1 INTRODUCTION

Blue carbon ecosystems (mangroves, salt marshes, and seagrass beds) cover only 0.07%–0.22% of the surface of the Earth but contain>50% of the carbon stored in marine sediment because they have high carbon sequestration rates (McLeod et al., 2011; Zhang et al., 2015; Spivak et al., 2019). Some of the sequestered carbon can be transported from marine sediment as dissolved organic carbon (DOC) and used in adjacent coastal ecosystems (Bouillon et al., 2008; Heck et al., 2008). DOC participates in various biogeochemical processes (including providing energy for microbial growth, participating in trace element and nutrient circulation, and affecting water transparency and heat exchange (Yamashita and Jaffé, 2008; Huang et al., 2017)) and plays an important role in regulating climate change by storing carbon as refractory DOC (RDOC) (Jiao et al., 2010).

Salt marshes are the most widely distributed blue carbon ecosystems. Carbon sequestered in salt marsh wetlands can be transported to coastal seawater through surface water export and subsurface water exchange (i.e., submarine groundwater discharge or porewater exchange) (Liu et al., 2021). It has been suggested that carbon transported through submarine groundwater discharge or pore water exchange is an important source of carbon to offshore water in some coastal ecosystems (Maher et al., 2013; Sadat-Noori et al., 2016; Chen et al., 2018, 2021). The various components of dissolved organic matter (DOM), e.g., amino acids, carbohydrates, and carboxyl-rich alicyclic molecules (CRAMs), range from bioavailable components with very fast turnovers to long-lived components and play different roles in carbon biogeochemistry (Bauer, 2002; Hansell, 2013). Dissolved labile DOC (LDOC) transported in submarine groundwater can be utilized by dinoflagellates in the absence of inorganic nutrients to cause red tide outbreaks (Lee and Kim, 2007). In contrast, CRAMs have residence times of up to thousands of years and are regarded as effectively sequestered forms of carbon (Jiao et al., 2018). Studies focused on RDOC such as CRAMs or LDOC such as amino acids and carbohydrates in aquatic ecosystems have been performed (Hertkorn et al., 2006; Jiao et al., 2018), but the characteristics of RDOC and LDOC in submarine groundwater are not well understood due to its own complexity and analytical limitations, which limits our understanding of the carbon cycle in coastal areas.

The structures and sources of DOM in submarine groundwater are usually studied by acquiring UV-vis absorption spectra and three-dimensional excitation-emission matrix (3-D EEM) spectra (Chen et al., 2010; Lu et al., 2020). For example, the fluorescence characteristics of shallow water on the Zhejiang coast, determined by UV-vis absorption and 3-D EEM spectroscopy, were found to indicate that submarine groundwater discharges were more likely than rivers to be the most important sources of chromophoric dissolved organic matter (CDOM) in the area (Gao et al., 2010). Chen et al. (2010) analyzed CDOM in groundwater from coastal marshes and wetlands in Florida by 3-D EEM spectroscopy and found that the CDOM was supplied from both the land and sea. In a study using 3-D EEM spectroscopy, the molecular weight and aromaticity of DOM in soil pore water was found to increase during transportation (Jiang et al., 2019). Nuclear magnetic resonance (NMR) spectroscopy can be used to determine the exact chemical structures of the major components of DOM and the contributions of different components to the total DOM (Aluwihare and Repeta, 1999; Hertkorn et al., 2013). In particular, CRAMs (the main components of RDOC) can be quantified by high-resolution NMR spectroscopy (Hertkorn et al., 2006; Lechtenfeld et al., 2015).

In this study, the optical properties and molecular compositions of DOM in submarine groundwater from salt marshes on Chongming Island, China, and in coastal water nearby were determined by UV-vis absorption spectroscopy, fluorescence spectroscopy, and NMR spectroscopy. The main aim was to identify the CDOM and DOM components in submarine groundwater and surface water in the salt marsh wetlands. This information will improve our understanding of the biogeochemical cycling of DOM in coastal areas, particularly salt marsh blue carbon ecosystems.

2 MATERIAL AND METHOD 2.1 Study area

Chongming Island is the third largest island in China, which located in the meso-tidal Changjiang (Yangtze) River estuary. The Changjiang River estuary has multi-order bifurcations. The south branch of the Changjiang River, along the south coast of Chongming Island, is dominated by freshwater discharges. The north branch of the Changjiang River, along the north coast of Chongming Island, has saline water intrusions (Zhao et al., 2015). Total runoff in the Changjiang River estuary in December 2017 was 3.93×10⁸ m³, which was 4.2% of total annual runoff. Average annual sediment discharge, measured at the Datong hydrological station (30°46′N, 117°37′E), between 2007 and 2017 was 1.27×10¹¹ kg (Changjiang Water Resources Commission, 2018). The sediment discharged by the Changjiang River has allowed extensive areas of shoals and tidal flats to develop. These have been colonized by various types of salt marsh vegetation. Chongming Dongtan National Nature Reserve, on the most eastern part of the Chongming Island shoreline, is the largest and most developed salt marsh wetland on Chongming Island, with a total area of ~4.69×10³ hm². The dominant native vegetation species in the Dongtan National Nature Reserve are herbs (Scirpus mariqueter) in the lower marshes and reeds (Phragmites australis) in the higher marshes (Huang et al., 2007; Gu et al., 2018).

2.2 Sampling

The samples were collected during December 2017. In total, 27 DOC samples (seventeen submarine groundwater samples, nine coastal surface water samples, and one seawater sample) and four DOM samples (two submarine groundwater samples, one coastal surface water sample, and one seawater sample) were collected (Fig. 1). The seawater sample was collected in February 2017. The submarine groundwater (well water and pore water) samples were collected from coastal wells and salt marshes on Chongming Island. Each pore water sample was collected at low tide using a push‐point piezometer and a peristaltic pump. Each well water sample was collected using an organic glass hydrophore (Chen et al., 2018). The coastal surface water samples were collected at three-hour intervals over two tide cycles at the time series station (31°28′51″N, 122°3′58″E) using a submersible pump. Freshwater sample collected from near Xuliujing on the Changjiang River estuary (Gao, 2018) was used as the freshwater end member. Seawater sample from the East China Sea was used as the seawater end member. The dissolved oxygen (DO), pH, salinity, and temperature for each sample were determined simultaneously in situ using a multiparameter meter (Multi 3430 WTW, Germany).

Each DOC and CDOM sample was passed through a filter immediately after being collected, then the filtrate was sealed in a pre-combusted (at 450 ℃ for 5 h) 20-mL glass vial and stored at -20 ℃ until analysis. Each DOM sample was passed through a filter into an acid-cleaned high-density polyethylene bottle and then adjusted to ~pH=2 by adding concentrated HCl (GC grade; Merck, Darmstadt, Germany). Solid-phase extraction (SPE) of DOM was performed following a standard protocol published by Dittmar et al. (2008). Briefly, a SPE cartridge (PPL, 1 g; Agilent Technologies, Santa Clara, CA, USA) was rinsed with methanol (GC grade; Merck) and then Milli-Q water adjusted to pH=2. The sample was then passed through the SPE cartridge at a flow rate<40 mL/min by applying a vacuum produced using a vacuum pump. DOC concentrations are generally higher in submarine groundwater than surface water, so only 3–5 L of each submarine groundwater sample were extracted, to avoid overloading the SPE cartridge.

2.3 Analytical measurement

The DOC concentrations were determined by performing three-five analyses of the sample using a TOC-VCPH analyzer (Shimadzu, Kyoto, Japan) that performed high-temperature catalytic oxidation (Zhang et al., 2019). The coefficients of variance were mostly<2%. The absorption spectra of the CDOM samples were acquired using a Cary 100 UV-vis spectrometer (Agilent Technologies) using a wavelength range of 200–800 nm at 1-nm increments (using a 1-cm path-length quartz cuvette). The measurements were baseline-corrected using a spectrum for Milli-Q water. Absorbance at 254 nm was determined and normalized to the DOC concentration to give the SUVA₂₅₄ value, which was considered to indicate aromatic and high-molecular-weight CDOM components (Weishaar et al., 2003). Absorbance at 350 nm (a₃₅₀) was determined and used to represent the abundance of CDOM, as was the case in previous studies (Helms et al., 2008; Gao et al., 2010; Zhou et al., 2013).

Fluorescence EEM spectra were acquired using a Hitachi F-4500 spectrofluorometer (Hitachi High-Technologies, Tokyo, Japan) with a 1-cm quartz cell. The parameters used to acquire the CDOM fluorescence spectra have been published previously (McKnight et al., 2001; Guéguen et al., 2007). Briefly, the excitation wavelength range was 220–450 nm at 5-nm intervals, and emission between 230 and 600 nm at 2-nm increments was recorded at a scanning speed of 12 000 nm/min. The excitation and emission bandwidths were both 5 nm. Milli-Q water, analyzed each day, was used as a blank, and the blank spectrum was subtracted from each sample spectrum to calibrate the Raman Effect. The fluorescent humification index (HIX), which indicates the degree of humification of DOM, was calculated by dividing the integrated fluorescence between 434 and 480 nm by the integrated fluorescence between 300 and 346 nm with excitation at 255 nm (Zsolnay et al., 1999). The fluorescent biological index (BIX) was defined as the ratio between fluorescence at 380 and 430 nm with excitation at 310 nm (Huguet et al., 2009).

The NMR analyses were performed using a method described by Hertkorn et al. (2006) and Zhang (2014). The mean DOC recovery was 40%. Briefly, 3–5 mg of a freeze-dried SPE-extracted DOM sample were dissolved in deuterated methanol and then analyzed using a Avance DRX 500 NMR spectrometer (Bruker, Billerica, MA, USA) equipped with a 5-mm broadband double-resonance probe at 298 K. A solution-state ¹H NMR spectrum was acquired by performing 3200 scans using an acquisition time of 3 s. The solvent signal was suppressed using the Bruker PRESAT system during the spectrum collection process. Mestrenova software was used to perform baseline correction and to integrate the spectrum.

2.4 Parallel factor analysis and NMR integration

Each EEM spectrum was modeled by performing parallel factor analysis (PARAFAC) using MATLAB 9.4 software and the DOMFluor toolbox (Stedmon and Bro, 2008), then the modeled components were matched in the OpenFluor database (Tucker congruence values >0.95; Murphy et al., 2014). The maximum fluorescence of a component was used to represent the fluorescence intensity of the component (in the Raman Unit, RU). The "total fluorescent dissolved organic matter (FDOM)" (∑F) was defined as the sum of all the PARAFAC components (∑F=∑1nCn). Four fluorescent components were obtained using the PARAFAC models, as shown in Table 1. The four fluorescent components could be divided into two categories, humic-like components (C1, C2, and C3) and a protein-like component (C4). C1 normally represents marine humic substances (Coble, 1996; Kulkarni et al., 2017). C2 is thought to represent allochthonous terrestrial-derived humic substances that are resistant to microbial degradation and optically active refractory DOM with turnover times between 400 and 600 a (Catalá et al., 2015; Li et al., 2015). C3 has been found to be ubiquitous terrestrial/autochthonous fulvic acid fluorophores (Stedmon and Markager, 2005; Yamashita et al., 2010). C4, which has previously been found in estuarine areas, represents freshly produced tryptophan‐like DOM composed of lignin-derived phenols and proteins produced by algae or microorganisms in submarine groundwater (Kothawala et al., 2014; Gonçalves-Araujo et al., 2015).

The relative abundances of the four groups of DOM constituents determined from the one-dimensional ¹H NMR spectra were quantified by integrating and normalizing the peak areas (Hertkorn et al., 2006; Lechtenfeld et al., 2015). The residual methanol peak was excluded during the integration process. All the hydrogen proton resonances came from four components, which were aromatic and alkenyl hydrogen atoms (10–5.3 ppm), carbohydrate hydrogen atoms (4.8–3.1 ppm), heteroatomic aliphatic hydrogen atoms (3.1–1.9 ppm), and aliphatic hydrogen atoms (1.9–0 ppm) (Hertkorn et al., 2006; Lechtenfeld et al., 2015). The CRAM concentration was determined by multiplying the CRAM (1.1–3.1 ppm) abundance by the DOC concentration (Lechtenfeld et al., 2015).

2.5 Statistical analysis

Statistical analyses were performed using SPSS 25 software (IBM, Armonk, NY, USA). P<0.05 was regarded as a statistically significant result.

3 RESULT AND DISCUSSION 3.1 Water chemistry and optical parameter 3.1.1 Water chemistry

The results of DO, pH, salinity, temperature, and DOC concentrations are shown in Table 2 and Fig. 2. The salinities of the submarine groundwater samples from different locations on Chongming Island were different. The south coast of Chongming Island is dominated by freshwater discharges from the Changjiang River, but the north coast is affected by saline water. The DOC concentrations in the well water and pore water samples were 61–494 and 135–1215 μmol/L, respectively. There were significant differences between the DOC concentrations in the submarine groundwater and coastal water (independent samples t-test, P<0.05). The mean DOC concentration in the submarine groundwater samples (386 μmol/L) was higher than that in the coastal water samples (91 μmol/L) (Fig. 2; Table 2). The DOC concentrations of submarine groundwater from Chongming Island were significantly higher (independent samples t-test, P<0.05) than the concentrations of 24–47 μmol/L previously found in submarine groundwater from a sandy beach at Hwasun Bay (Kim et al., 2013) and were significantly lower (independent samples t-test, P<0.001) than the concentrations of 790–1 910 μmol/L previously found in submarine groundwater from mangroves in the Maowei Sea in the dry season (Chen et al., 2018).

3.1.2 UV-Vis absorbance and fluorescence

The a₃₅₀ values for the well water, pore water, and coastal water samples were 0.67–5.42, 1.01–8.78, and 0.69–2.63/m, respectively (Fig. 3a). The CDOM abundance in submarine groundwater (4.89±2.66, n=17) was significantly higher (independent samples t-test, P<0.001) than that in coastal water (1.63±0.54, n=9) (Fig. 3a). The a₃₅₀ values in well water were significantly lower (independent samples t-test, P<0.05) than that in pore water (Fig. 3a). Similarly, submarine groundwater from a southern California salt marsh had a wide range of a₃₅₀ values (0.5–13.6/m) in the dry season, although the mean a₃₅₀ value (8.6/m) (Clark et al., 2014) was higher than that in our samples (4.89/m). The C1, C2, C3, C4, and ∑F values for the submarine groundwater and coastal water samples were significantly different (independent samples t-test, P<0.05) (Fig. 3). The mean FDOM abundances decreased in the order pore water (2.97 RU)>well water (1.90 RU)>coastal water (0.99 RU), which was consistent with the DOC concentrations and a₃₅₀ values. The high values for these parameters indicate that submarine groundwater may be an important reservoir of CDOM and FDOM in the salt marsh.

The SUVA₂₅₄ values for the samples are shown in Fig. 3d. The SUVA₂₅₄ data indicated that most of the CDOM was highly aromatic (>5 L/(m∙mg)). No significant difference was found between the SUVA₂₅₄ values for submarine groundwater and coastal water (independent samples t-test, P>0.05). The SUVA₂₅₄ values for coastal water decreased as the salinity increased (R=-0.703, P<0.05). However, no clear correlation was found between the SUVA₂₅₄ values for submarine groundwater and the salinity. This may have been related to the non-conservative behavior of DOC in subterranean estuaries (Chen et al., 2020). It has previously been found that the CDOM spectral properties deviated from the conservative mixing line in the saltwater freshwater mixing zone, indicating that different types of CDOM are supplied by different end members and that CDOM may be supplied through the release of organic matter by sediment or be produced in situ by microorganisms (Lu et al., 2015; Pain et al., 2019).

The HIX and the BIX values are shown in Fig. 3. The HIX value is generally used to indicate the degree of humification of CDOM (Zsolnay et al., 1999). A high HIX value (10) indicates strong CDOM humification, and a low HIX value (5) indicates autochthonous CDOM (Birdwell and Engel, 2010). It can be seen from Fig. 3g that most of the HIX values for submarine groundwater and coastal water were<5, although the HIX values were significantly higher (independent samples t-test, P<0.05) for submarine groundwater (3.62±1.28, n=17) than coastal water (2.58±0.11, n=9). This indicated that autochthonous CDOM was dominant in both submarine groundwater and coastal water. The BIX value is associated with microbially-derived and autochthonous CDOM (Hunt and Ohno, 2007; Wickland et al., 2007). A high BIX value (0.8–1.0) indicates predominantly autochthonous CDOM, and a low BIX value (<0.6) indicates little CDOM is produced (Huguet et al., 2009). The BIX values were high for both submarine groundwater (0.79±0.16) and coastal water (0.90±0.05), which agreed well with the HIX values. Degradation of plant detritus and microbial activities may have been the main sources of autogenous CDOM to submarine groundwater (Couturier et al., 2016). Autogenous CDOM may have been supplied to coastal water by algae (Suryaputra et al., 2015).

3.2 Molecular characterization of DOM 3.2.1 3-D EEM

The four fluorescent components and their relative contents are shown in Fig. 4. The submarine groundwater CDOM showed strong humic-like fluorescence with high maximum fluorescence values for components C1 and C3, but the coastal water CDOM was dominated by components C3 and C4 (Fig. 4b). The contributions of humic-like substances (C1, C2, and C3) to total fluorescence for the submarine groundwater samples were 63%–88% (mean 79%, n=17), which were higher than those for coastal water (71%–75%, mean 71%, n=9). The contributions of terrestrial humic-like substances (C2 and C3) to total fluorescence for the submarine groundwater were ~50%, indicating strong supplies of terrestrial humic substances. The contribution of protein-like substances (C4) to total fluorescence for coastal water (27%) was a little higher than that for submarine groundwater (21%). The seawater CDOM was dominated by protein-like components (59%). The submarine groundwater CDOM was generally affected both by marine and terrestrial substances, but terrestrial humic substances were dominant. Our results are consistent with the results of a previous study of coastal marshes and wetlands in Florida, in which submarine groundwater CDOM was generally found to be dominated by terrestrial humic-like components (Chen et al., 2010). In other studies, submarine groundwater CDOM was found to be affected by marine and terrestrial substances (Seidel et al., 2014, 2015).

3.2.2 NMR study

The ¹H NMR spectra used to identify the chemical environments of the protons in the DOM are shown in Fig. 5. Resonance at 3.31 ppm was for the solvent. There were four major functional forms of non-exchangeable hydrogen in all of the samples. Prominent resonances between 0.5 ppm and 1.9 ppm indicated hydrogen bound to saturated carbon with heteroatoms three or more bonds away. Resonance between 1.9 ppm and 3.1 ppm indicated heteroatoms two or more bonds away from protons. Resonances between 3.2 ppm and 5.0 ppm indicated hydrogen bound to carbon that was bound to oxygen. Weak resonances were found between 5.0 ppm and 10.0 ppm, suggesting the presence of olefinic and aromatic species (Hertkorn et al., 2006). The ¹H NMR spectra of the four DOM samples had similar main peaks. However, there were some subtle differences between the spectra for the different DOM samples, particularly in the molecular signals in the unsaturated proton region (δ_H>5 ppm). Particularly different peaks were found between 6.5 ppm and 8.5 ppm (enlarged in Fig. 5). The signals in this range included resonances of amino acid side chains and phenols (McCaul et al., 2011), which may have been substances forming part of fluorescent component C4 (proteins and lignin-derived phenols).

The relative abundances of the four DOM constituents are shown in Fig. 6a. The aromatic and olefinic constituents contributed<8% of the total for all samples, which agreed well with the results of previous studies of samples from various aquatic environments (Hertkorn et al., 2006; Lechtenfeld et al., 2015; Wang et al., 2018). However, purely aliphatic substances contributed almost half or more than half (45%–60%) of the ¹H resonances for all samples, and functionalized aliphatic substances contributed about half as much (23%–30%). Carbohydrates contributed less for the pore water samples (13.2%–14.8%) and coastal water samples (16.4%) than for the seawater samples (22.1%) or seawater from the Atlantic Ocean (20%–23%) (Hertkorn et al., 2013), but the contributions of aliphatic substances decreased in the order pore water>coastal water>seawater. Seidel et al. (2014) found, using ultrahigh-resolution Fourier transform ion cyclotron resonance mass spectrometry, and the tidal flats act as effective biofilters of labile DOM and bioavailable compounds such as dissolved carbohydrates. The bioavailability of DOM in groundwater has been investigated using bioactive compounds (e.g., amino acids and carbohydrates) in several studies, and relatively low bioavailability has been found for DOM in groundwater (Routh et al., 2001; Chapelle et al., 2009; Peter et al., 2012). The relatively low quantities of dissolved carbohydrates in pore water from subterranean estuaries on Chongming Island indicated that the DOM was weakly bioactive. Small molecules of plant origin in DOM are preferentially consumed by microorganisms and converted into larger molecules of microbial origin in pore water (Roth et al., 2019). In the anoxic intertidal zone of a salt marsh, relatively low carbohydrate concentrations may partly be caused by intense microbial activity.

Species classed as CRAMs are potentially refractory compounds that have been identified in DOM from various aquatic environments by multidimensional NMR and ultrahigh-resolution mass spectrometry (Hertkorn et al., 2006; Lam et al., 2007; Lechtenfeld et al., 2015). Prominent amino acid (LDOC) peaks between 1.1 ppm and 3.1 ppm were not found (Fig. 5), indicating that amino acids contributed little to the total integrals and signal overlaps (Lechtenfeld et al., 2015). Similar to the ¹H NMR spectrum of marine DOM, the smooth NMR resonance distribution and extensive signal overlap implies that DOM is dominated by refractory molecules (Hertkorn et al., 2013). It can be seen from Fig. 6a that the CRAM contribution was>60% for all samples. The CRAM contributions for the different samples were similar (64.5% for submarine groundwater, 61.7% for coastal water, and 66.8% for seawater), but the CRAM concentration was higher for submarine groundwater than coastal water and seawater (Fig. 6b) because of the high DOC concentration in the submarine groundwater. The CRAM contributions for the submarine groundwater were higher than previously found for DOM from freshwater from Lake Ontario (62%) (Lam et al., 2007) and the Penobscot River at Eddington (55%) but lower than for DOM from the Pacific Ocean (89%) (Cao et al., 2018). CRAM is a refractory component of DOM that is found in all aquatic ecosystems (Hertkorn et al., 2006; Lam et al., 2007; Cao et al., 2018). The 14C age and dominance of CRAM in DOM from the oceans indicated that terrestrial CRAM can be transported to the oceans and accumulate in the marine DOM pool (Cao et al., 2018). Estuaries and intertidal zones may be key to the transportation of terrestrial CRAM to the oceans.

3.3 Implications for coastal carbon processes and carbon sinks

Models of the mixing of freshwater end members and seawater end members have generally been used to study the behaviors of chemicals in estuarine and coastal environments (Santos et al., 2009; Gao et al., 2019; Reithmaier et al., 2021). We assumed that river water from near Xuliujing in the Changjiang River estuary could be used as the freshwater end member and seawater from the East China Sea could be used as the seawater end member in the mixing model. It can be seen from Fig. 2 that DOC is supplied in relatively large amounts through subterranean estuaries at salt marshes on Chongming Island because almost all the well water and pore water results were higher than the conservative mixing line. Similar results were found for salt marshes in southern California, and soil pore water was found to be the main CDOM reservoir in the salt marsh (Clark et al., 2014). Submarine groundwater discharges generally transport high DOC concentrations into coastal seas from salt marsh ecosystems such as the Chongming Island subterranean estuary. However, contradictory results have been found in a small number of studies, such as in a study of subterranean estuaries on Jeju Island (a volcanic island), in which DOM and CDOM were found to behave conservatively, perhaps because the submarine groundwater discharge rates were high (14–82 cm/d) (Kim et al., 2013).

High DOC concentrations in submarine groundwater are generally supplied by DOC leaching from sediment (Komada et al., 2004), sewage penetration (Yang et al., 2015), production by benthic algae (Suryaputra et al., 2015), and microbial activity (Couturier et al., 2016) in salt marsh environments. High CDOM abundances in submarine groundwater, particularly of the humic fraction of CDOM and CRAMs in DOM, do affect the carbon cycle in coastal areas through subterranean estuarine processes (Coble, 2007; Sadat-Noori et al., 2016; Chen et al., 2018). The humic fraction of CDOM and CRAMs with long turnaround times are proxies for slow cycling and recalcitrant DOM (Cory and Kaplan, 2012; Catalá et al., 2015). It would therefore be expected that DOM with a high CRAM content in submarine groundwater would make a strong contribution to the nearshore RDOC pool through submarine groundwater discharges and eventually accumulate in the ocean DOM pool (Fig. 7).

4 CONCLUSION

In the salt marsh subterranean estuaries on the Chongming Island, DOM behaves non-conservatively and is supplied by terrestrial humic substances. Although the CDOM of submarine groundwater is affected by both marine and terrestrial substances, the main component is terrestrial humic substances. DOM samples from the submarine groundwater has low carbohydrate contents (<20%) and high CRAM contents (>60%). The high RDOC concentrations in submarine groundwater indicates that DOM derived from submarine groundwater discharge may make important contributions to the offshore refractory carbon pool. In different seasons, DOM has different properties. The SPE method leads to the relative content of molecules with low molecular weight being underestimated by NMR spectroscopy. Studies with larger time scales and using more advanced characterization techniques should be performed to further improve our understanding of DOM migration and transformation processes in subterranean estuaries.

5 DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.

6 ACKNOWLEDGMENT

We thank Fule ZHANG and Xueqing YU from SKLEC/ECNU for their kind assistance in the field and laboratory. We also acknowledge the Editor-in-Chief Song SUN and two anonymous reviewers for their constructive comments that significantly improved the original manuscript.