Environmental Impact Assessment: Externalities of Automated Electric Vehicles for Public Transport

Abstract

This chapter provides a comprehensive analysis of the economic and environmental externalities associated with the use of automated minibuses in public transport, using a mixture of scenario planning and an externality model in the context of the AVENUE project cities. By analysing six different deployment scenarios, including the substitution of automated minibuses for buses and private cars, this study sheds light on the potential shifts in external costs and benefits. This chapter carefully assesses the impact of the deployment of automated minibuses on reducing external costs, taking into account factors such as energy efficiency, connectivity, automation features passenger numbers and vehicle utilisation rates. The results show that the environmental and economic outcomes of deploying automated minibuses depend significantly on the specific deployment strategies, highlighting scenarios in which automated minibuses could either reduce or exacerbate external costs. Through a detailed assessment of these scenarios, the chapter provides a nuanced understanding of how the strategic integration of automated minibuses into urban transport systems can influence the broader goals of economic sustainability and environmental protection. The study emphasises the importance of aligning automated minibus deployment strategies with city-specific goals and the broader sustainability agenda and provides valuable insights for policymakers, urban planners and transport stakeholders.

1 Introduction

The following assessment addresses six scenarios for potential deployment in AVENUE cities and the resulting external costs. It focuses on the potential increase or decrease in externalities as an indicator for the scenarios in which mobility should evolve. The summary of categories, impacts and methods using the work of van Essen et al. (2019), Jochem et al. (2016), Héran and Ravalet (2008), Fagnant and Kockelman (2018), and Shalkamy et al. (2015) is shown in Table 14.1.

Table 14.1 Recapitulation of the categories, effects and methods of externalities

In this chapter, the scenarios planning methodology (intuitive logic approach, key factors and driving forces) as well as the six selected scenarios (replace all buses, replace all cars, expand the network, targeted expansion of the network, robotaxis, AM in MaaS) are explained and the externality categories are introduced. First, the externalities and scenarios model is applied in detail to the case study of Geneva, where the mobility behaviour of the city, the results in terms of savings or losses per scenario and their implications are presented. Then, the results are discussed, such as the limitations of the analysis, comparison between the scenarios, potential policy recommendations, and rebound effects. Afterwards, it is applied to the other three cities of AVENUE (Luxembourg, Lyon and Copenhagen). In a final stage, a comparative analysis is conducted between the cities.

2 Scenarios

2.1 Methodology

Scenario planning is a way to imagine potential paths in the future (Derbyshire & Wright, 2017). The scenarios are built using the intuitive logic approach (ILA). ILA defines driving forces (political, economic, technological, ecological, social and legal) as well as key factors that help structure the scenarios. These factors are either quantitative and predictable, such as demographics, while others are qualitative and less predictable, such as user acceptance and policies (Huss & Honton, 1987; Lindgren & Bandhold, 2009). The ILA strength lies in its flexibility (Zmud et al., 2015).

For this study, the driving forces help answer how and why the AM might be deployed in each scenario; they are defined, based on Townsend (2014) and Milakis et al. (2017) scenarios, as follows:

• Technology advancement (automation technology and digital services).

• Urban policy (political agenda for mobility and sustainability).

• Transportation offer (trends of use and modes available).

• The users (most likely to use the AM in the scenario).

As for the key factors, they were determined through a deliberative process within the AVENUE team. The trial sites, the interdisciplinary nature of the project, and the work of (Beukers, 2019; Korbee et al., 2021; Viere et al., 2021) were used to select the following key factors:

• Whether the AM are replacing one mode of transport or multiple modes

• Whether AM supports or competes with public transport (PT) (replace buses modal share or not)

These two factors help estimate the modal shift within each scenario, which is crucial to estimating the future transportation performance in person-km (pkm) based on mobility censuses and the consequent externality estimations.

The setting of key forces and key factors is combined with the system boundaries of each scenario. Figure 14.1 shows the classification of the scenarios based on the key factors.

Fig. 14.1

The classification of scenarios based on key factors

The boundaries are defined by the key parameters of the vehicles, circulation specifications and AM service. Thus, the analysis is continued as a qualitative assessment that studies the direct and indirect consequences of the modal shift from each scenario using observations from AVENUE pilot studies and previous research on AV deployment. The overall structure is presented in Fig. 14.3, and it is the basis for the scenario description. The methodology helps limit the uncertainty of the scenarios by building plausible stories (Lindgren & Bandhold, 2009; Amer et al., 2013). See Fig. 14.2 for the steps of the scenario description and Fig. 14.3 for the methodology:

Fig. 14.2

Steps of scenario building

Fig. 14.3

Methodology of scenario description

2.2 The Scenario Description

Based on the methodology described, the analysis focuses on six scenarios that concern the deployment of AM in cities. At this level, the scenarios occur in a standard EU city (to be applied to Geneva, Lyon, Luxembourg and Copenhagen) and in the next decade, which is a safe estimation of when is it most likely to have AV technology on the roads (Milakis et al., 2017).

Out of the six scenarios, five focus on different modal shifts caused by the deployment of AM, while one focuses on robotaxis. Among the overall scenarios, four are set in the city centre, while two are in smaller urban dwellings (characterised by smaller urban densities) surrounding urban areas like villages, towns and suburban areas.

The scenario description using the proposed methodology helps pinpoint the advantages and disadvantages of each deployment strategy. It also seeks to learn about potential obstacles and catalysers to reduce the environmental footprint of introducing AV in urban areas. It is reinforced by a deliberative process with the AVENUE experts (Antonialli et al., 2021), where the most plausible scenarios are drafted based on the key factors and driving forces.

2.2.1 Scenario 1: Replace All Buses (Sc1)

This scenario occurs in the city centre. AM are implemented on a fixed schedule, yet they operate on flexible routes within a geofenced area.

According to Hansen et al. (2021), geofencing involves the utilisation of GPS or similar location-based technologies to define virtual boundaries or zones within which AM can function. These boundaries are typically determined by either the service provider or regulatory authorities. The minibuses are meant to support the public transport network, mainly rail transport, by replacing diesel buses. The AM are more flexible and less costly (Milakis et al., 2017).

Technological advancements such as achieving level 4/5 automation, improved sensory capabilities and AV platooning support the replacement (Wadud et al., 2016). The transportation system at the time of introducing the AM does not differ significantly from the status quo. The electrification process of the bus fleet is limited; motorised individual mobility is dominant.

The replacement might lead to a ripple effect that reduces car use and increases train ridership in the long term. Integrating electric powertrains in AM would lead to a decrease in both air pollution and greenhouse gas emissions compared to the typical buses. It is also wise to predict a change in the building environment, where more roads and train stations could be redesigned to assimilate the change. The city would assign more pick-up and drop-off points around the stations. Eventually, modernising the public transport and integrating AM in ITS will make the city centre more attractive, thus attracting more jobs and increasing the population (Narayanan et al., 2020).

2.2.2 Scenario 2: Replace All Cars (Sc2)

This scenario is also set in a high-density urban area such as a city centre. The AM are deployed to replace all cars trips. The service of the AM is on-demand, door-to-door, and running on flexible routes. The AM would have on average six people per vehicle; see Table 14.2. We assume no restrictions on waiting time (for a waiting time of more than 4 min, there is no need for an increase in fleet size to meet an increase in travel demand as well as empty runs based on Fagnant and Kockelman (2018).

Table 14.2 Occupancy factors (van Essen et al., 2019)

It is also set to support public transport and only replace cars by introducing policies restricting the circulation of internal combustion engine vehicles (ICEV). The scenario is supported by policies that mirror the current environmental agenda of sustainable cities, such as the zero-emission strategy or carbon tax, presenting the ultimate goal for sustainable mobility (Fournier et al., 2020; ITF, 2015; ITF, 2017; ITF, 2020).

The replacement of cars with automated vehicles such as the AM is bound to have a significant effect on accidents rates and traffic congestion, as explained by Childress et al., 2015; Alazzawi et al., 2018; Auld et al., 2017; Litman, 2021. Less AM are needed to meet the travel demand. This leads to an increase in road capacity and thus a decrease in road traffic, especially during peak hours (Wadud et al., 2016). On the one hand, the AM would be integrated better with other long-distance transportation such as trains as AM are used more by travellers. On the other hand, they might reduce the active transportation modal share because the AM provides on-demand and door-to-door short-distance trips (Milakis et al., 2017).

2.2.3 Scenario 3: Robotaxis (Sc3)

Robotaxis are described as shared automated vehicles in numerous studies (Alazzawi et al., 2018; Fagnant & Kockelman, 2018; Jones & Leibowicz, 2019; Litman, 2021). Although they might be comparable to the AM in terms of services (on-demand, door-to-door), they differ in the occupancy factors, speed, vehicle size and integration with PT (Merlin, 2017). The AM is a bigger vehicle that could carry up to 15 passengers, it rarely provides single-ride trips and it requires longer waiting times for pick-up. On the contrary, the robotaxis are destined mostly for single ridership. They are operated by private stakeholders, they could drive faster and they have reduced waiting times. They are convenient, especially if the passenger privileges their privacy (UITP, 2017).

For this scenario, the robotaxis serve the city centre as well. They do offer door-to-door and on-demand trips but no ridesharing services. They are competing with public transport, replacing more than one mode of transport (cars, buses and walking).

The regulatory conditions are described best as a “Laissez-faire” outcome. This means that there are no policies to regulate the AV market. Private stakeholders are seeking to maximise their profit which would have unpredictable consequences on sustainable mobility and people’s welfare (Niles, 2019). The regulatory conditions also translate into a deteriorating public transport offer that manifest in a high dependency on individual motorised mobility. A laissez-faire approach is likely to result in a rise in traffic volume and a potential increase in pkm. Concurrently, anticipated enhancements in network efficiency, often assumed, may not materialise without effective government intervention (Cohen & Cavoli, 2018).

The trend of AV markets driven by shareholders’ interest would create a race to optimise the services: higher speeds, less waiting times and more vehicles; therefore an increase in overall vehicle travelled km “VKM” is expected.

The long-term consequences of this scenario on the mobility system are the modal shift from active modes of transport and public transportation. Thus, this scenario is expected to cause induced demand as a rebound effect, and it could even reduce public transport ridership as it is very convenient (Litman, 2022; Niles, 2019; UITP, 2017). Furthermore, urban planning follows car-centric strategies, where the building environment is designed to accommodate private vehicles rather than the people. The spread of AV means new roadway design features such as improved lane markings, signs designed to be read electronically and wireless repeaters in tunnels to provide internet access (Childress et al., 2015).

2.2.4 Scenario 4: Expand the Network (Sc4)

In this scenario, the AM are deployed to support public transportation in low-density urban areas such as small villages and suburban areas where the PT offer is limited compared to more urbanised areas. The service provided by the AM is seamless intermodal trips. The modal shift is to replace one mode of transportation, mainly cars. The technological development is similar to previous scenarios. The policies in place that enable the expansion of the transportation network with AM are drafted to increase accessibility, reduce reliance on individual vehicles, and develop surrounding areas to cities. The modal shift is partially from cars to AM. The direct repercussions of the scenario, in terms of environmental and societal impacts, should be similar to the “replace all cars” scenario. The direct effect on mobility is that the AM is more susceptible to making empty runs and carrying fewer passengers on board. The AM are meant to attract car users. Thus they should provide services that could compete with the comfort and convenience of an individual vehicle.

For the indirect consequences, we predict that the AM would increase overall PT ridership. However, it is difficult to determine the effect on walking and biking, but as it offers short trips, it could also replace some walking and biking. The improvement in the public transport network outside of densely populated areas would make them more attractive to inhabitants.

2.2.5 Scenario 5: Targeted Expansion of the Public Transport Network (Sc5)

The targeted expansion scenario is similar to scenario 4 “expand the network”. In this scenario, the AM are supporting the public transport network. It also partially replaces car modal share. The prominent feature in this scenario is that it also replaces a share of the buses. Specifically, the AM are deployed to replace night buses and low occupancy buses. The bus service is at capacity (or even over capacity) during peak hours but underutilised during the day (Pyddoke, 2020).

Moreover, it runs frequently empty in areas with high car ownership (Adra et al. (2004). Hence, substituting these buses with AM could reduce the environmental impact. This scenario’s direct and indirect consequences are similar to scenario 4.

2.2.6 Scenario 6: AM in MaaS (Sc6)

The AM are deployed within MaaS to better provide on-demand services that bridge the first and last mile and provide seamless and intermodal trips. They are deployed in highly dense areas such as city centres. Their introduction aims to support PT. They are positioned to influence more than one mode of transport (cars, walking and biking). The technological innovations in AV are similar to the previous scenarios. However, there are significant advancements in digital on-demand services, interoperability, ticketing, utilising mobile apps, the cloud, ride-pooling and routing algorithms (Plested, 2021).

The regulations to support this deployment strategy rely on public and private collaboration for MaaS services, platform management, open API, and data sharing.

Other regulations in place are similar to scenario 2 “replacing all cars”, where the city is seeking to implement more sustainable practices in line with the sustainable urban mobility plan (SUMP) and smart city initiatives. They adopt fuel and parking measures and push and pull regulations (in line with transport demand management (TDM)) to prevent the use of ICEV in the city centre and reduce the environmental and societal impact. The public transportation offer is efficient and reliable. However, there are gaps connecting travellers to mobility hubs (e.g. tram and metro stations). Thus, the AM seeks to capture first- and last-mile travellers that would have driven to reach a train or a tram station. The modal shift to be studied in this scenario concerns the share of journeys within an intermodal trip that connects to or from a train/tram station. The deployment of AM would reduce emissions.

We consider that for the AM to meet the travel demand and remain competitive the waiting time is less than 4 min. Hence, there is an increase in VKM due to pooling and rerouting to pick-up and drop-off passengers (ITF, 2020; Jones & Leibowicz, 2019; Milakis et al., 2017; Moreno et al., 2018). Reduction in parking space is expected (Zakharenko, 2016).

The long-run consequences of this scenario is an increase in PT ridership, as the AM provide seamless intermodal and last-mile trips and is considered as a mobility gap filler. However, the convenience of the service could replace more short-distance trips from walking and biking (Shen et al., 2018).

2.2.7 General Comments

It is important to note that the speed in all scenarios is limited to 30 km/h for all road transportation. The average occupancy factors are those of the Delft report (van Essen et al., 2019) for traditional road transportation. Also, we follow the Huber et al. (2022) assessment where the AM occupancy factor is five passengers. We assume an occupancy factor of 1.2 for the robotaxis which is more comparable to a normal taxi where individual trips are significant. These factors vary in suburban scenarios to translate its deployment particularities. Table 14.2 presents the different occupancy factors.

2.3 Modal Shifts

The modal shift is essential to determine the direct and long-term consequences in the qualitative description above. Moreover, it is a key component to the externality estimations. Thus, some assumptions were needed to determine the AM modal share in each scenario. The modal shift explanation and resources for each scenario are in Table 14.3.

Table 14.3 Modal shift for the scenarios

2.4 The Representative Survey

The social impact assessment is based on surveys to study the potential acceptance of users and the trends related to AM use. In this assessment, we focus on the willingness questions that were part of the representative surveys. These specific questions help determine the potential modal shift in some scenarios.

Scenario: AM in MaaS: The modal share of cars to be absorbed by the AM is determined by using the question: “how willing are you to give up your car if AM offers a service that bridges the first and last mile?”. The modal share corresponds to the percentage of respondents who were very willing, who consider their residential area as a city centre and who use their cars daily.

Scenario: Expand the network and targeted expansion of the network: the modal share of cars to be absorbed by the AM is determined by using the question: “How willing are you to give up your car if AM is part of a seamless, intermodal trip?”. The modal share corresponds to the percentage of respondents who were very willing and who consider the area they live in to be either a big town or a small to medium village and who use their cars daily.

3 The Externality Methodology

3.1 Production Phase

The manufacturing of vehicles is a complex process; it accounts for steps from the extraction of raw materials to the production of the components (Chester, 2008; Pero et al., 2018).

The assessment focuses on the climate change impact on the production phase. The emissions are CO2, CH4, and N2O; it is based on climate avoidance costs. According to van Essen et al. (2019), the avoidance costs are determined by averaging values from the literature for the short and medium term (up to 2030) and the long term (2040–2060). The values used in this estimation are central and short- to medium-run climate avoidance which is 100 €/tCO2eq.

The marginal cost is estimated as follows:

$$ {\displaystyle \begin{array}{c}\mathrm{Marginal}\ \mathrm{cost}\ \mathrm{for}\ \mathrm{production}\;\left(\notin -\frac{\mathrm{cent}}{\mathrm{pkm}}\right)\equiv \frac{\begin{array}{l}\mathrm{total}\ \mathrm{emissions}\ \mathrm{of}\;\mathrm{CO}2\mathrm{e}\;\\ {}\mathrm{during}\ \mathrm{production}\;\left(\mathrm{tCO}2\mathrm{e}\right)\end{array}}{\begin{array}{l}\Big(\mathrm{expected}\ \mathrm{lifetime}\ \mathrm{mileage}\;\left(\mathrm{km}\right)\\ {}\times \mathrm{the}\ \mathrm{average}\ \mathrm{occupancy}\end{array}}\\ {}\times 100\times \mathrm{central}\ \mathrm{short}\;\mathrm{run}\;\mathrm{value}\;\left(\notin \right)\end{array}} $$

The total emissions, as well as the expected lifetime mileage, were determined using the ecoinvent database for the car and electric car, while for the bus, the values were taken from Chester and Horvath (2009). Finally, Huber et al. (2022) as well as Viere et al. (2021) study on the LCA of the AM were used. More details about the estimations are found in the environmental impact assessment chapter. The marginal costs are presented in Table 14.4. These values are applicable at the EU level.

Table 14.4 Marginal cost for the production phase

3.2 Parking Space

In the analysis, only the savings from parking spots for private vehicles are considered. The buses are usually stored and maintained in dedicated bus garages or depots, and they do not interfere with daily traffic. Although it raises an issue for land-use and transportation planning, it presents less nuisance to cities compared with private vehicles and thus, it is not covered in this assessment, similar to the AM (Lai et al., 2013).

To determine the overall parking space, the fleet size for cars and the AM is needed in each scenario. Thus, the fleet size calculator that is presented in the previous chapter is used. Moreover, we assume a linear relation between the parking space reduced and the modal share of car trips reduced.

4 Assumptions and Boundaries

To be able to calculate the external costs and build on work presented in Antonialli (2019) and Jaroudi et al. (2021) assessment, we assume the following:

The AM deployment could affect motorised mobility, public transport and active mobility (Janasz, 2018). The AM are introduced in mixed traffic (with no prior presence of automated technology on the roads).

Every additional person-kilometre travelled on the AM is travelled less on the other transportation modes. Therefore, the total transport performance remains identical compared to the reference scenario (Bubeck et al., 2014).

Active mobility’s negative externalities as walking and biking are considered negligible (Keall et al., 2018).

The study does not account for the increase in transport performance due to population increase and other sociodemographic changes; we are focused on the effect of unpredictable factors such as the integration within the transportation system and policies. That is why we chose to simplify the calculations by omitting their effects (Huss & Honton, 1987).

Intermodal trips consist of two or more modes of transport (car/bus/walking/biking+train) (Fraedrich et al., 2015), while monomodal or unimodal trips are considered as one mode.

Walking trips accounted as a mode of transport for intermodal trips from 600 m or more (more than 5 min) (Gebhardt et al., 2016).

The average speed of circulation for all vehicles ranges from 25 to 30 km/h.

The AM operate on “other urban roads” or “other interurban roads” based on the scenario, and we estimate that the average traffic flow¹ is near capacity.

5 Application of the Externalities Model to Geneva

In the following part, the assessment applies this model by adjusting the marginal costs (€-cent per pkm) to the context of the city and estimating the future transportation performance in person-km (PKM). It considers the four cities of AVENUE as a case study. First, Geneva is studied in detail to showcase the specifics of the scenarios and the externalities calculations. Second, the results are presented for Luxembourg, Lyon and Copenhagen, applying the steps shown in Fig. 14.4.

Fig. 14.4

Externalities calculations and data methodology

5.1 The Mobility Behaviour Profile in Geneva (National Census Data)

Geneva is the second most populated city in Switzerland, with 197,376 inhabitants. The “Mobilités 2030” long-term strategy aims to address mobility challenges in Geneva Canton, such as underdeveloped public transport in low-density areas, narrow city centre streets and the absence of tangential routes. The promotion of low-carbon mobility is part of the Climate Plan 2018–2022 in Geneva Canton. Automated driving could replace motorised individual mobility, promote active modes of transport and make the transportation system more robust against crises. The legal framework for autonomous vehicles is being developed, and the “Digital Switzerland” strategy will address the necessary digital key factors for autonomous driving implementation (Etat de Genève, 2013; Jenelius & Cebecauer, 2020; OFCOM, 2018).

5.2 Mobility Behaviour: The Reference Scenarios

To calculate the externalities, we refer to the current status with no AV on the roads, as mentioned in assumptions and boundaries. Based on the data available, we used the mobility census for Geneva in 2015. To distinguish between our urban and suburban scenarios, we present two reference scenarios; each is characterised by the modal split, average daily travelled distance per the mode of transportation, area in km2, population, and transportation performance (pkm). The first one is for the city of Geneva, while the second is for the second suburban ring, also where the Belle Idee pilot is operated.

The analysis relies on mobility behaviour data from a census (microrecensements mobilité et transports (MRMT)) realised between 2000 and 2015 by the statistical office of the Canton of Geneva (Montfort et al., 2019) as well as Geneva Public Transport-TPG (2016). The MRMT is an extension of a Suisse national census comprising the participation of 4500 residents. The census accounts for the evolution of mobility trends for 15 years (Montfort et al. (2019)). The analysis provides trip distribution and the average daily distance per mode of transport in 2015. These values help estimate the total pkm per mode of transport.

5.2.1 Reference Scenario (Sc01): City of Geneva

This scenario is the comparison point for the scenarios “replace all cars” and “replace all buses”, “Robotaxis” and “AM in MaaS”. The trip distribution and the average daily distance per the mode of transport in 2015 are computed in Table 14.5. The population is 197,376 in the city of Geneva (Rietschin, 2015). The area in the city of Geneve is 15.93 km2 (Service de la mensuration officielle, 2005). The overall daily trips amount to 711,000 trips (Montfort et al., 2019). The modal split is estimated using Montfort et al. (2019) and TPG (2016) analysis for the mobility in Geneva. These values are also used to estimate the new transport performance for the scenarios. The estimated transport performance and modal split is in Table 14.5.

Table 14.5 Mobility behaviour and transport performance in the city of Geneva 2015

Using annual transport performance from Table 14.5 and the marginal costs for buses and cars for Switzerland from Table 14.6, we estimate the total external costs for road passenger transport for Geneva in 2015.

Table 14.6 Marginal and total costs in €-cent per pkm in city of Geneva

The total external costs in 2015 are important because they represent the reference point that is used to compare the impacts of the AM introduction and calculate the potential increase or decrease in external costs (before and after the deployment of the AM).

5.2.2 Reference Scenario (Sc02): Second Couronne

This area has an overall population of 26,102 inhabitants and an area of 162 km2 (Federal Statistical Office, 2013; Service de la mensuration officielle, 2005). Based on Montfort et al. (2019), we estimate around 94,000 daily trips for all the inhabitants in the second couronne consisting of 11 municipalities. The modal split is in Fig. 14.5.

Fig. 14.5

Modal shift for the scenario: replacing all buses in Geneva

The mobility behaviour and the transportation performance in Table 14.7 present the values used to estimate the new pkm per mode of transportation used in the scenarios.

Table 14.7 The mobility behaviour and transport performance for the second suburban ring in Geneva, 2015

In this scenario, to better reflect the reduced public transportation offer, we adjust the occupancy factors. The bus has an average ten passengers on board. The estimation is based on the higher dependency on cars and a lower modal share of public transport, in general, in suburban areas compared to city centres (high-density population areas): The car modal share is 44% compared to 22.3% in the city centre. And the buses’ modal share is 10%, while it is 15% in the city of Geneva. We assume that the AM are circulating on interurban other roads. The marginal costs are adjusted accordingly and included in Table 14.8, as well as the total external costs (Total external costs (2015), Y = ∑_i∑_jxij, (2015)).

Table 14.8 Marginal and total costs in €-cent per pkm in second suburban ring in Geneva

These values are used to estimate the increase or decrease in external costs in the suburban scenarios: “expand the network” and “targeted expansion of the network”.

In the following part, the scenarios in the city centre are analysed. Afterwards, those in suburban parts are presented. The assessment relies on data from the marginal costs in Table 14.4, Table 14.6, and Table 14.8, scenario description and the reference scenarios (see Fig. 14.4) by applying the steps from the externalities model.

5.3 Scenarios in City Centre

5.3.1 Replace All Buses (Sc1)

Following, the scenario “replace all buses” is tested in the city of Geneva. First, the marginal costs for the vehicles are presented in Table 14.9 based on Table 14.6 and the marginal costs for the AM. Second, the AM are deployed to replace the 12.3% bus trips share. The estimated new modal share is shown in Fig. 14.5.

Table 14.9 Marginal costs (car, bus, and AM) in €-cent per pkm in Geneva city

Similarly, we assume that the AM conduct the trips of the buses with the same average distance of the bus trips from the reference scenario. This led to a prediction of the transport performance and, consequently, the total externalities.

Scenario (Sc1) leads to an increase in external costs compared to (Sc01), mostly due to congestion. The increase in externalities equals around +12 million euros; see Fig. 14.6. If the congestion is omitted, the scenario shows a decrease in externalities of −11 million euros. The wtt costs more in terms of external costs, but all the other environmental impacts are considered positive if the buses are replaced with AM (see Fig. 14.6). However, this is still offset by almost +23 million lost due to congestion. The congestion is also explained by the occupancy of an AM being lower than that of a bus, as well as the difference in the sizes of the vehicles (both factors interfere in the estimation of the marginal cost of congestion). In this scenario, there is no estimation for parking space saved since it focuses on bus replacement (Fig. 14.7).

Fig. 14.6

The increases or decreases in externalities with congestion in million euros in Sc1

Fig. 14.7

The increases or decreases in externalities without congestion in million euros in Sc1

5.3.2 Replace all Cars (Sc2)

The marginal costs are those in Table 14.9 (from Sc1). We replace the car modal share in the city of Geneva (i.e. 22.6%) with the traffic circumstances of 2015 (Fig. 14.8).

Fig. 14.8

Modal shift for scenario Sc2: replacing all cars in Geneva

Replacing all cars with AM leads to a reduction in all external cost categories except the wtt costs, where there is an increase of +3.4 million euros. The wtt emissions increase could be justified by reliance on electricity generated from fossil fuels (Kasten et al., 2016). In general, this scenario could save up to 308 million euros, with the majority coming from the congestion savings (−210 million euros); see Figs. 14.9 and 14.10.

Fig. 14.9

The increases or decreases in externalities with congestion in million euros in Sc2

Fig. 14.10

The increases or decreases in externalities without congestion in million euros in Sc2

The environmental impact of replacing all cars with the AM is reflected in gains of around 40 million euros. The findings are aligned with most studies that support a shift from individual motorised mobility toward electric and shared transportation. Using the fleet calculator, the estimated fleet size needed to replace all car trips in the city of Geneva with AM is around 1380 minibuses for a waiting time superior to 4 min. Finally, the savings in parking space is 0.65 km2, the equivalent of 64,824 parking spots.

5.3.3 Scenario 3: Robotaxis (Sc3)

This scenario focuses on AV’s integration in the transportation network in the form of robotaxis. The marginal costs for the robotaxis depend on van Essen et al. (2019) estimations for electric vehicles (EV). In this scenario, we estimate an occupancy factor for the robotaxi of 1.2. The marginal costs in the Delft assessment for the electric vehicle use an estimation of an occupancy factor of 1.6 (as well as for an ICEV, Table 14.2). Thus, the marginal costs for the different external costs categories are adjusted for an occupancy factor of 1.2 (new marginal cost for robotaxi = marginal cost for EV x 1.6/1.2).

The robotaxis’ marginal costs are presented in Table 14.10. Furthermore, to better translate the effect of the low occupancy factors, we assume that the robotaxis are operating on congested roads.

Table 14.10 Marginal costs for scenario robotaxis in €-cent per pkm

The robotaxis modal share is explained in Table 14.11, the new modal split of the scenario in Fig. 14.11. 20% of car modal share (~22%) means the AM will absorb around 5% of car trips and, similarly, a 6% reduction in bus share means around 1% of trips replaced by AM. Finally, for the walking trips, we assume the new modal share of walking is 37% (~48%–13% = 37%).

Table 14.11 Robotaxis modal share composition

Fig. 14.11

Modal shift for scenario Sc3: robotaxis in Geneva

The deployment of robotaxis in the city centre with a laissez-faire outcome causes a significant increase in external costs compared to Sc01. In fact, it is estimated to cost around +162 million euros in terms of environmental and societal impacts (see Fig. 14.12). Figure 14.13 is used to gauge the environmental impacts. The scenario also leads to decreased gains in the environmental and accidents categories except for the noise and wtt categories. The overall difference in external costs compared to Sc01 without the congestion is around one million euros.

Fig. 14.12

The increases or decreases in externalities with congestion in million euros in Sc3

Fig. 14.13

The increases or decreases in externalities without congestion in million euros in Sc3

Accounting for the increased travel demand, the fleet calculator by Fournier et al. (2020) shows a fleet of 1058 robotaxis for an on-demand service with a waiting time of 1 min. The 1058 robotaxis could cover 18.2% of all daily trips within a geofenced area of 15.93 km2. We follow the simulation results from (Bischoff & Maciejewski, 2016) since the simulation has similar conditions to Sc3; thus, one robotaxi replaces ten cars. Hence the new car fleet is 54,624 cars compared to the original 64,824 cars (Bischoff & Maciejewski, 2016). The 10,200 parking spots saved result in an additional 0.1 km2 of available space. We do not account for parking space for robotaxis based on numerous studies that assume the robotaxis are to be stored off roads in special facilities (AV parks) to be maintained and charged or would be circulating all day (Hayes, 2011; ITF, 2015; Nourinejad et al., 2018; Zhang & Guhathakurta, 2017). Moreover, these garages are not part of the public domain since they belong to the private stakeholders (the robotaxis are operated by private stakeholders; see 5.2.2, scenario 3 robotaxis); hence, they would not affect public infrastructure.

Finally, we test the results for when the robotaxis are a part of a pooling service. Based on Alazzawi et al. (2018), we account for an occupancy factor of 2.4 instead of 1.2. The subscenario with ridesharing registers an increase in externalities that accounts for around 76 million euros. Alternatively, if we consider Mosquet et al. (2015) estimation for an average of four people per vehicle, the increase is around 41 million euros. See Table 14.12 below.

Table 14.12 Scenario 3 robotaxis applied with and without ridesharing service in Geneva increase in externalities

This shows a correlation between the occupancy factors and the externalities. Hence, ridesharing could contribute to a better environmental and societal impact when deploying robotaxis.

5.3.4 Scenario 6: AM in MaaS (Sc6)

Table 14.9 includes the marginal costs used to calculate the externalities of integrating the AM in MaaS in the city centre of Geneva.

The modal split of Sc6 is explained in Table 14.3. The willingness to give up the cars for daily car users in the city centre of Geneva is used to estimate the modal share of cars that is absorbed by the AM. It is important to note that the sample for daily car users in the city of Geneva was limited, with only 34 respondents.

After conducting a social assessment, it was found that the level of willingness to use AM in Geneva was similar to that of Lyon. As a result, the sample used for the study included 81 individuals who use their cars daily and live in the city centre of both Lyon and Geneva combined.

From this sample, 23 respondents said they were very willing to give up their cars if the AM bridges first and last miles. Eventually, we consider that 28% of car trips could be replaced by the AM in MaaS, the equivalent of 6.42% of all trips. The AM is also replacing 3% from walking and 1.7% from biking trips, based on Table 14.3.

Table 14.13 and Fig. 14.14 show the composition of the modal share of the AM in this scenario.

Table 14.13 AM in MaaS modal share composition

Fig. 14.14

Modal shift for scenario Sc6: AM in MaaS in Geneva

The new modal split is shown in Fig. 14.14.

Following the same methodology to estimate the externalities, the scenario is predicted to lead to reductions in external costs that amount to −83.3 million euros compared to Sc01 (AM in MaaS will reduce the external costs from the reference scenario by 83.3 million euros). The reductions without congestion are around −13 million euros. Similarly to previous scenarios, the wtt impact when introducing the AM in densely populated areas is negative (leading to an increase in the wtt external costs). This is explained mostly by the effect of electricity production on air pollution. Congestion presents the bulk of the reductions, around −70 million euros. We estimate around −four million euros in savings for the categories of climate change, accidents and noise. See Fig. 14.15 for the results with congestion and Fig. 14.16 for results without congestion.

Fig. 14.15

The increases or decreases in externalities with congestion in million euros in Sc6

Fig. 14.16

The increases or decreases in externalities without congestion in million euros in Sc6

This contributes to savings in parking spaces of 0.04 km2 or around 4160 parking spots. Finally, we consider the results if the AM in MaaS in deployed in suburban areas. This could be interesting to see the difference in applying the same scenario in different settings (urban and suburban), and it highlights the importance of targeted and context-based deployment strategy. The analysis for suburban scenarios is applied in this case.

Based on the social survey, we can anticipate a replacement rate of cars by AM of 12.3%. The AM will also absorb 3% of walking and 1.7% of biking, similar to the analysis of the urban version of the scenario described previously. The decrease in externalities is estimated to be around 12 million euros. Without accounting for the congestion, we have a decrease of two million euros. See the comparative Table 14.14 below.

Table 14.14 Scenario 6 AM in MaaS applied in urban and suburban Geneva

5.4 Suburban Scenarios

For our two suburban scenarios, we use the mobility behaviour of the second suburban ring in Geneva (Sc02).

Table 14.8 is used to calculate the total externalities in the scenarios. The AM marginal costs are also adjusted to take into account that the AM will compensate for the shortage of public transportation in these areas. The occupancy factor is 8, and the marginal costs estimated in D8.4 have been adjusted accordingly in the Table 14.15 below.

Table 14.15 Marginal costs (car, bus and AM) in €-cent per pkm in second suburban ring of Geneva

5.4.1 Scenario 4: Expand the Network (Sc4)

To expand the network in suburban areas, the AM are deployed to replace a percentage of cars’ modal share. Thus, we rely on the explanation from Table 14.3 as well as why we used the sample of Lyon and Geneva from scenario 6.

Finally, the modal share of the AM equals 11.2%, absorbed from the cars’ modal share as presented in Fig. 14.17.

Fig. 14.17

Modal shift of Sc4: expand the network compared to Sc02

This modal shift leads to a decrease in externalities. We account for around −13 million euros. Consistent with the previous scenarios, congestion accounts for the highest impact, around −12.3 million euros; see Fig. 14.18. We register a slight reduction in the WTT category in this scenario. In general, there are reductions in all externalities categories. Without congestion, the reductions are around −three million euros; see Fig. 14.19.

Fig. 14.18

The increases or decreases in externalities with congestion in million euros in Sc4

Fig. 14.19

The increases or decreases in externalities without congestion in million euros in Sc4

The scenario presents reductions in external costs of −18 million euros for the city. If the congestion is not accounted for, the targeted expansion still leads to reductions of around −five million euros.

The fleet size for AM is estimated as 220, with a waiting time of 6 min and exit and entry time of 3 min (Fournier et al., 2020).

In conclusion, the savings in parking space equals 0.01 km2, around 1367 parking spots.

5.4.2 Scenario 5: Targeted Expansion (Sc5)

The targeted expansion scenario builds on the previous scenario. The same modal shift from cars to AM remains around 11.2%. However, in this case, we consider that the AM are targeted to also replace low occupancy buses and night buses. The demand for public transport fluctuates during the day and during its route. The notion of welfare optimisation justifies maximising the capacity of the bus (the bigger the bus, the better the ability to meet the peak time travel demands). Thus, public transport operators would justify running big buses (fitting up to 60 passengers). First, it is difficult to predict exactly when a bus is running empty (even during off-peak time, there might be exceptional demand). Second, operating another fleet of smaller vehicles would require more drivers and thus more costs than just keeping the big buses. However, as drivers are no longer necessary in an automated vehicle, these costs are eliminated. Furthermore, the on-demand feature makes the service customisable according to the demand (Pyddoke, 2020).

Determining the specific ratio of buses with low ridership is a complex process since the number varies unpredictably. That is why we consider Adra et al. (2004) study. In their analysis, empty running km for buses is around 11% based on data from Paris public transport system (RATP). Moreover, Mancret-Taylor and Boichon (2015) report stipulates around 4% of bus trips are taken between midnight and 5 a.m. Hence, we assume that more than 12% of all bus trips in suburban areas are replaced by AM trips. Finally, the modal share of AM in this scenario is 13.4%. Sc5 also registers reductions in external costs of around −15 million euros and −4 million euros without counting the external costs from congestions. The parking savings are similar to the previous scenario, “expand the network”; we estimate savings of 0.1 km2 (Figs. 14.20 and 14.21).

Fig. 14.20

The increases or decreases in externalities with congestion in million euros in Sc5

Fig. 14.21

The increases or decreases in externalities without congestion in million euros in Sc5

6 Discussion

To better demonstrate the scenarios and compare the results, Table 14.16 includes increases or decreases in external costs per scenario for each of the six scenarios analysed.

Table 14.16 The increases (+) and decreases (−) in external costs in the six scenarios of AM deployment in the Canton de Geneve

The two scenarios that incurred increases in external costs were “robotaxis” and “replacing all buses”. The first is set to compete with public transport, while the second is set to update the bus service. However, both demonstrate that replacing traditional public transportation without a planned strategy (or with a laissez-faire outcome) would worsen the impact of deploying AV in cities and diminish their potential benefits on the environment and society.

Replacing the bus scenario results matches the simulation of replacing buses with shared AV in Helsinki (ITF, 2017). Thus, it is not recommended to replace all the buses with AM but rather target the off-peak hours of bus trips like in scenario 5 “targeted expansion”, where the bus service is mixed with AM on-demand. The path to update the bus service would be better by replacing the fleet with electric buses.

The implementation of robotaxis exacerbates the shortcomings of the transportation system, as it perpetuates a model of individual mobility that has proven to be detrimental to the city’s environment.

Replacing all cars offers the greatest potential for savings, which aligns with many urban initiatives aimed at reducing car usage. This approach is in line with Geneva’s efforts to promote sustainable transportation, such as soft modes and public transit. Although replacing all cars clearly results in cost savings, it may be more difficult to implement in suburban areas, limiting its overall impact and potential (Duarte & Ratti, 2018; ITF, 2015; ITF, 2017). Moreover, it is more challenging to realise as it depends strongly on people’s acceptance of the technology and their willingness to get rid of their cars.

On a larger scale, MaaS could reduce environmental deterioration and increase the efficiency of the shuttles even if it absorbs some active mobility means. It is the most realistic and could provide better results. It could be part of an on-demand and door-to-door service across the canton of Geneva to reduce the long-distance trips conducted with cars and improve connectivity to mobility hubs.

Moreover, the suburban scenarios show optimistic results, which justifies the need to strengthen the transportation network where there are shortcomings of public transport. The gains might be limited compared to other scenarios gains (AM in MaaS or replacing all cars). Nevertheless, these deployment strategies would strengthen the overall transportation offer across the canton and encourage passengers to use trains more. This would lead to further gains in suburban and urban areas alike. Concurrently, high-density population areas usually contain more short-distance trips in smaller areas, as opposed to suburban and interurban areas. In this case study, the city of Geneva (one commune with a population density of 12,500 inhabitants/km2) accounts for 711,000 daily trips, while the second suburban couronne (11 communes with a population density of 161 inhabitants/km2) accounts for 94,000 daily trips. Hence, this makes the comparison between the two sets of scenarios biased. In order to maximise the savings, the AM could cover longer distances and connects communes (rather than operate exclusively within the commune).

If we focus on the externalities categories, we notice that the congestion presents the factor that leads to the biggest reduction in external costs (Sc 2, 4, 5 and 6) or the biggest increases (Sc 1 and Sc 3). It reflects the transport pricing and value of time. Replacing buses that usually operate within specific lanes with AM would potentially slow down the traffic flow, whereas deploying more individual vehicles like robotaxis would affect the traffic congestion.

This is incremental for policymakers as it showcases the perils of traffic jams as it worsens the traffic flow, which affects daily life and air pollution and GHG emissions. Reducing congestion is a leading cause of externalities gains in our scenarios. Dominating congestion externalities are aligned with Jochem et al. (2016) and van Essen et al.’s (2019) results for road traffic congestion.

The categories of accidents, air pollution, production and climate change show savings across the six scenarios. Hence, any introduction of AV, whether AM or robotaxis, will have positive impacts on accident rates and air pollution GHG emissions. The air pollution and GHG emission externalities during the wtt phase, on the other hand, show negative results for all urban scenarios. This is explained by the fact that the production of electricity for battery charging is strongly energy-intensive, and it involves air emissions, thus causing a not negligible environmental burden, as proved by Pero et al. (2018). Furthermore, the Robotaxis scenario is the only case where it causes a negative noise externality which is understandable since the marginal cost is the same as that of an ICEV. The decision not to distinguish between an EV and ICEV is due to the similarity in speed and the occupancy factors for both vehicles (Jochem et al., 2016).

Also, the introduction of AV in the city will, in general, lead to savings in parking spaces. The more cars are replaced, the more urban area is free. Free urban space means reshaping the built environment to be more green and liveable, which again aligns with the urban strategy of Geneva (Etat de Genève, 2013).

The direct costs of the deployment on a microscale for the pilots were analysed previously in Antonialli et al. (2021). Building on this analysis, we note that it is important to account for the effects of letting go of the drivers. The AM are operating with safety drivers on board in the AVENUE pilots. However, in the future, they will operate without human interference on board. This used a major advantage for replacing traditional public transport with AM as it will significantly cut the costs for the public transport operators, which would make it easier for them to adopt the technology and make their public transport more attractive and competitive. However, the layoffs mean an increase in the unemployment rate. For instance, Transport Public de Geneve (TPG), the public transport operator in Geneva, hires almost 1300 bus drivers. In Sc1, eliminating all buses might seem disadvantageous for the city in terms of environmental and societal impacts. It could reduce around 50 million euros of labour costs for TPG. However, this would have a more negative impact on the society by decreasing 1300 jobs, even if we account for the creation of new jobs: off-site safety operator positions are limited and would not compensate around 1000 posts, which would create labour market disruption and have a significant effect on the economy as well (Nikitas et al., 2021; Sousa et al., 2018).

To conclude, the externalities calculations supported most of the initial assumptions for the consequences of the deployment in the scenarios. It also helps clarify the ambiguity of the direct environmental effects of the AM. The analysis provides insights that would help policymakers decide on how to deploy the AM (or robotaxis) to support the prosperity of their cities.

7 The Model Applied to Other Cities

In this part, we apply selected scenarios to the AVENUE cities. First, we use the representative survey to filter which scenarios are interesting to be studied based on the sample of respondents that are using their cars daily in the geographic areas of the city centre, large towns and small to medium villages. For Luxembourg, the sample of respondents that are daily car users and live in the city was too small. Thus, we select the scenarios in the suburban and preurban parts. For Lyon, we follow the same analysis as for Geneva; we combine the samples for Lyon and Geneva. Finally, for Copenhagen, we focus on the urban scenarios (replacing all cars, replacing all buses, robotaxis and AM in MaaS) since the sample of respondents is significantly larger than those from small to medium villages.

7.1 Copenhagen

For this study, we focus on the municipality of Copenhagen for the four urban scenarios. The data used is based on the Danish national census for 2019 (Christiansen & Baescu, 2021). The modal share for the urban scenarios is from 2015 based on Kayser (2017). This represents the most recent data we could find that represents the mobility behaviour of the municipality of Copenhagen. The modal share is 26% private cars, 5% bus, 10% train, 36% biking, 22% walking and 1% other (including motorcycles). This is combined with the average distances per the mode of transport from the greater area of Copenhagen based on Christiansen and Baescu (2021).

Copenhagen municipality has around 623,000 residents and an area of 86.4 km2 (Statistics Denmark, 2021). The average daily trips per person are around 2.8. We account for a fleet of 252,600 cars and around 1000 buses (Movia, 2019).

The scenarios applied in this case are: “replacing all cars”, “replacing all buses”, “robotaxis”, and “AM in MaaS”. By using the same analysis and methods described for Geneva, we obtain the potential increase or decrease in external costs as presented in Table 14.17.

Table 14.17 The potential increases or decreases in external costs in Copenhagen for Sc1, Sc2, Sc3 and Sc6

The robotaxis scenario remains the one with the highest potential increase in external costs compared to the status quo, while “replacing all cars” is the one with the highest reductions in external costs.

7.2 Luxembourg

The scenario that we consider for Luxembourg is the communes of the Luxembourg canton without the city of Luxembourg. The area is 187 km2, and the population is 58,079 inhabitants (Ville de Luxembourg, 2021). The mobility behaviour data was limited. Therefore, we used averages for modal share from Luxembourg and average distances on EU level.

The externalities results following the estimations in marginal costs for Luxembourg, the transport performance and the predicted modal shift are presented in Table 14.18.

Table 14.18 The decrease in external costs in Luxembourg for Sc4 and Sc5

The results show that Luxembourg canton would reduce its external costs if it deploys the AM in the suburban parts.

7.3 Lyon

We focus on the area of the Greater Lyon (Metropole de Lyon) composed of the areas of Lyon-Villeurbanne (the centre Lyon) as the city centre and the ring of the Lyon Metropole (Couronne de la Metropole de Lyon). Thus, the reference scenarios are applied to these areas: Sc01 is the scenario of Lyon-Villeurbanne, and Sc02 is the scenario of the ring of the Lyon Metropole, both in 2015.

The transport performance and the mobility behaviour data are computed using a mobility survey for 2015, conducted by Lyon (Sytral, 2016).

The City of Lyon is one of the most populated cities in France, with a population of around 500,000 inhabitants and a surface of 47.9 km2. As for the mobility behaviour, using the average number of 3.3 daily trips for each inhabitant, we could estimate a sum of 1.65 million overall trips a day. In the city, the motorisation rate is 414 cars per 1000 residents. Thus, we count 207,000 cars. The modal split is 26% private cars, 9% bus, 16% tram metro, 3% biking, 45% walking and 1% other (including motorcycles)..

For the suburban ring, the population is around 885,927 inhabitants, and the surface is 486 km2.

The average number of daily trips is 3.43 for an average distance of 20 km compared to the 13 km average for the City of Lyon. In this area, there are around 30 million daily trips. The car fleet accounts for 575,853 cars. The modal split is 57% private cars, 5% bus, 9% tram metro, 1% biking, 27% walking and 1% other (including motorcycle).

We apply the scenarios and the externalities model as it was applied for Geneva.

The results are presented in Table 14.19 below.

Table 14.19 The potential increase or decrease in external costs in million euros Lyon for all six scenarios

The results here differ slightly from previous findings. In this case, updating the city’s bus fleet by replacing the vehicles with AM would lead to decreases in the external costs.

8 Conclusion

In general, the robotaxis scenario will always lead to increases in external costs when deployed in competition with public transport. A deployment with ridesharing could improve the balance. The environmental and accident categories record a slight decrease in externalities, but the decrease in external costs (without congestion) remains the lowest out of all the urban scenarios. The AM in MaaS, replacing all cars and the suburban scenarios consistently show decreases in external costs. Replacing all cars leads to the highest reduction in externalities (Table 14.20). While AM in MaaS is a user-centric approach pull strategy, replacing all cars is a typical push strategy which forbids cars in the city. Thus replacing all cars is more difficult to realise due to low social acceptance. For Copenhagen, the most appealing scenario seems to be to replace all car trips by the AM (see Table 14.21). The number of the city’s residents and the high average daily distance for car trips would justify this potential reduction.

Table 14.20 Summary of the results for AVENUE cities, increase (+) or decrease (−) in external costs for each scenario in million euros

Table 14.21 Summary of the results for AVENUE without congestion

Lyon shows that the emphasis should be on strengthening its public transport in suburban parts by deploying the AM serving seamless and intermodal services and also in the city, where the AM could be deployed within MaaS to bridge first and last-mile gaps and enhance connections to the rail stations.

Geneva would benefit from the AM deployment in the urban areas, mostly as seen in scenarios 2 and 6 since it has the highest reduction in externalities. It could reduce car access to the city centre and introduce the AM to support public transport and replace all car trips in these areas. Relying on the users’ acceptance, the most realistic approach is to introduce the AM gradually, first as part of a MaaS service. Where AM are filling existing mobility gaps rather than completely replacing individual mobility, as the users’ acceptance increases, passengers would switch more and more to the AM instead of relying on their cars. If this introduction is accompanied by urban policies such as road pricing and no-car zones, it will further deter citizens from using their cars in the city centre, which would increase the modal shift to the AM, active mobility and public transport.

Deploying the AM in suburban areas in Luxembourg also shows a reduction in externalities. This strategy fits with the canton’s plans to reduce car use and improve connections to train stations to better serve cross-border travellers.

The scenarios with the externalities calculations provide indicators on the recommended deployment strategy that would fit with a European city’s sustainable development. In conclusion, AM in MaaS presents a significant reduction in externalities, and it is a scenario supported by social acceptance as shown in the scenario description. This scenario also supports the public transport network, which makes it more sustainable. In contrast, the robotaxis scenario would increase externalities and compete with public transport. It is more convenient but also would lead to more congestion and pollution compared to the AM. However, the model has some limitations that are described in the following part.

8.1 Limitations of the Model

It is important to present the limitations of the model and where it falls short. It is a given that the accuracy of the results is dependent on the accuracy of the input data. For instance, Luxembourg and Copenhagen case studies would benefit greatly from updating the mobility behaviour data to better reflect the areas in question (city centre or suburban ring).

This study provides important results on AM as a new force capable of transforming the transportation sector. It foresees scenarios of embedding AM in societal and political contexts and their impacts.

It has the potential to be more robust if the following limitations are addressed:

The external noise costs depend on the population density, traffic status and time of day. However, the data on a country level for these specific contexts is limited. This requires more fine-tuning to reflect the contextual impact.

Similarly, monetising congestion is complicated as it depends on an EU-level study to reach national marginal costs (Jochem et al., 2016). The meta-analysis requires inputs of speed-flow functions, demand curves and value of time (VOT) at a large scale and poses a challenge to downscale to smaller cities.

The use of national-level marginal costs for city-level assessment is a limitation. For instance, air pollution national external costs might underestimate those on an urban level (PM emissions differ between rural and urban areas). This could be addressed using the European values for urban and rural parts in a sensitivity analysis.

We should also consider whether the use of AM will increase the use of traditional public transport and estimate this increase in modal share. Notably, the study of the induced demand requires more analysis since it is difficult to predict the exact numbers for this rebound effect.

An optimised assessment should also account for the potential demographic changes and their effect on VKM (in Geneva, between 2000 and 2015, there was an increase in travel demand by 10% attributed to the increase in the population, for example).

The reliance only on one source to estimate the marginal costs (the Delft handbook of external costs by van Essen et al. (2019) gives one version of the analysis: a sensitivity analysis using different marginal costs would make the model more robust.

A further limitation is accident marginal costs estimation when AV interacts with human driving. The methodology of calculation is still evolving with growing knowledge.

8.2 Rebound Effects

As mentioned in the limitations, the deployment of the scenarios should account for potential rebound effects. Since this analysis tries to imagine potential future scenarios based on current data and observations, it is limited to what could be extrapolated quantitatively. The study of rebound effects thus is better addressed qualitatively.

Hymel et al. (2010) describe the rebound effect as an increase in vehicle usage due to an unintended effect of raising fuel efficiency and decreasing the cost of car usage through policies and technological advancements. They also mention induced demand as a rebound effect to road transportation where an increase in infrastructure capacity attracts new traffic and causes it to reach its capacity, which was not the intended goal (Hymel et al., 2010). The AM would provide a convenient, affordable and safe option to passengers (Onat et al., 2023). Thus, it could lead to more trips, as it provides more trips to people who were not commuting with vehicles in the first place, such as children and the elderly. In addition, it might cause a secondary modal shift after its implementation. It could reduce active mobility and public transportation shares even further than first predicted (Childress et al., 2015; Fagnant & Kockelman, 2018; Zmud & Sener, 2017). This might lead to a vicious circle of deploying more vehicles to meet the new demand. Then, as an unintended effect, more people shift to using the AM. Thus, the operators will need to deploy even more vehicles to meet the increasing demand. Thus, the AM would aggravate the traffic congestion and increase the environmental footprint. Hence, the rebound effect undermines the gains from reducing the use of individual mobility by causing new external costs because of reducing walking, biking and public transport trips. Ergo, it is crucial that the deployment is accompanied by a regulatory framework to monitor the introduction of the AV in the transportation system and reduce potential rebound effects (Möller et al., 2019).

8.3 Policy Recommendations

To summarise replacing all cars has the best impact for reducing external costs but due to the expected problems with the citizen acceptance a political problem for implementing would probably raise. AM in MaaS seems thus to be the best solution for the cities as external costs can decrease substantially. At the same time, positive externalities through network effects can further be expected making the transport system more efficient for the TPOs and reliable for the citizen. The combination of AM and MaaS would enable a change of the mobility paradigm enabling sustainable mobility and a reliable and improved transport system. Last but not least, AM in MaaS would consequently enable SUMP without dos and don’ts. Traditional (unpopular) push and pull strategies (see, e.g. (TUMI, 2018)) could so become obsolete: the results of a representative survey among 1816 citizens (of which 1526 have privately owned vehicles) in Lyon, Copenhagen, Luxembourg and Geneva confirm that 45% of car drivers are “willing” (22%) or even “very willing” (23%) to give up using their own car to use AM to bridge the first and the last mile if this were available. If the service is on-demand and door-to-door, the acceptance could be even higher (Korbee et al., 2022).