Research on Inter-Provincial Transfer of CO2 Emissions from Transportation by Considering Fuzzy Parameter

CO2 reduction from transportation is exerting significant effects on global CO2 reduction. This industry contributes 23.96% of global CO2 emissions. In this research, an ecological network input–output interval fuzzy linear programming (EIFP) method is developed to clarify CO2 reduction responsibilities and depict transfer relationships of transportation. This method integrates input–output analysis (IOA), inexact rough interval fuzzy linear programming (IRFLP) and ecological network analysis (ENA) into a general framework. The proposed method is employed for calculating inter-provincial CO2 transfer under different situations in 30 provinces of China and further supporting the formulation of regional reduction policies. Results demonstrate that transportation energy demand of Beijing is dependent on imports, which indirectly increases CO2 reduction pressure in energy supply areas. Therefore, CO2 reduction responsibility should be traced to source and included in emission reduction plan of energy demand areas. In inter-provincial CO2 transfer relationships of natural gas, positive relationships account for a higher proportion; therefore, it is feasible to consider raising the proportion of natural gas in the future development direction of transportation. The achievements of this paper can provide scientific references for decision makers to formulate CO2 reduction policies in transportation.


Introduction
Clarifying CO 2 reduction responsibility of transportation is of considerable significance to global climate change and completing global CO 2 reduction objectives. CO 2 emissions of transportation take a large proportion in numerous regions of the world. The largest sources of transportation-related CO 2 emissions include passenger cars, mediumand heavy-duty trucks, and light-duty trucks, including sport utility vehicles, pickup trucks, and minivans. These sources account for over half of the emissions from transportation. The remaining emissions from transportation come from other modes of transportation, including commercial aircraft, ships, boats, and trains, as well as pipelines and lubricants. It was estimated that by 2018, U.S. transportation would attain 28% of CO 2 emissions, surpassing the power sector (27%) and developing into the largest CO 2 emissions in the United States [1]. Similarly, in the fuel type composition in FY 2017, gasoline contained in fuel oil took up the largest share (54%), followed by light oil (26%) and electricity consumed by railroad (11%). Furthermore, as of 2018, CO 2 emissions from transportation rose to 17% of total emissions in Tokyo, Japan [2]. As the most populous region in Greece, Attica emits 40% of the domestic CO 2 emissions of transportation [3]. In China, the largest carbon emitter around the world, CO 2 emissions of transportation obtained second place among six major sectors, and its energy consumption growth rate seeded first among six major sectors [4]. The further rise in total CO 2 emissions would be effectively controlled by CO 2 reduction from transportation, and macro-control policy is one of the basic measures on CO 2 reduction [5,6]. Emerging countries in Southeast Asia have also issued corresponding policies to cope with the increase in the number of motor vehicles and their negative impacts on air quality, transportation, energy security, livability, and greenhouse gas emissions [7]. In Dublin, Republic of Ireland, research shows that people prefer self-driving vehicles with clean fuel, and urban transport policies can leverage people's attitudes to promote sustainable urban transport [8]. A study in a specific area of Rzeszow, Poland, shows that converting 25% of existing vehicles into electric vehicles can reduce PM 10 emissions by about 30% [9]. In addition, with the increase in the proportion of electric vehicles, it is considerable to optimize the location planning and design of electric vehicle charging stations (EVCS S ). Reasonable EVCS S can greatly increase the number of electric vehicles in circulation [10]. Although the analytic hierarchy process and the spatial geographic analysis method can optimize the site selection of EVCS S , it still needs the support of relevant policies in the implementation process. Therefore, the development of relevant policies is necessary for transportation CO 2 emissions. However, traditional macro-control policies lack consideration of the responsibility for CO 2 reduction transferred during transportation [11]. Thus, ascertaining reduction responsibilities of different regions is crucial for accomplishing CO 2 reduction objective of transportation.
In the past decades, many scholars have done a number of studies to identify responsibility for CO 2 reduction. Nakano et al. adopted a bilateral trade method to define CO 2 share and solve the carbon trade problem of 17 sectors in 41 countries/regions [12]. Xu et al. considered the impact of the environmental indirect cost of labor input from the perspective of EOLI and calculated energy demand and CO 2 emission responsibility in Sino-US trade [13]. For transportation, Wang et al. calculated emission responsibility of transportation in China from 2000 to 2015 based on measured data [14], and Zhang et al. evaluated CO 2 emission responsibility of private and official cars according to the carbon footprint method [15]. It seems from these works that researchers have selected different methods to reveal the process of carbon transfers among regions or sectors and clarify responsibility for CO 2 emissions. However, compared with these methods, input-output analysis method is more accurate in quantifying CO 2 [17]. The IOA method proposed by Leontief in 1936, which could distinguish responsibility of end user by determining the size and direction of indirect emissions embodied in supply chain across geographical boundaries [18][19][20]. Therefore, compared with bilateral trade method and carbon footprint method, IOA is more advantageous in confirming CO 2 import and export volume of various sectors and promoting redistribution of CO 2 emissions; and this method is widely used in the calculation of transfer of CO 2 , virtual water, and others [21]. However, IOA is incapable of solving uncertainty problems. If the accuracy of input data is insufficient, it affects the validity of the calculation result to a certain extent. It is worth mentioning that fuzzy theory may boost the ability to judge information subjectively, which is useful for dealing with uncertainty problems. It combines quantitative analysis with qualitative analysis to reintroduce value judgments into the analysis of human spatial behavior, enabling natural language to be used in scientific and rigorous investigations and extended in geography, economics, mathematics, and other fields [22].
However, it is indispensable to further illustrate the interaction between each component in the system, but IOA and fuzzy theory is unable to represent relationships between the various components of system [23]. It is gratifying that the introduction of ENA makes up for the shortcomings and vividly illustrates the internal relationships of the system, which is a system-oriented analysis tool that can deeply understand complex system inter-actions by researching interdependence between different regions [24][25][26]. This method was applied to study different types of relationships between components by some scholars. In virtue of ENA, Zhang et al. described internal energy transfer relationships between social and economic sectors of Beijing [27]. Through network control analysis and network utility analysis, ways of virtual water cycle and interrelationships between pairs of departments are revealed by Fang et al. [28]. Wang et al. established a multi-regional ENA model to explore structural characteristics and inter-departmental interactions within urban agglomeration in the Beijing-Tianjin-Hebei region of China [29].
Therefore, the objective of this study is to present an ecological network input-output interval fuzzy linear programming (EIFP) method, effectively proposing CO 2 reduction advice for inter-provincial transfers implicit in transportation of China under different energy types. On the basis of EIFP, an inter-provincial transportation energy consumption-carbon linkage model (TECLM) combining IOA, IRFLP, and ENA is established. Respectively, the amount of inter-provincial CO 2 transfer in transportation is quantified by IOA. IRFLP is introduced to represent double uncertainty parameter, and the fault tolerance of data source is greatly reinforced through multiple representations of input data [30]. ENA visualized relationships of inter-provincial CO 2 transfer under each of energy types. Simulation results are consistent with the emissions of various administrative regions in China, which proves validity and practicability of the framework. The research achievements would provide more valuable suggestions for adjusting CO 2 reduction responsibility of transportation and formulating CO 2 reduction policies.

Methods
This research designed an ecological network input-output interval fuzzy linear programming (EIFP) method to consider CO 2 reduction responsibilities of 30 provinces in China. The research steps are shown in Figure 1. After processing, the energy consumption of transportation and the multi-regional input-output table are taken as the input data of this model. The research scope is divided into two categories and calculated by IOA and IRFLP. Finally, the results were inputted into the data analysis section. CO 2 emission factors and ENA were used to analyze the results and draw a conclusion. The first is to quantify inter-provincial CO 2 transfer by applying IOA. It is derived from general equilibrium theory and is a combination of system analysis methods and economic measurement methods. By establishing a linear model, these relationships among economic variables are regarded as linear functional relationships [31,32]. In inter-provincial transportation energy consumption-carbon linkage model (TECLM), the total output of each province is equal to the sum of its intermediate demand and the final demand. It can be seen from Table 1 that the number of provinces in model is m, and the number of departments in each province is n. The equation for each row is The direct input coefficient a rs ij , which represents the input of unit currency products produced by sector i in province r for sector j in province s.
Formula (1) can therefore be transformed into a form containing a rs ij , the converted representation is as follows: The matrix form of input-output model can be expressed as If this matrix is transformed into a demand-oriented form, it can be expressed as Sustainability 2021, 13, x FOR PEER REVIEW 4 of 21 The matrix form of input-output model can be expressed as If this matrix is transformed into a demand-oriented form, it can be expressed as . The relationship between final demand and output has been comprehensively shown in this matrix for all provinces.  The matrix form of input-output model can be expressed as If this matrix is transformed into a demand-oriented form, it can be expressed as . The relationship between final demand and output has been comprehensively shown in this matrix for all provinces.
. . . X m1 · · · X mm   . The relationship between final demand and output has been comprehensively shown in this matrix for all provinces. Then, obtaining a direct energy consumption coefficient is an important step to combine economic trade with energy consumption. E r (ton/10 thousand Yuan) refers to the direct energy consumption required to produce a unit product.
. The relationship between final demand and output has been comprehensively shown in this matrix for all provinces.
Therefore, the virtual energy transfer matrix among provinces can be obtained from the following energy consumption coefficient matrix and total output matrix [33]: . . .
The relationship between final demand and output has been comprehensively shown in this matrix for all provinces.
The relationship between fin demand and output has been comprehensively shown in this matrix for all provinces.
After the amount of CO 2 transfer among provinces is calculated by IOA, some CO 2 emission factors are established to reflect CO 2 emission responsibility of each province from different perspectives [34].
Secondly, IRFLP is adopted to characterize the results in a variety of situations. By improving selectivity, the effectiveness of results is enhanced. Generally, IRFLP model is expressed as follows: Restrictions: To a certain extent, the application of fuzzy interval solves the validity of calculation results when input data are not accurate enough.
The third step is to visualize relationships between components by operating ENA. It is a mathematical method spread to measure the control or advantage of a department over another department in its natural environment [35,36]. The interdependence of each element could be obtained by examining the contribution of each element to input and output of other elements [37]. According to mass balance theory, the sum of CO 2 flows into ith province should be equal to the sum of flows out of ith province.
The superscript of D (ranging from 0 to m) represents path length. D 0 stands for initial flow through each compartment, D 1 for direct utility relationship, and indirect interaction expressed by D m (m > 2).
The interaction mode between these elements can be judged by the sign of each element in matrix U and matrix D [38]. There are four types of utility relations: (+,−) stands for exploitation, (−,+) for exploited, (−,−) for competition, and (+,+) for mutualistic.

Study Area
As of June 2020, the number of motor vehicles in China has reached 360 million, making it the country with the largest number of motor vehicles in the world. CO 2 emissions from transportation in China are equal to total CO 2 produced by 36.3 million tons of standard coal. Although China has taken some measures to curtail CO 2 emissions from transportation, they are still not satisfactory. For example, first of all, the natural environment and the development level of transportation in different regions are distinct; therefore, it is impractical to implement a unified CO 2 reduction standard. Secondly, with a vast territory and frequent inter-provincial trades, it is difficult to quantify CO 2 transfer and clarify CO 2 reduction responsibility. Third, due to the insufficient accuracy of input data, expressing it more effectively and comprehensively is necessary.
Therefore, an ecological network input-output interval fuzzy linear programming (EIFP) method was proposed in this research to formulate CO 2 reduction responsibility. CO 2 emission and CO 2 transfer of each province were quantified accurately after the calculation of IOA. As IRFLP increased the selectivity of calculation process, the validity and comprehensiveness of calculation results were improved. The addition of ENA intuitively illustrated relationships between provinces and provided scientific basis for planning future development direction of transportation energy structure.

CO 2 Emissions and Transfers
Based on IOA method, CO 2 emissions and transfers among provinces and districts in China are obtained. Tables 2 and A1, respectively, illustrate two characteristics of CO 2 emissions from China's transportation: regionality of transfer direction and disparity of transfer volume. As shown in Table 2 Table A1 that these transfers of different provinces are thoroughly disparate. Specifically, large transfers are those such as from Shanghai to Inner Mongolia (i.e., 5507.30 tons) and from Jiangsu to Inner Mongolia (i.e., 4842.04 tons). Small transfers are those such as from Beijing to Qinghai (i.e., 0.26 tons) and from Tianjin to Hunan (i.e., 3.79 tons) and the transfer from Shanxi to Sichuan (i.e., 0.50 tons). CO 2 emission factors are helpful for evaluating CO 2 emissions situation in different provinces. Table A3 indicates that although developed regions have better economies, resources, and technologies, less developed regions bear more pollution. Particularly, the more developed regions such as Jiangsu, Hubei, and Guangdong have larger SSA values, indicating that these provinces have larger energy self-sufficiency rates. CSAI and CSAF in Inner Mongolia are smaller, revealing that CO 2 emissions from transportation have caused serious pollution to this province. From the perspective of DOA, Shanghai, Jiangsu, and Guangdong have larger DOAI, which states that these provinces emit more CO 2 to others for local transportation development. By observing COA and DOA, it can be noticed that provinces such as Inner Mongolia have larger COA values and smaller DOA values, which depicts that the main role of these provinces is to absorb CO 2 from the transportation of other provinces. Conversely, provinces such as Guangdong have smaller COA values and larger DOA values, explaining that imported energy accounts for a relatively high proportion of the energy consumption required by their transportation.

Inter-Provincial CO 2 Transfer of Different Energy Types
CO 2 emission of each kind of energy by each province and its proportion are calculated based on the energy consumption in each province and its corresponding emission coefficient. The results are presented in Table 3 (the corresponding relationship between code and province name is shown in Table A4). Subsequently, three different energy types were compared and analyzed, and the differences in CO 2 transfer of different types of energy were discussed. Through the comparative analysis of three energy groups in transportation, the differences of CO 2 emission transfer caused by different energy types are discussed. Figure 2 exhibits direct flow of CO 2 among provinces, and different provinces are represented by different colors. These letters in the outermost circle in each figure are abbreviations of provinces. The bottom thickness of each province represents the sum of the CO 2 flow between the province and other provinces, and the thicker the line is, the greater the amount of CO 2 transported. The higher the linear density is, the more CO 2 is transferred between regions. The specific value on each color represents the CO 2 flow of the province [39].   For example, in Figure 2b, the green line from E to S is thicker than the green line from E to R, which means that the transfers from Inner Mongolia to Guangdong (i.e., 2.28 thousand tons) are greater than those from Inner Mongolia to Hunan (i.e., 0.75 thousand tons). Figure 2a has a close connection, which proves that when three energy sources are considered at the same time, most provinces have mutual CO 2 transfers. The connection in Figure 2b is looser than that in Figure 2a, and the CO 2 emission flow path only exists in a few provinces, which indicates that the self-discharged CO 2 flow in this province is higher than that in other provinces. In other words, coal energy has a higher degree of self-use in the transportation industry. However, due to its small total amount, the use of coal does not interfere with the overall CO 2 emission flow, and the use of coal has obvious regional differences. Figure 2d shows that the use of natural gas in the transportation industry is relatively low; therefore, it has great development potential in the future. By comparing Figure 2b,d, it can be seen that coal and natural gas have a higher self-utilization rate than oil. The density of line in Figure 2c is similar to that of Figure 2a, which illustrates that the use of oil plays a pivotal role in transportation. In Figure 3b, the thickness of the green line is much larger than other lines, which means that coal energy in the Northeast has a high degree of self-sufficiency. Figure 3a,c are very close, which shows that oil occupies a dominant position in transportation. The transfer volume is low, but the density of these lines in Figure 3d is similar to that of Figure 3a, which illustrates that natural gas has a high circulation capacity and still has great development potential.
the density of these lines in Figure 3d is similar to that of Figure 3a, which illustrates that natural gas has a high circulation capacity and still has great development potential.

Provincial CO 2 Transfer under IRFLP Intervention
Inexact rough interval fuzzy linear programming (IRFLP) is applied to process the amount of CO 2 transfer in inter-provincial transportation. Definitely, two sets of fuzzy intervals are generated by obtaining two membership functions for each group of data through empirical method and cutting each group of functions when the membership is 0.9. Figure 4 indicates that part of calculation results of Shaanxi Province and other typical areas are divided into three groups according to the size of the transfer volume. It is worth mentioning that there are great differences in the CO 2 transfer between each pair of provinces. For example, CO 2 transfer between Shaanxi and Gansu is almost 100 times more than that between Shaanxi and Tianjin. In summary, through various characterizations of CO 2 transmission under different conditions, IRFLP enhances the validity and comprehensiveness of calculation results.

Provincial CO2 Transfer under IRFLP Intervention
Inexact rough interval fuzzy linear programming (IRFLP) is applied to process the amount of CO2 transfer in inter-provincial transportation. Definitely, two sets of fuzzy intervals are generated by obtaining two membership functions for each group of data through empirical method and cutting each group of functions when the membership is 0.9. Figure 4 indicates that part of calculation results of Shaanxi Province and other typical areas are divided into three groups according to the size of the transfer volume. It is worth mentioning that there are great differences in the CO2 transfer between each pair of provinces. For example, CO2 transfer between Shaanxi and Gansu is almost 100 times more than that between Shaanxi and Tianjin. In summary, through various characterizations of CO2 transmission under different conditions, IRFLP enhances the validity and comprehensiveness of calculation results.

ENA of CO2 Emissions in Different Provinces and Regions
Among all utility relationships, more mutualism and competition relationships should be developed. Because mutualism relationships are able to make the corresponding two parties mutually beneficial and for competition relationships, the negative impact can be offset by the interaction between provinces. What needs to be pointed out is that more attention should be paid to changing the relationship between exploitation and exploited. Figure

ENA of CO 2 Emissions in Different Provinces and Regions
Among all utility relationships, more mutualism and competition relationships should be developed. Because mutualism relationships are able to make the corresponding two parties mutually beneficial and for competition relationships, the negative impact can be offset by the interaction between provinces. What needs to be pointed out is that more attention should be paid to changing the relationship between exploitation and exploited. Figure 5A-D has 465 pairs of utility relationships for each energy type. In general, more exploitation relationships, exploited relationships, and competition relations are shown in the system but fewer mutualism relationships. Specifically, in the category of total energy, there are 124 pairs of exploitation relationships (26.67%), 149 pairs of exploited relationships (32.04%), 159 pairs of competition relationships (34.19%), and 37 pairs of mutualism relationships (7.96%). It is observed that competition relations dominate the category of coal. The oil group is similar to total energy group, which means that oil occupies a major position in the system of transportation. In the natural gas group, the proportion of competition relationships is higher than that of other categories, which means that increasing the proportion of natural gas consumption is a decisive direction for the development of transportation in future. Figure 5a-d has 28 pairs of utility relationships. For example, Figure 5a,d reveals that in the group of total energy, the utility relations between CC and SC, CC and NC are all exploited relationships (−,+), while in the natural gas group, the two pairs of utility relations have become competition relationships (−,−). It represents that between CC and SC, CC and NC, the negative impact of inter-provincial CO 2 transfer could be effectively offset by increasing natural gas consumption. In addition, simulation results show that relationships between components in ecological networks have not changed after operating IRFLP, which indicates that TECLM has a certain ability to deal with uncertainty. pairs of mutualism relationships (7.96%). It is observed that competition relations dominate the category of coal. The oil group is similar to total energy group, which means that oil occupies a major position in the system of transportation. In the natural gas group, the proportion of competition relationships is higher than that of other categories, which means that increasing the proportion of natural gas consumption is a decisive direction for the development of transportation in future. Figure 5a-d has 28 pairs of utility relationships. For example, Figure 5a,d reveals that in the group of total energy, the utility relations between CC and SC, CC and NC are all exploited relationships (−,+), while in the natural gas group, the two pairs of utility relations have become competition relationships (−,−). It represents that between CC and SC, CC and NC, the negative impact of inter-provincial CO2 transfer could be effectively offset by increasing natural gas consumption. In addition, simulation results show that relationships between components in ecological networks have not changed after operating IRFLP, which indicates that TECLM has a certain ability to deal with uncertainty.

CO2 Reduction Measures
The CO2 reduction measures in transportation can be divided into micro-perspectives and macro-perspectives. From micro-perspectives, first of all, electrification is an influential part of settling the growing emission challenges in transportation, because electrification eliminates exhaust emissions and realizes the potential for decarbonization of grid. Secondly, increase the proportion of the public transport market in passenger car market, and combine public transport with electrification to effectively control CO2 emissions. Third, incorporate new energy vehicles into the urban transportation network and boost utilization rate of shared cars and shared bicycles to reduce CO2 emissions. The second is to improve urban planning, build livable urban communities, and reduce traffic demand from the source as much as possible to achieve the goal of reducing CO2 emissions.

Conclusions
An ecological network input-output interval fuzzy linear programming (EIFP) method was established to clarify CO2 reduction responsibilities of different regions. Explicitly, this research quantified inter-provincial CO2 transfer of transportation; performed fuzzy processing on each CO2 transfer; and visualized mutual transfer relationships between provinces. On the basis of research results, different CO2 reduction recommendations are offered for different regions. For instance, Beijing should focus on the development of local renewable energy and other emerging energy sources to strengthen the ability to withstand energy crisis. Provinces with greater demand for fossil energy, such as Shandong, should introduce new policies to encourage all walks of life to vigorously develop energy-saving technologies by increasing investment in energy-saving technologies. What is more, input data of TECLM can be changed and applied to various regions. The EIFP method could further help to divide the cross-provincial CO2 reduction responsibility of transportation, which may provide effective theoretical foundation for decisionmakers to formulate CO2 reduction policies of transportation in China.

CO 2 Reduction Measures
The CO 2 reduction measures in transportation can be divided into micro-perspectives and macro-perspectives. From micro-perspectives, first of all, electrification is an influential part of settling the growing emission challenges in transportation, because electrification eliminates exhaust emissions and realizes the potential for decarbonization of grid. Secondly, increase the proportion of the public transport market in passenger car market, and combine public transport with electrification to effectively control CO 2 emissions. Third, incorporate new energy vehicles into the urban transportation network and boost utilization rate of shared cars and shared bicycles to reduce CO 2 emissions. The second is to improve urban planning, build livable urban communities, and reduce traffic demand from the source as much as possible to achieve the goal of reducing CO 2 emissions.

Conclusions
An ecological network input-output interval fuzzy linear programming (EIFP) method was established to clarify CO 2 reduction responsibilities of different regions. Explicitly, this research quantified inter-provincial CO 2 transfer of transportation; performed fuzzy processing on each CO 2 transfer; and visualized mutual transfer relationships between provinces. On the basis of research results, different CO 2 reduction recommendations are offered for different regions. For instance, Beijing should focus on the development of local renewable energy and other emerging energy sources to strengthen the ability to withstand energy crisis. Provinces with greater demand for fossil energy, such as Shandong, should introduce new policies to encourage all walks of life to vigorously develop energy-saving technologies by increasing investment in energy-saving technologies. What is more, input data of TECLM can be changed and applied to various regions. The EIFP method could further help to divide the cross-provincial CO 2 reduction responsibility of transportation, which may provide effective theoretical foundation for decision-makers to formulate CO 2 reduction policies of transportation in China.
Compared with previous studies, this study innovates the method and introduces inexact rough interval fuzzy linear programming (IRFLP) to enhance the ability to deal with data uncertainty. In addition, in terms of research content, this study adopted the "China Multi-Regional Input-Output Table 2015", focused on the interprovincial transfer of CO 2 in the transportation industry and future development planning, and established a relatively comprehensive modeling framework, which is relatively lacking in previous studies.
The limitations of this study are that it only studied the internal situation of transportation between 30 provinces and 7 regions. In terms of possible future recommendations of the study, we believe that it can be started from two aspects: the first is to expand the scope of research and explore the transfer of CO 2 in transportation between countries, and the second is to expand the category of research departments and cross-consider the mutual influence between departments, so as to get better research results.

Institutional Review Board Statement:
This study did not involve humans or animals.

Informed Consent Statement:
This study did not involve humans.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author. The data are not publicly available due to data which also forms part of an ongoing study.

Acknowledgments:
The authors are grateful to the editors and the anonymous reviewers for their insightful comments and suggestions.

Conflicts of Interest:
The authors declare no conflict of interest.

Notations y
Objective function G R ,H R Rough sets K R Decision variable z rs ij Intermediate input from sector i in province r to sector j in province s f rs i Input for the final demand of province s from sector i in province r X r i Total output of sector i in province r a rs ij Direct input from sector i in province r for the production of one unit of product by sector j in province s A rs Direct consumption coefficient matrix of province r to province s X rs Output of province r that is pulled by the total final demand that is used in province s (I − A) −1 Leontief inverse matrix