Abstract

The likelihood of future global water shortages is increasing and further development of existing operational hydrologic models is needed to maintain sustainable development of the ecological environment and human health. In order to quantitatively describe the water balance factors and transformation relations, the objective of this article is to develop a distributed hydrologic model that is capable of simulating the surface water (SW) and groundwater (GW) in irrigation areas. The model can be used as a tool for evaluating the long-term effects of water resource management. By coupling the Soil and Water Assessment Tool (SWAT) and MODFLOW models, a comprehensive hydrological model integrating SW and GW is constructed. The hydrologic response units for the SWAT model are exchanged with cells in the MODFLOW model. Taking the Heihe River Basin as the study area, 10 years of historical data are used to conduct an extensive sensitivity analysis on model parameters. The developed model is run for a 40-year prediction period. The application of the developed coupling model shows that since the construction of the Heihe reservoir, the average GW level in the study area has declined by 6.05 m. The model can accurately simulate and predict the dynamic changes in SW and GW in the downstream irrigation area of Heihe River Basin and provide a scientific basis for water management in an irrigation district.

INTRODUCTION

The warming of the global climate system in the past century is unequivocal, as evidenced by changes in climate with an increasing frequency, intensity, duration, and spatial extension of heat waves (Kassian et al. 2017). The stresses imposed on water resources increase as a consequence of continued population increases and economic development in many parts of the world, resulting in undesirable impacts on both surface and groundwater (GW) systems. The challenge is how to best communicate such complex matters to nonspecialists and how to best manage the interacting surface and GW resources of a region to prolong their usefulness for present and future generations (Sophocleous et al. 1999; Zhang et al. 2012).

Surface water (SW) and GW management are closely related to each other. The interaction between SW and GW has been a hot and difficult issue in hydrology and hydrogeology research. Previous simulation studies have separated SW and GW because of various conditions. With the development of technology, modeling of the coupling between GW and SW has become possible. SW and GW have a very close connection and are not independent parts of the whole hydrological cycle system (Yu & Rui 2007).

Freeze & Harlan (1969) were the first researchers to present a coupled hydrological response model at the basin scale. Pikul et al. (1974) found that the Richards equation is highly accurate for simulating GW level changes. The interaction between GW and SW was simulated by Li et al. (2013) in the Toronto area. Hu et al. (2007) reviewed coupled simulations of SW and GW around the world and analyzed their applicability in China. However, because of the distinct mechanisms of the processes involved and the simulation approaches, the simulation results vary greatly among different models. In addition, the structure of the model is very complex and requires a large number of parameters and input data, which results in limitations in application; the two must be studied in a unified way. The conversion between SW and GW is an important part of the hydrological cycle in the irrigation area, since the distributed model development is time consuming.

Relevant basin information from reliable data is essential to assess not only the current condition of water resources in a given basin but also to determine past trends and future possibilities. To explore options for the future, the prediction of the future water cycle requires the use of coupled models to explore trends in future hydrologic cycles and to find ways to adapt to these as much as possible to achieve sustainable development.

The Heihe River Basin is located within the Guanzhong Plain region in China, which is a typical semi-arid plain area. Heihe River Basin is the main drinking water source for over eight million residents in Xi'an city in Shaanxi province. Suitable management of water resources in the Heihe River Basin is very important for the development of society and the economy in Shaanxi province, China.

Therefore, the objective of this study is to develop a comprehensive model that can reasonably and accurately simulate the interaction of SW, GW, and hyporheic zones. The model should play a key role in water resource management and allow the Division of Water Resources to observe the effects of implemented water-management policies at the basin scale (Peng et al. 2017). The model should simulate the hydrological balance for a long time scale, determine the impact on water resources through effect tests, predict the effects of policy implementation on water resource management, and make detailed analyses of the impacts of agricultural and other land uses of water resources in the area (Huo & Li 2013; Huo et al. 2016). The model also needs to be user-friendly enough so that water resources policy-makers and other interested parties can easily evaluate long-term water resource management strategies and public health issues.

STUDY AREA AND DATASETS

Irrigation area in the downstream Heihe River

The Heihe River Basin is located in the north of the Qinling Mountains, in the territory of Zhouzhi County, Shaanxi Province, south of Weihe, China. It is located between longitudes of 107° 43′ and latitudes of 108° 24′ and 33° 42′ and 34° 13′. The Heihe River Basin covers an area of 2,258 km2 upstream from the Jinpen reservoir. The south is higher than the north Heihe River Basin; the terrain tilts from southwest to northeast. The upstream area is in the Qinling Mountains. The middle reaches the Qinling Mountains Alluvial-Proluvial Fan and downstream is the Weihe River terrace. The source of the Heihe originates in the valley near the town of Mafou in Zhouzhi County, and the river eventually flows into the Weihe River at a village named Shima (Xu et al. 2009). This paper focuses on the downstream irrigated region of the Heihe River (sub-basin 1 in Figure 1) and makes a case study to simulate the interaction between the SW and GW. The effects on GW levels and the ecological environment by the reservoir in the downstream irrigation region are analyzed.

Figure 1

Location map of the Heihe River Basin (Li et al. 2013). Please refer to the online version of this paper to see this figure in colour: http://dx.doi.org/10.2166/wcc.2019.042.

Figure 1

Location map of the Heihe River Basin (Li et al. 2013). Please refer to the online version of this paper to see this figure in colour: http://dx.doi.org/10.2166/wcc.2019.042.

Annual precipitation varies greatly in the Heihe River Basin. From July to October, heavy rainfall accounts for more than 60% of the annual precipitation, and the annual average precipitation is approximately 810 mm. The northern part of the remote mountainous area of Qinling Mountains has less than 800 mm precipitation, and the annual precipitation in the rest of the mountain area is about 1,000 mm. The elevation of the basin ranges from 260 to 3,754 m, and the Heihe Jinpen reservoir was built in 1999. Besides supplying drinking water for 8 million residents in Xi'an city, it also provides water for irrigation for a farmland area of 2.47 × 104 km2 (Li et al. 2013).

Datasets

The Heihe River Basin terrain attribute data were obtained by 1:1,000,000 DEM (Digital Elevation Model) extractions in the study area using the Heihe River Basin land-use map and soil type map. The watershed land use and soil were established using the attribute space database with the support of ESRI ArcGIS 9.3 software (CA, USA). Meteorological data consist of precipitation, maximum and minimum temperature, wind speed, and relative humidity measured daily from 2005 to 2013. There are 15 meteorological sites in Zhouzhi County, Shaanxi province (see Figure 1). The measured runoff data were taken from the Chenhe observatory in Shaanxi province for the period recorded from 2005 to 2013 (Qureshi et al. 2013; Huo et al. 2016).

METHODOLOGY

Outline of the combined model (SWATMOD)

In the simulation of GW models, the results were affected by a number of factors, with the main direct impacts coming from GW recharge and GW evaporation. However, it is usually difficult to estimate and assign these variables accurately. During the simulation, the processes used varied. We used MODFLOW instead of the GW simulation module in SWAT (Soil and Water Assessment Tool) to simulate GW, as it can simulate the GW change process in more detail, since the GW component in the SWAT model is not as detailed as that in the MODFLOW model and the distributed SW recharge in MODFLOW is not as detailed as that in SWAT. The linkage and conversion between the two programs are shown in Figure 2.

Figure 2

Schematic diagram of recharge calculation between SWAT and MODFLOW.

Figure 2

Schematic diagram of recharge calculation between SWAT and MODFLOW.

The linkage consists of modifying both the original SWAT and MODFLOW programs. The following part is a brief description of the method for GW partial conversion from the SWAT to the MODFLOW model.

In the SWAT model, the hydrological response unit (HRU) is the most essential computational unit of the model, reflecting the comprehensive impact of soil cover and land use in the sub-basin, and it is the basis of distributed simulation. Land-use and soil vector maps contain their own spatial locations and can be exported to SHP files in the SWAT preprocessing menu. Then, the two vectors are used to conduct a spatial overlay analysis in ArcGIS software, and SWAT generates the full HRU file. This manuscript contains information on the combination of soil and land-use attributes. Using this information, it is possible to create the reference information of the spatial location of the HRU in the basin.

As shown in Figure 2, the soil and land-use maps are exported to a shape file with its own spatial location numbers with the same sized cells, which are determined by DEM. Subsequently, the conversion interface in the ArcSWAT menu reads the HRU numbers in a sub-basin from the FULL HRU file in the ArcSWAT pre-process output and assigns these HRU numbers as spatial locations in the corresponding cells in MODFLOW. In Figure 2, the upper part shows the HRU spatial distribution in the SWAT model, while the lower part demonstrates the division of cells, which are the smallest computing unit in MODFLOW. The number in the grid corresponds to the HRU number of the SWAT model. The number zero designates the areas outside the boundary of the MODFLOW model; through location correspondence, the spatial distribution of the model elements from the SWAT model is assigned to MODFLOW to realize the coupling of SWAT and MODFLOW.

SWAT model calibration and test

Taking a downstream irrigation area in Heihe River Basin as an example, the calibration and verification of the SW-distributed SWAT model was conducted. Assessment of the model requires the objective function to be set. In this study, we used the water balance coefficient (rVol), efficiency coefficient (Ens), and correlation coefficient (r) (Moriasi 2007; Zhang et al. 2012).

The formulas for calculating rVol, r, and Ens are
formula
(1)
formula
(2)
formula
(3)
where Q0 is the measured runoff, Qp is the simulated runoff, Qavg is the average measured runoff, and Qpavg is the average simulated runoff.

Calibration was conducted for the period of 2005–2008; validation was conducted for the period of 2009–2013. The results of the model calibration are shown in Table 1. The monthly runoff data during the period 2009–2013 at Chenhe hydrological station was used to validate the model's predictability based on Ens, rVol, and r. The statistical results for the simulated vs. measured discharge at the Chenhe hydrological station during the validation period are shown in Table 2 and Figure 3. The statistical results are consistent with our previously defined ‘acceptable’ category, i.e., if the water balance coefficient is controlled within ±0.20, the correlation coefficient is not less than 0.80, and the efficiency coefficient is not less than 0.60, then the model result is termed ‘acceptable’ (Zhang et al. 2012). This indicates that the simulated flow rates are in good agreement with the measured ones. The parameters of the model are accurate and reasonable, and the model is stable and can be used for the simulation of water conversion in the irrigation district, and so, the SWAT model is appropriate for hydrological cycle simulation in the Heihe River Basin.

Table 1

Primary parameters for calibration of the model

NumberParameter_NameInterpretationInitial valueCalibrated valueRepresenta_ (value range)
CN2.mgt Scs runoff curve parameter 79 84 r_(1.5,14.0) 
ALPHA_BF.gw Base flow subsided parameter 0.05 0.31 v_(0.0,0.45) 
SOL_K(1).sol Saturated hydraulic conductivity parameter 90 58.32 r_(−0.8,0.8) 
GW_DELAY.gw Hysteresis parameters of GW 31 157.60 v_(122.0,374.0) 
SOL_BD(1).sol Moist bulk density 1.36 1.29 r_(1.1,2.5) 
ESCO.hru Soil evaporation compensation parameter 0.95 0.90 v_(0.8,0.9) 
CH_N2.rte Manning coefficient of the main river bed 0.01 0.24 v_(0.2,0.3) 
SFTMP.bsn Snowfall temperature parameter −1.40 v_(−4.2,1.0) 
SOL_AWC(1).sol Available soil water parameter 0.13 0.10 r_(0,0.5) 
10 GWQMN.gw Depth threshold of the base flow produced by the shallow aquifer 31 0.84 v_(0.3,1.2) 
11 CH_k2.rte Effective hydraulic conductivity in the main channel alluvium 2.85 v_(0,45.6) 
NumberParameter_NameInterpretationInitial valueCalibrated valueRepresenta_ (value range)
CN2.mgt Scs runoff curve parameter 79 84 r_(1.5,14.0) 
ALPHA_BF.gw Base flow subsided parameter 0.05 0.31 v_(0.0,0.45) 
SOL_K(1).sol Saturated hydraulic conductivity parameter 90 58.32 r_(−0.8,0.8) 
GW_DELAY.gw Hysteresis parameters of GW 31 157.60 v_(122.0,374.0) 
SOL_BD(1).sol Moist bulk density 1.36 1.29 r_(1.1,2.5) 
ESCO.hru Soil evaporation compensation parameter 0.95 0.90 v_(0.8,0.9) 
CH_N2.rte Manning coefficient of the main river bed 0.01 0.24 v_(0.2,0.3) 
SFTMP.bsn Snowfall temperature parameter −1.40 v_(−4.2,1.0) 
SOL_AWC(1).sol Available soil water parameter 0.13 0.10 r_(0,0.5) 
10 GWQMN.gw Depth threshold of the base flow produced by the shallow aquifer 31 0.84 v_(0.3,1.2) 
11 CH_k2.rte Effective hydraulic conductivity in the main channel alluvium 2.85 v_(0,45.6) 

GW, groundwater.

av_ means the existing parameter value is to be replaced by the given value, and r_ means the existing parameter value is multiplied by (1+ a given value).

Table 2

The simulated flow results for the Chenhe station

Station nameTimeRvolrEns
Chenhe hydrological station 01 January 2005/31 December 2008 (calibration period) 0.020 0.89 0.86 
01 January 2009–30 April 2013 (validation period) 0.019 0.90 0.82 
Station nameTimeRvolrEns
Chenhe hydrological station 01 January 2005/31 December 2008 (calibration period) 0.020 0.89 0.86 
01 January 2009–30 April 2013 (validation period) 0.019 0.90 0.82 
Figure 3

Flow simulation plots for the model during the calibration and validation period (2005–2013).

Figure 3

Flow simulation plots for the model during the calibration and validation period (2005–2013).

Groundwater model

MODFLOW was developed by the United States Geological Survey. It is used to simulate the GW flow and GW pollutant migration characteristics and development. A modular program structure is used in the MODFLOW program. One can use the model with all kinds of common boundary conditions that are usually encountered during actual situations. These include fixed or pressured heads, variable or constant fluxes, GW recharge/discharge, point withdrawals, and drains (Osman & Bruen 2002; Ye & Grimm 2013).

ArcGIS was utilized to import the soil (soil class) and land use (land-use class) shape files of Heihe irrigation areas downstream to generate the HRU distribution map by spatial overlay. In other words, one of the HRUs with the same land use type and soil type was used in each sub-basin, and then the spatial location in MODFLOW was determined, which can be easily understood by consulting Figure 2. The region which was extracted from the SWAT model was the study area, and this was imported into MODFLOW. The study area was separated into a square grid (100 m × 100 m) with 70 rows, 110 columns, 2 layers, and a depth of 80 m in MODFLOW software. According to the different hydrogeological conditions, the region was divided into two sub-regions: the river area, which was mainly affected by irrigation and weather, and the well irrigation area.

Boundary conditions

The first layer is an aquifer, and the lower boundary is an aquitard layer. The lower reaches north of the irrigation area, which is bounded by the Weihe levee and recharged by the Weihe River strongly permeable banks, are regarded as the river boundary. The southwestern and eastern boundaries are a catchment area which was extracted from the SWAT model. The SW and underground runoff flow to the inter watershed, which can be considered a no flow boundary. Vertical discharge is mainly evaporation and is well pumping. Pumping and pumping dates were used with the value of the SWAT model, according to the specific crop planting area, crop growth date, rainfall, and typical field survey data. Vertical recharge mainly relies on rainfall infiltration recharge, irrigation infiltration recharge, and channel infiltration recharge. The GW recharge values used in the MODFLOW model were the values of each HRU from the SWAT model calculation results. The distribution of daily recharge from the improved SWAT model was imported to MODFLOW by the HRU–Cells interface (Figure 2), and then the GW cycle was simulated with MODFLOW software.

The hydrogeological parameters included in the model were the horizontal hydraulic conductivity (Kh), vertical hydraulic conductivity (Kv), suspended solid (SS), water supply, and water storage rate of degree (Sy). The initial parameters were taken from the basic geological data and needed to be further corrected.

RESULTS AND DISCUSSION

Modeling results

Figure 3 shows the runoff simulation plots during the model calibration and validation period (2005–2013). In Figure 3, it can be seen that the maximum runoff and precipitation in the Heihe River Basin were concentrated in the wettest season (2011). Similarly, the minimum precipitation, surface runoff, and base flow all occur in the driest year (2008). The overall comparison shows that the average runoff in the past 9 years has changed a great deal, showing a downward trend in general. This was tested at the Chenhe hydrological station during the period of 2005–2013, and the annual and non-flood season discharges from the Heihe River Basin were shown to decrease significantly at the Chen River hydrological station. Precipitation fluctuations were the main causes of this change in the study area, and the results are essentially in agreement with previous studies (Bonfils et al. 2008; Wu 2009; Fezzi & Bateman 2015).

The average amounts of GW recharge (Recharge), potential evaporation (Pet), and rainfall were extracted from the above SWAT model for 1 year. Figure 4 shows the Recharge, Pet, and Rainfall dynamic changes in sub-basin 1 in 1 year. Figure 4 shows that Recharge, Pet, and Rainfall showed their maximum values in June, July, August, September, and October.

Figure 4

Recharge and potential evaporation (Pet) dynamic changes in the study area.

Figure 4

Recharge and potential evaporation (Pet) dynamic changes in the study area.

Figure 5 is the distribution map of the Recharge and Pet for October 2014. According to Figure 5(a), GW recharge values in sub-basins 15 and 22 were maximized. The second values were in the plain irrigation region in the lower reaches of Heihe River. Overall, the recharge in the Heihe River south bank was smaller than that on the north shore, and the differences between the different sub-basins were also relatively large. According to Figure 5(b), the Pet value was the largest in sub-basins 6, 7, and 42. The position of the Heihe reservoir was located in sub-basin 7, and the second Pet value was in the downstream plain area and south bank of the Heihe River. However, in contrast, in the mountain area, the Pet of the north bank was greater than that on the south bank of the Heihe River. Such a distribution may be related to the regional GW depth and the unsaturated zone lithology. Because the potential evaporation is influenced and restricted by climate factors such as temperature, ground temperature, precipitation, air pressure, water pressure, GW wind speed and depth, lithology, vegetation, and crops, the movement and change process of the potential evaporation are very complex; therefore, the concrete influence law requires further study.

Figure 5

Recharge and Pet on 10 October 2014.

Figure 5

Recharge and Pet on 10 October 2014.

Model calibration and verification

The corrected hydrogeological parameters are shown in Table 3. There were 18 observation wells in sub-basin 10. The locations of the observation wells in the study area are shown in Figure 6. Figure 7 shows a scatter map of the simulated GW levels and in situ observed values of the 18 observation wells from July to October 2014, and errors are shown in Table 4.

Table 3

The hydrogeological parameters after MODFLOW model calibration

LocationHorizontal conductivity (m/d)Vertical conductivity (m/d) Specific yield Storativity
River area 
 The first aquifer 8.3 0.83 0.2 0.0001 
 The second aquifer 0.01 0.001 0.08 0.0001 
Irrigated area 
 The first aquifer 0.5 0.04 0.0001 
 The second aquifer 0.01 0.001 0.08 0.0001 
LocationHorizontal conductivity (m/d)Vertical conductivity (m/d) Specific yield Storativity
River area 
 The first aquifer 8.3 0.83 0.2 0.0001 
 The second aquifer 0.01 0.001 0.08 0.0001 
Irrigated area 
 The first aquifer 0.5 0.04 0.0001 
 The second aquifer 0.01 0.001 0.08 0.0001 
Table 4

Underground water level and observation error indicators (MODFLOW model)

Error indicatorsAverage value (m)VAR (m)SDRAverage error value (m)
Simulated values 421 67.2 8.2 0.88 6.05 
Observed values 415 45.2 6.7 
Error indicatorsAverage value (m)VAR (m)SDRAverage error value (m)
Simulated values 421 67.2 8.2 0.88 6.05 
Observed values 415 45.2 6.7 
Figure 6

The spatial positions of the measured GW sites.

Figure 6

The spatial positions of the measured GW sites.

Figure 7

Scatter diagram of the simulated and measured water tables in sub-basin 1.

Figure 7

Scatter diagram of the simulated and measured water tables in sub-basin 1.

Figure 6 gives the results of the GW numerical simulation. It shows that the underground water level in sub-basin 1 is separated by Heihe River as the boundary. The GW level gradually escalates from southeast to northwest in the west of the Heihe River. However, from south to north, the GW level in the east of the Heihe River decreases gradually. The figure of the depth of GW was computed by DEM minus the GW level through using the GIS spatial analysis functions. The GW depths of the measured points were then extracted from the GW depth figure. Thus, a comparison and an analysis were conducted between the simulated values and measured values. In the GW depth figure, it can be seen that the table of GW near the Heihe River is very shallow, and the depth of the GW table gradually increases along the northwest and southeast directions. Through in situ investigation of the actual water wells, we found that the simulation value was consistent with the actual survey results.

On 18 October 2014, the GW level of sub-basin no. 1 in the Heihe River Basin was measured. The field survey was extremely necessary for understanding the overall situation of the research area and collecting first-hand accurate and reliable measured data to support the construction of the model. The preliminary investigation team did a great deal of preparatory activities, including the collection of information about the villages and towns in the Heihe River Basin and the contact information of the township leaders through the government website. By telephone communication, we were easily able to understand whether there were any irrigation wells or self-pressing wells. We were also able to determine the well distribution, water depth, etc. Then, using this information, the route of exploration was selected on the map, and the required time was roughly planned. Measurement of the GW level mainly used the measuring rope and the universal electric meter. The working principle was that when the measuring rope was continuously put down into the well, the metal head at the end touched the water, and the universal electric meter pointer on the surface was deflected so that the measuring rope was stopped in time. Through reading this level on the scale on the rope, the water level was obtained. The related instruments used were GPS and a compass. Through the field investigation, a total of 38 sets of measured GW level data were obtained.

Figure 7 shows the spots of the simulated GW level and the in situ observations of the GW level distributed evenly on either side of the 1:1 related line. This indicates that there was no systematic error in the modeling process. The correlation coefficients are all above 88% in Table 3, which indicates that the simulated underground water level values and measured values close in the same way. So, the parameters and hydrogeological conditions obtained from the modeling process are generally accurate, reasonable, and stable. The model can thus be used for the simulation of the irrigation water cycle.

From Figure 7 and Table 3, it can be seen that the average error of the simulated values and measured values was around 6.05 m. This means measured values were 6.05 m lower than simulated values. Why does such deviation appear? By comparing the simulation and measured processes, we found that the simulation process did not consider the influence of the reservoir (the location of the Heihe reservoir dam is shown in Figure 1 with the blue line). That is, the simulation was conducted in a similar manner to the natural situation. However, in fact, the influence of the Heihe reservoir should not be ignored. A possible reason for the existing deviation may be the direct impact of transferring water from the Heihe reservoir to Xi'an city. According to a field investigation on 18 October 2014, further from the reservoir downstream of the Heihe River, the water depths of observed wells are deeper. Several wells surrounding the reservoir had already dried up in downstream areas of the Heihe River. However, the conditions were better in several villages far from the reservoir in downstream areas of the Heihe River. It is possible that it was caused by the recharging of the Weihe River and the rainfall. The construction of the reservoir has had a great influence on the residents in surrounding towns and villages. In traditional times before the reservoir was built, the villagers generally relied on the well water to meet fundamental living requirements, and now, they have no alternative but to rely on the water tower to acquire water for living. In the meantime, the quality of water from deep wells has declined. According to the villagers, before the construction of the reservoir, the well water had a sweet taste and could be used for drinking, but after the reservoir was built, the water from pumping wells became murky, and it is now often mixed with a great deal of mud.

As is well known, the decline of the GW level will bring a series of ecological environment problems. Due to the irrigation, GW level continuing to decline, the water yield of agricultural motor pumping wells has declined quickly, even resulting in well pumps hanging in the air and leading to motor pumping wells being abandoned. This was proven from the field survey. On the other hand, the suitable soil moisture content guarantees normal crop growth, and the soil water content is mainly affected by the depth of the GW table. If the underground water levels continued to decline, soil moisture in the tillage layer will be greatly reduced.

For sub-basin 1, in 2014, the GW numerical simulation results show that although an average deviation error existed, the trend of the simulation of the regional GW level was consistent with the measured results. We should pay attention to the fact that this study is only limited to the data collected in this research process. If the underlying data changes, the conclusions of this research would change too. The coupling calculation model is beneficial as it can more accurately simulate and predict GW resources. Further research on the interaction between SW and GW should provide a scientific method for the planning and scientific management of regional water resources. This has practical significance for realizing the sustainable utilization of regional water. However, in the process of popularization and application, researchers should pay attention to the following problem: due to the divergent soil types, it is necessary to test different soil samples to measure the corresponding relation between the dielectric constant and the soil moisture to improve the accuracy of the water content measurement. This study only focused on the downstream irrigation area in Heihe River in Shaanxi province, China. Whether this method is suitable for other study areas still needs further study. When the integrated model considers the motion law of the aeration zone, the physical mechanism of the model will be refined. So, we need to deal with the aeration zone more perfectly.

CONCLUSION AND RECOMMENDATIONS

According to the characteristics of the irrigation area, integrated numerical modeling was carried out for irrigated area water resources management. Based on the definition of HRU, ArcGIS software was used to determine the spatial positions of the HRUs, and the HRUs–Cells interface was established. By correlating the HRUs in SWAT and the cells in MODFLOW, the mismatch problems were solved. Using the interface between the HRUs and cells, the distributed values of the GW daily recharge and Pet from the improved SWAT were imported into the corresponding module in MODFLOW, and then the GW cycle was simulated with MODFLOW software. Distributed model coupling between SW and GW in the irrigation area was achieved.

The results of the integrated simulation in the downstream irrigation area of Heihe River showed the following. (1) The simulation results of the regional GW level were basically consistent with the measured results. (2) According to the characteristics of the distributed SWAT model, the HRUs and cells were correlated in finite difference grids, and then the SWAT model calculation results were imported, which means that the spatial distribution characteristics of the model elements (GW recharge and diving evaporation) were imported into the MODFLOW GW recharge module (RCH) and the evaporation module. The coupling calculation model was thus achieved. After this, the model was calibrated and validated, and the results showed that the simulation accuracy was reasonable. So, the scientific methods were provided for the simulation and prediction of the conversion rules of SW–GW and the policy setting of water management. (3) Based on the hydrological and geological data analysis in the study areas, the GW recharge and temporal and spatial distribution of evaporation were simulated. Through a comparative analysis of the simulation results with the field survey results, it was found that since the construction of the Heihe reservoir, the average GW level in the study area has declined by 6.05 m.

Based on the above analysis, for downstream irrigation districts under the condition of climate change, in drought situations, the government of Xi'an city should follow the general plan of ‘exploring new water resources and saving water’ to make every effort to ensure urban water supply security. This includes stopping irrigation, conducting flood control operations, and carrying out reasonable development and utilization of regional underground water sources. In flooding situations, GW reservoirs and artificial ponds should be established by the government.

SW and GW are part of the global water cycle. The scope of research has been related to hydrogeology, hydrology, geochemistry, biology, meteorology, etc., presenting a complicated system model including various disciplines and integrating the complexities of the various components of the global water cycle. The formation of system models is an inevitable trend in the research and development of the interaction between SW and GW.

ACKNOWLEDGEMENTS

We are grateful to Professor Weibo Zhou who improved the manuscript; lecturer Jucui Wang, graduate student Boyang Liu, and Yanman Li from Chang'an University for their help in field experiments, and Maria Nazos and Wyn Andrews Richards from the University of Nebraska-Lincoln's Writing Center for improving the paper.

FUNDING

This work was supported by the National Natural Science Foundation of China (Grants No. 41672255, 41877232, 41402255, and 41790444) and High-tech research cultivation project (Grant No. 300102268202 and 300102299201).

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the authors.

REFERENCES

REFERENCES
Bonfils
C.
Santer
B. D.
Pierce
D. W.
Hidalgo
H. G.
Bala
G.
Das
T.
Barnett
T. P.
Cayan
D. R.
Doutriaux
C.
Wood
A. W.
2008
Detection and attribution of temperature changes in the mountainous western United States
.
Journal of Climate
21
,
6404
6424
.
Fezzi
C.
Bateman
I.
2015
The impact of climate change on agriculture: nonlinear effects and aggregation bias in ricardian models of farmland values
.
Journal of the Association of Environmental & Resource Economists
2
,
57
92
.
Hu
L.
Wang
Z.-J.
Zhao
J.-S.
Ma
Y.-H.
2007
Advances in the interactions and integrated model between surface water and groundwater
.
Shuili Xuebao
38
,
54
59
(in Chinese, with English abstract)
.
Huo
A.
Dang
J.
Song
J. X.
Chen
X. H.
Mao
H. R.
2016
Simulation modeling for water governance in basins based on surface water and groundwater
.
Agricultural Water Management
174
,
22
29
.
Kassian
L.
Tenywa
M.
Liwenga
E. T.
Dyer
K. W.
Bamutaze
Y.
2017
Implication of climate change and variability on stream flow in Iringa region, Tanzania
.
Journal of Water & Climate Change
8
(
2
),
336
347
.
Li
Y.
Huo
A.
Liu
R.
Chen
S.
Wang
X.
Li
J.
2013
Water resources responses to climate changes in Xi'an Heihe River Basin based on SWAT model
.
Journal of Water Resources Research
2
,
301
308
.
Peng
J.
Huo
A.
Cheng
Y.
Jian
D.
Hong
W.
Wang
X.
Li
C.
2017
Submersion simulation in a typical debris flow watershed of Jianzhuangchuan catchment, Loess Plateau
.
Environmental Earth Sciences
76
,
462
.
Pikul
M. F.
Street
R. L.
Remson
I.
1974
A numerical model based on coupled one-dimensional Richards and Boussinesq equations
.
Water Resources Research
10
,
295
302
.
Sophocleous
M.
Koelliker
J.
Govindaraju
R.
Birdie
T.
Ramireddygari
S.
Perkins
S.
1999
Integrated numerical modeling for basin-wide water management: the case of the Rattlesnake Creek basin in south-central Kansas
.
Journal of Hydrology
214
,
179
196
.
Wu
J.
2009
Analysis of characteristics of Xi'an Heihe watershed hydrology factors change
.
Water Conservancy Science and Technology and Economy
14
,
874
875
(in Chinese, with English abstract)
.
Xu
l.
Qi
Z.
Haijun
Z.
2009
Research status and progress of interaction and coupling modeling of surface water and groundwater
.
Water Resources Protection
25
,
82
85
(in Chinese with English abstract)
.
Yu
M.
Rui
X.-F.
2007
Study on integrated management model of groundwater and surface water for prevention of soil salinization
.
Water Rescource Protection
23
,
6
9
(in Chinese, with English abstract)
.
Zhang
Y.
Zhang
S.
Zhai
X.
Xia
J.
2012
Runoff variation in the three rivers source region and its response to climate change
.
Acta Geographica Sinica
67
,
71
82
.