Accessibility Analysis and Modeling in Public Transport Networks - A Raster based Approach Morten Fuglsang 1,2 , Henning Sten Hansen 2 & Bern Münier 1 1 Aarhus University – National Environmental research institute 2 Aalborg University Copenhagen
Jul 07, 2015
Accessibility Analysis and Modeling in Public Transport Networks - A
Raster based Approach
Morten Fuglsang1,2, Henning Sten Hansen2 & Bern Münier1
1Aarhus University – National Environmental research institute2Aalborg University Copenhagen
Structure of presentation
• Prensentation of topic
• Introduction to the case region
• The applied definition of accessibility
• The conducted modeling
• Results
• Evaluation
2
Project topic
• For the Pashmina project, CA based land use change modelling is to be conducted using the LUCIA Cellular autometa model.
• The project outlines policy scenarios for paradigme shifts in transportation for the next 40 year period.
• Modeling is to be conducted in Denmark and in up to two other European cases.
3
Project topic
• One of the ongoing tasks is to create a indicator of accessibility to jobs through public transportation.
• In order for it to be transferable, the data requirements was to be as simple as possible, to facilitate easy transfere to the other European regions.
4
Introduction
• ‘Outskirt Denmark’ is a popular theme in Danish politics, describing regions with :
3.Declining population
4.Poor number og jobs
5.Low service level in terms of schools, hospital services and public transportation
5
Study area
• The capitol region of Denmark
• 9200 km2 area coverage
• Large regional differences in terms of jobs and transportation services
6
Defining accessibility
• Components of accessibility:
2.Land use component
3.Transportation component
4.Temporal component
5.Individual component(Geurs, K. T & van Wee, B. 2004)
• We use three of the four components, and a gravity based accessibility measurement.
7
Data inputs
• Data used for the modeling:
2. Road dataset
3. Trainlines and stops
4. Buslines and stops
5. Metrolines and stops
6. Centers with population and number of jobs
8
Cost surface creation
• Average tavel-speeds was appended to the vector data
• Out of network travel was set to walking pace
9
Cost surface creation 2
• Data was rasterized to 100m resulution, and the differenet rasters where combined
• Finally the travelspeed was recalculated to CCT values
10
Juliao, R. P (1999)
GIS modeling
• Using python, a model was created, that combines the data with the information from the centers, and the coast surface
11
(Coffey, W & Shearmur R. G 2001)
GIS modeling
• The model creates a cost distance calculation for each center – calculating the centers contribution to the overall indicator.
• The center contributions are the summarized into one output layer
12
Result post processing
• Result showed som inconsitency in two areas with high low coverage and few stations/stops
• Furthermore the effect of the constraints was removed for the result
13
Results
• 285 centers was included in the calculation.
• High variability in terms of result scores
14
Results
• Based upon upon the entire dataset, mean accessibility was calculated
• Result classified into +/- 1 and 2 std. deviation.
15
Validation
• Municipality average was calculated based on the clssification scores
• Compared to commuting statistics.
16
Validation 2
17
Validation 3
• The cluster analysis highlights the regions where both accessibility and commuting is generaly low.
18
Strengs and weaknesses
+ Fast calculation time based on raster math
+ Low data requirements
+ Describes commuting trends
+ In line with commuting statistics
19
- Depiction of time in the model does not corrospond to traveltime
- Change of transportation mode and wait time is not modeled
- Region should and will be subdivided.
Conclusion
• Vector methods are much more precise, however this aproach was designed with low data requirements as main goal.
• The model illustrates the regional differences in service availability and align with commuting statistics
• Future work will focus on incorporating a more precise prediction of time to the raster based method.
20