The sampling sites chosen for field observation and microcosm incubation have been shown in Table 1. Notably, long-term surface water interchange management strategy has been implemented before sampling at Chu River riparian (CRR) and Fish pond_interchange (P_int) sampling sites. During a year of field observation (May 2020 to April 2021), the CH₄ emissions rate for the sampling site of P_int was found relatively low in summer and autumn, then it begins to rise in late autumn and reach a higher value during the winter and spring period. Similarly for CRR, the CH₄ emissions rate was determined as a high value in spring and exhibited an increasing trend in late autumn and winter. Peak emissions rate of P_int was observed to be high (2.79 mmol m⁻²·d⁻¹) in October, while the peak emissions rate of CRR was observed to be low (0.91 mmol m⁻²·d⁻¹) in April (Figure 1A), except for the period of nutrient replenishment for P_int in October and during the freezing periods (November to December), the emissions rate of CH₄ from the sample site of CRR was higher than those of sample site for P_int among monitoring period.

The temporal variability of CO₂ was performed for CRR and the highest CO₂ emissions rate (85.32 mmol m⁻²·d⁻¹) was found in November and found to be 1.32 folds higher than that of measured value for P_int. Further, it was determined that the peak CO₂ emissions rate (40.82 mmol m⁻²·d⁻¹) for P_int in February but highest emissions period ranged from September to next January. Although, CO₂ emissions rate for P_int was reported to be lower than that of CRR. In addition, N₂O emissions rate for P_int did not fluctuate much but highest value was 2.21 10^–2 mmol m⁻²·d⁻¹ in July and the lowest value was −2.30 10^–2 mmol m⁻²·d⁻¹ in November. Similarly for CRR, the N₂O emissions rate was observed to be higher from June to September and peaked (0.13 mmol m⁻²·d⁻¹) in September. Overall, field observation revealed that the annual accumulative CH₄, CO₂ and N₂O emissions for P_int were 185.95, 3,424.63, and 0.72 mmol m⁻² respectively whereas for CRR it was observed to be 167.60, 7,197.92, and 14.52 mmol m⁻², respectively (Figures 1B,C).

In microcosm incubation, four sediment samples were collected in situ. The triplicate samples of each treatment were established with different supernatant solutions that include deionized water, fish-pond-overlying water and artificial water. Fish-pond-overlying water was collected for P_int, while artificial water was prepared with a nitrogen/phosphorous (N/P) ratio of ∼1 from fish-pond-overlying water (for more details see supporting information in Supplementary Tables S1,S2). The microcosm incubation was operated for 50 days. During this course, all treatments showed an increasing trend in accumulative emissions. After 50-day of incubation, the GHG production rates approached a steady state (Figure 2). Moreover, two sediment samples collected from aquaculture ponds (P_int and Zero-interchange fish pond) had considerably greater accumulative emissions for both CO₂ and CH₄, which was higher than that of sediment obtained from urban rivers (CRR and Hongshan gate). Comparatively, incubation of deionized water treatments for Hongshan gate sediment had higher CO₂ emissions than P_int. For aquaculture pond sediments, the accumulative GHG emissions for P_int sediment had much lower than those of Zero-interchange fish pond sediment. Notably, CRR sediment had higher accumulative emissions ofCH₄ and N₂O than that of incubated sediment of Hongshan gate. Besides, incubated sediment for CRR was observed to be more N₂O than that of P_int sediment, shown in Supplementary Table S3.

From the three different applied supernatant solutions for sediments incubations, all treatments with artificial water emitted more CH₄ than the others. Except for Hongshan gate, the highest accumulative CO₂ emissions were observed from fish-pond-overlying water treatments. For the accumulative N₂O emissions, no statistical significance was found among river sediments. While artificial water treatment emits more N₂O than other treatments for aquaculture ponds sediments (Figure 2; Supplementary Table S3).

Stable carbon isotope analysis was performed for the collected gases in the samples on 29th day of incubation. The abundance ratios and isotopic fractionation factor (αc) were calculated from the thousand^th deviation of the actual sample from the standard sample. The isotopic fractionation factor of all samples was >1.055, (Supplementary Table S4), which indicates the occurrence of hydrogenotrophic methanogenesis (SI.Eq.S4).

During the microcosm incubation, the highest concentrations of ammonium nitrogen (NH₄ ⁺-N), nitrate nitrogen (NO₃ ⁻N), dissolved organic matter (DOC) and dissolved inorganic matter (DIC) were found in Zero-interchange fish pond sediment treatment. The DOC and DIC concentration in all treatments of P_int sediment were higher than those of CRR sediments. For aquaculture ponds, concentrations of NH₄ ⁺-N and NO₃ ⁻-N were lower than those of CRR sediment in deionized water treatment and fish-pond-overlying water treatment. Noticeably, the total phosphorus (TP) concentration in all treatments of river sediments was higher than those of aquaculture ponds sediments, and the peak value was found in the treatments of Hongshan gate sediment (Table 2).

Spearman’s correlation analysis was applied to the GHGs emissions rates and the concentrations of hydrochemical parameters, part of the correlations passed the significance test, but other correlations indicate the relations between factors as well (Figure 3). The results showed that the GHGs emissions rates were strongly negatively correlated with TP and C/N ratio, and weakly correlated with NH₄ ⁺-N, while all other parameters showed a positive trend of correlation. A strong positive correlation (R ²: 0.7, p < 0.05) was obtained between CH₄ emissions rates and N/P ratio while a positive correlation between CO₂ emissions rates and DOC, CO₂ emissions rates and N/P ratio at degree of 0.64 and 0.67 (p < 0.05), respectively. In addition, N₂O emissions rates had a positive correlation with NO₃ ⁻-N (R ²: 0.56, p < 0.05) and a negative correlation for TP (R ²: 0.45, p < 0.05).

Fluorescence-parallel factor analysis of water samples was performed using the DOMFluor toolbox. The result showed that a total of five organic components in all samples, where C₁ represents approximately equal humic acid and fulvic acid, C₂ represents fulvic acid, C₃ represents fulvic acid and a trace of humic acid, C₄ represents tyrosine and certain protein-like fractions, and C₅ represent aromatic protein and microbiological by-products (Figure 4). Notably, C₁, C_2, and C₃ were the components that existed in all sediment samples, while C₄ was found only in samples of Hongshan Gate sediment, and C₅ was found only in samples of the Zero-interchange fish pond, which means a similar distribution of components occurred in samples of CRR and P_int sediments. Figure 5 indicated that the relative content of C₁ increased before the 22nd day of incubation, from a mean value of all treatment groups of 31.25% to a mean value of 49.25%. However, decreased from the 15th day of incubation, the mean value at the end was 41.25%, whereas the variation trend of relative content of C₃ was diametrically opposite to that of C₁, decreasing from a mean value of 38.78% to a mean value of 19.5%, then starting increasing, the mean value at the end was of 26%, the relative content of C₂ was lower than the other two common fractions, varies from 28% to 36%.

In all treatments, the content of OM from the sediment was lower at the end of incubation than its initial value (Supplementary Table S5), proving that the incubation procedure significantly depleted the OM. In terms of sediments, the consumption rate of OM in descending order was Zero-interchange fish pond sediment, P_int sediment, CRR sediment and Hongshan Gate sediment. In terms of supernatants, the consumption rate of OM was found in descending order with deionized water treatment, fish-pond-overlying water treatment, and artificial water treatment.

During the field campaign, the period of high emissions was commonly found in the summer and autumn seasons. Moreover, CRR and P_int showed high CH₄ and CO₂ emissions periods between October to April (Yang et al., 2018b; Sieczko et al., 2020). It could be attributed to fishery management that bait, fresh fry, and fish bait were dosed into P_int in late summer while other aquaculture ponds popularly dosed for the same in the spring (Yang et al., 2017; 2018a). For P_int site, the input of nutrients in autumn and early winter ensured the metabolism and growth of organisms in pond even during the freezing season. Generally, it provides a substrate for the aerobic respiration of OM in the sediment which results in higher emissions of CH₄ and CO₂ in autumn and winter (Hill et al., 2017). The high loading of nutrients in P_int sediment might potentially be released into the water due to the long-term process of surface water interchange that accumulated into CRR. In addition, the observed increase in GHG emissions rates favors due to water pollution of CRR (Yu et al., 2013; Wang et al., 2015). Thus, surface water interchange management exhibits a similar emission pattern for CRR and P_int. The elevated N₂O emissions from June to October of CRR coincide with the results of previous studies on river riparians (Cole and Caraco, 2001). It could be due to input of nitrogen from household and industrial waste from upstream catchment of Chu River thereby increasing the N₂O emission rate. Further supportive made by rising groundwater levels over Chu riparian (Bange et al., 2019).

Field observation reveals that microcosm incubation experiments could be employed for further understanding the underlying mechanism of GHG emission patterns. In aspect of sediment chemical properties, in contrast to Hongshan gate sediment, the transportation of OM and nutrients through runoffs between P_int and CRR contribute high CO₂ and CH₄ emissions. Contrastingly, for Zero-interchange fish pond sediment, the process of surface water interchange reduces the accumulative GHG emissions of P_int sediment. In addition, CRR sediment had apparent accumulative N₂O emissions which could be attributed to the large input of surface runoffs (Fagervold et al., 2014; Bange et al., 2019).

From the perspective of applied overlying water, fish-pond-overlying water enriched with OM had the highest accumulative CO₂ emissions. However, the deposition of OM increases the carbon buried in benthic sediment and is further utilized in decomposition process (Hu et al., 2022). The presence of electron acceptors in fish-pond-overlying water, such as NO₃ ⁻ affects methanogenesis and thus CH₄ was found to be relatively low (Raghoebarsing et al., 2006). Besides, no clear CH₄ inhibition was found in artificial water treatment which might be due to excessive NH₄ ⁺-N. Generally, nitrogen and phosphorus inputs promoted the decomposition of OM, resulting in high GHG emissions from artificial water treatments (Tao et al., 2018). However, the substantial storage of carbon in both sediment and supernatant had made a positive impact to reduce CH₄ emissions, thus fish-pond-overlying water treatments of Zero-interchange fish pond sediment had the highest CH₄ emissions. The transformation between NH₄ ⁺-N and NO₃ ⁻ -N over the water-sediment interface is the key reason for N₂O emissions (Wang et al., 2007; Howard, 2015). The addition of NH₄ ⁺-N in artificial water increases the nitrification rate in the corresponding treatment which leads to higher N₂O flux within artificial water treatments. Subsequently, fish-pond-overlying water treatments exhibit high accumulative GHG emissions than that deionized water treatments which supports the viewpoint as a consequence of surface water interchange.

Acetotrophic methanogenesis and hydrogenotrophic methanogenesis are well-recognized for CH₄ production pathways under waterlogged conditions (Conrad et al., 2010; Mach et al., 2015). The present results are in agreement with CH₄ production mainly via hydrogenotrophic methanogens pathways which coincide with reported founding in incubated lake sediments (Avery et al., 2003). It is interpreted in the previous studies that the different energy sources may govern various methanogenic pathways (Glissmann et al., 2004. However, the present study ascertains no significant influence over the methanogenic pathway due to variation in overlaying water treatments.

As a consequence of a long-term surface water interchange, nutrient loading in the benthic sediment of CRR has increased accordingly, while that of sediment of P_int had been diluted which complies with the variation of GHG emission. The input of substantial industrial wastewater is responsible for the concentrations of NH₄ ⁺-N and NO₃ ⁻-N in treatments of CRR sediment and was found to be higher than those of P_int sediment (Rosamond et al., 2012; Zhang et al., 2021). The changes in the hydrochemical indicators altered the distribution of chemicals in the benthic sediment, which would have a potential impact on GHG emissions.

Apparently in this study, the variation of water chemistry has a strong impact on GHG emissions. Previous studies interpreted that transforming nitrogen would provide more substrate for microbial decomposition resulting significant amount of CH₄ and CO₂ emissions (Stiles et al., 2018). Besides, it has been reported that the presence of phosphate could either promote GHG production or inhibits methanogenesis, which is determined by total amount of loaded phosphor (Zhang et al., 2011). Current results had a positive correlation between GHG emission and TN loading, while TP was negatively correlated (Raghoebarsing et al., 2006). The higher concentration of TP in samples of Hongshan Gate sediment might inhibit the decomposition of OM, which inhibits the emissions of all three GHG (Zhang et al., 2011). Furthermore, it has been demonstrated that lower C/N ratios lead to the consumption of high amounts of organic nitrogen which promotes microbial activity leading to high GHG emissions (Zhang W et al., 2022).

Dissolved organic matter (DOM) is ubiquitous and it affects fundamental biogeochemical processes in freshwater ecosystems. In general, PARAFAC method was employed to interpret the key mechanism of the DOM and underlying GHG emission (Ding et al., 2022). Humic acid increases the ability of benthic sediment carbon burial and inhibits GHG emissions, while fulvic acid enhances the effectiveness of biological processes thus promoting the decomposition of OM. During the analysis, it was found that component C₁ had the highest proportion of fulvic acid, compared with other two common components (Yu et al., 2014). Since the proportion of component C₁ among the three common components was the highest in Zero-interchange fish pond sediment explicates the highest accumulative GHG emissions in the sediment. Component C₅ represents microbial by-products in Zero-interchange fish pond sediment which indicates the decomposition of OM resulting in high GHG emissions (Kimbrough and Dickhut, 2006). Theoretically, OM in urban river sediments ought to be homologous, and thus the process of surface water exchange might lead to the result that C₄ component was only derived in the sediment of Hongshan gate. These processes altered the composition of OM for CRR thus altering the potential of GHG emissions.

Fluorescence index (FI) and Humization index (HIX) are used frequently to evaluate DOM quality. Thus, FI and HIX were also used for DOM over incubation treatment, the results were given in Supplementary Table S6. Obtained FI value was higher than 1.5 in all treatments which indicates that the carbon within the studied incubation was mainly microbially active carbon, instead of carbon derived from terrestrial sources (Broder et al., 2017). In addition, obtained Humization index (HIX) value close to 1 for all treatments, indicates that DOM contains more humified compounds which potentially highlight an active carbon pool that contributes to GHG emission (D’Andrilli et al., 2022). It is suggested that the characterization of DOM is fundamental to understand carbon cycling and dynamics in aquatic ecosystems. Moreover, fluorescence-PARAFAC analysis provided the scientific basis for interpreting variation in GHG emissions therefore further studies need to be explored to fill the knowledge gap of organic carbon turnover at the terrestrial-aquatic interface.