USF Libraries
USF Digital Collections

Sensitivity Analysis of the Transit Boardings Estimation and Simulation Tool (TBEST) Model

MISSING IMAGE

Material Information

Title:
Sensitivity Analysis of the Transit Boardings Estimation and Simulation Tool (TBEST) Model
Physical Description:
Book
Language:
English
Creator:
Vuckovic, Dajana
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla.
Publication Date:

Subjects

Subjects / Keywords:
Service planning
Elasticities
Headway
Span of service
Routes
Dissertations, Academic -- Civil Engineering -- Masters -- USF   ( lcsh )
Genre:
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Abstract:
ABSTRACT: Public transportation, although modest in the United States carrying about 2 percent trips, still serves millions of people as the main and only means of transportation. Recently released data set by Census, the 2006 American Community Survey (ACS) shows the main mode of travel for work commute is not surprisingly the automobile with over 86 percent and public transportation with nearly 5 percent users. Transit agencies strive to provide effective, convenient, and desirable transport. Because of the constant changes in our environment, being able to predict the response of riders to different network or system changes is extremely useful. Ridership can be described as a function of the amount of service supplied such as frequency, span of service, and travel time. One of the methods for estimating ridership forecasts and evaluating ridership response is to use the new state-of-art software TBEST.^ TBEST stands for Transit Boardings Estimation and Simulation Tool and is the third generation of such transit models sponsored by the Florida Department of Transportation (FDOT). Designed for comprehensive transit network and short term transit planning, it offers great benefits to its users. TBEST is a user friendly, yet very advanced transit ridership forecasting graphical software which is interfaced with ArcGIS. This paper evaluates different sensitivity tests and compares the results to known industry used elasticities. Because the current TBEST experience is modest, the results will provide users with a general idea of the model's sensitivity and help in the process of model refinements. Sensitivity tests such as service frequency, span of service, service allocation, and travel time will be carried out in a systematic order for all six time periods as defined by TBEST.^ Results showed that TBEST Model is overestimating and is highly sensitive to headway changes, specifically headway decrease. The opposite effect of almost no sensitivity is shown for the in-vehicle travel times.
Thesis:
Thesis (M.S.C.E.)--University of South Florida, 2007.
Bibliography:
Includes bibliographical references.
System Details:
System requirements: World Wide Web browser and PDF reader.
System Details:
Mode of access: World Wide Web.
Statement of Responsibility:
by Dajana Vuckovic.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 59 pages.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 001966688
oclc - 263172301
usfldc doi - E14-SFE0002247
usfldc handle - e14.2247
System ID:
SFS0026565:00001


This item is only available as the following downloads:


Full Text

PAGE 1

Sensitivity Analysis of the Transit Boardings Estimation and Simulation Tool (TBEST) Model by Dajana Vuckovic A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science in Civil Engineering Department of Civil and E nvironmental Engineering College of Engineering University of South Florida Co-Major Professor: Steven E. Polzin, Ph.D. Co-Major Professor: Jian Lu, Ph.D. Xuehao Chu, Ph.D. Manjriker Gunaratne, Ph.D. Date of Approval: November 6, 2007 Keywords: service planning, elasticities headway, span of service, routes Copyright 2007, Dajana Vuckovic

PAGE 2

ACKNOWLEDGMENTS I thank Dr. Polzin for his guidance and s upport throughout this work. His knowledge and expertise in the field of transportation is much respected and appr eciated. Dr. Polzins mentorship, help, and support in the pursuit of my career goals have been enormous. Special thanks to Center for Urban Trans portation Research (CUT R) for providing me the opportunity to work as a graduate rese arch assistant while pursuing my Masters Degree. The knowledge and personal growth acquired at CUTR has been tremendous. I would also like to thank my graduate committ ee for their time and guidance. Both Dr. Lu and Dr. Gunaratne have made a positive im pact on my studies by providing their knowledge through teaching many classes I took. And without Dr. Chus expertise, this thesis would have not been possible as he provided so much to the development of TBEST. I am thankful to everyone that has pl ayed a part in comple tion of this thesis.

PAGE 3

i TABLE OF CONTENTS LIST OF TABLES.............................................................................................................iii LIST OF FIGURES............................................................................................................v ABSTRACT.......................................................................................................................vi CHAPTER 1 INTRODUCTION........................................................................................1 1.1 Background.......................................................................................................1 1.2 TBEST..............................................................................................................2 1.3 Objective and the Scope....................................................................................3 1.4 Methodology.....................................................................................................3 1.4.1 Transportation Elasticities.................................................................4 1.5 Outline of the Thesis.........................................................................................7 CHAPTER 2 LITERATURE REVIEW.............................................................................9 2.1 Ridership Forecasting a nd Service Planning Methods.....................................9 2.2 Transportation Elasticities..............................................................................10 CHAPTER 3 TBEST MODEL.........................................................................................18 3.1 Elements of TBEST........................................................................................18 3.2 Methodology...................................................................................................21 CHAPTER 4 BASE CASE...............................................................................................23 4.1 Existing Conditions.........................................................................................23 4.1.1 Demographics..................................................................................23 4.1.2 Transit Services................................................................................24 4.2 TBEST Network Development.......................................................................25 4.2.1 Routes..............................................................................................25 4.2.2 Stops.................................................................................................27 4.2.3 Network Attributes...........................................................................27 4.3 TBEST Calibration and Scaling......................................................................29 4.3.1 System Results.................................................................................30 CHAPTER 5 ALTERNATIVE SCE NARIOS AND MODEL RESULTS......................35 5.1 Alternative Scenarios......................................................................................35 5.1.1 Headway..........................................................................................35 5.1.2 Service Allocation............................................................................40 5.1.3 Span of Service................................................................................41 5.1.4 Travel Time......................................................................................42 CHAPTER 6 CONCLUSIONS AND FURTHER RESEARCH......................................45 6.1 Conclusions.....................................................................................................45 6.2 Further Research.............................................................................................47 REFERENCES.................................................................................................................49

PAGE 4

ii APPENDICES..................................................................................................................51 APPENDIX A TBEST Model..............................................................................52 APPENDIX B TBEST M odel Output Results......................................................55

PAGE 5

iii LIST OF TABLES Table 1 San Diego Vehicl e-Miles Elasticities..................................................................11 Table 2 In-Vehicle Travel Time Elasticities by Time Period...........................................12 Table 3 Bus Route or Small System Headwa y Elasticities Observed in 1960s/70s......13 Table 4 Headway Elastici ties by Time Period..................................................................15 Table 5 Recommended Tran sit Elasticity Values.............................................................16 Table 6 Transit Ridership Factors.....................................................................................17 Table 7 Definitions of Time Periods in TBEST...............................................................19 Table 8 Lakeland Area Mass Transi t District Route Network.........................................26 Table 9 Winter Haven Area Transit Route Network........................................................26 Table 10 Polk County Transit Services Route Network...................................................26 Table 11 Fare Structure for LAMT D, WHAT, and PCTSD Systems..............................27 Table 12 Special Generators.............................................................................................28 Table 13 Weekday Observed Versus TB EST Estimated Ridership by Route..................30 Table 14 Saturday Observed Versus TBEST Estimated Ridership by Route..................32 Table 15 Calibration Summary.........................................................................................33 Table 16 Polk County Alternatives...................................................................................35 Table 17 Alternative 1 Headway Scenarios......................................................................36 Table 18 Weekday Ridership Response to Headway Changes by Time Period...............36 Table 19 Saturday Ridership Response to Headway Changes by Time Period...............37

PAGE 6

ivTable 20 System Elasticities by Time Period...................................................................38 Table 21 Number of Stops in Each Service Level Category............................................39 Table 22 Decreasing Headway Elasticitie s Based on Current Level of Service...........39 Table 23 Increasing Headway Elasticities Based on Current Level of Service.............39 Table 24 Service Allocation and Frequency Analysis......................................................40 Table 25 Night Service Span Increase..............................................................................42 Table 26 Weekday Ridership Response to Travel Time Changes by Time Period..........42 Table 27 Saturday Ridership Response to Travel Time Changes by Time Period...........43 Table 28 Travel Time Elasticities.....................................................................................44 Table 29 Current Level of Service by Route....................................................................55 Table 30 AM Peak Ridership Re sponse to Headway by Route........................................56 Table 31 Off Peak Ridership Response to Headway by Route........................................57 Table 32 AM Peak Ridership Response to Travel Time by Route...................................58 Table 33 Off Peak Ridership Respon se to Travel Time by Route....................................59

PAGE 7

v LIST OF FIGURES Figure 1 Different Transit Elasticities.................................................................................7 Figure 2 TBEST Model User Interface.............................................................................20 Figure 3 Polk County Population......................................................................................23 Figure 4 Polk County Transit System Organization Chart...............................................25 Figure 5 Weekday Observed Versus TB EST Estimated Ridership by Route..................31 Figure 6 Saturday Observed Versus TBEST Estimated Ridership by Route...................33 Figure 7 Calibration Summary..........................................................................................34 Figure 8 Ridership Responses to Headway Changes by Time Period.............................38 Figure 9 Service Allocation Analysis Route 31................................................................41 Figure 10 Ridership Response to Trav el Time Changes by Time Period.........................43 Figure 11 Different Types of Reports...............................................................................52 Figure 12 Mapped Riders hip Output Sample...................................................................53 Figure 13 Model Output Su mmary Reports Sample.........................................................53 Figure 14 Alternative Scenario List..................................................................................54

PAGE 8

vi SENSITIVITY ANALYSIS OF THE TRANSIT BOARDINGS ESTIMATION AND SIMULATION TOOL (T-BEST) MODEL Dajana Vuckovic ABSTRACT Public transportation, although modest in the Un ited States carrying a bout 2 percent trips, still serves millions of people as the main and only means of transportation. Recently released data set by Census, the 2006 American Community Survey (ACS) shows the main mode of travel for work commute is not surprisingly the auto mobile with over 86 percent and public transportation with nearly 5 percent users. Transit agencies strive to provide effective, convenient, and desirable transport. Because of the constant changes in our environment, being able to predict the response of riders to different network or system changes is extremely useful. Riders hip can be described as a function of the amount of service supplied such as frequency, sp an of service, and travel time. One of the methods for estimating ridership forecasts and evaluating ridership response is to use the new state-of-art software TBEST. TBEST st ands for Transit Boardings Estimation and Simulation Tool and is the th ird generation of such transi t models sponsored by the Florida Department of Transportation (F DOT). Designed for comprehensive transit network and short term transit planning, it offers great benefits to its users. TBEST is a user friendly, yet very advanced transit ride rship forecasting graphical software which is

PAGE 9

vii interfaced with ArcGIS. This paper evaluates different sensitivity tests and compares the results to known industry used elasticities. Because the cu rrent TBEST experience is modest, the results will provide users with a general idea of the models sensitivity and help in the process of model refinements. Sens itivity tests such as service frequency, span of service, service allocation, and travel time will be carried out in a systematic order for all six time periods as defined by TBES T. Results showed that TBEST Model is overestimating and is highly sensitive to headway changes, specifically headway decrease. The opposite effect of almost no sens itivity is shown for the in-vehicle travel times.

PAGE 10

1 CHAPTER 1 INTRODUCTION 1.1 Background Public transportation, although modest in the United States carrying about 2 percent trips still serves millions of people as their main and sometimes only means of transportation. Recently released data se t by the Census, the 2006 American Community Survey (ACS) shows the main mode of travel for work commute is not surprisingly the automobile with over 86 percent and public transportation with nearly 5 percent users. What are the general goals and objectives of transit agencies and what is needed to meet those? Transit agencies strive to provide e ffective, convenient, a nd desirable transport. Our environment is constantly changing on soci al, demographic, and economic level. An important area of research is looking into how those changes effect transit patronage. Being able to predict the res ponse of riders to different ne twork or system changes is extremely useful. Those changes can be m odifications or improvements to the existing system or additions of completely new serv ices. Resource allocation often depends on accurate information on these impacts. Flor ida Department of Transportation (FDOT), Public Transit Office (PTO) recently passed a rule, Public Transit 14-73.00 that requires all transit agencies to submit Transit Deve lopment Plans (TDPs). TDP is a ten year planning document which among others includes an estimation of the communitys demand for transit service usi ng the planning tools provided by the Department, or a Department approved transit demand estimation technique with supporting demographic,

PAGE 11

2 land use, transportation, and transit data. The result of the transit demand estimation process shall be a ten-year a nnual projection of transit ride rship. This is one example where transit agencies are required to provid e ridership forecasts in order to receive funding. One of the approved methods for accomplishing this task of ridership forecasting is to use TBEST. Section 1.2 discusses what TBEST is. 1.2 TBEST TBEST stands for Transit Boardings Estim ation and Simulation Tool. This third generation transit model is a be tter and improved version of th e previous two Integrated Transit Demand and Supply Model (ITSUP) a nd Regional Transit Feasibility Analysis and Simulation Tool (RTFAST). Developed by the FDOT Public Transit Office, to provide support completing their TDPs, tran sit agencies can use TBEST as a tool to estimate their ridership forecasts. TBEST is user friendly, yet very advanced transit ridership forecasting graphical software which is interfaced with ArcGIS allowing transit agencies fairly easy manipulation of thei r network. As mentioned in section 1.1, being able to predict the ridership response to different system changes is important. Some typical system changes transit agencies and transportation planners are usually exploring are the service frequency, network coverage, fare pricing, span of service and speed. TBEST is capable of evaluating these variab les and how the ridership is impacted by each of these individually or in combination. Unlike many other transit planning models, TBEST has the capability of simulating riders hip at stop-level, t hus providing more detailed, accurate analysis. Stoplevel ridership can also be ag gregated to route, segment, and system level. Transit riders hip at the stop-level depends on a wide range of factors

PAGE 12

3 which TBEST incorporates. For example, dir ect versus transfer boardings, time of the day based analysis, socio-economic character istics, network connectivity and others. TBEST User Guide describes each of these in de tail and provides users with the complete TBEST methodology. More details on TBES T Model are provided in Chapter 3. 1.3 Objective and the Scope How travel demand is affected by transpor tation changes has been a growing field of interest among transportation professionals With this new software, we are a step closer to better and easier estimation and measurement of t hose changes. However, there are some caveats that need to be addressed. TBEST is new software and a planning tool, so the operation experience is modest. Alth ough there have been previous versions, continuous evaluations and updates of the mode l will be necessary. The overall objective of this paper is to help improve the model and provide users a general idea of the model sensitivities. This will be achieved through a nu mber of sensitivity tests. Polk County will be used as the case scenario. The results will be presented in terms of transportation elasticities or percent changes and will be co mpared to the known industry elasticities. 1.4 Methodology This paper attempts to analyze the TBEST Model sensitivities. As of now, Pennsylvania DOT has already initiated fo recasts with TBEST. In June 2007, two technical memos were produced. The first one described the network development for the two transit agencies used in this study, EM TA and Rabbit Transit. The second technical memo described the network calibration pr ocess. Completed in September, this Forecasting Short-Term Ridership report produced by Gannet Fleming for the

PAGE 13

4 Pennsylvania Department of Transportatio n included four scenarios that were developedand analyzed for both the EMTA and Rabbit Transit. The alternatives included headway adjustments, extending routes, addi ng new routes, modifying the route types and others. This paper uses Polk County as the base case for conducting a series of sensitivity tests with the TBEST Model. The series of tests can be grouped into four major categories: headway, service al location, service span, and trav el time changes. Different scenarios will be created for each of these categories for testing the sensitivity. The results will be analyzed and compared to known transit service frequency elasticities. Service allocation evaluation will include a series of model runs that will test how ridership responds to additional service versus service increases of the current system. Polk County currently does not offer Sunday se rvice. As part of the span of service analysis, Sunday service will be added and the results evaluated. Ridership response to service span changes will also be evaluate d by adding more arri vals during the night period. Each alternative scenar io developed will be further discussed in Chapter 5. 1.4.1 Transportation Elasticities Throughout the previous secti ons of this report there was much mention of the ridership response. How is this ridership re sponse measured and quantified? Law of Demand is a concept describing a pattern when the price of the good decreases, its consumption increases, and vice versa. Ec onomists measure price sensitivity using elasticities. Elasticity is defined as the percent cha nge in a consumption of a good, caused by a one percent change in its price. There are diffe rent forms or ways for

PAGE 14

5 calculating elasticity and those measures can be f ound throughout the transportation literature. Elasticities usually provide satisf actory results in assess ing ridership response. However, elasticities need to be used with caution. In order for elasti city measures to be applicable in transportation, the change must be a relative one that involves quantifiable percent increase or decrease in the system. So elasticities ca nnot be used to assess the ridership response of a new bus system. If the el asticity value is 1, we refer to that as the unit elasticity. Elasticity values greater than one are called elastic which means the price or service change causes more than proportional change in consumption. Elasticity values less than one are referred to as ine lastic, meaning the pr ice or service change causes less than proportional change in consumption. Transportation literature typically contai ns three different methods in computing elasticities: 1) Point elasticity 2) Arc elasticity 3) Shrinkage ratio Point elasticity is described as p = Q P x dP dQ where p is the elasticity at price P, and Q is the quantity demanded at that price. The most frequent form used in transportation is the Arc Elasticity Arc elasticity is defined as following: = P Q log log = 1 2 1 2 log log log log P P Q Q

PAGE 15

6 where = elasticity value Q1 and Q2 = demand before and after P1 and P2 = price or service before and after. When one value is zero, for example in the case of adopting or te rminating free use of transit, the mid-point arc elastic ity shown below must be used. = ) )( ( ) )( ( ) ( ) ( 2 / ) ( / 2 / ) (2 1 1 2 2 1 1 2 2 1 2 1 2 1 2 1Q Q P P P P Q Q Q Q P P P Q P P P Q Q Q The third form is the shrinkage ratio. More recently, this method has been used in road value pricing studies. Instead of the shrinkage ratio, the term thats been used is approximated point elasticity. = 1 1/ / P P Q Q 1 1 2 1 1 2/ ) ( / ) ( P P P Q Q Q When using elasticities, one needs to be cautious. Not only are there differences in nomenclature used throughout th e literature, but there are also differences in results one can get using different formulas. When the pe rcent change in transportation service is small, all the methods give approximately the same elasticity value. However, when there are large differences in service or fare change s, the results are very different. Figure 1 illustrates those differences when arc elasticity shrinkage ratio, and point elasticity is used for an initial point price elasticity of -0.30 (Mayworm, Lago, and McEnroe, 1980).

PAGE 16

7 Figure 1 Different Transit Elasticities 1.5 Outline of the Thesis This thesis contains si x chapters. Chapter 1 provi des the introduc tion of two areas: transit ridership respons e and the TBEST Model. The first section introduces the importance of accurately measuring transit ri dership response and the transit ridership forecasting requirements for certain grants. Foll owing that section is a brief introduction of the TBEST Model used in this paper an d the sensitivity tests performed. Chapter 2 consists of literature review broken down into the literature review of transit ridership forecasting and transit service planning methods The second part covers the research

PAGE 17

8 completed on the ridership response and re views known transit in dustry elasticities. Chapter 3 discusses the TBEST Model structur es, its features, and tools. Following the TBEST Model chapter, Chapter 4 will provid e the description of the base case, the network development, and the model calibra tion and scaling. Chapter 5 discusses the alternative scenarios developed and tested a nd provides the results of the model runs and the calculated elasticities. Also, here th e elasticities are compared to the known elasticities from the literature. Chapter 6 provides conclusions based on the model results. That chapter offers general conclusions and possible improvements for the TBEST Model.

PAGE 18

9 CHAPTER 2 LITERATURE REVIEW Literature review in this chapter consists of two parts. The first part covers the work done as it relates to the TBEST m odeling approach. The second part of the literature is a review of research in the area of ride rship response and transportation elasticities. Here some of th e known industry elasticities are presented. The second part of this chapter is broken down into sections as it relates to this paper. Much research has been done in this area and it is difficult to cover the whole body of literature. TBEST Model methodology and framework will be di scussed in detail in Chapter 3. 2.1 Ridership Forecasting and Service Planning Methods Budget preparation, a llocation of resources, and bett er service plan ning are all reasons why one may want know the impacts on ridership from certain service changes. Literature on assessing ridership impacts date s back to the early 1980s. The traditional four-step modeling process: trip generation, trip distribution, moda l choice, and traffic assignment does not work for assessing and eval uating transit ridershi p at the route level because of accuracy issues (Multisystems 1982) This model was designed for large scale changes and it typically uses zonal level data After realizing the difficulties in assessing the impacts of transit ridership with the four-step process, simpler models were developed. In 1984 Horowitz developed a simp lified version of the four-step process. After that some of the other early work has focused on route level analysis. Here many

PAGE 19

10 of the problems of the four-step process are avoided, but there are still issues to be considered. These direct demand models use homogeneous land use and everything else along the route. This type of homogeneity is very unlikely to be the case. Just recently, transit ridership modeling and forecasting has been analyzed at the segment level (Peng et al. 1997; Kimpel et al. 2000). In their simu ltaneous route-level transit patronage model they incorporate transit demand, supply and inter-route effects in a simultaneous system. The results of this model indicate that si multaneity exists between transit demand and supply and that there is also a strong inte rrelationship among routes. And while this approach avoids both problems the previous two had, it still assumes homogeneity along the route segments. This st udy also shows that single-equa tion model overestimates the effects of service increases because it assumes that all ridership incr ease on a new route is new ridership. This model could be improve d at the transit stop-level if reliable stoplevel data is available. There is not much literature that shows the attempts of modeling and forecasting ridership at the stop-level Kikuchi and Miljkovic in 2001 used fuzzy inference method to forecast ridership at the stop-level. The data used in this study was actual bus stops in Delaware. 2.2 Transportation Elasticities In 1981, Lago, Mayworm and McEnroe summar ized at that time the current state of knowledge on the transit serv ice elasticities from demons trations and demand models. In their Transit Service Elasticities Evidence from Demonstration and Demand Models paper, they group the transit se rvice elasticities into two broad categories. One is quasiexperimental, using data generated by a practic al demonstration of an actual change. The

PAGE 20

11 other one is non-experimental, re lying on some data where changes are part of historical data analysis. The quasi-experimental approach was used by Kemp (1979) in Atlanta and by Goodman, Green, and Beesley (1977) in San Diego. They used mo nthly data series. Among other things, they developed vehicle-m ile elasticities. The aggregate elasticities for the San Diego area varied from +0.75 to +0. 85. In Atlanta, the vehicle mile elasticities estimated by Kemp were +0.30. This difference can be attributed to more service being available and service expans ion occurred over a much sh orter time period. As with service frequency changes, these results sugge st that the response to increase in vehiclemiles of service depends on the initial amount of service provided. Other factors such as fares, auto availability, and size of the urba nized area can be as equally important (Lago 1981). Table 1 shows San Diego vehicle-mile elasticities estimated by Goodman, Green, and Beesley in 1977. Table 1 San Diego Vehicl e-Miles Elasticities Radial routes to CBD +0.65 Central-city routes +0.72 Suburban routes +1.01 One of the most important factors affec ting public transpor tation ridership is travel time. Measuring ridership response to travel-time changes is very difficult. Historically the only availabl e travel-time elasticities came form mode-choice and transitdemand models. Therefore those travel-time elasticities should be used with caution. The only evidence for the in-vehicle travel-time re sponse was obtained from an experiment in three cities. Table 2 displays these in-vehicle elasticities by time period.

PAGE 21

12Table 2 In-Vehicle Travel Time Elasticities by Time Period Time Period Elasticity Project Peak -0.29 0.13 (9 cases) Miami, Seattle, Boston Off Peak -0.83 (1 case) Seattle Aggregate value -0.35 0.21 (10 cases) All of the above Source: Wattleworth (1978), A.M. Voorhees and Associates (1973 ), Dupree and Pratt (1973), and Ecosometrics, Inc. (1980). Kraft and Domencich in 1970 observed higher el asticities during peak hours, especially for choice riders, in improved travel-time rath er than reduction in fares. Mullen (1975) analyzed some bus demonstr ation data in England and found that off peak headway elasticities are significantly higher than pe ak-period elasticities. One of the most comprehensive pieces of literatu re covering traveler responses to different transportation system changes are the TCRP Report 95 seri es. Since 1977, this re port has served as reference to many professionals. Th is part of the literature review will cover some of the elasticities presented in these reports. Also, Todd Litman produced two papers, Transportation Elasticities How Prices and Other Factors Affect Travel Behavior and Transit Price Elasticitie s and Cross Elasticities, which will be reviewed later in this section. Very common service changes a transit agency makes are scheduling and frequency modification. The objectives can ra nge from cost effectiveness to service quality. Service quality can be affected by either reducing passenger wait times or reducing wait time for transfers. Sometimes, ho wever, transit agencies are forced reduce the frequency due to funding. There are se veral types of sche duling and frequency changes a transit agency can make. Service frequency changes, service hours changes, and frequency changes with fare changes are so me of the types of m odifications agencies can make. These types of changes usually do not involve bus routing and coverage.

PAGE 22

13 Quantifying response of ridership to these changes is usually done using elasticities which was briefly introduced in th e pervious chapter. We know that increased transit frequency is expected to result in in creased ridership, and vi ce versa. Because of a wide variation in observed results it has b een suggested that service frequency and ridership changes may not be able to be re presented with a single numerical relationship (Holland 1974). More recent res earch indicated that frequency elasticities can be grouped in +0.3 or in +1.0 category. However, if one considers historical and current observations, the average service elasticity is around +0.5 (Pratt 2004). Historical data on service elasticities is presented in Table 3. Table 3 Bus Route or Small System Headwa y Elasticities Observed in 1960s/70s Massachusetts Demonstrations Headway Elasticity Months After Implementation Boston-Milford suburban route (new headway approx. hourly) -0.4 10-12 Uxbridge-Worcester suburban route (new headway hourly) -0.2 7-9 Adams-Williamstown city route (new headway approx. hourly) -0.6 1-3 Pittsfield city route (raised from 3 to 8 round trips daily) -0.7 1-3 Pittsfield city route (raised from 10 to 15 round trips daily) -0.6 1-3 Newburyport-Amesbury (depressed area) city route (new headway 30 min. peak/60 min. midday) -0.4 6-8 Fall River (depressed area) city service (overall 20 percent service increase) nil 4-6 Fitchburg-Leominster city route (new afternoon headway 10 min., to match morning) -0.3 6-8 Boston downtown distributor, Phase 1 (new midday headway 5 min., to match peak) -0.8 5-7 Boston downtown distributor, Phase 2 (new headway 4 min. base, 8 min. midday) -0.6 8-10

PAGE 23

14Table 3 Continued Boston rapid transit feeder route (new midday headway 5 min., to match peak) -0.1 4-6 Other Contemporary Findings Detroit city route (new headway 2 min. peak, 3.5 min. midday) -0.2 Chesapeake, VA, suburban service (new headway 35 to 42 min.) -0.8 Stevenage, England (peak period/off peak; new headway 5 min.) -0.4/-0.3 Madison, WI, circulator routes (Saturday/Sunday; new headway 20/30 minutes) -0.2/-0.6 Sources: Massachusetts Demonstrations Mass Transp ortation Commission et al. (1964). Massachusetts elasticity calculations Pratt, Pedersen and Mather (1977). Other Findings Holland (1974), Mayworm, Lago and McEnroe (1980). Some general conclusions can be made a bout the frequency el asticities based on historical and current research Elasticities tend to be hi gher in suburban systems than central cities. Also, it was observed that elas ticities are significantly higher in areas where the frequency was originally low. Service hours changes are very different from frequency changes, but their effect is often not identified separately (Pratt 2004). Service hours changes include increasing or decreasing span of service where the service during the day is either shortened or lengthened Another common change transit agencies implement is adding or eliminating days of service, usually Sunday operations. Bus headway elasticities discussed in Table 3 are also looked at in terms of the time of the day. Table 4 shows those results.

PAGE 24

15Table 4 Headway Elasticities by Time Period Time Period Number of Observations Arc (Mid-point) Elasticity Standard Deviation Peak Hours 3 -0.37 .19 Off Peak Hours 9 -0.46 .26 Weekends 4 -0.38 .17 All Hours 7 -0.47 .21 We note that there is higher elasticity in the off-peak period. This can be attributed to lesser service frequencies in th e off peak periods. Often, of f peak period travel can be related to choice riders. Anot her factor that needs to be considered when looking at the service frequency changes is socio-demographi cs. Public transportation mostly serves those that are dependent on transit, also know n as captive riders. Therefore, passengers that are most attracted by fr equency improvements tend to be choice riders and mostly in the middle to upper income groups (Holland 197 4). Two-year research on frequency and fare changes in the greater Dallas area reveled greater sensitivity to fares than service in the city center, and the opposite in the suburbs for both express and local service (Allen, 1991). In general, ridership app ears to be more sensitive to fare changes than frequency changes where frequency is high, and opposit e where service levels are low. Service restructuring of a transit system tries to improve the overall system effectiveness and productivity. Ridership surveys, transit planni ng models, and GIS application are some of the tools for measuring ridership response to service reconfiguration. Price Elasticities and Cross-Elasticities study by Todd Litman s uggests that the elas ticity of transit ridership with respect to fares is lower for captive riders than choice riders. Also, elasticities are about tw ice as high for travel during off peak than during peak times. This

PAGE 25

16 paper also suggests that because of high vari ability and uncertainty it is better to use ranges rather than point value when using el asticity analysis. The evidence shows that fare elasticity is usually in th e -0.2 to -0.5 range in the shor t run (first year), and increase to -0.6 to -0.9 over the long run (five to te n years). Table 5 summar izes general values found in this research. Table 5 Recommended Transit Elasticity Values Market Segment Short Term Long Term Transit ridership WRT transit fares Overall 0.2 to .5 0.6 to .9 Transit ridership WRT transit fares Peak .15 to .3 .4 to 0.6 Transit ridership WRT transit fares Off peak .3 to .6 .8 to .0 Transit ridership WRT transit fares Suburban Commuters .3 to 0.6 .8 to 1.0 Transit ridership WRT transit service Overall 0.50 to 0.7 0.7 to 1.1 Transit ridership WRT auto operating costs Overall 0.05 to 0.15 0.2 to 0.4 Automobile travel WRT transit costs Overall 0.03 to 0.1 0.15 to 0.3 Source: Todd Litman, 2004. In most cases the fare change happens because of change in operating cost. Fare changes can also be used to increase or decrease riders hip. For example, to alleviate peak periods or to shift/promote ridership to a less used period, one might implement a higher fare during those peak periods. The concept of tran sit pricing and fare changes is very simple, but the application can get complicated. The r eason is because of so many different types of fares or fare categories available. There are many ways of purchas ing the fare (single, multiple, or unlimited access). Another way of separating fares is in 1) Rider characteristics (student, military, disabled, etc.) 2) Trip characteristics (distance, duration, quality of service, and time period).

PAGE 26

17 With all those options, most if not all tran sit agencies can have around 10 different fare categories for the same trip. A lot of the data on ridership response to fare changes is very old. Some of the newer research fortunately shows the same result as the old data. If you reduce a fare, in order to complete the before-after analysis you would not only need to look at new ridership, but also those existing riders that used to use transit before the fare reduction. Getting this data may not be easy, as you would need to perform surveys, etc. The larger the city size, the sm aller the elasticity. In other words, users are not as sensitive of fare change s in large cities compared to smaller cities, probably due to other options and choices. Table 6 presents elas ticities with respect to some of the transit ridership factors, such as employ ment, population, headways, etc. Table 6 Transit Ridership Factors Factor Elasticity Regional Employment 0.25 Central City Population 0.61 Service (transit vehicle mileage) 0.71 Fare Price -0.32 Wait Time -0.30 Travel Time -0.60 Headways -0.20 Source: JHK, 1995; Kain and Liu, 1999 This table shows the elasticity of transit use with respect to various factors. We see that a 1 percent increase in regional em ployment is likely to incr ease transit ridership by 0.25 percent, while a 1 percent increase in fare prices will reduce ridership by 0.32 percent, all else being equal.

PAGE 27

18 CHAPTER 3 TBEST MODEL TBEST modeling software was briefly introduced in Chapter 1. This chapter will describe the model structure, model tools, methodology, and data requirements. TBEST was already introduced as the third generati on transit planning tool developed by the Florida Department of Transportation that provides users short term transit planning capability, but it is much more than that. Th e research team consists of Ram Pendyala, Xuehao Chu, and Steve Polzin, together w ith Gannet Flaming support for the software development. TBEST forecasts ridership at the stop-level based on socio-demographics and accessibility to transit. Its user friendly ArcGIS interface allows for fast learning and use of the model. Some of the elements of TBEST are presented and described below. 3.1 Elements of TBEST 1) Direct and Transfer Boardings 2) Time of Day Based Analysis 3) Spatial Accessibility (Socio -Economic Characteristics) 4) Time-Space Network Connectivity 5) Competing and Compleme ntary System Effects 6) GIS-Based Software Tool 7) Performance Measures

PAGE 28

19 One of the distinctive featur es of the TBEST Model is the fact that it distinguishes between direct and transfer boardings. Transit passengers ar e either transferring of boarding directly at any gi ven stop. Distinguishing betw een these two is important because it provides users better understand ing of the trip linking that is occurring. Methodology for distinguishing betw een direct and transfer board ings is as follows. First, one should consider two stops, one with transf er opportunity and one without any transfer options. Using the data from the non-transfer stops, TBEST estimates the direct boardings model, then that model is applied to the transfer stops to estimate the boardings at the transfer opportunity st ops. To estimate the transfer boardings, estimated direct boardings are subtracted from the total boa rdings. TBEST includes separate ridership estimation equations for each time of day and day of week. These times of day incorporated in TBEST are shown in Table 7. Different coefficients and equations were developed in order to account for the different ridership le vels in different periods. Table 7 Definitions of Time Periods in TBEST Period No. Period Name Variable Description 1 Weekday morning peak period 6:00 8:59 AM 2 Weekday Off peak period 9:00 AM 2:59 PM 3 Weekday evening peak period 3:00 5:59 PM 4 Weekday night period 6:00 PM 5:59 AM (next day) 5 Saturday 12 midnight 11:59 PM 6 Sunday 12 midnight 11:59 PM Various people characteristics can be attrib uted to different travel patterns. Such characteristics are age, income, auto availabi lity, work status, race, etc. TBEST uses a circular buffer area around each stop to identif y the market that has access to transit.

PAGE 29

20 Figure 2 shows the TBEST Model user interf ace. Appendix A includes screenshots of the models structure with descriptions. Figure 2 TBEST Model User Interface There are a number of database s that TBEST needs to run: 1) 2000 Census data with pre-formatted SF1 and SF3 variables 2) 2000 InfoUSA employment data grouped by commercial, industria l, and service 3) 2000 GDT street networks In addition, transit network and schedule data is stored in Microsoft SQL Server Express 2005. All Florida transit propert ies have already been code d and are ready for use. Main Menu Bar TBEST Tool Bar Status Bar System Options Window Attribute Data Window Mapping Window

PAGE 30

213.2 Methodology Section 3.2 provides the methodology used in TBEST and is based on the paper A Framework of Modeling and Forecasting Stop-Level Transit Patronage produced by Xuehao Chu, Steven E. Polzin, Ram Pendyala, and Ike Ubaka. As it was mentioned in the previous section, one of the key features of TBEST is that it distinguishes the direct boardings and transfer boardings. Therefore, the model stru cture consists of two sets of equations, direct and transfer. Direct boardings equation is as follows: N n X C R g Ds n s n s n s n s n s s n s n,..., 1 , 0 0 0 0 ,5 4 3 2 where s = index for any origin stop. n = index for any time period. N = number of time periods s nD= direct boarding at stop s during period n for the direction and along the route that define stop s. s nR= number of bus runs departing at stop s during period n for the direction and along the route that define stop s. sC= vector of buffer characteristics for stop s. These characteristics include the amount of population and employment as well as their characteristics. s n 20 = vector of accessibility to employment and population in the buffer areas of H2 stops during period n. s n 30 = vector of accessibility to employment and population in the buffer areas of H3 stops during period n. s n 40 = vector of accessibility to employment and population in the buffer areas of H4 stops during period n. s n 50 = vector of accessibility to employment and population in the overlapped buffer areas H5 stops and H2 stops during period n. s nX = vector of other stop and rout e characteristics during period n.

PAGE 31

22 The methodology used here addre sses three important features: 1) The model estimates and forecasts ri dership at the individual stop-level 2) The model separates direct from transit boardings 3) Inter-relationship of the transit netw ork is addressed by using the measure of accessibility to opportunities for potential activity The big advantage new fram ework involving stop-level boa rdings is the ability to capture inter-relationship of the transit network and with that, providing more accurate evaluation of the impact to service changes. The framework proposes the individual stops being defined by spatial location, route associ ation, and travel dire ction. There are two general component of th e transit acc essibility: 1) Access and egress to and from stops 2) Access from one stop to all other stops in the network The standard accessibility measure is us ed where one adds up all the weighted opportunities across all accessible stops. This framework also uses impedance measured by cost of travel. Five measures of accessi bility are used in this methodology. They include the measure of transfer potential from other routes at a subject stop, a measure of accessibility for three sets of accessible stops, an d also a measure of accessibility for the shared buffer areas between stops. One s hould use Appendix A of the TBEST User Guide for more detailed and complete description of the framework and TBEST methodology.

PAGE 32

23 CHAPTER 4 BASE CASE 4.1 Existing Conditions This section will describe the existing conditions of Polk County in terms of demographics and transit services. It is al ways important to understand the base case in order to analyze the alternatives and the resu lts one get from the alternative scenarios. 4.1.1 Demographics Located in central Florida, Polk County population ranks number nine in the state with 541,840 residents in 2005. Since 2000, th ere has been an increase of 57,916 new residents. Figure 3 shows Polk C ounty population from 1950 until 2005. 0 100,000 200,000 300,000 400,000 500,000 600,000 19501960197019801990200020012002200320042005 Estimates 2001-2005 Census Figure 3 Polk County Population

PAGE 33

24 The increase in population has been most lik ely due to affordable housing, and close proximity to the two major metropolitan area s, Tampa and Orlando. By 2030, University of Floridas Bureau of Economic and Busi ness Researchs (BEBR) shows projected growth in Polk County reaching 821,440 people. The two growing areas, Lakeland and Winter Haven are expected to merge and be identified as one in Census 2010. 4.1.2 Transit Services The transit system in Polk County consis ts of three agencies, the Lakeland Area Mass Transit District (LAMTD), Winter Haven Area Transit (WHAT), and the Polk County Transit Services Division (PCTSD). LAMTDs Citrus Connection operates 21 routes and a downtown trolley r oute. WHAT consists of 9 rout es. There are 7 routes that are operated by Citrus Connection and two routes operated by PCTSD. In addition PCTSD operates two rural routes. Those r outes are route number 25 which provides service between Fort Meade and Bartow via Homeland a nd route 35, which operates from Frostproof to Eagle Ridge Mall thr ough Lake Wales and Babson Park. Despite great coordination among these three different tran sit systems, duplicate administrative and operational functions, fragmented service, and complex roles and responsibilities are inevitable. In 2004, the Polk County Regional Transportation Organization (RTO) was created. Some of the RTS goals are to try to implement a strategy for the transition to a regional or countywide transit authority. Consolidating the th ree existing transit service providers into one system will greatly bene fit Polk County. Figure 4 shows the existing Polk County Transit System Organization chart. From this figure we see how complicated the system currently is.

PAGE 34

25 Figure 4 Polk County Transit System Organization Chart 4.2 TBEST Network Development TBEST contains a 2006 pre-coded networ k for all Florida agencies. However, Polk County routes were coded as circulators when they were actually radial routes. In order to get the most accurate results, the correct network was developed. This was done by re-digitizing all the incorrect routes to ra dial types and then adding the stops in the correct locations. 4.2.1 Routes As it was mentioned above, LAMTD operate s 21 routes plus a downtown trolley. WHAT has 9 routes, 7 of which are operate d by LAMTDs Citrus Connection and 2 by PCTS. Table 8, 9, and 10 lists the routs that were coded into TBEST for LAMTD, WHAT, and PCTSD systems respectively. WHAT PB Oversight RTO Board LAMTD Board Administration PCTSD LAMTD Operation PCTSD LAMTD Planning TPO PCTSD LAMTD Polk BoCC WHAT PB Oversight RTO Board LAMTD Board Administration PCTSD LAMTD Operation PCTSD LAMTD Planning TPO PCTSD LAMTD Polk BoCC LAMTD W HAT PCTSD

PAGE 35

26Table 8 Lakeland Area Mass Transit District Route Network Route Name Description Type LAMTD 10 Shuttle CIRCULATOR LAMTD 11 East Main/Combee Road CIRCULATOR LAMTD CONNECTOR 12 Winter Haven to Lakeland via Auburndale RADIAL LAMTD CONNECTOR 12 Lakeland to Winter Haven via Auburndale RADIAL LAMTD 20 Grove Park/Crystal Lake RADIAL LAMTD 21 Edgewood RADIAL LAMTD CONNECTOR 22XL Lakela nd to Bartow RADIAL WHAT CONNECTOR 22XW Bartow to Winter Haven RADIAL LAMTD 30 Cleveland Heights RADIAL LAMTD 31 South Florida Ave. RADIAL LAMTD 32 Medulla Loop/L akeside Village RADIAL LAMTD 37 South Lakeland RADIAL LAMTD 40 Ariana/ Beacon RADIAL LAMTD 41 Central Avenue RADIAL LAMTD 42 West Memorial RADIAL LAMTD 50 Kathleen RADIAL LAMTD 51 Lakeland Mall RADIAL LAMTD 52 North Florida Avenue RADIAL LAMTD 53 Lakeside Village CIRCULATOR LAMTD 56 Kathleen/Mall Hill Rd. RADIAL LAMTD 57 Kidron/Flightline RADIAL LAMTD Citrus Trolley Down town Trolley CIRCULATOR Table 9 Winter Haven Area Transit Route Network Route Name Description Type WHAT 10 Northside RADIAL WHAT 12 Lakeland to Winter Haven via Auburndale RADIAL WHAT 15 Haines City RADIAL WHAT 20 PCC / Hospital RADIAL WHAT CONNECTOR 22XW Bartow Express to Winter Haven RADIAL WHAT 30 Eagle Ridge Mall / Winter Haven RADIAL WHAT 40 Southside RADIAL WHAT 44 Southwest RADIAL WHAT 50 Westside RADIAL Table 10 Polk County Transi t Services Route Network PCTSD Route Name Description Type PCTSD 25 Fort Meade to Bartow RADIAL PCTSD 35 Frostproof to Eagle Ridge Mall RADIAL

PAGE 36

274.2.2 Stops Polk Transportation Planning Organization provided a bus stop inventory in the excel format. The inventory contained detaile d information such as location, stop name, and the X and Y coordinates. From those X and Y coordinates it was possible to import the stops layer through GIS in to TBEST. However, the stops needed to be coded into TBEST. 4.2.3 Network Attributes Network attributes such as arrivals, travel time, service span, fare structure, and growth rates are all variables th at needed to be input. The number of arrivals and service span was input for all time periods as de fined by TBEST using th e route schedules. TBEST calculates travel time by using the follo wing equation. Travel time is calculated from stop to stop. Arrivals definition can be described by the following equation: Table 11 shows the fixed route fares. Fare system/structure is the same for all three systems. Table 11 Fare Structure for LAMTD, WHAT, and PCTSD Systems Rider Type Fares Adult $ 1.00 Children under 5 Free Student (Grades 1-12) $ 0.75 Senior $ 0.50 Disabled $ 0.50 Period Time Defined in Hours of Number Headway 60 Arrivals

PAGE 37

28 Special generators are important in the sense that they can provide simple explanations for certain high or low ridership numbers. Special generators are stops that attract certain demographics. An example of those specia l generators and what TBEST uses are university, shopping mall, event center, park-n-ride, airpor t, and recreational park. TBEST equations treat all of these the same way. A list of special generators was provided by the Polk Transporta tion Planning Organization. This was then added into the TBEST Model. Table 12 Special Generators Special Generator Route Lakeland Regional Medical Center Lakeland Hills Blvd LAMTD 51, 52 Polk Community College/USF/Travis Technical Center US Hwy 98 LAMTD 22XL, 21 Kathleen High School US Hwy 92 (Memorial Blvd) LAMTD 42 Lakeland Government Center LAMTD 10, 11, 52 Florida Metropolitan University LAMTD 52 Florida Southern College LAMTD 30 Polk Community College Winter Haven WHAT 20 Winter Haven Hospital WHAT 20 Gil Jones Government Center WHAT 10, 15 Winter Haven High School WHAT 30

PAGE 38

294.3 TBEST Calibration and Scaling After the network development was complete, the next step was to calibrate the model and save it as base year. TBEST M odel calibration consists of five steps: 1) Entering stop-level observed ridership 2) Defining route collections 3) Entering collection-level observed ridership 4) Saving scenario as base year 5) Viewing collection-leve l calibration factors The first step is to collect the raw data such as ridership for each individual route. In this case, Polk County provided the monthly rout e ridership data by weekday and Saturday for all three transit systems. The data was then organized in the form that is appropriate for the TBEST Model. After the necessary info rmation was input into the model, the next step was to forecast the model using the default equations. Once the model run was complete, it was saved as Base Year and we were able to view our calibration results and scaling factors. TBEST scaling process was designed to automatically fit the calibrated model in order to replicate the actua l ridership data and to adjust for items not captured by the model coefficients. The scaling factors are then applied to all stops along individual routes for all future forecasting scenarios and model applications. There are two types of scaling TBEST allows users to apply. One is called the special generator scaling and the other one route-level scaling. Special generator scaling is applied to unique stops that generate ridership th at would not be reflected by the sociodemographics or service level data in the model. Examples of those unique stops are shown in Table 12. This calibration step of entering the unique st ops into the model is

PAGE 39

30 optional. The results of the route-level scali ng which was performed in this analysis are presented in the next section. 4.3.1 System Results Table 13 shows the weekday observed vers us TBEST estimated ridership by route as well as the scaling factor. Table 13 Weekday Observed Versus TBEST Estimated Ridership by Route Route Observed TBEST Estimated Scaling Factor CC_Route 10 153 38 4.08 CC_Route 11 311 106 2.95 CC_WHAT_Route 12 608 229 2.65 Route 25 (rural) 105 29 3.57 WHAT_Route 30 407 98 4.16 CC_Route 20 348 102 3.42 CC_Route 21 143 65 2.22 CC_Route 22XL 278 101 2.75 CC_Route 30 115 56 2.06 CC_Route 31 709 630 1.13 CC_Route 32 22 14 1.55 CC_Route 37 38 10 3.79 CC_Route 40 103 94 1.1 CC_Route 41 300 79 3.81 CC_Route 42 535 310 1.73 CC_Route 50 268 198 1.36 CC_Route 51 731 649 1.13 CC_Route 52 729 391 1.86 CC_Route 53 113 87 1.3 CC_Route 56 202 82 2.48 CC_Route 57 111 88 1.26 Route 35 (rural) 76 120 0.63 WHAT_Route 10 268 75 3.59 WHAT_Route 15 153 38 4.07 WHAT_Route 20 79 44 1.82 WHAT_Route 22XW 205 85 2.42 WHAT_Route 40 189 132 1.44 WHAT_Route 44 195 77 2.54 WHAT_Route 50 143 75 1.91 Figure 5 is a graphical presentation of the weekday observed versus TBEST Model estimated ridership.

PAGE 40

31 Weekday Observed vs. TBEST Estimated Ridership by Route 0 100 200 300 400 500 600 700 800CC_Rout e 1 0 CC_Rout e 1 1 CC_WHAT_Ro u te 12 R o ut e 25 ( rural) WHA T_Route 3 0 CC_Route 20 CC_Rout e 21 CC _R o ut e 22XL C C R ou t e 3 0 CC_Route 3 1 CC_Route 32 C C_Rout e 3 7 C C_R o ut e 4 0 C C _Rou t e 4 1 C C _Route 4 2 CC_Route 5 0 CC_Rout e 5 1 CC_Rout e 5 2 C C R ou t e 5 3 C C _Rou t e 5 6 CC_Route 57 Route 35 (rura l ) WHAT_Rout e 10 WHAT_Rout e 1 5 WHAT_Rout e 20 WHAT _R o ut e 22XW WHA T_Route 4 0 WHAT_ R ou te 4 4 WHAT_Rout e 5 0Ridership Observed TBEST Estimated Figure 5 Weekday Observed Versus TBEST Estimated Ridership by Route

PAGE 41

32 Table 14 shows the Saturday observed versus TBEST estimated ridership by route, as well as the scaling factor. Table 14 Saturday Observed Versus TBEST Estimated Ridership by Route Route Observed TBEST Estimated Scaling Factor CC_Route 10 107 28 3.77 CC_Route 11 185 66 2.81 CC_WHAT_Route 12 452 253 1.78 Route 25 (rural) 57 27 2.14 WHAT_Route 30 267 89 2.99 CC_Route 20 182 69 2.64 CC_Route 21 86 40 2.17 CC_Route 22XL 146 81 1.81 CC_Route 30 50 24 2.07 CC_Route 31 553 492 1.12 CC_Route 32 19 17 1.11 CC_Route 37 33 8 4.03 CC_Route 40 52 66 0.79 CC_Route 41 205 54 3.78 CC_Route 42 290 243 1.19 CC_Route 50 194 212 0.91 CC_Route 51 636 605 1.05 CC_Route 52 346 309 1.12 CC_Route 53 72 63 1.14 CC_Route 56 169 97 1.74 CC_Route 57 42 64 0.66 Route 35 (rural) 71 140 0.51 WHAT_Route 10 198 57 3.5 WHAT_Route 15 121 37 3.23 WHAT_Route 20 52 28 1.87 WHAT_Route 22XW 111 67 1.66 WHAT_Route 40 185 182 1.02 WHAT_Route 44 144 56 2.57 WHAT_Route 50 98 61 1.61 Figure 6 is a graphical presentation of th e Saturday observed versus TBEST Model estimated ridership.

PAGE 42

33 0 100 200 300 400 500 600 700C C R oute 10 C C _Route 11 CC_WHAT_Route 12 R out e 25 (r ural) WH AT_R out e 30 C C _Route 20 CC_Rout e 21 C C_R o ut e 22X L C C _Route 30 CC_Route 31 CC_Route 32 CC_Rout e 37 C C_R ou t e 40 C C _Route 4 1 CC_Route 42 CC_Rout e 50 C C_R ou t e 51 C C _Route 5 2 CC_Route 53 CC_Route 56 C C_R o ut e 57 Rout e 3 5 (rur al) W HAT_R out e 10 WHAT_Route 15 WHAT _R oute 2 0 WH AT_R oute 22X W W HAT_Rou t e 40 WHAT_R oute 44 WH A T _R oute 50 Observed TBEST Estimated Figure 6 Saturday Observed Versus TBEST Estimated Ridership by Route System results for both weekday and Satu rday observed versus TBEST estimated ridership are shown in Table 15 and Figure 7. From the calibration summary Figure 7, we see that the TBEST Model estimates Saturd ay ridership better than the weekday. Table 15 Calibration Summary Observed TBEST Estimated Scaling Factor Weekday 7,637 4,097 1.86 Saturday 5,123 3,535 1.45

PAGE 43

34 System Results0 1,500 3,000 4,500 6,000 7,500 9,000 WeekdaySaturdayRidership Observed TBEST Estimated Figure 7 Calibration Summary

PAGE 44

35 CHAPTER 5 ALTERNATIVE SC ENARIOS AND MODEL RESULTS 5.1 Alternative Scenarios Four major alternative scen ario categories were create d. Table 16 provides those scenarios their description. Alternative 1 i nvolved various headway changes. Alternative 2 tested different service a llocation changes. Se rvice span analysis involved adding Sunday service and other assessm ent. Alternative number 4, similar to headway, involved various travel time changes and the ridership response analysis to those. Results of the alternative scenarios were compared to the ba se year as it was described in Chapter 4. Table 16 Polk County Alternatives Alternative Description 1 Headway changes 2 Service allocation changes 3 Service span analysis Sunday service addition 4 Travel time changes 5.1.1 Headway Headway elasticities tend to vary significantly depending on different characteristics of routes, current level of serv ice, etc. A series of model runs was set up altering the headway by various amounts for each time period, as shown in Table 17.

PAGE 45

36 Table 17 Alternative 1 Headway Scenarios Headway Base -60% -50% -40% -20% 20% 40% 50% 60% 100% Elasticities were calculated for the system and for all the individua l routes. Appendix A contains some of the model outputs for each scenario create d. Average headway elasticities were calcu lated and compared to different time periods. Table 18 and 19 show the weekday and Saturday system percent cha nge of ridership for each time period for all nine scenarios. Table 18 Weekday Ridership Response to Headway Changes by Time Period AM Peak PM Peak Off Peak Night Headway Ridership % Ridership % Ridership % Ridership % Base 1120 1760 4598 158 -60% 7900 605% 10073 472% 23941 421% 764 384% -50% 4669 317% 6244 255% 16046 249% 624 295% -40% 3163 182% 4437 152% 11557 151% 525 232% -20% 1868 67% 2843 62% 7316 59% 174 10% 20% 642 -43% 1032 -41% 3294 -28% 146 -8% 40% 498 -56% 891 -49% 2447 -47% 115 -27% 50% 496 -56% 889 -49% 2263 -51% 115 -27% 60% 474 -58% 857 -51% 2202 -52% 115 -27% 100% 404 -64% 756 -57% 1370 -70% 27 -83%

PAGE 46

37 Table 19 Saturday Ridership Response to Headway Changes by Time Period Saturday Headway Ridership % Base 5123 -60% 20938 309% -50% 14452 182% -40% 11007 115% -20% 7005 37% 20% 3880 -24% 40% 3138 -39% 50% 2851 -44% 60% 2555 -50% 100% 1987 -61% Figure 8 is a plot of riders hip responses by time period to headway changes as described in Table 17. From this graph we see that th e ridership is not as responsive to headway increases as it is with the headway decreases. This may be attributed to captive riders. Because transit users that have no or limited choice to any alternat ive modes, increasing the headway may not cause much decrease in ri dership. Most of the literature described the elasticities being higher in the off peak periods. Off peak periods are associated with mostly recreational and choice travel. Table 20 presents the calculated elasticities for all time periods and comparison of elasticities for the weekday and Saturday period. Unexpectedly, we see higher elasticities in the peak periods than off peak.

PAGE 47

38 -80% -60% -40% -20% 0% 20% 40% 60% 80% 100% 120% 02,0004,0006,0008,00010,00012,00014,00016,00018,00020,000 RidershipHeadway AM Peak PM Peak OFF Peak Night Saturday Figure 8 Ridership Responses to Headway Changes by Time Period Table 20 System Elasticities by Time Period Headway AM PEAK PM Peak Off Peak Night Weekday Saturday Base -60% -2.13 -1.90 -1.80 -1.72 -1.88 -1.54 -50% -2.06 -1.83 -1.80 -1.98 -1.85 -1.50 -40% -2.03 -1.81 -1.80 -2.34 -1.85 -1.50 -20% -2.29 -2.15 -2.08 -0.42 -2.10 -1.40 20% -3.05 -2.93 -1.83 -0.44 -2.20 -1.52 40% -2.41 -2.02 -1.87 -0.95 -1.96 -1.46 50% -2.01 -1.69 -1.75 -0.79 -1.75 -1.45 60% -1.83 -1.53 -1.57 -0.68 -1.57 -1.48 100% -1.47 -1.22 -1.75 -2.57 -1.58 -1.37 Another comparison was done by the current level of service. Current level of service was defined by three levels high, medium, and low. High level of service was considered to be if the headway is between 30 and 40 minutes, medium if the headway is between 45 and 60 minutes, and low if the he adway is more than 60 minutes. The case scenario, Polk County, had the lowest h eadway of 30 minutes and highest of 120

PAGE 48

39 minutes. The headway is 60 minutes for over 80 percent of the stops. Current level of service by route as defined above is also presented in Appendix B, Table 29. Table below shows the number of stops in each service level category by time period. Table 21 Number of Stops in Each Service Level Category AM Peak Off Peak PM Peak Night Saturday High 405 364 364 100 364 Medium 1609 1393 1650 1658 1464 Low 81 338 81 0 265 Tables 22 and 23 present the elasticities calcul ated and sorted based on the current level of service. Here we see more familiar re sults when comparing to literature. Higher elasticities can be usually obser ved where the initial service is low. However decreasing headway by 50 percent indicates higher elastici ties where the initial service is high. The elasticities are visibly higher where the or iginal service was low when decreasing headway by 50 percent. Table 22 Decreasing Headway Elasticiti es Based on Current Level of Service Decreasing Headway by 50% Original LOS Route AM Peak PM Peak Off Peak Night Saturday High CC_Route 31 -2.35 -1.86 -1.86 -2.11 -1.52 Medium CC_Route 50 -2.00 -1.94 -1.74 -2.04 -1.37 Low CC_Route 32 -1.90 -1.78 -1.81 -1.53 Table 23 Increasing Headway Elasticiti es Based on Current Level of Service Increasing Headway by 50% Original LOS Route AM Peak PM Peak Off Peak Night Saturday High CC_Route 31 -1.89 -1.66 -1.79 -1.65 -1.45 Medium CC_Route 50 -2.45 -1.87 -1.85 -0.03 -1.17 Low CC_Route 32 -3.20 -2.82 -1.89 -1.53

PAGE 49

405.1.2 Service Allocation Service allocation analysis was performed as Alternative 2. Two routes were selected Route CC_Route_31 and the express route 22XL. In one alternative, Route_31 was copied and named Route_31A. On ly half the service of the Route_31 was placed on the Route_31A. Another alternat ive was created where the Route_31 was copied with the exact same amount of serv ice. The results were then compared to increasing the service frequency by 20, 40, 50 and 100 percent. The same procedure was done with the express Route_22XL. Table 24 displays these results. Table 24 Service Allocation and Frequency Analysis AM Peak Off Peak Ridership % Ridership % Base 1120 4598 20 % Frequency Increase 1733 55% 6092 32% 40 % Frequency Increase 2051 83% 8049 75% 50 % Frequency Increase 2250 101% 9481 106% 100% Frequency Increase 4670 317% 16046 249% Route 31A added with half service of Route 31 1234 10% 5105 11% Route 31A added with full service of Route 31 1442 29% 6040 31% Route 22XL_A added with half service of Route 22XL 1143 2% 4615 0.4% Route 22XL_A added with full service of Route 22XL 1199 7% 4674 2% Table 24 Continued PM Peak Night Ridership % Ridership % Base 1760 158 20 % Frequency Increase 2674 52% 174 10% 40 % Frequency Increase 2953 68% 211 33% 50 % Frequency Increase 3154 79% 521 230% 100% Frequency Increase 6244 255% 624 295% Route 31A added with half service of Route 31 1927 9% 215 36% Route 31A added with full service of Route 31 2247 28% 256 62% Route 22XL_A added with half service of Route 22XL 1808 3% 163 3% Route 22XL_A added with full service of Route 22XL 1890 7% 163 3%

PAGE 50

41 One would expect the same results when co mparing the 50 percent frequency increase with the increase of the additional route that is added. However, we see very different results. In the AM Peak period, 50 percen t frequency increase produced over 100 percent increase in ridership. However, the rout e addition produced only 10 percent increase for the Route 31 and an even smaller increase of 2 percent for the express route 22XL. Figure 9 displays those results. 0 1,000 2,000 3,000 4,000 5,000 6,000 7,000 8,000 AM PeakOFF PeakPM PeakNightSaturdayRidership Increase oo 50 % Frequency Increase Route 31A added with half service of Route 31 100% Frequency Increase Route 31A added with full service of Route 31 Figure 9 Service Allocati on Analysis Route 31 5.1.3 Span of Service Polk County currently operates from M onday through Saturday. For the span of service analysis, Sunday service was introdu ced. Sunday service scenario that was added replicated the Saturday servic e in terms of arrivals and speed. After the model run, the results produced an additional 3,297 riders. A nother scenario was crea ted, increasing the

PAGE 51

42 service span and arrivals for the night period. The number of arrivals at each stop was increased to 3. The speed for the night period was input the same as the off peak period. The headway was set to 40 minutes. Table 25 Night Service Span Increase AM Peak Off Peak PM Peak NIGHT Base 1120 4598 1760 158 NIGHT Service Increase 559 2535 910 472 % -50% -45% -48% 199% These changes produced results shown in Table 25. The night period produced an increase of almost 200 percent while during all other periods the ridership decreased by nearly 50 percent. Further analysis of appl ying the same amount of service increase to other time periods and then comparing the results would be beneficial. 5.1.4 Travel Time One of the most important factors affecti ng public transit ridership is the travel time. Eight scenarios were created to test th e sensitivity of travel time. This involved increasing and decreasing ridership by 10, 20, 30, and 50 percent for all time periods. Tables 26 and 27 show the ridership respons e to travel time changes for weekday and Saturday period. Table 26 Weekday Ridership Response to Travel Time Changes by Time Period Travel Time AM Peak Off Peak PM Peak Night Base 1120 4598 1760 158 -50% 1184 5.7% 4692 2.1% 1869 6.2% 169 7.0% -30% 1162 3.7% 4654 1.2% 1814 3.1% 161 2.1% -20% 1142 2.0% 4639 0.9% 1796 2.0% 161 1.8% -10% 1130 0.9% 4618 0.4% 1778 1.0% 160 1.1% 10% 1110 -0.9% 4580 -0.4% 1742 -1.0% 158 0.0% 20% 1101 -1.7% 4560 -0.8% 1724 -2.0% 157 -0.7% 30% 1097 -2.0% 4547 -1.1% 1710 -2.9% 156 -1.1% 50% 1072 -4.3% 4509 -1.9% 1680 -4.6% 153 -3.2%

PAGE 52

43Table 27 Saturday Ridership Response to Travel Time Changes by Time Period Travel Time Saturday Weekday Base 5123 7637 -50% 5228 2.0% 7914 3.6% -30% 5180 1.1% 7791 2.0% -20% 5162 0.8% 7737 1.3% -10% 5144 0.4% 7686 0.6% 10% 5104 -0.4% 7590 -0.6% 20% 5086 -0.7% 7542 -1.2% 30% 5070 -1.0% 7510 -1.7% 50% 5039 -1.6% 7412 -2.9% -60% -40% -20% 0% 20% 40% 60% 0100020003000400050006000 RidershipTravel Time Chang e AM Peak OFF Peak PM Peak Night Saturday Figure 10 Ridership Response to Trave l Time Changes by Time Period TBEST Model results for travel time varia tions show almost no sensitivity at all. Elasticities were calculated using the approxi mated point elasticity formula. Table 28 shows the results and we can see that the elas ticity is very low. The travel time analyzed in this section involved usi ng only in-vehicle time. Howe ver total trip times includes access, wait, in-vehicle, transfer, and egress time. One can expect that headway and outof-vehicle time elasticities to be similar. Literatu re suggests that evidence on the out-ofvehicle time elasticities (walk, wait, and transfer) as it relates to in-vehicle time is

PAGE 53

44 inconsistent. Quarmby (1967) estimated out-of-ve hicle time elasticities to be two to three times higher than in-vehicle time elasticities. Table 28 Travel Time Elasticities Travel Time AM Peak Off Peak PM Peak Night Saturday Weekday -50% -0.11 -0.04-0.12-0.14-0.04 -0.07 -30% -0.12 -0.04-0.10-0.07-0.04 -0.07 -20% -0.10 -0.04-0.10-0.09-0.04 -0.07 -10% -0.09 -0.04-0.10-0.11-0.04 -0.06 10% -0.09 -0.04-0.100.00-0.04 -0.06 20% -0.09 -0.04-0.10-0.03-0.04 -0.06 30% -0.07 -0.04-0.10-0.04-0.03 -0.06 50% -0.09 -0.04-0.09-0.06-0.03 -0.06

PAGE 54

45 CHAPTER 6 CONCLUSIONS AND FURTHER RESEARCH 6.1 Conclusions Being able to quantify the ridership res ponse to different service changes is an important area in transportation planning. Weather budget or service planning related, transit agencies will always look for better ways to analyze their network and transit system. This paper introduced new planning so ftware that is capable of capturing those ridership responses, TBEST. With TBEST, we are a step closer to better and easier analysis of the ridership response, riders hip forecasting, and network modeling. TBEST was used to produce general sensitivity of the model and to compare the results to transit elasticities shown in the literature. Sensitivity analysis was performed in four different areas. Headway, travel time, service span, and service allocation were analyzed. Some of the transit elasticities were described in the literature review and can be summarized as following. 1) Average service elasticity is around -0.5 2) Ridership response to serv ice changes is inelastic 3) Off peak period ridership is more responsive than peak period 4) Higher elasticities tend to occu r in the low service areas 5) Ridership is more responsive in head way improvements than travel time improvements

PAGE 55

46 Based on these few observations, there are numbe r of conclusions that can be drawn from the results. First alternative involved headway analysis From the results obtained, we see that TBEST overestimates ridership response. Th e headway elasticities calculated do not compare with the literature. Unreasonably high ridership increase occurs when the headway is decreased by even only 40 percent. In other words, decreasing headway from 60 minutes to 36 minutes resulted in 182 perc ent ridership increase in the AM peak period and over 200 percent ridership increase in the night period. Literature also suggests that the ridership is more sensitive to off peak periods then peak periods. The elasticities calculated by time period contradict that phe nomenon. However, many factors such as the socio-demographics, land use, and accessibility can influence these results. Further analysis of the headway elasticities revealed that increasing the headway does produce higher ridership response in the low service areas as show n in the literature. However, we see the opposite effect when decreasing the headway. This may be attributed to high area with captive riders. Service span analysis involved adding S unday service. This produced additional ridership of over 3,000 passengers. Additional analysis was performed by increasing the span if service the night period. An increase of almost 200 percent riders was observed. Literature does report that there are hi gh responses to extende d evening service, especially after the PM peak period. This may be due to the assuranc e to riders in case they are stranded at work due to a late meetin g, that the transit service is available to them in those situations. To better understa nd the model sensitivity to service span,

PAGE 56

47 further analysis of applying the same amount of service increase to other time periods and then comparing it to the results obtaine d for the night period is necessary. Service allocation analysis involved incr easing the frequency and adding a route with half the service of the route that it is mirroring. Same or simila r response is expected as we are adding the same amount of servi ce different way. TBEST however, produced unreasonable results of ridership being far more responsive to frequency improvements than route additions. Frequenc y increase of 50 percent produ ced ridership increase of over 100 percent whereas the route addition pr oduced only 10 percen t increase for route 31 and only 2 percent for the express route XL. Travel time sensitivity analysis was also performed. The results obtained show very little ridership respons e to travel time. Decreasing the travel time by 50 percent produced an increase of 7.0 percent in the night period and 5.7 percent during the AM peak period. Measuring ridership response to to tal travel time as well as in-vehicle travel time is difficult. Therefore the literature a nd evidence on these is modest. The literature that does exist on in vehicle travel time ri dership response, however, shows an aggregate value of -0.35. TBEST results do not compare to this, as the elasticities obtained are much lower. 6.2 Further Research TBEST is capable of providi ng ridership response to di fferent service changes. Based on the sensitivity results provided in this paper, TBEST overestimates and shows unreasonable estimates for the ridership response to headway changes. The elasticities are significantly higher than what th e literature suggests. Travel time has shown almost no

PAGE 57

48 sensitivity, which contradicts the literatu re. All the results, however, do provide a predictable pattern in terms of increase or decrease of ridershi p to certain service changes. Much of the analysis was time consuming. In addition to lengthy model run times, the ridership output reports available to dow nload are individual reports by time period. Having six time periods and multiple scen arios can lead to time consuming and cumbersome data collection and formatting. A possible software improvement suggestion would be to add a similar to batch model func tion. A batch reports function where a user could select the time periods to download in one spreadsheet would be beneficial for data analysis purposes. Common us e function in many computer applications is the undo function. TBEST does not have this, theref ore saving often is critical. Also high performance computers are recommended, especi ally for larger transit systems. Transit systems similar in size, socio-demographics and network to Polk County can expect same or similar results. The objective of this paper was to provi de users a general idea of the model sensitivities and to help further improvement of the model. This was achieved by the sensitivity analysis performed on some of the variables. In order to further improve the model, additional testing and sensitivity an alysis is needed. Because of unreasonable results especially in the frequency area, the equations and coefficien ts of the model need to be altered. Model methodology and dire ct boarding equation were introduced in Chapter 3. However, for satisfactory result s one needs full understanding of the model equations and coefficients in order to alter them. A new series of the model sensitivity analysis would then need to be performed.

PAGE 58

49REFERENCES Allen, J. B. (1991). Revenue and Ridership Impacts of D ART Service and Fare Adjustments. Chu, X., Polzin S., Pendyala, R., Ubaka I. (2006). A Framework Modeling and Forecasting Stop Level Transit Patronage. Dupree, J. and Pratt R. (1973). A Study of Low Cost Alternatives to Increase the Effectiveness of Existing Transportation Faciliti es Results of Case Studies and Analysis of Busway Applications. Ecosometrics, Inc. (1980). Patronage Impacts of Changes in Transit Fares and Services. Goodman, K. M., Green, M. A., and Beesley, M. E., (1977). The San Diego Transit Corporation: The Impacts of Fare an d Service Changes on Ridership and Deficits. Holland, D. K. (1974). A Review of Reports Relating to the Effects of Fare and Service Changes in Metropolitan Pub lic Transportation Systems. Horowitz AJ (1984). Simplifications for Single-Route Transit Ridership Forecasting Models. Transportation 12: 261-275. Kemp, M. A. (1981). A Simultaneous Equations Analys is of Route Demand and Supply, and Its Application to the San Diego Bus System. Litman, T (2004). Transit Price Elasticities and Cross-Elasticities. Litman, T (2007). Transportation Elasticities How Pr ices and Other Factors Affect Travel Behavior. Multisystems, Inc. (1982). Route-Level Demand Models: A Review. DOT-1-82-6, Urban Mass Transportation Administration, US Depa rtment of Transpor tation, Washington, D.C. Peng, Dueker, Strathman, Hopper (1997). A simultaneous route-le vel transit patronage model: demand, supply, and inter-route relationship.

PAGE 59

50 Quarmby, D. (1967). Travel Mode for the Journey to Work. Voorhees, A. and Associates (1973). Blue Streak Bus Rapi d Transit Demonstration Project. Wattleworth, J (1978). Evaluation of the Effects of the I-95 Miami Exclusive Bus/Carpool Lane Priority System on Vehi cular and Passenger Movement.

PAGE 60

51APPENDICES

PAGE 61

52APPENDIX A TBEST Model Appendix A contains the screenshots from TBEST showing many opportunities of the model. Figure 11 Different Types of Reports Route Report Stop Report Regional Analysis Report Segment Report

PAGE 62

53APPENDIX A (Continued) Figure 12 Mapped Ridership Output Sample Figure 13 Model Output Summary Reports Sample

PAGE 63

54APPENDIX A (Continued) Figure 14 Alternative Scenario List

PAGE 64

55APPENDIX B TBEST Model Output Results Appendix B presents some of the results ob tained from the model output. Table 29 shows the current level of service for Polk County, which was used in Chapter 5. Table 29 Current Level of Service by Route AM Peak Off Peak PM Peak Night Saturday CC_Route 10 60 60 60 60 60 CC_Route 11 60 60 60 60 60 CC_WHAT_Route 12 60 60 60 60 60 Route 25 (rural) 60 60 60 0 60 WHAT_Route 30 60 60 60 60 60 CC_Route 20 60 60 60 60 60 CC_Route 21 60 90 60 0 100 CC_Route 22XL 45 90 45 60 60 CC_Route 30 60 120 60 0 96 CC_Route 31 30 30 30 40 31 CC_Route 32 90 120 90 0 120 CC_Route 37 60 60 60 0 60 CC_Route 40 60 60 60 0 60 CC_Route 41 60 60 60 60 60 CC_Route 42 30 30 30 60 31 CC_Route 50 60 60 60 60 60 CC_Route 51 30 30 30 60 30 CC_Route 52 30 30 30 60 31 CC_Route 53 60 60 60 60 60 CC_Route 56 60 60 60 60 60 CC_Route 57 60 60 60 60 60 Route 35 (rural) 60 60 60 0 60 WHAT_Route 10 60 60 60 60 60 WHAT_Route 15 60 60 60 60 60 WHAT_Route 20 60 60 60 60 60 WHAT_Route 22XW 60 60 60 60 60 WHAT_Route 40 60 60 60 60 60 WHAT_Route 44 60 60 60 60 60 WHAT_Route 50 60 60 60 60 60

PAGE 65

56Table 30 AM Peak Ridership Re sponse to Headway by Route Route Base Alternatives 1 0.2 0.4 0.5 0.6 1.2 1.4 1.5 1.6 2 CC_Route 10 12 12 26 45 67 12 3 3 3 3 CC_Route 11 50 85 148 209 357 23 22 22 22 21 CC_WHAT_Route 12 85 145 238 351 630 41 39 39 38 38 Route 25 (rural) 18 31 47 66 114 8 8 8 8 8 WHAT_Route 30 68 115 177 250 431 30 30 30 30 30 CC_Route 20 61 105 186 270 463 29 28 28 27 27 CC_Route 21 19 19 42 72 109 19 5 5 5 5 CC_Route 22XL 82 124 244 327 505 47 39 38 19 19 CC_Route 30 33 60 101 150 270 15 15 15 15 14 CC_Route 31 89 150 278 453 753 63 41 41 41 25 CC_Route 32 4 5 10 16 25 4 1 1 1 1 CC_Route 37 7 8 16 27 40 7 2 2 2 2 CC_Route 40 18 31 49 71 120 8 8 8 8 8 CC_Route 41 46 81 133 196 344 22 22 22 21 20 CC_Route 42 86 147 237 354 585 57 38 38 38 22 CC_Route 50 34 60 96 135 230 13 13 13 13 12 CC_Route 51 74 130 220 341 596 50 33 33 33 20 CC_Route 52 111 189 303 454 753 77 52 52 51 31 CC_Route 53 2 2 8 8 8 2 2 2 2 0 CC_Route 56 28 49 76 106 180 13 13 12 12 12 CC_Route 57 15 28 53 78 135 6 6 6 6 6 Route 35 (rural) 3 3 6 11 16 3 1 1 1 1 WHAT_Route 10 36 53 92 139 237 23 15 15 15 15 WHAT_Route 15 26 44 67 94 161 12 12 12 12 12 WHAT_Route 20 6 7 15 25 38 6 2 2 2 2 WHAT_Route 22XW 29 51 79 114 198 13 13 13 12 12 WHAT_Route 40 23 40 62 88 153 10 10 10 10 10 WHAT_Route 44 33 58 91 129 224 16 16 16 16 16 WHAT_Route 50 23 40 64 92 159 11 11 11 11 11 Total System 1120 1868 3163 4670 7900 642 498 496 474 404

PAGE 66

57Table 31 Off Peak Ridership Re sponse to Headway by Route Route Base Alternatives 1 0.2 0.4 0.5 0.6 1.2 1.4 1.5 1.6 2 CC_Route 10 87 145 216 297 438 63 42 42 42 25 CC_Route 11 171 285 425 588 875 124 83 83 83 50 CC_WHAT_Route 12 376 627 947 1321 1979 271 181 181 178 107 Route 25 (rural) 66 110 164 227 337 48 32 32 32 20 WHAT_Route 30 238 395 595 828 1238 173 115 115 115 69 CC_Route 20 204 339 511 717 1101 146 99 98 92 57 CC_Route 21 68 102 186 236 353 41 41 41 20 20 CC_Route 22XL 90 134 246 313 467 54 53 53 26 26 CC_Route 30 35 59 91 129 216 17 17 17 17 16 CC_Route 31 459 688 1184 1667 2511 330 271 222 222 133 CC_Route 32 11 19 28 39 65 6 5 5 5 5 CC_Route 37 16 27 39 54 90 8 8 8 8 8 CC_Route 40 63 106 158 218 320 45 30 30 30 18 CC_Route 41 177 297 446 620 923 127 86 85 85 51 CC_Route 42 322 485 836 1190 1816 233 188 155 155 93 CC_Route 50 175 294 433 583 835 125 83 83 83 49 CC_Route 51 495 714 1177 1624 2337 369 310 261 260 169 CC_Route 52 463 693 1170 1641 2463 334 274 224 223 135 CC_Route 53 80 134 199 272 396 58 39 38 38 23 CC_Route 56 123 207 311 428 624 88 59 59 58 35 CC_Route 57 70 117 175 243 362 50 34 33 33 20 Route 35 (rural) 49 70 91 115 153 40 31 31 31 22 WHAT_Route 10 163 278 421 593 892 115 75 75 75 45 WHAT_Route 15 87 145 216 299 446 63 43 43 43 26 WHAT_Route 20 57 94 146 204 305 41 28 28 28 17 WHAT_Route 22XW 132 219 332 464 697 96 65 65 63 38 WHAT_Route 40 123 207 311 431 642 89 60 60 60 35 WHAT_Route 44 112 186 282 395 591 81 55 55 55 31 WHAT_Route 50 86 142 221 311 468 62 42 42 42 25 Total System 4598 7316 11557 16046 23941 3295 2448 2263 2202 1370

PAGE 67

58Table 32 AM Peak Ridership Resp onse to Travel Time by Route Route Base -10% -20% -30% -50% 10% 20% 30% 50% CC_Route 10 12 12 12 12 12 12 12 12 12 CC_Route 11 50 51 51 53 55 48 47 47 45 CC_WHAT_Route 12 85 86 87 89 91 84 83 83 81 Route 25 (rural) 18 18 19 19 19 18 18 18 18 WHAT_Route 30 68 68 69 69 70 67 67 67 66 CC_Route 20 61 61 62 63 65 61 60 60 59 CC_Route 21 19 19 19 19 19 19 19 19 18 CC_Route 22XL 82 83 84 86 87 82 81 81 75 CC_Route 30 33 33 36 36 37 33 32 32 32 CC_Route 31 89 89 90 92 93 89 88 89 86 CC_Route 32 4 4 4 5 5 4 4 4 4 CC_Route 37 7 7 7 8 8 7 7 7 7 CC_Route 40 18 18 18 18 18 17 17 17 17 CC_Route 41 46 47 47 48 49 46 45 45 44 CC_Route 42 86 86 87 87 90 85 85 84 82 CC_Route 50 34 34 35 36 37 33 33 33 32 CC_Route 51 74 74 76 77 79 73 72 71 70 CC_Route 52 111 111 112 114 116 110 108 108 105 CC_Route 53 2 2 2 2 2 2 2 2 2 CC_Route 56 28 28 28 29 29 28 27 27 27 CC_Route 57 15 15 15 16 17 15 15 15 15 Route 35 (rural) 3 3 3 3 3 3 3 3 3 WHAT_Route 10 36 36 36 36 37 36 36 36 36 WHAT_Route 15 26 26 26 26 27 26 26 26 25 WHAT_Route 20 6 6 7 7 7 6 6 6 6 WHAT_Route 22XW 29 30 30 31 31 29 28 28 28 WHAT_Route 40 23 23 23 24 24 23 22 23 22 WHAT_Route 44 33 34 34 35 35 33 33 33 32 WHAT_Route 50 23 24 24 24 24 23 23 23 23 System Total 1120 1130 1142 1162 1184 1110 1101 1097 1072

PAGE 68

59Table 33 Off Peak Ridership Resp onse to Travel Time by Route Route Base -10% -20% -30% -50% 10% 20% 30% 50% CC_Route 10 87 86 86 86 84 87 87 87 87 CC_Route 11 171 172 173 173 177 170 168 168 166 CC_WHAT_Route 12 376 377 379 381 384 374 373 371 368 Route 25 (rural) 66 67 67 67 67 66 66 66 65 WHAT_Route 30 238 238 239 240 242 236 236 235 231 CC_Route 20 204 204 205 205 205 204 204 204 203 CC_Route 21 68 68 69 69 69 68 68 68 68 CC_Route 22XL 90 90 90 90 90 90 90 90 89 CC_Route 30 35 35 35 36 36 35 35 35 34 CC_Route 31 459 459 462 461 463 458 456 455 452 CC_Route 32 11 11 11 11 12 11 11 11 11 CC_Route 37 16 16 16 16 16 16 16 16 16 CC_Route 40 63 63 63 63 63 63 63 63 63 CC_Route 41 177 178 179 179 180 176 175 175 173 CC_Route 42 322 327 328 330 337 320 317 317 312 CC_Route 50 175 176 177 178 181 174 173 172 171 CC_Route 51 495 499 502 505 512 493 489 488 483 CC_Route 52 463 466 467 469 472 461 458 457 452 CC_Route 53 80 81 81 82 82 80 79 79 78 CC_Route 56 123 124 124 125 127 123 122 121 120 CC_Route 57 70 70 70 70 70 70 69 69 69 Route 35 (rural) 49 50 50 50 50 49 49 49 49 WHAT_Route 10 163 164 169 169 169 163 163 162 161 WHAT_Route 15 87 88 88 88 89 87 87 86 86 WHAT_Route 20 57 57 57 57 57 57 57 57 56 WHAT_Route 22XW 132 132 133 133 134 132 131 131 129 WHAT_Route 40 123 123 123 124 124 122 122 122 121 WHAT_Route 44 112 112 113 113 114 112 111 111 110 WHAT_Route 50 86 86 86 86 87 86 85 84 84 System Total 4598 4618 4639 4654 4692 4580 4560 4547 4509


xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam Ka
controlfield tag 001 001966688
003 fts
005 20081023144116.0
006 m||||e|||d||||||||
007 cr mnu|||uuuuu
008 081023s2007 flu sbm 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0002247
035
(OCoLC)263172301
040
FHM
c FHM
049
FHMM
090
TA145 (ONLINE)
1 100
Vuckovic, Dajana.
0 245
Sensitivity Analysis of the Transit Boardings Estimation and Simulation Tool (TBEST) Model
h [electronic resource] /
by Dajana Vuckovic.
260
[Tampa, Fla.] :
b University of South Florida,
2007.
3 520
ABSTRACT: Public transportation, although modest in the United States carrying about 2 percent trips, still serves millions of people as the main and only means of transportation. Recently released data set by Census, the 2006 American Community Survey (ACS) shows the main mode of travel for work commute is not surprisingly the automobile with over 86 percent and public transportation with nearly 5 percent users. Transit agencies strive to provide effective, convenient, and desirable transport. Because of the constant changes in our environment, being able to predict the response of riders to different network or system changes is extremely useful. Ridership can be described as a function of the amount of service supplied such as frequency, span of service, and travel time. One of the methods for estimating ridership forecasts and evaluating ridership response is to use the new state-of-art software TBEST.^ TBEST stands for Transit Boardings Estimation and Simulation Tool and is the third generation of such transit models sponsored by the Florida Department of Transportation (FDOT). Designed for comprehensive transit network and short term transit planning, it offers great benefits to its users. TBEST is a user friendly, yet very advanced transit ridership forecasting graphical software which is interfaced with ArcGIS. This paper evaluates different sensitivity tests and compares the results to known industry used elasticities. Because the current TBEST experience is modest, the results will provide users with a general idea of the model's sensitivity and help in the process of model refinements. Sensitivity tests such as service frequency, span of service, service allocation, and travel time will be carried out in a systematic order for all six time periods as defined by TBEST.^ Results showed that TBEST Model is overestimating and is highly sensitive to headway changes, specifically headway decrease. The opposite effect of almost no sensitivity is shown for the in-vehicle travel times.
502
Thesis (M.S.C.E.)--University of South Florida, 2007.
504
Includes bibliographical references.
516
Text (Electronic thesis) in PDF format.
538
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
500
Title from PDF of title page.
Document formatted into pages; contains 59 pages.
590
Co-adviser: Steven E. Polzin, Ph.D.
Co-adviser: Jian Lu, Ph.D.
653
Service planning.
Elasticities.
Headway.
Span of service.
Routes.
690
Dissertations, Academic
z USF
x Civil Engineering
Masters.
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.2247