Hydrological Modeling of the Chikugo River Basin Using SWAT: Insights into Water Balance and Seasonal Variability

Abstract

Integrated hydrological modeling plays a crucial role in advancing sustainable water resource management, particularly in regions facing seasonal and extreme precipitation events. However, comprehensive studies that assess hydrological variability in temperate river basins remain limited. This study addresses this gap by evaluating the performance of the Soil and Water Assessment Tool (SWAT) in simulating streamflow, water balance, and seasonal hydrological dynamics in the Chikugo River Basin, Kyushu Island, Japan. The basin, originating from Mount Aso and draining into the Ariake Sea, is subject to frequent typhoons and intense rainfall, making it a critical case for sustainable water governance. Using the Sequential Uncertainty Fitting Version 2 (SUFI-2) approach, we calibrated the SWAT model over the period 2007–2021. Water balance analysis revealed that baseflow plays dominant roles in basin hydrology which is essential for agricultural and domestic water needs by providing a stable groundwater contribution despite increasing precipitation and varying water demand. These findings contribute to a deeper understanding of hydrological behavior in temperate catchments and offer a scientific foundation for sustainable water allocation, planning, and climate resilience strategies.

1. Introduction

Water resources are not only fundamental to human survival and agricultural productivity, but also for the health of marine ecosystems and industrial development. However, their availability is increasingly constrained by growing demands and environmental changes [1]. Hydrology, particularly the study of water balance and rainfall-runoff processes, is pivotal in effective water resource management, flood mitigation, and environmental planning [2]. However, these processes are under significant stress due to anthropogenic activities and environmental factors. Global climate change, extreme weather events, deforestation, urban expansion, and intensified agricultural practices have disrupted natural hydrological cycles, reducing groundwater replenishment, increasing surface runoff, and causing frequent floods and droughts [3]. These disruptions seriously threaten water security, particularly in regions with unpredictable or seasonal rainfall patterns, where fluctuations in water availability can exacerbate food insecurity and hinder sustainable development [4,5]. Furthermore, altered freshwater flows can have far-reaching effects downstream, impacting estuarine and marine ecosystems through changes in salinity, nutrient transport, and sediment delivery, ultimately influencing coastal biodiversity and ecological resilience. To address these challenges, a comprehensive understanding of hydrological processes is essential. This involves evaluating watershed characteristics through quantitative assessments of inputs (e.g., precipitation), outputs (e.g., runoff and evapotranspiration), and changes in water storage [6]. Such analyses are particularly critical in arid and semi-arid regions, where hydrological events are highly variable and heavily influenced by irregular precipitation patterns. However, even in temperate regions, such as the Chikugo River catchment in Japan, climate variability, rapid urbanization, and intensive agricultural practices have created complex hydrological dynamics that require urgent attention.

Streamflow modeling is a vital tool for addressing these issues because it supports reservoir management, water distribution planning, and flood and drought mitigation. Several hydrological models have been developed to simulate streamflow, including the Soil and Water Assessment Tool (SWAT), Python TOPographic Kinematic AP-proximation and Integration (PyTOPKAPI), and Hydrologic Modeling System-Hydrologic Engineering Center (HEC-HMS) [7,8,9]. Among the available hydrological models, the Soil and Water Assessment Tool (SWAT) stands out for its versatility, robustness, and broad applicability across diverse river basins and environmental settings [10,11,12]. Its strength lies in simulating key hydrological processes such as precipitation, runoff, evapotranspiration, and groundwater recharge while accounting for the effects of climate change, land use alterations, and urban development on watershed dynamics through integration with Geographic Information Systems (GIS) [13].

Previous studies in the Chikugo River Basin have primarily focused on flood flow analysis and pollutant load estimation using hydrological models such as HEC-HMS. For example, a study conducted by [14] simulated rainfall runoff and pollutant loads using a GIS-based distributed parameter model, emphasizing water quality indicators like total nitrogen and phosphorus. While valuable for watershed management, this study did not explore temporal and seasonal hydrological variability or the integration of water balance components within the context of current water demand and supply, which is essential for long-term planning. While GIS-based approaches have increasingly enhanced hydrological analysis at multiple scales in complex catchments, their application in the Chikugo catchment remains limited [15]. To address these gaps, this present study applies the SWAT model to the Chikugo River catchment, aiming to generate a comprehensive understanding of its hydrological behavior, especially focusing on baseflow or groundwater availability, which is essential to domestic and agricultural water demand management. Specifically, the study aims to (i) quantify the annual and seasonal key water balance components—evapotranspiration, surface runoff, baseflow, and recharge to the deep aquifer; (ii) assess factors contributing to water balance distribution; and (iii) estimate water availability across the study period (2007–2019) in the Chikugo catchment.

This approach addresses persistent knowledge gaps in integrated hydrological modeling for complex and data-scarce catchments such as the Chikugo River Basin, offering actionable insights for sustainable water resource management amid increasing urbanization and climate pressures. This enhanced understanding supports efforts to manage water availability and quality more effectively, particularly in regions where urban expansion poses risks to transboundary water security. Moreover, this analysis can be used in other catchments with similar environmental and data challenges. It helps in building practical and scalable plans for managing water resources in response to a changing climate and land use.

2. Materials and Methods

2.1. Study Area

The Chikugo River basin (Figure 1), encompassing approximately 2860 km² with an elevation ranging from 0 m above sea level (masl) in the downstream portion to 1524 masl in the upstream mountainous areas, is located in northern Kyushu Island, Japan. It extends across Fukuoka, Saga, Kumamoto, and Oita Prefectures, with most of the basin area located in Fukuoka and Saga. The Chikugo River, with a total length of 143 km, is the longest river on Kyushu Island. It originates from Mount Aso and discharges its water into the Ariake Sea. The upper reaches of the river are important for forestry, whereas the middle and lower reaches are crucial for local agriculture, providing irrigation water for 400 km² of rice fields on the Tsukushi Plain. The river is also vital for industry, with 20 electrical power plants situated along its banks, and is the major city of Kurume in Fukuoka Prefecture, thus contributing to the local economic sustainability of the study area. The mean freshwater inflow into the estuary is 54 m³/s during the dry season, and it exceeds 2800 m³/s during the rainy season [16]. The annual mean precipitation in the watershed is 2180 mm, reaching a maximum of 3000 mm in the mountainous areas.

The Chikugo River estuary is one of the most productive aquatic systems in Japan and supports numerous semi-endemic species [17]. The Chikugo River catchment, a vital water resource for the Fukuoka metropolitan area, faces significant challenges due to population growth, economic development, and climate variability. Industrialization and modernization have intensified water usage, necessitating careful planning to balance the competing demands of domestic, industrial, and agricultural consumers [18]. Historically, the region has experienced severe droughts caused by unprecedentedly low rainfall events, such as those in 1978 and 1994. More recently, abnormally low rainfall during critical periods, such as the autumn of 1999 and 2006 and the rainy seasons of 2002 and 2005, has led to sharp declines in the water volume of the Chikugo River in Senoshita station, prompting the implementation of drought countermeasures. The basin has also experienced extreme runoff events, such as the high historical water levels recorded at the Senoshita station in 2020, highlighting the region’s vulnerability to both floods and droughts.

Between 2006 and 2021, the Chikugo catchment experienced significant land use transformation, marked by rapid urban expansion and a corresponding decline in natural vegetation, as illustrated in Figure 2. Built-up areas nearly doubled from 231.4 km² (8.7%) to 456.3 km² (17.1%), while agricultural paddy fields and evergreen forests declined by 260.3 km² (−9.7%) and 190.6 km² (−7.1%), respectively [Table S5]. These changes reflect broader regional trends in Japan, where urban growth is driven by infrastructure development, rural depopulation, and the consolidation of urban centers. In the Chikugo catchment, this urban expansion primarily occurred at the expense of agricultural and forested lands, particularly evergreen forests, which are ecologically significant for their high-water retention and transpiration capacities.

2.2. General Data Processing

The Q-GIS 3.16 was used as an interface to run the SWAT 2012 model. In this study, key input data, including digital elevation model (DEM), soil, land use, and weather data, were processed using QSWAT to define the watershed, create hydrological response units (HRUs), and generate input files [Figure S2]. A 30-m resolution DEM from the United States Geological Survey (USGS) Shuttle Radar Topography Mission (SRTM) was used to delineate subbasins and stream networks. The soil map (Figure 1b) shows 12 soil types, including Andosols, Cambisols, Acrisols, and Gleysols as dominant soils [Table S4]. The land-use map (Figure 1c) shows that forests dominate the catchment (57.47%), primarily in the upper and midstream regions, whereas agricultural land (27%) is concentrated in lowland areas. Urban areas (10.41%) are mainly downstream, with the remaining 7% comprising grasslands, infrastructure, and water bodies. Slope classes were categorized into five ranges (0–10%, 10–20%, 20–30%, 30–45%, and >45%) to account for variations in runoff and erosion potential. The catchment was divided into 28 subbasins and 3151 HRUs. Land-use and soil data were obtained from the Ministry of Land, Infrastructure, and Transport (MLIT), whereas daily weather data (2004–2021) from 13 meteorological stations [Table S2] were sourced from the Japan Meteorological Agency (JMA). Streamflow data from the Senochita station (2004–2021) were used for calibration and validation and accessed via the MLIT water information system (

Table 1. Spatial datasets and their sources.

Data Set	Format	Temporal/ Spatial Resolution	Data Source
Digital Elevation Model (DEM)	raster	30 m	[19]
Soil type	shapefile	1:500,000	[20]
Land use (2014)	shapefile	100 m (grid cells)	[21]
Weather data	.csv	Daily (2004–2021)	[22]
Discharge (m3/s)	.csv	Daily (2004–2021)	[23]

).

2.3. Model Performance and Sensitivity Analysis

Model calibration and validation were conducted using the Sequential Uncertainty Fitting algorithm (SUFI-2) in the SWAT-CUP. The initial parameter ranges were defined based on the SWAT user manual and refined through iterative optimization. SUFI-2 employs a stochastic approach, running simulations per iteration to minimize errors and assess parameter uncertainty [24]. Calibration was primarily automatic, with manual adjustments made to avoid overfitting and ensure physical relevance. The simulation period ran from 2004 to 2021 and included a warming period (2004–2006). The calibration phase occurred between 2007 and 2014, and the validation phase lasted from 2015 to 2021. These two durations were selected based on the average daily observed streamflows during the two periods, which did not differ significantly. The SWAT model in this study replicated the rainfall-runoff connection using the modified Soil Conservation Service (SCS) curve number technique. Given the limited meteorological data in the research area, we employed the Hargreaves potential evapotranspiration (PET) approach, which is appropriate for areas with limited meteorological data [25]. To avoid over-parameterization and to determine the sensitive parameters, we performed a sensitivity analysis for several parameters that control the hydrological processes in the SWAT prior to the calibration phase [26]. In this study, we ran 100 simulations per iteration, and an uncertainty analysis was conducted together with calibration. The optimization process of the sensitivity analysis of the flow parameters is determined by multiple regression methods against the objective function that reverts the Latin hypercube-created parameters [24]. Approximately 17 parameters related to management, soil, and groundwater were tested using the absolute parameter range provided in the SWAT CUP and then ranked according to their sensitivity based on t-value and p-values. The model parameter with the largest t-distribution (i.e., the most sensitive) and the lowest p-value (i.e., the most significant) was assigned. Here, the parameter was significant if the p-values were close to zero. Subsequently, the best-fit value of each parameter was run once in the validation period. Model performance was evaluated using the coefficient of determination (R²), Nash-Sutcliffe Efficiency (NSE), and percentage bias (PBIAS). Satisfactory model performance was achieved, with R² and NSE values above 0.80 and PBIAS within acceptable ranges during both calibration and validation [27]. The formula for the three parameters (Equations (1)–(3)) was as follows:

$$R^2={\displaystyle \frac{{[{\sum }_{i=1}^n\left(O_i-\overline{O}\right)(P_i-\overline{P})]}^2}{{\sum }_{i=1}^n(O_i-\overline{O}){]}^2{\sum }_{i=1}^n(P_i-\overline{P}){]}^2}}$$

(1)

$$NSE=1-{\displaystyle \frac{{\sum }_{i=1}^n{(O_i-P_i)}^2}{{\sum }_{i=1}^n{(O_i-\overline{O})}^2}}$$

(2)

$$PBIAS=100 \times {\displaystyle \frac{{\sum }_{i=1}^n{O}_i-P_i}{{\sum }_{i=1}^n{O}_i}}$$

(3)

where O_i represents the observed data, P_i represents the simulated data, $\overline{O}$ is the mean of the observed data, and $\overline{P}$ is the mean of the simulated data.

2.4. Principal Component Analysis of Land Use, Soil Type, and Slope Band on Water Balance

Principal Component Analysis (PCA) is one of the most widely used statistical techniques across many research fields [28]. It simplifies complex data by reducing its dimensionality and transforming multidimensional datasets into a smaller number of principal components. This makes it easier to analyze and reveal the relationships within the data. In this study, PCA was used to examine the relationships between land use, soil type, slope band, surface runoff, groundwater recharge, evapotranspiration, and water yield. Our objective was to identify and compare the potential correlations among these factors. The analysis was conducted using Origin Pro 2025 software.

2.5. Assessment of Water Supply, Demographic Trends, and Domestic Water Demand

Data on the water supply population and domestic water consumption were obtained from the Water Resources Planning Division, Water Resources Department, Water Management and Land Conservation Bureau, MLIT [29]. Specifically, we used data on the average water supply per person per day and the water supply population, which are available on the MLIT website. The retrieved data only covered 2007–2019, as these were the years available at the time of the study. Using this data, we computed the water supply population in relation to domestic water consumption (measured in millimeters) and groundwater availability (in millimeters) as part of the water balance calculation. Furthermore, we analyzed domestic water consumption as a percentage of the total groundwater flow, which allowed us to better understand the relationship between water usage and groundwater resources.

3. Results and Discussion

3.1. Model Performance and Uncertainty Analyses

During the calibration period (2007–2014), the model demonstrated strong performance, with a Nash-Sutcliffe Efficiency (NSE) of 0.81 and a coefficient of determination (R²) of 0.81, indicating high accuracy in simulating streamflow [Figure S3]. The Percent Bias (PBIAS) was −3.6%, suggesting a slight underestimation of discharge. This underestimation is within acceptable limits and reflects a consistent trend where the simulated cumulative discharge was lower than the observed discharge. In the validation period (2015–2021), the model maintained its reliability, with an NSE of 0.80 and R² of 0.80 [Figure S4]. However, the fitting performance showed a slight decline, as indicated by an increased PBIAS of −5.7%. This further underestimation remains within acceptable thresholds, as the observed discharge exceeded the simulated discharge during both periods. The underestimation observed in both calibration and validation may be attributed to several factors. Notably, the model did not account for the contribution of reservoirs, as their influence was considered negligible for the scope of this study [30]. Additionally, semi-distributed models often face challenges related to input data quality, particularly in mountainous regions, where localized climatic variability can introduce errors in precipitation and temperature inputs. Despite these limitations, the model produced excellent calibration results, reflecting the quality of input data and the robustness of the model structure. As shown in

Table 2. Model performance statistics calibration and validation results.

Model Performance Evaluation and Uncertainty	Modeling Period	R2	NSE	PBIAS
Calibration	2007–2014	0.81	0.81	−3.6
Validation	2015–2021	0.80	0.80	−5.7

Note: NSE, Nash Sutcliffe efficiency; R², Regression; PBIAS, Percent of Bias.

, the negative PBIAS values confirm a consistent pattern of lower simulated discharge, which may result from unmodeled water losses, simplified land use representation, or limitations in capturing peak flow events. These insights reinforce the importance of refining model inputs and structure to improve simulation accuracy, especially under complex watershed conditions.

Figure 3 shows the relationship between the observed and simulated streamflows over the study period, providing insights into the model performance and hydrological dynamics in the Chikugo River basin. The simulated peak discharge exhibited significant variation when the model underestimated the peak flow during extreme flooding events. However, the observed and simulated discharges corresponded to the precipitation patterns of the basin. Overall, the SWAT model consistently produced high and moderate flows, particularly during rainy months. However, the streamflow was underestimated during the water stress months from October to November (following the monsoon season) and December to January. Asakura City experienced floods and severe heavy rainfall landslide disasters in 2017, owing to major extreme flow events in the summer of 2012 [31]. In 2020, another flood was considered to have the second-highest observed maximum daily peak discharge, reaching 6590 m³/s from 5 July to 7 July 2020, after the 1953 flood in the Chikugo catchment [32]. These extreme hydrological events not only pose severe risks to human settlements but may also disrupt downstream estuarine and marine ecosystems in the Ariake Sea. Sudden surges in freshwater inflow can alter salinity levels, sediment delivery, and nutrient dynamics, potentially affecting coastal biodiversity, primary productivity, and the ecological balance of the marine environment.

The observed and simulated temporal trends indicated the responsiveness of the basin to rainfall events, with sharp increases in streamflow following peak precipitation, reflecting the rapid runoff generation and transport characteristics of the basin [Figure S1]. The inaccuracy of peak discharge in the Chikugo catchment was possibly due to limited meteorological stations, the distance between stations, and uncertainty in GIS data for the spatial distribution of slope, land use, and soil [33]. Moreover, the ability of observed data to accurately record the peak discharge during extreme flooding periods may be limited, leading to uncertainty [34]. The major limitation was that rainfall intensity or duration was not accounted for. Runoff generated by short-duration, high-intensity storms may not all be largely simulated [35]. The agreement between the observed and simulated flows during low-flow periods indicated that the model captured baseflow contributions that are critical for maintaining streamflow during dry periods, thereby ensuring water availability for ecological and human needs [36].

3.2. Sensitive Parameter Analysis

Accurate parameterization is crucial in the calibration of a semi-distributed model characterized by many parameters and must be based on knowledge of the hydrological processes in the system [26]. The values of the 17 calibrated model parameters and their relative rank order of model performance based on the t-test and p-value are listed in

Table 3. List of sensitive parameters, minimum and maximum values, and statistical values according to rank.

Parameter	File Ext.	Method	Description	Min_ Value	Max_ Value	p-Value	T-stat	Rank
CN2	.mgt	r	Initial SCS-CN moisture condition II	−0.003	0.088	0.000	90.56	1
CH_N2	.hru	v	Manning’s “n” value for the main channel	0.040	0.078	0.000	−5.79	2
ALPHA_BF	.bsn	v	Baseflow alpha factor	0.500	0.600	0.009	2.68	3
LAT_TTIME	.hru	v	Lateral flow travel time	159	162	0.022	−2.33	4
CANMX	.hru	v	Maximum canopy storage (mm)	32	35	0.026	−2.26	5
CH_K2	.hru	v	Effective hydraulic conductivity in the main channel alluvium	28	29	0.058	−1.92	6
GWQMN	.gw	v	Threshold depth of water in the shallow aquifer required for return flow to occur	2230	2240	0.070	−1.83	7
GW_DELAY	.gw	v	Groundwater delay	440	445	0.121	−1.567	8
OV_N	.hru	r	Manning’s “n” value for overland flow	−0.20	−0.070	0.131	−1.526	9
ESCO	.hru	v	Soil evaporation compensation factor	0.85	0.95	0.2776	1.096	10
SOL_K	.sol	r	Soil saturated hydraulic conductivity	−0.057	0.129	0.288	−1.070	11
HRU_SLP	.hru	r	Average slope steepness	−0.560	−0.570	0.335	−0.971	12
GW_REVAP	.gw	v	Groundwater “revap” coefficient	9.1	9.3	0.418	0.813	13
SLSUBBSN	.hru	r	Average slope length	0.440	0.451	0.528	−0.634	14
SOL_AWC	.sol	r	Available water capacity of the soil layer	−0.609	−0.135	0.573	−0.565	15
REVAPMN	.gw	v	Threshold depth of water in the shallow aquifer for “revap” to occur	751	752	0.736	−0.338	16
SOL_BD	.sol	r	Moist bulk density	0.312	0.313	0.998	−8.843	17

Note: r indicates that the parameter value is changed relatively, and v indicates that the default parameter is replaced.

. The model parameters CN2, CH_N2, and ALPHA_BF were more significant (p < 0.01) than other parameters at the catchment scale. Among these, the SCS runoff curve number (CN2) was among the most sensitive parameters for the Chikugo catchment. The next sensitive parameter was CH_N2, which represents Manning’s “n” value for the main channel and is used to describe the channel’s roughness, which affects water flow velocity. This parameter is important in hydrological modeling because it influences the rate at which water flows through the channel network [37]. Parameters such as the baseflow recession coefficient (ALPHA_BF) are important in hydrological modeling, especially in temperate climate catchments such as the Chikugo River basin. Its sensitivity originates from its role in regulating the groundwater discharge into streams, which are affected by seasonal precipitation patterns and subsurface characteristics. Variations in ALPHA_BF significantly impact the simulation of baseflow and overall water balance in temperate regions, affecting streamflow predictions during the dry and wet seasons [38,39,40]. Accurate parameter calibration was critical for accurate hydrological modeling in the Chikugo River basin, where seasonal rainfall and groundwater interactions play significant roles in hydrology.

3.3. Analysis of Annual Water Balance Components

Figure 4 shows the annual temporal variability across the water balance components of the Chikugo River basin from 2007 to 2021, indicating the influence of climatic fluctuations and anthropogenic activities. Surface runoff varied significantly between years, ranging from approximately 236 mm in 2012 to 123 mm in 2013, highlighting its sensitivity to precipitation intensity [41]. This fluctuation reflects the basin’s responsiveness to precipitation intensity and distribution, with high runoff years often coinciding with intense rainfall events or reduced infiltration capacities due to land use changes. However, baseflow varied significantly over time, serving as an important component in maintaining streamflow during dry periods while also reflecting fluctuations in groundwater contributions and recharge patterns. Recharge to deep aquifers remained relatively stable over time, indicating a less immediate impact of short-term climatic fluctuations on deep groundwater storage, as in most Japanese catchments [42,43]. Recharge to deep aquifers varied only slightly in all years, indicating that long-term groundwater reserves remained stable despite interannual climatic changes.

Evapotranspiration (ET), which consistently accounted for the largest proportion of the water balance, highlighted the dominant role of vegetation and climatic factors in driving hydrological processes in the basin [44,45,46]. The dominance of ET, which consistently exceeded 43% of precipitation, demonstrated the basin’s semi-humid climate and the significant influence of vegetation and atmospheric moisture demand [47]. This high ET directly impacted water resource availability, limiting the proportion of accessible water for surface runoff and groundwater recharge, and necessitating efficient water management practices [48], particularly in the lowland areas, which are flatter and more developed agricultural land (27%) and urban zones (10.41%), playing a growing role in shaping water movement. Precipitation variability was critical in shaping the interannual trends of all components, with wetter years, such as 2012, 2016, and 2021, showing elevated values across ET, baseflow, and runoff. However, drier years, such as 2007, 2008, and 2017, exhibited suppressed values (Figure 4), particularly in surface runoff and baseflow patterns. The combined trends of baseflow and recharge to deep aquifers revealed the dynamics of groundwater contributions and storage, with higher baseflow years, such as 2012, 2016, 2020, and 2021 (847–913.8 mm), exhibiting increased groundwater contributions following significant recharge periods [49,50]. These hydrological shifts are closely tied to the catchment’s physical characteristics. Forests dominate the landscape, covering nearly 58% of the area, especially in the steep upland and midstream zones where slopes exceed 30% in about a third of the catchment. Forests typically promote infiltration and reduce runoff, but under heavy or prolonged rainfall conditions that are becoming more common, soils can become saturated, leading to increased runoff and baseflow. The catchment’s soils, including Andosols, Cambisols, Acrisols, and Gleysols, have diverse water retention and drainage properties. For example, Andosols and Acrisols hold moisture well but may slow drainage, contributing to higher baseflow and recharge under wet conditions. This relationship is crucial for understanding the sustainability of groundwater use, especially under scenarios of increased water demand or reduced rainfall [51].

3.4. Analysis of Seasonal Variation of Water Balance Components

The seasonal variation in water balance components for the Chikugo River basin (Figure 5) reveals distinct hydrological behavior across winter, spring, summer, and autumn, shaped by the interaction of climatic drivers and watershed characteristics. These seasonal dynamics are critical for understanding temporal water availability and guiding adaptive water resource management. Evapotranspiration (ET) dominated the water balance in all seasons, reflecting its central role in the basin’s hydrological cycle. The highest ET was observed in summer (354.9 mm), followed by spring (231.3 mm), autumn (202.7 mm), and winter (101.2 mm), consistent with seasonal variations in temperature and solar radiation [52,53,54]. This pattern highlights the strong influence of atmospheric demand and vegetation activity, particularly during the growing season. The elevated summer ET also indicates increased water loss from both soil and vegetation, which can reduce water availability for runoff and recharge, especially in agricultural zones where irrigation demand is high. Baseflow peaked in summer (379.1 mm), followed by autumn (186 mm), spring (179.7 mm), and winter (126.5 mm). This seasonal trend suggests that groundwater discharge into the river system is enhanced during warmer months, likely due to delayed recharge from spring precipitation and sustained soil moisture. The high summer baseflow also reflects the basin’s subsurface connectivity and the role of shallow aquifers in maintaining streamflow during periods of low rainfall. These dynamics are particularly important for sustaining ecological flows and water supply during dry seasons. In Kyushu Island, Japan, the planting season largely depends on the crop type and the region’s warm temperate climate. During the rice growing seasons (spring and summer), maintaining baseflow is important, especially outside the rainy season, as it ensures stable streamflow, which supports irrigation and buffers the effects of climate extremes, especially when surface runoff is unpredictable or ET is high. Therefore, effective water management allocation during this transition is essential to balance domestic and agricultural water demand, flood control, and groundwater recharge. Recharge to deep aquifers remained relatively stable across seasons, ranging from 40.8 mm in winter to 51.8 mm in autumn. This consistency indicates that deep groundwater systems are buffered against short-term climatic fluctuations, a characteristic typical of Japanese catchments with well-developed aquifer systems [55,56]. The slight seasonal variation may be attributed to differences in infiltration rates and soil moisture retention, but overall, the recharge stability supports long-term groundwater sustainability in the basin. Surface runoff (SURFQ) showed the most pronounced seasonal variation, with the highest value in summer (103.7 mm), followed by spring (37.3 mm), autumn (33.4 mm), and winter (23.9 mm). The substantial surface runoff during summer underscores the need for effective flood management strategies to mitigate the potential risks associated with intense rainfall and reduced infiltration. This reduced infiltration can result from high evapotranspiration rates, vegetation water uptake, and surface compaction caused by intense precipitation, all of which restrict water percolation despite increased rainfall [57,58]. By linking climatic variability to water balance components, the analysis supports the broader research objective of enhancing watershed resilience and informing seasonally adaptive water resource planning under changing climate conditions.

3.5. Relationship Between Land Use, Soil Type, and Slope Band on Water Balance

The PCA biplot in Figure 6 offers a clear visualization of the relationships among various environmental variables in the Chikugo catchment, revealing key insights into how soil types, land use, slope, and climatic factors interact and influence catchment characteristics. The first principal component (PC1), accounting for 29.54% of the variance, aligns closely with ETmm (evapotranspiration), GWQmm (groundwater quantity), and, to a lesser extent, soil type and land use, suggesting these variables collectively represent a gradient of hydrological and land surface processes. The second component (PC2), explaining 15.89% of the variance, aligns primarily with soil type and land use, and slightly with slope band, indicating a secondary gradient that differentiates catchment areas based on terrain and soil characteristics. In the context of the Chikugo catchment, where Andosols, Cambisols, Acrisols, and Gleysols are dominant soils, the clustering of soil type with land use along PC2 reflects the strong spatial heterogeneity of the catchment. The upper and midstream areas, dominated by forest cover (57.47%) on steeper slopes (with 34% of the catchment having slopes greater than 30%, likely correspond to soils such as Andosols and Cambisols, which favor forest growth and influence evapotranspiration and groundwater recharge dynamics. These forested areas show a clear association with higher PC2 scores, indicating distinct soil and land use patterns that impact hydrological processes. In contrast, the lowland areas are characterized by agricultural land (27%), and urban regions (10.41%) cluster on the opposite end of PC1, indicating a different hydrological regime likely influenced by soil types such as Gleysols and Acrisols that are more common in floodplains and low slopes (0–10%). This spatial distinction is critical, as agricultural and urban land uses generally reduce infiltration and increase surface runoff, thus impacting groundwater recharge and evapotranspiration patterns differently than forested regions.

The negative values of the slope band on PC2 suggest that slope influences soil and land use relationships, with steeper areas supporting forest soil and land use, while flatter areas support agricultural and urban land uses. This relationship is supported by a study conducted by Nang et al. [59] in the Takashi Catchment, which found that slope gradient has the most pronounced effect on bareland, agricultural, and rice land covers, resulting in 2.2 t ha⁻¹ year⁻¹ at slope > 45%. Climatic variables such as SURQmm (surface runoff) and WYLDmm (water yield) have negative values on PC1, indicating an inverse relationship with ETmm and groundwater quantities, which aligns with the catchment’s hydrological behavior, where higher runoff often corresponds with lower groundwater recharge. This relationship is further explained by previous studies by Guyo et al. [60] and Wetherl et al. [61], which highlights the interaction between ground and surface water and the role of vegetation and land cover in regulating watershed hydrology. Moreover, rapid urbanization and the impact of vegetation degradation on groundwater levels have led to changes in water balance due to increased impervious surfaces that enhance flooding and over extraction of groundwater [62,63]. Urban development accelerates runoff and limits infiltration, while agricultural practices can either support or hinder infiltration depending on soil management. According to Gatwaza et al. [64], surface runoff is the most sensitive component affected by land use changes. Other studies by Kimbi et al. [65] and Kibii et al. [66] have reported the combined effects of climate and land use changes, especially more intense and irregular rainfall, intensifying hydrological variability and pressure on water resources.

3.6. Domestic Water Use for Sustainable Water Resource Management

Domestic water use is a critical component of the overall water balance in the Chikugo River Basin, directly linking hydrological processes to human demand and long-term sustainability. The Chikugo River provides water essential for both domestic and agricultural use, benefiting the lives of many people. Domestic water intake from the river is distributed across three main areas: the Eastern Saga Water Supply Authority, the Greater Southern Fukuoka Water Supply Authority, and Fukuoka District Water Supply (Fukuoka Canal), also extending to other regions. According to data from MLIT, over 3.6 million people rely on the Chikugo River for their water supply. This represents approximately 83.3% of the population of Fukuoka Prefecture, 14.2% of the population of Saga Prefecture, 1% of the population of Kumamoto, and 1.6% of the population of Oita Prefecture. Domestic water use remained relatively stable, averaging 142.4 mm/year over the study period. The highest recorded consumption was 147.8 mm/year in 2011, whereas the lowest was 137 mm/year in 2014, exhibiting a 7.4% variation. The temporary decline during 2012–2014 may be attributed to improved water efficiency measures or climate-related factors affecting demand.

Groundwater availability showed significant interannual variability, fluctuating between 629.9 mm/year (2007) and 872.1 mm/year (2016), reflecting a 38.4% difference (

Table 4. Trends in Water Supply Population, Domestic Water use, and Groundwater Availability in the Chikugo catchment.

Year	Precipitation (mm)	Water Supply Population (million)	Domestic Water Consumption (mm)	Groundwater Availability (mm)	Domestic water Consumption As a Proportion of Total Groundwater Flow (%)
2007	1760.3	3.52	143.6	629.9	22.8
2008	1783.2	3.54	145.4	731.4	19.9
2009	1914.1	3.57	144.4	774.7	18.6
2010	1856.1	3.58	146.8	769.1	19.1
2011	2094.1	3.61	147.8	746.1	19.8
2012	2285.2	3.69	138.0	847.0	16.3
2013	2085.1	3.70	139.0	754.3	18.4
2014	1927.7	3.73	137.0	754.7	18.1
2015	2156.3	3.75	139.7	809.2	17.3
2016	2509.3	3.77	142.0	872.1	16.3
2017	1704.6	3.79	145.0	743.1	19.5
2018	2058.5	3.80	144.8	800.4	18.1
2019	1905.9	3.82	145.0	790.2	18.3

). The highest groundwater availability in 2016 (872.1 mm/year) coincided with an increase in domestic water use (142.0 mm/year), suggesting a potential correlation with recharge conditions and extraction levels. However, after 2016, groundwater availability declined by 14.8% in just 1 year (to 743.2 mm/year in 2017), remaining below 800 mm/year for the subsequent years. The increasing population in the Chikugo catchment, especially around rapidly urbanizing areas like Fukuoka City, has led to a gradual rise in water consumption. In this study, while groundwater availability has remained stable due to sufficient precipitation, the combined pressure from domestic, agricultural, and industrial demand underscores the urgency of integrating water management strategies that balance urban growth with sustainable groundwater use into hydrological assessments and planning frameworks. According to studies conducted in urban regions of Japan, such as Tokyo, Osaka, Chiba, and Toyama, improved water use efficiency and policy interventions have played a significant role in addressing the fragility of Japan’s water resource systems. These efforts highlight the importance of adopting more coordinated and flexible water management strategies to achieve sustainable outcomes [67,68,69].

4. Conclusions

The model demonstrated strong performance with Nash–Sutcliffe Efficiency (NSE) and coefficient of determination (R²) of 0.81 (calibration and validation) with less PBIAS, confirming its reliability for hydrological assessments in complex and data-scarce catchments. It also highlights the importance of baseflow, especially during the spring and early summer season, in maintaining the hydrological stability in the Chikugo catchment, particularly in meeting agricultural and domestic water demands across varying seasons. Moreover, the Principal Component analysis (PCA) reveals that evapotranspiration (ET), groundwater (GW), and land use are the most influential factors contributing to hydrological variability, with PC1 and PC2 explaining a combined 45.43% of the total variance. Additionally, domestic water use accounted for 16.3% to 22.8% of total groundwater flow, with notable temporal fluctuations, emphasizing the need for continuous monitoring, responsive water allocation, and adaptive water management. Overall, the findings contribute to advancing hydrological modeling in Kyushu regions, support climate-resilient water governance, and provide a scientific foundation for addressing transboundary water security challenges. Addressing existing data limitations and exploring future scenarios further will enhance the model’s reliability and support sustainable water resource management.