The effects of rainfall on groundwater hydrogeochemistry and chemical weathering

Due to the special hydrogeological conditions in karst areas, groundwater responds quickly to rainfall. The covariation of ion concentrations and spring discharge can help better understand the hydrogeochemical process of groundwater occurring in the heterogeneous karst aquifers. In this study, high-resolution monitoring of groundwater discharge, hydrochemistry, and stable isotopes was conducted at the Qingjiangyuan (QJY), a spring of the Qingjiang watershed in Hubei, China. The purpose is to investigate the changes in hydrogeochemical processes and chemical weathering under the influence of rainfall. The dynamics of spring discharge indicate the presence of pipelines and fissures of different sizes. According to the spring discharge attenuation curves, there are at least three medium types in the aquifer, which account for 45.7%, 34.2%, and 20.1% of the total groundwater. Pearson correlation analysis shows that the main sources of the solute in the QJY are carbonate minerals (mainly calcite and dolomite), evaporites (mainly gypsum and sylvite), celestite, and strontianite. Anthropogenic activities have less impact on groundwater solutes. Although carbonate minerals dominate the hydrochemistry, the changes in hydrogeochemical behavior caused by rainfall may come from gypsum, which is supported by the ion concentrations. At the early rainfall stage, Ca2+ concentration increased from 42.9 to 45.6 mg/L, followed by the SO42− from 15.2 to 16.6 mg/L. When the discharge increased to the maximum (2320 L/s), Ca2+ and SO42− showed opposite trends, decreased to 39.7 mg/L and 10.4 mg/L, respectively. The results also suggest that carbonate rocks and evaporites have important roles in hydrochemistry. The contributions of these three end-members were quantified based on the law of mass conservation. The proportions of carbonate weathering and evaporite weathering were 83.4% (85.2–80.3%) and 11.6% (6.9–18.0%), respectively, and rain was 5.0% (0.1–10.4%). These results were integrated into a hydrogeological conceptual model that explains the hydrogeochemical processes, including rock weathering, piston, and dilution effects caused by rainfall. The proposed conceptual model helps to improve the understanding of hydrogeochemical processes and chemical weathering in karst areas.


Introduction
Groundwater in karst aquifers is a significant drinking water resource, supplying approximately 25% of the world's population (White 2002;Bakalowicz 2005). Karst aquifers are highly heterogeneous due to irregular connections of karst pipelines and fissures (Herman et al. 2009;Frank et al. 2019). Rainwater is the primary source of karst groundwater, so karst groundwater systems are sensitive to rainfall, especially in the karst region of South China (Hu et al. 2008;Kurtulus and Razack 2010;Luo et al. 2017). Therefore, an intensive study of the hydrological and hydrochemical variations under the influence of rainfall is helpful for the conservation and management of karst groundwater resources (Zhao et al. 2010;Li et al. 2016).
The hydrogeochemical processes in karst groundwater include dissolution, mixing, and cation exchange (Wu et al. 2009;Zhao et al. 2010). In the present studies, the factors affecting hydrogeochemical processes include the mineralogical composition of the aquifers, chemical weathering, climate, and groundwater residence time (Umar et al. 2001;Liu et al. 2007). Minerals such as calcite, dolomite, and gypsum dissolve during rainfall infiltration with the highest dissolved load (Ma et al. 2011). CO 2 gas also dissolves during rainfall infiltration, while significant outgassing is in the discharge areas (Liu et al. 1995). It can be said that rainfall is an essential factor that drives karst hydrogeochemical processes.
Rainfall events can stimulate a variety of changing behaviors in hydrodynamics and hydrogeochemistry. High-resolution monitoring can reflect the subtle dynamics of recharge behavior and explain hydrogeochemical processes (Pu et al. 2011;Gassen et al. 2017). Using high-resolution monitoring methods in the study of the karst water system's response to rainfall can achieve better results. In the study of karst underground rivers in southwest China, Jiang et al. (2018) found that rainfall transported sulfuric acid and nitric acid introduced by urbanization to groundwater, which changed the hydrological process of groundwater and interfered with the carbon cycle.
Many studies characterized groundwater-bearing medium by revealing the attenuation of spring discharge during rainfall . The studies of Wang et al. (2021b) and Chang et al. (2021) showed that karst development could be determined based on spring discharge attenuation, and the proportion of different water-bearing mediums can also be calculated. The hydrogeological characteristics of groundwater in the different mediums are prerequisites for hydrogeochemical behavior (Luo et al. 2016c). Fresh rainwater enters underground and is efficiently recharged and transported through pipelines. The spring is a mixture of fast flow and matrix flow, and the groundwater in the matrix will be collected in the pipelines (Perrin et al. 2003;Jiang et al. 2018). The proportion of fast flow and matrix flow mixing in groundwater determines ion concentration and spring discharge dynamics. Therefore, the transportation of rainwater into the ground causes a covariation in ion concentration and spring discharge. Therefore, the combination of spring discharge, hydrochemistry, and stable isotopes can better understand the hydrogeochemical mechanisms in the hydrological perspective (Nisi et al. 2008;Wang et al. 2021a).
Elemental fractionation occurs during water-rock reactions, and stable isotopes can trace hydrogeochemical processes and quantify the contribution of different sources in the hydrochemical evolution (Meredith et al. 2009;Sun et al. 2021). In fact, the mixture of different potential sources, such as rain, carbonate rocks, silicate rocks, evaporites, and anthropogenic input, contributes significantly to the solutes in water (Millot et al. 2003;Moon et al. 2007;Zhang et al. 2019). However, anthropogenic input of SO 4 2− and Cl − may misestimate the effect of evaporite weathering. Water quality monitoring at QJY showed a high SO 4 2− concentration in groundwater, which is more likely to originate from gypsum (one of the evaporite minerals). Although some studies used stable isotopes to quantify SO 4 2− produced by evaporites, they did not emphasize the influence of evaporites on groundwater in karst areas that contained evaporites. In this study, high-resolution monitoring of macroelements and Sr were used to quantify the influence of evaporites and reveal the transport characteristics of evaporite minerals. The purposes of this study are to (1) understand the dynamics of hydrochemistry and karst spring discharge and its causes, (2) identify the groundwater solute sources and calculate their proportion, and (3) analyze the variations in hydrogeochemical behavior and chemical weathering under the influence of rainfall and discuss its mechanisms.

Hydrogeological background of the study area
The Qingjiang River is the second largest tributary of the Yangtze River in Hubei province, and the spring at the river's birthplace is called QJY. The QJY is located at the foot of the Qiyue Mountain on the northeastern edge of the Wuling Mountain area, with a spring outcrop height of 1220 m. The study area has a subtropical monsoon climate with an average annual temperature of 15.2 °C. There is abundant precipitation in the study area, with an average annual rainfall of 1561 mm, concentrated from April to September, accounting for 70% of the annual precipitation (Chen et al. 2012;Sun et al. 2016).
The distribution of strata in the study area is controlled by the Qiyue Fold and dips 20-40° to the east. The strata exposed from old to new are the upper Permian formation (P 2 ), the Triassic Daye formation (T 1 d), the Jialingjiang formation (T 2 j), the Badong formation (T 2 b), the Xujiahe formation (T 2 xj), and the Jurassic formation (J) (Fig. 1). The upper Permian formation (P 2 ), the Triassic Badong formation (T 2 b), the Xujiahe formation (T 2 xj), and the Jurassic formation (J) are mudstones and shales. The Triassic Daye formation (T 1 d) is carbonate rocks, while the Triassic Jialingjiang formation (T 2 j) is carbonate rocks and evaporites (Xu and Yan 2004). According to the field survey, the Jialingjiang formation is also the aquifer with the best karst development, and several caves are developed. The paleoclimatic environment during the Triassic Jialingjiang formation (T 2 j) deposition was hot and arid (Li et al. 2020). Therefore, the minerals in the Jialingjiang formation are mainly composed of calcite, dolomite, gypsum, sylvite, and minor amounts of celestite and strontianite (Wang et al. 2022).
QJY is developed in the Triassic Jialingjiang formation, one of the water sources for residents in the surrounding area, with less pollution and excellent water quality. The elevation of the recharge area is 1300-1500 m, and most of the recharge area is trough valleys, including the surrounding depressions and the mountain slope. The soil on the slopes is thin, generally less than 0.5 m, and carbonate rocks are exposed in some steep terrain; the thickness of soil in the depressions is more than 1 m, and almost all the flat areas are reclaimed for farmland, planting corn, tobacco and so on. There is an underground river entrance (LLD) at a distance of 5.4 km to the south of the QJY, which recharges the aquifer during rainfall and breaks off when there is no rain.

Hydrological monitoring
Hydrological monitoring of the QJY is conducted using an automated field device (Model 3001 LTC Levelogger, Solinst Canada Ltd.). This measuring device integrates builtin groundwater level, temperature, and conductivity sensors. Furthermore, the resolutions of the meter are 0.001 m, 0.1 °C, and 0.1 μS/cm, respectively. An automated monitoring device was installed on the river wall at the spring outlet of the QJY. The measured channel had only one source (QJY), and the spring entirely flowed into the channel. Thus, the flow in the channel and the spring were equal. Groundwater monitoring began in November 2019 and ended in November 2021 for two hydrologic years. The device automatically reads the water level, temperature, and conductivity of the spring with a frequency of 1 h. The spring discharge is calculated according to the Chézy formula, which is applied to the open channel, with the standard equation Eq. (1): in which ω is the water cross-section area. It is related to the width of the channel and the water level (h). The width of the channel is 3 m; h came from the monitoring of the device, which refers to the depth from the water surface to the bottom of the riverbed. In order to calibrate the water level measured by the device, firstly, the water depth was measured several times with a ruler. Then, a barometer was used to calibrate the water level measurement error caused by the change in barometric pressure. Parameter c is the Chézy coefficient, and the standard equation is Eq. (2): in which n is the roughness. There are gravels at the bottom of the riverbed, and n is chosen as 0.035 after referring to the studies of Jiang and Li (2010) and Mohammad et al. (2021). R is the hydraulic radius, which is also related to the channel width and water level. J is the hydraulic gradient, equal to the channel slope and taken as 0.002. After combining the known parameters, the regression equation (Eq. (3)) is derived, and the river depth can calculate the discharge of the spring.
where Q is discharge (L/s), and h is monitored water level (m).
The rainfall data were obtained from a tipping-bucket rain gauge installed on the roof of a building near the QJY, with a resolution of 0.2 mm. A portable water quality parameter analyzer (PH828, SMART SENSOR) was used to measure groundwater pH (with a resolution of 0.01 pH units) in the field.

Sampling and analytical procedures
Hydrochemistry and stable isotope monitoring were conducted from 09:30 on July 10 to 07:30 on July 15, 2021, during which rainfall occurred. Samples were collected every 2 h, encrypted to 1 h after the rainfall started, and the sampling frequency resumed after the rainfall stopped for 24 h. During the monitoring process, 73 groundwater and one rainwater samples were collected and tested within 5 days. Clean polyethylene (HDPE) bottles were rinsed 3 times with the collected groundwater before collection to ensure no bubbles were left in the bottles. All samples were filtered through 0.45 μm acetate fiber membrane filters. The bottles for cation analysis were acidified to pH less than 2 with ultra-purified HNO 3 , and samples were stored in a refrigerator at 4 °C before laboratory analysis. Cations were analyzed by ICP-OES (iCAP7600, Thermo Fisher Scientific, USA), and anions were analyzed by ion chromatography (IC-2200, Thermo Fisher Scientific, USA), with a resolution of 0.001 mg/L for all ions. In the sample analyzed for anions and cations, 10% were selected for duplicate testing to control the quality of the analytical results. HCO 3 − was measured by the potentiometric titrimetric method, and each sample was titrated three times with an average error less than 5%. The normalized index of charge balance (NICB) for all samples was within ± 5%, which indicates that the sample quality is acceptable. Stable hydrogen-oxygen isotopes were measured by a liquid isotope analyzer (IWA-45EP, LGR). The results were expressed by δ 18 O (with an accuracy of 0.1‰) and δD (with an accuracy of 0.5‰) of the international standard V-SMOW (Vienna Standard Mean Seawater). The analysis of cations, anions, and stable hydrogen-oxygen isotopes was completed in the Geological Survey Experimental Center of China University of Geosciences (Wuhan). The saturation index (SI) of calcite, dolomite, gypsum, celestite, and strontianite in groundwater was calculated using the PHREEQC Interactive 3.0 software.

Methods
The inverse modeling of parameters is commonly used in the calculation of chemical weathering (Gaillardet et al. 1999). This approach is based on the assumption that the sources of solutes include atmospheric input, evaporite dissolution, and weathering of carbonates and silicates. Based on the products of different sources, a set of equations for the chemical weathering mixing model is constructed. The main purpose of the model is to calculate the contribution of different sources to the dissolved load. The model consists of several mass budget equations, including Na-normalized elements of the major elements (Ca/Na, K/Na, Mg/Na, Cl/ Na, HCO 3 /Na, SO 4 /Na) and Sr isotope ratios. The mixing equations for the general form and Sr isotopes are as follows: where i is the rainwater, evaporites, silicate, and carbonate rocks. α i are the mixing ratios of different sources (the sum of α i is 1).
The advantage of the model is to consider the errors that exist for each parameter used in the model. The error reflects the knowledge we have of the parameter. The purpose of inverse modeling is to make all equations compatible (by adjusting the prior range of the parameters) and obtain new parameters (called posterior parameters) that are improved after the calculation (Gaillardet et al. 1999). The errors are reduced by iterative computation, making the results more accurate. However, the traditional inverse model has some inadequacies in the application of this study. First, rainfall causes flow path changes in the QJY, but the geological conditions remain constant overall. It is known that the Sr isotope is relatively stable and can better reflect the solute source. Therefore, Sr isotopes have been widely used in some basin studies to identify different chemical weathering sources (Gaillardet et al. 1999;Moon et al. 2007). Generally, the indicators may differ for the samples collected at different times, with the underlying causes lacking exploration.
With these aspects considered, the analysis and calculations were adjusted in this study. The classical hydrogeochemical method and the parametric inversion model were combined first (Fig. 2). Then, an attempt was made to simplify the set of equations for chemical weathering mixing in the inversion model. This improves the algorithm for calculating the parameter in the inverse model and uses the major elements to quantify different sources.

Temporal variations of spring discharge, temperature, and conductivity in the QJY
High-resolution monitoring (2019.11-2021.11) of spring discharge in the QJY showed that spring flows responded rapidly to rainfall events. Spring discharge increased within 3-7 h after the rainfall began and recovered after 5-7 days  . 3). The increase in spring discharge at the QJY generally occurred after continuous rainfall or rainfall intensity larger than 8 mm/h. It is obvious that there are seasonal variations in spring discharge, which is much higher in summer (from June to August) than in winter (from December to February). The spring discharge in summer remained at approximately 400 L/s and reached a maximum (6030 L/s) after the rain (July 16, 2020). In contrast, the spring discharge is about 200 L/s, and the maximum is only 1774 L/s (January 7, 2020) in winter with less rainfall. Groundwater temperature in the QJY was consistent with seasonal variations, with an average temperature of 11.7 °C in winter and 12.4 °C in summer. It should be noted that the influence of rainfall or snow on groundwater temperature cannot be ignored, which recharges warm in summer and cold in winter. Temperature variations on the rainfall event scale were observed in the monitoring of the QJY spring. Summer rainfall caused an increase in groundwater temperature, and winter rainfall (or snow) caused a decrease.
The monitoring results showed that seasonal variations and rainfall were responsible for the influence of spring temperatures. Overall, the spring temperature fluctuations were relatively small, less than 1.5 °C. The EC at the outlet of the QJY was ~ 200 μs/cm and lowered in summer than in other seasons. In contrast to the spring flow, the EC exhibited a rapid decrease followed by a slow increase after the rainfall started, and the EC decreased to 120 μs/cm under the influence of continuous rainfall (maximum rainfall intensity of 13 mm/h) from July 28 to August 7, 2020. The rapid decrease of EC was caused by rainwater entered underground through ponors or fissures and diluted groundwater. However, the response of EC to rainfall was 4~5 h later than that of spring discharge, which was attributed to the difference in pressure conduction and solute transportation.

Responses of spring discharge, hydrochemistry, and stable isotopes to rainfall in the QJY
Hydrochemistry and stable isotope monitoring began at 09:30 on July 10, 2021, and ended at 07:30 on July 15, 2021, during which two rainfall events occurred. There was no rainfall for over a week before the monitoring, so the variations in the QJY were only influenced by these two rainfall events. Figure 4 shows the spring discharge, hydrochemistry, and stable isotope variations in the QJY under the influence of rainfall. The first rainfall lasted from 16:30 to 20:30 on July 11, with a cumulative rainfall of 28.8 mm, and the maximum rainfall intensity was 18.2 mm/h. After the rainfall stopped, the spring discharge began to increase and reached its peak (1763 L/s) 3 h later. The second rainfall occurred from 06:30 to 11:30 on July 12, with a cumulative rainfall of 30.2 mm. Compared with the first rainfall, the intensity of the second rainfall was more even, with a maximum intensity of only 7.6 mm/h. Four hours after the second rainfall started, the spring discharge increased again and reached the second peak (2320 l/s) 9 h later. The spring discharge attenuation process lasted longer and took 55 h to return to the pre-rainfall level.
As shown in Fig. 4, the EC responded earlier than the discharge, with a rapid decline from 248 μs/cm to 217 μs/ cm at the end of the first rainfall. At first, the variations of EC were exactly opposite to the discharge. With the arrival of the second rainfall, EC experienced two declines. After the EC reached its minimum, it recovered to the pre-rainfall level within 38 h. Although the temperature of the QJY increased after the rainfall, it was not significant. The ion concentrations of Ca 2+ , SO 4 2− , and Sr 2+ responded similarly to the rainfall and showed a peak after the rainfall started. After the first rainfall stopped for 3 h, these ion concentrations decreased, with the Ca 2+ concentration decreased from 45.2 to 40.9 mg/L, SO 4 2− decreased from 16.57 to 11.59 mg/L, and Sr 2+ decreased from 0.312 to 0.245 mg/L. After the second rainfall occurred, they reached their minimum within 24 h and slowly recovered to the background values. During the whole monitoring, the concentrations of Mg 2+ , Na + , K + , Cl − , and NO 3 − remained relatively stable. Moreover, it is notable that the fluctuations of K + and Cl − seem to be correlated. The piper diagram (Fig. 5) shows that the groundwater hydrochemical type is the Ca-HCO 3 type. The hydrochemical type does not change with time, and the milligram equivalent percentage of Mg decreases, while SO 4 first decreases and then increases.
The temporal evolution in δD and δ 18 O at the QJY during rainfall is shown in Fig. 4, with δD ranging from Similarly, δD decreased twice under the influence of rainfall. The first decline occurred between the two rainfall events, from − 50.87 to − 51.67‰. The second occurred after the second rainfall stopped for some time, after which δD remained stable. Additionally, δ 18 O and δD decreased from − 8.03 to − 9.38‰ after the first rainfall event, but the second decrease did not occur. Figure 6 shows the temporal variations of hydrogen and oxygen isotopes during Fig. 4 Rainfall, discharge, EC, temperature, ion concentrations, δD, and δ 18 O of the QJY monitoring. As time passes, the distribution of points in the vertical direction gradually moves downward, indicating that δD has depleted significantly; the points are closer to each other in the horizontal direction, and δ 18 O becomes more concentrated. There are similar isotopic characteristics between groundwater and rainwater (δD = − 63.47‰, δ 18 O = − 8.65‰), suggesting that rainwater is the main source of groundwater.

Identifying the structure of the QJY karst groundwater system
High-resolution monitoring can objectively reflect the response of karst groundwater systems to rainfall events and help to determine the structure of the karst water system (Mohammadi and Shoja 2014;Chang et al. 2021). Meanwhile, continuous discharge and hydrochemistry monitoring can help us understand the relationship between hydrodynamics and hydrochemistry (Fiorillo and Guadagno 2012;Frank et al. 2019). Especially the spring discharge, its attenuation time and characteristics are controlled by the karst development and rainfall (Bakalowicz 2005;Amiel et al. 2010). Generally, spring discharge attenuation curves can identify different medium and their proportions in karst or pseudo-karst aquifers (Jukić and Denić-jukić 2015;Bon et al. 2016). The studies of McDonell (2003) and Peng and Wang (2012) showed that rainwater not only infiltrates through the surface but also directly enters into the ground through ponors and sinkholes and recharges groundwater. The soil thickness is relatively thin in the karst mountainous areas in Southwest China, and excessive rainwater caused by rainstorms cannot fully infiltrate into the subsurface, resulting in large amounts of surface water. This surface water is collected along with the topography and gathered underground through sinkholes, which provides rapid recharge (Jiang et al. 2018). In the absence of rain, spring water mainly consists of matrix flow, which is derived from the groundwater stored in the soil or fissures (Sasowsky and Wicks 2000;Frank et al. 2019).
The spring discharge in the QJY increases rapidly (usually within 3-7 h) after the rainfall stops, which indicates good connectivity from the recharge area to the discharge area, and karst is well-developed in the groundwater system. Meanwhile, the flood collected in the recharge area decreases sharply, and the spring discharge also decreases rapidly within 24 h. There were two spring discharge attenuations at the QJY during the monitoring, and they occurred after the two rainfall events stopped. The first attenuation occurred from 22:30 on July 11 to 10:30 on July 12, and the second occurred from 15:30 on July 12 to 1:30 on July 15. Here, we focus on the complete second attenuation, which can be seen in three phases (Fig. 7).
The first phase lasted from 15:30 on July 12 to 2:30 on July 13, and the duration of this phase was very short. The spring discharge varied drastically and decreased rapidly from 2320 to 1568 L/s. While Ca 2+ , SO 4 2− , and Sr 2+ concentrations decreased slowly and reached their minimum value at the end of this phase. This phase reflects  the rapid transport of groundwater through well-connected pipelines and fresh groundwater discharge before the sufficient reaction of water-rock interactions. The second phase lasted from July 13, 2:30, to July 13, 18:30, with the spring discharge slowly decreased to 1026 L/s. At this time, the spring discharge was still higher than that before the rain, and the ion concentrations remained low. This phase reflects groundwater discharge from the wide fissures, and the water-rock interactions enter a brief state of relative equilibrium. The third phase lasted the longest, from 18:30 on July 13 to 1:30 on July 15. During this phase, spring discharge attenuated at the slowest speed and ion concentrations recovered slowly. Fissures provide better storage of groundwater, so the trailing of discharge attenuation is related to the development of fissures in the aquifers (Luo et al. 2016b). In summary, the QJY karst groundwater system shows multiple mediums. In addition to pipelines and wide fractures, there also exist fissures with relatively poor connectivity.
According to the spring discharge attenuation curve integral calculation, the accumulative discharges of the three stages were 52,866 m 3 , 39,629 m 3 , and 23,269 m 3 . The results showed that groundwater from karst pipelines accounted for 45.7% of the total, while wide fissures occupied 34.2%, and poorly connected fissures accounted for 20.1%.

Source of solute in groundwater
The major sources of solutes in water include rock weathering, soil leaching, atmospheric inputs, and anthropogenic activities (Soulsby et al. 2007). Although the response of spring discharge indicates that groundwater primarily comes from atmospheric inputs, the influence of rock weathering on hydrochemistry is much more significant (Fig. 8). According to the Pearson correlation analysis of hydrochemistry ( Fig. 9), the following conclusions were obtained: (1) Discharge is negatively correlated with most of the indicators, indicating that the dilution effect is significant to the hydrochemistry in the QJY. (2) Ca 2+ has a positive correlation with Mg 2+ , HCO 3 − , and Sr 2+ , indicating that carbonate weathering dominates the hydrochemistry. A similar conclusion can be drawn based on the ion ratio relationship (Fig. 10). (3) Sr 2+ has a strong positive correlation with Ca 2+ , HCO 3 − , and SO 4 2− , suggesting that Sr 2+ in groundwater originates from celestite and strontianite. (4) Except for SO 4 2− , some pollution-related ions (NO 3 − and Cl − ) were less correlated with discharge, suggesting that the variation of SO 4 2− concentration during the rainfall event was not caused by human activities. (5) SO 4 2− has a positive correlation with Ca 2+ , Sr 2+ , HCO 3 − , and Cl − , indicating that SO 4 2− is probably derived from rock weathering.
As an important alkaline soil element, Sr is sensitive to hydrological dynamics. Mineral dissolution provides Sr 2+ to groundwater, and its concentration is determined Fig. 7 Spring discharge attenuation curve in the QJY (2021.7.10 9:30-2021.7.14 7:30) Fig. 8 Gibbs plots of groundwater samples by the groundwater residence time, so Sr can better reveal water-rock interactions (Bouchaou et al. 2008). After analyzing the curves of spring discharge, Ca 2+ , SO 4 2− , and Sr 2+ concentrations, we found that these ions reached their peak before the discharge started to increase. It means that the responses of Ca 2+ , SO 4 2− , and Sr 2+ are faster than the spring discharge. Rainwater infiltrated and pushed the groundwater stored in the surface karst zone and nearsurface fissures into the pipelines. The "old groundwater" was discharged first, after which the spring discharge increased rapidly. When the spring discharge reached its maximum, the ion concentrations also reached their minimum, which was attributed to the dilution of groundwater by fresh rainwater. Continuous rainfall makes groundwater velocity faster, resulting in shorter groundwater residence time in the aquifers. Hence, the water-rock interactions were insufficient, and the concentrations of Ca 2+ , SO 4 2− , and Sr 2+ in the QJY decreased (Fig. 11). It is notable that "old groundwater" was not observed during the second rainfall, and the ion concentrations continued to decline. Fig. 9 Heatmap of the correlation coefficient matrix (*p ≤ 0.05 and **p ≤ 0.01) Fig. 10 Mixing diagram of Na-normalized molar ratio in the dissolved phase of groundwater (the partition in Fig. 10 is adapted from Gaillardet et al. 1999) After the rainfall stopped, the ion concentrations slowly recovered to the background values within 72 h. During the recovery process, the variations of these ions' concentrations showed opposite characteristics to the spring discharge. In summary, the variations of Ca 2+ , SO 4 2− , and Sr 2+ concentrations were associated with the gypsum and celestite in the carbonate aquifers. In the case of the QJY, Ca 2+ and SO 4 2− also reflected the degree of water-rock interactions when human pollution was not severe. Besides Sr 2+ , Ca 2+ , and SO 4 2− can also reveal the water-rock interactions in the karst aquifers that are relatively rich in gypsum and celestite.
The studies of Agrawal et al. (1999) and Jeong (2001) showed that Cl − and NO 3 − originated from agricultural activities. After leaching through the soil and rainwater, it recharges to groundwater. There was no significant correlation between Cl − and NO 3 − in the QJY, while Cl − and K + showed a stronger correlation. Previous studies have reported the presence of sylvite in the evaporite strata of the study area, and it is inferred that sylvite provides Cl − to the groundwater (Li 1988). However, Cl − and K + only fluctuated during rainfall, indicating that the dissolution velocity of sylvite did not vary with hydrodynamics.
The δD in groundwater showed a decreasing trend and converged to rainfall (δD = − 63.47‰), which indicated that rainfall has a great influence on groundwater. Carbonate rocks are oxygen-rich rocks, and groundwater is enriched in δ 18 O after flowing through carbonate rocks, but δD does not vary appreciably (Luo et al. 2016a). When rainwater infiltrates underground, fresh groundwater dissolves the minerals in the aquifers. Simultaneously, isotopic fractionation occurs between groundwater and karst aquifers. Therefore, the δ 18 O in the groundwater of the QJY increased after both rainfall events.

Evolution of water-rock interactions during rainfall
The previous analysis showed that the QJY karst water system has multiple mediums, which correspond to different groundwater residence times. Climate altered groundwater flow paths, so spring discharge and hydrochemistry reflected groundwater from different mediums with time (Lee and Krothe 2001). Besides, precipitation and recharge also affect the hydrogeochemical processes in the aquifers (Musgrove et al. 2010).
The saturation indexes of calcite, dolomite, gypsum, celestite, and strontianite were calculated to illustrate the hydrogeochemical processes and water-rock interactions in the groundwater, and the results are shown in Fig. 12. In general, calcite dissolved "supersaturated"; dolomite dissolved "supersaturated" or in equilibrium; gypsum, celestite, and strontianite dissolved "unsaturated." According to the variations of the saturation index, the hydrogeochemical processes exhibited three stages. The first stage occurred before the rainfall. Calcite and dolomite were in the dissolved "supersaturation" state, and the saturation indexes were relatively stable. The second stage is between the two peaks of the spring discharge in the QJY. Calcite dissolved "supersaturated," and the saturation indexes decreased; dolomite dissolved in equilibrium; gypsum and celestite dissolved "unsaturated," with the saturation index increased. The groundwater gradually recovered in the third stage, with the saturation indexes of calcite and dolomite increased, gypsum and celestite decreased. Calcite, dolomite, and gypsum dissolved faster in the second stage, while it slowed down in the third stage.
The spring is in low-flow conditions before the rainfall begins, and most of the groundwater stored in the fissures is discharged through the pipelines. The piston effect occurs , and Sr 2+ ) for groundwater in the study area first at the beginning of the rainfall. When the rainwater infiltrates, groundwater in the fissures of the surface karst zone is pushed into the pipelines. In addition, rainwater dissolves carbon dioxide in the soil, which leads to a decrease in pH as well as an increase in mineral solubility (Liu et al. 2004). It was observed that the pH decreased from 8.13 to 8.07 after the rainfall began, which supported the contribution of soil CO 2 . Therefore, both "old groundwater" and soil CO 2 may lead to a peak in ion concentrations. Furthermore, the waterrock interactions experienced unexpected alterations under the influence of rainfall. Although solutes in groundwater were controlled by the dissolution of carbonate minerals, the variations in hydrochemistry caused by rainfall were mainly derived from gypsum (one of the evaporite minerals). The spring discharge increased tens of times, while the ion concentrations decreased by only one-tenth, which indicated that the minerals continued to dissolve and became faster as the spring discharge increased. Spring discharge is one of the most important factors affecting mineral dissolution, but the speed of mineral dissolution does not increase indefinitely. When the discharge exceeds a threshold value, the dilution effect begins to be significant.

Mixing calculations for different sources
The ion concentrations vary continuously with time under the influence of rainfall, which implies that groundwater from different sources also changes (Liu et al. 2004). Based on the previous analysis, it is assumed that the sources of solutes in the QJY include rains, carbonate rock, and evaporites (gypsum and sylvite). In contrast to the research of Moon et al. (2007), silicate rocks have less influence on solute and therefore were not considered. Because Na is not susceptible to nutrient cycling in ecosystems, the selected ions were Na-normalized to eliminate the effects of runoff and evaporation (Millot et al. 2003). However, K + and SO 4 2− in the QJY mainly originate from evaporites and are affected less by biological activity, so the mass conservation equations include K + and SO 4 2-. We simplified these equations (Eqs. (6)-(12)) to reduce the errors in the calculation results caused by ions at lower concentrations. Where x i denotes the mixing proportion of Na from different sources.
In this approach, all parameters are considered to be unknown. A set of a priori parameter values was chosen to constrain different sources. Some of these parameters ∑ (x) i = 1 Fig. 12 Saturation index of calcite, dolomite, gypsum, celestite, and strontianite in the QJY for the carbonate and evaporite were obtained from Millot et al. (2003), Chetelat et al. (2008), and Zhang et al. (2021). The rainwater collected in this monitoring was used as the parameters for rain. The parameters (Z/Na) i (Z = Ca, Mg, K, HCO 3 , SO 4 , Cl) and x i from different sources were solved by the commercial version of the 1stopt 9.0 software. The inversion of the parameters was based on 511 equations, which contained 6 × 73 mass balance equations of constant elements and 73 isotope constraint equations. After a series of successive iterations, 219 x i and 13 (Z/Na) i were obtained as the optimal solutions (Table 1).
To test the sensitivity of the inverse calculation, we changed the prior range. Two scenarios were set up in the sensitivity test of the rain parameters: (1) rainwater collected in the city near the QJY on August 9, 2020, and (2) rainwater collected in this monitoring. The contributions of rain (x rain ) to the dissolved cations were 7.3% (1.9-15%) and 5.0% (0.1-10.4%) in these two scenarios. Scenario 2 was chosen for the final run after optimization. Another sensitivity test for carbonate was set up with two scenarios: (3) with Mg/Na ranging from 12 to 28, which was referenced in numerous studies (Millot et al. 2003;Moon et al. 2007), and (4) another scenario was chosen for a wider range of Mg/Na, from 7 to 28. These two scenarios exhibited similar calculations, with the contribution of carbonate weathering (x carbonate ) being 83.4% (85.2-80.3%) and 81.5% (86.0-79.0%). Considering the hydrogeochemical characteristics of the QJY, we believe that scenario 3 is more suitable for this study. Blind use of the inversion calculations may produce erroneous results. However, the sensitivity test helps us to obtain accurate results by reducing the uncertainty of background information.
The contributions of rain, carbonate, and evaporite to dissolved cations are shown in Fig. 13. Moreover, carbonate weathering is the dominant source, accounting for 83.4% (85.2-80.3%). With respect to the other two sources, the contribution of rain is 5.0% (0.1-10.4%), and evaporite weathering is 11.6% (6.9-18.0%). Although rainwater provides large amounts of groundwater to the QJY karst water system, it contributes little to the solutes. This is attributed to the lower ion concentration in the rainwater. The results show that rain contributes only 2.6-0.2% to the dissolved cations in the early stages of the rainfall. As the dilution effect becomes significant, the contribution of rain increases to 7.2%.
During the evolution of hydrochemistry and spring discharge, the contribution of evaporites to dissolved cations varied, while carbonates did not vary markedly. Coincidentally, the contribution of evaporite to dissolved cations varied with the process of different phases. The proportion of evaporite stabilized at 14.5% (12.3~16.6%) before rainfall and increased to 17.0% (14.3~18.0%) thereafter. It seemed that the percentage of evaporites decreased as the spring discharge increased. When the spring discharge attenuated to the pre-rainfall level, the contribution of evaporite also recovered. This implies that evaporite weathering may be controlled by the medium. Groundwater flows more fluently in the pipelines and wide fissures, allowing soluble evaporite minerals to dissolve on the rock surface. Therefore, evaporite is relatively absent in these mediums, whereas the proportion of the matrix flow in the QJY is higher in the low-flow conditions, and evaporite contributes more to dissolved cations.

Hydrogeochemical conceptual model
The results of the study present a conceptual model of hydrogeochemistry. It summarizes the hydrogeochemical processes and different solute sources in a typical karst groundwater system with evaporite minerals after being recharged by pulsed rainfall. Different mediums and groundwater flow characteristics in them are important factors in the hydrogeochemical evolution. The conceptual model is divided into three phases.
In the absence of rain, the spring is in low-flow conditions (Fig. 14a). In such conditions, the hydraulic head of the pipeline is lower than the surrounding matrix. Therefore, the spring discharge is mainly composed of matrix flow at this time, i.e., groundwater in the fissures. Waterrock interactions are adequate in the fractures and provide abundant dissolved ions for groundwater. The concentrations of Ca 2+ , HCO 3 − and SO 4 2− in the spring remained high, which indicated that calcite and gypsum were the main sources of solutes in the groundwater. Furthermore, carbonate minerals dissolved "supersaturated," gypsum dissolved, and the contribution of rock weathering to dissolved cations remained stable.
The fast response of the spring after pulsed rainfall indicates that groundwater flows unimpeded in the karst water system. The spring is a mixture of three water types: groundwater originally stored in the fissures, rainwater directly injected into the ground through the ponor, and "new groundwater" recharged to the fissures by rainwater. The proportion of different water types varies accordingly, which is the fundamental cause of the variation in hydrochemistry. Before the spring reaches high-flow conditions, it first experiences the "old groundwater" discharge phase (Fig. 14b). Fresh rainwater infiltrates underground and pushes groundwater previously stored in fissures into the pipelines, also known as the piston effect. This results in a sudden increase in Ca 2+ , SO 4 2− , and Sr 2+ concentrations during the early rainfall events. Apparently, the variations in hydrochemistry in these circumstances are mainly caused by gypsum and celestite. The contribution of evaporite weathering to dissolved cations also increases. At the same time, the carbonate minerals are still dissolved "supersaturated." This means that in such karst aquifers with evaporite minerals, the amount of gypsum dissolved in groundwater is the main difference between "old groundwater" and "new groundwater." In high-flow conditions (Fig. 14c), the hydraulic head in the pipeline is higher than the surrounding matrix as fresh groundwater continues to sink into the pipeline. Groundwater transports along the direction of the pipeline development, and spring discharge increases rapidly. Water-rock interactions are more intense than at other times. The dissolution of carbonate minerals is the dominant hydrogeochemical process, which may be responsible for maintaining carbonate weathering to remain stable. However, the maximum intensity restricts the ion concentration, which results in lower ion concentrations in groundwater under the diluting effect of "fresh water." Meanwhile, the contribution of rain to dissolved cations increases slightly. It is possible that strong flow can dissolve soluble minerals around the pipelines in high-flow conditions. Therefore, gypsum and other soluble minerals are relatively absent in the pipelines, which is one of the reasons for the reduction in the proportion of evaporite weathering.

Conclusion
High-resolution monitoring was conducted in the QJY, a karst groundwater source in Hubei, China. A combination of discharge, hydrochemistry, and stable isotopes was used in this study to explore the response of the karst water system to rainfall. Although carbonate minerals provide large amounts of solutes to groundwater. Gypsum, as a less abundant mineral, needs to be further investigated for its importance in the karst aquifers.
The results show that (1) the spring discharge attenuation curve can be divided into three phases, which indicates the development of multiple mediums in the aquifer. (2) As the spring discharge increased, the intensity of water-rock interactions also enhanced, but not indefinitely. (3) A more important finding is that carbonate minerals provide large amounts of solutes to groundwater, but the variations in hydrochemistry caused by rainfall are associated with gypsum and celestite dissolution. (4) The dissolution of gypsum may be related to the residence time of groundwater, and in the karst aquifers with evaporite minerals, the amount of gypsum dissolved in the groundwater is the main difference between "old groundwater" and "new groundwater." The source of SO 4 2− was distinguished during the study and is mainly controlled by the dissolution of sulfur-containing minerals rather than anthropogenic activities. The advances in the study support a hydrogeological conceptual model that explains the observed hydrogeochemical behavior.