Policy Synergy Scenarios for Tokyo’s Passenger Transport and Urban Freight: An Integrated Multi-Model LEAP Assessment

Abstract

To identify the emission reduction potential and policy synergies of Tokyo’s road passenger and urban road freight transport under the “carbon neutrality target,” this paper constructs an assessment framework for megacities. First, based on macroeconomic socioeconomic variables (population, GDP, road length, and employment), regression equations are used to predict traffic turnover for different modes of transport from 2021 to 2050. Then, the prediction results are imported into the LEAP (Long-range Energy Alternatives Planning) model. By adjusting three policy levers—vehicle technology substitution (ZEV: EV/FCEV), energy intensity improvement, and upstream electricity and hydrogen supply decarbonization—a “single-factor vs. multi-factor (policy synergy)” scenario matrix is designed for comparison. The results show that the emission reduction potential of a single measure is limited; upstream decarbonization yields the greatest independent emission reduction effect, while the emission reduction effect of deploying zero-emission vehicles and improving energy efficiency alone is small. In the most ambitious composite scenario, emissions will decrease by approximately 83% by 2050 compared to the baseline scenario, with cumulative emissions decreasing by over 35%. Emissions from rail and taxis will approach zero, while buses and freight will remain the primary residual sources. This indicates that achieving net zero emissions in the transportation sector requires not only accelerated ZEV penetration but also the simultaneous decarbonization of electricity and hydrogen, as well as policy timing design oriented towards fleet replacement cycles. The integrated modeling and scenario analysis presented in this paper provide quantifiable evidence for the formulation of a medium- to long-term emissions reduction roadmap and the optimization of policy mix in Tokyo’s transportation sector.

1. Introduction

The transportation sector is one of the major sources of global carbon dioxide emissions, which has a serious impact on the global climate [1]. According to the IPCC (2022) [2], transport is one of the largest end-use sectors of global GHG emissions, accounting for about 15% of total emissions [3]. Official statistics from Japan show that the transportation sector accounts for 17% of the country’s total carbon emissions, with road transport accounting for 87.6% and freight transport for 39.2% of that [4,5]. This is largely due to heavy dependence on fossil fuels [6]. Empirical evidence from China indicates that, despite the rapid expansion of the new energy vehicle (NEV) market, the contribution of NEVs to reducing national crude oil imports has so far been modest, implying that transport decarbonization still requires broader systemic changes in the oil-dependent energy system [7]. Therefore, advancing environmental technologies and renewable energy through targeted policies is critical to transforming the sector and achieving zero-carbon urban transport [8,9]. Tokyo has proposed an overall goal and action framework to achieve net-zero emissions by 2050, providing a policy anchor for city-level emissions reduction pathways [10]. Against this backdrop, this paper focuses on a systemic emissions reduction assessment and scenario comparison for public passenger transport and urban road freight.

To achieve the carbon neutrality goal by 2050, the Japanese government and the Tokyo Metropolitan Government have introduced several policies and plans to accelerate the electrification of public transportation. For example, the Japanese government is promoting the popularization of electric vehicles (EVs) by adopting subsidies and improving infrastructure [11]. In addition to EVs, Japan has also been committed to the development of hydrogen fuel cell vehicles (FCEVs) [12], because green hydrogen is seen as a solution for building a zero-emission transportation system in the future [13]. In this context, it is necessary to establish a scientific model framework to estimate the carbon reduction potential to support the formulation and optimization of local climate strategies.

The Long-range Energy Alternatives Planning (LEAP) model has been widely applied in energy system and climate policy research, especially in the transportation sector [14]. However, existing LEAP-based studies have four main limitations. First, they neglect the driving forces of socioeconomic factors, and the forecasts of turnover volume typically use a fixed growth rate. Second, the scenario design is usually based on a single policy factor, rarely considering the synergistic effects of multiple policies. Third, zero-emission vehicles (ZEVs) are not differentiated according to vehicle technology and energy type (pure electric vehicles and hydrogen fuel cell vehicles). Finally, the reduction in energy intensity and emission factors is often based on subjective judgments, lacking objective technical and policy support, thus reducing the reliability and applicability of the studies.

In addition, theoretical and policy discussions related to urban transportation emissions reduction can be roughly divided into four development stages. As shown in Figure 1, early research provided a theoretical framework for emissions reduction in the urban transportation sector. Subsequent research shifted its focus to urban technological and energy transitions. After 2010, with increasing emphasis on carbon emissions reduction, international organizations and the Japanese government released a series of emissions reduction strategies and plans. In recent years, research on carbon neutrality scenarios and technology assessments for specific Japanese cities or sectors has been increasing.

However, as can be seen from Figure 1, previous studies have neglected how broader macro-level conditions (including changes in urban spatial structure, socio-economic transformation, and traveler behavior) affect transportation activities and their carbon emission consequences. Some recent studies have shown that urban carbon emissions do not evolve in a simple linear manner, and the relationship between urban spatial form and carbon emissions is highly nonlinear and spatially heterogeneous [15]. In addition, a study on social media machine learning in Japan’s three major metropolitan areas found that public sentiment fluctuates due to the influence of energy transition, highlighting the importance of integrating social cognition into the design of low-carbon city policies in Japan [16]. Therefore, in this study, socio-economic changes will be regarded as the main factor affecting urban carbon emissions, and the scenario setting will be closer to reality, rather than developing transportation demand and technological innovation independently of the social environment. In addition to interpreting data and policies, it is also necessary to combine theoretical analysis of passenger behavior and transportation-related social forms to assess technological and policy pathways while clarifying their social implications.

There are four main objectives of this research:

• Building a systematic forecasting and evaluation framework.
• Evaluating the impact of different policy pathways on carbon emissions in Tokyo’s transport sector.
• Identify the key drivers to achieve deep emissions reductions and carbon neutrality.
• Providing quantitative support and scenario analysis for Tokyo to achieve its carbon neutrality goal by 2050.

2. Materials and Methods

2.1. Theoretical Foundation

This study is based on the theoretical foundation of the Sustainable Development Goals (SDGs), urban energy transition and socio-technical transition theories, and the Avoid-Shift-Improve (ASI) framework widely used in the field of sustainable transportation. It combines statistical forecasting models with the LEAP model to provide conceptual support for assessing the emission reduction potential of various policy initiatives for passenger transport and road freight in Tokyo.

First, the Sustainable Development Goals (SDGs) and national carbon neutrality targets provide a normative direction for this study. SDG 11.2 calls for safe, economical, accessible, and sustainable transport systems for all by 2030, while SDG 7 and SDG 13 emphasize the importance of clean energy and climate action [17]. This means that urban transport decarbonization must not only be assessed based on CO₂ emission reduction potential but also ensure the service quality and accessibility of its transport system. By simulating CO₂ emissions from buses, railways, taxis, and urban freight under different policy combinations, this study quantifies and assesses the requirements for SDGs and carbon neutrality targets at the urban scale.

Secondly, theories of urban energy transition and social-technological transition are based on policy, institutional, and technological levels [18]. The multi-level perspective (MLP) views innovative technologies such as electric vehicles and hydrogen fuel cell vehicles as “niches,” interacting with systems comprising traditional energy transportation technologies and existing transportation networks [19]. From this perspective, this paper, based on the LEAP scenario design, reconstructs Tokyo’s transportation-energy system through different combinations of upstream decarbonization of electricity and hydrogen, promotion of zero-emission vehicles (ZEVs), and energy efficiency improvements.

Third, the Avoid–Shift–Improve (ASI) framework provides an analytical approach for sustainable transportation strategies [20,21]. “Avoid” refers to reducing unnecessary travel and freight demand; “Shift” encourages a shift to low-carbon modes such as public transport; and “Improve” focuses on reducing energy consumption and emissions per unit of transport services through cleaner vehicles and fuels. In this paper, ASI is primarily used to organize and interpret different policy levers, while quantitative modeling focuses on two mechanism dimensions that can be consistently characterized within the LEAP accounting framework: transformation and improvement. Specifically, Scenario A represents the replacement and penetration of zero-emission vehicles on the end-use side; Scenario B represents the decarbonization process of electricity and hydrogen energy in the upstream and downstream supply chains; and Scenario C represents the decline in the energy intensity of existing fossil fuel vehicles on the demand side. In contrast, explicit “Avoid” strategies such as demand management or compact urban planning rely more on behavioral and spatial responses, but it is difficult to generate quantitative parameters that can be consistently extrapolated over the long term and to conduct simulations under existing data conditions. Therefore, this paper does not include them in quantitative simulations but instead conducts qualitative analysis within the policy discussion.

By combining traffic demand forecasting models with the LEAP model, transportation activity is converted into CO₂ emissions for policy scenario analysis. Based on this, the “prediction–simulation–explanation” approach adopted in this paper is methodologically supported by the theoretical framework (Figure 2).

2.2. Data Source and Variable Settings

This section defines the dataset and variables used in the integrated forecasting–LEAP framework. Inputs include socioeconomic drivers (GDP, population, employment, road length, vehicle stocks) and activity/technology variables (turnover, EI, energy shares, EF). These inputs are subsequently used to forecast activity levels (Section 2.3) and to parameterize LEAP scenarios (Section 2.4).

Because the Japanese government’s official statistics on some socio-economic and transportation data for 2022–2024 are not complete and consistent, in order to ensure the consistency of the data in this study, 2021 is used as the base year, and the data for 2022 and beyond are all predicted values derived from statistical models, which are used as input and calculation for LEAP.

The data system of this study integrates official statistics, industry reports and corporate technical information, covering multiple dimensions such as socioeconomic development, traffic operation characteristics, energy use and vehicle technical parameters. The time span is 2001–2021, and the spatial scope is the urban transportation system of the Tokyo metropolitan area.

The main variables are divided into three categories:

• Socioeconomic variables (Data source: Tokyo Metropolitan Statistical Yearbook) [22]: GDP, population, employment, road length, vehicle stock.
• Transport activity variables: Annual turnover of railway, bus, taxi, ordinary truck and minivan.
• Energy and technology variables: Energy intensity (EI), energy structure (energy share), emission factor (EF), etc.

2.3. Variable Prediction Models

This section provides exogenous forecasts of socioeconomic factors and traffic turnover rates for the period 2022–2050. Time series models generate trajectories for road length, employed population, and truck stock, while regression/generalized linear models transform socioeconomic factors into turnover rates for each mode of transportation. The resulting activity paths are fixed inputs to LEAP; scenario differences are introduced only through assumptions about energy structure, energy efficiency and emission factors (Section 2.4).

In this research, population and GDP projections were sourced from official forecasts [23,24], but other socioeconomic variables and transport activity variables lacked official projections. Therefore, it was necessary to forecast these variables through modeling.

To quantify the impact of socioeconomic factors on transportation demand, a multi-model forecasting system was constructed (implemented in R, Version 4.5.0; R Core Team, R Foundation for Statistical Computing, Vienna, Austria; using RStudio IDE; Posit Software, PBC, Boston, MA, USA):

• Time series models were used to predict socioeconomic variables (Employment, road length and vehicle stock).
• Multiple regression models (MLR) and generalized linear models (GLM) were used to predict the activity levels (turnover) of various transportation modes.

2.3.1. Time Series Model

To enhance the robustness of forecasts for exogenous macroeconomic drivers, this study employs the Auto-ARIMA automatic order identification method (Hyndman–Khandakar procedure) to model and forecast annually (2001–2021 training, 2022–2050 forecast) for employment, road length, ordinary truck stock, and minivan stock.

The general expression of the model is:

$$\Phi \left(B\right){\left(1-B\right)}_{yt}^d=C+\Theta \left(B\right){\epsilon }_t$$

(1)

$$\mathrm{B} \mathrm{is} \mathrm{the} \mathrm{lag} \mathrm{operator}{ B}_{yt}=y_{\left(t-1\right)}$$

(2)

$$\Phi \left(B\right)=1-\varphi \text{\_}1 B-\dots -\varphi \text{\_}P B^P$$

(3)

$$\Theta \left(\beta \right)=1+{\theta }_{1}B+\dots +{\theta }_q{B}^q$$

(4)

${\epsilon }_t$ is the white noise disturbance term.

When d = 0, the constant term c represents the mean; when d = 1, c corresponds to the drift term, representing the linear change in the mean over time.

To verify the short-term extrapolation performance, a hold-out method was used. Specifically, data from 2001–2018 or 2013–2019 were used for training, and data from 2019–2021 or 2020–2021 were used for testing. Accuracy, stationarity, and residual tests were then performed to ensure that the prediction error was controlled within 10%. Data uncertainty was addressed by generating 80–95% confidence intervals.

2.3.2. Multiple Regression Models

To link socioeconomic development with the level of transportation activity, a multiple regression model for transportation turnover rate was established based on historical data from 2001 to 2021. The model aims to maintain the interpretability and attributability of scenario comparisons by inputting passenger/freight turnover as an exogenous driver into the LEAP demand side, comparing the impact of mechanisms such as energy structure (S), energy intensity (I), and upstream supply transformation (T) on emissions under a unified demand trajectory. These forecasts are then used as inputs to the LEAP model.

For the turnover of railway, bus, ordinary truck and minivan, a multiple linear regression model (MLR) was used to predict the turnover of railway, bus, ordinary truck and minivan. A generalized linear model (GLM) predicted the turnover of taxis. If a multiple linear regression model was used to predict the turnover of taxis, the turnover may become negative after a certain year. Therefore, a different model was used for taxis to ensure the accuracy of the prediction.

The general expression of MLR is:

$$A_{i,t}={\beta }_0+{\beta }_1{GDP}_t+{\beta }_2{Population}_t+{\beta }_3{Employment}_t+{\beta }_4{RoadLength}_t+{\epsilon }_t$$

(5)

where Ai,t is the annual turnover of the i-t mode of transportation, εt is the residual term.

The general expression of GLM is:

$$Ataxi,t\mid Xt\sim Gamma(\mu t,\varphi )$$

(6)

$$\log \left(\mu t\right)=\eta t=\alpha 0+\alpha 1GDPt+\alpha 2Populationt+\alpha 3Employmentt+\alpha 4RoadLengtht$$

(7)

where μt = E (Ataxi,t∣Xt) is the conditional mean, ϕ is the dispersion parameter, and ηt is the linear predictor. In other words, we assume a Gamma distribution for the response variable Ataxi,t with a log-link function.

For the MLR model, the accuracy is evaluated by adjusting R-square, R-square, and p-value. For the GLM model, the prediction accuracy is evaluated by calculating MAE, MAPE, and RMSE. Ensure the accuracy of the prediction.

2.4. LEAP Model Construction and Scenario Design

This section inputs the forecast results into the LEAP model (LEAP, Version 2024.5.0.9; Stockholm Environment Institute, Somerville, MA, USA) and defines the scenario matrix. It also specifies the system boundaries, computational logic, and policy levers for the scenario sequences (A/B/C and their combinations). These scenario settings determine the emissions results reported in Section 3.2.

2.4.1. System Boundary

(1) Spatial Scope: Tokyo Metropolitan Area.

(2) Sectors: railways, buses, taxis, ordinary trucks, and minivans.

(3) Time Scope: 2021 is the base year, and the forecast period is 2022–2050.

(4) Modeling boundary: exogenous-driver scenario framework.

2.4.2. Calculation Formula

The LEAP model’s emissions calculations are based on the formula:

$$E=A\times EI\times EF$$

(8)

where

E: Carbon dioxide emissions (unit: kg-CO₂);

A: Turnover (thousand person-km or thousand ton-km);

EI: Energy intensity per unit turnover;

EF: Emission factor per unit energy.

2.4.3. Scenario Design

The scenario design followed a hierarchical structure of “baseline-single policy-synergy,” aiming to evaluate the energy transformation path and emission reduction potential of Tokyo’s public transportation sector under different policy combinations. All scenarios shared the same socioeconomic projections and activity level inputs, differing only in their assumptions about technology, energy mix and emission factors. The core objective of this study is to identify and compare the “objective exogenous mechanisms” in transportation system emissions reduction that can be clearly characterized by scenario settings, namely the relative leverage and synergistic relationship between energy structure optimization (S), energy intensity improvement (I), and upstream supply transformation (T) on emissions changes. Based on this research objective, this paper adopts an exogenously driven scenario design: traffic activity levels are kept consistent as the exogenous input, thus allowing emissions differences between different scenarios to be primarily attributed to changes in the S–I–T mechanism setting. In other words, this study’s research question, scenario design, and LEAP simulation caliber are consistent, and the results are more suitable for mechanism identification and scenario comparison, rather than end-to-end prediction of actual policy effects including behavioral responses.

(1)

Baseline scenario (Business-as-Usual, BAU)

This study first defines a baseline scenario (BAU) as a reference: only key socioeconomic drivers evolve over time, while other parameters such as energy structure, energy intensity, and emissions factors remain at 2021 levels. It is important to note that the BAU scenario is not intended to depict the “most likely realistic policy continuation path,” but rather serves as a diagnostic benchmark to support clearer and more attributable comparisons of scenario levers A, B, and C. By keeping technology and supply-side parameters constant, differences between scenarios can be more directly attributed to changes in assumptions introduced by each lever, thereby improving the interpretability of the results and helping to identify dominant drivers and potential synergistic effects.

Compared to a policy-consistent baseline setting, this BAU may make relative differences more significant; however, this is a trade-off made to improve attribution clarity and comparative consistency, ensuring that scenario differences reflect the mechanisms of interest in the study. Therefore, the main role of the BAU scenario is to provide a stable and consistent mechanistic comparative reference for other policy scenarios.

(2)

Energy structure optimization scenario (A)

This scenario is set based on a series of policies and strategic plans designated by the Tokyo Metropolitan Government and the Japanese government, aiming to explore the substitution role of zero-emission vehicles (ZEVs, including electric vehicles and hydrogen fuel cell vehicles) in buses, taxis, and freight vehicles. A policy-driven ZEV proportion model was constructed using some of the annual targets set by the government [25,26,27] and vehicle age [28].

To effectively evaluate different ZEV development paths and amplify the emission reduction differences between them, we set up three sub-scenarios: A1 (hydrogen-first scenario, i.e., all ZEVs are hydrogen-powered vehicles), A2 (electricity-first scenario, i.e., all ZEVs are electric vehicles), and A3 (hybrid scenario, i.e., ZEVs consist of 50% hydrogen-powered vehicles and 50% electric vehicles). This design helps to determine the optimal ZEV deployment strategy.

Furthermore, Scenario A assumes that urban passenger and freight transport turnover remains at least at baseline levels. Therefore, ZEV replacement aims to improve environmental performance while maintaining or even enhancing service quality (e.g., quieter interiors and enhanced passenger experience with smart features), aligning with SDGs Objectives 7 and 11. In other words, this scenario focuses on how different ZEV replacement pathways can achieve urban transport decarbonization without compromising the availability and reliability of public transport and freight services.

(3)

Emission factors optimization scenario (B)

This scenario corresponds to the future development path of green electricity and green hydrogen, aiming to assess the impact of upstream energy purification on carbon emission reduction in the transportation sector. Based on energy development plans published by government departments, target values were set for several time nodes, including 2030, 2040, and 2050, and linear interpolation was used to simulate the decreasing trend of hydrogen and electricity emission factors.

According to MOEJ/NIES and REI data [29,30,31], this scenario assumes that by 2030 the emission factors will fall to 0.37 kg-CO₂/kWh (electricity) and 1.50 kg-CO₂/kg-H₂ (hydrogen). In 2040, the emission factor for hydrogen will continue to decrease to 0.60 kg-CO₂/kg-H₂. By 2050, the emission factors for both electricity and hydrogen will approach zero. This scenario aims to assess the role of cleaner energy systems in promoting carbon reduction in the transportation sector.

Since the scenario is designed to decarbonize the upstream energy system without restricting transportation activities, Scenario B meets the requirements of SDG 7 by increasing the proportion of green electricity and green hydrogen. At the same time, the scenario also ensures that emissions are significantly reduced without affecting the availability, affordability and reliability of transportation services, supporting SDG 11 and 13.

(4)

Energy intensity optimization scenario (C)

The purpose of this scenario is to assess the impact of declining energy intensity of traditional fossil fuel vehicles on carbon emissions in the transportation sector, driven by future policies and technologies. Therefore, this scenario will calculate the average annual rate of decline in unit energy consumption of fossil fuel vehicles from 2022 to 2050, based on the emission reduction targets set by the Ministry of Land, Infrastructure, Transport and Tourism (MLIT) [32], and use this as input data for LEAP.

Scenario C implies that emissions reductions are achieved through improved efficiency rather than service cuts, thereby helping to control operating costs and support affordable and reliable services. This aligns SDG 7 and 11 on energy efficiency, which aim to improve resource utilization efficiency in the transport sector while maintaining service quality.

(5)

Scenario combination and integrated analysis design

To comprehensively assess the impact of various policy pathways on carbon emissions from Tokyo’s public transportation system, this study combines individual policies to create multiple scenario combinations. These combined scenarios represent the synergistic effects among various policies, demonstrating the emission reduction potential of policy synergy and helping to identify key drivers of deep decarbonization.

By comparing single-factor and multi-factor scenarios, synergies are quantified, dominant mitigation drivers are identified, and net-zero feasibility is assessed, thereby informing policy priorities.

On the other hand, the projected traffic volumes in various scenario combinations will follow socioeconomic trends, rather than deliberately reducing traffic volumes to lower carbon emissions. Therefore, all scenario combinations are simulated under the premise of maintaining or improving accessibility and service quality and reconcile the climate mitigation goals under SDG 13 with the goals of SDG 7 and 11 regarding clean energy and inclusive, high-quality urban transport services.

All scenario combinations are shown in the

Table 1. Overview of scenario combinations.

Type	Scenario Codes
Single-Factor	A1, A2, A3, B, C
Two-Factor	A1B, A2B, A3B, A1C, A2C, A3C
Multi-Factor	A1BC, A2BC, A3BC

:

2.4.4. Sensitivity Analysis Settings

To quantify the impact of uncertainties in upstream decarbonization assumptions on emission pathways, this paper conducts a single-factor sensitivity analysis, applying ±20% perturbations to the emission factors for electricity and hydrogen production from 2022 to 2049 (fixed at 0 in 2050 to maintain the net-zero energy supply endpoint constraint). Given the large number of scenario combinations studied, and to avoid information overload in the Appendix E, the sensitivity analysis selects representative scenarios covering different policy combinations. Since the emission factors for electricity and hydrogen converge to 0 in 2050 as set, this paper uses emissions in 2040 (E_2040) and cumulative emissions from 2022 to 2050 (CumE_2022-2050) as the core indicators for the sensitivity analysis. Specific calculations and results are summarized in Appendix E.

3. Results

3.1. Socioeconomic Factors and Traffic Turnover Forecast Results

This section presents the prediction results of socioeconomic factors such as road length, employment, and truck stock using time series models, as well as the prediction results of turnover for various modes of transportation using regression models based on socioeconomic factors.

3.1.1. Forecasting Results of Time Series Models

The following figure shows the changing trends of employment, road length, ordinary truck and minivan stock from 2022 to 2050, using the time series model forecast. The accuracy, stationarity, and error diagnostic results for each ARIMA model are shown in Appendix A.7.

(1)

Employment

According to time series model projections (Figure 3), the number of employed people in Tokyo will continue to grow from 2021 to 2050, increasing from 8,146,000 in 2021 to 10,868,000 in 2050 (see Appendix A.4), with an average annual growth rate of approximately 1%. Stationarity tests indicate that the original series is non-stationary, but after first differencing, it can be considered basically stationary (see Appendix A.7). Model accuracy tests show that its error level on the test set is low, exhibiting good short-term fit performance; the residual diagnosis is generally acceptable, showing no significant systematic bias.

(2)

Road length

The road length prediction model was validated using the hold-out method: the training set was 2013–2019, and the test set was 2020–2021 (see Appendix A.7). Validation results show that the model has high prediction accuracy on the test set, indicating good short-term extrapolation performance (see Appendix A.7). Stationarity tests and residual diagnosis results show that the model specification is generally reasonable, and the residuals have no obvious systematic structure (see Appendix A.7). Subsequently, the model was re-estimated using the full sample from 2013–2021, and annual forecasts were made for 2022–2050 (see Appendix A.7).

The forecast results (Figure 4) show that the road length in Tokyo will slowly increase from 11,998,427 m in 2021 to 12,052,658 m in 2050 (see Appendix A.3), with an average annual growth rate of approximately 0.2%. This overall trend shows long-term low-speed growth, indicating that road construction is approaching saturation.

(3)

Ordinary truck stock

The model for predicting the stock of ordinary trucks was validated using the hold-out method: the training set was 2001–2018, and the test set was 2019–2021 (see Appendix A.7). Validation results show that the model has high prediction accuracy on the test set and low short-term prediction error (see Appendix A.7). The stationarity test and residual diagnosis results are generally acceptable, indicating that the model specification can well characterize the sequence features (see Appendix A.7). Subsequently, the model was re-estimated using the full sample from 2001–2021, and annual forecasts were made for 2022–2050 (see Appendix A.7).

The forecast results (Figure 5) show that the stock of ordinary trucks will be 86,075 in 2022, and is projected to decrease to 77,477 by 2030, and further decrease to 55,247 by 2050 (see Appendix A.5). The overall trend is a continuous and slow decline, with the rate of decline slowing down after 2030.

(4)

Minivan stock

The minivan stock prediction model was validated using the hold-out method: the training set was 2001–2018, and the test set was 2019–2021 (see Appendix A.7). Validation results show that the model has high prediction accuracy on the test set and low short-term prediction error (see Appendix A.7). The stationarity test and residual diagnosis results are generally acceptable, indicating that the model specification can well characterize the sequence features (see Appendix A.7). Subsequently, the model was re-estimated using the full sample from 2001–2021, and annual predictions were made for 2022–2050 (see Appendix A.7).

The prediction results (Figure 6) show that the stock of minivans in operation will be 169,480 in 2022, and is projected to decrease to 155,280 by 2030 and further decrease to 119,780 by 2050 (see Appendix A.6). The overall trend is one of continuous decline, and the rate of decline is gradually slowing down. By 2050, the number of small trucks will be reduced by about 30% compared to 2021, reflecting the long-term trend of traditional fuel-powered small trucks existing in the market.

3.1.2. Forecasting Results of Regression Models

After obtaining the prediction results of socioeconomic variables, this study further used MLR and GLM to estimate the activity level of each transportation mode from 2022 to 2050 (i.e., transport turnover, unit: 1000 person·km or ton·km).

Table 2. Regression equations and model accuracy evaluation for public transport turnover in Tokyo.

Regression Model	Transportation Mode	Regression Equation	Accuracy Testing
			Adjusted R2	R2	p-Value
MLR	Railway	Turnover = 2.045 × 107 + 647.3 × GDP	0.9890	0.9900	$3.06\times {10}^{-6}$
	Bus	Turnover = 73,070,000 + 70.54 × GDP − 7.069 × Road_length + 1562 × Employment	0.8466	0.8722	$6.078\times {10}^{-7}$
	Ordinary truck	Turnover = −3,605,000 + 86.96 × Stock	0.7004	0.7154	$1.37\times {10}^{-6}$
	Minivan	Turnover = −136,800 + 1.378 × Stock	0.8901	0.8956	$9.07\times {10}^{-11}$
			Accuracy Testing
			MAE	MAPE	RMSE
GLM	Taxi	Turnover = exp(19.99 − 8.072 × 10−7 × Population + 1.398 × 10−5 × GDP)	168,618.01	10.99%	205,007.77

shows the forecasting regression equations for various modes of transportation and the corresponding accuracy test results.

Table 3. Multicollinearity and robustness diagnostic results of regression models for various modes of transportation.

Modes	Maximum VIF	DW Value	Ljung–Box p-Value	ADF (p)	Conservative Inference Adjustment	Remark
Rail	6.31	1.86	0.19	0.95	HAC Steady SE	Residual stationarity is generally good
Bus	8.36	1.46	0.22	0.39	Newey–West correction	Slight autocorrelation exists, but the residuals are acceptable.
Taxi	7.28	1.16	0.058	/	Newey–West Remains Stable After Adjustment	The residuals showed slight autocorrelation but did not constitute a spurious regression.
Ordinary truck	1.00	1.16	0.27	/	Log-diff robustness test	Slight autocorrelation but acceptable residual randomness.
Minivan	1.00	2.89	0.09	/	No corrections needed	autocorrelation disappears.

shows the multicollinearity and robustness diagnostic results of the regression models for each mode of transportation.

Based on the equations in Table 2 and Table 3, combined with the prediction results of socio-economic data, the turnover volume of each public transportation department is predicted.

(1)

Railway turnover

Railway turnover and GDP show a robust and significant positive correlation: after controlling for other socioeconomic factors, GDP remains statistically significant, indicating its strong explanatory power for rail transit activity. Diagnostic results show some multicollinearity among the explanatory variables and mild autocorrelation in the residuals; however, after adjusting for inference using the Newey–West (HAC) robust standard error, the core conclusions remain consistent (see Table 3). Therefore, within the forecasting framework using GDP as the primary driver, rail transit turnover is projected to continue rising from 2022 to 2050, indicating that rail transit will continue to serve as a relatively stable growth subsystem within Tokyo’s public transportation system and provide a crucial foundation for the implementation of carbon reduction policies.

(2)

Bus turnover

In the final regression model, public transport turnover is primarily influenced by GDP, employment, and road length. GDP and employment have a positive impact on public transport turnover, indicating that increased economic activity and expanded employment will drive up public transport demand. Conversely, road length exhibits a moderating effect, suggesting that under the constraint of saturated road resources, the marginal driving force of road expansion on public transport turnover is limited, thus moderating the growth trend. Diagnostic results show some collinearity among the explanatory variables and mild autocorrelation in the residuals; however, after correction using the Newey–West (HAC) robust standard error, the main inferences remain consistent (see Table 3). Therefore, within the forecasting framework driven by GDP and employment growth, public transport turnover is projected to show a moderate increase from 2022 to 2050, but the overall growth potential is limited by road resource constraints.

(3)

Taxi turnover

A GLM is adopted instead of a standard MLR because taxi turnover exhibits distributional features and error structures that are not well captured by the classical OLS assumptions. In particular, the variance of turnover tends to change with its level (heteroskedasticity), and the relationship between turnover and socio-economic drivers is more plausibly nonlinear on the original scale. By using an appropriate link function and distribution family, the GLM provides more reliable coefficient inference and prediction performance for taxi turnover, while avoiding unrealistic assumptions such as normally distributed, homoscedastic residuals required by MLR.

Regression results show a significant positive correlation between taxi turnover and GDP, and a weak negative correlation with population-related variables. This reflects that the increasing diversification of per capita travel modes and the substitution effect of shared mobility have, to some extent, suppressed the demand for traditional taxis. Diagnostic results show a certain degree of collinearity among explanatory variables and mild autocorrelation in the residuals; however, after correction using the Newey–West (HAC) robust standard errors, the core inferences remain consistent (see Table 3). Based on socioeconomic projections from 2022 to 2050, the taxi turnover calculated by the GLM model exhibits an evolutionary characteristic of “decline in the early stage—stabilization in the middle stage,” reflecting the phased changes under the combined influence of economic growth and travel mode substitution effects.

(4)

Ordinary truck turnover

The turnover of ordinary trucks is significantly positively correlated with the stock of ordinary trucks in use, indicating that vehicle stock has a strong explanatory power for freight activities. Since the model uses a single independent variable, there is no multicollinearity problem; residual diagnosis shows a possible slight autocorrelation, but the conclusions remain consistent under robustness tests (see Table 3). Therefore, in the 2022–2050 forecast results, as the stock of ordinary trucks in use decreases, the turnover of ordinary trucks also shows a corresponding downward trend.

(5)

Minivan turnover

Minivan turnover and minivan stock show a significant and strong positive correlation, indicating that turnover is highly sensitive (and elastic) to changes in stock. Since the model is primarily driven by a single key explanatory variable, multicollinearity is not a problem; after outlier handling and robustness checks, the residual diagnosis results show that the inferences are stable and dependable (see Table 3). Therefore, in the 2022–2050 forecast, as minivan stock declines year by year, minivan turnover will also show a corresponding downward trend.

3.2. LEAP Model Scenario Analysis Results

The scenario comparison in this section is based on a unified exogenous turnover trajectory. Therefore, the differences mainly reflect the quantified shift/improve and upstream supply transformation mechanisms, rather than avoidance-type demand reduction. Simulation results based on the LEAP model (

Table 4. Comparison of cumulative emission reduction effects under various scenarios.

Scenarios	Emissions in 2050 (kt CO2)	Cumulative Emissions (kt CO2)	2050 Emission Reduction Ratio (VS BAU) (%)	Cumulative Emission Reduction Ratio (VS BAU) (%)
BAU	3044.90	93,296.10	/	/
A1	2824.40	89,963.80	7.24	3.57
A2	2913.20	90,845.20	4.33	2.63
A3	2854.80	90,051.90	6.24	3.48
B	1030.00	68,535.30	66.17	26.54
C	2834.70	88,801.70	6.90	4.82
A1B	680.20	63,627.30	77.66	31.80
A2B	670.80	63,955.20	77.97	31.45
A3B	680.50	63,531.70	77.65	31.90
A1C	2672.50	86,168.30	12.23	7.64
A2C	2761.30	87,049.40	9.31	6.70
A3C	2696.70	86,214.70	11.44	7.59
A1BC	528.30	59,831.80	82.65	35.87
A2BC	519.00	60,159.40	82.96	35.52
A3BC	522.30	59,694.50	82.85	36.02

) show that the carbon emission pathways of the Tokyo Metropolitan Transportation System between 2022 and 2050 diverge significantly under different policy scenarios. Under the baseline scenario (BAU), emissions in 2050 reach 3044.90 kt CO₂, with cumulative emissions reaching 93,296.10 kt CO₂, demonstrating that deep emission reductions are difficult to achieve without policy support. Among the single-factor scenarios, hydrogen energy priority (A1), electricity priority (A2), hydrogen and electricity parallel development (A3), and energy efficiency optimization (C) achieve smaller emission reductions, with declines of approximately 4.00–8.00% by 2050. Emission factor optimization (B) performs the best, with emissions dropping to 1030.00 kt CO₂ in 2050, a 66.17% reduction compared to the BAU scenario. Scenario A and Scenario B form a two-factor scenario (A × B), which offers more significant emission reduction benefits compared to other single-factor and two-factor scenarios, reducing emissions by approximately 78% by 2050 compared to the BAU scenario. The scenario with the greatest emission reduction potential is the three-factor scenario (A × B × C), which reduces emissions to 530.00 kt CO₂ by 2050, representing a reduction of approximately 83.0% compared to the BAU scenario. It is important to emphasize that all scenario comparisons in this section are based on the same exogenous turnover trajectory. Therefore, the differences in scenarios are mainly attributed to the differences in the setting of energy structure (S), energy intensity (I), and upstream supply transformation (T).

The “2050 emission reduction ratio (vs BAU)” is calculated as “Emissions in 2050” and “Cumulative emissions” are expressed in kilotons of CO₂ (kt CO₂). The “2050 emission reduction ratio (vs BAU)” is calculated as (E_BAU,2050 − E_scenario,2050)/E_BAU,2050 × 100.

The “Cumulative emission reduction ratio (vs BAU)” is calculated as (CumE_BAU − CumE_scenario)/CumE_BAU × 100, where cumulative emissions refer to the sum of emissions from 2022 to 2050.

All values are rounded to two decimal places.

3.2.1. Comparative Analysis of Single-Factor Scenarios

Figure 7 and Figure 8 show that under the baseline scenario (BAU), carbon emissions from Tokyo’s transportation sector remain essentially stable up to 2050, while the emission reduction effects under different single-policy scenarios vary significantly. The emission factor optimization scenario (B) achieves the most significant emission reductions, with a reduction of over 66.00% compared to BAU in 2050, demonstrating that upstream energy decarbonization is key to achieving deep emission reductions. The hydrogen and electrification pathway (A series) has a lower emission reduction efficiency, with A1, A2, and A3 decreasing by approximately 7.24%, 4.33%, and 6.24%, respectively. Due to the high proportion of fossil power and gray hydrogen, the carbon reduction effect of ZEV replacement in the public transportation sector is very limited. The energy efficiency optimization scenario (C) is effective in the early stages, but its long-term impact is limited, with a reduction of only approximately 6.90% compared to BAU in 2050. Overall, it is difficult to achieve zero carbon emissions in public transportation with a single policy driver.

3.2.2. Comparative Analysis of Multi-Factor Scenarios

Figure 9 illustrates the carbon emission pathways for the public transportation sector under a multi-factor policy stacking scenario. The results show that when hydrogen, electrification, and upstream energy decarbonization (A × B) are implemented together, emissions decline significantly, reaching about 78.00% reduction by 2050 compared to the baseline scenario (BAU). Adding energy efficiency optimization measures (A × B × C) achieves maximum emissions reduction, with emissions falling below 530.00 kt CO₂ in 2050 (about 83% below BAU), with cumulative reductions of about 33,500 kt CO₂. This demonstrates significant synergy between upstream clean energy supply and end-use technology substitution. The former expands the scope of emissions reductions by reducing emission factors, while the latter reduces final energy demand through energy substitution and efficiency improvements. Implementing these policies in combination would allow for sustained emissions reductions by 2050 and would be superior to any single policy path in bringing Tokyo closer to carbon neutrality.

3.2.3. Emission Reduction Characteristics by Sector

Table 5. CO₂ emission changes in various transport sectors under the A1BC scenario (2021–2050).

Sectors	CO2 Emissions (kt)
	2021	2025	2030	2035	2040	2045	2050	Cumulative Emission
Railway	1292.50	1680.50	1565.50	1202.30	820.90	418.60	0	31,574.60
Bus	188.10	444.90	439.80	390.40	354.30	344.30	338.90	11,539.00
Taxi	10.10	8.90	4.90	3.00	2.00	1.50	0.10	127.60
Ordinary truck	769.10	749.70	602.90	464.10	348.00	250.40	165.80	14,245.40
Minivan	112.80	137.30	103.20	72.10	50.80	34.90	23.50	2345.30
Total	2372.60	3021.30	2716.30	2131.90	1576.00	1049.70	528.30	59,831.90

shows the projected annual CO₂ emissions changes for various modes of transportation in Tokyo’s public transportation system from 2025 to 2050, with 2021 as the base year, under the A1BC scenario, and the projected cumulative emissions from 2021 to 2050. Since 2025, carbon emissions from all sectors have shown a continuous downward trend, but there are significant differences in the magnitude and potential of emission reductions among different modes of transportation. The railway sector, benefiting from the shift to cleaner electricity, is projected to achieve near-zero emissions by 2050, resulting in the most significant emission reduction. While taxis have low base carbon emissions and a negligible impact on overall carbon emissions, their continued downward trend indicates that ZEV replacement is an effective measure to reduce carbon emissions in the taxi sector. However, the emission reduction in the bus sector is much smaller, with emissions still reaching 338.90 kilotons by 2050, making it a major future source of emissions. Ordinary trucks and minivans will still not approach zero emissions by 2050, indicating that the road freight sector still faces significant challenges in energy transition and technological upgrading.

Furthermore, the total emissions for each sector in Table 5 for 2050 are consistent with the carbon emission projections for the A1BC scenario in Table 4, ensuring the consistency of the results of this study.

Figure 10 compares the sectoral emissions composition under BAU and Scenario A1BC in 2050. The comparison shows that under BAU, the main emission sources are electrification-related sectors (especially rail transit) and buses; while under Scenario A1BC, rail transit and taxi emissions can be reduced to near zero, with residual emissions significantly concentrated in public transport and road freight. This “structural shift” indicates that when upstream energy systems and end-use substitution advance simultaneously, the key to further approaching net zero emissions depends primarily on the speed of zero-emission vehicle proliferation, the maturity of energy replenishment systems, and the deep decarbonization capabilities of the public transport and freight sectors under operational constraints. Therefore, this paper clarifies the sectoral priority identification as the residual emissions set of the bus and freight sectors that are difficult to eliminate even under the strongest policy mix; this part is more likely to point to the bottlenecks in subsequent policies and key technological breakthroughs.

3.3. Univariate Sensitivity Analysis Results

After applying ±20% perturbations to the emission factors for electricity generation and hydrogen production from 2022 to 2049 (with 0 fixed in 2050), the changes in 2040 emissions (E_2040)** and cumulative emissions (CumE_2022–2050)** for scenarios B, A3B, and A3BC are relatively small: the relative range for E2040 is approximately 2.6–3.6%, and the relative range for CumE is approximately 2.1–2.6% (see Appendix E,

Table A27. Sensitivity of pathway emissions to upstream EF uncertainty (±20%, 2050 EF fixed at zero).

Scenarios	E_2040 Base (kt)	E_2040 Low (kt)	E_2040 High (kt)	Range (kt)	Range (%)	CumE_2022-2050 Base (kt)	CumE_2022-2050 Low (kt)	CumE_2022-2050 High (kt)	Range (kt)	Range (%)
B	1994.95	2024.30	1972.69	51.61	2.59	68,535.29	69,361.95	67,908.40	1453.54	2.12
A3B	1751.15	1783.09	1726.61	56.48	3.23	63,531.73	64,410.75	62,857.62	1553.13	2.44
A3BC	1586.42	1618.36	1561.87	56.48	3.56	59,694.50	60,573.51	59,020.38	1553.13	2.60

Note: Range (kt): The difference between E_Low and E_High. Range (%): The ratio of the difference between E_Low and E_High to E_Base.

). Under the low/baseline/high perturbations, the scenario ranking consistently remains A3BC < A3B < B, and the differences between scenarios are significantly greater than the uncertainty range, indicating that the main conclusion of this paper, “upstream decarbonization and synergistic substitution of electricity and hydrogen can significantly reduce emissions during the pathway period,” is robust within a reasonable range of upstream uncertainty.

4. Discussion and Conclusions

4.1. Timing Optimization of Multi-Policy Coordination

In this study, the BAU scenario was intentionally defined as a counterfactual “diagnostic baseline,” rather than a policy-consistent forecast, to clearly separate the marginal effects of the three measures in scenarios A, B, and C, and to more easily identify the main drivers behind the differences between scenarios. While this deliberately neutral baseline may exaggerate the percentage of emissions reductions, making it higher than a baseline that already incorporates existing policies and incremental improvements, the policy-consistent baseline primarily alters absolute emissions levels (and may narrow the relative gap). The comparative analysis results, particularly the qualitative ranking of the measures, will remain robust because all scenarios follow the same activity trajectory, differing only in the assumed measures.

According to the scenario quantification results (Table 4), implementing demand-side structural adjustments alone (Scenario A series) will contribute only a limited amount to emissions reduction by 2050 (approximately 4.33–7.24%), with a cumulative emissions reduction of only approximately 2.63–3.57%. In contrast, improving upstream emission factors (Scenario B) can achieve approximately 66.17% emissions reduction by 2050, with a cumulative emissions reduction of approximately 26.54%. Further combining technological substitution and energy efficiency improvements, the emissions reduction by 2050 under Scenario A × B can increase to approximately 77.65–77.97%, while Scenario A × B × C can reach approximately 82.65–82.96% (cumulative emissions reduction of approximately 35.52–36.02%). Therefore, a single policy cannot effectively achieve decarbonization, and end-use technology substitution (electric vehicles/fuel cell vehicles), decarbonization of upstream electricity and hydrogen supply, and improved vehicle energy efficiency are not simply additive but exhibit significant complementary and amplifying effects. This is consistent with the conclusions of Working Group III of the Intergovernmental Panel on Climate Change (IPCC) Sixth Assessment Report on the mix of emission reduction policies for transportation [33].

First, the emission factors of electricity and hydrogen fundamentally determine the actual emission reduction benefits of ZEV replacement policies. Based on the IEA’s emission factor database and reports, decarbonizing the power system and hydrogen production system can amplify the emission reduction benefits of ZEV replacement [34,35], based on a power system scenario analysis, pointed out that upstream low-carbon supply and end-use substitution must be designed in a coordinated manner; otherwise, they will fall into a dilemma of high cost or high emissions. This aligns closely with the findings of this study: only when low-carbon electricity and low-carbon hydrogen are feasible at the technological and policy levels can ZEV replacement of public transport and freight vehicles achieve effective carbon reduction benefits.

Secondly, Light-duty vehicles show clear fuel-saving effects through engine and transmission system upgrades and optimization of air resistance [36]; medium and heavy-duty vehicles can significantly reduce fuel consumption by improving thermal efficiency, optimizing transmission/rear axle and aerodynamics, strengthening tire management, and using low rolling resistance tires [37]; on the operational side, eco-driving/instant feedback can save an average of 5–7% fuel [38], and platooning can further reduce aerodynamic drag and fuel consumption under applicable operating conditions [37].

Therefore, the order of policy implementation should be emphasized: first, lock in the timetable and intensity of supply-side clean energy transformation, then accelerate the large-scale substitution of end-use vehicles, and at the same time, use the energy efficiency and operation optimization of fossil fuel vehicles as an amplifier to improve the stability and reliability of emission reduction [39,40]. The innovation of this study lies in its direct comparison of the emission reduction effects of single factors and multi-factor synergy between public transportation and road freight in Tokyo through quantitative scenario analysis, rather than merely relying on normative arguments.

To enhance policy operability, this study further organizes the implementing entities of different policy levers from the perspective of governance boundaries.

Table 6. Governance relevance of policy levers and implementation responsibilities (Tokyo vs. national level).

Policy Leverage	National Level	Tokyo Metropolitan Area	Scenario Correspondence
Decarbonization of power systems	Power structure objectives, market rules, power grid planning and investment, national carbon policy and subsidy framework.	Local renewable energy promotion, green electricity procurement for public buildings/public institutions, demand-side management and demonstration.	B (Electricity Emission Factor)
Decarbonization of Hydrogen Supply	Hydrogen Energy Strategy, Supply Chain Planning, Standards and Certification, National Subsidies/Demonstrations.	Site Layout and Permit Coordination, Demonstration Operations, Public Procurement-Driven.	B (Hydrogen Emission Factor)
Vehicle Technology and Access	Vehicle Regulations and Standards, Fuel Efficiency/Emission Controls, National Purchase Subsidies and Tax System.	Local Subsidies, Government Procurement Priority, Low Emission Zones/Delivery Management.	A (ZEVs replacement)
Infrastructure	National Subsidy Mechanism, Technical Standards, Cross-Regional Trunk Network Planning.	Site Selection, Permitting, Land Use Coordination, Public Station Renovation.	A (ZEVs replacement Feasibility)
Operational efficiency and demand management	Industry guidelines and standards, partial funding support.	Bus route and fleet scheduling optimization, urban logistics organization, congestion/parking management, public transportation guidance.	C (Energy efficiency)

summarizes the main division of responsibilities at the national and Tokyo Metropolitan levels and corresponds to the scenario variables (A/B/C) in this paper to support the discussion in the “Policy Coordination and Implementation Sequence” section.

4.2. Industry and Spatial Differences: Characteristics and Significance of Bus and Freight Emission Reduction in Tokyo

Based on the results of scenario A1BC (Table 5), railways are more likely to approach zero emissions due to their high proportion of clean electricity [41]; taxis have a limited impact on the overall emission reduction in the transportation sector due to their scale and replacement cycle. In contrast, the total emissions in 2050 are approximately 528.3 thousand tons, of which public transportation accounts for approximately 338.9 thousand tons and road freight (ordinary trucks) accounts for approximately 165.8 thousand tons, accounting for most of the remaining emissions. This indicates that in the near-net-zero-emission stage, the focus of emission reduction will be more on the technological substitution and operational optimization of public transportation and road freight systems [26].

(a) Buses and road freight require long operating hours and higher turnover rates, thus placing higher demands on range, refueling efficiency and vehicle availability.

(b) The distribution of charging piles and hydrogen refueling stations, vehicle scheduling, and the design of operating routes all depend on the path; if the initial layout is inadequate, it will be very difficult to adjust later [42].

(c) Buses and freight transport are primarily funded and operated by businesses and local governments, thus requiring a dynamic balance between capital expenditures, operating and maintenance costs, technological upgrades and innovations, and service quality [43].

Therefore, for public transportation, priority should be given to upgrading operating routes, stations, and charging facilities, followed by selecting different vehicle models based on different application scenarios [34,44]. Freight operations should be categorized into three types: trunk line transport, urban delivery, and last-mile delivery, with different truck models adapted to different scenarios and needs [45]. Furthermore, efficiency can be improved through measures such as off-peak charging, time-sharing operation, and optimized scheduling [34].

Furthermore, choosing Tokyo as the target city for this study is of great significance. First, as the core area of Japan, Tokyo occupies a central position in terms of population, GDP, passenger and freight transport. Tokyo’s transportation sector has long been one of the major sources of carbon emissions in the region. Therefore, Tokyo’s transportation sector is very suitable as a research subject for exploring deep emission reduction pathways [46]. Tokyo’s Zero Emission Tokyo Strategy sets a net-zero emission target for 2050 and a zero-emission vehicle orientation for public transport and freight, providing boundaries and standards for scenario design [47]. Third, Tokyo has introduced fuel cell buses, built multi-functional hydrogen refueling stations, and issued related policies to promote the demonstration application of hydrogen energy in public transportation, ports, and logistics [48], making the hydrogen transportation scenario not based on theoretical assumptions but on extrapolation based on existing conditions. Finally, compared to other cities or regions, Tokyo has a more comprehensive data statistics system, which is conducive to building a complete LEAP model; this model can also be applied to other megacities.

4.3. Demand and Substitution: Primarily Structural Adjustment

The regression results of this study confirm that socioeconomic factors such as GDP, population, employment, and road mileage have a significant impact on transportation demand, which is consistent with the conclusions of IPCC AR6 on the relationship between income and travel demand [33]. However, given that total travel volume and freight demand are unlikely to decrease significantly, effective reduction in emission intensity per unit can only be achieved by improving the attractiveness of railway, changing users’ travel choices, and reducing the share of other high-carbon travel modes.

The overall cost of travel (economic cost + time cost + reliability + comfort + accessibility) determines users’ choice of travel mode. Research from the ITF and OECD indicates that improving the overall cost of railway is more effective in enhancing emissions reduction than a single ZEV replacement subsidy policy [49]. Therefore, this argument provides a reference for future research. The research framework can be further improved by incorporating research on customer behavior.

4.4. Contributions and Limitations

Compared to previous similar studies, this study makes several significant improvements in terms of methodology and applicability.

Previous studies on transportation energy transition and emission reduction have mostly adopted top-down scenario assumptions, which have limited characterization of the relationship between socio-economic drivers and transportation demand. Furthermore, they have mostly been conducted on a national scale and lack research on transportation in specific cities.

For example, Meng et al. [11] also used LEAP to conduct a multi-scenario assessment of road traffic in Japan, comparing electrification and hydrogen energy pathways, but the evolution of transportation demand was mainly based on the preset growth rate and did not reflect the impact of socio-economic factors on traffic turnover. Ozawa et al. [50] constructed a carbon neutrality path for Japan by 2050 using an energy system model, highlighting that the decarbonization of the power system is the key to achieving carbon neutrality. However, they did not make detailed predictions about transportation demand, nor did they distinguish the differences in transportation structure among different types of cities. Instead, they focus more on the national energy supply-side layout, often neglecting the fact that traffic turnover is the key variable determining energy consumption and carbon emissions in the transportation sector. Studies such as those by Okuda et al. [51] and Hao et al. [52] have not systematically estimated the statistical relationship between traffic turnover and socioeconomic factors such as GDP, employment, and road length. This makes it impossible for changes in traffic conditions to be fully matched with socioeconomic developments.

This study uses Tokyo, a typical megacity, as a case study and incorporates key socioeconomic factors such as population, GDP, road length, and employment to predict the turnover of rail, bus, taxi, and road freight. A LEAP model is constructed to ensure that changes in traffic demand are based on the natural evolution of socioeconomic factors, rather than subjective assumptions. Furthermore, the emission factors and future trends of hydrogen and electricity are derived from official documents and authoritative research reports. Therefore, compared to previous studies, this research is closer to reality and has higher accuracy and applicability. Different carbon intensity scenarios for electricity and hydrogen are compared within the same urban framework to examine the decisive impact of “whether upstream is truly low carbon” on the emission reduction effects of public transport and freight. Based on these three points, this study more accurately reflects the characteristics of Tokyo’s transportation structure and policy constraints, providing a more explanatory and operational analytical framework for assessing the real emission reduction potential of hydrogen and electrified public transport routes in large cities.

However, this study still has certain limitations, which can be summarized in three aspects. First, exogenous demand trajectory settings and behavioral responses were not included. This study employs an exogenously driven scenario accounting framework, using passenger and freight turnover as exogenous inputs generated by a statistical forecasting model. This aims to more clearly identify the relative roles and synergistic relationships of mechanisms such as energy structure (S), energy intensity (I), and upstream supply transformation (T) on emissions under a unified demand trajectory. It should be noted that price elasticity, mode choice, induced demand/rebound effects, and demand adjustments caused by changes in land use and service supply in the real world may affect the absolute level of emissions in the long-term scenario. Due to the lack of behavioral parameters and spatial coupling modules that can be used for consistent long-term extrapolation, this paper does not explicitly model the endogenous behavioral feedback. Therefore, the results of this paper are more suitable to be understood as a “comparison of conditional emission reduction potential and mechanisms under a given demand trajectory,” rather than an end-to-end prediction of the actual policy effects after including behavioral responses. Second, at the model structure level, the framework of this study is still mainly based on “scenario-driven one-way calculation” and has not yet constructed an endogenous feedback mechanism between policy tools, infrastructure expansion, technology learning effects, and energy price changes. Therefore, it is difficult to characterize the impact mechanism with time-series characteristics, path-dependent characteristics, and dynamic coupling characteristics in the ZEV diffusion process. Third, based on the above limitations, the results of this study are more suitable for comparing the emission reduction potential of different policy and technology paths under the premise of consistent activity level assumptions, providing evidence to support government decision-making on relative advantages, disadvantages, and synergistic directions, rather than being interpreted as a precise prediction of future transportation carbon emissions. Meanwhile, although this study has provided prediction ranges for some exogenous variables and presented the outcome ranges for key scenarios by applying ±20% perturbations to upstream emission factors, it is important to emphasize that the LEAP scenario calculation process remains a deterministic calculation within the framework of this study. Future research will further explore multi-parameter joint uncertainty propagation to characterize the superposition and output distribution characteristics of key parameter uncertainties more systematically within the system.

Furthermore, demand-side “Avoid” measures (such as travel demand management, compact cities, remote work, and logistics integration) mainly influence traffic activity levels through behavioral and spatial structural mechanisms, typically requiring more granular travel/freight demand data and specialized demand models. Given the limitations in the availability of behavioral and spatial data in this study, forcibly using exogenous coefficients to define the demand decline path in LEAP may introduce strong subjectivity and additional uncertainties. Therefore, this paper explicitly defines “Avoid” as an important direction that exceeds the scope of this quantitative analysis and proposes corresponding extended paths for future research.

Future research can be further expanded in three directions. First, building upon the S–I–T mechanism accounting framework of this study, future research could further couple demand response (such as generalized travel cost/price elasticity, mode choice, and demand adjustments caused by changes in service levels) or simplified representations of land use-transport interaction, thereby endogenizing key behavioral feedback and providing emission ranges for “achievement effects.” Furthermore, scenario envelopes or sensitivity analysis could be used to incorporate rebound effects and demand management measures into uncertainty propagation, thus testing the robustness of the conclusions under different demand feedback intensities. Second, the construction progress and feasibility constraints of infrastructure such as charging facilities and hydrogen refueling stations can be more clearly incorporated into the analysis to define the upper limit of the ZEV promotion speed more realistically in public transportation and urban freight sectors. Third, the uncertainties of upstream energy system evolution and cost dynamics can be further systematically addressed, more clearly describing the impact of power structure evolution, hydrogen production path selection, and technological cost changes on emission factors and emission reduction effects; simultaneously, by using a multi-parameter joint uncertainty propagation method to jointly perturb key parameters and characterize the result distribution, a more comprehensive presentation of the possible emission outcome range of key scenarios can be obtained; based on this, this framework can also be extended to other megacities for comparative studies to improve the robustness and generalizability of the conclusions.

4.5. Transferability and Boundary Conditions

Although this paper uses Tokyo as the subject for parameter setting and calibration, some conclusions are transferable from a “mechanism level” perspective: First, the emission reduction benefits of end-use electrification and hydrogenation are highly dependent on upstream supply decarbonization. Therefore, the carbon intensity pathways of electricity and hydrogen supply will significantly constrain the actual emission reduction effects of end-use technology substitution in different cities. Second, under the same demand trajectory, a combined strategy that simultaneously covers supply decarbonization, end-use substitution, and efficiency improvement is often superior to a single lever, reflecting the importance of policy combination and timing.

Furthermore, the calculation results are sensitive to urban contexts. Tokyo has a mature rail and public transport system, high urban density, limited road expansion space, and local characteristics in its fleet structure and usage intensity, resulting in regional differences in upstream energy decarbonization pathways. For cities that rely more on private cars, have longer travel distances, different freight organization, or slower decarbonization in the power sector, the absolute emission reduction and strategy priorities may change.

Therefore, when applying this framework to other megacities, the following parameters should be reparametrized: (1) the mode structure and fleet structure of the baseline year; (2) the activity level and driving variable trajectory; (3) the carbon intensity pathway of electricity and hydrogen supply; and (4) the policy implementation constraints. After the above adaptation is completed, this framework is more suitable for comparative assessments of “mechanism comparison and synergy identification,” and it is not recommended to directly transplant the absolute emission values of Tokyo.