Variations in the Optical and Molecular Composition of Dissolved Organic Matter Exported from Coastal Wetlands

Abstract

Coastal wetlands regulate the transport and transformation of dissolved organic matter (DOM) between land and the ocean. It is known that DOM exported from tidal wetlands is chemically distinct from DOM in adjacent estuaries, yet its sources and compositional dynamics remain poorly resolved. Here, we investigated tidal variability in the optical and molecular composition of surface water DOM through absorbance and fluorescence spectroscopy and high-resolution mass spectrometry. Surface water samples were collected over multiple tidal cycles at three tidal marshes along a surface water salinity gradient in the Chesapeake Bay. At all sites, ebb tides consistently exported marsh-derived DOM enriched in terrestrial signatures with larger, more degraded, aromatic (e.g., tannin-like and condensed hydrocarbon-like) molecules. As water levels decreased during ebb tides, DOM composition gradually shifted from more protein-like and lipid-like to more condensed hydrocarbon-like substances, suggesting a sequential release of compositionally distinct DOM pools that may be distributed along the lateral gradient in the landscape. Tidal variability was minor at the low-salinity marsh due to limited hydrologic connectivity and greater riverine inputs. Our findings highlight the finer-scale tidal dynamics of surface water DOM, providing insights into how increasing hydrological alterations may affect coastal carbon cycling and downstream ecosystem functioning.

Introduction

Tidal marshes play a critical role in biogeochemical transformations and transport along the land-to-ocean continuum by mediating the quantity and composition of organic matter exported to coastal waters. , Guided by the outwelling hypothesis, numerous studies have quantified bulk dissolved and particulate carbon fluxes from tidal marshes to open water environments and generally reported that these systems act as net sources of carbon to adjacent estuaries. , Despite substantial efforts to measure bulk carbon fluxes and identify the drivers of their variability, little is known about the composition of the exported material. The biogeochemical role of dissolved organic matter (DOM) is not only defined by its abundance but also its composition, which influences degradation pathways, microbial utilization, and transport across terrestrial–aquatic interfaces. − Therefore, a deeper understanding of the chemical nature and spatiotemporal dynamics of DOM is essential for constraining the role of tidal marshes in coastal biogeochemical cycling and understanding the fate of marsh-derived DOM in coastal waters.

In tidal marsh-estuary systems, DOM is a complex mixture of allochthonous and autochthonous sources, , the composition and distribution of which are controlled by hydrologic processes, including tidal cycles, porewater exchange, and groundwater discharge, and by biogeochemical transformations such as physical processes (e.g., adsorption, flocculation), photochemical degradation, and microbial degradation. − Variations in DOM sources and compositions across tidal, seasonal, and spatial scales have been widely examined over the past few decades, most often using optical measurements such as absorbance and fluorescence spectroscopy. ,− In particular, parallel factor analysis applied to excitation–emission matrices (EEMs-PARAFAC) has been effective for statistically extracting independent fluorescence components, successfully capturing the export of marsh-derived DOM into adjacent estuaries with optical characteristics typically associated with terrestrial DOM, such as higher aromaticity, photoreactivity, and molecular weight, relative to estuarine DOM. ,,,− More recently, a few studies have employed advanced high-resolution analytical techniques, such as Fourier transform ion cyclotron resonance mass spectrometry (FT-ICR MS), to investigate DOM in these systems at the molecular level; , however, such applications have been limited to a low temporal resolution (high vs low tide), in part due to the high cost of the analysis and challenges with the complex data structure. Although limited in number, available studies have shown contrasts in the molecular composition of DOM between high and low tides, complementing findings from EEMs-PARAFAC. , However, to date, no study has yet attempted to resolve the “terrestrially derived” signal in marsh-derived DOM.

Terrestrial environments are inherently heterogeneous in space and time. Tidal wetlands are distinct from adjacent uplands in that they undergo periodic oxic–anoxic cycles, have higher primary productivity, and vegetation adapted to frequent inundation and periodic anoxia, all of which influence DOM composition. As seawater intrudes landward, upland coastal forests can gradually transition into wetlands, forming dynamic ecotones (i.e., transition zones). Within this upland-transition-wetland-estuary continuum, hydrological connectivity is regulated by fluctuating hydrologic regimes driven by both tidal inundation and groundwater hydrology, ultimately influencing the composition of land-derived DOM before it enters estuaries. ,, Based on this conceptualization of terrestrial landscape as a series of interacting subsystems, we may expect that exceptionally low tides, when the terrestrial signal is the strongest, mobilize DOM signatures from farther inland (e.g., upland and/or transition zones) through enhanced lateral transport. Little is known about how the spatially heterogeneous terrestrial DOM pools are transported and appear in estuarine surface waters under the combined influence of groundwater transport that is relatively persistent and slow (e.g., weeks to seasons time scale) vs tidal fluctuations of surface water that is more temporally dynamic (e.g., semidiurnal) at marsh-estuary interfaces. Differences in tidal amplitude, which determine the extent to which water levels fall during ebb tides, have been shown to influence the intensity of porewater export from marshes; , however, the chemical composition of DOM during these hot moments of export remains uncharacterized, opening the question of whether these higher fluxes are accompanied by shifts in DOM composition.

Chesapeake Bay is the largest estuary in the United States and hosts tidal marshes distributed along a salinity gradient from tidal freshwater marshes in the upper estuary to brackish and salt marshes toward the mouth of the estuary. Surface water salinity, as a proxy for a variety of chemical inputs to marshes, influences the dynamics of sediment chemistry, dominant anaerobic carbon metabolism pathways, plant diversity, microbial communities, subsequently shaping the accumulation, and composition of organic matter in marshes. ,− The river continuum concept and the literature that extended it to estuarine systems , offer a conceptual framework predicting shifts in geophysical conditions and biological responses along the river continuum. These works suggest decreasing terrestrial inputs and increasing autochthonous sources and marine influence from headwater toward the mouth of the estuary; however, empirical testing of this prediction, especially under dynamic tidal regimes, remains limited.

In this study, we examined tidal variations in the sources and composition of surface water DOM at three tidal marshes along a surface water salinity gradient (i.e., from oligohaline to polyhaline). We first evaluated the optical properties to confirm that our observations reflect previously documented tidal patterns. We then characterized molecular-level compositions to gain more detailed insights into how DOM varies across tidal and salinity gradients. By combining a higher temporal resolution sampling (hourly) with the analytical specificity of high-resolution mass spectrometry, we further resolved the tidally driven lateral transport of DOM across the terrestrial–aquatic interfaces. We hypothesized that (1) tidal marshes will export DOM with terrestrial characteristics, with a greater intensity of DOM from farther inland as it will enhance the lateral transport of DOM; and (2) marshes with higher salinity will exhibit greater variability in DOM composition due to greater marine influence and in situ primary production.

Materials and Methods

Site Description

This study was conducted at three tidal marsh sites (Sweet Hall Marsh, Kirkpatrick Marsh, and the Goodwin Islands) spanning a surface water salinity gradient from oligohaline to polyhaline in the Chesapeake Bay, located on the East Coast of the U.S., over the summer of 2024 (July to September) (Figure A). Sweet Hall Marsh (SWH) is an oligohaline tidal marsh located on the Pamunkey River in Virginia. Kirkpatrick Marsh (site code in this study: GCW), where Smithsonian Environmental Research Center’s Global Change Research Wetland (GCReW) is located, is a mesohaline tidal marsh on the Rhode River in Maryland. The Goodwin Islands (GWI), the polyhaline site, are located near the mouth of the York River. At each site, a transect along the terrestrial–aquatic interface from upland-transition-wetland-estuary was defined along the elevation and ecological gradient (Figure B). Uplands were characterized by higher elevations with mostly dry and oxic soils. Upland vegetation was dominated by pine forests in GWI, mixed deciduous forests in GCW, and a combination of deciduous and pine trees in SWH. , Wetlands were lower in elevation with predominantly anoxic marsh soils. Vegetations in higher salinity marshes (GCW and GWI) were dominated by salt-tolerant species such as Spartina patens and Distichlis spicata, while lower salinity marsh (SWH) had greater plant diversity, including Peltandra virginica, Carex stricta, Leersia oryzoides, Polygonum punctatum, and Polygonum arifolium. Transition zones between uplands and marshes were characterized by visually apparent stressed and dead trees with intermediate hypoxic or anoxic periods. Marsh soil types were Mispillion and Transquaking mucky peat in GCW, Axis fine sandy loam in GWI, and fine-textured silt and clay in SWH. Detailed soil characteristics along the transect in GCW and GWI are available in Patel et al. (2025).

1.

(A) Map of the Chesapeake Bay showing locations of the three sampling sites (SWH: Sweet Hall Marsh, GCW: Kirkpatrick Marsh, GWI: the Goodwin Islands). (B) Locations of groundwater end-member samples collected along the upland-to-wetland transect at each site (UP: upland, TR: transition zone, WC: wetland center, WE: wetland edge). Surface water salinity and elevation model data are from Chesapeake Bay Program and NOAA Office for Coastal Management, respectively. ,

Sample Collection and Bulk Measurements

Hourly surface water samples were collected from the center of each marsh tidal creek over a continuous 48 h period to capture the water flowing in and out of the marsh using ISCO 3700 autosamplers (Teledyne). The PTFE suction line with a weighted strainer was attached to a float so that it was roughly ∼10 cm below the water surface. The autosampler was programmed to rinse the suction line three times with surface water before collecting 1 L of water into an acid-washed bottle placed inside the autosampler chamber every hour, resulting in a total of 144 surface water samples (48 from each site). Samples were retrieved and replaced with a fresh set of 24 bottles after the first 24 h. Ice was placed inside the autosampler chamber to keep the collected samples cool. Due to a malfunction of the autosampler, minimal water was collected between the time points 25 and 28 (4 samples) at GWI; these samples were excluded from solid-phase extraction and subsequent molecular characterization. Water depth and salinity were recorded at 15 min intervals using EXO and SonTek sensors (YSI) deployed adjacent to the autosampler.

In addition to surface waters, groundwaters along the upland-to-wetland transect and estuarine water were collected as potential end-member samples at each site. Groundwater samples were collected by pumping from ∼1m deep wells installed at each subzone along the transect (i.e., upland, transition, wetland center, and wetland edge), except in the GCW upland where a 5m deep well was pumped due to the deep water table at this zone (Figure B). Estuarine water samples were collected far from the creek mouth to avoid immediate marsh influence. Two to three replicate water samples (1–2 L each) were collected from each end-member depending on water availability. In SWH, upland groundwater was insufficient for replicate sampling; therefore, additional groundwater was collected from a well in the adjacent forested swamp. To minimize mixing between end-members, terrestrial groundwater samples were collected during low tide, and estuarine water samples were collected during high tide.

Collected surface water and end-member samples were filtered immediately after collection or upon retrieval from the autosampler by passing through Sterivex filters (pore size 0.22 μm; Millipore); 0.45 μm prefilters were used when necessary. Filtered water samples were shipped to the laboratory on ice and kept at 4 °C until the bulk and optical analyses and solid-phase extraction, which were completed within 1 week.

Dissolved organic carbon (DOC) and total dissolved nitrogen (TDN) concentrations of the water samples were determined by the high-temperature catalytic oxidation method using a total organic carbon analyzer (TOC-L) equipped with a TNM-L unit and an ASI-L autosampler (Shimadzu). 50 μL sample was injected, acidified using 1 M HCl, and purged to eliminate inorganic carbon. Three measurements were taken for each sample. One additional measurement was taken if the coefficient of variation exceeded 2%. Calibration curves were generated for each analytical run using potassium hydrogen phthalate and potassium nitrate as standards. Check standards of known concentrations were analyzed every 10 samples to ensure analytical accuracy throughout the run.

Optical Analysis and PARAFAC Modeling

Absorbance spectra and three-dimensional excitation–emission matrices (EEMs) of filtered samples were obtained using an Aqualog (Horiba) with a 10 mm quartz cuvette (Starna Cells). Absorbance spectra of each sample were collected over a wavelength range of 230 to 800 nm. Samples were diluted if the absorbance exceeded 0.3 at 254 nm to meet the linearity in the Beer–Lambert law. Fluorescence emission was collected from 245 to 825 nm at 2.5 nm intervals, with excitation wavelengths between 230 and 800 nm in 3 nm increments. Integration time was determined to maintain the optimal signal range (20–40k) for each sample. Fresh Milli-Q water was analyzed daily and used as a blank to correct the data. Collected absorbance spectra and EEMs were processed in R (v4.3.2) using the fewsdom package (v1.1.5), which conducted blank subtraction, removal of Raman and Rayleigh scatterings, inner filter effect correction, Raman normalization, dilution correction, and DOC normalization (details in the Supporting Information). Commonly used absorbance and fluorescence indices, including specific UV absorbance at 254 nm (SUVA254), spectral slope ratio (SR), humification index (HIX), and freshness index (fresh), were calculated using the same R package. Parallel Factor Analysis (PARAFAC) modeling was performed on the processed EEMs following the tutorials in refs − using the staRdom package (v1.1.28). Five components, including three terrestrial humic-like components (C1, C3, and C4), one microbial humic-like component (C2), and one protein-like component (C5), were identified from the PARAFAC model (Table S2). The detailed information about calculating commonly used optical indices and PARAFAC modeling is available in the Supporting Information.

Solid-phase Extraction (SPE) and High-resolution Mass Spectrometry (HRMS)

Solid-phase extraction was performed following the workflow outlined in Dittmar et al. (2008). Each water sample was acidified and loaded onto a PPL cartridge (Agilent). DOM loaded on cartridges was eluted and reconstituted to a final concentration of 80 mgC/L in methanol/acetonitrile (80/20 v/v%) (Fisher Scientific, Optima LC/MS grade). Acidified Milli-Q water was extracted the same way as samples and used as process blanks. 5 μL of each reconstituted extract was injected into a Vanquish ultrahigh performance liquid chromatography (UHPLC) system coupled with an Orbitrap Exploris 240 mass spectrometer (Thermo Scientific). The injected sample was separated on a Hypersil Gold C18 column (150 mm × 2.1 mm, 3 μm; Thermo Scientific), and the eluate was electrosprayed at 3500 V in positive mode. The MS scans were acquired in the mass range of m/z 90–900 in profile mode, with an Orbitrap resolution of 120,000 at m/z 200. Each sample was injected three times to collect technical replicates.

The raw mass spectra were processed in Spyder (Python v3.11.9) using CoreMS (v3.3.0), an open-source mass spectrometry data analysis framework. Mass spectra were averaged at 0.5 min intervals between retention time (RT) 1 and 10 min, with a noise threshold of 10. Molecular formulas were determined with search criteria of C1–50, H4–200, O1–23, N0–2, S0–1 with mass error <0.5 ppm. The output feature quantification tables, including lists of features with their RT blocks, mass-to-charge ratios (m/z), intensities, and assigned molecular formulas, were further processed in R (v4.3.2) following the previously published nontargeted metabolomics data preparation and visualization protocol. Details regarding the SPE, LC–MS analysis, and MS data processing are provided in the Supporting Information.

Molecular property indices, such as nominal oxidation state of carbon (NOSC), modified aromaticity index (AI_mod), and double bond equivalents (DBE), for each molecular formula were calculated using the fticrrr workflow in R, which also included assignment of biochemical class of each formula based on its H/C and O/C ratios: lipid-like (0 < O/C ≤ 0.3 and 1.5 ≤ H/C ≤ 2.5), unsaturated hydrocarbon-like (0 ≤ O/C ≤ 0.125 and 0.8 ≤ H/C < 1.5), protein-like (0.3 < O/C ≤ 0.55 and 1.5 ≤ H/C ≤ 2.3), amino sugar-like (0.55 < O/C ≤ 0.7 and 1.5 ≤ H/C ≤ 2.2), carbohydrate-like (0.7 < O/C ≤ 1.5 and 1.5 < H/C ≤ 2.5), lignin-like (0.125 < O/C ≤ 0.65 and 0.8 ≤ H/C < 1.5), tannin-like (0.65 < O/C ≤ 1.1 and 0.8 ≤ H/C < 1.5), and condensed hydrocarbon-like (0 ≤ O/C ≤ 0.95 and 0.2 ≤ H/C < 0.8). Each assigned molecular formula may correspond to multiple structural isomers and thus should not be interpreted as representing a unique molecular structure. Similarly, the assignment of formulas to biochemical groups is based exclusively on H/C and O/C ratios; therefore, it should be considered as indicative of likely structural characteristics rather than specific chemical identities.

Intensity-weighted relative abundances of biomolecular classes were calculated by dividing the intensity of each class by the total intensity of all identified or subset features. All DOM samples were carbon-normalized prior to LC–MS analysis; thus, intensity-weighted relative abundances reflect proportions of observed classes relative to organic carbon in the sample. However, it should be noted that ionization bias may affect relative intensities of peaks as ionization efficiencies among molecules may vary.

Statistical Analysis and Visualization

Principal coordinate analysis (PCoA) and clustering of temporally dynamic molecular features were conducted following the aforementioned nontargeted metabolomics data preparation and visualization protocol. Before statistical analyses, missing values were considered as below the analytical detection limit and imputed using randomly generated values smaller than the smallest nonzero value in the blank-corrected dataset. For each site, water levels were first scaled from −1 to 1 with a mean of 0 (Figure S4A). Samples were initially binned into three groups–high tide (normalized water level > 0.4), mid-water level (−0.4 < normalized water level < 0.4), and low tide (normalized water level < −0.4). Adjustments were made based on compositional (dis)­similarities observed in the PCoA plots (Figure S4B) to ensure that the classification reflects actual compositional differences rather than arbitrary thresholds. For SWH, the same classification approach was initially attempted, but it did not yield meaningful results in the subsequent Kruskal–Wallis analysis due to weaker water-level-driven compositional changes (Figure S4B SWH). Therefore, samples with extremely low water levels were assigned to the low tide group, and all others were binned into the high tide group. Final sample counts were 11 (low) and 37 (high) for SWH, 10 (low)-22 (mid)-16 (high) for GCW, and 17 (low)-16 (mid)-11 (high) for GWI (Figure S4C). The nonparametric Kruskal–Wallis test was applied to individual features to evaluate differences in intensity across predefined groups of samples representing different water levels (high, mid, and low tides). Features with p < 0.05 were considered significantly different across water levels. Clustering of these significant features was then performed using the built-in k-means clustering function of the ComplexHeatmap package (v2.25.3), which simultaneously generated a heatmap to visualize their temporal patterns. Features on the heatmap (rows) were reordered to improve the visualization using the complete linkage method based on Spearman distance. The initial clustering separated low tide-associated features from high tide-associated features (Figure S5), and an additional posthoc clustering on the low tide-associated features further resolved them into two groupsL-Early ebb and L-Late ebb. At SWH, where temporal variability in molecular DOM composition across water levels was minimal (Figure S4B), only initial clustering was applied. Details of the clustering approach are available in the Supporting Information. Clustering robustness was assessed using a resampling approach (Jackknife resampling). One or two samples were removed at a time from the dataset and clustering was repeated (∼500 iterations per clustering). The recomputed clustering results were compared with the original clustering via adjusted Rand index, a similarity index ranging from 0 to 1, with value = 1 indicating two clustering results are identical. Across all analyses, the average similarity was 0.97 (Figure S6), indicating that our clustering approach was stable and reproducible.

Results and Discussion

Export of Marsh-derived DOM during Ebb Tides

At all sites, bulk properties (DOC and TDN concentrations) and optical indices (SUVA254, SR, HIX, and fresh) changed as a function of water level, showing significant linear relationships (p < 0.05), except the humification index (HIX) at SWH (Figure ). In general, DOC and TDN concentrations, SUVA254, and HIX were negatively correlated with water level, whereas SR and freshness index showed positive correlations. The higher DOC and TDN concentrations with larger molecular size (low SR) and higher aromaticity (high SUVA254) in low tide DOM are typical traits of terrestrially derived DOM ,− (Figure B) and were also observed in our terrestrial groundwater samples (Table S3). In contrast, high tide DOM samples were less degraded (high freshness index) and less aromatic (low SUVA254), and they were characterized by lower DOC and TDN concentrations and smaller molecular size (high SR) (Figure ). These patterns were also evident in our estuarine water samples, suggesting increased inputs of estuarine DOM during high tides (Table S3). Similar tidal patterns in bulk properties and optical indices have been reported previously. ,, Lower aromaticity in surface seawater has been attributed to greater photodegradation. The higher freshness index in our high tide DOM is consistent with previous observations of enhanced microbial activity in estuarine waters. ,

2.

(A) Variations in water depth and DOC concentrations of hourly collected surface water samples at the three sampling sites. (B) Pearson’s correlation coefficients between bulk measurements or optical indices and water depth by site. *p < 0.05, **p < 0.01, ***p < 0.001.

To further characterize the CDOM composition, we employed EEMs-PARAFAC analysis. Principal component analysis (PCA) of the relative abundances of PARAFAC components separated surface water samples along the water level gradient on the PC1 axis, which explained the largest variance (69%) in the surface water DOM composition (Figure A). High tide samples (negative PC1) showed higher loadings of C2 and C5 (Figure A and Table S2). C2 represents a marine humic-like component, typically indicative of DOM altered through photochemical and/or microbial processes. − C5 has been referred to as a protein-like fluorophore, often associated with biologically produced DOM from phytoplankton, algae, and/or other microbes. ,, In contrast, low tide samples (positive PC1) showed higher loadings of PARAFAC components C1, C3, and C4, all previously classified as terrestrial humic-like DOM, − based on long emission and excitation wavelengths (Em 434–520 nm) and a secondary excitation peak (Figure A and Table S2). These components are characterized by large molecular size and high hydrophobicity. Among them, the component with a shorter emission wavelength (e.g., C1; Table S2) has been suggested to originate from lignin degradation products derived from plant litter, whereas the component with a longer emission wavelength (e.g., C3) reflects more humified material common in sediments/soils. ,, These components have been reported to be negatively related to salinity at freshwater–seawater interfaces as terrestrially derived DOM mixes with and is diluted by marine DOM sources. ,,

3.

(A) Principal component analysis of relative abundances of PARAFAC components in surface water samples. PC1 and PC2 explain 69% and 29% of the variance, respectively. Gray arrows are loadings. (B) Principal coordinate analysis of dissolved organic matter composition based on Bray–Curtis dissimilarity matrix calculated from the HRMS data. Black arrows are intensity-weighted relative abundances of biochemical groups projected on the PCoA plot posthoc.

The contrasting DOC/TDN concentrations, optical indices, and associated fluorescent components at high and low tides align well with previously reported tidal DOM dynamics. ,, Consistent with those studies and the dominant net export of DOC previously reported at GCW, our results suggested that marshes export chemically distinct DOM during ebb tides to adjacent estuaries. Marsh-derived DOM was enriched in terrestrially derived materials, while high tide samples consisted of more protein-like and microbially influenced DOM (a detailed quantitative assessment of carbon fluxes from our studied wetlands will be presented in future work). This provides preliminary validation for our subsequent hypotheses regarding the lateral transport of terrestrial DOM at the landscape scale. However, optical properties alone cannot distinguish the variety of potential terrestrial sources such as groundwater transport from locations further upslope vs marsh-derived; therefore, we next employed mass spectrometry-based analysis, which provides greater molecular specificity and thus a higher resolution for identifying DOM sources and dynamics beyond the simpler terrestrial vs microbial DOM classification. ,−

The HRMS data of surface water DOM also captured both tidal and inter-site variations in molecular composition (Figure B); however, site-specific contrasts were more evident than tidal variations at the molecular level, compared to the optical measurements. The greatest variance in molecular composition (PCoA1 and PCoA2; total of ∼43% of variance; Figure B) appeared to separate the samples by site: the posthoc fitting of site differences onto the PCoA plot showed a much better correlation (R ² = 0.84, p < 0.001) than that with the water level (R ² = 0.05, p < 0.05). Specifically, surface water DOM at GWI was more enriched with lipid-like and unsaturated hydrocarbon-like molecules than the other two sites (p < 0.001), and their relative abundances increased with water level (p < 0.01) (Figures B and S3). At GCW, protein-like molecules were more abundant (p < 0.001), and their relative abundance decreased with the water level (p < 0.01) (Figures B and S3). At SWH, aromatic molecules such as lignin-like and condensed hydrocarbon-like molecules were more abundant than at the other two sites, and lignin-like molecules significantly increased with the water level (p < 0.001) (Figures B and S3). There was no significant relationship between the water level and the number of molecular formulas assigned, except for GCW where the number significantly increased with the water level.

These site-specific molecular patterns indicate that each site has a distinct surface water DOM molecular composition shaped by local inputs and processes (e.g., marsh inputs, DOM degradation, transport, and mixing) that can dominate over tidal signals, which are the primary focus of our study. HRMS detects thousands of molecular features, capturing the great complexity of DOM; however, not all features are responsive to tidal fluctuations. Biogeochemical processes such as photochemical and microbial transformations also occur continuously at varying rates in the marsh-estuary interfaces, making it even more challenging to pinpoint the molecules that are specifically mobilized by tidal water movement. To account for the site-specific heterogeneity and the limitations of overall molecular compositions, we applied a nontargeted approach to isolate the subset of molecular features that significantly varied with the water level at each site (see the Materials and methods “Statistical analysis and visualization”).

Variations in DOM Dynamics along the Salinity Gradient

Our nontargeted approach to identify tidally dynamic molecules (see the Materials and methods “Statistical analysis and visualization”) yielded two to three molecular feature clusters per site. At each site, high-tide-associated features were abundant during high tides and low-tide-associated features were abundant during low tides. At GCW and GWI, low tide-associated features were further separated into two subgroups, L-Early ebb and L-Late ebb, based on their temporal patterns: the intensity of “L-Early ebb” features increased earlier as the water level decreased, followed by the increase in “L-Late ebb” features.

The intensity-weighted average H/C and O/C ratios of the molecular clusters and their trajectories from high to low water levels (High to Low for SWH; High to L-Early to L-Late for GCW and GWI; Figure ) demonstrated that each site exhibits distinct tidal DOM dynamics. During high tides, when estuarine DOM contributed more strongly to surface waters, SWH surface water received DOM that was more oxidized, less saturated, and more aromatic than the other two sites (SWH High in Figure ). As the water level decreased at SWH, the surface water became enriched in more saturated and less oxidized molecules (SWH Low in Figure ). Compared to SWH, high-tide-associated features at GCW and GWI showed progressively lower oxidation and greater saturation (Figure ), indicating clear spatial differences in the estuarine DOM composition along the salinity gradient. As the water level fell at GCW and GWI, molecules with higher oxidation state and greater saturation increased in intensity (L-Early ebb in Figure ). With continued water-level decrease, both sites showed increasing contributions of more unsaturated (lower H/C) molecules; however, low tide DOM at GWI was more oxidized than that at GCW (L-Late ebb in Figure ), suggesting spatial variations in the marsh-derived DOM composition.

4.

(A) Intensity-weighted average H/C and O/C of water level-associated molecular feature clusters. (B) Molecular properties of all individual features in water level-associated molecular feature clusters. *p < 0.05, **p < 0.01, ***p < 0.001.

The distinct tidal compositional variability at SWH, in contrast to the other two sites, was also evident in the optical measurements. At SWH, HIX showed no significant correlation with the water level, and the range of its variation was also smaller than our analytical precision (±0.022) (Figure B). This indicates that humification levels did not measurably change across water levels at SWH. Instead, HIX remained consistently high at SWH, similar to values observed during low tides at the other two sites, suggesting a continuous supply of highly humified DOM regardless of tidal fluctuations (Figure B). Consistent with this pattern, the PCA of the PARAFAC components showed the lowest contribution of microbially derived, protein-like DOM (C5) during high tides at SWH (Figure A), and the SWH samples were well separated from the other two sites on the PCA plot (negative SWH on the PC2 axis with the other sites being positive on the PC2 axis).

Inconsistent patterns between optical indices, such as the decoupling of HIX from other bulk or optical measurements at SWH in our study, have been reported previously in a brackish estuary. In Wang et al. (2025), discrepancies between indices were attributed to additional biogeochemical processes, such as resuspension of sediments, rather than simple shifts in terrestrial and marine contributions. In our study, we attribute the unique pattern at SWH to its position within the estuary. Located in the upper reach of the estuary (Figure A), SWH is a tidal freshwater marsh with distinctive physical, chemical, and ecological characteristics that are substantially different from the higher-salinity marshes downstream. ,, The greater tidal amplitude and higher DOC concentrations at SWH (Figure A) are consistent with expectations for tidal freshwater marshes, which occur along narrower river channels in the upper reaches and receive constant riverine inputs with higher concentrations of dissolved and suspended matter. , High tide samples and estuarine water at SWH showed the highest DOC concentrations and the most terrestrial-like DOM characteristics compared to those at the other two sites (high SUVA254, HIX; low SR, fresh, PARAFAC component 5; greater aromatic molecules), reflecting strong riverine influence and inputs of plant- and soil-derived DOM and their degradation products − (Table S3, Figures B and ).

Variability between high and low tide samples was also distinctively small at SWH. Although the tidal range (i.e., the difference between the highest and lowest water levels) was the greatest at SWH, followed by GWI and GCW (Figure A), bulk C and N concentrations and optical indices were the least correlated with the water level at SWH (Figure B). Multivariate analyses also highlighted this distinction, as the SWH high-to-low tide DOM showed the least spread on the PCA and PCoA plots (Figure ). The tidal amplitude at the head of the estuary is primarily influenced by upstream water flow driven by tidal energy with less direct advection of saline waters compared to marshes near the mouth of the estuary. As a result, high tide waters introduced to the marsh are largely riverine at SWH. , Because compositional differences between marsh-derived and riverine DOM are smaller than those between marsh-derived and estuarine DOM at the higher-salinity marshes, tidal variability in DOM composition was more limited at SWH. In contrast, high tide DOM at the high-salinity site (GWI) was composed of smaller, less oxidized, more saturated molecules, with relatively more lipid-like and unsaturated hydrocarbon-like molecules that are structurally simpler than terrestrial DOM (Figures and S3), which indicates a stronger marine influence with greater primary production by microalgae and cyanobacteria. ,,, These suggest that while water levels govern temporal variability in DOM composition within each site (i.e., tidally driven exchanges between marsh-derived and estuarine DOM), this relationship does not scale proportionally with tidal ranges across sites, and site-specific factors need to be considered when understanding tidal dynamics of surface water DOM.

A comparative study of DOM properties in a tidal freshwater marsh and a brackish marsh (high tide salinity <0.5 ppt and ∼10 ppt, respectively) along a river-dominated estuary in China has reported contrasting tidal dynamics of DOM composition that aligns with our observations. Optical indices in their study suggested predominant riverine signatures (humic-like PARAFAC components) across tidal stages in the upper-reach freshwater marsh, whereas the brackish marsh DOM was more enriched with proteinaceous substances due to the greater oceanic inputs.

Beyond salinity, a range of other local environmental factors, such as vegetation, microbial communities, and soil properties, can further influence DOM compositions across marshes. Marsh age was another factor proposed to influence the DOM composition in the aforementioned study, which suggested that more degraded, aromatic DOM was exported from the older tidal freshwater marsh. Soil characteristics such as porosity and compaction have been proposed to affect oxygenation of tidal marsh soils, influencing DOM and nutrient fluxes. In line with their observation, we observed that marsh soils at SWH were finer-textured compared to the sandier soils in GWI. The limited water movement in SWH could have reduced the exchange of DOM-containing water between the marsh and surface waters, resulting in diminished hydrological connectivity, as shown by the small variations in DOC concentrations at SWH (Figure A). Such conditions may have promoted humification of DOM in the marsh by limiting oxygenation of the marsh soils and contributed to the accumulation of less bioavailable, humic-like material. Finer-textured soils may also enhance the selective preservation of certain molecular classes, further influencing DOM composition and reactivity.

Focusing only on the tidally dynamic portion of DOM (Figure ), DOM dynamically exported from SWH is compositionally different from that of the other two sites. Low tide-associated molecular features at SWH were characterized by larger molecular size and lower NOSC, AI_mod, and DBE, with a greater relative abundance of aliphatic compounds (H/C > 1.5) such as lipid-like molecules, compared to the higher salinity sites (Figure ). More reduced (lower NOSC), aliphatic DOM is likely preserved within the marsh under more prolonged anaerobic, finer-textured soil conditions and then exported from the marsh during low tides. Although it is a relatively small fraction of the marsh-derived DOM, this aerobically labile DOM is expected to be readily degraded by microbial communities in the oxic conditions once exported to the estuary. , Additionally, the larger molecules exported from SWH are more prone to photochemical and microbial degradation to smaller, more refractory molecules, according to the size-reactivity continuum. , Further work tracking the fate and reactivity of marsh-exported DOM across sites (e.g., incubation experiments) would clarify the cycling of the exported DOM and processes controlling its transformation and persistence in coastal waters.

Temporal Resolution of Marsh-derived DOM

We next examined the relative abundances of major biochemical groups across the water-level-associated clusters and evaluated their correspondence with the biochemical signatures of the landscape-level end-member samples. High tide-associated clusters and estuarine waters were enriched in lignin-like molecules (Figure ), consistent with typical estuarine DOM containing various photochemical and microbial degradation products. , As the water level decreased from high to mid-level (from high to L-Early ebb), molecular composition shifted toward higher proportions of lipid-like and protein-like molecules (Figure ). With further decreases in the water level (from L-Early ebb to L-Late ebb), the relative abundance of lipid-like and condensed hydrocarbon-like molecules increased. Aromatic compounds (lignin-like, tannin-like, and condensed hydrocarbon-like) were more prevalent in upland and transition zones, due presumably to the predominance of vascular plant inputs (Figure ), , although this pattern was absent in the GCW upland zone due to its much deeper groundwater level. Protein-like and lipid-like compounds were enriched in wetlands as those compounds are often associated with algal or marine sources. , Previously, GCW has been shown to trap algae and other suspended matter in the marsh during tidal cycles.

5.

Intensity-weighted relative abundance of biochemical classes in water level-associated molecular feature clusters (left) and end-members (right) by site.

As shown in Figure , the temporal patterns of L-Early ebb and L-Late ebb in response to tides were not identical (GCW and GWI). The intensities of both L-Early ebb and L-Late ebb increased during low tides, relative to high tides. However, the relative intensity of L-Early ebb increased earlier and reached a plateau, whereas the increase in L-Late ebb occurred later, peaking near the lowest water level at each site (Figure B). Together with their distinct chemical compositions discussed above, these patterns suggest that larger drops in the water level could trigger the export of compositionally distinct DOM pools, even at fine temporal scales.

6.

(A) Heatmaps showing temporal patterns of low tide-associated molecular features. Each row is an individual feature. For GCW and GWI, features were further clustered into two groups (L-Early ebb and L-Late ebb) using K-means clustering (k = 2). (B) Water level (gray lines) and intensity-weighted relative abundances of low tide-associated feature clusters determined in A (stacked bars). Biochemical composition of each cluster is provided in Figures and

This pattern was particularly evident at GCW (Figure B), where the sampled tides were asymmetric. Symmetric tides produce relatively consistent water levels at successive high and low tides (e.g., relatively symmetric tides at SWH; Figure A), whereas asymmetric tides generate unequal high and low water levels such as highest high tide and lower high tide (e.g., relatively asymmetric tides at GCW; Figure A). Our 48 h sampling at GCW captured four consecutive low tides (Figure B), each with a different water level. The relative proportions of low tide-associated clusters (“L-Early ebb” and “L-Late ebb”) varied across these four low tides (Figure B), demonstrating that the DOM composition can differ even among low tides. Previously, asymmetric tides have been shown to influence the intensity of terrestrial DOM exported to tidal creeks as the water residence time and the maximum high or low water levels vary. In Taillardat et al. (2018), the greatest porewater contribution during their sampling period was observed when the water level was the lowest, with no porewater input during the prior low tide. This was explained by the fact that the tide level was not low enough to drain the porewater during the preceding low tide, which allowed a longer contact time between the water and the sediments, leading to the greatest porewater contribution during the subsequent lowest low tide. Our observation at GCW is in line with their results, with additional evidence of molecular compositional variability in marsh-derived DOM depending on tidal regimes.

Previous work using stable carbon and nitrogen isotopes and biomarker analyses in four marshes in the Chesapeake Bay, including GCW, has shown that estuarine and riverine OM contributed more to marsh edges and surface soils, whereas marsh plant OM was more prevalent in subsurface soils. We propose that trapped algal and suspended matter closer to the wetland edge and surface soils was released first from the marsh as the water level decreased, followed by DOM from deeper sediments and/or farther inland, which was more enriched in lipid-like and condensed hydrocarbon-like molecules as water levels declined further. Our results therefore suggest that the molecular compositional difference is driven at least in part by changing contributions from interacting subzones of the landscape. However, this mechanism may not be exclusive, as vertical heterogeneity in DOM compositions with soil depth is possible. , A mechanistic understanding of whether the changes are driven by more vertical or lateral transport of OM requires further investigation, with more targeted sampling strategies.

Broader Implications

It is well documented that marshes export terrestrially-derived DOM to adjacent estuaries during ebb tides. Short-term, high-resolution (e.g., hourly) observations exist; however, these studies have largely relied on optical measurements, which provide only a limited binary framework (i.e., terrestrial vs microbial). A separate body of literature has reported that marine contributions to surface water DOM increase toward the mouth of an estuary. These studies have examined spatial variations in surface water DOM composition and its seasonal shifts (e.g., increased terrestrial DOM inputs linked to river discharge) using optical and molecular measurements, ,, but short-term tidal dynamics were generally neglected.

By integrating higher temporal, spatial, and analytical resolution measurements, our study showed that tidally driven water movements regulate DOM export from marshes to adjacent estuarine waters, and the compositions and dynamics of this export are shaped by the interplay of lateral hydrologic connectivity (e.g., upland-wetland-estuary continuum), longitudinal connectivity (e.g., along an estuarine salinity gradient), and local tidal regimes. These hydrological linkages are vulnerable to anthropogenic and natural perturbations, such as sea level rise, groundwater withdrawal, and altered river discharge; ,,, therefore, future shifts in water movement will likely modify not only the physical movement of water but also the quantity and composition of DOM exported from coastal wetlands. Our findings suggest that, at short-term and site-specific scales, alterations in tidal regimes would control the relative export of DOM along the landscape. Over longer time scales, as shown in our site-specific variability, sustained changes in tidal regimes (e.g., by sea level rise) can fully alter the dominant sources and chemical characteristics of marsh-derived DOM. As pressures on hydrological regimes are projected to increase, developing a deeper, more integrative understanding of DOM dynamics across spatial and temporal scales will be essential for predicting coastal carbon cycling and related impacts on marine food webs. ,

All data underlying this study are openly available in ESS-DIVE at 10.15485/3001080.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.est.5c11869.

• Experimental details, materials, and methods, including optical data processing and PARAFAC modeling; solid-phase extraction; LC–MS measurements and data processing; testing for potential storage effects in the autosampler; statistical analysis and visualization; calculation of commonly used optical indices; PARAFAC components identified in this study; dissolved organic carbon, total dissolved organic nitrogen concentrations, and optical indices of end-member samples; comparison between autosampler-stored and paired grab samples; tidal variations in relative abundances of biochemical groups (all assigned formulas); sample group assignments for clustering; initial clustering of significant molecular features; and validation of clustering robustness (PDF)

The authors declare no competing financial interest.