Urban Transitions Conference 2022

JIBE research was showcased at the Urban Transitions Conference 2022 held in Sitges, Barcelona, Spain. See abstracts below.


Built environment determinants of transport mode choice: Single and multilevel approaches

Dr Tayebeh Saghapour1, Dr Lucy Gunn1, Prof Gavin Turrell1, Carl Higgs1, Prof Billie Giles-Corti1, Dr James Woodcock2, Dr Belen Zapata-Diomedi1, Afshin Jafari1, Corin Staves2, Dr Dhirendra Singh1, Dr Alan Both1, A/Prof Liton Kamruzzaman3, Dr Labib S.M.2, Dr Haneen Khreis2, Dr Audrey de Nazelle4, Dr Ori Gudes5, Irena Itova6

1 RMIT University, Australia. 2 University of Cambridge, UK. 3 Monash University, Australia. 4 Imperial College London, UK. 5 University of New South Wales, Australia. 6 University of Westminster, UK


Globally, urban decentralisation in cities is characterised by low-density, single-use zoning, and disconnected street networks. Cities designed with these built environment (BE) characteristics are often dominated by the use of private cars as the main mode of transport, limiting opportunities for physical activity through active travel and contributing to greenhouse gas emissions. Scientific evidence supports that there is an association between the BE and transport mode choice, however, limited efforts have been made to consider the effect of clustering in the household travel survey data. Clustering can occur when survey participants are sampled from either the same household or the same neighbourhood, which increases the likelihood that they exhibit similar behaviours. Not accounting for clustering can yield inefficient standard errors within regression model estimation, impacting inference in identifying associations between the built environment and transport mode choices. This study tests the influence of clustering by fitting single-level and multi-level nested logit regression models to explore the impacts of population density, social infrastructure access (e.g., libraries, community centres, healthcare), and street connectivity on transport mode choice in Melbourne, Australia.  Households’ home locations from the Victorian Integrated Survey of Travel and Activity (VISTA, 2012-2016) were joined to BE indicators at the geocoded address point level. Results show that ignoring cluster membership in the single-level nested logit model leads to underestimating standard errors in mode choice models and increasing the likelihood of incorrect inference. For example, street connectivity and population density were statistically related to walking and bus use in the single-level model, while no significant associations were found for these variables in the multilevel model. Findings indicate that when applying mode choice modelling, more consideration of the structure of the data is required.


Area-based vs. route-based indicators of the built environment: What explains active transport behaviour better?

Dr Ori Gudes1, Prof Liton Kamruzzaman1, Mr Corin Stave2, Dr Alan Both3, Dr Irena Itova4, Dr S.M. Labib5, Prof James Woodcock2, Dr Dhirendra Singh3, Dr Lucy Gunn3, Dr Afshin Jafari3, Prof Billie Giles-Corti3, Dr Tayebeh Saghapour3, DR Jenna Pante2, Dr Haneen Khreis2, Dr Belen Zapata-Diomedi3

1 Monash University, Australia. 2 University of Cambridge, UK. 3 RMIT, Australia. 4 University of Westminster, UK. 5 Utrecht University, The Netherlands


Understanding the impact of Built Environment (BE) characteristics on active travel (walking & cycling) in cities has become increasingly important. There is growing evidence that active travel can have a range of positive outcomes for the wider community. Several studies have shown that active travel behaviour (frequency, duration, route choice, etc.) is affected by BE characteristics including access to public transport, population density, perceived pedestrian safety, street design, and others. Traditionally, BE indicators have been measured in area-based form, usually within a neighbourhood boundary.  Recent studies suggest a new methodology in which indicators are measured at the level of the street network link (route-based form). However, our knowledge about the ability to better explain active travel behaviour using route-based BE characteristics at the network level remains limited.

Our study aims to understand the difference in mode choice models for transport of using area-based vs route-based BE characteristics as explanatory variables. We hypothesise that a stronger influence of the BE in promoting active transport use can be observed using finer street-based BE measures rather than traditional aggregate area-based measures. To test this, we developed mode-choice models using both types of indicators to compare their statistical significance and influence. Our data sources are travel diary surveys from Melbourne and Manchester. These surveys include spatially precise origin and destination coordinates, which were used to link records with both types of BE indicators.

Our findings provided insights on which BE characteristics were found to be more influential on active transport behaviours. The findings verify our hypothesis, which indicates that BE interventions closer to street networks would produce better policy outcomes aimed at improving active transport use. Insights from this study can be used to inform active travel interventions and to better plan for the post-pandemic era, accommodating rapidly changing mobility patterns in our cities.


Exploring walking and cycling accessibility using network-based built environment indicators of safety, comfort, and attractiveness

Mr Corin Staves1, Mr SM Labib2, Ms Irena Itova1, Ms Qin Zhang3, Ms Jenna Panter1, Ms Aruna Sivakumar4, Ms Audrey de Nazelle4, Ms Belen Zapata-Diomedi5, Mr Ali Abbas1, Ms Tayebeh Saghapour5, Ms Lucy Gunn5, Mr John Gulliver6, Mr Afshin Jafari5, Mr Dhirendra Singh5, Mr Luke Knibbs7, Mr James Woodcock1, Mr Rolf Moeckel3, Ms Billie Giles-Corti5

1 University of Cambridge, UK. 2 Utrecht University, The Netherlands. 3 Technical University of Munich, Germany. 4 Imperial College London, UK. 5 Royal Melbourne Institute of Technology, Australia. 6 University of Leicester, UK. 7 University of Sydney, Australia


Accessibility describes the extent to which the built environment (BE) facilitates travel to activities such as workplaces, healthcare facilities, shops, and recreation grounds. Existing literature has identified accessibility as a key indicator of land value, residential location choice, and travel behaviour decisions. Although there is a growing body of literature on walking and cycling accessibility of urban regions, most rely on neighbourhood-based BE measures. This study expands on this existing literature by (1) using a detailed dataset of street network-based BE indicators and (2) using a synthetic population to explore accessibility inequalities at high spatial and demographic resolution.

For the Greater Manchester region, we calculate Hansen accessibilities which incorporate land use (i.e., the density and types of nearby activities) and the street network (i.e., the ability to reach them efficiently via comfortable, pleasant, and safe routes). For the land use component, attraction is calculated for different activity types (e.g., shopping, recreation, healthcare) using points of interest data. For the street network component, we develop composite disutility functions for walking and cycling which incorporate detailed network-based indicators including infrastructure quality and separation, surface type, gradient, car speeds and volumes, street lighting, number of crossings, vegetation, and green visibility. Our disutility functions vary for different user types (e.g., cyclists of different ages), acknowledging that some individuals are more willing to use certain streets that others may perceive as ‘unsafe’ or uncomfortable. We also develop a spatially detailed synthetic population for Greater Manchester using an iterative proportional updating method with data from the English census and housing surveys.

The results from this methodology are used to explore spatial variations in walking and cycling accessibility for different user and activity types. Using the synthetic population, we explore demographic variations in accessibility with age, sex, income, and household structure.