2.1.1. The Proposed Risk-Based Consensus-Based GDSS Process

The proposed flow diagram of the risk-based consensus-based GDSS process for the IUWM is represented in Figure 1, which includes the six phases of identification and selection, weighting, evaluation, aggregation and risk analysis, consensus-seeking, and ranking:

**Figure 1.** Proposed risk-based consensus-based group decision support system (GDSS) process for the integrated urban water management (IUWM). MOWA: modified ordered weighted averaging.

In the process, first, *m* initial water strategies are scored by each of the *p* stakeholders. Additionally, *<sup>n</sup>* initial criteria are weighted by each of the *<sup>p</sup>* stakeholders. Accordingly, *<sup>C</sup>* <sup>=</sup> *C* <sup>1</sup>, ... ,*C i* , ... ,*C n* is assumed as the set of initial criteria; *S* = *S* <sup>1</sup>, ... , *S j* , ... , *S m* is considered as the set of initial feasible strategies; and *Sth*. <sup>=</sup> *Sth*.1, ... , *Sth*.*k*, ... , *Sth*.*<sup>p</sup>* is the set of stakeholders. After selection of final strategies and criteria, *m* final strategies are evaluated with regard to *n* final criteria by each of the *p* stakeholders. For convenience, *C* = {*C*1, ... ,*Ci*, ... ,*Cn*} is assumed as the set of final criteria, and *<sup>S</sup>* <sup>=</sup> *S*1, ... , *Sj*, ... , *Sm* is considered as the set of final feasible strategies. Additionally, <sup>λ</sup> = λ1, λ2, ... , λ*<sup>p</sup> <sup>T</sup>* is the vector of the stakeholders' power weights, where <sup>λ</sup>*<sup>k</sup>* <sup>≥</sup> 0. In addition, *w*(*k*) = *<sup>w</sup>*(*k*) <sup>1</sup> , *<sup>w</sup>*(*k*) <sup>2</sup> , ... , *<sup>w</sup>*(*k*) *n T* is the vector of criteria weights in the *k*th stakeholder's viewpoint (*w*(*k*) *<sup>i</sup>* <sup>≥</sup> 0 , *<sup>n</sup> <sup>i</sup>*=<sup>1</sup> *<sup>w</sup>*(*k*) *<sup>i</sup>* = 1 , *k* = 1, 2, ... , *p*).

#### 2.1.2. Identification and Selection Phase

In Steps 1 and 2 of the proposed process (Figure 1), the stakeholders are identified to select final sustainable development criteria and choose final water strategies based on the stakeholders' group consensus.

In order to select the final appropriate criteria from a large number of criteria, first the Delphi methodology is used to extract the initial criteria from the large number of sustainable development criteria by obtaining the opinions of stakeholders through a survey process [15,38,39]. After that, considering the watershed facts comprises meteorological, hydrological, and hydrogeological characteristics of the watershed, priorities of the watershed, and concepts of sustainable development criteria, all stakeholders are asked about the relevant preferences of the initial criteria. The final sustainable development criteria are selected from the set of initial criteria based on the primitive consensus-based weighted Minkowski's method using Equation (1):

$$\text{Consensus}^{(G)}\big(\mathsf{C}'\_{\mathsf{i}'}\big) = 1 - \left\{ \sum\_{k=1}^{p} \left\{ \lambda\_k \times \left| w\_{\mathsf{i}'\mathsf{i}'}^{(G)} - w\_{\mathsf{i}'}^{\prime(k)} \right|^2 \right\}^{\frac{1}{2}}, \mathsf{i}' = 1, 2, \dots, \mathsf{n}' \tag{1}$$

where λ*<sup>k</sup>* is each stakeholder's final power weight. The power weight is determined primarily by the linguistic variable followed by defuzzifying the equivalent fuzzy number and obtaining each stakeholder's final power weight. *<sup>w</sup>* (*k*) *i* and *<sup>w</sup>* (*G*) *i* denote the preference values of the *i* th initial sustainable criteria based on the *k*th stakeholder's viewpoint and group viewpoint, respectively, where *<sup>w</sup>* (*k*) *i* is determined like the stakeholders' final power weights, and *<sup>w</sup>* (*G*) *i* <sup>=</sup> *<sup>p</sup> <sup>k</sup>*=<sup>1</sup> <sup>λ</sup>*<sup>k</sup>* <sup>×</sup> *<sup>w</sup>* (*k*) *i* . In addition, *Consensus*(*G*) *C i* is the consensus measurement for the *i* th initial criteria. According to the group consensus-seeking literature, a threshold level of agreement (TLA) is determined by group of stakeholders to control the final agreement level between the individual stakeholders' viewpoints and the overall group opinion related to the initial criteria. The criteria that satisfy the condition of *Consensus*(*G*) *C i* ≥ *TLA* are selected as the final sustainable water criteria and considered as the inputs of the risk-based GDSS model.

Regarding the generate water strategies in the group decision-making process, the design theory has been widely accepted, as it is one of the most frequently used methodologies [16,40]. Accordingly, the C-K theory (concepts–knowledge) has been considered as a generative process that allows stakeholders to describe and analyze innovative design processes for generating strategies [41,42]. For operationalizing the C-K theory, the method of K-C-P (knowledge-concepts-proposals) has been proposed to manage the GDSS design process, in which multiple stakeholders could be included [43].

In this study, all details about watershed conditions, including meteorological, hydrological, and hydrogeological characteristics of watershed, water resources, water demands, and properties of sustainable development criteria, are provided for stakeholders within the questionnaire during

the survey process [16]. Post-survey, all stakeholders are asked to comment about the initial water strategies. The final water strategies are chosen from the set of initial strategies according to the primitive consensus-based weighted Minkowski's method using Equation (2):

$$\text{Consensus}^{(G)}\left(S\_{\nearrow}^{\prime}\right) = 1 - \left\{ \sum\_{k=1}^{p} \left\{ \lambda\_k \times \left| a\_{\nearrow}^{(G)} - a\_{\nearrow}^{(k)} \right|^2 \right\} \right\}^{\frac{1}{2}}, j^{\prime} = 1, 2, \ldots, m^{\prime} \tag{2}$$

where *a* (*k*) *j* and *a* (*G*) *j* represent the preference values of the *j* th initial water strategy based on the *k*th stakeholder's viewpoint and the group viewpoint, respectively, where *a* (*k*) *j* is determined like the stakeholders' final power weights, and *a* (*G*) *j* <sup>=</sup> *<sup>p</sup> <sup>k</sup>*=<sup>1</sup> λ*<sup>k</sup>* × *a* (*k*) *j* . In addition, *Consensus*(*G*) *S j* is the consensus measurement for the *j* th initial water strategy. According to the group consensus-seeking literature, the strategies that satisfy the condition of *Consensus*(*G*) *S j* ≥ *TLA* are chosen as the final water strategies and considered as the inputs of the risk-based GDSS model. The other strategies are not chosen but have the chance to be reconsidered in the iterative process of the GDSS model. Additionally, these strategies could be analyzed in the K-C-P methodology for generating new strategies.

#### 2.1.3. Weighting Phase

Regarding Step 3, the criteria weights are determined. In the MCDM problems, several methods have been applied for calculating criteria weights [38,44]. One of the most commonly used methodologies is the entropy method, which represents the dispersion of a criterion in evaluations of strategy [39,45]. In this study, the entropy method is utilized to calculate the entropy weight of each criterion by using Equation (3):

$$u\_i^{(k)} = \frac{1 + K \sum\_{j=1}^{m} \left\{ \overline{a}\_{ij}^{(k)} \times \log \left( \overline{a}\_{ij}^{(k)} \right) \right\}}{\sum\_{i=1}^{n} \left\{ 1 + K \sum\_{j=1}^{m} \left\{ \overline{a}\_{ij}^{(k)} \times \log \left( \overline{a}\_{ij}^{(k)} \right) \right\} \right\}} \tag{3}$$

where *K* = 1/ log(*n*) is a constant value, *n* and *m* are the numbers of final criteria and final strategies, respectively, and *a* (*k*) *ij* is the normalized value of *a* (*k*) *ij* . *a* (*k*) *ij* represents the evaluation value of the *j*th strategy with respect to the *<sup>i</sup>*th criterion based on the *<sup>k</sup>*th stakeholder's viewpoint. *<sup>u</sup>*(*k*) *<sup>i</sup>* is the entropy weight of the *i*th criterion in the *k*th stakeholder's viewpoint.

In this paper, in addition to the entropy weight of each criterion as an objective weight, the linguistic importance degree of each criterion is also considered as a subjective weight, which represents the stakeholders' preferences related to that corresponding criterion.

In order to express the stakeholders' viewpoints, some of the methodologies have been proposed based on the fuzzy set theory and fuzzy logic [46]. In water resources management problems, the three types of response, such as crisp response, linguistic fuzzy response, and conditional fuzzy response, could be utilized for analyzing input values [47].

In this study, the linguistic fuzzy response is used for determining the importance degrees of criteria, which utilizes fuzzy membership functions and concludes accurate outputs [48]. Accordingly, each stakeholder determines the importance degree of each criterion by using one of the linguistic members from the set of S = (No importance, Very low importance, Low importance, Slightly low importance, Moderate importance, Slightly high importance, High importance, Very high importance, Perfect importance) [49,50]. The linguistic importance degrees of criteria are fuzzified by the trapezoidal-triangular fuzzy membership functions [51,52]. The trapezoidal-triangular fuzzy membership functions, which are used for importance degrees of criteria and the stakeholders' power weights, are represented in Figure 2:

**Figure 2.** Trapezoidal-triangular fuzzy membership functions of linguistic variables. No importance (NI), very low importance (VLI), low importance (LI), slightly low importance (SLI), moderate importance (MI), slightly high importance (SHI), high importance (HI), very high importance (VHI), and perfect importance (PI).

The fuzzified variables are defuzzified by using the centroid method [51,53,54]. Accordingly, the defuzzified importance degree of each criterion (subjective weight) is determined using Equation (4):

$$\boldsymbol{\dot{w}}\_{i}^{(k)} = \frac{\int \boldsymbol{\mu} \Big(\boldsymbol{\upnu}\_{i}^{(k)}\big) \times \boldsymbol{\upnu}\_{i}^{(k)} \times d \Big(\boldsymbol{\upnu}\_{i}^{(k)}\big)}{\int \boldsymbol{\upnu}\_{i}^{(k)}\big) \times d \Big(\boldsymbol{\upnu}\_{i}^{(k)}\big)}, \; i = 1, \ldots, n; \; k = 1, \ldots, p \tag{4}$$

where *w*´ (*k*) *<sup>i</sup>* is the defuzzified importance degree of the *i*th criterion in the *k*th stakeholder's viewpoint. μ *w*´ (*k*) *i* is the trapezoidal-triangular fuzzy membership function of *w*´ (*k*) *i* .

The linguistic variables and the defuzzified values are presented in Table 1.


**Table 1.** Linguistic variables and the equivalent fuzzy interval and defuzzified values [51].

The final weight of the *i*th criterion in the *k*th stakeholder's viewpoint is determined by using Equation (5):

$$w\_i^{(k)} = \frac{w\_i^{(k)} \times u\_i^{(k)}}{\sum\_{i=1}^n w\_i^{(k)} \times u\_i^{(k)}} \tag{5}$$

Like the process illustrated for determination of the subjective weight for each criterion, the final power weight for each stakeholder (λ*k*) is also determined using Equation (6):

*Water* **2020**, *12*, 1305

$$\lambda\_k = \frac{\int \mu(\lambda\_k) \times \lambda\_k \times d(\lambda\_k)}{\int \mu(\lambda\_k) \times d(\lambda\_k)}, \; i = 1, \dots, n; \; k = 1, \dots, p \tag{6}$$

#### 2.1.4. Evaluation Phase

Regarding Step 5, a decision matrix is formed for each stakeholder, in which *m* strategies are evaluated with regard to *n* criteria. Each element of decision matrix (*a* (*k*) *ij* ) represents the evaluation value of the *j*th strategy with regard to the *i*th criterion based on the *k*th stakeholder's viewpoint.

The stakeholders' decision matrices are then normalized by using the first type of linear normalization method, which is applicable for both the positive and negative criteria based on Equations (7) and (8), respectively [55]. *a* (*k*) *ij* is the normalized evaluation value of *a* (*k*) *ij* :

$$\overline{a}\_{ij}^{(k)} = \frac{a\_{ij}^{(k)}}{a\_i^{(k)\*}} \text{ where } a\_i^{(k)\*} = \max\_j \left\{ a\_{ij}^{(k)} \right\} \tag{7}$$

$$\overline{a}\_{ij}^{(k)} = \frac{a\_i^{(k) \sim}}{a\_{ij}^{(k)}} \text{ where } a\_i^{(k) \sim} = \min\_j \left\{ a\_{ij}^{(k)} \right\} \tag{8}$$

#### 2.1.5. Aggregation and Risk Analysis Phase

In GDSS for watershed management, a group of stakeholders have several risk-taking attitudes towards decision-making, which are expressed by linguistic phrases such as "selecting the more desirable water strategy based on satisfying all of criteria" in the completely risk-averse (completely conservative or completely pessimistic) viewpoint and "selecting the more desirable water strategy based on satisfying at least one criterion" in the completely risk-prone (completely nonconservative or completely optimistic) standpoint. In addition, the other risk-taking attitudes such as "most of, many of, half of, some of, and a few of" are applied between these two cases [13,56]. Accordingly, the risktaking degree of θ has been assigned for each of the risk-taking cases [57,58]. Several risk-taking cases, equivalent linguistic phrases, and the relevant risk-taking degrees are presented in Table 2.

**Table 2.** Risk-taking cases, equivalent linguistic phrases, and relevant risk-taking degrees [58].


Regarding Step 7, for each risk-taking case, a corresponding risk-based order weights vector of *v* = (*v*1, *v*2, ... , *vn*) *<sup>T</sup>*, *vi* <sup>≥</sup> 0, *<sup>n</sup> <sup>i</sup>*=<sup>1</sup> *vi* = 1 is determined. The order weights are determined for several risk cases and the relevant risk-taking degrees of θ based on the regular increasing monotone (RIM) fuzzy linguistic quantifier and using Equation (9) [13,56,58,59]:

$$v\_i = \left(\frac{i}{n}\right)^{(\frac{1}{\theta})-1} - \left(\frac{i-1}{n}\right)^{(\frac{1}{\theta})-1}, \ i = 1, 2, \dots, n \tag{9}$$

#### 2.1.6. External Aggregation

In the external aggregation, the order weights vector for each risk-taking case is utilized to calculate the scores of water strategies in each stakeholder's opinion. In order to complete Step 8, an *n*-dimensional function of *F* : *I <sup>n</sup>* → *J* is used for a weighted normalized matrix related to each stakeholder for aggregating its evaluation values within the first aggregation. In this function, *I* denotes the set of evaluation values of each strategy, and *J* represents the corresponding score.

Therefore, according to the external risk analysis through the EMOWA operator, the evaluation values of each strategy associated with each stakeholder are aggregated to calculate the score of that corresponding strategy in several risk-taking cases using Equation (10):

$$F\_{EMOWA}^{(k)}\Big(w\_1^{(k)}\overline{a}\_{1j}^{(k)},\ldots,w\_n^{(k)}\overline{a}\_{nj}^{(k)}\Big)\big(S\_j\big) = \sum\_{i=1}^n \left\{ \left\{ \left(\frac{i}{n}\right)^{(\frac{1}{\delta})-1} - \left(\frac{i-1}{n}\right)^{(\frac{1}{\delta})-1} \right\} \times b\_i^{(k)} \right\}\tag{10}$$

where *v* = (*v*1, *v*2, ... , *vn*) *<sup>T</sup>* is the risk-based order weights vector associated with *n* criteria, for which *vi* <sup>≥</sup> 0 , *<sup>n</sup> <sup>i</sup>*=<sup>1</sup> *vi* = 1. Additionally, *b* (*k*) *<sup>i</sup>* is the *<sup>i</sup>*th largest value of the *<sup>w</sup>*(*k*) <sup>1</sup> *a* (*k*) <sup>1</sup>*<sup>j</sup>* , *<sup>w</sup>*(*k*) <sup>2</sup> *a* (*k*) <sup>2</sup>*<sup>j</sup>* , ... , *<sup>w</sup>*(*k*) *<sup>n</sup> a* (*k*) *nj* vector related to each stakeholder's weighted normalized evaluation matrix. Finally, *<sup>F</sup>*(*k*) *EMOWA Sj* is the score of the *j*th strategy from the *k*th stakeholder's viewpoint. In Equation (10), the scores of strategies from each stakeholder's viewpoint is calculated for several risk-taking cases.

Regarding Step 9, in the second aggregation, a *p*-dimensional function of *F<sup>G</sup>* : *I <sup>p</sup>* → *J* is applied to a group of stakeholders for aggregating their scorings related to each strategy. In this function, *I* denotes the set of stakeholders' scorings related to each strategy, and *J* represents the corresponding group score.

Therefore, the second aggregation step is accomplished, in which the stakeholders' scorings related to each strategy are aggregated to calculate the group score of that corresponding strategy in several risk-taking cases by using Equation (11):

$$F\_{EMOWA}^G\{\mathcal{S}\_j\} = \sum\_{k=1}^p \left\{ \lambda\_k \sum\_{i=1}^n \left\{ \left\{ \left(\frac{i}{n}\right)^{(\frac{1}{\theta})-1} - \left(\frac{i-1}{n}\right)^{(\frac{1}{\theta})-1} \right\} \times b\_i^{(k)} \right\} \right\} \tag{11}$$

where λ*<sup>k</sup>* is the *k*th stakeholder's power weight, and *FG EMOWA Sj* is the score of the *j*th strategy from the viewpoint of the group. In Equation (11), the scores of strategies from the group of stakeholders' viewpoints is calculated for several risk-taking cases.

#### 2.1.7. Internal Aggregation

In the internal aggregation, the order weights vector for each risk-taking case is directly used to calculate the scores of water strategies in the group of stakeholders' viewpoints. Accordingly, in the one-step aggregation, an *n*-dimensional function of *FG* : *I <sup>p</sup>* → *J* is used for the group weighted normalized matrix related to the group of stakeholders for aggregating its evaluation values. In this function, *I* denotes the set of group evaluation values of each strategy, and *J*represents the corresponding group score.

Therefore, with respect to the internal risk analysis performed by the IMOWA operator, the evaluation values of each strategy associated with the group of stakeholders are aggregated to calculate the score of that corresponding strategy in several risk-taking cases using Equation (12):

$$F\_{\text{IMOWA}}^{(G)}\left(w\_1^{(G)}\overline{a}\_{1j}^{(G)},\ldots,w\_n^{(G)}\overline{a}\_{nj}^{(G)}\right)|\mathcal{S}\_j\rangle = \sum\_{i=1}^n \left\{ \left\{ \left(\frac{i}{n}\right)^{(\frac{1}{\mathfrak{d}})-1} - \left(\frac{i-1}{n}\right)^{(\frac{1}{\mathfrak{d}})-1} \right\} \times b\_i^{(G)} \right\}\tag{12}$$

where *v* = (*v*1, *v*2, ... , *vn*) *<sup>T</sup>* is the risk-based order weights vector related to *n* criteria, for which *vi* <sup>≥</sup> 0 , *<sup>n</sup> <sup>i</sup>*=<sup>1</sup> *vi* = 1. Additionally, *b* (*G*) *<sup>i</sup>* is the *<sup>i</sup>*th largest value of the *<sup>w</sup>*(*G*) <sup>1</sup> *a* (*G*) <sup>1</sup>*<sup>j</sup>* , *<sup>w</sup>*(*G*) <sup>2</sup> *a* (*G*) <sup>2</sup>*<sup>j</sup>* , ... , *<sup>w</sup>*(*G*) *<sup>n</sup> a* (*G*) *nj* vector related to the group weighted normalized evaluation matrix. Finally, *<sup>F</sup>*(*G*) *IMOWA Sj* is calculated as the score of the *j*th strategy from the group of stakeholders' viewpoints for several risk-taking cases.

#### 2.1.8. Group Consensus-Seeking Phase

Regarding Step 10 (Figure 1), group consensus should be controlled to confirm that a final agreement is reached among stakeholders about water strategies. Accordingly, the consensus measurement for each strategy is calculated in order to control the final agreement amongst stakeholders associated with all water strategies.

In recent years, various methodologies have been utilized for calculating consensus measurements. Most of the frequently used methodologies have been classified in the two general approaches [60–65]. The first approach has been developed based on the hard consensus, in which the consensus measurements are calculated concerning the similarity of individual preferences compared with the group opinion [5]. Next, this is compared with the threshold level of agreement (TLA) index. The second approach has been developed according to the soft consensus, in which the individuals change their opinions collaboratively, until a consensus is reached [66,67].

In this paper, a hard consensus approach is utilized for seeking consensus among stakeholders for the first implementation of the risk-based GDSS process. After the first implementation, a soft consensus approach is used in the iterative implementation of the risk-based GDSS process if a final agreement is not reached. First, the risk-based weighted Minkowski's method is applied to calculate the consensus measurements for water strategies. In this study, the Euclidean Minkowski's distance is used for calculating the consensus measurement of each strategy, which implies a simple squared weighting and the related parameter of *q* equals 2 (*q* = 2). Regarding the relationship between the Minkowski's parameter of *q* and the risk-taking degree of decision-making [68], the Euclidean Minkowski's method minimizes the distance between the individual viewpoints and the group opinion regarding water strategies leading to a consensus amongst the majority of stakeholders [69]. By using the Euclidean distance, the score of each water strategy (*F*(*k*) *EMOWA Sj* , *j* = 1, 2, ... , *m*), determined by individual stakeholders, is compared with the score of that corresponding strategy, determined by the group of stakeholders (*F*(*G*) *EMOWA Sj or F*(*G*) *IMOWA Sj* , *j* = 1, 2, ... , *m*). The consensus measurement for each water strategy is calculated based on the EMOWA and IMOWA results, using Equations (13) and (14):

$$\text{Consensus}\_{\text{EMOWA}}^{(G)}\{\mathcal{S}\_{j}\} = 1 - \left\{ \sum\_{k=1}^{p} \lambda\_{k} \times \left| F\_{\text{EMOWA}}^{(G)}\{\mathcal{S}\_{j}\} - F\_{\text{EMOWA}}^{(k)}\{\mathcal{S}\_{j}\} \right|^{2} \right\}^{\frac{1}{2}}, \ j = 1, 2, \ldots, m \tag{13}$$

$$\text{Consensus}\_{\text{IMOWA}}^{(G)}\{\mathcal{S}\_{j}\} = 1 - \left\{ \sum\_{k=1}^{p} \lambda\_{k} \times \left| F\_{\text{IMOWA}}^{(G)}\{\mathcal{S}\_{j}\} - F\_{\text{EMOWA}}^{(k)}\{\mathcal{S}\_{j}\} \right|^{2} \right\}^{\frac{1}{2}}, \ j = 1, 2, \ldots, m \tag{14}$$

where *Consensus*(*G*) *EMOWA Sj* and *Consensus*(*G*) *IMOWA Sj* are the consensus measurements for the *j*th water strategy, where its score is calculated based on using EMOWA or IMOWA in several risk-taking cases, respectively.

According to Equations (13) and (14), it is considered that the lower distances between the individual stakeholders' viewpoints and the overall group opinion associated with each water strategy leads to higher consensus measurement for that strategy.

To control the hard consensus in this study, the TLA index is determined as the linguistic variable of "slightly high" and defuzzified to the corresponding crisp value of 0.800. The consensus measurements for strategies are compared with the selected TLA. Accordingly, the final agreement amongst stakeholders is achieved when <sup>∀</sup> *<sup>j</sup>*,*Consensus*(*G*) *EMOWA Sj* ≥ *TLA* for external aggregation or <sup>∀</sup> *<sup>j</sup>*,*Consensus*(*G*) *IMOWA Sj* ≥ *TLA* for internal aggregation. Otherwise, the risk-based GDSS process is iterated, and the soft consensus approach is implemented. According to this issue, all stakeholders are asked about their preferences related to the generation of new strategies, considering the combination of rejected strategies. The generation of strategies' process could be modeled by the K-C-P methodology. The iterative risk-based GDSS process is then implemented based on the evaluation of newly generated strategies, the combined rejected strategies, and the previously agreed strategies with respect to the final selected criteria. This process is iterated until a sufficient level of agreement is achieved amongst all stakeholders.

Ultimately, after a final agreement among all stakeholders, the water strategies are ranked based on the group scores in the several risk-taking cases.

#### *2.2. Study Area*

The study of the risk-based GDSS model is performed on the Kashafroud urban watershed area, which is located in North-Eastern Iran with a longitude of 58◦20 up to 60◦08 and latitude of 35◦40 up to 36◦03 (Figure 3). The Kashafroud watershed is one of the largest and the most populated watersheds in Iran. The mean, minimum, and maximum watershed elevations above sea level are 1846 m, 390 m, and 3302 m, respectively. The watershed has a total area of 1,565,000 ha and a growing population that is estimated to reach 5,100,000 by 2040 [70]. The total urban water demand is predicted to reach 490 million cubic meters (MCM) by 2040. This watershed has a cold and arid climatic, and the mean annual precipitation is less than 250 mm [71].

**Figure 3.** Location of the Kashafroud watershed in North-Eastern Iran.

In recent years, the Kashafroud urban watershed has encountered challenges, including an increase in the variety of water demands, quantitative and qualitative degradation of water resources, and relevant conflicts among stakeholders [72,73]. In efforts to resolve the challenges, the integrated water resources management (IWRM) approach for the Kashafroud watershed was proposed by the Iran Ministry of Energy in 2010. Since 2015, the IWRM project for this watershed has been analyzed based on the MODSIM modeling by common collaboration between the ToossAb Water Engineering Consultant Company and Iran Water Resources Management Company. The summary of the average 40-year long-term hydrological and hydrogeological budget entail results from comprehensive studies performed for this project, including the meteorological and climatic, hydrologic, hydrogeologic, and socio-economic issues [74–77] (see Appendix A, Table A1). Additionally, for several urban, agricultural, industrial, and environmental water demands of the Kashafroud watershed, the current water consumptions have been specified, and the water demands of 2040 have been predicted [71,78–80] (see Appendix A, Table A2).

According to the detailed data obtained from reports and several analysis on the watershed data, the most competitive water strategies have been modeled by the collaboration of the Iran Water Resources Management Company and ToossAb consultant company based on the iterative calibration-validation process within the MODSIM modeling project [81].

However, a GDSS model should be developed for the Kashafroud watershed based on the IWRM project data and MODSIM modeling outputs considering the stakeholders' participation in a group MCDM process. This study proposes the risk-based GDSS model for evaluating the predefined water strategies with respect to the criteria while improving the properties of the risk-based operator for modeling GDSS, analyzing the effects of several risk-taking attitudes of stakeholders based on strategies ranking, and investigating the stakeholders' consensus.

### 2.2.1. Stakeholders

A thorough and extensive study was performed in efforts to analyze the risk-based consensus-based GDSS process for the Kashafroud watershed. The six most influential stakeholders in the urban watershed decision-making process, including governmental stakeholders and non-governmental organizations (NGOs), were selected based on the study. The governmental stakeholders' members include experts, deputies, and chief executive officers (CEOs). Details on the six identified stakeholders and the relevant members for Kashafroud urban watershed are presented in Table 3.


**Table 3.** List of stakeholders and the relevant members collaborated in the Kashafroud GDSS model. NGOs: non-governmental organizations.

## 2.2.2. Initial Criteria

In order to qualify the four sustainable development objectives for the Kashafroud watershed, including water resources sustainability, environmental sustainability, economic sustainability, and social sustainability, a detailed survey was distributed in the urban watershed to collect viewpoints from the relevant stakeholder members. The survey was conducted through one-on-one interviews, collaborative workshop meetings in the presence of all members, and responses from the provided questionnaires. Fifty-three multiple criteria in the four categories of sustainable development objectives were reviewed by the stakeholders in the primitive screening process. Considering the watershed conditions and related priorities, 21 criteria were voted as the initial criteria and are represented in Table 4. These criteria are defined according to reports provided by the United Nations Educational, Scientific, and Cultural Organization (UNESCO), International Association of Hydrogeologists, and the national reports provided by Iran Water Resources Management Company in the IWRM project [82–86].


**Table 4.** The initial sustainable development criteria for the Kashafroud watershed [82–85].

The initial criteria were weighted in the final screening process to select the final criteria based on a group consensus.

#### 2.2.3. Water Strategies

After the investigation of several water strategies by the Ministry of Energy and the watershed stakeholders, the most competitive urban water strategies were selected by the stakeholders for the IWRM project to make a decision about choosing the more desirable strategy within the Kashafroud watershed [81]. Table 5 presents the five final water strategies for the Kashafroud urban watershed classified by supply management, demand management, and combined supply-demand management.

The main reasons for the selection of these five competitive strategies by the stakeholders include:



**Table 5.** The urban water strategies for the Kashafroud watershed by the 2040 vision.

The existing water resources include the Ardak, Kardeh, Torogh, Dolatabad, Chalidarreh, and Esjil dams, as well as the groundwater reservoir. The under-studying supply management strategies include the utilization of purified wastewater on agricultural lands and the Idelik inter-basin water transfer. However, the Idelik project might cause conflicts between the stakeholders of the northern watershed and the Kashafroud watershed. The multi-criteria effects of this project and utilization of purified wastewater on agricultural lands are compared for the strategies *S*<sup>2</sup> and *S*3.

The Doosti Dam is considered as a structural supply management that plays an active role in supplying water for the Kashafroud watershed. However, this project has high operation and maintenance costs, and its implementation could lead to dependence on the transboundary river basin. Therefore, the stakeholders' approach is to substitute more reliable water strategies instead of the dam for providing urban water.

The under-studying demand management strategies include improving water network efficiency and modifying cropping patterns, which are typical for strategies *S*<sup>4</sup> and *S*5. The difference between these two strategies is that the strategy *S*<sup>4</sup> is considered as just a demand management approach with dependency on the Doosti Dam, while the strategy *S*<sup>5</sup> is considered as both a supply and demand management approach with no dependency on water transfer from the Doosti Dam.

The existing and under-studying strategies of the Kashafroud watershed are shown in Figure 4.

#### 2.2.4. Data Collection

In order to analyze the risk-based GDSS model for selecting the more desirable water strategy, two types of data were collected. The first type of data is related to the criteria, including the selection of final criteria and weighting of the final criteria. The second type of data is associated with the strategies, including the evaluation of strategies with respect to the final criteria.

Accordingly, for collecting the first type of data, a survey questionnaire was prepared, and the 21 members of the six stakeholders were interviewed to capture their viewpoints about the importance degrees of the initial criteria and the final criteria, using the linguistic answers (no importance, very low importance, low importance, slightly low importance, moderately importance, slightly high importance, high importance, very high importance, and perfect importance) (see Appendix B, Table A3)

For collecting the second type of data, the results of the MODSIM modeling project and the data from the meteorology and climatology, hydrology, hydrogeology, and socio-economic reports [74–77,79], as well as information from the reports of urban, agricultural, industrial, and environmental water demands for the Kashafroud watershed, were utilized for evaluation of the water strategies with respect to the sustainable development criteria (see Appendix A, Tables A1 and A2; see Appendix C, Table A6).

**Figure 4.** Location of Kashafroud in Iran and the existing and the under-studying water strategies.

#### 2.2.5. Final Criteria Selection

Regarding Step 2 of the GDSS model (Figure 1), in order to consider the four sustainable development objectives for evaluating the five water strategies, the final criteria should be selected from the initial criteria. The first step in the survey process is an interview with the stakeholders, where the definitions of the initial criteria are explained. Next, the provided questionnaires are completed by the 21 members of the six stakeholders, in which the linguistic importance degrees of the initial criteria are assigned (see Appendix B, Table A4). In the end, the members' viewpoints of each of the six stakeholder's community are aggregated. The aggregated results related to the six stakeholders on the defuzzified weights of the initial criteria and the stakeholders' defuzzified normalized weights are presented in Figure 5.

**Figure 5.** Assigned weights for the initial criteria by the stakeholders of the Kashafroud watershed.

The initial criteria weights are utilized to calculate the relevant group consensus measurements using Equation (1). The group consensus measurement results of the initial criteria are presented in Figure 6.

**Figure 6.** The group consensus measurements of the initial criteria for the Kashafroud watershed.

In order to select the final criteria from the 21 initial criteria, the group consensus measurements that are higher than the determined TLA (*TLA* = 0.800) are selected as the final criteria. According to Figure 6, the 10 black-filled criteria of *C* 1, *C* 2, *C* 5,*C* 9,*C* 12,*C* 14,*C* 16,*C* 17,*C* 18, and *C* <sup>21</sup> have been selected as the final sustainable development criteria, which are the most preferable criteria in the group viewpoints for the decision-making process within the watershed. The final selected criteria that are marked by *C*1, *C*2, *C*3,*C*4,*C*5,*C*6,*C*7,*C*8,*C*9, and *C*10, are defined in Table 6 [82–85].

**Table 6.** The final sustainable development criteria for the Kashafroud watershed by the 2040 vision. *TWW* is the total water withdrawal (includes urban, agricultural, and industrial and water withdrawal); *TWS* is the total water storage; *EGW* is the exploitation from groundwater; *TEWR* is the total exploitation of water resources (*TWW* + environmental water withdrawal); *SAWD* is the supplied agricultural water demand; *AWD* is the agricultural water demand; *SWR* is the surface water resources; *RGW* is the renewable groundwater resources; *P* is the population; *SPWD* is the supplied potable water demand; *PWD* is the potable water demand; *SIWD* is the supplied industrial water demand; *IWD* is the industrial water demand; *SEWD* is the supplied environmental water demand; *EWD* is the environmental water demand; *GWD* is the groundwater discharge; *GWR* is the groundwater recharge; *PS* is the purified sewerage; *US* and *IS* are the urban sewerage and industrial sewerage, respectively; and *B* and *C* are the amount of benefit and cost values of the implementation of the water strategies, respectively.

