Introduction

The urgent need for the decarbonization of the transportation sector has become a paramount concern for the United States in recent years1. In 2021, the United States transportation emissions accounted for 29% of the nation’s total greenhouse gas (GHG) emissions2. To achieve ambitious climate goals1, it has become imperative to address the emissions associated with the transportation sector comprehensively.

National efforts to mitigate GHG emissions from transportation in the United States have centered on three key objectives: increasing convenience, improving efficiency, and transitioning to cleaner options1. These endeavors are aimed at reducing the total vehicle kilometers traveled (VKT), increasing ridership, enhancing fuel economy, and adopting technologies that offer low emissions both presently and in the future1. However, this transformation has several challenges.

The United States currently has the highest passenger kilometers traveled (PKT) per capita globally, which is accompanied by low ridership of public transportation3. Additionally, the country maintains a substantial fleet of aging and inefficient vehicles, and a heavy reliance on petroleum-based fuels4. As such, addressing the transportation sector’s emissions necessitates a comprehensive understanding of the emissions profiles of average, current, and future vehicle models, as well as the identification of societal and technological solutions.

Several studies have assessed the GHG emissions associated with various modes of inland transportation. Noussan et al.5 reviewed and compared the GHG emissions factors of passenger transportation modes, finding that bicycles and rail had the lowest GHG emissions, while light-duty vehicles had the highest5. This study surveyed available literature from specific locales, capturing the mean rather than the United States average5. Creutzig6 compared the GHG emissions of active transportation modes to buses, rail, and cars in Germany and Israel, discovering that shared electric scooters had higher emissions than buses and rail but lower emissions than cars with one occupant6. Moreover, Creutzig 6 found private bicycles and electric scooters to have the lowest emissions among the modes considered, though these results do not use United States parameters6. For instance, Figueroa et al.7 noted that the energy intensity of buses per PKT in the United States is much higher than the rest of the world due to low ridership, highlighting the importance of evaluating GHG emissions specific to United States parameters7. These studies have been limited in scale and have not deliberately calculated the United States average, which is needed to compare and contextualize GHG emissions reductions possible through mode shifts and technology advancements on an impactful nationwide scale to inform transportation and energy sector policy.

This study aims to address these critical issues by leveraging data on the United States average passenger load factors, fuel economies across vehicle ages, and the technological composition of various inland transportation modes. It employs a robust methodology, including consistent assumptions for mode comparisons, Monte Carlo simulations to capture uncertainty, and the inclusion of dietary emissions associated with each mode. By comprehensively assessing the attributional life cycle emissions associated with each mode, this research provides valuable insights into GHG emissions in the United States transportation sector, providing a transparent breakdown of emissions for each mode, including both current conditions and potential future advancements. The assessment involves the determination of total PKT and GHG emissions for each transportation mode based on their current technological profiles. Moreover, the study explores current technology models and high ridership scenarios to identify technologies capable of reducing emissions. It also projects GHG emissions for 2050, assuming a decarbonized electricity grid, and highlights each technology’s potential for emissions reduction. Finally, the research contextualizes the results in terms of emissions reductions achievable through urban mode shifts and a transition to technologies reliant on decarbonized electricity. Overall, this study provides benchmark values for average GHG emissions of United States transportation technologies and projected advancements nationwide, enabling comprehensive emissions comparisons across scenarios to guide technology development and adoption efforts. The study finds that while shifting to public transportation and active modes can reduce GHG emissions, advancements in technology and electrification have a greater potential for reducing emissions by up to 85% by 2050. The results indicate that the majority of GHG emissions reductions can be achieved through fuel shifts alone, with mode shifts providing modest reductions in the near-term and minimal reductions in the long-term.

Results

Current transportation mode emissions

The study incorporates PKT data across various transportation modes, including light-duty vehicles, buses, rail, walking, traditional bicycles, electric bicycles, and electric scooters in the United States8,9,10,11,12. The attributional GHG emissions are derived using data from literature13,14,15,16,17,18,19,20,21,22, with adjustments made to ensure consistency across technologies and current usage (see Methods). These emissions encompass contributions from batteries, materials, and fuel utilization throughout the entire life cycle, from cradle-to-grave, of each transportation mode. Light-duty vehicles, buses, and rail are comprised of various technologies and fuels, such as gasoline, diesel, and electricity4,23. The emissions resulting from the additional caloric intake required for all modes is considered16,24,25,26. The total PKT and GHG emissions from each transportation mode are illustrated in Fig. 1, with values representing the mean of the Monte Carlo simulation and error bars indicating the 2.5th and 97.5th percentiles (see Methods). The results emphasize that light-duty vehicles are the predominant contributors to total GHG emissions in the United States.

Fig. 1: Passenger kilometers traveled (PKT) and greenhouse gas (GHG) emissions of transportation modes in the United States.
figure 1

Greenhouse gas emissions include all stages over the life cycle of each transportation mode. The values represent the mean of the Monte Carlo simulation, with error bars indicating the 2.5th and 97.5th percentiles. g grams, CO2e Carbon Dioxide equivalent.

Full size image

As shown in Fig. 1, light-duty vehicles represent an order of magnitude higher PKT and GHG emissions than all other modes combined. Notably, 93% of these light-duty vehicles run exclusively on gasoline-ethanol blends23, highlighting the limited deployment of alternative fuel sources. Electrified powertrains, including plug-in hybrid electric vehicles, hybrid electric vehicles, and electric vehicles, collectively account for 3.4% of the vehicle class population23, representing the initial transition toward cleaner technologies. Furthermore, although hybrid electric buses make up 18% of the bus fleet, a substantial 73% of buses still exclusively rely on petroleum-based fuels27. Buses represent a large share of PKT and GHG emissions in the United States, contributing 7% and 11% to the respective totals among the modes considered in this study. Rail, with limited PKT and emissions, accounts for less than 1% in both categories due to its limited deployment across the United States. Finally, active forms of transport, such as walking, traditional bicycles, electric bicycles, and electric scooters, collectively represent less than 1% of PKT and GHG emissions in the United States, emphasizing their minimal GHG emissions compared to traditional transport modes. These findings highlight the critical role of light-duty vehicles and buses in the context of the United States transportation emissions and the potential for gradual emissions reduction through the adoption of cleaner technologies, considering their long operational lifespans4,27.

Average emissions by transportation mode

A detailed breakdown of emissions throughout the entire life cycle for each transportation mode, comprising various vehicle and fuel technologies, is presented in Fig. 2, providing insights into the sources of emissions over the lifetime of these modes. To capture variability and uncertainty in average emissions estimates, the results are derived from a Monte Carlo analysis, with values representing the mean of the simulations and error bars indicating the 2.5th and 97.5th percentiles. These emissions are categorized into five distinct groups: battery emissions, which encompass the entire life cycle of the battery, including mineral extraction, manufacturing, and end-of-life; materials emissions, which encompass emissions associated with vehicle materials, spanning mineral extraction, manufacturing, maintenance, and end-of-life processes; fuel-embodied emissions, which incorporate upstream emissions stemming from fuel production and distribution; direct operational emissions, which primarily result from fuel combustion emitted at the tailpipe during vehicle operation; dietary emissions, which represent the United States average emissions from the caloric intake needed to fuel human effort for each form of travel. Light-duty vehicles are subdivided into multiple fuel types, considering their widespread use and diverse energy sources. The findings show GHG emissions from the United States transportation modes are largely from the use of energy regardless of the vehicle architecture. Moreover, the GHG emissions from all stages of the life cycle emanate primarily from fuel consumption associated with mineral extraction, cultivation, production, transportation, and combustion processes21.

Fig. 2: Average greenhouse gas emissions of transportation mode compositions in 2023.
figure 2

The values represent the mean of the Monte Carlo simulation, with error bars indicating the 2.5th and 97.5th percentiles. The average ridership for each mode is shown in parentheses next to the mean value. Average dietary emissions can be reduced through dietary change. g grams, CO2e Carbon Dioxide equivalent, PKT passenger kilometer traveled.

Full size image

Among the modes studied, buses emerge as the highest emitters on average due to their heavy reliance on petroleum-based fuels27 and inherent inefficiencies, attributed to large vehicle sizes coupled with relatively low average ridership4. In contrast, rail systems exhibit much lower mean emissions due to their higher efficiency, ridership, and utilization of electricity as fuel4.

Walking incurs notable emissions, primarily driven by carbon-intensive diets in the United States22 and the biomechanical inefficiencies of linear locomotion at the slowest travel speed of 4.8 kilometers per hour (kph)16,26. While traditional bicycles share similar fuel sources with walking, their greater efficiency through gyroscopic motion results in the lowest GHG emissions from a reduced metabolic equivalent of task (MET) and time per kilometer (km) traveled24. Electric bicycles offer a unique combination of electricity and human effort, resulting in lower combined fuel and dietary emissions compared to traditional bicycles, attributable to a lower MET and decreased time per km traveled24. However, due to higher battery and material emissions, electric bicycles have higher overall emissions per PKT than traditional bicycles. Although electric scooters have a lower MET (2.826) compared to walking (3.526), traditional bicycles (6.824), and electric bicycles (5.924), their slower travel speeds (8.4 kph28) result in similar dietary emissions per PKT as traditional bicycles. Consequently, electric scooters have the highest mean emissions among active travel modes.

Electric vehicles have the lowest total average GHG emissions among light-duty vehicles, despite having the highest combined battery and material emissions. Hybrid electric vehicles also demonstrate superior emissions performance compared to conventional gasoline vehicles, attributed to improved energy efficiencies from regenerative braking21.

Current vehicle technologies

The future emissions profiles of transportation modes hinge on technology choices, emphasizing the need for cleaner options. Figure 3 breaks down GHG emissions of new technologies in 2023, considering average and maximum ridership. This highlights how technology shifts and increased ridership can cut GHG emissions today. Additional light-duty vehicle technologies are included in Supplementary Fig. 1.

Fig. 3: Greenhouse gas emissions of 2023 transportation technology models.
figure 3

Results are presented for average ridership (left) and maximum ridership (right). The values represent the mean of the Monte Carlo simulation, with error bars indicating the 2.5th and 97.5th percentiles. The ridership for each mode is shown in parentheses next to the mean value. Average dietary emissions can be reduced through dietary change. g grams, CO2e Carbon Dioxide equivalent, SUV sport utility vehicle.

Full size image

In Fig. 3, new vehicle models exhibit lower emissions than their older counterparts (Fig. 2), attributed to improved fuel efficiency21. Notably, replacing conventional gasoline vehicles and hybrid electric vehicles with electric vehicles can dramatically cut emissions. Emissions from electric vehicle batteries, which occur prior to vehicle operation and at the end-of-life, make up just 18% of the total lifetime emissions of electric vehicles, and electric vehicles achieve emissions parity with conventional gasoline vehicles after covering 32 to 43 thousand VKT and with hybrid electric vehicles after 64 to 82 thousand VKT. Adopting smaller vehicles like cars require fewer materials and less energy per VKT than sport utility vehicles (SUVs) and pickup trucks, reducing battery and fuel demands per PKT in most cases21. Interestingly, gasoline SUVs have slightly lower emissions than gasoline cars per PKT due to a higher average passenger load factor than cars4.

Electric buses exhibit the third lowest mean emissions among all technologies in Fig. 3 under average ridership conditions. Under maximum ridership conditions29, electric buses could potentially achieve the lowest life cycle emissions among all technologies considered. Transit rail showcases the lowest mean emissions among rail types, exclusively relying on electricity for fuel, unlike intercity and commuter rail, which also use diesel4.

In addition to fuel types, operational characteristics dramatically influence emissions within each transportation mode. Personal electric scooters have 37% less emissions than dock-less electric scooters, as they avoid the high emissions associated with collection and redistribution. Similarly, personal electric bicycles exhibit the lowest emissions among electric bicycle types, avoiding the emissions associated with collection and redistribution. Dock-less electric bicycles have higher emissions than all other electric bicycles due to fewer daily PKTs, which result in inefficient material and battery usage, coupled with increased travel distance for collection and redistribution compared to station-based electric bicycles13.

The estimated GHG emissions for these transportation technologies can be compared to findings from other attributional life cycle assessments. Creutzig 6 similarly found that personal electric scooters have much lower emissions (45 g-CO2e per PKT) compared to shared electric scooters (120 g-CO2e per PKT)6. Blonde et al.25 and Noussan et al.5 identified traditional bicycles (21 g-CO2e per PKT) as having the lowest GHG intensity among travel modes when dietary emissions are included, though they did not consider walking and electric scooters5,25. Noussan et al.5 also reported substantially lower emissions for buses (48 g-CO2e per PKT) and rail (33 g-CO2e per PKT) than those shown in Fig. 3, likely due to different location-specific ridership and emissions factors5. Additionally, Noussan et al.5 reviewed light-duty vehicle GHG emissions, finding the mean higher than in Fig. 3, with 138 g-CO2e per PKT for electric cars and 235 g-CO2e per PKT for gasoline cars5, which can be attributed to advancements in vehicle energy efficiencies over time.

Impact of decarbonized electricity

The GHG emissions of more efficient transportation technologies coupled with a decarbonized electricity grid in the United States are explored in Fig. 4, representing potential GHG emissions in 2050 for each transportation technology under both average and maximum ridership scenarios. The decarbonization of electricity is integrated throughout the life cycle of each transportation mode21, influencing material, battery, and fuel-embodied emissions. Furthermore, projected enhancements in fuel economies are expected to reduce fuel consumption during operation21. However, even with remarkable improvements, combustion engines are forecasted to produce substantially higher emissions than electrified technologies over the next 27 years. Conversely, maximizing bus ridership, regardless of the fuel type, approaches the emissions levels of electric light-duty vehicles with average ridership.

Fig. 4: Greenhouse gas emissions of 2050 transportation technology models with a decarbonized electricity grid.
figure 4

Results are presented for average ridership (left) and maximum ridership (right). The values represent the mean of the Monte Carlo simulation, with error bars indicating the 2.5th and 97.5th percentiles. The ridership for each mode is shown in parentheses next to the mean value. Average dietary emissions can be reduced through dietary change. g grams, CO2e Carbon Dioxide equivalent, SUV sport utility vehicle.

Full size image

Among the technologies considered in Fig. 4, transit rail is shown to have the lowest emissions, primarily attributable to its reliance on electricity, absence of batteries, and efficient utilization of materials for passenger transportation. Other electrified transportation options also exhibit low fuel-embodied emissions but are burdened with substantial battery and material emissions. For instance, electric buses emerge as the second-lowest emissions option, underscoring the potential for substantial emissions reduction through the use of electrified public transportation. Furthermore, this study demonstrates that the emissions associated with dock-less electric scooters can approach those of personal electric scooters when electric vans, powered by decarbonized electricity, are employed for collection and redistribution. Nevertheless, under both average and maximum ridership, electric scooters have substantially higher GHG emissions than electric buses and electric vehicles due to large dietary emissions. Notably, the high average dietary emissions can be reduced by approximately 80% through dietary changes30.

The impact of uncertainty meaningfully affects the ranking of the cleanest transportation modes. Variability in input data, such as vehicle efficiency and operational characteristics, introduces a range of possible emission outcomes. Monte Carlo simulations help quantify this uncertainty by generating a distribution of potential average emissions for each mode. Consequently, while some modes may appear cleaner on average, the overlapping confidence intervals indicate that these rankings are not absolute. For instance, slight variations in dietary emissions or vehicle usage patterns can shift the relative positions of buses, light duty vehicles, and active transportation modes. This highlights the importance of considering uncertainty in environmental assessments to ensure that strategies remain effective under various real-world conditions.

Urban fleet transition

The impact of transitions and advancements is demonstrated through an urban fleet transition in Fig. 5, which features different combinations of mode shifts and electrification. The urban mode shift scenarios are derived from Fulton et al.31. Scenarios without a mode shift comprise 91% light-duty vehicles, 6% bus, 1% rail, 1% bicycle and scooter, and 1% walking. The mode shift scenarios encompass a transition away from light-duty vehicles (67%) and towards increased use of buses (17%), rail (3%), bicycles and scooters (10%), and walking (4%). Based on the ratio of new vehicles required according to Fulton et al.31 and the increase in bus and rail from the mode shifts, the load factors for bus and rail increase by a factor of 1.5 and 2.2, respectively31. Additionally, the current make-up of 2023 technologies from Fig. 3 is modeled and compared to 2050 technologies (Fig. 4), with buses and light-duty vehicles being fully electric in 2050. Notably, a limitation of the study is that the GHG emission factors represent national averages and are not specifically tailored to urban contexts. The results show that transitions to electrified powertrains and advancements in technologies by 2050 have a greater impact on average GHG emissions than urban mode shifts. The results are presented with and without dietary emissions.

Fig. 5: Average urban greenhouse gas emissions from various transportation scenarios.
figure 5

The scenarios include 2023 technology without mode shift, 2023 technology with mode shift, 2050 electric technology without mode shift, and 2050 electric technology with mode shift. The emissions are expressed as grams of Carbon Dioxide equivalent per passenger-kilometer traveled (g-CO2e per PKT). Values within brackets exclude dietary emissions.

Full size image

As shown in Fig. 5, the transition to electrified transportation technologies in 2050 dramatically reduces GHG emissions per PKT compared to 2023 technology, both with and without mode shift. Incorporating mode shifts in 2023 reduces GHG emissions, highlighting the benefits of reducing light-duty vehicle travel in the near-term. In 2050, including mode shifts does not alter GHG emissions when dietary emissions are taken into account, as electric vehicles exhibit lower total GHG emissions than walking and electric scooters and similar emissions to electric and traditional bicycles (Fig. 4).

Discussion

The decarbonization of the United States transportation sector necessitates substantial shifts in transportation behaviors and technological preferences. Despite buses having relatively high emissions due to low ridership, public transportation plays a vital role in ensuring mobility access, particularly for disadvantaged populations32. In the near-term, enhancing the ridership of public transportation modes like buses and rail not only curtails light-duty vehicle travel and its associated GHG emissions but also leads to GHG emissions reductions (Fig. 5) and financial savings per PKT for these modes. Shifts in transportation modes away from light-duty vehicles are largely constrained by locational factors, with the adoption of public transportation and active modes primarily limited to urban areas25,33. Moreover, public transportation has limited coverage across the United States34 and, if expanded, would require adequate ridership to achieve GHG emissions reductions relative to light-duty vehicles (Fig. 3).

Furthermore, active transportation modes can serve as primary options or seamlessly integrate with public transit systems to address connectivity challenges35. Active transportation also requires safe infrastructure, which may need to be expanded, and is best suited for short trip distances, limiting its usage33. Among active transport options, traditional bicycles emerge as having the lowest GHG emissions compared to walking, electric scooters, and electric bicycles, primarily due to the absence of a battery and limited material emissions. Furthermore, this study shows that active transportation modes currently have lower GHG emissions than the average light-duty vehicle fleet (Fig. 2). It is worth noting that promoting active transport modes extends beyond the scope of this study, as these modes may also provide health benefits through physical exercise.

Similarly, sustainable diets have the potential to both lower dietary GHG emissions and provide health benefits30. Emissions from dietary choices are highly variable, influenced by individual diets and geographical factors16,22,30,36. The compensatory dietary intake associated with active modes of transportation remains uncertain and varies widely5, with one study suggesting a range between 19% and 96%16. However, as an attributional life cycle assessment, this study fully considers the GHG intensity of the additional energy required for human effort for each transportation mode.

Electricity emissions also exhibit meaningful regional variations37,38, and this analysis is based on the United States average electricity mix. Moreover, individual technology emissions can vary by manufacturer39, but this study reveals that newer and more efficient technologies generally exhibit considerably lower emissions than the average vehicle composition.

The retirement of old and inefficient vehicles can substantially mitigate GHG emissions, particularly for light-duty vehicles and buses, which collectively account for 98% of the United States PKT among the modes assessed in this study. Given that buses have a minimum useful life of 12 years27 and more than half of light-duty vehicles remain in operation for 15 years or longer4, careful consideration of replacement technologies is of high importance. On an individual scale, using personal electric and traditional bicycles, riding public transportation, and carpooling can yield GHG emissions reductions in the near term (Fig. 3). On a societal scale, the primary strategies for decarbonizing the transportation sector should focus on the electrification of transportation and the decarbonization of electricity generation, rather than long-term mode shifts.

Methods

Life cycle assessment methodology was leveraged to evaluate the GHG emissions associated with various modes of inland transportation. Employing a cradle-to-grave system boundary and using a functional unit of one PKT, this analysis evaluated the emissions of several transportation options, including average light-duty vehicles, rail, buses, traditional bicycles, walking, electric bicycles, and electric scooters. The attributional GHG emissions were calculated across different time horizons: for currently deployed models, as well as for advanced technology models for 2023 and 2050. The emissions per VKT were adjusted based on the United States average passenger load factor per vehicle to provide a more accurate assessment of average emissions in the United States. Further, the emissions for each technology model were also evaluated with maximum ridership, as an absolute upper bound, to determine the potential for emissions reduction. These emissions were categorized into five distinct groups: battery, materials, fuel-embodied, direct operation, and dietary. Dietary emissions were included for the operator of each transportation technology. The values for key assumptions are summarized in Tables 1–5 below.

Table 1 Occupancy by vehicle type
Full size table
Table 2 Parameters to calculate dietary energy intake per kilometer traveled
Full size table
Table 3 Energy efficiencies for each vehicle type
Full size table
Table 4 Lifetime of vehicles
Full size table
Table 5 Greenhouse gas intensities of diets, electricity, and batteries, along with battery sizes for electric vehicles
Full size table

Passenger kilometers traveled

The study incorporated direct PKT data across a range of transportation modes, encompassing light-duty vehicles, rail, buses, traditional bicycles, and walking8. These data were compiled using the most recent information available before the onset of the COVID-19 pandemic, with data sources from 2017 for traditional bicycles and walking and 2019 for light-duty vehicles, rail, and buses8.

In contrast, the PKT values for electric scooters and electric bicycles were estimated based on more recent data from 2021 to account for emerging trends in their usage. Due to data collection limitations, these estimations were derived by multiplying the total number of trips, their corresponding average trip lengths, and a load factor of one passenger per vehicle. The PKT of personal electric bicycles was calculated from an estimated total of 4.1 million personal electric bicycles40, an average trip frequency of 202 per year11, and an average trip distance of 17 km11. Data on the number of trips and the average trip distances for station-based electric bicycles, dock-less electric bicycles, and dock-less electric scooters were acquired from the National Association of City Transportation Officials (2022)9. These figures revealed 2021 trip counts of 47 million, 2.5 million, and 62.5 million, respectively, with average distances of 2.3 km, 1.9 km, and 1.9 km, respectively9. Furthermore, the estimation of PKT for personal electric scooters was approximated based on findings from a survey on ownership preferences conducted by Heineke et al.12, which revealed that 22% of respondents purchased an electric scooter after initially using a shared one12. To reflect this, the PKT attributed to personal electric scooters was then calculated to be 28% of the PKT associated with dock-less electric scooters, assuming a similar usage pattern.

Average vehicle emissions

GHG emissions from vehicles within each transportation mode were determined through a weighted average methodology, encompassing various technologies and model years. In the case of buses and light-duty vehicles, emissions attributed to fuel and vehicle operations were averaged across multiple model years. However, for other transportation modes, including traditional bicycles, walking, electric bicycles, electric scooters, and rail, as well as for the battery and materials emissions categories across all modes, GHG emissions estimates were reliant on data from individual model years due to the unavailability of consistent historical data. Consistent 2023 United States average emissions values of 494 grams (g) of Carbon Dioxide equivalent (CO2e) per kilowatt-hour (kWh) of electricity consumed and 79 kilograms (kg) of CO2e per kWh of lithium-ion battery capacity in the vehicle were used for each electrified transportation mode other than electric buses21.

Emissions from light-duty vehicles were simulated using The Greenhouse gases, Regulated Emissions, and Energy use in Technologies (GREET) (2022) model21 and are shown in the Supplementary Data 1. The emissions calculations were weighted according to the proportion of vehicle sales, with cars accounting for 31%, SUVs for 52%, and pickup trucks for 17% of the market share41. To determine the average emissions, data on vehicle registrations categorized by fuel type were considered. The breakdown included gasoline (85%), hybrid electric (2%), 85% ethanol or less flex fuel (7%), diesel (3%), biodiesel (1%), electric (1%), and plug-in hybrid electric (1%)23. The fuel economy values spanning from model years 1993 through 2023 were modeled21 and a weighted average was computed for each type of light-duty vehicle, considering car and truck survival rates and the annual VKT for each vehicle age4.

In addition, emissions associated with batteries and materials were calculated using GREET (2022) for vehicles utilizing conventional materials with lifetimes of 279 thousand km for cars and 295 thousand km for SUVs and pickup trucks21. Electric vehicles were modeled to feature a 483 km (300 mile) range battery that lasted the entire vehicle lifespan21. Regardless of powertrain, to standardize the emissions, a passenger load factor was applied, with a value of 1.5 for cars, 1.7 for SUVs, and 1.8 for pickup trucks, converting the emissions to be per PKT4. The load factors for SUVs, as well as survival rates and annual VKT, were determined based on the proportion of SUVs classified as cars (23%) and trucks (77%)41. The load factor for the maximum ridership scenario was 5 for cars, SUVs, and pickup trucks.

The dietary emissions for light-duty vehicles were based on the MET of driving and speed of the vehicle. The average speed of the vehicle (45 kph) was derived from a combination of national average highway (77 kph) and city (34 kph) speeds4 and the amount of highway (25%) versus city (75%) VKT in 202242.

Emissions resulting from human effort were calculated using dietary GHG emissions from Bassi et al.22 of 2.54 kg-CO2e per capita per day (Fuelnormal), which captures the decline of dietary GHG emissions in the United States22. The standard MET value was modeled as 1.5 (METnormal), which is consistent with other studies16,25. Embodied emissions in fuel from human effort (Fuelactive) were calculated using Eq. 1, considering the MET value specific to each type of activity (METactive) and the speed of travel in kph.

$${{Fuel}}_{{active}}={{Fuel}}_{{normal}}* {{MET}}_{{active}}/\left({{MET}}_{{normal}}* 24* {speed}\right)$$
(1)

Embodied emissions in fuel and direct operation emissions were derived using the GREET (2022) model for transit buses21. These emissions were determined through a weighted average that considered various powertrain technologies and model years. Specifically, the analysis focused on transit buses, which had an average load factor of 7.5 passengers4 and a maximum load factor of 40 passengers29 per VKT. Due to a lack of load factor data for school and intercity bus types, these were not included in this study4. The composition of transit buses was distributed as follows: 42% diesel, 20% compressed natural gas, 18% diesel hybrid electric, 8% biodiesel, 2% gasoline, and 1% electric27. The allocation of model years for each fuel type was approximated based on historical bus sales, the historical composition of the fleet, and a minimum useful life of 12 years27. For gasoline buses, which were not covered in GREET (2022), embodied emissions in fuel and direct operation emissions were computed by utilizing the gasoline emissions per unit of energy from GREET (2022) and an average fuel economy of 5.2 km per gallon of gasoline43.

The emissions related to the battery and materials for each technology were simulated using a representative pickup and delivery truck model sourced from GREET (2022)21, with material emissions adjusted proportionally to a diesel bus weighing 11.6 tonnes44. In the case of compressed natural gas buses, material emissions were assumed to be 6% higher compared to diesel bus emissions based on O’Connell et al.18. Electric buses were modeled with a 200 kWh battery and considered to have one replacement during their operational lifespan45. Based on O’Connell et al.18, these buses were modeled to have an operational lifespan of 881 thousand VKT18, which is consistent with Federal Transit Administration guidelines found in Laver et al.46. The average bus GHG emissions of each fuel type are summarized in the Supplementary Data 1.

The dietary emissions for buses were based on the MET of the bus operator (2.526) and speed of the bus. The average speed of a bus (44 kph) was derived from a combination of average highway (77 kph) and city (34 kph) speeds4 and the amount of highway (23%) and city (77%) VKT in 202242.

The modeling of rail emissions was conducted independently for transit, commuter, and intercity rail systems, and the weighted average rail emissions were computed considering the PKT from each rail system. Material emissions data for transit and commuter rail were sourced from Chester and Horvath 14, while for intercity rail, the data was obtained from Chester and Horvath 15. As the referenced estimates did not account for end-of-life emissions, the material emissions were adjusted accordingly. For transit rail, an increase of 3.8% was applied while for commuter and intercity rail, a 3.5% adjustment was made. These modifications were derived from data on an electric and a hybrid electric pickup and delivery truck from GREET (2022)21.

Embodied emissions in fuel and direct operation emissions for each rail system were simulated using the GREET (2022) model with average4 and maximum ridership21. Maximum ridership reduced the average ridership emissions per PKT by 71% for commuter, 56% for intercity, and 76% for transit rail. For the transit rail system, it was further divided into light (13% of PKT) and heavy rail (87% of PKT), with their respective portions of PKT derived from Davis and Boundy4. In the case of commuter and intercity rail, embodied emissions in fuel and direct operation emissions were calculated by determining the proportion of VKT powered by electricity (commuter: 51%; intercity: 39%) and diesel (commuter: 49%; intercity: 61%), based on data from Davis and Boundy4. Emissions associated with batteries for rail systems were modeled as zero. Rail GHG emissions for each operating type are summarized in the Supplementary Data 1.

The dietary emissions for rail were based on the MET of the train operator and speed of the train categorized into two groups: 1) commuter and intercity; and 2) transit. The train operator was assumed to have the same 2.5 MET as a bus operator26. The average speed of the commuter and intercity rail was modeled to be 20 kph and the average speed of transit rail was 11 kph4.

This study also encompassed various modes of active transportation, including electric scooters, electric bicycles, traditional bicycles, and walking. Each mode was assumed to have a load factor of one passenger per VKT for both average and maximum ridership scenarios. The embodied emissions in fuel associated with each mode arose from human effort, electricity, infrastructure, and/or support vehicles. Infrastructure emissions pertained solely to station-based electric bicycles, amounting to 35 g-CO2e per PKT13. Notably, support vehicle emissions were negligible for walking, personal electric bicycles, and traditional bicycles. However, for dock-less electric scooters, dock-less electric bicycles, and station-based electric bicycles, substantial emissions were attributed to the collection and redistribution processes. These emissions were adapted from Kazmaier et al.19 for dock-less electric scooters19 and Luo et al.13 for dock-less electric bicycles and station-based electric bicycles13. The emissions were adjusted to use the cradle-to-grave emissions of a 2023 gasoline pickup truck (357 g-CO2e per km) as the basis for estimation13,19, resulting in emissions of 45 g-CO2e per PKT for dock-less electric scooters, 36 g-CO2e per PKT for dock-less electric bicycles, and 9.8 g-CO2e per PKT for station-based electric bicycles.

The electric scooters were exclusively powered by electricity, while walking and traditional bicycles relied solely on human effort for propulsion. The electric scooters were modeled to require an MET of 2.826 to operate with an average speed of 8.4 kph28. Electric bicycles combined human effort with electric power, traveling at 25 kph with 150 watts of electric power and human effort equal to 5.8 MET24. The electric bicycle consumed 6.1 watt-hours of electricity per PKT24, whereas the electric scooter consumed 29 watt-hours of electricity per PKT47. In contrast, traditional bicycles traveled at 21 kph with an MET of 6.724, while walking had a pace of 4.8 kph with an MET of 3.526. Notably, the MET values from Alessio et al.24 were selected for electric bicycles and traditional bicycles, as systematic reviews indicate that electric bicycles have an MET approximately 0.83 lower than traditional bicycles48. Alessio et al.24 also provided the necessary combination of MET, speed, and electrical output for electric bicycles, with an MET for traditional bicycles that is consistent with the literature26.

The material emissions associated with walking stemmed from shoes and amounted to 14 kg-CO2e with an assumed lifetime of 644 km (400 miles)17. Traditional bicycles had material emissions totaling 105 kg-CO2e20 and their assumed lifetime was 19 thousand km25. For both walking and bicycling, battery and operation emissions were zero.

In the case of electric scooters and electric bicycles, battery emissions were calculated based on lithium-ion batteries with capacities of 0.48 kWh49 and 0.96 kWh50, respectively. Material emissions for electric scooters and electric bicycles were estimated as 187 kg-CO2e per dock-less electric scooter19, 187 kg-CO2e per personal electric scooter19, 319 kg-CO2e per station-based electric bicycle13, 419 kg-CO2e per dock-less electric bicycle13, and 319 kg-CO2e per personal electric bicycle13.

As described in the Passenger kilometers traveled section, the distribution of PKT for each type of electric bicycle was approximated as follows: 99% personal, 0.046% dock-less, and 1.0% station-based. Electric scooters were approximated to be 22% personal and 78% dock-less. The lifetime PKT for each electric scooter and electric bicycle was modeled as follows: 47 thousand km for station-based electric bicycles13, 13 thousand km for dock-less electric bicycles13, 19 thousand km for personal electric bicycles25, and 6.5 thousand km for personal and dock-less electric scooters frames51, and 3.5 thousand km for personal and dock-less electric scooters batteries51. The lifespan of the electric scooters is modeled to be longer than in existing literature to account for the manufacturing improvements in the industry, as detailed in Hanson et al.51.

Current vehicle technology emissions

Based on the analysis in the Average vehicle emissions section, a breakdown of new technologies for each transportation mode was also included, with a focus on identifying the current technologies offering the lowest emissions in the future. In the case of electric scooters, electric bicycles, traditional bicycles, walking, and rail, the technologies pertinent to each mode were extracted, without specifying a particular year, due to the unavailability of historical emissions data.

In contrast, for light-duty vehicles, data extraction encompassed the 2023 models for each fuel type and vehicle size, while for buses, the 2023 models for each bus fuel type were extracted. Notably, the emissions for 2023 gasoline buses were determined by extrapolating from the gasoline emissions per unit of energy and applying a conversion based on diesel fuel economy, where a diesel bus travels 1.275 times farther than a gasoline bus per unit of energy52.

Future vehicle technology emissions

Improvements in vehicle efficiency and the transition to a decarbonized electric grid53, while maintaining consistent average and maximum passenger load factors, were modeled to assess the potential for emissions reduction across various technologies. In the GREET (2022) model, a fully decarbonized grid was assumed for the year 2050, impacting battery, material, and fuel-embodied emissions21. Notably, GHG emissions were nonzero (29 g-CO2e per kWh) for electricity due to the embodied emissions associated with electricity generation and distribution infrastructure21.

Each of the technologies from 2023 was considered in the 2050 scenario. For the projected improvements in vehicle fuel economies in 2050, data from the GREET (2022) model was utilized, covering light-duty vehicles, buses, and rail. The electricity consumption and emissions intensity of human-powered locomotion for electric scooters and electric bicycles remained unchanged due to the uncertainty of projections. However, the collection and redistribution emissions for electric scooters and electric bicycles were adjusted by modeling the support vehicle as a 2050 electric pickup truck, which exhibited lower emissions of 49 g-CO2e per VKT21.

Furthermore, lithium-ion battery emissions were determined to be 64 kg-CO2e per kWh of capacity from GREET (2022)21. Material emissions for each light-duty vehicle in 2050 were directly computed within the GREET (2022) model. For other technologies, reductions were approximated based on similar vehicles included in GREET (2022), with the percentage reduction measured from 2023 to 205021.

Material emissions for electric scooters and electric bicycles were calculated based on an average reduction of 35% compared to electric light-duty vehicles21. Transit rail and electric bus material emissions saw a 19% reduction based on an electric pickup and delivery truck. For hybrid electric bus battery and material emissions, as well as commuter and intercity rail material emissions, a 29% reduction in battery emissions and a 18% reduction in material emissions were assumed, modeled after a diesel-hybrid electric pickup and delivery truck. The reductions in battery (30%) and material (18%) emissions for diesel, biodiesel, gasoline, and natural gas buses were based on a diesel pickup and delivery truck21.

Uncertainty analysis

Due to variations in input values found in the literature, an uncertainty analysis was conducted. A sensitivity analysis using a ± 20% change was performed on 317 input values across 71 results to identify the five most sensitive inputs for each result. These inputs were then modeled in a Monte Carlo analysis, running 10 thousand simulations, which was validated as adequate based on the convergence of the mean. For PKT results, which had limited input variables, fewer than five input variables were included in the Monte Carlo analysis. When available, distributions were formulated based on the known confidence intervals from the original data sets; these variables and their distributions are listed in Supplementary Table 1. For the remaining inputs, distributions were set such that the standard deviation was 20% of the mean, adequately capturing the expected range of values for each input. These variables and their distributions are available in Supplementary Data 1.