A Long-Term Decarbonisation Modelling and Optimisation Approach for Transport Sector Planning Considering Modal Shift and Infrastructure Construction: A Case Study of China

Abstract

Reducing direct carbon emissions in the transport sector is crucial for carbon neutrality. It is a considerable challenge to achieve substantial CO₂ emissions reductions while satisfying rapidly growing traffic demands. Previous studies cannot be applied directly in long-term planning for the transport sector with rapid demand growth. To bridge this gap, a multi-regional model is proposed in this paper to quantify the optimal decarbonisation path for the transport sector in order to save costs. Considering modal shift and infrastructure construction, this model regards the transport sector as a whole and China is taken as a case study. The results show that electricity and hydrogen will be the major fuels of the transport sector in the future, accounting for 45 percent and 25 percent of fuel demands in 2060. This means that the electricity used by the transport sector accounts for 10 percent of the electricity consumed by the whole of society. The results reflect that freight transport has reached a CO₂ emissions peak, while passenger transport will reach its own CO₂ emissions peak around 2041. Giving priority to decarbonisation in freight transport can save 5 percent of the transition cost. The results also suggest that modal shift can save at most 7 percent of the transition cost.

1. Introduction

The transport sector is one of the largest sources of direct CO₂ emissions, responsible for 24 percent of global CO₂ emissions (through fuel combustion) in 2019 [1]. Efficient decarbonisation measures are expected to be implemented in the transport sector to meet the 1.5 °C target and achieve carbon neutrality, or the risk of a costly and disorderly transition will increase [2]. Due to rapid populational and economic growth, traffic demands are expected to rise continuously, posing great challenges for the low-carbon transition of the transport sector. Irrational planning and reckless movements towards transition in the transport sector could lead to unacceptable expenses. The costs of developing a low-carbon transport sector can be reduced significantly by quantifying the optimal scale of different types of traffic volumes and infrastructure.

Current studies mostly focus on a single transport subsector. The decarbonisation of the road subsector has been broadly studied formerly, for the reason that CO₂ emissions from the road subsector account for 75% of CO₂ emissions from the transport sector [3]. Maduekwe et al. [4], Fan et al. [5] and Tsita et al. [6] applied LEAP to design sustainable road transport systems in Lagos, Nigeria, Beijing, China and Greece, respectively. In addition to planning for the road subsector, several studies have been conducted to focus on the influence of specific low-carbon policies or technologies in the road subsector. Virley examined the influence of high oil prices on road transport CO₂ emissions in the UK [7] and Brand et al. explored the CO₂ emission reduction potential of policy instruments that influence consumers’ car preferences, such as the UK road tax [8], while Santos and Rembalski discussed UK subsidies for electric cars [9]. Dhar et al. explored the effect of electric cars on CO₂ emissions reduction in India using ANSWER-MARKAL [10]. Larsson et al. [11] and Li et al. [12] estimated the implications of fuel cell electric cars in Sweden and ASEAN countries separately, which have great potential to replace oil-fuelled cars in the future. Verger et al. overviewed the application of biomass-based fuels for heavy-duty transport [13] and Altun et al. examined the effect of using n-butanol in conventional diesel fuel–biodiesel blends on reducing engine exhaust emissions [14]. As emissions from aviation and shipping continue to grow rapidly, these hard-to-abate transport subsectors are also receiving increased attention [1]. Several studies verified the positive impact of carbon pricing on deepening emissions cuts in aviation [15,16]. Rehmatulla et al. analysed the obstacle of wind technologies on ships qualitatively and possible solutions [17], while Stolz et al. concentrated on sustainable fuels for shipping in Europe [18]. However, due to tight carbon budgets and premature utilisation of renewable technologies in each subsector, the discussion of decarbonisation in individual subsectors will inevitably result in high overall costs for the whole transport sector. It has been proved that the total costs may be significantly overestimated when decoupling a system into several subsectors [19]. Hence, to save transformation costs as much as possible, the transport sector should be viewed as a whole instead of being divided into separate subsectors. Routine analysis of the holistic transport sector is carried out with developed energy models, such as TIMES [20], LEAP [21] and EnergyPLAN. Zhang et al. compared the decarbonisation perspective of the transport sector in China and the USA using China-TIMES and US-TIMES [22], while Korkmaz et al. applied the TIMES PanEU model to establish a decarbonisation path for the transport sector in Europe [23]. LEAP and EnergyPLAN were used to evaluate the efficiency of low-carbon policies for the transport sector announced by South Korea [24] and to formulate emission-reduction policies for the transport sector in Iran [25].

Although the above research is not limited to only one subsector, modal shifts between different subsectors and the construction of infrastructures are not evaluated quantitatively. Modal shift refers to the use of vehicles from one subsector to replace the vehicles from another subsector and is also known as behaviour change. For instance, a passenger who is used to travelling by air may decide to travel by train. It has been proposed that modal shift is a means of achieving emissions reductions with extraordinary potential [26,27] and which it is necessary to consider for areas with high emission-reduction costs in the transport sector. Salvucci et al. applied elasticities of substitution to describe passenger transport modal shift and developed TIMES-DKEMS to assess the impact of policies on modal shift [28], while Strauss et al. evaluated the influence on decarbonisation in China of modal shift from air to high-speed trains [29]. Zhang et al. and Kurtuluş et al. analysed the impact and potential of modal shift on freight transport in Shenzhen, China, and in Turkey, respectively [30,31]. Similar to most of the previous studies considering modal shift, these studies did not pay enough attention to the construction of infrastructure, which is the main driver and constraint on modal selection [32]. Hickman et al. considered the transition path to low-carbon transport in London and Delhi [33], while Daly et al. reformed TIMES to optimise the costs of decarbonisation of the transport sector in Ireland and California [34]. Both sets of authors considered the construction of infrastructure, including rail tracks, and regarded the studied area as a point. However, since the length of rail tracks cannot be depicted as an integral multiple of the distance between two regions, the studies may have underestimated the cost of railway construction. Due to the fact that traffic demand in previous studies increased relatively slightly in the planning period [35,36,37], even if the description of infrastructure construction is simplified, there is little impact on the results. Nevertheless, the rapid or substantial growth of traffic demands cannot be avoided in the context of long-term planning in the transport sector. When considering modal shift, it is inevitable to describe the construction of infrastructure, while the majority of studies that have considered modal shift and infrastructure have greatly simplified the construction of distance-dependent infrastructures, such as railways.

According to the literature review, previous models cannot be applied directly to find the appropriate long-term low-carbon transition pathways for the transport sector. In this paper, a multi-regional model considering modal shift and infrastructure that can calculate the transmission pathway of the transport sector is proposed; the model covers four modes, including road, rail, aviation and shipping, both in passenger and freight transport. China is taken as a case study; its transport demand is considered to have large room for growth, with the transport sector accounting for about 9.12% of the country’s total CO₂ emissions and 19.42% of the country’s total end-use CO₂ emissions in 2019 [38]. Through the case study of China, by using this model, the following issues are expected to be resolved:

• The role of various fuels in transport in the movement towards low-carbon transition and the means of reducing CO₂ emissions in the terminal sector of transport;
• The means of minimizing the costs of the transition of the transport sector when faced with rapidly growing traffic demands;
• The benefits of favourable CO₂ reduction policies in different sectors when considering the transport sector as a whole.

The novelty of this work can be summarised in three points. Firstly, viewing the transport sector as an entirety, a multi-regional optimisation model using the superstructure method is established. This can be applied to analyse the long-term transition path of transport sectors in regions with snowballing traffic demands and unformed infrastructures and assess the construction of inter-regional infrastructures. Secondly, through the case study, a likely transitional route for the transport sector in China is described and the impact of different travel habits considering modal shift and diverse policies are analysed, which analysis can be used to explain the gravity of the prioritisation of emissions reduction in freight transport. Thirdly, the relationship between the power sector and the transport sector is pointed out quantitatively.

This paper is organised as follows: Section 1 clarifies the background of this study. Section 2 presents the method in detail. Section 3 describes the basic details of the case study. Section 4 presents the results and sensitivity analysis of essential parameters. Section 5 proposes policy recommendations and Section 6 summarises the main conclusions.

2. Methodology

2.1. Model Structure and Assumptions

A schematic diagram of the relevant concepts is shown in Figure 1 and the model structure is shown in Figure 2. Several concepts need to be clarified here. A city is the smallest unit in the spatial dimension of the model, which can be seen as a point. A region is composed of several cities. People and goods can move in a city or between different cities, which is regarded as inner-city and inter-city transport. Only trucks in freight transport, private cars, taxis, subways and buses in passenger transport are allowed to be utilised in cities. Inter-city transport includes communication between different cities in the same region and different cities in diverse regions. Transport between different cities in diverse regions can also be called inter-regional transport. All transport modes except taxis, buses and subways are allowed to be applied in inter-city transport. In this study, the construction of inter-regional rail tracks is evaluated significantly, which requires the description of the distances between regions, while the construction of rail tracks within one region is not considered. The distance between two regions is considered as the distance between the largest traffic point cities in the two regions in our model. For historical, social and economic reasons, the completeness of transport facilities varies between different cities. A traffic point city is a city with the most complete transport infrastructure and high passenger or freight throughput in a region.

2.2. Mathematical Formulation

Mathematical formulas for this optimisation model are presented in this section. The sets t, ser, f and r denote time, traffic mode type (there being 12 modes, as presented in Figure 2), fuel type and region, respectively. What is more, r and rr are different names for the set region. Parameters are represented with uppercase letters and variables are represented with lowercase letters. The meanings of the different symbols are listed in Abbreviations.

2.2.1. Objective Function

The objective function of the transport optimisation model is to minimise the total cost of the transport sector from 2020 to 2060. The total cost is composed of the infrastructure cost, purchase and maintenance costs of both freight and passenger transport and fuel costs and is calculated using Equation (1). Four parts of the total cost are presented in detail in Equations (2)–(5). In this study, infrastructure, such as charging piles, subway tracks, rail tracks (including both ordinary rail and high-speed rail), airports and ports, are considered, as Equation (2) shows. The annual cost of each item of infrastructure includes two parts: new construction cost and maintenance cost. This is the same as the cost composition of vehicles, as shown in Equations (3) and (4).

$$cost={{\displaystyle \sum }}_{t=2020}^{2060}\frac{inf{c}_t+buy{c}_t+main{c}_t+fuel{c}_t}{{\left(1+I\right)}^{t-2020}}$$

(1)

$$inf{c}_t=charge{c}_t+metro{c}_t+rail{c}_t+airport{c}_t+port{c}_t$$

(2)

$$buy{c}_t={{\displaystyle \sum }}_{ser,f}BU{Y}_{t,ser,f}\ast nn{b}_{t,ser,f}$$

(3)

$$main{c}_t={{\displaystyle \sum }}_{ser,f}MAI{N}_{t,ser,f}\ast \left(n{b}_{t,ser,f}-nn{b}_{t,ser,f}\right)$$

(4)

$$fuel{c}_t={{\displaystyle \sum }}_{r,rr,ser,f}t{s}_{t,r,rr,ser,f}\ast TRF{C}_{t,ser,f}\ast FCOS{T}_{t,r,f}$$

(5)

2.2.2. Supply and Demand Constraints

All demands in each transport mode are separated into two kinds: inescapable demands and changeable demands. It is considered that a certain proportion of demands are immutable in each traffic mode, which is the explanation of Equation (6). Only traffic modes with the same scope can convert mutually. In other words, in passenger transport, the modes of taxis, subways and buses can convert to each other, while the modes of coaches, trains, planes and boats can convert mutually. The four modes of freight transport can convert to each other. In this model, the participation of private cars in modal shift is not considered. Although there is a conversion between different modes, both freight and passenger transport should meet the requirement that total supply is not less than the total demand, as described in Equations (7)–(10). The constraint of travel time budget also needs to be considered in passenger transport. The total travel time budget can be computed using Equation (11), which has to meet the constraint of Equation (12), implying that the travel time budget for each person in a day is stationary every year, according to Tattini et al. [39]. Travel time budget makes the result of passenger transport mode conversion closer to reality.

$${{\displaystyle \sum }}_{f}t{s}_{t,r,rr,ser,f}\ge T{D}_{t,r,rr,ser}\ast FIXPR{O}_{ser}$$

(6)

$${{\displaystyle \sum }}_f\left(t{s}_{t,r,rr,taxi,f}+t{s}_{t,r,rr,subway,f}+t{s}_{t,r,rr,bus,f}\right)\ge T{D}_{t,r,rr,taxi}+T{D}_{t,r,rr,subway}+T{D}_{t,r,rr,bus}$$

(7)

$$\begin{gathered} {{\displaystyle \sum }}_f\left(t{s}_{t,r,rr,coach,f}+t{s}_{t,r,rr,train,f}+t{s}_{t,r,rr,plane,f}+t{s}_{t,r,rr,boat,f}\right)\ge \\ T{D}_{t,r,rr,coach}+T{D}_{t,r,rr,train}+T{D}_{t,r,rr,plane}+T{D}_{t,r,rr,boat} \end{gathered}$$

(8)

$${{\displaystyle \sum }}_{f}n{b}_{t,private car,f}\ge NB{D}_{t,r}$$

(9)

$$\begin{gathered} {{\displaystyle \sum }}_f\left(t{s}_{t,r,rr,truck,f}+t{s}_{t,r,rr,railway,f}+t{s}_{t,r,rr,aircraft,f}+t{s}_{t,r,rr,vessel,f}\right)\ge \\ T{D}_{t,r,rr,truck}+T{D}_{t,r,rr,railway}+T{D}_{t,r,rr,aircraft}+T{D}_{t,r,rr,vessel} \end{gathered}$$

(10)

$$totaltt{b}_t={{\displaystyle \sum }}_{r,rr,pas,f}t{s}_{t,r,rr,pas,f}/SPEE{D}_{pas,f}$$

(11)

$$totaltt{b}_t=TTB\ast PO{P}_t\ast 365$$

(12)

2.2.3. Vehicle Constraints

For all traffic modes except private cars, the relationship between the number of vehicles and turnover described in Equation (13) should be satisfied. Private cars are generally not used commercially and it is difficult to determine the parameter $LOA{D}_{private car,f}$. In our model, it is assumed that the annual mileage of each private car is determined, avoiding the model becoming a nonlinear model. As a result, the traffic turnover of private cars in a year is shown in Equation (14). Parameter $P$ means the average carrying capacity of a private car, which can be estimated by Equations (11) and (12) using the concept of $TTB$. The number of vehicles is presented in Equation (15). In this model, only oil-fuelled vehicles in the rail, aviation and shipping subsectors of passenger and freight transport can be decommissioned before their expiration date, implying that the number $retir{e}_{t,ser,f}$ is fixed to 0 for other means of transport.

$${{\displaystyle \sum }}_{r,rr}t{s}_{t,r,rr,ser,f}\le LOA{D}_{ser,f}\ast n{b}_{t,ser,f}$$

(13)

$${{\displaystyle \sum }}_{r,rr}t{s}_{t,r,rr,private car,f}=n{b}_{t,private car,f}\ast MILEAGE\ast P$$

(14)

$$n{b}_{t,ser,f}=n{b}_{t-1,ser,f}+nn{b}_{t,ser,f}-nn{b}_{t-tl{t}_{ser,f}-1,ser,f}-retir{e}_{t,ser,f}$$

(15)

2.2.4. Infrastructure Constraints

Infrastructure needs to match the traffic demand, as calculated in Equations (16)–(23). Parameter $PRO$ is the minimum ratio of the amount of infrastructure and traffic demand. Traffic demand can be the number of vehicles, the transport turnover or the capacity. The expiration date of infrastructure is not considered in order to simplify the model, for the reason that the lifespan of most transport facilities can be extended for a long time after regular maintenance. As a result, the number of specific transport facilities in year t equals the sum of newly added facilities in year t and maintaining facilities in year t − 1.

$$nbcharg{e}_t\ge PR{O}_{charge}\ast \left(n{b}_{t,private car,elec}+n{b}_{t,taxi,elec}+n{b}_{t,bus,elec}+n{b}_{t,coach,elec}+n{b}_{t,truck,elec}\right)$$

(16)

$$nbmetr{o}_{t,r}\ge PR{O}_{metro}\ast n{b}_{t,subway,r}$$

(17)

$$nbrai{l}_{t,r,rr,f}\ast DISTANC{E}_{r,rr}\ge PR{O}_{rail,pas}\ast \left(t{s}_{t,r,rr,train,f}+t{s}_{t,rr,r,train,f}\right)/DISTANC{E}_{r,rr}$$

(18)

$$nbrai{l}_{t,r,rr,f}\ast DISTANC{E}_{r,rr}\ge PR{O}_{rail,fre}\ast \left(t{s}_{t,r,rr,railway,f}+t{s}_{t,rr,r,railway,f}\right)/DISTANC{E}_{r,rr}$$

(19)

$$nbpor{t}_{t,r}\ge PR{O}_{port,pas}\ast \left(t{s}_{t,r,rr,boat,f}+t{s}_{t,rr,r,boat,f}\right)/DISTANC{E}_{r,rr}$$

(20)

$$nbpor{t}_{t,r}\ge PR{O}_{port,fre}\ast \left(t{s}_{t,r,rr,vessel,f}+t{s}_{t,rr,r,vessel,f}\right)/DISTANC{E}_{r,rr}$$

(21)

$$nbairpor{t}_{t,r}\ge PR{O}_{airport,pas}\ast \left(t{s}_{t,r,rr,plane,f}+t{s}_{t,rr,r,plane,f}\right)/DISTANC{E}_{r,rr}$$

(22)

$$nbairpor{t}_{t,r}\ge PR{O}_{airport,fre}\ast \left(t{s}_{t,r,rr,aircraft,f}+t{s}_{t,rr,r,aircraft,f}\right)/DISTANC{E}_{r,rr}$$

(23)

2.2.5. Maximum Speed of Modal Shift

In order to prevent a sudden increase in traffic supplementation, the maximum speed of modal shift is settled. This restriction has two practical implications. Firstly, the transformation of ideas takes time, meaning that the shift towards another mode cannot happen suddenly. Secondly, the number of vehicles that can be purchased is determined by the inventory, as it is impossible to satisfy an unexpected high demand immediately. The maximum speed of modal shift is set to be the increase in demand of one mode, which is depicted in Equation (24).

$$\mathsf{\Delta }{{\displaystyle \sum }}_{r,rr}t{s}_{t,r,rr,ser,f}\le {{\displaystyle \sum }}_{r,rr}\mathsf{\Delta }T{D}_{t,r,rr,ser}$$

(24)

2.2.6. Policy Constraints

Total CO₂ emissions from the transport sector can be calculated using Equation (25). CO₂ emissions from the transport sector are subject to policy constraints and cannot exceed the specified upper limit, as shown in Equation (26).

$$ycarbo{n}_t={{\displaystyle \sum }}_{r,rr,ser,f}t{s}_{t,r,rr,ser,f}\ast TRF{C}_{t,ser,f}\ast F{C}_f$$

(25)

$$ycarbo{n}_t\le TRCO{2}_t$$

(26)

3. Case Study

China is used as a case study for the proposed model because of its huge economic growth potential allowing for explosive growth in traffic demand. The model of the case study is coded on the platform of the General Algebraic Modelling System (GAMS). Critical input data and fundamental assumptions are clarified in this section.

3.1. Basic Assumptions and Scenario Settings

The mainland of China, excluding Macao, Hong Kong and Hainan province, is considered in this case study and is divided into nine regions. The optimisation time interval is from 2020 to 2060; several variables for 2020 have been rectified according to published statistical data [40].

Several scenarios have been set in this case study, while crucial settings for the basic scenario discussed in Section 4 can be described as follows. Firstly, CO₂ emissions from the transport sector in China reach a peak in 2030, with no more than 1091 million tonnes and at most 100 million tonnes in 2060. At the same time, we set 2030–2035 as the platform period for decarbonisation. Secondly, modal shift in the freight sector is ignored. Although the validation of modal shift has been proven in China [31], conversion is difficult. Over the past decade, numerous policies of decarbonisation have been announced in the transport sector in China, but the shift towards low-carbon density in freight transport is not evident. The proportion of shipping has increased by only 0.7% [41]. Thirdly, electric vehicles in aviation and shipping subsectors can only be applied after 2045. The input traffic demands are shown in

Table 1. Demand for freight transport (billion tonne-kilometres).

	2030	2040	2050	2060
Truck	9169	12,286	14,979	16,898
Railway	4875	6523	7708	8237
Aircraft	39	59	78	89
Vessel	11,735	12,754	13,726	14,635

and

Table 2. Demand for passenger transport (billion passenger-kilometres).

	2030	2040	2050	2060
Taxi	247	293	317	316
Subway	309	419	514	582
Bus	1014	1298	1512	1634
Coach	607	803	967	1080
Train	1566	1938	2171	2273
Plane	1659	2704	3483	3714
Boat	10	11	12	12

.

3.2. Historical Quantity and Service Year of Different Means of Transportation

Historical quantities for various means of transport were obtained through statistical yearbooks for China [42,43,44,45,46] and public data. Remarkably, the number of vehicles increased significantly around 2010, which may result in an extensive retirement henceforth.

The service years of different types of vehicles are listed in

Table A1. Maximum service life of vehicles.

	Gasoline	Diesel	Hybrid	Electricity	Gas	Hydrogen	Jet-Fuel	Bio-Jet-Fuel
Car	15	—	15	10	—	—	—	—
Taxi	15	—	—	8	8	—	—	—
Subway	—	—	—	30	—	—	—	—
Bus	—	10	—	10	10	10	—	—
Coach	—	10	—	10	10	10	—	—
Train	—	100	—	30	—	—	—	—
Plane	—	—	—	15	—	—	15	15
Boat	—	30	—	25	30	30	—	—
Truck	—	10	—	10	10	10	—	—
Railway	—	100	—	30	—	—	—	—
Aircraft	—	—	—	20	—	—	20	20
Vessel	—	35	—	30	35	35	—	—

in Appendix A, with reference to the relevant national standards.

3.3. The Ceiling of Carrying Capacity of Vehicles and Infrastructure

Few studies have discussed the upper limit of the carrying capacity of vehicles and infrastructure in one year, making it difficult to determine the parameter $LOA{D}_{ser,f}$ and $PRO$ in the case study. In this study, the determination of the above parameters refers to the current state of traffic development in China. Figures for the top ten crowded subway lines in China in 2019 were computed to settle the capacity of subways and figures for the top five busiest railway lines in China in 2019 were calculated to fix the capability of trains. The proportion of electric vehicles and charging piles is set at 3, according to the China Electric Vehicle Charging Infrastructure Promotion Alliance.

3.4. Investment Costs of Vehicles and Infrastructure

The current investment costs refer to public data, which are listed in

Table A2. Investment costs of vehicles in 2020 (CNY million).

	Gasoline	Diesel	Hybrid	Electricity	Gas	Hydrogen	Jet-Fuel	Bio-Jet-Fuel
Car	0.2	—	0.27	0.3	—	—	—	—
Taxi	0.2	—	—	0.3	0.2	—	—	—
Subway	—	—	—	40	—	—	—	—
Bus	—	0.4	—	1.7	0.5	2	—	—
Coach	—	0.4	—	1.7	0.5	2	—	—
Train	—	50	—	90 (high-speed 170)	—	—	—	—
Plane	—	—	—	—	—	—	385	385
Boat	—	0.8	—	—	12.6	—	—	—
Truck	—	0.3	—	1	0.4	1.2	—	—
Railway	—	10	—	50	—	—	—	—
Aircraft	—	—	—	—	—	—	467	385
Vessel	—	35	—	—	45.5	—	—	—

and

Table A3. Scope of the investment costs of infrastructure in 2020 (CNY million).

Metro Rail CNY M/KM	Rail CNY M/KM	Highrail CNY M/KM	Airport CNY M/Runway	Port CNY M/Berth	Charging Pile CNY M/Number
500–1500	68–316	102–900	10,000–20,000	200	0.03

Note: Infrastructure costs in different regions vary for topographical and economic reasons.

in Appendix A. The costs of vehicles are expected to decline with improvements in technology. However, the future cost of infrastructure is much more complex. On the one hand, as with vehicles, the maturity of technology will lead to a decrease in costs; on the other hand, primary infrastructure is often constructed on flat terrain, while subsequent infrastructures will be constructed in relatively built-up areas, resulting in higher costs. After comprehensive consideration, we assume that infrastructure costs in different years remain constant except for the costs of charging piles.

3.5. Forecast of Costs of Fuels

The price of primary fuels is regarded as invariant in our case study. We do not assume an increase in the cost of fossil fuels or a decrease in the cost of electricity, as we consider carbon tax and subsidies to be matters of policy instead of prerequisite assumptions. In addition, the future output of fossil energy is uncertain, especially after the outbreak of coronavirus disease 2019 (COVID-19), which makes it more difficult to predict future prices [47,48]. Furthermore, the increase in renewable electricity and its supporting infrastructure will increase the cost of electricity, which will introduce uncertainty regarding the cost of electricity. However, considering the development of technologies, the cost of hydrogen and bio-jet-fuel is set to decline by 2% per year. The detailed costs are listed in

Table A4. Costs of fuels in 2020 (CNY/kg).

	Gasoline	Diesel	Electricity	Gas	Hydrogen	Jet-Fuel	Bio-Jet-Fuel
Northeast	9.45	7.56	0.71	5.28	65	7	60
North	9.44	7.77	0.75	4.67	65	7	60
Central	9.45	7.71	0.67	5.23	65	7	60
East	9.39	7.7	0.64	5.89	65	7	60
South	9.45	7.72	0.65	4.50	65	7	60
Northwest	9.30	7.61	0.56	3	65	7	60
Xinjiang	9.34	7.61	0.45	1.86	65	7	60
Inner Mongolia	9.36	7.59	0.45	2.46	65	7	60
Tibet	9.60	7.79	0.66	5.95	65	7	60

.

3.6. Travel Time Budget

Schafer and Victor showed that, on average, a person spends about 1.1 h per day travelling across various societies and income classes [49]. However, this conclusion cannot be applied directly in our case study. Firstly, non-motorised travel modes, such as walking and biking, are not included in our case study due to a lack of statistics. Previous studies have shown that although non-motorised and motorised transport modes shift freely, the travel time budget for motorised transport modes is around 45 min [39], which can also be verified by statistics for China before 2020. Secondly, the influence of COVID-19 on the travel time budget of China cannot be ignored. During data calibration, it was found that the travel time budget in 2020 is far less than 45 min, owing to people being recommended to stay at home and restrict their movements. It is unknown how long the epidemic will affect China. However, the travel time budget in 2020 is the shortest in the next 40 years. Consequently, the travel time budget in our case study is constrained between the travel time budget in 2020 and 45 min. The average speed of different transport modes is listed in

Table A5. Modal speed (Km/h).

Car	Taxi	Subway	Bus	Coach	Train	Plane	Boat
55	38	35	18	45	60 (high-speed 220)	900	33

.

3.7. The Proportion of Specific Requirements in Passenger Transport

The review of scientific literature did not review any conclusions on the proportion of specific requirements in passenger transport. According to the characteristics of different transport modes, we set different values. The proportion of specific requirements for taxi, coach and plane is set as 0.6, while others are set as 0.4. Since this parameter is determined subjectively, we discussed its influence on this lately.

4. Results and Discussion

4.1. Optimal Structure of the Freight Sector

According to the results, the proportion of turnover powered by oil decreases in all kinds of modes of freight transport in China, as shown in Figure 3. This makes clear that although it is nearly a decade away from the peak of CO₂ emissions, the proportion of oil used in freight transport has no room to ascend. The role natural gas plays in road and shipping subsectors is divergent. Natural gas is a transitional fuel in trucks, while it plays a role as an alternative energy source until 2060 in vessels. Although the proportion of natural gas vehicles in vessels increases slowly, gas-fuelled turnover rises apparently with the enlargement of shipping demands. This shows the inevitability of the development of LNG powered ships. Gas-fuelled turnover in the shipping subsector can reach 4123 billion tonne-kilometres in 2060, accounting for 28.2%.

As CO₂-free fuels, electricity and hydrogen are the future for freight transport in China. Hydrogen can only be applied in road and shipping subsectors. Hydrogen-powered trucks are expected to carry 15 times more turnover than in 2020 and electric trucks are expected to carry 21 times more turnover than in 2020. Hydrogen-powered vessels complement the reduction in the number of diesel-powered vessels partially due to carbon reduction, but hydrogen cannot replace diesel due to the high costs. Electricity will be one of the major energy sources for the future freight transport sector, providing 35%, 75% and 0% of energy needs in 2030 and 76%, 100% and 12% in 2060 for the road, railway and shipping subsectors.

4.2. Optimal Structure of the Passenger Sector

Figure 4 and Figure 5 describe the inner-city passenger turnover and the structure of taxi and bus turnover. It can be observed, unexpectedly, that the turnover of buses rises incredibly due to modal shift. The reason for the slow growth of subway turnover is that the high cost of track construction limits the development of subways. However, bus turnover does not increase continuously. It reaches a peak in 2049 and then declines. Excessive bus travelling will add to travel time budgets, which is not in conformity with the law of constant travel time budgets. One unanticipated finding from Figure 5 is that oil use in both taxi and bus turnover increases first and then decreases, while the proportion of electricity shows an opposite trend. A possible explanation for this might be that passenger transport contributes less to CO₂ emissions and that the pressure to decarbonise is lighter. For the sake of saving costs, the freight transport should decarbonise first, while passenger transport should reduce its CO₂ emissions later. The decrease in the proportion of electricity-powered turnover implies that developing electric vehicles immediately is not a necessity, due to the high cost of electric vehicles and charging piles. This result is also reasonable. At present, many electric vehicles are promoted by demonstration projects or subsidies with which we are not concerned in our case study. The results suggest that electric vehicles are not economically competitive without subsidies at present. The trend for gas-fuelled taxi turnover changes three times. It was found that there was a boom in gas-fuelled taxies in the past decade and a rapid decline soon after, causing a downward trend from 2020 to 2035. After 2035, natural gas plays an essential role as a transitional fuel in taxi transport. Taxies are supposed to have 100% access to electricity in 2060.

The results for inter-city passenger turnover are shown in Figure 6 and the structure of each inter-city passenger transport mode is presented in Figure 7. Despite a large expansion in the aviation subsector in Figure 6, the rail subsector has replaced part of the demand. Compared with our forecast, the optimisation results for train turnover are higher. The optimisation result is 634 billion pkm higher in 2030 and 1894 billion pkm higher in 2060 than the prediction. Meanwhile, no new rail track construction is required at the resolution of our case study. Among the infrastructures considered, only airports in the eastern region of China are supposed to be expanded by 38% by 2060. These results suggest that the inter-city transport infrastructure of China is relatively advanced. A huge sum of money for building new infrastructure can be saved if existing infrastructures are reasonably utilised.

Similar to inner-city passenger transport, the use of fossil energy will not decline immediately. Although the use of diesel in the road subsector decreases significantly between 2020 and 2025, the use of natural gas increases largely at the same time. Gas-fuelled road transport reaches a peak at 289 billion pkm in 2024. The reason for the decline of electricity-powered road passenger vehicles is similar to the reason for that of electric taxis and buses. However, hydrogen seems to be competitive in coaches. Too many electric vehicles will lead to a surge in charging piles, increasing the total cost significantly. Among all hydrogen-powered road vehicles from passenger and freight transport, the cost per unit turnover of coaches is the lowest, making hydrogen more competitive than electricity in coaches. Contrary to expectations, diesel train turnover increases until 2040. However, this has little impact on CO₂ emissions, owing to the low carbon density of railway passenger transport. CO₂ emissions contributed by diesel-fuelled trains reach a peak at 5.1 million tonnes in 2039, which is only 0.5% of the total emissions of the transport sector for that year. Compared to aviation freight transport, air passenger transport has to eliminate jet-fuel-powered planes in 2060. There is no need to transform the fuel used in shipping passenger transport, for the reason that shipping passenger turnover is too low to be of concern. It is calculated that the highest shipping passenger turnover will not exceed 5 million pkm.

4.3. Optimal Structure of Private Cars

Figure 8 provides the optimisation results for the number of private cars. The number of gasoline-fuelled private cars increases until 2047 and then falls steeply, meaning that electric private cars are not fully economic competitive presently. What is more, it can be found that there is no need to heavily promote hybrid electric vehicles (HEVs), which is consistent with reality. Compared with electric cars, HEVs are only 10% cheaper and generate nearly 0.6 tonnes more CO₂ emissions per year. HEVs are only technological transitional products from gasoline cars to electric cars. After 2047, the number of private electric vehicles improves speedily, with 103 million in 2050 and 542 million in 2060, while gasoline-fuelled private cars are not allowed to be purchased.

4.4. CO₂ Emissions from the Transport Sector

Carbon neutrality is the major scenario in our case study. The only difference between the scenario of BAU and carbon neutrality is that there are no emission constraints in the BAU scenario. CO₂ emissions from the transport sector in China are supposed to reach a peak in 2030 and decrease to 100 million tonnes in 2060, as shown in Figure 9. The results show that although the emission peak year of the transport sector is 2030, CO₂ emissions from freight transport must be reduced from now on. The rate of decline for CO₂ emissions from the freight sector is even from 2020 to 2035 and from 2045 to 2060 and a rapid drop can be observed between 2035 and 2045, thanks to the fossil-fuelled freight vehicles being decommissioned on a large scale. Owing to the cost advantage and the lower CO₂ emissions of natural gas as compared to diesel, gas-fuelled trucks are preferred in the initial stage of optimisation. These newly added gas-fuelled trucks are out of service ten years later, leading to a large supply gap for road freight transport. Facing the challenge of decarbonisation, policymakers have to choose secondary energy-powered vehicles directly. The proportion of hydrogen-powered road transport in Figure 3a increases from 13% to 31% between 2035 and 2045, which can confirm our statement. However, different from the freight transport sector, CO₂ emissions from passenger transport increase first and then decrease. They reach a peak in 2041 at 552 million tonnes, the level falling to 41 million tonnes in 2060, while CO₂ emissions from the freight transport sector are 59 million tonnes in 2060. Notwithstanding CO₂ emissions from the passenger transport sector growing from 453 million tonnes to 540 million tonnes from 2030 to 2045, the total CO₂ emissions from the transport sector still decline by 309 million tonnes due to the drop in CO₂ emissions from the freight sector. Meanwhile, from Figure 9b, two major transitions in the CO₂ emission structure of the transport sector in China can be noticed. The major source of CO₂ emissions will gradually change to natural gas before 2035. With the reduction in natural gas consumption after 2035, CO₂ emissions will be dominated by oil again. What is more, the small increase in CO₂ emissions caused by oil between 2030 and 2040 is due to the increase in oil consumption for passenger transport, especially private cars, which can be verified in Figure 10. However, due to the decrease in the use of natural gas, this has not affected the downward trend for total CO₂ emissions.

End-use energy consumption in the transport sector in China is presented in Figure 10. Here, we use the electrothermal equivalent method to calculate electricity consumption. It can be seen that there is also a peak in energy consumption, although traffic demand continues to increase. Improved fuel efficiency and modal shift in passenger transport contributes to the decrease in end-use energy consumption. The peak of end-use energy consumption appears in 2037 with 667 million tce. In 2060, end-use energy consumption in the transport sector in China is only 442 million tce, lower than that in 2020. What is more, Figure 10 shows the significant transitional role of natural gas in the decarbonisation of the transport sector in China. The largest amount of natural gas will be used in 2030, reaching 233 billion m³, which is nearly six times that in 2019. However, only 7 billion m³ of gas is needed in 2060, which is much lower than the figure for 2019. Moreover, the extensive application of bio-jet-fuel is also noteworthy. Only a handful of routes in China use biomass kerosene, while 91% of the energy consumed by the aviation sector consists of bio-jet-fuel in 2060, based on our optimistic estimate of the cost of bio-jet-fuel. Additionally, hydrogen and electricity use in the transport sector will increase rapidly. Electricity becomes the major energy source for the transport system from 2051 onwards in China; this might be because hydrogen has not been competitive in price compared with electricity. We consider this conclusion to be realistic, for the reason that we regard all hydrogen as being produced by renewable electricity. Power demand and hydrogen demand for the transportation sector are shown in

Table 3. Power and hydrogen demand in the transport sector in China.

	Power Demand /Billion kWh	Hydrogen Demand /Billion kg	Equivalent Power Demand /Billion kWh
2030	410	6	682
2040	718	16	1361
2050	1101	24	1996
2060	1605	22	2391

. If hydrogen is produced by electricity totally and we assume that 45 kWh of electricity can develop a kilogram of hydrogen in 2020 and that 35 kWh of electricity can develop a kilogram of hydrogen in 2060, then 2391 billion kWh is needed for the transport sector. Due to the decline in the cost of charging piles after 2045, electric vehicles have greater cost advantages in the road subsector and occupy part of the original hydrogen-powered vehicle market. Consequently, the demand for hydrogen in the transport sector decreases slightly. In previous work in our group, over 25 trillion kWh of electricity was reckoned to be needed in China in 2060 when all hydrogen is generated by power [50]. Therefore, our results imply that the electricity that the transport sector needs in 2060 is nearly 10% of that needed by the whole of society, which is four times higher than the percentage in 2019. Decarbonisation in the transport sector will put unprecedented pressure on the power system.

4.5. The Costs of Decarbonisation for the Transport Sector

Figure 11 presents the results of the costs of decarbonisation in the transport sector in China. It can be found that the costs of the transport sector are rising, while the proportion goes down consistently. The cost of decarbonisation is about CNY 29.2 trillion in China, nearly three times the cost in 2020, while China only spends CNY 12.7 trillion to merely satisfy the demands of the transport sector. It can also be noticed that there is no big difference in costs between the two scenarios before 2035, meaning that the method of expansion selected in the first 15 years is particularly critical. Choosing an appropriate means of expansion can lay a solid foundation for further decarbonisation while saving costs. Meanwhile, three sharp increases in costs in the scenario of carbon neutrality can be observed in Figure 11. The reasons for these are diverse and they can be explained by reference to

Table 4. Proportion of different costs.

	Private Cars	Freight Vehicles	Passenger Vehicles (Public)	Fuel	Infrastructure
2025	52%	10%	11%	24%	3%
2030	45%	20%	10%	23%	2%
2035	48%	14%	11%	25%	2%
2040	41%	24%	9%	23%	3%
2045	39%	19%	13%	26%	3%
2050	37%	20%	8%	19%	16%
2055	34%	20%	11%	17%	18%
2060	49%	16%	9%	12%	14%

. The first rise between 2035 and 2040 is caused by freight vehicles. Due to the retirement of gas-fuelled trucks in significant quantities and the increase in the number of hydrogen-powered trucks, freight cost increased sharply. The second rise between 2045 and 2050 is caused by infrastructure. More charging piles are needed to support the large number of new electric vehicles in the road subsector. The third rise between 2055 and 2060 is caused by private cars. The cost of private cars consists of two parts in this period. One part is the increased demand for vehicles, and the other part is the demand for vehicles needed as replacements after the retirement of gasoline and electric vehicles, which explains the discrepancy in the costs of private cars before 2055.

4.6. Sensitivity Analysis

4.6.1. The Necessity of Considering Infrastructure Construction when Modelling

In order to illustrate the importance of considering transport facilities when modelling, two scenarios are compared in Figure 12. The major dissimilarities between the two scenarios concern planning for metros and electric road vehicles. Without considering metro tracks, the development of a subway will be overestimated. The difference between the two scenarios is 47 trillion pkm in 2060, implying that ignoring infrastructure construction leads to 60% more subways being planned. What is more, neglecting facilities also leads to more charging piles in reality. If infrastructure is not considered in the planning period, the number of charging piles required each year will be 45% higher than that on average when considering the facilities in the planning period. Above all, the cost of the whole transport sector each year is 7.3% higher when not considering infrastructure construction on average.

4.6.2. Impact of the Proportion of Specific Requirements in Passenger Transport on the Results

In Section 3.7, $FIXPR{O}_{taxi}$, $FIXPR{O}_{coach}$ and $FIXPR{O}_{plane}$ were set to be 0.6, due to their characteristics of being either convenient or fast, while other values were set to 0.4. In this section, diverse values are set to explore the impact of values on annual costs. The results are presented in Figure 13. It can be found that with increases in variable demand, annual cost generally decreases. If there is no variable demand in all passenger transport modes, the annual cost can be 7.5% lower on average, compared with no traffic mode conversion. Even if the transport sector does not carry out decarbonisation, 3.1% of costs can be saved every year if the transport modes can shift freely, compared with the scenario in which modes cannot shift. The sensitivity analysis for the parameter $FIXPR{O}_{ser}$ is of practical significance. The modification of people’s travelling concepts is beneficial for reducing the costs of the transport sector.

4.6.3. The Necessity of Giving Priority to Decarbonisation in the Freight Sector

The basic scenario discussed in Section 4.1, Section 4.2, Section 4.3, Section 4.4 and Section 4.5 indicates that giving priority to the decarbonisation of freight transport is economically efficient. However, the reality is that passenger transport and freight transport reduce CO₂ emissions simultaneously. In other words, the passenger transport also has decarbonisation requirements each year in China. In this section, the basic scenario is named the freight first scenario, while the other scenario with extra constraints on the passenger sector is regarded as a passenger first scenario. In the passenger first scenario, we simulated a policy that requires the number of electric private cars to increase by 10% every year and the passenger turnover undertaken by other electric vehicles to increase by 5% every year before 2045. At the same time, the passenger turnover of diesel-fuelled trains cannot increase. The annual cost of the transport sector under two different policies is presented in Figure 14. It must be admitted that the costs can be lower in the later stage of the planning period if the passenger first policy is implemented. However, compared with the freight first scenario, the passenger first scenario pays an additional 5% of the cost per year on average.

5. Policy Recommendations

Four policy recommendations can be summarised based on the analysis from Section 4.

Firstly, make full use of the existing inter-regional transport infrastructure rather than undertake large-scale expansion. Over the past two decades, China’s transport infrastructure has developed rapidly, especially rail-related infrastructure.By making full use of existing infrastructure, development costs can be saved and carbon emissions caused by massive infrastructure construction can be reduced.

Secondly, priority should be given to reducing CO₂ emissions from the freight sector rather than from the passenger sector. The freight sector is currently the largest source of carbon emissions in China’s transport system. If its CO₂ emissions are not reduced immediately, a greater cost will be paid later. The immediate implementation of emission reduction measures in the freight sector is also in line with the current requirements of the government.

Thirdly, guide passengers to choose more travel modes with low CO₂ emission density through propaganda or other means. Different transport modes have different marginal costs of CO₂ emission reduction. Choosing less carbon-intensive modes to travel could reduce the emission-reduction costs and take full advantage of existing railway infrastructure.

Fourthly, be alert to the rapid growth in power demand caused by the direct emissions reduction of the transport sector, which is a matter for policymakers in the power sector. The reduction in direct carbon emissions in the transport sector relies almost solely on secondary energy, such as electricity, which will lead to a huge increase in electricity consumption in the transport sector. Therefore, power sector planners need to take into account the substantial increase in power consumption in the terminal sector under the carbon-neutrality target.

6. Conclusions

In this study, a long-term multi-regional model is proposed to plan the transport sector, including both freight and passenger transport, which considers modal shift and infrastructure construction. This model can be used in the transport sector, where demand will increase alarmingly during the planning period. China is taken as a case study. Detailed development paths, energy consumption, CO₂ emissions, the composition of cost and sensitivity analysis have been discussed. The main conclusions are as follows:

Firstly, inter-regional infrastructure construction in China is ahead of schedule under the current spatial resolution. Only airports in East China have to be expanded by 2040. By 2060, the number of airport runways should be 138% of the current number. There is no expansion in other infrastructure considered in this paper. Nonetheless, infrastructure such as subway tracks and charging piles is far from sufficient. Ignoring infrastructure construction when modelling leads to an increase of over 7% in costs.

Secondly, oil consumption shows a downwards trend since 2021. The consumption of oil is only 0.4 trillion tce. Natural gas plays a transitional role in the transport sector, reaching a peak in 2030 and falling to 7 bcm in 2060. Electricity and hydrogen will be the chief fuels in the transport sector in the future. The promotion of electric vehicles is influenced by the price of charging piles. Electric cars will replace some hydrogen-powered cars when charging piles are cheap enough. Direct decarbonisation of the transport sector in China depends mainly on the utility of electricity and hydrogen, leading to an increase in power demand. In 2030, it is expected to use 0.4 trillion kWh and 6 billion kg of hydrogen. By 2060, the transport sector is expected to consume 2.4 trillion kWh of electricity and 22 billion kg of hydrogen.

Thirdly, CO₂ emissions from freight transport have now peaked and will fall to 59 million tonnes in 2060. Shipping contributes the most in 2060. The reduction in CO₂ emissions from freight transport occurs relatively quickly from 2035 to 2045, with an average annual decrease of 33 million tonnes. CO₂ emissions from passenger transport reach a peak of 552 million tonnes in 2041 and fall to 41 million tonnes by 2060.

Fourthly, the costs of decarbonisation of the transport sector rise from CNY 9 trillion per year to CNY 29 trillion per year in 2060. Increasing the changeable demand of passengers appropriately can reduce the cost effectively by up to 7%. More importantly, prioritising decarbonisation of the freight sector can also reduce costs by about 5%.

In future research, we suggest more elaborate spatial scales. This paper mainly focuses on inter-regional infrastructure construction due to the fact that in this study it is difficult to identify the distances between cities in one region. However, as urbanisation progresses, the traffic volume between cities in one region will further increase. When the inter-city infrastructure in one region reaches saturation point, the choice of future traffic modes needs to be carefully discussed.