2.1 Study site

The Han River (494 km), consisting of the North Han and South Han branches and the lower Han River (often called “Han River”), drains an area of 35 770 km² in the middle of the Korean Peninsula (Fig. 1; Seoul Metropolitan Government, 2017). The mean annual precipitation in the Han River basin was 1323 mm for the period from 1983 to 2014, with up to 70 % of the annual precipitation concentrated during 4 months from June to September. Major land uses in the basin include forests (73.6 %), croplands (14.1 %), urban and industrial areas (2.6 %), and other uses (9.7 %) (Fig. S1 in the Supplement). The highly urbanized metropolitan area along the lower reach has a large impermeable surface regarded as urban land use, accounting for 58 % of the total city area of Seoul (Seoul Metropolitan Government, 2017). Sampling sites were established along a North Han River section and the lower Han River, covering forested headwater (HR1) and agricultural streams along the upper reach through the middle and lower reaches impounded by dams and submerged weirs to the estuary.

Along the monitored reaches of the North Han River, four large multi-purpose hydroelectric dams have been constructed since 1943. The most upstream dam (Soyang; HR5) constructed in 1973 is the largest (height: 123 m; reservoir surface area: 70 km²). The most downstream dam constructed in 1973 (Paldang; HR10) located at the confluence of the North and South Han branches regulates river flow to the lower reach and supplies 2.6 million metric tons of water per day to the 25 million metropolitan population. Three major urban tributaries [Joongnang River (JN; HR12), Tan River (TC), and Anyang River (AY)] feed into the channelized lower Han River with submerged weirs at both ends. The tributary JN drains an area of 300 km² inhabited by 3.6 million, ∼45 % of which urban land use accounted for in 2014 (Seoul Metropolitan Government, 2017). Three WWTPs release treated sewage at the rate of 1.5 million m³ d⁻¹ to the tributary (JN), with the bulk (1.3 million m³ d⁻¹) discharged from the monitored WWTP near the mouth of JN, which employs tertiary treatments including modified Ludzack-Ettinger (MLE) and anaerobic/anoxic/oxic (A₂O) process (Ministry of Environment, 2015). Four WWTPs located within Seoul release effluents from tertiary treatments at the rate of 4.3 million m³ d⁻¹.

2.2 Sampling and field measurements

Basin-wide surveys were carried out at 15 sites encompassing major land uses and human-induced perturbations in the Han River basin during a summer monsoon period (July 2014) and in a dry period (May 2015) (Table 1; Figs. 1, S1). Compared to the upper reach (HR1–HR4) located in a heavily forested watershed with some scattered agricultural areas (Fig. S1), the impounded middle reach (HR5–HR11) and the lower reach receiving heavy loads of urban sewage (HR12–HR15) are subject to stronger anthropogenic perturbations. At six selected sites (HR1, HR2, HR4, HR8, HR11, and HR14), routine monitoring was continued at monthly intervals from July 2014 to July 2015, except for 2 winter months (January and February 2015). In addition to two-season surveys, 13 samplings were conducted at an outlet of the tributary JN (HR12) from May 2015 to December 2017. In November 2015 and May 2016, additional samples were collected at eight locations from a forested headwater stream to the outlet along the JN (Figs. 1 and S1). The eight sites were selected to cover the spatial pattern of land use, ranging from the forested upper reach to the increasingly urbanized downstream reaches (Fig. S1). The tributary samplings were complemented with four samplings at a discharge from a WWTP located a few km upstream of the JN outlet. To explore spatially resolved patterns of riverine concentrations of the three GHGs along the lower Han River, water and gas samples were collected at eight locations during a boat expedition on 10 June 2016 that also aimed at continuous underway measurements of pCO₂ using automated pCO₂ measurement systems as reported by Yoon et al. (2017).

Table 1Summary of sampling sites and the concentrations of the three GHGs measured during two basin-wide surveys at 15 sites and monthly monitoring at 6 sites in the Han River basin from July 2014 to July 2015. Monthly data are presented as means followed by ranges in parentheses. Note that additional data obtained at an urban tributary (HR 12) from May 2015 to December 2017 (N=13) are also presented as monthly data.

Download Print Version | Download XLSX

Water samples were collected from a depth of 10–20 cm below the water surface in the middle of the stream channel or > 2 m away from the river bank. Spot water samples for chemical analysis were collected through a peristaltic pump into acid-washed amber glass bottles. At the same sampling point we measured water temperature, pH, electrical conductivity (EC), and dissolved oxygen (DO) using a portable pH meter (Orion 5-Star Portable, Thermo Scientific, USA). Surface water concentrations of the three GHGs were measured by a manual headspace equilibration method (Kling et al., 1992; Yoon et al., 2016). Headspace equilibration was performed using a 60 mL polypropylene syringe to collect a 30 mL water sample and then a 30 mL ambient air sample. Another air sample was collected to measure gas concentrations in the ambient air. The syringe containing the water and air samples was shaken vigorously for 2 min; then ∼20 mL of the equilibrated air was stored in a pre-evacuated 12 mL Exetainer vial. The stored gas sample volume was larger than the vial volume to create overpressure and hence minimize gas concentration change associated with potential gas leakage. Vials had been flushed with high-purity N₂ gas before the filled N₂ gas was evacuated by a pump. Gas analysis was performed usually within 3 days after the sampling. Air temperature and barometric pressure were measured in situ by a portable sensor. Barometric pressure, together with water temperature, was used to calculate concentrations of the three GHGs from the gas concentrations of the equilibrated air and ambient air samples based on Henry's law (Hudson, 2004). Details on the calculation algorithms and Henry's law constants for the three GHGs are provided in Hudson (2004).

2.3 Laboratory analyses

The equilibrated headspace air sample, as well as a sample from the ambient air used for equilibration, was injected into a GC (7890A, Agilent, USA) equipped with an FID coupled with a methanizer (for analysis of CH₄ and CO₂), a µECD (for N₂O analysis), and a Supelco Hayesep Q 12 ft 1∕8 in. column for the simultaneous measurement of the three GHGs. High-purity N₂ gas (99.999 %) was used as carrier gas at a constant flow rate of 25 mL min⁻¹. The flow rate of the reference gas was 5 mL min⁻¹ for FID (N₂) and 2 mL min⁻¹ for µECD (Ar∕CH₄). Temperature settings include the inlet at 250^∘, the oven at 60^∘, the valve box at 100^∘, FID at 250^∘, and µECD at 300^∘. The volume of the sample loop was 1 mL. Standard reference gases of the three GHGs in N₂ balance (RIGAS Corporation, Korea) were used to calibrate the GC signals. The gas concentrations in these calibration standards were verified in the gas analysis laboratory of Korea Research Institute of Standards and Science (KRISS) that established the national measurement standards of gravimetrically prepared CO₂ in dry air and certified by a suite of gas analysis (Min et al., 2009). Additional headspace equilibration samples collected during the boat cruise in June 2016 were analyzed for stable C isotope ratios of CO₂ (${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{{\mathrm{CO}}_{\mathrm{2}}}$) and CH₄ (${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{{\mathrm{CH}}_{\mathrm{4}}}$) by a GasBench-IRMS (ThermoScientific, Bremen, Germany) at the UC Davis Stable Isotope Facility.

Water samples were filtered through pre-combusted (450 ^∘C) glass fiber filters (GF/F, Whatman; nominal pore size 0.7 µm) at the laboratory. The concentration of total suspended solid (TSS) was determined by filtering a known volume of water sample through a pre-weighed GF/F filter and then weighing the filter again after drying at 60 ^∘C for 48 h. Filtered water samples were analyzed for dissolved organic C, fluorescence excitation-emission matrices (EEMs), UV absorbance, total alkalinity (TA), dissolved ions, and chlorophyll a (Chl a). For quality control of all chemical analyses, standards with known concentrations and ultrapure water were analyzed for every batch of ten samples and triplicate analysis was performed for approximately 10 % of all analyzed samples to assess instrumental stability and accuracy.

DOC was measured by a total organic carbon (TOC) analyzer using high-temperature combustion of OM followed by thermal detection of CO₂ (TOC-V_CPH, Shimadzu, Japan). UV absorbance was measured across the wavelength range from 200 to 1100 nm using a UV-Vis spectrophotometer (8453, Agilent, USA). Fluorescence EEMs were collected on a fluorescence spectrophotometer (F7000, Hitachi, Japan) by simultaneous scanning over excitation wavelengths from 200 to 400 nm at 5 nm interval and emission wavelengths from 290 to 540 nm at 1 nm interval. Scan speed was 2400 nm min⁻¹ and the bandwidth was set to 5 nm for both excitation and emission. UV absorbance and fluorescence data were used to calculate specific UV absorbance at 254 nm (SUVA₂₅₄) (Weishaar et al., 2003), fluorescence index (FI) (McKnight et al., 2001), humification index (HIX) (Zsolnay et al., 1999), and common fluorescence EEM components (Fellman et al., 2010). Three fluorescent EEM components termed C1 (excitation/emission peaks at 325/467 nm), C2 (315/404 nm), and C3 (275/354 nm) were used to represent humic-like, microbial humic-like, and protein-like fluorescence, respectively (Fellman et al., 2010). TA was determined with 40–80 mL filtered samples on an automated electric titrator (EasyPlus Titrator Easy pH, Metrohm, Switzerland) based on the Gran titration method (Gran, 1952). Strong acid (0.01 N HCl) was used to titrate well beyond the equivalence point at pH between 4 and 3. Ions dissolved in water samples were analyzed on an ion chromatograph (883 Basic IC plus, Metrohm, Switzerland). Chl a was analyzed spectrophotometrically following filtration on GF/C filters and acetone extraction (American Public Health Association, 2005).

Dual carbon isotopes analyses were conducted for water samples collected from six monthly monitoring sites and the JN WWTP effluents. Some filtered water samples were kept frozen before analysis. For Δ¹⁴C-DOC analysis, each water sample was acidified by 40 % H₃PO₄ solution and sparged by helium gas to remove inorganic carbon. Then, the sample was oxidized with ultrahigh purity oxygen gas using UV lamp for 4 h (Raymond and Bauer, 2001). The oxidized CO₂ was separated cryogenically with liquid N₂ in a vacuum line and sealed in a pre-baked pyrex tube. The CO₂ samples in the sealed pyrex tubes were sent to the National Ocean Sciences Accelerator Mass Spectrometry (NOSAMS) facility at the Woods Hole Oceanographic Institution for dual isotope analysis (Δ¹⁴C and δ¹³C) where accelerated mass spectrometer and isotope ratio mass spectrometer were used for Δ¹⁴C and δ¹³C analysis, respectively (http://www.whoi.edu/nosams/home, last access: 25 October 2048). The CO₂ samples were transformed to graphite targets for Δ¹⁴C analysis by accelerated mass spectrometry.

2.4 Statistical analysis

Most dissolved GHGs and water quality data were not normally distributed, and their distribution patterns varied with river reach and measured parameter. Therefore, Kendall rank correlation, as a non-parametric test, was performed using the Kendall package of R (R Development Core Team, 2018) to investigate the relationships between water quality variables and dissolved GHGs both for the whole Han River basin and for each of the upper, middle, and lower reaches. For the monthly monitoring data, seasonal differences in the three GHGs were analyzed by the non-parametric Kruskal–Wallis analysis of variance followed by Dunn's multiple comparisons. For reach-specific analysis, sampling sites were grouped into three reaches, as shown on Fig. 1. The upper reach includes sites HR1 to HR4 (outlet of the major tributary to the Lake Soyang, HR5). The middle reach covers the first dam on the monitored transect (HR5) through three downstream dams and ends up at HR 11,which receives discharge from the most downstream dam (HR10). The lower reach includes the main stem (HR13, HR14, and HR15) and tributary sites [TC, JN (HR12), and AY] downstream of the first submerged weir in Seoul (Fig. 1). Relationships between the three GHGs and water quality measurements were examined by regression analysis conducted with the data pooled for the whole basin and each cluster of three reaches and urban tributaries. Spatial variations among three reaches were further examined by principal component analysis (PCA).

3.1 Longitudinal variations in GHG concentrations and DOM characteristics along the Han River

Two basin-wide surveys from the forested headwater stream (HR1) through the impounded middle reach to the highly urbanized lower reach, combined with the monthly monitoring at six selected sites, revealed distinct longitudinal patterns of pCO₂ and concentrations of CH₄ and N₂O along the Han River (Figs. 2 and 3; Tables 1, S1, S2, and S3). The levels of the three measured GHGs were generally low along the upper reach, although noticeable increases occurred in the agricultural stream (HR2) compared with the very low values found in the forested headwater stream (HR1) (Table 1; Figs. 2 and 3). The levels of pCO₂ in the middle reach ranged from 51 to 2465 µatm, with an overall mean value of 551 µatm for all measurements at 7 middle-reach sites (Tables 1, S1, S2, and S3). The pCO₂ values at the four dam sites, ranging from 51–761 µatm, averaged 304 µatm, lower than the level expected for atmospheric equilibrium (∼435 µatm). The mean pCO₂ was 1109 µatm in the upper reach (range: 163–2539 µatm) and 2587 µatm in the lower reach (range: 78–11298 µatm), both exceeding the average value of atmospheric equilibrium by a factor of 2.5 and 5.9, respectively. The levels of pCO₂ tended to be higher in summer than in other seasons at all monthly monitoring sites except HR8 and HR 11 located in the middle reach (Fig. S2).

Figure 2Spatial variations in pCO₂ (a), CH₄ (b), and N₂O (c) measured at 15 sites and 3 urban tributaries along the Han River. pCO₂ was measured in two basin-wide surveys – July 2014 and May 2015 (May 2015 data are modified from Yoon et al., 2017), but CH₄ and N₂O were measured only in May 2015. Additional measurements at the urban tributary JN (HR12; n=13) and two other tributaries, TC (n=4) and AN (n=2) are indicated by brown squares with standard deviations. Dashed horizontal lines denote the mean value of atmospheric equilibrium for pCO₂ (435 µatm), CH₄ (3.3 nmol L⁻¹), and N₂O (6.5 nmol L⁻¹). Four vertical lines on the middle reach indicate the location of dams. The yellow shade demarcated by two spiral lines indicate the lower-reach section separated by two submerged weirs.

Download

Figure 3Box plots showing spatial variations in the three GHGs (a–c), in situ measurements (d–f), DOC and optical characteristics (g–i), and concentrations of ${\mathrm{NO}}_{\mathrm{3}}^{-}$-N (j), ${\mathrm{PO}}_{\mathrm{4}}^{\mathrm{3}-}$ (k), and Chl a (l) measured at six monthly monitoring sites (July 2014–July 2015) and an urban tributary outlet (HR12; May 2015–December 2017). A significant downstream change between two successive sites is indicated by one or two asterisks placed on top of the downstream site (P<0.05 or P<0.01, respectively; Mann–Whitney test).

Download

Both CH₄ and N₂O were always supersaturated with respect to atmospheric equilibrium (CH₄: 3.3 nmol L⁻¹; N₂O: 6.5 nmol L⁻¹) in all waters except the forested stream HR1 (Table 1; Figs. 2 and 3). Compared with the very low concentrations measured at the forested stream (1.2–7.2 nmol L⁻¹), CH₄ concentrations were relatively high at the downstream agricultural site HR2 (41–1719 nmol L⁻¹) and three dam sites (HR6, HR7, and HR10; 693–748 nmol L⁻¹) (Table 1). N₂O concentrations also increased from the low range of 8.9–20 nmol L⁻¹ in the forest stream to 28–46 nmol L⁻¹ in the agricultural stream and 212 nmol L⁻¹ at a dam (HR6). Similar to pCO₂, CH₄ and N₂O exhibited very high values along the lower reach separated by two submerged weirs on both ends (HR12–HR14). The concentrations of CH₄ and N₂O at these four lower-reach sites averaged 1812 (112–6496) and 298 nmol L⁻¹ (22–1396), respectively. The concentrations of the three GHGs at some lower-reach sites were similar to or exceeded the levels found in the three tributaries draining the urban sub-catchments located in the Seoul metropolitan area (Fig. 2). There was no clear seasonality in CH₄ and N₂O across the sites, although the concentrations of these two gases at the lower-reach site HR14 tended to be higher in spring and summer than in fall and winter (Fig. S2).

The levels of DOC and FI exhibited overall downstream increases along the main stem toward the river mouth with the exception of relatively high FI values found in the agricultural stream (Fig. 3; Tables S1, S2, S3). When Mann–Whitney U tests were conducted to detect downstream changes between two successive sites, both DOC and FI were significantly different between the two main stem sites (HR11, HR14) and the urban tributary JN (HR 12). Downstream increases in FI from the mean of 1.26 at the forest stream (HR1) to 1.50 at the most downstream site (HR15) may indicate an increasing contribution from the autochthonous DOM fraction (McKnight et al., 2001). Consistent with the increasing downstream trend of FI, HIX, as a measure of humification in soil organic matter (Zsolnay et al., 1999), decreased from the mean of 7.67 at the forest stream to 2.06 at HR 14 (the most downstream monthly monitoring site) and 0.79 at the estuarine site HR15, exhibiting significant changes over the transitions from HR1 to HR2 and from HR4 to HR8 (Fig. 3, Tables S1, S2, S3). The concentrations of three measured nutrients (${\mathrm{NH}}_{\mathrm{4}}^{+}$, ${\mathrm{NO}}_{\mathrm{3}}^{-}$, and ${\mathrm{PO}}_{\mathrm{4}}^{\mathrm{3}-}$) were generally higher at the agricultural stream (HR2) and the lower-reach sites (HR12 and HR14) than at the other sites (Fig. 3; Tables S1, S2, S3). Given the relatively small proportion of tributary discharge in the main stem flow ranging from ∼5 % in the monsoon period to 12 % in dry seasons, the comparison of monthly water quality measurements between the six sites and the urban tributary (HR12) illustrates the disproportionate influence of urban tributary inputs on the downstream increases in the concentrations of DOC and nutrients observed in the lower reach (Fig. 3). The urban tributary generally exhibited high levels of DOC, FI, protein-like DOM fluorescence (C3/DOC), and the three major nutrients (${\mathrm{NH}}_{\mathrm{4}}^{+}$, ${\mathrm{NO}}_{\mathrm{3}}^{-}$, and ${\mathrm{PO}}_{\mathrm{4}}^{\mathrm{3}-}$), but relatively low values of pH, DO, HIX, and Chl a (Fig. 3; Tables S1, S2, S3). Downstream changes in all water quality measurements from HR11 downstream of the last cascade dam to HR14 in the middle of the lower reach reflected either very high or very low values observed in the tributary (HR12) compared to the values observed in the impounded middle reach.

When we aggregated measurements as the three main stem reaches and the group of the three urban tributaries, at least one of the three GHGs exhibited a significant negative relationship with pH (pCO₂) and DO (pCO₂ and CH₄), and a positive relationship with DOC (CH₄ and N₂O) (Fig. 4). The significant negative or positive relationships identified for each aggregated data set generally conformed to the overall trends shown for the whole basin (Fig. 4). A positive relationship between DO and N₂O established in the lower reach was noticeable given no significant relationship found for the other reaches. Reach-specific clustering of data was also found on a PCA scatter plot with two primary components accounting for 60.3 % of the total data variation (Fig. S3). While the middle-reach data on the PCA plot overlapped considerably with the large scatter shown by the upper reach, the majority of the lower-reach data and most tributary data points were separated from the regions covered by the upper and middle reaches.

Figure 4Relationships between water quality (pH, DO, and DOC) and dissolved concentrations of the three GHGs (pCO₂, CH₄, and N₂O) measured in the Han River basin. Regression analysis was conducted with data aggregated for each of the upper, middle, and lower reaches, and the group of the three urban tributaries (TC, JN, and AY). Only significant (P<0.05) relationships are indicated by the regression line through the plot.

Download

Figure 5Kendall rank correlations (tau) between GHGs and water quality components measured in the upper, middle and lower reaches of the Han River. Significant correlations at P<0.05 and P<0.01 are indicated by blue and red, respectively. Please note that correlation analysis results were not available for ${\mathrm{PO}}_{\mathrm{4}}^{\mathrm{3}-}$ in the upper and middle reaches due to many values below detection limit.

Download

Kendall rank correlation analyses conducted with all measurements grouped for each of three reaches revealed reach-specific patterns of significant correlations for the three GHGs (Fig. 5; Table S4). Significant negative correlations were found between pCO₂ and three water quality parameters – pH, DO, and Chl a in the lower reach. These negative correlations were either absent for N₂O or significant only with DO in the case of CH₄. Compared to no or weak correlations observed for the upper and middle reaches, some parameters measured in the lower reach, including water temperature, ${\mathrm{PO}}_{\mathrm{4}}^{\mathrm{3}-}$, C1/DOC, and C2/DOC, exhibited strong positive correlations with pCO₂. Values of pCO₂ measured at the middle reach sites had some negative correlations with pH, TA, water temperature, cations (Na⁺, K⁺, and Mg²⁺), and anions (Cl⁻ and ${\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}$). These shifting correlations across the reaches were also found for CH₄ and N₂O, but to varying degrees. CH₄ in the lower reach had significant, but relatively weak correlations with ${\mathrm{PO}}_{\mathrm{4}}^{\mathrm{3}-}$, C1/DOC, and C2/DOC, whereas N₂O showed strong positive correlations with parameters such as EC, TA, all measured cations, Cl⁻, ${\mathrm{NO}}_{\mathrm{3}}^{-}$, ${\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}$, DOC, FI, and C3/DOC. In the case of CH₄, stronger correlations were found in the upper reach, including pH, EC, TA, cations (K⁺, Ca²⁺, and Mg²⁺), anions (Cl⁻ and ${\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}$), FI, and C3/DOC, but these correlations were weaker or insignificant in the middle and lower reaches.

3.2 Longitudinal variations in GHG concentrations and DOM characteristics along an urban tributary

Two synoptic samplings along the urban tributary (JN), complemented with additional samplings at a tributary outlet (HR12) and an upstream WWTP discharge, revealed the dominant influence of WWTP effluents on the concentrations of the three GHGs, DOC, and nutrients measured at the tributary outlet (Fig. 6; Table S5). All of the three GHGs exhibited similar levels and variations in the WWTP effluents and the tributary outlet, indicating a strong contribution of treated wastewater to the tributary gas export to the lower Han River. Both pCO₂ and N₂O concentrations in the tributary abruptly increased along the terminal section downstream of the WWTP. In contrast, CH₄ concentrations were very low (4.7–10 nmol L⁻¹) at the forested headwater stream, exhibited large fluctuations along the middle reach (279–5838 nmol L⁻¹), and remained similar or decreased slightly ranging from 422 to 6496 nmol L⁻¹ in the terminal section downstream of the WWTP. The levels of the three gases displayed similar large temporal variations at the WWTP effluents and the tributary outlet (Fig. 6e, f, and g). The observed temporal variations in gas concentrations were not significantly related to fluctuations in temperature, precipitation, and streamflow measured near the tributary outlet site, although the three gases tended to be lower in concentration during wet summer months in 2016 and 2017 (Fig. 6d).

Figure 6Longitudinal variations in pCO₂ (a; data reported in Yoon et al., 2017), CH₄ (b), and N₂O (c) along the urban tributary (Joongnang River: JN) surveyed in November 2015 and May 2016 and temporal variations in hydroclimatic conditions (d; monthly precipitation and mean temperature and flow), pCO₂ (e), CH₄ (f), and N₂O (g) monitored at the JN outlet (HR12) and WWTP effluents from May 2015 to December 2017. Box plots shown on the left panel summarize the time series data obtained from the WWTP effluents and tributary outlet, as displayed on the right panel. Weather data were obtained from an automatic weather station adjacent to HR12, while flow was measured at a bridge upstream of HR12.

Download

3.3 C isotope ratios in DOM, CO₂, and CH₄

Despite large scatters on the dual plot of δ¹³C and Δ¹⁴C in DOM sampled across the Han River basin, the dual isotopic signatures for the urban tributary JN, WWTP effluents, and the most downstream site HR14 in the lower reach were clearly separated from the values found at the forested headwater stream HR1 (Fig. 7; Table S6). In both the wet-season (July 2014) and dry-season (May 2015) samplings, Δ¹⁴C was relatively high at the forested stream HR1, with a higher value in July 2014 (58.9 ‰; modern age) than in May 2015 (−29.6 ‰; 180 years old). Δ¹⁴C values exhibited an overall decreasing downstream trend toward the lower-reach site HR14 (July 2014: −78.2 ‰, 590 year old; May 2015: −87.9 ‰, 675 years old), with higher values observed across the main stem sites in July 2014 (−129.1 ‰ to 66.8 ‰) than in May 2015 (−113.5 ‰ to −29.6 ‰). An exceptionally high Δ¹⁴C value (66.8 ‰) was found in July 2014 at the second most downstream site (HR11) located 17 km downstream of the last of the cascade dams, indicating an influence of ¹⁴C-enriched DOM discharged from the upstream dams. The values of Δ¹⁴C measured in the urban tributary and WWTP effluent (around −100 ‰) were lower than those measured at all the main stem sites. δ¹³C was highly variable at two middle-reach sites and increased almost linearly from the second most downstream site through the urban tributary and WWTP effluents to the last site (Fig. 7).

Figure 7δ¹³C and Δ¹⁴C in DOM sampled at five Han River main stem sites, an urban tributary, and a WWTP discharge. Main stem (HR1, HR4, HR8, HR11, and HR14) and tributary (HR12) samples were collected in July 2014 and May 2015, whereas two WWTP effluent samples were obtained in June 2014 and November 2015. Reference values for DOM sources presented on the left plot are from Marwick et al. (2015) and references therein.

Download

Figure 8Spatial variations in pCO₂ and ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{{\mathrm{CO}}_{\mathrm{2}}}$ (a), CH₄ concentrations and ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{{\mathrm{CH}}_{\mathrm{4}}}$ (b), and N₂O and Chl a concentrations (c) measured at 8 main stem sites and 3 urban tributary outlets during a cruise expedition (June 2016) along the lower reach of the Han River. The grey shaded area indicates the main stem section receiving the inflow from the urban tributary JN (HR12). The four vertical dashed lines denote the locations of the four monthly monitoring sites along the lower reach. Data of pCO₂ and ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{{\mathrm{CO}}_{\mathrm{2}}}$ were modified from Yoon et al. (2017).

Download

The concentrations of the three GHGs and the values of δ¹³C in CO₂ and CH₄ (Fig. 8; Tables S7, S8) measured along a cruise transect exhibited large increases in gas concentrations and either gradual increases in ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{{\mathrm{CO}}_{\mathrm{2}}}$ or abrupt decreases in ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{{\mathrm{CH}}_{\mathrm{4}}}$ along the confluence of the urban tributary JN (HR12). ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{{\mathrm{CO}}_{\mathrm{2}}}$ values ranged from −20.9 ‰ to −16.7 ‰ (Table S8), which were distinctively higher than the range of δ¹³C_DOC measured at the same sites (−28.2 ‰ to −20.6 ‰) (Table S6). The gradual downstream increases in ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{{\mathrm{CO}}_{\mathrm{2}}}$ along the main stem transect reflected the tributary contributions to the main stem isotopic composition, because the values found in the two upstream tributaries (−18.2 ‰, −18.3 ‰) and a downstream tributary (−14.7 ‰) were higher than the upstream main stem values (Table S8). The downstream tributary had lower pCO₂, higher δ¹³CO₂, and higher concentrations of CH₄ and Chl a relative to the upstream tributaries (Fig. 8). The values of ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{{\mathrm{CH}}_{\mathrm{4}}}$ showed an overall increasing trend along the lower reach downstream of the urban tributary JN (Fig. 8; Table S8). While the tributary concentrations of N₂O were generally higher than those of the main stem sites, Chl a concentrations were lower at the two upstream tributaries (TC and JN), but higher at the downstream tributary AY than the values observed at the adjacent main stem sites (Fig. 8c).

4.1 Reach-specific patterns and controls of the three GHGs

Building on our previous report on CO₂ dynamics in the Han River basin (Yoon et al., 2016, 2017), this study provided a more comprehensive view of longitudinal patterns in the three GHGs across the upper, middle, and lower reaches affected by different types and magnitudes of anthropogenic perturbations (Figs. 2–4). The three gases exhibited large longitudinal variations resulting from gas-specific increases or decreases at impoundment-affected sites and localized concentration peaks along the lower reach downstream of polluted urban tributary inflows. Although pCO₂ at some impoundment-affected sites was very low, approaching or falling below the levels expected for atmospheric equilibrium, CO₂ and the other two gases were generally supersaturated with respect to atmospheric equilibrium across the river basin (Table 1; Fig. 2). The very high levels of pCO₂ observed in the main stem (up to 4132 µatm) and the three urban tributaries(up to 11970 µatm) of the lower Han River fall in the high ranges found in some polluted rivers in Europe (Kempe, 1984; Frankignoulle et al., 1998; Borges et al., 2006) and China (Ran et al., 2015; Liu et al., 2016; Wang et al., 2017). Consistent with these previous reports on pCO₂ and other studies reporting high levels of CH₄ and N₂O in polluted urban waters (Garnier et al., 2013; Yu et al., 2013; Smith et al., 2017; Wang et al., 2017b, 2018), the results observed in the lower Han River emphasize the dominant influence of urban tributaries carrying wastewater as a primary anthropogenic source of GHGs in the highly urbanized river system.

Different longitudinal patterns of the three measured GHGs (Fig. 2), together with reach-specific significant correlations between the GHGs and other measured water quality components (Fig. 5; Table S4), illustrate that the spatial distribution and temporal dynamics of the three gases in the Han River basin may be significantly different from those found in pristine or less anthropogenically modified river systems. The extent of major natural sources such as wetlands and inundated floodplains has been considered as a primary factor for longitudinal variations in CO₂ and CH₄ in large rivers (Richey et al., 1988; Borges et al., 2015; Abril et al., 2014). Those natural sources are rarely found in the Han River basin, where the middle and lower reaches have been modified substantially by man-made structures. This lack of natural sources, combined with the differential patterns of the three GHGs attributed to dams and urban wastewater, suggests that increased water retention time and nutrient enrichment may play crucial roles in the production and consumption of the three GHGs in this highly regulated river system (Crawford et al., 2016). It would be very challenging to identify the primary drivers out of the multiple factors involved in the production and consumption of the three GHGs in urbanized river systems, because some factors responsible for explaining the mechanism for two gases (e.g., CO₂ and CH₄) may be totally invalid for the other gas (e.g., N₂O) (Smith et al., 2017; Wang et al., 2017b). However, the observed longitudinal patterns of the three GHGs (Figs. 2–4), along with their correlations with specific sets of water quality components (Fig. 5), make one thing clear. The primary factors and mechanisms for the production and consumption of these three gases may change in response to longitudinal variations in dominant anthropogenic perturbations, often abruptly as shown by the localized concentration peaks of GHGs downstream of urban tributary inflows (Figs. 2 and 8).

The highly variable concentrations of CH₄ and N₂O along the middle reach in contrast to the consistently low levels of pCO₂ at the dam sites (Figs. 2–4) suggest that the rates of concomitant metabolic processes involved in the production and consumption of these gases in reservoir water and sediments may vary with predominant dam conditions such as water depth and sediment accumulation. The low values of pCO₂ measured at the impoundment-affected sites including site HR11 downstream of the last dam (HR10) indicate an enhanced planktonic CO₂ uptake, in agreement with the lowered pCO₂ levels as a consequence of increased primary productivity in some eutrophic impounded reaches of the Mississippi (Crawford et al., 2016), the Yangtze (Liu et al., 2016), and a Yellow River tributary (Ran et al., 2017). However, some previous studies have reported drastic increases in CO₂ and CH₄ emissions from the flooded vegetation and soils in the initial years following dam construction (Abril et al., 2005; Chen et al., 2009; Shi et al., 2017). Given the negative relationship established between reservoir age and emissions of CO₂ and CH₄ from a wide range of hydroelectric reservoirs (Barros et al., 2011), the initial pulse-like gas emissions from the newly constructed reservoirs might gradually decrease with increasing reservoir age. Several studies have reported significant negative relationships between Chl a and pCO₂ in eutrophic impounded rivers (Crawford et al., 2016; Liu et al., 2016; Ran et al., 2017). However, in the middle reach of the Han River, pCO₂ was not correlated with Chl a, but had some significant positive correlations with DOM optical properties such as HIX, C1/DOC, and C2/DOC (Fig. 4). While HIX and C1/DOC indicate the degree of humification and the proportion of terrestrial DOM components, respectively, C2/DOC represents the proportion of “microbial humic components” in the bulk DOM (Fellman et al., 2010; Parr et al., 2015). The observed concentration range of Chl a might be too narrow to detect any significant correlation with pCO₂. In contrast, the relatively high levels of pCO₂ concurred with strong optical intensities of terrestrial DOM components at some middle reach sites that are less affected by impoundments (e.g., HR 7 and HR8), resulting in the significant correlations between the relatively wide ranges of pCO₂ and DOM optical properties.

Although a small number of measurements of CH₄ and N₂O and their inconsistent spatial patterns observed along the impounded middle reach (Fig. 2; Tables S2, S3) require a cautious assessment of impoundment effects on the concentrations of the two gases, the higher values of CH₄ and N₂O in some dams and outflows compared to the upstream levels indicate dam-specific conditions driving the production and consumption of CH₄ and N₂O in reservoir water and sediments. Crawford et al. (2016) observed a weak summertime CO₂ sink due to enhanced photosynthesis but elevated concentrations and fluxes of CH₄ along the upper Mississippi River impounded by a series of low dams constructed for river navigation. They attributed the observed CH₄ supersaturation to anaerobic conditions in organic-rich sediments. Despite no correlation between CH₄ and DO levels in the middle reach (Fig. 5; Table S4), the fact that CH₄ levels were higher at the three shallower dam sites (HR6, HR7, and HR10) than at the most upstream dam (HR5) with a maximum depth of 110 m implies the balance between anaerobic CH₄ production in bottom sediment and aerobic and anaerobic CH₄ oxidation in the water column with a vertical gradient of O₂ availability as a driving force for the observed spatial variations in CH₄ concentrations (Roland et al., 2017). Shallow reservoirs and river inflows accumulating methanogenic sediments have been identified as CH₄ emission hot spots, in which relatively short retention time of sediment-derived CH₄ in aerobic water column can lead to reduced rates of CH₄ oxidation compared with deeper reservoirs (Maeck et al., 2013; Beaulieu et al., 2014). CH₄ concentrations in the middle reach exhibited a weak, but significant correlation with DOC concentrations (Fig. 5, Table S4). This correlation may indicate an active methanogenesis in anaerobic reservoir sediments that is often accompanied by increases in surface water DOC concentrations (Chen et al., 2009; Wang et al., 2017b). It is also possible that some local sources of organic wastes surrounding the reservoirs may have directly discharged wastewater rich in DOC and CH₄ (Bergier et al., 2014; Wang et al., 2017b). Higher CH₄ concentrations measured at warmer temperatures are also in agreement with the increased rates of CH₄ production observed during warm summer months in a shallow reservoir in Ohio, USA (Beaulieu et al., 2014). N enrichment, alone or in combination with anaerobic conditions favorable for denitrification, has been suggested as a key control on N₂O production in impoundments, although strictly anaerobic conditions might result in a more complete denitrification to N₂, producing little N₂O (Beaulieu et al., 2015; Wang et al., 2017b). The lack of clear impoundment effects on N₂O concentrations except for one reservoir (HR6; Fig. 2) can be explained by little N₂O production in the other reservoirs or the complex interplay between N₂O production from nitrification and denitrification and N₂O consumption under changing availability of O₂ (Beaulieu et al., 2015).

Large increases in GHG concentrations along the lower reach may be a combined result of the net in-stream production and direct inputs from WWTPs. Direct influences of wastewater-derived GHGs have been observed in urban rivers receiving WWTP effluents (Garnier et al., 2013; Yu et al., 2013; Burgos et al., 2015; Alshboul et al., 2016; Wang et al., 2017a; Yoon et al., 2017). For example, large pulsatile emissions of CH₄ and N₂O in the Guadalete River estuary in Spain were found near the discharge from a WWTP, as a combined result of direct gas emissions from WWTP effluents and indirect effects on the production of CH₄ and N₂O in the water channel and bottom sediments downstream (Burgos et al., 2015). Previously we used a mass balance approach based on three cruise underway measurements of pCO₂ and DOC and the estimated rates of CO₂ outgassing from the same reach of the lower Han River, and additional measurements of pCO₂ and DOC at two urban tributaries (TC and JN) to show that the two tributaries JN and TC delivering WWTP effluents accounted for up to 72 % of the CO₂ concentration measured at a downstream location of the lower reach (Yoon et al., 2017). When the rates of CO₂ production, consumption, and outgassing were estimated using the mass balance approach for a section upstream and two sections downstream of the two tributaries lower reach in June 2016 (Yoon et al., 2017), the amount of CO₂ produced from organic matter biodegradation was much greater than the amount of CO₂ consumed by phytoplankton and similar to the CO₂ outgassing to the atmosphere. In May 2015, when Chl a concentrations were much higher than in June 2016, the mass balance suggested that the bulk of CO₂ delivered by the tributaries might have been consumed by phytoplankton photosynthesis in the downstream section of the lower reach (Yoon et al., 2017). As observed in other polluted rivers enriched in labile DOM moieties derived from urban sewage (Guo et al., 2014), newly produced CO₂ from active microbial decomposition can sometimes exceed the phytoplankton uptake of CO₂ depending on the prevailing environmental conditions. By directly measuring δ¹³C in CO₂ respired by bacterioplankton in two streams and eight lakes in Canada, McCallister and del Giorgio (2008) showed that the production of CO₂ through bacterial degradation of terrigenous DOM decreased in sharp contrast to the increasing proportion of algal-derived DOC and CO₂ with increasing levels of Chl a. Their findings implied an algal-driven activation of metabolism in eutrophic freshwater systems, although it would require further research to verify the findings in the boreal freshwaters in temperate and other biomes. It remains largely unexplored how the balance between autotrophy and heterotrophy in many eutrophic river systems found across the temperate zones shifts in response to changing environmental conditions (Garnier and Billen, 2007).

Large downriver increases (Fig. 2) and temporal variability (Figs. 3 and S1) in CH₄ and N₂O along the lower reach may result from a combination of processes including direct WWTP discharge and in-stream production and consumption. As indicated by the similar ranges of CH₄ and N₂O found in the WWTP effluents and the tributary JN outlet (Fig. 6), the amount of CH₄ and N₂O discharged from the WWTP appeared to drive the magnitude and temporal variability of the tributary inputs to the lower reach. In the case of CH₄, however, the large spatial and temporal variations observed along the tributary upstream of the WWTP also indicate the potential role of bottom sediments as an upstream source of CH₄ (Stanley et al., 2016), although further research is needed to elucidate the in-stream production of CH₄ in the tributary. Despite the dominant influence of tributary exports, other riverine processes also need to be considered to explain the complex downstream spatial patterns of these gases, as indicated by the gas-specific sets of significant correlations with measured water quality components (Fig. 5). In contrast to the nonsignificant correlation between CH₄ and DO in the impounded middle reach, CH₄ concentrations in the lower reach exhibited a significant negative correlation with (DO) as well as positive correlations with ${\mathrm{PO}}_{\mathrm{4}}^{\mathrm{3}-}$, water temperature, C1/DOC, and C2/DOC. As observed in other urbanized river systems (Beaulieu et al., 2015; Smith et al., 2015; Wang et al., 2018), CH₄ correlated positively with ${\mathrm{PO}}_{\mathrm{4}}^{\mathrm{3}-}$, but negatively with DO, implying that the nutrient enrichment often leading to severe phytoplankton blooms during warm summer months may create favorable conditions for anaerobic methanogenesis in the lower reach that is almost impounded by the two submerged weirs. Out of many water quality components that were significantly correlated with N₂O (Fig. 5), ${\mathrm{NH}}_{\mathrm{4}}^{+}$ and ${\mathrm{NO}}_{\mathrm{3}}^{-}$ have been considered as two primary predictors of riverine N₂O, particularly in urban streams and rivers influenced by sewage (Yu et al., 2013, He et al., 2017; Smith et al., 2017). The lack of any significant correlation between N₂O and DO in the lower reach may indicate a potential N₂O production from nitrification and incomplete denitrification under widely varying levels of DO (Fig. 5).

4.2 Tracking sources of DOM and GHGs using C isotope ratios

The dual C isotope signatures of DOM measured at the five monthly monitoring sites and the outlet and WWTP discharge of the urban tributary JN overlapped considerably with the reported ranges for soil organic matter (SOM) and phytoplankton biomass (Fig. 7) (Raymond and Bauer, 2001; Marwick et al., 2015 and references therein). When Marwick et al. (2015) compiled 695 Δ¹⁴C_DOC data collected in a wide variety of rivers around the world, most data points fell in the range from −100 ‰ to +200 ‰ and 72 % of the Δ¹⁴C_DOC data indicated a modern age, contrasting with the much smaller proportion of “modern” POC (22 % of 483 Δ¹⁴C_POC measurements). The relatively large site-to-site variations in both δ¹³C_DOC and Δ¹⁴C_DOC make it very difficult to evaluate the relative contributions of allochthonous and autochthonous sources to the isotopic signatures of the bulk DOM. While δ¹³C_DOC was highly variable along the upper to middle reaches from HR2 to HR11, the values of δ¹³C_DOC at the most downstream site HR14 (−20.6 ‰ and −23.2 ‰) were distinctively higher than those measured at the forested headwater stream (−28.2 ‰ on both sampling dates) (Table S6). The DOM optical properties measured at HR14 were also significantly different from the high HIX and low FI values indicating the predominance of soil-derived DOM in the headwater stream (Fig. 3; Table S1). Taken together, the isotopic composition and optical properties of DOM in the lower reach may reflect the downstream addition of DOM components derived from anthropogenic sources such as WWTP effluents (δ¹³C_DOC around −26 ‰; Table S6) or plankton biomass (note the wide range of the plankton δ¹³C_DOC in Fig. 7).

The distinct seasonal differences in Δ¹⁴C_DOC across the five main stem sites illustrate that the age of DOM is generally younger during the monsoon period (July 2014; modern to 590 years BP) than in the dry season (May 2015; 180 to 675 years BP) (Fig. 7; Table S6). This seasonality might have resulted from an increased contribution of DOM components released from terrestrial sources during monsoon rainfalls. The enrichment of ¹⁴C in DOM identified in other temperate river systems has been attributed to a larger contribution from ¹⁴C-enriched litter and soil organic matter and shorter residence times of DOM in systems with higher annual precipitation (Raymond and Bauer, 2001; Butman et al., 2015). The longitudinal increase in DOM age from a modern age to 180 years BP at the forested headwater stream to 590–675 years BP at the most downstream site may reflect a preferential degradation of young, labile components during riverine DOM transport (Raymond and Bauer, 2001), but also indicates a significant contribution of aged DOM derived from downstream anthropogenic sources; for example, the age of DOM measured at the outlet and WWTP effluents of the urban tributary JN ranged from 765 to 1050 years BP (Table S6). The values of ¹³C_DOC (−25.8 ‰) and Δ¹⁴C_DOC (−97.9 ‰ and −113.5 ‰) measured in the WWTP effluents (Table S6) were similar to those reported by other studies (Griffith et al., 2009; Butman et al., 2015). WWTP effluents have been shown to contain aged organic matter with characteristic C isotopic composition (Griffith et al., 2009; Griffith and Raymond, 2011; Butman et al., 2015). As suggested by Griffith and Raymond (2011), aged DOM derived from the WWTP effluents (765–905 years BP; Table S6) may contain labile materials, which, mixed with other labile components from in-stream sources such as phytoplankton, can fuel the riverine heterotrophy along the lower reach.

While δ¹³C has usually been measured for the total DIC consisting of dissolved CO₂, ${\mathrm{HCO}}_{\mathrm{3}}^{-}$, and ${\mathrm{CO}}_{\mathrm{3}}^{\mathrm{2}-}$ to track downstream changes in DIC sources along streams and rivers (Barth et al., 2003; Schulte et al., 2011; Deirmendjian and Abril, 2018), our measurements of ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{{\mathrm{CO}}_{\mathrm{2}}}$ and ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{{\mathrm{CH}}_{\mathrm{4}}}$ (Fig. 8; Table S8), combined with the dual C isotope signatures of DOM (Fig. 7; Table S6), also provided some insights into the contributions of treated wastewater and enhanced outgassing and phytoplankton growth to the riverine isotopic composition of CO₂ and CH₄ in the eutrophic lower reach. The values of ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{{\mathrm{CO}}_{\mathrm{2}}}$ increasing from −20.9 ‰ at 76 km from the river mouth to −16.7 ‰ at 50 km fall within the range of δ¹³C measured for CO₂ dissolved in riverine and estuarine waters (−25 ‰ to −15 ‰) (Longinelli and Edmond, 1983; Maher et al., 2013). However, the values reported here are less negative than the ranges of δ¹³C measured directly for CO₂ respired by bacteria consuming organic matter of terrestrial and algal origin in two streams and eight lakes in Canada (−32.5 ‰ to −28.4 ‰) (McCallister and del Giorgio, 2008). The comparison of the lower-reach ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{{\mathrm{CO}}_{\mathrm{2}}}$ values observed > 10 km downstream of the tributary JN (−17.4 ‰ to −16.7 ‰; Table S8) with the δ¹³C_DOM measured at the forest stream HR1 (−28.2 ‰; Table S6) and the most downstream site HR 14 (−20.6 ‰ to −23.2 ‰; Table S6) suggests that the ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{{\mathrm{CO}}_{\mathrm{2}}}$ in the lower reach does not directly result from the isotopic composition of CO₂ originating from the microbial degradation of soil-derived DOM or the mixture of DOM components in the lower reach. In an inter-regional comparison of δ¹³C_DIC measured at 318 streams across Sweden, Campeau et al. (2017) estimated ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{{\mathrm{CO}}_{\mathrm{2}}}$ based on carbonate-equilibrium fractionation and ¹³C enrichment between the three DIC species (Zhang et al., 1995). The ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{{\mathrm{CO}}_{\mathrm{2}}}$ values estimated by Campeau et al. (2017) exhibited a significant negative relationship with DOC concentrations, with a particularly low median value of ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{{\mathrm{CO}}_{\mathrm{2}}}$ (−25.3 ‰) matching with a high median DOC concentration (28.7 mg L⁻¹) in headwater streams in a heavily forested region. The very low ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{{\mathrm{CO}}_{\mathrm{2}}}$ values in the DOC-rich forest streams in Sweden (Campeau et al., 2017) were close to the δ¹³C values for CO₂ released from bacterial DOM degradation (McCallister and del Giorgio, 2008), but much lower than those observed along the lower Han River. Therefore, we need to consider other sources than the soil-derived DOM to account for the isotopic composition of CO₂ dissolved in the lower reach. Although we did not measure ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{{\mathrm{CO}}_{\mathrm{2}}}$ and ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{{\mathrm{CH}}_{\mathrm{4}}}$ in WWTP effluents, the comparison of the ranges of ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{{\mathrm{CO}}_{\mathrm{2}}}$ and ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{{\mathrm{CH}}_{\mathrm{4}}}$ observed in the lower reach and the three urban tributaries delivering large loads of treated wastewater (Fig. 8; Table S8) suggests that WWTP-derived gases can directly influence the isotopic composition of the two gases dissolved in the main stem reach downstream of the tributary inflows.

We can also benefit from rich literature information on ${{\mathit{\delta }}^{\mathrm{13}}\mathrm{C}}_{\mathrm{DIC}}$ to further constrain the observed 4 ‰ downstream increase in ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{{\mathrm{CO}}_{\mathrm{2}}}$ along the lower reach (Fig. 8). Biogenic DIC originating from autotrophic respiration or mineralization of soil organic matter derived from C3 plants has a typical value around −27 ‰ but this δ¹³C_DIC value can increase up to −23 ‰ in soil solution as a result of 1 ‰–4 ‰ isotopic enrichment caused by dissolution and gas exchange between the soil and atmosphere (Doctor et al., 2008; Schulte et al., 2011). When this biogenic DIC mixes with atmospherically equilibrated DIC and geogenic DIC dissolved from carbonates, both of which have a typical δ¹³C value around 0 ‰, the resulting riverine δ¹³C_DIC usually ranges from −15 ‰ to −5 ‰ (Telmer and Veizer et al., 1999; Barth et al., 2003; Schulte et al., 2011). The values of ${{\mathit{\delta }}^{\mathrm{13}}\mathrm{C}}_{\mathrm{DIC}}$ estimated from our measurements of ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{{\mathrm{CO}}_{\mathrm{2}}}$ based on carbonate equilibrium fractionation and the enrichment factors between the DIC species determined by Zhang et al. (1995) ranged from −13.0 ‰ to −8.8 ‰, reflecting the influence of both biogenic and geogenic DIC. As mentioned earlier, the urban tributaries enriched in ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{{\mathrm{CO}}_{\mathrm{2}}}$ relative to the main stem might have contributed to the 4 ‰ downstream increase in the estimated δ¹³C_DIC. In addition, the enrichment may also reflect the net effect of a combination of concomitant processes including gas evasion to the atmosphere (Doctor et al., 2008) and photosynthesis (Finlay et al., 2004) that enrich the remaining DIC in ¹³C through the preferential removal of the lighter ¹²CO₂. The outgassing of CO₂ can result in a pronounced increase in δ¹³C_DIC, such as the 3 ‰–5 ‰ enrichment along a relatively short (0.5 km) transect of a forested headwater stream reported by Doctor et al. (2008). However, given the concomitant downstream increases in ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{{\mathrm{CO}}_{\mathrm{2}}}$ and Chl a concentrations, we cannot rule out a potential contribution of enhanced primary production in the eutrophic lower reach to the observed downstream¹³C enrichment. The high values of δ¹³C observed in the tributaries, especially at the most downstream tributary AY showing the highest levels of ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{{\mathrm{CO}}_{\mathrm{2}}}$ and Chl a, may also reflect that in-stream processes such as photosynthesis and CO₂ outgassing can modify the isotopic ratios of wastewater-derived CO₂, resulting in the observed inter-tributary variations in pCO₂ and ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{{\mathrm{CO}}_{\mathrm{2}}}$. Another potential source of isotopic enrichment might be the production of CH₄ from the reduction of CO₂ in the deeper water and bottom sediment, which consumes lighter C resulting in an enrichment of the remaining DIC pool (Barth et al., 2003). However, this mechanism might be less important than other described processes, because the longitudinal increase in δ¹³C in CH₄ (Fig. 8) indicates an active oxidation occurring through the shallow downstream reach.

While agricultural activities, including rice cultivation, animal husbandry, and N fertilization, represent the primary anthropogenic source of CH₄ and N₂O in anthropogenically impacted river systems (Silvennoinen et al., 2008; Garnier et al., 2013; Borges et al., 2018), the relative contribution of wastewater and landfills often increases drastically in urbanized watersheds (Yu et al., 2013; Smith et al., 2017; Wang et al., 2017b). Downstream changes in CH₄ concentrations along the lower reach reflect localized impacts of urban tributary inputs, which caused a large increase in the main stem CH₄ concentrations from 136 to 3088 nmol L⁻¹ (Table S7) and a pronounced decrease in ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{{\mathrm{CH}}_{\mathrm{4}}}$ from −36.6 ‰ to −48.6 ‰ (Table S8) across the inflows of the two tributaries TC and JN (Fig. 8). The concentrations of CH₄ peaked in the tributary inflow, but gradually decreased along the downstream reach toward the river mouth, indicating an efficient removal of CH₄ through oxidation and/or evasion to the atmosphere (Maher et al., 2013; Sawakuchi et al., 2016). Similar localized effects of WWTP effluents were observed in the lower reaches of Seine River, downstream of Paris and Rouen (Garnier et al., 2013). The contrasting downriver trends of decreasing CH₄ concentrations and ¹³C enrichment are consistent with the longitudinal patterns of CH₄ concentration and its stable C isotope ratios measured simultaneously in large rivers such as the Amazon River (Sawakuchi et al., 2016) and estuaries (Maher et al., 2013). Sawakuchi et al. (2016) found the increases in ${\mathit{\delta }}^{\mathrm{13}}{\mathrm{C}}_{{\mathrm{CH}}_{\mathrm{4}}}$ and the abundance of a genetic marker for methane-oxidizing bacteria (pmoA) in waters with lower CH₄ concentrations across the main stem and tributaries of the Amazon. They used stable isotopic mass balances of CH₄ in the water column and estimated that 17 %–100 % of CH₄ produced in the riverbed sediment may be oxidized during transport through water column to the atmosphere. During cruise expeditions employing a cavity ring-down spectroscope (CRDS) along a 15 km reach of the North Creek estuary in Australia, Maher et al. (2013) observed increasing δ¹³C values from −61.07 ‰ to −48.62 ‰ in contrast to large decreases in CH₄ concentrations from 74 to 2 nmol in the downstream direction. CH₄ oxidation was suggested as the primary driver of downstream increases in δ¹³C in the studied estuary with relatively low levels of anthropogenic pollution (Maher et al., 2013). The downriver patterns of CH₄ concentration and isotopic composition observed in this study also suggest that CH₄ oxidation in the well-mixed, shallow water, in combination with physical evasion to the atmosphere, may efficiently remove CH₄ derived from multiple sources including the urban tributaries enriched in CH₄ and riverbed sediments affected by the eutrophic water and frequent phytoplankton blooms.