Sensitivity of chemical weathering and dissolved carbon dynamics to hydrological conditions in a typical karst river

To better understand the mechanisms that hydrological conditions control chemical weathering and carbon dynamics in the large rivers, we investigated hydrochemistry and carbon isotopic compositions of dissolved inorganic carbon (DIC) based on high-frequency sampling in the Wujiang River draining the carbonate area in southwestern China. Concentrations of major dissolved solute do not strictly follow the dilution process with increasing discharge, and biogeochemical processes lead to variability in the concentration-discharge relationships. Temporal variations of dissolved solutes are closely related to weathering characteristics and hydrological conditions in the rainy seasons. The concentrations of dissolved carbon and the carbon isotopic compositions vary with discharge changes, suggesting that hydrological conditions and biogeochemical processes control dissolved carbon dynamics. Biological CO2 discharge and intense carbonate weathering by soil CO2 should be responsible for the carbon variability under various hydrological conditions during the high-flow season. The concentration of DICbio (DIC from biological sources) derived from a mixing model increases with increasing discharge, indicating that DICbio influx is the main driver of the chemostatic behaviors of riverine DIC in this typical karst river. The study highlights the sensitivity of chemical weathering and carbon dynamics to hydrological conditions in the riverine system.

with variations in hydrology 27,28 , as well as variations in sources and fluxes of dissolved carbon 26,29 . Detailed information on dissolved carbon dynamics with respect to hydrological conditions is still scarce, and relevant controlling processes are largely unknown. To constrain the DIC sources and clarify the relative biogeochemical processes, carbon isotopes can be used to identify the carbon sources and carbon biogeochemical processes 26,[30][31][32] .
The continuous exposure of carbonate rocks on the Yunnan-Guizhou plateau is the largest karst area in the world 8,10,25,26 . The Wujiang River as one of the largest rivers in the Yunnan-Guizhou plateau is an ideal river to study the chemical weathering and carbon dynamics in the carbonate-rich areas [8][9][10] . A series of field campaigns have been conducted to sample river water under different hydrological conditions. This study uses data collected through the field campaigns to investigate correlations among chemical weathering, dissolved carbon dynamics, biogeochemical processes and hydrology. Carbon isotopes of dissolved inorganic carbon denoted as δ 13 C DIC , are used to identify carbon sources and constrain their contributions, which can provide insights into the main factors controlling carbon dynamics over time.

Results
Hydrochemical characteristics. pH of the river water samples is mildly alkaline (7.93-8.30) within the range of the whole Wujiang basin (7.6 to 8.9) 8 . The electrical conductivity (EC) values range from 315 to 407 μ S/cm for the whole hydrological year (Table S1). Mean discharge-weighted concentrations can be calculated as Σ (Q i C i )/ Σ Q i , where the subscript i represents each sample during the hydrological year 33 . The total dissolved solids (TDS) range from 239 to 356 mg/L (Table S1), with an average of 265 mg/L, which is higher than the world average value (97 mg/L) 22 . In comparison to rivers draining carbonate terrain worldwide, the average TDS value of the Wujiang River is higher than that of the South Han River (174 mg/L) in South Korea 34 , the Ganges and Indus rivers (164 mg/L) draining the Himalayas 35,36 and the Mackenzie River (160 mg/L) draining the Rocky Mountains 37 . However, the TDS values are in the same range with that of the upper Yellow River (274 mg/L) 7 , the upper Xijiang River (297 mg/L) 25 but lower than that of the Houzhai River (441 mg/L) 26 in China. The total cationic charge (TZ + = Na + +K + +2Ca 2+ +2Mg 2+ , in μ eq/L) and total dissolved anions (TZ − = Cl − + 2SO 4 2− + HCO 3 − + NO 3 − , in μ eq/L) are well balanced, indicating that the unanalyzed ions play a minor role in charge balance. The mean discharge-weighted concentrations of major cations are as follows: Ca 2+ (1.33 mmol/L) > Mg 2+ (0.40 mmol/L) > Na + (0.17 mmol/L) > K + (0.04 mmol/L) (Fig. S1). The mean discharge-weighted concentrations of major anions are as follows: HCO 3 − (2.37 mmol/L) > SO 4 2− (0.44 mmol/L) > Cl − (0.13 mmol/L) (Fig. S1). The Wujiang River shows a dominance of Ca 2+ , Mg 2+ , HCO 3 − and SO 4 2− , which is similar to the characteristics of rivers draining karst areas 10,20,36,37 . Carbon characteristics. DIC is the sum of CO 2 (aq), carbonic acid (H 2 CO 3 ), HCO 3 − , and carbonate (CO 3 2− ) ions 26 . The DIC concentrations in the river water vary from 2303 to 2783 μ mol/L (Table S2), which is triple times higher than the world average concentration (852 μ mol/L) 19 . The dissolved organic carbon (DOC) has a narrow range from 0.89 mg/L to 1.32 mg/L (Table S2), with no significant temporal variations. The partial pressure of carbon dioxide (pCO 2 ) is a function of respiration, which can lead to increases in both riverine pCO 2 and the dissolution of carbonate 26 . The pCO 2 values range from 711 μ atm to 1619 μ atm (Table S2), which is much higher than the local atmospheric pCO 2 (349 μ atm) (Eq. S1). The value of δ 13 C DIC ranges from − 14.8‰ in the high-flow season to − 9.4‰ in the low-flow season (Table S2), with the average value of − 12.1‰.
Temporal variations in major elements. The concentrations of major elements show significant temporal variations in the Wujiang River (Fig. S2). The discharge (Q) is low and relatively stable from November 2013 to March 2014 and relatively high from April 2014 to October 2014, reaching a maximum in July 2014 (Fig. S2). Generally, all the major elements (except Si) exhibit slightly decreasing trends during the high-flow season due to dilution and reach the maxima in February and March during the low-discharge period (Fig. S2). However, Si shows an increasing trend in the high-flow season relative to the low-flow season (Fig. S2).
Relationship between elemental concentrations and discharge. Godsey et al. 33

and Clow and
Mast 38 have demonstrated that the concentrations of weathering products are negatively correlated with discharge and can be approximated as power-law functions: where a is a constant, and b represents the index that explains the deviation from chemostatic behavior 39 .
The regression coefficient b in the relationship between C and Q has a physical interpretation. If b = 0, the catchment behaves entire chemostatically; and if b = − 1, Q is the only controller on C, constant solute fluxes being diluted by variable fluxes of water 27 . However, when b > 0, no dilution effect is present, because large amounts of inputs are induced by high discharge. Significant relationships (R 2 values ranging from 0.35 to 0.65) between C and Q were found for the Wujiang River (Fig. 1). The slope values suggest that nearly all dissolved solutes (except Si) become diluted with increasing discharge and that the concentrations of Ca 2+ , Mg 2+ and HCO 3 − do not vary as much as those of Na + , K + , SO 4 2− and Cl − (Fig. 1). Although all these dissolved solutes except Si decrease with increasing discharge, they do not strictly follow the theoretical dilution curve. Godsey et al. 33 have postulated that concentrations are relatively constant in wide ranges of discharge, which maybe due to large amount of water stored in a catchment flows into the river induced by intense precipitation. Matrix porosity is widely distributed in the carbonate-rich areas, which stored large amount of "old" water. The near chemostatic behavior in the catchment may be ascribed to the carbonate-rich characteristics. The concentration of Si has a positive relationship with discharge ( Fig. 1), indicating that Si is not affected by dilution, and that multiple biogeochemical processes counteract the dilution effect.
Scientific RepoRts | 7:42944 | DOI: 10.1038/srep42944 Discussion Elemental ratio-discharge relationship. The dissolved loads of the river water are derived from atmosphere, rock weathering and anthropogenic pollutions 40 . The chemistry of river water in the Wujiang River is dominated by carbonate weathering (limestone and dolomite) (Fig. S3a). All of the samples have [Na + ]/[Cl − ] ratios exceeding 1 (Fig. S3b), indicating that silicate weathering is a clear source of major ions. There is no geological evidence for the exposure of evaporites in the river basin 8 , the contribution from evaporites is thus neglected. During carbonate weathering, the cations Ca 2+ and Mg 2+ are released into the dissolved phase. During silicate weathering, the cations Na + , K + , Ca 2+ and Mg 2+ as well as Si, are released into the dissolved phase. Carbonate weathering is the primary sources of Ca 2+ , Mg 2+ and HCO 3 − , and silicate weathering is the major source of Na + , K + and Si in the basin 8 . Cl − has two major sources: atmospheric and anthropogenic inputs 8 . Changes in DOC concentrations can be ascribed to mixing of multiple sources and biogeochemical processes 41 .
Elemental ratio-discharge relationships can be used to identify source changes and examine biogeochemical processes during various hydrological conditions 16 . Changes in elemental ratios are related to changes in the source or differential dissolution/precipitation rates between minerals, especially for carbonate and silicate minerals, with changing discharge 14,16 . As weathering progresses, saturation with respect to secondary silicate and the retention of silicate in the reservoir can buffer the concentration of dissolved Si, while the concentrations of dissolved cations that are not readily partitioned into secondary silicates continue to increase 16,42 . The variation in the ratio of (Na* + K)/Ca (Na* = [Na + ] − [Cl − ]) with changing discharge (Fig. 2a) suggests that the relative The theoretical dilution curve means that these elements are diluted by deionized water (b = − 1), and the theoretical dilution curve of rain water means that these elements are diluted by rain water, which is calculated by rainwater of Guiyang. contribution of silicate mineral dissolution to the dissolved loads changes with discharge, which is similar to the findings of other studies 38,43 . Si, Na + and K + are from same lithologic source, and variations in Si/(Na* + K + ) with discharge can be used to interpret the balance between secondary mineral precipitation and primary silicate weathering. The Si/(Na* + K) ratio can be regarded as a proxy commonly related to the "intensity" of silicate weathering 44 .
Because multiple concentration-discharge relationships occur with increasing discharge (Fig. 1), reactive transport can generate various behaviors in the ratio-discharge relationships. The observed variation in (Na* + K)/Ca with discharge in Wujiang River doesn't show a linear positive relationship or negative relationship. As the mean discharge of the Wujiang River in the monsoon season is approximately 3000 m 3 /s, we define that discharge is two times greater than the mean discharge (i.e., 6000 m 3 /s) as extremely high discharge. The (Na* + K)/Ca ratio decreases rapidly with increasing discharge at discharge below 6000 m 3 /s (Fig. 2a). Thus, the slope of the (Na* + K)/Ca ratio-discharge relationship suggests that the relative proportion of solutes derived from silicate weathering decreases with increasing discharge at discharges below 6000 m 3 /s. However, (Na* + K)/ Ca has a relative stable value with increasing discharge when the discharge is higher than 6000 m 3 /s (Fig. 2a). Under high discharge conditions with short fluid transit times, water cannot reach equilibrium with rocks 14,16 . With increasing discharge, carbonate minerals dissolve rapidly and can drastically alter the water chemistry composition relative to the silicate minerals 14,45 ; thus, the (Na* + K)/Ca ratio decreases with increasing discharge. When discharge is greater than 6000 m 3 /s, as most of dissolved solutes concentrations have relative constant values with increasing discharge (Fig. 1), the relative proportion of solutes derived from silicate weathering versus carbonate has a narrow range.
The power law exponent describing the relationship between Si and discharge shows a positive trend (Fig. 1). Dissolved Si concentration is maintained by equilibrium with respect to secondary silicate mineral and the retention of silicate in the reservoir 38,42,43 . The Si/(Na* + K) ratio increases with increasing discharge for discharge below 6000 m 3 /s, suggesting that the relative release of Si to Na* + K from primary silicates is greater than lower discharge and less dissolved Si is retained in the reservoir with increasing discharge for discharges below 6000 m 3 /s. Increasing discharge, decreases the transit time of fluids, leading to less time for the fluids and minerals to reach equilibrium with a secondary Si-bearing phase, and less time for retention in the reservoir in the upper reach. Thus, the Si/(Na* + K) ratios increase with increasing discharge for discharge below 6000 m 3 /s. Extreme discharge will shift deep flow paths to fast near-surface flow paths. Because of the less transit time of water, there is less time for biogeochemical processes such as ions exchange, biological uptake. So the Si/(Na* + K) ratio is near to the ratio from silicate weathering at extremly high discharge. Therefore, samples show relative stable Si/ (Na* + K) ratios with increasing discharge when the discharge is higher than 6000 m 3 /s. Chemical weathering fluxes affected by hydrological conditions. The temporal variability in discharge on the Yunnan-Guizhou Plateau in southwestern China mainly depends on rainfall associated with the monsoon climate. As discussed above, the major element dynamics are dominated by discharge, and hydrologic flushing further induces chemostatic behavior by increasing the reactive mineral surface area, which accelerates the mineral weathering 38 . In this study, a forward model is used to constrain the elemental sources (Eqs S1-9). Carbonate weathering fluxes (F Carb ) and silicate weathering fluxes (F Sil ) are calculated using Eqs S11 and 12, respectively.
The results show a broad range: F Carb ranges from 26.1 kg/s to 819.7 kg/s, F Sil varies from 3.7 kg/s to 149.8 kg/s. F Carb and F Sil have strong correlations with discharge ( Fig. 3a and b). The strong correlations between chemical weathering fluxes and discharge indicate that chemical weathering is dominated by hydrological conditions. Contours of different power law exponents spanning from dilution (− 1) to "chemostasis" (0) (Fig. 3a and b) illustrate the sensitivity of chemical weathering to discharge changes. The power law exponent between F Carb and discharge is approximately − 0.1. F Sil has a power law exponent of approximately − 0.2 at relatively low discharge rates (e.g., discharge of < 6000 m 3 /s) and approximately − 0.1 at extremely high discharge rates (e.g., discharge of > 6000 m 3 /s). Both F Carb and F Sil shows strong chemostatic responses to varying discharge, and F Carb shows a slightly more obvious chemostatic response. The near chemostatic behavior of chemical weathering fluxes responding to changing discharge may be ascribed to the hypothesis that fluids will quickly approach chemical equilibrium in rapidly eroding environments 16,43,46 . Variations in chemical weathering fluxes versus discharge can be ascribed to the dissolution kinetics, because the dominance of physical erosion during high discharge allows the more easily weatherable carbonate to dissolve 14 . And Importantly, our results characterize the sensitivity of chemical weathering fluxes to changes in discharge.
Response of dissolved carbon dynamics to hydrological changes. DIC is an important part of the total fluvial carbon to the ocean 47 . DOC is a significant constituent in aquatic ecosystems, and its concentrations in streams is influenced by both temperature and water flow pathway dynamics associated with changes in discharge 48 . The fluxes of DIC (F DIC ) and DOC (F DOC ) in rivers are strongly linked to climate condition ( Fig. 4a and b). As discussed above, DIC and DOC show strong chemostic response with respect to varying discharge. Both F DIC and F DOC have strong positive relationship with discharge ( Fig. 4a and b). The chemostic behaviors of DIC should be ascribed to primary production in the basin, as well as dissolution and precipitation. The power law exponents of both F DIC versus discharge and F DOC versus discharge are close to 0, indicating that the fluxes of dissolved carbon in the Wujiang River are dominated by hydrological conditions, and fluxes of DIC and DOC are sensitive to hydrological variability.
Biological CO 2 discharge, in situ biodegradation and photosynthesis is the primary driver of the pCO 2 in water 26,30 . Because of the relative low value of DOC and few amounts of aquatic plants in valley type of river channel, the effect of biodegradation and photosynthesis to pCO 2 in the Wujiang river could be neglected. Therefore, biological CO 2 discharge should be the main control on pCO 2 , especially during the flooding stage. The values of pCO 2 show a negative correlation with discharge (Fig. 4c) and a power-law dilution effect with increasing discharge. The pCO 2 values exhibit strong chemostatic behavior with respect to increasing discharge when the discharge is greater than 1500 m 3 /s. Biological CO 2 produced by microbiologic activities and plant respiration would increase under high temperature conditions in summer at Southwest China 26 . So biological CO 2 discharge is likely responsible for the chemostatic behavior. At extreme discharge rates, the fluid follows near-surface flow paths rather than deep flow paths, and biological CO 2 does not have enough time to react with rocks. Therefore, biological CO 2 can be transported to the river directly, which counteracts the dilution effect following periods of high discharge (Fig. 5). Li et al. 26 showed that soil CO 2 plays an important role in shifting δ 13 C DIC values in a small karst catchment. Epikarst aquifers water, with more negative δ 13 C DIC values than that in riverine water, have an important impact on karstic water carbon in carbonate-rich areas due to the active exchange between the surface water and subsurface flow water. Thus, the water stored in the matrix porosity and soil water with high contents of biological CO 2 would flow into the river induced by high discharge, leading to decrease of δ 13 C DIC in the Wujiang  Fig. 4d, there is a generally negative correlation between δ 13 C DIC values and discharge. The values of δ 13 C DIC in Wujiang River changing with hydrological variability could not be interpreted in terms of simple dilution. Chemical weathering enhanced by increasing discharge produces large amounts of DIC, which is more positive than those of biological CO 2 . Thus, the δ 13 C DIC does not respond dramatically to the increasing discharge. As seen in Fig. 4d, the δ 13 C DIC values exhibit a power-law mixture effect with increasing discharge, indicating that there is a power-law relationship between the δ 13 C DIC values and discharge, as follows:

River. As indicated in
where a, b and c are constants. Usually, c is equal to the δ 13 C value of biological CO 2 . The exponent in the power-law relationship depends on the amount of biological CO 2 . A power law exponent of zero indicates that dilution will not change the values of δ 13 C DIC and that biological CO 2 will not be directly discharged into the river under high discharge conditions. A power law exponent close to − 1 means that a large amount of biological CO 2 will be discharged into the river, shifting the δ 13 C DIC values.
The δ 13 C DIC values increase with increased (Na* + K)/Ca ratios, which represent the relative contribution of silicate weathering versus carbonate weathering (Fig. S4a). The results indicate that silicate weathering is not responsible for the lower values of δ 13 C DIC in the high-flow season due to relative low (Na* + K)/Ca ratios. Thus, more soil CO 2 dissolution following rain water discharge would drop δ 13 C-DIC value in the river and counteract the riverine pCO 2 at the same time in the high-flow season. There is a positive relationship between δ 13 C DIC values and SO 4 /DIC ratios (Fig. S4b), suggesting that the lower values of δ 13 C DIC are ascribed to the soil CO 2 discharge and the reduced ratio of carbonate weathering by H 2 SO 4 . In this study, the power law exponent of δ 13 C DIC versus discharge is close to − 0.1, indicating that relative high carbonate weathering rate and biological CO 2 contribution are stimulated by high discharge likely occurs. DIC sources. The sources of DIC can be constrained by δ 13 C DIC values due to the large difference between biological carbon and geological carbon 1,31,41 . As pCO 2 in riverine water is much higher than pCO 2 in the atmosphere (Fig. 4c), the contribution of atmospheric CO 2 to DIC can be neglected 5,25,30,32,49,50 . Thus, the riverine DIC has two major sources: geological source and biological source. As discussed by Telmer and Veizer 32 , carbon isotopes in marine limestones and dolostones deposited since the end of the Proterozoic show typical marine values close to 0‰. C3 plants dominate the study area 25 with the average δ 13 C value of − 27‰. Cerling et al. 51 have reported that there is carbon isotope fractionation of approximately 4.4‰ during the diffusion of CO 2 . Therefore, the carbon isotope of soil CO 2 in the Wujiang river basin is likely − 22.6‰. Given the different δ 13 C DIC of these two endmembers, the source contribution to riverine DIC can be calculated as follows: where δ 13 C geo and δ 13 C bio is the δ 13 C values of geological carbon and biological carbon, respectively, and F geo is the proportion of carbon from the geological source. In the case that the DIC is affected by carbonate precipitation and CO 2 degassing, the F geo is overestimated because of the isotope fractionation 52 . The contribution of DIC bio (DIC from biological sources) to DIC in river water increases from 41.5% to 65.4% based on the calculation (Eq. 3). Furthermore, the concentrations of DIC geo (DIC from geological sources) and DIC bio can be determined based on the relative proportions of DIC. The DIC geo concentrations show a power-dilution relationship with discharge (Fig. 6a), indicating that DIC geo exhibits strong chemostatic behavior with respect to discharge changes. Increasing carbonate weathering rates are likely responsible for this chemostatic situation. In contrast, DIC bio values show a positive relationship with discharge (Fig. 6b), suggesting that biological DIC influx is the main driver of the chemostic behavior of total DIC with increasing discharge. Atmospheric precipitation infiltrates into the soil and flush soil CO 2 into the river. Therefore, high discharge brings excessive biological DIC into the river, leading to increasing DIC bio concentrations with increasing discharge.
Fluvial DIC concentrations and δ 13 C DIC values primarily reflect the mixing of compositionally distinct endmembers ( Fig. 6a and b). Physical and biological processed take turns to alter the composition of the DIC pool during changing hydrological conditions (Fig. 5). δ 13 C DIC may be more sensitive than DIC concentrations to hydrological changes ( Fig. 4a and d), which is in agreement with Waldran et al. 53 . Clearly, physical and biological processes affect DIC concentrations during changing hydrological conditions. Therefore, continuous high-frequency monitoring during field programs should be conducted 53 . The results suggest that δ 13 C DIC values can be useful for revealing the response of biogeochemical processes to riverine hydrological conditions.

Methods
The study area. The Wujiang River basin (Fig. S5) is located in the center of the Southeast Asian Karst Region, the largest karst area in the world 8,25 . The Wujiang River is the largest tributary on the south bank of the upper Changjiang River, which is the 3 rd longest river in the world 8,9 . The drainage area is 87 920 km 2 , and the region features a warm subtropical climate 9 . The mean annual precipitation for the last several years has ranged from 850 to 1600 mm 8 , and the occurrence of a seasonal monsoon results in high precipitation during summer and low precipitation during winter 10 . Carbonate rocks are widely exposed in this area, with no significant outcrops of evaporites 8 . Sampling and analyses. The sampling site is located at the outlet of the Wujiang River (Fig. S5), where is 45 km away from the mainstream of the Changjiang River. The water samples for chemical and isotopic analyses were collected, monthly, from November 2013 to October 2014, i.e., throughout one entire hydrological year. Additional samples were collected during high-discharge periods. Samples were collected in the middle of the river by boat. pH, teeasured in the field. Alkalinity was determined with 0.02 μ M hydrochloric acid mperature(T) and Ec were mwithin 24 hours. The samples were filtered through 0.45 μ M cellulose-acetate filter paper. Major cations (K + , Na + , Ca 2+ and Mg 2+ ) and Si were acidified to pH = 2 with ultrapurified HNO 3 and measured via Inductively Couples Plasma-optical Emission Spectrometry (ICP-OES) (with an error of 3%). Anions (SO 4 2− , Cl − and NO 3 − ) were analyzed using a Diones ICS90 (with an error of 5%). The DOC was measured using an OI Analytical Aurora 1030 TOC analyzer. For the δ 13 C DIC analysis, the method of Li et al. 25 was used, with a precision of 0.2%. All these analyses were conducted at the State Key Laboratory of Environmental Geochemistry (Institute of Geochemistry, Chinese Academy of Sciences). Daily water discharge data were obtained online from the Ministry of Water Resources (http://www.hydroinfo.gov.cn/). Finally, pCO 2 , SIc and all DIC species were calculated based on mass action relationships and the relative equilibrium constants.