1. 14 th TRB National Planning Applications Conference May 5-9, 2013, Columbus, Ohio Rosella Picado Parsons Brinckerhoff

2. Background  Southeast Florida is home to 5.5 million people, spanning Miami-Dade, Broward and Palm Beach counties  Relatively disperse travel patterns with significant congestion on Turnpike and north-south freeways  Growing interest in improving transit, expand toll and managed lane infrastructure, mitigate adverse EJ impacts  Trip-based model reaching its limits, especially regarding variably-priced tolling, fare policies, spatial detail, EJ analysis

3. SE Florida ABM  C oordinated T ravel – R egional A ctivity-based M odeling P latform Family of ABMs  Main features:  Explicit intra-household interactions  Continuous temporal dimension (half-hour time periods)  Fine spatial dimension (12,000 MAZs)  Faithful transit access coding  Distributed values of time  Integration of location, time-of-day, and mode choice models

4. Model Development Strategy  Transfer the San Diego ABM  Adopt CT-RAMP structure and sub-models  Adopt model parameters estimated with San Diego household survey data  Update certain models to reflect SE Florida conditions:  input data availability (employment, population controls)  modal supply  trip assignment methods  ancillary models  Calibrate models to SE Florida travel patterns  Re-specify models that fail to perform well

5. Why Model Transfer?  Schedule:  To use the ABM in the development of the 2013 Long Range Transportation Plan  Approximately 18 months available for model development was insufficient time to estimate & validate all models  Data:  Quantity and quality of NHTS SE Florida sample may preclude statistically significant estimation of some models and/or population effects  Largely sufficient for calibration, with caveats

6. Data Limitations  Small sample size – 2,000 households  Some subareas within model region under- represented  Retired households over-sampled  College students and children under-represented  Missing data, ‘ungeocodable’ activity locations, etc.  Incomplete transit on-board survey

7. Assessing the Model Transfer Outcome  Evaluate initial estimated travel patterns against model calibration targets  Regional targets for important person markets  Sub-regional where data allow  Assess the magnitude of constant or parameter adjustments to match targets  Importance of model calibration targets  Based on NHTS and supplemented with other sources  Evaluated for reasonableness  Compared to targets from other regions

8. Work Location Model - initial results Person Type Avg. Length (mi.) Obs.Est. Full-time10.69.4 Part-time7.55.3 All9.98.7 Tour Frequency (%)

9. Work Location Model - calibrated Person Type Avg. Length (mi.) Obs.Est. Full-time10.610.2 Part-time7.57.0 All9.99.7 Tour Frequency (%)

10. School Location Model - initial Tour Frequency (%)

11. School Location Model - calibrated Tour Frequency (%)

12. Eating Out Location Model - initial Tour Frequency (%)

13. Daily Activity Pattern Model Target DAPModel Initial DAP Person typeMandatory Non Mandatory HomeMandatory Non Mandatory Home Full-time worker80%14%7%81%12%7% Part-time worker55%37%8%63%26%11% University student78%18%4%63%25%12% Non-working adult0%76%24%0%74%26% Non-working senior0%72%28%0%75%25% Driving age student89%4%6%92%3%5% Pre-driving student94%3%2%96%2% Pre-school35%43%22%43%41%15%

14. Daily Activity Pattern Model Target DAPModel Initial DAP Person typeMandatory Non Mandatory HomeMandatory Non Mandatory Home Full-time worker80%14%7%81%12%7% Part-time worker55%37%8%63%26%11% University student78%18%4%63%25%12% Non-working adult0%76%24%0%74%26% Non-working senior0%72%28%0%75%25% Driving age student89%4%6%92%3%5% Pre-driving student94%3%2%96%2% Pre-school35%43%22%43%41%15%

15. Non-Mandatory Tour Frequency Estimated Tour Frequency (%) Initial Calibrated

16. Non-Mandatory Tour Frequency Estimated Tour Frequency (%) Initial Calibrated

17. Work Departure and Arrival Times Initial

18. Shop Tour Departure Time Initial

19. Shop Tour Departure Time Calibrated

20. Work Tour Mode Choice TargetInitial Estimate auto sufficiency Tour Mode no veh.insuf.suf.totalno veh.insuf.suf.total Drive-Alone0%49%78%67%0%51%67%60% Shared 213%30%13%18%24%25%17%20% Shared 3+8%11%6%8%12%13% Walk11%3%0%1%31%6%1%4% Bike5%1%0% 18%2%0%1% Walk-Transit62%5%1%4%15%3%1%2% PNR-Transit0%1% 0% KNR-Transit2%1%0% Toll13%14% Local Bus68%51% Express Bus5%9% BRT1%13% Urban Rail21%23% Com Rail5%3%

21. Work Ahead  Finalize model calibration  Validation to traffic counts and transit boardings  Future year forecast and sensitivity tests

22. Conclusions / Lessons Learned  SANDAG CT-RAMP ABM is able to reproduce most regional travel patterns in SE Florida  Largest differences between observed and initial model forecasts:  non-mandatory tour location  CDAP and tour frequency for college students, part-time workers, pre-school children  Modest constant adjustments sufficient to calibrate the model

23. Conclusions / Lessons Learned  Supplemental data sources important to validate calibration targets and selected model outputs  Unable to observe transferability at high levels of disaggregation

24. Acknowledgments  Shi-Chiang Li, Florida DOT  Paul Larsen, Palm Beach MPO  Paul Flavien, Broward MPO  Larry Foutz, HNTB (formerly Miami-Dade MPO)  Ken Kaltenbach, The Corradino Group  Sung-Ryong Han, BCC Engineering  Bill Davidson, Ben Stabler, Jinghua Xu

25. Questions? Rosella Picado Parsons Brinckerhoff Seattle, WA picado@pbworld.compicado@pbworld.com | (206) 382-5227