Introduction

Managed aquifer recharge (MAR) was an increasingly common measure for increasing urban water supply and sustaining water resilience (Dillon 2005; Rodríguez-Escales et al. 2017). Excess water was stored in the aquifers for future consumption by MAR. And the major methods used for MAR include bank filtration, well injection, rainwater harvesting, and infiltration ponds (Ebrahim et al. 2016; Ganot et al. 2017). MAR provided a series of technical solutions to sustainably manage water resources by not only recharging groundwater but also displaying an integrated vision of water resources, following the EU Water Framework Directive approach (Escalante et al. 2019). The groundwater level dropped rapidly over 10 m per year in China. And groundwater overexploitation had occurred in 164 areas of China’s 31 provinces, affecting more than 180,000 km2 (Werner et al. 2013). Groundwater funnel formed gradually and expanded by the groundwater depletion, which caused frequent land subsidence, water quality deterioration, and seawater intrusion. To alleviate the abovementioned issues, a series of water management measures had been carried out in China, such as improving irrigation technology and water efficiency, implementing water price reforms, and water use rights. However, the lack of detailed information on the amount of groundwater extracted, and limited groundwater monitoring wells, as well as complex recharge water sources, make groundwater management still face severe challenges.

Water security associated with groundwater overexploitation was increasingly becoming a critical environmental concern. Growing studies had indicated that MAR could effectively improve water supply and surface water quality. Rainwater quality after being treated by constructed wetlands and MAR was studied in urban catchments (Page et al. 2010). The mechanism of excessive air formation during MAR (natural bank filtration) was investigated in Berlin (Massmann and Sültenfuß 2008). Leonard et al. (2015) analyzed the possibility of MAR-treated rainwater for potable and non-potable. Mankad et al. (2015) qualitatively studied the acceptance of MAR-treated urban rainwater for household use combined with psychological and policy-related factors. MAR risks were revealed using fault tree and probabilistic assessment methods at six recharge sites in the Mediterranean Basin (Rodríguez Escales et al. 2018). Fiorillo et al. (2019) analyzed the features of MAR in the Caposele Spring area, southern Italy, using a daily recharge model. MAR projects of two demonstration sites were studied according to the monitoring data in the Guadiana Basin, Spain (Fernández Escalante 2015). The 150,000-m3/day MAR project was proposed as an economic solution to solve the water supply and groundwater quality issues in Southwestern Florida, USA (Missimer et al. 2017). Dillon et al. (2019) assessed the efficacy of MAR and illustrated the development of MAR at the global scale over the past 60 years. To investigate the hydrodynamics of unsaturated flow and recharge efficiency, Liang et al. (2018) proposed a mathematical method for MAR induced by a Vadose zone infiltration well. Groundwater quality during MAR was sampled and monitored regularly at the Krasnohorska karst system in Slovakia (Malík et al. 2020). Urrutia et al. (2019) investigated the contribution of snowmelt and rainfall to aquifer recharge in the Andean Cordillera of northern Chile. Yin et al. (2011) analyzed the sustainable yield of karst water in Huaibei, China, using an artificial neural network model. The MAR system in karst spring catchment was simulated using the latest research version of MODFLOW-CFP (Xu et al. 2015). The influences of an inter-basin surface water transfer project on the MAR system and groundwater ecosystem were evaluated using a numerical simulation model in arid northwestern China (Zhu et al. 2018). Gong et al. (2016) developed a salt balance model in Badain Jaran Desert, China, to evaluate the salt accumulations in the groundwater-fed lakes. Lysimeter measurement and soil moisture models were used in the Huaihe River plain, China, to estimate groundwater recharge caused by precipitation and groundwater loss (Chen et al. 2008). The effects of clogging and dissolution in the MAR system were explored using hydrogeochemical and numerical simulation methods (Du et al. 2013; Ye et al. 2019).

Jinan is famous for its spring water and has been hailed as “Spring City.” The abundant spring water resources have made tremendous contributions to the socio-economic development in Jinan. Recently, the contradictions between water supply and demand have become increasingly prominent, with the rapid urbanization and increasing urban population (Jiao et al. 2017). In this case, Jinan has to exploit local groundwater to address the current situation of water shortages. However, a continuous decrease has occurred in groundwater level and free outflow rate of springs due to long-term overexploitation of urban groundwater resources (Wang et al. 2019). This situation has progressively become major threats to the protection of spring water resources and the safety of urban water supply in Jinan (Wang et al. 2017a, b). To realize the sustainable management of groundwater resources, it was urgent to carry out the MAR project in Jinan.

The aquifer system in Jinan spring catchment is a heterogeneous system, and the relationship between replenishment and discharge is extremely complex (Wang et al. 2015a, b; Gao and Tian 2018). And numerous empirical studies associated with groundwater management had been conducted in Jinan. The three-dimensional geological models were established to investigate the hydrogeological structure and groundwater flow characteristics in Jinan (Wu and Xu 2005; Kang et al. 2011; Wang et al. 2016; Luo et al. 2020). Qi et al. (2016) used the wavelet analysis method to predict the groundwater table in the Jinan spring catchment and then proposed sustainable groundwater extraction. To optimize the layout of groundwater recharge and extraction, the hydraulic connectivity in the Jinan spring catchment was explored (Wang et al. 2017a, b; Zhang et al. 2018). Hydrochemistry ions were used to study the recharge and discharge controls and groundwater flow paths in the Jinan spring catchment (Zhang et al. 2017b; Xing et al. 2018). Environmental isotope tracer tests were employed to investigate the main recharged water source for the Jinan spring catchment (Yin et al. 2017; Gao et al. 2014).

The aforementioned studies had achieved promising results and laid the foundation for optimizing regional MAR operations and maintaining water resilience. Although numerous micro and macro studies on MAR have been conducted, some investigations associated with multiple recharge water sources were scarce during the MAR process. Additionally, most current researches focus on the hydrogeochemistry of groundwater, almost ignoring the change in the state of groundwater quality. And there were few quantitative studies on the combination of comprehensive evaluation and high-risk indexes in groundwater quality. Therefore, it was necessary to investigate the impact of MAR with multi-source water on groundwater quality, especially in Jinan with groundwater depletion.

To achieve sustainable management of groundwater in Jinan, this study quantitatively investigated the impact of multi-source MAR on groundwater quality and high-risk indexes. The large-scale recharge experiments using multiple water sources were conducted in a natural river that the riverbed has high permeability. Groundwater and recharge water sources were sampled and tested to quantify the adverse effects of different recharge water sources on local groundwater quality, under a limited recharged quantity and period. An improved matter-element model was developed to implement a comprehensive assessment of groundwater quality, and the high-risk indexes in groundwater quality were determined during different water source recharge. The combination of two methods provides a new approach to investigate the MAR with multi-source water from multiple angles (the overall situation and high-risk indexes in groundwater quality). We hope this method could be extended to other MAR with multi-source water to optimize the operation of groundwater replenishment and reduce the damage of the recharged water source on the local aquifer system. Accordingly, the effects of MAR with multi-source water were assessed from the perspective of groundwater quality control, and countermeasures were then developed to provide a basis for formulating the high-efficiency and low-risk MAR mode.

Study area

Jinan spring catchment and Yufu River

Jinan’s terrain slopes from the south to the north with an elevation difference of more than 500 m (Yu et al. 2014). This unique geological structure makes the surface water and groundwater in Mount Tai flow into the northern urban area. Large quantities of groundwater are blocked by impermeable igneous rocks to form the high water pressure zones in northern Jinan (Xie 2017). After that, the groundwater flows out to form the Jinan spring groups under the action of strong water pressure (Fig. 1). As a symbol of Jinan’s social culture, the Jinan spring groups have made a historical contribution to the economic prosperity (Zhang et al. 2017a).

Fig. 1
figure 1

Jinan spring catchment and Yufu River (Zhang et al. 2018)

Full size image

To alleviate the pressure of spring water protection and urban water supply, the MAR projects in Yufu River have been carried out in Jinan. And the surface water in the Wohushan Reservoir (WR), the Yellow River (YR) and the South-to-North Water Diversion Project (SN) are used as the water sources for MAR, respectively. These MAR projects have played important roles in replenishing local groundwater and maintaining higher groundwater level of the Jinan spring groups. Although the MAR projects have been gradually implemented, there are still some challenges and obstacles. Due to the different recharge water sources and relatively complicated geological structure in Jinan, how to efficiently replenish local groundwater, avoid groundwater pollution, and sustain water resilience are the critical links.

Existing MAR projects in Yufu River

Yufu River is located in western Jinan. It originates from the mountain streams of Jinxiu, Jinyang, and Jinyun in the mountainous area of southern Jinan (Fig. 2). The rainfall and mountain streams are collected into the Wohushan Reservoir, and the Yufu River is then formed by the Wohushan Reservoir release (Hou et al. 2016). The study area is located in the upper reaches of the Yufu River. The riverbeds and river floodplains are mainly composed of loose Quaternary sediments (gravel, sand, and clay) with a thickness of about 24 m. And the strata are mainly Cambrian limestone with a thickness of about 150 m. This part of the Yufu River has a high perviousness based on its fully developed karst conditions. Consequently, the Yufu River plays an important role in groundwater replenishment and urban water supply in Jinan. Currently, the MAR projects in the Yufu River mainly include the following: (i) Yufu River is replenished by the WR release, and the water source of this MAR project is the local surface water in WR; (ii) surface water from YR is transferred to the upper reaches of Yufu River through the pumping station, and water source of this MAR project is from the YR; and (iii) surface water from the Jiping Main Canal is transported into the upper reaches of Yufu River employing water pipes, and water source of this MAR project is from the SN. The aforementioned MAR projects in Yufu River are shown in Fig. 2.

Fig. 2
figure 2

Existing MAR projects in Yufu River

Full size image

The water source and recharge quantity of MAR projects

As shown in Table 1, the water sources of MAR were from the WR, YR, and SN, respectively. The water quality and quantity of recharge water sources are two main reasons for groundwater quality changes after MAR. The recharge quantity was 1120–1170 (104m3) and the recharge period was from December to March of the next year in 2015–2018. Recharge quantity and period were well controlled in this study. So, the groundwater quality changes after recharge were mainly subjected to the water quality of recharge water source.

Table 1 Recharge water source and quantity in Yufu River from 2015 to 2018
Full size table

Monitoring wells and groundwater sampling

Eight monitoring wells (W1–W8) along the upper reaches of Yufu River were selected to monitor the groundwater quality. These monitoring wells consisted of three wells (W1–W3) in porous aquifer and five wells (W4–W8) in karst aquifer (Fig. 3). The depth of monitoring well was 24 m and 150 m in porous medium formation and karst formation, respectively. To determine the background level (BL) of groundwater quality, the groundwater in each monitoring well was sampled to test its water quality before MAR. Groundwater was directly sampled from a specific depth using the Discrete Interval Sampler, Model 425 (Solinst Canada Ltd., Georgetown, Ontario, Canada), and collected water could be highly representative of groundwater surrounding the monitoring well at that level (Parker and Clark 2004). The Discrete Interval Sampler was a passive sampler and was pressurized when lowered and upon retrieval to prevent water mixing at different levels (Dermatas et al. 2015). Furthermore, the water quality conditions of recharge water sources (WR, YR, and SN) were obtained through the sampling analysis.

Fig. 3
figure 3

The monitoring wells in Yufu River

Full size image

Materials and methods

Index system

Groundwater quality data from eight monitoring wells were served as the evaluation indexes. Environmental Quality Standards for Groundwater, China (GB/T14848-2017), were used as the evaluation standards. The groundwater quality was regularly sampled and tested from 2015 to 2018 during different water source recharge process. The test results showed that some water quality indexes in pore and karst water were detected, including total hardness (TH), nitrite nitrogen (NO2-N), nitrate nitrogen (NO3-N), ammonia nitrogen (NH4+), chlorides (Cl−), fluoride (F−), sulfate (SO42−), sodium ions (Na+), permanganate index (CODMn), pondus hydrogenii (PH), turbidity (TUB), and total dissolved solids (TDS). Previous research in Yufu River indicated that other groundwater quality indexes listed by Environmental Quality Standards for Groundwater, China (GB/T14848-2017), had less content or had not been detected in the study area, in addition to the above twelve groundwater quality indexes (Li et al. 2017b). So, the index system of groundwater quality was constructed using the above twelve indexes. The grades of groundwater quality were divided into five classifications (I, II, III, IV, and V) based on Environmental Quality Standards for Groundwater, China (GB/T14848-2017). Classifications of evaluation indexes are displayed in Table 2.

Table 2 Grades of groundwater quality in Environmental Quality Standards for Groundwater, China (GB/T14848-2017)
Full size table

Improved matter-element model

Matter-element model was presented by the Chinese researcher Cai Wen in 1983 (Cai 1999). Rules and methods for solving specific problems were qualitatively and quantitatively investigated according to the formalized logic tools (Li et al. 2017a; Liu et al. 2019). The matter-element method was the theoretical foundation, the logical unit was the matter-element, and a matter-element model was then developed by the characteristics and characteristic values of the evaluation object (Liu and Zou 2012). This model can quantify qualitative indexes and also deal well with the diverse and ambiguous evaluation objects.

Determination of the classical field, controlled field, and matter-element

The evaluation objectNhas a characteristiccwith the value of v. The ordered tripleR = (N, c, v) is composed of N, c, and v, which serves as the basic element for describing things. R is called the matter-element, and the three elements N, c, and v are used to establish the basis for matter-element theory. Assume that the object N can be described by m characteristics c1, c2, ⋯, cm, and the corresponding characteristic values are v1, v2, ⋯, vm. At this point, R is an m-dimensional matter-element, and defined as follows:

$$ R=\left(N,C,V\right)=\left[\begin{array}{c}{R}_1\\ {}{R}_2\\ {}\vdots \\ {}{R}_m\end{array}\right]=\left[\begin{array}{ccc}N& {c}_1& {v}_1\\ {}& {c}_2& {v}_2\\ {}& \vdots & \vdots \\ {}& {c}_m& {v}_m\end{array}\right] $$
(1)

where R is the matter-element, Ri is the sub-matter-element of R, N is the evaluation object, C = [c1, c2, ⋯, cm] is the eigenvector, V = [v1, v2, ⋯, vm] is the value corresponding to the eigenvector.

Suppose:

$$ {R}_j=\left[\begin{array}{ccc}{N}_j& {c}_1& {v}_{j1}\\ {}& {c}_2& {v}_{j2}\\ {}& \vdots & \vdots \\ {}& {c}_n& {v}_{jn}\end{array}\right]=\left[\begin{array}{ccc}{N}_j& {c}_1& \left[{a}_{j1,}{b}_{j1}\right]\\ {}& {c}_2& \left[{a}_{j2,}{b}_{j2}\right]\\ {}& \vdots & \vdots \\ {}& {c}_n& \left[{a}_{jn,}{b}_{jn}\right]\end{array}\right] $$
(2)

where Nj represents j evaluation grades, c1, c2, ⋯, cn are the characteristics of Nj, vj1, vj2, ⋯, vjn are the value ranges corresponding to the c1, c2, ⋯, cn, and vji(vji = [aji,bji]) is called the classical field.

Similarly,

$$ {R}_P=\left[\begin{array}{ccc}P& {c}_1& {v}_{p1}\\ {}& {c}_2& {v}_{p2}\\ {}& \vdots & \vdots \\ {}& {c}_n& {v}_{pn}\end{array}\right]=\left[\begin{array}{ccc}P& {c}_1& \left[{a}_{P1},{b}_{P1}\right]\\ {}& {c}_2& \left[{a}_{P2},{b}_{P2}\right]\\ {}& \vdots & \vdots \\ {}& {c}_n& \left[{a}_{pn},{b}_{pn}\right]\end{array}\right] $$
(3)

where P is all the grades in evaluation objects, vp1, vp2, ⋯, vpn are the value ranges corresponding to the c1, c2, ⋯, cn, and vPi (vpi = [api,bpi]) is called the controlled field of P.

The evaluation objects are represented by the matter-element.

$$ {R}_0=\left[\begin{array}{ccc}{P}_0& {c}_1& {v}_1\\ {}& {c}_2& {v}_2\\ {}& \vdots & \vdots \\ {}& {c}_n& {v}_n\end{array}\right] $$
(4)

where R0 is the matter-element to be evaluated, v1, v2, ⋯, vn are the specific data of P0 about c1, c2, ⋯, cn.

Calculation of the combined weights

The conventional combined weighting methods include multiplication and linear normalization. And they often have some limitations, such as the multiplication normalization method is only applicable to the cases where the number of indexes is large and the weight distribution between indexes is relatively uniform (Wang et al. 2009; Tang and Zhang 2013). It is difficult to determine accurately the preference coefficient of the linear weighting method, and it is usually applied based on experiences (Liu et al. 2020a, b; Zhang and Li 2020). To overcome these primary imperfections of conventional weighting methods, we propose a new combined weighting method based on the advantages of the subjective and objective weighting methods. There are two issues needed to be considered behind this combined weight. In the first place, the decision-makers’ preference for index attributes should be considered. The other point is that the information content provided by data itself should be used as much as possible. In view of this, the combined weighting method was proposed to make the weights and decision results become more reliable in the decision-making process. Detailed steps were as follows.

  1. (i)

    Definition of ideal solution

    Assume that the weights of evaluation indexes obtained by subjective weighting method (analytic hierarchy process (AHP)) are u = (u1, u2, …, un), and the weights obtained by objective weighting method (information entropy) are h = (h1, h2, …, hn), along with the combined weights obtained by the ideal solution method are w = (w1, w2, …, wn).

  2. (ii)

    Deviations of different weighting methods

    Unitization of weight vectors can be expressed as:

$$ {u}_j^{\prime }=\frac{u_j}{\sqrt{u_1^2+{u}_1^2+\cdots +{u}_n^2}};\kern0.33em {h}_j^{\prime }=\frac{h_j}{\sqrt{h_1^2+{h}_1^2+\cdots +{h}_n^2}};\kern0.66em {w}_j^{\prime }=\frac{w_j}{\sqrt{w_1^2+{w}_1^2+\cdots +{w}_n^2}} $$
(5)
  1. (iii)

    Calculation of combined weight

    \( {w}_1^{\hbox{'}},{w}_2^{\hbox{'}},\cdots, {w}_n^{\hbox{'}} \) can be obtained according to Eq. (6).

$$ {w}_j^{\prime }={\left[\frac{u_j^{\prime 2}+{h}_j^{\prime 2}}{2}\right]}^{\frac{1}{2}} $$
(6)

Normalizing \( {w}_1^{\hbox{'}},{w}_2^{\hbox{'}},\cdots, {w}_n^{\hbox{'}} \) to get the combined weight wj, the formula can be expressed as:

$$ {w}_j=\frac{w_j^{\prime }}{\sum \limits_{j=1}^n{w}_j^{\prime }},\left(j=1,2,\cdots n\right) $$
(7)

A detailed derivation process of the above model is described in the Supplementary Material.

Calculation of variable eigenvalues

The distance between matter-element and classical field is derived with Eq. (8).

$$ {\rho}_{\left({x}_0,{v}_{ji}\right)}=\left|{x}_0-\frac{1}{2}\left({a}_{ji}+{b}_{ji}\right)\right|-\frac{1}{2}\left({b}_{ji}-{a}_{ji}\right) $$
(8)

where \( {\rho}_{\left({x}_0,{v}_{ji}\right)} \) is the distance between matter-element and classical field, x0 is the point value on the real axis, aji and bji are the left and right endpoints of interval vji, respectively.

  1. (i)

    Calculation of the comprehensive correlation degree (CCD)

    The correlation function between evaluated matter-element and each grade can be established by Eq. (9).

$$ {K}_j\left({x}_0\right)=\left\{\begin{array}{c}\frac{\rho_{\left({x}_0,{v}_{ji}\right)}}{\rho_{\left({x}_0,{v}_{Pi}\right)}-{\rho}_{\left({x}_0,{v}_{ji}\right)}}\kern0.72em {x}_0\notin {v}_{ji}\\ {}-\frac{\rho_{\left({x}_0,{v}_{ji}\right)}}{\left|{v}_{ji}\right|}\kern3.12em {x}_0\in {v}_{ji}\end{array}\right. $$
(9)

where Kj(x0) expresses the value of correlation function for evaluated index x0 related to the jth grade, \( {\rho}_{\left({x}_0,{v}_{ji}\right)} \) represents the distance between the evaluated matter-element of index x0 and the interval vji in the classical field, \( {\rho}_{\left({x}_0,{v}_{pi}\right)} \) is the distance between the evaluated matter-element of index x0 and the interval vpi in the controlled field, and |vji| (|vji| = |bji − aji|) is the value range of the jth grade in the classical field.

Correlation function Kj(xi) is replaced by CCD.

$$ {K}_j\left({P}_0\right)=\sum \limits_{i=1}^n{w}_i{K}_j\left({x}_i\right) $$
(10)

where Kj(P0)expresses the CCD of evaluated matter-element P0 related to the jth grade, wi is the combined weight of evaluated indexes, Kj(xi) is the correlation function of the ith evaluated index related to the jth grade.

  1. (ii)

    Grade rating and variable eigenvalues calculation

    If Kj0(P0) = max Kj(P0), (j = 1, 2, …, m), then we conclude that the evaluated matter-element P0 belongs to the grade j0.

    The comprehensive grade of the evaluation object can be determined by \( \overline{K_j}\left({P}_0\right) \) set.

$$ \overline{K_j}\left({P}_0\right)=\frac{K_j\left({P}_0\right)-\underset{1\le j\le m}{\min }{K}_j\left({P}_0\right)}{\underset{1\le j\le m}{\max }{K}_j\left({P}_0\right)-\underset{1\le j\le m}{\min }{K}_j\left({P}_0\right)} $$
(11)

where \( \overline{K_j}\left({P}_0\right) \) is the CCD corresponding to the jth grade, \( \underset{1\le j\le m}{\min }{K}_j\left({P}_0\right) \) and \( \underset{1\le j\le m}{\max }{K}_j\left({P}_0\right) \) are the minimum and maximum values of CCD in all grades, respectively.

The variable eigenvalue (VE) of matter-element P0 is:

$$ {j}^{\ast }=\frac{\sum \limits_{j=1}^mj\cdotp \overline{K_j}\left({P}_0\right)}{\sum \limits_{j=1}^m\overline{K_j}\left({P}_0\right)} $$
(12)

where j is the evaluated comprehensive grade, j∗ is the VE of matter-element P0, and j∗ can be used as a reference for distinguishing the degree of the evaluated matter-element tend to adjacent grade.

Results

Hydrochemical types in MAR water source and BL of groundwater

The main ion compositions in MAR water source and BL of groundwater were analyzed using the Piper diagram (Fig. 4). The most abundant ions in BL of pore and karst water were Ca2+ that accounted for 60–70% of the total cation content, and HCO3− that was up to 70–80% of the total anion content. Accordingly, the hydrochemical type in BL of groundwater was of the HCO3-Ca type. This indicated that the ion compositions of groundwater were single before MAR, and the BL were relatively good in the study area. Compared with the ion compositions of BL, the ion changes in WR and YR primarily included that the Ca2+ and HCO3− content was decreased, and the Mg2+ and SO42− content was increased. In light of this, the hydrochemical type in WR and YR was of the HCO3·SO4-Ca·Mg type. When the SN was used as MAR water source, the most abundant cation was Na+ that accounted for 50–60% of the total cation content, and the relatively abundant anions were HCO3−, Cl−, and SO42−. So, the hydrochemical type was of the SO4·HCO3·Cl-Na·Ca type in SN. Overall, the hydrochemical types of pore and karst water were similar, and of the WR and YR were also close. However, the ion compositions in SN were quite different from the BL, WR, and YR.

Fig. 4
figure 4

The ion compositions in MAR water source and BL of groundwater

Full size image

Groundwater quality indexes and weights

The groundwater quality indexes (C1–C12) were sampled and tested from 2015 to 2018 during different water source recharge process. And the data of groundwater quality indexes are shown in Fig. 5. The mean and distribution range of groundwater quality indexes were the smallest in BL, compared with the groundwater quality during WR, YR, and SN recharge. This indicated that the BL of groundwater quality was relatively good, and the changes of BL were small before MAR. Overall, an obvious upward trend appeared in groundwater quality indexes during WR, YR and SN recharge process, and the groundwater quality gradually deteriorated during MAR. Moreover, the weights of the groundwater quality index were determined by AHP and information entropy method, respectively. The combined weights were then calculated according to Eqs. (5)–(7). Finally, the improved matter-element model was developed using combined weights. The combined weights of groundwater quality indexes are shown in Fig. 6.

Fig. 5
figure 5

Groundwater quality indexes

Full size image
Fig. 6
figure 6

Weights of groundwater quality index based on different weighting methods. And uj, hj, and wj were the weights obtained by the AHP, information entropy, and ideal solution methods, respectively

Full size image

Evaluation of groundwater quality during different water source recharge

The CCD of evaluated matter-element

The CCD of evaluated matter-element related to the classifications standard was calculated during different water source recharge. If the CCD associated with the comprehensive grade was the largest, the evaluation object belonged to this grade. The largest CCD of BL appeared in grades I and II (Fig. 7a). In light of this, the comprehensive grades of BL belonged to grades I and II. The CCD decreased in grade I, and gradually increased in grades II and III, during the WR recharge process (Fig. 7b). It showed that the groundwater quality of monitoring wells was getting worse, and its comprehensive grade was gradually approaching grades II and III. When the YR was used as the source of MAR, the CCD continuously decreased in grade I and increased in grades III and IV (Fig. 7c), meaning that a continuing downward trend occurred in groundwater quality. Compared with BL, WR, and YR, the CCD increased significantly in grades III, IV, and V, during the SN recharge process (Fig. 7d), that is, the comprehensive grade of groundwater quality was changing from grades I and II to grades III, IV and V. Overall, the CCD decreased in grades I and II, increased in III, IV, and V, when the sources of MAR were WR, YR, and SN.

Fig. 7
figure 7

The CCD of evaluated matter-element. a BL. b WR. c YR. d SN

Full size image

The comprehensive grade and VE of groundwater quality

The comprehensive grade of groundwater quality could be judged according to the CCD. Meanwhile, the trends in groundwater quality were further revealed by the changes in VE. Based on this, the influence of surface water sources on regional groundwater quality was evaluated in different water source replenishment processes. The higher the comprehensive grades and VE were, the poorer the groundwater quality. As shown in Figs. 8 and 9, the comprehensive grades were all grade II in BL, except monitoring well W6 was grade I. Moreover, the VE ranged from 2.050 to 2.064 with a mean value of 2.056 in porous aquifer (W1–W3) and ranged from 1.940 to 2.005 with a mean value of 1.979 in karst aquifer (W4–W8). In summary, the BL were relatively good, and karst water quality was slightly better than pore water quality.

Fig. 8
figure 8

The comprehensive grades of groundwater quality during different water source recharge

Full size image
Fig. 9
figure 9

The VE of groundwater quality during different water source recharge

Full size image

When the WR was as source of MAR, the comprehensive grades in W1 and W6 were increased to grades III and II, respectively, and the significant changes did not occur in the remaining wells. The VE ranged from 2.184 to 2.220 with a mean value of 2.205 in the porous aquifer, and ranged from 2.054 to 2.118 with a mean value of 2.098 in the karst aquifer. Overall, the VE in pore and karst water quality showed an upward trend after WR recharge, which indicated that a slightly deteriorated trend existed in groundwater quality.

During the YR recharge, the comprehensive grade was grade III in the porous aquifer, and was grade II in the karst aquifer. In addition, the VE ranged from 2.308 to 2.328 with a mean value of 2.321 in porous aquifer, and ranged from 2.208 to 2.258 with a mean value of 2.229 in the karst aquifer. Comparing with the BL and WR, the VE gradually increased. These indicated that the deteriorating trend occurred relative to the BL during the YR recharge process. Meanwhile, the YR had a greater adverse impact on regional groundwater quality than WR.

The comprehensive grades in porous and karst aquifer were all grade III, when SN was the recharge source. The VE ranged from 2.472 to 2.508 with a mean value of 2.488 in the porous aquifer, and ranged from 2.308 to 2.378 with a mean value of 2.330 in the karst aquifer. The comprehensive grades and VE were the largest during the SN recharge process, compared with WR and YR. These meant that the adverse impact of the SN on regional groundwater quality was greatest relative to WR and YR.

Above knowable, the comprehensive grades and VE of groundwater quality increased in turn and the local groundwater quality showed a deteriorating trend, when the water sources of MAR were WR, YR, and SN, respectively. Meanwhile, the adverse effects of different recharge water sources on regional groundwater quality were SN > YR > WR. Moreover, the VE of pore water quality was slightly larger than that of karst water quality during different water source recharge. This indicated that the karst water quality was better than the pore water quality.

Discussion

The dynamic evolution of VE

The dynamic evolution of VE was investigated based on Gaussian kernel density estimation (Fig. 10). A distribution curve with a single peak occurred in BL and double peaks appeared in WR, YR, and SN. The left peak of the bimodal distribution curve represented the karst aquifer (W4–W8), and the right peak was porous medium aquifer (W1–W3). The distribution curve changed from a single peak to double peaks, and the right part of curve extended to the right. In addition, the peak width gradually increased, and both the agglomeration center and right peak had a right-shift trend. These features indicated that the groundwater quality showed a downward trend during different water source recharge. Notably, the difference in groundwater quality between the porous medium aquifer and karst aquifer increased significantly after the SN recharge. The right-shift distance of the distribution curve in turn was SN > YR > WR > BL. Overall, the adverse effects of different recharge water sources on regional groundwater quality were SN > YR > WR > BL.

Fig. 10
figure 10

The Gaussian kernel density estimation of VE

Full size image

The high-risk indexes in groundwater quality

The grades corresponding to the maximum content of groundwater quality index in porous aquifer (W1–W3) during different water source recharge are shown in Fig. 11a. The pore water quality in BL was all grades I and II, except TDS. The contents of TH and NO3-N increased to grade III after WR recharge. An upward trend occurred in contents of TH, SO42−, and TUB after the YR recharge, wherein the TUB increased significantly to grade IV relative to the BL. TH, NO3-N, SO42−, TUB, and TDS all presented an upward trend, when SN was used as the recharge water source. And the contents of TH, TUB, and TDS increased significantly, compared with the BL.

Fig. 11
figure 11

Grades of the maximum content of groundwater quality index. a Grades of groundwater quality index in porous aquifer. b Grades of groundwater quality index in karst aquifer

Full size image

As shown in Fig. 11b, the karst water quality in BL was generally good, and the grades of groundwater quality indexes were all grades I or II, and grade III was not presented. This indicated that the BL of karst water quality were generally better than that of pore water quality. The contents of NO3-N, NH4+, and TDS increased to grade III after WR recharge. TH, NH4+, SO42−, and TDS showed an upward trend during the YR recharge, the contents of them increased to grade III. During the SN recharge, TH, NO3-N, NH4+, SO42−, and TDS displayed a growth trend, wherein the content of TDS had a huge rise to grade IV.

With the rapid socio-economic development, large-scale wastewater and fertilizers generated by production and living were discharged into natural waters, increasing the contents of TH and N in the rivers and lakes (Zhang et al. 2018). Therefore, the content of TH, NO2-N, NO3-N, and NH4+ in pore and karst water might be increased, when WR, YR, and SN were used as MAR water sources. F− and CODMn were less affected by the recharge water sources, the main reason was that the Quaternary sediments (sand and gravel) had a strong adsorption effect on F− and CODMn (Li et al. 2017b). Due to the high content of Cl−, SO42−, and Na+ in YR and SN, these ions in pore and karst water had an upward trend during the YR and SN recharge process. The TUB in groundwater is affected by the suspended solids in the recharged water source. TUB in pore water represented an increasing trend, while that in karst water had not changed significantly after YR and SN recharge. Additionally, an obvious upward trend appeared in the TDS of pore and karst water, especially after the SN recharge, due to the use of recharge water sources with high TDS.

In summary, a significant increase had occurred in the contents of some groundwater quality indexes during different water source recharge, which might further induce deterioration of regional groundwater quality. According to the aforementioned analysis, TH, NO3-N, NH4+, and TDS were the high-risk indexes of groundwater quality when the WR was used as the recharge water source. The high-risk indexes were TH, NH4+, SO42−, TUB, and TDS in the YR recharge process. And TH, NO3-N, NH4+, SO42−, TUB, and TDS were the high-risk indexes during the SN recharge. Accordingly, these high-risk indexes should be monitored emphatically to reduce the adverse effects on the regional aquifer system, when local groundwater was replenished using the aforementioned surface water sources.

Measures and suggestions

An efficient and low-risk mode for multi-water source recharge was explored by combining the aforesaid results with the existing MAR projects in Jinan. The following measures and suggestions were put forward.

  1. 1.

    The surface water in WR should be used preferentially to replenish the local groundwater resources when the water storage capacity of WR was sufficient. In contrast, if the water storage capacity of WR was insufficient, the YR should be preferentially used as recharge water source. Surface water in SN was considered to use, only when insufficient water supply exists in WR and YR.

  2. 2.

    Whether the replenished water sources were from the WR, the YR, or the SN, the regional groundwater quality will all be affected adversely during the recharge process. These may also lead to the accumulation of some potential high-risk indexes in the aquifer. Therefore, it is necessary to control strictly the proportion of total recharge quantity to regional groundwater resources. The contents of high-risk indexes in recharge water sources should be reduced as much as possible before recharge, and the abovementioned high-risk indexes in regional groundwater should be monitored emphatically during recharge process.

Study limitations

Due to the restricted spatial variability of the karst aquifer system and availability of data, this study also has some limitations. We only used the groundwater quality data of eight monitoring wells to evaluate the trends in groundwater quality during different water source recharge process, due to the limitation of the number of monitoring wells in the study area. In addition, as a seasonal river, the Yufu River without MAR generated surface runoff only in the flood season (July to August each year), and no surface runoff in the remaining months. During non-flood seasons, the groundwater level dropped below the bottom of the porous aquifer, and wells in porous medium formation were usually in a dry state. Therefore, this study focused on the investigation of groundwater quality when the groundwater level was higher than the bottom of the porous aquifer during MAR. Possible future researches based on this study include the following: (i) increasing the number of monitoring wells according to the detailed hydrogeological and geophysical investigations; (ii) increasing the types of groundwater quality indexes (such as heavy metals and antibiotics); (iii) adjusting the layout of monitoring wells combining with monitoring sites in pore and karst water; and (iv) investigating the hydrogeochemistry and seasonal variation of groundwater quality based on the regular monitoring of the karst aquifer.

So far, the study area still had not established a MAR scheme to achieve the conjunctive use of surface water and groundwater, and the joint management of water quality and quantity. Therefore, the dynamic changes between water quality and quantity cannot be combined effectively in the MAR with multi-source water. In future researches, the trends in groundwater quality should be investigated dynamically in conjunction with the recharge quantity of different water sources.

Conclusion

In this study, the Yufu River with multiple water sources was used as the research area to carry out large-scale recharge experiments. Under a limited recharged quantity and period, the water quality of groundwater and recharge water source was sampled and tested to study quantitatively the adverse effects of different recharge water sources on local groundwater quality. Meanwhile, the trends in regional groundwater quality were compared and analyzed to investigate the MAR with multi-source water. An improved matter-element model was developed to execute a comprehensive assessment of water quality, and the high-risk indexes in groundwater quality were determined during different water source recharge. The conclusions obtained were as follows.

The hydrochemical type in pore and karst water was of the HCO3-Ca type before recharge, which indicated that the ion composition was relatively simple, and the BL were relatively good. Furthermore, the hydrochemical type of the WR and YR were relatively close, both of which were of the HCO3•SO4-Ca•Mg type. The hydrochemical type of the SN was quite different from the BL, WR, and YR, and was of the SO4•HCO3•Cl-Na•Ca type. Overall, the water quality of the recharge water source was worse than the BL, and the SN was worst in MAR water sources. Under a limited recharged quantity and period, the adverse effects of different recharge water sources on regional groundwater quality were SN > YR > WR. Therefore, the long-term and large-scale use of three recharge water sources to replenish groundwater resources might deteriorate the groundwater environment quality and damage the aquifer system.

The combination of improved matter-element model and groundwater quality high-risk indexes provides a new approach to investigate the MAR with multi-source water from multiple perspectives (the overall situation of groundwater quality and the high-risk indexes). We hope this method could be extrapolated to other MAR with multi-source water to optimize the operation of groundwater replenishment and reduce the damage of the recharge water source on the local aquifer system.