1 / 54

A Tour-Based Travel Demand Model for the Ottawa-Gatineau Region Part 1 – Travel Generation

A Tour-Based Travel Demand Model for the Ottawa-Gatineau Region Part 1 – Travel Generation by P. Vovsha, V. Patterson, R. Donnelly, D. Stephens, P. Tremblay, D. Washnuk, L. Deneault Emme Users’ Conference, Toronto, October 2007. 2. Introduction. 3. Introduction. 1.2 M population. 4.

komala
Télécharger la présentation

A Tour-Based Travel Demand Model for the Ottawa-Gatineau Region Part 1 – Travel Generation

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. A Tour-Based Travel Demand Model for the Ottawa-Gatineau Region Part 1 – Travel Generation by P. Vovsha, V. Patterson, R. Donnelly, D. Stephens, P. Tremblay, D. Washnuk, L. Deneault Emme Users’ Conference, Toronto, October 2007

  2. 2 Introduction

  3. 3 Introduction • 1.2 M population

  4. 4 Introduction • 1.2 M population

  5. for travel generation and spatial distribution • daily tours • for mode choice and traffic / transit assignments • both AM and PM trips 5 Major Structural Features • Advanced tour-based structure • draws on experience of first activity-based models • implementable in Emme in aggregate fashion • Conventional trip-based structure • other periods can be added in a future version

  6. j i • Tours (P-A) j i 2. Directional half-tours (O-D) k j i 3. Chained trips (O-D) k 6 Adopted Tour-Based Concept • Dealing with both tours and trips

  7. consistency of time of day (TOD)-specific trip matrices (AM and PM) • all TOD periods derived from the same daily source • consistency between outbound and inbound trip generation and distribution 7 Adopted Tour-Based Concept • Advantages taken • Advantages not taken • consistency of mode choice across TOD periods and outbound and inbound trips

  8. 8 Design of the Core Travel Model

  9. 9 Design of the Core Travel Model Tour Generation

  10. 10 Design of the Core Travel Model Tour / Trip Distribution

  11. 11 Design of the Core Travel Model Trip Mode Choice

  12. 12 Population Synthesizer

  13. External marginal controls for each traffic zone • household distribution by size • household distribution by housing type • total labour force in the zone • population distribution over 6 age ranges 13 Population Synthesizer • The only non-Emme component (JAVA) • List of 23,868 individual households in 556 traffic zones • IPF applied to the individual household weight from the O-D survey • Production of joint household distribution in each traffic zone by 42 feasible combinations of: • 6 household size categories: 1, 2, 3, 4, 5, 6+ • 4 household worker categories: 0, 1, 2, 3+ • 2 housing types: 1=detached, 2=townhouses, apartments

  14. 14 Population Segmentation Households are further allocated to 4 car sufficiency groups Number of cars in household Number of workers in household 0 1 2 3+ 0 Zero High High High 1 Zero Balanced High High 2 Zero Low Balanced High 3+ Zero Low Low Balanced

  15. 15 Population Segmentation Population segmentation results in: Sub-model Segment Car Ownership Tour Generation Non-motorized Time of DayChoice Tour / Trip Distribution Mode Choice 6 HH size X X 4 worker X X 2 housing X X 4 car sufficiency X X X X X Total 42 168 4 4 4 4

  16. 16 Travel Segmentation • 5 travel purposes • Work : workplace, work-related • School : high school,18 or younger • University : university, college / CEGEP, other schools for 19+ • Maintenance : shopping / banking, medical, pick up / drop off • Discretionary : leisure / sport, eating out, visiting relatives and friends

  17. 17 Travel Segmentation Observed frequency by purpose

  18. AM PM AM 6:30-8:59 AM 6:30-8:59 AM AM AM AM AM PM AM AM AM PM 15:30-18:29 PM PM PM PM 18 Travel Segmentation • Time of Day Correspondence Early Early Early 4:00-6:29 Early Early Midday Early 15 tour TOD combinations by outbound (→) & inbound (←) directions Early Late 5 trip TODs Midday Midday 9:00-15:29 Late Midday Midday Midday Midday Late Late 18:29-28:00 Late Late Late

  19. 19 Travel Segmentation Travel segmentation results in: Sub-model Segment Car Ownership Tour Generation Non-motorized Time of DayChoice Tour / Trip Distribution Mode Choice 5 purposes X X X X X 15 tour TODs X X 9 relevant 5 trip TODS X 2 relevant X 2 relevant X 9 modes Total - 5 5 75 45 90

  20. 20 Combined Segmentation Together, population and travel segmentations result in: Sub-model Segment Car Ownership Tour Generation Non-motorized Time of DayChoice Tour / Trip Distribution Mode Choice Population 42 168 4 4 4 4 - 5 5 75 45 90 Travel 42 840 20 300 180 360 Total

  21. good enough for some models (generation) • more variables needed for others (distribution, mode choice) • models applied to list of individual households, persons • unlimited segmentation / variables 21 Further Segmentation Microsimulation ? • Conventional aggregate / zonal structure limits segmentation : • 999 matrices is nearly not enough ! • Individual microsimulation :

  22. retail • service • public offices • private offices • education • health • industry • school • university • % low income • % detached houses 22 Land Use / Socio-Economic Data • Employment • Shopping • gross leasable area • Enrollment • Households

  23. 3 spatial levels tested statistically • 556 traffic zones • 94 super-zones • 26 districts • Measures • population density • employment density • retail employment density 23 Derivative Density Measures

  24. 24 Model Components Tour Production • Household-based linear regression model • Segmented by 5 purposes, 42 HH compositions (HH size, # of workers, housing type) and 4 car sufficiency groups • Includes derivative HH composition variables • # of non-workers with no worker (e.g. retirees, students) • # of non-workers with 1 worker (e.g. stay-at-home) • # of non-workers with 2+ workers (e.g. children) • Sensitive to density measures at different spatial levels

  25. Major variables 25 Model Components Tour Attraction • Linear regression model

  26. Binary logit choice model • motorized travel • non-motorized travel 26 Model Components Pre-Mode Choice • Fully segmented by 5 purposes • Applied separately for HH daily tour productions and zonal land use attractions • Production side segmented by 4 HH car sufficiency groups

  27. 27 Model Components Time of Day Choice • Multinomial logit model with 15 TOD alternatives • Fully segmented by 5 purposes • Applied separately for HH tour productions and zonal land use attractions • Production side : • extended to 60 alternatives by inclusion of 4 stop-frequency sub-alternatives • segmented by 4 car sufficiency groups • Attraction side : • driven by land use (employment) mix • sensitive to location / density measures

  28. Work Half-Tours 28 Model Validation

  29. University Half-Tours 29 Model Validation

  30. 30 Model Validation School Half-Tours

  31. 31 Model Validation Maintenance Half-Tours

  32. 32 Model Validation Discretionary Half-Tours

  33. 33 To be continued …

  34. A Tour-Based Travel Demand Model for the Ottawa-Gatineau Region Part 2 – Distribution and Mode Choice by P. Vovsha, V. Patterson, D. Stephens, P. Tremblay Emme Users’ Conference, Toronto, October 2007

  35. 2 Introduction

  36. j i • Tours (P-A) j i 2. Directional half-tours (O-D) k j i 3. Chained trips (O-D) k 3 Distribution 3 Steps of Matrix Construction

  37. 4 Distribution Step 1 of Matrix Construction • Tour matrices in P-A format • Hybrid balancing-gravity model derived from maximum entropy principle • Seed matrix prepared from the O-D survey by “smoothing” (to avoid lumpiness)

  38. O-D Survey Impedance cij Seed Matrix sij Gravity Model Proportional Balancing Combined Model 5 Distribution • Hybrid Balancing – Gravity

  39. 4 steps of smoothing : – aggregate O-D survey matrix to superzone level – calculate traffic zone marginals – calculate zone-to-zone gravity proportions – redistribute aggregate matrix by gravity proportions within each superzone-to-superzone cell • 3 properties of a smoothed matrix : – identical to O-D survey at superzone aggregation – almost identical to O-D survey for zonal marginals – smooth and logical at zone-to-zone level 6 Distribution • Matrix Smoothing • Matrices from O-D survey are “lumpy” (expansion factor = 20) and cannot be used as zone-to-zone seed matrices

  40. Half-tour matrices in O-D format, by direction • outbound ( ij ) • inbound / transposed ( ji ) • Each direction processed by stop frequency • direct half-tours correspond to trips ( ij or ji ) • half-tours with stops are broken into chained trips ( ik, kj or jk, ki ) 7 Distribution Steps 2 and 3 of Matrix Construction

  41. Half-tour matrix : Multinomial logit stop-location model : Combined utility function (based on both impedances and stop attractions) : 3.23 3.21 3.21 1st trip leg matrix : 2nd trip leg matrix : 8 Distribution • Chained Trip Distribution

  42. 9 Mode Choice Estimation Motorized Trips

  43. 10 Mode Choice Estimation

  44. 11 Mode Choice Estimation

  45. 12 Mode Choice Estimation

  46. differential auto time coefficient – free flow time – congestion delay (unreliability effect) 13 Mode Choice Estimation • No separate bus and rail sub-nests because of too few rail observations • Interesting and non-standard variables • Transitway / rail share of total transit distance, as a reliability bonus

  47. Willingness to Pay 14 Mode Choice Estimation

  48. Transitway / Rail Reliability 15 Mode Choice Estimation

  49. Transit time weights : – walk : 1.2 (university, PM) – 2.8 (work, AM) – wait : 2.2 (school, AM) – 3.4 (maintenance, AM) – transfer, mins : 4.0 (work, PM) – 9.0 (work, AM) • Strong nesting – substitution of transit modes – limited substitution between auto driver and passenger 16 Mode Choice Estimation • Other Main Findings • Strong and consistent impact of car sufficiency and density • Significant variation by purpose and somewhat between AM and PM

  50. Modelled explicitly with specific mode choice parameters – P+R bus, K+R bus – P+R rail, K+R rail • Trip matrices are broken by module 3.23 into mode trip legs (auto / transit) for assignment – AM : auto  transit – PM : transit  auto 17 Mode Choice Estimation • Treatment of Bi-Modals • LOS skims are created using module 3.23, with explicit identification of parking lots (all-or-nothing) • Parking lot capacity constraint can be introduced

More Related