Assessing the Impact of Urbanization and Climate Change on Hydrological Processes in a Suburban Catchment

Abstract

Global urbanization, population growth, and climate change have considerably impacted water resources, making sustainable water resource management (WRM) essential. Understanding the changes in hydrological components is important for effective WRM, particularly in cities such as Higashi-Hiroshima, which is known for its saké brewing industry. This study used the Soil and Water Assessment Tool (SWAT) with Hydrological Response Units (HRUs) to achieve high spatial precision in assessing the impacts of land use change and climate variability on hydrological components in a suburban catchment in western Japan. Over the 30-year study period (1980s–2000s), land use change was the main driver of hydrological variability, whereas climate change played a minor role. Increased surface runoff, along with decrease in groundwater recharge, evapotranspiration, and baseflow, resulted in an overall reduction in water yield, with a 34.9% decrease in groundwater recharge attributed to the transformation of paddy fields into residential areas. Sustainable WRM practices, including water conservation, recharge zone protection, and green infrastructure, are recommended to balance urban development with water sustainability. These findings offer valuable insights into the strategies for managing water resources in rapidly urbanizing regions worldwide, emphasizing the need for an integrated WRM system that considers both land use and climate change impacts.

1. Introduction

Global urbanization projections indicate that by 2050, 60 to 90% of the global population will reside in cities [1,2]. This rapid surge in urbanization driven by population growth and socioeconomic conditions has heightened the demand for water, thereby straining both surface and subsurface water resources [3,4,5]. Anthropogenic activities, including land use and climate change, are critical factors affecting catchment hydrological processes [6,7]. Urbanization is a key driver of land use transformation, considerably modifying the topography and vegetation cover by altering the physical structure of land surfaces [8,9,10]. These changes increase impervious surfaces, reduce infiltration capacity, and alter water movement, leading to variations in hydrological balance parameters, such as groundwater recharge, runoff, and evapotranspiration [11,12,13]. Additionally, while global warming-induced changes in precipitation and temperature patterns can be predicted, the uncertainty surrounding these predictions complicates the accurate modeling of their effects on individual hydrological components of catchments [14,15]. Understanding these effects requires hydrological models that simulate changes in hydrological components over time and space and quantify these changes [16].

Models are widely used to evaluate hydrological processes in watersheds and examine the effects of land use change, climate variations, and urbanization on catchment hydrology. Numerous studies have employed different models to investigate the hydrological impacts of climate and land use change, as discussed by Xiao et al. [17]. Among these models, the Soil and Water Assessment Tool (SWAT) is the most widely used one [17,18]. While these studies demonstrated that both land use and climate change can considerably affect hydrological parameters, the relative influence of each factor varied considerably across different spatial scales. Some studies emphasize the greater impact of climate change [15,19,20,21,22], whereas others highlight the dominant role of land use change [2,23,24,25]. Given these contrasting observations, investigating the effects of urbanization-induced climate change and land use on catchment hydrology at different spatial resolutions is crucial for developing effective water resource management (WRM) policies.

Globally, groundwater accounts for 97% of the freshwater and is essential for the public water supply, particularly in areas with unpredictable rainfall and limited surface water resources [26]. However, urbanization and industrial activities are increasingly placing pressure on groundwater systems, thereby affecting both their quantity and quality. In Japan, rapid urbanization and industrialization have heightened environmental pressures on groundwater both regionally and locally [27]. Higashi-Hiroshima, located in western Japan and renowned for its saké (traditional Japanese wine) brewing industry, is experiencing rapid urban growth [28]. This has led to notable changes in land use owing to the increase in population (Figure S1); various development projects associated with residential and industrial expansion place further stress on groundwater resources [28]. Previous research has established the effectiveness of the SWAT model in examining the impact of environmental factors on hydrological processes in Japanese catchments [2,10,29,30,31,32,33]. These studies predominantly focused on large catchments; however, small catchments, due to their limited buffering capacity, steep slopes, and relatively short hydrological response times, are more prone to experiencing dramatic changes in their water balance due to a higher percentage of the catchment being altered [29]. Consequently, it is important to conduct research on the impacts of urbanization and climate change on small catchments with high spatial precision. Hydrological response units (HRUs) in the SWAT model represent the smallest units for calculating hydrological characteristics, offering a more precise representation of the effects of land use and climate change on all hydrological components, particularly in smaller catchments. Visualizing spatial variability at the HRU scale provides detailed insights into localized impacts, enabling targeted management practices and a more accurate calibration of hydrological models. This level of detail is crucial for understanding the specific effects of land use and climate change, which might be overlooked when analyzed at broader catchment or sub-basin scales.

Therefore, this study applied the SWAT model, utilizing HRU for high spatial precision, to evaluate the ipacts of land use and climate change on hydrological components in a suburban catchment in western Japan. The specific objectives were to (i) model and assess the historical impacts of land use change (LUC) on the hydrological response of the catchment over the past three decades (1980s–2000s); (ii) evaluate the spatial variations in hydrological components due to land use and climate change; (iii) provide practical recommendations for sustainable groundwater management based on the observed impacts. This study has valuable implications for hydrological management in other rapidly urbanizing regions facing similar environmental challenges.

2. Materials and Methods

2.1. Study Area

The Kurose River Catchment (KRC), covering an area of approximately 39.9 km² and featuring a main river channel that extends approximately 14.3 km in length, is situated in Higashi-Hiroshima City within Hiroshima Prefecture in western Japan, between the coordinates 34°25′30″ N latitude and 132°44′33″ E longitude (Figure 1).

This area features an elevation ranging from 204 to 642 m above sea level. The catchment climate is generally warm and temperate, with significant rainfall. However, the temperature and rainfall patterns have changed since the 1970s owing to the changes in climatic conditions [34]. The catchment records an average annual rainfall of 1467 mm, with 78% of the total precipitation occurring during July and August. The average annual temperature in the catchment is 13.7 °C. On average, the temperatures are highest in August (25.8 °C) and lowest in January at approximately 2.3 °C [34]. The average annual temperature of the catchment increased by 0.4 °C, while its precipitation reduced by 67.9 mm between 1979 and 2010, as shown in Figure S2.

Owing to the presence of abundant groundwater in the study area, nine saké factories are located there. As approximately 80% of saké production depends on the availability of water for rice cultivation [35], groundwater resources in this area are crucial for the region’s sake industry and local development. From 1980 to 2010, the population of Higashi-Hiroshima City increased from 74,235 to 129,744, representing an increase of approximately 75% (Figure S1) [28]. Land use maps from 1987 and 2009 were used to develop two simulation scenarios: one representing the 1980s (covering the years 1983–1993) and the other representing the 2000s (covering the years 2000–2007). These maps captured key land cover changes in the catchment (Figure 2 and Table S1). In the 1980s, the catchment had more than 18.6 km² of forest cover, accounting for approximately 47.3% of the catchment area. Rice paddies covered 13.6 km², accounting for 34.7%, whereas residential land was limited to 5 km², representing 12.8% of the catchment area. By the 2000s, residential areas had expanded to 13.6 km², accounting for 34.5% of the catchment area, with an urbanization rate of approximately 7.4% per decade. Conversely, areas covered by paddy fields and forests decreased by 14.8% and 4.7%, respectively (Figure 2). This approach allows the assessment of historical land use changes and their impacts on hydrological processes. The relocation of Hiroshima University to Saijo town in the late 1970s likely contributed significantly to the accelerated urban growth in this region compared with that in other areas of Higashi-Hiroshima (Figure S1).

2.2. Soil and Water Assessment Tool (SWAT)

SWAT, a hydrological modeling tool developed by the USDA Agricultural Research Service, was designed to assess the effects of various land management practices on water resources, sediment, and agricultural outputs in diverse and multifaceted watersheds [36]. This model effectively handled different soils, land uses, and management strategies across extended timeframes, making it highly suitable for analyzing large and small watersheds. The SWAT model simulates hydrological processes in a watershed by dividing the catchment into sub-catchments (Figure S3), which are further subdivided into HRUs. Each HRU represents a unique combination of land use, soil type, and slope, and the water balance is calculated for each HRU during the land phase of the hydrological cycle [36]. The main hydrological components modeled during this phase included canopy interception, infiltration, evapotranspiration, lateral subsurface flow, surface runoff, groundwater recharge, and stream flow. This study focused on the land phase water balance, where land use and climate change exert the greatest influence. The water balance equation that governs the land phase of the SWAT hydrological cycle in a time interval from day 1 to day t is expressed by [37]:

$${SW}_t={SW}_o+{Ʃ}_{i=1}^t\left(R_{day}-Q_{surf}-E_a-W_{seep}-Q_{gw}\right)$$

(1)

where SW_t is the soil water content at the end of the t-th day (mm), SW_o is the initial soil water content at the beginning of the first day of the considered time interval (mm), t is the time in days, R_day is the daily precipitation (mm), Q_surf is the daily surface runoff (mm), E_a is the daily evaporation (mm), W_seep is the daily amount of water entering the vadose zone from the soil profile (mm), and Q_gw is the daily groundwater flow (mm). The modified curve number method of the Soil Conservation Service (SCS) was used to estimate surface runoff; however, to calculate evapotranspiration from the three alternative methods provided by SWAT, the Penman–Monteith method was used based on available data. Although the SWAT simulates surface processes, its ability to simulate groundwater flow is limited because it is aggregated or lumped [38]. Groundwater was recharged from the vadose zone by percolation to the bottom of the soil profile. In SWAT, groundwater is partitioned into two aquifer systems: a shallow, unconfined aquifer that contributes to return flow to streams within the watershed and a deep, confined aquifer that does not contribute anything to streams [37]. Therefore, SWAT simulates percolation at the HRU scale, mapped as groundwater recharge [38], using Equation (2), in which the model can further apportion part of the recharge to deep aquifers [37].

$${\omega }_{rchrg,i}=\left(1-exp\left[{\displaystyle \raisebox{1ex}{$-1$}\!\left/ \!\raisebox{-1ex}{${\delta }_{gw}$}\right.}\right]\right){\omega }_{seep}+exp\left[{\displaystyle \raisebox{1ex}{$-1$}\!\left/ \!\raisebox{-1ex}{${\delta }_{gw}$}\right.}\right]{\omega }_{rchrg,i-1}$$

(2)

Here, ω_rchrg,i is the amount of water entering the aquifer on day i (mm), δ_gw (days) is the lapse period, which refers to the time delay (in days) that groundwater takes to move through the geologic material before reaching the aquifer, ω_seep represents the amount of water exiting the bottom of the soil profile on day i, and ω_rchrg,i−1 is the amount of recharge entering the aquifer on day_i−1 (mm). Further details of the water balance parameters can be found in the SWAT theoretical document [37].

2.3. Input Data and Model Set-Up

The Kurose River Catchment (KRC) was modeled using the ArcSWAT_64 interface in ArcMap 10.6, with spatial datasets obtained from open-source websites (Table S2). For modeling purposes, data from 1987 and 2009 were used to represent the 1980s and 2000s, respectively (Figure 2). The primary land use categories identified were forests, paddy fields, and residential areas. It was assumed that residential areas included pervious spaces such as house yards, leading to the classification of residential parameters as urban residential medium density (URMD) in the SWAT model (Table S1) [39]. According to Wang et al. [39], URMD classification assumes an average impervious area of 38% (15 km²), which matches the settlement conditions of the study area. The catchment was delineated using a 30 m resolution Digital Elevation Model and subdivided into seven sub-basins. These sub-basins were further divided into 223 HRUs for the 1980s and 279 HRUs for the 2000s (Figure S3), allowing for spatial variability in land use, soil, and slope to be captured, providing more accurate hydrological simulations. The HRUs were defined with 0% thresholds for land use, soil, and slope to ensure full catchment coverage in the water balance calculation. The HRUs were connected to the SWAT weather database sourced from a meteorological station located approximately 1 km southwest of the study area (Figure 1). Meteorological data, including precipitation, maximum and minimum temperatures, wind speed, solar radiation, and relative humidity from 1980 to 2007, were used to represent the two historical periods. Although the meteorological data were sourced from a single weather station, the spatial variability in land use, topography, and soil types were captured using the SWAT model HRUs. The HRU framework divides a catchment into smaller units with homogeneous characteristics, allowing the model to account for local variability in hydrological responses.

Although the meteorological data were sourced from a single weather station, we recognize the limitations of this study. Although this station provides the best available data, the temperature and precipitation data, particularly temperature data, might not fully represent the variability within the catchment, especially given the presence of two hills, as shown in Figure 1. These features could contribute to the microclimatic variations that are not captured by a single weather station. We have included this data source for practicality, but we believe that future studies could benefit from utilizing gridded climate data or multiple stations to better account for spatial variability in temperature. For a catchment of this size (39.9 km²) and given the simplicity of the overall topography, using data from a single station should provide a reasonable ap-proximation of general climatic conditions across the study area. However, we caution that spatial variability in temperature could introduce minor uncertainties into our model results, particularly for simulations sensitive to temperature fluctuations. During this period, the daily river discharge along the Kurose River was continuously monitored using data loggers. However, some discharge data from the 2000s are missing during the simulation period. The lack of localized meteorological and irrigation data may limit the accuracy of the model during calibration. To address this issue, we employed the automatic irrigation technique of the SWAT model was employed, which simulates irrigation based on crop water demand and availability [37]. This feature automatically calculates and applies irrigation to rice and agricultural fields from reservoirs and channels to meet the water requirements of the crops, thereby compensating for the absence of precise irrigation data.

2.4. SWAT Calibration, Validation, and Model Evaluation

The parameters for the SWAT calibration were determined for both the 1980s and the 2000s based on the methodology outlined by Abbaspour et al. [40]. The intervals for calibration and validation were selected based on the availability and completeness of the observed streamflow data, ensuring sufficient variation in hydrological conditions for robust model calibration. For the 1980s, seven years (1983–1989) were used for calibration, and four years (1990–1993) were used for validation. For the 2000s, five years (2000–2004) were used for calibration, and three years (2005–2007) were used for validation. These intervals were chosen to ensure that both the calibration and validation periods captured a range of hydrological conditions, allowing for more reliable model performance across different timeframes. For each model run, the initial three years of simulations were treated as warm-up periods and excluded from both the calibration and validation phases. Sensitivity analysis, along with the calibration and validation processes, was performed daily using Sequential Uncertainty Fitting 2 (SUFI-2) in SWAT-CUP, which automatically minimizes errors while providing insights into parameter uncertainty, enabling a more objective calibration process [39].

Parameter adjustments were applied to minimize the uncertainty and improve the accuracy of simulations. The initial parameter ranges were defined using the minimum and maximum values specified in the SWAT user manual and further refined through optimization using the SUFI-2 algorithm [40]. SUFI-2 employs a stochastic approach to minimize errors and assess parameter uncertainty, with each iteration consisting of 1000 simulations. The parameter ranges were progressively narrowed based on the best-fitting solutions until a satisfactory agreement between the measured and simulated streamflow was achieved. Calibration was conducted automatically, and manual adjustments were made only when necessary to avoid the extreme pursuit of mathematical indicators, which can sometimes occur in automatic calibration, potentially ignoring the physical meaning of actual hydrological processes. Calibration concluded that a set of parameter ranges consistently yielded fitted parameter values. If the calibration results were unsatisfactory, parameter ranges were reselected (Table S3). The accuracy of the simulation was verified through validation, during which the model was executed using calibrated parameters without additional modifications, and the results were evaluated against the remaining observational data. Similar parameter values were applied for the 1980s and 2000s to ensure consistency between the calibration phases. This approach was adopted to allow for direct comparisons of hydrological responses under different land use conditions, minimizing the impact of varying parameter ranges on model performance [2]. Special attention was paid to parameters related to surface runoff, groundwater flow, and evapotranspiration, considering the observed urbanization trend in the catchment over the past 30 years. Improper calibration can result in poor forecasting performance, particularly when predicting peak flows during extreme rainfall events. Nonetheless, certain parameters, such as the SCS runoff curve number (CN2) and soil evaporation compensation factor (ESCO), exhibited different responses during the calibration process in both simulation scenarios. Specific land use classes were assigned different parameter ranges to account for the substantial land use changes between 1987 and 2009, particularly the conversion of paddy fields into residential areas. These changes influenced how the catchment responded to hydrological processes such as runoff and evaporation. The ranges used for these parameters are detailed in Table S3, which shows the optimization ranges based on land use types (rice paddies, agricultural land, residential areas, and forest). These land-use-specific ranges reflect the physical differences in how each type of land cover influences runoff generation and water retention. For instance, the CN2 values for residential areas were higher, reflecting increased impervious surfaces, whereas the CN2 values for forested areas were lower due to higher infiltration rates. Similarly, ESCO values were adjusted to account for the differences in vegetation cover and evaporation potential across land use types.

To judge the goodness of fit of the model calibration and validation performance, two statistical indices, the coefficient of determination (R²) [41] and the Nash–Sutcliffe efficiency (NSE) [42], were calculated, as shown in Equations (3) and (4). In this study, the NSE and R² values were considered adequate based on the size of the dataset, model complexity, and number of calibrated parameters. Further refinement of the thresholds is necessary for larger datasets or more complex models. R² represents the square of the correlation coefficient, which reflects the strength of the relationship between the observed and simulated data, with values ranging from zero to one. A value of 1 indicates perfect alignment between the two datasets, whereas a value of 0 indicates no correlation. However, relying solely on R² can be misleading, as the model can still produce high R² values despite systematic over- or under-prediction. NSE was also used to provide a more accurate evaluation. NSE measures the model’s predictive accuracy, with values ranging from −∞ to 1. An NSE value of 1 is ideal, and models with NSE values greater than 0.5 are typically considered satisfactory [42].

$${\mathrm{R}}^2={\displaystyle \frac{{\sum }_i{\left[\left(X-X^{\prime }\right)\left(Y-Y^{\prime }\right)\right]}^2}{{\sum }_i{\left(X-X^{\prime }\right)}^2{\sum }_i{\left(Y-Y^{\prime }\right)}^2}}$$

(3)

$$\mathrm{N}\mathrm{S}\mathrm{E}=1-[{\displaystyle \frac{{\sum }_{\mathrm{i}}{\left(\mathrm{X}-\mathrm{Y}\right)}^2}{\sum _{\mathrm{i}}{\left(\mathrm{X}-{\mathrm{X}}^{\prime }\right)}^2}}]$$

(4)

where X is the observed data; Y is the SWAT simulation result; X′ and Y′ are the means of the observed and simulated data, respectively; and i is the number of observed and simulated data points, respectively.

2.5. Model Output and Statistical Methods

To assess the effects of land use and climate change on the study catchment, HRUs for major land use types were extracted from the calibrated SWAT model, and statistical analyses were performed on the relevant hydrological components [2]. Based on Equation (5), the rates of change in the selected hydrological components in the catchment were calculated for the land use scenarios of 1987 and 2009 under fixed climatic conditions.

$$R_{i,H}=\left({\displaystyle \frac{H_{i,year2}}{H_{i,year1}}}-1\right)\times 100$$

(5)

where year1 and year2 represent the first and second years associated with the first and second land use scenarios, respectively; R_i,H is the rate of change of the hydrological component H (H = actual evapotranspiration, surface runoff, groundwater recharge, and base flow) in catchment i; H_i, year1 and H_i, year2 represent the amounts of hydrological component H in catchment i in year1 and year2, respectively. Additionally, statistical analysis using linear regression was used to determine the relationship between precipitation and simulated groundwater recharge (Equation (6)):

$$y=b+kx$$

(6)

where y is the simulated groundwater recharge in the HRU (mm), x is the observed precipitation (mm), k is the slope of the line indicating the ratio of groundwater recharge to precipitation, and b is the intercept representing the baseline level of groundwater recharge when precipitation is zero. The linear model in Equation (6) provides a simplified depiction of the relationship between precipitation and groundwater recharge. While it offers useful insights for predictive modeling, it does not account for nonlinear factors such as soil saturation thresholds, evaporation, and land cover variability. Consequently, recharge may not occur consistently during low-rainfall events.

3. Results and Discussion

3.1. SWAT Model Performance, Uncertainty, and Simulation Accuracy

The sensitivity of the parameters was evaluated using both t-statistics and p-values, with lower p-values and higher absolute t-values indicating greater sensitivity. Parameters with p-values < 0.05, such as CN2 and ESCO, were considered highly sensitive. Initially, larger parameter uncertainties were assumed to fit the observed data within the 95% Prediction Uncertainty (95PPU) band [36]. However, after refining the parameter ranges, we evaluated the performance of the fitted parameters using Nash–Sutcliffe Efficiency (NSE) values during the calibration and validation periods (

Table 1. Daily calibration and validation statistics of the SWAT model in the Kurose River catchment.

Period	Period Range	R2	NSE
1980s (calibration)	1983–1989	0.79	0.71
1980s (validation)	1990–1993	0.83	0.78
2000s (calibration)	2000–2004	0.74	0.65
2000s (validation)	2005–2007	0.85	0.74

) [36,42].

Based on the relative sensitivity values during model calibration, 13 parameters related to surface runoff (CN2), groundwater (ALPHA_BF, GW_DELAY, GWQMN, GW_REVAP, and ALPHA_BNK), soil evaporation (ESCO), and soil properties (SOL_AWC, SOL_K, and SOL_BD) were considered relevant for estimating daily streamflow (Table S2). In the KRC, the SCS curve number (CN2) for moisture condition II emerged as the most influential parameter for estimating surface runoff during discharge simulations. CN2 governs the proportion of water that infiltrates the soil or becomes surface runoff through overland flow. Higher CN values indicate reduced infiltration and increased surface runoff [43]. The CN2 values displayed significant variations across different land use types during model calibration. Additionally, the ESCO.hru parameter was found to be highly sensitive to actual evapotranspiration rates.

The model performance for both the 1980s and the 2000s was assessed by comparing the observed and simulated flow discharge with the results presented in Table 1. For the 1980s, the calibration yielded NSE and R² values of 0.71 and 0.79, respectively, indicating good model performance [42]. In the 2000s, the calibration also produced satisfactory results, with NSE and R² values of 0.65 and 0.74, respectively. During the validation phase, both metrics further improved and remained within acceptable ranges (Table 1), thereby confirming the accuracy and reliability of the model. The similarity between the calibration and validation NSE and R² values suggests that the model residuals are closely aligned with the uncertainties inherent in forecasting. This indicates that the model effectively captured the primary hydrological processes influencing the catchment. According to Moriasi et al. [42], NSE values greater than 0.5 and R² values above 0.6 are considered satisfactory, confirming the reliability of the model. However, some uncertainties remain, particularly during peak events, when the model overestimates or underestimates extremes. It is important to note that the thresholds used (NSE > 0.50 and R² > 0.60), as outlined by Moriasi et al. [42], were appropriate considering the dataset size and model complexity. Given the small scale of the KRC (39.9 km²) and the available dataset, these thresholds provide a reasonable standard for assessing model performance. This approach is consistent with those of previous studies [32,44,45]. For example, Ridwansyah et al. [44] used only NSE and R² to evaluate the performance of the SWAT model for the Cimanuk Watershed, Indonesia, which is a semi-distributed catchment and demonstrated that these two statistical indices are sufficient for smaller or moderately complex watersheds. In contrast, studies on larger and more complex catchments, such as Guyo et al. [32], who analyzed the Asahi River Catchment (1634 km²), required additional indices, such as the percentage of bias (PBIAS), to account for the spatial variability in groundwater dynamics across diverse slope gradients. Similarly, Chen and Nakatsugawa [45] employed more advanced indices like Kling–Gupta Efficiency (KGE) for larger and more complex hydrological systems. Although stricter thresholds and additional metrics may be required for larger datasets or more complex models, the criteria used in our study aligned well with the hydrological modeling standards for studies of this nature. The alignment between the calibration and validation results further supported the robustness of the model, although some uncertainties remained, particularly during extreme flow events.

The hydrographs showed good alignment between the observed and simulated streamflow during both the calibration and validation periods (Figure S4), indicating that the model performed well in replicating the hydrological patterns of the KRC. The peak flow discrepancies between the observed and simulated data were minimal and did not significantly affect the overall model performance. In addition, automatic irrigation used in this study may not have fully reflected the actual water budget during crop growth. However, because of the drastic transformation of cultivated land into residential areas in the basin during the 2000s, the potential impact of discrepancies in irrigation data can be considered negligible [39]. Overall, the hydrograph analysis, combined with satisfactory statistical performance metrics, indicated that the model was well calibrated and validated for the KRC, providing accurate streamflow simulations suitable for assessing water availability during the historical periods examined.

3.2. Land Use and Climate Change Effect on Catchment Hydrology

3.2.1. Changes in Hydrological Components at the Catchment Scale

The mean annual precipitation in the catchment area was 1573 mm/y in the 1980s and 1375 mm in the 2000s. Simulated water balance components, such as surface runoff (SurfQ), baseflow (comprising lateral flow (LatQ) and groundwater discharge (GW_Q)), groundwater recharge to the deep aquifer (DA_RCHG), evapotranspiration (ET), and water yield (WYLD) are presented as mean annual values (Figure 3) and expressed as percentages of the average annual rainfall (Table S4) for both simulation periods. These key hydrological parameters have been evaluated over the past three decades to understand the impacts of land use and climate change.

Water yield, which is crucial for water management in the KRC, exhibited a 17.2% decrease in the 2009 land use scenario compared with 1987. Figure 3a shows that the ET component had the largest share of the water balance among the water balance components. Although the visual representation shows an apparent increase in the proportion of ET in the 2000s, this reflects the relative contribution of ET to the total water budget, which decreased due to reduced rainfall. In absolute terms, the ET decreased from 699 mm in the 1980s to 664 mm in the 2000s, representing a 5.1% reduction (Table S4). This decline can be attributed to a 4.7% reduction in forest cover between the two periods (Figure 2), which is consistent with the findings of Kundu et al. [46]. Surface runoff increased significantly, rising from 92 mm in the 1980s to 213 mm in the 2000s, marking a 131.7% increase. In contrast, average annual groundwater recharge and baseflow decreased by 34.9% and 41.4%, respectively (Table S4). Deforestation and urbanization significantly alter the water balance, reduce groundwater recharge, and increase surface runoff, as seen in the Pra River Basin [47]. These changes in hydrological components align with other studies [48,49] that attribute such alterations to rapid urbanization. For example, Wang et al. [48] noted that, in the Xitiaoxi River watershed, urban expansion at the expense of forest cover played a pivotal role in reducing groundwater recharge. Surface runoff, which is highly sensitive to land use change, tends to escalate with urbanization, further reducing groundwater recharge [50,51]. CN2, a significant parameter influencing surface runoff, was identified as a key factor associated with changes in land use [2]. The CN2 values increased from 47.4 in the 1980s to 65.6 in the 2000s, reflecting reduced infiltration rates and heightened overland flows due to land use alterations. This observed increase in surface runoff, especially during intense rainfall events, raises concerns regarding flooding, which poses a potential threat to current and future water resources in the KRC. Moreover, changes in land use patterns and rainfall intensity make catchments susceptible to soil erosion [33,51]. Overall, the combined effects of reduced ET and increased surface runoff due to urbanization lowered groundwater recharge, potentially leading to issues such as reduced baseflow in rivers and decreased water availability for agricultural, industrial, and other uses in catchments.

3.2.2. Changes in Hydrological Components per Land Use

We compared the mean annual water balance components across three primary land use types (paddy fields, forests, and residential areas) during the 1980s and the 2000s. As shown in Figure 3b, the groundwater recharge varied significantly across these land uses. In forested areas, the mean annual recharge decreased from 168 mm in the 1980s to 123 mm in the 2000s. Residential areas, which had higher recharge in the 1980s (202 mm) than forest areas, experienced a significant decline to 88 mm in the 2000s. Paddy fields consistently exhibited the highest recharge rates, decreasing from 214 mm in the 1980s to 180 mm in the 2000s (Figure 3b). The reduction in groundwater recharge within residential areas is attributed to the expansion of residential land, which grew from 12.8% in the 1980s to 34.5% in the 2000s, with more than 45% of paddy fields being converted to residential zones (Figure 2) due to urbanization. Tashie et al. [52] highlighted that natural ecosystem with higher infiltration rates differ from with urban soils, which are more impervious and have limited infiltration. Rice paddies consistently maintained the highest recharge rate, consistent with the findings of previous studies [53,54].

In the 2000s, surface runoff in residential areas increased substantially from 212 mm to 482 mm, contrasting with insignificant changes in rice paddies and forest areas (Figure 3b). This trend aligns with other studies that describe urban areas as impervious, promoting surface runoff while reducing infiltration, which affects hydrological balance and water resource availability [55]. Rice paddies and forests facilitate slow water infiltration through the streamflow when precipitation is not intercepted or evaporated. Vegetation plays a crucial role in enhancing groundwater saturation, reducing runoff, and recharging groundwater systems. Conversely, residential areas with compacted impervious soils have a limited capacity to absorb and retain water, restricting groundwater percolation [52]. Consequently, during storm events in the catchment, surface runoff escalates significantly, increasing the likelihood of flooding due to extensive concrete coverage [56]. In addition, there is an elevated risk of groundwater contamination by nutrients from agricultural runoff in catchments, particularly in areas susceptible to soil erosion [57].

The average annual evapotranspiration (ET) was generally higher in the 1980s than in the 2000s for all land use types, with forests showing the highest ET values compared to paddy fields and residential areas. In the forests, the ET values decreased slightly from 716 mm (46% of the mean annual rainfall) in the 1980s to 706 mm (52%) in the 2000s. Residential areas recorded the lowest ET values, with 603 mm (39%) in the 1980s, falling to 589 mm (43%) in the 2000s. The higher ET ratios in the 2000s suggest that changes in climatic conditions may contribute to the urban heat island effect in the future. Reduced ET values in residential areas may be associated with reduced soil moisture from surface sealing, which limits evaporation. Climate change can significantly affect water resources by altering ET, precipitation patterns, and air temperature, leading to greater spatial and temporal variability in the hydrological cycle [58,59,60].

3.3. Groundwater Recharge and Evapotranspiration under Land Use and Climate Change

3.3.1. Groundwater Recharge Rates in Major Land Use Types

The effective use of groundwater is crucial for shaping urban development, and has significant implications for groundwater ecosystems. According to Bucton et al. [61], precipitation and temperature are the key drivers of the notable changes in groundwater patterns in Cambodia. A comparison of annual precipitation and groundwater recharge across different land use types in the KRC (Figure 4 and

Table 2. Estimated groundwater recharge at an average precipitation of 1400 mm/yr in the KRC across different land use types during both the 1980s and 2000s simulation scenarios.

	Mean Annual (mm)	% Based on Rainfall	Mean Annual (mm)	% Based on Rainfall	Mean Annual (mm)	% Based on Rainfall
Land use	Residential		Forest		Paddy field
1980s	639	45.7	615	44.0	656	46.9
2000s	228	16.4	536	38.3	621	44.4

) highlights significant trends.

Figure 4 shows high R² values (>0.95) for paddy fields and forest areas, while residential areas exhibited more variations, with R² values of 0.99 in the 1980s and 0.87 in the 2000s. Direct recharge at the land surface remained consistent in the paddy fields and forest areas, suggesting similar recharge characteristics. However, the stark difference in gradients between the 1980s and the 2000s in residential areas suggests that non-precipitation factors influenced recharge. The intensification of urbanization in the 2000s resulted in a sharp decline in recharge ratios from 0.66 to 0.23. Precipitation contributions to recharge differed across land uses, with paddy fields consistently contributing approximately 45%, forest areas approximately 38%, and a significant reduction in residential areas from 45.7% in the 1980s to 16.7% in the 2000s (Table 2). Paddy fields have the highest average annual groundwater recharge, underscoring their essential role in sustaining groundwater resources. Overall, increased precipitation was correlated with a heightened potential for groundwater recharge in the KRC. Nevertheless, the swift pace of urbanization, coupled with the effects of climate change, has significantly disrupted the groundwater quantity in this region.

3.3.2. Spatial Variations in Groundwater Recharge at the HRU Scale

Visualizing the spatial variation in hydrological components at the HRU scale offers significant insights into the impacts of land use changes on catchment hydrology over time. Notable spatial differences in groundwater have been observed in the study area over the past 30 years (Figure 5). The average annual groundwater recharge was calculated to be 688 mm (50% of the mean annual rainfall) during the 1980s, decreasing to 415 mm (40%) in the 2000s. In the 1980s (Figure 5a), the central and southern regions of the catchment exhibited higher recharge rates (700–900 mm/year), primarily because of the presence of extensive rice paddies (Figure 2).

The average annual groundwater recharge was calculated to be 688 mm (50% of the mean annual rainfall) during the 1980s, decreasing to 415 mm (40%) in the 2000s. In the 1980s (Figure 5a), the central and southern regions of the catchment exhibited higher recharge rates (700–900 mm/year), primarily because of the presence of extensive rice paddies (Figure 2). These areas contain permeable alluvial soils, facilitating infiltration and supporting elevated groundwater recharge. In the 2000s, a noticeable decrease in recharge values was observed, particularly in areas with previously showing high values (Figure 5b). Urbanization has led to an increase in impervious surfaces, which reduces infiltration and subsequently lowers groundwater recharge. From the 1980s to the 2000s, the catchment experienced more than 80% reduction in groundwater recharge (Figure 5c), underscoring the impact of urbanization on groundwater resources in the KRC. The observed spatial variation in groundwater recharge in the KRC can be attributed to both climatic changes and significant land use alterations over the past three decades. As shown in Figure S2, the significant rise in temperature and decline in rainfall partially explain the substantial variation in the hydrological components observed in the study area. Reduced rainfall directly affects the amount of water available for recharge, whereas increased temperatures can enhance evapotranspiration, further reducing the water available for infiltration. Thus, in addition to climatic influences, land use change has critically altered the hydrological regime of the KRC. The conversion of permeable agricultural land to impermeable urban areas significantly reduces groundwater recharge by limiting rainwater infiltration, highlighting the profound impact of land use change on groundwater resources. In contrast, Minnig et al. [62] observed that urbanized watersheds may sometimes exhibit higher groundwater recharge than rural watersheds owing to reduced evapotranspiration and increased leakage from wastewater systems. Factors such as geology, topography, and soil also influence the groundwater recharge variability in the study area.

3.3.3. Spatial Variations in ET at the HRU Scale (Relative Urbanization Effect)

Trees play a significant role in regulating local climate and hydrological processes through the ET mechanism. Figure 6 and Figure 7 illustrate the spatial distribution and changes in ET for the entire catchment, specifically for forested areas, at the HRU scale.

Forests can enhance groundwater recharge by increasing infiltration rates and reducing surface runoff [53,63]. However, they may also decrease recharge by elevating ET, a process influenced by tree age and temperature [10], and influence water loss by interception, which limits water availability for groundwater recharge [53]. As shown in Figure 6a, during the 1980s, higher ET values were observed in the central and southern parts of the catchment, which are dominated by forest vegetation and agricultural activities. However, a noticeable decrease in ET was observed in many areas of the catchment during the 2000s (Figure 6b) compared to the 1980s. This reduction is particularly evident in areas that have undergone significant urbanization, where vegetation and agricultural land have been converted to residential and commercial zones, as shown in Figure 2. In response to this reduction in catchment ET, more water might be runoff, as less water is returned to the atmosphere, further reducing the amount of water available in the catchment. The spatial variation in forest ET (Figure 7) followed a trend similar to that of the catchment ET.

The noticeable shift in ET during the 2000s, with some areas showing increased ET, was likely due to forest growth and maturity. The relative increase in the ET (Figure 7c) in some forested areas suggests that mature trees have higher transpiration rates, which can reduce the amount of water available for groundwater recharge [10]. However, forests also enhance infiltration rates, which can mitigate this effect to some extent through transpiration, cooling the environment by taking up water through roots, moving it through the stem, and evaporating it through the leaf stomata, thereby mitigating the urban heat island effect [64]. With more than 80% of the catchment area showing a decrease in forest ET values during the 2000s, the reduction in vegetative cover diminished the cooling effect of ET. A comparison with the Zuli River Basin (ZRB) by Liu et al. [65] showed similar ET patterns, with forests maintaining higher ET levels than residential areas in both catchments. In the ZRB, afforestation increased ET, whereas urbanization reduced it. Similarly, in the KRC, forest areas had higher ET (46–52% of rainfall), and residential areas showed lower ET due to surface sealing. Both studies underscore that urbanization consistently reduces ET, while forests help maintain higher ET rates, emphasizing the importance of preserving forest cover to support the hydrological balance.

3.4. Sustainable Management of Groundwater Resources Amidst Environmental Changes

The City of Higashi-Hiroshima is underlain by aquifers with substantial potential for groundwater resources that support the region’s brewing industry. However, these resources are threatened by the cumulative impact of environmental changes on catchment hydrology. Most studies investigating the impacts of land use and climate change on hydrological regimes have focused on either land use or climate change, with few studies addressing both simultaneously.

Table 3. Change in water balance components and their relative percentage change due to land use change, climate change, and their combined hydrologic impact reported in previous studies worldwide. In the ‘key results’ column, the environmental factors are identified. Highlighted factors are those observed to be the dominant contributor to changes in catchment hydrology.

Reference	Study Area	Area (km2)	Study Period	Spatial Distribution	Change Rate (%) in Water Balance Components			Key Results
					SurfQ	ET	RCHG
This study	Kurose River catchment	39.9	1980–2009 (30 years)	HRUs	131.7	−5.1	−34.9	Land use change (rice paddies to urban) + climate change
[21]	Middle Tapi basin, India	32,927	1994–2013 (20 years)	Subbasin	3.1	0.8	−0.7	Climate variability
[24]	Muco watershed, Chile	651	1982–2016 (35 years)	Subbasin	14.1	2.4	−11	Land use change and climate change
[23]	Little Ruaha River catchment, Tanzania	6370	1990–2015 (25 years)	Subbasin	6.1	−0.1	−0.8	Land use change (forest conversion to agriculture)
[49]	Upper Teles Pires basin, Brazil	37,500	1986–2014 (29 years)	None	30.8	−2.5	5.3	Land use change (increase cultivated land)
[60]	Kaptagat catchment, Kenya	269	1989–2019 (30 years)	None	−1.4	N/A	−32.7	Land use change (increase settlement and decrease forest cover) and climate variability (decrease and intense rainfall events)
[11]	Upper Baro basin, Ethopia	23,362	1987–2017 (31 years)	Subbasin	5.6	−1.3	−9.1	Land use change (drastic decrease in grassland and shrubland with an increase in agricultural land and settlement)
[2]	Yamato River catchment, Japan	1077	1970–2010 (50 years)	None	144.8	5	−5.7	Land use change (increase urban percent imperviousness due to urbanization)
[7]	Andalien basin, Chile	742	1984–2013 (30 years)	None	3.43	4.2	−15.6	Land use change (Increase Forest planation coverage)
[48]	Xitiaoxi River watershed, China	66,128	1980–2015 (35 years)	Subbasin	11.9	−0.7	−16.5	Land use/cover change (conversion of forest-grass land and agricultural land to urban land)
[62]	Dübendorf, Switzerland	13.6	1980–2009 (30 years)	None	19.7	−8	5.6	Land use change (urban expansion)

Note: Not all studies assessed the three water balance components, and an N/A, representing ‘not applicable’, directly follows the code in such cases. Rate of change (%) in water balance components: (−) represents a decrease, whereas no negative sign indicates an increase.

summarizes the findings from related studies worldwide, indicating that the groundwater recharge rates in other catchments are generally lower than those observed in this study. This discrepancy may be attributed to variations in catchment size and socioeconomic factors, including land use policies, population density, agricultural production, and urbanization. Most studies listed in Table 3 attributed changes in recharge rates to urban expansion, agricultural conversion, and forest cover.

In contrast, in this study, the conversion of paddy fields to urban settlements led to these changes, with climate change being recognized as a secondary contributor. Kibii et al. [60] reported a similar decline in recharge rates driven by land use and climate change. Other studies, such as Sajjad et al. [9], have revealed the impact of urbanization and vegetation degradation on groundwater levels in Pakistan. The findings in the KRC are consistent with the increased water demand in recent years, reduced recharge from shifting land use, and decreased ET due to rising temperatures and urbanization. Because paddy fields are the primary recharge zones in catchments, protecting them from urban development is essential. While urbanization was the primary driver of reduced groundwater recharge, climate change also altered hydrological processes. The reduction in annual rainfall and increased temperatures led to lower ET rates in residential areas and decreased infiltration, highlighting the combined impact of climate variability and land use changes on groundwater recharge. In light of these findings, conservation efforts, such as the “Satoyama Activities to Cultivate Groundwater in the Saké Capital of Saijo”, have been initiated to promote sustainable water use and ensure the protection of groundwater recharge areas [66]. These initiatives are crucial in mitigating the pressures of rapid urbanization and industrial expansion while ensuring the long-term sustainability of groundwater resources in the KRC.

4. Conclusions

This study applied the Soil and Water Assessment Tool (SWAT) at the Hydrological Response Unit scale to analyze the hydrological impacts of land use changes and climate variability in the Kurose River Catchment over three decades (1980s–2000s). Our findings revealed that land use changes, particularly the conversion of paddy fields into residential areas, were the dominant drivers of hydrological alterations, leading to a significant 34.9% reduction in groundwater recharge. The impacts of climate change were present but were comparatively minor during the study period. Additionally, the combination of increased temperature and reduced evapotranspiration in the catchment likely exacerbates the urban heat island effect, resulting in urban areas becoming warmer than rural areas. This localized warming could further strain water resources because elevated temperatures increase both evaporation rates and water demand.

Projections based on current trends highlight the risks of future challenges to groundwater resources if management strategies are not strengthened. While conservation efforts are underway, further steps are needed to sustain groundwater resources, such as curbing urban encroachment into critical recharge zones, like paddy fields, and enhancing vegetation cover in the catchment to support groundwater recharge. These findings highlight the need to balance urban development with sustainable water management practices in the face of climate change and land use shifts. While this study successfully modeled groundwater recharge using land use and climate data, the absence of well data for the study period limited the direct validation of the modeled recharge rates. Incorporating well data in future studies would provide a stronger validation of the model results, offering greater confidence in the trends related to groundwater recharge. In addition, the use of multiple weather stations or high-resolution gridded data can improve model precision by capturing the spatial climate variability more accurately.