Improving Agricultural Green Supply Chain Management by a Novel Integrated Fuzzy-Delphi and Grey-WINGS Model

Abstract

This study suggests a novel hybrid model for calculating the interrelationships between factors by integrating the Fuzzy set, Delphi, the Grey theory, and Weighted Influence Nonlinear Gauge System (WINGS) approaches in agricultural green supply chain management (AGSCM). Fuzzy Delphi helps to select 12 indicators from 19 factors by defuzzification for ambiguity associated with subjective judgment by 10 experts in data collection. Grey WINGS can illustrate the relationships, direction, and strength of factors simultaneously, which illustrates that environmental law, green consciousness, product quality, and price are the most significant factors of AGSCM. The results can help operators not only to analyze these key influencing factors, but also to understand the complex cause-and-effect relationships between these factors. This integrated model will hopefully provide a useful tool to agricultural policy makers and decision makers for sustainable development.

1. Introduction

Agricultural green supply chain management (AGSCM) aims to transform environmental constraints into advantages and opportunities, such as eco-brand, green consumption, and sustainable development, which is difficult to optimize influenced by complex and interactive factors intrinsic in an ever-changing complex environment, which includes global warming, the COVID-19 pandemic, and environmental pollution. In particular, agriculture is one of the largest sources of methane emissions, with the Food and Agriculture Organization (FAO) stating that the emission of greenhouse gas will increase by 30% by 2050 [1]. With the increasing concern for green and sustainable development, more and more consumers are forcing the traditional supply chain reform to become environmentally conscious with components, such as biological pesticides, renewable energy, recyclable packaging, environmentally friendly fertilizers, and so on [1,2].

As we know, the yield of agricultural products is particularly influenced by many uncertain challenges related to environmental, political, economic, social, technical, and legal dimensions, which has become a major issue affecting human beings in recent years [3]. A growing global population and a deteriorating environment have led to an increased focus on agricultural supply chains, such as resource constraints and environmental pollution [4]. With the growing environmental awareness, decision makers must take environmental factors into account in supply chain management. The implementation of environmental and social performance expands the scope of legal, social, technical, economic, and ethical properties in green supply chain management (GSCM) [5]. Furthermore, the performance of GSCM combines environmental, social, and economic dimensions, which must be considered in many interrelated operations, such as planning, production, packaging, transportation, storage, processing, distributing, publicity, and sales [6,7,8]. Sustainability has become a necessary obligation for enterprise development. Enterprises need to take responsibility for social and environmental issues in supply chain management [9]. However, AGSCM has become more difficult with the spread of the COVID-19 pandemic, global warming, extreme climate, and environmental pollution across the world.

Although there have been a few attempts to study agricultural green supply chain management [10,11,12], these studies mainly studied the factors which are independent of each other as a prerequisite assumption but ignored the interrelationships within them. This assumption may limit the development of AGSCM and the improvement of economics. However, there are many uncertain complex hierarchical factors affecting AGSCM, such as perishability, seasonality, customers’ demand, and supply relationships [13]. In order to improve development of AGSCM within the restrictions of available natural resources, the decision support model must be concentrated on the real-world scenario and integrated with complicated methods to evaluate performance and the relationship of every factor [14].

Multiple-Criteria Decision-Making (MCDM) methods are designed to address complex decision-making difficulties by analyzing the structure of criteria, alternatives, and decision-makers’ preference, which are suitable for assisting managers, practitioners, and developers in selecting the best options within various conflicting criteria. Saaty introduced the Analytic Hierarchy Process (AHP) as a popular MCDM approach in 1980. The hierarchical structure of AHP makes it possible to visualize the factors influencing the alternatives. Analytic Network Process (ANP) is an amplification of AHP which can take into account the intricate interdependence of decision factors in a hierarchical structure [15]. To deal with the uncertain situation, the fuzzy AHP and ANP have been used in many domains [16,17]. So, the hybrid MCDM methods have the advantage to accomplish analysis the imprecise, incomplete, or uncertain information.

In contrast to the methods mentioned above, the Decision-Making Trial and Evaluation Laboratory (DEMATEL) is an advanced and sophisticated decision-making method for addressing interdependencies by visualizing the causal interactions of indicators proposed by Gabus and Fontela [18]. DEMATEL uses mathematical tools to comprehend various specialists’ perspectives on associated factors, as well as logical correlations and direct effects between these factors [19], which has been widely used in supply chain management (SCM) [20,21,22]. Michnik developed the Weighted Influence Nonlinear Gauge System (WINGS) approach from DEMATEL [23]. With interdependencies of factors in MCDM situations, DEMATEL simulates the direction and strength of the impact. Furthermore, WINGS simulates both the intensity and direction of the influence, in addition to the strength of the criterion, which could be utilized as a theoretical basis for AGSCM. However, classical DEMATEL and WINGS methods ignore the vagueness and uncertainty of human judgment that are so prevalent in real life. Regarding this problem, the Grey theory may successfully handle the ambiguities inherent in human subjective judgement while acquiring accurate results with a moderate data sample.

The contribution of this paper can be summarized as follows.

• A fuzzy-Delphi and grey-WINGS approach to decision theory, which can be utilized to analyze different group choices, ambiguity, and complex interrelationships in evaluation problems, is presented in this study. The combination of a fuzzy set and grey theory can provide a more realistic representation of human judgement under ambiguous and subjective conditions.
• The target of this study is conducive to the improvement of AGSCM by applying the current assessment approach to provide a more accurate and objective prioritization tool for AGSCM in a hazy and diverse environment. The approach is intended to assist AGSCM designers in identifying the most critical factors with the highest potential.
• The fuzzy-Delphi and grey-WINGS method integrates four techniques, which have not been combined for illustrating mutual relationships of factors in previous studies. According to the results analysis, this research contributes significantly to improving AGSCM by providing policy and management implications.

The remainder is arranged as follows. Section 2 contains literature reviews. Section 3 consists of materials and methods. Section 4 includes research results. Section 5 is discussions. Section 6 includes conclusions.

2. Literature Review

2.1. Agriculture Green Supply Chain Management

GSM is always known as the environmental supply chain based on green manufacturing theory, which was first introduced by the Manufacturing Research Society of Michigan State University. The enterprises, merchants, and farmers within the supply chain can gain benefit from GSCM and use it as a valuable resource to improve their environmental performance [24] because it is a vital important management system involving suppliers, production plants, distributors, and customers, with the aim of minimizing the negative impacts and maximizing efficiency of resource utilization through the improvement of the whole implementation incorporating environmental concepts [25,26].

Agriculture is one of the industries that is most affected by climate. There is a clear relationship between agricultural productivity and climate fluctuations, which is especially complex and unique in developing countries [27]. Moreover, agricultural products have several specific characteristics that make agricultural supply chain management (ASCM) more complicated due to factors associated with seasonality, environment, and perishability when compared with typical supply chains [28]. In order to maintain environmental sustainability, the ‘green’ concept integrates environmental and ecological concerns, which has a significant impact on the environment including pollution, emissions, the health hazard to human beings, etc. [29,30,31,32,33,34,35]. Therefore, AGSCM has been established as an important discipline of sustainable operations management, which must be paid more attention. As more and more environmental regulations are published, AGSCM plays a proactive role in improving environmental performance and economic stability [36,37].

Environment, strategy, and logistics are the three critical components of AGSCM, involving proactive measures such as recycling, reprocessing, and monitoring of environmental standards [38,39]. In order to improve sustainable development, it is necessary that the product, package, and purchase must meet green standards. All supply chain participants must be proactive and work together to minimize negative environmental effects [40].

2.2. The Influence Factors of AGSCM

AGSCM incorporates environmental and economic elements, which are important challenges with the limitation of resources for minimizing environmental negative consequences and enhancing economic stability [36,41]. We reviewed papers with GSCM and AGSCM from 2010 to 2022. According to the operation of the supply chain, the forces driving all farmers, stakeholders, and customers should be associated with activities in AGSCM. It can be concluded that the factors influencing AGSCM include customer and stakeholder requirements and competitive advantage, both of which come from economic and social factors [42,43]. These are the motivations for successful implementation of AGSCM. On the other hand, regulation and market pressure could force companies to adopt the rules of AGSCM in the pursuit of environmental performance, such as environmental laws, competitors’ pressure, suppliers’ requirements, and customers’ awareness [44,45,46]. Meanwhile, barriers are factors that hamper the implementation process of AGSCM. Some of the important barriers are cost and risk, lack of government support, financial constraints, poor supplier commitment, lack of legitimacy, technology, and resistance from the stakeholders [47,48,49].

2.3. Hybrid Methodology and Applications

Compared to traditional SCM, the ASCM is more difficult to measure due to the issues associated with environmental factors. Some structural methodologies have been extended in ASCM, such as AHP [50], ANP [51], and Technique for Order Preference by Similarity to an Ideal Solution (TOPSIS) [52] in

Table 1. Overview of relevant studies with MCDM methods.

Findings	Approach	Relevant Literature
Delphi and Fuzzy AHP methods are constructed to estimate the factors of green design, purchasing, production, warehousing, and logistics in supply chain practices.	Delphi, Fuzzy AHP	[50]
The research discusses the four main criteria, including product quality, production cost, customer requirements, and delivering time to select an effective supplier by Fuzzy ANP.	Fuzzy ANP	[51]
This article studied green supplier selection with the criteria of service, quality, price, and environment by using fuzzy TOPSIS.	Fuzzy TOPSIS	[52]
The case study developed a hybrid model with AHP and TOPSIS to evaluate a supply chain perspective within economic, environmental, and social dimensions.	Fuzzy AHP, TOPSIS	[59]
This work uncovered ten main factors to sustainable initiatives for ASCM, such as government pressure, stakeholder requirements, monitoring and auditing, competitive advantages, cost, and benefits, by using Fuzzy DEMATEL.	Fuzzy DEMATEL	[26]
This method proposes a model combing DEMATEL and ANP to assess indicators such as services, technology, environmental, financial, and economic dimension in sustainable supplier selection.	Fuzzy ANP, DEMATEL	[53]

DEMATEL: Decision-Making Trial and Evaluation Laboratory; TOPSIS: Technique for Order Preference by Similarity to an Ideal Solution; ANP: Analytic Network Process; AHP: Analytic Hierarchy Process.

. Furthermore, AGSCM is also a typical MCDM problem that requires estimation of factors based on complex objective and subjective information. In order to achieve accurate and scientific evaluation results, the fuzzy set theory might be a major tool which was initiated as a mathematical tool to handle ambiguity and fuzzy information influenced by subjective judgments. These sources of imprecision contain incomplete, nonobtainable, unquantifiable, and partial information, which exist in real life. Therefore, the fuzzy set can be employed in some decision models to analyze the factors of supply chain management [53]. In order to exploit the ambiguity and variety in articulating preferences of decisionmakers, grey theory has been utilized to assemble group fuzzy evaluations, and it can handle the preferences of various decisionmakers [54,55]. Compared with the fuzzy logit, grey theory can handle uncertainty problems with discrete values and imperfect knowledge by creating a flexible choice model with interval numbers. Its significant advantage is the capability to obtain accurate results with limited data under conditions of high variable variability [56]. By combining linguistic variables, the grey set theory can be used to evaluate uncertain conceptions related to people’s subjective judgments. The implications of the grey set theory will be more significant, especially when experts make decisions based on inadequate information or when they are conscious that they lack knowledge in some scenarios. Numerous effective applications of grey system theory have been made in several fields, including business, geography, medicine, agriculture, and disaster preparedness [57,58]. So, grey system theory has been improved as an efficient approach to unresolved and ambiguous issues in recent years.

The Delphi technique is a qualitative approach for gathering the opinions of a diverse group on a specific topic, which was proposed by the RAND Corporation. Because traditional Delphi techniques cannot deal with ambiguity, a fuzzy-Delphi method, which can handle the ambiguity and uncertainty inherent with the data, was combined by Ishikawa et al. (1993) [60]. Various applications have been employed in supply chain performance, agricultural cost, design analysis of products, healthcare, and construction [61,62,63,64,65,66]. Moreover, to analyze the complex intertwined relationships between influencing factors, scholars proposed several powerful methods including ANP [15,51], DEMATEL [56,59], and WINGS [23,67]. ANP is just a generalized version of the analytical hierarchy process proposed by Saaty, which illustrates general relations among the indicators, whereas the AHP emphasizes hierarchical relations between decision levels [68]. The ANP uses ratio scale measurements through comparisons, but unlike the AHP, it does not impose a fixed hierarchical structure. Both methods have a prerequisite assumption as no influence between criteria. ANP has been widely applied in various situations, such as location selection, project selection, and supplier selection [69,70,71].

DEMATEL is used to translate the interrelationships between the criteria into an understandable structural model, which was established by the Battelle Memorial Institute of the Geneva Research Center. The numbers measuring the level of influence can construct the matrices or digraphs to illustrate the interrelationship between criteria and identify the core criterion to express the performance of variables, which could also eliminate overfitting for assessment [21]. Being an update of DEMATEL, WINGS takes over the superiority of DEMATEL, including the ability to handle complex problems with various factors and the simplicity of its mathematical procedures [23]. However, it also has its own special characteristics. WINGS measures the operating factors’ strength and the level of its influence, whereas DEMATEL only considers the latter. So, an improved version of WINGS can be used to evaluate the interrelationships between criteria more powerful. Especially when the criteria are distinct, it has been demonstrated that WINGS simplifies the additive agglomeration as shown in

Table 2. Comparations within different approaches.

Approach	Interdependencies of Factors	Intensity of Impact	The Strength of Factors	Group Fuzzy Assessments
FAHP	-	-	s	-
FANP	-	-	s	-
FDEMATEL	s	s	-	-
Grey–WINGS	s	s	s	s

WINGS: Weighted Influence Nonlinear Gauge System.

[72].

Based on the above analysis, the DEMATEL and WINGS techniques are superior than other traditional methods, since the input values can immediately enter the matrix, which has the advantage in calculations over the AHP/ANP approach with pairwise comparisons. However, WINGS method is superior than the classical DEMATEL method, which considers the strength of the standard., as well as the interrelationship between criteria. Unlike the previously mentioned approaches, this study combines the WINGS and DELPHI methods with fuzzy and grey theory to handle the fuzzy decision environment. Therefore, the suggested model of this paper is more accurate in describing the subjective information and more practicable in analyzing the difficult assessment problems with simple calculation. In addition, there is no instance integrating the grey theory, the fuzzy set, the WINGS, and the DELPHI approaches, which involves the ambiguity and uncertainty during the evaluation process.

3. Materials and Methods

The proposed model combining fuzzy Delphi and Grey WINGS contains two phases as in Figure 1. Firstly, identifying and finalizing the factors of AGSCM. Secondly, a cause-and-effect analysis of the components that have been selected will demonstrate how they interact.

3.1. Influencing Factors of AGSCM

Based on the status of the AGSCM and structural analysis approaches that have been applied to supply chain management, an evaluation method of the influencing elements of AGSCM has been constructed. We chose 19 factors from three dimensions including government, economy, and society, including green consciousness, competitive pressure, government subsidies, produce quality, customer demand, environmental laws, logistics, renewable material, green operation, technology, waste reduction, price of product, cost, stockholders’ requirement, monitoring, social responsibilities, infrastructure, income level, and reusable packaging.

3.2. Fuzzy Delphi

The theory of fuzzy sets proposed by Zadeh to describe the ambiguity of human cognitive processes formed the basis of the fuzzy-Delphi technique. A triangular fuzzy number can be presented like $\tilde{\lambda }=\left(l,o,k\right)$, where $l\le o\le k$. Then, the membership function is:

$${\theta }_{\tilde{\lambda }}=\{\begin{array}{l}\frac{x-l}{o-l},\hspace{1em}x\in (l,o)\\ \frac{k-x}{k-o},\hspace{1em}x\in (o,k)\\ 0,\hspace{1em}\hspace{1em}x\in (-\infty ,l)\cup (k,\infty )\\ 1,\hspace{1em}\hspace{1em}x=o\end{array}$$

(1)

The basic operations show as:

$$\begin{array}{l}(1){\tilde{\lambda }}_1+{\tilde{\lambda }}_2=\left(l_1,o_1,k_1\right)+\left(l_2,o_2,k_2\right)=\left(l_1+l_2,o_1+o_2,k_1+k_2\right);\\ (2){\tilde{\lambda }}_1-{\tilde{\lambda }}_2=\left(l_1,o_1,k_1\right)-\left(l_2,o_2,k_2\right)=\left(l_1-l_2,o_1-o_2,k_1-k_2\right);\\ (3){\tilde{\lambda }}_1\times {\tilde{\lambda }}_2=\left(l_1,o_1,k_1\right)\times \left(l_2,o_2,k_2\right)=\left(l_1{l}_2,o_1{o}_2,k_1{k}_2\right);\\ (4){\tilde{\lambda }}_1÷{\tilde{\lambda }}_2=\left(l_1,o_1,k_1\right)÷\left(l_2,o_2,k_2\right)=\left(l_1/k_2,o_1/o_2,k_1/l_2\right).\end{array}$$

(2)

where $l_1,l_2>0;o_1,o_2>0;k_1,k_2>0.$

The following are all fuzzy-Delphi steps:

Step 1: This process involves identifying and categorizing numerous factors that are relevant to the field under research.

Step 2: Once the criteria have been established, the experts are given the questionnaire detailing the criteria to compare by using the linguistic scale listed in

Table 3. Linguistic scale with triangular fuzzy number.

Linguistic Values	Numbers	Corresponding Triangular Fuzzy Number
Very unimportant	1	(0.1,0.1,0.3)
Unimportant	3	(0.1,0.3,0.5)
Normal	5	(0.3,0.5,0.7)
Important	7	(0.5,0.7,0.9)
Very important	9	(0.7,0.9,0.9)

. Fuzzy numbers could be transformed from experts’ evaluations for each criterion. A fuzzy number referring to the cth factor suggested by the ath expert is expressed as:

$$e_{ca}=(l_{ca},o_{ca},k_{ca});c=1,2...p;a=1,2...q$$

(3)

where p and q are the number of criteria and experts.

The fuzzy number for each criterion could be estimated using triangular fuzzy numbers (E), as stated in Equation (4), which integrates the evaluations from all q experts as follows:

$$E_c=(l_{cD},o_{cM},k_{cH})=\left({\min }_q{l}_{cD}^q,{({\prod }_{a=1}^q{o}_{cM}^a)}^{1/q},{\max }_q{k}_{cH}^q\right)$$

(4)

Step 3: The fuzzy number of each assessment factor should be defuzzied using the Simple Center of Gravity (SCGM) approach to obtain the final value of each factor, which is the most prevalent approach for defuzzification [73]. This stage of SCGM involves computing the defuzzification value G using the mean approach as shown below:

$$G_c=(l_{cD}+o_{cM}+k_{cH})/3$$

(5)

Step 4: A threshold value (β) must be defined to choose the most significant criteria from the expert group in order to create the list of criteria. The final step is to construct the final list of criteria based on the following threshold criteria: The criterion is chosen if G ≥ β, and the criterion is omitted if G ≤ β.

3.3. Fuzzy-Delphi Grey-WINGS Model

The main steps can be described as:

Step 1. Determine selection criteria by using the fuzzy-Delphi method.

Numerous factors relevant to AGSCM are estimated by experts. After gathering expert opinions from surveys, the triangle fuzzy numbers are utilized to determine selection criteria through the Delphi method.

Step 2. Construct an initial strength–influence matrix for all experts.

Table 4. Grey linguistic scales.

Linguistic Variables	Influence Number	Related Grey Numbers
None (N)	0	[0.0,0.0]
Low (L)	1	[0.0,0.25]
Medium (M)	2	[0.25,0.5]
High (H)	3	[0.5,0.75]
Very high (VH)	4	[0.75,1.0]

displays the language evaluation and the related grey numbers, which could measure factor x impact over factor y using an integer scale ranging from 0 to 4, indicating “no influence”, “low influence”, “medium influence”, “high influence”, and “very high influence” between factors.

Step 3. Compute the corresponding grey matrix for the strength–influence matrix.

The ratings on the integer scale can be transformed into corresponding grey scales that give an upper range and a lower range of values. Based on the obtained grey values, the initial relation matrices are transformed into grey relation matrices, as $\otimes r_{xy}^a=\left[\underset{\_}{\otimes }{r}_{{}_{xy}}^a,\overline{\otimes }{r}_{xy}^a\right]$, where x,y indicate the criterion, and a indicates the ath expert, 1 ≤ a ≤ q; 1 ≤ x ≤ p; 1 ≤ y ≤ p.

Step 4. Calculate the average grey strength–influence matrix.

The average grey strength–influence matrix $\left[\otimes {\stackrel{⌢}{r}}_{xy}^{}\right]$ can be computed by q grey relation matrices,

$$\otimes {\stackrel{⌢}{r}}_{xy}^{}=\left(\frac{{\displaystyle {\sum }_a\underset{\_}{\otimes }{r}_{xy}^a}}{q},\frac{{\displaystyle {\sum }_a\overline{\otimes }{r}_{xy}^a}}{q}\right)$$

(6)

Step 5. Obtain the crisp strength–influence matrix.

(1) Standardization of the grey number:

$$\overline{\otimes }{\tilde{r}}_{xy}=\left(\overline{\otimes }{\stackrel{⌢}{r}}_{xy}-\min \overline{\otimes }{\stackrel{⌢}{r}}_{xy}\right)/\left(\max \overline{\otimes }{\stackrel{⌢}{r}}_{xy}-\min \underset{\_}{\otimes }{\stackrel{⌢}{r}}_{xy}\right)$$

(7)

$$\underset{\_}{\otimes }{\tilde{r}}_{xy}=\left(\underset{\_}{\otimes }{\stackrel{⌢}{r}}_{xy}-\min \underset{\_}{\otimes }{\stackrel{⌢}{r}}_{xy}\right)/\left(\max \overline{\otimes }{\stackrel{⌢}{r}}_{xy}-\min \underset{\_}{\otimes }{\stackrel{⌢}{r}}_{xy}\right)$$

(8)

(2) Normalization of the crisp values:

$$t_{xy}=\left(\underset{\_}{\otimes }{\tilde{r}}_{xy}\left(1-\underset{\_}{\otimes }{\tilde{r}}_{xy}\right)+\left(\overline{\otimes }{\tilde{r}}_{xy}\times \overline{\otimes }{\tilde{r}}_{xy}\right)\right)/\left(1-\underset{\_}{\otimes }{\tilde{r}}_{xy}+\overline{\otimes }{\tilde{r}}_{xy}\right)$$

(9)

(3) Calculate the accurate total crisp values.

$$f_{xy}=\min \overline{\otimes }{\tilde{r}}_{xy}+t_{xy}\left(\max \overline{\otimes }{\tilde{r}}_{xy}-\min \underset{\_}{\otimes }{\tilde{r}}_{xy}\right)$$

(10)

and $F=\left[f_{xy}\right]$

Step 6. Obtain the normalized strength–influence matrix.

$$W=\frac{1}{{\displaystyle \sum _{x=1}^p{\displaystyle \sum _{y=1}^p{F}_{xy}}}}\mathrm{and}B=W\times Fx,y\in \left\{1,2,...,p\right\}$$

(11)

The element of matrix $B$ is between 0 and 1.

Step 7. Acquire the total strength–influence matrix.

The matrix $Z$ is obtained by:

$$Z=B{\left(I-B\right)}^{-1}$$

(12)

where $Z=\left[z_{ca}\right]$, and I presents an identity matrix.

Step 8. Sum of rows and columns in matrix Z.

The sum of rows (T) and columns (L) in matrix Z can be calculated as:

$$T=[T_c]={\displaystyle {\sum }_{c=1}^p{z}_{ca}},c=1,2,...,p$$

(13)

$$L=[L_a]={\displaystyle {\sum }_{a=1}^q{z}_{ca}},a=1,2,...,q$$

(14)

T depicts the whole influence of component c as a cause affecting remaining components, while L illustrates an effect as the whole influence from other components impacting component a.

Step 9. Set up cause–effect relationship diagram.

Using the values obtained through Equations (13) and (14), a causal diagram is set up. The total impacts the given and received values by factor x, which represents the degree of prominence in the overall system.

The sum (T + L) presents the total effects by factor x, which represents the degree of prominence in the overall system, while (T − L) illustrates the net effect of factor x on the overall system. Factor x is the net cause if (T − L) is positive. Then, factor x is the net effect if (T − L) is negative.

Step 10. As shown below, a threshold value (β) is established to eliminate minor effects.

$$\beta =\raisebox{1ex}{${\displaystyle {\sum }_{x=1}^p{\displaystyle {\sum }_{y=1}^p\left[z_{xy}\right]}}$}\!\left/ \!\raisebox{-1ex}{$N$}\right.$$

(15)

where N is the number of factors in matrix Z.

4. Results

4.1. Data Collection and Fuzzy Delphi

The main steps can be described as:

For the purpose of gathering data, 10 experts from agricultural businesses and academics were engaged. The experts team consisted of 2 professors within the agriculture field, 2 agricultural consultants, 2 agricultural supply chain managers, 2 rural cooperative managers and 2 farmers, who all have an experience of more than 12 years.

Table 5. Information of the experts.

No	Gender	Position	Work Experience
Exp 1	Male	Professor	20
Exp 2	Female	Professor	22
Exp 3	Male	Rural cooperative manager	18
Exp 4	Male	Rural cooperative manager	15
Exp 5	Male	Supply chain manager	16
Exp 6	Male	Supply chain manager	18
Exp 7	Female	Farmer	23
Exp 8	Male	Farmer	25
Exp 9	Male	Agricultural consultant	12
Exp 10	Female	Agricultural consultant	14

Exp: expert.

depicts the details of these experts. The data are gathered and assessed in two stages, which are described below:

The fuzzy–Delphi method was used to select only those indicators significant to AGSCM that were determined through interviews and a literature review. Ten experts were given the same questionnaire based on the identified indicators, and they were asked to evaluate each factor in relation to the AGSCM by the linguistic scale shown in Table 3. Additionally, by applying the transforming procedures above, the values were converted into triangular fuzzy number to aggregate the fuzzy values of all 19 elements using Equations (3) and (5).

To select the more significant factors, the threshold defuzzification value (β) was chosen at 0.60 in this paper to determine whether to accept or reject a factor, which is larger than the normal value (0.56) for the nine-fuzzy scale [73]. Based on this threshold value of defuzzification, a total of 12 factors with values greater than 0.60 were selected, and 7 factors less than 0.60 were rejected.

Table 6. Finalizing factors using fuzzy Delphi.

No	Factors	Fuzzy Weight	Defuzzification	Selection	Codes
1	Green consciousness	(0.3,0.76,0.9)	0.65	√	F1
2	Competitive pressure	(0.1,0.38,0.9)	0.46	-	-
3	Government subsidies	(0.3,0.60,0.9)	0.60	√	F2
4	Produce Quality	(0.5,0.81,0.9)	0.74	√	F3
5	Customers’ demand	(0.3,0.77,0.9)	0.66	√	F4
6	Environmental laws	(0.3,0.71,0.9)	0.64	√	F5
7	Logistics	(0.1,0.30,0.9)	0.43	-	-
8	Renewable material	(0.1,0.28,0.7)	0.36	-	-
9	Green operation	(0.3,0.68,0.9)	0.63	√	F6
10	Technology	(0.3,0.62,0.9)	0.61	√	F7
11	Reducing waste	(0.1,0.43,0.9)	0.48	-	-
12	Price of product	(0.3,0.7,0.9)	0.63	√	F8
13	Cost	(0.3,0.77,0.9)	0.66	√	F9
14	Stockholders’ requirement	(0.3,0.64,0.9)	0.61	√	F10
15	Monitoring	(0.3,0.69,0.9)	0.63	√	F11
16	Social responsibilities	(0.1,0.48,0.9)	0.49	-	-
17	Infrastructure	(0.1,0.42,0.9)	0.47	-	-
18	Income level	(0.3,0.68,0.9)	0.63	√	F12
19	Reusable packaging	(0.1,0.28,0.7)	0.36	-	-

Codes are the abbreviations of Factors.

lists all the selected and rejected factors.

4.2. Grey WINGS Analysis

The impact factors of AGSCM were empirically investigated using the grey-WINGS method. The significance among the 12 factors was evaluated by experts. At this stage, the grey-WINGS approach was utilized by the same 10 experts to obtain the final interrelationships and cause–effect linkages between the factors. The following sections cover the implementation of the grey-WINGS approach:

Step 1: Using the linguistic scale provided in Table 4, experts were asked to build a strength–influence matrix for factors in the AGSCM. The grey initial strength relationship matrix from the No.1 expert is displayed in

Table 7. The grey initial strength relationship matrix from Exp 1.

	F1	F2	F3	F4	F5	F6	F7	F8	F9	F10	F11	F12
F1	(0.75,1)	(0.75,1)	(0,0.25)	(0.5,0.75)	(0.25,0.5)	(0.25,0.5)	(0.25,0.5)	(0.25,0.5)	(0.75,1)	(0.25,0.5)	(0,0.25)	(0,0.25)
F2	(0.5,0.75)	(0.75,1)	(0,0.25)	(0,0.25)	(0,0.25)	(0.25,0.5)	(0.25,0.5)	(0,0.25)	(0,0.25)	(0,0.25)	(0,0.25)	(0.25,0.5)
F3	(0.5,0.75)	(0.5,0.75)	(0,0.25)	(0.5,0.75)	(0.5,0.75)	(0.25,0.5)	(0.25,0.5)	(0,0.25)	(0.25,0.5)	(0.25,0.5)	(0,0.25)	(0.25,0.5)
F4	(0.75,1)	(0.75,1)	(0,0.25)	(0.75,1)	(0,0.25)	(0.25,0.5)	(0.25,0.5)	(0.25,0.5)	(0.25,0.5)	(0.25,0.5)	(0,0.25)	(0.25,0.5)
F5	(0.75,1)	(0.75,1)	(0.75,1)	(0.75,1)	(0,0)	(0.5,0.75)	(0.5,0.75)	(0.5,0.75)	(0.75,1)	(0.5,0.75)	(0,0.25)	(0.25,0.5)
F6	(0.75,1)	(0.75,1)	(0.75,1)	(0.75,1)	(0.5,0.75)	(0.75,1)	(0.25,0.5)	(0.25,0.5)	(0.5,0.75)	(0.25,0.5)	(0,0.25)	(0.25,0.5)
F7	(0.75,1)	(0.75,1)	(0.25,0.5)	(0.75,1)	(0.5,0.75)	(0.5,0.75)	(0.5,0.75)	(0.5,0.75)	(0.75,1)	(0.25,0.5)	(0.25,0.5)	(0.5,0.75)
F8	(0.75,1)	(0.75,1)	(0.75,1)	(0.5,0.75)	(0,0.25)	(0,0.25)	(0.5,0.75)	(0.5,0.75)	(0.75,1)	(0.25,0.5)	(0,0.25)	(0,0.25)
F9	(0.75,1)	(0.75,1)	(0.25,0.5)	(0.5,0.75)	(0.25,0.5)	(0.5,0.75)	(0.5,0.75)	(0.25,0.5)	(0.75,1)	(0.25,0.5)	(0.25,0.5)	(0.5,0.75)
F10	(0.75,1)	(0.75,1)	(0.25,0.5)	(0.75,1)	(0.25,0.5)	(0.5,0.75)	(0.5,0.75)	(0.5,0.75)	(0.5,0.75)	(0,0.25)	(0,0.25)	(0.25,0.5)
F11	(0.75,1)	(0.75,1)	(0.25,0.5)	(0.75,1)	(0.25,0.5)	(0.25,0.5)	(0.25,0.5)	(0.5,0.75)	(0.25,0.5)	(0.25,0.5)	(0,0.25)	(0.25,0.5)
F12	(0.75,1)	(0.75,1)	(0,0.25)	(0.75,1)	(0,0.25)	(0,0.25)	(0,0.25)	(0.5,0.75)	(0.25,0.5)	(0,0.25)	(0,0.25)	(0,0.25)

.

Step 2: The average grey strength–influence matrix shows as

Table 8. Average grey matrix of expert evaluations.

	F1	F2	F3	F4	F5	F6	F7	F8	F9	F10	F11	F12
F1	[0.5,0.725]	[0.475,0.725]	[0.375,0.625]	[0.35,0.575]	[0.4,0.6]	[0.45,0.7]	[0.375,0.575]	[0.4,0.65]	[0.525,0.75]	[0.55,0.8]	[0.4,0.625]	[0.375,0.6]
F2	[0.4,0.625]	[0.6,0.825]	[0.375,0.625]	[0.275,0.525]	[0.4,0.65]	[0.425,0.675]	[0.325,0.575]	[0.3,0.5]	[0.375,0.625]	[0.45,0.7]	[0.375,0.6]	[0.4,0.65]
F3	[0.5,0.725]	[0.5,0.75]	[0.525,0.775]	[0.425,0.675]	[0.4,0.625]	[0.4,0.65]	[0.325,0.55]	[0.275,0.5]	[0.45,0.7]	[0.525,0.775]	[0.35,0.55]	[0.375,0.6]
F4	[0.475,0.725]	[0.475,0.725]	[0.425,0.675]	[0.575,0.825]	[0.35,0.6]	[0.35,0.6]	[0.25,0.5]	[0.425,0.675]	[0.3,0.55]	[0.375,0.6]	[0.35,0.575]	[0.375,0.625]
F5	[0.475,0.7]	[0.5,0.75]	[0.55,0.8]	[0.5,0.725]	[0.45,0.675]	[0.525,0.775]	[0.4,0.65]	[0.375,0.625]	[0.425,0.675]	[0.4,0.625]	[0.35,0.55]	[0.375,0.625]
F6	[0.4,0.65]	[0.35,0.575]	[0.45,0.7]	[0.35,0.6]	[0.3,0.525]	[0.5,0.75]	[0.325,0.575]	[0.35,0.6]	[0.4,0.625]	[0.375,0.625]	[0.375,0.6]	[0.225,0.45]
F7	[0.5,0.725]	[0.5,0.75]	[0.425,0.65]	[0.425,0.675]	[0.425,0.675]	[0.425,0.675]	[0.375,0.625]	[0.35,0.575]	[0.45,0.7]	[0.375,0.6]	[0.3,0.55]	[0.4,0.625]
F8	[0.65,0.9]	[0.6,0.825]	[0.525,0.775]	[0.475,0.725]	[0.3,0.55]	[0.35,0.6]	[0.45,0.7]	[0.475,0.725]	[0.45,0.675]	[0.375,0.625]	[0.25,0.5]	[0.35,0.6]
F9	[0.4,0.6]	[0.5,0.75]	[0.375,0.625]	[0.475,0.725]	[0.25,0.475]	[0.45,0.7]	[0.35,0.6]	[0.275,0.5]	[0.6,0.85]	[0.375,0.625]	[0.275,0.475]	[0.35,0.6]
F10	[0.4,0.65]	[0.4,0.625]	[0.425,0.65]	[0.45,0.675]	[0.225,0.45]	[0.225,0.475]	[0.225,0.45]	[0.35,0.6]	[0.525,0.775]	[0.475,0.725]	[0.325,0.55]	[0.275,0.525]
F11	[0.525,0.775]	[0.525,0.75]	[0.375,0.625]	[0.55,0.8]	[0.275,0.525]	[0.35,0.6]	[0.35,0.6]	[0.325,0.55]	[0.275,0.475]	[0.375,0.625]	[0.375,0.575]	[0.275,0.525]
F12	[0.45,0.675]	[0.55,0.8]	[0.375,0.625]	[0.4,0.65]	[0.25,0.45]	[0.4,0.65]	[0.325,0.55]	[0.325,0.55]	[0.4,0.65]	[0.425,0.675]	[0.4,0.65]	[0.5,0.75]

. After averaging the grey initial values by Equation (6), the standardization of the grey numbers can be obtained by using Equations (7) and (8), which transform the values into the standard interval form in Table 8. Most values contain the interval [0.4,0.6]. The biggest value is [0.65,0.9], while [0.225,0.45] is the smallest interval number.

Step 3. The crisp strength–influence matrix is shown in

Table 9. The crisp strength–influence matrix.

	F1	F2	F3	F4	F5	F6	F7	F8	F9	F10	F11	F12
F1	0.1271	0.1468	0.0000	0.0720	0.1846	0.2231	0.1585	0.1477	0.2966	0.2061	0.1641	0.1633
F2	0.0014	0.2770	0.0000	0.0000	0.2061	0.1983	0.1177	0.0278	0.1270	0.0892	0.1367	0.2004
F3	0.1271	0.1759	0.1453	0.1488	0.1944	0.1736	0.1108	0.0000	0.2189	0.1769	0.1039	0.1633
F4	0.1036	0.1468	0.0484	0.2975	0.1477	0.1240	0.0305	0.1769	0.0352	0.0000	0.1094	0.1719
F5	0.0957	0.1759	0.1696	0.2137	0.2500	0.2975	0.2050	0.1184	0.1883	0.0277	0.1039	0.1719
F6	0.0055	0.0000	0.0727	0.0744	0.0833	0.2727	0.1177	0.0892	0.1489	0.0015	0.1367	0.0000
F7	0.1271	0.1759	0.0469	0.1488	0.2354	0.1983	0.1759	0.0833	0.2189	0.0000	0.0595	0.1905
F8	0.3327	0.2770	0.1453	0.1983	0.0892	0.1240	0.2632	0.2354	0.2079	0.0015	0.0016	0.1435
F9	0.0000	0.1759	0.0000	0.1983	0.0278	0.2231	0.1468	0.0000	0.4026	0.0015	0.0273	0.1435
F10	0.0055	0.0554	0.0469	0.1665	0.0000	0.0000	0.0000	0.0892	0.3108	0.1184	0.0820	0.0581
F11	0.1691	0.1939	0.0000	0.2727	0.0599	0.1240	0.1468	0.0556	0.0000	0.0015	0.1294	0.0581
F12	0.0643	0.2341	0.0000	0.1240	0.0278	0.1736	0.1108	0.0556	0.1577	0.0600	0.1753	0.3142

, which is established from the average grey strength–influence matrix. The interval numbers can integrate most information, and the further analysis needs to convert the interval form to a crisp value by using Equations (9) and (10). As a result, the total crisp values are calculated as shown in Table 9, which retain four decimals for ensuring accuracy.

Step 4. The normalized strength–influence matrix was created in

Table 10. The normalized crisp strength–influence matrix.

	F1	F2	F3	F4	F5	F6	F7	F8	F9	F10	F11	F12
F1	0.0070	0.0081	0.0000	0.0040	0.0102	0.0123	0.0088	0.0082	0.0164	0.0114	0.0091	0.0090
F2	0.0001	0.0153	0.0000	0.0000	0.0114	0.0110	0.0065	0.0015	0.0070	0.0049	0.0076	0.0111
F3	0.0070	0.0097	0.0080	0.0082	0.0107	0.0096	0.0061	0.0000	0.0121	0.0098	0.0057	0.0090
F4	0.0057	0.0081	0.0027	0.0164	0.0082	0.0069	0.0017	0.0098	0.0019	0.0000	0.0060	0.0095
F5	0.0053	0.0097	0.0094	0.0118	0.0138	0.0164	0.0113	0.0065	0.0104	0.0015	0.0057	0.0095
F6	0.0003	0.0000	0.0040	0.0041	0.0046	0.0151	0.0065	0.0049	0.0082	0.0001	0.0076	0.0000
F7	0.0070	0.0097	0.0026	0.0082	0.0130	0.0110	0.0097	0.0046	0.0121	0.0000	0.0033	0.0105
F8	0.0184	0.0153	0.0080	0.0110	0.0049	0.0069	0.0145	0.0130	0.0115	0.0001	0.0001	0.0079
F9	0.0000	0.0097	0.0000	0.0110	0.0015	0.0123	0.0081	0.0000	0.0223	0.0001	0.0015	0.0079
F10	0.0003	0.0031	0.0026	0.0092	0.0000	0.0000	0.0000	0.0049	0.0172	0.0065	0.0045	0.0032
F11	0.0093	0.0107	0.0000	0.0151	0.0033	0.0069	0.0081	0.0031	0.0000	0.0001	0.0072	0.0032
F12	0.0036	0.0129	0.0000	0.0069	0.0015	0.0096	0.0061	0.0031	0.0087	0.0033	0.0097	0.0174

, containing the positive numbers which are less than 1 since it is necessary to control different variables within the same scale through the process of standardization as listed in Equation (11).

Step 5. The total strength–influence matrix was calculated between 0.0001 and 0.0232 as shown in

Table 11. The total strength–influence matrix.

	F1	F2	F3	F4	F5	F6	F7	F8	F9	F10	F11	F12
F1	0.0075	0.0091	0.0003	0.0050	0.0109	0.0135	0.0096	0.0087	0.0177	0.0117	0.0097	0.0099
F2	0.0004	0.0161	0.0002	0.0006	0.0120	0.0119	0.0071	0.0018	0.0078	0.0051	0.0081	0.0118
F3	0.0074	0.0106	0.0083	0.0091	0.0114	0.0107	0.0068	0.0004	0.0133	0.0101	0.0063	0.0099
F4	0.0062	0.0089	0.0030	0.0172	0.0088	0.0078	0.0023	0.0103	0.0026	0.0002	0.0065	0.0103
F5	0.0059	0.0108	0.0098	0.0129	0.0147	0.0178	0.0122	0.0071	0.0116	0.0018	0.0064	0.0105
F6	0.0006	0.0005	0.0042	0.0047	0.0050	0.0158	0.0070	0.0052	0.0088	0.0002	0.0079	0.0004
F7	0.0074	0.0107	0.0029	0.0090	0.0138	0.0122	0.0105	0.0051	0.0132	0.0002	0.0039	0.0114
F8	0.0191	0.0166	0.0083	0.0119	0.0059	0.0082	0.0155	0.0137	0.0129	0.0005	0.0008	0.0090
F9	0.0002	0.0104	0.0001	0.0117	0.0020	0.0132	0.0086	0.0003	0.0232	0.0002	0.0019	0.0086
F10	0.0005	0.0036	0.0027	0.0098	0.0002	0.0005	0.0004	0.0051	0.0179	0.0066	0.0047	0.0037
F11	0.0097	0.0114	0.0002	0.0157	0.0039	0.0077	0.0086	0.0035	0.0006	0.0003	0.0077	0.0038
F12	0.0040	0.0138	0.0001	0.0076	0.0020	0.0105	0.0067	0.0035	0.0095	0.0035	0.0103	0.0182

after utilizing Equation (12). The diagonal line represents the strength of the factor itself, while the other positions represent the degree of influence of the factor influence on other factors.

Step 6. Sums of the rows T and columns L are obtained by Equations (13) and (14) in a total strength–influence matrix. Looking through the T column in

Table 12. Prominence and relation of value elements.

Factors	T	L	T + L	T − L
F1	0.1136	0.0689	0.1825	0.0447
F2	0.0829	0.1225	0.2055	−0.0396
F3	0.1043	0.0402	0.1444	0.0641
F4	0.0841	0.1153	0.1994	−0.0311
F5	0.1215	0.0906	0.2121	0.0309
F6	0.0601	0.1296	0.1897	−0.0694
F7	0.1002	0.0954	0.1955	0.0048
F8	0.1224	0.0647	0.1872	0.0577
F9	0.0805	0.1391	0.2196	−0.0586
F10	0.0558	0.0404	0.0962	0.0154
F11	0.0730	0.0743	0.1473	−0.0012
F12	0.0898	0.1073	0.1971	−0.0176

, F8 has the maximum value of 0.1224, and F10 has the minimum value of 0.0558. In L column, the max value is 0.1391 for F9, and the min value is 0.0402 for F13. In addition, (T + L) values are utilized to measure the degree of prominence, and (T − L) values are computed to identify cause and effect factors in Table 12. The max value of (T + L) is F9 with 0.2196, and the min is F10 with 0.0962. On the other hand, F3 has the max value of (T − L) as 0.0641, and F6 has the min value as −0.0694. Then, a causal graph is shown in Figure 2 by placing the (T + L) data set on the horizontal axis and the (T − L) data set on the vertical axis.

Step 7. A threshold value (β) was computed using Equation (15). An interaction matrix that depicts the interrelationships between factors is created by the values greater than β as in

Table 13. Interaction matrix of factors.

	F1	F2	F3	F4	F5	F6	F7	F8	F9	F10	F11	F12
F1		∆			∆	∆	∆	∆	∆	∆	∆	∆
F2		∆			∆	∆			∆		∆	∆
F3		∆	∆	∆	∆	∆			∆	∆		∆
F4		∆		∆	∆	∆		∆				∆
F5		∆	∆	∆	∆	∆	∆		∆			∆
F6						∆			∆		∆
F7		∆		∆	∆	∆	∆		∆			∆
F8	∆	∆	∆	∆		∆	∆	∆	∆			∆
F9		∆		∆		∆	∆		∆			∆
F10				∆					∆
F11	∆	∆		∆		∆	∆				∆
F12		∆		∆		∆			∆		∆	∆

∆ presents the interrelationship between factors.

. F6, F2, F9, F4, and F12 have more than nine interactions, respectively, whereas F1, F10, F3, and F8 have less than three interactions. Furthermore, the network diagram of interrelationships among factors can be illustrated in Figure 3.

5. Discussion

In most cases, we encounter complex MCDM problems in which the factors are mutually influenced by each other. Due to the dependencies between various factors, it is not true that any one factor can improve the entire system. Therefore, it is necessary to identify the interrelationship of the factors in the causal group that can be improved and thus influence the entire system. Considering the above situation, this study proposes a novel combination of fuzzy-Delphi and grey-WINGS techniques to illustrate the causal relationships among the factors of AGSCM. To select the relatively more important factors, a threshold of 0.6 was set in the fuzzy-Delphi method. Furthermore, utilizing the integrated grey WINGS approach, the causal relationships between the factors can be identified by aggregating the group subjective assessment from various decisionmakers. As a result, the integrated fuzzy-DELPHI grey-WINGS methods can make a significant contribution to the MCDM employed in the AGSCM.

Based on the values of (T + L) in Table 12, the factors are prioritized as F9 > F5 > F2 > F4 > F12 > F7 > F6 > F8 > F1 > F11 > F3 > F10. Moreover, the ranking of cause–effect relationships is based on (T − L) values. Qualitative and prioritized ranking of the factors in the causal group helps to identify about how much influence each factor has. Based on positive and negative signs, the factors can be categorized into two parts as causal and effect factors in Table 12. The causal factors can be sorted as F3 > F8 > F1 > F5 > F10 > F7, and the ranking of effect factors is obtained as F11 > F12 > F4 > F2 > F9 > F6. Through Table 12 and Figure 3, produce Quality (F3) was found to be the prime causal factor with a value of 0.0641. Price of product (F8) and green consciousness (F1) followed the primary factor with values 0.0577 and 0.0447. The environmental laws (F5), stockholders’ requirement (F10), and technology (F7), also can be categorized as driver factors, since the values are 0.0309, 0.0154, and 0.0048, which are greater than 0. These factors’ impacts are higher than other factors, such as monitoring (F11), income level (F12), customers’ demand (F4), government subsidies (F2), cost (F9), and green operation (F6). In order to demonstrate the advantage of this model, the result of DEMATEL was calculated to compare with WINGS, which is derived from DEMATEL. As shown in

Table 14. The values calculated by DEMATEL.

Factors	T	L	T + L	T − L
F1	0.0986	0.0858	0.1844	0.0129
F2	0.0887	0.0989	0.1876	−0.0102
F3	0.0942	0.0907	0.1849	0.0035
F4	0.0894	0.0905	0.1799	−0.0012
F5	0.1000	0.0866	0.1866	0.0133
F6	0.0835	0.0891	0.1726	−0.0056
F7	0.0950	0.0863	0.1813	0.0087
F8	0.0977	0.0906	0.1883	0.0071
F9	0.0858	0.0918	0.1777	−0.0060
F10	0.0815	0.0892	0.1708	−0.0077
F11	0.0867	0.0937	0.1803	−0.0070
F12	0.0887	0.0964	0.1850	−0.0077

, most causal and effect factors are the same except for F10, which is the same factor with min T + L value between the two methods. Furthermore, F11, F7, F1, F12, F2, and F5 have a similar sequence to T + L values, but the other factors are different in both methods. The discrepancy is caused by the assumption that the WINGS considers the strength of the indicator itself, while DEMATEL omits these ingredients, which lacks a certain degree of accuracy.

Further analysis should be performed by categorizing all the factors into various quadrants, with factors above the X-axis being prominent as causal factors, and factors below the X-axis being effectors due to their dependence on causal factors. As illustrated in Figure 2, all the factors can be classified into four distinct clusters, where quadrant 1 is the least relevant factor or the least important factor. Monitoring (F11) lies in this group. Quadrant 2 is the causal group of factors that have a driving effect on other factors, but a weaker driving effect. Stockholders’ requirement (F10) and product quality (F3) belong to this area. The shareholders generally set the goals of corporate development based on their requirements, which in turn influence various activities, including production, sales, and management operations. The next quadrant 3 is the most important and critical factor in the causal group. Green consciousness (F1), product price (F8), environmental law (F5), and technology (F7) belong to this group, thus indicating their importance to AGSCM. As discussed above, these factors have a high degree of prominence and relationship, which are priorities in AGSCM, since they can dominate other influencing factors. The fourth quadrant is for factors of high importance in the effect group, which require immediate management attention and control to improve AGSCM. Green operations (F6), cost (F9), government subsidies (F2), customer demand (F4), and income level (F12) are in this area, which integrates the activities of various parties, such as government, consumers, and companies for improving the development of AGSCM.

6. Conclusions

This study concentrates on the hierarchical evaluation structure in a complete model and proposes a novel approach using fuzzy Delphi and grey WINGS to resolve the interrelationships and incomplete information to acquire the strength and relationship between the factors of AGSCM. The practical implications and insightful conclusions of this study can be explained as follows:

With the globalization of climate change, food crisis, and the issue of the vulnerability of the agricultural supply chain, AGSCM is a complex MCDM project, which requires high priority by any organizations that are facing competition and pressure from enterprises, society, and governments. Therefore, the AGSCM needs to be improved through the optimization of influencing factors. To meet the requirements of green development, managers and policy makers strike a balance between efficiency and redundancy in the AGSCM. It is very important for the top managers to actively focus on the critical factors.

In this paper, identifying the critical factors and the corresponding causal relationships in AGSCM is the purpose. These findings suggest some preliminary guidance for the successful implementation of AGSCM. In this paper, the novel integrated method utilizes a structural modeling tool based on fuzzy Delphi and grey WINGS to evaluate the various factors of AGSCM. The fuzzy-Delphi technique is a qualitative approach for gathering opinions from various participants, which can capture the ambiguity and uncertainty in the data. By combining grey systems theory with this method, it is quite practical for integrating the preferences and views of different experts. Through the causal diagram, the factors can be divided into cause-and-effect groups. From a research perspective, this approach is valuable for assessing the relative impact and strength of the various relationships in MCDM.

The implementation of the proposed model illustrates some perspectives on the actual application and management implications of AGSCM. Some fundamental factors have been found to adjust plan and solutions. Furthermore, the cause-and-effect relationships can help to identify the factors that practitioners and researchers need to consider in AGSCM.

Product quality (F3), price of product (F8), green consciousness (F1), and environmental law (F5) are the most vulnerable causal factors of AGSCM, which need more attention. Product price (F8) and quality (F3) are the eternal concerns of consumers. Product quality (F3) is one of the main tools for marketers to position themselves in the market, which has two components: level and consistency. Agricultural product quality means the ability of an agricultural product to perform its function, including its nutrition, taste, safety, and other attributes. Price of product (F8) is the basis for establishing a diversified market mechanism, designing an efficient incentive mechanism and playing an important role in positive incentive effect, which is related to the whole process of production and marketing. Reducing the cost of green agricultural products can improve the operation of AGSCM. Environmental laws (F5) and green consciousness (F1) are the important factors for improvement of AGSCM, which refer to the activities to reduce and minimize environmental pollution of various factors. Furthermore, green consciousness (F1) improves the social image and environmental performance with new life cycle assessment, which would influence stockholders’ perceptions. Environmental laws (F5) can guide agricultural production operators to scientific planting, breeding, application of pesticides, fertilizers, and other agricultural inputs. Moreover, the agricultural nonpoint pollution and other agricultural waste can also be reduced, so that AGSCM performance could be greatly developed.

Consumer demand (F4) is the number of items which consumers are able and willing to buy with any given price. The former is influenced by the level of demand for the good, the price of the good, and the price of the substitute good, while the latter is influenced by the consumer’s willingness to buy and the actual income level. Thus, it can be stated that the price of the agricultural product determines the quantity of consumer demand. Stockholders’ requirements (F10) are directly associated with activities of green product and process in AGSCM, as well as require incorporating green innovation for modifying product green operation, cost control, and satisfying customers’ demand.

Cost(F9) is the economic value of the resources consumed to produce and sell a certain type and quantity of products measured in money. The cost of agricultural products is influenced by a variety of factors, which require focusing on. Moreover, technology (F7) is an important support to improve agricultural production capacity and competitiveness. Agricultural technology is an irreplaceable and important guarantee for the promotion of supply chain management, which is an important support to promote the development of the agricultural economy. It is necessary to strengthen government support for agricultural technology promotion, deepen the reform of the agricultural technology promotion mechanism, innovate in the agricultural technology promotion organization, and form a socialized agricultural technology service system, which is necessary to adapt to the development of AGSCM.

Government subsidies (F2) can improve the efficiency of the entire green agricultural production, thus promoting the motivation of agricultural supply chain participants to utilize green technology and supply green agricultural products. Moreover, since government subsidies can compensate some costs of green product producers, these producers can offer green products at lower prices. For the whole society, government subsidies for green agricultural products improve the willingness of consumers to pay for green consumption and increase the consumer surplus that consumers can obtain by consuming green agricultural products. Monitoring (F11) refers to the management of political, economic, and social public affairs by the relevant departments, which can supervise and manage the behavior of the subjects at all levels in the green agricultural supply chain through laws and regulations. Monitoring is not only conducive to maintaining fair development rules, but also can create a harmonious and stable social environment, thus making the green supply chain develop in a better and healthier way.

In summary, all participants of AGSCM can analyze each influencing factor and its supporting causes, or they can identify the causal links of each influencing factor through a cause–effect diagram. This can help them identify and categorize those factors and their relationships that need more attention.

This paper has some limitations. Firstly, though a sizable number of specialists took part in the investigation, there might still be some bias in the experts’ assessments, and more experts can be invited to verify the statistical results of this study. Secondly, we have considered 19 factors of AGSCM, and more factors can be added at the expense of complexity. From this study, future studies could use other MCDM approaches, such as DEMATEL and ANP, and results can be compared to check the accuracy of grey WINGS. Furthermore, this proposed method could be extended to other MCDM problems in different industries, such as healthcare, the environment, pollution, transportation, etc.