbon budgets from other coastal settings. They are of similar magnitude to the 4.25–15.8% reported for the organically enriched, semi-enclosed Jinhae Bay (An et al., 2025b), and well below the 42% buried of the 42 mol C m^-2 y^-1 delivered annually to an intertidal mudflat of the Scheldt Estuary (Middelburg et al., 1995). A comparable dominance of mineralization has also been reported against a different reference flux, primary production, at a nearby S. japonica marsh and mudflat in Ganghwa-do, Korea (An et al., 2024) and in global mangrove soils (Alongi, 2020).

Two methodological limitations bound this interpretation. OC_min is a contemporary process rate scaled from short-term incubations, whereas OC_seq integrates decades of net accumulation, so the two are not directly commensurable; and incubation-based estimates are sensitive to factors such as incubation duration (Alongi, 2020) and may represent upper bounds on in situ rates, while the SJ sedimentation rate is a lower bound (Section 2.5). We therefore regard the order-of-magnitude separation, rather than any specific efficiency, as the robust result; it is large enough that neither correction would reverse it.

Finally, this sediment-focused comparison does not by itself establish these habitats as net sources of CO₂ to the atmosphere, which would require a full net ecosystem carbon balance incorporating in situ primary production and lateral DIC, DOC, and POC export, a substantial term in Garolim Bay (An et al., 2025a). Within these bounds, the similar burial efficiencies at the vegetated and unvegetated sites and the sharp summer intensification of mineralization at both indicate that OC_min is not merely a vegetation-associated modifier of net storage but a first-order flux governing the sedimentary carbon budget, and that stock inventories or production-based estimates alone cannot resolve the carbon dynamics of these habitats. Because a single core site was sampled in each habitat, however, the contrasts reported here should be regarded as indicative of the difference between these two locations rather than as a general property of vegetated and unvegetated tidal flats; establishing the generality of the pattern would require replication across multiple marsh-mudflat pairs.

5 Conclusions

In the intertidal sediments of Garolim Bay, the S. japonica marsh supported a higher supply and microbial turnover of OC than the adjacent unvegetated mud flat. OC_min and δ¹³C–DIC-based Keeling plot analyses converged to show that summer mineralization at SJ was concentrated in rapid, near-surface respiration, consistent with a greater contribution of relatively fresh organic matter, whereas the DIC produced at UMF was isotopically more offset from bulk sedimentary OC; neither site showed evidence of methanogenesis as a dominant pathway. Direct measurement of dissolved Ca²⁺ during the incubations showed that carbonate dissolution accounted for only a small fraction of DIC production, indicating that this offset is not primarily an inorganic signal. OC_sto and OC_seq at both sites nonetheless remained well below global blue carbon means, and mineralization exceeded sequestration by three- to tenfold in both seasons, corresponding to annual burial efficiencies of 12-14%; stock- or production-based estimates alone are therefore insufficient to characterize the carbon balance of these habitats. These findings underscore the need to pair carbon stock inventories with direct measurement of mineralization when evaluating the blue carbon potential of vegetated and unvegetated intertidal wetlands, including bare tidal flats now being considered for carbon accounting frameworks.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmars.2026.1964660/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Hermano Melo Queiroz, University of São Paulo, Brazil

Reviewed by: Ruiting Gu, East China Normal University, China

Yijing Wu, Tongji University, China