Introduction

The latest assessment report points out that the global carbon budget imbalance was + 0.5 Pg C/year on average for 2008–2017 (Le Quéré et al. 2018). Thus, balancing the global carbon budget is still a problem for determining the looping of global carbon (Le Quéré et al. 2018). Recently, inland waters (rivers, streams, reservoirs, and lakes) have been recognized as an important part of the global carbon cycle because they link the terrestrial and marine carbon reservoirs (Cole et al. 2007; Raymond et al. 2013; Regnier et al. 2013; Holgerson and Raymond 2016; Hotchkiss et al. 2018; Tranvik et al. 2018). Carbon dioxide (CO2) emissions from inland waters may represent a significant CO2 source to the atmosphere because inland water bodies are often higher CO2 partial pressure (pCO2) than atmospheric CO2 equilibrium. A global estimate of 2.1 Pg C/year of CO2 emission flux from inland waters has been recently reported (Raymond et al. 2013), which may account for some of the imbalance in the global carbon flux between the different carbon reservoirs (Cole et al. 2007; Battin et al. 2009; Aufdenkampe et al. 2011; Tranvik et al. 2018).

In recent years, the influence of groundwater discharge on CO2 emissions from inland waters has received significant attention. Groundwater can transport high concentrations of dissolved carbon (inorganic and organic), which originates from biogenic and/or geologic carbon, to the surface and possibly change the carbon balance of the aquatic and surrounding terrestrial ecosystems (Johnson et al. 2008; Oviedo-Vargas et al. 2015; Looman et al. 2016; Oviedo-Vargas et al. 2016; Pu et al. 2017; Duvert et al. 2018). Continental karst water bodies are an important category of inland water bodies on earth because they are typically carbon-rich bodies. Many studies have shown that karst systems play an important role in regulating the regional and global carbon cycles through carbonate mineral dissolution and precipitation (Yuan 1997; Jiang and Yuan 1999; Liu and J. 2000; Martin 2013, 2017; Liu et al. 2018), which is a large fraction of the residual land sink (Le Quéré et al. 2018). In karst areas, many surface streams (headwaters) are only fed by a karst spring or a subterranean stream, which constitutes a groundwater-stream continuum. The carbon-rich nature of karst groundwater results in the high carbon contents of surface streams and the high CO2 gradient between the water and air, and thus, karst groundwater influences carbon exchange processes. It has been suggested that karst groundwater carrying soil-respired CO2 may degas some of this CO2 back into the atmosphere when it is discharged from the karst aquifers (Curl 2012). Several studies have estimated the CO2 emission flux on certain time-scales and the carbon sink effect in small karst groundwater-stream continuums (de Montety et al. 2011; Jiang et al. 2013; Khadka et al. 2014; Liu et al. 2015; Yang et al. 2015; Pu et al. 2017). Other studies have reported that the pCO2 and the dissolved inorganic carbon (DIC) of karst groundwater-stream continuums decrease farther downstream because CO2 is emitted as the water travels away from the high CO2 sources (Drysdale et al. 2003; Doctor et al. 2008; Wang et al. 2010; de Montety et al. 2011). However, research on the spatial variability of CO2 emissions from the karst groundwater outlet to the stream mouth is sorely lacking. In particular, few studies have been conducted on the spatial heterogeneity of CO2 emission longitudinally along a single karst stream.

Accurately quantifying CO2 emission from a karst groundwater-stream continuum is important for assessing the carbon sink effect because karst systems are being recognized as a key component of the global carbon cycle (Yuan 1997; Larson 2011; Curl 2012; Martin 2017; Liu et al. 2018). The objectives of this study are to confirm the high spatial and seasonal heterogeneity of the pCO2 and CO2 emissions of a karst groundwater-stream continuum and to determine how these variations affect the carbon budget of this continuum. This study is based on data collected from a 1300-km typical tropical karst groundwater-stream continuum in southern China. The data collected provide relatively high spatial scale of the variations in CO2 emissions and contribute to a deeper understanding of dynamic CO2 processes in karst groundwater-stream continuums. Our results will aid in the accurate upscaling of CO2 emissions across different karst groundwater-stream continuums and the accurate estimation of the amount of carbon sequestered in karst streams.

Materials and methods

Study area

The Guancun surface stream (GSS) is located in Daliang Town, Rong’an County, Guangxi Zhuang Autonomous Region, China (Fig. 1). The GSS is a typical subtropical headwater stream. It is almost exclusively fed by the Guancun underground stream (GUS) in the upper Devonian (D3r) limestone aquifer since no surface tributaries flow into the GSS (Pu et al. 2017; Pu et al. 2019). The outlet of the GUS is the head of the GSS. Thus, they constitute a typical karst groundwater-surface stream continuum (Fig. 1). The length and average width of the GSS are 1320 m and 3.5 m, respectively. The water level of the GSS is controlled by the GUS discharge and exhibits large fluctuations due to monsoon precipitation (Fig. 2). At base level, the water depth is shallow (0.2–1.2 m).

Fig. 1
figure 1

Study area. a Google Earth image of the location of the Guancun Stream in SW China. b Photographs showing the landscape of the sample sites, the groundwater outlet (GC1, B-1), the middle of the stream (GC2, B-3 and B-4), and the stream mouth (GC4, B-2), respectively. c Map of the surface stream flow route and the sample sites in the study area. (Modified from Google Earth 2015)

Full size image
Fig. 2
figure 2

Variation in the hydro-meteorological parameter. a Variations in the daily discharge of the GSS from August 18, 2015 to July 9, 2017. b Variations in the monthly cumulative precipitation and the monthly average air temperature from July, 2014 to July, 2017. The black diamonds in Fig. 2a are partial sampling trips during the study period. The gray-shaded area shows the extreme El Niño year in 2015 (Ma and Ye 2016)

Full size image

The study area is in an eastern Asian monsoon climate, so it is characterized by a cold–dry winter from late November through March and a hot–rainy summer from April through October. The annual average air temperature is 19.7 °C and the multi-annual average precipitation is 1726 mm, 72% of which occurs in the wet season from late April to early September. As a typical monsoon region, the air temperature and precipitation in the GSS catchment co-vary, i.e., they are both high in the wet season and low in the dry season (Fig. 2).

In this study, four monitoring sites were chosen along the GSS channel for sampling (Fig. 1). The most upstream site, site GC1 (24° 52′ 10″ N, 109° 20′ 07″ E), is located at the GUS outlet, which is characterized by typical karst groundwater. There were two monitoring sites in the middle of the GSS, site GC2 (24° 51′ 53″ N, 109° 20′ 03″ E) and site GC3 (24° 51′ 45″ N, 109° 20′ 05″ E). The downstream site, site GC4 (24° 51′ 32″ N, 109° 20′ 01″ E), is near the stream mouth, approximately 1.30 km downstream of site GC1. According to our manual field measurements taken in the middle of the channel, the distance from site GC1 to site GC2, site GC2 to site GC3, and site GC3 to site GC4 is 560 m, 276 m, and 456 m, respectively (Fig. 1).

Field sampling and lab analyses

Monthly field measurements were conducted at the four monitoring sites (GC1, GC2, GC3, and GC4) during the 2014–2017 period. Hydrochemical variables, including water temperature (WT), pH, and dissolved oxygen (DO), were tested monthly in situ at the four sites using multi-parameter meters (WTW 3430, WTW GmbH, Weilheim, Germany). The resolutions of the water temperature, pH, and dissolved oxygen (DO) were 0.1 °C, 0.004 pH units, and 0.01 mg/L, respectively. The pH probe was calibrated prior to deployment using pH standards (4 and 7) according to the manufacturer’s specifications. The DO probe was calibrated using water saturated air.

Water samples were collected at all four sites from midstream using a peristaltic pump. The water was pumped from 0.1 m below the surface. Unfiltered water samples were immediately titrated to determine the total alkalinity with an accuracy of 0.05 mmol/L using a portable testing kit (Merck KGaA Co., Germany). All of the water samples were immediately filtered through 0.45 μm cellulose acetate membranes for ion analysis and were subsequently stored in acid-washed high-density polyethylene (HDPE) bottles. The samples filtered for cation analysis (Ca2+, Mg2+, K+, and Na+) were acidified with trace-metal grade nitric acid (7 M HNO3) to a pH of < 2.0. The samples filtered for analysis of the stable carbon isotopes of the dissolved inorganic carbon (δ13CDIC) were collected in acid-washed dry HDPE bottles and three drops of HgCl2 were added in order to prevent microbial activity. A portable cooler was used to store all of the samples in the field. All of the water samples were stored at 4 °C in a refrigerator in the laboratory until they were analyzed.

According to procedures based on the APHA 2012 methods (Rice et al. 2012), the major anions and cations were measured using an automated Dionex ICS-900 ion chromatograph and ICP–OES (IRIS Intrepid II XSP, Thermo Fisher Scientific, USA), respectively. Calculated analytical errors were within ± 5%. The δ13CDIC values of the water samples were analyzed using a MAT-253 mass spectrometer coupled with a Gas Bench II automated device. The results are expressed as δ13CDIC (‰) with respect to the Vienna Pee Dee Belemnite (V-PDB) standard with an analytical precision of ± 0.15‰. All of the lab analyses were conducted in the Environmental and Geochemical Analysis Laboratory at the Institute of Karst Geology, Chinese Academy of Geological Science (Pu et al. 2017; Zhang et al. 2017).

The rainfall and air temperature were measured using an on-site Vantage Pro 2 weather station (Davis Instruments Corp., USA) during the study period. The resolutions of the rainfall and air temperature were 0.2 mm and 0.1 °C, respectively. Continuous hydrological monitoring started on August 18, 2015, which have 10 months lag compared to the hydrochemical monitoring (Fig. 2a). The daily discharge of the GSS from August 18, 2015 to July 9, 2017 was obtained from the water-level measurements at the gauging station and the water-level-discharge formula.

The CO2SYS program, which is based on the measured alkalinity, pH, and water temperature, was used to calculate the aqueous partial pressure of CO2 (pCO2) and the inorganic carbon species including HCO3−, dissolved CO2, and CO32− using the carbonic acid dissociation constants and the CO2 solubility (Lewis and Wallace 2006). The DIC concentration was calculated using the formula DIC = H2CO3 + HCO3− + CO32−. The CO2SYS software is commonly used to obtain the pCO2 and inorganic carbon species in freshwater and seawater (Butman and Raymond 2011; Ran et al. 2015b; Ran et al. 2017b; Deirmendjian and Abril 2018; Li et al. 2018a).

CO2 evasion flux monitoring

The floating chamber method (FC) was employed to study the CO2 evasion flux across the water–air interface (Fig. 1). The floating chamber is a very popular and relatively low-cost instrument that can determine the diffusive flux at the surface of aquatic ecosystems (Matthews et al. 2003; UNESCO/IHA GHG 2010; Khadka et al. 2014; Lorke et al. 2015). A home-made floating chamber with a volume of ~ 28.3 L and a surface area of 0.0707 m2 was constructed from a polyurethane foam layer sandwiched between two stainless steel cylinders. About 5 cm of the chamber is placed under water and the level is maintained using a styrofoam sheet. At each site, a floating chamber was placed at midstream in the GSS. Prior to deployment, the chambers were placed upside down for a few minutes to allow them to equilibrate with the local air. The air samples were drawn from the chamber via a Tygon tube after deployment using an aspirator pump. The gas samples were stored in a Devex polymer-aluminum bag (volume 1 L) at room temperature. The gas samples were analyzed for CO2 concentration within 48 h of collection using gas chromatography (Agilent-7890) with a resolution of 0.01 ppmv. The five concentrations collected at ~ 0, 5, 10, 15, and 30 min enabled us to calculate the CO2 flux using linear regression. Only sites with a correlation coefficient higher than 0.90 for CO2 were used.

The flux was calculated using the equation (UNESCO/IHA GHG 2010): CO2 flux (mg m−2 h−1) = (S × F1 × F2 × V)/(Sf × F3), where S is the slope from graph of concentration versus time in ppm/min, F1 is the conversion factor from ppm to μg/m3 (1798.45 for CO2), F2 is the conversion factor from minutes to hour (60), V is the volume of air trapped in the chamber (m3), Sf is the surface of the floating chamber over the water (m2), and F3 is the conversion factor from μg to mg (1000).

Results

Hydro-meteorological characteristic

During the monitoring period (August, 18, 2015–July 9, 2017), the GSS discharge ranges from 6.1 to 8069.4 L/s with an average value of 1295.5 L/s and a high coefficient of variation of 128.3% (Fig. 2a). Generally, the GSS discharge increased following precipitation with a relatively short lag time. The GSS was dominantly recharged by karst groundwater from the GUS and mirrored the hydrological variation of the GUS karst aquifer. A significant seasonal variation in the discharge of the GSS was observed during the study period, with a higher discharge in the rainy season and a lower discharge in the dry season, which is consistent with the monsoon climate of the study area. In 2015, an abnormal flood event occurred in the GSS in November and December, 2015 due to abnormal precipitation events caused by extreme El Niño effects (Ma and Ye 2016) (Fig. 2a). In 2016–2017, the GSS had a normal hydrological curve with two higher discharge events in April–July and late August–early October and two low discharge periods in January–March and August (Fig. 2a).

During the study period, the average monthly air temperature in the GSS area ranged from 10.3 to 29.0 °C with the higher value occurring in summer and the lower value in winter (Fig. 2a). The monthly precipitation ranged from 8.9 to 513.4 mm (Fig. 2a). Generally, more precipitation occurred in the summer than in the winter due to the Asian monsoon climate. In 2015, an extreme El Niño event impacted the study area and resulted in extreme precipitation events. The total yearly precipitation was 1903 mm in 2015 with the highest value occurring in May 2015 (513.4 mm) and the second highest occurring in November 2015 (278.8 mm).

Variability of the WT, pH, DO, inorganic carbon species, and δ13CDIC

The water temperature (WT) varied from 14.8 to 24.9 °C in the GSS. The WT of site GC1 is inherited from the GUS karst aquifer, ranging from 19.0 to 21.7 °C with a low coefficient of variation (CV) of 3.35%. The WT increased from site GC1 to site GC4. The downstream sites GC2, GC3, and GC4 had relatively high CVs of 7.09%, 10.27%, and 12.42%, respectively (Table 1, Fig. 3) and were likely affected by the ambient air temperature. The pH of the water ranged from 7.223 to 8.281 in the GSS. A gradual increase in pH occurred from site GC1 to site GC4 during the study period (Table 1, Fig. 3). The highest average pH value occurred at site GC4 (8.025), while the lowest occurred at site GC1 (7.504). The DO concentration varied from 6.14 to 16.72 mg/L during the study period. The lowest DO value was observed at site GC1. The other three sites, GC2, GC3, and GC4, had higher DO concentrations and larger variation ranges than those of site GC1 (Table 1).

Table 1 Summary of the statistics of the DIC species, water temperature, pH, dissolved oxygen, SIc, pCO2, δ13CDIC, and the CO2 flux at the four sites along the GSS
Full size table
Fig. 3
figure 3

Spatio-temporal heterogeneity of the DIC species, pCO2, δ13CDIC, CO2 emission flux, water temperature (WT), and pH from 2014 to 2017 in the GSS. Box plots of the 25th and 75th percentiles and mean values; whiskers represent minimum and maximum values. The Rs and the Ds represent the rainy season and the dry season, respectively

Full size image

The inorganic carbon species were primarily DIC, HCO3−, and dissolved CO2 (Rice et al. 2012). Site GC1 had the highest DIC, HCO3−, and dissolved CO2 concentration, which varied from 180.38 to 348.02 mg/L, from 2789.71 to 5091.10 μmol/L, and from 154.04 to 657.50 μmol/L, respectively, throughout the study period due to recharge from karst groundwater. The DIC, HCO3−, and dissolved CO2 decreased downstream from site GC1 to site GC4, demonstrating significant spatial variation (p < 0.01 by ANOVA) (Table 1). Since the pH value ranged from 7.223 to 8.281 in the GSS, HCO3− was the dominant component of the DIC in all of the water samples from the GSS, accounting for 93.0%, 95.9%, 97.2%, and 97.4% of the DIC on average for sites GC1, GC2, GC3, and GC4, respectively.

The δ13CDIC values ranged from − 14.44 to − 9.53‰ (mean = − 12.82‰) at site GC1 and were lower than those of the downstream sites (p < 0.05 by ANOVA) (Table 1). Generally, the mean δ13CDIC value increased from site GC1 to site GC2 (Fig. 3). The lowest mean δ13CDIC values occurred at site GC1, which was directly impacted by karst groundwater. The δ13CDIC values of site GC2 ranged from − 16.61 to − 9.05‰ (mean = − 12.29‰). The highest mean δ13CDIC values (mean = − 11.94‰) occurred at site GC3 ranging from − 14.45 to − 9.03‰. The δ13CDIC values of site GC4 ranged from − 16.07 to − 9.55‰ with a mean of − 12.05‰.

pCO2 and CO2 emission variability

All of the water samples had pCO2 values higher than the atmospheric value (~ 420 μatm) and were supersaturated with CO2 (Table 1). The pCO2 of site GC1 had the highest mean value (8818.4 μatm) and ranged from 4073.8 to 16,595.9 μatm. pCO2 decreased significantly from site GC1 to site GC4 (p < 0.01 by ANOVA) (Fig. 3). The mean pCO2 values of sites GC2, GC3, and GC4 were 3817.3 μatm, 2732.9 μatm, and 2752.7 μatm, respectively.

The CO2 emission flux varied significantly from 16.5 to 1125.3 mg m−2 h−1 during the study period. Site GC1 had the largest mean CO2 emission flux (423.4 mg m−2 h−1) and ranged from 35.7 to 950.0 mg m−2 h−1, which is about 26 times fluctuation. A significant decrease in the total CO2 emission flux occurred from site GC1 to site GC2 (p < 0.01 by ANOVA). The mean CO2 emission flux decreased from 411.1 mg m−2 h−1 at site GC2 to 328.4 mg m−2 h−1 at site GC3 to 257.0 mg m−2 h−1 at site GC4. The amplitude of the variations in the CO2 emission flux at sites GC2, GC3, and GC4 were significantly lower than that of site GC1.

Discussion

High seasonal and spatial heterogeneity of the pCO2 and CO2 emission flux

We determined that the pCO2 and CO2 emission flux of the GSS had a high seasonal heterogeneity (Fig. 3). Generally, the pCO2 fluctuated from 1513.6 μatm (at site GC3) in the dry season to 16,595.9 μatm (at site GC1) in the rainy season. The mean pCO2 values were the highest at site GC1 in both seasons, whereas the lowest value was observed at site GC3. Higher pCO2 values were observed in the rainy season than in the dry season at all of the sites except for site GC1 (Fig. 3, Table 1). Soil CO2 input, respiration of organic carbon, and CO2 influx from the ambient air above the surface stream will increase the amount of CO2 and the pCO2, while in-stream photosynthesis and CO2 emission from the water to the ambient air will decrease pCO2 (Wallin et al. 2013; Peter et al. 2014; Marx et al. 2017; Pu et al. 2017; Deirmendjian and Abril 2018). The pCO2 of site GC1 is impacted by the karst aquifer because it is the outlet of the GUS. During the rainy season, greater amounts of soil CO2 are produced by strong organic carbon degradation flows into the karst aquifer, resulting in an increase in the pCO2 of the groundwater, which has been reported in numerous studies of subtropical karst areas in southern China (Yang et al. 2012; Pu et al. 2014; Zhao et al. 2015). However, hydrological processes likely disturb this trend (Sun et al. 2007; Yao et al. 2007; Peter et al. 2014; Almeida et al. 2017). During the rainy season, a lot of surface water carrying less soil CO2 flows directly into the aquifer through sinkholes, fractures, fissures, or shafts, which could significantly dilute the karst groundwater and decrease the pCO2 level (Liu et al. 2007; Pu et al. 2014; van Geldern et al. 2015; Marx et al. 2017). The good Pearson correlation coefficient (r = − 0.80, p < 0.05, Table 2) between the discharge and pCO2 values of site GC1 during rainy season indicates that discharge has an impact on the variation of pCO2. This also results in a larger range and lower mean value of the pCO2, HCO3− concentration, dissolved CO2 content, DIC concentration, and δ13CDIC values and a higher mean pH (Fig. 3, Table 2). During the dry season, due to the low air temperature and low discharge in the study area, although soil CO2 recharge was low, the GUS discharge significantly decreased, which resulted in a relative high mean pCO2 at site GC1.

Table 2 Pearson correlation coefficients between discharge, pCO2, CO2 emission flux, and selected variables (significance levels: p < 0.05)
Full size table

The downstream sites (GC2, GC3, and GC4) had higher pCO2 values in the rainy season than in the dry season (p < 0.05 by ANOVA). This phenomenon was directly controlled by the soil CO2 influx and the in-stream respiration/decomposition of organic matter. As was previously mentioned, a lot of soil CO2 could flow into the GUS along with the rainwater and increase the pCO2 value of the stream (Yang et al. 2012; Marx et al. 2017; Campeau et al. 2018). In addition, a higher water temperature and stronger solar radiation can promote in situ respiration and produce more CO2 in the stream (Marx et al. 2017). Thus, the pCO2 level increased in the GSS during the rainy season. This process was also observed in several large subtropical rivers: the lower Red River (Le et al. 2018), the Xijiang River (Yao et al. 2007), and the Guijiang River (Zhang et al. 2017). Due to the influence of the soil CO2 influx and the respiration of organic matter on the pCO2 values of the GSS, elevated discharge played a minor role at sites GC2, GC3, and GC4 (Table 2).

CO2 emission flux varied seasonally, ranging from 16.5 mg m−2 h−1 at site GC4 in the dry season to 1125.3 mg m−2 h−1 at site GC3 in the dry season. However, the mean CO2 emission flux exhibits a regular pattern characterized by higher values during the rainy season and lower values during the dry season at all four of the sites along the GSS (Fig. 3). Usually, the CO2 emission flux is controlled by hydrological processes (lotic or lentic water) and the CO2 concentration gradient between the water and the ambient air (Raymond et al. 2013; Long et al. 2015; Gomez-Gener et al. 2016; Looman et al. 2016; Marx et al. 2017). Generally, the air pCO2 is approximately 445 μatm 1.5 m above the stream surface and changes slightly throughout the year (Pu et al. 2017). Our data indicate that the mean water pCO2 of the water was about 6–21 times higher than that of the ambient air. Thus, the higher water pCO2 levels during the rainy season are favorable to CO2 evasion due to the large CO2 concentration gradient between the water and the ambient air. The elevated discharge in the rainy season could change the hydraulic status and disturb the surface water, which would accelerate the emission of CO2 (Long et al. 2015; Looman et al. 2016; Almeida et al. 2017). This is supported by the good Pearson correlation coefficient between the discharge and the CO2 emission flux (Table 2). It is notable that there were several high CO2 emission events during the dry season (Fig. 3), and some of these events approached or exceeded the highest values for the rainy season. We determined that these high values occurred in November, 2015 during a rare flood event caused by extreme precipitation (278.8 mm) due to El Niño (Fig. 2). This flood changed the hydrological status and stimulated CO2 emission, which resulted in the higher values.

This study demonstrates that the high spatial heterogeneity of the pCO2 and CO2 emission can occur within a distance of only ~ 1300 m of stream. Generally, a clear decrease in pCO2 occurred from site GC1 to site GC4 in both the rainy and dry seasons (p < 0.05 by ANOVA). The rate of decrease of pCO2 was 9.63 μatm/m from site GC1 to site GC2, 2.89 μatm/m from site GC2 to site GC3, and 0.12 μatm/m from site GC3 to site GC4, illustrating the complex spatial variability of the GSS. Although the total mean CO2 emission flux decreased significantly from site GC1 downstream to site GC4 (Table 1), different patterns of change were observed for the rainy and dry seasons. In the rainy season, no significant changes in the CO2 emission flux occurred from site GC1 to site GC2 (p = 0.590 by ANOVA) or from site GC3 to site GC4 (p = 0.856 by ANOVA). Significant changes in the CO2 emission flux only occurred from site GC2 to site GC3 (p < 0.05 by ANOVA) during the rainy season. In the dry season, except for the anomaly high value in November, 2015, the CO2 emission flux changed significantly from site GC1 to site GC2 (p < 0.05 by ANOVA), from site GC2 site to site GC3 (p < 0.05 by ANOVA), and from site GC3 to site GC4 (p < 0.05 by ANOVA). Thus, the CO2 emission flux of the GSS exhibits a complex spatial variability in different seasons.

The decrease in pCO2 from site GC1 to GC4 is most likely controlled by two processes: (1) assimilation of CO2 by aquatic plants during photosynthesis and (2) the emission of CO2 (Marx et al. 2017). The sharp decrease in pCO2 from site GC1 to site GC2 was likely caused by a large amount of CO2 emission coupled with the high water discharge of the rainy season (Fig. 3). Usually, the DO content can be used as a proxy of the intensity of in-stream photosynthesis (Demars et al. 2015; Pu et al. 2017). Significant in-stream photosynthesis in the GSS has been reported by a previous study (Pu et al. 2017). Therefore, during the rainy season, an increase in the mean DO value from site GC1 (mean 7.77 mg/L) to site GC2 (mean 10.10 mg/L) (Table 1) implies that the photosynthesis of the aquatic plants in the GSS assimilates the CO2 and DIC in the water and releases DO into the water (Demars et al. 2015), which also decreases the pCO2 level. During the dry season, the significant decrease in the pCO2 level in the GSS from site GC1 to site GC2 was also caused by a large amount of CO2 emission and the use of CO2 and DIC in in-stream photosynthesis, which the mean DO value increased from 7.63 to 11.73 mg/L (Fig. 3). An increase in the amount of photosynthesis and CO2 emission from site GC1 to site GC2 likely caused increases in pH and δ13CDIC and decreases in the HCO3−, dissolved CO2, and DIC contents (Fig. 3). After strong CO2 occurred at sites GC1 and GC2, stream pCO2 reached a relatively low level. Hence, CO2 emissions at site GC3 and GC4 were far low than at site GC2 during the rainy and dry seasons (Fig. 3). The smaller difference in the pCO2 values of sites GC3 and GC4 resulted in smaller difference in their CO2 emission fluxes (Fig. 3).

Numerous previous studies have concluded that many large rivers exhibit significant spatial and seasonal variability in pCO2 and CO2 emissions as a result of changes in the channel, lake or reservoir control, metropolitan influences, and special climate conditions (Yao et al. 2007; Li et al. 2013; Raymond et al. 2013; Ran et al. 2015a; Ran et al. 2015b; Almeida et al. 2017; Ran et al. 2017b; Wang et al. 2017; Zhang et al. 2017; Le et al. 2018). In addition, several studies have determined that CO2-rich groundwater recharge along the channel at the headwater of the system or into forest creeks likely causes significantly spatial and seasonal variability in the pCO2 and CO2 emissions (Kokic et al. 2015; Looman et al. 2016; Schelker et al. 2016; Marx et al. 2017; Deirmendjian and Abril 2018; Duvert et al. 2018). Our results reveal that pronounced spatial and seasonal variability of pCO2 and CO2 emission flux also occurred within ~ 1300 m of a small karst groundwater-stream continuum, with a 68.8% decrease in the mean pCO2 from site GC1 (groundwater outlet) to site GC4 (farthest downstream). In particular, in the first 560 m, the mean pCO2 decreased by 56% from site GC1 to site GC2, which accounts for 81.4% of the total decrease. The high spatial variability is consistent with the findings of several previous studies. Venkiteswaran et al. (2014) pointed out that most of the CO2 originating from groundwater is lost by emission within a short distance, causing pCO2 to decrease sharply. Johnson et al. (2008) found that in Amazonian headwater streams, over 90% of the CO2 was emitted within 20 m downstream of a groundwater seepage area. Öquist et al. (2009) found that in a Swedish boreal catchment, 65% of the DIC export to the stream was emitted to the atmosphere as CO2 within 200 m of the groundwater entering the stream. Duvert et al. (2018) compiled spatially groundwater-derived pCO2 data for 15 streams and creeks in tropical, temperate, and boreal areas and determined that the quantity of CO2 dissolved in the water decreased by 64 to > 92% (median 76%) in the first 100 m downstream of the groundwater outflow area. These findings suggest that CO2 emission in a groundwater-stream continuum occurs very soon after the groundwater enters the surface stream. Therefore, this study highlights the fact that the measured CO2 emission and pCO2 levels decrease along the length of streams and C loss is underestimated when it is calculated from the concentrations of downstream sampling points. Thus, high temporal-spatial resolution sampling along a groundwater-stream continuum is required to constrain the CO2 water dynamics.

Magnitude of CO2 evasion in the GSS

The pCO2 and CO2 emission flux of the GSS is comparable to the values observed in other streams and rivers in tropical, subtropical, temperate, and continental climates (Table 3). Due to the high spatial variability of the CO2 emission flux observed in this study, we separated the four monitoring sites into two groups: the upstream sites (sites GC1 and GC2) and the downstream sites (sites GC3 and GC4). Generally, the pCO2 values of the upstream sites were significantly higher than most of the values reported for major rivers in Asia, i.e., Yellow River’s main stream (3687 μatm), the Yangtze River (1297–2826 μatm), and the Lower Mekong River (1090 μatm), as well as other rivers around the world such as the Amazon (3230 μatm or 4350 μatm) and Zambezi (2475–3730 μatm) Rivers. These values also significantly exceed those of several subtropical rivers, i.e., the Guijiang, Daning, and Xijiang Rivers and most temperate and continental rivers, i.e., the Mississippi, Hudson, Wuding, and Tigris (Table 3). The pCO2 values of the downstream sites had higher values (> 2500 μatm), which also exceeded those of several subtropical rivers, i.e., the Yangtze, Guijiang, Daning, Xijiang, and Lower Mekong Rivers, and several temperate rivers, i.e., the Mississippi, Hudson, and Wuding Rivers (Table 3). The pCO2 values of the vast majority of the rivers and streams listed were higher than that of atmospheric CO2 equilibrium (Table 3), suggesting that these rivers and streams are oversaturated in CO2 and act as CO2 sources to the atmosphere. Our estimated CO2 emission fluxes for the upstream sites are fall within the middle of the previously reported range of flux (Table 3). However, the estimated CO2 emission fluxes of the downstream sites fall within the lower end of the published rates (< 200 mmol m−2 day−1) (Table 3). A subtropical river, the Guijiang River, has a negative CO2 flux in the summer and winter, implying that the river directly absorbs ambient atmospheric CO2 and is a carbon sink (Table 3). However, the CO2 emissions of all of the sites along the GSS were positive. The highest CO2 emission flux was found in the main stream of the Yellow River (886.2 and 661.9 mmol m−2 day−1 for the rainy and the dry seasons, respectively). Temperate rivers (Mississippi River), subtropical rivers (Daning and Zambezi River), tropical river (Amazon River), and subtropical stream (Buffalo Bayou and Spring Creek) tend to have CO2 emission flux 1.1–1.5 times higher than the upstream sites and CO2 emission fluxes 1.5–2.3 times higher than the downstream sites in the GSS (Table 3). In general, although the GSS is a small karst stream with a ~ 1300 m length, its CO2 emission flux is comparable to that of some large rivers due to the local enrichment of karst groundwater with high CO2 or DIC concentrations.

Table 3 Comparison of our data with the results of previous studies pCO2 and CO2 emission flux studies in different climate conditions (S, summer; W, winter; A, autumn; D, dry season; We, wet season)
Full size table

CO2 emission flux and the DIC flux of the stream

The discharge, CO2 emission, and DIC data of the four sites from October 2015 to December 2016 were used to calculate the CO2 emission flux and the DIC flux of the stream. The distance between two neighboring sample sites was manually measured in the field at midstream. The surface area of the stream was estimated by mapping each stream sector between two sample sites using the polygon applications in Google Earth Pro. Then, each sector was multiplied by the corresponding average flux of the two neighboring sample sites and the results were added together to calculate the overall CO2 emission flux (t CO2/year) (Teodoru et al. 2015). The average annual dissolved inorganic carbon (FDIC) fluxes of the different sectors of the GSS were calculated from the mean annual discharge (Qm), the instantaneous DIC concentrations (DICi), and the instantaneous discharge rates (Qi) (Brunet et al. 2009) using the following equation: FDIC = 365Qm\( \left[\frac{\sum \limits_{\mathrm{i}=1}^{\mathrm{n}}{\mathrm{DIC}}_{\mathrm{i}}{\mathrm{Q}}_{\mathrm{i}}}{\sum \limits_{\mathrm{i}=1}^{\mathrm{n}}{\mathrm{Q}}_{\mathrm{i}}}\right] \). Since the GSS is a karst groundwater-fed stream with a short stream channel, this study assumes that the discharge does not significantly change from site GC1 to site GC4, i.e., within the length studied. Thus, the same instantaneous discharge was used for all four of the sites in each monitoring period to calculate the DIC flux.

The calculated results indicate that the total amount of CO2 released into the atmosphere from the GSS was 21.75 t CO2/year. This value is equal to the sum of the three sectors, i.e., 13.16 t CO2/year from site GC1 to site GC2, 3.62 t CO2/year from site GC2 to site GC3 and 4.97 t CO2/year from site GC3 to site GC3. The calculated DIC yield from the GC1 site was 1.24 × 104 t DIC/year, which represented DIC export from the GUS karst system. The calculated DIC fluxes of sites GC2, GC3, and GC4 are 1.17 × 104 t DIC/year, 1.15 × 104 t DIC/year, and 1.14 × 104 t DIC/year, respectively, showing a decreasing trend due to carbon loss through CO2 emission, carbon assimilation by aquatic phototrophs, or/and calcite precipitation in the GSS (Pu et al. 2017). In this study, we calculated the approximate carbon loss between two neighboring sample sites. The DIC loss flux was 0.07 × 104 t DIC/year from site GC1 to site GC2, 0.02 × 104 t DIC/year from site GC2 to site GC3, and 0.01 × 104 t DIC/year from site GC3 to site GC4. Thus, the carbon loss ratio due to CO2 emission (CO2 emission flux/DIC loss flux) was 1.17% from site GC1 to site GC2, 2.96% from site GC2 to site GC3, and 5.62% from site GC3 to site GC4. Overall, the CO2 emission flux from the GSS only accounts for 1.71–5.62% (mean = 3.43%) of the stream’s DIC loss flux, which falls within the lower end of the range of published ratios (2–30%; Liu et al. 2010). The estimated CO2 emission ratio indicates that the proportion of CO2 lost from the GSS was low, implying that most of the DIC loss was the result of carbon assimilation by aquatic phototrophs or/and calcite precipitation, which is consistent with the results of Pu et al. (2017).

Further implications

Demonstrating the high spatio-temporal heterogeneity of pCO2 and CO2 emissions contributes to our understanding of carbon cycle processes and to improving the accuracy of the carbon budget in karst catchments. Although several new techniques and methods, e.g., high-resolution online monitoring equipment, floating autochambers, and thin boundary layer models, were used to study the CO2 emissions from streams and rivers, accurately estimating of the CO2 emission flux is still difficult due to high spatio-temporal heterogeneities. This study provides a new point of view regarding the complexity of the spatio-temporal variations of CO2 emission in a karst groundwater-stream continuum. The obvious seasonal heterogeneity of the pCO2 and CO2 emission flux implies that monitoring should be conducted over as many seasons as possible, and should at least include the dry and rainy seasons. Specially, the pCO2 and CO2 emission flux of some extreme climate events should be taken into account. If sampling was conducted during the dry season, a low pCO2 value and CO2 emission flux would be obtained, whereas if sampling was conducted during the rainy season, the annual CO2 emission could be overestimated. If sampling was conducted during or immediately after extreme rainfall events, abnormal pCO2 and CO2 emission value would be obtained, resulting in the overestimation of the CO2 emission flux. Complex spatial heterogeneity significantly affects the accuracy of CO2 emission estimates in karst catchments. The upstream pCO2 value and CO2 emission values, especially near the karst groundwater outlet, are significant higher than the downstream values. If sampling is conducted near the karst groundwater outlet, the pCO2 and CO2 emission values will be overestimated, whereas, if sampling is conducted downstream, the pCO2 and CO2 emission values will be underestimated. Therefore, based on the results of our study, in a groundwater-stream continuum, monitoring sites employing the floating chamber method should be located in the upstream, midstream, and downstream areas. This will provide a relatively accurately estimate of the CO2 emission flux and allow for the accurate determination of the carbon budget of a groundwater-stream continuum. Therefore, when designing a sampling campaign for a stream or river, it is necessary to include as many time points and sampling sites as possible to accurately calculate the CO2 emission flux, rather than sampling once at only a few or one sampling site.

Recently, several studies have concluded that since land karst system are carbon sinks, it is uncertain if degassing of from rivers and streams (especially the headwater of the system) can be ignored (Cole et al. 2007; Aufdenkampe et al. 2011; Curl 2012; Marx et al. 2017). In terms of a karst groundwater-stream continuum, the results of our study of the Guancun stream provide important information that sheds light on how regional karst groundwater influences CO2 emissions from streams on the spatio-temporal scale and increases our understanding of the C source/sink status of karst systems. The Guancun stream is not a unique case, given that the hydrogeological factors (karst groundwater-stream continuum) cause the high CO2 emission flux on the spatio-temporal scale. In karst areas, streams such as the Guancun stream may represent relatively common hotspots of CO2 emission, and thus, their emission fluxes merit a closer investigation. In addition, the proportion of CO2 returned to the atmosphere from the GSS was low, implying that the inorganic carbon exported from the karst system was not fully returned to the atmosphere by groundwater-stream continuum in the form of CO2 gas and some of the inorganic carbon was fixed by aquatic phototrophs. Overall, the accurate assessment of the CO2 emission flux in a karst groundwater-stream continuum needs to take into consideration the high spatio-temporal heterogeneity in order to reduce the bias of the CO2 emission flux and to improve the catchment CO2 budget balance.

Conclusions

Quantifying the CO2 emission flux from a stream or river is important to the global carbon balance. However, the high spatial and seasonal heterogeneity of pCO2 and CO2 emissions restricts the accuracy of quantifying the CO2 emission flux. This study demonstrates that several physio-chemical parameters, such as pCO2 and CO2 emissions, exhibit high spatial and seasonal heterogeneity within a distance of only ~ 1300 m in a small karst groundwater-stream continuum. A significant decrease in the pCO2 and CO2 emissions occurred from site GC1 (mean 8818.4 μatm for pCO2 and mean 423.4 mg m−2 h−1 for CO2 emission) to site GC4 (mean 2752.7 μatm for pCO2 and 257.0 mg m−2 h−1 for CO2 emission). Except during extreme rainfall events, higher pCO2 and CO2 emission values were observed in the rainy season than in the dry season at most sites. The calculated results show that the total CO2 released to the atmosphere from the GSS was 21.75 t CO2/year, which accounts for 1.71–5.62% (mean 3.43%) of the stream’s DIC loss flux. Thus, this study highlights the importance of considering spatio-temporal heterogeneities when assessing the CO2 emission flux from a stream or river, especially in a groundwater-stream continuum. This study also provides a sampling and monitoring framework that can reduce the potential biases of CO2 emission assessments.