https://orcid.org/0000-0002-8833-873X
Abstract
The hydrological dynamics and sources of karst tidal springs are difficult to capture and quantify due to fluctuations in their flow velocity. In this study, the Laolongshui (LLS) karst tidal spring in the Maocun underground river basin of Guilin City was taken as an example, with a 2-year continuous monitoring of the electrical conductivity, and water level, as well as water chemical analysis. The results showed the following: (1) discovered the variation pattern of LLS water level in different seasons, the water level fluctuates regularly on a daily scale, with a rise and fall time of 43.6 min after heavy rainfall in the rainy season, while in the dry season, it lasts for about 74–80 h. (2) Four peaks were extracted from the frequency distribution of electrical conductivity, representing the response of springwater under different rainfall conditions. (3) The annual average frequencies of the occurrence of P1, P2, P3, and P4 in terms of time are 53.82, 39.29, 6.18, and 0.72%, respectively. The results provide a new method for analyzing groundwater's source and dynamic changes in karst areas.
HIGHLIGHTS
Two years of automatic continuous monitoring illustrated the water level fluctuation of a typical karst tidal spring.
Explaining the forming mechanism of karst tidal springs.
The frequency distribution of conductivity quantified the source of karst water.
Providing a fast and low-cost new method for identifying karst water sources.
Enriching groundwater investigation methods and understanding of underground structures.
INTRODUCTION
Globally, the exposed area of carbonate rocks covers about 12% of the land area (Cao et al. 2003), and about 25% of the world's population depends wholly or partially on drinking water from karst aquifers (Dar et al. 2014). In many regions, karst water is the main source of drinking water (Vasić et al. 2019; Moldovan et al. 2020). Spring refers to the phenomenon of groundwater gushing out from the surface at the intersection of an aquifer or aquifer channel with the ground (Zhou & Li 2022), while a karst spring refers to a spring formed in karst areas due to the dissolution of soluble salt rocks, where surface water quickly emerges in the form of a spring after infiltrating the cave (Wang et al. 2008), and their water sources reflect the geological background of the spring basin and the surface and groundwater processes. The media of aquifers in karst areas show heterogeneity (Goldscheider & Drew. 2007), and due to the complexity of the medium composition, fracture development, and uneven distribution (Hartmann et al. 2014; Medici et al. 2023), the karst water migration law is more complex than that of pore water (Zhang et al. 2019). In addition, the complex physical environment and water movement patterns in karst areas also increase the difficulty of identifying the source of karst springwater (Palmer. 2010). The source of springwater recharge is mainly atmospheric precipitation, but the depth, recharge direction, and recharge size of different spring groups are not clear (Hou 2019), so clarifying the sources and dynamics of karst springwater is still a difficult problem.
Affected by special hydrogeological conditions, the hydrodynamic process and material cycle process in the karst basin of southwestern China are complex, and it is difficult to analyze the hydrodynamic process in the area by using a single water parameter. At present, researchers mainly use water chemistry and electrical conductivity to identify the source of springwater components. Electrical conductivity is an important indicator in water quality analysis, often used to evaluate soil salinization and groundwater quality (Yu et al. 2020) and also used to indirectly estimate the total concentration of ionic components in water (Chen 2000). The electrical conductivity of water is mainly determined by the type of ions dissolved in water, their concentration, and the water temperature (Gao et al. 2006). Water chemistry data are often used to characterize the internal hydrodynamic characteristics of karst aquifers (Richieri et al. 2023), using natural components or indicators dissolved in water as natural tracers (Jiang et al. 2015). Bakalowicz (Bakalowicz 1979) argued that different patterns of conductance frequency distribution (CFD) in karst springs reflect the geochemical movement of different water masses through the aquifer, and the average conductance of a single water mass or type depends on its source and residence time. The so-called frequency distribution is to divide data into several groups according to certain rules. The number of data falling within each group is called frequency, and the ratio of the frequency of each group to the total number of data is called frequency. The frequency distribution can be determined by the frequency of data occurrence within each small range. Therefore, through CFD, the contribution rates of water from different sources can be more intuitively observed. At present, this method has been widely applied by scholars from various countries in the study of karst water systems (Guo et al. 2018).
Karst tidal springs, on the other hand, are sometimes stable and sometimes swollen due to their fluctuating flow (Guo et al. 2022a, 2022b), making their hydrological dynamics and sources more difficult to capture and quantify. Karst tidal springs refer to karst springs that are periodically discharged from groundwater under the control of special karst pipelines, also known as intermittent karst springs (Yang & Tan 1992), siphon springs (Sanz et al. 2016), periodic springs (Kansou & Bredeweg 2014), and rhythmic karst springs (Xiao & Zhang 2021). Karst siphon pipes are generally developed and are the decisive factor in the formation of karst tidal springs (Mudry et al. 2014). Intermittent upward conditions of the Earth's crust, easy development of dark rivers into cascade pipes (equivalent to dark river fractures), and a lack of branches are necessary conditions for the formation of tidal springs. Studies have proven (Dong 1983) that the cause of tidal springs is not the tidal influence of the Sun and Moon, but the siphoning of karst pipe flow. Zou (1993) successfully simulated the working principle of a karst groundwater siphon circulation zone through dynamic model tests.
The karst spring of Laolongshui in the underground river basin of Maocun, Guilin, Guangxi, China is a typical karst tidal spring (Wang 2005). Also, the dynamic change law of its springwater and its response to rainfall cannot be fully explored using hydrogeological surveys, tracer experiments, or geophysical exploration in the early stage. Hydrogeological surveys are commonly used to delineate groundwater types and clarify the composition of water bearing-rock formations (Quan & Yin 2023; Wang 2023); tracer experiments and geophysical exploration are mainly used to reveal the spatial structural characteristics of karst aquifer systems (Luan et al. 2021; Guo et al. 2022a, 2022b). Thus, there is room for improvement to explore the dynamics of the karst tidal spring. The goal of this study was to propose a faster and lower-cost method for the source investigation and dynamic change analysis of water resources in karst areas, especially for karst tidal springs. To achieve this goal, the specific objectives of this study are as follows: (1) a high-resolution observation of Laolongshui spring was set up to monitor the water level and electrical conductivity for two years to clarify the dynamic changes of the springwater. (2) The dynamic changes of springwater through 2 years of water level were clarified. (3) Through 2 years of automatic and continuous monitoring of electrical conductivity, the source of the components of the springwater in Laolongshui was determined using the frequency distribution of electrical conductivity. (4) Using CFD, the water mass contribution ratio of karst pipelines in the entire hydrological year was quantified under different rainfall intensities.
OVERVIEW OF THE RESEARCH AREA
Laolongshui is located in Shanwan Village, Lingchuan County, southeast of Guilin City, Guangxi, China, 30 km away from Guilin. It has a subtropical monsoon climate (Yang et al. 2012), and rainfall is mainly concentrated from April to August every year. There are four distinct seasons. Summers are hot and humid, and heavy rainfall is frequent; autumn rainfall decreases rapidly (Cao et al. 2011). The annual average temperature is 18.6°C, and the annual rainfall amount is 1,600–2,200 mm. Rainfall is affected by monsoon activity, and the spatial and temporal distributions are extremely uneven. Summer flooding and winter drought often occur.
. Laolongshui is located on the intersection fault between the sand shale of the Yingtang Formation of the Middle Devonian and the limestone of the Donggangling Formation of the Middle Devonian, and springwater gushes out from the karst crevices of the gray-white clouds of the Donggangling Formation. 
Geological background map of the study area.
RESEARCH METHODS
High-resolution monitoring
The Laolongshui water level data were recorded using a HOBO U20L-04 automatic water level recorder, from ONSET, and the electrical conductivity data were obtained using a HOBO U24-002 electrical conductivity detector, made in the United States. The automatic monitoring interval was set to 15 min, and data acquisition and instrument calibration were carried out every month using the HOBO special software. For meteorological data collection, a HOBO U30 weather station, made in the United States, was used to measure various parameters such as wind speed, wind direction, air temperature, humidity, total solar radiation, rainfall, and air pressure. The automatic sampling interval was set to 15 min.
On-site monitoring
For this study, on-site water quality analysis of the springwater in Laolongshui was conducted from August 2020 to July 2022. At the same time, the Beidiping karst spring (in the same basin's dolomite formation), Bianyan (receiving allogenic water), and the Xiaolongbei surface stream (in the clastic rock area) were compared, at intervals of generally 30 days. A total of 11 samples were taken. A French PONSEL portable water quality analyzer was used for on-site detection of the electrical conductivity, pH, and temperature of the water. At the same time, on-site detection of the Ca2+and 
content in springwater was carried out using a German Merck portable reagent kit. Water samples were collected in 600-ml polyethylene plastic bottles and sent to the laboratory of the Institute of Karst Geology, Chinese Academy of Geological Sciences. ICP-OES was used to detect the plasma concentrations of K+, Na+, Ca2+, Mg2+, Cl−, 
, and 
, these seven ions are the most widely distributed in groundwater and largely determine the basic characteristics of groundwater chemistry. They are also the most important basis for distinguishing groundwater hydrochemical types.
Data analysis
CorelDraw2019, Origin2018, and Excel2016 were used for data plotting, processing, and statistical analysis. To obtain the saturation indices of calcite (SIc) and dolomite (SId), we used the WATSPEC program (Wigley 1977), which requires at least nine parameters: the water temperature, pH, and concentrations of seven major ions (Ca2+, K+, Na+, Mg2+, 
, Cl−, and 
).
CFD decomposition method and principle
Assuming that the same type of water source has a normally distributed electrical conductivity, in addition, the electrical conductivity distribution of a single component of karst pipeline runoff is also normal without considering its source (Massei et al. 2007). CFD represents the superposition of electrical conductivity groups, which may include a combination of the overall electrical conductivity of two or more water groups (Guo et al. 2018). The CFD shape of a hydrological year approximates a histogram, and Origin2018's statistical program is used to find the probability density function that best corresponds to the histogram. Using Origin2018's peak fitting program, the probability density function is decomposed into its normally distributed component population. Using the residual method, all patterns or peaks that make up the original CFD are determined, including peak patterns that are not visible in the original curve. Since the basic distribution is normal, it is possible to determine the average and variance of the specific electrical conductivity of each water type determined through decomposition, as well as the proportion of that water type to the entire CFD. It is worth noting that because electrical conductivity is a reflection of ion charge rather than ion concentration unless the geochemical composition of different water types is similar, the proportion of CFD represented by each peak does not represent the volume proportion of the total flow formed by each peak. The proportion of peaks reflects the contribution rate of different water masses throughout the hydrological year. Laolongshui is a surface karst spring that is sensitive to precipitation and has significant differences in electrical conductivity. The peak ratios of various CFD peaks can reflect the contribution rate of water masses in karst pipelines under different rainfall intensities throughout the hydrological year.
RESULTS
Daily updates
Dynamic changes in the daily rainfall and water level.
Hydrochemical analysis
Beidiping is a karst spring originating from the dolomite area, with a maximum electrical conductivity of 533 μS/cm. Although Bianyan is located in the dolomite area, it has been supplied by allogenic water from upstream Xiaolongbei for a long period, with a maximum electrical conductivity of 228 μS/cm. The electrical conductivity of Xiaolongbei, located in the non-karst area upstream of Laolongshui, is generally between 11 and 57 μS/cm, while the electrical conductivity of rainwater in the study area is between 5.3 and 11.5 μS/cm. From Table 1, it can be seen that the variation characteristics of the average electrical conductivity values of each spring and rainwater are as follows: Beidiping (462.56 μS/cm) > Laolongshui (296.46 μS/cm) > Bianyan (174.84 μS/cm) > Xiaolongbei (26.86 μS/cm) > Rain (8.10 μS/cm). The variation characteristics of the average 
concentration are as follows: Beidiping (5.23 mg/L) > Laolongshui (3.41 mg/L) > Bianyan (1.95 mg/L) > Xiaolongbei (0.29 mg/L) > Rain (0.02 mg/L). The variation characteristics of the average Ca2+ concentration are as follows: Beidiping (90.49 mg/L) > Laolongshui (43.67 mg/L) > Bianyan (24.19 mg/L) > Xiaolongbei (2.64 mg/L) > Rain (0.44 mg/L). The variation characteristics of the average Mg 2+ concentration are as follows: Laolongshui (18.62 mg/L) > Beidiping (12.16 mg/L) > Bianyan (9.10 mg/L) > Xiaolongbei (1.30 mg/L) > Rain (0.39 mg/L).
Hydrochemical comparison of springwater with different geological backgrounds and rainwater
 |
| |
, accounting for 72–94% of the total number of anions. According to Schöeller's classification, Xiaolongbei's hydrochemical types are HCO3-Ca-Mg, and HCO3- (Na + Ca). The cations of Bianyan are mainly Ca2+and Mg2+, accounting for 68–81% and 18–30%, respectively, while 
is the main anion, accounting for 97–98%. According to Schöeller's classification, the hydrochemical type of Bianyan is HCO3-Ca-Mg. The main cations of Laolongshui are Ca2+ and Mg2+, accounting for 65–81% and 19–34%, respectively, while the anion is 
, accounting for 94–98%. According to Schöeller's classification, the hydrochemical type of Laolongshui is HCO3-Ca-Mg. Bediping hydrochemistry mainly comprises Ca2+ and 
, representing 85–91% cationic charge and 96–99% anionic charge, respectively. 
Piper three-linear diagram of the hydrochemistry of springs in different geological backgrounds and rainwater.
Piper three-linear diagram of the hydrochemistry of springs in different geological backgrounds and rainwater.
Quantitative analysis of the CFD
Proportions of peak area, annual average flow, and total runoff
| Year | P1 | P2 | P3 | P4 | Annual average flow (l/s) | Annual total runoff (100 million liters) |
|---|---|---|---|---|---|---|
| 2020.8–2021.7 | 47.28% | 49.65% | 2.26% | 0.81% | 6.73 | 2.12 |
| 2021.8–2022.7 | 60.35% | 28.93% | 10.09% | 0.62% | 9.67 | 3.05 |
| Average | 53.82% | 39.29% | 6.18% | 0.72% | 8.20 | 2.58 |
| Year | P1 | P2 | P3 | P4 | Annual average flow (l/s) | Annual total runoff (100 million liters) |
|---|---|---|---|---|---|---|
| 2020.8–2021.7 | 47.28% | 49.65% | 2.26% | 0.81% | 6.73 | 2.12 |
| 2021.8–2022.7 | 60.35% | 28.93% | 10.09% | 0.62% | 9.67 | 3.05 |
| Average | 53.82% | 39.29% | 6.18% | 0.72% | 8.20 | 2.58 |
CFD for Laolongshui.
CFD for Beidiping and Bianyan in 2021.
Analysis of the causes of tidal springs
Schematic diagram of the siphon structure (cited from (Zou 1993)). (1) The water supply pipeline, (2) relatively closed karst reservoir, (3) karst siphon pipeline, and (4) karst fissure lower than the siphon pipeline; H1 is the starting water level of the siphon and H2 is the stopping water level of the siphon; Q1, Q2, Q3, and Q4 are the flow rates during the operation of each pipeline; (a) is the lowest point at the inlet of the siphon, (b) is the highest point of the siphon, and (c) is the dew point of springwater, at the outlet of the siphon; h1 represents the height difference between points (a) and (b); and h2 represents the height difference between points (b) and (c).
Schematic diagram of the siphon structure (cited from (Zou 1993)). (1) The water supply pipeline, (2) relatively closed karst reservoir, (3) karst siphon pipeline, and (4) karst fissure lower than the siphon pipeline; H1 is the starting water level of the siphon and H2 is the stopping water level of the siphon; Q1, Q2, Q3, and Q4 are the flow rates during the operation of each pipeline; (a) is the lowest point at the inlet of the siphon, (b) is the highest point of the siphon, and (c) is the dew point of springwater, at the outlet of the siphon; h1 represents the height difference between points (a) and (b); and h2 represents the height difference between points (b) and (c).
Daily changes in the spring water level.
DISCUSSION
Analysis of water sources in tidal springs
Massei et al. (2007) analyzed four hydrological year data collected in Barton Springs, Austin, Texas, regarding the CFD and found that although the overall shape of CFD varies year by year, it can always be divided into the same group of normally distributed populations. Each population represents the type of water generated by a specific aquifer functional pattern. Therefore, to identify the proportions of water from different sources in springwater, the frequency distribution method in statistics is used to process electrical conductivity data. Guo et al. (2018) found through experiments that the decomposition results of CFD for four hydrological years at the same research point were similar, as they could all be decomposed into a series of similar peaks, indicating that the annual electrical conductivity distribution in the same region is roughly similar. The frequency distribution curve of electrical conductivity in Figure 6 shows four peaks, and the area of the peaks reflects the amount of electrical conductivity data within a certain range. The larger the peak, the greater the data volume, indicating a higher proportion from this water source. Each peak represents a water source that is influenced to varying degrees by factors such as the structure of the aquifer system, rainfall, and the intensity of water–rock interaction.
Previous studies have shown that Beidiping is a karst spring without an allogenic water supply (Huang et al. 2017). The CFD analysis results for Laolongshui and Beidiping are similar but differ significantly from those for Bianyan, which does have an allogenic water supply, indicating that the water supply mode of Laolongshui is the same as that of Beidiping. Meanwhile, the hydrochemistry of Laolongshui is closest to that of Beidiping, indicating that Laolongshui has no allogenic water supply. In addition, Huang et al. (2017) added tracers to Xiaolongbei upstream of Laolongshui but did not receive them in Laolongshui, which also confirmed this. Therefore, we believe that there are two main Laolongshui water source components: one is the water stored in the karst aquifer system, and the second is rainwater, which only arises from differences in seasonal rainfall, resulting in different degrees of dilution of springwater and different water–rock interaction times. Through comprehensive consideration of the hydrogeological background, rainfall, and upstream land use types in the Laolongshui area, we believe that P1 corresponds to the part where the water storage space inside the karst is slowly released during the dry season without rainfall. This part has a slow water flow rate and, after sufficient water–rock interaction has the highest electrical conductivity. The electrical conductivity of P2 is higher than that of P3 but lower than that of P1, indicating a prolonged and small amount of rainwater infiltration during the rainy season. The flow rate and velocity of this water are relatively small, and it has a smaller dilution effect on the internal karst water, so its electrical conductivity is also higher. P3 represents the high summer rainfall, when rainwater quickly infiltrates through the upper karst fissures, resulting in a high flow rate and a significant dilution effect on the water in the internal pipelines. Additionally, the water–rock interaction time is short, resulting in lower electrical conductivity. P4 represents the rapid infiltration of a large amount of rainwater within a short time during summer rainstorms. The internal water is diluted by a large amount of low-electrical conductivity rainwater within a short time, and the water–rock interaction time is very short, so the electrical conductivity of this part is the lowest. The electrical conductivity of the allogenic water from Xiaolongbei is the same (Table 1). Due to the short duration of rainfall at this level, the proportion of this water in the overall source is also the lowest. From the perspective of electrical conductivity values alone, it is easy to believe that this water is supplied by upstream allogenic water. However, by comparing its electrical conductivity distribution map and tracer experimental results, this source was excluded.
Analysis of the causes of wave peaks in the CFD curve
The electrical conductivity of groundwater is significantly influenced by karst hydrogeological conditions, which can comprehensively reflect the regional hydrodynamic conditions and lithological characteristics. The electrical conductivity of karst springwater flow has always been a fundamental variable for characterizing karst systems (Chang et al. 2021); it reflects the total amount of dissolved solids in the water. In karst water systems, dissolved solids are mainly controlled by the balance of calcium carbonate (Massei et al. 2007). Electrical conductivity reflects the degree of interaction between groundwater and rock, and its dynamic characteristics and distribution patterns are an important indicator (Smith et al. 2004). Changes in spring electrical conductivity provide information on local recharge and conduit flow in karst aquifers (Birk et al. 2004). The CFD curve can be decomposed into different peaks, which are determined by factors such as the pipeline size, the characteristics of each water source component, and the water area at the spring mouth (Guo et al. 2018). For example, the Lingshui Karst Spring in Nanning, Guangxi, has multiple water sources with significant electrical conductivity differences, and its CFD decomposes into four peaks without sufficient mixing. The Guancun underground river has a huge pipeline space, and its CFD has only one peak (Guo et al. 2018) because there are huge karst pipelines and caves upstream of the underground river with enough space to fully mix water from various sources (Huang et al. 2017). Similarly, Bianyan is the entrance where Xiaolongbei's surface water flows into the underground river in Maocun, mainly through fissure flow without sufficient mixing of various water sources. The spring area is equivalent to that of Laolongshui, but due to Bianyan's ability to continuously receive an allogenic supply with significant differences in electrical conductivity, its CFD is not as large as that for a karst spring with changes in rainfall and can only be decomposed into three peaks.
The electrical conductivity changes for Laolongshui and Beidiping are similar to those for Lingshui, with four peaks in the decomposition. The changes are mainly affected by rainfall, and the rainfall in the study area shows obvious seasonal characteristics. Therefore, the peaks decomposed from the Laolongshui CFD curve represent the degree of influence of different rainfall conditions on spring electrical conductivity.
CONCLUSION
Laolongshui is a typical karst tidal spring, and its hydrological process and formation are related to its unique geological structure. Through long-term monitoring, the dynamic changes of Laolongshui have been grasped, and the tidal patterns and causes have been clarified. This provides a certain data foundation for the development and utilization of groundwater resources and also has a certain significance for responding to geological disasters.
CFD is a more realistic and robust method for analyzing long-term electrical conductivity data, and it is easier to implement due to its low cost and ease of operation. The CFD of Laolongshui in two hydrological years has decomposed into four peaks (P1–P4), with significant differences in electrical conductivity values and peak positions, reflecting the degree of influence of different rainfall on water masses in karst pipelines. The source of the springwater in Laolongshui was quantified through CFD, mainly from the water stored in the karst aquifer system and rainwater. The electrical conductivity of the springwater is regulated by rainfall. Among the four peaks decomposed from the frequency distribution curve of electrical conductivity, P1 represents the slow release of water storage space inside the karst during the dry season without rainfall, and P2-P4 represents the dilution degree of springwater electrical conductivity or the duration of water rock interaction under different rainfall conditions. In addition, comparing the frequency distribution characteristics of the electrical conductivity of karst springs that receive and do not receive allogical water recharge in the same basin, it is confirmed that there is no allogenic water recharge in Laolongshui. This study found that CFD can effectively identify water masses with significant differences in karst water systems. This article suggests that this method can also help to identify the water source components of karst springs injected with different allogenic water sources.
ACKNOWLEDGEMENTS
This study was funded by National Natural Science Foundation of China (No. 2022YFF1300705, 42277077), Natural Science Foundation of Gunagxi (No. 2022GXNSFAA035569, Guike-AD21196005, 2021GXNSFBA075013), the Project of the China Geological Survey (No. DD20230547), and the Science and Technology Base and the Guilin Science Research and Technology Development Plan Project (2020010905).
DATA AVAILABILITY STATEMENT
All relevant data are included in the paper or its Supplementary Information.
CONFLICT OF INTEREST
The authors declare there is no conflict.
