Risk assessment of water treatment chemical supply disruptions in China using a synthetic water quality dataset and the Best-Worst Method

1 Introduction

Water treatment systems depend on a continuous supply of critical chemicals such as coagulants, disinfectants, and pH regulators to ensure safe drinking water and regulatory compliance (Bandh and Mushtaq, 2025; Kekes et al., 2023; Zhao et al., 2023). Disruptions in chemical supply chains pose significant risks to operational stability, public health, and service continuity (Al-Khatib et al., 2024; Feng et al., 2023; Luo et al., 2018). In recent years, rising global uncertainty, including logistical constraints, market volatility, and environmental pressures, has intensified water utilities’ vulnerability to supply disruptions (Berglund et al., 2022; Peng et al., 2024; Wang L. et al., 2025). These challenges are particularly relevant in China, where large-scale and regionally diverse water systems require reliable and resilient supply mechanisms (Ren et al., 2020; Li et al., 2021; X et al., 2025).

Traditional approaches to managing chemical supply risks often rely on static strategies, such as safety stock or single-supplier contracts (Foerstl et al., 2010; Yang et al., 2026; Chen et al., 2026). However, these approaches are limited in their ability to address dynamic and uncertain disruption scenarios (Zheng D. et al., 2026; Zhan et al., 2024; Wang X.-Y. et al., 2025). As a result, there is a growing need for data-driven decision support frameworks that integrate quantitative datasets with structured evaluation methods (Hu et al., 2025; Hu et al., 2026; Zheng Y. et al., 2026). Synthetic water quality datasets provide an opportunity to test such frameworks under controlled, reproducible conditions (Kamble and Gunasekaran, 2020; Chen et al., 2022; Li et al., 2024).

Multi-criteria decision-making methods have been widely applied in infrastructure and environmental management to support complex decision processes involving multiple conflicting objectives. Rezaei (2015) originally proposed the Best Worst Method as a multi-criteria decision method for determining criterion weights through comparisons with the best and worst criteria (Rezaei, 2015). By combining data-driven indicators with structured weighting techniques, it becomes possible to evaluate mitigation strategies under different disruption scenarios in a transparent and reproducible way. This work develops an integrated framework that combines a synthetic water quality dataset, indicator-based evaluation, and multi-criteria decision analysis to assess mitigation strategies for chemical supply disruptions. The framework supports decision-making under uncertainty and identifies strategies that enhance system resilience while maintaining compliance and service continuity.

1.1 Literature review

Water quality assessment has traditionally relied on physicochemical, chemical, and biological indicators aggregated into composite indices or decision criteria. Juwana et al. (2012) reviewed indicator-based approaches for water sustainability assessment and showed that water indicators can characterize environmental conditions and support water resource management decisions (Juwana et al., 2012). More recent research has emphasized the importance of parameter selection, weighting, normalization, and aggregation in constructing water quality indices (Akhtar et al., 2021). Uddin et al. (2021) reviewed major WQI models and identified parameter selection, weighting, aggregation, and uncertainty as important sources of variation in water quality assessment (Uddin et al., 2021). In China, water quality studies have also used WQI and related indicator systems to evaluate spatial and temporal variations in river and surface water quality. For example, national-scale research has examined water quality patterns and their drivers using large monitoring datasets, while other studies have applied WQI-based approaches to specific river systems and water diversion projects. These studies establish a strong precedent for transforming multiple water quality measurements into quantitative indicators for comparative assessment.

The variables considered in the present study are also consistent with established water quality assessment practice. Previous water quality assessments have used parameters such as pH, turbidity, dissolved oxygen, ammonia nitrogen, nitrate, total phosphorus, chemical oxygen demand, biological oxygen demand, heavy metals, and microbial indicators. For example, the South to North Water Diversion Project has been evaluated using multiple water quality parameters, with turbidity, dissolved oxygen, ammonium nitrogen, nitrate, and chemical oxygen demand identified as important contributors to the WQI. National-scale research on Chinese rivers has also used large datasets to examine water quality impairment and its environmental drivers. These studies show that the individual water quality variables used in the present dataset are relevant to water quality assessment (Nong et al., 2020; Huang et al., 2021). The present study therefore does not introduce new water quality parameters. Instead, the contribution lies in transforming these established variables into performance indicators for a chemical supply disruption decision model.

The four evaluation criteria adopted in this study are supported by both Chinese regulatory requirements and previous resilience research. Compliance and public health are directly related to drinking water safety and regulatory water quality requirements under GB 5749–2022 (Liu, 2025). Service continuity is a key requirement of water supply systems because continuous and stable water supply is required during normal operation and emergency conditions (Alsancak et al., 2022). Supply vulnerability is recognized in water supply resilience assessment as a measure of a system’s exposure to and sensitivity to uncertain external conditions (Wang L. et al., 2025). Recoverability represents a water supply system’s ability to restore performance after disruption and is recognized as an important aspect of resilience assessment (Alsancak et al., 2022; Mefteh et al., 2022). These criteria therefore represent four distinct consequences of chemical supply disruption: protection of water quality and public health, continuity of service, exposure to disruption, and restoration after disruption. Together, they provide a basis for assessing mitigation strategies from both water quality and system resilience perspectives. Table 1 provides a comparative summary of previous studies, including their main research problems, water quality criteria, methods, datasets, and relevance to the present study.

TABLE 1. Comparison of previous studies and the present study in relation to water quality criteria, decision methods, datasets, and research objectives.

The specific synthetic China Water Pollution Monitoring Dataset used in this study should be distinguished from the empirical Chinese water quality datasets reported in previous studies. The dataset is used as a controlled numerical source for methodological evaluation. To the best of the authors’ knowledge, no previous peer-reviewed study has used this specific dataset to assess water treatment chemical supply disruption or select mitigation strategies. Therefore, its use in the present study supports methodological evaluation rather than an empirical assessment of Chinese water utility performance.

2 Methodology

Chemical supply disruptions pose increasing risks to regulatory compliance and service continuity in water utilities, necessitating robust, data-driven decision-support approaches. Data are collected and pre-processed, with primary inputs obtained from the synthetic China Water Pollution Monitoring Dataset. The relevant variables are transformed into quantitative indicators representing compliance and public health, service continuity, supply vulnerability, and recoverability. The dataset serves as a controlled numerical data source for methodological evaluation and does not represent direct measurements from operating Chinese water utilities. The relative importance of these criteria is determined using the Best-Worst Method, and a data-driven scoring model is formulated to integrate normalized indicator values with the derived weights to evaluate mitigation strategies. The analysis is conducted under defined disruption scenarios, and sensitivity analysis is performed to assess the robustness of the results. Figure 1 schematically illustrates the overall methodological procedure, showing the sequence of data preprocessing, criteria weighting, strategy evaluation, and validation.

2.1 Research framework and problem definition

The decision problem is to evaluate and rank mitigation strategies for water treatment chemical supply disruptions under uncertainty. The objective is to minimize operational and public health risks while maintaining regulatory compliance and service continuity. The system is modeled as a supply-dependent water utility network in which chemical availability directly affects treatment performance and service reliability. The system boundary includes upstream chemical supply, treatment operations, and downstream service delivery outcomes. External factors such as geopolitical shocks or global market dynamics are treated as exogenous inputs represented through disruption scenarios. Internal system responses are represented through performance criteria derived from water quality indicators contained in the synthetic dataset (see Figure 2).

A data-driven analytical framework integrates the synthetic water quality dataset with decision modeling techniques. The primary dataset is the synthetic China Water Pollution Monitoring Dataset. It provides water quality values that are used as proxy inputs for treatment performance and compliance assessment. Because the dataset does not include direct chemical supply or mitigation strategy data, strategy effects are represented through predefined disruption-scenario adjustments. The decision space is defined by a finite set of mitigation strategies, denoted as Equation 1:

S = S 1 , S 2 , … , S m (1)

where each S i represents a distinct strategy such as dual sourcing, safety stock, or supplier prequalification. The evaluation is conducted across a set of criteria, expressed as Equation 2:

C = C 1 , C 2 , C 3 , C 4 (2)

where C 1, C 2, C 3, and C 4 correspond to compliance and public health, service continuity, supply vulnerability, and recoverability, respectively. For each strategy S i performance is represented by a vector of scenario-adjusted indicators rather than by direct observations of strategy implementation. Because the dataset does not contain variables describing chemical supply availability, supplier capacity, inventory levels, procurement arrangements, or implementation of the mitigation strategies, the available water quality variables are first transformed into normalized performance indicators. The mitigation strategies are then represented through predefined disruption-scenario adjustments applied to these indicators as Equation 3:

I i = I i 1 , I i 2 , I i 3 , I i 4 (3)

Let I i j denote the performance of strategy under criterion j. These values are obtained from the normalized water quality indicators and the strategy-specific scenario adjustment defined in Section 2.6. Thus, the water quality dataset provides the baseline system performance, while the scenario assumptions provide the modeled effect of each mitigation strategy. Uncertainty is incorporated through disruption scenarios representing different levels of supply disruption severity. The interaction between strategies and disruption scenarios is therefore evaluated as a scenario-based comparative assessment rather than as a causal estimate derived from observed supply chain data.

The indicator I i j value is not directly observed for the mitigation strategies. Instead, it is obtained by applying the strategy- and scenario-specific adjustment defined in Section 2.6 to the corresponding baseline indicator derived from the synthetic water quality dataset. This distinction matters because the dataset contains water quality variables but not direct measurements of chemical supply availability or mitigation strategy implementation. Uncertainty is incorporated through disruption scenarios, defined as Equation 4:

D = D 1 , D 2 , … , D k (4)

where each D k represents a distinct level of supply disruption severity. The interaction between strategies and scenarios is explicitly modeled to assess robustness. The overall framework is structured as a sequential process in which raw data are transformed into indicators, criteria weights are derived using the Best Worst Method, and strategies are evaluated through a composite scoring model. The output is a ranking of mitigation strategies under varying disruption conditions, supported by sensitivity and robustness analysis.

2.2 Data sources and preprocessing

The quantitative analysis is based on one primary data source, namely, the synthetic China Water Pollution Monitoring Dataset. The dataset provides structured water quality values representing multiple monitoring locations and time periods in China. It serves as a controlled data source for methodological evaluation rather than a collection of direct field measurements from Chinese water utilities. The variables include physicochemical indicators such as dissolved oxygen, ammonia nitrogen, pH, turbidity, nitrate, total phosphorus, COD, BOD, lead, coliform count, water temperature, and water quality index. These variables serve as baseline inputs for indicator construction and subsequent scenario-based evaluation. No independent utility level, chemical supply, supplier, inventory, logistics, or operational dataset is integrated into the quantitative analysis. The downloaded dataset was first screened for missing values and physically invalid observations. No missing values were identified in the variables used for the analysis; therefore, no imputation was performed. Statistical outliers were examined using the z-score criterion, but observations were not removed solely because they were statistically extreme. For concentration variables with a physically non negative domain, including ammonia nitrogen, total phosphorus, and lead, negative values were classified as invalid observations rather than statistical outliers. These observations were excluded from subsequent indicator calculations and were not replaced by zero or by a positive concentration value. No winsorization was applied to these physically invalid observations. This procedure follows the principle that environmental outliers should be removed only when a scientific basis exists and that negative values should be specifically screened according to the nature of each parameter. After validity screening, the retained variables were normalized using the min-max transformation described in Equation 6. Thus, the preprocessing workflow distinguishes statistical outliers from physically invalid concentrations: statistical extremes are retained unless a documented error is identified, whereas negative concentrations for parameters with a nonnegative physical domain are excluded from subsequent calculations.

The z-score criterion is statistically extreme as in Equation 5:

z i j = x i j − μ j σ j (5)

where μ j and σ j denote the mean and standard deviation of the variable j. Observations exceeding a predefined threshold are treated through winsorization or removal. Third, variable selection is conducted to retain only the indicators relevant to the defined evaluation criteria. Redundant or weakly informative variables are excluded to improve model interpretability and reduce noise.

Finally, all selected variables are normalized to ensure comparability across different scales. A min-max normalization scheme is adopted as Equation 6:

x i j n o r m = x i j − min x j max x j − min x j (6)

where x i j n o r m is the normalized value of variable j for observation i. This transformation maps all variables into the interval 0 , 1, which is required for subsequent aggregation and scoring.

2.3 Criteria identification and indicator construction

The evaluation criteria represent the main consequences of chemical supply disruptions for drinking water safety and water supply resilience. Four criteria are considered: compliance and public health, service continuity, supply vulnerability, and recoverability. Compliance and public health are based on GB 5749–2022, which specifies drinking water quality requirements to protect human health and ensure safe drinking water (Liu, 2025). Service continuity represents the ability of a water supply system to maintain safe and stable service during disruption. This criterion is consistent with Chinese water supply requirements for continuous and stable water supply (Alsancak et al., 2022). Supply vulnerability represents the water supply system’s exposure to and sensitivity to external disruptions and dependencies. Vulnerability is a recognized dimension in water supply resilience assessment (Ren et al., 2020). Recoverability represents the system’s ability to restore performance after a disruption and is consistent with established resilience assessment approaches for water supply systems (Leštáková et al., 2024; Liu et al., 2021). Compliance and public health refer to the extent to which treated water meets regulatory standards and protects human health. This criterion is represented by indicators derived from pollutant concentration levels and exceedance frequencies. In this study, service continuity is represented by the temporal stability of water quality-related performance indicators. The synthetic dataset does not include direct measurements of plant uptime and physical water supply continuity, so these are not treated as directly observed variables.

Supply vulnerability represents the modeled sensitivity of treatment performance to chemical supply disruption. Because the dataset does not include direct supply chain variables, this criterion is represented through proxy indicators derived from the variability of relevant water quality parameters. Recoverability is represented through scenario-derived indicators that quantify the restoration of the normalized performance score after a simulated disruption. Recovery time is not directly available in the dataset and is therefore not treated as an observed variable. The indicators are divided into two groups. Dataset-derived indicators are calculated directly from the available water quality variables after preprocessing and normalization. Scenario-derived indicators are calculated from the baseline indicator values and the strategy-specific disruption adjustments described in Section 2.6. This distinction prevents water quality variables from being interpreted as direct measurements of chemical supply, plant operation, inventory, or recovery time.

All indicators are constructed from variables contained in the synthetic China Water Pollution Monitoring Dataset. We transform raw variables into normalized indicators to ensure comparability across criteria. Because the dataset does not contain direct information on chemical consumption, supplier dependence, inventory levels, or chemical availability, we represent supply vulnerability through proxy indicators based on the variability and exceedance behavior of selected water quality parameters. Scenario-derived indicators are subsequently calculated from baseline performance and predefined disruption adjustments.

Raw variables are transformed into normalized indicators to ensure comparability across criteria. For each observation i and criterion j, an aggregated indicator is defined as Equation 7:

I i j = 1 n j ∑ k = 1 n j x i j k n o r m (7)

where x i j k n o r m denotes the normalized value of variable k associated with criterion j, and n j is the number of variables mapped to that criterion. The constructed indicators are organized into a decision matrix as in Equation 8:

I = I 11 I 12 I 13 I 14 I 21 I 22 I 23 I 24 ⋮ ⋮ ⋮ ⋮ I m 1 I m 2 I m 3 I m 4 (8)

which serves as the input for subsequent weighting and evaluation procedures.

Dataset-derived indicators are calculated from the available water quality variables, whereas scenario-derived indicators are calculated from baseline performance and the predefined disruption adjustments (see Table 2) (Ye et al., 2026). The indicators are classified by their source and construction method. Dataset variables are obtained directly from the synthetic China Water Pollution Monitoring Dataset, dataset-derived indicators are calculated from the available water quality variables, and scenario-derived indicators are calculated from baseline performance and the predefined disruption assumptions. Regulatory standards are used only to define applicable compliance thresholds and do not constitute an additional quantitative data source.

TABLE 2. Definition and construction of indicators used in the decision model.

Figure 3 presents a schematic overview of the relationship between criteria and constructed indicators. This structure ensures that raw monitoring data are systematically translated into measurable components of system performance and resilience.

2.4 Criteria weighting using the Best Worst Method

The relative importance of the evaluation criteria is determined using the Best Worst Method. This method is selected because it produces consistent weights with a limited number of pairwise comparisons. A decision maker is required to identify the most important criterion, denoted as the best criterion C B and the least important criterion, denoted as the worst criterion C W, from the set defined in Section 2.3. Let n denote the total number of evaluation criteria. Based on this selection, we construct two preference vectors. The first vector represents the preference of the best criterion over all criteria as Equation 9:

A B = a B 1 , a B 2 , … , a B n (9)

where a B j indicates the preference of the best criterion C B over criterion C j, expressed on a predefined scale. The second vector represents the preference of each criterion over the worst criterion as Equation 10:

A W = a 1 W , a 2 W , … , a n W (10)

where a j W indicates the preference of criterion C j over the worst criterion C W. Let w j denote the weight of criterion C j, w B denote the weight of the best criterion C B, and w W denote the weight of the worst criterion C W. The optimal weights are obtained by solving the following optimization problem as Equation 11:

min w , ξ ξ (11)

subject to Equations 12–14:

∣ w B w j − a B j ∣ ≤ ξ , ∀ j ∈ 1 , 2 , … , n (12)

∣ w j w W − a j W ∣ ≤ ξ , ∀ j ∈ 1 , 2 , … , n (13)

∑ j = 1 n w j = 1 , w j ≥ 0 (14)

The objective is to minimize the maximum absolute deviation ξ, which ensures consistency between the derived weights and the pairwise comparisons. The solution yields a unique set of optimal weights as Equation 15:

w = w 1 , w 2 , … , w n (15)

A consistency ratio is also computed to assess the reliability of the comparisons. Lower values of ξ indicate higher consistency in the decision maker’s judgments. The resulting weights are used in the subsequent scoring model to aggregate indicator values and evaluate mitigation strategies under different disruption scenarios.

2.5 Data driven scoring model for strategy evaluation

A quantitative scoring model evaluates mitigation strategies by integrating normalized indicators with the weights derived from the Best Worst Method. The model provides a unified performance measure across all criteria defined in Section 2.3. Let S i denote strategy i, where i ∈ 1 , 2 , … , m and m is the total number of mitigation strategies. Let C j denote criterion j, where j ∈ 1 , 2 , … , n and n is the total number of evaluation criteria. Let I i j denote the normalized indicator value of strategy S i under criterion C j, and let w j denote the weight of criterion C j. Let S c o r e i denote the overall performance score of strategy S i. The overall performance score is computed as a weighted aggregation in Equation 16:

S c o r e i = ∑ j = 1 n w j I i j * , i ∈ 1 , 2 , … , m (16)

This formulation ensures that each criterion contributes to the final score according to its relative importance. All indicators are normalized to a common scale. For criteria with a positive performance direction, the normalized indicator is used directly. For criteria with a negative performance direction, the normalized indicator is transformed before aggregation so that higher values consistently represent better performance. The post-adjustment indicator is denoted by I i j * and is defined in Equation 17:

I i j * = 1 − I i j (17)

where I i j * denotes the post adjustment indicator value. For a criterion with a positive performance direction, I i j * = I i j. For a criterion with a negative performance direction, I i j * is obtained using the transformation in Equation 17. Therefore, I i j * represents the indicator value used in the final weighted aggregation. The scoring model is applied under each disruption scenario. Let K denote the total number of disruption scenarios and let k ∈ 1 , 2 , … , K denote the scenario index. The scenario specific performance score of strategy S i is calculated as Equation 18:

S c o r e i k = ∑ j = 1 n w j I i j * k , i ∈ 1 , 2 , … , m , k ∈ 1 , 2 , … , K (18)

where S c o r e i k denotes the overall performance score of strategy S i under disruption scenario k. Strategies are ranked based on their composite scores, with higher values indicating better overall performance. The model enables a systematic comparison of mitigation strategies by combining synthetic water quality data from the China Water Pollution Monitoring Dataset with structured decision weights. The resulting scores represent model-based evaluations under controlled synthetic conditions and should not be interpreted as empirical measurements of actual water utility performance. This data-driven formulation ensures that the evaluation process remains transparent, reproducible, and consistent across different disruption conditions.

2.6 Mitigation strategy definition and scenario design

A set of mitigation strategies is defined to reduce the impact of chemical supply disruptions on water treatment operations. The strategies are selected based on practical relevance and applicability within water utilities in China. Three representative strategies are considered, namely, dual sourcing, safety stock, and supplier prequalification. The available dataset does not directly observe these three strategies. Therefore, their effects are represented through a scenario-based proxy mapping. The water quality variables provide the baseline performance conditions, while each mitigation strategy is represented by an assumed reduction in the effective impact of supply disruption. Dual sourcing represents the availability of multiple chemical suppliers and is therefore modeled as reduced dependence on a single supplier. Safety stock represents the availability of additional chemical inventory and is modeled as a temporary buffer against supply interruption. Supplier prequalification represents the prior assessment of alternative suppliers and is modeled as a reduction in disruption impact through improved switching capacity. Let d s denote the baseline disruption intensity under scenario s, and let r j s denote the assumed reduction in disruption impact associated with strategy j under scenario s. The effective disruption intensity is defined as Equation 19:where 0 ≤ r j s ≤ 1. A higher value of r j s represents a greater assumed reduction in the effect of supply disruption. The adjusted indicator is then calculated from the corresponding baseline indicator and the effective disruption intensity. The strategy-specific effective disruption intensity d j k is used to determine the corresponding adjustment coefficient δ j k in Equation 22. Thus, the scenario adjustment applied to each strategy is not identical across strategies.

d j s = d s 1 − r j s (19)

Dual sourcing is defined as using multiple suppliers for critical treatment chemicals to reduce dependence on a single source. Safety stock is defined as maintaining additional inventory to buffer against supply interruptions. Supplier prequalification is defined as the prior evaluation and approval of alternative suppliers to enable rapid switching during disruptions. The strategy set is expressed as Equation 20:

S = S 1 , S 2 , S 3 (20)

where S 1, S 2, and S 3 correspond to dual sourcing, safety stock, and supplier prequalification, respectively. Disruption scenarios represent increasing levels of chemical supply stress and are defined as controlled model assumptions rather than observations from the synthetic water quality dataset. The severe disruption scenario is assigned an intensity of 0.50 and represents a major reduction in effective chemical availability. These values are used as predefined scenario parameters for controlled model evaluation and do not represent direct observations of chemical shortages, supplier data, or historical utility records. Three severity levels are considered: low, moderate, and severe. The scenario levels represent progressively larger reductions in effective chemical availability and treatment system performance. Their conceptual basis is informed by documented water treatment chemical supply disruptions, including production losses, logistics constraints, force majeure events, and regional or widespread shortages reported for critical treatment chemicals. The U.S. Environmental Protection Agency has documented these disruption conditions and classified historical chemical supply disruption experience into graded severity and likelihood categories (Mefteh et al., 2022). The present study does not use these external cases as quantitative observations. Instead, they provide the rationale for testing progressively more severe controlled scenarios. These scenarios are defined by changes in chemical availability and their simulated effects on water quality indicators derived from the synthetic China Water Pollution Monitoring Dataset. The scenario set is defined as Equation 21:

D = D 1 , D 2 , D 3 (21)

where D 1, D 2, and D 3 represent low, moderate, and severe disruption conditions, respectively. The calibration procedure therefore consists of assigning progressively increasing disruption coefficients to the three predefined scenarios and examining their effects on the resulting strategy scores through the sensitivity and robustness analyses.

For each strategy i and disruption scenario k, the effect of supply disruption is represented by a criterion-specific, strategy-specific adjustment. Let I i j denote the baseline normalized value of criterion C j for strategy S i, and let the corresponding scenario adjustment be associated with strategy S i under scenario k. The scenario-adjusted indicator is defined as Equation 22:

X i j k a d j = X i j 0 1 − δ i j k (22)

where δ k represents the disruption intensity associated with scenario D k. Higher values of δ k correspond to more severe disruptions. Where 0 ≤ δ j k ≤ 1. δ i j k represents the assumed proportional performance adjustment for criterion i under strategy j and disruption scenario k. Therefore, the adjustment coefficient can differ across criteria and across strategies under the same disruption scenario. The scenario-adjusted composite score is then calculated using the criterion-specific adjusted values in Equation 23:

S j k = ∑ i = 1 m X i j k adj (23)

Thus, disruption severity changes the criterion-level performance of each strategy, while differences in the criterion-specific adjustment coefficients allow the relative ranking of strategies to change across scenarios. A common adjustment coefficient applied uniformly across all criteria and strategies would produce proportional changes and would not explain the criterion-level differences reported in Table 7. Supplementary Table S6 reports the resulting strategy-specific composite scores for the low, moderate, and severe disruption scenarios. These values demonstrate the different modeled performance trajectories of the three mitigation strategies across disruption scenarios. Supplementary Table S7 reports the criterion-specific, strategy-specific disruption adjustment coefficients used for the three mitigation strategies under low, moderate, and severe disruption. These coefficients are applied in Equation 22 to obtain the scenario-adjusted criterion values reported in Table 7.

Each mitigation strategy is evaluated across all defined scenarios, enabling a comparative assessment of performance across different risk levels. This structure ensures that both the strategy’s effectiveness and the system’s resilience are examined in a consistent, controlled manner. Table 3 defines the conceptual mapping used in the model. The strategy effects are scenario assumptions, not direct measurements from the water quality dataset.

TABLE 3. Operational mapping of mitigation strategies to scenario adjustments.

3 Results and discussion

3.1 Descriptive statistics and indicator analysis

Descriptive statistics assess the variability of the water quality variables used to construct the indicators. The dataset contains 3,000 synthetic records. Table 4 presents the summary statistics.

TABLE 4. Descriptive statistics of the original synthetic water quality variables before validity screening.

Table 4 shows that most variables exhibit moderate variability, which is appropriate for discriminative analysis. The pH values are centered around neutrality, which indicates generally stable water conditions. Dissolved oxygen levels show a reasonable spread, reflecting variation in synthetic water quality across the simulated locations. Turbidity and COD exhibit relatively high standard deviations, representing variability in the synthetic water quality conditions used for indicator construction. The wide range observed in turbidity values suggests that this variable is particularly informative for compliance and public health assessment.

Nutrient-related variables such as nitrate, ammonia, and phosphorus show moderate dispersion. This supports their use in representing pollution load and treatment effectiveness. The original dataset contains several physically invalid negative concentration values. These values are excluded during the validity screening described in Section 2.2 and are therefore not used in the subsequent indicator construction. The water quality index shows a broad distribution, ranging from 0 to approximately 100, with a mean close to 50. This indicates a balanced dataset with both low- and high-quality observations, which is suitable for robust indicator construction. Table 5 summarizes the constructed indicators after normalization and aggregation.

TABLE 5. Summary of normalized indicators representing compliance and public health, service continuity, supply vulnerability, and recoverability.

The values in Table 5 indicate that service continuity shows relatively low variability, suggesting stable system performance over time. Compliance and public health exhibit moderate dispersion, reflecting differences in pollution levels across regions. Supply vulnerability shows moderate values, indicating that the system is neither highly resilient nor highly sensitive to chemical disruptions. Recoverability shows moderate performance, which suggests that restoration processes are present but not optimal.

3.2 Criteria weights derived from the Best Worst Method

The relative importance of the evaluation criteria is determined using the Best Worst Method, following the procedure described in Section 2.4. A structured preference scale from 1 to 9 is adopted, where higher values indicate a stronger preference. This scale enables consistent, discriminative comparisons between criteria. Compliance and public health are the most important criteria, while supply vulnerability is the least important. The pairwise comparison process represents the decision priorities adopted within the proposed framework, with regulatory compliance and human health protection assigned greater importance. Supplementary Tables S1-S5 provide the complete baseline and scenario-specific decision matrices, including the normalized criterion values and BWM weighted scores. These matrices provide the numerical basis for reproducing the composite strategy scores reported in Table 6.

TABLE 6. Criteria weights and ranking results obtained using the Best Worst Method.

The results indicate a clear prioritization structure. Compliance and public health account for the largest share, close to forty percent of the total weight. This confirms that water quality compliance remains the primary concern in disruption scenarios. Service continuity ranks second, reflecting the importance of uninterrupted supply and stable system operation. Recoverability is assigned a moderate weight, which indicates that the ability to restore system performance after disruption is important but not dominant. Supply vulnerability receives the lowest weight, suggesting that its influence is indirect and largely captured through its impact on other criteria.

The consistency of the pairwise comparisons is evaluated through the optimal deviation value obtained from the Best Worst Method optimization model. The computed deviation value is approximately 0.05, close to zero and indicating a high level of internal consistency in the decision maker’s judgments. According to established BWM guidelines, such a low deviation implies that the pairwise comparisons satisfy the consistency conditions with minimal contradiction. This result confirms that the preference structure between the best, intermediate, and worst criteria is logically coherent and mathematically stable. Consequently, the derived weight vector can be considered reliable for subsequent aggregation and strategy evaluation. The robustness of the weighting scheme is particularly important in this study, as it directly influences the ranking of mitigation strategies under different disruption scenarios. To further support this interpretation, Figure 4 provides a schematic representation of the comparison structure used in the BWM approach. The figure illustrates how the best criterion serves as a reference point for upward comparisons, while the worst criterion serves as a reference point for downward comparisons. This dual-reference mechanism reduces the decision maker’s cognitive burden and improves the consistency of the elicited preferences.

Figure 5 quantitatively represents the derived criteria weights and their relative contributions to the decision framework. The results show that compliance and public health account for the largest share of the total weight, approximately 39 percent, which indicates their dominant role in the evaluation process. This outcome reflects China’s strict regulatory requirements and the direct link between water quality and human health outcomes. Service continuity is assigned a weight of approximately 26 percent, making it the second most important criterion. This indicates that maintaining stable, uninterrupted treatment operations is considered a critical objective, particularly under disruptive conditions. Recoverability carries a weight of about 21 percent, suggesting that the ability to restore system performance after a disturbance is relevant but secondary to immediate operational and compliance concerns.

Supply vulnerability receives the lowest weight, approximately 14 percent. This result indicates that vulnerability is treated as an indirect factor, whose effects are largely captured through its influence on compliance and continuity. Despite its lower weight, it remains an essential component of the framework, as it reflects exposure to disruption risks. The distribution shown in Figure 5 also demonstrates a balanced weighting structure, with no extreme concentration on a single criterion. The gradual decrease in weights across criteria indicates a rational prioritization scheme rather than an arbitrary assignment. This supports the validity of the weighting results and confirms that the decision framework appropriately reflects both regulatory priorities and operational considerations.

3.3 Strategy evaluation under disruption scenarios

The data-driven scoring model defined in Section 2.5 is applied to evaluate mitigation-strategy performance under the defined disruption scenarios. The evaluation uses normalized indicators derived from the synthetic China Water Pollution Monitoring Dataset, the criteria weights obtained from the Best Worst Method, and the strategy-specific scenario adjustments defined in Section 2.6.

The criterion-level results show that all three strategies experience declining performance as disruption severity increases. Dual sourcing maintains the highest overall criterion profile across the tested scenarios, particularly for compliance and service continuity. Safety stock performs relatively well under low disruption but declines more sharply as disruption severity increases. Supplier prequalification shows intermediate performance under low disruption and becomes relatively stronger under moderate and severe disruption. The composite scores are obtained by applying the BWM criteria weights to the four criterion scores. Supplementary Table S6 also reports the complete strategy-specific composite scores under low, moderate, and severe disruption scenarios (see Table 7).

TABLE 7. Strategy-specific criterion scores under different disruption scenarios.

The difference between safety stock and supplier prequalification under severe disruption can be interpreted through supply chain ripple effect and resilience theory. A supply disruption at an upstream chemical supplier can propagate through subsequent supply chain stages and affect treatment continuity, particularly when the system depends on a limited number of supply nodes (Katsaliaki et al., 2022; Dolgui et al., 2018). Safety stock provides a time-limited buffer because its protective effect depends on the quantity available before the disruption and the duration of the interruption. As disruption severity and duration increase, the available inventory buffer can be progressively depleted, which explains the sharper decline in the modeled safety stock score. Previous research also shows that inventory can mitigate the initial ripple effect but that its effectiveness depends on disruption characteristics and inventory levels (Kek et al., 2022). Supplier prequalification provides a different resilience mechanism because it increases the capacity to switch toward alternative suppliers when the original supply node is disrupted. This flexibility can reduce dependence on a single disrupted node and limit disruption propagation through the supply network. Research on supply chain network resilience also identifies multiple suppliers and flexible supply arrangements as resilience investments that can support recovery from longer disruptions (Kek et al., 2022). Therefore, under the present scenario assumptions, the relative advantage of supplier prequalification over safety stock under severe disruption is attributed to the different resilience mechanisms represented in the model: safety stock provides temporary buffering, whereas supplier prequalification provides switching flexibility. This interpretation describes the mechanism represented by the model and does not constitute empirical evidence of superior performance in actual water treatment utilities.

The analysis considers three strategies, namely, dual sourcing, safety stock, and supplier prequalification, under low, moderate, and severe disruption conditions. Table 8 presents the computed composite scores for each strategy across all scenarios.

TABLE 8. Strategy performance scores under different disruption scenarios.

The results in Table 8 show that all strategies experience a decline in performance as disruption severity increases. This trend reflects the increasing stress imposed on the system and the reduced effectiveness of mitigation measures under extreme conditions. Under low-disruption conditions, safety stock and dual sourcing perform comparably, with dual sourcing achieving the highest score. Under the specified scenario assumptions, both the modeled inventory buffering and supply diversification strategies obtain relatively high composite scores under limited disruption conditions. This result represents a scenario-based comparison rather than empirical evidence of their actual effectiveness (see Table 9).

TABLE 9. Ranking of mitigation strategies under different disruption scenarios.

Under moderate disruption conditions, dual sourcing maintains the highest score, followed by supplier prequalification. This suggests that flexibility in supplier selection becomes more important as uncertainty increases. Safety stock declines more sharply, indicating limitations in relying solely on stored inventory during prolonged disruptions. Under severe disruption conditions, dual sourcing continues to outperform the other strategies, although the performance gap narrows. Supplier prequalification performs relatively well compared to safety stock, which declines the most. Under the specified scenario assumptions, the modeled adaptive sourcing strategies obtain higher composite scores than the modeled inventory-based strategy under severe disruption conditions. To provide a visual comparison, Figure 6 illustrates how strategy scores vary across disruption scenarios.

Figure 6 confirms the overall decline in composite performance as disruption severity increases. Dual sourcing retains the highest modeled score across all scenarios, while safety stock shows the largest decline. Supplier prequalification shows intermediate performance and becomes relatively stronger under more severe disruption conditions.

Although its initial performance is slightly lower than safety stock under low disruption, its relative decline is smaller, and it outperforms safety stock under severe disruption conditions. This indicates that flexibility in supplier selection becomes increasingly valuable as uncertainty rises. The convergence of scores under severe disruption further suggests that strategy effectiveness becomes constrained under extreme conditions, although dual sourcing still retains a clear advantage. Overall, the numerical differences observed in Figure 6 indicate that the modeled adaptive and diversified supply strategies obtain higher composite scores under the specified scenario assumptions.

3.4 Comparative ranking of mitigation strategies

The ranking remains stable across the tested scenarios. Dual sourcing ranks first under all disruption conditions. Supplier prequalification moves from third place under low disruption to second place under moderate and severe disruption, whereas safety stock moves from second to third place. This pattern indicates that supplier flexibility becomes relatively more important as disruption severity increases. Because the strategy effects are based on predefined scenario assumptions. Interpret these rankings as model-based comparisons rather than empirical estimates of strategy effectiveness.

The ranking results confirm that dual sourcing remains the highest-ranked strategy across all scenarios. Supplier prequalification moves from third to second place as disruption severity increases, whereas safety stock moves from second to third place. This pattern indicates that supplier flexibility becomes relatively more important under severe disruption conditions. Because the strategy effects are based on predefined scenario assumptions, these rankings should be interpreted as model-based comparisons rather than empirical estimates of strategy effectiveness. The ability to rapidly switch to alternative qualified suppliers provides adaptive capacity, which is particularly valuable under high uncertainty.

Safety stock ranks second under low-disruption conditions but declines to third under moderate and severe scenarios. This change reflects its limited scalability in prolonged or severe disruptions. Within the modeled scenario structure, stored inventory provides a short-term buffer against supply interruption, but its modeled contribution decreases as disruption intensity increases. To visualize the ranking patterns, Figure 7 presents the comparative ranking of strategies across scenarios.

Figure 7 shows that dual sourcing maintains a constant top position across all disruption scenarios, indicating stable, dominant performance independent of disruption severity. This consistent ranking is supported by its composite scores, which remain the highest across scenarios, reflecting its ability to mitigate supply risk through diversification. In contrast, the relative positions of safety stock and supplier prequalification vary as disruption intensity increases. Under low-disruption conditions, safety stock ranks second, with a score of 0.71, compared to 0.69 for supplier prequalification. This indicates that inventory buffering is effective when disruptions are limited in duration and magnitude. However, under moderate disruption, supplier prequalification surpasses safety stock, with scores of 0.66 and 0.65, respectively, reversing the ranking. This shift becomes more pronounced under severe disruption, where supplier prequalification scores 0.61, compared to 0.58 for safety stock. The widening gap reflects the decreasing effectiveness of static inventory strategies and the increasing importance of flexibility and supplier adaptability.

The ranking pattern across scenarios indicates that strategy performance depends heavily on disruption severity. The stability of dual sourcing suggests that diversification provides consistent protection against uncertainty. The decline in the safety stock ranking demonstrates that reliance on stored resources is insufficient during prolonged or severe disruptions. The improvement in the supplier prequalification ranking highlights the value of adaptive response mechanisms that enable rapid adjustment to changing supply conditions. The overall ranking, which places dual sourcing first, supplier prequalification second, and safety stock third, indicates that the strategies associated with diversification and supplier flexibility obtain higher composite scores under the specified scenario assumptions. These rankings do not establish causal effectiveness because direct implementation data for the three strategies are not available.

3.5 Sensitivity analysis of criteria weights

A sensitivity analysis examines the stability of the ranking results under variations in criteria weights derived from the Best Worst Method. The objective is to assess how changes in the relative importance of criteria affect the evaluation outcomes and to verify the robustness of the proposed decision framework. The analysis varies each criterion weight within a predefined range while maintaining the normalization constraint. Specifically, each weight is adjusted within ±20 percent of its baseline value, and the remaining weights are proportionally rescaled. This approach ensures that the total weight remains 1 while allowing controlled perturbations of individual criteria. Table 10 summarizes the sensitivity analysis results.

TABLE 10. Sensitivity of strategy ranking under variations in criteria weights.

The results in Table 10 indicate that the strategy ranking remains unchanged across all tested cases. Dual sourcing consistently retains the first position, while supplier prequalification and safety stock maintain the second and third positions, respectively. This stability shows that the ranking is not sensitive to moderate changes in criterion importance. To provide a visual interpretation, Figure 8 illustrates the variation of strategy scores under weight perturbations.

Figure 8 shows that variation in criteria weights produces only limited changes in the composite scores across all strategies, indicating a high level of robustness in the evaluation model. Dual sourcing consistently achieves the highest scores across all weight scenarios, ranging from approximately 0.66 in the base case to 0.68 when the compliance weight is increased, and gradually decreasing to 0.64 when the vulnerability weight is emphasized. This narrow variation confirms that its performance is not highly sensitive to changes in decision priorities. Supplier prequalification exhibits moderate variation, with scores ranging from 0.59 to 0.62. Its performance slightly improves when the compliance and service continuity weights increase, suggesting it benefits from criteria associated with operational flexibility and quality performance. Safety stock shows the lowest scores across all cases, ranging from 0.58 in the base case to 0.54 under increased vulnerability weight, indicating a consistent disadvantage relative to the other strategies.

The relative differences between strategies remain stable across all scenarios, with no overlap in score ranges that would cause rank reversals. The gap between dual sourcing and the other strategies remains evident, particularly under scenarios that emphasize compliance and service continuity. The decline in scores under increased vulnerability weight reflects all strategies’ sensitivity to disruption exposure, although the magnitude of change remains limited. Overall, the results show that the proposed model maintains consistent performance evaluation despite variations in criteria importance, confirming its robustness and reliability for decision support.

Figure 9 shows that the ranking of mitigation strategies remains unchanged across all weight-variation scenarios, confirming the robustness of the proposed evaluation framework. Dual sourcing consistently ranks first in every case, indicating that its superior performance is not dependent on a specific weighting structure. Supplier prequalification maintains the second rank across all scenarios, while safety stock consistently ranks third. This absence of rank reversal shows that the relative performance differences between strategies are large enough to withstand variations in criterion importance. The stability observed across all cases suggests that the decision model is not sensitive to moderate uncertainty in weight assignment. Even when individual criteria such as compliance, service continuity, recoverability, or vulnerability are emphasized, the overall ranking structure remains preserved. This indicates that dual sourcing’s dominance is driven by balanced performance across all criteria rather than reliance on a single factor. The consistent second-place ranking of supplier prequalification further confirms its role as a flexible, adaptive strategy, particularly when decision priorities change. In contrast, the persistent third position of the safety stock reflects its relatively weaker performance across multiple criteria, especially under conditions requiring adaptability.

3.6 Robustness analysis under scenario variations

A robustness analysis evaluates the stability of strategy performance under variations in disruption severity beyond the baseline scenarios defined in Section 2.6. The objective is to examine whether the ranking and relative performance of mitigation strategies remain consistent when disruption intensity assumptions are modified. Additional scenario variations are introduced by adjusting the disruption-intensity levels for low, moderate, and severe conditions. These variations simulate both less severe and more extreme supply shocks affecting water treatment operations. The evaluation uses the same normalized indicators derived from the China Water Pollution Monitoring Dataset and the criteria weights obtained from the Best Worst Method. Table 11 summarizes the robustness analysis results.

TABLE 11. Strategy performance and ranking under alternative disruption intensity scenarios.

The robustness analysis shows that the principal ranking pattern is preserved across the additional disruption intensity scenarios. Dual sourcing remains the highest ranked strategy, while supplier prequalification remains ahead of safety stock at higher disruption intensities. These results indicate that the ranking is stable beyond the baseline scenarios considered in Section 3.3. However, as disruption intensity increases, supplier prequalification obtains a higher composite score than safety stock under the specified scenario assumptions. This result indicates a higher modeled performance for supplier flexibility under the tested conditions rather than an empirically estimated causal effect. To visualize these trends, Figure 10 shows how strategy performance varies across different disruption intensities.

Figure 10 confirms a systematic decline in composite performance as disruption intensity increases. Dual sourcing retains the highest score across all scenarios, while safety stock shows the largest decline. Supplier prequalification remains intermediate and becomes relatively stronger as disruption severity increases. These results confirm that the ranking remains stable under the additional disruption intensity scenarios.

In contrast, safety stock exhibits the steepest decline among all strategies, with scores dropping from 0.73 under lower intensity to 0.65, then to 0.54, and finally to 0.49 under extreme disruption. This sharper reduction highlights its limited capability to handle prolonged or severe supply interruptions. The narrowing gap between strategies at higher-intensity levels suggests that extreme-disruption conditions constrain the effectiveness of all mitigation measures. However, dual sourcing retains a consistent advantage, which reinforces its role as the most robust strategy. The differences in performance trajectories indicate that strategies based on diversification and flexibility are more resilient than those based on static resource allocation. The consistent superiority of dual sourcing across all intensity levels confirms its suitability as a primary mitigation approach, while the stable performance of supplier prequalification highlights its importance as a complementary strategy.

3.7 Discussion

The results provide important insights into the role of data-driven and multi-criteria approaches in assessing disruptions to water treatment chemical supply. The dominance of compliance and public health in the weighting structure aligns with recent studies that emphasize regulatory compliance as the primary driver of water system decision-making. For example, recent work integrating the Best Worst Method with other multi-criteria techniques for water treatment systems also identifies compliance-related criteria as the most critical factor influencing mitigation strategies (Zheng D. et al., 2026; Guo et al., 2025). This alignment indicates that the prioritization structure adopted in this study is consistent with priorities reported in the water treatment and resilience literature (Luo et al., 2018; Chung et al., 2015). However, because the present analysis uses synthetic data, this consistency does not constitute empirical validation of operational or regulatory conditions in Chinese water utilities.

The ranking results indicate that, under the specified scenario assumptions, strategies based on diversification and supplier flexibility obtain higher composite scores than the modeled inventory-based strategy. Because the dataset does not contain direct measurements of strategy implementation, these results should be interpreted as scenario based comparative findings rather than empirical estimates of strategy effectiveness. The superior performance of dual sourcing across all disruption scenarios is consistent with the broader supply chain resilience literature, in which diversification is recognized as a key mechanism to enhance system adaptability and reduce dependency risk. Supply chain resilience is commonly defined as the ability to persist, adapt, and recover under disruption, which inherently favors strategies that enable flexibility and redundancy. Similarly, empirical studies show that resilience-oriented strategies that incorporate multiple sourcing channels and adaptive responses perform better under high-uncertainty conditions (Singh et al., 2023).

The modeled limitations of safety stock under severe disruption conditions are consistent with findings reported in previous supply chain resilience studies. Static buffering strategies are effective for short-term disruptions but tend to fail under prolonged or systemic shocks due to depletion and lack of adaptability. This behavior has been widely reported in disruption studies, where inventory-based approaches are shown to provide temporary resilience but do not address structural vulnerabilities in supply networks (Thakur-Weigold and Miroudot, 2024). In contrast, supplier prequalification obtains a higher modeled score under higher levels of disruption, which is consistent with the assumed role of supplier flexibility in the scenario design.

The robustness and sensitivity analysis results further strengthen the validity of the proposed framework. The absence of rank reversal under both weight variation and scenario changes indicates that the model captures stable relationships between criteria and strategy performance. This finding addresses a key limitation identified in the literature, where many resilience assessment models rely on subjective or unstable metrics. Previous research has highlighted that resilience indicators often capture only partial system properties rather than overall system behavior, leading to inconsistent evaluations (Leštáková et al., 2024). The stability observed in this study suggests that integrating normalized data with structured weighting improves the reliability of the evaluation process.

3.8 Practical implications and limitations

From a practical perspective, the findings suggest that water utilities may benefit from combining supplier diversification with supplier prequalification when preparing for chemical supply disruptions. Safety stock may provide a useful short-term buffer, but its modeled benefit decreases as disruption severity increases. These implications remain conditional on the scenario assumptions and require validation with utility-level operational data. The practical interpretation of these results is subject to important limitations. The synthetic dataset does not contain direct observations of chemical availability, supplier capacity, inventory levels, transportation constraints, procurement arrangements, or implementation of the mitigation strategies. Therefore, the strategy effects are represented through predefined scenario assumptions rather than empirically estimated relationships. The resulting rankings describe the proposed decision model’s behavior under the specified assumptions and should not be interpreted as direct estimates of strategy effectiveness in Chinese water utilities.

The criteria weights are also based on expert judgment, and the disruption scenarios use predefined intensity levels. In addition, uncertainty in the water quality index is not propagated through the final strategy scores. The transferability of the results is therefore limited by the synthetic nature of the dataset, the simplified scenario design, and the selected strategy set. Practical application requires validation with utility level data on water quality, chemical consumption, supplier arrangements, inventory, logistics, and treatment operations. Future studies should also examine dynamic disruption and recovery processes and compare the results with independently collected operational data (Ivanov and Tu, 2025; Dolgui and Ivanov, 2021).

4 Conclusion

This study evaluates the performance of three mitigation strategies for water treatment chemical supply disruptions under low, moderate, and severe disruption scenarios. The analysis combines synthetic water quality indicators, BWM-derived criteria weights, and strategy-specific disruption coefficients to assess the relative performance of dual sourcing, safety stock, and supplier prequalification. The following results summarize the main quantitative findings and the contribution of the proposed decision model.

Dual sourcing obtains the highest composite performance across the tested scenarios, while safety stock shows the largest decline as disruption severity increases. Supplier prequalification performs relatively better under moderate and severe disruption than under low disruption. These results are model-based outcomes generated from the specified synthetic dataset, criteria weights, and scenario assumptions. They do not represent empirical measurements of actual water utility performance. Sensitivity and robustness analyses further indicate that the ranking pattern remains stable within the tested parameter ranges.

The study’s main contribution is integrating quantitative water quality indicators with structured BWM weighting and scenario-based disruption assessment to evaluate chemical supply mitigation strategies. The results demonstrate the applicability of the proposed approach for comparative decision analysis under controlled conditions. The findings should be interpreted as methodological evidence rather than empirical estimates of actual performance in water utilities.

Declaration of Competing Interests: The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fenvs.2026.1952439/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Antonio Albuquerque, University of Beira Interior, Portugal

Reviewed by: Ying Zhu, Xi’an University of Architecture and Technology, China

Farshad Jahangiri, Sharif University of Technology, Iran

Sibel Eriskan, Mehmet Akif Ersoy University, Türkiye