Groundwater Vulnerability Assessment in the Huangshui River Basin Under Representative Environmental Change

Abstract

The Huangshui River Basin is located in the transition zone between the Loess Plateau and the Qinghai–Tibet Plateau, characterized by a fragile hydrological and ecological environment. Groundwater serves as a vital water source for local economic development and human livelihood. With the acceleration of urbanisation and climate change, groundwater resources face challenges such as pollution and over-exploitation. This study employs an improved DRASTIC model, tailored to the characteristics of the groundwater system in the Huangshui River Valley of the upper Yellow River, to integrate groundwater resources, groundwater environment, and ecological environment systems. Improving the DRASTIC model for groundwater vulnerability assessment. A two-tiered evaluation system with nine indicator parameters was proposed, including six groundwater quality vulnerability indicators and five groundwater quantity vulnerability indicators. Fuzzy analytic hierarchy process and entropy weight method were used to determine the weights, and Geographic Information System (GIS) spatial analysis was employed to evaluate groundwater vulnerability in the Huangshui River basin in 2006 and 2021. The results indicate that the proportion of areas with high groundwater quality vulnerability increased from 10.7% in 2006 to 31.57% in 2021, while the proportion of areas with high groundwater quantity vulnerability decreased from 22.33% to 14.02%. Overall, groundwater quality vulnerability in the Huangshui River basin is increasing, while groundwater quantity vulnerability is decreasing. Based on the evaluation results of water quality and quantity vulnerability, protection zoning maps for water quality and quantity were compiled, and preventive measures and recommendations for water quality and quantity protection zones were proposed. Human activities have a significant impact on groundwater vulnerability, with land use types and groundwater extraction coefficients having the highest weights. This study provides a scientific basis for the protection and sustainable use of groundwater in the Huangshui River basin.

1. Introduction

Groundwater, as an important freshwater resource, plays an irreplaceable role in maintaining ecological balance and supporting socio-economic development, particularly in arid and semi-arid regions [1,2]. Since the 1970s, various scholars have proposed their own interpretations of groundwater vulnerability. Barbulescu defined groundwater vulnerability as the likelihood of pollutants penetrating and spreading from the surface to the groundwater layer under natural conditions [3]. Vierhuff defined groundwater vulnerability as the complexity of surface and subsurface conditions that influence the entry of pollutants into aquifers [4]. Duijvenbooden and Waegeningh introduced the concepts of ‘aquifer pollution vulnerability’ and ‘groundwater pollution risk’ in 1987 [5]. Some researchers correctly linked aquifer vulnerability to the characteristics of the layer above the saturated zone to reduce the impact of pollutants [6,7,8]. Baig et al. regarded it as a property of the groundwater system that depends on the sensitivity of the material in permitting the degradation of the saturated zone by pollutant substances originating from human activities [9]. Due to the enhanced interaction of human activities, groundwater pollution is a critical environmental issue [10,11].

Methods for evaluating groundwater vulnerability have become increasingly diverse as the number of study areas has grown, primarily including overlay index methods, process simulation methods, and statistical methods. Researchers have developed various typical models based on different methods, such as DRASTIC, GOD, SINTACS, SEEPAGE, COP, and PLEIK [12,13,14,15,16]. Among these, the DRASTIC model, integrated with the ARCGIS platform, has been widely applied for groundwater vulnerability assessment and grading in countries and regions such as the United States, Canada, South Africa, and the European Union. As research on groundwater vulnerability has progressed, numerous scholars both in China and abroad have proposed enhanced versions of the DRASTIC model, aiming to improve its accuracy and adaptability under diverse hydrogeological conditions [17]. This widespread methodological evolution reflects the model’s broad acceptance and practical applicability in the global academic community. However, methods for assessing the vulnerability of karst aquifers vary significantly [18]. In 2009, Umar conducted a groundwater vulnerability assessment of the Yamuna-Krihni sub-basin in the central plains of the Ganges River. Due to the small topographical variations, the parameter T (topographical slope) was removed from the DRASTIC index and replaced with parameter L (land use), which reflects the impact of land use on water quality. Through the study of the aquifers in the Ganges Plain, it was concluded that depth to water table (D), hydraulic conductivity (C), impact of the vadose zone (I), and land use (LU) are the primary factors influencing vulnerability [19]. In 2012, Alam et al. used an improved DRASTIC-LU model to assess groundwater vulnerability to pollution in the central Ganges Plain, creating a vulnerability zoning map. The evaluation criteria included soil medium type, depth, net recharge, evaporation zone influence, terrain slope, aquifer medium type, aquifer conductivity, and land use [20]. This model incorporated land use as an important indicator, considering the impact of human activities on groundwater vulnerability. In 2015, Bartzas et al. considered pesticide DRASTIC and sensitivity index (SI) methods within a GIS framework to assess groundwater vulnerability in the Albenga agricultural area of northern Italy. The SI was obtained by removing S, I, and C from DRASTIC, including land use parameters (LU), which includes the impact of agricultural activities (especially nitrates) on water quality [21]. Additionally, determining indicator weights is a critical step in establishing an evaluation indicator system. To enhance the objectivity of vulnerability indices, reduce the subjectivity and uncertainty of evaluation models, and improve the accuracy of the evaluation system, many scholars have combined multiple methods in their research to maximise the accuracy of the evaluation system. In 2023, Abu Sayed et al. [22] used Geographic Information Systems (GIS) and pollution risk indices such as DRASTIC and GOD to comprehensively analyse industrialised suburban areas in Bangladesh, demonstrating significant variations in vulnerability across the study area. In 2021, Maryam Torkashvand [23] and in 2023, Karimzadeh Motlagh [24] conducted separate studies on the Gazvin aquifer in Iran and the Najafabad Plain, the former employed a new multi-attribute decision-making method, SWARA (Step-Wise Weighted Assessment Ratio Analysis), to enhance the DRASTIC rate and weights, while the latter designed a novel approach combining the DRASTIC model and machine learning algorithms to assess groundwater pollution risks. Through studies in different regions, the applicability and effectiveness of various evaluation methods under diverse environmental conditions were validated, contributing to the further refinement of evaluation methods and enhancing their universality.

The temporal and spatial distribution of groundwater resources in Qinghai Province is severely mismatched with the demands of land-resource development and socio-economic growth. Temporally, monsoon-dominated recharge is concentrated in June–September, whereas municipal–industrial and irrigation uses are year-round and peak in late winter and spring; the limited effective storage of valley-fill aquifers cannot fully buffer this lag, so deficits and head declines recur in dry and heating seasons. Spatially, effective recharge is focused on piedmont and upper alluvial-fan zones, while withdrawals are concentrated along the Xining–Haitong urban corridor and in the Qaidam salt-lake industrial zone, where aquifers are thin or laterally discontinuous and impervious surface expansion limits local infiltration [25,26]. The Huangshui River Basin is located at the transition zone between the Qinghai–Tibet Plateau and the Loess Plateau, serving as a crucial ecological barrier, meaning a landscape belt that maintains regional ecological security by conserving headwater runoff and baseflow, retaining soil on loessal slopes, limiting desertification and dust export, and providing habitat connectivity through its coupled groundwater and surface water system, and as an economic development core region in China’s northwestern arid zone. In recent years, under the dual impact of global climate change and human activities, groundwater resources in the basin have faced severe challenges: on the one hand, accelerated urbanisation has led to a surge in water demand, resulting in widespread groundwater over-exploitation, which has triggered issues such as continuous water level decline and ground subsidence; on the other hand, agricultural non-point source pollution and industrial wastewater discharge have caused groundwater quality vulnerability deterioration, threatening drinking water safety [27,28,29]. Groundwater vulnerability assessment is a crucial tool for quantifying the sensitivity of groundwater systems to external stresses, providing a scientific basis for water resource protection and management. The traditional DRASTIC model does not consider the influence of human activities, such as land use changes, agriculture, and urbanization, which can significantly alter the dynamics of groundwater recharge, contamination, and availability. In this study, the land use type indicator was incorporated into the modified DRASTIC model to address these human influences, thereby providing a more accurate and region-specific assessment of groundwater vulnerability, particularly in areas undergoing significant environmental and anthropogenic changes, such as the Huangshui River Basin.

This study is based on a comprehensive dataset that includes hydrological, meteorological, hydrogeological, land use, and human engineering activity data from the Huangshui River Basin. Both groundwater quality vulnerability and quantity are considered critical factors in assessing groundwater vulnerability, as they are essential for understanding the groundwater system dynamics in the region. The DRASTIC model was used to identify key indicators influencing both groundwater quality vulnerability and quantity vulnerability in the study area. To enhance the accuracy and minimize subjectivity in indicator weighting, fuzzy AHP and entropy weighting methods were integrated. This combination ensures a balanced approach that incorporates expert judgment and objective data variability, improving the overall reliability of the assessment. A modified DRASTIC model was established for groundwater vulnerability assessment, and a tailored indicator system was developed specifically for the Huangshui River Basin. The modified model was applied to conduct single-indicator evaluations using ArcGIS, and spatial overlay analysis was employed to assess groundwater vulnerability in terms of both water quality and quantity for the years 2006 and 2021. A groundwater vulnerability zoning map, ranging from Grade I to Grade V, was produced, and trends in groundwater vulnerability under changing environmental conditions, such as urbanization and climate change, were analyzed.

2. Materials and Methods

2.1. Study Area

The Huangshui River Basin (100°42′–103°04′ E, 36°02′–37°28′ N) is located in the eastern part of Qinghai Province and serves as a critical ecological transition zone between the Qinghai–Tibet Plateau and the Loess Plateau (Figure 1). The basin covers a total area of 1.6 × 10⁴ km², with a topography that is higher in the northwest and lower in the southeast, averaging an elevation of 3000–4000 m. This has formed a unique landform structure characterised by ‘three mountains flanking two valleys.’ The region has a complex geological structure, and since the Cenozoic Era, it has been subjected to the compressional forces of the Indian Plate, resulting in a series of fault-block basins that provide excellent groundwater storage spaces. The climate of the basin is typical of a highland continental climate, with an average annual temperature of −3–8 °C. Precipitation varies significantly across the region: 250–350 mm in the river valleys and up to 500–700 mm in the mountainous areas. The Huangshui River, a first-order tributary of the Yellow River, has a total length of 342 km and is fed by 73 tributaries, forming a feather-like river system. The groundwater system primarily receives replenishment from precipitation infiltration, river seepage, and snowmelt, forming an aquifer system dominated by pore water in loose rock (accounting for 60%) and supplemented by bedrock fracture water and karst water. Among these, the Quaternary sand and gravel aquifer in the river valley has a thickness of 15–25 m, with individual wells yielding over 1000 m³/d, making it the primary extraction zone. The study area covers Xining and Haidong, core economic zones of Qinghai Province, with a GDP of 215.1 billion yuan in 2020 and a permanent population of 4.47 million (75% of the province’s total). Over the past 15 years, the urbanisation rate has increased by 9.4%, and the expansion of construction land has led to significant changes in land use in the river valley area. According to monitoring data, the groundwater level in the Xining Basin decreased by 6.8 m between 1985 and 2022, and nitrate (NO₃⁻) concentrations exceeded the standard (>50 mg/L) in agricultural areas such as Ledu. Concurrently, the ‘Yin Da Ji Huang’ water diversion project transfers about 250 million cubic metres of surface water per year from the Datong River in Qinghai to the Huangshui River Basin, modifying surface flows, seasonal availability, and recharge via canal seepage and irrigation return. This redistribution of water under natural and human controls makes the basin an ideal region for studying groundwater vulnerability under changing environmental conditions (Figure 1).

2.2. Data Sources and Processing

This study compiled all indicator datasets required by the improved DRASTIC framework into a unified geospatial database and processed them to a common 30 m grid to ensure comparability and reproducibility. Groundwater level depth (D) was derived from 44 monitoring wells across the Huangshui River Basin, with depth surfaces for 2006 and 2021 generated by ordinary Kriging in ArcGIS after harmonizing coordinates and extents, screening outliers using median ± 3 MAD against field logs, de-trending, and fitting spherical, exponential, and Gaussian semivariograms; the final model for each year was selected by leave-one-out cross-validation and we report RMSE, mean error, and standardized RMSE. Net recharge (R) was computed from a basin-consistent water-balance formulation that integrates riverbed infiltration, canal leakage, field irrigation seepage, and precipitation infiltration using land use and soil-texture-dependent infiltration coefficients calibrated against multi-year water-table variations at the same 44 wells to keep mass balance within plausible bounds. Aquifer media (A) were interpreted from hydrogeological drilling logs and existing aquifer maps, with permeability information embedded in the media classes in order to avoid double-counting with the hydraulic-conductivity-related indicators in the improved design. Soil media (S) were obtained from the Resource and Environmental Science Data Centre, converted to the study projection, topology-checked, reclassified by texture groups, and scored by permeability. Topographic slope (T) was calculated from a 30 m DEM using a 3 × 3 window, inspected for artifacts, and reclassified with a natural breaks method. Land use (L) for 2006 and 2021 used CLCD 30 m products from Wuhan University, clipped to the basin, topology-checked, reclassified to assessment categories, and filtered with a 3 × 3 mode operator in highly fragmented mosaics to reduce edge effects. Aquifer thickness (H) was compiled from borehole records and groundwater well profiles, interpolated along structural trends and then classified; groundwater water-bearing capacity (W) was mapped by vectorizing specific discharge grades and spring yields and converting water-rich zones into assessment classes; the exploitable modulus (M) was digitized from provincial groundwater resource assessments and reconciled with local borehole information before vectorization and grading. All layers were co-registered in a single projection, continuous variables were resampled bilinearly and categorical variables by nearest neighbor, class scores followed Table 1 and Table 2, and a data-inventory and processing log with layer versions and timestamps was maintained to ensure traceability.

The selection of 2006 and 2021 as the reference years for this study is based on data reliability, policy transitions, and representative environmental change. Specifically, 2006 marks the earliest year with complete and standardized groundwater monitoring records across the basin, following the provincial groundwater resource assessment. In contrast, 2021 coincides with the implementation of China’s 14th Five-Year Plan and a new phase of land use reform, groundwater regulation, and ecological restoration efforts. The 15-year interval captures significant anthropogenic and climatic changes, enabling a meaningful comparison of groundwater vulnerability patterns and evolution trends under shifting environmental pressures.

2.3. Improving the DRASTIC Evaluation Model

The DRASTIC evaluation model is a widely used groundwater vulnerability assessment model both domestically and internationally [30,31,32,33], developed jointly by the National Well Water Association (NWWA) and the U.S. Environmental Protection Agency (USEPA) in 1985. It was initially used to assess the sensitivity of groundwater pollution in 40 counties in the United States. Later, experts integrated relevant parameters on a geographic information system (GIS) platform, further developing the model to accommodate larger hydrogeological units. The DRASTIC evaluation method considers seven indicators to quantitatively assess the degree of groundwater vulnerability, namely groundwater depth (D), vertical net recharge (R), aquifer medium (A), soil medium (S), terrain slope (T), unsaturated zone medium type (I), and hydraulic conductivity (C). The first letters of these indicators form the acronym DRASTIC. Each indicator is further divided into different numerical ranges and quantified into indicator values. Additionally, the weighting of each indicator was determined in accordance with its relative contribution to groundwater vulnerability, as quantified through a combined approach of expert judgment and objective statistical methods. This ensured that the assigned weights accurately reflect the indicator’s influence within the overall evaluation framework. Finally, the comprehensive indicator value is calculated using a weighted method, which represents the groundwater vulnerability index and reflects. The DRASTIC model employs a weighted overlay method, categorising each indicator into different categories (qualitative indicators) and intervals (quantitative indicators). For each interval of each indicator, scores ranging from 1 to 10 are assigned from weak to strong. Then, based on the relative importance of each indicator’s impact on groundwater vulnerability, weight values ranging from 1 to 5 are assigned, with higher values indicating greater impact. Using the overlay index method, the scores are weighted and summed to obtain the groundwater vulnerability score and vulnerability grade classification.

The DRASTIC groundwater vulnerability index (DI) is determined by the following equation:

$$DI={\displaystyle \sum _{i=1}^7{p}_i{\omega }_i}$$

(1)

where P_i is the score of the i-th indicator in the region, and ω_i is the weight of the i-th indicator.

According to the classification and valuation of groundwater vulnerability assessment indicators in the ‘Guidelines for the Division of Groundwater Pollution Prevention and Control Zones’ issued by the Ministry of Ecology and Environment in 2019, the interval scores for each indicator of the DRASTIC model are shown in Table 1.

Table 1. DRASTIC model indicators and weights.

Score	Groundwater Depth (m)	Net Recharge (mm)	Aquifer Medium	Soil Medium	Terrain Slope (°)	Packing Gas Medium Type	Hydraulic Conductivity (m/d)
10	(0, 2]	>235	gravel	thin or missing	(0, 2]	cobble-gravel	>81.5
9	(2, 4]	(216, 235]	sand-gravel	cobble-gravel	(2, 3]	sand-gravel	(71.5, 81.5]
8	(4, 6]	(178, 216]	coarse sand	gravel and coarse sand	(3, 4]	coarse sand	(61.1, 71.5]
7	(6, 8]	(147, 178]	medium sand	silt and fine sand	(4, 5]	medium sand	(40.7, 61.1]
6	(8, 10]	(117, 147]	fine sand	cohesive clay	(5, 6]	fine sand	(34.6, 40.7]
5	(10, 15]	(92, 117]	silty fine sand	sandy soil	(6, 7]	silty fine sand	(28.5, 34.6]
4	(15, 20]	(71, 92]	silt	loam	(7, 8]	silt	(20.3, 28.5]
3	(20, 25]	(51, 71]	sandy silt	silty loam	(8, 9]	sandy silt	(12.2, 20.3]
2	(25, 30]	(0, 51]	silty clay	clay	(9, 10]	silty clay	(4.1, 12.2]
1	>30	0	clay	rock	>10	clay	<4.1

Table 1. DRASTIC model indicators and weights.

Score	Groundwater Depth (m)	Net Recharge (mm)	Aquifer Medium	Soil Medium	Terrain Slope (°)	Packing Gas Medium Type	Hydraulic Conductivity (m/d)
10	(0, 2]	>235	gravel	thin or missing	(0, 2]	cobble-gravel	>81.5
9	(2, 4]	(216, 235]	sand-gravel	cobble-gravel	(2, 3]	sand-gravel	(71.5, 81.5]
8	(4, 6]	(178, 216]	coarse sand	gravel and coarse sand	(3, 4]	coarse sand	(61.1, 71.5]
7	(6, 8]	(147, 178]	medium sand	silt and fine sand	(4, 5]	medium sand	(40.7, 61.1]
6	(8, 10]	(117, 147]	fine sand	cohesive clay	(5, 6]	fine sand	(34.6, 40.7]
5	(10, 15]	(92, 117]	silty fine sand	sandy soil	(6, 7]	silty fine sand	(28.5, 34.6]
4	(15, 20]	(71, 92]	silt	loam	(7, 8]	silt	(20.3, 28.5]
3	(20, 25]	(51, 71]	sandy silt	silty loam	(8, 9]	sandy silt	(12.2, 20.3]
2	(25, 30]	(0, 51]	silty clay	clay	(9, 10]	silty clay	(4.1, 12.2]
1	>30	0	clay	rock	>10	clay	<4.1

2.3.1. Selection of Water Quality Vulnerability Evaluation Indicators

The Huangshui River basin is located in a highland semi-arid region. The traditional DRASTIC model has limitations for groundwater quality vulnerability in this basin because groundwater occurs mainly in Quaternary unconsolidated aquifers and the controlling processes differ from humid settings. Based on previous analyses of groundwater evolution and its drivers in the basin [34,35], we adopt an improved scheme that matches the dominant mechanisms and the available data. For water quality vulnerability we use depth to water D to represent travel time, net recharge R to represent downward flux, aquifer media A to represent permeability pathways, soil media S to represent near surface filtration, topographic slope T to represent infiltration propensity, impact of the vadose zone I to represent attenuation and delay across the unsaturated column, and land use L to represent anthropogenic loading and infiltration routes. For water quantity vulnerability we use R, A, aquifer thickness H, water bearing capacity W and exploitable modulus M to represent supply, transmissive capacity, storage, yield potential and sustainable availability. Hydraulic conductivity information is embedded within A to prevent double-counting with H and with permeability proxies. The choice reflects the measured unsaturated thickness of 5.7 to 26 m with soil thickness usually under 1 m, the hydrostratigraphy of the valley fill and fans, and the spatial pattern of human activity in the corridor.

2.3.2. Selection of Water Quantity Vulnerability Evaluation Indicators

From the perspectives of water quality and quantity, the vulnerability of groundwater quality vulnerability and quantity are two distinct attributes. In regions with high aquifer permeability and abundant recharge, water quality vulnerability is relatively higher, as pollutants can easily enter the aquifer through various pathways, such as precipitation runoff. These regions have favourable recharge conditions and abundant groundwater resources, providing favourable conditions for the development and utilisation of groundwater. Therefore, a combined assessment of water quality and quantity vulnerability can help strengthen the integrated management of groundwater protection and utilisation. Typically, areas with low water quantity vulnerability are often also regions with high water quality vulnerability that require strict protection. Based on the groundwater resources and human activities in the Huangshui River basin, the selected indicators for groundwater quantity vulnerability in this study are net recharge (R), aquifer thickness (H), aquifer medium (A), water content (W), and exploitable modulus (M). Based on data from 2006 and 2021 for relevant indicators in the Huangshui River basin, the dynamic indicators selected for this water quantity vulnerability assessment are net recharge (R) and groundwater exploitable modulus (M).

2.4. Determination of Evaluation Indicator Weights

Determining the weights of each indicator is a critical step in deriving the final score. In the traditional DRASTIC model, the weights of the seven parameters are fixed. However, in real-world scenarios, the weights of vulnerability indicators for different regions may vary due to differences in geological conditions. Previous studies on groundwater vulnerability assessment have commonly employed methods that subjectively classify and assign weights to evaluation factors, such as the Analytic Hierarchy Process (AHP) [36] and fuzzy mathematics theory [37]. These methods lack objectivity. To ensure that the determination of indicator weights is as objective and reasonable as possible, this study references previous research findings and combines the actual influence of each indicator on groundwater vulnerability. The weights of the vulnerability indicators are calculated using the entropy weight method [38] and the fuzzy hierarchical analysis method [39], and then the subjective weights and objective weights are integrated into composite weights using the principle of information entropy [40].

2.4.1. Fuzzy Analytic Hierarchy Process

The Fuzzy Hierarchical Analysis Method (FAHP) was proposed by some scholars to address the limitations of T.L. Saaty’s Hierarchical Analysis Method in practical applications [41]. The process involves first using the AHP to decompose the indicators into multiple hierarchical levels of objectives, forming a judgment matrix. Subsequently, the fuzzy comprehensive analysis method is applied to conduct a detailed analysis of the membership degrees of each indicator, ultimately yielding the evaluation weights. The method proposed in this paper combines the structured advantages of the Hierarchical Analysis Method with the uncertainty handling methods of fuzzy mathematics. The main steps for determining weights using this method are as follows:

(1)

Establish the hierarchical structure of evaluation indicators;

(2)

Construct a fuzzy judgment matrix;

(3)

Calculate the weights of each indicator:

$${\omega }_i={\displaystyle \frac{1}{n}}-{\displaystyle \frac{1}{2\alpha }}+{\displaystyle \frac{{\displaystyle \sum _{k=1}^n{r}_{ik}}}{n\alpha }}, i = 1, 2,\dots , n$$

(2)

${\omega }_i$ for indicator weighting, $\alpha$ = ${\displaystyle \frac{n-1}{2}}$.

(4)

Weight normalization:

$${\omega }_i^{\prime }={\displaystyle \frac{{\omega }_i}{{\displaystyle \sum _{i=1}^n{\omega }_i}}}, i = 1, 2,\dots , n$$

(3)

(5)

Consistency check:

$$CI={\displaystyle \frac{{\lambda }_{\max }-n}{n-1}}$$

(4)

${\lambda }_{\max }$ for the maximum eigenvector of the matrix; n is the order of the matrix.

$$CR={\displaystyle \frac{CI}{RI}}$$

(5)

Look up the corresponding average consistency index RI. When n = 1, RI is 0.00; when n = 2, RI is 0.00; when n = 3, RI is 0.52; when n = 4, RI is 0.89; when n = 5, RI is 1.12; when n = 6, RI is 1.26; when n = 7, RI is 1.36; when n = 8, RI is 1.41; when n = 9, RI is 1.46, providing reference data for the average consistency index. If CR < 1, the judgement matrix is considered to have passed the consistency test.

2.4.2. Entropy Weighting Method

The entropy weight method is a technique for calculating weights based on the degree of data dispersion [42]. According to information entropy theory, the smaller the entropy value, the more significant the comparative role of the evaluation indicator in distinguishing between different samples. During the weight allocation process in constructing the evaluation system, indicators that play a greater comparative role are assigned higher weights [43]. The calculation steps of this method are as follows:

(1)

Construct a groundwater vulnerability evaluation indicator matrix, assuming there are n samples and m evaluation indicators, with the original data matrix being X_ij (i = 1, 2, …, n; j = 1, 2, …, m);

(2)

Indicator normalisation processing:

$$Y_{ij}={\displaystyle \frac{X_{ij}-\min \left(X_j\right)}{\max \left(X_j\right)-\min \left(X_j\right)}}, Y_{ij}={\displaystyle \frac{\max \left(X_j\right)-X_{ij}}{\max \left(X_j\right)-\min \left(X_j\right)}}$$

(6)

$Y_{ij}$ for standardised values, $X_{ij}$ for raw values, $\max \left(X_i\right)$ for the maximum value in the data, $\min \left(X_i\right)$ for the minimum value in the data. Evaluation indicators are divided into positive indicators and negative indicators. Positive indicators are those where a higher score is better, while negative indicators are those where a lower score is better.

(3)

Calculate the proportion of the i-th evaluation indicator in the j-th sample area to be evaluated:

$$R_{ij}={\displaystyle \frac{Y_{ij}}{{\displaystyle \sum _{i=1}^n{Y}_{ij}}}}$$

(7)

(4)

Calculation of entropy value for the j-th evaluation indicator:

$$H_j=-k{\displaystyle \sum _{i=1}^n{R}_{ij}\cdot }\ln R_{ij}$$

(8)

$$k={\displaystyle \frac{1}{\ln n}}$$

(5)

Calculation of the weight of the j-th evaluation indicator:

$$W_j={\displaystyle \frac{1-H_j}{{\displaystyle \sum _{j=1}^x\left(1-H_j\right)}}}$$

(9)

2.4.3. Combination Weight

Combine subjective weights and objective weights into composite weights, and calculate the composite weights of the two methods according to Formula (10).

$$W_{cn}={\displaystyle \frac{{\left({\omega }_{an}\cdot {\omega }_{pn}\right)}^{0.5}}{{{\displaystyle \sum _{n=1}^i\left({\omega }_{an}\cdot {\omega }_{pn}\right)}}^{0.5}}} n=1,2,\dots ,i$$

(10)

The combined weights of the indicators are $W_{cn}=\left(w_{c1},w_{c2},\dots ,w_{ci}\right)$ (i is the number of indicators), ${\omega }_{an}$ are the weights determined by the fuzzy hierarchical analysis method, and ${\omega }_{pn}$ are the weights calculated by the entropy weight method.

2.4.4. Groundwater Depth Factor and Kriging Interpolation

Groundwater depth is a key factor in the DRASTIC model, as it determines the vertical distance that potential contaminants must travel before reaching the aquifer. In this study, groundwater depth data were obtained from hydrogeological survey reports and validated well monitoring records for the Huangshui River Basin. To ensure spatial continuity and accuracy, we applied Kriging interpolation, which is recognized for its statistical robustness and ability to account for spatial autocorrelation. Compared with deterministic interpolation techniques such as Inverse Distance Weighting (IDW), Kriging produced a smoother and more reliable spatial distribution. The interpolated groundwater depth ranged from approximately X m to Y m, with shallower water tables (<Z m) concentrated in low-lying agricultural zones, and deeper levels (>W m) observed in the upstream mountainous area. This parameter was found to exert a strong influence on the overall vulnerability index, underscoring its significance in aquifer protection planning.

2.5. Parameter Rationale and Expected Influence

To enhance interpretability and ensure consistency between the indicator system and the ensuing maps, we summarize the physical meaning and expected influence direction of each parameter in the context of the Huangshui Basin. For water-quality vulnerability, shallower depth to water, greater net recharge, coarser aquifer and soil media, gentler slopes, and land uses associated with higher pollutant loads typically increase the likelihood and rate of downward migration and reduce natural attenuation, therefore raising vulnerability. For water-quantity vulnerability, larger net recharge, higher-permeability aquifer media, greater aquifer thickness, stronger water-bearing capacity, and a higher exploitable modulus generally indicate stronger replenishment and storage buffering under pumping stress and hence lower vulnerability.

To address the subjectivity inherent in the traditional DRASTIC model, we employed a combination of the Fuzzy Analytic Hierarchy Process (FAHP) and the Entropy Weight Method (EWM) to determine the indicator weights. The EWM objectively quantifies the degree of dispersion in the data and reflects the discriminative power of each indicator, while FAHP incorporates expert knowledge with fuzzy logic to handle uncertainty and reduce individual bias. To integrate these two complementary approaches, we adopted a geometric averaging method to derive the final combined weights. This hybrid strategy ensures a balanced representation of both subjective expert judgment and objective data variability, effectively enhancing the scientific robustness and reliability of the evaluation system. Such combination weighting techniques have been widely validated and applied in environmental vulnerability assessments, groundwater risk evaluations, and multi-criteria decision-making contexts.

3. Results and Discussion

3.1. Groundwater Vulnerability Indicator

3.1.1. Groundwater Depth (D)

Groundwater depth is a key indicator for assessing water quality vulnerability. Based on the DRACTIC model principle, the shallower the groundwater depth (i.e., the closer the aquifer is to the ground surface), the shorter the pollutant migration pathways and the shorter the natural purification time, resulting in higher vulnerability and thus higher score values. The specific scoring criteria are shown in Table 2. When applied to the Huangshui River basin, monitoring in 2006 indicated that groundwater depth was primarily shallow to medium (concentrated in the 5.5–9.5 m range), with overall high vulnerability scores. Spatially, vulnerability increased from mountainous areas to river valleys, with high-vulnerability zones (high scores) concentrated in the middle and lower reaches of the river valley plains (e.g., Minhe, Ledu, and Ping’an), while low-vulnerability areas (low scores) are located in the upstream mountainous transition zone (e.g., Haiyan, Chengguan, and Datong). By 2021, although the distribution trend of depth remained similar (with the peak range shifting to 11.5–14 m), the overall depth increased, reflected in the vulnerability scores: the area of high-score zones significantly decreased, the area of low-score zones notably expanded, and the proportion of intermediate-score zones increased. Although the spatial pattern of increasing vulnerability from mountainous areas to river valleys persists, in densely populated river valley water source areas, underground water levels are relatively low due to mining activities, resulting in slightly lower vulnerability scores compared to surrounding regions, as shown in Figure 2.

Table 2. Groundwater vulnerability assessment index classification and scoring criteria.

Water Quality Vulnerability Index Classification and Scoring Criteria
Score	Groundwater Depth (m)	Net Recharge (mm)	Aquifer Media	Soil Media	Terrain Slope (°)	Land Use Type
1	>26	<30			50	Wetland
2	22.5–26	30–45	Argillaceous sand-cobble-gravel	Silty clay	29	Snowland
3	19.5–22.5	45–60			25	Forestland
4	16.6–19.5	60–90	Mud-bearing sand-gravel-cobble	Loamy clay	21	Shrubland
5	14–16.5	90–110		Clay loam	18	Grassland
6	11.5–14	110–140	Sand-gravel	Sandy loam	15	Wasteland
7	9.5–11.5	140–180			12	Water body
8	7.5–9.5	180–210	Sand-gravel-cobble	Silty gravelly loam	9	Arable land
9	5.7–7.5	210–240			6	Artificial surface
10	<5.7	>240	Gravel-cobble	Fine silty sand	3
Water quantity vulnerability index classification and scoring criteria
Score	Net supply (mm)	Aquifer media	Thickness of aquifer (m)	Water abundance grade	Groundwater exploitable modulus (×104 m3/km2·a)
1	>240		80
2	210–240	Gravel-cobble	60	1	>3.5
3	180–210		50
4	140–180	Sand-gravel-cobble	40	2	3–3.5
5	110–140		30
6	90–110	Sand-gravel	25	3	1–3
7	60–90		18
8	45–60	Mud-bearing sand-gravel-cobble	15	4	0.5–1
9	30–45		12
10	<30	Argillaceous sand-cobble-gravel		5	<0.5

Table 2. Groundwater vulnerability assessment index classification and scoring criteria.

Water Quality Vulnerability Index Classification and Scoring Criteria
Score	Groundwater Depth (m)	Net Recharge (mm)	Aquifer Media	Soil Media	Terrain Slope (°)	Land Use Type
1	>26	<30			50	Wetland
2	22.5–26	30–45	Argillaceous sand-cobble-gravel	Silty clay	29	Snowland
3	19.5–22.5	45–60			25	Forestland
4	16.6–19.5	60–90	Mud-bearing sand-gravel-cobble	Loamy clay	21	Shrubland
5	14–16.5	90–110		Clay loam	18	Grassland
6	11.5–14	110–140	Sand-gravel	Sandy loam	15	Wasteland
7	9.5–11.5	140–180			12	Water body
8	7.5–9.5	180–210	Sand-gravel-cobble	Silty gravelly loam	9	Arable land
9	5.7–7.5	210–240			6	Artificial surface
10	<5.7	>240	Gravel-cobble	Fine silty sand	3
Water quantity vulnerability index classification and scoring criteria
Score	Net supply (mm)	Aquifer media	Thickness of aquifer (m)	Water abundance grade	Groundwater exploitable modulus (×104 m3/km2·a)
1	>240		80
2	210–240	Gravel-cobble	60	1	>3.5
3	180–210		50
4	140–180	Sand-gravel-cobble	40	2	3–3.5
5	110–140		30
6	90–110	Sand-gravel	25	3	1–3
7	60–90		18
8	45–60	Mud-bearing sand-gravel-cobble	15	4	0.5–1
9	30–45		12
10	<30	Argillaceous sand-cobble-gravel		5	<0.5

3.1.2. Net Recharge (R)

Groundwater net recharge has an opposite effect on water quality and quantity vulnerability: increased recharge accelerates pollutant dispersion (increasing water quality vulnerability) while enhancing system resilience (reducing water quantity vulnerability). Based on the DASTIC model scoring rules, in 2006, water quality vulnerability in the Huangshui River basin exhibited a “high in river valleys, low in mountainous areas” pattern—the middle reaches of the river valley (such as the main stem of the Huangshui River) have high vulnerability (8–10 points) due to loose aquifers with high permeability and large recharge rates; mountainous areas, constrained by terrain (low precipitation infiltration), account for 49.2% of the region with low vulnerability (1–3 points). Water quantity vulnerability exhibited a ‘high in mountainous areas, low in river valleys’ pattern, with mountainous areas (such as Huan Yuan and Haiyan) having weak recharge capacity (9–10 points, high vulnerability), while river valleys (around Xining) were supported by diverse recharge sources (1–3 points, low vulnerability). Specific scoring criteria are shown in Table 2. By 2021, the total water supply in the entire basin had significantly increased (with the proportion in the 60–110 mm/a range reaching 60%), leading to an overall rise in water quality vulnerability: the high-vulnerability zone expanded to 780 km² (including river valleys in Datong and Xining), the medium-vulnerability zone (4–7 points) increased to 11,244 km² and became the dominant zone, and low vulnerability areas sharply decreased to 9.54%; water quantity vulnerability decreased overall: high vulnerability areas (8–10 points) retreated to the basin margins (25.16% of the total area), medium vulnerability areas (4–7 points) expanded to 70% (mountainous areas on both sides of river valleys), and low vulnerability areas slightly increased. Spatially, water quality vulnerability remains concentrated in river valleys as high-risk cores (due to strong hydrodynamic diffusion), as shown in Figure 2; water quantity vulnerability is most pronounced in mountainous areas (due to low terrain-controlled water retention efficiency), as shown in Figure 3.

3.1.3. Aquifer Medium Type (A)

Aquifer media exert a bidirectional influence on groundwater vulnerability through differences in permeability: from a water quality perspective, larger particles (e.g., gravel) and more developed pores (higher permeability) result in faster contaminant migration and higher water quality vulnerability (scoring up to 10 points). Conversely, fine-grained media with high clay content (e.g., clayey gravel) impede contaminant transport and lower vulnerability (scoring 2 points). From a water quantity perspective, the logic is reversed: high-permeability media (gravel and cobble) facilitate rapid recharge and post-extraction runoff recovery, resulting in low water quantity vulnerability (scoring 2 points), while low-permeability media (muddy sand and gravel) cause slow water flow and difficult water level recovery, leading to significantly higher water quantity vulnerability (scoring 10 points), specific scoring criteria are shown in Table 2. The Huangshui River basin is dominated by gravel and sand (29.14%) and clayey gravel and sand (25.42%) as the main media, with a spatial distribution showing a regular correlation: high water quality vulnerability scoring zones (8–10 points) are concentrated in the river valley core areas (such as the Datong and Zhonghu tributary regions) and mountainous peripheries, due to the formation of coarse-grained, highly permeable channels by river alluvial deposits; however, high water quantity vulnerability zones (8–10 points, accounting for 40.7% of the entire basin) are located in the upstream mountainous areas and hill belts, primarily composed of clayey, low-permeability media. This spatial antagonistic feature indicates that while the coarse-grained medium in the river valley exacerbates water quality risks (scores of 8–10), its high permeability supports water quantity recovery, resulting in low water quantity vulnerability (score of 2). Conversely, the fine-grained medium in the mountainous areas impedes pollutant dispersion (low water quality vulnerability) but, due to its poor permeability, hinders water quantity recovery (high water quantity vulnerability), as shown in Figure 2 and Figure 3.

3.1.4. Soil Medium Type (S)

Soil media regulate the migration of pollutants through their particle structure and density. When soil layers are thick and particles are closely packed, water flow paths become tortuous and complex, significantly prolonging the infiltration time of pollutants and reducing the risk of groundwater contamination, thereby decreasing vulnerability. Conversely, vulnerability increases under less favourable conditions. The soil in the Huangshui River basin is primarily clay loam (53.04%, mainly grey calcareous soil and black calcareous soil) and sandy loam (31.9%, mainly chestnut calcareous soil and dark chestnut calcareous soil). Based on particle characteristics and clay content, it is classified into six categories and assigned water quality vulnerability scores (10 points for high permeability, 2 points for low permeability), with specific scoring criteria shown in Table 2. Spatial distribution indicates that high-vulnerability zones (score 10) are concentrated in the northern Daban Mountain and southern Laojishan Mountain areas with exposed bedrock, where extremely high soil permeability (fine sandy loam) creates rapid pollution pathways; moderate-vulnerability zones (score 5, clay loam) cover the Xining, Datong, and other major river valleys, with moderate permeability; low-vulnerability zones (score 2, clayey silt) are distributed in the eastern and northern high-altitude mountainous areas, where clay minerals are concentrated, resulting in extremely low permeability. Notably, high-altitude mountainous areas (rated 8–10) and river valley belts (rated 4–5) exhibit a significant permeability gradient, reflecting the control of topography-weathering processes on soil retention capacity: cold mountainous areas promote clay formation (low permeability), while river valley deposits and bedrock zones favour the development of highly permeable soils, as shown in Figure 2.

3.1.5. Terrain Slope (T)

Terrain slope significantly influences water quality vulnerability by controlling surface water infiltration processes—the steeper the slope (e.g., >25°), the faster precipitation forms surface runoff, resulting in shorter pollutant retention times and reduced infiltration rates, thereby lowering groundwater vulnerability (scores 1–5); conversely, gentler slopes (e.g., <3°) prolong surface water retention times and facilitate adequate infiltration, significantly increasing vulnerability (scores 9–10). Specific scoring criteria are detailed in Table 2. The Huangshui River basin has complex topography. Slope analysis based on a 30-metre resolution DEM shows that the 6–9° slope range has the highest proportion (18.36%), followed by steep slopes (29–50°, 12.82%), while gentle slopes (0–3°) account for only 9.8%. The spatial scoring pattern exhibits a three-tiered differentiation: (1) High-vulnerability zones (9–10 points) are concentrated in river valley plains, where slopes < 6° provide terrain advantages for maximising infiltration; (2) Moderate-vulnerability zones (6–8 points) are distributed in the transition zones between river valleys and mountains (e.g., around Huanyuan and Datong), where slopes of 6–15° reduce infiltration capacity; (3) low-vulnerability zones (1–5 points) occupy the high-altitude mountainous areas surrounding the watershed, where steep slopes > 25° strongly inhibit pollutant infiltration. This ‘high vulnerability in river valleys and low vulnerability in mountainous areas’ gradient pattern fundamentally reflects the regulation of hydrodynamic processes by slope: gentle terrain prolongs the contact time of pollutants, while steep terrain accelerates their removal, as illustrated in Figure 2.

3.1.6. Land Use Type (L)

Land use types regulate groundwater vulnerability through the intensity of human activities. Human-dense areas (arable land, artificial surfaces) exhibit high vulnerability (scores of 8–9) due to wastewater discharge, agricultural irrigation, and industrial pollution, while natural vegetation areas (forests, grasslands) reduce pollution risks through ecological barriers (scores of 3–5). Specific scoring criteria are presented in Table 2. The Huangshui River basin is dominated by grasslands (70.27% in 2006, increasing to 73.04% in 2021), followed by farmland (21.49% in 2006, decreasing to 17.32% in 2021). In 2006, spatial differentiation of vulnerability was significant: The river valley areas with artificial land surfaces (such as Xining) formed a 9-point extremely high vulnerability core; farmland (river valley plains and gentle slope mountain areas) constituted an 8-point high vulnerability zone; mountain grasslands, which account for the majority of the basin, received a 5-point moderate rating due to vegetation coverage; and remote high-altitude natural areas (dominated by forests and shrubs) were classified as 3-point low vulnerability zones. By 2021, human activities had intensified and reshaped the landscape: the expansion of artificial surfaces significantly enlarged the 9-point highly vulnerable zone in river valleys; although the reduction in farmland caused the 8-point zone to shrink, this was offset by the rise in urbanisation vulnerability; the enhanced ecological functions of mountain grasslands drove the expansion of the 5-point moderately vulnerable zone, reflecting local improvements in vulnerability. The 15-year evolution revealed the core contradiction: river valley urbanisation (artificial land surface +1.12%) and agricultural intensification continued to increase pollution risks, while mountainous area ecological protection (grassland +2.77%) enhanced groundwater protection capacity, highlighting the decisive role of human activity gradients in vulnerability, as shown in Figure 2.

3.1.7. Aquifer Thickness (H)

Aquifer thickness (distance between the water table and the impermeable base) directly determines groundwater storage capacity and recharge potential—the greater the thickness (e.g., 60–80 m), the greater the storage capacity and lateral recharge potential, and the lower the water volume vulnerability (score 1); the smaller the thickness (e.g., <15 m), the weaker the regulatory capacity, and the significantly higher the water volume vulnerability (score 9). Specific evaluation criteria are presented in Table 2. In the Huangshui River basin, thickness primarily ranges from 15 to 25 m (accounting for 53.91% of the entire basin), with the highest proportion (30.73%) in the 15–18 m range, followed by 23.18% in the 18–25 m range, while aquifers thicker than 25 m account for only 25.2% (including the extremely thick zone of 60–80 m at 2.03%). Spatial vulnerability exhibits a ‘low in mountainous areas, high in river valleys’ pattern: (1) High-vulnerability zones (8–9 points, accounting for 21.85%) are concentrated in river valley sections such as Ping’an, Ledu, and Minhe, where local uplift of the basement has caused aquifers to thin (<18 m); (2) Moderate-vulnerability zones (4–7 points, accounting for 45.1%) are distributed in Haiyan, Huangyuan to Xining, with a thickness of 15–40 m providing limited regulatory capacity; (3) low-vulnerability zones (1–3 points, accounting for 12.6%) are located in the Daxing and Menyuan depression areas, where aquifers with a thickness of 60–80 m offer strong recovery resilience. This distribution reveals the core geological control: the uplifted basement zone weakens water storage capacity, while the depression zones enhance water resource resilience, as shown in Figure 3.

3.1.8. Groundwater Water-Bearing Capacity (W)

Aquifer water-bearing capacity is characterised by single-well discharge rate—the higher the discharge rate (e.g., >5000 m³/d), the stronger the water yield capacity and recovery potential, and the lower the water quantity vulnerability (score: 2 points); conversely, the lower the discharge rate (e.g., <10 m³/d), the greater the risk of depletion, and the significantly higher the vulnerability (score: 10 points). Specific scoring criteria are presented in Table 2. The Huangshui River basin is dominated by medium to low water-bearing zones: the 100–1000 m³/d range (30.96%) and the 10–100 m³/d range (33.7%) together account for 64.66%, while high-water-rich areas (>1000 m³/d) account for only 7.35% (of which extremely high-water-rich areas > 5000 m³/d account for just 1.07%). Spatial vulnerability exhibits a significant ‘high in mountainous areas, low in river valleys’ pattern: (1) High-vulnerability zones (8–10 points, accounting for 92.64%) are widely distributed in high-altitude mountainous areas (spring water discharge < 100 m³/d), particularly in the northern mountainous regions (rated 8) and the middle and lower reaches of river valleys (rated 10), constrained by steep terrain and low-permeability aquifers; (2) Moderately vulnerable zones (2–6 points) are concentrated in river valley plains, with core river valleys such as Xining, Duoba, and Datong forming extremely low-vulnerability zones (score 2) due to their flat terrain and favourable water collection conditions, forming extremely low-vulnerability zones (score 2, with groundwater discharge > 5000 m³/d), while the marginal mountainous areas with gentle slopes are predominantly medium-vulnerability zones (scores 4–6). The distribution of groundwater-rich areas is fundamentally controlled by the dual factors of aquifer permeability (bedrock vs. loose sediments) and topography (rapid mountain runoff vs. strong river valley runoff), as detailed in Figure 3.

3.1.9. Groundwater Exploitable Modulus (M)

The exploitable modulus (annual exploitable volume per unit area) directly characterises the groundwater development potential and system resilience—the larger the modulus (e.g., >3.5 × 10⁴ m³/km²·a), indicates high resource abundance, strong sustainability of extraction, and lower water quantity vulnerability (score 2 points); a smaller modulus (e.g., <0.5 × 10⁴ m³/km²·a) indicates a high risk of resource depletion, and significantly increased vulnerability (score 10). Specific scoring criteria are presented in Table 2. In 2006, the Huangshui River basin was dominated by medium to high modulus values: 1–3.5 × 10⁴ m³/km²·a (54.17%) and >3.5 × 10⁴ m³/km²·a (43.46%), together covering 97.63% of the area. Spatial vulnerability exhibits a ‘low in basins, high at edges’ pattern: the Xining Basin (Daxing, Zhonghu, etc.) forms a large area of low vulnerability (score 2, 43.46%) due to abundant resources; the middle and lower reaches (Haiyan to Minhe) constitute a medium-risk zone (score 6, 54.17%); and the marginal areas of Guide and Jianzha have scattered high-risk zones (scores 8–10, 0.93%). By 2021, long-term exploitation has led to an overall decline in the modulus: the high-potential zone (>3.5 × 10⁴ m³/km²·a) has disappeared, with the dominant type shifting to 3–3.5 × 10⁴ m³/km²·a (57.95%, rated as 4), indicating a general increase in vulnerability. Specifically: the area around Xining decreased from 2 points to 4 points (57.95%); the Ping’an-Minhe Basin remained at 6 points, maintaining moderate vulnerability (20.45%); Huanxian, Hualong, and other areas rose to 8 points, indicating high vulnerability (20.99%); only the Qinghai Lake return flow area retained a small area of 2 points, indicating low vulnerability (0.6%). Fifteen years of evolution reveal the core crisis: human extraction has continuously weakened groundwater resilience, particularly in the Xining Basin, where resource depletion has triggered vulnerability upgrades. It is urgent to regulate development intensity to maintain water resource sustainability, as detailed in Figure 3.

3.2. Vulnerability Indicator Weighting Results

Each evaluation indicator was classified according to the scoring criteria in Table 2. Then, using the spatial analysis function of ARCGIS 10.8 software, a fishing net was created with a grid size of 3 km × 3 km. Using the editing function, incomplete grids were deleted to obtain 1663 square grids covering the entire basin. Using the spatial connection tool, the grid layer was spatially connected with the raster and vector layers. The ‘Raster to Point’ function was then used to assign the values on the grid surfaces to grid points. Finally, the score values for each indicator were exported for each grid point, and the score feature values were used to represent the values within each cell. This provides the data foundation for the subsequent dimensionless processing of weights in the entropy method.

3.2.1. Water Quality Vulnerability Indicator Weights

According to Table 1, a preliminary judgement matrix was established by comparing the relative importance of the six indicators in pairs. The six indicators of water quality vulnerability were entered into SPSSAU. Referring to the DRASTIC model for the judgement of groundwater vulnerability indicator weights, and combining the actual conditions of the Huangshui River basin, the scale was continuously adjusted to obtain a fuzzy complementary judgement matrix that passed the consistency test, as shown in

Table 3. Water quality fuzzy complementary judgement matrix.

Indicator	Groundwater Depth	Net Recharge	Aquifer Medium	Soil Medium	Terrain Slope	Land Use Type
Groundwater depth	0.5	0.4	0.3	0.2	0.3	0.3
Net recharge	0.6	0.5	0.4	0.6	0.6	0.6
Aquifer medium	0.7	0.6	0.5	0.6	0.4	0.6
Soil medium	0.8	0.4	0.4	0.5	0.6	0.3
Terrain slope	0.7	0.4	0.6	0.4	0.5	0.4
Land use type	0.7	0.4	0.4	0.7	0.6	0.5

.

Based on the above fuzzy complementary judgement matrix and Equations (2) and (3), the weights of the six indicators (groundwater depth, net recharge, aquifer medium type, soil medium type, terrain slope, and land use type) and the normalised weights of these indicators were calculated. The weights and normalised weights are identical, with the following values: groundwater depth 13.333%, net recharge 17.667%, aquifer medium type 18.000%, soil medium type 16.667%, terrain slope 16.667%, and land use type 17.667%. These values are used to display the weights of each indicator and the results of the normalisation process. According to Equation (5), the consistency check index (CR) value is calculated to be 0.0948, which is less than 0.1. Therefore, the judgment matrix passes the consistency check, and the weights of the indicators are considered reasonable.

According to Section 3.2, the indicator values of the 1663 evaluation units in the study area were statistically analysed. Using the entropy weight method, the water quality vulnerability evaluation index was calculated, with the following components: groundwater depth (weight 0.087), net recharge (weight 0.097), terrain slope (weight 0.121), land use type (weight 0.204), soil type (weight 0.090), and aquifer medium type (weight 0.401).

According to Equation (10), for the weighting of water quality vulnerability indicators, the geometric mean of the six indicators calculated using the fuzzy hierarchical analysis method and the entropy weight method are as follows: the combined weight of groundwater depth is 0.11, the combined weight of net recharge is 0.14, the combined weight of aquifer medium type is 0.15, soil medium type composite weight 0.19, terrain slope composite weight 0.13, and land use type composite weight 0.28. Land use type has the highest weight at 0.28, followed by soil medium at 0.19, and the lowest weight is groundwater depth at 0.11. This indicates that human activities have the greatest impact on groundwater quality vulnerability in the Huangshui River basin.

3.2.2. Water Vulnerability Indicator Weighting

According to Table 1, the five indicators of water vulnerability were entered into SPSSAU. Referring to the DRASTIC model for the weighting of groundwater vulnerability indicators and combining the actual conditions of the Huangshui River basin, a fuzzy complementary judgement matrix was made, as shown in

Table 4. Water quantity fuzzy complementary judgement matrix.

Indicator	Net Recharge	Aquifer Thickness	Aquifer Medium	Groundwater Water-Bearing Capacity	Groundwater Exploitable Modulus
Net recharge	0.5	0.7	0.8	0.6	0.6
Aquifer thickness	0.3	0.5	0.6	0.2	0.4
Aquifer medium	0.2	0.4	0.5	0.4	0.3
Groundwater water-bearing capacity	0.4	0.8	0.6	0.5	0.3
Groundwater Exploitable Modulus	0.4	0.6	0.7	0.7	0.5

.

Based on the above fuzzy complementary judgement matrix and Equations (2) and (3), the weights and normalised weights of the five water vulnerability indicators (net recharge, aquifer thickness, aquifer medium type, groundwater salinity, and groundwater exploitable modulus) are obtained. The weights and normalised weights are consistent, with the following values: net recharge 23.500%, aquifer thickness 17.500%, aquifer medium type 16.500%, groundwater salinity 20.500%, and groundwater extractivity 22.000%. These values are used to present the weights of each indicator and the results of the normalisation process. According to Equation (5), the consistency check index CR value is calculated to be 0.0948, which is less than 0.1. Therefore, the decision matrix passes the consistency check, and the weights of the indicators are reasonable.

According to Section 3.2, the characteristic values of the water quantity vulnerability indicators for the 1663 evaluation units in the study area were statistically analysed. Following the steps of the entropy weight method, the indicators were input into the calculation software to obtain the water quantity vulnerability evaluation weight indicators, which include groundwater exploitable modulus (0.124698), aquifer thickness (0.180284), net recharge (0.205737), water richness grade (0.256361), and aquifer medium type (0.23292). Among these, the water richness grade has the highest value of 0.256, followed by groundwater recharge at 0.206, and the groundwater recharge coefficient has the lowest value of 0.1245.

According to Equation (10), the geometric mean weights of the five indicators for water quantity vulnerability calculated using the fuzzy hierarchical analysis method and the entropy weight method are as follows: net recharge weight 0.17, aquifer thickness weight 0.18, aquifer medium type weight 0.19, groundwater salinity weight 0.23, groundwater extractivity coefficient composite weight 0.23. The groundwater salinity and extractivity coefficient weights are the largest, both at 0.23, indicating that human activity factors also have a significant impact on water quantity vulnerability.

3.3. Sensitivity Analysis

The single-parameter sensitivity [44,45] quantifies the actual impact of parameter changes on the results of groundwater vulnerability assessments. It is widely used to evaluate the influence of each parameter on groundwater vulnerability. The formula for calculating the effective weight of parameters is as follows:

$$W={\displaystyle \frac{p_r \times p_w}{DI}}$$

(11)

W represents the effective weight of the parameter; p_r represents the score of the indicator, p_w represents the theoretical weight of the indicator; DI represents the vulnerability score.

In ArcGIS, the field calculator was used to calculate the effective weights of water quality and water quantity vulnerability indicators using a single-parameter sensitivity analysis method. The comparison with the combined weights is shown in

Table 5. Single-parameter sensitivity analysis weights for water quality and quantity vulnerability.

Water Quality Evaluation Indicators	Theoretical Weight	Effective Weighting
		Maximum Value	Minimum Value	Average Value
Groundwater depth	0.11	0.210191	0.016975	0.11
Net recharge	0.14	0.276134	0.040698	0.16
Aquifer medium	0.15	0.306748	0.044118	0.18
Soil medium	0.19	0.310966	0.052342	0.18
Terrain slope	0.13	0.273109	0.038179	0.16
Land use type	0.28	0.501119	0.130638	0.32
Water Quantity Evaluation Indicators	Theoretical Weight	Effective Weighting
		Maximum Value	Minimum Value	Average Value
Net recharge	0.17	0.327196	0.043628	0.19
Aquifer thickness	0.19	0.375833	0.051491	0.21
Aquifer medium	0.18	0.336898	0.025714	0.18
Groundwater water-bearing capacity	0.23	0.459082	0.068554	0.26
Groundwater Exploitable Modulus	0.23	0.371717	0.076412	0.22

.

The sensitivity analysis of water quality vulnerability indicators’ weights shows that land use type has the greatest impact on water quality vulnerability, followed by aquifer medium type and soil medium type, each accounting for 18%, followed by net recharge, terrain slope, and groundwater depth; the six indicators calculated using the combination of fuzzy cluster analysis and entropy weighting methods account for 23%, 23%, 19%, 18%, and 17%, respectively, land use type has the highest weight, followed by soil medium type, aquifer medium type, net recharge, terrain slope, and groundwater depth. This is consistent with the results of the single-parameter sensitivity analysis, indicating that the calculated theoretical weights can be used as weights for evaluating water quality vulnerability indicators.

The results of the effective weight calculation show that there are significant differences in the weights of various indicators. Water richness has the greatest impact on water quantity vulnerability, followed by groundwater exploitable modulus, aquifer medium type, net recharge, and aquifer thickness, accounting for 26%, 22%, 21%, 19%, and 18%, respectively. The combined weights obtained using fuzzy stratification analysis and entropy weighting methods indicate that water richness and groundwater exploitable modulus have the largest weights, followed by aquifer medium type, aquifer thickness, and net recharge, accounting for 23%, 23%, 19%, 18%, and 17%, respectively. The results are generally consistent with the single-parameter sensitivity analysis, with net recharge and aquifer thickness showing some differences. However, the effective weights calculated for aquifer thickness are the same as those obtained from the combined weights. Therefore, it can be concluded that the weights of the water quantity vulnerability indicators are generally consistent with the actual influence of each indicator on groundwater water quantity vulnerability, validating the rationality of the weight allocation.

3.4. Groundwater Vulnerability Assessment Zoning and Protection Measures

3.4.1. Trend Analysis of Water Quality Vulnerability Zoning and Evaluation Results

Based on the score distribution maps for each indicator of water quality vulnerability in Section 3.1, the score maps for groundwater depth, net recharge, terrain slope, and land use are raster data, while the score maps for aquifer medium type and soil medium type are vector data. A comprehensive score is calculated for the six indicators. First, the four raster datasets should be converted into vector data using the ‘Raster to Polyhedra’ function. Then, using the ‘Analysis Tools’ module in ARCGIS’s ‘ArcToolbox,’ the vector maps of the seven evaluation indicators should be overlaid. Finally, by intersecting the attribute values, the groundwater quality vulnerability index overlay map is generated. The comprehensive vulnerability scores obtained are categorised using the natural breakpoint method. Scores are divided into five categories from low to high: scores < 4.5 are classified as low vulnerability, 4.5–5.1 as moderate vulnerability, 5.1–5.7 as high vulnerability, 5.7–6.3 as very high vulnerability, and >6.3 as extremely high vulnerability. The zoning results for 2006 and 2021 are shown in Figure 4.

The 2006 water quality vulnerability zoning map indicates that low-vulnerability areas are primarily distributed in the mountainous regions of the upper reaches of the basin and parts of Minhe County, where the terrain is mountainous with high elevations and minimal human activity, resulting in limited groundwater pollution. Moderately low-vulnerability areas are located on both sides of the low-vulnerability zones, characterized by steep terrain and precipitation primarily discharged as surface runoff, with limited human activity, with relatively low vulnerability; moderately vulnerable areas are located along the mountainous edges of river valleys such as Huan Yuan, Duo Ba, Ping An, and Le Du, with varied terrain and some pollutant discharge and localized human activity; highly vulnerable areas surround the periphery of high-vulnerability zones, primarily near river systems, forming a gentle slope zone where river valleys transition to mountains. Agricultural pollution and urban wastewater migrate with surface runoff into river valleys and infiltrate groundwater, increasing vulnerability. High-vulnerability areas are primarily river valley plains with high overlap with river systems, located in the core of the Huangshui River valley plain, characterized by flat terrain and dense human activity.

Figure 4 shows that low-vulnerability areas decreased significantly in 2021, mostly distributed in the northern mountainous areas of the watershed, with a small portion located in the transitional zones of mountainous river valleys near Xining, Ping’an, and Minhe River Valley; urbanisation has led to increased agricultural and industrial activities in densely populated areas, exacerbating pollutant emissions and raising groundwater vulnerability; low-vulnerability areas have expanded downstream; and the area of low-vulnerability regions has also decreased significantly, mainly distributed in the mountainous areas on the periphery of the river basin and the high-altitude areas of the river basin, such as Haiyan County, Huangzhong District, and Hualong Hui Autonomous County; the area of medium-vulnerability regions has increased significantly and is distributed relatively scattered, mostly located in mountainous areas near rivers; high-vulnerability regions are distributed around the periphery of high-vulnerability regions and are relatively scattered, mainly concentrated on the edge of the Huangshui River valley plain. Agricultural activities and urban expansion have led to the infiltration of pollutants into groundwater, exacerbating vulnerability; high-vulnerability areas have increased significantly compared to 2006, with dense human activities and concentrated pollutant emissions. Except for the river valley area, the mountainous areas on both sides of the river valley and those near the river valley have all become high-vulnerability areas.

According to the area distribution of water quality vulnerability zoning maps for 2006 and 2021 shown in Figure 5, overall, the groundwater quality vulnerability of the Huangshui River in 2021 was generally higher than that in 2006. In 2006, the regions with the largest vulnerability areas were the low vulnerability zone and the relatively low vulnerability zone, accounting for 27.27% and 26.84% of the basin area, respectively, with areas of 4382.5 km² and 4313.7 km².

In 2021, the regions with the same vulnerability levels accounted for 2.04% and 14.45% of the basin area, respectively, with areas of 327 km² and 2275 km². In 2006, the areas of medium vulnerability, relatively high vulnerability, and high vulnerability regions accounted for 16.86%, 18.32%, and 10.7% of the basin area, respectively. In 2021, the areas of these three vulnerability regions all increased. Among them, the high vulnerability region accounted for the highest proportion in 2021, reaching 31.57%, the medium vulnerability region rose to 24.86%, and the relatively high vulnerability region increased to 27.39%.

3.4.2. Trend Analysis of Water Quantity Vulnerability Zoning and Evaluation Results

For the five indicators selected for water quantity vulnerability, based on the score distribution plots for each indicator in Section 3.1, the raster data were vectorised. The intersect tool was then used to perform a comprehensive overlay analysis for each indicator. After multiplying each indicator by its corresponding weight, the comprehensive water quantity vulnerability scores for 2006 and 2021 were generated. Based on the score intervals, the score intervals were divided into five grades using the natural breakpoint method, from lowest to highest: <4.46 is the low vulnerability zone, 4.46–5.53 is the moderately low vulnerability zone, 5.53–6.45 is the moderate vulnerability zone, 6.45–7.38 is the moderately high vulnerability zone, >7.38 is the high vulnerability zone. This resulted in the water quantity vulnerability zoning map of the Huangshui River basin, as shown in Figure 6.

Figure 6 indicates that in 2006, the overall vulnerability of the watershed was predominantly moderate, with areas of high and relatively high vulnerability accounting for a higher proportion than those of low and relatively low vulnerability. For mountainous areas, regions with relatively high and high vulnerability were primarily distributed in the mountainous areas of the middle and lower reaches of the watershed and the mountainous areas along both sides of the river in the southern part of the upper reaches. These regions have limited groundwater resources in the river valleys, resulting in relatively high vulnerability. Low and very low vulnerability areas were mainly distributed in the mountainous areas in the middle and northern parts of the basin, where groundwater in the river valleys is abundant, resulting in lower vulnerability in the surrounding mountainous areas. For the river valleys, areas with high and medium vulnerability were concentrated in the upper reaches of the Huangshui River (Haiyan, Huangyuan) and the lower reaches (Ping’an, Ledu, Minhe) of the Huangshui River, while low and relatively low vulnerability areas are primarily distributed in the water-rich, flat areas of the middle reaches of the Huangshui River (Duoba, Zongzhai, Xining, Zhonghu, Datong).

In 2021, a significant trend emerged, with an increase in the proportion of areas with higher vulnerability and lower vulnerability, and a decrease in areas with high vulnerability and low vulnerability. Overall vulnerability concentrated in lower, medium, and higher vulnerability zones. Low-vulnerability areas remained concentrated in the flat areas along the middle reaches of the Huangshui River and to the north, but the area of low-vulnerability zones decreased significantly, while the area of lower-vulnerability zones increased slightly. The distribution of medium-vulnerability zones remained largely unchanged, with areas remaining largely the same as in 2006. High-vulnerability areas expanded, extending to the northern part of the basin around the Baoku River and the southern mountainous areas previously classified as high-vulnerability zones. High-vulnerability areas are primarily concentrated in the mountainous regions of the upper basin (Haiyan and Huanxian) and parts of the lower river valleys (Minhe). In Ping’an and Ledu districts, high-vulnerability areas are more dispersed, mainly in mountainous areas near valleys.

The area proportions of each vulnerability level in 2006 and 2021 are shown in Figure 7. It indicates that the water quantity vulnerability of the Huangshui River Basin presents temporal and spatial dynamic changes. From 2006 to 2021, the distribution of vulnerability levels tended to be more complex, with the areas of low vulnerability and high vulnerability decreasing, while the areas of relatively low vulnerability and relatively high vulnerability increasing.

In 2006, the medium vulnerability area in the basin accounted for 32.3% with an area of 5185 km². The low vulnerability area had the smallest proportion, accounting for 6.24% with an area of 1002 km². The relatively high and high vulnerability areas together accounted for 48.18% with a total area of 7732 km². Therefore, the groundwater quantity vulnerability in the basin was mainly dominated by medium and high vulnerability. In 2021, the relatively high vulnerability area had the largest proportion in the basin, which increased significantly compared with 2006, rising from 25.85% to 33.86% with an area of 5435 km². The proportion of the medium vulnerability area was 32.48%, showing little change compared with 2006. The proportion of the low vulnerability area dropped to 3.65% with an area of only 586 km². The proportion of the high vulnerability area decreased from 22.3% to 14.02% with the area reducing to 2251 km². Overall, the water quantity vulnerability in the basin in 2021 was mainly medium and relatively high vulnerability. The relatively low vulnerability area increased slightly, and the high vulnerability area decreased significantly, indicating that the ecological protection and governance measures for the high vulnerability areas in the basin have achieved initial results.

4. Conclusions

This study focuses on groundwater in the Huangshui River basin, extensively collecting data and information on the natural geography, surface hydrology, and hydrogeology of the study area. The years 2006 and 2021 were selected as representative years. An improved DRASTIC model was used to establish a dual-indicator system for water quality and quantity, and groundwater vulnerability was assessed under changing environmental conditions. The main conclusions are as follows:

(1)

An improved DRASTIC model was constructed, proposing a dual-level indicator system for water quantity and water quality. The groundwater water quality evaluation indicators include six items, while the groundwater water quantity vulnerability indicators include five items. For the dual attributes of water quality and water quantity, scoring intervals were defined for the indicators, with scores ranging from 1 to 10, and the distribution characteristics of each interval were analysed.

(2)

The combined weights of the corresponding indicators were calculated using fuzzy hierarchical analysis and entropy weighting methods. The weights corresponding to the groundwater water quality vulnerability indicators D, R, A, S, T, and L are 0.11, 0.14, 0.15, 0.19, 0.13, and 0.28, respectively; The weights corresponding to the groundwater quantity vulnerability indicators R, A, H, W, and M are 0.17, 0.18, 0.19, 0.23, and 0.23, respectively. Vulnerability score maps for each interval of the single indicators were drawn. Through single-parameter sensitivity analysis, the results showed good consistency, and the weight calculation results were reasonable.

(3)

Comprehensive evaluation of groundwater quality vulnerability and quantity vulnerability, with vulnerability zoning maps drawn. In 2006, the areas of high, moderately high, moderate, moderately low, and low vulnerability zones in the Huangshui River basin accounted for 10.7%, 18.32%, 16.86%, 26.84%, and 27.27%, respectively; by 2021, the area of high vulnerability zones had increased to 31.57%, indicating an overall upward trend in groundwater quality vulnerability. The results of the groundwater quantity vulnerability assessment indicate that in 2006, the areas classified as high, moderately high, moderate, moderately low, and low vulnerability accounted for 22.33%, 25.85%, 32.30%, 13.28%, and 6.24% of the basin, respectively. In 2021, the basin’s water quantity vulnerability was primarily characterised by moderate and high vulnerability, with a significant reduction in high-vulnerability areas, indicating a decreasing trend in groundwater quantity vulnerability in the Huangshui River basin.

This study constructed an indicator system and conducted a zoned evaluation of groundwater quality vulnerability and quantity vulnerability in the Huangshui River Basin based on an improved DRASTIC model. However, there are certain limitations in the temporal scope and validation of the results. To ensure the sustainable management of groundwater resources in the region, future research should focus on enhancing real-time monitoring technologies for dynamic evaluations, incorporating long-term hydrological data to track changes over time. Additionally, verification of evaluation results through collaboration with relevant departments, such as water conservancy, land resources, and environmental protection agencies, will help ensure that the model’s predictions align with actual field conditions. Furthermore, future studies should consider the influence of climate change and human activities, including urbanization and agriculture, to provide more accurate assessments and inform more effective management strategies for groundwater resources in the region.