Abstract

In 2020, the mayor of Paris announced the ambitious goal to redesign the city for pedestrians to reach amenities and facilities for daily use within a 15-minute walk. Other cities, such as Melbourne or Portland, have also set themselves this goal. This objective has propelled older initiatives to improve the walkability and pedestrian friendliness of cities to the influential sphere of the tabloid press, giving a new boost to sustainable urban planning paradigms that emphasize the long-neglected importance of active mobility. In this context, researchers in public health and spatial sciences have worked on measurement tools of walkability for almost two decades now. Based on the widely acknowledged theories of the three Ds (density, diversity, and design), multicriteria indices, such as the IPEN walkability index, have shown how city structures can be assessed for their framing conditions for people to walk for a large share of activities. Such tools, however, have been difficult to operate for urban planners due to demanding requirements for data sources and software methods, including GIS network analysis capabilities. Therefore, this paper aims to provide a new assessment method for the three current foci of research on the mobility transition: the 15-minute city, the one-hour metropolis, and car-dependent structures. Based on an open-source tool, we could classify the prevailing mobility structure on the neighborhood level. The application of this tool allows for data-driven management to implement and evaluate plans and urban design policies to identify and assess the “15-minute city.” We will showcase for the cities of Paris, Melbourne, and Portland how outcomes of the tool help to unearth formerly unseen city live dimensions and legitimize urban design projects to mitigate and monitor deficiencies.

Keywords: sustainable mobility, walkability, public transport, car-dependency, mobility assessment.

Introduction

Urban mobility structures and transport systems are of central importance for sustainable development. On the one hand, they are the “maker and breaker of cities” (Clark 1958)when they support or limit the mobility of people and transportation of goods. On the other hand, the dominance of motorized forms of transport is causing stress to city environments far beyond the carrying capacity of environmental systems. This applies to local emissions as well as cumulative downstream effects contributing to climate change. As a side effect, people’s health suffers from exposure to air pollution and noise as well as inactive lifestyles (WBGU 2016; Naess 2006).

The ubiquitous dominance of private motorized mobility is determined as a driver for negative urban developments in terms of urban sprawl and functionally segregated areas (WBGU 2016; Kosonen 2015, 534-35). The redesign of urban structures to support nonmotorized and active forms of travel is therefore a dominant objective for a socioecological mobility transition in the twenty-first century. To this end, transport planning is increasingly adopting a comprehensive understanding of mobility as a complex sociotechnical system. Political, academic, and civic initiatives advocate a redesign of urban structures for people-friendly forms of transport. At the same time, critical researchers point to a mismatch between the sustainable mobility rhetoric and actual developments on the ground. This is obvious when measures such as carbon emissions per person continue to increase in cities trumpeting their sustainable development policies. The effects are more difficult to assess when showcase examples of sustainable neighborhoods focus on privileged local conditions and neglect their embeddedness in a regional and national transport system.

Against this background, our paper introduces new assessment methods for the three current foci of research on the mobility transition: the 15-minute city and the one-hour metropolis with a positive association and car-dependent structures with a negative association (see, e.g., Kosonen 2015, 521). Whereas the 15-minute city is largely concerned with local neighborhood access to services and amenities by active modes of mobility (walking, cycling), the one-hour metropolis relates to mode integration based on an excellent public transport infrastructure for a city region. Car dependency, by contrast, is defined by “the lack of alternative transport modes to the car in terms of time, cost and effort in accessing destinations” (Wiersma, Bertolini, and Harms 2021, 1314).

The opening sections describe the need for assessment methods based on a critical review of existing concepts and theories in the literature and in planning practice. This review is accompanied by a description of technical innovations in spatial analysis and supporting data infrastructures to advance measurement approaches. The following sections show sample implementations and interpret the results of a new measurement approach targeting pioneers of the 15-minute city idea: Paris, Melbourne, and Portland. The paper concludes with a critical discussion of the objectiveness of this approach in evaluating the policy objectives with regards to urban mobility structures in these cities.

Do digital planning tools reinforce path dependencies in transport planning?

Digital tools in transport planning have evolved substantially since their conception in the famous Chicago and Detroit transportation studies of the 1950s. At the time, pioneering models in the United States (and later in the United Kingdom) were mostly concerned with infrastructure demand for a rapidly increasing number of automobiles that required new infrastructure. Refined software packages for transport modeling evolved in due course, with a focus to integrate all transport options in order to comprehensively predict transport demand for future mobility needs. The main inputs were observed activities across spatial units of different land uses (e.g., residential, business, recreational, mixed use) that entail the movement of people and goods and the characteristic properties of such units to attract trips of various purposes (e.g., working, shopping, education, living) and at different times (morning and evening peaks, weekends, vacation periods). The data is sourced from household surveys where people state their current mobility patterns, which in turn is used to calibrate transport modeling scenarios for future transport demand (Boyce and Williams 2015).

This latter point is crucial in understanding why such modeling approaches potentially lead to problematic path dependencies in transport planning: under quasilinear conditions of growth and increasing wealth, application of today’s mobility preferences to a growing number of users must ultimately lead to capacity shortages and a requirement to enhance capacities in the future. Transport modeling software either acts on this requirement with incremental infrastructure interventions until an equilibrium between demand and capacity is achieved or introduces assumptions on behavioral change to acknowledge the limits to extrapolate behavioral patterns of the past into the future (Dalvi 2021).

In this context, an article by Duranton and Turner (2011) comparing detailed data on road capacities and corresponding traffic numbers sparked a wide debate on path dependencies in transport planning. The authors’ intention was to empirically test the “induced demand” hypothesis in transport planning postulated as early as 1972 (Koppelman 1972). This hypothesis states that any road widening that increases capacity generates new demand and will be consumed shortly thereafter. Duranton and Turner (2011, 2618) found significant evidence for the United States that this is the case: “people drive more when the stock of roads in their city increases; commercial driving and trucking increase with a city’s stock of roads; and people migrate to cities that are relatively well provided with roads.”

Interestingly, Duranton and Turner (2011, 2634) also found evidence in their analysis that investments in public transport do not provide the expected road system relief targeted by corresponding transport policies: while “public transit serves to free up road capacity by taking drivers off the roads and putting them in buses or trains,” vehicle drivers who change to buses or trains simply make room for new drivers and vehicles (i.e., demand is induced).

Duranton and Turner (2011) mention people’s relocation to places attractive for motorized transport as a possible cause for their observations. They also hint at misconceptions in influential transport policies that one-sidedly continue to argue for more road space and public transport options to relieve congestion. Following this line of thinking, a model designed by Randelhoff (2016) adds additional explanatory value. His “cycle of structural land use change with increasing car ownership” attributes the “induced demand” to increasing distances and mobility needs that have developed over time (see the highlighted text in Figure 1). Car ownership and the car-oriented transport policies of the past have resulted in a path-dependent competitive advantage for motorized transport. “Induced demand” can therefore be seen as a self-reinforcing cyclical dynamic between the land use structures made possible by motorized transport and the continued demand for new transport infrastructure (UN-HABITAT 2013, 197).

Litman (2017) uses two revelatory citations by influential transport officials to pinpoint the root causes of such vicious cycles. Paradigms of the past that are still deeply ingrained in road engineering guidelines frequently follow the logic of “increasing highway capacity is equivalent to giving bigger shoes to growing children.” This stance is contrasted by a view that acknowledges limits to growth when ‘widening roads to ease congestion is like trying to cure obesity by loosening your ’belt’ (Litman 2017, 24). Such contrasting viewpoints can be seen as value-oriented interpretations of priorities for the role of transport in a socio-technical system. The choice of digital planning tools can reinforce such priorities in decision-making processes, for example when providing path-dependent planning options for deliberation without further consideration of alternatives. Where critical views on path-dependent growth are clearly on the rise (Banister 2008, 78-79), the role of digital planning tools is only starting to become the subject of critical reflection. Authors such as Pavlovskaya (2018, 13) point to an underexposed proliferation of future development options with neoliberal logics but also emphasize the potential of spatial ontologies in maps to facilitate social transformation.

In this context, advocates of socioecological transformation are calling for a fundamental mobility transition to a transport system that attaches high value to social equity and sustainably mitigates adverse health and environmental impacts (Banister 2018; Schneidewind 2018). Showcase examples of sustainable mobility in new local developments are frequently cited as blueprints for people-oriented mobility in cities such as Freiburg, Germany (Barton 2016, 69-72). Critical commentators, however, find this praise misplaced. It ignores significant rebound effects, such as the dynamics of displacement and the heightened segregation of transport-inducing land uses in the city region (Mössner, Freytag, and Miller 2018, 6-8).

This point highlights the contradictions between academic descriptions of sustainable mobility and their current impact, looking beyond prototypical best-case examples. A recent contribution by Nikolaeva et al. (2019, 346) points to the logics of scarcity as a driver of mobility planning, where savings of oil, finance, space and time motivate transport policymakers in Western countries to call for greener, smarter, and cheaper mobilities. Such scarcities could theoretically be managed by an austerity approach aimed at redistributing remaining mobility options fairly, giving highest priority to carbon reductions and just transitions. In reality, however, opposing viewpoints attributed to neoliberal politics lead to “lock-in” situations in democratic decision-making processes, with path-dependent mobility patterns prevailing. A new alternative advocated by Nikolaeva et al. (2019, 353) is the “commoning” approach: “commoning mobility can […] be understood as a process that encompasses governance shifts to more communal and democratic forms while also seeking to move beyond small‐scale, niche interventions and projects.” The authors argue that this is necessary for a truly transformative mobility transition where the relationship between humans and mobility is reconfigured toward “shared responsibilities for what mobility does to societies and communities” (Nikolaeva et al. 2019, 356).

The commoning approach could help inspire a paradigm shift at the policy level. At the same time, it remains to be defined what commoning actually means for planning practice. In this context, Rammert (2021) posits that mobility governance in Germany has a persistent institutional bias toward transport planning for economic growth. He strongly argues for integrated measurement methods able to assess a mobility system in a way similar to the human development index or other socioeconomic indicators. In order to conceptualize this approach, Rammert (2021, 250, 257-259) deconstructs individual mobility as the sum of structural framework conditions, individual preconditions for mobility preferences, and subjective perceptions.

This logic resembles the so-called “person-environment fit” that first emerged in developmental psychology in the towards the end of the twentieth century. When interpreted for behavioral change in mobility, the theory acknowledges the role of lifestyles on the mobility preferences of different groups of people: “person-environment fit theory focuses on the interaction between characteristics of the individual and the environment, whereby the individual not only influences his or her environment, but the environment also affects the individual […]. The adequacy of this fit between a person and the environment can affect the person´s motivation, behavior, and overall mental and physical health” (Holmbeck et al. 2007, 33). In this sense, Rammert’s (2021) structural framework conditions stand for the environment, whereas his dimensions of the individual preconditions for mobility preferences and subjective perceptions characterize the person.

The added value of the developmental psychology perspective lies in an understanding of people’s capacities to adapt to their transport environment and—conversely—the need to reconfigure the transport system in line with people’s needs and preferences. A prominent example is the “15-minute city” that has recently emerged as a model for urban development. This concept attaches greatest value to pedestrian accessibility to resources at a neighborhood level. It has also been labeled as a “planning eutopia” (Pozoukidou and Chatziyiannaki 2021, 20) and is strongly contested by urban development theorists, such as Edward Glaeser: “I am very worried that a focus on enabling upper-middle income people to walk around in their nice little 15-minute neighborhood precludes the far larger issue, which is how do we make sure our cities once again become places of opportunity for everyone? I am only interested in urban planning concepts that fundamentally solve that and I cannot see how the 15-minute city does” (London School of Economics and Political Science 2021).

At the same time, Glaeser suggests adopting the most valuable features of the 15-minute city concept—high levels of accessibility and less driving, supported by congestion pricing and tightened on-street parking regulations (London School of Economics and Political Science 2021). Pozoukidou and Chatziyiannaki (2021) express their reservations from a methodological point of view, looking at the assumptions of researchers when measuring the pedestrian friendliness of urban structures. Deconstructing different measurement algorithms, they reapply them in Vancouver, Canada. The results show that such “walkability tools” are inconclusive in areas outside the most prominent walkable neighborhoods. The authors attribute this shortcoming to the lack of empirical evidence that assumed dimensions of walkability actually exert on people’s travel behavior (e.g., street connectivity, land use mix).

This finding is in line with the research conducted by one of this paper’s authors in Stuttgart, Germany. Standard walkability tools do not explain mobility behavior, with people living in walkable neighborhoods not necessarily walking more (Reyer et al. 2014, 5859-60). This observation inspired additional research funded by the German Research Foundation and aimed at collecting empirical evidence on the mobility preferences of social groups in cooperation with psychologists.1 The objective of this project is to provide input for the development of an assessment tool for the person-environment fit at neighborhood level.

For the purpose of this paper, a provisional version of this tool has been modified to incorporate assessments of areas accessible with high-quality public transport and areas that are car-dependent. The following sections explain the methodological approach and show results for prominent examples of the 15-minute city in three cities on three continents: Paris, Portland, and Melbourne. The research design provides an initial assessment for the status quo.

Research design: Concepts, data, and methods

The empirical section of this paper unfolds this assessment based on the work of Newman and Kenworthy (1999) and Wiersma (2020), who describe car dependency as the dominance of automobiles regarding land use, infrastructure, and transportation. Drawing on other work, we conceptualize this definition with a measurement of car dependency. This includes comparative assessments of accessibility and quality of public transport as well as pedestrian accessibility to important infrastructures (Fina 2015; Siedentop, Roos, and Fina 2013, 330-32).

The research design aims to reapply these methods for a classification of three different inner-city types, in line with the debates outlined in the previous section: the walking city, the transit city, and the automobile city. Newman, Kosonen, and Kenworthy (2016) describe the main characteristics of these types (see Table 1).

In order to classify a neighborhood’s prevailing mobility structure, the method requires data sets suited to our analysis methods. We make use of OS-WALK-EU, an open-source tool initially designed to measure walkability at neighborhood level in the person-environment fit project described in the previous section. The current version of this tool2 rearranges indicators for walkability to combine aspects of pedestrian accessibility (e.g., facilities and services) with information on the built environment (e.g., slopes, density, recreational areas).

In a first adaptation step for the case study cities, we set accessibility thresholds in the tool to a 15-minute walking radius around the geometric centroids of equally spaced grid cells of 500x500 meters, a size that stands as a proxy for local neighborhood accessibilities. All data sources and analysis components are open source and available at no cost. For the processing and network calculations we make use of QGIS and the OpenRouteService. The source data was extracted from user generated data in the OpenStreetMap (OSM) data repository. All following indicators are calculated based on the resulting isochrones. The tool was further enhanced to incorporate public transport timetable information available from online local transport authority repositories complying with the general transit feed specification (Google 2020). All used indicators and components are listed in Table 2 and explained in the following subsections in more detail. 3

Proximity to facilities and services

The local provision of facilities and services is an essential feature of neighborhood walkability and a key requirement of the 15-minute city (Moreno et al. 2021, 97; Reyer et al. 2014, 5854-55). Many day-to-day trips can be categorized under the “visit-live-work” triangle (Dovey and Pafka 2017, 255-56). Whereas places of residence and work are highly individual, “visit” destinations are frequently public and subject to planning decisions. Some measurement approaches use shop floor space as a proxy for “visit” activities, focusing on the trip purpose of shopping (Dovey and Pafka 2017, 251; Frank et al. 2006, 77). In our approach, pedestrian accessibility means proximity to a variety of facilities and services (see Table 3). The data source for the locations is OSM. Their importance for residents can differ with distance and type of facility or service. For this reason, some methods make use of distance calculations and distance decay functions to weight the accessibility of nearby facilities and services (Walk Score 2011; Otsuka et al. 2021). Our measurement concept simplifies this logic by a count and weighting of facilities and services by type within a 15-minute walking radius.

Table 3 shows the categories, their importance, moderation by variety, and maximum scores of facilities and services. The service quality of supermarkets, for instance, is very important (highest value of 5 in the importance column), with one facility (variety column) being sufficient to reach this maximum score. In contrast, single entertainment facilities are less important (lowest value of 0.5 in the importance column), but they can add up to a maximum score of 2, provided that there are at least four entertainment facilities (see the value of four in the variety column) within the search radius. This assessment logic has been adopted from empirical evidence on the relevance of facilities and services for residents’ (daily) activities (see, e.g., Ahlmeyer and Wittowksy 2018, 532; Regionalverband Ruhn 2017, 5; Walk Score 2011).

Functional land use mix

Proximity to facilities and services relates to daily errands in the neighborhood. The mix of land uses is a complementary indicator. A high mix frequently reduces the need to travel longer distances for all trip purposes in the “visit-live-work” triangle: “an effective mix shortens the distances between where we are and where we need to be” (Dovey and Pafka 2017, 249). To rate the land use mix (source: OSM), we decided to use Shannon’s Diversity Index (SHDI). Originating in ecology to assess biodiversity, this indicator has seen a multitude of adaptations for other purposes in spatial analysis (Cegielska et al. 2019, 2; McGarigal 2015; Nagendra 2002, 176-79). It includes two diversity components: richness and evenness (Spellerberg and Fedor 2003, 177-178). Richness refers to the number of species (here: land use types) in a given area or community, while evenness concerns the population in a land use class and its distribution over all classes (here: land use configuration). To calculate the SHDI, we use equation (1):

$H^{'} = \ - \sum_{}^{}{p_{\text{i\ }}*\ln{p_{\text{i\ }},\ with\ p_{i} = \frac{n_{i}}{N}\ }}$ (1)

where N equals the number of blocks and n_i the number of blocks belonging to land use type i. When all individuals are equally distributed over all blocks, the result is the maximum value of H’, which ranges from zero to five. To better interpret the results, we use the true diversity. Here, H’ is the power for the base e (Eulerian number), which changes the value range from one to the number of land uses. As we make use of seven land use categories, the SHDI varies between one and seven. One stands for a perfect monofunctional and seven for the most diverse land use mix (see Table 4).

Green spaces

Previous research confirmed that, in addition to the indicators shown so far, a neighborhood’s design and attractiveness, together with the streetscape, influence walkability (Adkins et al. 2012, 501-3; Frank et al. 2009, 924-25; Vale 2015). We proxy this aspect with an assessment of green spaces as a fundamental building block of neighborhood attractivity. Green spaces and their function as a recreational resource for residents are also positively associated with physical and mental health and motivate social interaction (Picavet et al. 2016; Ward Thompson et al. 2012, 226-27). The calculation procedure for this indicator computes the share of green areas and woods within a 15-minute walking radius. Table 5 shows how these shares contribute to the overall score in the assessment.

Pedestrian network

The directness and connectivity of the pedestrian network have also frequently been described as an influential dimension for walkable neighborhoods. Gori, Nigro, and Petrelli (2014, 38) consider this “[…] a key component in the development of a more sustainable mobility system […].” We make use of the pedestrian radius (Pr) as a proxy for street connectivity and the directness of pedestrian routes. Pr sets the area of the 15-minute walking radius in relation to the area of the highest possible connectivity, which is a perfect circle around the starting point (here: centroid of 500x500 meter grid cells).

For the computation, we use equation (2):

$Pr = \frac{A}{\pi*\ {(v*\ \frac{15}{60})}^{2}}$ (2)

where A represents the area of the 15-minute walking polygon4 and v the walking speed in km per hour. The calculation is based on an average speed of 5 km per hour, an average value for walking speeds of pedestrians in cities. Table 6 shows the value range for this indicator.

Access to high-frequency public transport

Besides the dimensions of walkability presented with the indicators above, we also aim to integrate dimensions of public transport attractivity for the transit neighborhood mobility type. This is done with data derived from the transit feeds of local transport authorities published under the general transit feed specification, a standard that Google originally designed to inform people in online routing applications. The standard provides for georeferenced data sets with information on the mode of transport and its frequency (Google 2020). We adopt a concept developed by the Regionalverband Ruhr (2017, 24-25) and Eichhorn et al. (2020, 8-11) to evaluate the quality of public transport services based on two indicators shown in Table 7. The type and frequency values are multiplied and add up to the total score, which ranges from 1 to 18.

Classifying the neighborhoods

The five indicators described above allow for a typification of mobility structures into three categories: the walking neighborhood, the transit neighborhood, and the car (-dependent) neighborhood. This is done by summing up all walkability indicator scores with resulting value ranges from 1 to 40. For the classification we set threshold values:

• The walking neighborhood is defined by high walkability values. We set a minimum value of 32, which corresponds to 80% of the maximum.

• The transit neighborhood is characterized by accessibility to high-frequency public transport. This requires at least one stop with a minimum value of 8 in the 15-minute walking radius. This is equivalent to a subway station with a minimum frequency of 15 minutes or a rail station with a minimum frequency of 30 minutes.

• The walking/transit neighborhood fulfills the criteria of both types.

• The car (-dependent) neighborhood is defined by a lack of alternative transport options. As we focus on the three main transport modes (walking, transit, and car), a (populated) neighborhood is car dependent when neither of the abovementioned criteria are met.

Besides mapping these types, we add population numbers and newly developed built-up areas to the analysis. This helps to understand how many people reside within each neighborhood type and where population and settlement growth has occurred. The data is sourced from the Global Human Settlement Layer, which provides harmonized information on population and built-up area worldwide. In this context, the term “built-up” refers to all sealed areas (including residential, commercial, industrial, and transport infrastructure). Based on census and remote sensing data, this information is available in a 250x250 meter grid (European Commission 2018; 2019). Values for population and built-up area are aggregated and transposed into the 500x500 meter grid used in our study. In order to assess population and settlement trends in each neighborhood type, we calculate growth rates between 2000 and 2015.

Application for the case studies

We applied the methodological approach described above in three cities strongly advocating the 15-minute city (in the case of Melbourne and Portland: 20 minutes) in their transport policies: Paris, Portland, and Melbourne. The final results of our spatial analysis are shown in Figure 4. Besides mapping the neighborhood types on the 500 m grid, we also calculated the distribution of inhabitants living in walking neighborhoods, transit neighborhoods, or car-dependent neighborhoods in two variations. First, we calculated the population share within city boundaries to verify whether we could classify the inner city as a walking, transit, or a car-dependent city (see Figure 2). In a second analysis, we calculated travel time polygons from the city center (by car) to classify the outskirts based on neighborhood type changes from the city center (max. 5 minutes driving time from the city center) to the suburban fringe (max. 20 minutes driving time from the city center) (Figures 5–7). The use of static buffers would also have been possible. However, we think that the travel time polygon can better represent the urban structures and the spatial connection to the suburban fringe. In addition, we plot the distribution of new population and built-up area in each neighborhood type in Figure 3. Due to the high overlap of walking neighborhoods and walking/transit neighborhoods, we make no distinctions between these types in the graphs. For simplicity, both are classified as walking neighborhoods. The results are interpreted in the following section for each city.

Paris

Paris has recently received considerable attention in the media and in academia for its ambitious goals of retrofitting urban structures under the 15-minute city concept to make it a “city of proximities” (O'Sullivan and Bliss 2020). Our analysis shows that the city center already has a remarkably high number of cells with a high walkability score (Figures 2 and 4). More than 80% of residents in the core city live in highly walkable neighborhoods. These areas are also well endowed with high-frequency public transport. Neighborhoods that are not walkable but have good access to public transport account for about 17.8% of the total population, while just 1.2% live in car-dependent neighborhoods. These areas show concentrations on the southeastern and western part of the city. The three bordering départements differ. In Hauts-de-Seine, the predominant neighborhood type is walking, while there are many car-dependent neighborhoods in Seine-Saint-Denis and Val-de-Marne. From 2000 to 2015, only about 30% of new developed land has been in walking (6.2%) or transit (24.0%) neighborhoods (see Figure 3). However, the analysis also shows that the population growth in car-dependent areas (14.5%) is significantly lower than in walking (28%) and transit (57.5%) neighborhoods. This result hints at a decoupling of new land take from population growth.

The assessment of urban structures is also reproduced when looking at the distribution of residents by dominant mobility type in the commuter sheds of Paris. Figure 4 illustrates the travel time by car from the city center in minutes on the x-axis and the total share of population in percent on the y-axis. The colors represent the dominant mobility types: walking in green, transit in blue, and car in red. We observe that the walking neighborhood is the dominant type in the 5- and 10-minute commuter shed areas. This proportion gives way to higher transit neighborhood shares, up to almost 70% in the 15-minute catchment area. One ring further out, the number of residents in car-dependent neighborhoods increases, although the transit neighborhood remains dominant.

Melbourne

Melbourne integrated the concept of a “20-minute neighborhood” and “living locally” in its long-term planning strategy Plan Melbourne 2017–2050 (Victoria State Government 2019). The plan is to create more inclusive, vibrant, sustainable, and healthy neighborhoods by enhancing walkability and reducing the length of daily trips. In contrast to other concepts, the city stresses the importance of public transport as an efficient connection for work and higher-order services. Implementation in the city has already started with a pilot program, including several case studies.

Our analysis of Melbourne reveals a high share of very walkable areas and neighborhoods with access to high-frequency public transport within the city boundary. Around 35% of inhabitants live in a walking neighborhood and 62% in a transit neighborhood (see Figure 2). Thus, only about 3% of the city population is dependent on cars. The spatial distribution of the dominant mobility types is highly clustered: highly walkable areas are located in the inner city and subcenters in suburban areas. This spatial cluster is entirely surrounded by transit neighborhoods. Car-dependent areas are concentrated in locations on the city outskirts. The visual dominance of the transit neighborhoods is also reflected in the distribution of new population from 2000 to 2015. About 57% of the increase occurred in transit areas, while just 10% of the new population lives in walking neighborhoods. However, the situation is different in the newly developed urban areas. The share of built-up areas in car-dependent neighborhoods is 95%.

This pattern is also visible when considering the population share for each neighborhood type in the travel-time-based commuter sheds. The share of inhabitants living in walking neighborhoods in the 5-minute commuter shed is very high (70%) but goes down rapidly as distance to the city center increases. In the 5–10-minute ring, the proportion is 18%, but it goes down to a negligibly low share in the 10–20-minute ring. In this context, Figure 4 shows the cells along radial rail-bound transport axes from the city center to the outskirts. This pattern explains the shares in the second and third ring, with a dominance of transit neighborhoods of around 75%. This value drops to approximately 40% further out. The proportion of car-dependent residents increases steadily with distance from the center. In the 20-minute catchment area, car-dependent neighborhoods are home to nearly 60% of residents belonging to this mobility type.

Portland

Portland, Oregon, has a long history of promoting walkable and cyclable neighborhoods. In the United States, it is one of the pioneers of urban sustainability policies in general and of the 15-minute city in particular (O'Sullivan and Bliss 2020), with the city presenting an analysis of walking accessibility to commercial services and amenities as early as 2010 (City of Portland 2010). The plans were published under the “20-minute neighborhood” label back then and have since been integrated into the Portland plan for Future Possibilities and Choices.

For the case study on Portland, we had to cope with the limitation that the domestic border of the Portland commuter shed extends from Oregon into Washington State. This caused inconsistencies in the availability of public transport timetable data. For this reason, cells with implausible results in the north were excluded from the commuter shed classification.

Figure 4 shows that, compared to Paris, there is a geographically smaller cluster in the Portland city center with highly walkable areas. This includes Downtown, Old Town, and Pearl District and some smaller areas on the southern side of the Oregon River. Starting from the cluster in the western part of the city, the public transport network radiates in all directions except the southwest with neighboring car-dependent cells. The classification procedure identified some elongated transit neighborhoods in the rest of the city region. Besides some smaller subcenters and isolated structures with high walkability and transit accessibility, the rest of the city’s neighborhoods are classified as car dependent. This is also shown when looking at the population distribution: almost 65% of residents live in car-dependent neighborhoods but just 4.3% in walking neighborhoods. This observation is substantiated by the results shown in of Figure 3. Less than 1% of new built-up areas was developed in walkable neighborhoods and only 13% in transit neighborhoods. For new population, we can observe slightly higher shares in walkable (1.8%) and transit (21.8%) areas. Overall, recent development in Portland concentrates in car-dependent areas.

The results of the commuter shed analysis come as no surprise. In the 5-minute catchment area, the share of highly walkable neighborhoods is about 25%, while transit neighborhoods have the dominant share, covering 70% of residents. Patterns change with increasing distance from the city center. The share of inhabitants living in car-dependent neighborhoods exceeds 60% of the total population in the 10-minute commuter shed, going up further to 80% in the maximum travel time ring. An analysis of the individual components of our assessment shows that a lack of facilities and services leads to classifications other than the walking neighborhood in the outer rings. The average walkability value per cell in the first ring is about 26. In the commuter sheds further out, this value drops to 16 (10-minute commuter shed), 13 (15-minute commuter shed), and 11 (20-minute commuter shed).

Discussion

Our study provides a new method for classifying neighborhoods according to their dominant mobility types: high walkability, high-frequency public transport, or car dependency. For the three cities of Paris, Portland, and Melbourne, we were able to identify similarities but also significant differences.

Our main findings can be summarized as follows:

• All cities have highly walkable areas in the city center. However, outside the city center, results differ greatly.

• Paris, one of the most densely populated and oldest cities in Europe, is very walkable within the metropolitan area. Just 1.2% of the population in the core city and 14.5% in the metropolitan region live in car-dependent neighborhoods. Paris is therefore very close to its goal of being a 15-minute city.

• Melbourne, one the largest cities in Australia, has a lower share of inhabitants living in walkable neighborhoods and a slightly higher share living in car-dependent neighborhoods than Paris; 2.9% of the Melbourne population lives in car-dependent areas. Therefore, it is also very close to its goal of being a walkable city.

• In Portland, one of the role models of walkability in the United States, major shares of the population living outside the city center live in car-dependent neighborhoods, which affects about 65% of the population. Portland is by far the most car-dependent city region in this comparison.

Overall, our results are consistent with the findings of other studies, despite some differences in methodology. The high results for walkable neighborhoods in Paris are confirmed in studies that rank it as one of the most walkable cities (Carrington 2020; Institute for Transportation & Development Policy 2020). Due to its historic development as a high-density metropolis, Paris has a more mature urban fabric than Melbourne or Portland. Car-dependent areas are primarily located in suburbs outside the metropolitan region, as visible In the 20-minute catchment area (Motte-Baumvol, Massot, and Byrd 2010, 607-8). Due to a well-developed regional public transport network, transit neighborhoods dominate outside the core city. It is noticeably here where the highest population growth between 2000 and 2015 occurred (see Figure 3), giving effect to the planning paradigm to strengthen public transport and to focus on transit-oriented development (TOD). This paradigm has been pursued for decades and is also part of the strategic plan for greater Paris (Debrincat 2015; Desjardins 2018). Turning to Melbourne, Giles-Corti et al. (2014) developed and applied a walkability index. Although the set of indicators and assessment methods differ greatly, the authors come to similar conclusions, confirming the city’s high degree of walkability. Dodson and Sipe (2008) and Jeffrey et al. (2019) identified areas with a high car dependency in the outer suburbs and around the metropolitan fringe. To some extent, these findings are also reflected in Figure 2. The main difference is that we find larger areas with public transport options, meaning that not that many neighborhoods in the outer areas are car dependent. This also represents the city’s approach to sustainable planning. In its vision for 2030, Melbourne focused on restructuring and strengthening public transport, which also contains the promotion of TOD (State of Victoria 2002). Our results in Figure 2 show that these planning efforts are already visible from 2000 to 2015. However, the outcome is limited to population growth in existing areas of sustainable transport options. In newly built-up areas, car dependency dominates. The success of the “20-minute neighborhood” goals in the Melbourne strategic plan must therefore be monitored over the long term.

Portland presented its own findings on the 20-minute neighborhood a decade ago (City of Portland 2010). Based on its methodology, a high share of inner-city areas is walkable due to the high proximity to facilities and services, particularly in Downtown, Old Town, and Pearl District. Despite Portland’s international reputation for pioneering urban sustainability practices, our results reveal shortcomings in the mobility structures outside the inner core when compared to Paris and Melbourne. We posit that this finding can be explained by the observation that car use and car dependency in U.S. cities are generally higher than in European and Australian cities (Dodson and Sipe 2008, 393-96; Newman and Kenworthy 1999, 26). Our comparison of three global cities therefore provides a somewhat different picture for the status quo in Portland when compared to the self-picture painted by the city’s sustainable transport concepts such as the “20-minute neighborhood” or TOD (City of Portland 2010; Oregon Metro 2020). Figure 2 reveals that most of the population and built-up growth between 2000 and 2015 concentrates in car-dependent areas.

At this point, we would like to remind readers that our results reflect on the structural preconditions for mobility and not on individual mobility preferences. It is important to note this focus when we talk about deficits in the urban fabric. Overall, the results seem to be robust and consistent with other research findings. Nevertheless, we have some limitations to discuss. The simplified measurement methods—compared to other studies we are currently conducting in European city regions—generalize at the expense of more refined weightings of individual indicator components. High-frequency public transport is a case in point. It can theoretically be assessed by such quality criteria as good connectivity, opportunities for changing train/bus/etc., and design elements. Such characteristics can certainly vary across international cities. At this point, we have not yet integrated all of these criteria. This indicator is therefore significantly simplified compared to our walkability assessment, which analyzes routes to infrastructures and sets different weightings. We are aware that there is some potential for optimization, opening up opportunities for further research. For further methodological development, additional components from other methodological approaches could be integrated, such as elements of the Public Transport Access Level of London’s transport authorities (Transport for London 2015).

In addition, other transport modes, such as cycling, need to be included in the future. This could possibly affect the results in Portland, with its high share of bike commuting, at least in comparison to other major U.S. cities (O'Sullivan and Bliss 2020). Furthermore, alternative transport modes, such as car-sharing and e-mobility options, are becoming increasingly important (WBGU 2016, 154). It should also be stated that our results are only as valid as the data sets used. This applies especially to the quality of OSM data. While it can vary locally, it is generally good in world cities such as Paris, Portland, and Melbourne, which feature large numbers of contributors. On the upside, the simplification we offer in this approach provides possibilities for a fast and potentially worldwide classification of mobility structures. Ease of use and simplicity are useful for an intuitive understanding of the structural predicament influencing the mobility transition.

Referring back to the theoretical sections of this paper, we would like to recap the criticism of the concept of the 15-minute city (London School of Economics and Political Science 2021; Pozoukidou and Chatziyiannaky 2021). The idea is that reducing trips could also reduce the amount of business opportunities and social interaction in the city region and lead to further segregation between privileged and disadvantaged households. Whether and to what extent inequalities exist should be examined and monitored using valid socioeconomic data. In this way, it is also possible to check in the long term whether such concepts have negative social effects. In addition, locally improved accessibility can lead to higher property values and downstream effects, such as gentrification and social inequalities (Pozoukidou and Chatziyiannaki 2021, 20-22). Our methodological approach is designed to provide planners and policymakers with assessments identifying shortages in the supply of mobility options, which in turn serve urban structures characterized by the supply of facilities and services as well as public transport.

In summary, we present our results as an explorative component for a more comprehensive system for measuring the mobility transition as proclaimed by Rammert (2021). Our analysis is an option to initially measure the structural conditions he calls for, although we are currently unable to cover his additional dimensions of “individual preconditions for mobility preferences” and “subjective perceptions.” Our analysis is currently limited to an assessment of the status quo. Future monitoring applications, however, can use this as the baseline to monitor if the “cycle of structural land use changes with increasing car ownership” and its resulting car dependency (as shown in Figure 2) can be redeemed for sustainable transport modes.

Conclusion

In this paper, we presented a methodological approach contributing to the current discussion on implementing a mobility transition aimed at achieving carbon reductions and greater equity in transport. The methodology classifies urban structures and their mobility options into walking neighborhoods, transit neighborhoods, and car-dependent neighborhoods. In this way, strategic approaches as requested by, for example, Newman, Kosonen, and Kenworthy (2016, 450-51) can be localized at a small-scale level. The underlying spatial assessment logic is derived from an enhancement of walkability and transit indicators from the literature. This approach enables evaluations of transport policies and remaining challenges in urban mobility structures. The results show to what degree three case-study cities already have high-walkability urban structures, as advocated in the concept of the 15-minute city or, in contrast, the car-independent city. Our assessment shows that despite similar policy objectives, mobility structures especially outside inner core cities differ greatly. However, the methodological approach unfolds its greatest potential when long-term developments are monitored in addition to the status quo. In this way, the success of sustainable mobility planning can be evaluated and counterproductive path dependencies and related negative outcomes can possibly be identified at an early stage.

The literature research presented in this paper emphasizes that an effective mobility transition is hindered by many path dependencies deeply ingrained in transport planning procedures. Academia and transport policymakers are calling for new or modified concepts to accelerate the transition. Measurement methods and monitoring systems such as the one presented here are of fundamental importance in this process. In future research projects, we plan to enlarge and validate the sample with empirical data, to integrate mobility preferences and individual perceptions, and to combine our results with socioeconomic indicators on transport equity.