The study site is located on Spiekeroog, a barrier island in the southern North Sea, adjacent to the German coast of Lower Saxony (Figure 1a). As one of Germany's least developed East Frisian Islands with a low number of permanent residents (≈800 inhabitants) and negligible agricultural or industrial activity, this pristine environment offers exceptionally favorable conditions for studying natural biogeochemical processes with little anthropogenic interference [CWSS (Common Wadden Sea Secretariat, Wilhelmshaven, Germany) and WHNPG (World Heritage Nomination Project Group), 2008]. Spanning ~10 km east to west and 2 km north to south, the island covers an area of 21.3 km² (Zielinski et al., 2022). It experiences high-energy, meso-tidal conditions with a mean tidal range of 2.7 m and a mean significant wave height of 1.4 m (Herrling and Winter, 2015). The island's geology comprises medium to fine sands from the Pliocene, Pleistocene, and Holocene epochs, occasionally containing gravel, clay, peat, and shales, with a model-based estimation of the hydraulic conductivity of 11 m d⁻¹ (Beck et al., 2017; Grünenbaum et al., 2020). A clay layer, ranging from 1.5 to 15 m in thickness, lies beneath the main dune area, and constrains the islands' freshwater lens at ~40 m below sea level, although its extent beneath the beach aquifer is unknown (Röper et al., 2012). The average annual precipitation is ~800 mm, with groundwater recharge rates ranging from 300 to 400 mm annually (Röper et al., 2012; Seibert et al., 2018). Groundwater temperatures in the freshwater lens remain around 10°C (Röper et al., 2014; Seibert et al., 2018), while seawater temperatures vary seasonally from 0 to 4°C in winter 16 to 20°C in summer (Reuter et al., 2009). Along the northern shoreline of Spiekeroog, three multilevel (ML) observation wells were installed in 2022 with filter depths at 6, 12, 18, and 24 m below the beach surface (Massmann et al., 2023). The beach stretch measures ~200–300 m, with well ML1 located near the dune base, well ML2 near the high-water line, and well ML3 near the low-water line (Figure 1). Previous studies of the shallow (upper 2 m) STE in the same location revealed distinct redox conditions, ranging from oxic near the dune base to sub- to anoxic toward the low-water line, with additional anoxic freshwater discharge from deeper layers of the freshwater lens into the North Sea (Reckhardt et al., 2017). Numerical models further demonstrated dynamic variations in redox conditions under high-energy beaches like Spiekeroog, extending to depths of 20 m due to constantly changing beach morphodynamics (Greskowiak et al., 2023; Figure 1c). A first survey of the newly installed wells by Massmann et al. (2023) reported distinct redox and DOC distribution patterns. ML1 and ML2 exhibited mostly oxic conditions above 18 m, transitioning to anoxic conditions from this depth. In contrast, ML3, near the low water line, was anoxic at all investigated depths. DOC concentrations increased from the 6 m well depth toward mid-depths (12–18 m) in ML1 and in the deepest part (18 and 24 m) of ML3, while ML2 showed less variability in DOC concentrations across depths (Reckhardt et al., 2024).

Solid-phase extraction and ultrahigh-resolution mass spectrometry were employed to investigate the molecular composition of sediment leachates, porewater before and after Fe-DOM coagulation, and seawater DOM. For the sediment leachates, ~45 mL of water and HCl leachates were adjusted to pH = 2 with HCl and sodium hydroxide and desalted and concentrated using 100 mg Agilent Bond Elut PPL cartridges, following Dittmar et al. (2008). Prior to extraction, the cartridges were soaked and repeatedly rinsed with methanol and acidified ultrapure water (pH = 2). After passing the samples through, the resins were washed with acidified ultrapure water (pH = 2), dried under N₂ gas and DOM was sequentially eluted with 5 mL Optima-grade methanol and 3 mL dichloromethane. The elutes were mixed, dried under argon, reconstituted in 2 mL methanol, and stored at −20°C. It should be noted that the leachates were not filtered prior to the extraction step. Although the PPL resin retains only DOM (i.e., dissolved molecules and small colloids) in its pore space, it is still possible that some suspended solids were captured on the 20 μm frit or between the resin grains. Therefore, some DOM may have been leached from these particles during elution (Waska et al., 2015). In contrast to porewater and seawater extractions, a dichloromethane elution step was included for the sediment leachates. Previous sediment leaching experiments yielded lower SPE-PPL recoveries compared to those from water samples (Dittmar et al., 2008; Zhou et al., 2024) which we attributed to a higher affinity of more aromatic, hydrophobic mineral-associated compounds to the PPL resin that could not be mobilized by methanol. However, extraction efficiencies including the additional dichloromethane elution were still rather low compared to those for porewater and seawater DOM samples and ranged from 4% to 63% (average: 17% ± 20%) for water leachates and from 3% to 43% (average: 12% ± 9%) for HCl leachates, based on DOC. For bulk porewater DOM, porewater Fe-DOM, and seawater, ~200 mL of filtered and acidified (pH = 2) samples (n = 75) were desalted and concentrated using 200 mg Agilent Bond Elut PPL cartridges, following Dittmar et al. (2008). Samples were eluted with Optima-grade methanol and stored at −20°C, with average extraction efficiencies ranging between 48% and 66% (Average of 53% ± 11% based on DOC).

Extracts were analyzed using a 15 Tesla Bruker solariX XR Fourier transform ion cyclotron resonance mass spectrometer with electrospray ionization (ESI-FT-ICR-MS) in negative mode. Samples were adjusted to 2.5 ppm DOC in a 1:1 v/v methanol: ultrapure water matrix, filtered (0.2 μm PTFE), and directly infused with a flow rate of 20 μL min⁻¹. Ion accumulation time was 0.2 s, and 200 broadband mass scans were accumulated per sample. An in-house deep-sea reference (Green et al., 2014) was analyzed daily to monitor instrument drift. Mass spectra were internally calibrated (<0.1 ppm error) and processed using the open-source program ICBM-OCEAN (Merder et al., 2020) for noise removal, peak alignment, and molecular formula assignment (<0.5 ppm error, C_{1 − 100}H_{1 − 200}O_{1 − 70}N_{0 − 6}S_{0 − 2}P_{0 − 1}.). To ensure unique formula assignments, the ICBM-OCEAN program applies a hierarchy of knock-out criteria, prioritizing isotopolog verification, homologous series length, and assignment error. All assigned unique molecular formulae were categorized into distinct molecular groups based on their elemental ratios. The resulting cross tables containing molecular formulae, chemical characteristics, and sample-specific signal intensities were then exported from ICBM-OCEAN for further processing.

Two main datasets were assembled for DOM molecular composition and environmental parameters (salinity, oxygen concentration, pH): (1) sediment leachates and (2) Fe-DOM coprecipitation data (bulk and removed), including porewater and seawater samples. Data processing was conducted individually for each dataset. Data crosstables were then filtered to retain only monoisotopic masses for assigned molecular formulae, enhancing the prominence of less prevalent compounds by removing ¹³C-containing formulae before statistical analysis. Samples were normalized based on their intensity sums, and cumulative sums of molecular properties (elemental composition, compound group classification) were calculated from the relative signal intensity distributions. Five distinct DOM molecular indices from existing literature were retrieved and applied:

In line with Reckhardt et al. (2024), the porewaters of the investigated STE wells contained varying proportions of fresh groundwater and recirculating seawater under oxic to anoxic conditions (Figure 3). Spatially, ML1 showed salinities ranging from ~2 to 12 (average = 7) and remained oxic to a depth of 18 m (Figures 3a, b). ML2 exhibited salinities between ~17 and 30 (average = 23) and was oxic down to 12 m. ML3 displayed higher salinities from 6 to 12 m (15–28) but lower salinities below 18 m (11–15), with anoxic conditions observed throughout. Dissolved Fe was detectable at depths ≥18 m in ML1 and ML2, while it was measurable at all depths in ML3 (Figure 3c). Overall, porewater pH was slightly alkaline to mildly basic (7.6–8.4) in the investigated wells and was highest at ML1 at 6 m (Figure 3d). FDOM values varied across stations, with the highest average observed at ML3 (80 ± 3 ppb QSE), approximately double that of ML1 (45 ± 12 ppb QSE) and ML2 (39 ± 9 ppb QSE; Figure 3e). At ML3, FDOM was lower in the upper 12 m and higher in deeper, low-salinity waters. DOC concentrations followed a similar trend, with the highest average recorded at ML3 (119 ± 25 μM), exceeding ML1 (92 ± 27 μM) and ML2 (86 ± 5 μM). At ML3, DOC was stratified, showing lower concentrations at 6–12 m and higher concentrations at 18–24 m. All porewater DOC concentrations were lower than seawater levels (162 ± 24 μM) from the same campaign (Figure 3f). Groundwater salinity and redox gradients at ML1, and to a lesser extent at ML2, were consistent across the investigated months and depths compared to ML3 (Figure 3). DOC concentrations at ML1 and ML2 were highest in September, lowest in October, and intermediate in December. In contrast, FDOM exhibited a different trend, with the highest values recorded in December at ML1 and in October at ML2. At ML3, Fe and DOC concentrations, as well as FDOM values, particularly at 12 m across the investigated months, covaried with the porewater salinity gradient above and below 12 m (Figure 3a).

Grain sizes at depths of 6, 12, 18, and 24 m below the beach surface varied across ML1, ML2, and ML3 (Figures 4a–c), with coarser grain sizes observed at greater depths. At ML1, fine sand predominated, with median sizes ranging from 145 to 302 μm. ML2 showed a slight trend toward coarser grain sizes, from fine-medium sand in the upper layers (198–208 μm) to medium sand at greater depths (380–381 μm). At ML3, grain size gradually increased from fine sand in the upper layers (183–209 μm) to medium sand at deeper levels (329–353 μm). Gravel was present at 18 and 24 m depths in ML2 and ML3 but was removed prior to measurement, preventing quantification of the gravel fraction.

Total organic carbon (TOC) concentrations ranged from below the detection limit to 1,370 ppm and varied across stations and depths (Figures 4d–f). The highest levels were observed at ML1, followed by ML3 and ML2. With the exception of ML2, where TOC was highest in the coarse 24 m fraction, TOC concentrations were generally higher in fine-grained sediments and lower in the coarse 24 m fraction. In comparison, TOC was about 200–4,200 times higher than DOC in the ambient porewater over the 3 months studied (September, October, and December 2022; Supplementary Table 1).

The concentrations of leachable dissolved organic carbon (LOC) were slightly higher in HCl leachates (13 ± 5 to 50 ± 2 ppm) than in water leachates (7 ± 4 to 44 ± 4 ppm) and decreased progressively from ML1 to ML3 in both cases (Figures 4g–i). Although no consistent depth-related trends were observed for LOC concentrations, their general patterns mirrored those of TOC. Normalized to bulk sediment concentrations, leachable LOC from both HCl and water leachates accounted for ~4%−14% of the TOC in the bulk sediments (Supplementary Table 1). Furthermore, loosely bound DOC and DOC associated with the poorly crystalline Fe phases were up to ~130 and 160 times higher than porewater DOC, respectively (Supplementary Table 1).

Fe concentrations in the HCl leachates ranged from 8 to 550 ppm (Figures 4j–l). No consistent depth-related trends were observed for Fe; however, its concentration generally decreased from ML1 to ML3. With the exception of ML2, Fe profiles at ML1 and ML3 mirrored those of TOC and LOC, with higher Fe concentrations found in finer sediments. Additionally, Fe/LOC ratios for HCl leachates at ML1 and ML3 generally mirrored the depth profiles of Fe and LOC concentrations, while discrepancies were observed at ML2 (Supplementary Table 1). At ML1 and ML2, the highest Fe/LOC ratios occurred at 18–24 m, and the lowest at 6 m. At ML3, the highest ratio was observed at 12 m, and the lowest at 18 m. No consistent depth trend was observed across the stations (Supplementary Table 1).

Overall, sea- and porewaters in the deep STE of Spiekeroog had higher SPE_DOC/SPE_TDN ratios compared to both water and HCl leachates (Table 1). Porewater also exhibited the highest molecular diversity indicated by numbers of assigned molecular formulae, surpassing both seawater and leachates. Water leachates had the lowest molecular diversity, except at ML1, where it was slightly higher than in HCl leachates. Porewater DOM had the highest molecular mass, exceeding seawater and leachates. Molecular diversity and mass increased with depth at ML1 and ML2 but decreased at ML3. Depth trends for water leachates were inconsistent, although molecular diversity and mass decreased from ML1 to ML3. In HCl leachates, depth and spatial trends were inconsistent at ML2 and ML1; however, at ML1, both molecular diversity and mass increased with depth.

Elemental stoichiometries of water leachates exhibited the highest H/C ratios and MLB_L values, followed by HCl leachates, and then sea- and porewaters (Table 1; Figures 5a–c). In contrast, sea- and porewaters had the highest O/C ratios, AI_mod, and DBE values, followed by HCl and water leachates. Spatial variation in H/C and O/C ratios, AI_mod, and DBE values in porewater and leachates were not associated with specific depths or stations, although MLB_L in porewater was higher at ML1 compared to ML2 and ML3. Among the molecular indices, the terrestrial index (I_Terr) was highest in HCl leachates and showed the strongest variation between depths and stations (Figures 5d–f). In contrast, I_Terr values in porewater and seawater were relatively low and remained fairly constant across the spatial scale. The degradation index (I_DEG) was lowest in sediment leachates and highest in STE porewaters, with intermediate values for seawater (Figures 5g–i). Consistent with the lower H/C ratios, sea- and porewaters had higher proportions of aromatic compounds and were dominated by highly unsaturated compounds compared to water and HCl leachates (Supplementary Figure 1). In contrast, the sediment leachates were primarily composed of unsaturated compounds (Supplementary Figure 1). No clear depth or beach cross-sectional trend was observed, except for water leachates at ML2, where aromatic and unsaturated compound classes increased with sampling depth.

Van Krevelen plots (Figure 6) of unique molecular formulas for each leachate type revealed distinct compositional differences. Water leachates were characterized by a higher proportion of oxygen- and hydrogen-rich labile compounds (upper right quadrant), with decreasing abundance from ML1 to ML3. HCl leachates exhibited a more diverse composition, with a higher proportion of oxygen-poor aromatic compounds (lower left quadrant) which decreased from ML1 to ML3 and increased with depth, along with a substantial amount of oxygen-poor, hydrogen-rich formulas (upper left quadrant).

The Fe-DOM coprecipitation experiment revealed complex spatiotemporal variations in DOM removal efficacy across anoxic porewater samples (i.e., samples with detectable concentrations of Fe²⁺; Reckhardt et al., 2024). This was particularly evident at depths of 18–24 m at ML1 and ML2, and 6–24 m at ML3 (Figures 3b, c). In these anoxic porewaters, pH levels ranged from 7.6 to 8.2 (Figure 3d). Initial DOC concentrations prior to aeration ranged from ~75 to 153 μM, while Fe concentrations ranged from 9 to 21 μM, resulting in molar Fe/DOC ratios of 0.01 to 0.22. FDOM values before aeration ranged from 24 ppb QSE to 114 ppb QSE, with maximum values observed at 18–24 m at ML3 (Supplementary Table 2).

Following aeration, FDOM removal exhibited a spatiotemporal gradient: highest in September (~50%), moderate in October (~11%), and minimal in December (~7%; Figures 7a–c). Spatially, the greatest FDOM reductions occurred at ML1 and ML3 (≥12 m), with intermediate reductions at ML2. DOC concentrations showed minimal changes after aeration, with notable decreases in September (~12% at ML1; ~5% at ML3). In October and December, slight increases in DOC concentrations were observed. Overall, changes of <10% were considered negligible, as they were below the precision of the TOC analyzer used (Figures 7d–f). In line with FDOM trends, molecular DOM composition in porewater also changed following aeration, showing both spatial and temporal differences (Tables 2–4; Figure 8). In September, removal primarily affected highly unsaturated and saturated compounds, resulting in decreases of ~5% and 52%, respectively (Figures 8a–d; Table 2). This selective removal led to slight reductions in molecular mass, O/C ratio, DBE, and AI_mod values of up to 5%. The highest removal of highly unsaturated compounds occurred at ML1 and ML2, with intermediate values at ML3. This differential removal caused a ~65% increase in unsaturated compounds and a ~49% increase in aromatic compounds in the remaining DOM fraction (Table 2). In October, the removal shifted toward aromatic compounds (~27%) and highly unsaturated compounds (~14%), lowering AI_mod and DBE values by up to 14% (Figures 8e–h; Table 3). The greatest removal was observed at ML1, while the lowest was at ML3, resulting in a ~119% enrichment of unsaturated compounds in the remaining DOM fraction. In December, the removal was less pronounced compared to October and September, primarily affecting aromatic and unsaturated compounds, which decreased by up to 39% and 42%, respectively (Figures 8i–l; Table 4). This resulted in a decrease in AI_mod by up to 14%. The highest removal occurred at ML2 and ML3 (18 and 24 m), leading to an enrichment of saturated compounds at these depths. Despite a higher percentage of saturated compounds being removed, their relative contribution remained low (<0.5%) in the bulk samples compared to other compound classes (Supplementary Figure 1).

In high-energy subterranean estuaries (STEs), the dynamic redox conditions, influenced by advection, enable iron to play a crucial yet intricate role in the transformation of dissolved organic matter (DOM). Iron alternates between its Fe³⁺ and Fe²⁺ states, acting as an electron acceptor during organic matter decomposition (Reckhardt et al., 2017; Zhou et al., 2023). From our experimental leaching experiment (Figure 2), we observed that DOM linked to poorly crystalline iron (extracted using 0.5 M HCl) showed higher levels of aromatic, oxygen-rich, and highly unsaturated compounds compared to porewater, along with some saturated labile compounds (Figure 6). The enrichment of aromatic, oxygen-rich, and highly unsaturated compounds was most pronounced in deep anoxic zones (Figure 6), consistent with previous iron oxyhydroxide adsorption studies (Lv et al., 2016; Linkhorst et al., 2017; Zhou et al., 2024). Furthermore, high leachable DOC concentrations in HCl extracts at ML1 and ML3 corresponded with elevated leachable Fe concentrations, Fe/DOC ratios, and finer grain sizes. Conversely, coarser sediments with lower Fe concentrations exhibited reduced leachable DOC in HCl extracts (Figure 4 and Supplementary Table 1). This pattern might be attributed to finer, clay-rich sediments enhancing organic matter retention. Iron, often associated with clay-rich sediments (Ilgen et al., 2019), likely contributes to organic matter preservation through various mechanisms. For instance, clay minerals can restrict oxygen penetration, creating anoxic microenvironments that slow down organic matter decomposition (Hulthe et al., 1998). Additionally, clay minerals provide extensive surface areas for the adsorption of organic molecules, increasing the sediment's organic matter retention capacity (Mayer, 1994a,b; Hulthe et al., 1998; Kaiser and Zech, 2000). The layered structure of clay minerals can also physically protect organic matter from microbial degradation (Mayer, 1994b). The lower degradation states of HCl leachates compared to the ambient porewater DOM support the protective effects of fine sediments against microbial degradation or DOM oxidation (Figure 5).

Additionally, during reoxygenation events triggered by porewater advection from oxic zones, dissolved Fe²⁺ can be reoxidized and coprecipitate with DOM. The resulting Fe³⁺ oxyhydroxides, with their large surface area, can further adsorb DOM (Chen et al., 2014; Zhou et al., 2018). Under reducing conditions, DOM bound to Fe³⁺-oxyhydroxides may be remobilized through microbial Fe reduction and release (Hu et al., 2023). We observed higher DOC and Fe²⁺ concentrations in anoxic porewaters at ML1 and ML2 (18–24 m) and ML3 (6–24 m) compared to oxic zones (Figure 3). At our study site, oxygen and nitrate penetrate deep into the sediments (Massmann et al., 2023). We hypothesize that frequent O₂ intrusion into deeper sediments, driven by porewater advection, may restrict the release of Fe²⁺ from sediments into porewater, thereby limiting the release of Fe³⁺-bound DOM.

In line with Zhou et al. (2024), we propose that poorly crystalline iron phases in deep beach sediments function as intermediate storage for terrestrial organic carbon in the deep STE. The quantity of DOM associated with poorly crystalline Fe phases was ~160 times greater than the ambient porewater DOM (Supplementary Table 1), indicating their significance as an intermediate reservoir for terrestrial DOM. Moreover, the loosely bound-labile DOM fraction of the sediments was about 130 times higher than the ambient porewater DOM, suggesting that advective flow in the deep STE may readily mobilize this DOM fraction. We also propose that Fe oxides/hydroxides in clay-rich minerals serve as key adsorption sites for terrigenous organic matter within deep STE sediments, facilitating long-term preservation through organo-mineral associations. In contrast, the labile DOM fraction—shown by Abarike et al. (2024) to rapidly stimulate microbial activity upon release—is readily mobilized by advective flow, creating a dynamic balance between carbon retention and transport. This dual mechanism establishes the STE as a biogeochemical reactor where: (1) mineral-bound DOM accumulates as a refractory reservoir, and (2) labile fractions either fuel subsurface microbial communities or are exported to coastal waters. The predominance of fine-grained, Fe-rich sediments enhances this selective preservation, while porewater flow governs the mobilization and fate of reactive DOM fractions.

In contrast to the findings of Linkhorst et al. (2017), who reported locally distinct reductions in DOC following Fe-DOM coprecipitation in shallow STEs (<5 m), our experimental study—conducted at the same site but focusing on the deep subsurface (>5 m)—yielded less consistent results. Anoxic porewater samples (18–24 m at ML1/ML2, 6–24 m at ML3; Figures 3) collected during three sampling periods showed little DOC changes after aeration within the precision range of the TOC analyzer ( ≤ 10%; Figures 7a–c). This suggests that Fe-DOM coprecipitation in the aqueous phase has a much lower impact on organic carbon mobility in deeper STEs compared to shallower depths. Reported dissolved iron and DOC concentrations in the upper meters of the porewater exfiltration zone at Spiekeroog beach were much higher (Reckhardt et al., 2015, 2017; Beck et al., 2017) than what we found in the deep STE porewaters, enhancing the Fe-DOM coprecipitation potential in the upper STE. In our study, there was no clear relationship between abundance of iron, DOC or iron-DOC molar ratios and observed removal of DOC. However, a neutral to alkaline pH of the porewater, together with Fe/DOC ratios of <0.5, indicate that the Fe-DOM coprecipitation potential could be overall low in this environment (Supplementary Table 2, Nierop et al., 2002; Du et al., 2018).

Despite the low Fe/DOC ratios and low amounts of removed DOC, some molecular changes were observed after coprecipitation. FDOM values decreased up to nearly 50% after coagulation, especially at ML3 (Figures 7d–f). The FDOM measured here is indicative of humic-like terrestrial and aquatic DOM. In previous studies, its abundance was significantly positively correlated with porewater dissolved iron concentrations, leading to the suggestion that reductive dissolution of iron oxides may lead to FDOM release into STE porewaters (Waska et al., 2019; Sirois et al., 2018). Our study results imply that the reverse reaction, e.g., immobilization of Fe and FDOM, can take place at STE redox interfaces as well. Still, not all FDOM was removed during coagulation. Within the defined excitation/emission spectra, FDOM fractions still encompass a diverse range of compounds, which may not all bind to oxidized iron. In line with FDOM reduction, throughout the STE, oxygen-rich aromatic and highly unsaturated compounds were removed as well (Figure 8). Spatially, ML1 and ML2 exhibited substantial removal of aromatic and unsaturated DOM fractions, whereas ML3 showed a more variable response (Figure 8; Tables 2–4). Although the observed fractionations in part supported the findings of Riedel et al. (2013) and Linkhorst et al. (2017), they too varied substantially across different seasons and stations.

While research on the composition, persistence, and inventory of sediment-bound organic matter both from the marine and terrestrial realm gained considerable traction in recent years (e.g., Zhao et al., 2018; Zhou et al., 2024), Fe-DOM coprecipitation experiments to date are fewer and still mainly employ terrestrial natural organic matter (Du et al., 2018). One comparative study found that marine (estuarine) DOM precipitating with ferric hydroxide was more saturated (higher H/C) than precipitates from terrestrially derived SRNOM (Suwannee River Natural Organic Matter) reference material (Li et al., 2019). For a shallow brackish hydrothermal vent system with high Fe²⁺ fluxes, Gomez-Saez et al. (2015) conducted precipitation experiments similar to ours. They found that while DOC and DOM from the fluids and adjacent seawater did not reveal strong fractionation patterns, the organic matter re-released from Fe-DOM precipitates was distinctly different and had higher O/C ratios than compounds which remained in the fluid. Compared to terrestrial DOM, marine DOM has smaller molecular sizes and overall different moieties involved in Fe complexation, and Fe-DOM binding could lead to an increase in Fe solubility, instead of Fe-DOM coprecipitation. For example, Fe-binding ligands in marine DOM were likened to fulvic acids in their acid-base properties (Lodeiro et al., 2020), and fulvic acids are characterized by higher solubility and lower Fe-coagulation potential than humic acids, which are common in terrestrial aquatic systems (Du et al., 2018).

Recent experimental work revealed that the amount of carboxyl groups which occur ubiquitously in natural DOM, can increase the affinity of a molecule to precipitate with ferrihydrite (Curti et al., 2021). We cannot determine structural entities like carboxyl groups from FT-ICR-MS-derived molecular formulae, but O/C ratios, as well as the number of carboxyl groups, are predicted to increase in the DOM pool with increasing degradation (Lechtenfeld et al., 2014). Therefore, one could speculate that (i) increasing O/C ratios as a function of DOM aging might increase the intensity of Fe-DOM coprecipitation of natural organic matter, (ii) mineral-bound organic matter contains elevated O/C ratios compared to ambient DOM, and that (iii) O/C ratios should decrease after Fe-DOM coprecipitation due to the removal of carboxyl-rich compounds. However, in our study, pore- and seawaters, rather than HCl sediment leachate samples, had the highest O/C ratios amongst all investigated sample types (Table 1), and although O/C ratios decreased slightly in porewater after Fe-DOM coprecipitation, this trend was not persistent (Tables 2–4). All in all, our and previous research indicate that the factors found to govern Fe-DOM coprecipitation in the terrestrial environment may not be entirely transferrable to marine systems, both due to intrinsic DOM properties and environmental conditions.

From our experimental study, we found a remarkable change in FDOM distribution patterns across different months (September, October and December 2022) and a strong decrease in FDOM values after Fe-DOM coprecipitation (Figures 7a–c). Since we only observed high FDOM values in one of the three campaigns, we cannot know whether it was a unique occurrence or may happen more often during the year. The September 2022 campaign was the first where all three wells were installed and ready to sample, and ML3 had to this point only been installed for a month, which may have caused a disruption of sediment stratigraphy and flow regime, and perhaps suspension of organic-rich particles. However, environmental parameters like salinity and oxygen, DOC, and dissolved iron concentrations were in the same range as in the October and December campaigns (Figure 3), indicating that the samples extracted from the wells were in equilibrium with the surrounding groundwater. Nevertheless, long-term observations will be necessary to confirm the extraordinary FDOM dynamics found here.

In addition to FDOM, spatio-temporal variations in the molecular changes of DOM following coprecipitation with Fe³⁺ (Figure 8) were found in all three campaigns. This pattern could be attributed to temporal shifts in DOM signatures across different zones of the deep STE, which may be driven by the intense mixing of fresh groundwater and seawater in high-energy STEs (Greskowiak et al., 2023). For example, in September, samples exhibited the highest removal of oxygenated and highly unsaturated compounds (Figures 8a–c). Conversely, substantial removal of aromatic compounds was observed in October (Figures 8d–f). In December, almost no removal of specific molecular formulae was found after coagulation (Figures 8d–i). These observed differences in DOM fractions removed through Fe-DOM coprecipitation were consistent with overall changes in the molecular composition of the whole DOM geometabolome across the three different campaigns (Figure 8). This suggests that the temporal, transport-driven dynamics of DOM composition influence the Fe-DOM coprecipitation process within the STE. Moreover, these temporal shifts in DOM regimes not only influence the types of organic compounds preferentially removed but also impact the overall quantity of DOM that undergoes Fe-DOM coprecipitation at any given time. The dynamic interplay between changing DOM inputs and iron-mediated coagulation suggests that the STE functions as a temporally variable filter for organic matter transport from land to sea. The efficiency and selectivity of this filtration process are likely to fluctuate in response to temporal and event-driven changes in DOM inputs, potentially leading to variations in both the quantity and quality of organic matter ultimately exported to coastal waters. Ultimately, while environmental parameters (Figure 3) did not vary significantly across the three campaigns, we observed molecular fractionation of DOM composition following aeration, which differed across the studied months (Figure 8). Therefore, we recommend that future studies prioritize investigating the spatio-temporal dynamics of DOM composition in the deep STE to better understand its reactivity with iron under changing redox conditions, typical of high-energy STEs (Greskowiak et al., 2023).