Impact of vehicle electrification on global supply chains and emission transfer

Numerous countries plan to promote increased levels of vehicle electrification. This study demonstrates that, when considering the life cycle of automobiles, higher levels of vehicle electrification over the next 30 years in Japan would generate higher carbon emissions, preventing the country from meeting carbon reduction and neutrality targets in 2030 and 2050, respectively. In 2020, 2030, and 2050, domestic emissions could decrease to 92.5, 72.9, and 49.6 Mt, respectively, while emission transfers to other countries could reach 12.1 Mt (11.5% of the global carbon footprint), 10.4 Mt (12.5%), and 8.7 Mt (14.9%). The results indicate that even if the widespread use of alternative fuel vehicles could reduce domestic emissions, such emissions might be transferred to foreign countries, and blanket promotion of automobile electrification should be avoided. Instead, governments and the automotive industry should take responsibility for global and whole life-cycle emissions of vehicles, not only domestic tank-to-wheel emissions. These results provide baseline information for use in the recommendation measures and international rules to be adopted by the automobile industry stakeholders and policymakers.


Introduction
In 2021, the global battery electric vehicle (BEV) fleet was 16.5 million vehicles, three times that in 2018 (International Energy Agency: IEA 2022a). Subsidy policies for consumers purchasing EVs in various countries are leading to significant impacts on such expansion (e.g. Europe: Münzel et al 2019; Canada: Azarafshar and Vermeulen 2020; China: Sheldon and Dua 2020; U.S. : Xing et al 2021). Some countries have announced, or are considering, a ban on the sale of internal-combustion engine vehicles (ICEVs) 4 over the next 15 years (i.e. Norway in 2025; UK in 2030; Japan in 2035) (IEA 2022b). Policymakers in these countries believe that the electrification of vehicles contributes to the reduction of CO 2 emissions. However, in countries that are highly dependent on thermal power generation, the diffusion of alternative fuel vehicles (AFVs), including BEVs, does not contribute to emission reductions or reduces emissions at a very high cost (e.g. Japan The gap between national policies and existing literature regarding emission reductions through vehicle electrification should be addressed because policymakers only consider emissions during the driving phase of a vehicle (particularly tank-towheel). Since AFVs are powered by electricity or hydrogen, emissions during the tank-to-wheel phase are very low (IEA 2020a). However, AFVs use batteries for driving, which compared with ICEVs, require more energy and material input during their manufacture, in addition to during the production of electricity and hydrogen needed for driving (well-totank). In particular, emissions from the well-to-tank phase of a vehicle strongly depend on the energy mix of a country. Therefore, the life-cycle CO 2 emissions (carbon footprint) of AFVs vary widely depending on the country in which the vehicle operates (Knobloch et al 2020). Therefore, policymakers should consider emissions throughout the life cycle of a vehicle, and not just during its driving phase, when discussing vehicle electrification.
The high functionality and electrification of vehicles have led to increased energy and material inputs to vehicles (Los et al 2015, Tokito 2018. Furthermore, because the automotive supply chain is globally distributed, supply chain management at a global scale is critical for the sustainability of the automotive industry (Kagawa et al 2015). Therefore, environmental measures in the automotive industry must address the management of green supply chains at the global scale, in addition to improvements to traditional fuel efficiency (Zhu et al 2007). Examples of green supply chains include eco-design through procurement from environmentally advanced suppliers and energy conservation at business sites. To be able to delineate effective emission reduction measures, we must identify how vehicle electrification impacts the global supply chain (GSC); however, existing lifecycle studies of AFVs have not considered this issue (Knobloch et al 2020, Kito et al 2022. Here, based on the environmental input-output life-cycle assessment (LCA) (Joshi 1999, Wiedmann andMinx 2008), we estimated the carbon footprint induced directly and indirectly through the life cycles of ICEVs and AFVs. The global multiregional inputoutput (MRIO) model was used to estimate the carbon footprint at a global scale (Stadler et al 2018). As a case study, we focused on Japan, which has a relatively high fuel efficiency for ICEVs and is highly dependent on thermal power generation (70.5% of electricity generated in 2020; table S1) (IEA 2020b). The main objectives of the present study were to (1) analyse how changes to the penetration level of AFVs and energy mix in Japan impact the carbon footprint of automobiles, (2) assess whether the emission reduction target for 2030 (−46% emission reduction compared with that in 2013) and carbon neutrality in 2050 could be met, and (3) estimate domestic emissions and emissions transferred to other countries from 2020 to 2050 to forecast the structure of the future supply chain.
Decarbonisation of the power generation sector is crucial for reducing emissions effectively through vehicle electrification (Knobloch et al 2020, Kito et al 2022. The analytical model used in the present study was similar to that used in the studies by Knobloch et al (2020) and Kito et al (2022). Similar to the aims of previous studies, herein, we hypothesised that the electrification of vehicles could result in higher emissions in the countries that have not yet decarbonised their power sector. This study contributes to the existing literature by (1) extending the carbon footprint analysis of electric vehicles to the global scale and (2) examining how a country's transition from ICEVs to AFVs impacts the GSC of automobiles and their emission transfer to other countries.

Methods
Based on the top-down environmental input-output LCA (Hendrickson et al 1998, Hawkins et al 2007, we calculated the carbon footprint of passenger cars registered in Japan from 1995 to 2020, and then estimated it for 2020-2050. We assumed that AFVs were hybrid electric vehicle (HEV), Plug-in HEV (PHEV), BEV, and fuel cell electric vehicle (FCEV). Following Washizu and Nakano (2022), we assumed that the producer's price for an FCEV would be similar to that for a BEV. The system boundaries of the present work had five components: (1) production of goods and services associated with the manufacturing of new cars, including resource mining and parts manufacturing; (2) refinement of gasoline and/or power generation consumed by in-use cars (well-to-tank); (3) burning of gasoline by ICEV, HEV, and PHEV in use (tank-to-wheel); (4) battery replacement for AFVs; and (5) car disposal (figure S1). Details and limitations of the methodology, equations, and data sources are provided in the supplementary materials.

Product lifetime
Product lifetime is a key parameter influencing the stock and flow of products (Babbitt et al 2009, Miller et al 2016. Using a lifetime function, we implemented a dynamic LCA of the product (Kagawa et al 2011, Kito 2021) that considered its aging. In the present study, following Oguchi and Fuse (2015), we assumed that the cumulative scrappage rate of cars followed the Weibull distribution function (see equations (1)-(4) and table S2 in supplementary materials).

Dynamic stock-flow model
The physical lifetime of a product describes its survival from production to final disposal . We assumed that all ICEVs and AFVs registered during the study period  had the same physical lifetime, irrespective of the year of manufacture. Based on the product lifetime function and dynamic stock-flow model (Nakamoto et al 2019, Nakamoto 2020, we calculated the number of new car sales, in-use cars, and disposed cars for 1995-2020, and then estimated it for 2020-2050 (see equations (5) and (6) in the supplementary materials).
Annual energy (gasoline, electricity, and hydrogen) consumption of cars was estimated by multiplying the annual driving distance by the average fuel economy (table S3). Based on existing data (Ministry of Land, Infrastructure, Transport and Tourism 2021), we assumed that all cars would have the same annual travel distance of 10 000 km. Supporting the IEA (2020a), PHEV fuel economies with 40% total mileage were gasoline-based and 60% electricitybased. Direct CO 2 emissions of a car during the driving phase (tank-to-wheel) were calculated by multiplying annual gasoline consumption by direct CO 2 emission intensity (i.e. direct CO 2 generated per unit of gasoline combustion on the road) (see equations (7) and (8) and table S4 in supplementary materials).

MRIO model
Environmentally extended input-output analysis (Leontief 1970) is an application of input-output analysis to environmental impact assessment. Additionally, many studies on GHG emissions use MRIO analysis, which considers the GSC (e.g. Lenzen 2018, Wood et al 2019). Using the MRIO analysis, we calculated global CO 2 emissions that are directly and indirectly induced by the global final demand of cars and auto-related products and services, and visualised emission transfers from the materials input into Japanese automobiles to each emitter country through the GSC (see equation (9) in supplementary materials). The MRIO database used herein, EXIOBASE 3.8, contains domestic and international monetary transactions for 200 products across 49 countries and regions (Stadler et al 2021).
The input requirements of cars and auto-related energy, products, and services were estimated by multiplying the inputs to car manufacturing (inventory data of ICEV and AFVs) (Washizu and Nakano 2022) with the number of new car sales. The input requirements associated with new car sales could be obtained as values (tables S5 and S6). The inputs to the well-totank and battery replacement phases were set based on fuel consumption, energy mix, energy cost, and battery cost. We assumed that the replacement of AFV batteries would be conducted every 8 years after a new car is registered. By multiplying the disposal cost of an ICEV and an AFV with the number of disposed cars, we obtained the inputs to the scrapping phase of endof-life cars (see equations (10)-(12) in supplementary materials).

Scenario analysis
Scenario analyses were conducted to estimate how various environmental changes and uncertainties related to the electrification of automobiles impacted the carbon footprint of automobiles. These scenarios included (1) three scenarios at the proportion of the new vehicle sales of AFVs, (2) four scenarios for energy mix, and (3) three scenarios for technological innovation. The scenarios were implemented either independently or in combination.

New AFV sales
A scenario analysis to estimate how changes to the proportion of AFVs in new vehicle sales impacted the carbon footprint of automobiles in Japan was conducted using three scenarios (table S7). Based on the IEA (2021a), the proportion of ICEVs and AFVs in new vehicle sales each year from 2020 to 2050 was assumed to change as follows: Business as usual (BAU): vehicle electrification would continue at a low level, because the proportion of new vehicle sales of AFVs follows the government's target (IEA 2021a).
Widespread: the proportion of AFVs in new vehicle sales would be higher than the government target, and the electrification of automobiles would progress at a high level.
2035 ICEV ban: the proportion of AFVs in new vehicle sales would be considerably higher than the government target, and the proportion of ICEVs in new vehicle sales after 2035 would be zero, indicating a high level of vehicle electrification.

Energy mix
Based on the IEA (2021b), the four scenarios investigated for the energy mix in Japan were (1) stated policies scenario (STEPS), (2) announced pledges scenario (APS), (3) sustainable development scenario, and (4) Net Zero Emissions by 2050 scenario (NZE). We set the baseline energy mix for Japan over the study period to STEPS. Electricity was assumed to be directly supplied during the vehicle manufacturing and the tank-to-wheel (e.g. charging of electric vehicles) phases (see equations (13) and (14), and table S1 in supplementary materials).

Technological innovation
Four scenarios were established to represent expected technological innovations that accompany vehicle electrification: Improved emission intensity: emission intensity for all countries and regions would improve by 30% for the #120: electrical machinery and apparatus sector (especially vehicle-mounted storage batteries), accounting for a large proportion of the raw materials used to construct AFVs. This would be achieved by reducing the environmental impact of battery manufacture and replacement (Majeau-Bettez et al 2011) through creating a green supply chain and battery innovation (e.g. development of alternative batteries).
More efficient battery design (EBD): most automakers warranty the repair and replacement of parts if battery capacity falls below a certain level within 8 years or 160 000 km from the registration of a new car, whichever comes first (IEA 2020a). In such a scenario, the battery replacement cycle for AFVs would be extended from 8 to 16 years through technological innovation and improved battery durability.
Improved fuel economy: following the Japan Automobile Manufacturers Association (2022a) guidelines, improvement to the fuel economy of ICEV, HEV, and PHEV would be maintained until 2030, 2035, and 2050, respectively. FCEV would maintain its current fuel economy until 2050 (table  S3).
Vehicle lifetime extension: the average physical lifetime of all vehicle types would be improved by 30% through improved durability, promotion of used vehicles, changes to consumer behaviour, and expansion of the repair market (Kagawa et al 2013).

Carbon footprint of an automobile
In Japan, the carbon footprint of a BEV is larger than that of an ICEV under the current energy mix (figure 1). Compared with that for 2020, in the energy mix for 2035, emissions for the manufacturing and well-to-tank phases (generation of electricity to drive PHEVs and BEVs) decreased through the decarbonisation of the power generation sector (table S1). The carbon footprint of ICEVs in the energy mix for 2020 was 22.6 t, of which 84.6% was emitted during the driving phase (well-to-tank: 3.7 t; tank-to-wheel: 15.4 t). Additionally, the carbon footprint of HEVs was the lowest for the energy mix in 2020. Compared with the IEA (2020), Japan has a relatively low average annual vehicle mileage (Kawamoto et al) and high fuel efficiency of vehicles (Knobloch et al 2020), resulting in low emissions during driving.
In comparison, the carbon footprint of BEVs in the energy mix for 2020 was 29.9 t, including manufacturing (9.6 t; 32.0% of the total), well-totank (16.3 t, 54.3%) and battery replacement (4.1 t, 13.7%). As with PHEVs, the carbon footprint of BEVs noticeably declined through changes to the energy mix. With higher decarbonisation levels in the power generation sector, the carbon footprint of PHEVs and BEVs could be reduced to 14.3 and 15.9 tons, respectively, by 2035. While reduction is smaller than that of ICEVs and HEVs, the emissions from the manufacturing phase of BEVs were approximately two times higher than those of ICEVs. Even if the power generation sector was decarbonised, the reduction in the manufacturing phase remained minimal (table  S8). Therefore, different types of emission reduction measures are required to reduce the carbon footprint of BEVs.
FCEVs had the largest carbon footprint among all fuel types. Automobile companies must reduce the environmental impact of FCEVs through technological innovations, such as efficient vehicle design and the establishment and diffusion of green hydrogen production technology.

Annual carbon footprint of automobiles 2020-2050
In 2020, 60 million cars were in use in Japan. Compared with the actual data for the same year, the value in the current study is close (table S5). The percentage of each fuel type was 79.7%, 19.8%, 0.3%, 0.3%, and 0% for ICEVs, HEVs, PHEVs, BEVs, and FCEVs, respectively (figures 2(a)-(c)) (Japan Automobile Manufacturers Association 2022b). In 2050, the number of in-use cars was predicted to decrease by two-thirds of that in 2020 (e.g. to 41 million cars). The decline was attributed to a decline in the number of households and population (Kito et al 2022) (table S5). Notably, the number of cars in use each year under the two scenarios (widespread and 2035 ICEV ban) was consistent with that under the BAU scenario because it was assumed that the number of cars in use would not change, even if the sales ratio of ICEVs and AFVs changed.
Under the BAU scenario, the annual carbon footprint peaked at 105 Mt in 2020 and decreased to 58 Mt by 2050. In 2050, ICEVs and HEVs were estimated to account for 46.7% and 40.4% of the total fleet, respectively. Approximately half of emissions were predicted to be derived from the combustion of gasoline (tank-to-wheel) by ICEVs, HEVs, and PHEVs ( figure 2(g)).
The percentage of ICEVs in the fleet by 2050 was estimated to be 4.9%, with AFVs (mainly HEVs and BEVs) being prevalent ( figure 2(b)). The annual carbon footprint in 2050 was estimated to be 58 Mt, similar to that under the BAU scenario (figures 2(e) and (h)). Emission would likely decline owing to reduced gasoline combustion through vehicle electrification being offset by increased emissions from vehicle and battery manufacture of AFVs, and from the power generation sector.
Under the 2035 ICEV ban scenario, PHEVs and BEVs accounted for 27.4 and 64.0% of the number of in-use cars in 2050, respectively, achieving full vehicle electrification (figure 2(c)). The transitional period of vehicle electrification (around 2030) had a very large annual carbon footprint compared with that under BAU. Additionally, the annual carbon footprint in 2050 was 63 Mt, an increase of 9% compared with that under BAU (figures 2(f) and (i)). The result was attributed to a greater increase in emissions from the increased demand for vehicles, batteries, and electricity for AFVs than the reduction in emissions from reduced gasoline combustion due to the electrification of vehicles.
Under the three scenarios, the 2030 reduction target and 2050 carbon neutrality target were not achieved (figures 2(d)-(i)). In contrast, a high level of vehicle electrification would increase emissions relative to BAU, supporting Knobloch et al (2020) and Kito et al (2022). In Japan, the reduction in emissions during the driving phase of the electrification of vehicles is expected to be smaller due to the short

Global distribution of the carbon footprint associated with automobiles in Japan
To reduce the carbon footprint of PHEVs, BEVs, and FCEVs, which are expected to be widely distributed in the future, it is essential to reduce the environmental burden at the vehicle and battery manufacturing phases. It is also critical to decarbonise the power generation sector. Consequently, a different approach to conventional environmental measures  (table 1). However, emission transfers to other countries were 12. Mt (11.5% of the global carbon footprint) in 2020. Under the BAU scenario, emission transfers were 10. Mt (12.5%) and 8.7 Mt (14.9%) in 2030 and 2050, respectively. Under the widespread scenario, emission transfers were 11.5 Mt (13.5%) and 12.9 Mt (21.6%) in 2030   3  and table 1). Emission-inducing countries changed to other countries (especially China, Southeast Asia, Australia, the U.S., and South American countries) by 2050, with more emissions being induced in more countries as electrification progressed at a higher level. For example, under the 2035 ICEV ban scenario, Japan induced 6.9 and 8.5 Mt of emissions in China in 2030 and 2050, respectively, which corresponded to 6.7 and 13.1% of the global carbon footprint induced by automobiles in Japan for the same years, respectively.
Policymakers and the automotive industry should be aware that even if the widespread use of AFVs could reduce domestic emissions, such emissions might be transferred to foreign countries through the associated transformation of GSC, hindering the achievement of their emission targets. In this context, emission-inducing countries should reduce emissions of products that induce significant emissions in other countries, such as electronics and batteries, through a clean development mechanism and joint implementation program. Simultaneously, emissioninduced countries should reduce emissions of the such products through cleaner production. Figure 4 shows the emission transfer between the primary input of large emission-inducing products to automobiles in Japan (origins) and emission-induced countries and regions (destination). In 2020, automobiles in Japan were estimated to mainly induce emissions to WWM ( figure 3(a)). The main emission sources were identified as motor gasoline (#67), basic iron, steel, and ferro-alloys and first products thereof (#104), electrical machinery and apparatus (#120), and motor vehicles (#123). With progress in Figure 4. Transfer of emissions globally associated with automobiles from Japan under the three scenarios in 2020, 2030, and 2050. Note: Each circle represents the top 50 emission transfers in each year for each scenario. The length of the arc reflects the percentage that each sector (product or region) contributes to automobile manufacturing emissions. The width of the ribbon connecting the product and region reflects the emissions induced from the product origin to the destination region. The colour reflects the destination. Colours at each end of the ribbon reflect the origin and destination, respectively. For example, in (a) the colour of the ribbon connecting WWM (destination, purple) and sector 67 (origin, blue) is purple, while the colours of the destination and origin edges are blue and purple, respectively. the electrification of automobiles, GSC is undergoing a major transformation, and the presence of electronic components in East Asia (mainly in China and Taiwan) is increasing, instead of gasoline in oilproducing countries. Such a trend would be higher under the 2035 ICEV ban scenario, in which the penetration of AFVs would proceed at a high level.

Strategies for achieving effective decarbonisation through vehicle electrification
Although the electrification of automobiles does not contribute toward reducing emissions in the short term (rather, it causes emissions to increase), it could achieve noticeable emission reductions in the medium to long term (figure 5). When the penetration of AFVs was higher, the reduction effect through the greening of the power generation sector grew. Under the widespread scenario, the cumulative carbon footprint in 2050 was less than that of BAU because the energy mix shifted from STEPS to APS (figure 5(e)). However, under the 2035 ICEV ban scenario, the cumulative carbon footprint in 2050 was negligibly below that of the BAU, by achieving the NZE energy mix (figure 5(f)). Thus, higher levels of vehicle electrification must be accompanied with a higher level of decarbonisation in the power generation sector to reduce emissions.
The decarbonisation of the power generation sector (NZE) could result in an annual global carbon footprint of 29.7 Mt in 2050 if combined with green supply chain management (emission intensity improvement (EII)), EBD, fuel economy improvement (FEI), and vehicle lifetime extension (VLE) (figure 5(c)). This value was approximately half that estimated under BAU (58.0 Mt) in the same year. Thus, emissions could be noticeably reduced by conventional environmental measures focused on the tank-to-wheel phase (i.e. FEI) and comprehensive environmental measures covering the entire vehicle life cycle (manufacturing: green supply chain management, well-to-tank, decarbonisation of the power generation sector, battery replacement, battery innovation and durability improvement).

Policy application
While the demand for gasoline is expected to decrease due to the electrification of automobiles, the demand for products that are components of vehicles and batteries for AFVs is expected to increase significantly. To reduce emissions from automobiles effectively, . The sixth basic energy plan in Japan (Agency for Natural Resources and Energy 2021) sets the share of renewable energy in the energy mix in 2030 at 36%-38% under an ambitious outlook to thoroughly expand the use of renewable energy. This corresponds to STEPS, the lowest decarbonisation level of the energy scenario in this study. Accordingly, a very high level of decarbonisation of the power generation sector, as in the NZE energy mix, is unlikely to be feasible at least by 2030.
Thus, the blanket promotion of automobile electrification should be avoided. With an imbalanced energy mix and vehicle electrification, transition strategies that inherently achieve emission reduction will have unintended consequences. Each country should promote its own optimal electrification of vehicles, considering future energy policies, diffusion structure of AFVs, and the carbon footprint of vehicles.
The carbon footprint of PHEVs and BEVs could be reduced through reducing emissions in the wellto-tank phase by decarbonising the power generation sector. However, because this action alone does not contribute toward reducing emissions during the manufacturing phase of BEVs (table S8), strategies should target the supply chain to reduce emissions through vehicle electrification. The global carbon footprint analysis of automobiles based on the MRIO model in the current study showed that the electrification of automobiles could contribute toward reducing domestic emissions (especially in the tank-to-wheel phase). Therefore, it could meet national emission reduction targets but might also increase emissions induced in other countries (figures 3 and 4). The phenomenon would occur because the percentage of foreign-induced emissions in the global carbon footprint of a vehicle was estimated at 11 and 15% for ICEVs and HEVs, respectively, but was relatively large for PHEVs, BEVs, and FCEVs (20%) (table S8).
For governments and automotive industries in countries that induce high emissions throughout the life cycle of their vehicles, we propose a framework in which environmental measures in the automotive industry are implemented globally, and throughout the automobile life cycle. Specifically, governments and the automotive industry should take responsibility for global and whole life-cycle emissions of vehicles, not just domestic tank-to-wheel emissions. Furthermore, automakers should identify and disclose the life-cycle emissions of their vehicles. Therefore, the design of incentives for car companies to build a green supply chain (e.g. border carbon tax) would have a greater presence, in parallel to promoting decarbonisation in industries with high emission intensity (e.g. steel).

Conclusion
The study utilises a modelling framework that considers the entire life cycle of electric vehicles, rather than only one phase, and evaluated the consequences of the entire process, allowing us to provide sound policy recommendations at both national and international scales. We demonstrate that, when considering the life cycle of automobiles, higher levels of vehicle electrification over the next 30 years in Japan would generate higher carbon emissions, preventing the country from meeting carbon reduction and neutrality targets in 2030 and 2050, respectively, under various scenarios. Moreover, the electrification of vehicles would increase emissions induced in other countries, hindering the achievement of their netzero emission targets. Electrification is not necessarily effective at reducing emissions in Japan or other countries, and blanket promotion of automobile electrification should be avoided.
The present study provides several suggestions for the future directions of the dynamic life-cycle analysis of automobiles. The first is the development of an analytical model that considers rare metals (e.g. cobalt, lithium, and nickel) used in the batteries of electric vehicles. Because rare metals present clear geopolitical risks (Zhou et al 2022), stable resource supplies and the development of alternative materials are urgent issues (Tokito et al 2016, Sen et al 2019. An analytical framework that considers the risks and constraints to resource supply is required to enhance the feasibility of the diffusion structure and emission reduction of electric vehicles. Second, the electrification of automobiles has reduced dependence on unstable countries, such as the Middle East, but has increased induced emissions in most other countries. We recommend examining changes to the carbon footprint, and its uncertainty in the event of structural changes to GSCs (e.g. reshoring and friendshoring of production sites in the electrical machinery sector), due to unstable conditions, high resource prices, and supply constraints. Finally, we assume that the production structure related to the input requirements of automobiles and the energy mix of other countries remain constant during the study period, which is the limitation of the present study. Therefore, the reduction effect of electrification of automobiles is expected to become larger as a result of these improvements, and a more detailed analysis that considers these factors will be explored in future work.

Data availability statement
All data that support the findings of this study are included within the article (and any supplementary files).