Journal of Hydrology 519 (2014) 833–847
Contents lists available at ScienceDirect
Journal of Hydrology journal homepage: www.elsevier.com/locate/jhydrol
Combined use of tracer approach and numerical simulation to estimate groundwater recharge in an alluvial aquifer system: A case study of Nasunogahara area, central Japan Yaping Liu a,b,c,⇑, Tsutomu Yamanaka d, Xun Zhou c, Fuqiang Tian a, Wenchao Ma b a
Department of Hydraulic Engineering, Tsinghua University, Beijing 100084, China Graduate School of Life and Environmental Sciences, University of Tsukuba, Ibaraki 3058572, Japan School of Water Resources and Environment, China University of Geosciences (Beijing), Beijing 100083, China d Faculty of Life and Environmental Sciences, University of Tsukuba, Ibaraki 3058572, Japan b c
a r t i c l e
i n f o
Article history: Received 31 January 2013 Received in revised form 24 June 2014 Accepted 5 August 2014 Available online 14 August 2014 This manuscript was handled by Laurent Charlet, Editor-in-Chief, with the assistance of Chong-Yu Xu, Associate Editor Keywords: Groundwater recharge Numerical simulation Tracer approach Stable isotopes Alluvial aquifer system
s u m m a r y In this study, we simulate the spatial and temporal distribution of groundwater recharge in an alluvial aquifer system in the Nasunogahara area of Japan. Natural stable isotopes (18O, D) were considered as additional calibration targets in a numerical model. The reliability of the model outputs was further validated by comparing the results from the numerical simulation and an independent tracer approach. The results indicated that the calibrated model can effectively simulate the spatial and temporal characteristics of the contribution ratios of recharge sources to groundwater in the Nasunogahara area. However, the tracer approach (i.e., end member mixing analysis) provided more reliable results at point scale, particularly for the estimated contribution ratios of paddy field water. The precipitation in the Nasunogahara area is the major recharge source; its mean contribution ratio is 58% for a one-year period over the entire alluvial fan. River seepage is significant in the upstream area of the alluvial fan, and the contribution ratio of river waters along the river channels in the upstream area increases during the wet season. Paddy field water is a highly important recharge source in the midstream and downstream areas of the alluvial fan, and the contribution ratio of paddy field water obviously increases from dry season to wet season because of irrigation. This study demonstrates that combined use of the tracer approach and numerical simulation with stable isotopes as additional calibration targets can eliminate their respective limitations and can assist in better understanding the groundwater recharge mechanism in alluvial aquifer systems. Ó 2014 Elsevier B.V. All rights reserved.
1. Introduction Groundwater is a major source of drinking water for more than 1.5 109 people worldwide. Groundwater is nearly the sole water source for several cities, including Jakarta, Lima, and Mexico City (Sampat, 2000). Alluvial deposits, which have substantial thickness and high porosity, contain abundant groundwater resources. Many alluvial aquifers have an important role in supplying drinking water and irrigation water worldwide, such as those in the North China Plain in China (Foster et al., 2004; He et al., 2011) and Hueco Bolson in the United States and Mexico (Eastoe et al., 2010). However, excessive groundwater withdrawal to meet increasing ⇑ Corresponding author at: Institute of Hydrology and Water Resources, Department of Hydraulic Engineering, Tsinghua University, Beijing 100084, China. E-mail addresses:
[email protected] (Y. Liu),
[email protected] (T. Yamanaka),
[email protected] (X. Zhou),
[email protected] (F. Tian),
[email protected] (W. Ma). http://dx.doi.org/10.1016/j.jhydrol.2014.08.017 0022-1694/Ó 2014 Elsevier B.V. All rights reserved.
demand for alluvial groundwater because of population increase and economic growth has resulted in declining water tables, land subsidence and groundwater salinization (Famiglietti et al., 2011; Galloway and Burbey, 2011; Liu et al., 2012). The renewal ability of the groundwater system, which is a key parameter determining the sustainable yield of such a system, highly depends on the quantity and quality of aquifer recharge (Mikita et al., 2011; Yamanaka et al., 2011). Thus, a thorough understanding of recharge and flow paths in alluvial aquifers is necessary in water resource management to ensure sustainability on a local or regional scale (Choi et al., 2010; Foster et al., 2004). Numerical modeling provides an efficient method of clarifying the flow path and seasonal dynamics of groundwater, as well as quantifying the recharge amount of each recharge source. However, the reliability of recharge estimated by models depends on the accuracy of the measured and interpolated hydraulic parameters (Scanlon et al., 2002; Sibanda et al., 2009). Therefore, multiple methods or objective functions have been used to calibrate
834
Y. Liu et al. / Journal of Hydrology 519 (2014) 833–847
numerical models to achieve robust results. For example, Kim et al. (1999) determined the best calibration method among a set of possible options with the help of four sets of parameter values (i.e., steady-state heads, transient heads, head gradients, and flow path information). Sanford et al. (2004) used 14C activities and the location of hydrochemical zones as additional calibration targets to determine the model parameters for the Middle Rio Grande Basin in the United States. Dahan et al. (2004) used a multi-variable mixing cell model that represents the hydrochemical approach to calibrate a groundwater flow model, and pointed out that using both hydraulic and hydrochemical tools can improve understanding of the hydrologic system. The simultaneous use of hydrochemical tools and groundwater flow models can also provide insight into strengths and limitations of each method (Carroll et al., 2008). The stable oxygen and hydrogen isotopes (18O and D) are kind of environmental isotopes, which widely exist in natural waters. The isotopic compositions (d18O and dD) in water are affected by meteorological process (temperature, humidity, and so on), but not as much by reactions with geologic materials, making 18O and D suitable for investigating the provenance of groundwater (Clark and Fritz, 1997). The d18O and dD values are particularly useful for tracing water movement at river banks because two main water sources are typically present: (1) river water, which is depleted of heavy isotopes and originates upstream, and (2) groundwater, which comes mainly from local rainfall (Lambs, 2004). When the d values are negligibly affected by evaporation processes and the effects of past climate regimes on d signatures are ignored, 18O and D can be considered as conservative tracers (Carroll et al., 2008). Although d values are not technically concentrations, they can be considered as concentrations because d values scale linearly with concentration. However, using d values as simulation terms to calibrate numerical simulation results is highly limited. These studies are mainly restricted to steady-state conditions (Stichler et al., 2008), simple model (Yamanaka and Wakui, 2009), or coastal wetland system (Reynolds and Marimuthu, 2006). The present study aims to: (1) reveal the spatial and temporal distribution of the contribution ratios of groundwater recharge sources in an alluvial aquifer system through combined use of a tracer approach and a numerical simulation; (2) improve the model calibration by adding stable isotopes as additional calibration targets of hydrometric data; and (3) clarify the capabilities and limitations of the tracer approach and numerical simulation applied to compute the contribution ratio of groundwater recharge sources. The Nasunogahara area in Japan is used as a case study (Fig. 1). A hydrochemical and stable isotopic study was carried out in the middle of the alluvial fan. The results indicate that precipitation, river water, and paddy field water are the three main recharge sources of groundwater. The hydrochemical and stable isotopic study was also estimated the contribution ratios of each source to well waters through a three-end-member mixing analysis (Wakui and Yamanaka, 2006). According to the hydrochemistry and isotope characteristics of the sampled groundwater, the alluvial fan was generally divided into three sub-areas: Sabi River-influenced area (Fig. 2a), Naka River-influenced area (Fig. 2b) and Houki River-influenced area (Fig. 2c). The spatial changes of groundwater recharge/ discharge fluxes along the Sabi River were characterized by a compartmental mixing cell model with stable isotopes as calibration terms (Yamanaka and Wakui, 2009). The compartmental mixing cell model revealed that precipitation was the most dominant recharge source to groundwater at the non-paddy area, while river water infiltration at the fan apex was important in driving groundwater flow system. However, paddy field infiltration is relatively small even during irrigation period. Elhassan et al. (2001, 2003, 2006) examined the groundwater budget of unconfined aquifer in the Nasunogahara area and analyzed the influence of paddy field
irrigation on groundwater budget using a modified tank model with a two-dimensional groundwater flow model. However, the detailed spatial and temporal distribution of the groundwater recharge in this study area still needs to be elucidated with the help of a three-dimensional groundwater flow model.
2. Site description The Nasunogahara area, one of the largest alluvial fans in Japan, is a compound alluvial fan formed by the Naka River, Houki River, Sabi River, and Kuma River (Fig. 1). The alluvial fan is bound by the Houki River on the west and south, the Naka River on the east, and the Shimotsuke Mountains on the northwest. The annual mean precipitation is 1533(±280) mm, and approximately 83% of the annual precipitation occurs during the wet season from April to October according to climatic normals (1981–2010) calculated from the observed data of Japan Meteorological Agency (http:// www.jma.go.jp/jma/index.html). The annual mean temperature in the area from 2004 to 2006 was 12.3(±0.4) °C, (which was close to the long-term annual mean temperature 11.7(±0.8) °C. The annual mean precipitation was 1553(±260) mm, which was also close to the long-term annual mean precipitation of 1533 mm. The largest monthly mean precipitation of 338(±46) mm occurred in July; the second highest monthly mean temperature of 23.0(±0.9) °C also occurred in the same month. The lowest monthly mean precipitation 20(±6) mm and the lowest monthly mean temperature 0.6(±0.3) °C occurred in January. Thus, the simulated results in July were used to represent the wet season conditions, while the outputs in January were used to represent the dry season conditions. Thornthwaite method was used to estimate the potential evapotranspiration, and the calculated annual mean potential evapotranspiration was 654(±11) mm in the study area (Thornthwaite, 1948). The precipitation and temperature data were obtained from the Kuroiso Station (36°58.90 N, 140°01.10 E, and 343 m above mean sea level (a.m.s.l.)). The alluvial aquifers in the Nasunogahara area are mainly composed of gravel and sand, whereas aquitard consists of pumice and clay (Fig. 3). The northwestern part of the study area is bounded by an impermeable fault that behaves as an aquiclude. The impermeable fault separates the alluvial aquifers from mountain aquifers/ mountain blocks. Given the high permeability of the river beds, the middle reaches of the Sabi River dries up and only has water on rainy days during the wet season. Downstream, the Sabi River receives groundwater discharge and becomes a permanent river reach again. The Houki River, a branch of the Sabi River, is also an intermittent river. The river bed and river banks of the Houki River are covered by concrete. The stable isotopic analysis showed that the influence of the Houki River on groundwater was limited (Wakui and Yamanaka, 2006). The seepage of the Houki River is expected to be limited for the entire study area; thus, it is not considered in this study. Moreover, many springs with large discharge occur downstream of the Nasunogahara area. Forest, agricultural, and residential areas are the three major land use/land cover types in the Nasunogahara area. Rice is the major farm crop. Approximately 40% of the land (i.e., 164 km2) is used as paddy field, but most of the paddies are distributed midstream and downstream of the Nasunogahara area, as well as along rivers (Fig. 4a). The irrigation period usually begins in April and ends in early August. The groundwater within the study area is abstracted for local consumption; however, none of these wells are metered. Moreover, no large water consumption wells are found for industrial and agricultural use in this area. Therefore, groundwater abstraction is expected to be extremely small and does not have a significant effect on the groundwater system in the area.
Y. Liu et al. / Journal of Hydrology 519 (2014) 833–847
835
Fig. 1. Map of the Nasunogahara area showing groundwater sampling locations during periods from March 2004 to February 2005 (filled circle) and from April 2006 to October 2006 (open square).
3. Methodologies
3.2. Model description
3.1. Samplings and analysis
The well-established US Geological Survey computer code MODFLOW (McDonald and Harbaugh, 1988; Harbaugh et al., 2000) was used in modeling groundwater flow in the Nasunogahara area. The model was constructed using Visual MODLFOW Pro (Ver. 4.2; Waterloo Hydrogeologic, Inc.) as the graphical user interface. To simulate solute transport in aquifers, a MT3DMS code (Zheng and Wang, 1999), an improved version of Modular 3-D Transport model (MT3D) (Zheng, 1990) was used to simulate transport of multispecies, included with the Visual MODFLOW Pro. The MT3DMS code solves the transport equation after the flow solution has been calculated from MODFLOW. The general advective–dispersive equation describing the fate and transport of the contaminants of species k in the 3-D transient groundwater flow systems is as represented by
Field works were conducted monthly from March 2004 to February 2005 and nearly bimonthly from April 2006 to October 2006 for sampling well, river, and paddy field waters (Wakui and Yamanaka, 2006; Wakui, 2007). Precipitation samples were collected monthly from March 2004 to October 2006. All samples were collected for 18O and D compositions analysis. The compositions of 18O and D were determined by mass spectrometry using a stable isotope ratio mass spectrometer (MAT252, Thermo Finnigan) from the Hydrology Laboratory, University of Tsukuba, and expressed by d18O and dD, respectively, relative to the Vienna Standard Mean Ocean Water. The measurement accuracy was ±0.1% for d18O and ±1% for dD. Water tables were also measured for all of the sampling wells during the fieldwork periods.
836
Y. Liu et al. / Journal of Hydrology 519 (2014) 833–847
(a)
(b)
(c)
Fig. 2. Plot of the annual (arithmetic) mean d values in river, paddy field, and ground waters, as well as the weighted mean (weighted by monthly precipitation) d values in precipitation (modified from Wakui and Yamanaka, 2006). (a) The Sabi River-influenced area, the filled gray triangle represents the isotopic signature of Kuma River water; (b) the Naka-River influenced area; and (c) the Houki-River influenced area.
@ @C k hDij @xi @xj
!
@ @ðhC k Þ ðhv i C k Þ þ qs C ks þ RRn ¼ @xi @t
ð1Þ
where Ck is the dissolved concentration of species k (ML3); h is the porosity of the subsurface medium, which is dimensionless; t is time (T); xi is the distance along the respective Cartesian coordinate axis (L); Dij is the hydrodynamic dispersion coefficient tensor (L2 T1); Vi is the seepage or linear pore water velocity (L T1) related to the specific discharge or Darcy flux through the relationship of Vi = qi/h; qs is the volumetric flow rate per unit volume of aquifer representing fluid sources (positive) and sinks (negative) (T1); Cks is the concentration of the source or sink flux for species P k (ML3); and Rn is the chemical reaction term (ML3 T1). 3.3. Model construction and parameters settings In this study, 18O and D were considered as two contaminants, and the absolute d value is inputted into the transport model as contaminant concentration. Because the d18O and dD in water are not as much by reactions with geologic materials (Clark and Fritz, 1997), we assume that d18O and dD are conservative tracers along the flow-paths (Stichler et al., 2008), i.e., no isotope fractionation occurs during the process of water transfer from recharge sources to observation wells. Thus, the chemical reaction term should be ignored in the simulation. Given that the mountain block recharge is prevented by impermeable fault in the Nasunogahara area, the groundwater system in the alluvial sediments, without considering the mountain block recharge, becomes the focus of the simulation in this study. The modeled area is 20 km 30 km long, which contains the entire
Nasunogahara area shown in Fig. 1. The simulation used variable cell areas of 250 m 250 m, 250 m 500 m, and 500 m 500 m, and contained 29,520 cells in 82 rows, 60 columns, and 6 layers. Most wells in the Nasunogahara area have been drilled into the unconfined aquifer, and data on deep wells in this area are not available. Therefore, layer 1 was considered as the focused simulation and calibration layer, which corresponds to the unconfined aquifer, where gravel and sand are dominant. Layer 2 corresponds to the first aquitard where pumice and clay with sand and gravel are present. Layers 3–6 correspond to the confined aquifers/aquitards. Surrounding mountains and hills were assigned as inactive cells. Based on the local hydrogeology setting and certain boring logs, variable thickness ranging from 2 m to 107 m was assigned to layer 1 in the active model domain. Uniform thicknesses were assigned to layers 2– 6, which were 10, 20, 15, 30, and 50 m, respectively. The Naka and Houki rivers were assigned as constant head boundaries using the time-variant specified-head package (Leake and Prudic, 1991) in the model. The daily observed river stage in the Kurobane gaging station (140°070 0700 E, 36°510 2500 N) of the Naka River during the simulation period was used to simulate the hydraulic head of the constant head boundaries. The Sabi River was considered as a stream boundary using the streamflow-routing package (Prudic, 1989). The entire Sabi River was divided into eight segments, and the different streambed hydraulic conductivities were set for different segments. The inflow of the first uppermost segment of the stream boundary was estimated by a sample river tank model (Wakui, 2007; Yamanaka and Wakui, 2009). Springs were simulated as a sink term of groundwater using drain boundaries with fixed head in time and variable in space. Given that an impermeable fault separates the Nasunogahara area from
Y. Liu et al. / Journal of Hydrology 519 (2014) 833–847
837
Fig. 3. (a) Geologic map with (b) cross section A–A0 in the Nasunogahara area.
the Shimotsuke Mountains in the northwest area of the model domain, a no-flow boundary was assigned for the northwestern boundary of the model domain (Elhassan et al., 2001; Fujinawa, 1981). The present study assumed that the bedrock or deep aquifers below layer 6 do not contribute to groundwater flow in the simulated aquifers. Therefore, a no-flow boundary was adopted for the bottom of the model. The groundwater recharge from the paddy field during the irrigation period was considered as paddy field water recharge, whereas that during the rest of the year was considered as precipitation recharge. The daily net recharges of precipitation and paddy field water were estimated by a tank model, and the precipitation, evapotranspiration, spatial distribution of paddies, as well as irrigation time were considered to estimate the recharge rates. The parameters used in the tank model were from previous studies (Wakui, 2007; Yamanaka and Wakui, 2009). The recharge at
paddies and non-paddy fields was estimated for each 1 km 1 km mesh on a daily time step. According to the estimated net recharge from both paddy and non-paddy fields, the model domain of MODFLOW was divided into four recharge zones (Fig. 4a). The estimated recharge rates at the four recharge zones are shown in Fig. 4b. The recharge package was used to assign the corresponding recharge to the uppermost active aquifer of the model. In the transport model, the Sabi River was set as a constant concentration boundary, which acts as a source that provides solute mass to the model domain in the form of a known concentration. Point source boundaries corresponding to constant head boundaries were assigned to the Naka and Houki rivers. The point source boundary condition specifies the concentration of each species entering or leaving the model through the constant head boundary condition grid cells specified in the flow model. For the impermeable and drainage boundaries, a default setting where the
Y. Liu et al. / Journal of Hydrology 519 (2014) 833–847
(b)
Recharge (mm)
Recharge (mm)
Recharge (mm)
Recharge (mm)
(a)
Rainfall / ET (mm)
838
Fig. 4. (a) Land use map with mesh division for recharge estimation in the tank model and zone division for recharge input in MODFLOW, and (b) estimated recharge rates from paddy and non-paddy fields at the four recharge zones. Monthly mean precipitation and potential evapotranspiration estimated by Thornthwaite method are also added for reference.
concentration gradient is equal to zero was accepted. The recharge concentration boundary condition (i.e., the source term of the governing equation), which specifies the concentration of each species accompanying the recharge flux specified in the flow model, was assigned to the recharge boundary. A steady state simulation was conducted to achieve the initial conditions of the transient simulation. The observed hydraulic heads and d values of 18O and D in February 2004 were used as initial values in the model under steady state condition. The calculated hydraulic head and d values were then applied as initial conditions in the model under transient condition. Considering the heterogenetic and anisotropic properties of the aquifers, layer 1 was subdivided into 21 parameter zones based on local
hydrogeological conditions (Watanabe and Sagehashi, 1960, 1962; Watanabe et al., 1960). For the other layers, one parameter zone was set and accompanied by one layer. This study assumed that the horizontal conductivity (Kx, Ky) is isotropic and vertical conductivity (Kz) is less than the horizontal conductivity in each parameter zone. The hydraulic conductivities, specific yield, and specific storage parameters were initially adopted from previous studies (Johnson, 1967; Elhassan et al., 2001; Kelly, 2004; Wu et al., 2010; Mikita et al., 2011). Effective porosity, streambed hydraulic conductivity, and drain conductance were set by following previous studies (Freeze and Cherry, 1979; Zheng and Bennett, 2002; Wakui, 2007). Other parameters, such as hydrodynamic dispersion coefficient, were obtained from the system defaults. The
839
Y. Liu et al. / Journal of Hydrology 519 (2014) 833–847
Table 1 Annual mean d values with standard deviations of sampling points in the study area, and the estimated contribution ratios by the tracer approach with their potential standard errors. Sampling points
d18O (‰)
dD (‰)
R_Precipitation (%)
R_River (%)
R_Paddy (%)
G1 G3 G4 G5 G6 G7 G30 G31 G44 G45 Precipitationa Sabi River Naka River Houki River Paddy field water
9.0(±0.7) 9.2(±0.1) 8.1(±1.0) 8.9(±0.2) 8.7(±0.2) 8.7(±0.2) 8.2(±0.2) 8.8(±0.2) 8.7(±0.1) 8.3(±0.2) 8.4(±2.4) 9.6(±0.5) 9.2(±0.2) 9.4(±0.5) 5.0(±1.7)
59(±3) 60(±1) 55(±3) 58(±2) 58(±1) 57(±2) 55(±1) 57(±1) 57(±2) 55(±2) 55(±18) 62(±4) 60(±2) 61(±4) 40(±8)
44(±53) 33(±46) 0(±56) 49(±51) 23(±54) 73(±52) 95(±47) 63(±63) 67(±52) 97(±42)
55(±39) 67(±33) 68(±41) 49(±39) 65(±39) 27(±38) 0(±42) 37(±58) 33(±46) 0(±37)
1(±15) 0(±13) 32(±16) 2(±15) 12(±15) 0(±14) 5(±16) 0(±19) 0(±12) 3(±14)
a Weighted mean value (weighted by monthly precipitation), R is contribution ratio, values in the brackets represent standard deviations of the d values and standard errors of the estimated contribution ratios.
parameters were then calibrated by trial-and-error method with the help of Visual PEST (Doherty, 1998), a graphical nonlinear parameter estimation and predictive analysis package.
4. Results and discussion 4.1. Tracer approach Fig. 2 shows a plot of the annual (arithmetic) mean d values in river, paddy field, and ground waters, as well as the weighted mean (weighted by monthly precipitation) d values in precipitation. Data for all groundwater samples are plotted within a triangle with
(a) 320
end-members that are constituted by river water, precipitation and paddy field water, indicating that river water, precipitation, and paddy field water are the three recharge sources of groundwater. The d values of precipitation used in this study are the weighted mean values, not the arithmetic mean value used in the previous study (Wakui and Yamanaka, 2006). The contribution ratio from each of the potential sources can be estimated through the end-member mixing analysis. The standard error of the estimated ratio, which is probably caused by the uncertain composition/concentration of the end members and the resultant mixtures (i.e., groundwater samples), can be computed by error propagation analysis (Phillips and Gregg, 2001; Liu and Yamanaka, 2012). The standard deviation of the d values of
Head
300
Y= X
280
260
240
N=119 220
220
240
260
280
300
320
Observed (m a.m.s.l.)
Simulated (‰)
(b)
Fig. 5. Comparison between the simulated and observed hydraulic head, d18O, and dD during the simulation periods (a) from March 2004 to February 2005, and (b) from April 2006 to October 2006 (N represents the number of calibration points).
2.05 1.94 N represents the number of calibration points, ME represents mean error, MAE represents mean absolute error, and RMS represents root mean square error.
Normalized RMS (%) RMS (‰) MAE (‰)
1.39 1.28 0.03 0.40 10 9 0.48 0.26 0.24 0.17 0.01 0.02 1 1
ME (‰) RMS (m) MAE (m)
1.03 1.61
dD (‰)
Normalized RMS (%) RMS (‰) MAE (‰) ME (‰) ME (m)
Normalized RMS (%)
d18O Head Simulation periods
Table 2 Calibration results for the numerical model.
Water tables in the study area were measured, and samples were obtained from 10 wells from March 2004 to February 2005 with monthly interval and from 17 wells from April 2006 to October 2006 with nearly bimonthly interval (Fig. 1). Both the observed hydrometric data and the measured isotopic data in the two investigation periods were used to calibrate the water flow and transport models. All of the observation wells were in the unconfined aquifer (i.e., layer 1), and the water tables ranged from 1.9 m to 24 m below the ground surface. Fig. 5 shows the comparison between the simulated and observed hydraulic head and the d values in the two simulation periods. The average difference between the simulated and observed hydraulic head and d values can be examined by mean error (simulated minus observed), mean absolute error, root mean square error (RMS), and normalized RMS (Anderson and Woessner, 1992; Zhang and Hiscock, 2011; Surinaidu et al., 2013). To compare our results with those of previous studies, all four parameters were used as the goodness-of-fit measure (Table 2). The RMS for the 119 calibration points from March 2004 to February 2005 was 1.03 m for head, 0.48‰ for d18O, and 2.05‰ for dD, whereas the normalized RMS was 1% for head, 10% for d18O, and 12% for dD. The RMS for the 65 calibration points from April 2004 to October 2005 was 1.61 m for head, 0.26‰ for d18O, and 1.94‰ for dD, whereas the normalized RMS was 1% for head, 9% for d18O, and 11% for dD. The simulated heads and d values were slightly lower than the observed values. The goodness-of-fit was future tested by the reasonable match between the temporal variations of the observed and simulated hydraulic head and d values during the two simulation periods (Fig. 6). The observed d values at G1 and G4 wells in July 2004 were slightly high. These values were not reproduced by the model probably because of the spatial difference in spatiotemporal resolution between the model and observation. The simulated values were daily mean values of the isotopic composition of groundwater averaged over the cell, whereas the observed data represented the instantaneous values of the isotopic values at each sampling point. Notably, the d values of spring water (9.0‰, 58‰) adjacent to the G1 well were slightly different from that of the G1 well water (6.6‰, 51‰) and were closer to the simulation d values (8.6‰, 57‰) at the G1 well in July 2004. The model results exhibited reasonable agreement between the simulated and observed head as well as isotopic data over the two simulation periods. Thus, the model was considered to be calibrated reasonably well and the calibrated model can be used to simulate the groundwater flow and the 18O and D transport in the aquifer systems of the Nasunogahara area. The determined parameters through hydraulic and isotopic calibrations are summarized in Table 3. The spatial distribution of the observed and the simulated d values was shown in Fig. 7. The observed d values agree with the spatial distribution of the simulated values during both wet and dry seasons. The isotopic values of the groundwater in the upstream area are lower than that in the downstream area and increase with increasing distance from the river channels. However, the increase was not smooth and the d values in number of areas were
0.81 1.08
4.2. Model calibration and sensitivity analysis
0.10 0.18
groundwater and recharge sources was used as a measure of the uncertainty of the end members. The estimated contribution ratios and their potential standard errors are shown in Table 1. The estimated contribution ratio ranged from 0% to 97% for precipitation, 0% to 68% for river water, and 0% to 32% for paddy field water. The standard errors ranged from 12% to 63% for all the estimated contribution ratios. The intermonthly variation of the d values of precipitation is probably the main reason for the relatively large errors of the estimated contribution ratios.
12 11
Y. Liu et al. / Journal of Hydrology 519 (2014) 833–847
March 2004–February 2005 (N = 119) April–October 2006 (N = 65)
840
Y. Liu et al. / Journal of Hydrology 519 (2014) 833–847
841
(a)
Fig. 6. Temporal variations of the simulated and observed hydraulic head, d18O, and dD during the simulation periods (a) from March 2004 to February 2005, and (b) from April 2006 to October 2006.
obviously higher than those in surrounding areas. Surface elevation variation within the model domain was 518 m; thus the isotopic altitude effect of precipitation within the model domain was not significant and not considered in the model. The lower d values in the upstream area were probably caused by seepage from rivers and the obviously high d values in certain areas were probably caused by paddy field water infiltration. A sensitivity analysis was conducted to determine which parameter was mostly constrained by the calibration data and which one had the least uncertainty. Four primary parameters were varied linearly over a selected range (e.g., parameter multipliers shown in the x-axis of Fig. 8) to determine the effects on this calibration target (Zhang and Hiscock, 2010). These parameters are Kx and Ky of layer 1, Kz of layer 1, Kx and Ky of the deep layers (layers 2–5), and Kz of the deep layers. The calibration target normalized RMS value does not depend on the units, and was therefore recorded to describe the sensitivity of the solution to the particular parameters. The results of the sensitivity analysis suggested that the flow model solution was sensitive to all the four groups of parameters, but the transport model was less sensitive to the Kz of layer 1. However, the variation trends of the normalized RMS for the hydraulic head and isotopic values were not always consistent when the parameters increase or decrease relative to
the calibrated values. For example, the normalized RMS for hydraulic head and dD increased when the Kz of the deep layers increased, whereas the normalized RMS for d18O decreased. The sensitivities of the flow and transport model to the Kx, Ky and Kz of the deep layers implied that the hydraulic conductivity of these layers can be constrained by the measured data to a certain extent. The mechanism of influence of the deep aquifers on the shallow aquifers will be another study with more detailed data on the deep aquifers. In addition, the isotopic values were slightly less sensitive to the parameter variations than the hydraulic head, which may be attributed to the fact that the transport equation was solved after the flow solution. The dD values were also slightly more sensitive than the d18O values. 4.3. Contribution ratio of each recharge source estimated by the model When groundwater is a mixture of different sources with fixed composition, the contribution ratio (i.e., proportion) of each source can be estimated by the following equation:
Ri ¼ C i =ðC 1 þ C 2 þ þ C i þ þ C n Þ
ð2Þ
where Ri is the contribution ratio of the recharge source i; C is the simulated concentration of each recharge source; subscripts
842
Y. Liu et al. / Journal of Hydrology 519 (2014) 833–847
(b)
Fig. 6 (continued)
Table 3 Summary of the final values of hydraulic and transport parameters in the calibrated model. Parameter
Layer
Final values in the calibrated model
Range of parameter values used during calibration
Unit
Kx, Ky Kz Specific storage Specific yield Streambed hydraulic conductivity Drain conductance (simulating springs) Effective porosity
1–6 1–6 2–6 1–6 1 1 1–6
3.5–311 6.9 104–0.14 5 106–9 104 6.5 104–0.1 6.9–38 0.024–0.24 0.1–0.2
0.02–400 3 105–40 5 106–1.4 103 0–0.27 0.02–400 0.002–0.45 0.1–0.5
m/d m/d 1/m m/d 1/d
Y. Liu et al. / Journal of Hydrology 519 (2014) 833–847
843
Fig. 7. The spatial distribution of the observed and the simulated d18O and dD values during (a) wet season (July 2004) and (b) dry season (January 2005).
1, 2, . . .i,. . .n represent the 1st, 2nd, . . .i-th,. . ., n-th recharge source of groundwater, and the local groundwater system has a total of n recharge sources. The main recharge sources of groundwater in the study area can be classified into three: precipitation, river water, and paddy field water. The MT3DMS with calibrated transport and flow parameters was used to compute the contribution ratios of the recharge sources of groundwater. The initial concentration of groundwater was set to zero. The recharge of the source water into the aquifer for one species corresponding to one recharge source (i.e., Ci = Ck) was represented in the transport model as a constant tracer mass source, with the concentration Cks = 100% corresponding to the proportion of the source water (Stichler et al., 2008; Yamanaka et al., 2011). Sasaki et al. (1958) pointed out that groundwater velocity in the gravel layer of the study area is 15–21 m per day and the groundwater flow takes 389–546 days from the Nasu canals (river waters) to the springs (discharge points). Moreover, a study demonstrated that the groundwater may be transported from the fan
apex to distal areas in 20–30 years in an alluvial fan in the Atacama Desert, Chile (Houston, 2002). Tritium results of the groundwater samples from the unconfined aquifers provided evidence that the mean residence time of groundwater is less than 36 years in the alluvial fan zone in the Heihe River Basin, China (Chen et al., 2006). Therefore, a 30-year period was selected to run the model for computing the contribution ratio R. After running for 30 years, the contribution ratio of each recharge source to the unconfined aquifer was calculated based on Eq. (2). The contribution ratio of river water to groundwater is nearly symmetrically distributed along the Sabi River in the Nasunogahara area, as show in Fig. 9. Almost the entire alluvial fan was recharged by river waters, except for a small linear area of 7 km2, where the contribution ratio of river water was less than 1%. The distribution of river-recharged groundwater with over 50% contribution ratios extended to 2.5 km from the channel upstream, but less than 1 km downstream. The results indicated that good connections exist between groundwater and rivers in the study area,
844
Y. Liu et al. / Journal of Hydrology 519 (2014) 833–847
particularly in the upstream area, and the connections between them are not only from quantity aspect but also from quality aspect. The contribution ratio of river water decreased in the upstream and midstream areas along the channel of the Sabi River because of the low river flow (or even absence of flows) during the dry season, whereas the spatial distribution pattern of the contribution ratio of river water in the alluvial fan did not vary in different seasons. The contribution ratio of precipitation in most areas, except for the cells adjacent to river channels, was larger than 50%, particularly in the upstream area and the fan toe, indicating that the contribution from precipitation was the dominant recharge source at the zones between river channels. The seasonal variation of the contribution ratio of precipitation was less than 20% in most areas of the alluvial fan, except for the cells along the middle reach of the Sabi River. The recharge of paddy field water to groundwater mainly occurred midstream and downstream of the Nasunogahara area, which concurs with the spatial distribution of the paddies (Fig. 4a). Moreover, the contribution ratio of paddy field water increased during the irrigation period (wet season), whereas that of precipitation decreased. The seasonal variation of the contribution ratio of paddy field water at point scale was less than 20% in most areas. The contribution ratio of paddy field water was less than 50% at point scale, even during the irrigation period. However, influence area of paddy field water infiltration is around 300 km2, which is three quarters of the total area of the alluvial fan, indicating that paddy field water is a highly important recharge source for groundwater in the study area. The results also imply that the paddy field water is an important potential groundwater pollution source because groundwater can be contaminated by agricultural activities such as fertilizer and pesticide applications. The limited distribution of paddy field and the short irrigation period are probably the two main factors that restrict the paddy field water contribution to groundwater. The downstream of the study area is the main discharge zone of the local groundwater system. The local groundwater discharge system also potentially restricts the contribution of paddy field water to groundwater. The mean contribution ratios of the recharge sources of the unconfined groundwater in the entire model domain during a year-long period from March 2004 to February 2005 were as follows: precipitation 58(±2)%, river water 28(±0)%, and paddy field water 14(±2)%. The monthly mean contribution ratio of paddy field water increased from March to July and then gradually decreased from August to February, as indicated in Fig. 10. By contrast, the monthly mean contribution ratio of precipitation decreased from March to July, and then gradually increased from August to February. The monthly mean contribution ratio of river water was relatively stable, but the contribution ratio of river water increased along the river channel in the river-recharge-dominated area (i.e., the area where the contribution ratio of river water was P50%) from January to July, and remained stable or slightly decreased in other areas (Fig. 9). Given the changes in river water contribution, the contribution ratio of precipitation from January to July decreased along river channels, except Kuma River, in which the infiltration was not considered in this study. The precipitation contribution ratio also decreased in the paddy field area and increased in the non-paddy field area from January to July. The seasonal variations of the contribution ratios of river water were larger in the upstream and midstream areas, whereas those of the paddy field water were larger in the midstream and downstream areas.
in the middle stream of the Nasunogahara area (Wakui and Yamanaka, 2006), and an error propagation analysis (Phillips and Gregg, 2001) was used to compute the possible standard errors of the contribution ratios. The estimated contribution ratios to wells by the tracer approach with possible errors and the calculated values by the numerical simulation are compared in Fig. 11. The estimated contribution ratios by the numerical simulation approach are included within the error range by the tracer approach, indicating that the two results are generally both reliable. Although the difference in the estimated contribution ratios between the two approaches could be as much as 40% at G6, the river contribution ratio estimated with the simulation approach is almost within the error range estimated with the tracer approach. The river water contribution at G6 is probably overestimated with the tracer approach because of the influence of the Kuma River; the calculated value with the simulation approach
4.4. Comparison between the results from the numerical simulation and the tracer approach As mentioned in Section 4.1, a tracer approach was used to estimate the contribution ratios of the recharge sources to well waters
Fig. 8. The results of the sensitivity analysis of the calibrated flow and transport model.
Y. Liu et al. / Journal of Hydrology 519 (2014) 833–847
(a)
River
Precipitation
845
Paddy field water
(b)
Fig. 9. The spatial distribution of the contribution ratio of river water, precipitation and paddy field water to groundwater during (a) wet season (July 2004) and (b) dry season (January 2005).
should be the reasonable contribution ratio of Sabi River water to groundwater at G6. The estimated contribution ratio of paddy field water by the simulation approach is approximately 20% lower than that of the tracer approach at G4 and is approximately 20% higher at G7 (Fig. 11c). The water of G4 was obtained from a well near a paddy field, whereas that of G7 was obtained from a well that is several hundred meters away from a paddy field. Thus, the contribution ratio of paddy field water computed by the tracer approach is probably more reliable than that computed via the simulation approach. The difference in estimated contribution ratios of paddy field water is probably caused by (1) the input error of location and area of paddy field to the numerical model and (2) the same setting of hydraulic properties for paddy and non-paddy fields in the numerical model. Although a 20% error is observed between the estimated contribution ratio of paddy field water by the simulation approach and the tracer approach at certain points, the numerical model can still simulate the spatial variations of paddy field water infiltration with approximate 80% accuracy. The estimated contribution ratio of paddy field water from the numerical simulation is less accurate than that from the tracer approach at point scale, but the estimated contribution ratio of river water to groundwater at certain points from the numerical simulation is more accurate than that from the tracer approach
Fig. 10. The monthly mean contribution ratios of the recharge sources to groundwater in the entire model domain during a year-long period from March 2004 to February 2005.
(such as G6 in Fig. 11a). In addition, the computed contribution ratio by the tracer approach has a certain degree of uncertainty and the error is even larger than 60% in some cases (G31 in Fig. 11a). Moreover, the numerical simulation can provide detailed spatial and temporal characteristics of groundwater recharge in
846
Y. Liu et al. / Journal of Hydrology 519 (2014) 833–847
(a)
(b)
(c)
Fig. 11. Comparison of the estimated contribution ratios of (a) river, (b) precipitation and (c) paddy field water to wells by the tracer approach with possible errors and the calculated values by the numerical simulation.
the entire study area, whereas the tracer approach cannot. Therefore, combined use of the tracer and numerical approaches can mitigate the defects of each method and improve our understanding of the general characteristics of groundwater recharge in the Nasunogahara area. The results of this study imply that the combined use of the tracer approach and the numerical simulation is better than the single use of either of the two to understand the groundwater recharge mechanisms in alluvial aquifer systems. Close interaction between groundwater and rivers has been observed in the Nasunogahara area. The decrease/increase in river flow may cause a corresponding decrease/increase in water tables, and the pollution of river water may induce groundwater pollution. Land use (i.e., paddies) is a highly important factor that affects the groundwater recharge system in the study area. The influence of paddy field irrigation should also be considered for groundwater conservation. The findings in this study may inspire local managers to formulate appropriate policy for water resource management and protection.
characteristics of groundwater recharge and stable isotopes (18O, D) in the study area. The calibrated model was validated by comparing the computed contribution ratios of the recharge resources to groundwater with that of the tracer approach. The following can be concluded from this study. 1. A model using both stable isotopes and hydraulic heads as calibration objectives can provide more reliable results than a model that uses only hydraulic heads as calibration objectives. However, the isotopic values are slightly less sensitive to parameter variations than the hydraulic head, and dD values are more sensitive than the d18O values. 2. The calibrated model with stable isotopes as additional calibration targets can well represent the spatial and temporal characteristics of contribution ratios of recharge sources to groundwater in the Nasunogahara area, but has approximately 20% error in estimating the contribution ratio of paddy field water at point scale. By contrast, the tracer approach provided better results in estimating the contribution ratio of recharge sources at point scale, particularly for paddy field water. Thus, the combined use of these two methods can provide more reliable results and eliminate the limitations of both methods. 3. The characteristics of groundwater recharge in the Nasunogahara area are as follows: (1) Precipitation is the major recharge source and its mean contribution ratio is 58% for a year-long period over the entire alluvial fan. (2) The distribution of contribution ratio of river water is nearly symmetrically distributed along the Sabi River. The river water seepage is significant, particularly in the upstream area, and the mean contribution ratio of river water is 28%. (3) The paddy field water recharge mainly occurred in the midstream and downstream areas. The mean contribution ratio of paddy field water is 14%, whereas approximately three quarters of the alluvial fan was recharged by the paddy field water. 4. The contribution ratio of the paddy field water increased from the dry to the wet season (irrigation period), but the seasonal variation at point scale was less than 20% in most areas. The contribution ratio of precipitation decreased as the paddy field water recharge increased. The seasonal variation of the mean contribution ratio of the river water was relatively stable (61%), but the contribution ratio of river water along the river channels at point scale in the upstream and midstream areas slightly increased during the wet season.
Acknowledgements The research was supported by the State Scholarship Fund of the China Scholarship Council, China (File No. 2009640008), the China Postdoctoral Science Foundation, China (Grant No. 2013M540960) and the special study fund of Terrestrial Environment Research Center, University of Tsukuba, Japan. The authors would like to thank two anonymous reviewers for their helpful comments regarding this manuscript. References
5. Conclusions The tracer approach, using end-member mixing analysis, is useful for estimating the contribution ratio of recharge sources to groundwater at point scale in the Nasunogahara area. However, the possible error of the estimated contribution ratio is slightly large (ranging from 12% to 63%). A numerical model that includes a groundwater flow model (MODFLOW) and a transport model (MT3DMS) was used to simulate the spatial and temporal
Anderson, M.P., Woessner, W.W., 1992. Applied Groundwater Modeling: Simulation of Flow and Advective Transport. Academic Press, San Diego, California. Carroll, R.H., Pohll, G.M., Earman, S., Hershey, R.L., 2008. A comparison of groundwater fluxes computed with MODFLOW and a mixing model using deuterium: application to the eastern Nevada Test Site and vicinity. J. Hydrol. 361, 371–385. Chen, Z., Nie, Z., Zhang, G., Wan, L., Shen, J., 2006. Environmental isotopic study on the recharge and residence time of groundwater in the Heihe River Basin, northwestern China. Hydrogeol. J. 14, 1635–1651. Choi, B.Y., Yun, S.T., Mayer, B., Chae, G.T., Kim, K.H., Kim, K., Koh, Y.K., 2010. Identification of groundwater recharge sources and processes in a
Y. Liu et al. / Journal of Hydrology 519 (2014) 833–847 heterogeneous alluvial aquifer: results from multi-level monitoring of hydrochemistry and environmental isotopes in a riverside agricultural area in Korea. Hydrol. Process. 24, 317–330. Clark, I., Fritz, P., 1997. Environmental Isotopes in Hydrogeology. CRC Press, New York, p. 328. Dahan, O., McGraw, D., Adar, E., Pohll, G., Bohm, B., Thomas, J., 2004. Multi-variable mixing cell model as a calibration and validation tool for hydrogeologic groundwater modeling. J. Hydrol. 293 (1–4), 115–136. Doherty, J., 1998. Visual PEST: Graphical Model Independent Parameter Estimation. Watermark Computing and Waterloo Hydrogeologic Inc. Eastoe, C.J., Hutchison, W.R., Hibbs, B.J., Hawley, J., Hogan, J.F., 2010. Interaction of a river with an alluvial basin aquifer: stable isotopes, salinity and water budgets. J. Hydrol. 395, 67–78. Elhassan, A.M., Goto, A., Mizutani, M., 2001. Combining a tank model with a groundwater model for simulating regional groundwater flow in an alluvial fan. Irrig. Drainage Rural Eng. J. 215, 21–29. Elhassan, A.M., Goto, A., Mizutani, M., 2003. Effect of conjunctive use of water for paddy field irrigation on groundwater budget in an alluvial fan. Agric. Eng. Int: the CIGR J. Sci. Res. Develop. V, Manuscript LW03002. Elhassan, A.M., Goto, A., Mizutani, M., 2006. Simulated effects of changing land use on shallow aquifer of Nasunogahara basin, Japan. An American Society of Agricultural and Biological Engineers meeting presentation, paper No. 062285. Famiglietti, J.S., Lo, M., Ho, S.L., Bethune, J., Anderson, K.J., Syed, T.H., Swenson, S.C., de Linage, C.R., Rodell, M., 2011. Satellites measure recent rates of groundwater depletion in California’s Central Valley. Geophys. Res. Lett. 38, L03403. http:// dx.doi.org/10.1029/2010GL046442. Foster, S., Garduno, H., Evans, R., Olson, D., Tian, Y., Zhang, W., Han, Z., 2004. Quaternary Aquifer of the North China Plain-assessing and achieving groundwater resource sustainability. Hydrogeol. J. 12, 81–93. Freeze, R.A., Cherry, J.A., 1979. Groundwater. Prentice Hall, Englewood Cliffs. Fujinawa, K., 1981. The role of water for agricultural use as the source of groundwater recharge; analysis of groundwater in Nasuno-ga-hara Basin by a mathematical model. Bull. Natl. Res. Inst. Agric. Eng. 21, 127–141 (Japanese with English abstract). Galloway, D.L., Burbey, T.J., 2011. Review: regional land subsidence accompanying groundwater extraction. Hydrogeol. J. 19, 1459–1486. Harbaugh, A.W., Banta, E.R., Hill, M.C., McDonald, M.G., 2000. MODFLOW-2000, the U.S. Geological Survey modular ground-water model – User guide to modularization concepts and the Ground-Water flow process. U.S. Geological Survey Open-File Report 00–92. 121 pp. He, Z., Wang, Y., Tian, F., Ni, G., 2011. A simulation on the recovery of groundwater level in Beijing Plain after the implementation of South-to-North water transferring project. In: Water Resources and Environmental Protection (ISWREP), 2011 International Symposium on 1, 266–270. Houston, J., 2002. Groundwater recharge through an alluvial fan in the Atacama Desert, northern Chile: mechanisms, magnitudes and causes. Hydrol. Process. 16, 3019–3035. Johnson, A.I., 1967. Specific yield – compilation of specific yields for various materials. U.S. Geological Survey Water Supply Paper 1662-D. 74 pp. Kelly, B.P., 2004. Simulation of ground-water flow, contributing recharge areas, and ground-water travel time in the Missouri River alluvial aquifer near Ft. Leavenworth, Kansas: U.S. Geological Survey Scientific Investigations Report 2004–5215. 69 pp. Kim, K., Anderson, M.P., Bowser, C.J., 1999. Model calibration with multiple targets: a case study. Ground Water 37 (3), 345–351. Lambs, L., 2004. Interactions between groundwater and surface water at river banks and the confluence of rivers. J. Hydrol. 288 (3–4), 312–326. Leake, S.A., Prudic, D.E., 1991. Documentation of a computer program to simulate aquifer-system compaction using the modular finite-difference ground-water flow model: U.S. Geological Survey Techniques of Water-Resources Investigations 06–A2. 68 pp. Liu, Y., Yamanaka, T., 2012. Tracing groundwater recharge sources in a mountainplain transitional area using stable isotopes and hydrochemistry. J. Hydrol. 464–465, 116–126. Liu, D., Tian, F., Hu, H., Lin, M., Cong, Z., 2012. Ecohydrological evolution model on riparian vegetation in hyperarid regions and its validation in the lower reach of Tarim River. Hydrol. Process. 26, 2049–2060. McDonald M.G., Harbaugh A.W., 1988. A modular three-dimensional finitedifference ground-water flow model. U.S. Geological Survey Techniques of Water-Resources Investigations, Book 6, Chapter A1. 586 pp.
847
Mikita, M., Yamanaka, T., Lorphensri, O., 2011. Anthropogenic changes in a confined groundwater flow system in the Bangkok basin, Thailand, part I: was groundwater-recharge enhanced? Hydrol. Process. 25 (17), 2726–2733. Phillips, D.L., Gregg, J.W., 2001. Uncertainty in source partitioning using stable isotopes. Oecologia 127 (2), 171–179. Prudic, D.E., 1989. Documentation of a computer program to simulate streamaquifer relations using a modular, finite-difference, ground-water flow model: U.S. Geological Survey Open-File Report 88–729. 113 pp. Reynolds, D.A., Marimuthu, S., 2006. Deuterium composition and flow path analysis as additional calibration targets to calibrate groundwater flow simulation in a coastal wetlands system. Hydrogeol. J. 15 (3), 515–535. Sampat, P., 2000. Groundwater shock: the polluting of the world’s major freshwater stores. World Watch, January/February, 9–22. Sanford, W., Plummer, L.N., McAda, D., Bexfield, L., Anderholm, S., 2004. Hydrochemical tracers in the middle Rio Grande Basin, USA: 2. Calibration of a groundwater-flow model. Hydrogeol. J. 12, 389–407. Sasaki, M., Ajisaka, T., Okamoto, A., 1958. Hydrogeology of the Nasu plain, Tochigi Prefecture. J. Geography 67, 59–73 (Japanese with English abstract). Scanlon, B.R., Healy, R.W., Cook, P.G., 2002. Choosing appropriate techniques for quantifying groundwater recharge. Hydrogeol. J. 10, 18–39. Sibanda, T., Nonner, J.C., Uhlenbrook, S., 2009. Comparison of groundwater recharge estimation methods for the semi-arid Nyamandhlovu area, Zimbabwe. Hydrogeol. J. 17, 1427–1441. Stichler, W., Maloszewski, P., Bertleff, B., Watzel, R., 2008. Use of environmental isotopes to define the capture zone of a drinking water supply situated near a dredge lake. J. Hydrol. 362 (3–4), 220–233. Surinaidu, L., Bacon, C.G.D., Pavelic, P., 2013. Agricultural groundwater management in the Upper Bhima Basin, India: current status and future scenarios. Hydrol. Earth System Sci. 17, 507–517. Thornthwaite, C.W., 1948. An approach toward a rational classification of climate. Geograph. Rev. 38, 55–94. Wakui, H., 2007. Numerical simulation on isotopic changes in groundwater in the Nasu Fan, Tochigi Prefecture. Master thesis, University of Tsukuba, Japan. 73 pp. Wakui, H., Yamanaka, T., 2006. Sources of groundwater recharge and their local differences in the central part of Nasu fan as revealed by stable isotope. Groundwater Hydrology (J. Japanese Assoc. Hydrol. Sci.) 48 (2), 263–277 (Japanese with English abstract). Watanabe, K., Sagehashi, N., 1960. Geologic history of the Nasu Plain, Tochigi Prefecture. J. Geol. Soc. Japan 66, 147–156 (Japanese with English abstract). Watanabe, K., Sagehashi, N., 1962. ‘‘Kwanto Loam’’ of the Nasu Plain, Tochigi Prefecture. J. Geol. Soc. Japan 68, 451–461 (Japanese with English abstract). Watanabe, K., Sagehashi, N., Shindo, S., 1960. Geologic structure of the Nasu Plain, Tochigi Prefecture, with special reference to the ancestral river course of the Naka River. J. Geol. Soc. Japan 66, 113–122 (Japanese with English abstract). Wu, Y., Zhang, Y., Wen, X., Su, J., 2010. Hydrologic cycle and water resource modeling for the Heihe River basin in Northwestern China. Science Press, Beijing (Chinese). Yamanaka, T., Wakui, H., 2009. Intensive groundwater-surface water interaction in an alluvial fan: assessment using a numerical model and isotopic tracer. In: Taniguchi, T. et al. (Eds.), From Headwaters to the Ocean: Hydrological Changes and Watershed Management. Taylor & Francis Group, London, pp. 321–327. Yamanaka, T., Mikita, M., Lorphensri, O., Shimada, J., Kagabu, M., Ikawa, R., Toshio, N., Tsujimura, M., 2011. Anthropogenic changes in a confined groundwater flow system in the Bangkok Basin, Thailand, part II: how much water has been renewed? Hydrol. Process. 25 (17), 2734–2741. Zhang, H., Hiscock, K.M., 2010. Modelling the impact of forest cover on groundwater resources: a case study of the Sherwood Sandstone aquifer in the East Midlands, UK. J. Hydrol. 392, 136–149. Zhang, H., Hiscock, K.M., 2011. Modelling the effect of forest cover in mitigating nitrate contamination of groundwater: a case study of the Sherwood Sandstone aquifer in the East Midlands, UK. J. Hydrol. 399, 212–225. Zheng, C., 1990. MT3D: a modular three-dimensional transport model for simulation of advection, dispersion and chemical reactions of contaminants in groundwater systems. Report to the U.S. Environmental Protection Agency. 170 pp. Zheng, C., Bennett, G.D., 2002. Applied contaminant transport modeling, second ed. John Wiley & Sons Inc, New York. Zheng, C., Wang, P.P., 1999. MT3DMS: a modular three-dimensional multispecies transport model for simulation of advection, dispersion, and chemical reactions of contaminants in groundwater systems: documentation and user’s guide. Contract Report SERDP-99-1. U.S. Army Engineer Research and Development Center, Vicksburg, Mississippi. 202 pp.
