1 Introduction

Inland river systems are important places for the understanding of global carbon cycle, due to their disproportionate contribution to inorganic carbon fluxes. With only 0.57% of the global land area, inland rivers annually transport about 0.4 Pg inorganic carbon to the ocean transversely and release about 1.8 Pg CO₂ to the atmosphere vertically, and their sum is close to the annual net absorption of the global terrestrial system (about 2.6 Pg C) (Bauer et al., 2013; Raymond et al., 2013; Wehrli et al., 2013;

Poulter et al., 2014; Grill et al., 2019). However, highly natural biogeochemical processes and enhanced anthropogenic perturbations (such as urban expansion and artificial storage) results in a large uncertainty in above estimates in inland rivers, and thus the studies on the controls of riverine carbonate system attract increasing attention in recent years to mitigate global climate change (Lauerwald et al., 2015; Marx et al., 2017; Gao et al., 2021; Ran et al., 2021).

Compared with the most concerned big rivers with large volume, small watershed rivers usually had a more sensitive carbonate system due to their weak dilution effect and the fragile ability to resist environmental interference (Ran et al., 2015; Wang et al., 2021). Though natural factors (such as carbonate and silicate weathering, exogenous input of CO₂ from soil and groundwater) formed their initial dissolved inorganic carbon (DIC) at headwater or upstream (Moquet et al., 2011; Song et al., 2021; Zhang et al., 2021), the changes of eco-environment and land use type would be of extremely importance to DIC variability at the mid-downstream of small watershed rivers (e.g., Hafsi et al., 2016; Zhong et al., 2018; Chen et al., 2021; Wang et al., 2021). In the humid karst areas in Southwestern China, Wang et al. (2021) found that sewage input increased the CO₂ concentration in two small rivers flowing through urban areas by 2–2.5 times than the suburb areas, and Chen et al. (2021) emphasized the difference in the impact of farmland and urban land on DIC increase in three small rivers. Similar urbanization effect on riverine CO₂ was also reported in Han River in South Korea and some small rivers in Southwest Germany (Alshboul et al., 2017; Yoon et al., 2017; Zhong et al., 2018). Thus, for an in-depth understanding of the inorganic carbon cycle in rivers, studies in typical small basin rivers have inherent advantages due to their large mutability to natural changes and anthropogenic disturbances, especially the rivers in arid or semi-arid areas with less rains.

The Loess Plateau is located in the north of central China, accounting for 6.6% of China's total land area. The loess is an aeolian deposit and has a highly homogeneous material composition with 5% – 20% carbonate content (Yang et al., 1985). Thus, due to a less vegetation coverage on the Loess Plateau, the strong mechanical erosion and chemical weathering resulted in a high concertation of riverine DIC (about 35 mg L⁻¹) in the Yellow River, the sixth longest river in the world (Cai et al., 2008; Fan et al., 2014). Meanwhile, arid and semi-arid climate cause the rivers on the Loess Plateau to be featured with small river discharge, and the Yellow River only has a discharge accounting for about 4.3% of Yangtze River (China) or about 7.7% of the Mississippi River (USA) (Wang et al., 2016). Fragile ecological environment and small river runoff had become the typical characteristics for the rivers on the Loess Plateau, which would further highlight the impact of anthropogenic interference such as damming, agricultural activities and urban construction on riverine DIC. However, the current research on river carbon transport on the Loess Plateau was limited, especially in the small watershed rivers (Zhang et al., 2015; Wu et al., 2016), which greatly restricted the accurate estimation of river CO₂ budget. The Fenhe River in Shanxi Province in the eastern area of the Loess Plateau is a tributary of the middle reaches of the Yellow River, and shows high human interference. In 2022, the total population and GDP within the Fenhe River Basin accounted for 39% and 50% of the total amount in Shanxi Province, respectively (Shanxi Provincial Bureau of Statistics, 2022). In this study, through seasonal sampling surveys, we investigated the spatiotemporal distributions of riverine DIC and their controls, revealed the important impact of aquatic production induced by damming as well as urban sewage input, and assessed the dynamic change of riverine CO₂. The results of this study will be helpful to clarify the control mechanism of riverine CO₂ system under high urbanization in semi-arid regions.

2 Materials and Methods 2.1 Study Area

This study was conducted in the Fenhe River (35˚ – 39˚N and 110˚ – 114˚E), which is the second longest tributary of the Yellow River and located in the east of the Loess Plateau with an extensive loess cover. The Fenhe River originates in the Mount Guancen, the northern Lvliang Mountains, with a total length of 713 km and a drainage area of 39721 km². From north to south, it gradually transited from mountainous river (the upstream, stations F1 – F8) to plain river (the mid-downstream, stations F9 – F27), with an elevation drop of about 1000 m (Fig. 1). In the upstream, except for the reservoir area, the river width is generally less than 50 m and the water depth is < 3 m. In the mid-downstream, a maximum width of 500 m and a maximum depth of 6 m showed. Dominated by a temperate semi-arid monsoon climate, the multi-year (1956 – 2016) average annual precipitation and runoff were 509 mm and 9.18×10⁸ m³, respectively (Yellow River Conservancy Commission of China, 2022). Also, nearly 50% of the annual rainfall concentrated in July and August in hot and rainy summer, while in cold winter some upstream sections of the river would be completely frozen.

The Fenhe River is the largest river in Shanxi Province, one of the major coal production areas in China. The upstream of the Fenhe River (stations F1 – F8) are the important drinking water source, with a good vegetation coverage dominated by forests and grassland (Fig. 1c). In contrast, the middle (stations F9 – F19) and lower reaches (stations F20 – F27) flow through the provincial capital (Taiyuan City) and economic core areas with an about 13.5 million population and an about 130-billion-dollar GDP in 2021 (Shanxi Provincial Bureau of Statistics, 2021), and received about 50% of the total provincial sewage, showing an intense human disturbance. On the rock distribution, the Fenhe River Basin is covered by Quaternary loess with uneven thickness, especially the middle and lower reaches. Triassic and Jurassic sandstone and shale are also exposed on both sides of the upstream valley. Also, Cambrian and Ordovician limestone and Carboniferous coal measures minerals are widely distributed in the basin (Fig. 1a).

2.2 Station Setting and Sample Processing

In this study, two sampling campaigns were conducted in August and November 2021 in the Fenhe River. During each campaign, 27 stations were visited (Fig. 1) and river water samples at a depth of about 0.3 m were collected using a plexiglass vertical water sampler. Notably, river water ran dry at station F1 in August and therefore no water sample was collected at this station.

Water temperature (Tw) and dissolved oxygen saturation (DO%) were determined via a YSI oxygen analyzer (Pro20i, YSI Corporation, Yellow Spring, OH, USA), and the Winkler titration method was also applied for DO calibration (nominal precision: 0.1%) (Grasshoff et al., 1999). pH was measured repeatedly at least three times at 25℃ (pH_@25℃) in the laboratory using a pH Benchtop Meter (FE28, Mettler Toledo Corporation, Switzerland) with a precision of ± 0.01, and the NBS (pH = 4.01, 7.00, and 9.21 at 25℃) buffers were used for pH calibration. The total suspended solids (TSS) collected on 0.45 μm filter membranes with known weight was dried at 55℃, and the weight difference was the TSS concentration of the sample. DIC and its stable carbon isotope (δ¹³C_DIC) samples were filtered with 0.45 μm disposable syringe filters and stored at 4℃ in 20 mL high borosilicate glass bottles until laboratory analysis after adding saturated mercuric chloride solution (Li et al., 2017). The concentrations of DIC were measured by using a total organic carbon analyzer (Multi N/C 3100, Analttik Jena Corporation, Germany) and Certified Reference Materials (CRMs) from A. G. Dickson's laboratory were used for DIC quality control with an uncertainty of about 0.06 mg L⁻¹ (Li et al., 2022). δ¹³C_DIC values were measured using a Gasbench Ⅱ Extraction Line coupled with a Finnigan MAT 253 Mass Spectrometer (Thermo Electron Corporation, USA), with a standard deviation of 0.05‰ (n = 4) (Yang et al., 2018). Concentrations of major cations (Na⁺, Ca²⁺, Mg²⁺ and K⁺) and anions (Cl⁻, NO₃⁻ and SO₄²⁻) for the water samples were determined by using an Ion Chromatography (HIC-20Asuper, Shimadzu Corporation, Japan) with uncertainties of 5%. Also, in situ pH, the molar concentration of HCO₃⁻ and the partial pressure of CO₂ (pCO₂) were calculated from DIC, pH_@25℃ and in situ Tw based on the CO2SYS program (Version 2.3) (Pierrot and Wallace, 2006). The dissociation constants (K1 and K2) of carbonic acid determined by Millero (1979) were used.

3 Results 3.1 Characteristics of Water Chemistry

The concentrations of in situ pH, DO% and TSS in the Fenhe River showed large seasonal differences (Fig. 2). The river water in warm August (26.5 ± 4.1℃) showed a property of high pH (8.33 ± 0.31), high DO% (143.1% ± 34.9%) and low TSS (36.1 ± 27.7 mg L⁻¹), while the cold November (7.1 ± 2.0℃) showed low pH (8.21 ± 0.21), low DO% (103.7% ± 7.2%) and high TSS (93.7 ± 115.9 mg L⁻¹). Spatially, the river water in the Fenhe River was weakly alkaline, and in situ pH values in both August and November were higher than 7.90, except the station F11 (pH < 7.70) at the downstream of the Taiyuan City. Meanwhile, pH was often accompanied by high DO%, such as the pH values (about 8.30) in the upstream of the Fenhe River with a high DO% of about 110% in cold November and the higher pH (about 8.60) in the downstream with a higher DO% of about 160% in warm August. The TSS also had an obvious spatial variation. In November, TSS increased along the river generally and the highest TSS (at station F27) reached 393.3 mg L⁻¹. In comparison, the TSS in August had a relatively small fluctuation (< 100 mg L⁻¹).

The total dissolved cation charge (TZ⁺ = Na⁺ + K⁺ + 2Mg²⁺ + 2Ca²⁺) and total dissolved anion charge (TZ⁻ = Cl⁻ + NO₃⁻ + 2SO₄²⁻ + HCO₃⁻) of each sample were basically balanced, and most samples had normalized inorganic charge balances (NICBs) within ± 10%, indicating the small influence from other unknown ions. The cation equivalent concentration in the Fenhe River ranged between 3660 μEq L⁻¹ and 17270 μEq L⁻¹ with an average of 11370 μEq L⁻¹, which was largely higher than the average value (1125 μEq L⁻¹) of global rivers (Meybeck, 2003). The molar concentrations of major cation decreased in the order: Na⁺ > Ca²⁺ > Mg²⁺ > K⁺. In August, Na⁺ and [Ca²⁺ + Mg²⁺] accounted for 50.8% and 45.1% of total cations, respectively. In November, Na⁺ and [Ca²⁺ + Mg²⁺] accounted for 46.2% and 51.4% of total cations, respectively (Fig. 4a). Spatially, the major cationic concentrations showed an obvious increase at the station F11 which located in the downstream of the Taiyuan City, and especially in November the increased values of Ca²⁺ (> 2.85 mmol L⁻¹) and Mg²⁺ (> 1.50 mmol L⁻¹) sustained to the outlet (station F27) of the Fenhe River (Fig. 3a). Notably, the molar concentrations of major cation in headwater (stations F1-F2) decreased in the order: Ca²⁺ > Mg²⁺ > Na⁺ > K⁺, and the concentration of [Ca²⁺ + Mg²⁺] accounted for > 85% of total cations (Fig. 4a), indicating an obvious effect of carbonate weathering. The molar concentrations of major anion in August decreased in the order of Cl⁻ > HCO₃⁻ > SO₄²⁻ > NO₃⁻, with Cl⁻, HCO₃⁻ and SO₄²⁻ accounting for 37.9%, 31.1% and 28.4% of total anions, respectively (Fig. 4b). In comparison, the molar concentrations of major anion in November decreased in the order of HCO₃⁻ > Cl⁻ > SO₄²⁻ > NO₃⁻, with HCO₃⁻, Cl⁻ and SO₄²⁻ accounting for 43.7%, 27.7% and 25.4% of total anions, respectively. Spatially, like the cationic distribution, an obvious increase in the major anionic concentrations appeared at the station F11, and the increased values (> 3.9 mmol L⁻¹ in HCO₃⁻, > 2.9 mmol L⁻¹ in Cl⁻, > 2.7 mmol L⁻¹ in SO₄²⁻) in November extended to the outlet (station F27) of the Fenhe River (Fig. 3b). Additionally, the molar concentrations of major anion in headwater decreased in the order of HCO₃⁻ > SO₄²⁻ > Cl⁻ > NO₃⁻ (Fig. 4b), with HCO₃⁻ dominating and contributing > 77.4% of total anions in both August and November.

3.2 Riverine DIC, δ¹³C_DIC and pCO₂

In warm August, the concentrations of riverine DIC, δ¹³C_DIC and pCO₂ in the Fenhe River were 29.5 ± 12.1 mg L⁻¹, −5.73‰ ± 2.22‰ and 1167 ± 1547 μatm, respectively (Fig. 5a). Spatially, the carbonate values showed a relatively minor fluctuation in the upstream of the Fenhe River, and the average levels of DIC, δ¹³C_DIC and pCO₂ were 36.6 mg L⁻¹, −7.46‰ and 962 μatm, respectively. However, when the river entered the Taiyuan City between stations F8 and F11, the most acute carbonate gradient in the whole basin appeared. The DIC and pCO₂ inthe urban constructed wetland (stations F9 and F10) decreased to 25.5 mg L⁻¹ and 504 μatm with a more enriched δ¹³C_DIC (−4.03‰). In contrast, a sudden increase in DIC (> 57 mg L⁻¹) and pCO₂ (> 8000 μatm) appeared at the downstream of Taiyuan City (station F11) with a more depleted δ¹³C_DIC (< −10.5‰). In the later river reach, DIC and pCO₂ experienced a large decrease again, and the values at more than half of the sampling stations were lower than the levels in the upstream of the Fenhe River while the δ¹³C_DIC was more positive with the highest value of −1.43‰. In comparison, in cold November, the DIC, δ¹³C_DIC and pCO₂ in the upstream of the Fenhe River showed a similar variation and level with those in August (Fig. 5b). However, after the river flowed out the station F11, high DIC (> 47 mg L⁻¹) and depleted δ¹³C_DIC (< −7.90‰) were extended to the outlet (station F27) of the Fenhe River.

4 Discussion 4.1 DIC Originated from Carbonate Weathering Involved by H₂CO₃ and H₂SO₄

The effects of carbonate weathering on riverine DIC and its spatial variation have been reported in other watersheds globally (Martin, 2017; Rosentreter and Eyre, 2020; Shan et al., 2021; Song et al., 2021), since weathering of carbonate can be 12 times faster than that of granite (Meybeck, 1987). In the Fenhe Basin with a minor silicate rocks cover (Fig. 1a), the carbonate rocks area mainly composed of limestone and calcite minerals in the upper reaches with a karst land type (Yang et al., 2017), and carbonate minerals in the mid-downstream with a broad loess cover (Meng et al., 2018). At the headstream of the Fenhe River, the average DIC concentration reached 37 mg L⁻¹ (Fig. 5), indicating that carbonate weathering would be an important source for the high DIC concentration.

In carbonate weathering, the produced DIC mainly comes from carbonate itself and soil CO₂. In the Fenhe Basin, C3 plants and C4 plants are widely distributed, such as naked oats (C3) and sorghum (C4) in the upstream, and wheat (C3) and corn (C4) in middle and lower reaches (Shanxi Provincial Bureau of Statistics, 2021). The average δ¹³C value of C3 and C4 plants were −27.0‰ and −12.5‰, respectively (Cerling et al., 1991; Clark and Fritz, 1997). Thus, the δ¹³C value of soil CO₂ dissolved in water should be −17‰ for C3 plants and be more positive from coupled C3 and C4 plants after considering isotopic fractionation (Deuser and Degens, 1967; Telmer and Veizer, 1999; Zhong et al., 2018). In comparison, marine carbonate δ¹³C has more enriched value with an average of about 0 (Clark and Fritz, 1997).

Generally, carbonate rocks are predominantly dissolved by H₂CO₃ and produce Ca²⁺ and Mg²⁺ in carbonate weathering. In this case, half of total DIC is derived from carbonate rocks and the other half is from soil CO₂ (C3 plants), producing an intermediate δ¹³C value of −8.5‰ and a 1:2 molar ratio of [Ca²⁺ + Mg²⁺]/[HCO₃⁻] according to Eq. (1), without considering the effect of CO₂ degassing. However, when carbonate rocks are weathered by H₂SO₄, total DIC is 100% derived from minerals, producing an enriched δ¹³C value of 0‰ and a 1:1 molar ratio of [Ca²⁺+Mg²⁺]/[HCO₃⁻] according to Eq. (2).

$ {\rm{Ca}}_{x}{\rm{Mg}}_{1−x}{\rm{CO}}_{3} + {\rm{CO}}_{2} + {\rm{H}}_{2}{\rm{O}} ↔ x{\rm{Ca}}^{2+} + (1 − x){\rm{Mg}}^{2+} + {\rm{2HCO}}_{3}^{−}, $

(1)

$ {\rm{2Ca}}_{x}{\rm{Mg}}_{1−x}{\rm{CO}}_{3} + {\rm{H}}_{2}{\rm{SO}}_{4} ↔ 2x{\rm{Ca}}^{2+} + 2(1 − x){\rm{Mg}}^{2+} + {\rm{SO}}_{4}^{2−} + {\rm{2HCO}}_{3}^{−}. $

(2)

In the upstream of the Fenhe River, the molar ratio of [Ca²⁺+Mg²⁺]/[HCO₃⁻] was between 0.5 and 1 in both August and November (Fig. 6a), and the average δ¹³C_DIC value (−7.46‰ in August and −7.12‰ in November) was also between −8.5‰ and 0 (Fig. 6b). This is indicated that the riverine DIC in the upstream mainly originated from the weathering of carbonate rocks involved by both H₂CO₃ and H₂SO₄. This situation may be due to the extensive exploitation of coal resources which is rich in SO₄²⁻ (Yang et al., 2017). In fact, many coal mining sites were around the Fenhe River, and the annual volume of coal mining in 2020 in Shanxi Province where the Fenhe River is located had reached 1.08 × 10⁹ tons, accounting for 28% of the total coal output in China (National Bureau of Statistics of China, 2022). Given the favorable natural background at the upstream, the SO₄²⁻ input from coal mining may be through various ways such as dry settlement, runoff input and groundwater input, and thus related further researches are needed in the future. In contrast, in the mid-downstream of the Fenhe River, the molar ratio of [Ca²⁺ + Mg²⁺]/[HCO₃⁻] showed a value of about 1 in November and even > 1 in August (Fig. 6a). Moreover, [Ca²⁺ + Mg²⁺] concentrations in mid-down- stream were obviously higher than those in the upstream generally in both August and November, indicating that some [Ca²⁺ + Mg²⁺] could also come from dissolution of evaporite rocks (such as gypsum) in addition to carbonate weathering in the mid-downstream. The widely distributed loess rich in sulfate evaporite (CaSO₄ and MgSO₄) in the mid-downstream of the Fenhe River (Fig. 1a) seemed to confirm this possible source. However, we found that Ca²⁺ and Mg²⁺ concentrations in the mid-downstream increased in a sudden and sharp form from station F11 which was in the downstream of Taiyuan City and received large amounts of sewage input (Fig. 3). According to the high ions concentrations of collected sewage near station F11 (Table 1), anthropogenic direct input may be an important source of high [Ca²⁺ + Mg²⁺] in the middownstream of the Fenhe River (Meng et al., 2017; Yang et al., 2017; Wang et al., 2022). For the variation of [HCO₃⁻] in the mid-downstream, [HCO₃⁻] in November were higher than that in the upstream and the opposite situation occurred in August (Fig. 3b). Obviously, in addition to the carbonate weathering, [HCO₃⁻] could come from other sources in November, while it may be consumed by some processes (such as CO₂ consumption from aquatic production) in August. As shown in Fig. 6b, more depleted δ¹³C_DIC (mean = −8.27‰) appeared in November and then more enriched δ¹³C_DIC (mean = −5.09‰) were in August in the mid-downstream of the Fenhe River, which also proved the seasonal difference of the factors controlling DIC.

4.2 Importance of Aquatic Photosynthesis

Biological production (photosynthesis) will decrease DIC and release oxygen through the consumption of CO₂. Conversely, respiration will increase DIC and consume oxygen. In general, the non-free flowing water body is easy to be autotrophic, due to stable water conditions and the accumulation of nutrients; the free-flowing water body tends to be a slightly heterotrophic or biological balanced status (Halbedel and Koschorreck, 2013; Maavara et al., 2017; Ran et al., 2017; Jiang et al., 2021). Under semi-arid climate, the Fenhe River Basin supplied 37.3% (2.71 × 10⁹ m³) of total water consumption in Shanxi Province, and water storage projects that hindered water flow existed widely (Shanxi Provincial Department of Water Resources, 2021). First, in the upstream of the Fenhe River as the important drinking source, two large reservoirs (61 m-high Fenhe Reservoir and 88 m-high Fenhe Reservoir 2) were located between the stations F5 (Goukou Country) and F8 (Lancun) (Fig. 1a). Based on the runoff data at hydrological station from the Yellow River Conservancy Commission of China (http://xxfb.mwr.cn/index.html), the upstream of the Fenhe River had a small runoff, especially the areas at the downstream of the reservoirs with a runoff of < 4 m³ s⁻¹ (Fig. 7). However, due to strict protection of water resources and less nutrient input, the DO at most stations in the upstream was slightly oversaturated with a weak production, in addition to the areas affected by reservoirs with the highest DO% (135.7% in warm August, Fig. 2c). Thus, DIC in the upstream of the Fenhe River showed a relatively minor spatial fluctuation (Fig. 5), and it basically reflected a natural result caused by carbonate weathering.

Second, in the mid-downstream of the Fenhe River which was the core economic area with a large population, though no large reservoirs existed, many small weirs and dams were implemented for agricultural water withdrawals or urban water use, such as eleven rubber dams with a height of about 3.5 m in the urban constructed wetland in Taiyuan City (between stations F8 and F11). Meanwhile, the runoff in August was less than one twentieth of that in November in the middle and lower reaches, and even became 0 m³ s⁻¹ near the outlet (F26, Hejin) (Fig. 7), indicating that the river kept a more static status in summer but had a relatively good fluidity in autumn. This situation was also reflected in the TSS parameters in the middownstream. The low TSS in the low-runoff August was due to the gravity sedimentation of particles, while high TSS in the high-runoff November was owing to the intense erosion of loess (Fig. 2d). Thus, large numbers of lentic areas were observed in the mid-downstream in warm August, and DO% in 90% of sampling stations was more than 120% with the highest value of 270% (station F22), indicating a strong aquatic photosynthesis (Fig. 2c). Moreover, CO₂ consumption from the strong production in August caused the riverine DIC in mid-downstream to be obviously lower than that in the upstream of the Fenhe River, which was also the main factor for the obvious positive deviation of [Ca²⁺ + Mg²⁺]/[HCO₃⁻] to the 1:1 ratio line (Fig. 6a). In comparison, at the mid-down- stream of the Fenhe River in cold November, the stations F9 and F10 with the highest DO% (125.2%) in the urban constructed wetland in Taiyuan City, a riverside park with a width of about 500 m and a length of about 30 km, sustained the lowest DIC (22.4 mg L⁻¹) while DIC in most other stations with a slightly oversaturated DO showed a small fluctuation (about 52 mg L⁻¹) (Figs. 2c and 5b). During photosynthesis, the light ¹²CO₂ would be preferentially absorbed and heavy δ¹³C_DIC would be produced, thus δ¹³C_DIC in water would be enriched and more positive (Parker et al., 2005). As shown in Fig. 8a, δ¹³C_DIC had a significant negative correlation (p < 0.01) with DIC in the Fenhe River, and heavy δ¹³C_DIC values accompanied with low DIC concentrations, which appeared in the high DO% conditions. Overall, aquatic photosynthesis played a significant role in the spatial variation in DIC, especially in warm August.

Note that the large consumption of CO₂ and increase in pH caused by strong aquatic photosynthesis usually decrease the calcite solubility, thus calcite precipitation could easily occur (Hammes and Verstraete, 2002; Rogerson et al., 2008; Liu et al., 2018). In mid-down- stream of the Fenhe River dominated by strong photosynthesis during warm and the low-runoff August, pH reached 8.33 ± 0.35. Calcite precipitation also would be importance to the obvious decrease in DIC, which was demonstrated by the well positive correlation between Ca²⁺ and DIC (Fig. 9). Similar situations also showed in Ichetucknee River in USA, Wujiang and Pearl River in China (de Montety et al., 2011; Yang et al., 2016; Wang et al., 2019). Meanwhile, we found that runoff variation in the Fenhe River showed an abnormal situation since rivers in China usually have a high river discharge in summer due to the effect of the southeast monsoon. In fact, the Fenhe River itself had a relatively small annual average runoff (about 1.7 × 10⁹ m³), less than 4% of the main stream of the Yellow River (about 4.4 × 10¹⁰ m³) in 2021 (Yellow River Conservancy Commission of China, 2021). Therefore, the large amounts of water withdrawals would further intensify the reduction of runoff, especially the mid-downstream flowing through the provincial capital (Taiyuan City) (Fig. 1c). In 2021, the total amount of water resources in the downstream of the Fenhe River is 1.61×10⁹ m³, of which 57% are taken by human activities (Shanxi Provincial Department of Water Resources, 2021). Meanwhile, many dams or weirs are in a drainage state to avoid flood events in summer when the 70% of the annual precipitation is concentrated. However, there was no obvious rainfall before and after our investigations. In contrast, water supplement works (Project of leading Yellow River into Fenhe River) have been operating in the upstream of the Fenhe River to avoid flow interruption during autumn and winter with less rainfall. Compared with the rivers in humid areas, dam construction and excessive water withdrawals caused more nonfree flowing rivers in arid or semi-arid areas which cover 45% of the Earth's land surface (Poulter et al., 2014; Grill et al., 2019). Recently, Ran et al (2021) found that riverine CO₂ efflux in Northwest China with less rains decreased by 56% in last 30 years due to damming and water withdrawals. More research is needed to identify whether the riverine inorganic carbon system in arid areas is particularly vulnerable to rapid urbanization and dam construction in subsequent years, even in rainy summer. Correspondingly, the impact extent of urbanization and damming on river carbonate systems is urgently explored through more advanced methods and in-depth investigations.

4.3 Impact of Human Direct Input

Globally, the impact of anthropogenic direct input on riverine DIC has received increasing attention (Raymond et al., 2008; Yang et al., 2018; Gao et al., 2021), especially sewage input from urban areas (Yoon et al., 2017; Park et al., 2018, 2021; Chanda et al., 2020). Taiyuan City between station F8 and F11 is the capital of Shanxi Province with a 5.3 million population and a 62 billion $ GDP in 2021. Obviously, the effects of urban sewage input cannot be ignored. Due to degradation of organic matter, whether the wastewater is treated or not by the wastewater treatment plants (WWTPs), it usually has properties of high DIC and depleted δ¹³C. Wachniew et al. (2006) found that the massive decomposition of organic matter in untreated sewage would cause δ¹³C_DIC to be decreased by about 4‰. Recently, Liu et al. (2019) found that the depleted δ¹³C_DIC values in effluents from WWTPs in Qingdao City (China) ranged from − 13.4‰ to − 10.2‰. The average DIC in the treated sewage collected near station F11 during our investigations in August and November 2021 was up to 55.8 mg L⁻¹ with a depleted δ¹³C of − 9.86‰ (Fig. 6b), indicating that sewage input would obviously increase the riverine DIC.

Note that the sewage input is a continuous process and it may be a major factor causing high concentration of DIC in the mid-downstream which flowed through many cities. Especially in the November with weak biological activities, the increased DIC levels caused by urban sewage input even extended to the outlet of the Fenhe River (Fig. 5). In 2018, the amount of sewage received in the Fenhe River Basin reached 0.35 billion tons, accounting for 44.5% of the total sewage discharge in Shanxi Province (Shanxi Provincial Department of Water Resources, 2018). And the sewage is mainly discharged into the middle and lower reaches of the Fenhe River especially in some urban areas near stations F11, F15, F21, and F24 – F26. To evaluate the contribution of sewage input to riverine DIC, the δ¹³C_DIC values at the station F11 could approximately reflect the mixing signal between the river water of the urban wetland in the upstream and sewage (Fig. 6b). This was because that the existence of cement or stone berms on both river banks would limit input of other carbonate from the basin. Thus, we can estimate the individual contribution of mixed sources through the endmember mixing model (Raymond et al., 2004; Wang et al., 2013):

$ {\rm{\delta }}^{13}{\rm{C}}_{{\rm{DIC-Mix}}} = {\rm{α}}_{{\rm{wet}}} {\rm{\delta }}^{13}{\rm{C}}_{{\rm{DIC-wet}}} + {\rm{α}}_{{\rm{sew}}} {\rm{\delta }}^{13}{\rm{C}}_{{\rm{DIC-sew}}}, $

(3)

$ {\rm{α_{wet} + α_{sew} = 1,}} $

(4)

where α_wet and α_sew denote the contribution ratio of the river water of the urban wetland in the upstream and sewage to riverine δ¹³C_DIC. Here, the δ¹³C_DIC value in the urban park reach (− 3.5‰ in August and − 5.0‰ in November) could be the endmember value in the river water flowing through urban wetland. The endmember value of sewage should be the mixing value of untreated sewage and treated sewage due to the value of about −11.0‰ at the station F11. We then selected the average δ¹³C_DIC value (− 12.18‰) of treated sewage (− 9.86‰, Table 1) in this study and untreated sewage (− 14.5‰) from Xi'an City which is also located in the eastern area of the Loess Plateau (Guo et al., 2013). Thus, the evaluated contribution of sewage input to DIC at station F11 accounted for 84% in August and 70% in November, showing an important contribution to riverine DIC. Fortunately, the reduction of high DIC concentration in urban sewage has been gradually concerned to achieve the negative carbon emission, such as adding alkaline materials (NaOH or Ca(OH)₂) to the effluent of wastewater treatment plants (Cai and Jiao, 2022), and surely further relevant technical research is also necessary. Additionally, large amounts of organic matters brought by sewage also increased the riverine DIC levels due to enhanced aquatic respiration. As shown in Fig. 2c, the station F11 in both August and November showed an unsaturated DO (< 86%).

As discussed above, carbonate weathering is the underlying factor of the high DIC in the Fenhe River initially. However, aquatic photosynthesis caused by widespread damming played a dominant role in the spatial distribution of DIC, especially in the warm and low-runoff August. Meanwhile, sewage input showed an obvious increase effect on riverine DIC in the urban downstream.

4.4 Controls of Surface pCO₂

When pCO₂ in river is higher than that of the ambient atmosphere, dissolved CO₂ will diffuse out of the water and the river becomes a source of atmospheric CO₂ (Ran et al., 2017). On average, pCO₂ value in the Fenhe river was 1167 μatm in August and 1439 μatm in November, which was obviously lower than that in the Yellow River (2810 μatm, Ran et al., 2015) and global average pCO₂ of rivers (2400 μatm, Lauerwald et al., 2015), and the river was overall an atmospheric CO₂ source (Fig. 5). It is worth noting that pCO₂ in the mid-downstream in August was obviously lower than that in November, and the river even became an atmospheric CO₂ sink at some areas.

The factors affecting DIC will also appear on pCO₂ which is directly determined by the component of CO₂* (dissolved CO₂ and H₂CO₃). At the headwater, carbonate weathering resulted in a high DIC concentration (about 37 mg L⁻¹) initially (Fig. 5). Thus, when other processes had minor effects, the river would show higher pCO₂ level than the ambient atmosphere, like other headwater and upstream of rivers with a broad carbonate rocks cover (Wang et al., 2020; Zhang et al., 2019). Meanwhile, due to the closure of large reservoirs in the upstream and dams or weirs for irrigation in the mid-downstream of the Fenhe River, aquatic photosynthesis was dominant and oversaturated DO appeared in most of sampling stations (Fig. 2c), which would reduce pCO₂. In the upstream with a weak production demonstrated by the slightly oversaturated DO (about 104% in August and about 116% in November), pCO₂ showed a relatively small seasonal variation (Figs. 2c and 5). In the middle and lower streams, sewage input and aquatic respiration caused a sudden increase in pCO₂ at the station F11, while aquatic photosynthesis decreased the riverine CO₂ levels obviously after the station F11. Especially in the warm and low-runoff August, strong aquatic photosynthesis demonstrated by the oversaturated DO (about 154.5%) caused the riverine pCO₂ at most sampling stations to be lower than that in the upstream, and CO₂ sink appeared at stations F16 − F18 and F21 − F22 (Fig. 5a). In comparison, the pCO₂ in cold November showed a relatively minor fluctuation with an average of 1801 ± 307 μatm due to the slightly oversaturated DO (about 104%). Overall, ln pCO₂ had a significant negative relationship with δ¹³C_DIC (p < 0.01), and heavy δ¹³C_DIC values accompanied with low ln pCO₂ concentrations in the high-DO% conditions (Fig. 8b). This indicated that biological effect dominated by aquatic photosynthesis played a major influence on the spatial variation of riverine pCO₂, though carbonate weathering resulted in the initial pCO₂ level in the Fenhe River. In fact, the river surface area in dry Huang-HuaiHai region in China including this study area had shrunk by 17%, while the reservoir water surface had expanded by one-third between the 1980s and 2010s, which caused an about 77% decrease in CO₂ efflux (Ran et al., 2021). Meanwhile, massive constructed wetland through damming was implemented to optimize water quality and alleviate anthropogenic pollution in urban areas and their downstream, which further exacerbated the appearance of non-free flowing river with an eutrophic risk (Grill et al., 2019; Chen et al., 2021; Bao et al., 2022). This probably means the decrease in riverine CO₂ caused by damming and constructed wetland in arid and semi-arid areas will be a growing low-carbon hotspot on the way of achieving carbon peak and further net-zero carbon emissions (i.e., carbon neutrality) followed by the Pairs Agreement and COP26 in Glasgow in 2021.

5 Conclusions

This study revealed the significant seasonality and spatiality of riverine carbonate system and their controls in an urbanized river (the Fenhe River) on the Loess Plateau. Under semi-arid climate, riverine CO₂ system variability highlighted obvious effects of anthropogenic perturbations. In the upstream, DIC were mainly originated from carbonate weathering involved by H₂CO₃ and H₂SO₄, with a high DIC level (about 37 mg L⁻¹) and an oversaturated pCO₂ (about 800 μatm) initially. Meanwhile, watershed damming for water withdrawals and constructed wetland caused high aquatic photosynthesis, leading to enriched δ¹³C_DIC and an obvious decrease in pCO₂. In warm August, some river segments in the mid-down- stream even acted as an atmospheric CO₂ sink. Notably, wastewater input resulted in a sudden increase in DIC (> 55 mg L⁻¹) and pCO₂ (> 4500 μatm) in the downstream of the Taiyuan City, and in cold November the increased DIC even extended to the outlet of the Fenhe River. Obviously, compared with the rivers in humid areas, the rivers in arid or semi-arid areas would show more sensitiveness in riverine CO₂ system variations as well as air-water CO₂ fluxes estimates, especially the urban rivers highly affected by damming, constructed wetland and sewage input. Therefore, high spatiotemporal observations were recommended to clarify the anthropogenic stressors on riverine inorganic carbon system and achieve more CO₂ absorption caused by aquatic photosynthesis in arid and semi-arid areas covering 45% of global land.

Acknowledgements

This work was jointly supported by the National Natural Science Foundation of China (NSFC) (No. 41376123), the Youth Project of Shanxi Basic Research (Nos. 20210302124317, 201901D211383), the Research and Promotion Project of Water Conservancy Science and Technology in Shanxi Province (No. 2023GM41), and the Science and Technology Innovation Fund of Shanxi Agricultural University (No. 2018YJ21). We thank Dr. Meifang Zhang for the sampling and measuring work. Also, we gratefully acknowledge financial support for study abroad and international cooperation from Shanxi Agricultural University.