1. Introduction
Groundwater is being overexploited in many regions of the world due to increasing demand for water resources brought about by rapid economic development and population growth [
1,
2]. It causes a variety of problems such as drawdown of groundwater levels, drying up of aquifers [
3], increase in the groundwater cones of depression [
4], and land subsidence [
5]. Intensive irrigation can lead to the development of salinity problems, and the extraordinary rise of piezometric surface in aquifers may induce groundwater inundation [
6], resulting in several damage processes such as building foundation destabilization, groundwater infiltration and pollutant remobilization [
7]. Thus, considerable attempts have been made to control groundwater exploitation and water level [
8]. In China, the State Council stipulated the implementation of controlled groundwater exploitation in those groundwater overexploited regions [
9].
Many models have been developed to optimize groundwater exploitation and address the impact of groundwater overexploitation. Commonly used methods for groundwater simulation include finite difference method, finite element method, boundary element method and finite volume method. Mehl and Hill [
10] proposed a new method of local grid refinement for two-dimensional block-centered finite-difference meshes in the context of steady-state groundwater flow modeling. Wang et al. [
11] proposed a groundwater flow domain decomposition model coupling the boundary and finite element methods. However, Anderson and Woessner [
12] pointed out the accuracy and reliability of groundwater numerical models depended critically not only on the simulation method but also on the properly generalized conceptual hydrogeological model. Recent decades have also witnessed significant progress in the development of groundwater simulation software based on the conceptual hydrogeological model, such as the groundwater simulation modeling system (GMS) [
13], modular three-dimensional finite difference groundwater flow model (MODFLOW) [
14], finite element groundwater flow modeling software (FEMWATER) [
15], and finite element subsurface flow system (FEFLOW) [
16]. These models have achieved remarkable success in investigating groundwater levels [
17], groundwater mass balance [
18], salt transport in coastal aquifers [
19,
20], groundwater quantity balance [
21], impact of predicted climate changes on groundwater flow systems [
22], sustainable groundwater management [
23,
24], and groundwater irrigation [
25]. However, given the “natural-artificial” dualistic characteristics of the water cycle system [
26], water resources allocation can have significant impacts on the groundwater recharge and discharge, resulting in significant changes in groundwater exploitation and consequently changes in water level. Thus, groundwater numerical models coupled with optimal allocation of water resources can provide a more effective way to simulate groundwater in complex regions.
Artificial fish swarm algorithm [
27], multi-objective optimization [
28,
29,
30], interval-parameter multi-stage stochastic programming model [
31], ant colony optimization [
32], support vector machines and genetic algorithms [
33] have often been used in coupling groundwater-surface water models. These models make it possible for the dynamic allocation of water resources in different regions in reference year [
34]. Lu et al. [
35] showed that the coupling of water resources allocation models and groundwater numerical models reduced the allowable and overexploitable quantity of groundwater for quantifying the groundwater allowable withdrawal accurately in planning years.
Groundwater level can be indicative of groundwater quantity and underflow, and thus, it is an important index in groundwater management. Knowledge of spatial and temporal changes in groundwater levels following the optimal allocation of water resources is essential to better understand the stability of groundwater environment [
36]. Jang et al. [
37] quantified the recovery of groundwater levels in townships when groundwater for drinking and agricultural demands was replaced by surface water based on the groundwater-surface water coupled model. In order to more accurately control the groundwater level in the canal-well irrigation district, Su et al. [
38] simulated the spatiotemporal groundwater depth in planning year using the optimal allocation model of water resources coupled with the groundwater numerical model. Stefania et al. [
39] modeled groundwater/surface-water interactions in an Alpine valley (the Aosta Plain, NW Italy) to investigate the effects of groundwater abstraction on surface-water resources.
In summery, previous studies have focused mainly on the modification of numerical modeling and groundwater exploitation calculation [
40,
41,
42], while changes in future groundwater levels were seldom considered. Thus, this study aims to consider changes in future groundwater levels, as well as changes in groundwater exploitation in water resources allocation. A water resources allocation model is constructed to predict future groundwater recharge and discharge, and the results are input into the groundwater numerical model to forecast changes in water levels. The established groundwater numerical model will be adopted to feedback the allocation results. Groundwater exploitation and water level are quantified by using multiple-iterated technique to achieve dual control. Also, groundwater allowable withdrawal and critical groundwater level are used as feedback factors to optimize the model. However, changes of groundwater exploitation and water level will affect the natural equilibrium state of groundwater system and environment. Thus, the proposed multiple-iterated dual control model makes contributions to hydrogeology.
The groundwater over-exploited region in Shenyang of northeast China is selected as the study area. The main purposes of this study are to (1) evaluate the performance of groundwater numerical model in simulating water level and calculate the critical groundwater level; (2) simulate spatial and temporal changes in groundwater levels and propose a scheme for sustainable groundwater management; (3) investigate spatial and temporal changes in groundwater levels under different precipitation conditions in the future; and (4) evaluate the performance of the multiple-iterated dual control model in controlling groundwater exploitation and water level.
2. Study Area and Methods
2.1. Study Area
The groundwater over-exploited region in Shenyang, the capital city of Liaoning Province in northeastern China, is selected as the study area (geographical coordinates, 41°11′51″–43°02′13″ N and 122°25′09″–123°48′24″ E, see
Figure 1). It is a plain area administratively divided into Urban District, Shenbei New District, Sujiatun District, Hunnan New District, Yuhong District and Development Zone with a total area of 2318 km
2. There are a total of 36 municipal water sources in the study area, as shown in
Figure 1, which are the main source of groundwater.
Figure 2a shows changes in groundwater exploitation and average water level in the study area over the period 2004–2013. Clearly, groundwater exploitation increases from 2005 to 2007 and then decreases from 2007 to 2009. However, it is noted that groundwater exploitation increases again to a maximum of 403.01 million m
3 in 2010. Groundwater level decreases to a minimum of 31.07 m in 2010, after which it increases continuously. Two water transfer projects (Dahuofang and Liaoxibei) were constructed in 2011, and the government of Liaoning Province had taken measures to close many groundwater sources to stop decline of water level due to groundwater overexploitation. Now, the transferred water are used instead of groundwater since 2012.
Figure 2b shows spatial distribution of groundwater level and location of monitoring wells in 2013. The spatial distribution of groundwater level is plotted based on the monitored groundwater level by the 114 monitoring wells. Due to the influence of topography, groundwater level is generally high in the northeast but low in the southwest. It is important to note that despite the decrease in groundwater exploitation quantity, groundwater overexploitation remains a serious problem in 2013, resulting in the formation of three cones of depression with a maximum groundwater depth of 20.77 m.
2.2. Research Framework and Simulation Model
Figure 3 shows the multiple-iterated dual control model for groundwater exploitation and water level, which consists of the following four modules, including optimal allocation module of water resources, groundwater allowable withdrawal calculation module, groundwater simulation module, and groundwater level control module. This study focuses mainly on the dual control for groundwater exploitation and water level and the interactions of the four modules, so the procedures for development of water resources allocation model can be referred to Zhou [
43]. The details are provided in the
Appendix A.
The model focuses on the coupling of water resources allocation and groundwater numerical simulation. The surface water supply and groundwater exploitation are coupled in the water resources allocation model to calculate field infiltration and well irrigation regression recharge of the groundwater numerical model. Water supply, demand and deficit are analyzed by different water demand schemes, and the rational allocation of water resources is put forward. The future groundwater level can be predicted by data extraction and interaction, and dual control for groundwater exploitation and water level can be achieved based on groundwater allowable withdrawal and critical groundwater level.
2.2.1. Optimal Allocation Module of Water Resources
The optimal allocation model of water resources consists of a number of objective functions, constraints and water balance equations. In the model long-time series hydrological and economic data and ecological water requirements in different years serve as the inputs. The calculation unit is based on the geographical locations of water resources and administrative regions. The constraints are the water balance constraints, and the objective is to maximize the net benefit of water supply and minimize water loss. The model is solved by the mathematical planning. In this study, the General Algebraic Modeling System (GAMS 2.5) [
44] is used to establish and solve the model.
(1) Objective function
The objective is to maximize the net benefit, minimize water loss, and ensure the precedence of water sources for water supply, as described in Equation (1).
where
is the vector composed of decision variables;
S is the feasible set of decision variables composed of different constraints; and
is the objective for the development of society and economy, ecological environment and water resources.
(2) Constraints and water balance equations
The constraints and water balance equations are described in Wei et al. [
45]. The groundwater supply constraints and balance equations in the calculation unit are shown in Equation (2).
where
XZGC,
XZGI,
XZGE,
XZGA, and
XZGR are the groundwater supply for urban domestic use, industrial use, urban ecological use, agricultural use, and rural domestic use (million m
3), respectively; PZGTU is the exploitable coefficient of groundwater in the calculation period;
PZGW is the annual groundwater availability (million m
3); tm is the calculation period; and J is the calculation unit.
In general, the main data of this module include the predicted water demand, water supply of municipal water sources, and model parameter. The supplemental materials and the first water allocation scheme are listed in
Table A1,
Table A2,
Table A3,
Table A4,
Table A5,
Table A6,
Table A7,
Table A8,
Table A9 and
Table A10.
2.2.2. Groundwater Allowable Withdrawal Calculation Module
Groundwater allowable withdrawal is used to judge whether groundwater in a given region is exploited reasonably. It is used as a feedback of the water resources allocation model. In the study area, the shallow groundwater is the main water source and it is phreatic water. The allowable withdrawal of shallow groundwater is defined as the maximum quantity of groundwater that can be extracted from the aquifer without causing environmental and geologic impacts on the premise of economic and technical feasibility. It can be calculated by mining coefficient method based on groundwater balance method [
46] (Equation (3)).
where
W is the groundwater allowable withdrawal (million m
3);
Wr is the quantity of groundwater recharged by precipitation infiltration, mountain and plain area infiltration, reservoir and riverway leakage, and other recharges (million m
3);
We is the discharge of groundwater resulting from lateral discharge, artificial exploitation, phreatic water evaporation, and other discharges (million m
3); Δ
S is the changes in groundwater level (m);
µ is the specific yield of phreatic water aquifer;
F is the area of equilibrium region (km
2); Δ
t is the time span of equilibrium period;
ρ is the exploitable coefficient, which is related to the long-term series groundwater data, aquifer type, and mining conditions. The calculation of
ρ is described in our previous study, and the iterative calculation of groundwater allowable withdrawal is listed in
Table A11.
2.2.3. Groundwater Simulation Module
GMS 7.1 consisting of several modules such as MODFLOW, FEMWATER, and MODPATH is used in this study to develop the groundwater simulation model, and the equations are composed of fundamental differential equations describing the three-dimensional unsteady groundwater flow in porous media, boundary conditions, and initial constraints (Equation (4)). The MODFLOW module is used for groundwater simulation in this study.
where
K is the aquifer permeability coefficient;
Kx, Ky, and
Kz are the component of the permeability coefficient along the
x,
y and
z directions, respectively (m/d);
W is the source term per unit volume (m
3/d);
μ is the specific yield of the phreatic water aquifer;
H is the groundwater level (m);
H0 is the initial water level (m);
B is the aquifer floor elevation (m);
q is the discharge per unit width under the second type boundary conditions (m
3/d/m);
x,
y and
z are the coordinates (m);
n is the inner normal on the boundary; and Γ1 and Γ2 are the first and second type boundary, respectively.
The accuracy of groundwater simulation is determined based on the average relative error (ARE), mean absolute error (MAE), and root-mean-square error (RMSE) (Equations (5)–(7)).
where
Hi and
Hi′ are the observed and calculated groundwater level (m), respectively, and
n is the length of sample series.
2.2.4. Groundwater Level Control Module
A critical groundwater level
Hc is set in this study in order to prevent potential environmental and geological impacts resulting from too high or too low groundwater levels, and it is established according to the function of different groundwater systems in different areas. The upper and lower limits of the critical groundwater level (
Hup and
Hlow) are determined for each region to control groundwater exploitation and water level more reasonably (Equations (8) and (9)).
where
H1 is the critical groundwater level for frost heaving and boiling (m),
Hfrost is the frost line (m),
H2 is the critical groundwater level for soil salinization (m),
H3 is the anti-floating design water level for underground orbit traffic (subway) (m),
H4 is the waterproof design water level for underground structures (m),
Huph is the historical maximum water level (m),
Hupl is the maximum water level in the recent 3–5 years (m), M is the aquifer thickness (m), and
h is the ground elevation (m), respectively.
The average critical groundwater level is calculated from groundwater levels of all monitoring wells by the Thiessen Polygons method described in Equation (10) [
47].
where
and
A are the average critical groundwater level (m) and area (km
2) for different regions, respectively;
ai is the area of the
ith calculation unit of the
ith Thiessen polygon,
i = 1, 2, …,
n (km
2);
Hci is the critical groundwater level of the
ith calculation unit of the
ith Thiessen polygon,
i = 1, 2, …,
n (m); and
n is the number of Thiessen polygons.
2.3. Multiple Iteration Processes
The multiple-iterated dual control model is derived from the coupling of the optimal allocation model of water resources and the groundwater numerical model:
Step 1: The topological and recharge-discharge relations among different water systems, water conservancy projects, and water users in reference year are analyzed. The objective functions and constraints are determined, and the optimal allocation model of water resources is established for the calculation of water demand-supply balance in planning year, which is referred to as the “first allocation scheme” in this study.
Step 2: The results obtained from the first allocation scheme are substituted into Equation (3) to calculate the groundwater allowable withdrawal Wj (where j is the number of iterations, j = 0, 1, 2, …, n). If |Wj+1 − Wj|/Wj ≤ ρ (ρ = 0.02), go to Step 3; otherwise let j = j + 1 and return to Step 1. This process is repeated until satisfactory results are obtained, which is referred to as the “second allocation scheme” in this study.
Step 3: The aquifer, boundary conditions, and hydrogeological parameters in the study area are generalized, and the conceptual hydrogeological model and groundwater numerical model are established. Subsequently, groundwater levels are calibrated and verified, and hydrogeological parameters are determined.
Step 4: The results obtained from the second allocation scheme are input into the verified groundwater numerical model to simulate changes in groundwater levels Hj (where j is the number of iterations, j = 1, 2, …, n). If Hlow ≤ Hj ≤ Hup and |Wj+1 − Wj|/Wj ≤ ρ, stop calculation; otherwise, adjust groundwater exploitation Qj and optimize water resources allocation again until the following three requirements are met: (1) the total groundwater supply is lower than groundwater allowable withdrawal and the total water consumption is lower than that mandated by government regulations, (2) water supply-demand balance and groundwater recharge-discharge balance are realized; and (3) the water deficit ratio of each unit is no more than 5%. This scheme is referred to as the “third allocation scheme” in this study.
2.4. Data Collection
The input data of the optimal allocation model of water resources include corrected monthly precipitation and runoff in 1956–2013, river flow data, groundwater recharge in 1980–2013, and social and economic data in water demand. The input data of the groundwater numerical model include water levels of monitoring wells, groundwater exploitation of municipal water sources, and groundwater recharge and discharge in 2007–2013. These data are provided by the water management institutes of Shenyang. Main model parameters include river parameters, water supply channel parameters, irrigation water use efficiency parameters, and reservoir parameters, which are determined by field research and expert consultation. Hydrogeological parameters and recharge coefficients of river and field infiltration are calibrated by the model.
4. Conclusions
In this study, a multiple-iterated dual control model is proposed for short-term and long-term groundwater resources management. This model integrates the optimal allocation model of water resources and the groundwater numerical model, thus making it possible to achieve dual control of groundwater exploitation and water level. It has been successfully applied to the groundwater simulation in Shenyang of Liaoning Province, China. The following conclusions could be drawn:
There is a good agreement between calculated and observed groundwater levels for Yuhong and Railway Machinery School wells over the period 2007–2012, indicating that the groundwater numerical model performs well in simulating groundwater levels with high accuracy.
The optimal allocation of water resources makes it possible for the attainment of water supply–demand balance and groundwater recharge–discharge balance. As a result, the groundwater exploitation reduces from 290.33 million m3 in 2013 to 87.05 million m3 in 2020 and 116.76 million m3 in 2030, respectively.
The controlled exploitation of groundwater results in a disappearance of cones of depression and a rapid recovery of groundwater levels in normal years. The average groundwater level increases from 34.27 m in 2013 to 36.63 m in 2035 in wet years, recovers to 34.72 m in normal years, and reduces to 31.77 m in dry years, respectively. The groundwater exploitation is controlled between the groundwater allowable withdrawal and the maximum groundwater exploitation, and the groundwater levels are controlled within the critical groundwater level.
Water demand predictions of social economy development and ecological environment consider sustainable development in the economy. The optimal allocation model of water resources realizes water supply–demand balance. Regional water resources are rationally allocated in order to achieve better economic and social development. When groundwater is overexploited, it recovers slowly. The multiple-iterated dual control model can be used in overexploitation areas in which surface and transferred water can be used to replace groundwater, and it contributes significantly to economic development, environmental protection, and sustainable groundwater exploitation.
However, there are some limitations in this model. For instance, although future precipitation changes and groundwater exploitation schemes have been considered in this model, some other uncertainties, such as temperature and groundwater quality, are not considered. This deserves further research in future studies.