Shared pooled mobility essential complement to decarbonize China’s transport sector until 2060

This work aims to develop a scenario-based assessment framework of the passenger car fleet, taking shared socioeconomic pathways (SSPs) as a starting point. Figure 1 illustrates the general modeling process and main exogenous inputs. The evaluation framework consists of 3 main modules: fleet estimation, CO₂ emission assessment, and scenario analysis. The fleet module simulates the annual car stocks. An extended Gompertz model is used to estimate future passenger car stock of 31 provinces in China (Dargay et al. 2007). The CO₂ emission model quantifies the direct and indirect emissions (i.e., vehicle manufacturing, fuel production, and fuel use) for EVs and ICEVs. Finally, three main scenarios are built to evaluate policy tool effects. On the basis of these results, the carbon emission of passenger cars in China from 2010 to 2100 is evaluated at the provincial level.

The shared socioeconomic pathways (SSPs) are used to quantify socioeconomic components, such as GDP, population, and urban share. The SSPs represent five global development pathways, describing the future evolution of key aspects of society that would together imply a range of challenges for mitigating and adapting to climate change (Riahi et al. 2017; Kasera et al. 2016). SSPs have been widely used in forecasting future energy demand and GHG emissions (Xing et al. 2015; Milovanoff et al. 2020; Wolfram et al. 2021). The five global development pathways are SSP1 (Sustainability), SSP2 (Middle of the Road), SSP3 (Regional Rivalry), SSP4 (Inequality), and SSP5 (Fossil-fueled Development). A brief description of SSP1-5 is shown in the supplementary information (Table S1). The SSP database only provides national-level economic forecast data. China’s provincial GDP data is from Jiang et al. (2018), the population data is from Jiang et al. (2017), and the urbanization rates data is from Chen et al. (2020). We briefly show the changes in population and urbanization rates under SSPs in the supplementary information (Table S2 and Fig S1). As mainly referenced, we use SSP1, which is regarded as the most sustainable path among all SSPs, to then demonstrate the differential impact of passenger cars and carbon emissions in three scenarios (BAU, EV, and SEV—see Fig. 1).

In the “fleet” module, we developed a demand-based extended Gompertz function to simulate car stocks retrospectively (1995–2019) and prospectively (2020– 2100) for 31 provinces in China. The historical data (such as passenger car ownership, car sales, GDP, and population) are from national statistics (National Bureau of Statistics of China 2020) and the China Automotive Technology And Research Center (CATARC 2020).

The traditional Gompertz function (Eq. (1)) is widely applied to measure the S-shaped relationship between vehicle ownership and per-capita income (Dargay and Gately 1999; Dargay et al. 2007; Zheng et al. 2015; Zeng et al. 2016; Ma et al. 2019). Compared to other more intricate models such as Dynamic Simulation (Levin et al. 2017) or Linear Programming (Jones and Leibowicz 2019; Zhang et al. 2022) which offer greater flexibility in handling diverse factors but with greater data and computational intensity requirements, the Gompertz function can simply facilitate a clearer understanding and provide vehicle ownership prediction trends based on economic conditions, proving valuable for policymakers and stakeholders who may not have an extensive technical background (Shen et al. 2023). Further, the Gompertz function allows for the adjustment of parameters to match the unique characteristics of each region.where \({V_{it}}\) is the vehicle ownership (per 1000 people) in province i in year t, \({GDP}_{it}\) is the GDP in province i in year t, \(\gamma\) > 0 is the saturation level, and \(\alpha <0\) and \(\beta <0\) are parameters defining the shape of the curve.

Figure 2 shows passenger car ownership per thousand people based on China’s national and provincial GDP per capita from 1995 to 2019. The passenger car ownership is around 109–250 per thousand people in 2019. The passenger car ownership in areas with good economic development such as Beijing and Zhejiang is significantly higher than that of other provinces such as Gansu and Jiangxi, major destinations for migrant workers in the west. For example, while the passenger car ownership in Zhejiang is 249 vehicles per thousand people, Gansu has the lowest car ownership rate, with only 109 vehicles per thousand people. Due to the strict vehicle purchase limitation policy, there is a flattening of the curve in Beijing, Tianjin, and Shanghai. However, most provinces in China are still in the initial stage of the “S shape,” which means that the passenger car stock will continue to grow. The differences between different provinces are obvious.

As acknowledged, the conventional Gompertz function is limited to capturing the long-term correlation between vehicle ownership and per-capita income, while lacking the capacity to depict the dynamics of new vehicle sales and retirements. We are actively addressing this concern in the formulation of our Gompertz model. Firstly, following Dargay et al. (2007), we built an extended Gompertz function that explicitly models the vehicle saturation level as a function of observable country characteristics: urbanization and population density. Although there are many other factors that influence the modal shift on the usage of cars such as travel patterns and driver characteristics, such factors are difficult to take into account as they would require more data that are mostly unavailable for all provinces. Then, we postulated a simple partial adjustment mechanism in the econometric model to quantify the short-term relationship between the current period’s vehicle count and that of the preceding period (Eq. (2)). In this way, we used demographic and economic data to reassess vehicle production and usage from the demand side. We verified the validity and reliability of the extended Gompertz function and related parameters by comparing predicted values and historical data in the supplementary information (Fig S2-S3).where adjustment parameters θ_R > 0, θ_F > 0, \(\alpha <0\), \(\lambda <0\), and \(\varphi <0\) are constrained to be the same for all provinces, while \({\beta _i}\) < 0 is allowed to vary across provinces. Based on the provincial historical data in China and iterative least squares regression coefficients, we obtain parameters as shown in the supplementary information (Table S3‐S4). \({V_{it}}\) is the vehicle ownership (in numbers per 1000 people) in province i in year t.We assume that the maximum saturation level (\(\gamma_{max}\)) will be that estimated for Beijing because other provinces are less urbanized and less densely populated than Beijing. Under the government policy of restricting car purchases, research generally estimated that the saturation level in cities such as Beijing, Shanghai, or Tianjin will be approximately 250–400 (Zeng et al. 2016; Zheng et al. 2015; Gan et al. 2020; Huo and Wang 2012). In this study, we assumed \(\gamma_{max}\) to be 400 vehicles per 1000 people in Beijing. Furthermore, the saturation level in other provinces can be calculated according to the difference in the population density and the level of urbanization from Beijing (Eq. (3)where \(\overline {{{D_{it}}}}\) is the population density and \(\overline {{{U_{it}}}}\) is the level of urbanization.

Income asymmetry (i.e., many provinces have experienced negative as well as positive GDP per-capita growth over the period) is taken into consideration in the extended Gompertz function in Eq. (2) as \({\theta _R}{R_{it}}+{\theta _F}{F_{it}}\) (Dargay et al. 2007) where θ = θ_R for income increases and θ = θ_F for income decreases. R_it and F_it are shown in Eq. (4):

Three main scenarios for passenger car fleet are developed to evaluate the impact of two policy tools (i.e., shift to EVs and reduce car stock with shared mobility) on decarbonizing the passenger transport sector of China. Table 1 presents their descriptions. Specific details about the scenarios are shown in the supplementary information (section SI4).To ensure the robustness of the scenarios, we carry out a sensitivity analysis (from 2030 to 2050) because the prohibition time of fuel vehicles has a high degree of uncertainty in various provinces.

On-road CO₂ emissions from EVs and ICEVs are determined by car stock, traveled distance, and carbon emission factors. The on-road CO₂ emissions can be calculated by Eq. (5).where \(C{O_2}_{usage}\) is the on-road CO₂ emissions for passenger cars. i stands for province i. j stands for vehicle type j (\(j=\left\{ICEV, EV\right\}\)). VKT represents the vehicle kilometers (km) traveled. VKT is an important parameter reflecting vehicle usage. The VKT of provinces in China in 2015 are from Liu et al. (2017). Due to the unavailability of provincial-level VKT time series data, we use the national average growth rate of VKT to calculate the future VKT of passenger vehicles in each province. According to Huo et al. 2012; iCET 2018; and Zhang 2024, the annual VKT of private passenger vehicles in China has decreased from 12,489 to 11,660 km between 2015 and 2020 with an average annual decrease rate of 0.93%. We further estimate the VKT of ICEVs and EVs at the provincial level based on the provincial vehicle ownership level, as shown in the supplementary information (Fig. S4). Importantly, the effect of shared mobility on VKT is complex, which depends on many factors such as the substitution of public versus private transport modes and the occupancy rate of substituted modes. Limited researches analyze the changes in VKT for shared mobility (Narayanan et al. 2020). We have compiled some of the related studies and displayed them in the supplementary information (Table S5). Shared mobility fleets are predicted to generate an additional 10~20% of travel miles (VKT/VMT). To account for this uncertainty, we set two different growth rates (10% and 20%) in the VKT of shared mobility (Fig. S5).

EMF represents the emission factors (g CO₂/km). While driving, EVs emit CO₂ emissions from electricity generation and ICEVs emit CO₂ emissions from fuel production and fuel consumption. The detailed information on the calculation of on-road CO₂ emissions is summarized in the supplementary information (section SI5).

The specific greenhouse gas (GHG) emission reduction potential of electric vehicles not only depends on the carbon intensity of electricity usage, but also on the production process of vehicles. In determining vehicle emissions, it is important to consider not only direct emissions during vehicle use, but also implicit emissions resulting from vehicle production. Vehicles generate massive indirect emissions during the manufacturing stage in the industry sector (Hill et al. 2019; Milovanoff et al. 2021). The exact GHG emissions from the production phase of an EV are about 13.0 t CO₂eq, 24% larger than those of an ICEV (10.5 t CO₂eq) in China in 2015 (Qiao et al. 2019). This part of the emissions may limit the emission reduction potential of EVs. The CO₂ emissions from production will be given by the number of new vehicles produced as shown in Eq. (6).where \(C{O_2}_{production}\) is the CO₂ emission of passenger vehicle production in province i and vehicle type j. VP represents new vehicles produced. The detailed information on the calculation of vehicle production CO₂ emissions is summarized in the supplementary information (section SI 5.7).

Total CO₂ emission incorporates production and usage emissions as shown in Eq. (7).

There will be a significant increase in the demand for passenger vehicles in all provinces over the next few decades. The model results show two main characteristics of future changes in passenger cars in China. First, there are strong regional differences. Figure 8 shows the change of provincial passenger vehicle stock in the BAU scenario under SSP1. The numbers of passenger cars in the southeast coastal provinces are significantly higher than those of other regions, especially in Shandong and Guangdong, where the numbers of passenger cars have exceeded 20 million in 2020. Further, vehicle ownership in various provinces continued to rise in the BAU scenario. By 2060, there will be 9 provinces with more than 20 million passenger cars (i.e., Jiangsu, Zhejiang, Shandong, Henan, Hebei, Guangdong, Sichuan, Anhui, and Hunan), most of which will be located in the southeast coastal areas. The demand for passenger cars in these regions is based on larger population and higher economic turnover. Special attention should be paid to provinces such as Guangdong, Henan, and Hunan. These provinces not only have large stocks, but also display strong dynamic increase.

However, although China is the largest EV market in the world, the EV industry in the provinces has just started. The CO₂ emissions from passenger cars, EV stocks, and EV shares in 31 provinces are shown in Fig. 9. In provinces where vehicle ownership continues to grow, EV development is still very slow at this stage. Guangdong and Shandong are the top 2 highest emission releasers in passenger cars with more than 80 Mt CO₂ in 2020 (Fig. 9a). Although their EV holdings are also higher than other provinces (Fig. 9b), the share of EVs is only 1−2% of their total passenger cars (Fig. 9c). The EV scenario is far away from current trends. In the EV scenario, we assume that the EV sales shares in 31 provinces in China will reach 100% in 2030, which is the most optimistic policy scenario. According to the current EV market penetration, electric vehicles need to be heavily promoted in all provinces over the next ten years. The EV share of passenger cars in Shanghai, Beijing, Tianjin, and Hainan is more than 4%, while other provinces’ numbers are below 2% (Fig. 9c). These four provinces can be used as pilot provinces for banning the sale of ICEVs.

Figure 10 displays provinces’ passenger transport sector emission in 2030 and 2060 under three scenarios. From the perspective of passenger vehicle emissions, provinces differ greatly in their emissions. The south-eastern provinces, which still have a high demand for vehicles, have significantly higher emissions than other provinces. Guangdong will be the highest passenger car emission producer in China with 86 Mt CO₂ by 2030 under the BAU scenario, which is 39 times that of Tibet (2.2 Mt CO₂). Figure 10 clearly shows that EVs and shared mobility can effectively reduce emissions from passenger transport in various provinces. In Guangdong province, GHG emissions will fall to 55 Mt CO₂ by 2060 under the BAU scenario. Under the EV and SEV scenarios, it will further reduce emissions by 72% and 83%, respectively. Overall, south-eastern provinces will benefit more from the development of EVs and shared mobility. They have higher potential in reducing emission levels.

China is expected to achieve 2030 carbon peak from passenger cars under the BAU scenario with 1.3 Gt CO₂ in SSP1 with decarbonization (sustainable path). Large-scale shift with the EV promotion can advance passenger car emissions to peak around 2026. However, EVs will not provide a solution for near-term CO₂ mitigation because EVs cannot significantly reduce peak emissions.

Deep decarbonization to net-zero by 2060 in the passenger transport sector of China is easier with a comprehensive climate policy (including electrification, sharing mobility to reduce car use, and improving vehicle efficiency and fuel carbon intensity) instead of a single EV policy. The single EV policy (shift ICEVs to EVs) can reduce GHG emissions by 71% in 2060. There are still 302 Mt CO₂ of emissions that need to be resolved by carbon sinks or negative emissions technology. If electric vehicles are also shared, GHG emissions will be reduced by 83% in 2060 compared to the BAU scenario. The SEV scenario explores the extreme setting of a case when all private trips change to shared trips, a fairly optimistic assumption. In contrast, the scenarios assume conservative values in decarbonization of the electricity grid. One interpretation is that combined intermediate shared mobility and electrification strategies can achieve between 71 and 83% GHG emission reductions between 2020 and 2060 on their own (around 302 Mt CO₂ to 181 Mt CO₂ left by 2060). More aggressive phase out of coal and scale up of renewable energy could then support the decarbonization of the residual emissions. Importantly, because of reduced demand for transport sector electrification, additional energy demand can be more easily covered by renewable energy sources (Creutzig et al. 2021; p. 1).

Additionally, analysis shows that variation of SSP has little implication for mitigation scenarios (supplementary information, Fig. S12–S13). Specifically, variation in SSPs results in GHG emissions varying little in either the BAU or mitigation scenarios (maximally 15% in 2060). Introducing technological (EV) and mobility system change (SEV) is responsible for the impactful change in GHG emission trajectories.

How fast can transport electrification and shared mobility contribute to China’s “30·60” goals depends on their share in total passenger car market. Without limiting the number of vehicles, 500 million EVs would need to be on the road in China in 2050 to achieve max emission reduction in 2060, up to 94% of the total passenger cars, a considerable challenge. In addition, the emission reduction potential in the large-scale promotion of EVs is limited due to higher production-related emissions. Improving vehicle efficiency and carbon intensity in EV production should be the basis for the large-scale promotion of EVs. Also, the increase in power consumption caused by the development of EVs has an impact on the power system. This will also put substantial pressure on the power system. Additional electric vehicle growth could result in substantial material concerns. Milovanoff et al. (2020) estimated that 8% and 29% of the identified world terrestrial resources of lithium and cobalt, respectively, would be needed for the USA alone to achieve its climate goals by electrification. China has a larger passenger car system, meaning that replacing ICEVs with EVs will require more resource consumption. Therefore, the reduction in the passenger car ownership and usage is still critical to reduce their carbon emissions. The development of shared mobility can reduce the vehicle required for emission reduction, which will greatly reduce vehicle material demand.

Electric vehicles and shared mobility can effectively reduce carbon emissions from passenger transport in various provinces. In terms of regional differences, passenger car stock and carbon emission in the southeast coastal provinces is higher than those of other regions under all scenarios. At the same time, they also have higher potential in reducing emission levels. Currently, most provinces have a very low EV share (generally below 2%). The electric vehicle industry will still be in a stage of rapid development in the next decade. In addition, affected by the performance of EVs, it is difficult to use EVs in cold regions (Heilongjiang, Jilin, Liaoning) and western regions (Gansu, Qinghai, Ningxia). The provincial promotion in EVs will hence be an important research topic. The EV share of passenger cars in Shanghai, Beijing, Tianjin, and Hainan is higher. These four provinces can be used as first pilot provinces in banning the sale of ICEVs.