Transportation leadership you can trust. presented to TRB Planning Applications Conference presented by Vamsee Modugula Cambridge Systematics, Inc. May 9, 2007 Evaluating the Effects of Transit Crowding - Transbay Ridership Forecasting Model

1 Why Model Transit Crowding? Models forecast travel demand (irrespective of capacity?) Highway models account for congestion by increasing travel time Transit models account for parking capacity by increasing drive time to station No direct modeling of increased wait time due to crowding Crowding affects ridership on competing transit modes

2 Transbay Ridership Study - Overview Determine future transit ridership at Transbay Terminal AC Transit (Bus bay requirements) Analyze the impact of capacity constraints on Transit More accurate ridership estimates with improved travel forecasting tools Provide analysis needed for the TIFIA loan application Project study team included TJPA, AC Transit, BART, MTC and WTA

3 Innovative Features of This Project New Mode choice model with detailed transit modes New capability to model Transit crowding Model passenger perception that travel time is more onerous when they have to stand or when the vehicle is crowded Increased wait times when passengers are unable to board a crowded vehicle Apply a range of capacity assumptions for BART Analyse ridership and traffic volumes for Peak Hours

4 Traditional Method: “best path” between home neighborhood & work Better analysis of competing transit modes New Method: multiple paths” + detailed treatment of each transit option in mode choice models Full treatment of access/egress & transfers

5 Mode Choice Motorized Bicycle Walk Drive Alone Shared Ride 2 Shared Ride 3+ Walk- Access Drive- Access Local Bus Express Bus LRT Commuter Rail BART Ferry Local Bus Express Bus LRT Commuter Rail BART Ferry Transit Mode Choice Structure for the Trans Bay Ridership Model

6 Base year model validation - Daily Validation for the Bay Bridge Corridor 2005 Observed 2005 Modeled Difference % Diff Target Auto 267, ,344 8,4003%+/- 15% or +/- 500 Bart 288, ,484 3,0041%+/- 15% or +/- 500 Express Bus 11,841 13,151 1,31111%+/- 15% or +/- 500 Ferry 3,302 3, %+/- 15% or +/- 500 Total 571, ,692 13,1252%+/- 15% or +/- 500

7 Base year model validation – Freeway Volumes BridgeDirAM Peak Period PM Peak Period Daily Volume AM Peak HourPM Peak Hour 2005 OBSERVED AUTO VOLUMES BY TIMEPERIOD Bay BridgeEB25,66536,537136,1317,2519,685 Bay BridgeWB34,51930,138131,8139,3508, MODEL ESTIMATED AUTO VOLUMES BY TIMEPERIOD Bay BridgeEB25,97539,310140,8287,57911,964 Bay BridgeWB43,62727,680135,51611,2068,447 DIFFERENCE BETWEEN OBSERVED AND MODEL VOLUMES Bay BridgeEB3102,7734, ,279 Bay BridgeWB9,108-2,4593,7031,857282

8 Base year model validation – Peak Hour BART LINE2005 Observed 2005 Modeled Difference% DiffTarget AM Peak Hour Dublin-SFO 8,904 10,105 1,20113%+/- 15% Fremont-Daly City 6,280 6, %+/- 15% Richmond - Daly City 8,413 9, %+/- 15% Pittsburg - Daly City 11,157 11, %+/- 15% PM Peak Hour Dublin-SFO 8,680 9,739 1,05912%+/- 15% Fremont-Daly City 5,839 5,845 60%+/- 15% Richmond - Daly City 8,422 8, %+/- 15% Pittsburg - Daly City 10,966 10,609 (356)-3%+/- 15%

9 Transit crowding model When trains are too crowded, riders can: Wait for next train Switch to bus or ferry Switch to auto Includes feedback to mode choice models Longer travel times for riders who would experience over- crowding

10 Transit Crowd Modeling Balance transit demand against capacity by applying: In-vehicle Travel Time Adjutment

11 Transit Crowd Modeling Wait Time Adjustment Based on probability to board a transit line Stochastic Assignment to reallocate ridership based on capacity Trip Tables with excess demand Trip tables with perceived travel times and actual wait times

12 Projected Growth in Travel in the Corridor Growth in jobs faster than population in San Francisco County Huge increases in commuter and total trips in the corridor Increased Transbay Travel demand – Can BART meet demand for service TypeIncrease from Population in SF16% Employment in SF44% Commuter Trips to SF51% Total Trips to SF43%

MODEL – CAPACITY ASSUMPTIONS (Peak Hour Peak Direction) (Low - High) 2030 (Low - High) BART CAPACITY Trains per hour Train Frequency2.7 min2.5 – 2.15 min2.15 – 1.9 min Seats per car6856 Consist910 Passengers per car90100 Passengers per train Potential passengers/Hour18,00024,000 – 28,00028,000 – 32,000

MODEL –CAPACITY ASSUMPTIONS (Peak Hour Peak Direction) (Low - high) 2030 (Low- High) AC TRANSIT CAPACITY Buses per hour Average seats per bus5065 Passengers per bus5065 Potential passengers/Hour4, – 11,375 FERRY CAPACITY Boats per hour599 Passengers per Boat

15 Year 2030 Transit Ridership – PM Peak Hour (Low Bart Capacity Alternative) CapacityBaselineScenarioChange Daly City - Fremont5,0004,300 0 SF Airport-Dublin5,0006,5005,500 -1,000 Concord - Daly City13,00011,200 0 Colma – Richmond5,0006,1005, TOTAL BART28,00028,10026,800-1,400 Diversion from Bart to Other Modes TOTAL AC Transit7,8006,9007, TOTAL Ferry2,0001,9002, Total (All Modes)39,60036,80036,

16 Year 2030 Transit Ridership – PM Peak Hour (High Bart Capacity Alternative) CapacityBaselineScenarioChange Daly City - Fremont 5,750 4,300 0 SF Airport-Dublin 5,750 6,5006, Concord - Daly City 14,750 11,200 0 Colma – Richmond 5,750 6,1005, TOTAL BART 32,000 28,10027, Diversion from Bart to Other Modes TOTAL AC Transit7,8006,9007, TOTAL Ferry2,0001,9002, Total (All Modes)47,20036,80026,

17 Year 2015 Transit Ridership – PM Peak Hour (Low Bart Capacity Alternative) CapacityBaselineScenarioChange Daly City - Fremont4,3003,400 0 SF Airport-Dublin4,3004,7004, Concord - Daly City11,1007,500 0 Colma – Richmond4,3004,000 0 TOTAL BART 24,00019,60019, Diversion from Bart to Other Modes TOTAL AC Transit7,8005,8006, TOTAL Ferry2,0001,4001, Total (All Modes)33,60026,80026,700-0

18 Year 2015 Transit Ridership – PM Peak Hour (High Bart Capacity Alternative) CapacityBaselineScenarioChange Daly City - Fremont 5,0003,4003,40 0 SF Airport-Dublin5,0004,700 0 Concord - Daly City13,0007,500 0 Colma – Richmond5,0004,000 0 TOTAL BART 28,00019,600 0 Diversion from Bart to Other Modes TOTAL AC Transit7,8005,800 0 TOTAL Ferry2,0001,400 0 Total (All Modes)39,60026,700 0

19 Conclusions This analysis did not include contraints on station platform capacity Transit Crowd Modeling is a good way of forecasting realistic transit ridership Assists transit operators in preparing service plans for future years.