Global trajectories for urban passenger transport decarbonisation: a policy-based modelling approach

Abstract

Understanding the impact of public policy measures on urban transport outcomes is essential for the development of evidence-based transport decarbonisation strategies. This paper presents a modelling framework to estimate the evolution of urban passenger travel and related social externalities for each of the 9234 urban areas in the world. It builds on a hybrid four-step model and systems dynamic modelling approach considering socio-economic projections and forecasts of urban growth, transport supply provision and policy assumptions to estimate transport outcomes for 19 world regions. As a result of its detailed representation of 18 transport modes and 20 policy measures, the model can be used to identify effective levers for reducing carbon emissions while improving urban public health, safety, and access to opportunities. An Increased Ambition policy scenario explores the impact of greater climate ambition on global carbon emissions from transport and other transport externalities relative to the status quo.

Introduction

The urban passenger transport subsector is an essential component of land use, development, and urbanisation studies. According to the United Nations Department of Economic and Social Affairs (UN DESA)¹, the global urban population is expected to grow by 68% between 2015 and 2050. Combined with rapid urban economic growth, this trend will exert intense pressure on urban travel demand, especially in emerging economies. During the same 35-year period, the International Transport Forum (ITF) estimates that global urban passenger travel demand will increase by 75%², including a tripling of demand in Sub-Saharan Africa and a doubling of demand in Southeast Asia. This tremendous growth in demand will substantially impact transport system design and operation, local economic output, and environmental pollution and social well-being. Quantitative modelling of the dynamics of the urban passenger transport sector is critical for better anticipation of worldwide challenges and opportunities, and for enabling the design of efficient policies to achieve sustainability and wellbeing in urban areas.

There is a longstanding practice of creating aspirational narratives and policy scenarios that would achieve international goals for human development and climate change mitigation; many such efforts have been cited as supporting evidence for national and local policy implementation^3,4,5,6,7. When it comes to transport, many policy scenarios focus on improving the fuel efficiency of the vehicle fleet and shifting travel to more climate-friendly travel modes⁸. While decarbonisation is often the focus of scenario planning, Creutzig et al.⁹, and Fu et al.¹⁰, demonstrate that demand-side solutions for transport-related carbon mitigation are typically complementary to a high quality of life on various measures, which require some adaptation of lifestyle and consumption patterns.

There have also been notable efforts to model policy scenarios for the urban transport sector in particular, to inform public policy design, although rarely at the global level. Winkler et al.¹¹, model transport policy pathways to align London’s transport sector with Paris Climate Agreement-compliant carbon budgets. Another recent study explored local-level policy scenarios for decarbonising the transport sector in New York City¹². Cities can vary enormously, however, with respect to their geography, demographics, and the characteristics of their transport systems. Identical policies may have diverging effects in different cities; research has found that the impact of urban transport decarbonisation policies on carbon emissions from transport can vary from high to negligible depending on the city, and welfare impacts can flip from positive to negative¹³. Evaluating the impact of urban transport policies on global initiatives such as SDGs or the Paris Climate Agreement, therefore, benefits from a model with a global scope.

The proposed transport modelling framework aims to quantify trajectories for travel demand, carbon dioxide (CO₂) emissions, and other externalities from urban passenger transport, as well as the mitigation potential of ambitious policy packages. An Increased Ambition (IA) policy scenario is proposed, presenting an alternative pathway in which current decarbonisation efforts are accelerated while incorporating existing economic, technological, political and demand-based constraints. As the modelling framework includes performance indicators related to quality of life, the impact of transport decarbonisation policies on the well-being of urban residents and visitors is also evaluated. These indicators include measures of travel affordability, traffic safety, air pollutants, urban space consumption and transport resilience, among others.

Demand forecasts are generated from the ITF Global Urban Passenger Transport Model (GUM) to evaluate the worldwide impact of the proposed ambitious transport policies. The GUM, a component of the ITF Modelling Framework (PASTA)⁴, combines global coverage of travel occurring within the world’s 9234 urban areas (more precisely, the 9234 macro-functional urban areas defined by Moreno-Monroy et al.¹⁴), with sufficient detail to enable a strategic functional scope. It is intended to be used for assessing travel demand evolution and well-being over time under alternative policy measures and technology development scenarios. PASTA-GUM estimates travel demand at a trip-based disaggregate level, allowing it to capture policies that affect travel generation (e.g. teleworking), trip lengths (e.g. mixed-use development) and the characteristics of transport supply (e.g. the expansion of a public transport network) in each city. Previous strategic urban transport models have been developed at the local, regional and national levels; however, the PASTA-GUM design is innovative in its capacity to model travel in all worldwide urban areas at the trip level. It can differentiate policy impacts by global region and city typology, thus enabling tailored policy prescriptions specific to local urban contexts such as small, dense cities or rapidly growing cities in middle-income countries.

The remainder of this paper describes a bundle of ambitious urban passenger transport policies, how they can be incorporated into a global transport model, and their potential for contributing to sustainable development goals and international carbon emissions targets. Moreover, it demonstrates how a geographically expansive, yet highly detailed modelling framework can provide insights for the combined impact of local public policies that, when aggregated together, help to achieve broad international goals.

Results

The PASTA-GUM model used to evaluate urban passenger transport scenarios (exogenous factors and policy measures) belongs to a broader family of transport models. Like other transport models, PASTA-GUM reflects the impact of demographic, economic and infrastructure evolutions on travel demand dynamics. These dynamics are then converted into pollutant emissions by linking the model to the vehicle emissions profiles produced by the ITF Vehicle Fleet model⁴.

A systems dynamics stock-and-flow formulation for the evolution of urban demographics, land use, and transport supply is linked to traditional four-step model methods, including econometric models for travel generation and discrete choice models for destination and mode choice. This combination of systems dynamics and four-step modelling methods provides a tractable framework for modelling demographic, land use and transport development at the global scale while retaining a detailed representation of travel behaviour. These transport modelling methods are largely derived from concepts described in Ortúzar and Willumsen¹⁵.

In 2017, the ITF introduced an early global urban passenger transport model¹⁶, which was used to support the analysis conducted in the 2017 ITF Transport Outlook¹⁷. PASTA-GUM, inspired by the earlier model, involves substantial improvements, and its potential uses have been expanded to enable more refined policy impact analysis. The main model characteristics, assumptions, and mechanisms for representing future mobility demand and related emissions are described in detail in the Methods section.

An increased ambition scenario

The scenario design process begins with a Business-as-Usual scenario that represents the trajectory of current political commitments. This is then used to highlight the potential for improvement under an Increased Ambition scenario.

The assumptions underlying the policies in the Increased Ambition scenario and their level of implementation are outlined here to facilitate analysis of the results. The approach employed in this paper is not ‘target-based’ but rather proposes a very ambitious yet potentially feasible scenario under stronger political and economic commitments to mitigating climate change¹⁸. It takes an all-of-the-above approach to mitigating emissions, including ambitious technology and demand management measures at the maximum feasible level for each global region, as determined by local experts.

First, it is important to define what a ‘policy scenario’ represents. A scenario can have different meanings that generally depend on which inputs drive the model results. Typically, many policy scenarios are linked to assumptions regarding the expected growth of Gross Domestic Product (GDP), primarily when used to develop national budgets. While socio-economic inputs are essential in the PASTA-GUM model, its primary aim is to assess the impact of policies and technology developments on urban passenger transport demand and related emissions. In this case, the scenario is defined by the degree of implementation of each policy considered by the model.

In the second step, one must determine the set of relevant policies to include in the scenario. The Increased Ambition measures were identified principally via a technical workshop conducted in January and February 2019 comprised of 25 international urban transport experts. Collectively, these experts identified a series of high-impact policy measures that contribute to urban transport decarbonisation. The experts also determined the mechanism for incorporating these measures into the PASTA-GUM modelling framework. This second discussion was critical, as there are many different methods of representing a policy by adjusting input parameters, subject to the constraints of a model that is aggregated at the urban area level. For example, generic measures such as road pricing can take many different forms, such as fixed roadway tolls, dynamic tolling, and distance-based charging. The responses from these experts were merged with the results of a global expert survey and findings from subsequent ITF studies. The result was the definition of 19 policy measures divided among five categories: Infrastructure enhancement, Economic instruments, Regulatory instruments, Stimulation of innovation and development, and Employment trends.

The third step consists of determining the measures’ implementation level for each scenario. While some quantified objectives and commitments are available for particular policies in some regions of the world, large gaps remain. Therefore, the exercise became entirely qualitative. To be able to set the level of implementation of each measure for each of the 19 modelling regions of the model, a survey was distributed to the academic, governmental and private industry expert network of the ITF. This survey was conducted for the first time in 2020 and a second time in 2022 to support the definition of the ITF Transport Outlook 2021¹⁹ and 2023² analyses. A total of 146 and 85 respondents, respectively, filled in the surveys across the world. An additional post-processing standardisation process was conducted to ensure the consistency of the expected level of policy implementations across time periods and regions.

The outcome of the scenario definition process is presented in Table 1. More information on how the scenario measures are translated into changes in the modelling parameters is available in Supplementary Table 1. The change in modelling parameters is generally implemented on a linear basis across the modelling years from the 2022 baseline to 2060, although the initial implementation of certain forward-looking measures does not begin until later years in certain regions (e.g. carbon pricing). Results are presented for the nine aggregated world regions determined by similar GDP per capita levels and relatively homogenous mobility supply and demand:

• 1. Europe—the 27 European Union member states, Ukraine, UK, Switzerland and Türkiye.

• 2. East and Northeast Asia (ENEA)—China, Japan and South Korea.

• 3. Latin America and the Caribbean (LAC)—Mexico, Central America, South America and the Caribbean.

• 4. Middle East and North Africa (MENA)—including Middle Eastern countries, Algeria, Djibouti, Egypt, Libya, Mauritania and Morocco.

• 5. Southeast Asia (SEA).

• 6. Sub-Saharan Africa (SSA).

• 7. South and Southwest Asia (SSWA)—Afghanistan, Bangladesh, Bhutan, India, Iran, the Maldives, Nepal, Pakistan and Sri Lanka.

• 8. Transition Economies and the Pacific (TAP)—Russia, North Korea, Pacific Island states and former Soviet Republics in Central Asia.

• 9. A group of high-income English-speaking countries—USA, Canada, Australia and New Zealand (UCAN).

Table 1 Level of measure implementation for the business-as-usual and increased ambition scenarios in 2060 relative to the 2022 baseline

Under the Increased Ambition scenario, the infrastructure development measures are much more aggressive, especially in emerging economies where public transport networks are expected to triple. Infrastructure measures are less pronounced in regions with large existing networks, such as Europe, ENEA and UCAN, where technology adoption and demand management are the primary drivers of decarbonisation. Public transport service improvement is nearly homogeneous across regions. However, considering that networks and services are already much higher in richer economies, the result is that a similar relative improvement leads to a better level of service in these regions.

The economic, regulatory and innovation measures show greater ambition in richer economies than in emerging ones. This comes from the perspective that emerging economies are already facing demographic and economic pressures and are less likely to accept the political and financial cost of implementing such measures. It is worth noting that while implementing one measure at its maximum feasible level may be feasible in isolation, setting all the measures at their maximum level may overestimate achievable decarbonisation outcomes somewhat. These policies should therefore be considered an upper bound on decarbonisation potential for urban mobility within the social, political and economic constraints of each region.

Potential for decarbonisation

The Increased Ambition policy scenario produces substantial changes in travel behaviour and externalities relative to the Business-as-Usual scenario. The PASTA-GUM model permits evaluation of the policy impacts across several dimensions: geography, city type, trip length, trip purpose, and traveller demographics.

Figure 1 presents the results disaggregated by geography, showing forecasts for urban passenger transport demand (in passenger-kilometres, or PKMs) and related well-to-wheel CO₂ emissions from transport for four large global regions: Europe and UCAN, East and Northeast Asia, Latin America, and the Rest of the World (MENA, SEA, SSA, SSWA and TAP). The ‘Rest of the World’ combined region is used for convenience in this paper to report and visualise the aggregate results for many rapidly growing and urbanising countries across Africa and Central, South and Southwest Asia. The disaggregate results for each of the nine global modelling regions are available as part of PASTA-GUM outputs, and previous iterations of the results can be explored at itf-oecd.org/itf-transport-outlook-2023. The sensitivity of the results to modelling inputs is discussed in the Supplementary Information section.

Fig. 1: The Increased Ambition scenario enables lower emission levels that are decoupled from passenger demand.

The evolution of passenger demand and CO₂ emission projections from urban passenger transport in four world regions from 2025 to 2060. a East and Northeast Asia, b Europe and the United States, Canada, Australia and New Zealand (UCAN). c Latin America and the Caribbean. d All remaining world regions. Continuous green lines represent well-to-wheel carbon dioxide (CO₂) emissions under the Business-as-Usual scenario, continuous blue lines represent passenger demand expressed in passenger-kilometres (PKM) under the Business-as-Usual scenario, dashed green lines represent well-to-wheel CO₂ emissions under the Increased Ambition scenario, dashed blue lines represent passenger demand expressed in PKM under the Increased Ambition scenario. All results are normalised with the 2025 value equal to 1.

First, the Increased Ambition scenario shows a significant reduction in demand compared with the Business-as-Usual trajectory. The Increased Ambition policy measures reduce passenger demand by 36, 26, 19 and 27% in ENEA, Europe, UCAN, Latin America, and the Rest of the World, respectively, in 2060. While the relative difference between scenarios is highest for ENEA, the absolute reduction in passenger-kilometres for the Rest of the World under the Increased Ambition scenario is greater than that of all other regions combined.

Emissions from the urban passenger transport sector are expected to decrease significantly by 2060 in Europe and UCAN, as well as ENEA, under both scenarios. This outcome may appear rather optimistic, but it is worth noting that urban passenger transport is expected to be the easiest subsector to decarbonise within the overall transport sector²⁰. Urban transport generally has a higher political salience as its externalities are more apparent to citizens than non-urban transport subsectors. Thanks to this higher salience, urban passenger transport is usually under more scrutiny, and more robust demand management measures have a greater political feasibility.

The level of decarbonisation reached by the Increased Ambition scenario is notable, with Europe, UCAN and ENEA approaching net-zero annual CO₂ emissions from urban passenger transport by 2050 and Latin America by 2060. The principal drivers of the remaining emissions in the Global South are limited expectations for the implementation of strong demand management measures, and slow technology improvement caused by the inertia of vehicle fleet turnover in these urban areas, which are often comprised heavily of older, used second-hand vehicles imported from ENEA, Europe and UCAN²¹. While this decarbonisation trajectory is much closer to the Paris Agreement than the Business-as-Usual scenario, it is not entirely sufficient to meet targets set out for the transport sector to limit warming to 1.5 °C^22,23.

The model also allows for differentiation by city size and population. Figure 2 presents the demand and CO₂ emissions forecasts for cities within different population categories. Note that the allocation of cities is determined by the 2022 population and fixed over time to create a consistent basis for evaluation. From 2025 to 2050, demand for transport grows most quickly in cities above 1.5 million inhabitants due to ongoing urbanisation and economic concentration in the largest cities. In addition to population growth, trip distances rise more quickly in large cities as activities and residences are distributed over a larger geographic area. The land use planning and transit-oriented development policy measures included in the Increased Ambition scenario allow residents to access various economic and recreational activities closer to home, mitigating the rise in trip distances and thus overall travel demand.

Fig. 2: The pressure from increasing transport demand in more populated urban areas also provides a higher decarbonisation potential.

The evolution of passenger demand and CO₂ emission projections from urban passenger transport for urban areas in four different population categories (2022 population). a Urban areas with fewer than 500 thousand inhabitants in 2022. b Urban areas with 500 thousand to 1.5 million inhabitants in 2022. c Urban areas with 1.5–5 million inhabitants in 2022. d Urban areas with 5 million inhabitants or more in 2022. Continuous green lines represent well-to-wheel carbon dioxide (CO₂) emissions under the Business-as-Usual scenario, continuous blue lines represent passenger demand expressed in passenger-kilometres (PKM) under the Business-as-Usual scenario, dashed green lines represent well-to-wheel CO₂ emissions under the Increased Ambition scenario, dashed blue lines represent passenger demand expressed in PKM under the Increased Ambition scenario.

While the total volume of the demand is increasing, the modal and trip length composition of the demand is also much different, as displayed in Figs. 3 and 4. The effect of the Increased Ambition measures not only reduces the demand for urban mobility but also relies on achieving it with a shift in behaviour.

Fig. 3: The increased ambition implies a decrease in private vehicles use balanced by a stronger uptake of active and public transport across all urban areas.

Mode splits for urban passenger transport (in passenger-kilometres) in 2060 under the Business-as-Usual and the Increased Ambition scenarios, for urban areas in four different population categories (2022 population). a Urban areas with fewer than 500 thousand inhabitants in 2022. b Urban areas with 500 thousand to 1.5 million inhabitants in 2022. c Urban areas with 1.5–5 million inhabitants in 2022. d Urban areas with 5 million inhabitants or more in 2022. The dark blue slice represents the passenger-kilometre share of active mode trips (walking, cycling, bikesharing, scooter sharing), the green slice represents the share of public transport trips (metro, rail, light rail transit, bus, bus rapid transit), the grey slice represents the share of shared service trips (taxi, ride sharing, taxibus), the light blue slice represents the share of shared vehicle service trips (car sharing, motorcycle sharing), the blue–green slice represents the share of paratransit trips (bus-based paratransit, motorcycle-based paratransit) and the purple slice represents the share of private vehicle trips (car, motorcycle).

Fig. 4: The overall growth in demand is much more visible for longer trips which must benefit from public transport alternatives to achieve an increased ambition.

Passenger transport demand expressed in trillion passenger-kilometres for 2025 and 2060, by trip distance category, under the Increased Ambition scenario. The dark blue segment of the column represents the average demand for active modes (walking, cycling, bikesharing, scooter sharing), the blue-green segment represents the demand for paratransit (bus-based paratransit, motorcycle-based paratransit), the purple segment represents the demand for private vehicle trips (car, motorcycle), the green segment represents the demand for public transport (metro, rail, light rail transit, bus, bus rapid transit), the grey segment represents the demand for shared services (taxi, ride sharing, taxibus), and the light blue segment represents the demand for shared vehicle services (car sharing, motorcycle sharing).

Private vehicles represent 51% of all passenger-kilometres in 2060 under the Business-as-Usual policy assumptions, but the share would be reduced to 46% under the Increased Ambition scenario. This decrease is achieved due to a heavier share of active modes, especially among shorter trips. Despite a substantial increase in overall demand for trips under 10 kilometres, the demand for short private vehicle trips declines, especially for trips between 1 and 2.5 km in length. Active transport accounts for the majority of travel from 0 to 1 km and the largest share of passenger-kilometres from 1 to 2.5 km in length by 2060. The shift to active transport under the Increased Ambition policy scenario is most pronounced in cities under 1.5 million inhabitants, where short trips are a greater share of overall travel demand.

Much like active travel for shorter trips, public transport captures a greater share of longer intra-urban trips in the Increased Ambition scenario. Much of the growth in passenger-kilometres for trips over 10 km is captured by public transport, leading to a much higher mode share for public transport, especially in the largest cities where public transport services have greater coverage and frequency. The difference in public transport mode share between scenarios is driven by a combination of ‘pull’ measures (e.g. service improvements, integrated ticketing) and ‘push’ measures (e.g. parking pricing). Many paratransit systems are expected to be converted into regulated shared mobility or public transport services, leading to a decrease in their modal share.

Finally, the impacts of policy measures on travel behaviour can be disaggregated by income percentiles. Figure 5 compares the per-capita urban travel demand by global income percentile, in 2060, between the two policy scenarios. In the Business-as-Usual (BAU) scenario, transport demand, and in particular, demand for private vehicle travel, is strongly correlated with income.

Fig. 5: Reaching the increased ambition requires reducing the individual transport demand especially for higher income households.

Individual urban passenger transport demand, expressed in passenger-kilometres per capita, in 2060, compared between the Business-as-Usual (BAU) scenario and the Increased Ambition (IA) scenario, by income percentile category. The dark blue segment of the column represents the average demand for active modes (walking, cycling, bikesharing, scooter sharing), the blue-green segment represents the demand for paratransit (bus-based paratransit, motorcycle-based paratransit), the purple segment represents the demand for private vehicle trips (car, motorcycle), the green segment represents the demand for public transport (metro, rail, light rail transit, bus, bus rapid transit), the grey segment represents the demand for shared services (taxi, ride sharing, taxibus), and the light blue segment represents the demand for shared vehicle services (car sharing, motorcycle sharing).

The richest 5% of the global population travel 20% more than any other group, and about 50% more than the bottom half of the income distribution. The demand management measures under the Increased Ambition scenario largely restrict the private vehicle use of travellers in the upper half of global income, producing a more even distribution of urban travel demand across income groups. Note that these results do not include inter-urban and international travel, where demand and emissions are much more heavily skewed towards the richest segments of the population due to vast disparities in air travel demand.

Co-benefits of decarbonisation measures

Aside from the major and more traditional output of the model, several performance indicators are computed to measure the impact of passenger transport policies on urban quality of life. The capacity to produce these indicators is derived from the disaggregation of PASTA-GUM results into detailed modal, population, city and road categories, and the many intermediate outputs. These crucial indicators quantify the effect of the Increased Ambition scenario with respect to eight different urban transport externalities or co-benefits, split into three categories. For health and safety, there are indicators for local air pollution and road conflicts. For the effectiveness and efficiency of transport, indicators for access to opportunities, traffic congestion, road space consumed by passenger vehicles, and the availability of modal alternatives are computed. Finally, indicators for affordability and transport exclusion are produced to measure the impact of policies on costs to users. The estimated difference in each of these indicators between the Business-as-Usual and Increased Ambition scenarios in 2060 is presented in Figs. 6, 7 and 8 below for four different global regions.

Fig. 6: The increased ambition almost fully removes local air pollutant emissions. It also significantly improves safety in Latin America and the Caribbean.

a Comparison of local air pollutants emissions from urban passenger transport—black carbon (BC), carbon monoxide (CO), ammonia (NH3), nitrous oxide (NOx), particulate matter (PMse), sulphur dioxide (SO2) and volatile organic compounds (VOC)—expressed in kg of emissions per square kilometre in 2060 between scenarios, by world region. b Comparison of the safety indicator expressed in number of conflicts per billion passenger-kilometres in 2060 between scenarios, by world region. The dark blue bar represents the value for the Business-as-Usual scenario, and the green bar represents the value for the Increased Ambition scenario. The world regions are grouped: East and Northeast Asia (ENEA), Europe, the United States, Canada, Australia and New Zealand (Europe + UCAN), Latin America and the Caribbean (LAC), and the Rest of the World.

Fig. 7: An increased ambition tackling congestion and increasing resilience worldwide while slightly improving connectivity and reducing vehicle space consumption.

a Comparison of the connectivity indicator expressed in average travel time to travel the average radius of a city in 2060, between scenarios, by world region. b Comparison of the modal resilience indicator expressed in percentage of trips replaceable by public transport in 2060, between scenarios, by world region. c Comparison of the congestion indicator expressed in percentage of consumed road capacity in 2060 between scenarios, by world region. d Comparison of the vehicle space consumption indicator expressed in square metres of vehicles per thousand passenger-kilometres (PKM) in 2060, between scenarios, by world region. The dark blue bar represents the value for the Business-as-Usual scenario, the green bar represents the value for the Increased Ambition scenario. The world region are grouped: East and Northeast Asia (ENEA), Europe, the United States, Canada, Australia and New Zealand (Europe + UCAN), Latin America and the Caribbean (LAC), rest of the world.

Fig. 8: The combination of ‘push’ measures reducing public transport and active mode trip costs, and ‘pull’ measures penalising private vehicle trip costs are essential for the Increased Ambition scenario but generate modal exclusion.

a Comparison of the trip cost for public transport and active modes indicator expressed in average trip cost expressed in 2022 USD, in 2060, between scenarios, by world region. b Comparison of the trip cost for private vehicles indicator expressed in average trip cost expressed in 2022 USD, in 2060, between scenarios, by world region. c Comparison of the mode choice exclusion expressed in the average share of trips that are not affordable for all individuals in 2060, between scenarios, by world region. The dark blue bar represents the value for the Business-as-Usual scenario, the green bar represents the value for the Increased Ambition scenario. The world region are grouped: East and Northeast Asia (ENEA), Europe, the United States, Canada, Australia and New Zealand (Europe + UCAN), Latin America and the Caribbean (LAC), rest of the world.

Local air pollution is measured by the concentration of emissions of the most harmful pollutants from road vehicles. Much like CO₂ emissions, the PASTA-GUM model uses air pollutant emissions profiles per kilometre-travelled for road-based modes, which are combined with the estimated travel by mode to compute the total emissions by mass for each urban area under each scenario. The harmful pollutants measured include black carbon (BC), carbon monoxide (CO), ammonia (NH3), nitrous oxide (NOx), particulate matter (PMse), sulphur dioxide (SO2) and volatile organic compounds (VOC), all of which are known to have adverse effects on public health². Tailpipe emissions profiles for passenger cars, light commercial vehicles, buses and heavy vehicles are provided by the International Council on Clean Transportation (ICCT). Note that tailpipe emissions do not include pollutants generated by wear on vehicle components such as tyres, or pollutant emissions from 2- and 3-wheelers due to a lack of available data.

The model computes the mass and concentration of each pollutant separately, although they are aggregated in Fig. 6 for brevity. Under the Increased Ambition scenario, air pollution from tailpipe emissions is expected to be nearly eliminated in Europe + UCAN, and ENEA by 2060. This trend is driven by an aggressive transition of the urban fleet to electric vehicles, which do not produce tailpipe emissions (although they do emit pollutants from vehicle component wear). In the Rest of the world, despite lower electric vehicle penetration rates than other regions, tailpipe pollutant concentrations in 2060 are expected to be reduced by over 95 per cent relative to the Business-as-Usual scenario. This indicator illustrates how transport policies can support public health outcomes, and similar estimates could be incorporated into policy cost-benefit analyses using estimates of the monetary value of reduced urban air pollution (e.g. Muller and Mendelsohn²⁴).

PASTA-GUM also calculates the estimated exposure to potential conflicts between vulnerable road users (pedestrians and cyclists) and passenger vehicles to evaluate the impact on road safety as a proximate measure of crash risk. The calculation is based on travel volumes for each mode, the difference in speed between modes, and the longitudinal separation/segregation between modes²⁵. The number of conflicts is expected to decline somewhat in ENEA and the Rest of the World under the Increased Ambition scenario due to improved pedestrian and cyclist infrastructure and a mode shift away from passenger cars. The impact on LAC is more dramatic: a decline of over 40% in conflicts per passenger kilometre, representing a substantial improvement in road safety. Much of this trend results from the more ambitious expansion of separated infrastructure for pedestrians and cyclists. However, risks remain high in the auto-dominated cities outside of ENEA, and there is further room for improvement.

Providing convenient access to opportunities is the essential function of an urban passenger transport system. To measure the effect of transport policies on travel times, the model computes the average amount of time needed to travel from the geographic centre of each city to the perimeter; in other words, the maximum amount of time needed to access any destination or opportunity in the city when starting from the centre. To differentiate it from the well-established concept of ‘accessibility’ that also accounts for the proximity of activity locations, this measure is referred to as ‘connectivity’. Ambitious transport policies provide modestly improved passenger connectivity in Europe and UCAN, ENEA and in the Rest of the World by 2060, as shown in Fig. 7. Connectivity in Latin America and the Caribbean, which is better than other regions, is expected to worsen somewhat under the Increased Ambition scenario as the impact of vehicle restriction measures is larger than the benefits of greater public transport investment.

PASTA-GUM estimates the modal resilience of the passenger transport system to road and active transport infrastructure disruptions by computing the share of trips that can be completed by public transport in equal or less time relative to the original mode. This provides an indicator of the availability of public transport as a substitute for private vehicles and active travel, which is a proxy for modal and social robustness. In all regions, the added investment in public transport infrastructure, services, and ticketing under the Increased Ambition scenario results in a greater share of trips that could be replaced by public transport with no increase in travel time. In Europe and UCAN, the share of trips that could be replaced by public transport is more than 30 percentage points greater under the Increased Ambition policies, reaching above 99% of urban trips.

The congestion indicator evaluates the urban transport system’s average utilisation relative to the day’s capacity. High congestion results in travel delays that can hinder economic productivity and well-being. By shifting travel to public transport and active modes, all regions could reduce urban passenger congestion despite very limited increases in road capacity. Demographic changes and a shift to shared and public transport cut congestion by half for the ENEA region, and similar amounts for LAC, Europe and UCAN. In the Rest of the World, rapid population growth in many urban areas partially offsets ambitious policies that encourage more efficient use of the passenger transport network, resulting in a smaller impact.

The vehicle space consumption indicator evaluates the amount of urban area that is consumed by the movement and storage of passenger transport vehicles. Minimising the urban space consumed by transport can create opportunities to expand other land uses such as parks and public services. This indicator combines the space needed for each vehicle to move safely on the roadway and the space needed for parking, normalised by total travel demand. These measures are computed based on the size and speed of the vehicle and the average proportion of the time spent in motion²⁵. In Europe, UCAN and the Rest of the World, policies to decarbonise the transport sector have a limited impact on space consumption, while they produce relatively substantial benefits for LAC and ENEA. These policies therefore represent a valuable opportunity to introduce new open spaces and public services into many dense and congested cities in Latin America, the Caribbean and East Asia.

PASTA-GUM can also estimate each trip’s financial burden to gauge the affordability of travel in each city by mode type. Monetary costs involve fares for shared modes such as public transit and taxis, and maintenance and operating costs for private modes. The cost is then normalised by GDP per capita to measure an average trip’s affordability. Figure 8 shows how transport policies aimed at decarbonisation can also produce a dramatically less expensive public and active transport system in all regions. Public transport (PT) and active modes include traditional public transport (i.e. bus, metro, rail, paratransit) and active modes such as cycling, bikesharing, and walking. The average cost per trip is weighted by demand for each mode. Costs for public transport and active travel in Europe and UCAN are reduced by half relative to the Business-as-Usual scenario. Similar, although less pronounced impacts are expected in other regions, ensuring that sustainable transport remains affordable worldwide.

Affordability of car travel declines in ENEA, LAC, and Europe and UCAN under the Increased Ambition scenario, as escalating road user charges that internalise the public costs of driving are added to the cost of driving. These charges more than compensate for the reduction in vehicle operating costs associated with greater electric vehicle market penetration. Private vehicles include private cars, motorcycles, taxis, ridesharing and car sharing. The effect of the Increased Ambition scenario is most noticeable in ENEA and Europe and UCAN, where the highest road user charges are applied, per Table 1. In the Rest of the World, relatively low road user charges, electric vehicle adoption, and the impact of larger shared fleets, as well as carpooling incentives, contribute to more affordable private vehicle travel.

Finally, the mode choice exclusion indicator represents the percentage of urban trips for which one or more transport mode is not available to the traveller due to the cost. Transport modes are considered to be unaffordable if the monetary cost of using that mode for daily commuting exceeds 15% of the population group’s annual income. For example, if the annual monetary cost of driving to work in one city is 6000 EUR, then driving is considered to be unaffordable for the percentage of the city population whose annual income is below 40,000 EUR.

In almost all cases, these are trips where private transport is considered to be too expensive and, therefore, another mode is selected. These results demonstrate that Increased Ambition policies restrict lower-income groups from using private transport to a greater extent than the Business-as-Usual policy trajectory. Complementary measures such as a minimum mobility guarantee, which has been explored in several European regions²⁶, are needed to alleviate these challenges and ensure that everyone has access to safe, convenient and affordable urban transport. Public revenue from schemes that internalise the social costs of driving can be used to fund the additional transport services that are likely to be necessary to achieve broad mobility guarantees.

An aggregated modal exclusion indicator can obscure the distributional impact of pricing on different income groups within a city, and which transport modes are unaffordable. Figure 9 presents the share of the population in each region that would be required to spend more than 15% of their income on daily commuting by bus and by private car. Bear in mind that this analysis is exclusive to urban areas, where parking prices and tolls for private cars can already be quite expensive. The results indicate that there is little concern related to the affordability of bus transport for commuting. The Increased Ambition scenario, by replacing some informal bus services with regulated services, increases has a minor impact on the affordability of those services in the Rest of the World. For private cars, however, the increase in operating costs due to internalisation of externalities described in Fig. 8 results in a larger share of the population being excluded from daily commuting to the city centre by private car, especially in Europe, UCAN and East and Northeast Asia.

Fig. 9: Following its impact on prices, the improved ambition scenario drives modal exclusion up which carries a potential social risk, especially for car use.

a Comparison of the bus exclusion indicator expressed in the share of population price burdened if using the bus for daily commuting in 2060, between scenarios, by world region. b Comparison of the private car exclusion indicator expressed in the share of population price burdened if using the private car for daily commuting in 2060, between scenarios, by world region. The dark blue bar represents the value for the Business-as-Usual scenario, the green bar represents the value for the Increased Ambition scenario. The world regions are grouped: East and Northeast Asia (ENEA), Europe, the United States, Canada, Australia and New Zealand (Europe + UCAN), Latin America and the Caribbean (LAC), rest of the world.

Discussion

Modelling a sustainable pathway towards a future that combines high well-being with low energy and resource demand is one approach to the complex challenge of defining and evaluating future policy outcomes. Instead of focusing on the impacts exclusively, this research started by proposing a method for defining policy measures and their maximal level of implementation worldwide to build a consistent implementation trajectory. This first step sheds light on the different perspectives towards the potential for robust policy measures based on the different political, social and economic constraints and the existing policy and mobility environments across world regions. The wealthier regions typically exhibit a stronger potential for implementing demand management measures and service improvements as opposed to a stronger potential for infrastructure development in emerging economies.

This Increased Ambition policy scenario was then evaluated using PASTA-GUM. The main output in terms of potential for CO₂ emission reductions, the evolution of the passenger demand volume and of the modal mix was analysed in a second step. The output showed that the scenario was indeed capable of achieving the initial ambitions of the scenario successfully: a significant decrease in emissions in all world regions fuelled by a 35% decrease in total demand volumes compared with the current trajectory, and a reassignment of the modal mix from high private vehicle use towards more use of active, public transport and shared modes, accompanied with a lower paratransit mode share.

In the third step, the analysis focused on the well-being co-benefits of the Increased Ambition scenario that complement the decarbonisation and low-demand objectives. The Increased Ambition scenario showed potential for nearly eliminating local tailpipe pollutant emissions, improving congestion, accessibility, and resilience worldwide. Additional improvements in vehicle space consumption, safety, and affordability can be observed in certain regions, but are not always sufficient to mitigate the expected evolution of urban mobility systems. As a result, these indicators are not always significantly improved under the Increased Ambition scenario.

While this research proved successful in designing and assessing the Increased Ambition scenario based on an ambitious but feasible policy agenda, several questions are raised regarding the context and limitations of this exercise. An immediate limitation in the interpretation of the results is the accuracy of the global scope. While the urban mobility estimates are all calibrated and validated at the world region level, they cannot be as precise as the ones from more disaggregated national or city-based models. This limitation is also true for the evaluation of policy efficiency: if a policy is efficient in the average case, that does not necessarily indicate that it is efficient in every geographical, political, social and economic setting. The current exercise can only advocate for the overall average potential of policy measures, but cannot replace a local feasibility study analysis for implementing targeted policy packages.

Another limitation is that the outputs measure only the direct impacts of policy reforms and infrastructure investments on transport sector outcomes. It does not include the secondary socio-economic benefits that are typically associated with a transition to a more sustainable urban mobility system, such as improved public health and increased labour productivity²⁵. Communicating the wider benefits of sustainable urban mobility transitions has proven to be an effective tool for policymakers to build public support for such measures^26,27,28. However, these secondary impacts typically have complex relationships with many urban characteristics beyond the transport sector (e.g. obesity rates, labour market composition) and are therefore extremely difficult to estimate for all urban areas across the globe with the data currently available.

The interest of policymakers in the current type of policy modelling exercise generally is to support the identification of optimal policy pathways for reaching climate goals with the lowest financial or political cost. As such, there are often requests for presenting the costs of policies in relation to their decarbonisation potential. This approach is challenging because it is very difficult to capture all the political, social and financial costs of a measure, especially as it depends heavily on how, where and when it is implemented. Typically, the development of zero-emission vehicles is often a measure considered to have a strong decarbonisation potential. However, the cost of financing the development of these vehicles is rarely accounted for, as it is largely financed by the private sector, in addition to public subsidies for research and development activities. In the end, it is difficult to state whether the development of zero-emission vehicles is efficient, as the allocation of these private and public funds to other measures is often not evaluated. The transferability of the financing of these funds for zero-emission vehicles to other measures is also arguable, as they are the result of a contextual setting that may not hold for other measures.

With the stipulations above in mind, the approximate annual cost of the public investments proposed in the Increased Ambition scenario was calculated using region-specific infrastructure and vehicle costs. For low-income countries, the required investment in urban transport under the Increased Ambition is estimated at 2.5% of GDP, which is somewhat less than the 2.9% of GDP required under the Business as Usual scenario. When existing infrastructure and services are quite limited, it requires less investment to accommodate sustainable urban mobility than to build the heavy infrastructure required for widespread use of private cars. For lower-middle, upper-middle and high-income countries, the required investment ranges from 1.0 to 1.3% of GDP, which is slightly higher than under the Business as Usual Scenario (0.8–1.0% of GDP).

To another extent, current methods for assessing the potential Increased Ambition scenario include a bias leading to overly optimistic scenarios. It is the authors’ observation that the exercise of assessing the potential level of implementation of each measure individually is relatively effective when considering each measure independently. There is, however, an intuitive bias towards the systematic inclusion of low levels of implementation of each measure in the scenario development phase, when many are perhaps not likely to be implemented at all due to political constraints that persist regardless of the level of implementation. In addition, the exercise of stating how far each measure can be pushed also becomes biased when considering the whole bundle of measures: implementing one measure at its maximum feasible level may be feasible in isolation, but setting all the measures at their maximum level is likely too optimistic. While the current modelling exercise still features this bias, the authors aim to introduce bundle penalties in the future that would slightly decrease the impact of each policy measure when many are implemented at an ambitious level.

Other biases may have been introduced into the scenario development process as an unintended consequence of the timing and distribution of the expert survey. Experts in all global regions responded to the 2020 and 2022 surveys used to set the level of implementations of each measure, but more than half of the survey responses were from European experts, with 17% from Latin America, 14% from Asia and Oceania, 13% from Africa and 4% from North America. While regional expertise was used to set each region’s implementation levels, with fewer numbers of experts, an awareness of the full range of decarbonisation solutions and policy ambition across each region may be lacking.

The timing of the surveys may have resulted in an overly optimistic view of the impact of teleworking. At the time, teleworking rates were very high and non-work travel remained low due to persistent pandemic-related restrictions and concerns. Teleworking has since abated, and studies have found the effect of teleworking on total household trip rates and PKM to be lower than expected across a range of contexts^29,30,31. The authors intend to revisit assumptions about teleworking in future scenarios.

Concerning the difference in decarbonisation potential between transport subsectors, urban passenger transport is generally perceived to higher potential for decarbonisation than the non-urban passenger and the freight subsectors. In addition to being more politically salient, urban passenger transport generally involves shorter travel distances than the other subsectors, which makes a shift from polluting to cleaner modes of transport easier. Finally, the population concentration of urban areas makes shared transport modes more effective as a decarbonisation measure.

Regarding the impact of the different policy measures on the outcome of the Increased Ambition scenario, the current modelling exercise does not measure individual policy impacts and therefore cannot determine which have the greatest contribution towards decarbonisation. Technology improvements have a strong impact on the decarbonisation of private motorised modes, as can be intuited from the very low emission levels despite a significant volume of private vehicle use. Yet, it is important to consider that the demand management measures can reduce total passenger demand, and that private vehicle mode share decreases between 2025 and 2060. While meeting the challenge of decarbonisation, zero-emission vehicles do not cause congestion to disappear or necessarily improve accessibility, for instance. Evidently, replacing fossil fuel vehicles with a zero-emissions vehicle does not meet the standards of a ‘high-with-low’ future that provides high well-being with low energy and material resource consumption while limiting climate change.

To achieve a truly sustainable future, decarbonisation must be seen as an opportunity to find new solutions for adapting behaviour to upgrade the quality of life. While technological advancements are part of the solution, they cannot be the only lever considered, especially as they are deployed more quickly in richer economies and have a slower penetration in emerging economies via second-hand vehicle fleets. All of society, from individuals to policymakers, organisations and governments, must put forward and promote equitable urban mobility demand management measures on which to build the foundations for healthy cities.

Methods

The scenario analysis was conducted with PASTA-GUM. PASTA-GUM is a strategic global model aiming at assessing the medium to long-term trajectories of urban passenger transport demand and related emissions and the impact of different policy pathways on these worldwide. When considered jointly with the non-urban passenger, freight and vehicle fleet global models of the ITF PASTA framework⁴, it enables tracking the success of a set of policies for achieving international climate goals such as the one set in the 2016 Paris Climate Agreement of UNFCCC. The following subsections detail the scope of PASTA-GUM, how it is positioned within the transport policy modelling literature, how it represents the evolution of mobility and how it includes policy measures.

Model scope

Figure 10 provides a brief summary of the scope of PASTA-GUM. The model focuses on the representation of all passenger trips happening within the boundaries of 9234 macro–Functional Urban Areas (mFUAs). These aggregations of the Functional Urban Areas (FUAs) defined by Schiavina et al.³², include all urban areas with over 50,000 inhabitants in the world in 2015. Each of these mFUAs belongs to a country within one of the 19 modelling regions. The modelling regions were defined to represent specific areas where transport behaviour is relatively homogenous and can lead to an identical calibration of sub-models. The geographical unit of the model is the mFUA, with additional descriptors enabling to distinguish between its high-density centre and the peri-urban area.

Fig. 10: PASTA-GUM: a global yet detailed model for assessing the socioeconomic and environmental impact of policy measures on urban passenger transport.

The figure details the scope of the Global Urban passenger Model (GUM) within the ITF Policy Ambitions and Sustainable Transport Assessment (PASTA) modelling framework developed by the International Transport Forum (ITF).

Complementary with the geographical scope, the population considered within the model encompasses the tourist and residential population, the latter being disaggregated between 10 population categories mixing 2 gender categories (female, male) and 5 age categories (0–19 years old, 20–34 years old, 35–54 years old, 55–69 years old, 70 years old and over). This disaggregation enables representation of the different mobility behaviours of each population and analysis of the different impacts of policies on these population segments, especially for equity considerations. Analysis of mobility behaviour by age and gender can be found in previous ITF publications^2,17,19, but it is omitted from this paper for the sake of brevity.

In addition, a third segmentation by 6 trip distance categories (0–1, 1–2.5, 2.5–5, 5–10, 10–20, 20 km and above) stemmed from the need to represent varying modal alternatives based on the average trip distance, instead of working with overall average trip distances. Each of the trips considered is assigned one of the 18 modes considered in the model, generally aggregated into six modal groups for better result readability: Active mobility (walk, bike, scooter sharing, bikesharing), Public Transport (commuter rail, metro, light rail transit, bus rapid transit, bus), Shared mobility (taxi, ride sharing, on-demand taxibus), Private vehicles (motorcycles, cars), Paratransit (informal buses, informal two and three-wheelers), Shared vehicles (shared motorcycles, shared cars). The representation of shared mobility is one of the most detailed available in transport models.

The four major outputs of the model are the evolution of the total number of trips, the passenger demand in passenger-kilometres, the vehicle demand in vehicle-kilometres, and the related CO₂ and local pollutant (BC, NOx, SO₂, NH₃, PM_se and VOC) emissions in tonnes of CO₂ equivalent and kilograms of pollutants. Other indicators of resilience, accessibility, space consumption, road safety, congestion and affordability are also available. All the outputs can be disaggregated by the geographical, population, distance and mode segments, even though aggregations are generally made for readability purposes and because the model is calibrated at the modelling region level.

The key data sources used to initialise and calibrate PASTA-GUM are listed in Table 2. A significant effort was required to adjust these varied data sources to the spatial aggregation of the mFUAs.

Table 2 List of primary data inputs for PASTA-GUM

Comparison with other transport sector models

To understand the methods underpinning PASTA-GUM, it is helpful to position it in the context of other urban passenger transport models in terms of geographic scope, functional scope, and methodological approach. PASTA-GUM is classified as a global, strategic, trip-based behavioural model. To the best of the author’s knowledge, the ITF model is unique in its capacity to model urban passenger transport at a strategic and trip-based level for each city across the globe.

There are many urban transport models with a strategic functional scope. They are often developed at the city level to analyse the impact of specific infrastructure projects or decarbonisation policies on urban passenger transport. Notable examples include Martinez and Viegas³³, which develops an agent-based model for predicting the effects of a new automated shared mobility system in Lisbon. Aziz et al.³⁴, also use an agent-based model approach to quantify the benefits of improved active travel infrastructure in New York City. Taking a behavioural approach using elasticities and discrete choice models, Olszewki and Xie³⁵ estimate the impact of different road pricing schemes on traffic congestion in Singapore. Microsimulation can also be used for city-level strategic transport models; this approach has been used to design models for Toronto³⁶ and Beijing³⁷, among others. Integrated land-use and transport models are another popular approach that considers long-term land use changes resulting from transport policies and their implications for carbon emissions at the city level³⁸. Finally, rather than representing overall passenger transport flows, Reul et al.³⁹, simulate a synthetic city to test the evolution of its demand and emissions under different policies.

Other strategic models have been extended to cover multiple urban areas in the same region or country. Brand et al.⁴⁰, develop a strategic model for the overall transport sector in the UK using disaggregate vehicle stocks and average travel distances by vehicle category. Similarly, Matsuhashi and Ariga⁴¹ assess the emission reduction potential of compact urbanisation for cities across Japan using a strategic transport model. Li et al. forecast future travel demand for 288 cities across China with travel behaviour models to compare the outcome of national sustainable transport policies for different city typologies⁴². Other examples include models of urban travel demand for cities in the United States⁴³, Denmark⁴⁴ and Germany⁴⁵.

With regards to methodological approach, PASTA-GUM is a behavioural (or sectoral) model because it models the trips that originate from passenger travel decisions, which are based primarily on exogenous determinants of demand⁴⁶. Behavioural models generally estimate aggregate travel behaviour by sequentially implementing different choice models for the parameters of each trip. The sequential approach is known as the ‘four-step model’, which is suited for macro analysis and changes that would not result in great systemic changes in the city (e.g. destination location, mode availability)¹⁵. There are also activity-based models that estimate several decisions simultaneously (destination choice, mode choice and route choice). Certain activity-based models also include trip-chaining behaviour wherein a traveller may choose to pursue multiple activities at different destinations as part of the same trip, thus introducing dependence between the choice of travel modes for those activities⁴⁷. Activity-based models, when integrating land use, can help policymakers understand how changes in land use and transport dynamics influence travel patterns⁴⁸. Trouvé and Leurent⁴⁹ demonstrate how different travel behaviour model structures can be applied to the same urban area, depending on the purpose of the analysis and desired insights.

Behavioural models are different from agent-based microsimulation (ABM) models, which are focused on detailed descriptions of transport movements and often include the effects of complex interactions between agents or vehicles in the model. Even city-level ABM models require both granular calibration data and substantial computational resources, making the approach infeasible for a global-scale model. Finally, there are also system dynamic models that measure variations of outputs (e.g. travel activity in passenger-km) based on aggregate changes or elasticities of variables that have been identified as drivers of change⁵⁰.

PASTA-GUM has a global geographic scope, making it unique among strategic, trip-based behavioural models. To the best of the authors’ knowledge, there are no other global four-step models for urban passenger transport. Other global transport models are generally components of a transport-energy model whose main purpose is to estimate future energy consumption. Such models use a purely systems dynamics approach based on vehicle stocks where the activity, scrappage and renewal of vehicles are estimated and aggregated to predict total travel demand. Vehicle fleets or population are the unit of analysis, with activity rates for each person or vehicle aggregated to determine overall energy consumption dynamics. This method is well-suited for applications in the energy sector where aggregate demand is sufficient to generate actionable energy consumption insights, but it does not account for policies that might change trip lengths or the characteristics of transport supply by mode in each city. Leading examples of such global energy-transport models include:

• The Global Change Assessment Model Transport Sector Model (GCAM-Transport) developed by the University of California, Davis and the Pacific Northwest National Laboratory⁵¹.

• The MESSAGE-Transport model developed by the International Institute for Applied Systems Analysis (IIASA)⁵².

• The Mobility Model (MoMo) developed by the International Energy Agency (IEA)⁵³.

• The Transport Roadmap Model developed by the International Council on Clean Transportation (ICCT)⁴⁶.

A fourth global transport-energy model is the AIM/Transport model by Mittal et al.⁵⁴, which provides a high-level estimate of urban and non-urban passenger transport demand. Like the models listed above, it also uses an activity-based estimation method rather than a trip-based estimation method and produces results for several global regions rather than individual urban areas. Tjandra et al.⁵⁵, compare the results from five similar global transport models with regional outputs, including the GCAM-Transport and MoMo models. Yeh et al.⁵⁶, provide an excellent comparison of the four widely used global transport models listed above, finding them to be relatively consistent with one another. The magnitudes and trends for passenger travel demand and carbon emissions in these models are also consistent with the most recently published ITF Outlook results².

There have been case-specific applications of the global transport-energy modelling frameworks described above. ‘Three Revolutions in Urban Transportation’ model created by Institute for Transportation and Development Policy (ITDP) and the University of California, Davis⁵⁷ is an extension of the MoMo model that analyses urban transport under scenarios related to automation, electrification and adoption of shared mobility. Similarly, a case-specific version of the ITF global urban passenger transport model was used to evaluate how transport policy can reduce CO₂ emissions from shared mobility services, and the results were compared across 247 cities⁵⁸. Taking a qualitative approach to applying previous models, Miskolczi et al.⁵⁹, review and synthesise a broad range of past literature to identify several common scenarios and estimate their potential for decarbonising the urban passenger transport sector.

Global transport-energy models, including the MESSAGE-Transport, GCAM-Transport and the AIM/Transport models described above, are also used as components of an Integrated Assessment Model (IAM)⁶⁰. IAMs combine outputs from diverse fields and economic sectors to analyse total emissions and energy consumption from transport, land use, agriculture, forestry and other activities. These integrated models are generally used to evaluate the global impacts of low-carbon pathways and policies^61,62 and do not consider the trip and city-level detail that PASTA-GUM provides.

Note that the choice of methodological approach can also affect the definition of ‘urban’ passenger trips. The PASTA-GUM model considers travel demand at the trip level, so it naturally includes all trips that are made within urban areas. However, by taking a vehicle-based approach, the MoMo model and similar transport-energy models can only differentiate urban and non-urban trips by vehicle type, rather than trip location. As a result, all travel activity by urban vehicle types is considered to be urban travel, even when it takes place outside of an urban area. By taking a trip-based perspective, the PASTA-GUM model can ultimately provide a more accurate definition of urban passenger transport.

Model structure and sources

The ITF global urban passenger model is organised around several interconnected components represented in Fig. 11. Every 5 years from 2015 to 2060, the model represents the evolution of the mFUAs and their transport supply characteristics, which in turn lead to travel demand evolution and related emissions. Outputs are also available for 2019 and 2022 in order to better analyse the impact of the COVID-19 pandemic crisis.

Fig. 11: PASTA-GUM: a modelling framework to simulate urban transport.

The figure details the different sub-components of the ITF Global Urban Passenger Transport Model (PASTA-GUM) model and the relationships between them. Grey blocks represent model input, the yellow block highlight the calibration data, blue blocks represent intermediate endogenous modelling, and green blocks represent model output.

The input blocks in grey are aggregations of the input data categories, which are entirely exogenous to the model. Base year transport supply data is used to establish the initial transport supply characteristics. The supply data for future years is endogenous to the model; it evolves based on changes to other indicators. Projections of socio-economic characteristics from the base year until 2060 are the second group of exogenous input data. Lastly, the scenario inputs include different future transport and land-use policy measures, and societal trends or technology assumptions impacting the development of metropolitan areas or transport systems. This input is used to test the impact of different scenarios on future transport demand profiles and related emissions.

The calibration block in yellow illustrates that the base year demand data input is not used as a direct input within the model, but rather as calibration values. The model parameters are set to reproduce these travel demand values based on the base year data inputs available. Expert analyses on the evolution of the model results are also conducted to ensure a proper calibration over time. This includes comparisons with other international studies on urban passenger transport and results of previous ITF models.

Both of the demand submodels shown in Fig. 10 were calibrated using available data from cities around the world. The trip generation submodel was calibrated using data on passenger kilometres and the number of trips for 24 global cities from Google Environmental Insights Explorer⁶³. These estimates are collected and validated by Google and made available to governments and international organisations. The calibration was limited to these cities due to data availability and restrictions on the number of data requests per user. They are, however, relatively diverse cities in terms of urban mobility systems and geography. The cities used for trip generation calibration are presented in Table 3.

Table 3 List of cities used for calibration of the trip generation submodules

The mode choice submodel was calibrated using a selection of urban mobility surveys from a much larger set of cities: 245 cities across 55 countries. This collection of cities is relatively diverse in terms of city size and region, and features 10 cities in Africa and 24 cities in Latin America, which are often underrepresented in urban mobility surveys. The mode share for walking, cycling, public transport and private car was first extracted from each of the mobility surveys. The region-specific mode choice model parameters were then adjusted to reduce the difference between the model-predicted values and the calibration data. The calibration process was terminated with an average root mean square deviation of 0.5 percentage points across all cities and mode categories.

The core model component blocks in blue are implemented within the modelling process and are therefore endogenous.

First, a demographic module estimates the evolution of the population and its composition based on a survival stock model approach. Second, the characteristics of urban areas are updated, beginning with spatial geographic features (i.e. area, density), which impact transport supply and trip distance distribution. Several types of regression were developed to represent different sensitivities to changes in average Gross Metropolitan Product (GMP) per capita, population densities, based on thresholds dependent on the modelling region. Especially, the representation of the road network is based on five road categories ranging from pedestrian roads to urban highways, and the public transport infrastructures are distinguished by type of public transport systems: commuter rail, metro, light rail transit and bus rapid transit. Advanced features are set up to enable the apparition or disappearance of mass transit and paratransit modes based on thresholds of GMP per capita, population density and mFUA population. The mFUA characteristics then affect mode characteristics (reliability, travel time, access and egress time, waiting time, infrastructure connectivity, travel cost, parking cost, number of transfers) in turn. These two blocks represent the systems dynamics components of the model.

Third, a typical travel demand generation step is executed within the trip generation and mode choice blocks for each population category. The trip generation is based on a regression model that includes GDP per capita and population category as explanatory variables, while the mode choice is based on a multinomial logit (MNL) discrete choice model sensitive to modal characteristics that range from time and cost to service coverage and availability, individual-specific population category and policy measures. The initial availability of a mode alternative for the mode choice within an mFUA and for a distance bin is determined by the existing transport supply and mode applicability by distance bin, varying based on the age and gender population segments. The affordability of each transport alternative for different income distributions is calculated to assess whether the cost of a mode might render it unavailable to certain users. The modal cost calculation includes the vehicle acquisition component for individual transport and the user fees or service fares. Unaffordable trips are eliminated from the choice set to account for the impact of financial constraints when choosing transport alternatives. The final mode shares are then computed following the MNL discrete choice.

After the travel demand modules, the trip number and passenger demand are directly accessible. Data on the average occupancy rates for the base rate and assumptions regarding their evolution up to 2060 enable directly converting the passenger demand into a vehicle demand. The latest stage, converting vehicle demand into emissions, is conducted via the ITF global vehicle fleet model, taking into account all the demand from the global demand models of the PASTA framework. While historically produced by MoMo from the IEA, the vehicle fleet model enables an endogenous simulation of the penetration of new technologies in the vehicle fleet, and thus better refining the overall quality of the emission output. It also extends the capacity to represent local pollutants by including BC, NOx, SO₂, NH₃, PM_se and VOC.

The policy measures developed in the model are conceived as perturbations from the regular evolution simulated by the model. Each policy measure will impact the evolution of the related variables and submodules. The impact of each policy measure on the model parameters is detailed in Supplementary Table 1. As such, PASTA-GUM enables testing the combined effect of policy measures as well as the individual impact of each. The whole set of policy measures defines the scenario that is simulated in the model. Another traditional scenario developed at the ITF includes a Business-as-Usual scenario reflecting the pathway stemming from current commitments or policies expected by experts.