Groundwater is an important water resource in arid and semi-arid regions, and the widespread over-exploitation of groundwater is a global issue (Foster and Chilto et al., 2003; Wada et al., 2012; Dollet et al., 2012; Gleeson et al., 2015). The global groundwater exploitation rate was estimated at about 1500 km³ per year, which was much larger than the natural recharge rate of groundwater (Dollet et al., 2012). This problem has become more severe with increasing population growth and lack of understanding of the regional water cycle. Surface water and groundwater are important components of the water cycle in a basin (Kalbus et al., 2006). Clearly, understanding of the mutual transformation between surface water and groundwater is critical to utilize scientifically water resources in any basin.

The measurements of water isotopes (δ¹⁸O, δD and³H) combined with hydrochemical evidences are effective ways to study the water cycle. The tracing techniques play an important role in the study of groundwater recharge sources and mutual conversion between surface water and groundwater in a basin (Perry et al., 2009; Huang et al., 2010; Li et al., 2015; Yao et al., 2016; Xu et al 2017). Firstly, hydrogen-oxygen stable isotopes (δD and δ¹⁸O) are widely used in the research of hydrological cycles in natural and urban environments for better water resources management (Yi et al., 2010; Yuan and Mayer 2012; Gibson 2016; Tipple et al., 2017). Furthermore, radioisotopes (³H) are commonly used in groundwater dating which is important for groundwater management. They are part of water molecules (HTO) and are effective tracers for modern hydrological processes over the past 60 years. Groundwater age can be used to determine groundwater recharge areas, recharge rates, estimate velocities and describe flow characteristics (Liu et al., 2014; Kamtchueng et al., 2015; Cao et al., 2018). Groundwater age is also an effective and informative way to map groundwater renewability and to describe resource attributes, thus facilitating the sustainable use of groundwater resources (Huang et al., 2017; Zhang et al., 2017). ³H is also used to calculate groundwater recharge rates and renewability (Morgenstern and Daughney, 2012; Yangui et al., 2012; Gusyev et al., 2016; Ansari et al., 2017).

Until now, previous research on groundwater utilization in China mainly concentrated on the North China Plain and the arid regions in the northwest. Zhangjiakou is located at the intersection of the North China Plain, Mongolian Plateau and Loess Plateau. It is also an important ecological conservation area and water source for Beijing and Tianjin, with the function of balancing natural ecology (Tian et al., 2012). The water resources of Zhangjiakou are mainly replenished by atmospheric precipitation, decreasing rainfall during the last 20 years caused repeated droughts and serious water resource crisis in this area. In recent years, with the rapid development of social economy, the gap between the demand for water and the shortage of water resources has become increasingly tremendous. The utilization of surface water resources has been unable to meet the development needs of Zhangjiakou city, thus increasing the exploitation and utilization of groundwater. According to the survey, at present, 70% of the groundwater in the Zhangjiakou area is used for agricultural irrigation purposes, most of which is the results of Cao et al. (2019). The shortage of water resources and the irrational utilization have become the most important challenge to the sustainable development of Zhangjiakou’s economy.

However, only a few studies have been done on the groundwater recharge and circulation processes in the Zhangjiakou aquifer system where is mainly in the upper reaches of the Yongding River, which provide a broad understanding of the mechanism of local recharge sources and zones (Chen et al., 2017; Feng et al., 2013; Hou et al., 2019). Even the previous studies of Wang et al. (2015) and Li et al. (2017) in the upper Yongding river basin did not reach a concensus how fast the groundwater is replenished in this area, Thus, in order to understand the recharge mechanism and spatial characteristics of the aquifer as well as the relationship between surface water and groundwater in the study area, we focused on a region-wide survey for groundwater recharging characteristics in Zhangjiakou, including Zhangbei, Yanghe and Sanggan river basins, based on an integrated tracing approach. This case study was specifically relevant to the sustainable development of the northwestern part of Beijing-Tianjin-Hebei region.

Accordingly, the objectives of this study were (1) to investigate water chemical composition characteristics and water isotope spatial composition characteristics of surface water and groundwater, (2) to calculate the residence time of the groundwater in Zhangjiakou area, (3) to estimate the main recharge sources of groundwater and the transformation relationship between surface water and groundwater. The findings will provide further understanding of the regional water cycle to benefit the rational utilization of water resources in the northwestern part of Beijing-Tianjin-Hebei region.

Zhangjiakou City is located in the northwestern part of Hebei Province, bordering Inner Mongolia Autonomous Region in the north and west, bordering Shanxi Province in the southwest, and connecting Chengde city, Beijing municipality and Baoding city in the east and southeast. The geographical range is between 113°50°-116°30° E and 39°30°-42°10°N. Zhangjiakou city is 289.2 km long from north to south and 216.2 km wide from east to west. It covers an area of about 36,965 km², including 23,149 km² in the Bashang plateau and 13,816 km² in the Baxia plain. The terrain is relatively flat with an elevation of about 1400 m. The area beyond Bashang plateau accounts for about two-thirds of the total study area. There are many large mountains with altitudes ranging 1000-2000 m. The mountains in the area are formed by a structural cut to form beaded inter-montane basins. The altitudes of the basins range from 500 m to 1000 m. The larger ones are: Chaigoubao-Xuanhua Basin, Yuxian-Yangyuan Basin, and Zhulu-Huailai Basin. There are large rivers in each basin, and relatively fertile cultivated land is distributed in the basins (Sun et al., 2014). The study area is located in the mid-latitude area and belongs to the cold temperate continental monsoon climate. It is windy and rainy in spring and autumn, long and cold in winter, and short and hot in summer. The average annual temperature in the whole district is 5-6℃, and the rainfall mostly concentrates in July-August of each year (Shao et al., 2012). The average annual precipitation is 403.1 mm with the average evaporation of 1446.9 mm.

The Zhangjiakou area is located in the Yanliao stratigraphic belt. From the old to the new exposures, there are the Archean, Proterozoic, Cambrian, Ordovician, Carboniferous, Permian, Triassic, Jurassic, Tertiary, and Quaternary strata. The lithology revealed in the Bashang plateau is mainly Holocene alluvial deposits, Late Paleozoic granites, and ancient Paleogene Shiji group red clay; the main lithology of the Yanghe river basin is the Upper Pleistocene Malan Formation and the Late Jurassic granite, Middle Jurassic Anshan Formation, and volcanic breccia; the Sanggan river basin mainly exposes Quaternary alluvial deposits, the Middle Jurassic Anshan Formation and volcanic breccia; the main exposed lithology of the Qingshui river basin is Jurassic andesite, volcanic breccia, schist and marble.

In order to study the interaction between surface and groundwater in the study area, we determined sampling points based on existing hydrogeological maps and our field surveys (Figure 1). We collected samples from river water, precipitation and groundwater in July 2018 (rainy season), October 2018 (dry season) and April 2019 (snowmelt), respectively. We took 9 samples from the Bashang plateau (Zhangbei), 76 samples from the Qingshui river basin, 36 samples from the Yanghe river basin and 12 samples from the Sanggan river basin. A total of 133 samples were collected in this study, 52 of them are precipitation samples (they were all collected from the Chongli and Chaigoubaodong hydrological monitoring stations), 54 of them are groundwater samples and 27 are surface water samples. However, only the groundwater samples collected in the rainy season are selected to analyze the concentration characteristics of tritium isotope. The groundwater was divided into shallow groundwater (<60 m) and deep groundwater (>60 m) for sampling. Most of the groundwater samples collected in this study were from the deep groundwater due to few shallow groundwater existed. Before sampling, water was released for 30 min using a pump. All water samples were collected in pre-cleaned polypropylene bottles (20 mL for stable isotopic analysis and 500 mL for tritium analysis), and then protected from radiation. Additional parameters, such as pH, temperature (T), electrical conductivity (EC, automatic temperature compensation to 25℃) and total dissolved solids (TDS) were measured in the field using French PONSEL portable multi-parameter water quality instrument. The precision reached 0.01 units, 0.1℃, and 1 μs cm^-1 for EC. HCO₃^- concentration was measured by the alkalinity meter of Merck Company in Germany. Before titration, samples were percolated with a 0.22 µm aqueous phase needle filter. Each sample was titrated for 2-3 times with an average error of < 5% and an accuracy of 0.1 mmol·L^-1.

The samples were placed in a refrigerator at 4℃ before testing. The analysis of oxygen and hydrogen isotopes (δD, δ¹⁸O) was carried out by L2140-i liquid water vapor isotope analyzer (Picarro, USA). The analysis errors of δD and δ¹⁸O were less than 0.5‰ and 0.1‰ respectively, and the results were given relative to V-SMOW (Vienna Standard Mean Ocean Water) standard. Anions (SO₄^2-, Cl^-, NO₃^-, F^-) were analyzed by Dionex ICS3000 ion chromatograph with an accuracy of 0.01 mg·L^-1, while cations (K⁺, Na⁺, Ca²⁺, Mg²⁺) were analyzed by Aquion ICS ion chromatograph (Thermo Fisher, USA) with an accuracy of 0.01 mg·L^-1. Furthermore, the tritium isotope (³H) was analyzed at the Institute of Karst Geology, Chinese Academy of Geological Sciences. The³H isotopes were electrolytically enriched and measured using the liquid scintillation counting method with Quantulus-1220 (Pharmacia LKB, Sweden). The results were reported in tritium units (TU), with an analytical precision of ± 2 TU.

The estimation of the mean residence time of the groundwater was done according to the tritium input into the groundwater and the residual tritium measured in the groundwater (Clark and Fritz, 1997). We calculated the mean residence time which was equal to the age of groundwater based on the exponential-piston flow model, which is suitable for aquifers that have regions of confined and unconfined flow model, the measured³H concentration at time t (C_t) is connected with the input of³H (C_i) via the convolution integral:

where τ represents the residence time, t-τ is the time when the water recharged, λ is the radioactive decay constant of tracers (³H, λ = 0.55764 a^-1), and g(τ) is the transit time distribution function of flow paths and residence times in the aquifer. For the exponential-piston flow model, g(τ) is defined as follows (Cartwright and Morgenstern, 2016):

The letter T is meaning of transit time (a) and f is meaning of the ratio of the total volume to the exponential volume in the above equation. When the f = 0 the exponential-piston flow model is equivalent to the piston flow model and when the f= 1 the exponential-piston flow model is equivalent to the exponential model (Maloszewski and Zuber, 1982; Zuber et al., 2005).

When the exponential-piston flow model is used to calculate the mean residence time, the unknown parameters need to be determined first, which is done by the lumped parameter method. Simulations are calibrated to fit the measured³H output composition (Jurgens et al., 2016; Chatterjee et al., 2018; Hagedorn et al., 2018). This can be accomplished by numerically integrating the convolution integral (Equation 1). For calculations we used the FLOWPC software (Maloszewski and Zuber 1996).

The physical and chemical parameters of river water and groundwater are shown in Table 1. The pH of the river water varied from 7.02 to 8.89, averaging as 8.07, which was generally weakly alkaline. The pH value of the river water during the wet season was lower than that of the river water during the dry season. The groundwater pH ranged from 6.49 to 8.37, averaging as 7.66, where the groundwater pH was higher during the dry season. In general, the pH value of river water was higher than that of groundwater. The possible reason might be that the presence of aquatic plants or algae in river water caused the pH value of water to rise (Wang et al., 2017). In this study, the TDS value of river water varied from 154 mg·L^-1 to 936 mg·L^-1, with an average of 476.76 mg·L^-1, which signifies fresh water quality. The mean value of river water TDS during the wet season was 296.88 mg·L^-1. Generally, it was lower than the salinity value in the dry season (mean value 574.25 mg·L^-1). The low alkaline and TDS values in river water samples indicate that the river water may be affected by glacial snowmelt (Wang et al., 2016). Additionally, the TDS values of groundwater varied from 139 mg·L^-1 to 1782 mg·L^-1, averaging as 525.8 mg·L^-1, which was slightly higher than the TDS values of the river. The maximum value of the TDS was found at the QJSW sampling point located in the Sanggan river basin, which is close to the salt water standard. Definitely, this was partly related to local soil salinization. Besides, TDS and EC data revealed a very good linear relationship (Figure 2) with R² = 0.94 (p<0.001), indicating that the parameters measured in the field and lab matched very well. Thus it is justified to use TDS data to represent corresponding changes of EC.

Figure 3 showed the main cations of the river water were Mg²⁺ and Na⁺, which accounted for 41.85% and 34.94% of the total amount of cations, respectively. Ca²⁺ followed with an average proportion of 21.19%. The order of cation concentrations was Mg²⁺>Na⁺>Ca²⁺>K⁺. The HCO₃^- was the dominant anion in the river water, accounting for 15.96% - 97.60% of the total anions, with an average ratio of 54.05%. Cl^- followed with an average of 21.98%. The order of anion concentrations was HCO₃^->Cl^->SO₄^2->NO₃^->F^-. Thus, the water chemistry of the river basin were mainly of the HCO₃-Mg·Na type and HCO₃·Cl-Na type.

Figure 4 depicts the triangle diagram of groundwater anion and cation in each basin of Zhangjiakou. The main cation in the Zhangbei district were Mg²⁺ and Na⁺, with the average proportion being 40.27% and 38.56%, respectively. The main anions were HCO₃^- and Cl^-, with an average proportion of 51.99% and 23.92%, respectively. The water chemistry type was mainly HCO₃·Cl-Na and HCO₃·Cl-Mg type. The dominant cations of the groundwater in the Qingshui river basin were Mg²⁺ and Ca²⁺, with an average proportion of 56.39% and 25.46%, respectively, the main anions were HCO₃^- and SO₄^2-, with an average proportion of 64.54% and 13.05%, respectively. The main prevalent chemistry type was HCO₃-Mg·Ca. The dominant cation of the groundwater in the Yanghe river basin was Mg²⁺, with an average proportion of 50.66%. Ca²⁺ and Na⁺ showed almost similar proportions with the average contents of 52.68 mg·L^-1 and 56.37 mg·L^-1, respectively. Anions were mainly HCO₃^- and SO₄^2-, with the average proportion of 67.19% and 13.70%, indicating the HCO₃-Mg type for the water chemistry. The concentration of Na⁺ and Cl^- of the groundwater in the Sanggan river basin is abnormally high. The average Na⁺ concentration is 564.85 mg·L^-1, which is the maximum in the Zhangjiakou area. The main cations are Na⁺ and Mg²⁺, with the average proportion of 44.76% and 36.11%, respectively, followed by Ca²⁺, whose average proportion is 17.90%; the dominant anion was HCO₃^-, and the Cl^- and SO₄^2- had the similar proportions of 17.40% and 17.78%, respectively. Collectively, the water chemistry type mainly expressed as HCO₃·Cl·SO₄-Na and HCO₃·Cl·SO₄-Mg type.

The values of δD and δ¹⁸O of the atmospheric precipitation exhibit notable seasonal variation (Table 2). Generally, the δD and δ¹⁸O values were at their maximum in spring, their minimum in winter, and their medium in summer and autumn. This pattern occurs mainly because in spring the water vapor source has a relatively high humidity which produces isotopic depletion. During the winter, the water vapor sources for precipitation are mainly cold and dry continental air masses (Zhu et al., 2017). The δD ranged -140.370‰ to +22.058‰, and the δ¹⁸O was -20.393‰ to -0.479‰. According to the global reference values from the International Atomic Energy Agency, the δD was -350‰ to +50‰, the δ¹⁸O was -50‰ to +10‰. In China, the δD was -280‰ to +24‰, the δ¹⁸O was -35.5‰ to +2.5‰ (Tian et al., 2001). It is shown that the δD and δ¹⁸O values of precipitation in the Zhangjiakou area were within the expected range.

Based on the δD and δ¹⁸O data of precipitation collected from June 2018 to May 2019 Local Meteoric Water Lines (LMWL) were established as δD = 7.30 δ¹⁸O+11.11 (R² = 0.96, 0.96, n = 52; Figure 5) for the study area. Compared to the GMWL (δD = 8 δ¹⁸O + 10) (Craig, 1961; Zheng et al., 1975) and the China atmospheric water line (δD = 7.9 δ¹⁸O + 8.2) (Zheng et al., 1982), both the slope (7.30) and the constant term (11.11) were similar, yet with a slightly lower slope and slightly higher constant term, showing an arid area feature.

In sub-arid regions, evapotranspiration and subcloud evaporation have obvious impacts on the stable isotopes of local precipitation (Zemp et al., 2014; Wang et al., 2016b). In general, the lower values of δD and δ¹⁸O might be caused by the recycling of evaporated moisture, while the subcloud evaporation leads to enrichment (Pang et al., 2011). Therefore, the LWML in Zhangjiakou area had the smaller slope compared with GMWL, indicating that rainfall isotopic enrichment due to subcloud evaporation over-compensates for isotopic depletion by moisture recycling.

The average values of δ¹⁸O and δD are shown in Table 1. The isotopic values of the river water samples varied from -86.53‰ to -58.37‰ for δD, averaging as -71.50‰; the variation of δ¹⁸O value was -12.54‰ to -7.09‰, averaging as -10.08‰. According to the δD and δ¹⁸O data of river water samples the Surface Water Line (SWL) equation was modeled as δD =4.43 δ¹⁸O - 26.80 (R² = 0.82, n = 27) for Zhangjiakou area. The groundwater isotopes of the δD values varied from -98.63‰ to -62.65‰, averaging as -79.39‰, the δ¹⁸O values varied from -14.62‰ to -7.92‰ with an average of -11.59‰. A cross plot of δ¹⁸O and δD of the groundwater samples is used to develop a Groundwater Line (GWL) as δD = 4.92 δ¹⁸O - 22.39 (R² = 0.85, n = 56) for the study area (Figure 6). The slope of the GWL was 4.92, and hence much lower than that of the LMWL (7.30) but was slightly higher than that of the SWL (4.43), which is a typical phenomenon in semi-arid areas (Wen et al., 2005). The similarity compositions among surface water and groundwater suggested a good hydrological connection in the study area. Furthermore, the slopes of the SWL and GWL were close to each other and less than LMWL, implying the groundwater received recharge mostly from surface waters and partly from direct precipitation (Hao et al., 2019).

According to the data of groundwater collected from each basin in the Zhangjiakou area, a GWL was established for each region as follows: δD = 4.34 δ¹⁸O - 28.58 (R² = 0.88) [Zhangbei], δD = 4.99 δ¹⁸O - 26.03 (R²=0.85) [Sanggan], δD = 3.79 δ¹⁸O -35.49 (R² = 0.85) [Qingshui], δD = 4.12 δ¹⁸O - 29.54 (R² = 0.88) [Yanghe]. Furthermore, the slope of Sanggan GWL (4.99) was slightly higher than that of GWL (4.92), the GWL slope for the other three basins was smaller than that of GWL, but all the values of slope were close to that of the GWL. Both regression lines for SWL and GWL have a lower slope than the GMWL and a negative intercept, suggesting an important effect of evaporative enrichment on groundwater (Wassenaar et al., 2011).

The d-excess is a useful parameter and commonly applied to study the sources of water vapor and the evaporation effect during rainfall (Gat 1983; Clark and Fritz, 1997). The deuterium excess (d-excess), defined by d-excess = δD - 8×δ¹⁸O quantifies the surplus deuterium about Craig’s line (Dansgaard, 1964). In the Zhangjiakou area the d-excess of precipitation samples varied largely from -4.16‰ to +37.71‰, averaging as 18.45‰ (Table 1). The d-excess values of river water samples varied from -4.40‰ to +28.30‰, averaging as 9.14‰. The d-excess of groundwater samples also varied largely from +0.09‰ to +22.20‰, averaging 13.36‰. Most samples had higher d-excess values (10.75‰ to 37.71‰) than the intercept of GMWL (10‰), except the values measured during snowmelt (-4.40‰ to 9.97‰). This indicates that the primary evaporation was controlled by the low humidity of the vapor sources (Kendall and McDonnell, 1998). Both d-excess in samples of QJSW (4.13‰) and DTW (5.13‰) were relatively lower, where are from artesian wells in the Sanggan river basin. The two particular groundwater presented high TDS (1446.5 mg/L and 1588 mg/L), high concentration of Cl^- and Na⁺ in the study area. They might be attributed by the water-rock interactions with a deep groundwater circulation, which allowed these two samples characterized by a δ¹⁸O-shift-like change.

Tritium (³H) is a useful radioistope to distinguish groundwater recharged during the pre-bomb time from younger water with its half-life of 12.43 a, especially to identify the modern recharge of groundwater (Clark and Fritz, 1997; Newman et al., 2010; Clark, 2015). The³H concentration in groundwater varied from detection limit (<2 TU) to 10.57 TU, averaging as 4.97 TU (Figure 7). The sample of ZBM in Zhangbei had the largest³H content (10.57 TU) in the study area, indicating the groundwater in ZBM may be mainly affected by atmospheric precipitation. Apart from this, the values of³H in the rest samples were less than 8 TU. Most samples exhibited higher³H concentrations than 2 TU, suggesting that the groundwater is recharged by modern water of less than 60 years. These³H concentrations varied vertically with depth, and showed a clear decline with depth (Figure 7). The groundwater samples collected above 50 m showed a much broader range values of³H concentration (2.06 to 10.57 TU, 5.62±2.70), the groundwater samples between 50 to 100 m showed a smaller range of³H concentration variation (<2 to 6.94 TU, 4.32±2.10), the groundwater samples collected from more than 100 m showed a minimum range of³H concentration variation (<2 to 4.36, 2.68±1.05). The³H concentration in Qingshui river basin has the largest range of variation, while the³H concentration in Sanggan river basin presented the smallest range varying from detection limit (<2 TU) to 2.06 TU. The³H values of 3 samples from depths below 100 m were <2 TU, which implies that the³H concentration of groundwater at 100 m is the background level of stratum. Groundwater had a relatively low tritium in the confined aquifers, indicating that the groundwater might be recharged before 1952 (Guo et al., 2019).

For the Zhangjiakou area, the best fit of the exponential piston flow model was found for f ranging between 0.7 and 0.9 and the transit time t. Accordingly, the calculated groundwater age varied between 25 and 60 years (Figure 8). The corresponding groundwater age of most samples was more than 60 years. The sample of ZBM in Zhangbei might be modern water due to its lower determined age of 22 years (Figure 8).

The mean residence time of groundwater can be used to evaluate the recharge rates based on the assumption that the groundwater flow paths in the aquifers are steep so that groundwater residence times correlate with depth without lateral position (Ian and Uwe, 2012). The relationship between groundwater residence time and the depth (Figure 8) indicates that the interpretation of recharge rates from residence times is valid. Particularly, the³H residence time of groundwater samples was positively correlated with depth from the Yanghe river basin and Qingshui river basin. The vertical velocity of groundwater from these two basins was calculated to be 1.1 to 2.26 m/a. In turn, the corresponding recharge rates were estimated to be 0.99-0.203 m/a (99-203 mm/a) taking into account of their investigated soil porosity of 0.09 (The data from Geological Survey report for Zhangjiakou). Soil texture and porosity in the Zhangbei and in the Sanggan river basin are different from the Yanghe river basin and Qingshui river basin (Geological survey report for Zhangjiakou). Hence, the vertical velocity was estimated to be 1.71 to 2.1 m/a, and the recharge rates were 0.034-0.042 m/a (34-42 mm/a) in the Zhangbei and the Sanggan river basin. The groundwater age and recharge rates are very important indicators to evaluate renewability regime of groundwater, and this can provide useful information for the sustainable use of groundwater.

This study explained the spatial and temporal variation of the stable isotope (δD, δ¹⁸O) composition in groundwater, river and precipitation and adopted the radioisotope (³H) to estimate the age of groundwater in the Zhangjiakou area. Furthermore, the mean residence times and recharge rates were calculated using the exponential-piston flow model. Results showed that the LMWL exhibited different slopes and intercepts compared with the GMWL, partly due to isotopic depletion caused by over-compensation of subcloud evaporation. The groundwater in the region is strongly connected with the local meteoric water. Based on the tritium concentrations varying with depth, the age of groundwater ranged from 25 years to more than 60 years and groundwater generally above 100 m might be the modern water, younger than 60 years. However, we assumed groundwater below 100 m to be a mixture of modern and old water. Furthermore, the recharge rates ranged from 0.034 to 0.203 m/a. This study has still several limitations of insufficient data such as more isotopic data and potential influencing factors since the discharge of groundwater is complex, more isotopic sampling and analysis of potential influencing factors are needed in further study. More studies on this subject are therefore necessary to manage precisely groundwater resources in the Zhangjiakou area, North China.