Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.
Development of a Pedestrian
Demand EstimationTool
Kelly J Clifton, PhD
Framework and Methods
Outline
• Background
• Project, methods,
zones, & data
• Pedestrian index of
the environment (PIE)
• I:Trip generation
• I...
BACKGROUND
Background
4
Why model pedestrian travel?
health & safety
new data
mode shifts
greenhouse
gas emissions
plan for pedestria...
Background
Background – Project – PIE – Walk Model – DC Model – Conclusion 5
1. Generation
2. Distribution
3. Mode choice
...
Background
• Walking behavior data
– improved travel surveys, pedestrian count data collection
• Built environment data
– ...
PROJECT, METHODS, ZONES & DATA
Project overview
• Partnered with Metro: metropolitan
planning organization for Portland, OR
• Two research projects
• Imp...
Current 4-step method
9
Trip Distribution or
Destination Choice (TAZ)
Mode Choice (TAZ)
Trip Assignment
PedestrianTrips
Al...
New MoPeD method
10
TAZ = transportation analysis zone
PAZ = pedestrian analysis zone
Trip Generation (PAZ)
Trip Distribut...
Utilizes spatially fine-grained archived
information on the built environment
MoPeD Contributions
11Background – Project –...
Pedestrian analysis zones
Background – Project – PIE – Walk Model – DC Model – Conclusion 12
264 feet = 80 m ≈ 1 minute wa...
Travel survey data
• Oregon Household Activity Survey (OHAS)
– Household-based survey
– One-day travel diary
• Portland re...
PEDESTRIAN INDEX OF THE
ENVIRONMENT (PIE)
14
Pedestrian environment
Pedestrian Index of the Environment (PIE)
20–100 score = calibrated ∑(6 dimensions)
Background – Pr...
Visualizing PIE
17
100 – Downtown core
80 – Major neighborhood centers
Downtown
Lloyd District
Background – Project – PIE ...
Visualizing PIE
18
70 – Suburban downtowns
60 – Residential inner-city neighborhoods
Laurelhurst
Gresham
Background – Proj...
Visualizing PIE
19
50 – Suburban shopping malls
40 – Suburban neighborhoods/subdivisions
Clackamas Town Center
Aloha
Backg...
Visualizing PIE
20
30 – Isolated business and light industry
20 – Rural, undeveloped, forested
Forest Park
N. Marine Drive...
I. TRIP GENERATION
21
22
TAZ = transportation analysis zone
PAZ = pedestrian analysis zone
Trip Generation (PAZ)
Trip Distribution or
Destinatio...
Trip Generation
23Background – Project – PIE – Walk Model – DC Model – Conclusion
Metro currently has 8 trip production mo...
24
II. WALK MODE SPLIT
25
26
TAZ = transportation analysis zone
PAZ = pedestrian analysis zone
Trip Generation (PAZ)
Trip Distribution or
Destinatio...
Walk mode split
Prob(walk) = f(traveler characteristics, PIE)
Data: 2011 OHAS, Production trip ends,
90% sample
Method: bi...
28
Walk mode split modelsII
Background – Project – PIE – Walk Model – DC Model – Conclusion
Traveler characteristics: Hous...
Walk model results
29
II
Background – Project – PIE – Walk Model – DC Model – Conclusion
Traveler characteristics:
+ posit...
Mode SplitValidation
Model
HBW HBO NHB
Observed Walk Mode Share 2.9% 9.4% 6.7%
Predicted Walk Mode Share 3.0% 9.5% 8.6%
30...
31Background – Project – PIE – Walk Model – DC Model – Conclusion
Walk model application
32
TAZ = transportation analysis zone
PAZ = pedestrian analysis zone
Trip Generation (PAZ)
Trip Distribution or
Destinatio...
III. DESTINATION CHOICE
33
34
TAZ = transportation analysis zone
PAZ = pedestrian analysis zone
Trip Generation (PAZ)
Trip Distribution or
Destinatio...
PedestrianTrips
Destination Choice (PAZ)
Prob(dest.) = function of…
– network distance
– size / # of destinations
– pedest...
36
Destination choice
Background – Project – PIE – Walk Model – DC Model – Conclusion
Destination Choice
superPAZ:
– a grid of
5 × 5 = 25 PAZs
Choice set generation:
– Random sample of 10 superPAZs within 3 m...
Destination Choice
38Background – Project – PIE – Walk Model – DC Model – Conclusion
Key variables
Impedance Size
Pedestri...
Destination Choice
39
III
Impedance ∆ odds of walking to destination
+ 1 mile of distance
by auto own.:
by children:
by ch...
Destination Choice
40
III
Ped. supports ∆ odds of walking to destination
+ 10 points PIE: 16–34% increase (*)
presence of ...
Destination Choice
Background – Project – PIE – Walk Model – DC Model – Conclusion 41
III
Destination Choice
Background – Project – PIE – Walk Model – DC Model – Conclusion 42
III
PIE = 75 PIE = 85
Destination choice
43Background – Project – PIE – Walk Model – DC Model – Conclusion
ModelValidation – % Correct Destinati...
Destination Choice
44Background – Project – PIE – Walk Model – DC Model – Conclusion
ModelValidation – Avg. Distance Walked
45Background – Project – PIE – Walk Model – DC Model – Conclusion
Destination Choice
CONCLUSIONS & FUTURE WORK
46
Conclusions
• Nests within current model but can be used alone
• Pedestrian scale analysis (PAZs)
• Pedestrian-relevant va...
Future work
Before application:
• Relate PIE more explicitly to policy changes
• Forecasting inputs
• Test method in other...
Future work
Research & Model Improvements:
Trip Generation
– Multinomial Logit model
• Destination Choice
– Allocate from ...
Questions?
Project info & reports:
http://trec.pdx.edu/research/project/510
http://trec.pdx.edu/research/project/677
Kelly...
Development of a Pedestrian Demand Estimation Tool
Upcoming SlideShare
Loading in …5
×

Development of a Pedestrian Demand Estimation Tool

900 views

Published on

Kelly Clifton, Portland State University

Published in: Education
  • Be the first to comment

  • Be the first to like this

Development of a Pedestrian Demand Estimation Tool

  1. 1. Development of a Pedestrian Demand EstimationTool Kelly J Clifton, PhD Framework and Methods
  2. 2. Outline • Background • Project, methods, zones, & data • Pedestrian index of the environment (PIE) • I:Trip generation • II:Walk mode split • III: Pedestrian destination choice • Conclusions & future work 2 Adapted from: http://www.flickr.com/photos/takomabibelot/3223617185 Background – Project – PIE – Walk Model – DC Model – Conclusion
  3. 3. BACKGROUND
  4. 4. Background 4 Why model pedestrian travel? health & safety new data mode shifts greenhouse gas emissions plan for pedestrian investments & non-motorized facilities
  5. 5. Background Background – Project – PIE – Walk Model – DC Model – Conclusion 5 1. Generation 2. Distribution 3. Mode choice 4. Assignment Trip-based model sequence How do travel models estimate walking? Source: Singleton, P. A., & Clifton, K. J. (2013). Pedestrians in regional travel demand forecasting models: State-of-the-practice. • Among 48 large MPOs in US: – 38% did not estimate walking – 33% estimated non-motorized (walking + bicycling) travel – 29% estimated walking • Lacking pedestrian built environment measures & small spatial units
  6. 6. Background • Walking behavior data – improved travel surveys, pedestrian count data collection • Built environment data – archived spatial datasets, GIS processing • Travel demand models – smaller zones, complete networks, computer power • Walking behavior research – more knowledge and studies Background – Project – PIE – Walk Model – DC Model – Conclusion 6 What are some opportunities?
  7. 7. PROJECT, METHODS, ZONES & DATA
  8. 8. Project overview • Partnered with Metro: metropolitan planning organization for Portland, OR • Two research projects • Improve representation of pedestrian environment in current 4-step method Background – Project – PIE – Walk Model – DC Model – Conclusion 8
  9. 9. Current 4-step method 9 Trip Distribution or Destination Choice (TAZ) Mode Choice (TAZ) Trip Assignment PedestrianTrips AllTrips PedestrianTrips VehicularTrips TAZ = transportation analysis zone Trip Generation (TAZ) Background – Project – PIE – Walk Model – DC Model – Conclusion
  10. 10. New MoPeD method 10 TAZ = transportation analysis zone PAZ = pedestrian analysis zone Trip Generation (PAZ) Trip Distribution or Destination Choice (TAZ) Mode Choice (TAZ) Trip AssignmentPedestrianTrips Walk Mode Split (PAZ) Destination Choice (PAZ) I III AllTrips PedestrianTrips VehicularTrips Background – Project – PIE – Walk Model – DC Model – Conclusion
  11. 11. Utilizes spatially fine-grained archived information on the built environment MoPeD Contributions 11Background – Project – PIE – Walk Model – DC Model – Conclusion Operates at a smaller spatial scale, more relevant to pedestrians (PAZ) Incorporates knowledge of influences on pedestrian travel behavior Designed to work with regional travel demand model or as standalone tool
  12. 12. Pedestrian analysis zones Background – Project – PIE – Walk Model – DC Model – Conclusion 12 264 feet = 80 m ≈ 1 minute walk Metro: ~2,000TAZs  ~1.5 million PAZs TAZs PAZs Home-based work trip productions
  13. 13. Travel survey data • Oregon Household Activity Survey (OHAS) – Household-based survey – One-day travel diary • Portland region dataset (2011) – 6,100 households – 13,400 people – 56,000 trips ÷ 4,500 walk trips ≈ 8% walk mode share Background – Project – PIE – Walk Model – DC Model – Conclusion 13
  14. 14. PEDESTRIAN INDEX OF THE ENVIRONMENT (PIE) 14
  15. 15. Pedestrian environment Pedestrian Index of the Environment (PIE) 20–100 score = calibrated ∑(6 dimensions) Background – Project – PIE – Walk Model – DC Model – Conclusion 15 ULI = Urban Living Infrastructure: pedestrian-friendly shopping and service destinations used in daily life. People & job density Transit access Block size Sidewalk extent Comfortable facilities Urban living infrastructure
  16. 16. Visualizing PIE 17 100 – Downtown core 80 – Major neighborhood centers Downtown Lloyd District Background – Project – PIE – Walk Model – DC Model – Conclusion
  17. 17. Visualizing PIE 18 70 – Suburban downtowns 60 – Residential inner-city neighborhoods Laurelhurst Gresham Background – Project – PIE – Walk Model – DC Model – Conclusion
  18. 18. Visualizing PIE 19 50 – Suburban shopping malls 40 – Suburban neighborhoods/subdivisions Clackamas Town Center Aloha Background – Project – PIE – Walk Model – DC Model – Conclusion
  19. 19. Visualizing PIE 20 30 – Isolated business and light industry 20 – Rural, undeveloped, forested Forest Park N. Marine Drive Background – Project – PIE – Walk Model – DC Model – Conclusion
  20. 20. I. TRIP GENERATION 21
  21. 21. 22 TAZ = transportation analysis zone PAZ = pedestrian analysis zone Trip Generation (PAZ) Trip Distribution or Destination Choice (TAZ) Mode Choice (TAZ) Trip AssignmentPedestrianTrips Walk Mode Split (PAZ) Destination Choice (PAZ) I III AllTrips PedestrianTrips VehicularTrips Background – Project – PIE – Walk Model – DC Model – Conclusion Trip Generation
  22. 22. Trip Generation 23Background – Project – PIE – Walk Model – DC Model – Conclusion Metro currently has 8 trip production models applied to ~2,000TAZs: – HBW – Home-based work; – HBshop – Home-based shopping; – HBrec – Home-based recreation; – HBoth – Home-based other (excludes school and college); – NHBW – Non-home-based work; – NHBNW – Non-home-based non-work; – HBcoll – Home-based college; and – HBsch – Home-based school. After testing for scalability, we applied the same models to our pedestrian scale ~1.5M PAZs
  23. 23. 24
  24. 24. II. WALK MODE SPLIT 25
  25. 25. 26 TAZ = transportation analysis zone PAZ = pedestrian analysis zone Trip Generation (PAZ) Trip Distribution or Destination Choice (TAZ) Mode Choice (TAZ) Trip AssignmentPedestrianTrips Walk Mode Split (PAZ) Destination Choice (PAZ) I III AllTrips PedestrianTrips VehicularTrips Background – Project – PIE – Walk Model – DC Model – Conclusion Walk mode split
  26. 26. Walk mode split Prob(walk) = f(traveler characteristics, PIE) Data: 2011 OHAS, Production trip ends, 90% sample Method: binary logit model Spatial unit: pedestrian analysis zone (PAZ) Background – Project – PIE – Walk Model – DC Model – Conclusion 27 Walk Mode Split (PAZ) PedestrianTrips VehicularTrips II
  27. 27. 28 Walk mode split modelsII Background – Project – PIE – Walk Model – DC Model – Conclusion Traveler characteristics: Household size, income, age, # of workers, # children, # vehicles Built environment: PIE
  28. 28. Walk model results 29 II Background – Project – PIE – Walk Model – DC Model – Conclusion Traveler characteristics: + positively related to walking – negatively related to walking number of children in HH age of household head HH vehicle ownership Ped. Environment: ∆ odds of choosing to walk + 10 points PIE 43% increase (HBW) 54% increase (HBNW) 67% increase (NHB) Pseudo R2 0.137 (HBNW) – 0.253 (NHB)
  29. 29. Mode SplitValidation Model HBW HBO NHB Observed Walk Mode Share 2.9% 9.4% 6.7% Predicted Walk Mode Share 3.0% 9.5% 8.6% 30 1. Apply the final model equations to trips in the validation sample (10% of data) and calculate the walk probability for each trip; 2. Average the probabilities to get the predicted walk mode share of trip ends (called sample enumeration) Background – Project – PIE – Walk Model – DC Model – Conclusion
  30. 30. 31Background – Project – PIE – Walk Model – DC Model – Conclusion Walk model application
  31. 31. 32 TAZ = transportation analysis zone PAZ = pedestrian analysis zone Trip Generation (PAZ) Trip Distribution or Destination Choice (TAZ) Mode Choice (TAZ) Trip AssignmentPedestrianTrips Walk Mode Split (PAZ) Destination Choice (PAZ) I III AllTrips PedestrianTrips VehicularTrips Walk mode split Background – Project – PIE – Walk Model – DC Model – Conclusion
  32. 32. III. DESTINATION CHOICE 33
  33. 33. 34 TAZ = transportation analysis zone PAZ = pedestrian analysis zone Trip Generation (PAZ) Trip Distribution or Destination Choice (TAZ) Mode Choice (TAZ) Trip AssignmentPedestrianTrips Walk Mode Split (PAZ) Destination Choice (PAZ) I III AllTrips PedestrianTrips VehicularTrips Background – Project – PIE – Walk Model – DC Model – Conclusion Destination choice
  34. 34. PedestrianTrips Destination Choice (PAZ) Prob(dest.) = function of… – network distance – size / # of destinations – pedestrian environment – traveler characteristics Data: 2011 OHAS Method: multinomial logit model Spatial unit: super-pedestrian analysis zone Six trip types: home-based: work (HBW), shopping (HBS), recreation (HBR), & other (HBO); non-home-based: work (NHBW) and non-work (NHBNW) Destination choice 35 III Background – Project – PIE – Walk Model – DC Model – Conclusion
  35. 35. 36 Destination choice Background – Project – PIE – Walk Model – DC Model – Conclusion
  36. 36. Destination Choice superPAZ: – a grid of 5 × 5 = 25 PAZs Choice set generation: – Random sample of 10 superPAZs within 3 miles – 99% of OHAS walk trips < 3 miles (4.8 km) Background – Project – PIE – Walk Model – DC Model – Conclusion 37 III
  37. 37. Destination Choice 38Background – Project – PIE – Walk Model – DC Model – Conclusion Key variables Impedance Size Pedestrian supports Pedestrian barriers Traveler attributes Additional variables network distance btw. zones employment by category, households PIE, parks slope, freeway, industrial LUs III auto own., children
  38. 38. Destination Choice 39 III Impedance ∆ odds of walking to destination + 1 mile of distance by auto own.: by children: by children: 76–86% decrease (*) -62% (no), -74% (yes) (HBW) -78% (no), -83% (yes) (HBR) -78% (no), -90% (yes) (HBS) Size ∆ odds of walking to destination 2 × # destinations minimum: maximum: 28–42% increase (†) 4% increase (HBR) 88% increase (HBS) † Except for HBR and HBS. Background – Project – PIE – Walk Model – DC Model – Conclusion * Except for HBW, HBR, and HBS.
  39. 39. Destination Choice 40 III Ped. supports ∆ odds of walking to destination + 10 points PIE: 16–34% increase (*) presence of park: 58% increase (HBR) Ped. Barriers ∆ odds of walking to destination + 1° mean slope: 14–35% decrease (2,3,4) presence of freeway: 64% decrease (2) + 1% industrial jobs: 33–82% decrease (1,2,3,4) Pseudo R2 0.416 (HBR) – 0.680 (HBS) Background – Project – PIE – Walk Model – DC Model – Conclusion * Except for HBS and HBR. 1 HBW, 2 HBS, 3 HBO, 4 NHBW.
  40. 40. Destination Choice Background – Project – PIE – Walk Model – DC Model – Conclusion 41 III
  41. 41. Destination Choice Background – Project – PIE – Walk Model – DC Model – Conclusion 42 III PIE = 75 PIE = 85
  42. 42. Destination choice 43Background – Project – PIE – Walk Model – DC Model – Conclusion ModelValidation – % Correct Destination
  43. 43. Destination Choice 44Background – Project – PIE – Walk Model – DC Model – Conclusion ModelValidation – Avg. Distance Walked
  44. 44. 45Background – Project – PIE – Walk Model – DC Model – Conclusion Destination Choice
  45. 45. CONCLUSIONS & FUTURE WORK 46
  46. 46. Conclusions • Nests within current model but can be used alone • Pedestrian scale analysis (PAZs) • Pedestrian-relevant variables (PIE) • One of the first studies to examine pedestrian destination choice in modeling framework • Highlights policy relevant variables: distance, size, pedestrian supports & barriers Background – Project – PIE – Walk Model – DC Model – Conclusion 47
  47. 47. Future work Before application: • Relate PIE more explicitly to policy changes • Forecasting inputs • Test method in other area(s)/regions – Examine relationships in other contexts – Assess PIE’s transferability • Provide agency guidance for implementation Background – Project – PIE – Walk Model – DC Model – Conclusion 48
  48. 48. Future work Research & Model Improvements: Trip Generation – Multinomial Logit model • Destination Choice – Allocate from superPAZ to PAZ level – Explore non-linear effects & other interactions • Route choices or potential pathways – Need fundamental research to improve understanding 49Background – Project – PIE – Walk Model – DC Model – Conclusion
  49. 49. Questions? Project info & reports: http://trec.pdx.edu/research/project/510 http://trec.pdx.edu/research/project/677 Kelly J. Clifton, PhD kclifton@pdx.edu Patrick A. Singleton Portland State University Christopher Muhs DKS & Associates Robert Schneider, PhD Univ. Wisconsin–Milwaukee 50Background – Project – PIE – Walk Model – DC Model – Conclusion

×