Journal of Hydrology 500 (2013) 1–11

Contents lists available at ScienceDirect

Journal of Hydrology

journal homepage: www.elsevier .com/ locate / jhydrol

Upward recharge through groundwater depression cone in piedmontplain of North China Plain

0022-1694/$ - see front matter � 2013 Elsevier B.V. All rights reserved.http://dx.doi.org/10.1016/j.jhydrol.2013.06.056

⇑ Corresponding author. Address: Wucheng Road No. 36, Xiaodian District,Taiyuan 030006, China. Tel.: +86 155 3698 8152; fax: +86 0351 7010600.

E-mail addresses: rqyuan@sxu.edu.cn, rqyuan@aliyun.com (R. Yuan).

Ruiqiang Yuan a,⇑, Xianfang Song b, Dongmei Han b, Liang Zhang b, Shiqin Wang c

a College of Environment and Resources, Shanxi University, Taiyuan 030006, Chinab Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing 100101, Chinac Faculty of Horticulture, Chiba University, Matsudo 271-8510, Japan

a r t i c l e i n f o

Article history:Received 7 February 2013Received in revised form 15 June 2013Accepted 30 June 2013Available online 10 July 2013This manuscript was handled by Peter K.Kitanidis, Editor-in-Chief, with theassistance of Markus Tuller, Associate Editor

Keywords:Cone of depressionGroundwaterRechargeStatistical analysesNorth China Plain

s u m m a r y

Whether a recharge was induced by groundwater depression cones is a crucial issue for water resourcemanagement. In the North China Plain, shallow groundwater had been over-pumped since 1970s and manygroundwater depression cones formed. The groundwater depression cone, Daceying–Machang, occurredeven in the piedmont plain. In the area, water levels of deep and shallow groundwater were observed since2005 and field survey was conducted in the dry season 2010. The upward recharge induced by the depres-sion cone is verified based on water level records, major ions, 2H, 18O and kinds of statistical analyses. SinceAugust 2006, the water level of the deep groundwater ascended by 1.9 mm/d. High correlations (r = 0.86,s = 0.67) between the water level series of shallow and deep groundwater were found by two distinct cor-relation analyses only in the center of depression cone. Further, the reversion of hydraulic gradient of thedepression cone occurred in dry seasons since September 2008. Hydro-geochemical features of the shallowgroundwater are consistent with deep groundwater in the center of depression cone, which was demon-strated by the fuzzy C-means clustering based on principal components and paired t test, respectively. Itis concluded that the deep groundwater recharged the shallow groundwater from the center of the depres-sion cone. As a result, the groundwater mixture occurred that improves the quality of the shallow ground-water. Seasonally changed flow of shallow groundwater enhanced the mixture. The persistent over-pumping of shallow groundwater and the large elevation difference (around 1200 m) between the rechargezone and the discharge point of deep groundwater facilitate the upward recharge.

� 2013 Elsevier B.V. All rights reserved.

1. Introduction Furthermore, depression cones could impose great influences

Large quantity of continuously groundwater pumping togetherwith the absence of any integrated water resources managementplan in an area forms regional groundwater depression cone andleads to a series of environmental and geological problems. Pachecoet al. (2006) demonstrated a widespread association of ground fail-ure with water table declines. Zhang et al. (2007) illuminated that anidentical hydrostratigraphic unit could present different deformationcharacteristics, such as elasticity, elasto-plasticity, and visco-elasto-plasticity, at different sites of the cone of depression or indifferent periods. By analyzing the volumetric evolution of the coneof depression, Rhode et al. (2007) illustrated the nature of volumet-ric weighted mean transmissivity within the cone of depression as afunction of time. Shi et al. (2008, 2012) simulated regional land sub-sidence and indicated that about 3.08 � 107 m3/yr groundwatercould be provided as emergency water source while meeting theland subsidence control target of 10 mm/a in Suzhou, China.

on the hydrodynamic and hydrochemical fields of a groundwatersystem. Petalas and Lambrakis (2006) reported cation exchangephenomena and the degradation of the groundwater quality duringsalinization processes resulted by the permanent presence of a re-verse regional cone of depression in the coastal area. Sun et al.(2007) analyzed the evolution of depression cones in Yinchuan,an inland city of China and identified that confined water wasmixed with phreatic water and the water quality was deteriorated.Currell et al. (2010) found downward vertical hydraulic gradientsin a cone of depression promoting downwards leakage of shallowwater and high nitrate concentrations in deep groundwater inYuncheng, China. Samborska and Halas (2010) illustrated dissolu-tion of pyrite and its weathering products have a significant influ-ence on chemical composition of water derived from the center ofthe depression cone, due to the enlarged aeration zone creatingoxidation conditions in southern Poland. Nath et al. (2008) foundthat there is no conspicuous relationship between high groundwa-ter As concentration and high groundwater abstraction, althoughthe cone of depression has enlarged over 2 years in West Bengalaquifers India.

2 R. Yuan et al. / Journal of Hydrology 500 (2013) 1–11

In the North China Plain (NCP), population, economic activities,and agricultural production have increased greatly over the last dec-ades, which results in a growing water demand (Foster et al., 2004).Except for a few streams flowing seasonally, groundwater is themain source for industrial, agricultural, and domestic water supply.As a result, the aquifer systems were over-exploited which led to aregional decline of groundwater levels and the local formation ofdepression cones of the potentiometric surface in the area of largecities since the 1970s (Rohden et al., 2010). In order to achieve a sus-tainability of groundwater resource, many studies have addressedthe water balance and recharge mechanisms (Chen et al., 2004,2005; Kendy et al., 2004; Song et al., 2009, 2011; Yang and Tian,2009; Yuan et al., 2011, 2012). The groundwater of the confinedaquifer has been dated (Chen et al., 2003, 2005; Lu et al., 2008; Roh-den et al., 2010). However, depression cones could change thehydrodynamic and hydrochemical fields of groundwater and com-plicate the groundwater flow systems. Consequently, the rechargeinduced by groundwater depression cones should be noticed.

Groundwater isotopes combined with chemistry can produce areliable conceptual model of a groundwater flow system. In thisstudy, the authors attempt to verify the existing of the vertical re-charge induced by a depression cone and to identify influences onthe groundwater systems on the basis of observations of water le-vel and stable isotope compositions and major ion contents. It isexpected that this study will enhance the understanding of thecomplication of the groundwater system impacted by a depressioncone and support the sustainable management and protection ofgroundwater resource in the NCP.

2. Study area

2.1. General settings

The study area lies in the piedmont plain of the North ChinaPlain. The study area covers an area of about 1500 km2 includingBaoding city, Mancheng county, Wanxian county, Qingyuan

Fig. 1. The sampling sites and DEM model (ASTER) of the stud

county, Xushui county and Wangdu county, as illustrated inFig. 1. The climate is continental semiarid with a mean annual tem-perature of about 13 �C. The mean annual precipitation during1955–2009 at Baoding was 531 mm according to monitoring datafrom the Beijing Climate Center (http://www.bcc.cma.gov.cn).About 70% of the annual precipitation falls in the monsoon seasonfrom June to August. Rivers dried out since the 1970s. The landcover is mainly farmland.

Baoding is one of the serious water shortage cities in China. Theamount of water resource occupation per capital of Baoding re-leased by Baoding Institute of Hydrology and Water Resources Sur-vey is just 273 m3/yr in 2005. Groundwater in the unconfined partsof the piedmont plain was strongly exploited over the last severaldecades. As a result, a depression cone of the shallow groundwater,with the name of Daceying–Machang, was formed. The center ofthe cone, Dongdianzhuang, is located in the west of Baoding, wherewater level decline reaches 44 m. The depression cone occupies anarea extent of approximately 650 km2, located mainly in Manch-eng and surrounded by Wanxian, Wangdu, Qingyuan, Baoding,and Xushui.

2.2. Hydrogeological settings

The unconsolidated sediments of Quaternary constitute themain stratigraphy of the study area (Fig. 2). According to the fea-tures of the stratigraphy, it can be divided into four aquifer groups.The first and second aquifer groups (I and II) include aquifer ofHolocene Qh4 with the depth of 10–20 m, the upper PleistoceneQp3 with the depth of 50–70 m and the mid Pleistocene Qp2 withdepth of 80–160 m. The third aquifer group (III) is the Lower Pleis-tocene Qp1 aquifer group with depth of 200–400 m. The fourthaquifer group (?) is Tertiary stratigraphy (Wang et al., 2009; Moiwoet al., 2010). The piedmont plain consists of many alluvial fanswhere the first and second aquifer groups have a close hydraulicconnection (Rohden et al., 2010). Therefore, the two aquifer groups

y area. The ASTER GDEM is a product of METI and NASA.

Fig. 2. The sketch map of geological section along O–O0 in Fig. 1 (according to Zhang et al., 2009).

R. Yuan et al. / Journal of Hydrology 500 (2013) 1–11 3

are considered as the shallow groundwater, while the third aquifergroup is taken as the deep groundwater.

3. Methodology

3.1. Field survey and observation

The shallow and the deep groundwater systems are the mainfocus of the research. Dilution of chemical components is negligi-ble in the end of dry seasons due to very limited precipitation.Therefore, field survey was carried out in May 2010. Two samplingtransections were set crossing the depression cone area. One isalong the elevation gradient (MQ), and the other is vertical to theelevation gradient (XW). Groundwater was surveyed in differentsites along the transections (Fig. 1). Water level was measured be-fore pumping. Then temperature, electrical conductance (EC) andpH value of water samples were measured in situ (DKK. TOA Cor-poration, model: WM-22EP). Shallow groundwater and deepgroundwater were sampled in the same site. Shallow groundwaterwas collected from domestic wells, and deep groundwater wassampled from deep tube wells which were drilled by the local gov-ernment for drinking purpose. Fifteen shallow groundwater sam-ples and twelve deep groundwater samples were collected.

Groundwater was sampled and filtered immediately through a0.45 lm cellulose-ester membrane into three 60 ml and one100 ml high density polyethylene bottles, which were filled tooverflowing and capped. The samples in the 100 ml bottles wereused for titration of bicarbonate on the sampling day. The samplesfor cation analysis were acidified immediately (pH � 2) using highpurity, concentrated HNO3. The other samples were analyzed foranions, dD, and d18O. Chemical and isotopic compositions of watersamples were analyzed in the laboratory of Institute of GeographicSciences and Natural Resources Research, Chinese Academy of Sci-ences. The chemical composition was characterized by ICP-OES forcations (PerkinEler Optime 5300DV) and Ion Chromatography (Shi-madzu LC-10A) for anions with analytical precision of ±1 mg/L.Hydrogen and oxygen isotopes compositions of the water sampleswere analyzed by the Isotope Ratio Mass Spectrometer (FinniganMAT-253) with TC/EA method. The dD and d18O values are reportedas per mill (‰) deviations from the international standard

V-SMOW (Vienna Standard Mean Ocean Water). The dD and d18Omeasurements were reproducible to ±1.0‰ and ±0.2‰

respectively.Water level was monitored in the depression cone area from

2005 to 2011. Two observation boreholes were installed for theshallow groundwater. One is located around the center of thedepression cone (DC01), and the other is seventeen km away fromthe center (DC02). A deep groundwater observation borehole(DC03) was installed thirteen kilometers away from the center(Fig. 1). Water level was read every 60 min by the automaticinstruments (KADEC MIZU II, Japan).

3.2. Water level verification based on correlation analyses and trendanalyses

A close hydraulic connection between two aquifers usually pro-duces similar fluctuations and trends of water level. The similarityof fluctuations can be tested by correlation analyses. The confirma-tion of a trend can be gained by Mann–Kendall test and Sen’s slopemethod.

Correlation analyses were performed in terms of the Pearsoncorrelation coefficient (r) and the Kendall rank correlation coeffi-cient (s). The Pearson correlation coefficient is widely used as ameasure of the strength of linear dependence between two vari-ables. The Kendall rank correlation coefficient is a measure of rankcorrelation: that is, the similarity of the orderings of the data whenranked by each of the quantities. The formula for r is

r ¼ 1n� 1

Xn

i¼1

Xi � X�

sX

!Yi � Y

�

sY

!

where n is the sample size; Xi and Yi (i = 1, 2, 3, . . . , n) are samples oftwo variables; X

�and Y

�are sample means; sX and sY are sample stan-

dard deviations. The formula for s is (adjusted for ties)

s ¼ nc � ndffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiðn0 � t1Þðn0 � t2Þ

pwhere n0 = n(n � 1)/2 is the total number pair combinations; nc isthe number of concordant pairs; nd is the number of discordant

4 R. Yuan et al. / Journal of Hydrology 500 (2013) 1–11

pairs; t1 and t2 are numbers of ties for the samples of two variables,respectively. The two coefficients must be in the range from �1 to 1.

Trend analyses were implemented using the non-parametricMann–Kendall (MK) test and Sen’s slope, respectively. In theMann–Kendall test, the Z-statistic is used to test the null hypothe-sis, H0 that the data is randomly ordered in time, against the alter-native hypothesis, H1, where there is an increasing or decreasingmonotonic trend (Kuo et al., 2011; Martinez et al., 2012; Oguntun-de et al., 2011; Zhang et al., 2011b). Then the true slope of an exist-ing trend was estimated using the Sen’s slope method (Dinpashohet al., 2011). The procedures of MK trend test and Sen’s slope fol-low Kendall (1975) and Sen (1968), respectively.

The non-parametric Mann–Kendall-Sneyers test (Moraes et al.,1998) was applied to locate the approximate starting point of atrend within the data series. The null hypothesis meant that thesample under investigation did not have a beginning developingtrend. When the statistic value rejected the null hypothesis, thereis a change trend. The formula determining the test statistic wasilluminated by Zhang et al. (2011a).

3.3. Geochemical verification: a fuzzy C-means clustering based onprincipal components

The principal component analysis (PCA) was carried out on thebasis of hydrochemical variables of water samples. The objective ofPCA is reducing the possibly correlated variables into a smallernumber of uncorrelated variables, called principal components(Crespo et al., 2012). PCA method reveals the internal structureof the data in a way that best explains the variance in the data.Generally, the first few principal components could explain themost part of variance in the data. Therefore, the first few principalcomponents enhance the major features of the data and facilitateclustering to find near-optimal solutions.

The fuzzy C-means clustering (FCM) was performed based onthe first few principal components. FCM calculates the degree ofmembership of each site into each cluster and allows sites to be-long to more than one cluster simultaneously (Güler et al., 2012).It opposed to traditional clustering which results in mutuallyexclusive clusters (Maharaj and D’Urso, 2011). For this reason,the FCM is very suitable for the mixing case.

The FCM algorithm assumes the attributes are from a vectorspace. The objective is to achieve a minimized total intra-clustervariance function Dv

Dv ¼XK

k¼1

Xxj2Sk

jxj � Ckj2

where Ck the mean point (centroid) of all the points in cluster k; K isthe total number of clusters; Sk is the set of points in the kth cluster;xj is the standardized vector for site j. The FCM algorithm starts bymaking an initial set of k groups. It then calculates the mean pointof each set. The next step is construction of a new partition by asso-ciating each point with the closest centroid. Then the centroids arerecalculated for the new clusters and the algorithm is repeated byalternate application of these two steps until convergence. Partialmembership is permitted in FCM, meaning that each point has a de-gree of membership in each of the clusters. Thus points on the edgeof a cluster may be in that cluster to a lesser degree than points inthe center of a cluster.

The degree of belonging of site i in the kth cluster is equal to theinverse of the distance of site i to the centroid of cluster

bkðiÞ ¼1

dðCk; iÞ

where bk(i) is the degree of belonging of site i in the kth cluster;d(Ck, i) is the distance of site i to the centroid of cluster k. The

coefficients are normalized so that the sum of membership of onesite of interest to all different clusters is unity.

8iXK

k¼1

UkðiÞ ¼ 1

!

where Uk(i) normalized coefficient of site i in the kth cluster. Eachsite is assigned to the cluster with which it has the highest degreeof membership. By the cluster analysis, the geochemical impactand the degree of the impact between groundwater systems couldbe verified.

3.4. Further geochemical verification: Paired t test

Deep groundwater and shallow groundwater were sampled inthe same site nearly simultaneously. Samples of different ground-water systems in the same site form a pair. For this case, the pairedt test can offer a further verification of geochemical similarity ofdifferent groundwater systems.

The paired t test typically consists of a sample of matched pairsof similar units. It is most commonly applied when the sampleswould follow a normal distribution. A non-parametric test, onesample Kolmogorov–Smirnov test could serve as the test for nor-mality of the sample distribution under the null hypothesis thatthe sample is drawn from the reference distribution. The nullhypothesis of the paired t test is that the means of two normallydistributed populations are equal. The statistic of the paired t testis calculated by

t ¼ D�

SD

where t is the statistic for equality of means, D�

is the mean of thedifferences between the paired data; SD is standard error of the dif-ferences (David and Gunnink, 1997). One sample Kolmogorov–Smirnov test and paired t test were performed in SPSS software.

4. Results

4.1. Distribution of depth to water level and variance of water level

According to Baoding Institute of Hydrology and Water Re-sources Survey, the average depth to water level in the study areachanged between 21.34 m and 24.15 m and the mean in the centerof the cone varied between 43.76 m and 44.95 m in 2010. Thedepth to water level measured during the field survey varied be-tween 21.44 m and 39.78 m. The shallowest one was observed inthe site of G7 that is 25.5 km away from the center of the cone.While the deepest one appeared in the site of G3 that is 3.9 kmaway from the center. The areal distribution of the depth estimatedby tension spline interpolation (Fig. 3) is in accordance with the re-search result of Lv et al. (2010). The cone of depression is situatedin the area of Mancheng. It is confirmed that G4, G5, G6, G7, G9,G10 and G15 are located on the edge of the depression cone withan average depth to water level of 22.6 m and other sites lie inthe cone with the mean depth of 33.1 m.

The observation period of water level is from March 2005 toMay 2011 (Figs. 4 and 5). About half year’s records before March2010 missed due to failure of batteries in the instruments. In addi-tion, water level lower than �12.55 m was not recorded for DC01before July 2007, because the probe was hanged too high in thewell pipe.

The records show that the shallow groundwater has evidentlydaily fluctuations. The well of DC01 with a depth of 70 m is locatedin the pluvial fan of Jie River that is the water source area for Bao-ding city. As a result of periodic development of groundwater,

Fig. 3. The areal distribution of depth to water level estimated by tension spline interpolation.

Fig. 4. The variance of water level of the observation wells during the period from March 2005 to October 2009. The observation boreholes of DC01 and DC02 were installedfor the shallow groundwater. The observation borehole of DC03 was installed for the deep groundwater. Water level was read every 60 min by the automatic instruments.Water level lower than �12.55 m was not recorded for DC01 before July 2007.

R. Yuan et al. / Journal of Hydrology 500 (2013) 1–11 5

water level reached the climax in 7 am and arrived at the bottom in5 pm. Then the water level remained stable until the pumping stop

Fig. 5. The variance of water level of the observation wells during the period fromMarch 2010 to May 2011.

in about 6 pm which shows that the shallow groundwater was re-charged continuously. The well of DC02 with a depth more than40 m is a domestic well. The daily variation of water level changedlittle besides some sudden drops occurred in 12 am or 8 pm. Thedeep groundwater has a stable water head which varied dailywithin 8 cm on average. At the same time, the mean daily varia-tions are 1.05 m and 0.63 m for DC01 and DC02 respectively.

In a whole hydrological year, groundwater regime shows twocycles in the observation wells. Water level fell down in Apriland rose in July In October water level descended again. In Decem-ber the second ascending occurred. The same dynamic regimesindicate the similar recharge and discharge in the shallow andthe deep groundwater. It is suggested that plenty rainfall in rainyseason and smelted snow in early April formed two rapid lateralrecharges to the groundwater systems.

During the observation period, the water level shows anascending trend in DC01 and DC03, while the water level presentsa descending trend in DC02. The average water levels are �8.4 m,�5.3 m and 2.9 m in DC01, DC02, and DC03 respectively. Thehydraulic head of the deep groundwater is far higher than that of

Table 1Hydrochemical and stable isotopic characteristics of shallow groundwater samples.

Sites Altitude(m)

Well depth(m)

pH EC (lS/cm)

Ca2+

(mg/L)Mg2+

(mg/L)Na+

(mg/L)K+ (mg/L)

HCO�3(mg/L)

Cl�

(mg/L)SO2�

4

(mg/L)

NO�3(mg/L)

dD(‰)

d18O(‰)

G1 96 80 8.11 824 52.4 43.1 34.0 1.9 334.4 43.9 33.7 37.0 �59.2 �8.09G2 32 80 8.07 530 69.8 29.2 8.7 1.1 304.4 15.7 9.8 14.1 �64.1 �9.44G3 26 80 8.34 544 34.9 28.4 7.6 2.7 211.3 16.8 13.2 0.0 �62.6 �10.45G4 13 60 7.25 1349 60.6 67.9 104.6 5.0 295.0 143.2 117.7 78.2 �56.6 �7.16G5 13 50 7.54 1077 47.2 73.8 52.2 0.8 409.6 94.4 37.3 42.5 �64.7 �9.00G6 13 80 8.09 752 42.0 46.8 52.5 0.8 368.9 31.6 39.7 2.2 �55.7 �9.98G7 13 36 8.06 748 40.6 40.7 48.7 0.7 318.9 25.1 39.4 6.1 �60.0 �7.44G9 13 40 8.24 873 49.3 47.6 28.2 0.9 363.0 61.0 36.8 2.8 �63.4 �7.61G10 13 36 8.18 1057 50.5 74.7 70.9 1.4 473.1 72.1 55.0 9.7 �62.5 �8.08G10’ 13 80 8.24 1004 47.8 75.3 61.7 1.2 415.1 87.8 42.5 11.7 �67.1 �7.78G11 18 40 8.01 605 72.2 36.6 16.4 1.7 365.5 18.5 10.2 10.4 �62.3 �8.81G12 23 80 8.16 634 50.7 36.2 15.5 1.0 326.7 19.3 6.7 9.7 �62.2 �7.84G13 29 60 8.24 541 44.2 24.2 12.7 1.2 259.3 11.8 5.2 10.6 �64.0 �8.30G14 33 30 8.12 770 56.7 43.2 25.6 0.9 276.1 53.2 30.8 32.6 �61.9 �8.06G15 45 60 8.03 827 85.9 23.0 24.9 0.8 304.6 32.8 44.1 76.2 �63.2 �6.93Mean 8.0 802 54 46 38 1 335 48 35 23 �62.0 �8.33

Table 2Hydrochemical and stable isotopic characteristics of deep groundwater samples.

Sites Altitude(m)

Well depth(m)

pH EC (lS/cm)

Ca2+

(mg/L)Mg2+

(mg/L)Na+

(mg/L)K+ (mg/L)

HCO�3(mg/L)

Cl�

(mg/L)SO2�

4

(mg/L)

NO�3(mg/L)

dD(‰)

d18O(‰)

G3d 26 120 8.13 549 38.2 27.6 6.0 1.0 250.2 11.8 10.1 5.4 �65.2 �10.46G4d 13 300 7.95 625 57.1 18.4 111.8 1.8 412.5 26.7 48.2 9.9 �79.8 �9.27G5d 13 360 8.01 516 42.1 17.5 50.5 1.4 259.3 10.4 42.1 1.5 �80.6 �8.58G6d 13 300 8.34 543 20.3 9.5 72.0 1.3 230.7 18.9 29.3 6.0 �70.6 �11.95G8d 11 180 8.28 413 23.7 12.5 51.3 2.2 193.1 7.7 60.6 0.0 �70.4 �10.42G8d’ 11 400 8.38 495 25.6 13.4 69.2 2.2 175.0 16.6 74.1 5.6 �76.9 �10.07G9d 13 300 8.25 583 42.8 24.6 60.1 1.3 306.2 16.8 30.4 7.9 �75.9 �10.19G10d 13 150 8.28 777 42.4 50.8 47.1 0.8 371.6 20.6 48.7 4.6 �66.9 �8.24G11d 18 120 8 548 31.7 29.2 20.3 1.3 306.5 5.4 4.8 3.9 �62.1 �10.04G12d 23 170 8.26 497 55.4 20.0 31.5 1.2 311.1 5.2 7.7 6.6 �69.6 �8.34G13d 29 150 8.31 424 52.2 20.9 12.1 1.5 253.8 7.6 5.1 9.0 �63.7 �8.26G14d 33 160 8.27 401 53.0 15.3 9.9 1.2 206.7 8.3 11.6 8.9 �66.9 �8.15Mean 8.2 531 40 22 45 1 273 13 31 6 �70.7 �9.50

6 R. Yuan et al. / Journal of Hydrology 500 (2013) 1–11

the shallow groundwater. The hydraulic condition for the upwardrecharge is satisfied. The most important thing is that water levelin DC01 was higher than that in DC02 from September 2008 toApril 2009 and from August 2010 to April 2011. During the twoperiods (dry seasons), the hydraulic gradient in the cone of depres-sion was reversed, which means the sink becomes a source.

Fig. 6. The plot of EC versus dD of water samples.

4.2. Geochemical features

The hydrochemical and stable isotopic characteristics of watersamples are shown in Tables 1 and 2. The average of pH is8.0 ± 0.29 in the shallow groundwater, while the mean is8.2 ± 0.14 in the deep groundwater. In general, the pH value ofshallow groundwater is lower than the one of deep groundwater.The averages of EC and dD are 802 ± 258 lS/cm and �62 ± 3‰ inthe shallow groundwater, while the means are 531 ± 103 lS/cmand �71 ± 6‰ in the deep groundwater. The shallow groundwateris originally saltier and isotopic heavier than the deep groundwa-ter. Mg2+ is one of dominated cations in the shallow groundwater,and Na+ concentration is relatively high in the deep groundwater.In addition, the contents of Cl� and NO�3 mainly from anthropo-genic input are about four times higher in the shallow groundwa-ter (in average 48 mg/L and 23 mg/L, respectively) than those inthe deep groundwater. The compositions of major ions are mainlyMg�Ca–HCO3 and Ca�Mg–HCO3 type in the shallow and Na�Ca–HCO3 and Ca�Mg–HCO3 type in the deep groundwater respectively(Fig. 8). Consistent with the results of former researches (Songet al., 2011; Yuan et al., 2011; Zhuang et al., 2011), the deepgroundwater has high pH values and light isotopic compositions,

while the shallow groundwater has high EC values and contentsof Cl� and NO�3 .

However, the geochemical similarity between the shallow andthe deep groundwater is evident. More than half of the samplesfrom shallow groundwater have the similar pH values with thedeep groundwater. Many samples of the shallow groundwaterhave the similar EC values, chemical types and isotopic composi-tions with the deep groundwater (Figs. 6–8). The distribution ofdata points is loose especially for deep groundwater, although a

Fig. 7. The plot of d18O versus dD of water samples. The local meteoric water line(LMWL) follows Yuan et al. (2012).

R. Yuan et al. / Journal of Hydrology 500 (2013) 1–11 7

part of the points follows the LMWL line closely in Fig. 7. In deepaquifers, there are fossil groundwater that have obviously differentisotopic compositions with modern precipitation (Chen et al.,2005). Also, isotope exchanges that would cause data points awayfrom the meteorological water line could happen by interactions ofwater and rock in deep aquifers. In this case, it seems that the deepgroundwater induced a loose distribution of data points of shallowgroundwater.

The variances of major ion contents of the shallow groundwaterwere presented along the transections for illumination of evolutionof the shallow groundwater (Fig. 9). In the cone of depression, thedeep groundwater has an average of EC about 531 lS/cm and dis-tinctly low ion contents of NO�3 , Cl� and Mg2+. When the deepgroundwater recharge the overlying shallow groundwater underthe upward vertical hydraulic gradient, the EC value and the corre-sponding ion concentrations in the shallow groundwater de-creased greatly. Along the MQ transection, EC value of theshallow groundwater decreased from 824 lS/cm in the edge of

Fig. 8. The Piper plot

the cone (G1) to 544 lS/cm in the center of the cone (G3), and thenincreased to 1349 lS/cm and 1077 lS/cm in G4 and G5 where isclose to the metropolis Baoding. After G5, the EC value almostmaintained stable with a little variation between 752 lS/cm and748 lS/cm. The variations of contents of Ca2+, Mg2+, Na+, HCO�3 ,SO2�

4 and Cl� follow the same changing trend and give an explana-tion to the EC fluctuation. The same evolution was observed whenthe XW transection extends through the cone of depression, ioncontents and also EC values were dropped evidently into a low va-lue interval (G11, G12 and G13).

In the center area of the cone, the major ion contents and EC va-lue of shallow groundwater (G3) are very similar with the localdeep groundwater (G3d) but obviously lower than the averagesof the shallow groundwater (Tables 1 and 2). In addition, the isoto-pic composition of G3 (d18O �10.45‰, dD �62.6‰) is almost thesame with G3d (d18O �10.46‰, dD �65.2‰) considering the ra-tional measurement errors, which suggests the same originationof waters. It is suggested that deep groundwater has occupiedthe phreatic aquifer in the center of the depression cone.

The difference of d18O between the G3d (�10.46‰) and the G1(�8.09‰) reveals that the recharge zone elevation of the deepgroundwater is higher than 1200 m based on the altitude gradientof d18O �0.2‰/100 m (Liu et al., 2010). The large elevation differ-ence between the recharge zone and the discharge point of thedeep groundwater offers the upward vertical hydraulic gradient.The great upward vertical hydraulic gradient promotes the upwardrecharge.

5. Discussion

5.1. Evidence from water level analyses for the upward recharge

In the center of the depression cone, the water level is very low(average �8.4 m, minimum �15.2 m), while the water level ofdeep groundwater is relative high (average 2.9 m, maximum4.3 m). Under the hydraulic condition, it is possible that shallowgroundwater was recharged by the deep groundwater. If the re-charge happened, the fluctuation and even the trend of water levelseries of shallow groundwater will be controlled mainly by the

of water samples.

Fig. 9. The variances of major ion contents of the shallow groundwater along (a) theMQ transection and (b) the XW transection.

8 R. Yuan et al. / Journal of Hydrology 500 (2013) 1–11

deep groundwater. The deduction was verified by correlation anal-yses and trend analyses of time series of water level.

Daily water level series were analyzed. The time series of dailywater level were produced by extracting the water level on 4 amwhen the water level had been recovered and remain stable. ThePearson correlation analysis was executed within ±30 time lags(day) and the max correlation coefficient was obtained. The resultshows that there is a strong hydraulic connection between thedeep and the shallow groundwater in the center of the depressioncone. The strong hydraulic connection is presented by high corre-lation coefficients (r = 0.86, s = 0.67 with no time lags) betweenwater level time series of DC01 and DC03 (Table 3). Furthermore,the correlation is weak between DC01 and DC02 which means dif-ferent mechanisms of water level fluctuation in different sites ofthe depression cone. It is suggested that the fluctuation of water

Table 3The results of correlation analysis of water level series.

Observation wells Periods Number of data

DC01–DC02 2007/7–2009/8 7782010/3–2011/5 413

DC01–DC03 2007/7–2009/8 7782010/3–2011/5 412

DC02–DC03 2005/4–2009/9 16032010/3–2011/5 412

level of shallow groundwater in the center of depression conewas controlled by deep groundwater.

The variation trends were evaluated based on the daily series ofwater level. First, the non-parametric Mann–Kendall–Sneyers testwas used to locate the approximate starting point of a trend(Fig. 10). The forward sequence curve does not intersect the back-ward sequence curve within the confidence interval ±1.96(a = 0.05%) for the DC01 and DC02 time series. It is confirmed thatno changing point exists in the series and variation trend of water le-vel is monotonous. However, there is a changing point occurred onAugust 2006 for the DC03 time series. Therefore, it is necessary totest the trend before and after the changing point respectively. Thesignificance and magnitude of water level series trends were deter-mined using the non-parametric Mann–Kendall test and Sen’s slopemethod, respectively. The results indicate that there are differenttrends in different sites of the cone of depression. In the center re-gion of the depression cone (DC01), the water level of shallowgroundwater presented an upward trend with a slope about 6.9and 7.0 mm/d. At the same time, the water level of deep groundwa-ter also showed an upward trend with a slope about 1.9 and 1.0 mm/d in the corresponding periods (Table 4). By contrast, the water levelof shallow groundwater exhibited a downward trend at the site ofDC02 from April 2005 to September 2009, and then it varied withno significant trends from March 2010 to May 2011. The stronginfluence of deep groundwater on the water level of shallow ground-water in the center of the depression cone is confirmed.

The impacts of deep groundwater on the fluctuation and thetrend of water level of shallow groundwater are the results of the re-charge from deep groundwater to shallow groundwater in the cen-ter region of the depression cone. The shallow groundwater wasover-pumped for more than 40 years, and the water level declinedpersistently. At the same time, the water level of deep groundwaterascended at a rate of 1.9 mm/d since August 2006. The largesthydraulic head difference between the deep and the shallowgroundwater occurred in the center of the depression cone. Finally,the shallow groundwater was recharged by the deep groundwater.As a result, the water level of shallow groundwater in the center ofthe depression cone presents an ascending trend and a close corre-lation with the deep groundwater (Tables 3 and 4). The sink becamea source. Since September 2008, the water level of DC01 was higherthan that of DC02 in dry seasons when the hydraulic gradient in thecone of depression was reversed. The shallow groundwater flowedfrom the center of the cone to the edge which makes the declinetrend of DC02 ceased (Table 4) and the correlation between thetwo sites became stronger (the time lags of max r decreased from7 days to 2 days, shown in Table 3). As a result of the enlarged re-charge, the ascending trend of the deep groundwater slowed downfrom 1.9 mm/d to 1.0 mm/d (Table 4).

5.2. Evidence from integrated analyses of geochemical data for theupward recharge

As a result of the upward recharge, the hydro-geochemistry ofthe shallow groundwater would be changed considering the

Pearson correlation Kendall correlation

Lags (day) Max r s

7 0.64 0.422 0.65 0.43

0 0.86 0.671 0.70 0.46

�30 �0.26 �0.19�1 0.25 0.15

Fig. 10. Mann–Kendall–Sneyers test for detecting change points in daily water level series with forward sequence curve (ufk, solid line) and backward sequence curve (ubk,dashed line). The horizontal lines represent the 5% significance level. (a) DC01, (b) DC02, and (c) DC03.

Table 4The results of trend analysis of water level series.

Observation well Periods Number of data Z value Sen’s slope (mm/d) Trend

DC01 2007/7–2009/8 778 18.99 6.9 Upward trend detected2010/3–2011/5 413 8.62 7.0 Upward trend detected

DC02 2005/4–2009/9 1603 �31.73 �1.9 Downward trend detected2010/3–2011/5 413 0.48 No significant trend

DC03 2005/3–2006/7 502 �1.19 No significant trend2006/8–2009/10 1163 32.47 1.9 Upward trend detected2010/3–2011/5 412 8.80 1.0 Upward trend detected

R. Yuan et al. / Journal of Hydrology 500 (2013) 1–11 9

different geochemical characteristics between the shallow and thedeep groundwater. The fuzzy C-means clustering based on princi-pal components was employed to clarify the influences and to as-sure the recharge further. First, the PCA analysis was carried out forthe twelve variables (pH, EC, dD, d18O, Ca2+, Mg2+, Na+, K+, HCO�3 ,SO2�

4 , Cl�, NO�3 ) of water samples to achieve principal components.The results show that EC, HCO�3 , Cl� and Mg2+ have a strong effecton the value of the first principal component. The first principalcomponent explains 93.3% of the variance of the original variables.It is confident the first principal component could present compre-hensively the geochemical features.

The FCM clustering for two groups was carried out based on thefirst principal component. The iterative procedure of FCM algo-rithm generated a fuzzy membership matrix (Table 5). A fuzzymembership is a probability of belonging to clusters for a sample.All deep groundwater samples are classified into group B. In addi-tion, five samples out of fifteen samples of shallow groundwateralso belong to group B. The rest samples of shallow groundwaterare classified into group A. The five samples are all located in thearea of the depression cone. It is confirmed that shallow ground-water in the area of depression cone is similar with the deepgroundwater in hydro-geochemistry. The result also suggests that

Table 5The results of the fuzzy C-means clustering based on the first principal component.

Site Type Membership matrix Uk(i)

Group A Group B

G100 Sa 0.988 0.012G10 S 0.961 0.039G5 S 0.949 0.051G9 S 0.949 0.051G15 S 0.843 0.157G1 S 0.837 0.163G4 S 0.808 0.192G14 S 0.629 0.371G6 S 0.543 0.457G7 S 0.523 0.477G10d Db 0.342 0.658G12 S 0.089 0.911G4d D 0.070 0.930G14d D 0.056 0.944G8d D 0.048 0.952G13d D 0.042 0.958G11 S 0.039 0.961G3 S 0.030 0.970G9d D 0.017 0.983G8d’ D 0.007 0.993G12d D 0.007 0.993G5d D 0.002 0.998G3d D 0.001 0.999G11d D 0.001 0.999G13 S 0.000 1.000c

G2 S 0.000 1.000G6d D 0.000 1.000

a Membership of 1.000 means very close to 1.b S represents shallow groundwater.c D represents deep groundwater.

10 R. Yuan et al. / Journal of Hydrology 500 (2013) 1–11

shallow groundwater is evidently impacted by the recharge ofdeep groundwater in the region of depression cone.

For a further validation of the geochemical similarity, paired ttest was carried out. In this case, the shallow and the deep ground-water sampled in the same sites are the matched pairs and thevariables (Ca2+, Mg2+, Na+, K+, HCO�3 , SO2�

4 , Cl�, NO�3 , dD, d18O) arethe units. The null hypothesis of the test is that the geochemicaldifferences between the shallow groundwater and the deepgroundwater are insignificant. The result of the one-sample Kol-mogorov–Smirnov test confirms that all variables fit the normaldistribution well at the significant level of 0.05. Finally, four pairsincluding G3/G3d, G11/G11d, G13/G13d, and G12/G12d werefound. The result of paired t test shows that the null hypothesiscannot be rejected for those variables besides Cl� at the significantlevel of 0.05 (Table 6). It is verified that there is a very similarhydrochemistry between the shallow and the deep groundwaterin the area of depression cone. The content of Cl� in shallowgroundwater is related to anthropogenic input at individual points.

Table 6The results of the paired t test.

Mean Std. error mean ta p Value (2-tailed)

HCO�3 10.3 20.1 0.513 0.644Ca2+ 6.1 11.5 0.533 0.631NO�3 1.5 2.5 0.579 0.603K+ 0.4 0.5 0.869 0.449Na+ �4.4 4.0 �1.095 0.353

SO2�4

1.9 1.5 1.307 0.282

dD 2.4 1.8 1.316 0.280d18O 0.4 0.3 1.442 0.245Mg2+ 6.9 3.4 2.050 0.133Cl� 9.1 2.6 3.485 0.040

a The degree of freedom is 3.

Consequently, it labeled the differences in hydrochemistry be-tween the shallow and the deep groundwater.

Geochemical differences between the shallow and the deepgroundwater are originally evident as mentioned before. However,the hydrochemical features of the shallow groundwater within thedepression cone are significantly similar with the deep groundwa-ter. While the geochemical features of the shallow groundwater inthe margin of the depression cone are significantly distinct withthe deep groundwater. The phenomenon offers a geochemicalproof confirming the recharge from the deep groundwater to theshallow groundwater in the area of the depression cone.

6. Conclusions

Due to persistent over-pumping of the shallow groundwater forindustrial, agriculture and domestic consuming, the groundwaterdepression cone (Daceying–Machang) occurred and expandedeven in the piedmont plain of the North China Plain. The existenceof Daceying–Machang depression cone induced the recharge fromdeep groundwater to shallow groundwater in the center of thedepression cone, which changed hydrodynamic and hydro-geo-chemical fields of the shallow groundwater greatly. Since 2008,the hydraulic gradient of the depression cone was reversed indry seasons which enhances the groundwater mixture and compli-cates the flow in the depression cone in spatial–temporal scales.The high contents of Cl� and NO�3 in shallow groundwater were di-luted. Water quality of the shallow groundwater is improved. Thepersistent over-pumping of shallow groundwater and the largeelevation difference (around 1200 m) between the recharge zoneand the discharge point of deep groundwater facilitate the upwardrecharge.

Statistical analyses, such as fuzzy C-means clustering based onprincipal components, paired t test, trend analysis and correlationanalysis are very useful and economical tools to verify the interac-tion between groundwater systems and could be considered as animportant assist for the routine analyses of groundwaterhydrology.

Acknowledgements

This work was financially supported by Key Program of Na-tional Natural Science Foundation of China (No. 40830636) andthe State Basic Research Development Program (973 Program) ofChina (No. 489 2010CB428805). The authors would like to thankthe Water and Soil Conservation Station of Baoding City for givingsupport to the project, with special mention to Guangying Zhang,Shengbao Wang for their help on field surveys. The authors givethe sincere thanks to Dr. Xiting Long, Dr. Bing Zhang and Dr. YileiYu for their earnest field work.

References

Chen, J. et al., 2003. Use of water balance calculation and tritium to examine thedropdown of groundwater table in the piedmont of the North China Plain (NCP).Environmental Geology 44, 564–571.

Chen, J. et al., 2004. Spatial geochemical and isotopic characteristics associated withgroundwater flow in the North China Plain. Hydrological Processes 18, 3133–3146.

Chen, Z., Nie, Z., Zhaoji, Z., Jixiang, Q., Yunju, N., 2005. Isotopes and sustainability ofground water resources, North China Plain. Ground Water 43 (4), 485–493.

Crespo, P.J. et al., 2012. Identifying controls of the rainfall–runoff response of smallcatchments in the tropical Andes (Ecuador). Journal of Hydrology 407 (1–2),164–174.

Currell, M.J., Cartwright, I., Bradley, D.C., Han, D., 2010. Recharge history andcontrols on groundwater quality in the Yuncheng Basin, north China. Journal ofHydrology 385 (1–2), 216–229.

David, H.A., Gunnink, J.L., 1997. The paired t test under artificial pairing. TheAmerican Statistician 51 (1), 9–12.

R. Yuan et al. / Journal of Hydrology 500 (2013) 1–11 11

Dinpashoh, Y., Jhajharia, D., Fakheri-Fard, A., Singh, V.P., Kahya, E., 2011. Trends inreference crop evapotranspiration over Iran. Journal of Hydrology 399 (3–4),422–433.

Foster, S. et al., 2004. Quaternary Aquifer of the North China Plain-assessing andachieving groundwater resource sustainability. Hydrogeology Journal 12, 81–93.

Güler, C., Kurt, M.A., Alpaslan, M., Akbulut, C., 2012. Assessment of the impact ofanthropogenic activities on the groundwater hydrology and chemistry in Tarsuscoastal plain (Mersin, SE Turkey) using fuzzy clustering, multivariate statisticsand GIS techniques. Journal of Hydrology 414–415, 435–451.

Kendall, M.G., 1975. Rank Correlation Methods. Charles Griffin, London.Kendy, E., Zhang, Y., Liu, C., Wang, J., Steenhuis, T., 2004. Groundwater recharge

from irrigated cropland in the North China Plain: case study of LuanchengCounty, Hebei Province, 1949–2000. Hydrological Processes 18, 2289–2302.

Kuo, Y.M., Chu, H.J., Pan, T.Y., Yu, H.L., 2011. Investigating common trends of annualmaximum rainfalls during heavy rainfall events in southern Taiwan. Journal ofHydrology 409 (3–4), 749–758.

Liu, J. et al., 2010. Characteristics of d18O in precipitation over Eastern MonsoonChina and the water vapor sources. Chinese Science Bulletin 55 (2), 200–211.

Lu, Y. et al., 2008. Spatial characteristics of water quality, stable isotopes and tritiumassociated with groundwater flow in the Hutuo River alluvial fan plain of theNorth China Plain. Hydrogeology Journal 16, 1003–1015.

Lv, C., Jia, S., Ji, Z., 2010. Dynamics and causes of groundwater table change in plainarea of Baiyangdian Basin in last 30 years. South to North Water Transfers andWater Science & Technology 8 (1), 65–68 (in Chinese).

Maharaj, E.A., D’Urso, P., 2011. Fuzzy clustering of time series in the frequencydomain. Information Sciences 181 (7), 1187–1211.

Martinez, C.J., Maleski, J.J., Miller, M.F., 2012. Trends in precipitation andtemperature in Florida, USA. Journal of Hydrology 452–453, 259–281.

Moiwo, J.P., Yang, Y., Li, H., Han, S., Yang, Y., 2010. Impact of water resourceexploitation on the hydrology and water storage in Baiyangdian Lake.Hydrological Processes 24 (21), 3026–3039.

Moraes, J.M. et al., 1998. Trends in hydrological parameters of a southern Brazilianwatershed and its relation to human induced changes. Water ResourcesManagement 12, 295–311.

Nath, B., Stüben, D., Mallik, S.B., Chatterjee, D., Charlet, L., 2008. Mobility of arsenicin West Bengal aquifers conducting low and high groundwater arsenic. Part I:comparative hydrochemical and hydrogeological characteristics. AppliedGeochemistry 23, 977–995.

Oguntunde, P.G., Abiodun, B.J., Lischeid, G., 2011. Rainfall trends in Nigeria, 1901–2000. Journal of Hydrology 411 (3–4), 207–218.

Pacheco, J. et al., 2006. Delimitation of ground failure zones due to land subsidenceusing gravity data and finite element modeling in the Querétaro valley, México.Engineering Geology 84 (3–4), 143–160.

Petalas, C., Lambrakis, N., 2006. Simulation of intense salinization phenomena incoastal aquifers – the case of the coastal aquifers of Thrace. Journal of Hydrology324 (1–2), 51–64.

Rhode, K.L., Osiensky, J.L., Miller, S.M., 2007. Numerical evaluation of volumetricweighted mean transmissivity estimates in laterally heterogeneous aquifers.Journal of Hydrology 347 (3–4), 381–390.

Rohden, C.V., Kreuzer, A., Chen, Z., Kipfer, R., Aeschbach-Hertig, W., 2010.Characterizing the recharge regime of the strongly exploited aquifers of theNorth China Plain by environmental tracers. Water Resources Research 46,W05511.

Samborska, K., Halas, S., 2010. 34S and 18O in dissolved sulfate as tracers ofhydrogeochemical evolution of the Triassic carbonate aquifer exposed tointense groundwater exploitation (Olkusz–Zawiercie region, southern Poland).Applied Geochemistry 25 (9), 1397–1414.

Sen, P.K., 1968. Estimates of the regression coefficient based on Kendall’s tau.Journal of American Statistical Association 63, 1379–1389.

Shi, X. et al., 2008. Regional land subsidence simulation in Su-Xi-Chang area andShanghai City, China. Engineering Geology 100 (1–2), 27–42.

Shi, X. et al., 2012. Sustainable development and utilization of groundwaterresources considering land subsidence in Suzhou, China. Engineering Geology124, 77–89.

Song, X. et al., 2009. A study of soil water movement combining soil water potentialwith stable isotopes at two sites of shallow groundwater areas in the NorthChina Plain. Hydrological Processes 23, 1376–1388.

Song, X. et al., 2011. Relationships between precipitation, soil water andgroundwater at Chongling catchment with the typical vegetation cover in theTaihang mountainous region, China. Environmental Earth Sciences 62 (4), 787–796.

Sun, Y., Qian, H., Wu, X., 2007. Hydrogeochemical characteristics of groundwaterdepression cones in Yinchuan City, Northwest China. Chinese Journal ofGeochemistry 26 (4), 350–355.

Wang, S. et al., 2009. Shallow groundwater dynamics in North China Plain. Journalof Geographical Sciences 19 (2), 175–188.

Yang, Y., Tian, F., 2009. Abrupt change of runoff and its major driving factors inHaihe River Catchment, China. Journal of Hydrology 374 (3–4), 373–383.

Yuan, R. et al., 2011. Using major ions and stable isotopes to characterize rechargeregime of a fault-influenced aquifer in Beiyishui River Watershed, North ChinaPlain. Journal of Hydrology 405 (3–4), 512–521.

Yuan, R. et al., 2012. Rate and historical change of direct recharge from precipitationconstrained by unsaturated zone profiles of chloride and oxygen-18 in dry riverbed of North China Plain. Hydrological Processes 26 (9), 1291–1301.

Zhang, Y. et al., 2007. Characteristics of aquifer system deformation in the SouthernYangtse Delta, China. Engineering Geology 90 (3–4), 160–173.

Zhang, Z., Hong, F., Chen, Z., Zhao, Z., Xie, Z., Wang, Y., Miao, J., Yang, L., Shao, J., Jin,M., Xu, G., Yang, Q., 2009. Investigation and assessment of sustainableutilization of groundwater resources in the North China Plain. GeologicalPublishing House, Beijing (in Chinese).

Zhang, Y. et al., 2011a. Analysis of impacts of climate variability and human activityon streamflow for a river basin in northeast China. Journal of Hydrology 410 (3–4), 239–247.

Zhang, Q. et al., 2011b. Precipitation and streamflow changes in China: changingpatterns, causes and implications. Journal of Hydrology 410 (3–4), 204–216.

Zhuang, C. et al., 2011. Impacts of human activities on the hydrology of BaiyangdianLake, China. Environmental Earth Sciences 26 (7), 1343–1350.