1. Multi-Modal Public Transportation Accessibility and the ARCADIA Project Duncan Alexander Smith CASA Seminar, 17/11/10

2. Multi-Modal Public Transportation Accessibility in GIS Introduce ARCADIA- importance of accessibility in land use transport modelling. Accessibility model- network representation and analysis Results and applications – travel time mapping, commuting sustainability, flooding impacts. Approach limitations- dealing with capacity, wider project challenges.

3. ARCADIA Project Aims- “To provide system-scale understanding of the inter- relationships between climate impacts, the urban economy, land use, transport and the built environment.” Collaborative project between several universities including Newcastle, Cambridge and UCL.

4. Interface for testing of policy options Working with key London stakeholders Climate impacts assessment and adaptation planning Analyse risks of Flooding Drought Urban heat and health impacts Test adaptation options Greenhouse gas emissions assessment module Multi-sectoral emissions accounting tool Detailed sub-modules for transport (personal and freight) Analyse of city-scale energy policies Land use and spatial interaction module Employment Multi-modal transport Developed land cover Population Land use planning constraints and attractors MDM-E3 Multi-sector city-scale economics module Dynamic resource interactions between sectors Specialist energy sector module Demographic scenarios City-scale climate scenarios Temperature Precipitation Sea level rise Storm surge City-scale climate scenarios Temperature Precipitation Sea level rise Storm surge Urban Integrated Assessment Facility

5. Importance of Accessibility Modelling Key areas where accessibility/travel costs applied- Modelling future geography of firms and residents Input to land use -transport model, relevant to SCALE, GENESIS also. Sustainability Impacts of Transportation Energy and carbon emissions from travel patterns. Climate Event Impacts Estimating accessibility impacts of transport infrastructure failures from climate events (principally flooding).

6. Travel Cost Measures Various aspects. Joan discussing car travel, focus here on public transport. Behavioural costs- Travel Time- walk time, wait time, network service time, interchange. Generalised Cost- include monetary costs, weight journey components. Sustainability costs- Mode choice, distance travelled Energy use, Carbon emissions Demand/Assignment costs- Service choices, flows, congestion.

7. Existing Accessibility and Transport Models Government/Consultancy Accessibility Models TfL Capital Model, various Consultancy Models. Journey Planning Websites MySociety Travel Time mapping work. Need Flexibility & Control - T ravel cost measures Network representation (future scenarios) Zones, 2000*2000 matrix Demographic, economic disaggregation Study area of South East…

8. Multi-Modal Public Transport Often in transport modelling ‘multi-modal’ translates as treating PT networks separately. Multiple modes regularly combined in one trip in London context. 2008 Weekday Millions of Public Transport Journeys (TfL) (Analysis of LTDS, NTS needed for in-depth interchange patterns) RailUGBusTotal Trips Main Mode2.22.13.57.8 Journey Stages2.43.25.711.3

9. Public Transport Modes Included Journeys 1.Pedestrian 2.Underground 3.DLR, Overground 4.Mainline Rail 5.Bus 6.Bicycle 7.Car

10. Data Sources, Level of Detail Station Locations Entry and exit interchange points. NAPTAN publicly available. Infrastructure Geography Measure distances between stations. Ordnance Survey provide streets, mainline rail. Timetable Information Available from NPTDR. Highly detailed, all services and timetables. Simplified into AM peak frequency.

11. Street Network Representation Use OS ITN street network (remove motorways) Standard walk speed: 5km/h (too slow for London?)

12. Underground-DLR-Overground Network Representation Service based representation to model interchanges and wait times. Topological network, use timetable data.

13. Mainline Rail Representation More complicated relationship with stops, e.g. express services. Each service given separate link structure with frequency and speed classification tailored to service. Over 150 rail services into London, rather than use timetable simplified into speed groups to measure travel time. Speed Classes: Rail Speed GroupSpeedExamples High Speed170HS1 (St Pancras – Ashford) Moderate High150Great North Eastern Moderate115-150Great Western, Western Main Line Local90Guildford, Maidstone, Croydon Low-congested30-65Thameslink

14. Rail Link Time = Link Distance / Service Speed + Station Delay (default 2.5 min) Time = 0 Time = 12.5 Time = 25 Time = 37.5 Time = 17.5 Local Service, Speed 50 kmph Express Service, Speed 100 kmph

15. Rail Network

16. Bus Network Representation Should also be serviced based. Hundreds of services with more variable route patterns. Requires fully automated method to generate network geometry from the timetable and stop databases. Not complete. Currently using a simple proxy network using location and density of stops. No interchanges modelled, aggregate speeds for Inner London, Outer London, ROSE.

17. Bus Network Representation

18. Database Attribute Structure Mode, Service ID, Link Time, Wait Time, Link Length Service IDLink Traversal TimeInterchange Wait Time 1Street Length / 5 kmph0 2,LineID,BranchIDTimetable30 / Service Frequency 3,LineID,BranchIDTimetable30 / Service Frequency 4,TerminusID,ServIDLink Length / Speed Class + stop delay30 / Service Frequency 5,ServiceIDLink Length / Speed Class + stop delayStop density classification 6Street Length / 12 kmph- 7Street Length / ITIS speed-

19. Identifying Interchange Wait Times Rule for when service wait times are included: Journey Time = Total Link Time + Total Interchange Delay If fromEdge.ServiceID = toEdge.ServiceID Then Interchange Delay = 0 Else Interchange Delay = toEdge.WaitTime Service ID: 01 Link Time: 02 Wait Time 04 Service ID: 02 Link Time: 02 Wait Time 04 Time = 0 Time = 2 Time = 4 Time = 8

20. Interfaces to the Model Mapping Interface Through ArcMap Visual, interactive, (slow!). ExampleExample Toolbox / Scripting Interface Faster (ish…), matrix calculations. Interface Selections- Choose which travel cost measure to minimise (travel time, generalised cost) Choose which modes to be active (Select Query by ServiceID). Choose origins and destinations.

21. Travel Time Mapping- Bank

22. Travel Time Mapping- Kings Cross

23. Travel Time Mapping- Paddington

24. Travel Time Mapping- Croydon

25. Travel Time Mapping- Reading

26. Travel Time Mapping- Romford

27. Generalised Costs and Disaggregation By applying an average cost per unit distance for each mode, calculate basic generalised cost. Also standard weightings for journey stages. Standard Money-Time Conversion DfT WEBTAG guidance, £5.04 per hour. Income and Demographic Disaggregation Conversions between time and money highly dependent on income, need to perform demographic disaggregation first, yet to do. 1 2 30 mins £10 GC= £12.52 60 mins £3 GC=£8.04

28. Applying the Model to Commuting & Sustainability Analysis Census Commuting Data Comprehensive sample flows, origins, destinations, main mode. Assign to Network Services Calculate network distances, generalised costs and secondary modes, carbon emissions... Absolute Assignment More realistic to assign probabilistically, logit model etc.

29. Applying the Model to Flooding Impacts Geography of Service Closures Can remove services from model, based on spatial extent of event. ‘Damage functions’ estimate severity, temporal dimension. Underground highly exposed to flooding events, acts as drainage network. Capacity Limitation How realistic to recalculate accessibility without capacity modelled?

30. The Challenge of Modelling PT Capacity Capacity Major Constraint Tube strikes graphic illustration of capacity determining journey patterns in London. Road, bus travel at standstill, even pedestrian street network near capacity. Could assign capacity to limits to service networks, validate against census commuting flows. Major task, have to be realistic about time-frame of project.

31. Conclusions Achieved Multi-Modal PT Accessibility Model Captures interchanges, more realistic journey routes and times. Service basis successful. Representational detail against performance trade-off, rather slow performance in ArcGIS, much faster without street network. Challenge of handling capacity issues. Radial Nature of PT Infrastructure Mapping highlights longstanding issue of orbital connections in Outer London, OMA. Car travel will be significantly faster. Future Tasks Commuting analysis based on accessibility model. Generalised costs. Comparison against car speeds, mode choice. Useful for more automated methods for generating networks, particularly for bus. Formal validation against other models, journey planners.

32. Thank you for listening! Welcome comments and questions. Contact Email: duncan.a.smith@ucl.ac.ukduncan.a.smith@ucl.ac.uk

33. Urban Modelling Framework, Dynamics

34. Entrance and Exit Times Time for Accessing Platforms Creating Entrance and Exit Links to Stations