Selection of Hydrologically Vulnerable Areas in Urban Regions Using Techniques for Order Preference by Similarity to Ideal Solution

Abstract

Hydrologically vulnerable areas should be identified for sustainable urban watershed management, flood mitigation, and climate-resilient infrastructure planning. However, assessing hydrological vulnerability in complex urban environments requires a comprehensive framework that integrates hydrological components and considers spatial heterogeneity. Thus, this study proposes an objective, data-driven method for identifying hydrologically vulnerable areas in urban regions using multicriteria decision-making (MCDM). The MCDM technique is used to rank the hydrological health of subwatersheds in an urbanizing watershed. Entropy-based weights are assigned to key hydrological indicators, which are computed using the soil and water assessment tool. Entropy-based weighting reveals that groundwater-related components contribute more to overall vulnerability than surface runoff. According to initial MCDM analysis, the most vulnerable areas are those in the upper reaches of the watershed, where steep slopes accelerate runoff and limit infiltration. This confounding influence of elevation is addressed by implementing topographic normalization and reevaluating subwatershed vulnerability while controlling for elevation bias. The findings underscore the importance of incorporating both hydrological and topographical factors into urban watershed vulnerability assessment and demonstrate the applicability of entropy-weighted MCDM to complex, data-scarce urban environments. The proposed framework is a replicable decision support tool for prioritizing hydrologically sensitive areas in intervention planning.

1. Introduction

Rapid urbanization has significantly altered hydrological processes in many watersheds, intensifying surface runoff, reducing infiltration, and exacerbating flood risks [1]. These changes highlight the urgent need to assess hydrological vulnerability in urban regions to support sustainable watershed management and climate-resilient infrastructure planning [2]. However, evaluating such vulnerability in complex urban environments requires a systematic framework that incorporates diverse hydrological components and accounts for spatial heterogeneity. In response, recent studies have increasingly applied multicriteria decision-making (MCDM) approaches to integrate key indicators and identify priority areas for intervention [3].

In Korea, vulnerability assessment studies typically aim to inform policy formulation and implementation through evaluations based on various indicators. Indicator-based methods are easy to understand and facilitate comparisons between regions; however, their results are significantly influenced by factors such as indicator selection, standardization, and weighting [4]. Given the critical role of indicator weighting in determining the final ranking of alternatives, the weight assignment method must be scientific and rational for sound decision-making. Various techniques can be used to determine indicator weights, potentially resulting in different values for the same indicators [5,6,7]. In hydrological vulnerability studies, the Delphi technique has been adopted multiple times [8,9], but it is subjective, relying on survey results from experts or stakeholders. By contrast, the entropy method enables objective analysis by considering only the quantitative characteristics of the collected data [10,11].

In the technique for order preference by similarity to ideal solution (TOPSIS) evaluation method, rankings are calculated based on the distance/proximity between analytical factors, which is used by decision-makers to select alternatives [12]. Hajkowicz and Collins. [13] investigated numerous decision-making techniques in water management research and found that uncertainties in input data and indicator weights can be addressed by combining multicriteria decision-making (MCDM) techniques with fuzzy concepts. TOPSIS [12] is a frequently used MCDM method. Several studies in the Korean context have applied TOPSIS. For example, Lee et al. [14] analyzed various drought-related factors using TOPSIS to establish a drought vulnerability assessment framework, and Kim et al. [15] used it with fuzzy techniques to evaluate flood vulnerability in Korea using climate change factors. Park et al. [16] compared TOPSIS results based on subjective and objective weights to assess water use vulnerability across unit watersheds. For watershed management, TOPSIS has been used to prioritize subwatersheds for nonpoint-source pollution control in the Seonakdong River [17] and select monitoring networks in the Nakdong River basin [18]. Recently, TOPSIS has been applied in diverse fields beyond hydrology and water quality, such as the selection of alternative fuels for ships [19] and the prioritization of industrial accident prevention factors [20].

Hydrological vulnerability in urban watersheds is also strongly influenced by physical and climatic factors such as groundwater dynamics, drought conditions, and topography. Groundwater-related components, including percolation, groundwater recharge, and baseflow, are critical for sustaining streamflow during dry periods and mitigating drought impacts, particularly in regions with variable precipitation regimes [21]. Studies have shown that reductions in groundwater recharge can exacerbate hydrological stress and increase the frequency and severity of droughts in urbanizing basins [22]. Topography, especially slope and elevation, plays a decisive role in controlling runoff–infiltration partitioning and the spatial distribution of hydrological processes [23]. Steeper slopes tend to accelerate surface runoff, reduce infiltration, and limit groundwater replenishment, thereby increasing vulnerability. Integrating these hydrological considerations into decision–support frameworks enables a more accurate identification of vulnerable areas, as demonstrated in prior studies that combined MCDM approaches with drought risk assessment, groundwater monitoring, and terrain analysis [24]. This study builds on such approaches by explicitly incorporating groundwater and topographic characteristics alongside surface hydrological indicators to improve the robustness of vulnerability assessment in urban watersheds.

This study aims to select priority management areas or vulnerable regions for policy derivation using MCDM, using the entropy method for objective (rather than subjective) weight determination. Given its extensive use in assessing vulnerabilities related to water resources, water use, water quality, and climate change, as demonstrated in prior studies, TOPSIS is appropriate for evaluating hydrological phenomena in this study. Hydrologically vulnerable areas in the upstream Yangjaecheon watershed, encompassing Gwacheon City, are assessed by extracting factors related to hydrological soundness using the soil and water assessment tool (SWAT) model, which analyzes water circulation, and applying the MCDM method TOPSIS.

2. Materials and Methods

2.1. Methods for Evaluating Vulnerable Areas

The process of this study (Figure 1) is broadly divided into two main components: analysis of vulnerable areas and SWAT model analysis for generating factors for vulnerability assessment.

First, indicators for evaluating the vulnerability of the hydrological elements within the watershed are selected and normalized through standardization. These indicators are derived from the output values of the SWAT model [25]. The SWAT model is constructed and simulated over a long period, and the resulting hydrological data from the watershed serve as input for vulnerability assessment [25,26]. The normalized data enable the identification of characteristics specific to each indicator within the watershed. The normalized data are then weighted using the entropy method, and the weighted indicators are evaluated using TOPSIS to assess vulnerable areas [25]. With the weights calculated via the entropy method, reliable indicators within the watershed can be analyzed and applied. Additionally, these results can be compared with those obtained using equal weighting to identify differences based on land use types [26].

2.2. Case Study

The study area comprises the upstream Yangjaecheon watershed, including Gwacheon City. The watershed covers an area of approximately 39.86 km², encompassing most of Gwacheon City (Figure 2). Gwacheon City is in the midwestern part of Gyeonggi Province, bordered by Seoul Special City to the north and adjacent to Anyang City, Seongnam City, and Uiwang City in Gyeonggi Province. The area of Gwacheon City is 35.87 km², accounting for 0.035% of South Korea’s total land area (100,443.6 km²) and 0.35% of Gyeonggi Province’s area (10,199.5 km²). Yangjaecheon, which runs through the center of the city, originates from Gwanaksan Mountain and flows northeast, eventually joining Tancheon, a river that traverses Seocho-gu and Gangnam-gu in Seoul Special City.

2.3. SWAT Model

The SWAT model is a widely used hydrological model developed by the US Department of Agriculture Agricultural Research Service [27]. It has been evolving since its development in 1994. The model enables the long-term simulation of hydrological and water quality processes in large watersheds with complex land use types. It can analyze water circulation, sediment transport, and the behavior of organic substances, considering land use and soil characteristics. Furthermore, this model can simulate runoff based on daily rainfall and meteorological data. The SWAT model comprises four submodules: hydrology, soil erosion, nutrients/pesticides, and main channel processes. The hydrology submodel calculates water balance, including daily surface runoff, lateral flow, infiltration, groundwater, and evapotranspiration, using a storage equation [28].

Runoff resulting from rainfall is calculated, and the SWAT model, composed of various submodules, facilitates the efficient analysis of hydrological and water quality aspects across applications. It has been extensively used in studies worldwide and in practical applications [29,30,31]. Water balance analysis is performed using the SWAT model based on the water balance equation, as shown in Equation (1) [28].

$${SW}_t={SW}_0+\sum _{i=0}^t{(R}_{day}-Q_{sur f}-E_a-P_{erc}-Q_{gw})$$

(1)

where ${SW}_t$ is the final soil moisture content (mm), ${SW}_0$ is the initial soil moisture content (mm) on day $i$, $t$ is time (days), $R_{day}$ is the precipitation (mm) on day $i$, $Q_{surf}$ is the surface runoff (mm) on day $i$, $E_a$ is the evapotranspiration (mm) on day $i$, $P_{erc}$ is the total amount of water percolating from the soil layer to the aquifer (mm) on day $i$, and $Q_{gw}$ is the return flow (mm) on day $i$.

2.4. Selection of Indicators

In this study, evaluation indices are calculated, and vulnerability is assessed for 25 subwatersheds considering the hydrological elements among the watershed soundness evaluation factors proposed by the US Environmental Protection Agency (EPA).

Table 1. Vulnerability evaluation factors.

Criterion	Method	SWAT Model Variable (hru.out)	Data
Hydrology	Total water yield (mm)	WYLD	SWAT simulation result
	Surface processes Surface runoff (mm)	SURQ (mm)
	Soil water dynamics Infiltration (mm) Soil moisture (mm) Lateral flow (mm)	DAILYCN SW (mm) LATQ (mm)
	Groundwater dynamics Percolation (mm) Groundwater recharge (mm) Baseflow (mm)	PERC (mm) GW_RCHG (mm) GWQ (mm)

presents the selected hydrological indicators [32,33]. The SWAT simulation results are utilized to evaluate vulnerability in terms of hydrology, and the following indicators are selected: total water yield, surface processes (surface runoff), soil water dynamics (infiltration, soil moisture, and lateral flow), and groundwater dynamics (percolation, groundwater recharge, and baseflow). These indicators are evaluated by normalizing their data to calculate a watershed vulnerability index, which is then used to identify vulnerable areas.

2.5. Normalization

The datasets corresponding to these evaluation factors have different units and properties. Hence, they have to be normalized before they can be evaluated. Normalization methods include Z-score standardization, rescaling, the distance-to-reference-country method, and categorical scaling [34]. This study adopts the dimension index method, a widely used rescaling technique where data are ranked in a linear sequence based on their order within the entire data range, as expressed by Equation (2).

$$X_i={\displaystyle \frac{x_i-x_{min}}{x_{max}-x_{min}}}\times 100$$

(2)

where $X_i$ is the normalized value of the ith observation, $x_i$ is the original data value, $x_{max}$ is the maximum value of the dataset, and $x_{min}$ is the minimum value of the dataset [34].

2.6. Entropy-Based Weighting

In this study, the entropy method is used to determine the weights of the evaluation indicators. Originally proposed by Shannon in 1948, the entropy method is a straightforward approach to decision-making. This method is particularly useful for evaluating data with inherent uncertainty because it quantifies the degree of uncertainty in the information used to assess each indicator [35,36].

The entropy method enables objective weight estimation by relying solely on the informational content of indicators (i.e., independent of subjective judgments) [35]. Entropy-based weights are calculated by structuring a data matrix where each column represents the values of a particular indicator, followed by a normalization process [34]. Once the data are normalized, the entropy values for the indicators are calculated and then used to derive the final entropy-based weights. The weight computation steps are detailed as follows:

2.6.1. Construction of Indicator Data Matrix

The first step is the construction of a data matrix where each row represents an alternative (e.g., a subwatershed) and each column corresponds to an evaluation indicator. This matrix is the foundation of the subsequent normalization and entropy calculations, with each cell containing the value of a specific indicator for a given alternative [37].

$$R=\left[r_{ij}\right](r_{ij} :normalized value)$$

(3)

where $i$ is an index denoting an alternative (e.g., subwatersheds), $i=\mathrm{1,2},\dots ,m$, and $j$ is an index representing an indicator, $j=\mathrm{1,2},\dots ,n$.

Each element of the normalized matrix $rij$is calculated as follows:

$$r_{ij}={\displaystyle \frac{x_{ij}}{\sum _{i=1}^m{x}_{ij}}}(i=1,2,\dots ,m;j=1,2,\dots ,n)$$

(4)

where $xij$ is the original value of the $j$th indicator for the $i$th alternative and $rij$ is the normalized value of $xij$.

2.6.2. Calculation of Entropy for Each Indicator

Once the normalized decision matrix $R=\left[rij\right]$ is obtained, the entropy $H_i$ for each indicator $i$ is calculated as follows [38,39]:

$$H_i= -k\sum _{i=1}^n{f}_{ij}\mathit{log}{f}_{ij} (k={\displaystyle \frac{1}{\mathit{ln}n}} )$$

(5)

where $H_i$ is the entropy value of the $i$th indicator, $f_{ij}$ is the proportion of the $i$th alternative with respect to the $j$th indicator, $n$ is the number of alternatives, and $k$ is a normalization constant ensuring that $H_i$ ∈ [0, 1]. The value $f_{ij}$ is computed as follows [40]:

$$f_{ij}={\displaystyle \frac{r_{ij}}{\sum _{j=1}^n{r}_{ij}}}$$

(6)

2.6.3. Calculation of Degree of Diversification and Final Weight Determination

Once the entropy $H_i$ of each indicator is calculated, the degree of diversification $d_i$ is determined as follows:

$$d_i=1-H_i.$$

(7)

This value represents the contrast intensity or the informational utility of each indicator. A larger $d_i$ indicates that the indicator can differentiate the alternatives better. The normalized weight $w_{ij}$ for each indicator is then computed as follows [41]:

$$w_{ij}={\displaystyle \frac{1-H_i}{\sum _{i=1}^m(1-H_i)}} (0\le W_i\le 1,\sum _{i=1}^m{W}_i=1)$$

(8)

2.7. TOPSIS Analysis

MCDM is used to assess multiple alternatives based on several criteria. Each criterion is assigned a weight reflecting its relative importance, and the goal is to identify the optimal alternative or rank the options. This approach is widely adopted in various research fields for selection among competing alternatives. The MCDM technique used in this study is TOPSIS, where the best alternative is identified by calculating the positive ideal solution (PIS) and the negative ideal solution (NIS) and then evaluating the distance of each alternative from these ideal points. Alternatives closer to the PIS are more desirable, whereas those closer to the NIS are less favorable. This technique enables objective, systematic evaluation and has been successfully applied to a wide range of decision-making problems [15].

The TOPSIS method consists of the following steps [12,42,43,44]:

Construction of Decision Matrix

$$R=\left[r_{ij}\right]$$

(9)

where $r_{ij}$ is the normalized value of the $j$th criterion for the $i$th alternative ($i$ = 1, 2, …,$m$ is the alternative index, and $j$ = 1, 2, …,$n$ is the criterion index).

Application of Weights to Normalized Matrix

$$v_{ij}=w_j{r}_{ij}$$

(10)

where $v_{ij}$ is the weighted normalized value and $w_j$ is the weight of the $j$th criterion.

Determination of PIS and NIS

$$\begin{gathered} A^{*}=v_1{,v}_2,\dots ,v_n^{*} \\ {A}^{-}=v_1^{-},v_2^{-},\dots ,v_n^{-} \\ {v}_j^{*}=max{v}_{ij},v_j^{-}=min{v}_{ij} \end{gathered}$$

(11)

where $A^{*}$ is the PIS (best performance for each criterion), and A⁻ is the NIS (worst performance for each criterion).

Calculation of Distance of Each Alternative from PIS and NIS

$$\begin{gathered} S_i^{*}=\sqrt{\sum {\left(v_j^{*}-v_{ij}\right)}^2} \\ {S}_i^{-}=\sqrt{\sum {\left(v_j^{-}-v_{ij}\right)}^2} \end{gathered}$$

(12)

where $S_i^{*}$ is the Euclidean distance of alternative i from the PIS and $S_i^{-}$ is the Euclidean distance of alternative i from the NIS.

Calculation of Relative Closeness to the Ideal Solution

$$C_i^{*}={\displaystyle \frac{S_i^{-}}{S_i^{*}+S_i^{-}}}$$

(13)

where $C_i^{*}$ is the relative closeness coefficient of alternative i to the ideal solution; the closer $C_i^{*}$ is to 1, the better the performance of the alternative. The alternatives are ranked from most to least preferable based on these values.

3. Results and Discussion

3.1. Development of the SWAT Model

A watershed model was developed for the target watershed using the SWAT model for hydrological data analysis. The simulation period covered 10 years, from 2014 to 2023, and the hydrological output data generated during this period were collected for analysis. Model construction required various input datasets, including meteorological, topographical, and observational datasets. Meteorological data were obtained from the Seoul Meteorological Observatory: daily precipitation, temperature (mean, maximum, and minimum), average wind speed, relative humidity, and solar radiation. These datasets are publicly available through the Korea Meteorological Administration Open Data Portal (https://data.kma.go.kr, accessed on 13 June 2025).

Topographical data were sourced from a digital elevation model (DEM) with a 30 m resolution, based on latitude and longitude grids. These data were obtained from the National Geographic Information Institute of the Ministry of Land, Infrastructure, and Transport (https://www.ngii.go.kr/, accessed on 13 June 2025) [45]. Additionally, land cover data based on 2023 classifications were collected from the Environmental Geographic Information Service of the Ministry of Environment (http://egis.me.go.kr, accessed on 13 June 2025) [46]. Soil property data were derived from detailed soil maps provided by the Soil Environmental Information System of the Rural Development Administration (http://soil.rda.go.kr, accessed on 13 June 2025) [47], and these were used to construct the model’s soil input layer. The daily discharge volume from the Gwacheon wastewater treatment plant, which discharges into the Yangjaecheon Stream, was obtained using data from the National Pollutant Source Survey of the Ministry of Environment and incorporated into the model as point-source pollution data. These data are accessible through the Water Environment Information System (https://water.nier.go.kr, accessed on 13 June 2025) maintained by the National Institute of Environmental Research. For streamflow calibration, observed flow data from the Yangjaecheon 1 station, located at Umyeon Bridge, were collected and utilized.

Figure 3 presents the geographical data used in the model setup. The watershed was divided into 25 subbasins, and the land use in each subbasin was classified into four categories: transportation, forest, residential, and others.

Table 2. Assessment factor of vulnerability.

Subbasin No.	Land Use Rate (%)				Subbasin No.	Land Use Rate (%)
	Transportation	Forest	Residential	Others		Transportation	Forest	Residential	Others
1	71.18	24.88	0.74	3.20	14	78.40	18.97	0.04	2.59
2	73.84	22.57	1.73	1.86	15	82.59	4.56	0.25	12.60
3	73.34	22.41	1.66	2.59	16	32.46	63.87	0.22	3.45
4	96.56	0.41	0.73	2.30	17	0.30	97.94	0.00	1.76
5	74.72	23.11	0.76	1.41	18	49.23	46.61	1.90	2.26
6	97.05	0.13	0.27	2.55	19	0.23	98.03	0.00	1.74
7	96.96	0.00	0.34	2.70	20	0.02	98.44	0.00	1.54
8	72.05	21.85	0.73	5.38	21	33.19	63.49	0.53	2.79
9	0.08	97.61	0.00	2.31	22	0.20	97.31	0.00	2.49
10	81.38	17.31	0.14	1.17	23	65.57	25.53	0.25	8.65
11	99.17	0.00	0.00	0.83	24	26.46	64.36	0.93	8.25
12	67.00	27.75	0.27	4.98	25	94.52	1.29	0.56	3.61
13	32.58	64.09	0.11	3.22	Avg.	48.30	47.71	0.52	3.47

presents the land use distribution per subbasin.

3.2. Normalization of Indicators

The eight indicators were normalized for analysis. During normalization, the vulnerability characteristics of each indicator were considered by conducting a correlation analysis among the indicators to assess their positive (+) and negative (−) relationships. Pearson correlation analysis was performed on the eight indicators, and the results are presented in Figure 4. The following indicators exhibited positive correlations with each other: total water yield (WYLD), surface runoff (SURQ), soil moisture (SW), percolation (PERC), infiltration (CN), groundwater recharge (GW_RCHG), and baseflow (GWQ). By contrast, lateral flow (LATQ) showed negative correlations with all other indicators.

These results indicate that high-surface-runoff areas also tend to have high infiltration, percolation, and groundwater recharge. However, as lateral flow increases, the volume of water infiltrating the soil or recharging the groundwater tends to decrease. Consequently, a higher lateral flow was interpreted as a sign of lower hydrological integrity.

The curve number (CN), which is related to infiltration, showed a strong positive correlation with both surface runoff and soil moisture but a relatively weak correlation with baseflow. Percolation, groundwater recharge, and groundwater storage exhibited strong correlations with each other, indicating that percolated water contributes to groundwater recharge, which in turn becomes baseflow.

Figure 5 illustrates the spatial distributions of the normalized indicators per subwatershed. The eight indicators in Figure 5 were all normalized using the dimension index method described in Equation (2), rescaling the values to a 0–100 range. This ensured that the spatial patterns reflect relative hydrological conditions without being affected by differences in units or scales. The blue (red) regions indicate areas with high (low) indicator values, reflecting good (poor) hydrological conditions. Due to its inverse relationship with the other indicators, as found through the correlation analysis, lateral flow (LATQ) was normalized such that higher values corresponded to lower indicator scores.

Based on this analysis, the subwatersheds with poor performance per indicator were identified as follows:

Total water yield (WYLD): Subwatersheds 9, 10

Surface runoff (SURQ): 9, 10, 13, 16, 17, 20, 22, 24

Soil moisture (SW): 9, 17, 20, 22, 24

Lateral flow (LATQ): 14

Curve number (CN): 9, 12, 14, 17, 20, 22

Percolation (PERC): 17, 20, 22

Groundwater recharge (GW_RCHG): 17, 20, 24

Baseflow (GW): 9, 10, 17, 20, 24

These findings were compared with the DEM (Figure 3a). The upper watershed areas, located at higher elevations, have poorer hydrological characteristics, whereas the midstream–downstream regions demonstrate better hydrological conditions.

Being at higher elevations, the upper watershed areas experience greater runoff during rainfall events; the reduced amount of water infiltrating into the soil contributes to weaker groundwater-related performance. Conversely, the midstream and downstream reaches exhibit favorable hydrological characteristics, likely due to sufficient infiltration and groundwater recharge.

3.3. Entropy Analysis

Entropy analysis was conducted to examine the changes in the indicator weights. The original (pre-change) weights, calculated using the equal-weight method, are presented in the upper rows of

Table 3. Weight changes according to the entropy method.

Method	Factor	Average	Transportation	Forest	Residential
Equal weights	Total water yield	0.065	0.053	0.087	0.050
	Surface runoff	0.079	0.064	0.109	0.053
	Soil moisture	0.071	0.056	0.099	0.050
	Lateral flow	0.045	0.039	0.052	0.050
	CN	0.059	0.047	0.085	0.029
	Percolation	0.078	0.068	0.093	0.078
	Groundwater recharge	0.078	0.068	0.093	0.078
	Groundwater	0.078	0.067	0.093	0.078
Entropy method	Total water yield	0.062	0.050	0.083	0.048
	Surface runoff	0.082	0.066	0.114	0.055
	Soil moisture	0.072	0.056	0.101	0.051
	Lateral flow	0.037	0.033	0.043	0.042
	CN	0.055	0.044	0.080	0.027
	Percolation	0.083	0.072	0.099	0.084
	Groundwater recharge	0.083	0.072	0.099	0.084
	Groundwater	0.083	0.072	0.099	0.084

for each land use type (average, transportation, forest, and residential). These were compared with the updated weights derived from the entropy method (Figure 6). Compared with the equal-weight case, the indicator values for surface runoff, percolation, groundwater recharge, and baseflow increased on average when the entropy-derived weights were applied.

Therefore, under the given conditions, these indicators may have a greater influence on hydrological health compared with the others. In particular, percolation, groundwater recharge, and baseflow are important indicators for assessing drought severity within a watershed and thus can be effectively used to identify drought-prone areas. According to the analysis per land use category, the impact of percolation, groundwater recharge, and baseflow is most significant in forested areas, whereas transportation areas exhibit the least influence. Therefore, among the indicators affecting hydrology, as identified through entropy analysis, those related to groundwater play a more dominant role in shaping the hydrological characteristics of the watershed.

3.4. Analysis of Vulnerable Area

The normalized indicator data and TOPSIS were used to identify hydrologically vulnerable areas.

Figure 7a and

Table 4. Results of the selection of vulnerable areas.

Original Ranking							Reevaluated Ranking
Rank	Subbasin No.	Elevation Percentile	Rank	Subbasin No.	Elevation Percentile		Rank	Subbasin No.	Rank	Subbasin No.
1	20	100.00	14	23	58.16	Re evaluation	1	5	14	3
2	14	99.78	15	2	37.09		2	10	15	7
3	17	84.42	16	16	29.94		3	9	16	6
4	5	56.08	17	18	29.49		4	12	17	8
5	24	75.65	18	1	23.99		5	13	18	25
6	10	46.19	19	3	43.72		6	19	19	11
7	9	30.17	20	7	16.52		7	21	20	15
8	12	60.54	21	6	16.44		8	4	Excluded	14, 17, 20, 22, 24
9	22	70.21	22	8	28.39		9	23
10	13	28.21	23	25	21.90		10	2
11	19	25.92	24	11	32.84		11	16
12	21	32.09	25	15	20.26		12	18
13	4	16.03					13	1

present the TOPSIS vulnerability assessment results. The most vulnerable subwatersheds were subbasins 20 > 14 > 17. Most of these top-ranked subwatersheds are in the upper watershed region. As shown in Figure 5, these upper areas exhibit poor performance across most hydrological indicators, which is attributed to their topographical characteristics. These upstream regions have steep slopes that limit water infiltration into the soil or groundwater during rainfall events, causing rapid moisture loss through lateral flow or surface runoff. Therefore, the vulnerability observed in these areas is significantly influenced by both hydrological factors and topographic conditions.

According to the elevation percentile values presented in Table 4, subbasins 20, 14, and 17—all identified as highly vulnerable—belong to the top 70th percentile or higher in elevation. Reevaluation was conducted by excluding areas above the 68th percentile in elevation to control for this topographic bias. The threshold of 68% was chosen based on the statistical principle that approximately 68.27% of data in a normal distribution lie within one standard deviation of the mean, implying that topographic extremes significantly influence elevation values above this range. The high-elevation subwatersheds (above the 68th percentile) were excluded, and the previously top-ranked subbasins 14, 17, 20, 22, and 24 were removed from the vulnerability list. The reevaluation then identified subbasins 5, 10, and 9 as the new most vulnerable areas. These subwatersheds are more representative of hydrological vulnerability, independent of elevation bias.

Given their performance across the eight previously selected hydrological indicators, these areas were determined to be hydrologically vulnerable, particularly due to their poor results in terms of groundwater recharge, percolation, and baseflow. Hence, these subwatersheds are critical for evaluating the overall hydrological health of the watershed.

4. Conclusions

This study evaluated a method for prioritizing hydrologically vulnerable areas in urban environments by assigning weights to hydrological indicators using the entropy method and ranking them with the TOPSIS approach. The target area was the upper Yangjaecheon watershed, including Gwacheon City, for which a SWAT model was developed to estimate hydrological indicators. The selected indicators included total water yield, surface processes, soil water dynamics (infiltration, soil moisture, and lateral flow), and groundwater dynamics (percolation, groundwater recharge, and baseflow). The entropy method assigned higher weights to groundwater-related factors such as percolation, groundwater recharge, and baseflow compared with equal weighting, indicating their significant influence on hydrological conditions in the study area. Moreover, these indicators can also serve as proxies for drought assessment, aiding in the identification of drought-prone regions within a watershed.

The TOPSIS results showed that most of the highly vulnerable subwatersheds are located in the upper reaches of the basin, dominated by steep slopes and forested land cover. These areas performed poorly in key indicators such as percolation, groundwater recharge, and baseflow, consistent with the entropy analysis findings.

Higher weights assigned to groundwater-related indicators—such as percolation, groundwater recharge, and baseflow—reflect the particular hydrological dynamics of the upstream Yangjaecheon watershed. The basin’s steep slopes and predominantly forested upper reaches promote rapid surface runoff and limit infiltration in certain areas, making the locations where infiltration and groundwater recharge occur especially critical for sustaining streamflow during dry periods. As a result, variations in groundwater-related processes have a disproportionately large influence on overall hydrological stability in this watershed. While similar patterns have been observed in other monsoon-influenced, topographically diverse basins (e.g., Taylor et al., 2013; Shope, 2016 [22,23]), the magnitude of the weight increase is context-specific and should not be assumed universal. In flatter, less permeable, or more uniformly urbanized watersheds, other indicators—such as surface runoff or soil water storage—may carry greater importance in vulnerability assessments.

However, the results were strongly influenced by topographical characteristics, which limited the ability to identify vulnerable areas accurately based solely on hydrological indicators. To address this, the analysis was refined by standardizing watershed elevation and excluding subwatersheds above the 68th percentile of elevation. This adjustment excluded subbasins 14, 17, 20, 22, and 24, while subbasins 5, 10, and 9 emerged as new priority areas that are more directly influenced by hydrological factors rather than topographic effects.

While the proposed method provides a systematic approach to identifying hydrologically vulnerable areas, several limitations should be noted. In addition to the influence of topographical characteristics, the reliability of the results is constrained by the availability and resolution of spatial datasets. For example, detailed information on surface sealing, imperviousness, and the temporal dynamics of urban development was not available at sufficiently fine scales, which may reduce the precision of hydrological indicator estimation, particularly in highly urbanized subwatersheds where small-scale land cover changes can substantially affect infiltration and runoff.

Furthermore, the analysis relied primarily on model-simulated hydrological indicators without integrating certain site-specific variables such as detailed soil hydraulic properties, land management practices, or anthropogenic alterations to the watershed. The approach also did not incorporate long-term climate variability or projected climate change impacts, which could influence vulnerability patterns over time. Future research should address these limitations by integrating additional datasets (e.g., remote sensing products, groundwater observation networks, and socio-economic vulnerability indicators), applying the methodology under different climate scenarios, and incorporating higher-resolution spatial data. Expanding the framework to other urban and peri-urban watersheds would also help evaluate its generalizability and robustness across varying hydrological and topographical contexts.