Comprehensive Sustainable Assessment and Prioritization of Different Railway Projects Based on a Hybrid MCDM Model

Abstract

Railways play an essential role in China’s development, driving economic and social progress while negatively impacting the environment. Assessing different railway projects’ sustainability can maximize the positive effects and minimize the side effects. This article aims to propose a methodological framework containing a series of indicators to evaluate the sustainability of four railway projects in China. A hybrid MCDM (multi-criteria decision-making) method was adopted to analyze the interrelationships among the dimensions and indicators and then present the projects’ differences. The weights of each indicator in the research were studied according to experts’ advice, and the same indicators were of different importance when placed in different research environments, such as developing and developed countries. This helps to establish a developing-country-centered evaluation indicator system. After comparison, it can be seen that Railway A is the best in terms of sustainability, mainly because it is in an economically developed region of China that is rich in construction resources, and is less constrained by natural conditions than other projects. The results show that although environmental sustainability has been improved over the past few years, its low level of influence still urges people to invest more attention and take more actions with significant effects in this respect. Furthermore, railways show significant differences in economic sustainability, while the differences in social sustainability have reduced considerably. This study provides further insight into the investigated measures and proposes recommendations for improving railway transport in China.

1. Introduction

Transportation plays a vital role in the development of society, and railways are generally considered to be among the most sustainable developments for ground transportation [1]. The railway industry is developing rapidly and has made remarkable achievements in China. By 2020, China’s railway operating mileage had reached 146,300 km. The national railway network density is 152.3 km/million square kilometers, ranking first in the world [2]. Railways have genuinely become the nation’s bloodlines. However, railways promote economic growth by transporting people and goods and providing access to employment, while they inevitably affect the environment and society. According to the United Nations, the transport sector in individual countries accounts for about one-quarter of all their energy-related greenhouse gas emissions. To minimize the side effects of railways, the Chinese government has proposed a series of documents urging people to make efforts to achieve sustainable transportation and presenting some specific requirements involving social development, economic development, and environmental protection [3,4]; thus, the sustainable development of the railways has become a national strategy.

The transports system’s holistic sustainability assessment and optimization are essential for making the most of its positive effects and reducing its negative ones [5,6]. Many organizations and scholars have proposed standardized approaches to measure sustainability, and establishing an appropriate indicator system is the basis for such assessment. These organizations and scholars are mainly based in developed countries with high standard-of-living requirements in recent years. However, there are still non-negligible gaps between developed and developing countries, with differences in their development strategies and goals. This raises the question of whether developing countries should slow down their economic development to pursue the more demanding ecological goals of environmental protection. Therefore, in selecting indicators, it is essential to pay attention to the applicability of these indicators in developing countries such as China.

Many methods and models have been used for the sustainable assessment of transportation, such as BP neural networks, TOPSIS, DEA, Pearson correlation, etc. [7,8]. Some methods focus on analyzing the interrelationships between indicators rather than the sustainability assessment of railway projects, while other methods only present the weights of indicators or the ranking of different projects instead of giving actual optimization suggestions. Therefore, choosing an appropriate method is significant in proposing practical recommendations for policymakers and other stakeholders.

This study aims to develop a methodological framework for assessing the sustainability of railway projects. The implementation of this framework requires three stages: indicator selection, weighting and ranking, and empirical application. Numerous indicators were found in the literature, but only 13 key indicators attributed to three dimensions—economic, environmental, and social—were finally selected. These indicators are representative and available in developing countries and could help to establish a new developing-country-centered indicator system. In the next stage, a hybrid MCDM model was implemented. The DEMATEL method was used to find the degree of influence of each indicator and dimension. The DANP was used to obtain the global and local weights of dimensions and indicators. Finally, the performance scores and the gaps in the aspiration levels of alternatives were derived via the VIKOR method. The advantage of the VIKOR method is that not only can the ranking of the options be obtained, but their potential improvements can also be presented with this method, which could provide guidance for policymakers. In the last stage, four railway projects in different regions in China were chosen to validate the reliability of the proposed framework. The study’s results illustrate the differences in the sustainable development of railway projects in different areas, which could help policymakers, construction companies, and other stakeholders to optimize the sustainable development of transport and eliminate the inequity of railway development in different regions of China.

2. Literature Review

Sustainable development (SD) is a new trend that meets the world’s development needs and provides a new development direction for transportation. The WCED defines sustainable development as development that “meets the needs of the present without compromising the ability of future generations to meet their own needs”. According to the 2030 Agenda for Sustainable Development, one of the top priorities in the next decade is to achieve sustainability in transportation. Sustainable transportation can be defined as the transportation of goods and people to ensure economic and social development while reducing undesirable impacts on the environment and society [9,10,11]. Sustainability in transportation has been widely recognized to consist of three dimensions: environmental, social, and economic sustainability [12]. Even though railways are considered to be among the most promising sustainable transport modes, their performance still needs to improve—especially in developing countries. Given such a need, the question of achieving sustainable goals for the railway industry remains very worthwhile.

To achieve sustainability goals, transportation systems should be assessed and quantified through helpful indicators [13]. Litman proposed a method for the selection of comprehensive indicators for sustainable transport planning [14]. Awasthi et al. [15,16,17] developed a series of indicators to assess the sustainability of urban transportation. The United Nations proposed 17 Sustainable Development Goals (SDGs) in 2015, and each SDG was subdivided into a few targets that could be quantified and monitored through a series of indicators, some of which are directly related to transportation. As for the railway industry, previous studies on indicator systems play a very positive role in reference and guidance. Rao used some synthetic sustainability indices for evaluating high-speed intercity railways [18]. Garg and Kashav established a framework to assess the sustainability of initiatives in the containerized freight railways of India using indicators [19], Azzouz and Jack developed a benchmarking and reporting framework for HSRs through indicators [20]. Due to the complexity of transport systems, assessing their sustainability using a comprehensive set of indicators—the number of which varies in different studies—is more scientifically valid.

On the other hand, the sustainability indicator systems differ between developed countries and developing countries. Many scholars and organizations reside in developed countries with a high standard of living. Some sustainability indicators underestimate developing countries’ sustainability performance. For example, Esty et al. pointed out that the Environmental Sustainability Index (ESI) favors developed countries due to their economic status and high income support for environmental pollution control [21]. Mori and Christodoulou also observed a similar trend in the Environmental Policy Index (EPI) [22]. Esty et al. found that the indicator “environmental health” is highly related to wealth, which may lead to an underestimation of sustainable performance in developing countries that have not invested enough in pollutant control or water clean-up [21]. In this case, good economic performance can offset bad environmental performance, and vice versa.

Developed countries likely tend to be highly evaluated because of their high income. This principle is also valid for railway sustainability assessment. Thus, it is necessary to establish a suitable sustainability indicator system for developing countries. Scholars have made some progress in China, especially in the sustainability of urbanization [23,24,25]. As for railways, Liu et al. formalized an integrated metric system measuring the performance of urban sustainability in China [26]. Qiu et al. established a sustainable evaluation indicator system for railway tunnels in China [27]. Li et al. conducted a sustainable competitiveness analysis of China Railway in the context of the Belt and Road Initiative through some indicators covering a few dimensions [28].

A variety of methods and tools have been used to analyze the indicators, see

Table 1. Summary of the methods and tools used in the literature.

Author	Year	Study Location	Method/Tool	Research Topic
Kijewska et al. [29]	2018	Szczecin, Poland	AHP and DEMATEL	Distribution of goods
Moslem, Sarbast, and Szabolcs Duleba [30]	2019	Mersin, Turkey	Fuzzy AHP	Urban transport development
Moslem et al. [31]	2019	Mersin, Turkey	Fuzzy AHP and Interval AHP	Transport development decision
Singh et al. [32]	2020	Delhi, India.	AHP	Selection of a sustainable transport system
Seker, Sukran, and Nezir Aydin [33]	2020	Turkey	IVIF-AHP and CODAS	Sustainable public transportation system
Rivero Gutiérrez et al. [34]	2021	Madrid, Spain	AHP	Sustainable urban public transport systems
Sharma et al. [35]	2018	India	AHP-MABAC	Smart railway stations
Manzolli et al. [36]	2021	Europe	PROMETHEE	Urban mobility
Morfoulaki, Maria, and Jason Papathanasiou [37]	2021	Greece	PROMETHEE	Sustainable urban mobility planning
Bojković, Nataša, Ivan Anić, and Snežana Pejčić-Tarle [38]	2010	Bulgaria, the Czech Republic, Latvia, Lithuania, Hungary, Poland, Romania, Slovakia, Croatia & Turkey	ELECTRE	Transport sustainability evaluation
Romero-Ania et al. [39]	2021	Madrid, Spain	ELECTRE	Sustainable urban public transport Systems
Djordjević et al. [8]	2021	Europe	Correlation analysis and DEA	Railway sustainability monitoring
Liu et al. [26]	2022	China	EFA and CFA	Urban sustainability
Al-Thawadi et al. [40]	2020	Qatar	LCA	Sustainable urban transportation
Sharma et al. [41]	2022	India	DMS	Railway reservation system

. MCDM provides a framework for constructing decision problems and methods for generating preferences for alternative solutions, and has attracted consistent attention from researchers in transport sustainability assessments [12,15]. AHP/ANP and their hybrids are the most frequently used methods [29,30,31,32,33,34,35]. Moreover, some other MCDM methods have been adopted in research, such as PROMETHEE [36,37], ELECTRE [38,39], DEA [8], etc. In addition to MCDM methods, other approaches have also been used in transport sustainability assessment. Liu et al. adopted EFA and CFA to measure the performance of urban sustainability [26]. Al-Thawadi et al. used life-cycle assessment (LCA) to evaluate passenger transport sustainability [40]. Sharma et al. adopted data mining scaffolding (DMS) to improve the efficiency of railway reservation [41]. These methods and tools contribute to complete data mining and allow an excellent and specific analysis of the interrelationships between the factors and the weighting ratios.

Based on the literature review above, it is apparent that there remain two significant research gaps:

(1)

The topic of measuring the sustainability of Chinese railway projects on a national scale has not been well explored, and an indicator system has not yet been established. Many frameworks for the evaluation and analysis of transport sustainability originated from Western-civilization-oriented concepts of modernization and development. There is a considerable gap in GDP per capita between developing and developed countries, so the indicator systems in different countries should pay attention to the choice of each indicator or its weight according to the actual situation. Moreover, some indicator systems include dozens or even hundreds of indicators, which are difficult to manipulate for comparison in real-life cases. The indicators should be crucial, representative, and easy to apply.

(2)

Some research has only assessed railways’ sustainability or ranked different alternatives, without highlighting ideas and paths for optimization. However, in real-life cases, policymakers and stakeholders value these optimization suggestions more than the study process. Thus, the results of the evaluation need to be explained clearly, and policy suggestions need to be proposed to help promote sustainable development.

To fill in the research gaps above, a methodological framework was established for assessing multiple railway projects in different areas of China. The framework consists of the indicators most frequently used in the literature and screened or recreated according to China’s development needs and the actual situation. DEMATEL and ANP were used to determine the interrelationships among dimensions and indicators. Furthermore, the VIKOR method was adopted to rank the railway projects and provide some improvement strategies by comparing the gaps to aspiration levels, which could help in furthering the sustainable development of railway projects in developing countries.

3. Materials and Methods

As mentioned above, previous studies generally used MCDM methods such as AHP/ANP and their hybrids, TOPSIS, PROMETHEE, ELECTRE, etc. Some studies did not suggest improvement strategies for optimization. However, in real-world cases, policymakers, industry practitioners, and related stakeholders take these suggestions very seriously. Therefore, this article uses a framework that combines the DANP method with a modified VIKOR method.

DEMATEL is a fast-developing decision-making method that has been adopted in many fields. Chang et al. used fuzzy DEMATEL to improve performance in SCM supplier selection [42]. Yazdi et al. used an integrated DEMATEL method in safety management in the high-tech industry [43]. Dalvi-Esfahani et al. analyzed social media addiction using DEMATEL [44]. Garg applied the DEMATEL technique to determine the interdependence among e-waste mitigation strategies (MSs) by cause–effect analysis [45]. Singh et al. analyzed barriers to Green Lean practices in manufacturing industries using DEMATEL [46]. In recent years, DEMATEL has also been widely used in combination with traditional ANP methods because it can compensate for some shortcomings and deficiencies of the traditional ANP method. DANP has received widespread recognition. Khan et al. analyzed critical success factors for circular economy transition through a DANP approach [47]. Gigović et al. adopted a hybrid DANP model to identify the locations of wind farms [48]. Hamedi et al. used DANP to find and prioritize human factors affecting entrepreneurial resilience [49]. Kumar and Anbanandam used grey DANP to analyze interrelationships and prioritize the factors influencing sustainable intermodal freight transport systems [50].

However, like other MCDM methods, DANP only provides the ranking of objectives, without any detailed optimization direction for improvement. A modified VIKOR method has been used to solve this problem. The introduction of the modified VIKOR method is based on the nature of decision-making. The alternatives often show inconsistent performance in different criteria when facing complex evaluations. Decision-makers have to compromise to reach a final decision. For example, Railway A might outperform Railway B in indicator j, but Railway A might have worse performance in indicator i. The modified VIKOR method aims to determine the desired solution and choose the alternative with the shortest distance to the aspiration level. The DANP–VIKOR method can give both the ranking of railway projects and the optimization direction for each project, which are crucial for policymakers. DANP–VIKOR has become a major method for decision-makers to evaluate multiple objectives. Hsu et al. selected the best vendor for conducting recycling of materials based on a hybrid MCDM model [51]. Chiu et al. used the DANP–VIKOR method to improve e-store business [52]. Rao used the DANP–VIKOR method to analyze the sustainability indicators in high-speed railways [18]. Lin et al. evaluated smart office buildings from a sustainability perspective through the modified VIKOR method [53].

The proposed methodological framework can be divided into several steps: The first step is to select the key indicators. The next step is to use a hybrid MCDM model combining the DEMATEL method with the ANP method to determine the interrelationships and the weights of the indicators and dimensions. Finally, the modified VIKOR method ranks railway projects and provides improvement strategies for each project. The modeling processes are shown in Figure 1.

3.1. Exploration of Indicators

Indicators are fundamental in building a systemic assessment framework, because they can quantify evaluation and translate abstract concepts into concrete criteria [13]. The indicators should be clearly defined, accessible, quantifiable, standardized for comparison purposes, and reflective of various aspects of the study [54,55,56]. This study collated relevant railway sustainability assessment indicators from previous literature and used the top 20% of the most cited indicators. Considering the actual research situation and the difficulty of data collection, and after careful screening, 13 key indicators were finally selected.

3.1.1. Economic Dimension

The analysis of economic impacts has always been taken as the central part of assessing the sustainability of transportation. Some indicators have become standard measures that are used by many scholars, such as average fee, total investment, passenger loading factor, satisfaction, construction fee, cost-to-benefit ratio, income, etc. [8,18,20,57,58,59]. Punctuality and reliability are widely considered among economic effects [18,60,61], while Djordjević et al. regarded them as aspects of service quality in social consequences [8].

3.1.2. Environmental Dimension

Environmental indicators are significant, since there is some controversy about the impact of transportation on the environment. Generally, environmental indicators are divided into three main aspects: pollution production, energy consumption, and protection. Greenhouse gas emissions are the main cause of global warming—the primary environmental challenge in the 21st century—so it is crucial to monitor the emissions of CO₂, NO₂, and other gases. Moreover, other pollution emitted by vehicles—such as PM_2.5, PM₁₀, CH₄, SO₂, No_x, and noise pollution—should be seriously considered. Energy consumption has mainly referred to fossil fuel consumption in the past, but with people’s growing awareness of using clear and eco-friendly energy, electricity and other new energy sources are becoming important energy alternatives and analyzing their impacts on energy consumption is a growing trend [6,62,63]. Furthermore, to assess the protection of the environment, many indicators—including biodiversity, land use, impacts on habitats, vegetation coverage, soil erosion rates, and others—should be considered.

3.1.3. Social Dimension

Social impacts are related to people more directly, covering almost every activity in people’s daily lives. Security and safety are two key indicators that are vital in transit to improve the safety of transportation users, including drivers, passengers, pedestrians, etc. Accessibility and transportation efficiency (sometimes including punctuality and reliability) are taken to make all transportation more accessible and user-friendly and to improve service quality. In addition to these indicators that are relevant to people’s daily lives, many indicators consider the cultural aspects of preserving monuments and heritage, such as cultural heritage preservation. Moreover, transportation is evaluated to estimate its impacts on promoting social equity. Thus, affordability, accessibility for disabled people, and transportation subsidy rates in remote areas are important indicators [6,14,18,57].

3.2. Key Indicators and Their Classifications

According to the 2/8 rules, we chose the top 20% of the most frequent indicators for further analysis [64]. Not all of these indicators are helpful, due to the lack of data, the different situations in the study setting, etc. Therefore, the determination of indicators should use the following criteria [8]:

• Indicators can describe railway objectives or issues.
• Data for indicators are available in databases or can be collected from railway undertakings or operators.
• Indicators can be classified into any dimension of sustainability (i.e., economic, environmental, and social).
• Indicators should be relevant to the study case and its location.
• Indicators should be understandable and readable to all potential users.

Based on the above principles, the indicators in the literature were screened and classified, and 13 key indicators falling into three dimensions were finally chosen, as shown in

Table 2. Key indicators and dimension.

Dimension	Item	Indicators	Measurement	References
Economic dimension	Ec1	Costs	Average operation cost (CNY·km−1)	[59,65]
	Ec2	Mobility	Average velocity (km·h−1)	[54,66]
	Ec3	Reliability	Average punctuality rate (%)	[61,66]
	Ec4	Productivity	Daily ridership of railway (passengers·day−1)	[18,57]
Environmental dimension	En1	GHG (greenhouse gas) emissions	The emissions of CO2/NOX/SOX and other GHGs (g·passenger-km−1)	[14,54,57,60,67]
	En2	Total energy consumption	Total energy consumption rate (kJ·passenger-km−1)	[14,57,60,66,67]
	En3	Land use	Land space used for running the transportation service (passenger-km·land-use area−1)	[14,57,60]
	En4	Renewable energy	The portion of renewable energy consumption of total energy (%)	[14,66]
	En5	Noise	Noise levels at the railway operation area (dBA)	[14,60,66]
Social dimension	So1	Accessibility	Access to residential areas, activity areas, and other transportation modes	[54,60,66]
	So2	Safety	Total number of fatalities/injuries caused to people	[14,57,66,68]
	So3	Satisfaction	Overall customer satisfaction rate (%)	[14,60,66]
	So4	Employment	Total number and rate of newly hired staff	[45,69,70]

.

3.3. DEMATEL

DEMATEL (decision-making trial and evaluation laboratory) is a decision-making method developed by the Science and Human Affairs Program of the Battelle Memorial Institute of Geneva (United Nations) from 1972 to 1976. With DEMATEL, the causality and correlation between various factors are expected to emerge, and the indicators with substantial weighting or significant correlations can be tracked when reviewing key indicators. According to Gabus and Fontela [71] and other researchers [42,43,44,45,46], the DEMATEL method involves five steps (see Appendix B.1, stage 1).

3.4. DANP (DEMATEL-Based ANP)

The sustainability of a railway project usually consists of multiple dimensions/clusters, and the weight of influence of each indicator and dimension should be calculated. In the traditional ANP method, normalization is achieved by dividing indicators in the column by clusters so that each cluster can reach the same weight. However, the influence of one cluster on another might differ, so the assumption that each cluster is of equal weight is not suitable. The combination of DEMATEL and ANP can improve the traditional ANP method, since DEMATEL can determine the causality between indicators. DANP is a scientific tool to analyze the interaction and interdependence among the dimensions and indicators in real-world problems. DANP consists of two steps [47,48,49,50] (see Appendix B.2, stage 2).

3.5. Modified VIKOR

The VIKOR (Vlse Kriterijumska Optimizacija Kompromisno Resenje) method was developed to solve multi-criteria optimization problems of complex systems. The traditional VIKOR method mainly focuses on selecting and sorting alternatives. It can present the positive solution (represented by 1) and negative solution (represented by 0) while measuring how closely the alternatives can achieve the ideal solution. Thus, it might be hard to propose reasonable improvement strategies, due to the specific features of the traditional VIKOR method. Therefore, in 2007, Opricovic and Tzeng proposed an extended VIKOR method to improve the traditional VIKOR method based on the “aspiration level” research [72]. This optimized method allows decision-makers to assess the performance of different systems on a scale of 0–10, and has been used in a wide variety of applications. The modified VIKOR method is conducted in three steps [18,51,52,53,72] (see Appendix B.3, stage 3).

4. Case Study

4.1. Background and Problem Description

The construction of railways has developed rapidly in China, and China has become the pioneer in the railway industry—especially in terms of high-speed railways. Four railway projects in different regions of China were chosen to conduct a case study. Railway A is a high-speed railway that passes through the north and east of China; the area through which it passes is one of the areas with active socioeconomic development in China. Railway B is one of the four major projects of the new century in China, and has the highest altitude and longest line of any plateau railway in the world, connecting the central part of China with the southwest region. Railway C is the world’s longest completed high-speed railway that is open to traffic at this time, and is located in Northwest China; it is the critical project of China’s medium- and long-term railway network planning and an essential part of the railway corridor of the Asia–Europe Continental Bridge. Railway D is a high-speed railway located in Northeast China, and is the first in China and the world to put a new high-speed railway in an alpine region. The locations and routes of the railways are presented in

Table 3. Locations and routes of the railway projects.

	Railway A	Railway B	Railway C	Railway D
Location	East China	Central and Southwest China	Northwest China	Northeast China
Route	Beijing–Langfang–Tianjin–Cangzhou–Dezhou–Jinan–Tai’an–Qufu–Tengzhou–Zaozhuang–Xuzhou–Suzhou–Bengbu–Chuzhou–Nanjing–Zhenjiang–Changzhou–Wuxi–Suzhou–Shanghai	Xining–Golmud–Nagqu–Lhasa	Xining–Zhangye–Jiuquan–Jiayuguan–Hami–Turpan–Urumqi	Beijing–Chengde–Chaoyang–Fuxin–Jinzhou–Shenyang

.

4.2. Data Collection

Various resources were used to collect the data related to these four railway projects, such as feasibility reports of railways, environmental evaluation reports, news reports, the Railway Statistical Bulletin released by the National Railway Administration, etc. Moreover, a questionnaire was designed and allocated to experts to determine their opinions about the interrelationships among indicators and dimensions as well as the performance scores in each indicator. The experts who participated in the questionnaire were a mix of academics and people with extensive working experience in the railway industry.

4.3. Measuring Relationships among Dimensions and Indicators Using DEMATEL

In this study, the DEMATEL method was adopted first, and the experts (information listed in Appendix A) were asked to determine the degree of influence of each indicator on the other indicators. The average initial influence matrix $A$ could be obtained pairwise by comparing indicators, see

Table 4. The initial average influence matrix $A$.

Indicators	Ec1	Ec2	Ec3	Ec4	En1	En2	En3	En4	En5	So1	So2	So3	So4
Ec1	0.000	3.316	2.474	2.750	3.000	3.000	2.895	2.789	2.421	2.263	2.833	2.944	3.111
Ec2	2.526	0.000	3.000	2.650	2.526	2.789	1.947	2.053	2.000	2.737	3.111	2.889	1.333
Ec3	2.526	3.053	0.000	2.150	1.632	1.474	1.368	1.526	1.000	2.368	2.111	3.833	1.222
Ec4	2.632	2.368	2.842	0.000	2.000	2.158	1.789	1.947	2.211	2.684	2.389	2.833	2.444
En1	2.842	2.211	1.737	1.400	0.000	3.368	1.421	3.158	1.053	1.105	1.222	2.111	1.500
En2	3.158	2.263	1.632	1.700	2.842	0.000	1.105	3.105	1.263	1.053	1.611	2.444	1.000
En3	3.263	1.842	1.474	1.450	1.737	1.526	0.000	1.211	1.421	2.211	1.556	2.056	1.333
En4	3.053	1.737	1.526	1.050	3.263	2.947	1.211	0.000	1.368	1.632	1.722	2.167	1.500
En5	2.316	2.105	1.368	1.750	1.158	1.263	1.263	1.421	0.000	1.421	1.556	3.167	1.222
So1	2.737	2.737	2.053	2.500	1.158	1.579	2.053	1.421	1.211	0.000	2.611	3.222	1.722
So2	3.263	2.842	1.842	2.200	1.211	1.368	1.421	1.421	1.211	2.632	0.000	3.833	2.611
So3	2.474	2.895	3.158	2.700	1.842	1.684	1.579	1.895	2.474	2.789	3.167	0.000	2.556
So4	3.000	1.474	1.579	2.000	1.158	1.158	1.000	1.421	1.421	1.579	2.333	2.889	0.000

Note: $\frac{1}{n\left(n-1\right)}{{\displaystyle \sum }}_{i=1}^n{{\displaystyle \sum }}_{j=1}^n\frac{\left|a_{ij}^n-a_{ij}^{n-1}\right|}{a_{ij}^n}\times 100\%=2.6\%<5\%$, where $a_{ij}^n$ and $a_{ij}^{n-1}$ denote the average influence of criterion i on criterion j according to experts p and p − 1, respectively; n denotes the number of criteria. Thus, the results above are significant at the 97.4% confidence level, which is greater than the 95% level that is used to test for significance.

. The significance reached the 97.4% confidence level—greater than 95%—so the experts’ opinions were consistent and credible. Then, the initial average matrix was normalized from Equations (A1)–(A3) to obtain the normalized direct influence matrix Z, see

Table 5. The normalized direct influence matrix Z.

Indicators	Ec1	Ec2	Ec3	Ec4	En1	En2	En3	En4	En5	So1	So2	So3	So4
Ec1	0.000	0.098	0.073	0.081	0.089	0.089	0.086	0.083	0.072	0.067	0.084	0.086	0.092
Ec2	0.075	0.000	0.101	0.090	0.085	0.094	0.066	0.069	0.068	0.093	0.105	0.084	0.045
Ec3	0.075	0.106	0.000	0.088	0.067	0.061	0.056	0.063	0.041	0.097	0.081	0.111	0.050
Ec4	0.078	0.082	0.115	0.000	0.071	0.076	0.063	0.069	0.078	0.095	0.084	0.082	0.086
En1	0.084	0.077	0.070	0.058	0.000	0.139	0.061	0.135	0.046	0.045	0.047	0.061	0.065
En2	0.093	0.078	0.066	0.070	0.121	0.000	0.048	0.133	0.055	0.043	0.061	0.071	0.043
En3	0.097	0.064	0.060	0.060	0.074	0.063	0.000	0.052	0.067	0.090	0.059	0.060	0.062
En4	0.090	0.060	0.062	0.043	0.139	0.121	0.052	0.000	0.059	0.067	0.066	0.063	0.065
En5	0.069	0.073	0.055	0.072	0.049	0.052	0.063	0.061	0.000	0.058	0.059	0.092	0.057
So1	0.081	0.095	0.082	0.100	0.046	0.063	0.082	0.057	0.048	0.000	0.100	0.094	0.069
So2	0.097	0.099	0.075	0.085	0.047	0.053	0.055	0.055	0.047	0.102	0.000	0.111	0.101
So3	0.073	0.099	0.108	0.092	0.063	0.058	0.054	0.065	0.085	0.095	0.108	0.000	0.087
So4	0.089	0.051	0.064	0.082	0.049	0.048	0.048	0.061	0.068	0.065	0.089	0.084	0.000

.

Next, the total influence matrices T_C and T_D were derived using Equation (A4), T_C is shown in

Table 6. The total influence matrix T_C for the indicators.

Indicators	Ec1	Ec2	Ec3	Ec4	En1	En2	En3	En4	En5	So1	So2	So3	So4
Ec1	0.808	0.893	0.835	0.830	0.820	0.833	0.686	0.816	0.673	0.813	0.852	0.890	0.763
Ec2	0.865	0.794	0.849	0.827	0.806	0.826	0.661	0.794	0.659	0.825	0.859	0.879	0.713
Ec3	0.811	0.837	0.708	0.778	0.740	0.747	0.613	0.737	0.597	0.780	0.789	0.849	0.673
Ec4	0.867	0.868	0.860	0.745	0.791	0.808	0.658	0.791	0.668	0.826	0.842	0.878	0.748
En1	0.803	0.790	0.752	0.729	0.670	0.803	0.602	0.790	0.586	0.713	0.737	0.783	0.667
En2	0.810	0.793	0.750	0.740	0.778	0.681	0.591	0.788	0.594	0.712	0.751	0.793	0.650
En3	0.752	0.721	0.686	0.677	0.674	0.676	0.501	0.657	0.560	0.698	0.692	0.724	0.616
En4	0.806	0.776	0.743	0.716	0.789	0.787	0.593	0.669	0.596	0.729	0.752	0.784	0.667
En5	0.693	0.695	0.652	0.657	0.622	0.634	0.533	0.632	0.473	0.641	0.661	0.718	0.584
So1	0.828	0.838	0.794	0.798	0.732	0.758	0.643	0.741	0.612	0.703	0.816	0.845	0.699
So2	0.849	0.849	0.795	0.793	0.738	0.756	0.626	0.746	0.617	0.802	0.734	0.868	0.734
So3	0.871	0.891	0.863	0.838	0.790	0.799	0.656	0.793	0.680	0.836	0.871	0.811	0.756
So4	0.740	0.706	0.687	0.693	0.647	0.656	0.542	0.658	0.558	0.674	0.715	0.741	0.556

. After that, the r and c in the total influence matrix were calculated using Equations (A7) and (A8), illustrating the cause and effect values of the indicators and dimensions, and the INRM (Figure 2, Figure 3, Figure 4 and Figure 5) was drawn to visualize the complex relationships.

Table 7. The cause and effect values for T_C.

Indicators	r	c	r + c	r − c
Ec1	10.512	10.503	21.015	0.010
Ec2	10.357	10.450	20.807	−0.093
Ec3	9.660	9.974	19.634	−0.314
Ec4	10.349	9.821	20.170	0.528
En1	9.425	9.598	19.023	−0.173
En2	9.431	9.765	19.196	−0.335
En3	8.634	7.905	16.540	0.729
En4	9.408	9.610	19.018	−0.202
En5	8.194	7.875	16.069	0.320
So1	9.808	9.752	19.560	0.056
So2	9.908	10.072	19.979	−0.164
So3	10.455	10.562	21.017	−0.108
So4	8.572	8.826	17.397	−0.254

demonstrates how the cause and effect relationships impact the indicators. Taking the economic dimension as an example, the indicators can be prioritized as E_c1 > E_c2 > E_c4 > E_c3 based on the (r + c) values. The most important indicator in the economic dimension is cost, with a value of 21.015, while reliability in the economic dimension influences the other indicators the least, with a value of 19.634. Additionally, the values of E_c2 and E_c3 are negative. These two indicators are most likely affected by other indicators and belong to the affected group.

Likewise,

Table 8. The total influence matrix T_D for the dimensions and their cause and effect values.

Dimensions	Economic	Environmental	Social	r	r + c	r − c
Economic	0.823	0.736	0.811	2.371	4.733	0.008
Environmental	0.737	0.651	0.704	2.092	4.172	0.011
Social	0.802	0.693	0.760	2.255	4.530	−0.020
c	2.362	2.081	2.275	-	-	-

shows the cause and effect relationships between dimensions. The three dimensions can be prioritized in the order economic > social > environmental based on their (r + c) values. The economic dimension has the most substantial influence in term of the strength of its relationships, with a value of 4.733, while the environmental dimension influences other dimensions to the lowest degree, with a value of 4.172. When considering the value of (r − c), the social dimension is more likely to be affected by other dimensions, which means that it belongs to the affected group.

4.4. Weighting Each Indicator and Dimension Using the DANP Technique

Our research combined the DEMATEL and ANP methods to obtain the adjusted influence weights. By using Equations (A9)–(A11), the total influence matrix T_C was normalized to obtain the ${\mathit{T}}_{\mathit{C}}^{\mathit{n}\mathit{o}\mathit{r}}$ and the unweighted supermatrix ${\mathit{W}}_{\mathit{c}}$. Then, the normalized total matrix ${\mathit{T}}_{\mathit{D}}^{\mathit{N}\mathit{o}\mathit{r}}$ and the weighted supermatrix ${\mathit{W}}_{\mathit{D}}^{\mathit{\alpha }}$ could be obtained by Equations (A12) and (A13), respectively. Finally, the limit supermatrix $\mathit{W}=\underset{\mathit{\phi }\to \infty }{\mathbf{lim}}{\left({\mathit{W}}_{\mathit{D}}^{\mathit{\alpha }}\right)}^{\mathit{\phi }}$ was converted into a long-term and stable supermatrix. Through this matrix, the global and local weights of all indicators and their ranks could be confirmed.

The weights and priority of the dimensions and indicators of the railways were obtained by using the DANP method, and the detailed calculation process is shown in

Table 9. The normalized total influence matrix ${\mathit{T}}_{\mathit{C}}^{\mathit{n}\mathit{o}\mathit{r}}$.

Indicators	Ec1	Ec2	Ec3	Ec4	En1	En2	En3	En4	En5	So1	So2	So3	So4
Ec1	0.240	0.265	0.248	0.247	0.214	0.218	0.179	0.213	0.176	0.245	0.257	0.268	0.230
Ec2	0.259	0.238	0.255	0.248	0.215	0.221	0.176	0.212	0.176	0.252	0.262	0.268	0.218
Ec3	0.259	0.267	0.226	0.248	0.216	0.218	0.178	0.215	0.174	0.252	0.255	0.274	0.218
Ec4	0.260	0.260	0.257	0.223	0.213	0.218	0.177	0.213	0.180	0.251	0.256	0.266	0.227
En1	0.261	0.257	0.245	0.237	0.194	0.233	0.174	0.229	0.170	0.246	0.254	0.270	0.230
En2	0.262	0.256	0.242	0.239	0.227	0.198	0.172	0.230	0.173	0.245	0.258	0.273	0.224
En3	0.265	0.254	0.242	0.239	0.220	0.220	0.163	0.214	0.183	0.256	0.254	0.265	0.226
En4	0.265	0.255	0.244	0.235	0.230	0.229	0.173	0.195	0.174	0.249	0.256	0.267	0.227
En5	0.257	0.258	0.242	0.243	0.215	0.219	0.184	0.218	0.163	0.246	0.254	0.276	0.224
So1	0.254	0.257	0.244	0.245	0.210	0.217	0.185	0.212	0.176	0.230	0.266	0.276	0.228
So2	0.258	0.258	0.242	0.241	0.212	0.217	0.180	0.214	0.177	0.256	0.234	0.277	0.234
So3	0.251	0.257	0.249	0.242	0.212	0.215	0.177	0.213	0.183	0.255	0.266	0.248	0.231
So4	0.262	0.250	0.243	0.245	0.211	0.214	0.177	0.215	0.182	0.251	0.266	0.276	0.207

,

Table 10. The unweighted supermatrix ${\mathit{W}}_{\mathit{c}}$.

Indicators	Ec1	Ec2	Ec3	Ec4	En1	En2	En3	En4	En5	So1	So2	So3	So4
Ec1	0.240	0.259	0.259	0.260	0.261	0.262	0.265	0.265	0.257	0.254	0.258	0.251	0.262
Ec2	0.265	0.238	0.267	0.260	0.257	0.256	0.254	0.255	0.258	0.257	0.258	0.257	0.250
Ec3	0.248	0.255	0.226	0.257	0.245	0.242	0.242	0.244	0.242	0.244	0.242	0.249	0.243
Ec4	0.247	0.248	0.248	0.223	0.237	0.239	0.239	0.235	0.243	0.245	0.241	0.242	0.245
En1	0.214	0.215	0.216	0.213	0.194	0.227	0.220	0.230	0.215	0.210	0.212	0.212	0.211
En2	0.218	0.221	0.218	0.218	0.233	0.198	0.220	0.229	0.219	0.217	0.217	0.215	0.214
En3	0.179	0.176	0.178	0.177	0.174	0.172	0.163	0.173	0.184	0.185	0.180	0.177	0.177
En4	0.213	0.212	0.215	0.213	0.229	0.230	0.214	0.195	0.218	0.212	0.214	0.213	0.215
En5	0.176	0.176	0.174	0.180	0.170	0.173	0.183	0.174	0.163	0.176	0.177	0.183	0.182
So1	0.245	0.252	0.252	0.251	0.246	0.245	0.256	0.249	0.246	0.230	0.256	0.255	0.251
So2	0.257	0.262	0.255	0.256	0.254	0.258	0.254	0.256	0.254	0.266	0.234	0.266	0.266
So3	0.268	0.268	0.274	0.266	0.270	0.273	0.265	0.267	0.276	0.276	0.277	0.248	0.276
So4	0.230	0.218	0.218	0.227	0.230	0.224	0.226	0.227	0.224	0.228	0.234	0.231	0.207

,

Table 11. The normalized total influence matrix ${\mathit{T}}_{\mathit{D}}^{\mathit{N}\mathit{o}\mathit{r}}$.

Dimensions	Economic	Environmental	Social
Economic	0.347	0.311	0.342
Environmental	0.352	0.311	0.336
Social	0.356	0.307	0.337

,

Table 12. The weighted supermatrix ${\mathit{W}}_{\mathit{c}}$.

Indicators	Ec1	Ec2	Ec3	Ec4	En1	En2	En3	En4	En5	So1	So2	So3	So4
Ec1	0.083	0.090	0.090	0.090	0.092	0.092	0.093	0.093	0.091	0.090	0.092	0.089	0.093
Ec2	0.092	0.083	0.093	0.090	0.091	0.090	0.089	0.090	0.091	0.091	0.092	0.091	0.089
Ec3	0.086	0.089	0.078	0.089	0.086	0.085	0.085	0.086	0.085	0.087	0.086	0.089	0.086
Ec4	0.086	0.086	0.086	0.077	0.083	0.084	0.084	0.083	0.086	0.087	0.086	0.086	0.087
En1	0.066	0.067	0.067	0.066	0.060	0.071	0.068	0.072	0.067	0.065	0.065	0.065	0.065
En2	0.068	0.069	0.068	0.068	0.073	0.062	0.068	0.071	0.068	0.067	0.067	0.066	0.066
En3	0.056	0.055	0.055	0.055	0.054	0.054	0.051	0.054	0.057	0.057	0.055	0.054	0.054
En4	0.066	0.066	0.067	0.066	0.071	0.072	0.067	0.061	0.068	0.065	0.066	0.065	0.066
En5	0.055	0.055	0.054	0.056	0.053	0.054	0.057	0.054	0.051	0.054	0.054	0.056	0.056
So1	0.084	0.086	0.086	0.086	0.083	0.082	0.086	0.084	0.083	0.078	0.086	0.086	0.085
So2	0.088	0.090	0.087	0.088	0.085	0.087	0.085	0.086	0.085	0.090	0.079	0.090	0.090
So3	0.092	0.092	0.094	0.091	0.091	0.092	0.089	0.090	0.093	0.093	0.093	0.084	0.093
So4	0.079	0.075	0.075	0.078	0.077	0.075	0.076	0.076	0.075	0.077	0.079	0.078	0.070

and

Table 13. The limit supermatrix $\mathit{W}=\underset{\mathit{\phi }\to \infty }{\mathit{lim}}{\left({\mathit{W}}_{\mathit{D}}^{\mathit{\alpha }}\right)}^{\mathit{\phi }}$.

Indicators	Ec1	Ec2	Ec3	Ec4	En1	En2	En3	En4	En5	So1	So2	So3	So4
Ec1	0.091	0.091	0.091	0.091	0.091	0.091	0.091	0.091	0.091	0.091	0.091	0.091	0.091
Ec2	0.090	0.090	0.090	0.090	0.090	0.090	0.090	0.090	0.090	0.090	0.090	0.090	0.090
Ec3	0.086	0.086	0.086	0.086	0.086	0.086	0.086	0.086	0.086	0.086	0.086	0.086	0.086
Ec4	0.085	0.085	0.085	0.085	0.085	0.085	0.085	0.085	0.085	0.085	0.085	0.085	0.085
En1	0.066	0.066	0.066	0.066	0.066	0.066	0.066	0.066	0.066	0.066	0.066	0.066	0.066
En2	0.068	0.068	0.068	0.068	0.068	0.068	0.068	0.068	0.068	0.068	0.068	0.068	0.068
En3	0.055	0.055	0.055	0.055	0.055	0.055	0.055	0.055	0.055	0.055	0.055	0.055	0.055
En4	0.066	0.066	0.066	0.066	0.066	0.066	0.066	0.066	0.066	0.066	0.066	0.066	0.066
En5	0.055	0.055	0.055	0.055	0.055	0.055	0.055	0.055	0.055	0.055	0.055	0.055	0.055
So1	0.084	0.084	0.084	0.084	0.084	0.084	0.084	0.084	0.084	0.084	0.084	0.084	0.084
So2	0.087	0.087	0.087	0.087	0.087	0.087	0.087	0.087	0.087	0.087	0.087	0.087	0.087
So3	0.091	0.091	0.091	0.091	0.091	0.091	0.091	0.091	0.091	0.091	0.091	0.091	0.091
So4	0.076	0.076	0.076	0.076	0.076	0.076	0.076	0.076	0.076	0.076	0.076	0.076	0.076

.

Table 14. The performance scores and gaps to the aspiration levels of different railways.

Dimension and Indicators	Local Weight	Global Weight	Performance Score				Gap to Aspiration Level
			A	B	C	D	A	B	C	D
Economic	0.347		8.443	5.315	5.645	7.956	0.156	0.469	0.435	0.204
Ec1. Costs	0.240	0.083	8.815	6.930	7.633	6.861	0.119	0.307	0.237	0.314
Ec2. Mobility	0.265	0.092	9.125	6.344	8.103	7.334	0.088	0.366	0.190	0.267
Ec3. Reliability	0.248	0.086	9.126	7.938	8.827	7.146	0.087	0.206	0.117	0.285
Ec4. Productivity	0.247	0.086	9.518	6.310	6.433	7.100	0.048	0.369	0.357	0.290
Environmental	0.311		8.643	7.512	6.215	7.423	0.136	0.149	0.248	0.258
En1. GHG emissions	0.214	0.066	8.521	7.561	6.354	7.112	0.148	0.244	0.365	0.289
En2. Total energy consumption	0.218	0.068	7.613	6.215	6.656	7.164	0.239	0.379	0.334	0.284
En3. Land use	0.179	0.056	8.714	6.935	5.441	6.243	0.129	0.307	0.456	0.376
En4. Renewable energy	0.213	0.066	7.312	5.234	4.314	5.808	0.269	0.477	0.569	0.419
En5. Noise	0.176	0.055	9.313	7.452	6.314	5.912	0.069	0.255	0.369	0.409
Social	0.342		9.213	8.314	7.245	7.313	0.079	0.169	0.276	0.269
So1. Accessibility	0.245	0.084	9.213	6.412	7.312	7.564	0.079	0.359	0.269	0.244
So2. Safety	0.257	0.088	9.412	8.907	7.322	8.009	0.059	0.109	0.268	0.199
So3. Satisfaction	0.268	0.092	9.465	7.313	7.153	5.812	0.054	0.269	0.285	0.419
So4. Employment	0.230	0.079	8.789	6.314	7.213	8.012	0.121	0.369	0.279	0.199
Total performance			8.243	6.443	6.354	7.616
Total gap							0.176	0.356	0.365	0.238
							Scores (ranking)
$\mathrm{Average} \mathrm{gap} (S_k$)							0.110	0.306	0.302	0.301
Max gap ($Q_k$)							0.269	0.477	0.569	0.419
$H_k$ ($\alpha =0.5$)							0.1896 (1)	0.3914 (3)	0.4353 (4)	0.3603 (2)

lists the global and local weights of all indicators and dimensions. It can be seen that the economic dimension ranks first in the local weighting, followed by the social and environmental dimensions. Moreover, the priority of indicators in each dimension was extended from the local weights. Taking the economic dimension as an example, mobility is the priority.

4.5. Evaluating the Performance of Railways Based on the VIKOR Method

Experts were asked to score each railway project’s performance according to related data on these projects. They gave the scores on a scale of 0–10, with 0 indicating very bad and 10 indicating the best. Then, the average scores were used and applied with the VIKOR model to determine the gaps from the intended level, total gap, average gap, and maximal gaps of different railway projects, as shown in Table 14.

4.6. Sensitivity Analysis

Sensitivity analysis was conducted to verify the stability and validity of the proposed framework in different scenarios. According to previous research [33,73], it is necessary to change the value of $\alpha$ in the range of 0–1 to examine the effect of the comprehensive index on the result and the optimal answer. The results of the sensitivity analysis are presented in

Table 15. The results of the sensitivity analysis.

		Railway A	Railway B	Railway C	Railway D
	Average gap ($S_k$)	0.110	0.306	0.302	0.301
	Max gap ($Q_k$)	0.269	0.477	0.569	0.419
Scenario 1	$\alpha =0.1$	0.2530 (1)	0.4596 (3)	0.5419 (4)	0.4074 (2)
Scenario 2	$\alpha =0.2$	0.2371 (1)	0.4425 (3)	0.5153 (4)	0.3957 (2)
Scenario 3	$\alpha =0.3$	0.2213 (1)	0.4255 (3)	0.4886 (4)	0.3839 (2)
Scenario 4	$\alpha =0.4$	0.2055 (1)	0.4085 (3)	0.4620 (4)	0.3721 (2)
Scenario 5	$\alpha =0.6$	0.1738 (1)	0.3744 (3)	0.4087 (4)	0.3486 (2)
Scenario 7	$\alpha =0.7$	0.1580 (1)	0.3574 (3)	0.3820 (4)	0.3368 (2)
Scenario 8	$\alpha =0.8$	0.1421 (1)	0.3403 (3)	0.3554 (4)	0.3250 (2)
Scenario 9	$\alpha =0.9$	0.1263 (1)	0.3233 (3)	0.3287 (4)	0.3132 (2)

. The results are consistent in all scenarios, demonstrating that the proposed approach is robust and efficient in dealing with MCDM problems.

4.7. Comparative Analysis

It is very important to be sure of whether or not the results are valid, and the calculations should be performed using other methods. We selected TOPSIS and COPRAS, consistent with some previous articles [74,75,76]. TOPSIS and VIKOR are based on an aggregating functions representing ‘‘closeness to the ideal’’, which originate in the compromise programming method. COPRAS is one of the most frequently used MCDM methods. The results in

Table 16. The results of the TOPSIS method.

	Railway A	Railway B	Railway C	Railway D
${\mathit{C}}_{\mathit{i}}$	0.50483628	0.4362	0.4360	0.4373
Rank	1	3	4	2

and

Table 17. The results of the COPRAS method.

	Railway A	Railway B	Railway C	Railway D
${\mathit{Q}}_{\mathit{i}}$	0.2994	0.2332	0.2330	0.2349
${\mathit{U}}_{\mathit{i}}$	1	0.7789	0.7782	0.7846
Rank	1	3	4	2

show that the ranking of TOPSIS and COPRAS is consistent with that of VIKOR, which could prove the robustness of the results in this research.

5. Results and Discussion

As the existing indicators formed a relatively standardized system, a corresponding standard procedure was developed for data disclosure. Still, there were some discoveries. According to the results of DANP, the weights of the three dimensions are as follows: economic (0.347), social (0.342), and environmental (0.311)—which means that experts still regard the economic dimension as the most important part of the evaluation, and that the economic dimension has more influence over the other two dimensions, which is consistent with some previous findings [18,77]. However, in this study, the experts considered the social dimension to be an affected dimension, meaning that the economic and environmental factors would influence it. Meanwhile, in some other research, the environmental dimension was the affected one [18]. The indicators with high weights were mobility (0.092), satisfaction (0.092), safety (0.088), reliability (0.086), and productivity (0.086). This indicates that the average velocity of railways is very important in different regions. This differs from the findings of a previous study [18], and may be closely related to the huge size of China’s territory, as people travel a long distance from the departure to their destination, and the speed determines the efficiency of travel.

Moreover, transportation is developed with the goal of serving people, and satisfaction is an intuitive feedback indicator, so it occupies a rather important position in the assessment. Safety, reliability, and productivity had similar importance compared with other evaluation research [8,14,18]. The indicators in the environmental dimension had a low degree of influence. This indicates two facts: First, people are now considering the economic and social dimensions more because they are more directly related to people’s daily lives. Although the railway industry has begun to emphasize the importance of environmental sustainability, as China is still a developing country, the focus of railway development remains on the economic dimension. Second, the environmental dimension is influenced more by the features of railway operation and the energy structure at present, so people can only urge the adoption of new energy and reduction in GHG emissions, meaning that the effectiveness of the adoption of environmental measures may not be achieved for a short period.

According to the results of VIKOR, the total performance scores of the four railway projects (A–D) were 8.243, 6.433, 6.354, and 7.616. Railway A was the best, followed by railways D, B, and C. This may relate to the spatial distribution of each project. The performance of the railway projects was similar to the findings of some previous studies, but new discoveries were also made. Railway A passes through China’s eastern and central regions, while railway B is in the west. These two railway projects showed a significant difference in economic sustainability but a slight difference in social sustainability. Additionally, according to Stefaniec et al., in 2015, the inland transport in the eastern and western zones were very polarized in terms of social sustainability, while the central zone contained provinces with a medium level of social sustainability [78]. The border regions without access to the sea present very low economic sustainability, and the innermost areas are generally rather low. The results of this research confirm the part of economic sustainability. However, the differences between the east and west regions in terms of social sustainability have been widely reduced in the past seven years. Moreover, Railway C runs through the northwest area of China, while Railway D is in the northeast. Stefaniec et al. pointed out that inland transport in the west and northeast parts of China is characterized by low performance in environmental sustainability. Still, Railways C and D demonstrate that environmental sustainability is promoted in both regions [78]. Despite the ranking of the four railway projects, their gaps from their desired levels could provide detailed improvement suggestions for each railway. For example, although Railway A was the best in the assessment, the average gap to its aspiration level was 0.110. To minimize this gap, some improvement suggestions can be adopted, such as promoting the use of new energy sources (E_n4), reducing total energy consumption (E_n2), reducing GHG emissions (E_n1), etc. The maximum gaps all occur in renewable energy, which could be the most crucial indicator for improving railways’ sustainability.

Based on the above results, some managerial and policy implication suggestions can be proposed. In selecting indicators, many indicators characteristic of developing countries could not be used for the sustainability evaluation of railway projects due to missing data. These missing data may cause many valuable indicators to lose their research significance, especially when establishing a developing-country-centered indicator system. Government departments should urge relevant parties to disclose valid and real data promptly to provide feedback to policymakers and facilitate adjustments to improve applicable development policies.

The interplay between the three dimensions was explored in this study, and some new ideas were provided, such as increasing investment in ecological and environmental protection, improving railway capacity, promoting the use of new energy, etc. The low level of influence of the environmental dimension deserves to be considered, as it indicates that the consideration of environmental protection is still lacking in current railway construction. There is still a need to constantly update railway development policies, construction concepts, purposes, and measures. Second, the DANP weights implied the degree of influence of the indicators. Optimizing indicators with high impact weights may have a more critical effect on the sustainability of overall railway projects.

Moreover, the sustainability assessment of the four projects reflected the sustainable development of transportation in different regions of China. With the introduction of a series of national policies focusing on supporting and developing the western region over the past few years, transport sustainability in the western region has made some progress. However, there is still plenty of room for improvement in the western region—especially in terms of economic sustainability—compared to the central and eastern areas. The railway development policy of each site should be tailored to the regional characteristics, considering the actual situation. The above recommendations could help to eliminate the disparities in regional railway development and, finally, achieve the Sustainable Development Goals.

6. Conclusions

This study adopted a hybrid MCDM model to evaluate the sustainability of railway projects in different regions of China. The indicators used in the study were those that appear very frequently in the existing literature and apply to developing countries. The weights of each indicator were initially investigated and subsequently calculated based on the questionnaire results in the follow-up expert interviews. It is worth noting that some common indicators have already formed a standardized system in developed-country-oriented evaluations, which may be reflected in their different weights in the assessment of developing countries. The findings of this study could provide a practical and comprehensive guideline for policymakers in evaluating and improving transportation systems. This research indicates some regions’ current situation and characteristics of sustainable transportation development, and could help to eliminate regional inequalities in railway development, achieve sustainability goals, and promote synergistic development across regions in developing countries.

There are also some limitations to this study. One is that only a small number of experts were surveyed, rather than an industry-wide survey covering all levels. Although the consistency of the survey passed the test, there are inevitably some subjective differences in awareness. Future research should expand the scope of the survey and solicit more people’s opinions so that the whole assessment can be more objective, professional, and scientific. Secondly, this study only compares railway projects’ sustainability in different regions and does not consider differences in the construction time, construction techniques, or construction purpose. Furthermore, the access to data is limited because much information was unavailable due to confidentiality issues. In future studies, adding indicators that consider construction time, construction technology, and construction purpose for further sustainability assessment would be significant.