Making Headw a y s
Smart-Card Fare Collection and Bus Dwell Times in Los Angeles
Daniel Shockley, MURP ‘15
Introduction Delay can force a transit agency to run more busses in order to maintain a constant level of service. As the marginal cost of providing more service (one more bus) is very high, transit agencies are keen to control it. The easiest form of delay to manipulate is dwell time at stops, that is, the amount of time a bus waits for passengers to get off, on, pay their fare, and navigate through the crowds already on board. Managing dwell time to the extent possible is a critical aspect of transit operations planning. Smart-card fare payment promises to do just that. On the assumption that paying with a card is faster per person than cash and coins, smart cards may reduce the amount of time per person to board and pay their fare; multiplied by thousands of runs in a day this can stabilize the variability of dwell time and improve schedule adherence, while also providing for a positive customer experience. Many agencies and regional planning organizations around the world have invested large sums of money in order to obtain these benefits.
Delay outside of stops - congestion, etc
Dwell time is when a vehicle waits at a stop to accommodate passenger activity.
Research Question
Managing the variances in this delay is one of the few that agencies can manipulate directly.
This study will model the impact of the Transit Access Pass (TAP) smart card on Los Angeles County Metropolitan Transportation Authority (Metro) busses using regression analysis with automatically generated data in order to determine the relative significance and effect of the TAP card on Metro bus dwell time.
What role does fare payment have to play relative to the other causes, and how can the TAP card help manage it?
Dwell time at a stop is related to passenger activity, including fare payment.
Findings
Methods The data were obtained from two on-board systems: the Universal Fare System (UFS) and the Automatic Passenger Counter (APC). The UFS notes cash and TAP transactions while the operator manually enters bicycle and wheelchair boardings. The APC has sensors mounted the doors that count passenger movement, while noting GPS location coordinates at each stop. Initially, the data are split into two separate datasets based on service type in order to account for the different operating environments of the TAP card typically, rapid-service routes have more passenger activity but fewer stops compared to locals that stop very frequently but take on only a few passengers. Route Local 120 Rapid 720
Service Type Many stops, infrequent Limited stops, very frequent
Avg. Weekday Trips
Avg. Saturday Trips
Avg. Sunday Trips
4,120
2,111
1,638
40,790
28,569
22,457
Next, each set was purged of outliers that were irrelevant to either fare processing or passenger-related activity. To account for fare records created after doors closing a “Cooldown Time” of 15 seconds was determined and UFS records were matched on the basis of their time stamp falling between the door open time and the “Cooldown” time.
UFS
APC
Cash,TAP, paper pass transactions. Bicycle and wheelchair loading. Observations of fare evasion.
Date/time doors are open/closed. Passenger boarding/alighting. Passengers currently on bus. GPS location of bus.
Fare evasion related records deleted.
Timepoint/layover/terminal stops, equipment malfunction, and abnormally long dwell time (>180 sec) deleted.
After the UFS data were related to the APC records of dwell time, the ratio of TAP-related payments to non-TAP related payments were calculated. A “TAP-Related Payment” is defined as any fare transaction that involves a TAP card, including using stored value to pay a regular fare (eg. $1.75), or using a multi-day pass stored on the card. A “Non-TAP Related Fare” is defined as any transaction that does not involve a TAP card, such as paying in cash. Finally, it is possible to purchase a pass on a TAP card from the bus operator, so these transactions were singled out and considered separately, as the process to purchase a new pass is not well known and would conceivably take longer.
Initial overview of route 720 and route 120 indicate relatively large shares of TAP card usage on each.
Selection/Development of Variables
Metro Rapid 720 TAP to Non-TAP
A number of variables have been selected initially to describe the primary components of passenger service time: fare payment, passenger volume, and unusual boardings.
Preliminary Results: Correlation to Dwell Times
Dwell Time: The dependent variable. Measured as the difference in time in seconds between the doors opening and closing. Fare Processing: Primarily concerned with describing fare activity, these variables are the prime indicators of TAP card use. N_NOTAP: The number of fare records with no TAP card. N_TAP: The number of fare records with a TAP card. N_TAPSALEOFPASS: Fare records buying pass with TAP. TAPFACTOR: Percentage of fare records using TAP. N_INCOOLD: Number of fare records recorded in the “cooldown” time.
Passenger Activity: The most direct cause of dwell time variance. ONs/OFFs: Boarding and alighting passengers. PAXACT: Passenger activity - sum of ons and offs. PAXSERVTIME: Time per passenger to board. Dwell divided by ONs. LOAD: Number of passengers on board the bus. LOADFACT: Ratio of passengers to seats. Values approaching 1.8 represent passenger congestion. D_LOAD: Dwell-affecting load; number of passnegers on bus prior to stopping. LOAD minus ONS. D_LOADFACT: Ratio of dwell-affecting load to seats.
Unusual Boardings: Passengers in wheelchairs and bicyclists, although infrequent, can contribute to large variances in dwell times. After processing, the APC and UFS data are combined a relational database with a one-to-many relationship on the basis of Vehicle ID, date, and time (the only fields the databases have in common). This provides for over 99,801 unique observations of dwell time, taking into account not only smart card usage, but passenger activity, vehicle loading, time of day, type of service, and type of vehicle.
N_BIKE: Number of bicycle loadings and unloading. N_WC: Number of passengers in wheelchairs boarding and alighting.
TAP-Related Fare, 68%
TAP-Related Fare, 48%
Non-TAP Related Fare, 31%
Non-TAP Related Fare, 49%
Sale of a Pass on TAP Card, 1%
Sale of a Pass on TAP Card, 2%
Metro Local 120 TAP to Non-TAP
In a bivariate correlation matrix, TAPFACTOR (the percentage of fare records using TAP in one dwell time observation) was found to have a slight, but statistically significant, correlation with dwell time (r = .220), however higher values of TAPFACTOR are associated with longer dwell times, when we would expect it to be opposite. This may be because other factors of dwell time are not yet accounted for, so here TAPFACTOR is standing in as a weak proxy for passenger activity. Preliminary Results: Differences in Mean Dwell Time As expected, NIGHT and PEAKH had statistically significant differences in mean dwell times for both Route 120 and 720. For NIGHT, this may be because there are less people boarding during the night time, and thus faster dwell times. For PEAKH, more people boarding the bus at these times will result in longer dwell times.
Next Steps The final step in the analysis will be to estimate a regression model to determine the weight of each variable in the variability of dwell time. This will be done after selecting the appropriate variables, and accounting for multicollinearity among them. Once the effect of TAP can be seen simultaenously with the other variables, its importance can be determined.
Acknowledgements and Credits In addition, extraneous factors to dwell time were accounted for. The variables PEAKH (Peak Hour) and NIGHT were created based on the time of day. PEAKH describes wether or not the dwell was during rush hour, and NIGHT describes wether or not the dwell was at night.
Prepared By: Daniel Shockley, UCLA Urban Planning Master’s Candidate, 2015 Client: Julia Salinas, Metro Faculty Advisor: Brian D. Taylor Image credits: (Bicycle) Edward Boatman, (Bus Stop) Marc Serre, (Currency) Creative Stall, (Time) Wayne Middleton, (Passenger) Alexander Wiefel, (Credit Card) Hugo Medeiros, (Wheelchair) Alex Qunito, - From the Noun Project - nounproject.com