John Douglas Hunt. Giovanni Circella* Alan Thomas Brownlee. Kevin Stefan. Michael McCoy. HBA Specto Inc., Calgary (Canada)

Transcription

1 Simplified Model of Local Transit Services Giovanni Circella* University of California, Davis (USA) Kevin Stefan HBA Specto Inc., Calgary (Canada) Michael McCoy University of California, Davis (USA) John Douglas Hunt University of Calgary (Canada) Alan Thomas Brownlee HBA Specto Inc., Calgary (Canada) *Corresponding author: 90 TH TRANSPORTATION RESEARCH BOARD MEETING

2 Aim of the study To develop a robust methodology for the synthetic representation of local transit services in the State of California to use in the California i Statewide Travel Demand Model (CSTDM). G. Circella 2

3 Models Short Distance Personal Travel Model (SDPM) Long Distance Personal Travel Model (LDPM) Short Distance Commercial Vehicle Model (SDCVM) Long Dist. Comm. Veh. Model (LDCVM) External Travel Model (ETM) Objectives To provide a realistic i representation of local transit in the Short Distance PT Model (component of the CSTDM) To avoid the explicit coding of all local transit lines in the State To reduce the efforts required for the update of the PT network, when defining future scenarios in the CSTDM G. Circella 3

4 Public Transportation in the CSTDM Public transportation is an important component of any transportation system Two main components in the short distance PT: rail and buses Limited railway services in California: 3 IC railways 4 commuter railways Some light rail and subway systems G. Circella 4

5 Public Transportation in the CSTDM Limited railway services in California (3 IC railways, 4 commuter railways, a few light rail and subway systems) Limited modifications in the Rail network due to the large investments for railways, the fixed routes (tracks) and high sunk costs On the contrary, Bus services are quite common in the State (50+ operators, bus lines) Bus services change frequently to adapt to travelers demand, changes in LU, funding and subsidies ( need of parsimony in reducing the requirements for network updates for future scenarios). G. Circella 5

6 Bus Services in California Downtown Sacramento, 2008 G. Circella 6

7 Bus Services in California (2) G. Circella 7

8 Local Transit services In a Statewide model, there is limited convenience in explicitly coding every bus line in the State Very high costs of development and implementation of the model Difficulties i in keeping the network updated dwith ihthe modifications introduced by local operators Not justified by the scale oftheprojectand the small mode share for PT in California ( limited relevance for CSTDM project) G. Circella 8

9 Modeling approach The synthetic methodology expresses local bus attributes as a function of other variables in the model. The approach uses functions estimated using observed data. Based on a service variable, as well as network and zonal properties. Public transportation characteristics are strongly correlated with relevant features of the road network (distances, average speed) and the land use. The methodology allows the evaluation of policy response, while keeping the modeling efforts consistent with the benefit. G. Circella 9

10 Theoretical Framework Traditionally, public transportation have had a more important role in higher density and more historical cities. Additional impact ofhistorical variables: traditions, habits, urban form, years in which the urban area was developed. Demand for transit affected by the transit supply. Role of subsidies and funding. The distance from the transit stations is a determinant of transit patronage. Number of transit users affected by the number of employees/workplaces in the central business district (CBD) (Hendrickson, 1985). G. Circella 10

11 Source: Commuting in America III (TRB, 2006) G. Circella 11

12 x x x xx x xx CANADA x AUSTRALIA USA EUROPE x x x ASIA G. Circella 12

13 Transit Supply and Urban Density Limited experience in literature on the relations between transit supply and land use and sociodemographic variables. HigherGDPisassociated is with better services. Transit supply is better in the Capitals (Albalate and Bel, 2010). Regional patterns are observed, and better services are usually found din more historical i regions with higher h density. Role of subsidies and policies towards PT. Integration of services and coordination with pedestrian and bike networks (Buehler, 2009)....More space for research in this field! G. Circella 13

14 Local Transit in the CSTDM framework Two different functions for In Vhil Vehicle Time (IVT) and Out of of Vehicle time (OVT) for local transit Different models by time of the day (4 time periods) Division of the State of California in 32 Catchment Areas The catchment areas account for the different levels of service and the easiness to transfer from one line to another Some TAZs are not served by any local ltransit service G. Circella 14

15 The catchment areas Catchment areas define the area of influence (catchment) of bus operators and the easiness to transfer from a bus line to another The Local Transit Functions use 4 key inputs: Transfer Area: area in which a person can travel Service Area: area where transit services are generally operated by one operator Level of Service: it represents the quality of local bus services provided in each service area Fare: value (in dollars) paid by the travelers to use the transit services G. Circella 15

16 The catchment areas (2) Each TAZ is assigned to a catchment area, depending on the proximity from the centroid to the closest bus lines (<3mi) If the test above fails (large distance from the centroid) but at least one bus line passes through the TAZ, then the TAZ is included d in that catchment area Otherwise, the TAZ is not included in any catchment area (no transit services) G. Circella 16

17 The catchment areas (3) Transfer Service Area Area Agency Level of Service 1 1 Del Norte County Public Transit Humboldt Transit Authority Lassen Transit Service Agency Redding Area Bus Authority Butte County Transit, Chico Area Transit, Oroville Area Transit Gold Country Stage (Nevada County) Sacramento Regional Transit District Yolobus, Unitrans Placer County Transit, Roseville Transit El Dorado County Transit Authority Sonoma County Transit, Santa Rosa CityBus, Petaluma Transit San Mateo County Transit District (SamTrans) San Francisco Municipal Railways (Muni) Alameda-Contra Costa Transit District (AC Transit) Santa Clara Valley Transportation Authority (VTA) Golden Gate Transportation District 46.1 Central Contra Costa Transit Authority, Eastern Contra Costa Transit Authority, Western Contra Costa Transit Authority Livermore / Amador Valley Transit Authority Vallejo Transit, Fairfield and Suisun Transit, Benicia Breeze The VINE (Napa County) San Joaquin Regional Transit District i t Modesto Area Express Note: first 10 Transfer Areas reported in the table above (from a total of 32). G. Circella 17

18 The catchment areas (4) 23. Los Angeles 8. San Francisco Bay Area 26. San Diego G. Circella 18

19 Local Transit Functions: Main assumptions IVT is affected by the speed of the private vehicles that share the road HOVlanes (where available) can be used by buses IVT correlated with HOV3 congested speed Variables depend don the LOS (investments in the transit system and local conditions) OVT depends on the distance and the densities of the areas of origin and destination G. Circella 19

20 The Data Observed data collected from internet through the Google platform (information to the users from by Graphserver: python code available to test the data (tool for transit agencies) to post on the Google Platform Use of the Google dt data to extract trecords with the required information (IVT, OVT, catchment area, geographical locations of origins and destinations, time of the day) Data are merged with information available from other sources (e.g. Caltrans, other CSTDM components) to create the datasets for the estimation of the IVT and OVT functions for local transit G. Circella 20

21 The sample Total sample size: 91,074 records collected for inter zonal trips in the State of California 29 service areas covered in the sample Final sample with 4 time periods during the day: AM peak Midday PM peak Night (Off peak) G. Circella 21

22 The database format G. Circella 22

23 In Vehicle Time (IVT) Function (2) Assumption of dependence on HOV3 Time. Use of a quadratic form suggested by the data, to represent the diminishing effect of HOV3 time for longer trips. Function depends on time of the day and on the LOS: G. Circella 23

24 Final IVT function: Peak Time PEAK IVT MODEL N= Unstandardized Coefficients Standardized Coefficients B Std. Error Beta t Sig. HOV3_Ti me HOV3_ LOSHOV Adjusted R R R Square Square Std. Error of the Estimate Function estimated for HOV3time < 65 min Need to cap the function for higher values of time G. Circella 24

25 Final IVT function: Peak Time (2) PEAK IVT MODEL N= Unstandardized Coefficients Standardized Coefficients B Std. Error Beta t Sig. HOV3_Ti me HOV3_ LOSHOV Adjusted R R R Square Square Std. Error of the Estimate G. Circella 25

26 Final IVT function: Off Peak Time OFF PEAK IVT Model N= Unstandardized Coefficients Standardized Coefficients B Std. Error Beta t Sig. HOV3_Ti me HOV3_ LOSHOV R R Square Adjusted R Square Std. Error of the Estimate G. Circella 26

27 IVT: Peak and Off Peak G. Circella 27

28 Out of Vehicle Time (OVT) Function Assumption of dependence on HOV3 Distance. Function depends on time of the day and on the LOS. Function depends on the population and employment densities at the origin and destination: ) G. Circella 28

29 Final OVT functional forms PEAK WaitTime N= Model Unstandardized Coefficients Standardized Coefficients B Std. Error Beta t Sig. SQLOS LOSDIST E SQP2EDENS E 137 R R Square Adjusted R Square Std. Error of the Estimate OFF PEAK WaitTime N= Standardiz ed Coefficien Model Unstandardized Coefficients ts t Sig. B Std. Error Beta SQLOS LOSDIST SQP2EDENS E 24 R R Square Adjusted R Square Std. Error of the Estimate G. Circella 29

30 OVT: Peak and Off Peak G. Circella 30

31 Implementation in the CUBE script G. Circella 31

32 The local transit model in CUBE G. Circella 32

33 The control file for catchment and fares in CUBE CUBE runs the local transit model with 4 script files (the same for all scenarios) The LocalBus_{year}.dbf file provides information for catchment areas, LOS and fares for each scenario. G. Circella 33

34 Conclusions The use of the hybrid approach (synthetic methodology) for local transit provides a robust representation of bus services in the Short Distance Personal Travel Model of the CSTDM Framework. The methodology eliminates the need for explicit coding of all bus slinks and connections in California. In the model, local transit attributes are expressed in dependence of othertransportation transportation andland use variables. Two equations are estimated for the In Vehicle Time (IVT) and the Out Vehicle Time (OVT). Models are sensitive to different level of service and local conditions through the definition of the catchment areas for local transit operators. Different models are estimated for Peak and Off Peak time of the day, to account for variability in services by time of the day. G. Circella 34

35 Conclusions (2) The model is estimated using observed data collected from the internet platform Google (total sample: 90k records) All models have a very good goodness of fit, with R square > 0.9 for IVT (> > 0.8 for OVT). The approach optimizes the use of resources, providing a robust model for local transit without extensive network coding in a Statewide Travel Demand Model. The model is appropriate also in consideration of the limited mode share of bus services in California. The synthetic methodologyallows several possibilities ofpolicy testing through the simulation of their impact on the LOS of the operators. The proposed approach allows easier maintenance of the transit network, withreduced efforts required for the definition offuturefuture scenarios. G. Circella 35

36 Additional information available on For any question, please contact: Dr. Giovanni CIRCELLA Urban Land Use and Transportation Center University of California, Davis Office Phone: (530) e mail: gcircella@ucdavis.edu