Paper 62 PEDESTRIAN DEMAND MODELLING OF LARGE CITIES: AN APPLIED EXAMPLE FROM LONDON SPATIAL ANALYSIS Working Paper Series Working Jake Desyllas Elspeth Duxbury John Ward CENTRE FOR ADVANCED CENTRE FOR Andrew Smith Centre for Advanced Spatial Analysis University College London 1-19 Torrington Place Gower Street London WC1E 6BT [t] +44 (0) 20 7679 1782 [f] +44 (0) 20 7813 2843 [e] [email protected] [w] www.casa.ucl.ac.uk http//www.casa.ucl.ac.uk/working_papers/paper62.pdf Date: June 2003 ISSN: 1467-1298 © Copyright CASA, UCL Pedestrian Demand Modelling of Large Cities: An Applied Example from London Dr. Jake DESYLLAS*, Elspeth DUXBURY*, John Ward, Andrew Smith** Centre for Advanced Spatial Analysis, University College London, 1-19 Torrington Place, London WC1E 6BT, UK *Intelligent Space Partnership, 68 Great Eastern Street, London EC2A 3JT **Transport for London Street Management, Windsor House, 50 Victoria Street London SW1H 0TL PEDESTRIAN DEMAND MODELLING OF LARGE CITIES: AN APPLIED EXAMPLE FROM LONDON Dr. Jake DESYLLAS Elspeth DUXBURY Partner Partner Intelligent Space Partnership Intelligent Space Partnership 68 Great Eastern Street 68 Great Eastern Street London EC2A 3JT London EC2A 3JT Tel: +44 20 7739 9729 Tel: +44 20 7739 9729 Fax: +44 20 7739 9547 Fax: +44 20 7739 9547 E-mail:[email protected] E-mail:[email protected] Web: http://www.intelligentspace.com Web: http://www.intelligentspace.com John WARD Andrew SMITH Research Assistant (and Consultant ISP) Senior Service Development and Performance Centre for Advanced Spatial Analysis (CASA) Officer University College London Transport for London Street Management 1-19 Torrington Place Windsor House London WC1E 6BT 50 Victoria Street Tel: +44 207 679 1782 London SW1H 0TL Fax: +44 207 813 2843 Phone: [+44] (0)20 7941 4382 Email: [email protected] Fax: [+44] (0)20 7941 4356 Web: www.casa.ucl.ac.uk/people/John.html E-mail:[email protected] Abstract This paper introduces a methodology for the development of city wide pedestrian demand models and shows its application to London. The approach used for modelling is Multiple Regression Analysis of independent variables against the dependent variable of observed pedestrian flows. The test samples were from manual observation studies of average total pedestrian flow per hour on 237 sample sites. The model will provide predicted flow values for all 7,526 street segments in the 25 square kilometres of Central London. It has been independently validated by Transport for London and is being tested against further observation data. The longer term aim is to extend the model to the entire greater London area and to incorporate additional policy levers for use as a transport planning and evaluation tool. Keywords pedestrian model, land use transportation planning, Visibility Graph Analysis, movement, dynamics, GIS, London 1 1. INTRODUCTION The aim of transport modelling is to predict patterns of movement and the functioning of movement systems, yet research in this field until now has been almost exclusively focussed on motorised transport to the exclusion of other modes. A keyword search of the engineering INSPEC database shows 53 times more articles published since 1969 with ‘vehicle' in the title than 'pedestrian’ (counts of 10,211 and 192 respectively). Perhaps the neglect of pedestrians in the research arose because modelling started at the same time as automobile dependence became a key feature of transport, so attention was focussed on understanding vehicular traffic. With the continuing rise of the car in the last 25 years, walking fell in importance from 35% of all trips in England to 26%. With this rising automobile dependence, the disparity in research between motorised and non motorised modes became ingrained in policy. Gemzøe points out that "there is no city that has a 'pedestrian department', recording the numbers, flow and behaviour of people on foot on the same regular basis as traffic departments record the vehicular traffic, so the pedestrians tend to be invisible in the planning process - because there are no data about them" (2001). Despite its decline, walking still accounts for over 80% of all journeys made under a mile in length (National Statistics, 2001). In big cities, walking is a key component of transport: 40% of all inner London trips are by foot (DETR 1999), yet walking has been almost completely ignored by modellers until recently. As Brög noted; "Walking is neglected in transport policy and planning because it is often not considered in traditional transport models. But even if it is included in behavioural transport surveys, the methods applied are very often inadequate and insufficient to show its relevance for everyday mobility. And from this neglect ... walking is underestimated for transport and town planning." (2001). The transport planners who must regulate and manage our urban movement networks do not currently have modelling tools to help them understand and therefore plan for pedestrian flows. However, there has been growing political pressure to develop more sustainable transport polices in response to automobile dependence and this is beginning to change the agenda for transport modelling. The International Council for Local Environmental Initiatives (ICLEI) formed in 1990, has had some influence on local government planning policy through the Local Agenda 21 campaign ‘dedicated to sustainability through participatory multi-stakeholder sustainable development planning’ (ICLEI 2000). The ‘visions of transport in 2030’ section of the OECD guidelines on Environmentally Sustainable Transport (OECD 2000) includes ‘a focus on reducing the number of long-distance trips and on much greater use of non-motorised means for short distance trips, with a large increase in the provision of supporting infrastructure for non motorized travel’ To achieve this goal, forms of travel such as walking, cycling and the use of public transport must be encouraged, in order to reduce motor traffic and its adverse effects and reduce dependency on fossil fuels. The move to more sustainable transport has led to specific policy changes in London. A key aspect of the Mayor's Transport Strategy (Livingstone 2001) is to "make London one of the most walking friendly cities for pedestrians by 2015". The Central London Partnership (CLP) which represents businesses and Stakeholders in London, considers it essential to promote an increase in walking for short to medium length trips (2001). Transport for London (TfL), which is the strategic transport authority for the city, is tasked with developing a methodology to measure progress against these objectives. In order to meet these objectives, Transport for London and the Central London Partnership commissioned a pedestrian modelling consultancy called Intelligent Space Partnership (ISP) to develop a model of total walking volumes for every node on the street network in Central London. This paper introduces the project and the framework for its policy implementation. The main aim 2 is to outline the new approach that has been adopted in developing a large scale pedestrian model, but some initial findings are also presented. 2. REQUIREMENTS OF THE PEDESTRIAN MODEL The model has been designed to meet the requirements of a strategic transport authority and avoid some of the previous problems of large scale urban modelling (Lee 1972). The requirements and approach adopted are summarised as follows: • The kind of model required is a comprehensive pedestrian demand model. Therefore the output variable was pedestrian flows (measured as average people per hour) on every node in the road network as defined by ordnance survey road centre lines. • The modelling has to be based on standard, best practice statistical techniques. The approach adopted is the standard Multiple Regression Analysis (MRA) technique of statistical inference for modelling the relationship between a series of independent variables on a dependent variable (Rawlings 1988). • The model has to be empirically tested from the start and capable of ongoing tests. The required tests show how well the predicted values correlate with observed values, identifying the accuracy of the model. This has been achieved by constructing the model with MRA using observed data and testing it against pedestrian flows in further observation studies. • The model has to be open to scrutiny and the modelling results had to be independently verified by the transport authority. Using the standard technique of MRA makes this relatively straightforward, as all modelling data was provided for independent verification by TfL engineers. • The model composition has to be extendible, to provide a flexible framework for testing the influence of different factors on pedestrian flows in the future. If new policies need to be tested in the future, these can be accommodated in the modelling process as new variables to be tested in the MRA. • The model has to be applicable at an urban-wide scale in order to be useful as a policy evaluation tool. The modelled area is 25 square kilometres and incorporates all 7,526 street segments within this region. This is considered large for both pedestrian models and some road network models such as Saturn (Van Vliet 1982) designed for 100 to 150 intersections. This ruled out some of the microsimulation approaches (Hoogendorn et al. 2000) as these are not designed for large scale spatial systems. Scalability has been a key issue in the choice of all model components. In summary, the approach adopted was to focus on the development of a flexible and testable pedestrian modelling framework using MRA. The plan for progressive improvement of the model is to implement continuous retrials for new areas or new policy levers, and to use independent validation by TfL. Modelling components developed in this project are also being used by TfL's internal research programme on pedestrian modelling, which is in development phase. With every 3 extension of the model, the importance of all factors will be retested to ensure that the best balance between accuracy and applicability of the model design.