2012 ESRI UC

Key Factors Affecting Journey to Work in Melbourne using Geographically Weighted Regression

Dr Daniel Paez Universidad de los Andes

Daniel Paez dpaez@uniandes.edu.co

This paper presents key factors that affect journey to work in Melbourne

Introduction

Research context

Methodology and model

Results and conclusions

Daniel Paez dpaez@uniandes.edu.co

This paper presents key factors that affect journey to work in Melbourne

Introduction

Research context

Methodology and model

Results and conclusions

Daniel Paez dpaez@uniandes.edu.co

This paper presents results from an empirical analysis using census data and GIS
Traffic congestion is widely recognised as a growing problem focussed on peak periods and commuting travel by cars effectiveness of measures depends greatly on our understanding of the factors driving car use This paper presents the results of an empirical analysis using geographic information systems Census data The focus of the work is car commuting in Melbourne, Australia. Analysis part of a wider study of car dependence
Daniel Paez dpaez@uniandes.edu.co

This paper presents key factors that affect journey to work in Melbourne

Introduction

Research context

Methodology and model

Results and conclusions

Daniel Paez dpaez@uniandes.edu.co

There are some common factors in previous research but local conditions change their relative importance

Based on a literature review, factors affecting car choice for JTW are:
car ownership access to transit distance from the CBD CBD employee share To some extent urban residential and employment density

However there is some variation in specific factors

degree to which each is of influence local conditions
Daniel Paez dpaez@uniandes.edu.co

A geographic weighted regression explored the spatial pattern of regression results

Characteristics This technique allows local as opposed to global models of relationships to be measured and mapped Main assumption is that spatial phenomena (e.g. car dependency) will vary across a landscape This tool generates a separate regression equation for every feature analyzed in a sample dataset as a means to address spatial variation

Daniel Paez dpaez@uniandes.edu.co 7

Geographic weighted regression has both advantages and disadvantages

Advantages Regression-based models largely ignore spatial variation, much to the detriment of spatially varying relationships (e.g. proximity to train stations) In most cases improves the certainty of the model Disadvantages Not commonly used and, therefore, results are complex to analyze The spatial distribution of coefficients is developed finding smooth patterns and they should only be used to identify trends and not predictions

Daniel Paez dpaez@uniandes.edu.co

This paper presents key factors that affect journey to work in Melbourne

Introduction

Research context

Methodology and model

Results and conclusions

Daniel Paez dpaez@uniandes.edu.co

A four step methodology was used in this research

Daniel Paez dpaez@uniandes.edu.co

Eleven variables were explored as factors affecting JTW in Melbourne

Variable
Public Transport Supply Public Transport Ranking Residential Density Distance to Rail Station Distance to Melbourne CBD Distance to Local Activity Centre Distance to Arterial Road Distance to Highway Provision of Roads

Description
Estimation of the level of service of Public Transport in each CCD based on access, frequency and spam Relative ranking of each CCD in relation to PT supply Total number of people divided by area

Distance to Business Zone Linear distance to the closest business 1 planning zone
Linear distance to the closest metropolitan train station Total distance using the shortest path in the road network Linear distance to the closest activity centre (CAD, PAC or MAC) Linear distance to the closest arterial road Linear distance to the closest Highway Level of service in relation to roads within the CCD Ratio between the Cycling network in the area and total number

Daniel of persons Provision of Cycling Paez dpaez@uniandes.edu.co

This paper presents key factors that affect journey to work in Melbourne

Introduction

Research context

Methodology and model

Results and conclusions

Daniel Paez dpaez@uniandes.edu.co

The standard regression model produced a limited result

SUMMARY OUTPUT Regression Statistics Multiple R 0.756081099 R Square 0.571658628 Adjusted R Square 0.570957196 Standard Error 0.083805347 Observations 5506

Daniel Paez dpaez@uniandes.edu.co

The SWR improved fit of the model significantly

NAME Bandwidth ResidualSquares Eec>veNumber Sigma AICc R2 R2Adjusted VALUE 0.27319257992 22.34178023090 58.97956002890 0.06404413809 -14590.88225500000 0.75207654510 0.74943758074

Daniel Paez dpaez@uniandes.edu.co

In provision of public transport appears to be more relevant in south west areas

Geographic distribution of the regression coefficient for PT supply

Daniel Paez dpaez@uniandes.edu.co

Proximity to the CBD is relevant to car use across Melbourne with the highest impact in the Inner and Northern areas
Geographic distribution of the regression coefficient for Distance to CBD

Vehicle Usage for JTW

10

Distance to CBD

Daniel Paez dpaez@uniandes.edu.co

High residential densities have an impact on reduced car use in growth areas and in the inner city
Geographic distribution of the regression coefficient for population density

Importance of population density in relation to driving choice

Distance to CBD

Daniel Paez dpaez@uniandes.edu.co