Spatial and temporal distribution characteristics of pCO2 and CO2 evasion in karst rivers under the influence of urbanization

Surface rivers play an essential role in carbon cycle processes in karst regions. However, the CO2 diffusion flux from karst rivers under the influence of urbanization has been scarcely examined in the literature. Along these lines, in this work, the CO2 partial pressure (pCO2) and its degassing in a typical karst river (Nanming River and its tributaries) were thoroughly investigated, which are significantly affected by urbanization in Southwest China. From the acquired results, it was demonstrated that the average values of pCO2 in the main stream of the Nanming River in the wet season, dry season, and flat season were 1975.77 ± 714.45 μatm, 1116.08 ± 454.24 μatm, and 976.89 ± 746.37 μatm, respectively. On the other hand, the tributary showed mean pCO2 values of 1770.46 ± 1120.79 μatm, 1638.13 ± 1121.82 μatm, and 1107.74 ± 824.03 μatm in the three different hydrographic periods. Overall, the pCO2 of the Nanming River basin decreased in the following order: wet season > dry season > flat season, while the mainstream of the Nanming River was slightly higher than that of the tributaries in the wet season. However, it was lower than that of the tributaries in the dry and flat seasons. Additionally, more than 90% of the samples displayed a supersaturated state of CO2, which acted as an important source of CO2 in the atmosphere. From a spatial point of view, pCO2 tended to be higher in the western region than in the eastern region, higher in the middle than in the immediate vicinity, and higher in the south during the three seasons. The higher urban areas showed also relatively higher pCO2 than the lower urban areas. The urban land along with the Nanming River’s mainstream exhibited a weaker correlation with pCO2 than the urban land along with the main tributaries due to the mainstream’s regular management in recent years. Moreover, the pCO2 was mainly influenced by the carbonate rocks dissolution, aquatic organism metabolic processes, and human activities. In the Nanming River basin, the average CO2 diffusion fluxes in the wet season, dry season, and flat season were 147.02 ± 100.3 mmol·m−2·d−1, 76.02 ± 67.45 mmol·m−2·d−1, and 119.28 ± 168.22 mmol·m−2·d−1, respectively, which indicates high potential CO2 emissions. In addition, it was found that urban construction could increase the pCO2 of karst rivers and cause an increase in the CO2 release flux during the regional urban expansion. In view of the increasing intensive and extensive urbanization in karst regions, our findings are helpful to elucidate the characteristics of carbon dioxide emissions from karst rivers under the disturbance of human activities and further deepen the understanding of the carbon balance in karst river basins.

by terrestrial ecosystems (2.6 Pg C/year) and the uptake of carbon by oceans (2.3 Pg C/year) . Thus, rivers are regarded as one of the important sources of atmospheric CO 2 and play an essential role in the regional and global carbon cycle. As a result, an accurate assessment of river CO 2 emission intensity can provide valuable insights for understanding the process of the global carbon cycle. However, urbanization has steadily increased in recent years, changing the ecosystem structure and function of rivers, as well as the physical and chemical mechanisms that regulate inorganic carbon release and storage in rivers (Cole et al. 2001;Wang et al. 2017a, b;Wu et al. 2015), thus affecting the carbon emission intensity of the rivers. According to the literature, the CO 2 emission intensity of rivers that is affected by urbanization is significantly higher than that of rivers under the natural background Wang et al. 2021). Consequently, with the intensification of global urban construction, urban rivers may be a potential heat source for CO 2 emissions, affecting the current estimates of riverine greenhouse gas emission fluxes from global rivers.
River partial pressure of CO 2 (pCO 2 ) is considered one of the most important factors controlling the release of CO 2 to the atmosphere from the surface water. When the pCO 2 in the river is higher than the atmospheric pCO 2 , the river releases CO 2 into the atmosphere. At this time, the river is a "carbon source," and in contrast, it is a "carbon sink." Raymond et al. (2013) noted that up to 95% of rivers are in a supersaturated pCO 2 state, with an average value of 3100 μatm, which is characterized by carbon sources. The production, transport, and emission of CO 2 in rivers are the result of the combined actions from internal and external environments, involving a variety of physical, chemical, and biogeochemical processes (Alin et al. 2011). Previously reported works in the literature have shown that the CO 2 in rivers under natural background mainly comes from the degradation of the internally dissolved organic carbon and the input of soil CO 2 (Hotchkiss et al. 2015). Meanwhile, river pCO 2 is also influenced by rock weathering, biogeochemical processes, and anthropogenic activities (Cartwright 2010;Liu and Han. 2021). On top of that, it has been proved that domestic and industrial wastewater in combination with soil erosion due to human activities often increase organic matter and nutrient (TN, TP) inputs, thereby significantly increasing pCO 2 (Li and Bush 2015) . This disturbed watershed environment plays a vital role in driving the heterogeneity of CO 2 emissions (Yoon et al. 2017). Thus, the various control factors of pCO 2 vary widely among the different river systems. Therefore, a deep understanding of the factors affecting pCO 2 in typical rivers and the CO 2 fluxes at the water-air interface are important for elucidating the regional and global carbon balance of inland waters.
The main lithology in karst areas that consists of carbonate rocks and carbonate dissolution ((Ca 1-x Mg x ) CO 3 + CO 2 + H 2 O → (1-x)Ca 2+ + xMg 2+ + 2HCO 3 − ) can convert soil or atmospheric CO 2 into HCO 3 − stored in karst river water. Therefore, rivers are thought to play a significant role in the karst region's carbon cycle. Meanwhile, these processes cause karst river water to have a high concentration of dissolved inorganic carbon (DIC), which also leads to a relatively higher intensity of CO 2 gas exchange with the atmosphere (Liu et al. 2007). Hence, the accurate evaluation of the CO 2 exchange flux at the water-air interface of karst rivers becomes the key to accurately evaluating the carbon sink effect of karstification. The previously reported works on pCO 2 and its diffusive flux in karst rivers have mainly focused on the natural environmental background (Abril et al. 2014;Johnson et al. 2008). However, karstification creates well-developed funnels and sinkholes, which makes it easy for surface pollutants to enter karst rivers, resulting thus in an increase in the flux of CO 2 emissions from rivers (Wang et al. 2021;Zeng et al. 2019). Nevertheless, the response of karst river pCO 2 to urbanization is still limited, while the temporal and spatial patterns of karst riverine CO 2 emissions and the relevant driving factors in urbanized areas are still unclear.
Southwest China is the largest karst contiguous distribution area in the world. In recent years, with the acceleration of urbanization, the surface water environment in this area has been affected to varying degrees (Shi et al. 2022). Guizhou is located in the center of karst in Southwest China, and the urbanization rate of its provincial capital, Guiyang, has increased from 68.13% in 2010 to 80.07% in 2020. The urbanization rate is the highest in the province, and the area of construction land has increased by more than 15,000 hm 2 in the past 10 years. The Nanming River is the largest river in Guiyang, running through the whole city, and its mainstream flows through the highly urbanized area, and most of its tributaries flow through the suburbs. The mainstream and tributaries create a distinctive contrast in urban gradients. Therefore, the Nanming River and its tributaries were taken as the research object in this work. More specifically, the river water samples were collected during the three hydrological periods (wet season, dry season, and flat season) to reveal the spatiotemporal distribution characteristics and the factors affecting pCO 2 . Thus, the source-sink effect of CO 2 in the river was clarified, and its diffusion flux was estimated. Our work provides valuable insights for studying the carbon cycling in karst regions, with a view to providing scientific data support for the accurate estimation of carbon release from karst rivers.

Study site
The Nanming River in Guiyang city has a total length of approximately 42 km and a river width of 30-100 m, with a drainage area of approximately 1164.59 km 2 . The main tributaries are the Xiaohuang River, Mati River, Xiaoche River, Shixi River, Songxi River, Toubao River, Yudong River, Maodong River, and Dazhai River. The river width varies from 20 to 40 m, and the Toubao River is the largest tributary, accounting for 35.8% of the watershed area. The most significant characteristics of the subtropical monsoon humid climate in the study area are the average annual temperature value of 15.3 °C, the 1147 h of sunlight, and the average annual rainfall of 1130 mm. The three seasons are flat (March to May), wet (June to October), and dry (November to February of the following year), and the average precipitation of each hydrological period was 271.2 mm, 694.4 mm, and 107.2 mm, respectively. Since most of the precipitation in the basin falls during the rainy season, the river flows are greatly increased during this period, while during the flat and dry seasons' rainfall decreases and some sections of the river even dry up. Additionally, the formation lithology of the study area is mainly composed of limestone of the Carboniferous and Devonian systems and sand shale of the Triassic system. Moreover, due to the suitable temperature, precipitation, and other climatic conditions, the Nanming River basin is rich in aquatic plants, with a total of 31 species, mainly emerging plants, submerged plants, and algae, as well as common plants including Potamogeton crispus, Potamogeton malainus, Potamogeton pectinatus, Ceratophyllum demersum, Otteliaacuminate, and Vallisneria asiatica (Qin et al. 2007).

Sampling procedures and analysis
River samples were collected from 52 sites (20 in the mainstream and 32 in the tributaries) in July 2021 (wet season), November 2021 (dry season), and March 2022 (flat seasons), which basically covered the whole basin (Fig. 1b). Samplings were performed from 9:00 am to 18:00 pm, and the data were collected from 0.2 m below the river surface.
At each sampling site, a portable multiparameter water quality analyzer (Multi 3630, WTW, Germany) was used to measure water temperature (T), pH, dissolved oxygen (DO), and electrical conductivity (EC), while the test accuracy was ± 0.1 ℃, ± 0.01, ± 0.01 mg/L, and ± 1 μs/cm. The alkalinity (Alk) was estimated after titration with dilute hydrochloric acid (0.01 mol/L) within 24 h of sampling. The river water was filtered through a 0.45-μm aqueous filter and stored in centrifuge tubes. The samples used for the determination of cations need to be acidified by adding premium pure concentrated nitric acid to pH ≤ 2. Moreover, the samples used for the determination of DOC should be acidified by adding high-grade pure phosphoric acid to pH ≤ 2 and then stored in a brown bottle to keep away from light. The others were directly packaged after filtration for the determination of anions and nutrients. All the samples for analyses were kept at the temperature value of 4 °C until analysis.
Anions (F − , Cl − , NO 3 − , SO 4 2− ) and cations (K + , Ca 2+ , Na + , Mg 2+ ) were analyzed by performing ion chromatograph (ICS-1100, DIONEX, USA) and atomic absorption spectrometer (TAS-990, Pgeneral, China) measurements, respectively, whereas the detection limit of each ion was 0.01 mg/L. The measured error percentage of the charge between anions and cations was within ± 10%, which meets the requirements. The river nutrients (total nitrogen (TN), and total phosphorus (TP)) were measured by using a multiparameter water quality meter (GL-900, Glkrui, China) according to the method given by the National Environmental Protection Agency (2002), where 10% of parallel double samples were measured for each index, and the standard deviation was within 5%. Dissolved organic carbon (DOC) was analyzed with a total organic carbon analyzer (Vario TOC, Elmenter, Germany). Each water sample was measured three times, and the standard deviation was within 5%.

pCO 2 and water-air CO 2 flux calculations
The formula for calculating pCO 2 in water is as follows (Li et al. 2012): where pCO 2 is the partial pressure of dissolved CO 2 in rivers (μatm); K 1 and KCO 2 represent the equilibrium constants of H 2 CO 3 and CO 2 , respectively; and (HCO 3 − ) and (H + ) denote the molar concentrations of HCO 3 − and H + in the water (μmol/L), respectively.
Currently, the main methods for monitoring the CO 2 diffusion fluxes at the water-gas interface are the thin boundary layer model (TBL) and floating chamber (FC). However, there are differences in results between the methods due to theoretical principles, and there is no available approach in the literature yielding the best results (Zhang. 2018). The Nanming River basin is located in the karst mountainous area, with high topography and steep channel slopes. As a result, rapid flow and floating chambers are developed that are susceptible to hydrodynamic disturbances. Meanwhile, after the management of the urban rivers, which make the river water flow far from the shore, it is not easy to put floating boxes in the urban area. Consequently, the TBL method was used to calculate the CO 2 diffusion flux at the river-air interface, by employing the following formula (Cole and Caraco 1998;Parkhurst 1999): where F c is the water-air interface diffusion flux (mmol·m −2 ·d −1 ). The condition Fc > 0 means that the river (1) pCO 2 = HCO 3 − H + ∕K 1 K CO 2 (2) F c = k 600 × K H × pCO 2 water − p air releases CO 2 into the atmosphere; otherwise, it means that it absorbs CO 2 . pCO 2 refers to the partial pressure of CO 2 in the surface water (μatm); p air denotes the atmospheric CO 2 partial pressure (μatm); K H represents Henry's constant (mol/(L·atm)), which can be calculated by Formula (3) (Plummer and Busenberg 1982); and the gas transfer coefficient, k (cm/h), which mainly depends on the wind speed and local temperature distribution. According to the literature, the k 600 value is closely related to the flow velocity and the channel gradient of the river (Alin et al. 2011). In this work, Formula (4) of Cole and Wanninkhof was adopted to calculate (Wanninkhof 1992;Cole and Caraco 1998). The meteorological observation data in this work were obtained from the National Basic Meteorological Station in Guiyang (WMOID: 57,816), provided by the Meteorological Bureau of Guizhou Province. The average daily wind speeds in the wet season and dry season were 1.6 m/s and 1.5 m/s within 1-day sampling period, respectively, while the average daily wind speeds in the flat season were 2.9 m/s and 3.8 m/s during the 2-day sampling period, respectively. where T w refers tp the water temperature (K) where U 10 indicates the wind speed at 10 m above the water surface (m/s), S c denotes the Schmidt constant of CO 2 , depending on the temperature t (℃), which was calculated by Eq. (5) (Demarty et al. 2010). The exponent n depends on U 10 , when U 10 > 3 m/s, n = 0.5, and when U 10 < 3 m/s, n = 0.66.

Data analysis
Before performing statistical analysis, all data were tested for normality and homogeneity using Shapiro-Wilk and Levene's tests. Particularly, the pCO 2 was calculated with the hydrogeochemical computer software PHREEQC v.2 (Parkhurst 1999) using the PHREEQC database. The solution conditions in the input file included the pH value, temperature, and concentrations of major ions (K + , Ca 2+ , Na + , Mg 2+ , Cl − , SO 4 2− , NO 3 − ), and the output file contained the values of the CO 2 partial pressure. The histograms were generated using Origin Lab 2021 software package, whereas the IBM SPSS Statistics 20 software package was used to analyze the Pearson correlations. As far as the spatiotemporal distribution and sampling diagrams are concerned, they were drawn by using the Arc-GIS 10.2 software with Kriging.

Types of land use
In general, a higher proportion of construction land in the land use status map lead to a bigger degree of urbanization. As can be observed from Fig. 1, the urban construction in the Nanming River basin was mainly distributed in the mainstream and the major tributary of the Taobao River. Hence, it can be inferred that the urbanization areas were mainly concentrated in the middle section of the river, and the urbanization intensity of the mainstream areas was greater than that of the tributary areas. Grassland, forest, and cultivated land make up the majority of the other land use classifications in the studied area, accounting for approximately 80.26% (3) log 10 K H = 108.3865 + 0.01985076T w − 6919.53∕T w − 40.45154log 10 T w + 669365∕T w 2 (4) k 600 = (2.07 + 0.215 U 10 of the overall watershed. In addition, the mainstream of the Nanming River and the Toubao River showed a trend of first increasing urbanization and then decreasing along with the flow direction from upstream to downstream (Fig. 2).

Variation characteristics of the physical and chemical parameters of rivers
(1) Variation of water chemistry parameters The spatiotemporal pattern of the physical and chemical parameters in the study area are depicted in Fig. 3. The average water temperature values of the Nanming River and its tributaries in the wet, dry, and flat seasons were 23.13 ± 1.93 ℃, 14.18 ± 1.15 ℃, and 17.03 ± 2.02 ℃, respectively, with obvious seasonal changes (Table S1). On top of that, the average pH values in the basin were 8.01 ± 0.23, 8.19 ± 0.22, and 8.20 ± 0.29 in the wet, dry, and flat periods, respectively, and were overall alkaline. Spatially, the pH value was lower in the south and higher in the northwest during the wet season, the dry season was low in the middle and east, and high in the northeast region. During the flat season, it was also lower in the middle and east and higher in the southwest region. Meanwhile, the pH value of the three hydrological periods showed that the mainstream was larger than the tributaries. The DO concentrations in the wet, dry, and flat seasons were 7.72 ± 0.91 mg/L, 9.23 ± 0.77 mg/L, and 9.57 ± 2.30 mg/L, respectively. DO was lower in the wet season, which could be related to the decrease in the solubility of the oxygen with higher water temperature. Its spatial distribution tended to be higher in the southeast region and lower in the west during the wet season in this region, and the dry season was high around and low in the middle region. The trend was higher in the southwest region and lower in the northeast during the flat season in this region. On the whole, the DO was relatively low in the urbanized areas in the three seasons, indicating that the mainstream of the Nanming River and the Taobao River were disturbed to varying degrees in the urbanized areas.
The mean values of EC in the wet, dry, and flat seasons were 541.26 ± 66.97 μs/cm, 579.87 ± 83.83 μs/cm, and 598.44 ± 89.55 μs/cm, respectively, with no significant seasonal differences. Spatially, the urbanized region showed high values in the three seasons. Generally, the changes in EC are mainly affected by the types and concentrations of anions and cations in water. More specifically, the average values of HCO 3 − and Ca 2+ with the highest concentrations of anions and cations in the basin were 213.69 ± 20.23 mg/L and 97.16 ± 13.32 mg/L, respectively. During the wet season, the average values in the dry season were 225.06 ± 22.80 mg/L and 105.31 ± 18.62 mg/L, respectively, while the average values of the flat season were 147.52 ± 39.33 mg/L and 75.53 ± 23.32 mg/L, respectively. However, the spatial distribution of those ions varied in each hydrological period and was not similar to the EC changes.
(2) Changes in TN, TP, and DOC concentrations The concentrations of nutrients (TN, TP) and DOC in rivers are mainly related to exogenous inputs and endogenous releases, which are important indicators for evaluating the degree of river pollution. The TN concentration in the watershed reached the minimum value (3.24 ± 2.07 mg/L) in the flat season, which was 2 times lower than in the wet season (6.75 ± 3.18 mg/L), and the dry season (6.43 ± 3.65 mg/L). However, the TP concentration reached the maximum value in the dry season (1.43 ± 1.65 mg/L), which was 2 times higher than in the wet season (0.68 ± 0.56 mg/L) and 7 times higher than in the flat season (0.28 ± 0.46 mg/L). Overall, the TN and TP contents were relatively low in the flat season. By comparing these concentrations with the Chinese National Water Quality Standards (China 2002), it can be concluded that the TN concentration of 64% of the points and the TP concentration of 20% of the points exceeded the Class V limit (most polluted water) during the flat season, while more than 90% of the TN and TP concentrations exceeded the V limit in the wet and dry seasons.
The average concentration of DOC in the flat season (9.10 ± 2.15 mg/L) was twice that of the wet season (5.39 ± 2.15 mg/L) and the dry season (4.61 ± 2.29 mg/L). Spatially, TN and TP generally showed a trend of being high in the west and low in the east region in the three seasons, among which TN was the most obvious. Nevertheless, the changes in DOC were relatively complex in the three seasons (Fig. 4). From the above-mentioned results, it can be argued that there were significant seasonal differences in TN, TP, and DOC in the basin, while nitrogen, phosphorus, and organic carbon in the river could receive more from exogenous input. Meanwhile, the TN and TP in the river section flowing through the urban area were higher than those of the other river sections.

Temporal and spatial variation characteristics of pCO 2
The variation range of pCO 2 was 371.54 ~ 5754.4 μatm during the wet season, with an average value of 1852.58 ± 983.87 μatm. Additionally, its spatial distribution was generally higher in the west than in the east region. The variation range of the dry season was 281.84 ~ 6309.57 μatm, with an average value of 1437.34 ± 958.29 μatm, which was higher in the middle and near the southwest region. The variation range of the flat season was 147.91 ~ 3890.45 μatm, the average value was 1055.40 ± 796.46 μatm, and the spatial distribution characteristics were similar to those of the dry season. In general, the pCO 2 of the river in the study area was the Fig. 2 Land use ratio of the Nanming River and Taobao River. According to the previous research method (Gu et al. 2022;Shi et al. 2017;Zhang et al. 2019a, b), a circle with a diameter of 1 km is designated as a buffer zone upstream of the sampling point to extract the land use of the sampling point highest in the wet season, followed by the dry season, and the lowest in the flat season. The seasonal differences were obvious, and the standard deviation values were all large, indicating the existence of a significant spatial variation (Fig. 5). The average pCO 2 of the three hydrological periods was 2.5 ~ 4.5 times higher than that of the atmosphere (410 μatm), and the river as a whole presented the characteristics of a carbon source. However, the pCO 2 during the wet and flat seasons was lower than that of the atmospheric pCO 2 at a small number of sampled points in this work. In addition, most of the high-value areas of pCO 2 in the three seasons were shown in urban areas, while the low-value areas all appeared in grassland, and forestland. This result indicates that pCO 2 responds to the urbanization gradient between different rivers and in the upper, middle, and lower reaches of the same river (Fig. 5). Therefore, higher urbanization could lead to a higher pCO 2 .
The pCO 2 of the different rivers in the three seasons is displayed in Fig. 6. Except for the Shixi River, Songxi River, Maodong River, and Dazhai River, the pCO 2 values of the other rivers were the highest in the wet season, followed by the flat season and lowest in the dry season. Some of the sampling points along with the Xiaohuang River, Mati River, Xiaoche River, and Shixi River show carbon sink characteristics during the flat season, while parts of the sampling points along with the Songxi River show carbon sink characteristics during the wet season, whereas the rest of the rivers act as sources of atmospheric CO 2 . The Xiaohuang River, which is a tributary in the southern region of the basin, exhibited the highest pCO 2 value (average of 2839.83 ± 1492.62 μatm), which was twice the pCO 2 of the mainstream (average of 1356.25 ± 787.24 μatm). Due to the construction in this area during the sampling time, the Xiaohuang River was seriously polluted by turbidity and there were two sewage treatment plants in the upper reaches of the river (Fig. 1), resulting in a high pCO 2 . The lowest pCO 2 value in the basin was the Yudong River in the eastern region (average of 562.30 ± 60.11 μatm) since the river was located in the jurisdiction of the Yudongxia Reservoir and was less affected by human activities and urbanization. The average pCO 2 of the Taobao River, the largest tributary in the basin, was 1287.86 ± 692.07 μatm, which was slightly lower than the pCO 2 of the mainstream.

Seasonal variation of pCO 2 and its response to land use
Generally, biogeochemical processes, hydrogeological feature changes, and human activities are considered the main factors affecting the pCO 2 of rivers (Andrews et al. 2021;Le et al. 2018). The physiochemical parameters and the concentration of major ions are closely related to the photosynthesis and respiration processes of aquatic plants in rivers (Barth et al. 2003;Robinson et al. 1999), while the DIC from the weathering of carbonate rocks and infiltration of soil CO 2 also affect the distribution of CO 2 concentrations in karst rivers (Cartwright 2010;Polsenaere and Abril 2012). From the concentrations of the main ions and nutrients, obvious seasonal changes in the Nanming River basin can be observed. Thereby, the pCO 2 of the river would show significant temporal and spatial differences in the studied area. The pCO 2 values of the mainstream of the Nanming River occurred in the following order (Fig. 7a) (Fig. 7b). These research results are consistent with a previously reported work on the Red River (Le et al. 2018) and Jinshui River . This is mainly related to the following reasons. First, heavy precipitation in the wet season could increase the input of terrigenous organic matter and nutrients (Hafsi et al. 2016;Park et al. 2018;Teodoru et al. 2015). Moreover, the wet season in the studied area takes place mainly in summer. As a result, the increase in the temperature could significantly enhance the soil respiration adjacent to the river and increase the input of soil CO 2 in the terrestrial area. The increase in the terrigenous organic matter and nutrients will lead to the enhancement of the life activities of aquatic animals, plants, and microorganisms, and then the CO 2 emission will increase accordingly (Maranger et al. 2005). Second, the rivers in the study area originate from mountainous areas and the flow of rivers increases in the wet season, which leads to the frequent occurrence of falling water and ripples during the wet season, thus enhancing the turbulence of the river water. This effect will make the dissolved inorganic carbon unstable and converted to CO 2 . Finally, turbulence in the wet season is not conducive to the growth of aquatic photosynthetic organisms, resulting in lower primary productivity of rivers and less CO 2 consumption. Therefore, the dilution effect brought by rainwater is relatively weak in combination with the above-mentioned outcomes, while pCO 2 is higher in the wet season and pCO 2 is positively correlated with the flow of the river (Ran et al. 2017).
Additionally, the pCO 2 of the mainstream was higher than that of the tributaries during the wet season, which may be due to the large scale and flow of the mainstream (Hope et al. 2004), which increased the pCO 2 of the main stream. Simultaneously, the urbanization degree of the mainstream area was higher that of the tributary area, which is also one of the influencing factors (Tang et al. 2021). However, the   pCO 2 of the main stream was smaller than that of the tributaries in the dry season and the pCO 2 of the main stream was equivalent to that of tributaries in the flat season, which is mainly related to the fact that the Taobao River occupies a large area of land for river development and construction when the water flow is not turbulent. Therefore, the riparian ecosystem was destroyed in the dry and flat season, the input of organic matter and nutrients was increased, and the in situ respiration of microorganisms was enhanced, resulting in high pCO 2 (Maranger et al. 2005). Overall, the pCO 2 of the main stream and Taobao River showed a trend of first decreasing, then increasing, and finally decreasing along with flow direction (Fig. 7). The variation rule of the Taobao River was more significant than that of the main stream, which was basically consistent with the variation rule of the urbanization degree of the Toubao River in Fig. 2. However, although the upper reaches of the main stream were less urbanized than the middle reaches, the upper reaches of the main stream still exhibited higher pCO 2 values during the wet season. This may be related to the flood discharge of the reservoir during the wet season, where usually, the upper reaches of the river with large undulating terrain would cause falling water and then enhanced CO 2 outgassing (Wang et al. 2011). Based on the land use data obtained in Sect. 2.1 and the pCO 2 value of the two rivers, the correlation coefficients between the land use ratio of the Nanming River and the Toubao River and their corresponding pCO 2 values are presented in Table 1. As can be seen, there was no significant correlation between the pCO 2 value of the Nanming River and the various land use types, which could be related to the fact that the Nanming River had been subjected to regular artificial management in recent years. The water quality of the central urban section of the Nanming River improved significantly, and the coverage rate of the submerged plants in this region reached more than 70% (Wang et al. 2020). However, the pCO 2 value of the Toubao River had a positive correlation with urbanized land, which was significantly positively correlated in the dry season and the flat season. On the contrary, the correlation was weak in the wet season, indicating that the pCO 2 was relatively less affected by urbanization during this period. In conclusion, the increase in urban land was an important factor affecting the pCO 2 of rivers in the study area.
Previous works in the literature have shown that land use change could significantly affect river pCO 2 (Andrews et al. 2021). Thus, each sampling point was classified based on the land use changes and the proportion of urban land area that was more than 20% as high urbanization was defined. The opposite case was low urbanization, while the forest and grassland areas account for more than 50% of the high forest area, and the opposite is the low forest area. The results are illustrated in Fig. 8. The high urbanization area and the low forest river section had relatively high pCO 2 among the three seasons. According to the literature, urbanization could accelerate the direct input of terrestrial soil CO 2 to rivers (Romain et al. 2002). Meanwhile, construction land is more prone to nonpoint source pollution than other types of land use, which could lead to the respiration of rivers being enhanced. Thus, pCO 2 would be accordingly increased (Lambert et al. 2017). The forest grassland can not only absorb nutrients, organic matter, and carbon sequestration (Chen et al. 2018), but also intercept surface runoff, and even hinder the direct contact between terrestrial soil and rivers in the wet season. Thereby, the transport of nutrients and organic matter to rivers is reduced, which eventually leads to weakened heterotrophic respiration in rivers and reduced  pCO 2 (Zhang et al. 2019a, b). In addition, compared with the other land use types, forest soils have the lowest decomposition capacity for organic carbon (TOC), which could also reduce land-to-river CO 2 input (Borges et al. 2019). However, the lack of vegetation, such as forests in urban areas, will lead to large-scale exposure of rocks, which will also enhance chemical weathering of rocks and ultimately increase pCO 2 (Romain et al. 2002).

Relations between pCO 2 and environmental variables
pCO 2 in rivers is considered an important indicator of CO 2 emissions, which can be controlled by water temperature, CO 2 concentration, and Alk. Among them, the water temperature will affect the heat exchange process of the water-air interface, and also would affect the dissolution of carbonates, as well as the processes of biological photosynthesis, respiration, and mineralization (Wang et al. 2011;Ding et al. 2015;Barth et al. 2003). Nevertheless, under the influence of urbanization, these biogeochemical processes and the physical and chemical parameters of rivers can be easily changed, which in turn affect the fluctuation of pCO 2 . The correlation analysis results of pCO 2 and the main environmental factors in the Nanming River and its tributaries are presented in Table 2. A significant negative correlation between pCO 2 and pH in the three seasons of the basin was observed, which is in direct line with the previously reported results in the literature (Le et al. 2018;Gu et al. 2022;Li and Zhang 2013).
Since the pH is closely related to the dynamic equilibrium of carbonate in aquatic ecosystems, an increase in CO 2 in water results in a growth of H + and then a decrease in the pH value. In addition, the pCO 2 of the three hydrological periods was correlated with DO, HCO 3 − , and F C , and especially the significant positive correlation between pCO 2 and F C . pCO 2 is one of the important indicators that control the release of CO 2 from the water surface to the atmosphere (Qian et al. 2017), while F C refers to the specific amount of CO 2 released or absorbed by the water into the atmosphere. According to Eqs. (2), (3), and (4), F C is mainly affected by the water temperature, wind speed, and pCO 2 value of the water. In accordance with the meteorological observation data, the wind speed and water temperature in the study area exhibited negligible fluctuation in each hydrological period. As a result, F C is mainly affected by pCO 2 . O 2 is an important substance involved in the transformation of substances during the process of CO 2 production, and sensitive elements and their compounds affecting redox are distributed in the water column. Although a significant negative correlation between pCO 2 and DO was found, this is related to aquatic organisms' metabolic processes. More specifically, because aquatic photosynthetic organisms, as producers in the water column, can use dissolved CO 2 directly or use HCO 3 − for photosynthesis based on the CCM (Carbon Concentration Mechanism) mechanism, these processes led to an increase of the DO levels in the river water and respond to lower pCO 2 (Zhang et al. 2019a, b). When the pH value of the river water is between 6 and 9, DIC mainly exists in the form of HCO 3 − (Jiang 2013). A significant positive correlation between pCO 2 and HCO 3 − was also extracted. Generally, the HCO 3 − can be converted into CO 2 (reaction HCO 3 − ⇌ CO 2 + H 2 O is going forwards) in karst water. When more HCO 3 − is introduced into the received river, the pCO 2 could be increased (Atekwana and Krishnamurthy 1998). Since HCO 3 − mainly comes from carbonate rock and soil CO 2 in karst basins, accounting for 80% and 20%, respectively (Xuan et al. 2020), both HCO 3 − and pCO 2 have a common growth trend that could result from the weathering of carbonate rocks in the study area.
The correlation between EC, Ca 2+ , and pCO 2 was not obvious in the three hydrological periods. EC is mainly related to the ion concentration in the water. As was mentioned above, the spatial distributions of EC, Ca 2+ , and HCO 3 − were not similar in the three hydrological periods, which also indicates that the rivers were affected by different degrees of human activities in each hydrological period. As a result, the input of exogenous acid (nitric acid, sulfuric  (Jiang. 2013). Nitrogen and phosphorus mainly control the trophic state of aquatic ecosystems. Particularly, they are the main substrates for microbial metabolism, and they have a regulatory role in the biological factors of CO 2 exchange at the water-air interface (Li and Bush 2015;Wang et al. 2007). A reasonable amount of nutrient input will promote the growth of photosynthetic organisms and thus reduce the concentration of CO 2 in the water. However, too much nutrient input will increase respiration and promote the production of CO 2 in the river water (Wang et al. 2017a, b;Wang et al. 2017a, b). The pCO 2 has a significant positive correlation with TN and TP only during the flat season (Table 2), which could suggest that anthropogenic activities during the wet and dry seasons are not the main drivers of river pCO 2 in the studied area, or that other environmental processes mask the impact of anthropogenic inputs, such as dilution effects caused by increased rainfall and enhanced aquatic biological activity. It also indicates the complexity of factors influencing pCO 2 in karst rivers under the influence of urban construction. The concrete reasons with regard to pCO 2 and nutrients remain to be further studied.
In riverine ecosystems, DOC can be converted to DIC through microbial uptake or abiotic degradation, such as photochemistry, and DIC can be converted to DOC through photosynthesis by primary producers (Marx et al. 2017). However, there is no significant correlation between pCO 2 and DOC in this study area. This result suggests that DOC in rivers in karst urban areas is not exclusively from endogenous organic matter, and that urban rivers generally have relatively high DOC due to the influence of watershed surface and point sources (Huang et al. 2020). In particular, urban wastewater treatment plants' discharge was treated effluent with a high concentration of DOC (Yoon et al. 2017;Yang et al. 2018). On the contrary, domestic sewage treatment plants are widely distributed throughout the study area. The daily scale of sewage treatment in Guiyang City has reached 1 million tons. Nevertheless, this is yet to be studied in depth to clarify the proportion of its contribution.

Seasonal variation in river CO 2 diffusion flux
The variation range of CO 2 diffusion flux F c at the water-air interface of the river in the study area was − 3.57 to approximately 519.13 mmol·m −2 ·d −1 in the wet season, with an average value of 147.02 ± 100.30 mmol·m −2 ·d −1 . In addition, the variation range of the dry season was − 9.65 to approximately 399.69 mmol·m −2 ·d −1 , with an average value of 76.02 ± 67.45 mmol·m −2 ·d −1 , and the variation range of the flat season was − 55.27 to approximately 747.93 mmol·m −2 ·d −1 , with an average value of 119.28 ± 168.22 mmol·m −2 ·d −1 . Among them, the wet season was highest, followed by the dry season, the lowest in the flat season, and the seasonal difference was significant. Spatially, the wet season was higher in the west than in the east region, and the dry and flat seasons were higher in the middle and near the southwest region (Fig. 9), which is basically the same as pCO 2 . In the whole watershed, F c was negative at 1 point in the wet season, F c was negative at 2 points in the dry season, and the other 98% of the sampling points were positive. However, the range and standard deviation of the flux in the flat season were higher than those in the wet and dry seasons, indicating that the interference of F c is the most complicated in the flat season, and F c showed a large difference compared with other seasons. Besides, during the flat season, 7 points showed a negative F c all of which were distributed in the upper reaches of the Nanming River, accounting for 14% of the total sampling points. This may be due to the weak disturbance of anthropogenic activities in the area and the favorable growth of aquatic plants, which is related to their strong photosynthesis. In general, the karst surface river water-air interface CO 2 flux was the source of atmospheric CO 2 in each season, and its negative points were basically distributed in areas with low urbanization intensity, which also proves that the urbanization process would increase river CO 2 emissions. In addition, the CO 2 flux in the Nanming River basin was higher than the global average of 75.5 mmol·m −2 ·d −1 (Lauerwald et al. 2015) in the three seasons, while its range value was wide and the standard deviation was large. This result was caused by the relatively high pCO 2 value and wide range of variation in the study area, which is also related to the fact that the study area was located in a rugged mountain area. Since the chemical weathering and mechanical erosion are higher and the CO 2 diffusion rate is faster in mountainous areas (Meybeck et al. 1989;Wang et al. 2017a, b), the weathering rate of karst basin is much higher than that of other non-karst basins (Han and Liu. 2004). Therefore, rivers in the mountainous region of the study area would accelerate the weathering of carbonate rocks and CO 2 diffusion in the basin, resulting in significant differences in CO 2 emissions from rivers in different seasons.
As can be seen from Table 3, on the whole, most of the world's rivers show a supersaturation of CO 2 , indicating atmospheric CO 2 sources. At the same time, different regions and environmental backgrounds suggest that the river CO 2 diffusion flux varies, which means that the factors affecting river pCO 2 are complex, posing a challenge to accurately assessing the status of the global river carbon budget. Compared with rivers in non-karst regions, the rivers in karst areas had higher CO 2 diffusion flux as a whole, and the variation range was relatively larger. On the one hand, the weathering of carbonate rocks brings a large amount of inorganic carbon, which could increase the carbon emission intensity of water in karst areas (Han and Liu. 2004;Liu et al. 2007). On the other hand, the inorganic carbon brought by karstification can also provide a carbon source for the growth of aquatic plants, stimulating the growth of aquatic plants , and reducing CO 2 emissions in rivers. Table 3 also reveals that the Maotiao River, Sancha River, and Lijiang River were close to the CO 2 flux results in this work. Although the Xijiang River, Daning River, Longchuan River, Red River, and Nanming River are karst rivers, the CO 2 fluxes of the first four are greater than the results of this work. This is because   (Ferrón et al. 2007) the lengths and flow scales of the Xijiang River, Daning River, Longchuan River, and Red River were much larger than those of the Nanming River, and the Xijiang River flows through highly urbanized areas, such as the Pearl River Delta, the Daning River as a famous tourist attraction was disturbed by human activities. Moreover, there was developed agriculture in the Red River and Longchuan River, so these three rivers had high pCO 2 . In addition, the CO 2 flux of the Maling River, which is a tributary of Wanfeng Lake Reservoir in the karst area, was much lower than the results of this work. This effect might be related to the fact that the river flows through the natural environment and is less affected by human activities. Meanwhile, the CO 2 diffusion flux of rivers in karst areas was also higher than that of rivers in non-karst urban areas, which might be related to the relatively lower dissolution of carbonate rocks in these areas. On top of that, the CO 2 diffusion flux of non-karst rivers (with less human activity disturbance) was the lowest. In conclusion, karst rivers affected by the progress of urbanization have greater carbon storage and carbon emission potential.

Limitations and uncertainties
The CO 2 flux value was calculated in this work through the model method and the influencing factors of CO 2 in karst urban rivers were discussed. However, there is a certain proportion of agricultural land in the study area (Fig. 1b), and the irrigation of the cultivated land agriculture will change the hydrological channel connecting the river to the land, increasing the contact between the river and the soil. Meanwhile, the use of fertilizers and pesticides will also increase the pCO 2 level of the river (Lambert et al. 2017). Therefore, it is necessary to take into account the impact of agricultural activities on the pCO 2 of rivers in further research. In addition, it has been demonstrated that the CO 2 released by rivers at night is much greater than that during the daytime (Gómez-Gener et al. 2021), while the samples in this work were collected during the daytime. Thus, night sampling should be strengthened in an attempt to increase the scientificity and reliability of the data.

Conclusion
In this work, the spatial and seasonal changes in pCO 2 in the Nanming River were revealed and its tributaries were affected by urbanization and the CO 2 diffusion flux. As a whole, more than 90% of the sampling points were in a supersaturated state, showing the characteristics of carbon sources. The overall variation range of pCO 2 was 147.91 ~ 6309.57 μatm.
The pCO 2 of the river was also the highest in the wet season (1852.58 ± 983.87 μatm), followed by the dry season (1437.34 ± 958.29 μatm), and the lowest in the flat season (1055.40 ± 796.46 μatm). The CO 2 fluxes ranged from − 55.27 to 747.93 mmol·m −2 ·d −1 , and the average values were 147.02 ± 100.30 mmol·m −2 ·d −1 , 76.02 ± 67.45 mmol·m −2 ·d −1 , and 119.28 ± 168.22 mmol·m −2 ·d −1 , respectively. Compared with the main stream, the pCO 2 of the largest tributary, the Taobao River, was more significantly affected by urbanized land use, which implies that a higher degree of urbanization led to bigger pCO 2 . In addition, the correlation analysis results showed that pCO 2 was mainly affected by the weathering of carbonate rocks, aquatic organism metabolic processes, and pollutant emissions. From our analysis, it was demonstrated that karst rivers had a huge CO 2 emission potential under the influence of urbanization. In the future, it is necessary to strengthen the management of river water environments in karst urban areas, improve the river water quality, and reduce the emission intensity of CO 2 in rivers.