Contrasting fate and GHG emissions of algal vs. manure DOM under light and microbes in lakes

Summary

The escalating dissolved organic matter (DOM) inputs from algal (ADOM) and livestock manure (MDOM) sources are altering lake ecosystems, yet their fate under light and microbes in steppe lakes is unclear. Using microcosms, we found that photodegradation suppressed ADOM bioavailability by 4.8% via substrate competition and reactive oxygen species (ROS) inhibition (especially singlet oxygen), while light enhanced MDOM bioavailability by 5.4%. ADOM mineralization released 1.6× more CO₂ than MDOM. Multi-group structural equation modeling (SEM) indicated that efficient ADOM microbial transformation drove higher CO₂ emissions, whereas light-induced ROS shifted microbial communities and transformation pathways. ADOM and MDOM inputs predominantly enhanced CO₂ emissions and water browning, respectively. These processes are interconnected—exogenous MDOM can promote algal blooms and subsequent greenhouse gas release. For eutrophic steppe lakes, integrated management combining exogenous load reduction and endogenous bloom suppression is critical to mitigate ecological risks and greenhouse effects.

Graphical abstract

Highlights

• •Light oppositely regulates ADOM (inhibited) and MDOM (promoted) biodegradation
• •Singlet oxygen (¹O₂) selectively inhibits ADOM-degrading bacteria
• •ADOM mineralization releases more CO₂ than MDOM via efficient microbial processing
• •Light alters DOM transformation pathways to CO₂ through ROS and substrate shifts

Microbiology; Aquatic science; Aquatic biology; Environmental Science

Introduction

Lakes are pivotal components of the global carbon cycle, functioning as critical hubs for the transport, transformation, and long-term storage of carbon.^1,2,3 Dissolved organic matter (DOM) represents the most dynamic pool within this cycle, mediating essential geochemical processes and exerting a profound influence on aquatic ecosystem function through its interactions with microorganisms and light.^4,5 The environmental fate of DOM—whether it is rapidly mineralized to greenhouse gases (GHGs) or processed into stable, recalcitrant forms—directly controls the climatic feedback and carbon sequestration potential of inland waters.^6,7,8 Critically, this fate is not merely a biogeochemical endpoint but a process that generates byproducts—including GHGs and reactive oxygen species (ROS)—that exert profound influences on aquatic biogeochemistry, ecosystem health, and climate feedback.^9,10

Human activities are dramatically altering the quantity and character of DOM entering lakes. The intensification of agricultural practices and climatic warming has fueled widespread algal blooms, making algal-derived DOM (ADOM) a dominant and growing autochthonous source with high bioavailability.¹¹ Endogenous ADOM is generally characterized by a higher contribution of relatively labile, protein- and amino-acid-like components with high bioavailability, making it readily accessible to microbial degradation.¹² Concurrently, global livestock production, which relies heavily on grassland-based systems covering ∼25%–30% of ice-free land, generates vast quantities of manure.¹³ The runoff and leaching of this manure introduces a massive, allochthonous flux of manure-derived DOM (MDOM) into aquatic networks. In contrast to ADOM, MDOM is typically more humic-like, with higher molecular weight, stronger light absorption, and higher aromaticity, rendering it more recalcitrant to microbial attack.¹⁴ This input acts as a significant vector for nutrient pollution and is estimated to contribute substantially to anthropogenic GHG emissions, with livestock systems responsible for 4.5–6.5 billion tons of CO₂ equivalent annually.¹³ While the role of MDOM in promoting heavy metal mobilization has been investigated,^15,16,17 its fundamental role within the aquatic carbon cycle has been systematically overlooked.

In sunlit surface waters, DOM undergoes coupled photochemical and microbial transformations, a nexus that ultimately determines its environmental impact and hazard potential.⁶ The composition of DOM, a complex mixture from diverse sources, is a key control on the quantity and mix of GHGs released during its degradation.^18,19 Photodegradation can prime DOM for microbial respiration, enhancing CO₂ production by generating labile, low-molecular-weight compounds.^20,21 However, this same process also photochemically produces ROS—including hydroxyl radicals (·OH), singlet oxygen (¹O₂), and excited triplet-state chromophoric DOM (³CDOM∗)—which play a dual role in aquatic photochemistry. On one hand, ROS such as ¹O₂ can oxidize proteins and lipids, inducing microbial stress and potentially altering community function, thereby inhibiting carbon mineralization. On the other hand, ROS also contribute to the oxidative degradation of organic pollutants and natural water self-purification.^22,23,24 The net effect of light—whether it amplifies or attenuates DOM mineralization and GHG emissions—is thus a balance between this priming and inhibition, a balance that may be critically dependent on the intrinsic chemical composition of the DOM source. A mechanistic understanding of this source-dependent response is essential for accurately forecasting the environmental risks posed by different DOM inputs.

Despite the recognized importance of both ADOM and MDOM in lake carbon cycling, their distinct molecular compositions and environmental fates have not been systematically compared with a single experimental framework. In particular, the differential effects of ADOM and MDOM on CO₂ emission potential and the role of ROS as a source-dependent modulator of microbial carbon processing remain poorly understood. Filling this knowledge gap is essential for predicting how ongoing eutrophication and agricultural intensification will alter the GHG budgets of northern steppe lakes. Therefore, the central scientific question we address is how do the contrasting intrinsic properties of autochthonous ADOM versus allochthonous MDOM differentially regulate CO₂ emission potential and ROS-mediated microbial degradation under coupled photochemical and microbial processes? Specifically, we test the following hypotheses: (1) ADOM exhibits higher bioavailability and thus greater CO₂ emission potential than MDOM; (2) light enhances microbial mineralization of MDOM, while exerting a weaker or inhibitory effect on ADOM; and (3) ROS production and its consequent microbial impacts are source-dependent, serving as a key modulator of carbon processing that differs markedly between ADOM and MDOM.

To test these hypotheses, we conducted an 85-day controlled incubation experiment comparing ADOM and MDOM under four critical treatments: dark control, light-only (photodegradation), microbial-only (biodegradation), and coupled light-microbial exposure. By integrating bulk DOC analysis, GHG flux measurements, ROS quantification, microbial functional gene abundance (via qPCR), and molecular-level DOM characterization, we aimed to (1) quantify how DOM source regulates the net effect of light on microbial carbon mineralization, (2) differentiate the CO₂ emission potentials attributable to DOM source and degradation pathway, and (3) elucidate the pathways through which ROS production mediates microbial carbon processing and GHG fluxes. To our knowledge, this is the first study to systematically compare ADOM and MDOM within a single experimental framework, integrating DOM molecular characterization, ROS quantification, and microbial community analysis to unravel source-dependent photochemical-microbial interactions. This study provides a comparative risk assessment of the transformation fates of two major DOM sources, offering new insights into their distinct roles in aquatic carbon cycling and GHG emissions.

Results

Two-phase degradation kinetics and endpoint removal efficiencies of ADOM and MDOM

The degradation of dissolved organic carbon (DOC) from both ADOM and MDOM over the 85-day incubation followed a distinct two-phase pattern, with a rapid initial phase followed by a slower second phase, across all treatments (Figure 1; Table 1). These biphasic dynamics were well captured by the first-order exponential decay model, as evidenced by high coefficients of determination (R² > 0.69 for phase 1 and R² >0.71 for phase 2 in microbially active treatments).

Figure 1.

Temporal changes in DOC concentrations

Changes in the DOC concentration of ADOM (A) and MDOM (B) over time. Data are represented as mean ± standard error of the mean (n = 3). Error bars are smaller than the symbols at most time points. The light orange background indicates the rapid degradation phase, the light blue indicates the slow degradation phase, and the gray dashed line marks the boundary between the two phases.

Table 1.

Two-phase degradation rate constants and model fit for ADOM and MDOM under four treatments during 85-day incubation

DOM_source	Treatment	Phase 1_rate	Phase 1_R2	Phase 2_rate	Phase 2_R2
ADOM	control	0.30 ± 0.22	0.38	−0.13 ± 0.08	0.60
ADOM	light	0.27 ± 0.14	0.56	−0.08 ± 0.02	0.89
ADOM	microbial	2.44 ± 0.95	0.69	0.08 ± 0.01	0.94
ADOM	coupling	2.38 ± 0.63	0.83	0.17 ± 0.05	0.88
MDOM	control	0.22 ± 0.09	0.58	−0.05 ± 0.03	0.71
MDOM	light	0.42 ± 0.11	0.79	0.07 ± 0.04	0.72
MDOM	microbial	1.52 ± 0.44	0.75	0.05 ± 0.01	0.98
MDOM	coupling	1.40 ± 0.42	0.74	0.22 ± 0.05	0.95

Data are represented as mean ± standard error of the mean (n = 3). See also Table S1.

For ADOM, the microbial treatment exhibited the fastest phase 1 rate constant (2.44 ± 0.95 day⁻¹), nearly an order of magnitude higher than that of the control (0.30 ± 0.22 day⁻¹) or light-alone (0.27 ± 0.14 day⁻¹) treatments. The coupling treatment slightly reduced the phase 1 rate (2.38 ± 0.63 day⁻¹) compared to the microbial treatment, but the difference was not pronounced. Phase 2 rates were substantially lower in all treatments, with the microbial treatment showing a rate of 0.08 ± 0.02 day⁻¹, while the coupling treatment yielded a slightly higher phase 2 rate of 0.17 ± 0.05 day⁻¹, possibly reflecting a shift in the composition of remaining DOC. In contrast, the control and light-alone treatments displayed slightly negative phase 2 rates, suggesting negligible net degradation or minor analytical variability within the detection limits.

For MDOM, the phase 1 degradation rate was also highest in the microbial treatment (1.52 ± 0.44 day⁻¹), followed by the coupling treatment (1.40 ± 0.42 day⁻¹). Light-alone (0.42 ± 0.11 day⁻¹) moderately accelerated the initial decay relative to the control (0.22 ± 0.09 day⁻¹), indicating a certain photochemical lability of MDOM. The phase 2 rate of the coupling treatment (0.22 ± 0.05 day⁻¹) was notably higher than that of the microbial treatment (0.05 ± 0.01 day⁻¹), implying that the combination of light and microbes facilitated the breakdown of more recalcitrant components in the later stage.

When the 85-day endpoint degradation efficiencies were corrected for the background lake water contribution, pronounced and contrasting effects of light on microbial DOC removal were revealed between the two DOM sources. For ADOM, microbial degradation alone was highly efficient, removing 91.9% ± 0.3% of the initial DOC after 85 days. However, the simultaneous exposure to light in the coupling treatment significantly inhibited this microbial process, reducing the degradation efficiency to 87.1% ± 1.2% (one-way ANOVA, p = 0.019), corresponding to a 4.8% inhibition. In stark contrast, for MDOM, light significantly enhanced microbial degradation. While the microbial-alone treatment achieved 69.5% ± 0.3% removal, the coupling treatment reached 74.9% ± 1.6% (one-way ANOVA, p = 0.024), representing a 5.4% promotion. The control and light-alone treatments exhibited much lower endpoint removals (ADOM: near-zero or slightly negative; MDOM: 5.3% and 19.4%, respectively), confirming that both biotic and photochemical processes alone were insufficient to substantially mineralize the bulk DOC pool without the synergistic or antagonistic interaction between light and microbes.

Characterization of DOM molecular composition

The initial molecular composition of ADOM and MDOM is presented in Figures 2A and 2B, with 2,608 and 5,236 molecules identified, respectively (Table S2). Although the two sources differed significantly in their initial composition (Figure S1), both were dominated by three major compound classes in decreasing relative abundance: highly unsaturated structures (HUS), aliphatics, and peptides (Table S2). During incubation, the trends in these major components varied depending on the treatment, while the overall patterns were largely consistent between the two DOM sources.

Figure 2.

Van Krevelen diagrams showing initial DOM molecular composition and its variation

(A and B) Initial molecular formulas identified in ADOM and MDOM, respectively. See also Table S2.

(C) Temporal variation in compound classes across different treatments. See also Figures S1 and S3.

(D) Temporal variation of the carboxyl-rich alicyclic molecule (CRAM)-like region under different treatments. Dashed lines delineate molecular formula categories: condensed aromatic-like (CAS), aromatic-like (AS), highly unsaturated-like (HUS), aliphatic-like, and peptide-like. The solid yellow shaded area represents the CRAM-like region.

In (C) and (D), the x axis labels denote the sample name, formatted as [day][treatment][DOM source]. For example, “D10_Lig_ADOM” refers to the ADOM sample from the light treatment on day 10. Lig, light treatment; Mic, microbial treatment; Cou, coupling treatment.

In the light treatment, the relative intensity of more bio-labile components (aliphatics + peptides) increased, whereas that of the more refractory HUS decreased. In contrast, the microbial treatment showed an opposite trend: the relative intensity of aliphatics + peptides decreased, while HUS increased. The coupling treatment largely followed the same pattern as the microbial group. With respect to carboxyl-rich alicyclic molecules (CRAMs), their relative abundance decreased over time in the light and coupling treatments but increased in the microbial treatment. After 85 days of incubation, the final molecular composition under all treatments showed varying degrees of convergence compared to the initial state (Figure S1).

Bacterial abundance and community composition

Bacterial abundance in both the microbial and coupling treatments showed a consistent pattern of initial growth followed by decline for both ADOM and MDOM (Figure S2). ADOM treatments peaked earlier (day 10) than MDOM, and the average abundance was significantly higher in the microbial-only group (p = 0.022, Figure S2A). For MDOM, peak abundance occurred earlier in the microbial group (day 5) than in the coupling group (day 10), with the latter reaching a higher peak value. Microbial activity in the control and light-alone groups was effectively suppressed, as confirmed by plate culture. Although a minor increase was observed in the MDOM light group, its abundance remained over four orders of magnitude lower than in the biotic treatments.

Significant shifts in bacterial community composition at the phylum level were observed throughout the incubation (Figure 3A). Initially, the ADOM-associated community was dominated by Proteobacteria (89.25%). In contrast, the starting MDOM community exhibited higher complexity, primarily composed of Proteobacteria (50.72%), Patescibacteria (17.87%), and Verrucomicrobiota (16.90%). During the microbial-only incubation, communities from both DOM sources became less diverse when microbial biomass peaked at the mid-incubation stage, with Proteobacteria becoming overwhelmingly dominant. By the end of the incubation, community complexity partially recovered: in ADOM, Proteobacteria and Planctomycetota were the dominant phyla, whereas in MDOM, Proteobacteria and Patescibacteria prevailed. Under the coupling treatment, Bacteroidota emerged as the second most abundant phylum alongside Proteobacteria. The relative abundance of Bacteroidota increased notably during the mid-incubation period when microbial biomass was highest. Non-metric multidimensional scaling (NMDS) analysis based on phylum-level composition (Figure 3B) confirmed that the initially distinct communities of ADOM and MDOM gradually converged over time within each treatment. Nevertheless, clear separations were maintained between the microbial-only and coupling treatments throughout the experiment.

Figure 3.

Changes in the microbial community composition during the incubation

(A) Stacked bar plot showing the relative abundance of major bacterial phyla.

(B) NMDS ordination plot based on the phylum-level community composition.

In (A), the x axis labels indicate the sample names, formatted as [day][treatment][DOM source]. For example, “D0_Mic_ADOM” represents the ADOM sample from the microbial treatment on day 0. Mic, microbial treatment; Cou, coupling treatment. In (B), the plot illustrates the shifts in samples across different incubation periods within the NMDS space. The size of the circles corresponds to the incubation time, with larger diameters indicating longer incubation periods. The sample naming convention in the legend follows the same format as in (A). See also Figure S2.

Changes in ROS concentrations

The cumulative concentrations of ¹O₂, ³CDOM∗, and ·OH followed similar temporal patterns across DOM sources and ROS types (Figure 4). ROS production increased rapidly during the first 10–20 days, after which the accumulation rate slowed. Notably, while ADOM showed a more pronounced leveling off, MDOM continued to generate ROS at a low but sustained rate thereafter. Minimal ROS were detected in the sterile dark controls, consistent with the established findings.²⁵ Low but measurable levels of all three ROS were observed in the microbial treatments. Substantially higher ROS accumulation occurred in both the light and coupling treatments. Notably, the dominant treatment differed between DOM sources: for ADOM, ROS concentrations were higher in the light group than in the coupling group, whereas for MDOM, the coupling group produced more ROS than the light group. Overall, MDOM demonstrated a greater ROS-production potential than ADOM across treatments.

Figure 4.

Accumulated concentrations of ROS

Accumulated concentrations of ¹O₂ (A and B), ³CDOM (C and D), and ·OH (E and F) in ADOM (left column) and MDOM (right column) over time. Data are represented as mean ± standard error of the mean (n = 3).

Changes in GHG flux

For both ADOM and MDOM, the four treatments generally functioned as net sinks for CO₂. To investigate the impact of DOM input on CO₂ flux (FCO₂), we focus on the FCO₂ values corrected for the background lake water signal. Between days 2 and 85, FCO₂ in all four treatments for both ADOM and MDOM showed a generally consistent trend of continuous decline (Figures 5A and 5B). The control groups exhibited minor flux variations, with values oscillating between positive and negative. The light groups showed greater variation, displaying positive values before day 10 and switching to negative thereafter. In contrast, both the microbial and coupling groups consistently yielded positive CO₂ fluxes, except for the ADOM microbial group on day 85; fluxes in these groups exhibited a relatively smooth or gradual decline over time.

Figure 5.

CO₂ flux and cumulative release

Timecourse of background-subtracted, DOC-normalized CO₂ flux (A and B), and cumulative release (C and D) for ADOM (A and C) and MDOM (B and D) across four treatments. FCO₂ (mmol m⁻² day⁻¹) is the background-corrected CO₂ flux normalized to initial DOC concentration; cumulative release (mmol m⁻²) represents the time-integrated sum of FCO₂ from day 0 to each sampling time point (i.e., cumulative emission up to that day). Data are represented as mean ± standard error of the mean (n = 3).

Regarding the mean flux, for ADOM, the control, light, microbial, and coupling groups were −0.148 ± 0.268, 0.703 ± 0.532, 1.808 ± 0.294, and 1.755 ± 0.333 mmol m⁻² day⁻¹, respectively. No significant differences were found among the light, microbial, and coupling groups (p > 0.05), though all were significantly higher than the control (p < 0.05). For MDOM, the corresponding mean fluxes were −0.237 ± 0.190, 0.461 ± 0.333, 1.378 ± 0.210, and 1.182 ± 0.207 mmol m⁻² day⁻¹. Similarly, the light, microbial, and coupling groups showed no significant inter-group differences (p > 0.05) but were significantly higher than the control (p < 0.05). Notably, the negative background-corrected FCO₂ in the control treatments (e.g., −0.148 ± 0.268 mmol m⁻² day⁻¹ for ADOM) reflects abiotic CO₂ uptake under alkaline pH conditions (pH > 8.6) in the absence of biological activity, where the carbonate equilibrium favors CO₂ dissolution. In contrast, the positive values observed in the light, microbe, and coupling treatments confirm that DOM addition drives net CO₂ release when biological or photochemical mineralization is active. Overall, FCO₂ was generally greater for ADOM than for MDOM across corresponding treatments. Specifically, the carbon-normalized mean FCO₂, averaged across the microbial and coupling treatments, was 1.6 times higher than that in the corresponding MDOM groups (0.021 ± 0.003 vs. 0.013 ± 0.002 mmol m⁻² day⁻¹, p = 0.01), indicating a significantly greater CO₂ emission potential from ADOM compared to MDOM. Consistent with this rate difference, the cumulative CO₂ release (Figures 5A and 5B) over the entire incubation period showed a similar numerical trend (1.11 ± 0.22 vs. 0.68 ± 0.10 mmol m⁻²), though this difference did not reach statistical significance (p = 0.1085).

FCO₂ was closely coupled with microbial activity and the kinetics of DOM degradation, with greater CO₂ release observed during the rapid DOC degradation phase for both ADOM and MDOM. Specifically, for ADOM, the mean FCO₂ in the microbial group during the rapid degradation phase was 2.696 ± 0.186 mmol m⁻² day⁻¹, significantly higher than that during the slow degradation phase (0.624 ± 0.368 mmol m⁻² day⁻¹, p = 0.002). Similarly, in the coupling group, FCO₂ during the rapid phase (2.522 ± 0.390 mmol m⁻² day⁻¹) was significantly higher than that during the slow phase (0.733 ± 0.378 mmol m⁻² day⁻¹, p = 0.022). For MDOM, the microbial group showed no significant difference between the rapid (1.783 ± 0.134 mmol m⁻² day⁻¹) and slow phases (0.367 ± 0.452 mmol m⁻² day⁻¹, p = 0.137), whereas the coupling group exhibited significantly higher FCO₂ during the rapid phase (1.621 ± 0.166 mmol m⁻² day⁻¹) compared to the slow phase (0.086 ± 0.268 mmol m⁻² day⁻¹, p = 0.005).

Direct and indirect effects of DOM sources and treatments on FCO₂

Multi-group structural equation modeling (SEM) was employed to investigate the drivers of FCO₂, utilizing data on microbiome composition and coupled processes. The baseline model revealed direct controls on FCO₂: a negative effect by pH and positive effects by both DOC concentration and log-transformed bacterial abundance (LogBA). All three ROS influenced FCO₂ indirectly through bacterial mediation. Specifically, ³CDOM∗ exhibited a positive indirect effect, whereas ¹O₂ and ·OH showed negative indirect effects (Figure 6A). The baseline model demonstrated excellent fit to the data (comparative fit index, CFI = 0.98; Tucker-Lewis index, TLI = 0.97; root mean square error of approximation, RMSEA = 0.05; p = 0.27). This model explained a substantial proportion of the variance in the endogenous variables, with R² values of 0.51 for LogBA and 0.30 for FCO₂ (Figure 6A). These R² values indicate that the selected predictors accounted for 51% of the variation in bacterial abundance and 30% of the variation in FCO₂, respectively.

Figure 6.

Structural equation models illustrating direct and indirect effects on CO₂ flux

(A) Results of the baseline structural equation modeling.

(B) Path comparison between multi-group SEMs based on DOM sources and treatments.

See also Table S3. Solid red and blue arrows represent significant positive and negative effects, respectively; arrow width is proportional to the effect size. Numbers adjacent to colored arrows indicate standardized path coefficients. Gray dashed arrows are for visual guidance only. Significance levels: ∗p < 0.05, ∗∗p < 0.01, and ∗∗∗p < 0.001.

For the multi-group comparisons, the DOM source grouping model (ADOM vs. MDOM) showed excellent fit (CFI = 0.99, TLI = 0.99, RMSEA = 0.01, p = 0.44), while the treatment grouping model (microbial vs. coupling) showed acceptable fit given the reduced sample size per subgroup (CFI = 0.93, TLI = 0.84, RMSEA = 0.12, p = 0.08; Table S3).Measurement invariance tests indicated that the DOM source (ADOM vs. MDOM) significantly altered the model structure (Δχ² (8) = 18.85, p = 0.016), with a particularly notable difference in the strength of the FCO₂ ∼ DOC and path pH ∼ DOC (p = 0.027 and 0.047, respectively; Figure 6B). Correspondingly, the treatment (microbial vs. coupling) also had a significant effect on the model (Δχ² (8) = 19.48, p = 0.012). Further analysis pinpointed significant differences in two key paths: FCO₂ ∼ DOC and LogBA ∼ ¹O₂ (p = 0.018 and 0.021, respectively; Figure 6B).

Discussion

Source-dependent dual pathways of photodegradation on microbial DOM processing

Light influences microbial degradation of DOM through both direct and indirect pathways, with the net effect being strongly dependent on DOM sources.^20,26 Direct effects include UV-induced damage to microorganisms and photochemical alteration of substrates, while indirect effects primarily involve the photoproduction of ROS that subsequently affect microbial activity.^20,27 These mechanisms typically co-occur.²⁷ The net photo effect on DOM biodegradation is largely determined by whether it ultimately promotes or inhibits the growth and respiration of indigenous microbial communities.^19,20 Although substrate alteration mediated by direct photochemistry can be decisive,^19,20 the indirect effects mediated by ROS are also significant.^28,29

For ADOM, photodegradation and microbial utilization targeted a highly overlapping pool of molecules, creating a competitive relationship for substrates (Figure S3A). This competition was compounded by the photoproduction of ROS. Multi-group SEM revealed that the effect of ¹O₂ on bacteria was source dependent, showing a significant negative effect specifically in the ADOM system (Figure 6B; Table S3). Mechanistically, this ¹O₂-mediated inhibition in ADOM can be attributed to two non-mutually exclusive pathways. First, ¹O₂ is known to oxidize peptide-like compounds—which are abundant in ADOM—potentially generating toxic byproducts or directly damaging microbial cell membranes and enzymes, thereby inducing bacterial stress or mortality.³⁰ Second, ¹O₂ may competitively consume these same peptide-like substrates that would otherwise be available for bacterial assimilation, effectively reducing the carbon and energy supply for microbial growth.³¹ These mechanisms are expected to be specific to ADOM, as MDOM is intrinsically poorer in protein-like components and less susceptible to such ¹O₂-mediated stress (as discussed further).This dual stress of substrate competition and ROS inhibition likely explains why microbial abundance in the coupling treatment was significantly lower than in the microbial treatment (p = 0.022, Figure S2A) and accounts for the significant reduction in the net degradation rate observed under coupling conditions.

In contrast, for MDOM, the substrate competition with microbes was weaker than for ADOM (Figure S3B). Furthermore, light exposure appeared to prime MDOM for enhanced microbial utilization. Fourier transform ion cyclotron resonance mass spectrometry (FT-ICR MS) results indicated that a greater proportion of bioavailable molecules (e.g., aliphatics) were generated from MDOM upon irradiation compared to ADOM (Figure 2). The absence of a significant difference in average microbial abundance between the MDOM microbial and coupling treatments (p = 0.214, Figure S2B) further supports that microbial growth was not inhibited by light in the MDOM coupling system. Consistent with this, multi-group SEM indicated no significant inhibitory effect of ¹O₂ on the bacterial community in the MDOM system (Table S3). This photochemical facilitation, combined with the lack of strong ¹O₂-mediated inhibition, resulted in the observed synergistic enhancement of degradation in the coupling treatment for MDOM.

The molecular-level mechanisms underlying these source-dependent photochemical-microbial interactions are explored in detail in a complementary study, which identifies lignin-like molecules as the key drivers of degradation pathway divergence.³²

Differential effects of DOM source and treatment on FCO₂ regulation mechanisms

Based on multi-group SEM, this study further reveals significant differences in the regulatory mechanisms of FCO₂ by the DOM source and treatment. Although no statistically significant differences in FCO₂ were observed between the treatment groups, the DOM source directly led to a clear divergence in flux levels—the carbon-normalized FCO₂ of the ADOM group was 1.6 times that of the MDOM group. The SEM results provided a mechanistic explanation for this phenomenon from the perspective of pathways. The stepwise invariance testing framework (from the unconstrained configural model to the constrained structural invariance model) allows us to move beyond detecting overall structural differences and to pinpoint exactly which pathways drive the observed divergences between groups.

For comparison between ADOM and MDOM sources, model fitting indicated significant structural differences between the two groups (Δχ² = 18.85, p = 0.016). Within the structural invariance step, where individual path coefficients were constrained to be equal across groups, the test revealed that the most pronounced divergence lay in the DOC∼FCO₂ pathway. Specifically, the direct positive effect of DOC on FCO₂ was stronger in the ADOM group (path coefficient = 0.727) compared to that of the MDOM group (0.410), indicating that ADOM possesses higher bioavailability and conversion efficiency, making it more readily metabolized by microbes and converted to CO₂.^33,34,35 Additionally, the direct effect of bacterial abundance (LogBA) on FCO₂ was more pronounced in the ADOM group, further supporting the mechanistic advantage of algal-derived carbon in driving emissions through microbial processes.

For comparison between the microbial and coupling treatments, significant structural differences were also found between the two groups (Δχ² = 19.48, p = 0.012). The stepwise invariance tests further identified the FCO₂∼DOC (p = 0.018) and LogBA∼¹O₂ (p = 0.021) pathways as the primary sources of structural divergence. The FCO₂∼DOC pathway showed a markedly stronger positive effect in the coupling treatment (0.478, p = 0.002) than in the microbial treatment (0.001, p = 0.995), indicating that photochemical processes substantially enhance the conversion efficiency of DOC to CO₂. For the LogBA∼¹O₂ pathway, the coupling treatment exhibited a significant negative effect of ¹O₂ on bacterial abundance (−0.617, p = 0.002), whereas no such effect was observed in the microbial treatment (0.110, p = 0.531). As discussed earlier, this ¹O₂-mediated inhibition involves oxidative stress and substrate competition, mechanisms particularly pronounced for ADOM. The convergence of these two pathways reveals a systemic shift: light simultaneously enhances carbon conversion efficiency while suppressing bacterial abundance—a counterintuitive combination, suggesting that photomodified substrates and/or metabolic stimulation of the remaining community sustain CO₂ production despite reduced microbial biomass. In contrast, the microbial treatment was dominated by a stronger negative pH effect on FCO₂ (−0.669 vs. −0.357), indicating a purely chemically controlled system without photochemical intervention.⁶

These results collectively indicate that although the mean flux did not differ significantly between treatments, the key pathways regulating FCO₂ and their strengths varied distinctly. In particular, the significant advantage of ADOM in the DOC∼FCO₂ and LogBA∼FCO₂ pathways mechanistically explains its higher CO₂ emission potential. This finding highlights that when assessing inland water carbon cycling, relying solely on flux measurements may obscure critical process differences; it is essential to integrate the analysis of DOM sources and process mechanisms. This mechanistic insight, enabled by the stepwise invariance testing approach, demonstrates the added value of SEM beyond traditional comparative flux analysis.

Environmental risks and management implications

The mechanisms described earlier position ADOM not merely as a substantial carbon pool but as a highly efficient, light-regulated “biogeochemical reactor” in eutrophic waters, capable of driving sustained, high-volume CO₂ emissions and constituting a direct positive feedback to climate warming. In contrast, MDOM, akin to broader terrestrial inputs from plants and soils, exhibits relatively lower degradability compared to ADOM,^36,37 with approximately 30% of its initial carbon remaining as a recalcitrant residual pool after 85 days of incubation. Although the majority of MDOM was mineralized over the course of the experiment—consistent with the dynamic nature of even recalcitrant DOM in aquatic systems over multi-month timescales—the substantially lower CO₂ emission potential and the persistence of a non-negligible refractory fraction suggest that MDOM inputs exert a less pronounced positive feedback to atmospheric CO₂ than ADOM. In relative terms, MDOM therefore acts as a carbon sink enhancer compared to the highly labile ADOM, which was nearly fully mineralized and emitted as CO₂. This contrasting behavior between the two DOM sources creates a functional “double-edged sword” in aquatic carbon cycling: while MDOM tends to partially stabilize carbon through residual refractory organic matter, ADOM readily drives carbon emissions under eutrophic conditions.

Nevertheless, the potential benefits of such carbon sequestration are increasingly jeopardized by large-scale anthropogenic pressures.^38,39 The widespread browning of northern lakes, driven primarily by increased terrestrial DOM inputs, reduces light penetration and suppresses benthic productivity.^40,41 However, the concurrent input of nutrients, combined with sufficient light availability in the upper water layers, can stimulate pelagic primary production and even trigger phytoplankton blooms—particularly those dominated by cyanobacteria—thereby significantly elevating ecological risks.³⁹ In vast grassland ecosystems, this vulnerability is further amplified: widespread overgrazing exacerbates soil erosion and grassland degradation, increasing the flux of nutrients and organic matter exported to water bodies.^13,42 These allochthonous inputs can further fuel cyanobacterial blooms, which, through the subsequent production and rapid mineralization of ADOM, stimulate GHG emissions,^43,44,45 forming a vicious cycle of “eutrophication-bloom-ADOM production-GHG release.”

Therefore, source-specific management of DOM is imperative for grassland lakes facing dual threats from external manure inputs and internal algal blooms. Our findings suggest two complementary, DOM-oriented intervention strategies to break the previously identified vicious cycle.

• (1)Reduce exogenous MDOM loading by strengthening livestock manure collection and treatment, combined with science-based grazing intensity controls to curb overgrazing-induced soil erosion and diffuse nutrient runoff. This approach directly limits the entry of terrestrial, humic-rich organic matter and its associated nutrients into lake systems.
• (2)Suppress endogenous ADOM production through in situ algal bloom mitigation measures (e.g., nutrient inactivation or targeted bloom control), thereby directly interrupting the massive generation of labile, protein-like ADOM that serves as the primary substrate for light-stimulated CO₂ emissions.

Adopting this dual-pronged strategy—integrating exogenous load reduction with endogenous bloom suppression—is critical for mitigating interconnected carbon-climate feedbacks and enhancing ecosystem resilience. Importantly, this framework moves beyond generic watershed management by targeting the distinct biogeochemical behaviors of allochthonous versus autochthonous DOM, as elucidated by our mechanistic incubation study.

Limitations of the study

This study has several limitations. Although MDOM serves as a useful proxy for livestock-influenced terrestrial DOM inputs, its representativeness of broader soil-derived DOM is limited. Furthermore, while the laboratory conditions were designed to clarify key mechanisms, they inevitably simplify the complexity of natural environments. Importantly, this study does not provide an in-depth exploration of the interactions among DOM molecular composition, specific microbial communities, and GHG emissions. In particular, our 16S rRNA amplicon sequencing provides taxonomic resolution but does not directly inform on functional gene abundance (e.g., hydrolase or oxidase genes involved in organic carbon degradation). Direct measurements of functional gene expression—such as qPCR targeting key carbon degradation genes or metagenomic approaches—would be necessary to definitively link the observed taxonomic shifts to specific metabolic pathways driving CO₂ production. Future research should address these aspects by linking the microscale dynamics of molecules and microorganisms with macroscale ecosystem functions. Such efforts will contribute to a more comprehensive understanding of DOM biogeochemical cycling and support strategies for mitigating climate change.

Additionally, our ROS quantification using chemical probe methods has several methodological constraints. Specifically, the furfuryl alcohol (FFA) probe used for ¹O₂ quantification can also react with ·OH, which may lead to overestimation of ¹O₂ contributions if ·OH is not adequately quenched.^46,47 Moreover, the calculated ROS values represent semi-quantitative indicators rather than true steady-state concentrations, as rigorous steady-state determination would require more complex kinetic modeling accounting for multiple production and consumption pathways.⁴⁸ Nevertheless, the consistent patterns of ROS production across treatments provide meaningful comparative insights into the source-dependent photochemical reactivity of DOM, and our SEM-based conclusions rely on these relative differences rather than absolute concentration values.

In summary, this study demonstrates that the fate of DOM in lake ecosystems is strongly dependent on its source, with light playing a decisive regulatory role in both degradation pathways and GHG emissions. Specifically, light suppressed the microbial degradation of endogenous ADOM while enhancing the degradation of exogenous MDOM. Among ROS, ¹O₂ significantly reduced the abundance of ADOM-degrading microorganisms without affecting MDOM degraders. The efficient microbial transformation of ADOM resulted in substantially greater CO₂ release. These findings highlight that, for eutrophic steppe lakes like Hulun Lake, an integrated management strategy—combining exogenous load reduction with endogenous bloom suppression—is crucial for mitigating both ecological risks and greenhouse effects.

Resource availability

Lead contact

Requests for further information and resources should be directed to and will be fulfilled by the lead contact, Yulong Tao (taoyulong_1982@163.com).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• •The 16S rRNA amplicon sequencing data reported in this paper have been deposited in the Genome Sequence Archive (GSA) at the National Genomics Data Center, China National Center for Bioinformation, under accession number GSA: CRA038938. They are publicly accessible at https://ngdc.cncb.ac.cn/gsa/browse/CRA038938.
• •This paper does not report original code.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

This work was supported by the Science and Technology Project of the Inner Mongolia Autonomous Region (grant no. 2021CG0037).

Author contributions

Conceptualization, Y. Tao; methodology, Y. Tao; investigation, Y. Tao, Q.W., M.B., Y. Tian, and L.L.; writing – original draft, Y. Tao; writing – review and editing, Y. Tao; data curation, Y. Tao; visualization, Y. Tao

Declaration of interests

The authors declare no competing interests.

Declaration of generative AI and AI-assisted technologies in the writing process

During the preparation of this work, the authors used DeepSeek in order to improve the language and readability of the manuscript. After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Bacterial and virus strains
Freshwater microbial consortium	This paper (collected from Hulun Lake)	N/A
Biological samples
Algal biomass (for ADOM extraction)	Hulun Lake basin, collected in August 2023	N/A
Livestock manure (cattle, sheep, horse)	Hulun Lake basin, collected in August 2023	N/A
Lake water samples (for LDOM and inoculum)	Hulun Lake, collected in October 2023	N/A
Chemicals, peptides, and recombinant proteins
Algal-derived dissolved organic matter (ADOM) leachate	This paper (extracted from algal biomass)	N/A
Manure-derived dissolved organic matter (MDOM) leachate	This paper (extracted from manure samples)	N/A
Critical commercial assays
ALFA-SEQ Advanced Soil DNA Kit	Findrop Biosafety Technology, Guangzhou, China	Cat# DC309-09
Deposited data
16S rRNA gene amplicon sequencing data	This paper	GSA: CRA038938https://ngdc.cncb.ac.cn/gsa/browse/CRA038938
Software and algorithms
FastQC	Babraham Bioinformatics	https://www.bioinformatics.babraham.ac.uk/projects/fastqc/
DADA2 (v1.16)	Callahan et al., 2016	https://benjjneb.github.io/dada2/
SILVA database (Release 138)	SILVA	https://www.arb-silva.de/
FT-ICR MS data processing algorithm	Referenced in text	An et al.49; Spencer et al.50
Other
Customized LED solar simulator	This paper	N/A
Shimadzu TOC-L Analyzer	Shimadzu Corporation	https://www.shimadzu.com/

Experimental model and study participant details

Algal biomass (for ADOM extraction)

Fresh algal biomass was collected from Hulun Lake and its basin in August 2023. The collected biomass was immediately frozen, freeze-dried, and ground into powder for DOM leachate preparation. No specific genotypes, developmental stages, or sexes were applicable due to the mixed natural phytoplankton composition.

Livestock manure (for MDOM extraction)

Fresh manure samples from cattle, sheep, and horses were collected from pastoral areas surrounding the Hulun Lake basin in August 2023. The manure samples were mixed at a mass ratio of 2:1:1 (cattle:sheep:horse) based on local livestock excretion patterns, then freeze-dried and ground for DOM leachate preparation. No experimental animals were directly used in this study.

Microbial inoculum

Lake water collected from Hulun Lake in October 2023 was filtered through a 0.45 μm membrane to retain the indigenous microbial community and acclimated at room temperature for 3 days to prepare the microbial inoculum. The consortium represents the natural lake bacterioplankton community; no specific strains were isolated or characterized prior to inoculation.

Lake water basal medium

Surface water (200 L) was collected from the lake center in October 2023, a period when freshly produced algal DOM from spring had largely degraded. The water was filtered through a 0.22 μm membrane to remove large particles and microorganisms, sterilized, and stored at 4°C as the basal medium for incubation experiments.

Ethics statement and sex/gender influence

This study did not involve live vertebrate animals, human participants, or protected species; all samples were environmental collections (water, plant biomass, and livestock waste). Therefore, no ethical approval or institutional oversight was required. As the study uses environmental microbial communities and organic matter extracts without sexual reproduction or differentiation, the influence of sex and gender on the results is not applicable and does not affect the generalizability of the findings.

Method details

Overall methodology

This study investigated the photochemical and microbial degradation of DOM from different sources in Hulun Lake, Inner Mongolia, China. Surface water samples were collected in October 2023, and DOM leachates were prepared from endogenous (algae) and exogenous (livestock manure) sources collected in August 2023. A microcosm incubation experiment was designed with four treatment groups (control, light only, microbial only, and coupled light-microbial) and three DOM types (original lake water (LDOM), lake water amended with algal DOM (ADOM), and lake water amended with manure DOM (MDOM)), each with three replicates. Incubations were conducted for 85 days under simulated solar radiation with a 12-h light/dark cycle. Water samples were analyzed for DOC, ROS, and CO₂ fluxes. Microbial abundance was quantified using qPCR, and bacterial community composition was determined through 16S rRNA gene amplicon sequencing. The molecular characterization of DOM was performed using Fourier Transform Ion Cyclotron Resonance Mass Spectrometry (FT-ICR MS). This experimental design enabled the systematic comparison of degradation mechanisms and environmental effects of DOM from different sources under simulated bloom conditions.

Study area and sample collection

This study was conducted in Hulun Lake (48.550°–49.333° N, 116.967°–117.800° E), the fifth largest lake in northern China, located in the western part of the Hulunbuir Grassland, Inner Mongolia Autonomous Region. This lake represents the easternmost extension of the vast Eurasian Steppe ecoregion. It has an irregular oblique rectangular shape with an average surface area of approximately 2,054 km² and an average depth of 5–6 m. The region experiences an arid to semi-arid continental temperate monsoon climate, with an average annual precipitation of about 260 mm and a high evaporation rate of approximately 1,100 mm. The sunshine percentage is around 70%.¹⁵ Affected by climate change and human activities such as overgrazing, the lake is moderately eutrophic and frequently experiences cyanobacterial blooms in summer.⁴²

Water, dissolved organic matter (DOM), and microbial inoculum were collected from Hulun Lake. Water and DOM sample collection: In October 2023, 200 L of surface water was collected from the lake center, a period when freshly produced algal DOM from spring had largely degraded. The water was filtered through a 0.22 μm membrane to remove large particles and microorganisms, sterilized, and stored at 4°C as the basal medium for incubation experiments. Algal samples for endogenous DOM and livestock manure (from cattle, sheep, and horses) for exogenous DOM were collected from the lake and its basin in August 2023. DOM leachate preparation: Freeze-dried algal and manure samples were immersed in ultrapure water. After ultrasonication and horizontal oscillation, the mixture was filtered through a 0.22 μm membrane to obtain the DOM leachate. Manure samples were mixed at a mass ratio of 2:1:1 (cattle: sheep: horse) based on local livestock excretion patterns. All leachates were stored at 4°C until use. Microbial inoculum preparation: Lake water was filtered through a 0.45 μm membrane and acclimated at room temperature for 3 days to prepare the microbial inoculum.²⁵

Microcosm incubation experiment design

Northern lakes are predominantly influenced by terrestrial-derived DOM, with this contribution showing a sustained increasing trend.³⁹ Typical DOC concentrations in northern lakes generally range between 2–60 mg L⁻¹.⁵¹ In lakes such as Daihai and Juyanhai in Inner Mongolia, DOC concentrations can reach approximately 60 mg L⁻¹.⁵² Specifically, in Hulun Lake, typical DOC concentrations vary between 6–43 mg L⁻¹.⁵¹ However, during algal bloom periods, DOM concentrations can significantly exceed these levels. Based on our routine monitoring, DOC concentrations in Hulun Lake approached 52.5–59.3 mg L⁻¹ at multiple sites in August 2022. Moreover, the DOC concentration in the adjacent Lake Buir reached 87.7 mg L⁻¹ in July 2022.

Against the backdrop of increasing intensity of cyanobacterial blooms and terrestrial inputs,^11,53,54 to simulate the high DOM concentrations observed during cyanobacterial blooms in the Hulun Lake basin and to comparatively investigate the degradation mechanisms and environmental effects of DOM from different sources, we designed microcosm experiments with elevated initial DOM concentrations: 82–86 mg L⁻¹ for ADOM and 95–97 mg L⁻¹ for MDOM. These concentrations were selected to represent realistic scenarios during intense bloom events while enabling clear differentiation of treatment effects. The ADOM concentration was set to represent realistic extreme bloom scenarios, approaching the maximum recorded value of 87.7 mg L⁻¹ in the adjacent Lake Buir. The MDOM concentration was configured to be as close as possible to the ADOM level to enable a meaningful comparison of their degradation behaviors, despite the practical difficulty of achieving exact equivalence in initial concentrations due to differences in their extraction and preparation. While MDOM at such high concentrations may not represent a common field scenario, this study frames it as a stress-test scenario to systematically reveal the differential mechanisms by which these two major DOM sources respond to light and microbial degradation. This approach does not aim to mimic average environmental conditions, but rather to elucidate the intrinsic reactivity differences between allochthonous and autochthonous DOM under controlled, comparable starting conditions—a strategy consistent with established microcosm studies investigating DOM degradation mechanisms. The experiment was designed to disentangle the individual and combined effects of photodegradation and microbial degradation on DOM. Four treatment groups were established (Table S1): (1) Control (dark, sterile conditions), (2) Light treatment (simulated solar radiation), (3) Microbial treatment (dark with active microbial community), and (4) Coupling treatment (combined light and microbial exposure). Each treatment included three DOM types: original lake water (LDOM), lake water amended with ADOM, and lake water amended with MDOM, with three replicates per type.

All procedures were conducted under a laminar flow hood using sterile glassware. For each microcosm, 10–15 mL of LDOM, ADOM, or MDOM leachate was added to an Erlenmeyer flask, followed by lake water to a final volume of 250 mL. The leachate constituted less than 10% of the total volume to minimize volume effects. A customized LED panel served as the light source, providing an intensity of 60,000–70,000 lx, calibrated based on 30 consecutive days of summer sunlight measurements. The light spectrum included UVA (380 nm, 4%) and UVB (280 nm, 1%) components.¹⁵ A 12-h light/dark cycle was maintained. Incubations were carried out in a temperature- and humidity-controlled room. The experiment lasted 85 days with destructive sampling, involving a total of 270 quartz Erlenmeyer flasks.

Temperature, pH, dissolved oxygen (DO), DOC, and ROS were measured on days 0, 2, 5, 10, 20, 30, 60, and 85. The greenhouse gas (CO₂) was analyzed on days 2, 5, 10, 20, 30, 60, and 85. Microbial abundance was assessed on days 0 and 85 for the control and light groups and on days 0, 2, 5, 10, 20, 30, and 85 for the microbial and coupling groups. Microbial culture tests using nutrient agar plates were conducted at each interval to verify bacterial contamination in the control and light groups, with contaminated samples excluded from the analysis. Ultra-high-resolution mass spectrometry was used to determine the molecular structure of DOM and microbial community composition on days 0, 10, and 85, with additional analyses on days 2 and 5 depending on microbial abundance. No day 0 samples were collected from the light and coupling groups because their initial conditions were identical to those of the control and microbial groups, respectively.

Based on the research objectives, the results and discussion focus on ADOM and MDOM, while the lake water serving as the substrate was used only for DOC concentration and CO2 flux correction.

Chemical analysis

The water temperature, pH, and DO were measured using a portable instrument (WTW). The DOC content was determined using a Shimadzu TOC-L Total Organic Carbon Analyzer in the NPOC mode. GHG (CO₂) was collected in tinfoil bags and analyzed by gas chromatography, with fluxes calculated using the UNESCO/IHA GHG Measurement Guidelines for Freshwater Reservoirs,⁵⁵ specifically as follows:

$$Flux=k_x\left(C_{water}-C_{air}\right)$$

where Flux is expressed in mmol of CO₂

C_air is the concentration of gas in the air, obtained with:

$$C_{air}\left[umole·L^{-1}\right]=K_0\left[umole·{atm}^{-1}\right]\ast {pCO}_{2\left(air\right)}\left[uatm\right]$$

where pCO_2(air)[uatm] is the partial pressure of CO₂ in the ambient air, measured by gas chromatography, T_K is the temperature in Kelvin.

C_water is the concentration of gas in the water (obtained by the product of gas partial pressure (p^Gas) times the gas solubility (K₀) at a given temperature) obtained with:

$$C_{water}\left[umole·L^{-1}\right]=K_0\left[umole·{atm}^{-1}\right]\ast p^{Gas}\left[uatm\right]$$

p^Gas[ppm] is the partial pressure of CO₂ in the headspace, measured by gas chromatography.

$$\ln K_0\left({CO}_2\right)=-58.0931+90.5069\ast \left(\frac{100}{T_K}\right)+22.294\ast \ln \frac{T_K}{100}+s\ast \left(0.027766-0.025888\ast \left(\frac{T_K}{100}\right)+0.0050578\ast {\left(\frac{T_K}{100}\right)}^2\right)$$

where K₀(CO₂) is the solubility of CO₂ in water expressed in mole·L⁻¹ ·atm⁻¹, T_K is the temperature in Kelvin, and s is the salinity in parts per thousands. Hence, in freshwater systems, only the first partof the equation is necessary as s = 0.

$$k_x=k_{600}{\left(\frac{S_C}{600}\right)}^{-x}$$

where k_x is the gas exchange coefficient expressed in cm·h⁻¹, x is equal to 0.66 for wind speed ≤3 m s⁻¹ and is equal to 0.5 for wind speed >3 m s⁻¹, and S_C is the Schmidt number for CO₂ which is dependent on temperature (t) in degree Celcius:

$$S_C\left({CO}_2\right)=1911.1-118.11t+3.4527t^2-0.04132t^3$$

and:

$$k_{600}=2.07+\left(0.215\ast U_{10}^{1.7}\right)$$

where k₆₀₀ is the gas exchange coefficient expressed in cm·h⁻¹ normalised for CO₂ at 20°C in freshwater with a Schmidt number of 600. Various k₆₀₀ equations exist in the literature and weredeveloped for specific study sites or region.

and:

$$U_{10}=1.22\ast U_1$$

where U₁ is the wind speed at the water surface and U₁₀ is the frictionless wind speed at 10 m expressed in m·s⁻¹.

The resulting final number should be multiplied by 0.24 to obtain a flux in mmole·m⁻² ·d⁻¹.

Molecular characterization of DOM

The molecular composition of DOM was characterized using a Fourier Transform Ion Cyclotron Resonance Mass Spectrometer (FT-ICR MS, Bruker Solari X) with an electrospray ionization (ESI) source operating in the negative ion mode. The key detection parameters included a continuous injection rate of 120 μL h⁻¹, capillary inlet voltage of 4.0 kV, ion accumulation time of 0.06 s, mass range of 100–1000 Da, and 4M 32-bit data sampling with 300-fold time-domain signal superposition to enhance the signal-to-noise ratio. Prior to analysis, the instrument was calibrated using 10 mmol/L sodium formate, and internal standard calibration with soluble organic matter of known molecular formulas was performed post-detection. The mass errors were controlled within 1 ppm using calibration. The formula assignments were performed using a well-established algorithm based on stoichiometric parameters, including the H/C and O/C ratios as well as the modified aromaticity.^49,50 The identified molecular formulas were further classified into six structural groups: condensed aromatic-like structures (CAS, AImod>0.67), aromatic-like structures (AS, 0.5 < AImod ≤ 0.67), highly unsaturated structures (HUS, AImod<0.5, H/C<1.5, O/C≤0.9), aliphatic-like formulas (1.5 ≤ H/C ≤ 2.0, O/C < 0.9, N = 0), peptide-like formulas (1.5 ≤ H/C≤ 2.0, O/C < 0.9, N > 0), and sugar-like formulas (O/C > 0.9). Additionally, carboxylic-rich alicyclic molecules (CRAM) were identified using the double-bond equivalents per carbon (DBE/C = 0.30–0.68), DBE per hydrogen (DBE/H = 0.20–0.95), and DBE per oxygen (DBE/O = 0.77–1.75).

Microbial abundance and community

An adequate volume of each water sample was filtered through 0.22 μm polycarbonate membranes. All filters were stored at −80 °C until subsequent DNA extraction. Total genomic DNA was extracted using the ALFA-SEQ Advanced Soil DNA Kit (Findrop Biosafety Technology, Guangzhou, China). The V3–V4 hypervariable region of the bacterial 16S rRNA gene was amplified by polymerase chain reaction (PCR) using the universal primers 338F (5′-ACTCCTACGGGAGGCAGCA-3′) and 806R (5′-GGACTACHVGGGTWTCTAAT-3′).⁵⁶ Amplicons were purified and sequenced on an Illumina NovaSeq 6000 platform (paired-end, 2 × 250 bp). Raw sequencing reads were quality-filtered and trimmed using FastQC. Amplicon sequence variants (ASVs) were then generated and chimera-checked using the DADA2 pipeline. Taxonomic classification of ASVs was performed against the SILVA 138 SSU rRNA reference database. Beta diversity of the bacterial communities at the phylum level was examined using non-metric multidimensional scaling (NMDS) based on Bray-Curtis distances. Separately, the absolute abundance of bacterial 16S rRNA genes in each sample was quantified via quantitative PCR (qPCR) on a LineGene 9600 system.

Analysis of ROS

The ROS levels were measured according to the method described by Song et al.²⁵ Water samples which were taken on days 2, 5, 10, 20, 30, 60, and 85 were analyzed for the contents of three kinds of ROS (·OH, ¹O₂ and ³CDOM∗). Water samples(20 mL) were spiked with 2 mM of benzene (·OH probe when phenol formation is monitored), 1 mM of furfuryl alcohol (FFA, the probe of ¹O₂), and 1 mM of 2, 4, 6-trimethyphenol (TMP, the probe of ³CDOM∗). The control group and microbial group were placed in a 25 °C constant temperature shaker with a set frequency of 170 and shaken for 24 h. The light and combined groups were placed under lamp and irradiated for 24 h. Changes of reactive transients (³CDOM∗, ¹O₂ and ·OH) were indirect represented by the changing of TMP, FFA and benzene according to the reactions (as shown in Equations 1, 2, and 3). The monitoring of TMP, FFA, and phenol during the degradation experiment was carried out by high-performance liquid chromatography with diode array detection (HPLC-DAD, Shimadzu). All samples were eluted with constant mixtures (60:40) of methanol and aqueous H₃PO₄ (pH 2.8) at a flow rate of 1.0 mL· min-1. The TMP retention time was 20.3 min and the detection wavelength was at 210 nm. For the analysis of FFA, the retention time was 2.6 min and the detection wavelength was at 210 nm. The phenol retention time was 4.3 min and the detection wavelength was at 210 nm.

$${}^{3}CDO{M}^{\ast }+TMP\to products$$ (Equation 1)

$${}^1{O}_2+FFA\to products$$ (Equation 2)

$$·OH+Benzene―\left(95\%yield\right)\to Phenol$$ (Equation 3)

The concentrations of ROS were calculated as follows (Equations 4, 5, and 6):

$$C_{3CDO{M}^{\ast }}=C_{TMP,initial}-C_{TMP,fina\mathrm{l}}$$ (Equation 4)

$$C_{1O2}=C_{FFA,initial}-C_{FFA,final}$$ (Equation 5)

$$C_{3CDO{M}^{\ast }}=\left(C_{Phenol,final}-C_{Phenol,initial}\right)\ast 95\%$$ (Equation 6)

The obtained ROS values should be interpreted as semi-quantitative indicators of relative production across treatments rather than absolute steady-state concentrations (see limitations of the study section for detailed discussion).

Quantification and statistical analysis

Data analysis was conducted using SPSS 23 and Excel 2021, with data visualization performed using OriginPro 2024 and R 4.5.1. Group comparisons were conducted following a one-way ANOVA framework. After evaluating the assumption of homogeneity of variances using Levene’s test, the data were analyzed accordingly: when variances were homogeneous, the standard one-way ANOVA was applied, followed by the LSD post-hoc test for pairwise comparisons; when the assumption of equal variances was violated, the Welch’s ANOVA, which does not require variance homogeneity, was employed. To test whether the structural relationships among variables differed across groups (ADOM vs. MDOM; microbial vs. coupling), we conducted multi-group SEM using the lavaan package in R. A stepwise approach was used to assess measurement and structural invariance. We began by establishing a configural invariance model (baseline), where all parameters were freely estimated across groups. Subsequently, we tested for structural invariance by constraining all regression paths to be equal across groups using the group.equal = c(“regressions”) argument. The fit of these nested models was compared using the chi-square difference test (Δχ²), where a significant p-value (p < 0.05) indicated that the equality constraints significantly worsened model fit, suggesting structural differences between groups. Model fit for each step was evaluated using the Comparative Fit Index (CFI), Tucker-Lewis Index (TLI), Root-Mean-Square Error of Approximation (RMSEA), and the chi-square test.