Impacts of Urbanization and Climate Variability on Groundwater Environment in a Basin Scale

Abstract

Globally, groundwater resources are experiencing a decline in hydraulic heads resulting from the dual effects of urbanization and climate change, highlighting the need for integrated and sustainable water resources management. Urban development in the cities of Kansai region, western Japan, presents a significant challenge to the sustainability of groundwater resources. This study aims to assess the combined influence of urbanization and climate change on the groundwater resources of the Nara Basin using MODFLOW 6 for two distinct periods: The Pre-Urbanization Period (PreUP: 1980–1988), and the Post-Urbanization Period (PostUP, 2000–2008) with an emphasis on spatiotemporal distribution of recharge in a multi-layer aquifer system. Simulated hydraulic heads were evaluated under three different recharge scenarios: uniformly, spatiotemporally and spatially distributed. The uniform recharge scenario both overestimates and underestimates hydraulic heads, while the spatially distributed scenario produced a simulated heads distribution similar to the spatiotemporally distributed recharge scenario, underscoring the importance of incorporating spatiotemporal variability in recharge input for accurate groundwater flow simulation. Moreover, our results highlight the relevance of spatial distribution of recharge input than temporal distribution. Our findings indicate a significant decrease in hydraulic heads of approximately 5 m from the PreUP to PostUP in the unconfined aquifer, primarily driven by changes in land use and climate. In contrast, the average head decline in deep confined aquifers is about 4 m and is mainly influenced by long-term climatic variations. The impervious land use types experienced more decline in hydraulic heads than the permeable areas under changing climate because of the impedance to infiltration and percolation exacerbating the climate variability effect. These changes in hydraulic heads were particularly evident in the interactions between surface and groundwater. The cumulative volume of groundwater discharge to the river decreased by 27%, while the river seepage into the aquifer increased by 16%. Sustainable groundwater resources management under conditions of urbanization and climate change necessitates a holistic and integrated approach.

1. Introduction

Many regions around the world are currently experiencing a rapid decline in groundwater resources due to population growth, climatic pressures and variability [1,2]. Over the past 50 years, the global population has doubled from approximately 4.02 billion to 8.06 billion [3]. While urban areas have historically accommodated a dense population within limited space [4]; recent decades have witnessed rapid expansion of urban cities, which has significantly impacted the global environment and intensified pressure on groundwater resources [5,6,7,8,9]. One of the key impacts is the alteration of groundwater percolation rates due to the widespread development of impervious surfaces [10]. Consequently, urban growth and associated land cover changes profoundly influence hydrologic processes, including evapotranspiration, and infiltration, affecting both surface and subsurface water systems.

The impact of climate variability has been deduced to directly affect surface water features such as streams and lakes driven by variations in climatic variables including rainfall and temperature [11]. However, its influence on groundwater resources is more complex, involving intricate and often indirect connections [11]. Variations in precipitation and air temperature can influence aquifer recharge and evapotranspiration rates [11,12,13]. An exhaustive evaluation of the impacts of climate variability on groundwater resources requires accurate estimations of critical meteorological variables and groundwater recharge rates [11]. Likewise, determining the spatial and temporal distribution of groundwater recharge presents significant challenges due to factors such as land use and hydrogeological variability. Consequently, estimating groundwater recharge is essential for efficient integrated river basin management strategies targeted towards safeguarding the scarce groundwater resources. Numerous studies conducted over the past two decades have assessed how climate change would affect hydraulic heads and regional groundwater recharge in various regions of the world, indicating that these changes may have either favorable or adverse consequences [14,15,16,17]. Based on 16 Global Climate Models (GCM), deep aquifer recharge will either rise or decline by approximately 10% by 2050 in the High Plains aquifer, USA [15]. Research conducted on the Mancha Oriental aquifer system in Spain, which assessed the response of groundwater resources to changing climate and land use, revealed a significant reduction in groundwater recharge for all scenarios where climate change has a greater impact than land use change [18].

Various models have been employed to evaluate the effects of climatic variability and land use alterations on surface and groundwater resources, such as hydrological models [10,19,20,21,22], groundwater numerical models [23,24], and integrated model such as SWAT-MODLOW [1,25,26]. MODFLOW is a three-dimensional groundwater flow model that uses the finite-difference method for the simulation of groundwater flow [27]. Coupled hydrological and hydrogeological models are commonly used to simulate both the surface and groundwater systems; however, the inherent assumption in these models makes it difficult to adjust and assess certain parameters and variables [18]. If appropriately constructed and calibrated, numerical models that include spatiotemporal heterogeneity offer the most reliable method of estimating the effects of changing land use and climate on groundwater systems [18]. However, in previous studies, the combined effects of urbanization and climate variability on the groundwater environment have not been thoroughly examined, particularly when considering the spatiotemporal variation in groundwater recharge.

Therefore, this study aims to evaluate the combined influence of urbanization and climate variability on the groundwater environment at the basin scale between two periods (Pre-Urbanization Period (PreUP): 1980–1988, and Post-Urbanization Period (PostUP): 2000–2008) using MODFLOW 6 with an emphasis on spatiotemporal distribution of recharge in a multi-layer aquifer system. The specific objectives include: (1) Assessment of the optimal recharge scenario for a precise and representative hydraulic heads simulation in a multi-layer aquifer system considering three scenarios; uniform, spatiotemporally distributed, and spatially distributed, (2) Evaluation of the influence of urbanization and climate variability on surface water and groundwater interactions over the entire basin, (3) Comparative analysis of the dynamics of hydraulic heads observed in permeable and impermeable zones of the basin.

2. Study Area

2.1. Geography of Nara Basin

Nara Basin (34°23′7″~34°46′48″ N, 135°39′15″ E~135°56′50″ E), a landlocked catchment located in Nara Prefecture of Western Japan and surrounded by the Ikoma and Kongo Mountains to the west, Ryūmon and Kasagi Mountains to the South and East, respectively (Figure 1). The basin boundary, as defined by Wang et al. [21] covers approximately 714 km² with about 56% and 45% coverage for the plains and mountains, respectively [21]. The elevation ranges from 200 to 1000 m on the hills, whereas the plain area spans approximately 16 km from east to west and 25 km from north to south (Figure 1). The Nara Plain encompasses roughly 300 km², with elevations ranging from 40 to 100 m [28]. The long-term (30 years) average annual precipitation and mean temperature are about 1450 mm/year and 14 °C respectively across the entire basin [29]. The Nara Basin experiences a distinct rainfall pattern that is uneven both spatially and temporally; the rainy seasons occur between June and July, while the typhoon seasons with intensified rainfall occur between September and October [29].

The main river in the Nara Basin is the Yamato River, extending 68 km in length with a catchment area of 1077 km² [19]. It originates from the Kasagi Mountain with steep flow and subsequently flows into the Nara plains, connecting with tributaries downstream. Traversing the central region of the Nara Basin, the Yamato River flows gently westward, thereby depositing its sediments and exiting into the Osaka plains downstream [28,30]. The steep gradient of the Yamato River upstream leads to rapid water movement, which swiftly transitions into its middle course, coupled with low water retention capacity during significant flooding [28]. Conversely, during prolonged dry seasons, the river tends to dry up rapidly due to its interaction with underlying groundwater resources [28].

2.2. Hydrogeology of Nara Basin

The Nara Basin is primarily characterized by a sequence of north-south movements in the earth’s crust that took place during the late Pliocene and Pleistocene epochs, along with east-west compressions that followed the Pleistocene period [31]. The geological structure of the Nara Basin consists of bedrock from the Ryoke belt (granite), which outcrops in the mountains and is overlain by sedimentary layers from the Upper and Lower Osaka Group (Plio-Pleistocene) including interbedded layers of sand, gravel, and marine and nonmarine clay conglomerates (Figure 2) [30]. The alternating beds of marine and nonmarine clay, sand and gravel form various hydrogeological units (aquitards and aquifers) that define the groundwater resources in Nara Basin. Above the Osaka Group are the terrace deposits along the foothills, whereas alluvial deposits dominate the surface of the plains and cover more than 60% of the basin. In addition, several faults can be identified in the Nara Basin including the Nara-Toen fault in the northeastern part, and the southwestern Kongo fault spanning from north to south which forms the rift valley of the Nara Plain [28,30].

Both shallow and deep groundwater resources are spatially distributed within the Nara plains and in fractured rocks in the southern part of Nara Prefecture. Percolating surface water reaches the deep aquifers in the northern part of Nara [32]. Moreover, along the foothills of the Ikoma, Katsuragi and Kongo mountains, deep groundwater resources in confined aquifers are present due to the prevalence of gravels and coarse terrace deposits in this region [28]. Groundwater discharge occurs in the center of the basin along the Yamato River due to low hydraulic heads and the connectivity of the river with the basin’s water table. The hydrogeological units are separated by both marine and nonmarine clays in different layers in the subsurface, where the alluvial plain of sand and gravel serves as the basin’s major aquifers [30,33].

2.3. Land Use Classification of Nara Basin

The dominant land use types in the Nara Plain are urban and rice paddy. For the purpose of analysis, the various land use types are subdivided into pervious and impervious surfaces as shown in

Table 1. Land use classification based on permeability.

Pervious Surfaces	Impervious Surfaces
Rice Paddy	Built-up (urban)
Croplands	Barren-Land (Rocks)
Orchards	Roads
Tree Fields	Public Area
Forest	Other sites

. Other sites include areas such as athletic fields, airports, horse racing tracks, baseball stadiums, schools and man-made land which are mostly composed of impervious surfaces [34]. Agriculture in the Nara Basin has advanced swiftly since ancient times, establishing Nara as an important agricultural region of Western Japan [21]. However, urban development is rapidly growing at a rate exceeding agricultural expansion (Figure 3). From 1987 to 2006, the built-up area transitioned from the second most prevalent land use type to the first, experiencing an approximate increase of 38%. In contrast, the area dedicated to rice paddy cultivation declined by approximately 20%. Surrounding the Nara plain are the mountains classified into four forest regions with varying vegetation distribution, which includes evergreen coniferous vegetation, broad-leaved evergreen vegetation types, shrub and bamboo forests [19,35,36].

Prior studies conducted in the Nara Basin reveal drastic change in land use due to urbanization between 1976 and 2016 and a corresponding increase in the percent imperviousness of urban areas (PIU), thereby resulting in a rapid decline in the recharge rates across the basin [10]. The PIU of the northern and southern parts of Nara Basin increased by 14% and 16%, respectively, from the 1970s to the 2010s. The impervious land use types (Built-up, Roads, Public areas) produce the lowest groundwater recharge, with values approximately 10% lower than those in pervious land use types (Rice paddy fields, and cropland). Additionally, the hydraulic heads and estimated recharge rates in the Nara Basin have been evaluated using various methods, including numerical models, water balance method coupled with groundwater temperature profiles, heat-flow techniques and hydrogeological models [30,33,37,38]. However, no study to date has examined the combined impact of land use and climate variability on hydraulic heads in the Nara Basin.

3. Methods

3.1. Data Collection

For this study, hydrological, geological and groundwater abstraction datasets were collected. The SRTM 1 Arc-Second Global dataset was obtained from the United States Geological Survey (USGS) and used as elevation data [39]. The surface geological and soil maps were obtained from the Ministry of Land, Infrastructure, Transport, and Tourism of Japan (MLIT) [40]. Additionally, hydrological datasets, including river stages for the simulated rivers in the Nara Basin were obtained from the Nara Prefecture River Information System (a subsidiary of MLIT) [41]. High-resolution land use raster (with a spatial resolution of 30 m) was obtained from Japan Aerospace Exploration Agency (JAXA) Earth Observation Research Center (EORC) [42] and from MLIT [34].

The recharge datasets were inherited from previous research focused on the long-term water balance of the Yamato River Catchment, which utilized the Soil and Water Assessment Tool (SWAT) to analyze the spatial and temporal distribution of both surface and groundwater recharge influenced by urbanization and forest growth [10,19]. This study analyzed two periods, from 1989 to 1998 and from 1999 to 2008, and recorded shallow and deep percolation outputs from SWAT on a subbasin scale. These data were incorporated into the groundwater flow model to assess the impact of urbanization on groundwater flow dynamics. Furthermore, to compare the effects of spatiotemporally distributed versus constant recharge, a uniform annual recharge rate of 450 mm/year was adopted as a benchmark value for simulation purposes [38].

Groundwater abstraction datasets were obtained from the individual city office in charge of water supply in the Nara Basin [43,44]. The locations of the pumping wells are shown in Figure 2 above, with depths ranging from 2 to 50 m, including the industrial areas with wells pumping at different rates ranging from 100 to 7000 m³/day from the deep confined aquifers. In addition, the regions surrounding the Yamato River at the center of the basin have the highest level of groundwater abstraction.

Borehole logs along vertical cross-sections were obtained from the Geospatial Information Authority of Japan [45] to delineate hydrostratigraphic units with similar hydrogeologic properties in the Nara Basin (Figure 4).

3.2. Groundwater Flow Model

3.2.1. Conceptual Model of Nara Basin

The conceptual model of Nara Basin was designed based on the available geological, hydrological and hydrostratigraphic datasets. The surface geological map unveils the bedrock in the northeastern–southeastern part of the basin along the Nara-Toen fault; therefore, a no-flow boundary was assigned. Also, another no-flow boundary condition was assigned to the western part of the basin due to the impermeable surfaces of the Ikoma and Kongo mountains. Whereas the groundwater-divide between the Nara Basin and the Kizu River Basin, as indicated by the regional groundwater head table, demonstrates a hydraulic no-flow boundary in the northern region [30]. A defined head boundary was assigned at the outlet of the Yamato River, which was also adopted in previous studies conducted in this basin [30,33].

The hydrostratigraphic units of the subsurface including aquifers and aquitards in the Nara Basin were developed by examining, analyzing and simplifying the lithologic descriptions contained in 10 borehole logs [45] along the west-east cross-section of the Nara Plain, as shown in Figure 4 above. The hydrostratigraphic layers identified from the careful examination of the cross-sections include three layers of deep confined aquifers, one shallow confined aquifer and an unconfined aquifer, separated by marine and nonmarine clay layers. The clay layers function as confining beds (Aquitards). The thickness of the unconfined aquifer (UCA), shallow confined aquifer (SCA), and the three deep confined aquifers (DCA) ranges from 6–20 m, 30–40 m, 55 m, 40 m, and 70 m, respectively in the Nara Plain. Additionally, the aquitards increase in layer thickness and layer continuity with depth, while the lateral variations of layer thickness decrease significantly from the SCA to DCA.

The rivers were simulated using the head-dependent flux boundary condition, particularly the Stream Flow Routing Package, while the specified flux boundary condition (recharge package) was used for the recharge. The river stage datasets, reach, width and hydraulic conductivity of the riverbed obtained from MLIT [46] were assigned for the computation of riverbed conductance.

3.2.2. Construction of the Numerical Model

A three-dimensional numerical groundwater flow model of the Nara Basin was constructed using MODFLOW 6, an open-source software developed by the United States Geological Survey (USGS) [27], and coupled with the ModelMuse graphical user interface (GUI), which facilities input file preparation and visualization of the model result. MODFLOW 6 is a fully distributed model that uses the generalized control-volume finite-difference method for solving groundwater flow equations. It is an object-oriented, modular platform leveraging an adaptable and extendable architecture to combine groundwater flow and transport operations [27]. Furthermore, a unique feature of MODFLOW 6, which distinguishes it from previous versions, is the Discretization by Vertices (DISV) method used for unstructured grids to model faults, fractures, and complex geology.

The study area (the Nara Basin) was discretized into a model grid size of 300 × 300 m, 145 rows, 85 columns, and 9 layers, consisting of 5 aquifers and 4 confining beds (aquitards), as shown in Figure 4. The total number of active cells over the entire basin is 49,231 cells. The model top was defined using the SRTM 1 Arc-Second Global dataset obtained from the USGS [39], while the thickness of the underlying layers was determined from cross-sectional interpretations of over 80 borehole logs collected across the basin (Figure 4). Initial hydraulic heads were defined as the top surface of the model for a steady-state groundwater flow model, the output of which was subsequently utilized for the transient groundwater flow model simulated over the period 2000–2008. The transient simulation was designed to assess annual variations in groundwater flow dynamics, which were considered when defining stress periods and time step.

The MODFLOW packages used include the Multi-Aquifer Well (MAW) Package for simulating municipal and industrial groundwater abstraction, the Streamflow Routing (SFR) Package for upstream and downstream rivers in a connected manner, the Recharge (RCH) package, and the Time-variant Specified-Head (CHD) package. Model input parameters are the horizontal hydraulic conductivity, river heads, well pumping rate, riverbed hydraulic conductivity, groundwater observations, and the spatiotemporal recharge rate [10,38]. Furthermore, the SFR package was used to assess the interactions between surface water and groundwater in the Nara Basin.

3.2.3. Model Calibration, Performance Evaluation and Sensitivity Analysis

Model calibration was performed using hydraulic head datasets from observation wells distributed across the Nara Plain, as shown in Figure 2. Daily groundwater head data were collected from 13 observation sites and subsequently aggregated into annual time series from 2000 to 2008 [46].

In this study, manual calibration coupled with Independent Parameter Estimation (PEST), was performed by varying the hydraulic conductivities, river conductivity, specific yield and specific storage iteratively until a good fit was achieved. PEST uses inverse modelling and an advanced regularization technique for the estimation of model parameters with the aim of minimizing the differences between measured and simulated hydraulic heads [47]. The steady-state model was initially calibrated using long-term averages of observed hydraulic heads, which was followed by the calibration of the transient-state model.

While the recharge was kept constant during calibration, it was later used in scenario analyses to examine the effect of different recharge inputs on hydraulic heads in the Nara basin. At each phase of the calibration process, simulated heads were compared with the observed heads using the following model performance metrics: Root Mean Square Error (RMSE), Correlation coefficient (R²), and Water Balance Error (WBE). The calibration process was halted after attaining a satisfactory fit between the simulated and observed heads.

$$\mathrm{R}\mathrm{o}\mathrm{o}\mathrm{t} \mathrm{M}\mathrm{e}\mathrm{a}\mathrm{n} \mathrm{S}\mathrm{q}\mathrm{u}\mathrm{a}\mathrm{r}\mathrm{e} \mathrm{E}\mathrm{r}\mathrm{r}\mathrm{o}\mathrm{r} \mathrm{R}\mathrm{M}\mathrm{S}\mathrm{E}=\sqrt{{\displaystyle \frac{{\sum }_{j=1}^n({h_{obs}-h_{sim})}_j^2}{n}}},$$

(1)

$$\mathrm{C}\mathrm{o}\mathrm{e}\mathrm{f}\mathrm{f}\mathrm{i}\mathrm{c}\mathrm{i}\mathrm{e}\mathrm{n}\mathrm{t} \mathrm{o}\mathrm{f} \mathrm{D}\mathrm{e}\mathrm{t}\mathrm{e}\mathrm{r}\mathrm{m}\mathrm{i}\mathrm{n}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n} {\mathrm{R}}^2=1-{\displaystyle \frac{\sum {(h_{obs}-h_{sim})}^2}{\sum {(h_{obs}-{\overline{h}}_{obs})}^2}},$$

(2)

$$\mathrm{W}\mathrm{a}\mathrm{t}\mathrm{e}\mathrm{r} \mathrm{B}\mathrm{a}\mathrm{l}\mathrm{a}\mathrm{n}\mathrm{c}\mathrm{e} \mathrm{E}\mathrm{r}\mathrm{r}\mathrm{o}\mathrm{r} \mathrm{W}\mathrm{B}\mathrm{E} \left(\mathrm{\%}\right)=\left({\displaystyle \frac{Inflow-Outflow-∆Storage}{Inflow}}\right)\times 100.$$

(3)

where h_obs, h_sim and n are the observed head, simulated head, and the total number of observed data, respectively.

3.2.4. Scenario Analysis for Varying Recharge Input

The impact of varying recharge input on the groundwater system was examined based on three scenarios: uniform recharge scenario (Scenario A), spatiotemporally distributed recharge scenario (Scenario B), and spatially distributed temporally averaged scenario (Scenario C). For scenario A, an estimated annual groundwater recharge rate of 450 mm/year was used, as derived from the study by Taniguchi (1994) [38]. This study employed a type-curve approach based on temperature-depth profiles of groundwater to estimate the water balance across the Nara Basin. This recharge rate was applied uniformly across the entire basin and simulation period.

The spatial and temporal distribution of groundwater recharge datasets for scenario B was obtained from previous research that analyzed the impacts of urbanization and forest growth on the long-term water balance in the Yamato River Catchment using the Soil and Water Assessment Tool (SWAT) [10,19]. This previous study employed two periods, 1989–1998 and 1999–2008, and produced subbasin scale estimates of shallow and deep aquifer recharge. In the present study, these SWAT-derived recharge datasets were integrated into the groundwater flow model to assess the impact of urbanization on hydraulic heads and surface water-groundwater interactions. The SWAT hydrological model output for the Yamato River Catchment, derived from 39 subbasins, was retrieved and subsequently clipped to match the spatial extent of the Nara Basin, revealing both spatial distribution and annual fluctuations in groundwater recharge rates over the simulation period. For scenario C, a 9-years average of the scenario B dataset was used, resulting in spatially distributed recharge with no temporal distribution. The spatial variation of average recharge during the PostUP shows the highest recharge rate in the paddy fields, followed by other pervious land use types, including croplands and orchards, and the lowest in the urban areas. In contrast, PreUP forest land use exhibits the highest recharge rates, followed by the paddy fields and other pervious surfaces, with the lowest recharge in urban and impervious areas. This pattern is attributed to the presence of large macropores in forested areas, which enhance groundwater infiltration and recharge [19].

To evaluate the relative importance of spatial and temporal variation in recharge input across the basin, the simulated head difference between scenario A and scenario B $\left(A-B\right)$, and scenario B and scenario C $\left(B-C\right)$ were analyzed.

3.2.5. Evaluation of the Impacts of Climate Variability and Urbanization

The groundwater flow model was simulated for two different periods PostUP (2000–2008), and PreUP (1980–1988), to evaluate the influence of urbanization and climate variability on hydraulic heads and surface water and groundwater interactions in Nara Basin. The baseline simulation period is PostUP. However, the calibrated transient model was also simulated for PreUP with different land use and climate conditions to assess both the effect of climate variability and urbanization on the baseline simulation period. Based on data obtained from the Japan Meteorological Agency [29], the PreUP period experienced higher rainfall than the PostUP period, reflecting a shift towards drier conditions. Total rainfall over the Nara Basin decreased by 1.5% between the two periods. Furthermore, the 9-years averaged daily mean, minimum and maximum temperature over the basin between the PreUP and PostUP increased by 8%, 12%, and 7.5%, respectively. Solar radiation increased by about 4%, from 12.6 MJ/m² to 13.1 MJ/m², which corresponds to the rise in mean temperature and may have resulted in a proportionate increase in evapotranspiration and a decrease in groundwater recharge. Wang et al. [10] estimated a decadal rise in total evapotranspiration in the forested regions (Ikoma, Kasuga, and Kongo) of 33%, 6% and 15% between PreUP and PostUP due to climate change, forest growth and anthropogenic influences.

According to Wang et al. [10], groundwater recharge to deep aquifers (excluding baseflow) decreased between the PreUP and PostUP by 28%, from 123 mm/year to about 94 mm/year; however, surface runoff increased by 27% between the two periods. It is evident that alterations in climatic parameters, specifically a decrease in rainfall and an increase in temperature of 1.5% and 8%, respectively, cannot alone explain the 28% decline in groundwater recharge to the deep aquifer; hence, changes in land use, notably urbanization, also significantly contribute. The increase in surface runoff can be attributed to urban expansion in the Nara Basin and the rise in percent imperviousness of urban areas over the past 50 years [10]. This highlights the need to assess the combined impact of climate variability and urban expansion.

Pixel analysis and zonal statistical summaries were performed to assess the combined impact of urbanization and climate variability on hydraulic heads in Nara Basin. Pixels that maintained the same land use between the two simulated periods were analyzed to isolate the stand-alone impact of climate change, while pixels that transitioned to urban land use were analyzed to evaluate the combined effects of climate variability and urbanization.

4. Results and Discussion

4.1. Estimation of Hydraulic Heads

The comparison between the measured hydraulic heads and the simulated heads at the observation wells (Figure 1), based on the spatiotemporally distributed recharge scenario, is illustrated in the scatter plot below (Figure 5). The result shows a good fit between the observed and the simulated heads with a high R² value of 0.95, and a low RMSE value of 3.97 m, indicating that the model is well-calibrated. This performance presents an acceptable correlation with observed values and is significantly better than the prior work by Pathak et al. [30], who recorded an RMSE value of 9.7 m. More than two-thirds of the residuals calculated by subtracting the simulated heads from the observed heads fall within the range of −3.5 to +3.5 m. The model’s strong performance is attributed to the availability of detailed and spatially representative input data in these regions, which contributed to the accurate simulation of the observed hydraulic heads.

The model slightly overestimated hydraulic heads in most of the observation wells. Overestimated points with residuals greater than 3.5 m were observed at Imaizumi (OBS_4), Nishikujo (OBS_13) and Oji_Temple (OBS_7) observation points (Figure 5). These overestimations could be due to uncertainties in the no-flow boundary conditions, unreported groundwater abstraction, and the potential overestimation of groundwater recharge in these regions. These residuals were notably larger prior to calibration and were rectified by adjusting the uncertain model parameters outlined in the preceding section. In Nishikujo, the overestimation may be linked to unreported groundwater abstraction by farmers for rice paddy irrigation. Although the Oji_Temple and Imaizumi locations are close to the hilly regions of Mt. Ikoma and Mt. Kongo, classified as no-flow boundaries based on surface geological map, the land use maps indicate that those mountainous areas are forested and experience high evapotranspiration rates [19], which may affect recharge estimation rates.

Underestimated hydraulic heads with residuals below −3 m are limited to Bojo_Castle (OBS_1) and Tenrikita (OBS_10) (Figure 5) due to the underestimation of recharge, especially in OBS_1, which is located in the fan-shaped area of Mt. Katsuragi capable of storing large recharge due to its coarse-grained conglomerates [32].

Groundwater is being pumped out from both shallow and deep wells for domestic and industrial use in Nara Basin, especially in Yamatokoriyama city, with the aid of abstraction wells, as shown in Figure 2. This region is known for over-abstraction beyond the aquifer’s recovery rate, leading to a significant decline in groundwater pressure heads. Figure 6 illustrates a spiral-shaped blue hue seen in the deep aquifers (DCA1-3), indicating severe aquifer depletion exceeding natural recharge. Additionally, pumping activities alter the groundwater flow, leading to a steeper hydraulic gradient in the depleted region.

4.2. Scenario Analysis for the Impact of Varying Recharge Input on Hydraulic Heads

The results of the simulated hydraulic heads for the three distinct recharge scenarios (A–C) during the period from 2000 to 2008 are presented in Figure 6. However, results along the northeastern and southern boundaries were excluded due to the uncertainty in the simulated heads, which may have been influenced by the no-flow boundary conditions at the boundaries of the basin. These no-flow boundaries restrict lateral groundwater movement, causing accumulation in adjacent cells, resulting in unrealistic hydraulic head values.

Across all layers and scenarios, the simulated hydraulic heads generally followed the topography of the basin, with lower heads in the central region and higher heads in the northern and southern parts of the regions. This distribution creates a hydraulic gradient that drives groundwater flow from the north and south towards the midwestern part of the basin. The pattern of simulated hydraulic heads remained consistent across all layers and scenarios, aligning with previous findings in the basin by Pathak (2015) [30].

Scenario A produced the highest hydraulic heads in all the aquifers, with maximum hydraulic heads of 188 m, 173 m, 145 m, 114 m, and 105 m, respectively, for all five aquifers, from UCA to DCA-3 (Figure 6a). RMSE of around 5.8 m was detected when compared with observed heads, exceeding the values of the other two scenarios (Scenario B and C: 3.97 m), indicating an overestimation of hydraulic heads in the forested areas of the basin and an underestimation in the urban areas with impervious surfaces. In contrast, scenario B, which accounted for the spatial and temporal distribution of recharge, produced a more accurate simulation of hydraulic heads with a root mean square error of 3.9 m when compared to observed values (Figure 6b). The average hydraulic head difference between scenarios A and B (A minus B) is about 3.6 m, with a minimum of −13 m and a maximum of 30 m at the eastern and northern boundaries of the basin, as shown in Figure 7A. Overestimation of hydraulic heads in scenario A is prevalent generally over the basin because of the variation in land use types over the basin, except for the eastern side, where underestimated heads could be observed. The forested land use type shows the greatest head difference, around 13 m, attributable to high recharge in forested areas resulting from macropores formed by broad roots, which renders the uniform recharge strategy inadequate.

Pixels with permeable land use classifications, such as paddy fields, croplands, and water bodies, exhibited the lowest overestimated mean head difference of 1–3 m, as these areas serve as recharge zones with high recharge rates. In contrast, pixels characterized by impermeable land use types, including urban areas, public spaces, recreational zones, roadways, and barren ground (rocks), exhibited greater overestimated mean head difference of 3 to 9 m due to minimal or nonexistent recharge, which was overestimated by scenario A.

The underestimated region in the eastern part of the basin, a paddy field subbasin at the base of Mt. Kasuga may receive additional recharge via flow through fractures and faults which are factors not captured in the uniform recharge scenario. Additionally, the simulated average head difference between the two scenarios increases by about 5 m from the UCA to the DCA in the deepest aquifer, as deep aquifer recharge in the Nara Basin is lower than that in the unconfined and shallow aquifers, which scenario A overestimates. The uniform recharge scenario both overestimates and underestimates the hydraulic heads and underscores the necessity of considering spatial and temporal variations of recharge input for accurate groundwater flow simulation.

The spatially distributed scenario C (with no temporal change), which accounted for the 9 years averages of recharge over the basin produced a simulated heads distribution similar to scenario B. To evaluate the impact of temporal variations of recharge, the head difference between scenario B and scenario C was calculated. The results, as shown in Figure 7B, indicate a mean hydraulic head difference of 0.4 m across the basin, with a maximum and minimum head difference of 0.5 m and −8 m, respectively, in the top aquifer; however, the head difference becomes negligible transitioning from the top aquifer to the deep aquifers. The maximum head difference (underestimated head) is prevalent in the forested pixels which have been mentioned earlier to have been influenced by forest growth that occurred over the decades, causing evapotranspiration increase and the significant temporal variation in recharge which was not accounted for in scenario C. Other land use types experience a mean head difference ranging from −1.2 to −0.2 m, due to a less significant temporal shift over the decade. Hence, spatial variability in recharge has a greater influence on simulated hydraulic heads than temporal variability. The result obtained from this study also aligned with transient groundwater flow research conducted on the Grand Forks unconfined aquifer, which revealed a greater effect of the spatially distributed recharge scenario than the temporal distribution effect when compared with the constant scenario [17]. Similarly, Awan et al. found that integrating hydrological models, water balance models, and remote sensing techniques improved recharge estimates by taking spatial variability into account [48]. However, in karst regions, the effect of climate variability is more pronounced in regions characterized by subsurface heterogeneity compared to regions with subsurface homogeneity [49].

Therefore, groundwater models are powerful tools that could either reproduce the true state of groundwater resources in a basin or an unrealistic overly or underly estimated state. This underscores the importance of integrating spatiotemporal variability in input datasets when using numerical models to assess groundwater systems.

4.3. Impact of Recharge Input on Surface Water and Groundwater (SW-GW) Interactions

The interaction between the river and aquifer across three distinct scenarios has been investigated by estimating the volume of water that flows from the river to the aquifer and the groundwater discharge into the river. The results, as shown in Figure 8, reveal that groundwater discharge from the unconfined aquifer is significantly greater than river seepage into the aquifer in all scenarios with groundwater discharge volumes of 1400–2200 million cubic meters (MCM), while the river seepage ranges from 469–545 MCM. Recharge of the intermontane aquifer is primarily by precipitation, as shown in Figure 8, where rainwater recharge exceeds the river seepage into the aquifer. Both river seepage and groundwater discharge to the river are higher in scenario A compared to scenarios B and C, due to the overestimation of recharge in the uniform recharge scenario. In comparison with scenarios B and C, the rainwater recharge was overestimated by about 55%, resulting in a 48% and 16% increase in groundwater discharge to the river and river seepage to the aquifer, respectively in scenario A. Moreover, there is little to no difference between the spatiotemporally distributed and spatially distributed scenarios. Therefore, it can be concluded that spatial variability in recharge has a greater impact on the surface -water groundwater interactions than temporal variability.

4.4. Effect of Climate Variability and Land Use Change on Hydraulic Heads and SW-GW Interactions

Hydraulic heads of the basin were simulated using spatiotemporally distributed recharge for two distinct periods: the historical climate (PreUP) and the contemporary climate (PostUP) with substantial land use alterations. Urbanization plays a critical role in the decline of hydraulic heads due to the increasing impervious surfaces, which impede groundwater recharge. Nara Basin experienced tremendous urbanization between 1970 and 2010, with a 20% increase in impervious surface coverage, resulting in a significant decline in the ratio of groundwater recharge to precipitation [10]. The average decrease in hydraulic heads (PreUP minus PostUP) over the basin in all the layers from unconfined to deep confined aquifers are 5, 4.3, 4.2, 4.1, and 4 m, respectively (Figure 9a). Slight changes in hydraulic head are observed in the Nara plain near the Yamato River, while substantial declines occur in the forested mountain regions. In UCA and SCA, a maximum head difference of about 15 m was observed, as shown in Figure 9a,b. Figure 9a displays all pixels across the basin, while Figure 9b illustrates pixels that retained the same land use type. These changes are concentrated in the forested areas along the eastern, northern and western boundaries of the basin (Figure 3). The decline is likely driven by increasing evapotranspiration over the Ikoma, Sakurai, Kongo and Kasuga forested areas between 1980 and 2010 [19]. Similar findings were reported by Toure et al. [50] in the Kiela basin, Southern Mali, using Global Climate Models and MODFLOW revealed an average hydraulic head drop of 15 m from the past to future climate scenarios (1970–2010 vs. 2010–2050) due to reduced groundwater recharge. Between the two different periods (PreUP: 1980–1988 and PostUP: 2000–2008), the cumulative volume of groundwater discharge to the river decreased by 27%, while the river seepage into the aquifer increased by 16%. This shift reflects the severe lowering of hydraulic heads along certain river sections, primarily driven by reduced recharge due to increased evaporation and urban expansion, which in turn enhanced river seepage into the aquifers.

Furthermore, analysis of the pixels that maintained the same land use between the two simulated periods was conducted to examine the stand-alone impact of climate change over the basin (Figure 9b). The premise of this analysis is that, since there was no land use change in this pixel, the only factor affecting groundwater levels must be the alteration in recharge due to climate variability. The result reveals an average simulated head decline of about 3.3 m in the UCA, 3 m in the SCA, and declines of 2.9 m in the DCA1-3 (Figure 9b). Therefore, the impact of reduced recharge resulting from climate variability is more significant in the UCA and SCA compared to the DCA, particularly in the impervious land use types with a mean decline in hydraulic heads between 2.25 and 3.61 m, while the pervious land use types experienced just a slight decline in hydraulic head with values ranging from 1.3 to 1.7 m. The impervious surfaces, including urban zones, public spaces, and barren terrain (rocks), intensify the influence of climatic variability on water availability for recharge by preventing infiltration, whereas the pervious layers are only affected by climate variability. Deep aquifers are more affected by groundwater over abstraction, unlike the unconfined and shallow aquifers which are highly influenced by climate variability, including changes in temperature and precipitation [51,52].

To assess the combined influence of urbanization and climate change on hydraulic heads, we analyzed the pixels that experienced a change in land use to urban type. The average decline in simulated heads was about 0.5 m in the UCA and SCA and 0.4 m in the DCA. However, among the various land use types that changed to urban between PreUP and PostUP, cropland, orchard, public-area and forest land use had the greatest mean head difference of 4 m, 3.7 m, 3.7 m, and 5 m respectively in UCA and 3.4 m, 3.7 m, 4.3 m and 5.1 m respectively in SCA. The mean hydraulic head difference is greatest when forested land use types, which are known for their high infiltration and percolation rates due to macropores created by root channels and organic matter, change to urban land use. This transition implies a shift from permeable macropores to impermeable concrete surfaces, significantly affecting groundwater recharge and hydraulic heads. Besides, cultivated lands and public areas exhibit significant declines in hydraulic heads due to urbanization. The former is attributed to irrigation percolation and recharge via drains and canals, while the latter results from parks and public gardens with permeable surfaces that facilitate groundwater recharge, which are all eradicated due to urban expansion, thereby negatively influencing the groundwater resources. Typically, in Japan, orchards and rice paddy are often cultivated at the bases of hills, which are highly permeable areas due to the dominance of permeable soils formed by terrace deposits, gravels and sands. Therefore, the expansion of urban areas to these regions reduces natural infiltration, thereby posing a serious threat to groundwater resources.

Additionally, climate variabilities including decreased precipitation by 1.5%, increased temperature and solar radiation by 8% and 3.8%, respectively, play a crucial role in increasing the rate of evapotranspiration over the basin, thereby reducing the amount of recharge to shallow and deep aquifers. The paddy fields, which are often inundated or flooded during the planting season also witnessed a 1.5 m and 2 m drop in hydraulic heads in UCA and SCA, respectively, after urbanization. The slight hydraulic head decline due to urbanization from paddy fields relative to the croplands and forests is attributable to the partial urbanization observed in the paddy field regions, as illustrated in Figure 3b, where urban and paddy field pixels are interwoven. This indicates that the paddy field pixels that transitioned to urban areas between PreUP and PostUP are limited and are still affected by the adjacent permeable land use pixels surrounding them.

Therefore, urbanization and climate change both influence the hydraulic head changes in all aquifers, with the latter having a more significant impact. Shallow aquifers are more sensitive to urbanization and climate variability, while deep aquifers are more influenced by over-abstraction. Previous research conducted by Pulido-Velazquez et al. [18] deduced that climate change appears to be the primary factor influencing alterations in groundwater recharge, as the variations among climate scenarios with identical land uses exceed those caused by the land use scenarios. In contrast, according to a three-decade study of the Kurose River Catchment (1980s–2000s), the main factors driving hydrological modifications were urbanization, particularly the conversion of paddy fields into built-up areas, which led to a significant 34.9% decrease in groundwater recharge [22]. This research, compared with the previous studies, evaluated the overall effect of climate variability and land use change on hydraulic heads and surface and groundwater interactions for a multiple aquifers intermontane basin.

4.5. Limitations and Recommendations

The main sources of uncertainty in the calibrated model are the limited hydrological and hydrogeological datasets available for this research. Particularly, the lack of upstream river stage measurements for the simulated period, and limited information about the river conductivities in the Nara Basin. This limitation was addressed through inverse modelling utilizing the observed groundwater heads available both upstream and downstream. The impact of agricultural activity was not evaluated in this study due to limited groundwater abstraction data for irrigation [53].

The no-flux boundary conditions along the model borders assigned based on the surface geological map of the study area could also be the reason for significant groundwater head changes between the past climate and simulated period in the forested areas. Japanese mountains are mostly covered by forest which can potentially influence the shallower aquifer through the hydrological processes (e.g., groundwater recharge, evapotranspiration); therefore, the mountain areas are not completely a no-flow boundary, and shallower layers should be considered in the simulation.

In addition, future climate scenarios under various Shared Socioeconomic Pathways (SSPs) were not considered in this study.

Therefore, subsequent research should incorporate the effects of future climate change and anthropogenic activities, including urbanization and agriculture on groundwater environments at the basin scale.

5. Conclusions

This research analyzed the impacts of urbanization and climate variability on hydraulic heads applying a three-dimensional groundwater flow model.

In evaluating the impact of groundwater recharge inputs on our groundwater flow model, three scenarios were analyzed. The findings highlight the relevance of spatial distribution of recharge in comparison to temporal distribution for both hydraulic heads and SW-GW interactions. The constant recharge scenario both overestimates and underestimates groundwater heads, emphasizing the need to account for spatial variations in recharge input for groundwater flow simulation.

The results indicate a more significant decrease in groundwater heads in the unconfined aquifer, attributed to the combined effects of urbanization (increase in impervious surfaces) and climate variability. The impervious land use types experienced more decline in hydraulic heads than the permeable areas under changing climate because of the impedance to infiltration and percolation exacerbating the climate variability effect. The forested areas stand as an exception because of the increased evapotranspiration due to forest growth, and the economically influenced anthropogenic activities, resulting in a significant decline in hydraulic heads between the PreUP and PostUP. In addition, the impact of urbanization and climate variability on hydraulic heads in previously forested and pervious areas (especially croplands, and orchards cultivated on hilly terrains) is significantly higher than in impervious areas because of their high recharge potential. In contrast, declines in the heads of deep confined aquifers are primarily influenced by climatic variations over several years. These hydraulic head declines were notably evident in the SW-GW interactions, resulting in more recharge to the aquifer from the river. Therefore, the expansion of urban areas, coupled with climate variability, has a significant impact on groundwater resources and presents a dual threat to groundwater sustainability. Urban infrastructure constrains recharge potential, whereas climate variability diminishes the dependability of natural replenishment. Our findings were focused on the changes in hydraulic heads caused by urbanization and climate variabilities; however, future research should incorporate other anthropogenic activities, including groundwater over-abstraction and future climate change scenarios on groundwater heads that are influenced by societal transformations.