Dissolved organic matter (DOM) is a critical flux of organic carbon transported to the ocean from continents, yet how its photoreactivity varies in estuaries remains poorly understood. Traditional approaches overlook compositional relationships between DOM molecular formulas (MFs) and therefore fail to resolve the molecular ambiguity where identical formulas potentially exhibit divergent reactivity across different environments. Here, we embedded domain knowledge across multiple critical dimensions into a customized two-step graph network framework to represent MFs detected by ultra-high-resolution mass spectrometry. The framework was trained on irradiation experiments from major estuaries in China and includes a primary network to capture global chemical relationships and secondary networks to represent sample-specific local features. We further applied these graph representations to develop machine learning models able to predict the photoreactivity of molecules in field samples indicative of their transformation potential, which we validated on non-irradiated samples. This graph-based approach can effectively distinguish identical MFs under different systems and represent their environmental dependence. Results reveal that estuaries significantly modulate the photochemical reactivity of MFs and that the modern, photo-labile component of terrestrial carbon flux is transient in estuarine systems, highlighting the role of photochemical mechanisms in transforming DOM towards a more marine-like signature. This study demonstrates that biogeochemistry-informed graph-based models can not only characterize DOM dynamics in estuarine environments through which massive amounts of terrestrial carbon traverse en route to the ocean, but also contribute to more accurate ecological forecasting of carbon dynamics across the land-ocean continuum under changing environmental conditions.
Seasonal variability in dissolved trace-metal behavior and river-to-sea fluxes remains insufficiently constrained in large river estuaries, limiting our understanding of how estuarine processing modifies river-borne metal signals. This study investigates three seasonal cruises to examine the distributions, estuarine behavior, and fluxes of dissolved Fe, Ni, Cu, Cd, and Pb in the Changjiang Estuary. Dissolved Fe consistently exhibited strong non-conservative removal, with removal rates ranging from 57 to 85%. Conversely, dNi and dCd displayed persistent addition, with dCd showing an exceptional addition rate of 1520% in spring. Dissolved Cu showed distinct seasonal variability, shifting from non-conservative addition in spring and autumn to near-conservative mixing during the high-discharge summer cruise, which may reflect the combined effects of river discharge and particle–solution exchange in the estuary. Dissolved Pb behaved quasi-conservatively; the scattering data may result from localized anthropogenic inputs. Accounting for seasonal variability, the estimated annual net outputs of the five dissolved trace metals spanned approximately 105–107 mol year−1, with dCu and dNi showing the largest fluxes. Comparing these results with 1980s data reveals a dramatic decline in dFe flux, with 2–4-fold increases in dNi, dCu, and dCd fluxes, likely driven by rapid industrialization over recent decades. The Changjiang contributes ~70% of the combined riverine dCu and dCd inputs to the Northwest Pacific marginal seas, highlighting the importance of accounting for seasonal estuarine processing when constraining regional river-to-ocean trace-metal fluxes.
Suspended particulate organic matter (POM) plays a central role in carbon cycling and trophic transfer in Antarctic waters, yet its sources, composition, and degree of transformation remain poorly constrained in nearshore Antarctic Peninsula. Here, we characterize the elemental (C and N), isotopic (δ13C and δ15N), and lipid biomarker (alcohols, sterols, and fatty acids) composition of suspended POM collected in Admiralty Bay and the Bransfield Strait, south of King George Island, during late austral spring 2019. Surface samples were collected at 12 stations, whereas DCM samples were collected at 6 of these stations. Suspended particulate matter concentrations ranged from 5.4 to 21.9 mg L−1 and showed marked variability between the two sampling campaigns. Molar C/N ratios and lipid biomarker distributions indicate a predominance of marine-derived, phytoplankton-associated OM, largely linked to diatom-rich assemblages with minor contributions from other marine sources. Diagnostic lipid biomarker ratios further suggest that suspended POM was relatively fresh and had undergone limited early diagenetic alteration. Paired comparisons between surface and DCM samples revealed no statistically significant depth-related differences, indicating limited vertical variability in OM composition within the upper water column. The results show that physical forcing primarily regulates particle loads, whereas biological production exerts stronger control on OM composition and quality, highlighting the sensitivity of Antarctic coastal carbon cycling to ongoing environmental changes along the Antarctic Peninsula.
The Tatar Trough lies off the eastern coast of Far Eastern Russia, extending into the northern part of the East Sea (also known as the Sea of Japan). This region is characterized by active tectonics and widespread gas seepage, making it an ideal natural laboratory for investigating the biogeochemical dynamics of gas-rich sediments. In this study, we analyzed two sediment cores, LV67-07HC and LV67-19HC, which were recovered from an active fault zone on the eastern slope during the 2014 SSGH expedition. Using a combination of lipid and nucleic acid analyses along with complementary measurements such as gas and porewater composition, we assessed the ecological roles of the archaeal communities inhabiting these seep environments. Both cores exhibited elevated concentrations of light hydrocarbon gases (primarily methane, CH4) and carbon dioxide, along with a distinct sulfate-methane transition zone (SMTZ). However, notable differences in porewater geochemistry, lipid biomarkers, and microbial community composition were observed between the two cores, reflecting variations in CH4 flux and subsurface fluid migration pathways within the underlying coal-bearing strata. In core LV67-07HC (91-185 cm depth), we observed notably high concentrations of 13C-depleted archaeal lipids in sediments surrounding carbonate accretions. These features suggest enhanced anaerobic methanotrophic activity under relatively higher CH4 flux conditions. Archaeal 16S rRNA gene sequences were dominated by members of the ANME-1 clade, consistent with previous observations in CH4-rich seep environments, and further constrained here within a site-specific, multi-proxy framework. To the best of our knowledge, this study provides new insights into how spatial differences in CH4 sources and subsurface fluid-migration pathways influence carbon cycling and archaeal community structure in CH4-rich continental margin sediments.
