Full Issue 15(4)

Journal of Public Transportation, Dec 2012

Published on 12/01/12

A PDF file should load here. If you do not see its contents the file may be temporarily unavailable at the journal website or you do not have a PDF plug-in installed and enabled in your browser.

Alternatively, you can download the file locally and open with any standalone PDF reader:

https://scholarcommons.usf.edu/cgi/viewcontent.cgi?article=1084&context=jpt

Full Issue 15(4)

All articles should be approximately 1077-291X Christopher M. Hewitt W. E. (Ted) Hewitt 0 1 2 3 4 5 6 0 e Effect of Proximity to Urban Rail on Housing Prices in Ottawa 1 Planning Public Transport Networks- e Neglected Influence of Topography 2 Central Business Districts and Transit Ridership: A Reexamination of the Relationship in the United States 3 Simulated Analysis of Exclusive Bus Lanes on Expressways: Case Study in Beijing , China 4 Transit Operator Evaluation of ree Wheelchair Securement Systems in a Large Accessible Transit Vehicle 5 Cost Estimation of Fare-Free ADA Complementary Paratransit Service in Illinois , USA 6 Lisa Ravenscroft, Assistant to the Editor Center for Urban Transportation Research (CUTR) University of South Florida Fax: (813 , USA Jeffrey R; Brown Dristi Neog - 2 0 1 2 Linda van Roosmalen Douglas Hobson Patricia Karg Emily DeLeo Erik Porach Lin Zhu Lei Yu Xu-Mei Chen Ji-Fu Guo Steven E. Polzin, Ph.D., P.E. University of South Florida Lawrence Schulman LS Associates George Smerk, D.B.A. Indiana University Vukan R. Vuchic, Ph.D., P.E. University of Pennsylvania The contents of this document reflect the views of the authors, who are responsible for the facts and the accuracy of the information presented herein. This document is disseminated under the sponsorship of the U.S. Department of Transportation, University Research Institute Program, in the interest of information exchange. The U.S. Government assumes no liability for the contents or use thereof. Be sure to include the author’s complete contact information, including email address, mailing address, telephone, and fax number. Submit manuscripts to the Assistant to the Editor, as indicated above. Steven E. Polzin, Ph.D., P.E. University of South Florida Lawrence Schulman LS Associates George Smerk, D.B.A. Indiana University Vukan R. Vuchic, Ph.D., P.E. University of Pennsylvania The contents of this document reflect the views of the authors, who are responsible for the facts and the accuracy of the information presented herein. This document is disseminated under the sponsorship of the U.S. Department of Transportation, University Research Institute Program, in the interest of information exchange. The U.S. Government assumes no liability for the contents or use thereof. Be sure to include the author’s complete contact information, including email address, mailing address, telephone, and fax number. Submit manuscripts to the Assistant to the Editor, as indicated above. Journal fo Public Transportation The Journal of Public Transportation is published quarterly by National Center for Transit Research Center for Urban Transportation Research University of South Florida • College of Engineering 4202 East Fowler Avenue, CUT100 Tampa, Florida 33620-5375 Phone: (813) 974-3120 Fax: (813) 974-5168 Email: Website: www.nctr.usf.edu/jpt/journal.htm © 2012 Center for Urban Transportation Research CONTENTS Central Business Districts and Transit Ridership: A Reexamination of the Relationship in the United States Jeffrey R. Brown, Florida State University Dristi Neog, Sushant School of Art and Architecture Abstract Many scholars claim that public transit’s long-term ridership decline can be attributed to the decentralization of U.S. metropolitan areas and the decline of the central business district (CBD) as their primary economic engine. However, recent research has begun to challenge this view and has prompted this reexamination. Using multivariate analysis, we examine the relationship between the strength of the CBD and transit ridership in all U.S. metropolitan areas with more than 500,000 persons in 2000, while controlling for other factors thought to influence bus and rail transit ridership. We find no relationship between the strength of the CBD and transit ridership, which suggests that other factors are much more important contributors to transit ridership. Introduction Most scholars argue that public transit’s long-term ridership decline is associated with the decentralization of U.S. metropolitan areas and the decline of the central business district (CBD) as their primary economic engine. Recent research suggests that this relationship remains strong, although some scholars have begun to challenge this view by noting circumstances where transit agencies are increasing ridership in decentralized urban areas. These recent research developments have prompted us to reexamine the relationship between the strength of the CBD and transit ridership (measured as transit journey-to-work mode share by bus and/or rail transit modes), while controlling for other factors thought to influence ridership. The Relationship between Transit Ridership and the CBD Transit ridership is one of the most frequently studied phenomena in transportation, and a large literature has emerged that seeks to explain it. The literature divides explanations for ridership (and ridership change) into two broad categories: external factors and internal factors. External factors include urban structure, population change, regional economic conditions, household auto ownership levels, and urban population density, all factors over which transit managers have no control. Internal factors include fare and service policies over which transit managers exercise some control. Traditional View Our particular interest in this study is the role of urban structure in explaining variation in transit ridership, and there is an extensive literature on this topic. Most of the literature focuses on the relationship between transit ridership and the relative strength of the CBD as a locus of regional economic activity. Scholars writing in this topic area tend to view the CBD and the CBD-bound commuter as the most important market for public transit (Pucher and Renne 2003; Pushkarev and Zupan 1977; Pushkarev and Zupan 1980). Mierzejewski and Ball (1990) found support for this view in their survey of transit users, which found that 82 percent of choice riders worked in the CBD of their metropolitan area. Studies of the post-war decline in U.S. transit use frequently cite the decline of the CBD and the decentralization of population and employment as major causal factors (Ferreri 1992; Jones 1985; Meyer, Kain, and Wohl 1965; Meyer and GómezIbáñez 1981). A number of scholars have used statistical analysis to examine this relationship, when controlling for the influence of other variables. Most of these authors have found strong connections between the strength of the CBD (or its corollary, the degree of decentralization) and transit ridership. Hendrickson’s work (1986) is one example of these studies. He examined the relationship between transit ridership and both the size and strength of the CBD and total population for 25 U.S. metropolitan areas in 1970 and 1980. He found strong, statistically-significant associations between the strength of the CBD and his transit ridership measures. However, his multivariate models failed to control for other important variables, such as fares, service quality, regional economic conditions, and auto ownership, which might also affect transit ridership. He also included New York, an outlier that accounts for 40 percent of all U.S. transit patronage, in his models, which undoubtedly influenced his results. Both Gómez-Ibáñez (1996) and Kain (1997) performed time-series multivariate analysis to examine the relationship between urban structure and transit ridership in individual metropolitan areas. Gómez-Ibáñez (1996) examined ridership change between 1970 and 1990 in Boston. He estimated multivariate models that examined ridership as a function of the number of jobs in Boston (his urban structure variable), per-capita income, fare, service miles, and a dummy variable for 1980–1981, a period during which transit service was significantly reduced. He found that a 1 percent decline in the percent of jobs in the city of Boston was associated with between a 1.24 percent and 1.75 percent decline in ridership, when controlling for the influence of these other variables. However, his definition of employment is problematic and measures jobs located throughout the city of Boston as opposed to jobs inside the CBDs of Boston and Cambridge, which he had originally hoped to measure. Kain (1997) examined ridership change between 1972 and 1993 in Atlanta. He employed a secular trend variable that functions as an indirect measure of urban decentralization and found that average fares, service levels, total metropolitan employment, and the trend variable were the explanatory variables with the strongest influence on transit ridership. Work by Beesley and Kemp (1987), Heilbrun (1987), Pisarski (1996), and Taylor (1991) provides additional scholarly support for the notion that transit ridership is strongly linked to the strength of the CBD and the degree of urban decentralization. More Nuanced Views However, more recent studies describe a more nuanced relationship between urban structure and transit ridership. In a nine-city case study, Thompson and Matoff (2003) found that transit agencies that altered their service to better serve the dispersed destination patterns that characterized their metropolitan areas increased their ridership. Brown and Thompson (2008a) found similar results in a national study of transit service productivity in 2000. They estimated models predicting service productivity (the ratio of ridership to service) as a function of the strength of the CBD, service orientation, service coverage, fares, fuel prices, auto ownership, regional unemployment rate, West region (a dummy variable), ratio of rail service to total service, and ratio of peak service to off-peak service. They found no relationship between the strength of the CBD and transit productivity when these other factors were included. However, productivity—not ridership—was the focus of their study. Ridership is the focus of recent work by Brown and Thompson (2008b) in Atlanta. In a study that updates Kain’s earlier analysis, they estimate a time-series model that predicts ridership (measured as passenger miles per capita) as a function of service, fare, motor fuel price, a dummy variable for the 1996 Olympics, and three urban structure variables (percent of MSA [metropolitan statistical area] employment inside the transit service area but outside the CBD, the ratio of employment outside the transit service area to employment inside the transit service area, and the ratio of population outside the transit service area to population inside the transit service area). They find that transit ridership is associated with fares, service, and the two employment variables. Transit ridership is positively associated with the percent of MSA employment inside the transit service area (but outside the CBD) and negatively associated with the ratio of employment outside the service area to employment inside the service area. They found that transit ridership is not associated with the strength of the CBD itself, when these other variables are taken into account. These more nuanced findings prompted our desire to reexamine the link between the strength of the CBD and transit ridership. Our work builds on Hendrickson’s (1986) earlier study and addresses some of the limitations of his work. We examine the relationship between transit ridership and the strength of the CBD in 2000, while also controlling for other factors that the literature suggests influence transit ridership. The literature suggests that the key external factors (those outside the control of transit managers) include motor fuel prices (as a surrogate for the overall cost of auto use) (Kain 1997; Pucher 2002), regional unemployment rates (Kain and Liu 1999; Pucher 2002), and the percent of households in the MSA that do not own an automobile (Kain and Liu 1999; Kitamura 1989; Taylor and Miller 2003). The literature suggests that the key internal factors (those within the control of transit managers) include fares (Kain and Liu 1999; McCollom and Pratt 2004; McLeod et al. 1991; Kohn 2000; Stanley and Hyman 2005) and service quality (such as frequency, coverage, and reliability) (Kohn 2000; Pucher 2002; Stanley and Hyman 2005; Taylor and Miller 2003; Thompson and Brown 2006). Data and Methodology The geographic unit for our analysis is the MSA. Other studies have selected individual transit systems (Hartgen and Kinnamon 1999) or urbanized areas (Taylor and Miller 2003) as the unit of analysis, but we rejected these approaches for two reasons. We rejected using individual agencies as our unit of observation because we are interested in the effect of urban structure and, in particular, the strength of the CBD on overall transit ridership in the metropolitan area without regard to which transit agency might transport the riders. We rejected using urbanized areas as our unit of analysis because in many metropolitan areas major transit operators provide service across multiple urbanized areas. Attributing service and ridership data to the proper urbanized area in such circumstances is difficult and subject to significant attribution error. We selected the MSA as the geographic unit that would minimize attribution error, and we aggregated all transit variables to this geographic unit. We defined the MSAs to include the areas identified by the Office of Management and Budget (OMB 2005). We examine the relationship between the strength of the CBD and transit ridership in all U.S. MSAs with more than 500,000 persons, of which there are 82 in the United States as of the 2000 Census. Two are very large MSAs (population in excess of 10 million persons), 8 are large MSAs (population between 5 million and 10 million), 43 are medium MSAs (population between 1 million and 5 million), and 29 are small MSAs (population between 500,000 and 1 million). We stratify the MSAs into three population size groups. The first group contains all 82 MSAs, the second group contains the 43 medium MSAs, and the third group contains the 29 small MSAs. We stratified our MSAs by population size because there are significant differences in the values of our dependent variable from one MSA size category to the next, as we will discuss shortly. We selected the medium MSA and small MSA groups as specific objects of examination because these groups are large enough to permit the use of multivariate statistical analysis. We included the “all MSA” group as a roundabout method of examining the relationship between the urban structure variable and transit ridership in the very large and large MSAs. By comparing the models for the medium and small MSAs to those for the entire dataset and noting the differences in the behavior of the explanatory variables, we are able to gain some insight into the determinants of transit ridership in these 10 largest MSAs. Our analysis covers the year 2000. We obtained data from the U.S. Bureau of Economic Analysis, U.S. Bureau of Labor Statistics, U.S. Census Bureau, and National Transit Database. Data from the U.S. Bureau of Economic Analysis included employment and population (by county) for each MSA (U.S. Bureau of Economic Analysis 2006a; U.S. Bureau of Economic Analysis 2006b). Data from the U.S. Bureau of Labor Statistics included MSA unemployment rates (our measure of MSA economic conditions), consumer price index (used to adjust all money variables to 2005 dollars), and motor fuel price index (used as our measure of the cost of using an automobile) (U.S. Bureau of Labor Statistics 2005a; U.S. Bureau of Labor Statistics 2005b; U.S. Bureau of Labor Statistics 2005c). Data from the U.S. Census Bureau included CBD employment, transit journey to work mode share, and the percent of MSA households that do not own an automobile (U.S. Census Bureau 2000). We obtained all three variables using the Census Transportation Planning Package (CTPP) software. We defined the CBD for each MSA as encompassing the census tracts identified in the 1982 Census of Retail Trade, but we made minor definitional adjustments after consulting local government and metropolitan planning organization websites in each of the MSAs (U.S. Census Bureau 1982). We obtained transit data from the National Transit Database using the Florida Department of Transportation’s (FDOT) Florida Transit Information System (FTIS) software (FDOT 2005). We extracted agency-specific data and aggregated it into MSA-level data for our analysis. The data we obtained include passenger kilometers, vehicle kilometers, route kilometers, and fare revenue variables. We used the combination of these transit variables and other variables discussed above to construct three ratio variables: (1) service coverage (ratio of route kilometers to population), (2) service frequency (ratio of vehicle kilometers to route kilometers), and (3) fare revenue per passenger kilometer (a proxy for average passenger fare; adjusted to 2005 dollars). Measure of Urban Centralization versus Decentralization Our urban structure variable is the share of MSA employment in the CBD for each MSA (CBD employment divided by total metropolitan employment). Table 1 lists CBD employment, total metropolitan employment, and CBD employment share (by MSA) in 2000. In 2000, Greenville, South Carolina, had the weakest CBD (0.68 percent of MSA employment), while New Orleans, Louisiana, had the strongest CBD (10.75 percent of MSA employment). The median MSA had 4.86 percent of its MSA employment inside its CBD in 2000. We selected employment, as opposed to population, as our measure of centralization versus decentralization for three reasons. First, employment decentralization is the focus of most of the literature on urban decentralization and transit ridership that we discussed earlier in the paper. Second, recent studies have found a closer connection between transit ridership and employment than between ridership and population (Brown and Thompson 2008b). Third, employment tends to be collocated with most other travel destinations, which is why it is used as a proxy for these other destinations in most travel demand models used by transportation planners. We decided to express CBD employment as a percent variable, as opposed to number of jobs in CBD, because CBD size (expressed in count form) is correlated with total MSA population and with many other variables that we wished to examine. Measure of transit ridership We measured transit ridership as transit journey-to-work (commute) mode share. This variable measures the percent of work trips made by public transit, and hence is focused solely on commute travel. We hypothesized that this variable would be more strongly influenced by the strength of the CBD than a more general ridership measure, such as passenger kilometers per capita, because the CBD is primarily a destination for work trips. Table 2 reports the 2000 values for transit journey-to-work (commute) mode share by MSA. The smallest reported value for 2000 is found for McAllen, Texas (0.32 percent), while New York has the highest reported value (24.7 percent). The median MSA had a transit commute mode share of 1.98 percent in 2000. We found significant differences in transit commute mode share among MSAs in our four population size groups. The median value for MSAs in the very large MSA group (population over 10 million, 14.7% mode share) is 60 percent higher than the corresponding value for the large MSA group (population from 5 million to 10 million, 8.8% mode share). The median values for our smaller population groups are much lower than these values. The median value for our medium MSAs (population 1 million to 5 million, 2.4% mode share) is nearly twice as large as that for the small MSA group (population from 500,000 to 1 million, 1.2% mode share). These differences reinforced our decision to stratify the MSAs by group size for our multivariate analysis. Hypotheses The literature suggests that transit ridership is tied to a metropolitan area’s urban structure and, in particular, to the strength of the CBD as a locus of economic activity. The purpose of this paper is to test this hypothesis, while also controlling for other internal and external factors that are hypothesized to influence transit ridership. We include the following variables in each of our models: 1. Percent of MSA employment in the CBD. This variable is our CBD strength variable and can be used to measure the degree of employment centralization or decentralization in the MSA. Based on the literature, we would expect to find a positive relationship between the percent of MSA employment in the CBD and transit ridership. 2. Fare per passenger kilometer (adjusted to 2005 dollars). This is a variable that is at least partially under the control of transit agency managers. We expect that MSAs where transit agencies have higher fares will have lower ridership. 3. Service frequency (ratio of vehicle kilometers to route kilometers). This is a variable that is at least partially under the control of transit agency managers. We expect that MSAs where transit agencies offer more frequent service will have higher ridership. 4. Service coverage (ratio of route kilometers to population). This is a variable that is at least partially under the control of transit agency managers. We Central Business Districts and Transit Ridership: A Reexamination of the Relationship in the U. S. Table 2. Transit Journey-to-Work (Commute) Mode Share (by MSA) in 2000 9 expect that MSAs where transit agencies offer more service coverage will have higher ridership. 5. Percent of MSA households that do not own an automobile. This is an external variable (i.e., not under the control of agency managers) that may influence transit ridership. Based on the literature discussed earlier, we expect that MSAs that have higher percentage of carless households will have higher levels of transit ridership. 6. MSA unemployment rate. This is an external variable that may influence transit ridership. We expect that MSAs with higher unemployment rates will have lower ridership, because riders would have less need to use transit to reach jobs. 7. Fuel price index. This is an external variable that may influence transit ridership. We use this variable as a general proxy for the cost of using an automobile. We expect that MSAs with high fuel prices will have high transit ridership. Model Specification We estimated three cross-sectional multivariate ordinary least squares regression models to test our hypotheses. We estimate separate models for all MSAs, medium MSAs, and small MSAs. Through comparison with the medium MSA and small MSA models, we can treat the all MSA model as a pseudo-model for the very large and large MSAs. In evaluating the explanatory variables in each of the models, we are interested in the presence (or lack thereof) of statistical relationships and the practical importance of the statistical association. To measure practical importance, we use elasticity. In order to obtain elasticities, we transformed all the variables into their natural log forms. After this transformation, the coefficients for each explanatory variable can be read as the elasticity of the transit ridership variable with respect to the explanatory variable. We report descriptive statistics for our transformed variables in Table 3. We tested the use of MSA population as a control variable, but decided not to include it because it was not statistically significant in any of our preliminary models. Our MSA stratification appears to have accounted for the variation in transit ridership (by population size group) discussed earlier in the paper. We also tested the percent of MSA population made up of recent immigrants in our preliminary tests but decided not to include it because it was not correlated with our transit ridership variables. We suspect this is due to the wide dispersion of immigrant populations throughout the United States. We considered the inclusion of a variable measuring density, but decided not to include such a variable because only metropolitan-scale measures of density (urbanized area density, MSA density) were available for all 82 MSAs. Multivariate Analysis of Transit Ridership As noted above, we transformed all variables into their natural log forms in order to observe simultaneously 1) the statistical significance of the relationship between each explanatory variable and our dependent variable (when controlling for all other explanatory variables) and 2) the elasticity of the dependent variable with respect to the explanatory variable. The unstandardized coefficients in the tables can be read directly as elasticities. Statistical tests revealed no multicollinearity issues among the variables in our models. Test results are shown under the collinearity statistics columns of each model table. The model results for all 82 MSAs are shown in Table 4. This “all MSAs” model has very high R squared and F statistics, indicating that the model has strong explanatory power. The key insight from the model is the absence of a statistical relationship between the strength of the CBD and transit commute mode share, when other explanatory variables are taken into consideration. This finding thus differs from the traditional view in the literature that posits a strong link between transit ridership and the strength of the CBD. There are four explanatory variables that have statistically-significant relationships with transit commute mode share (at the 0.05 significance level). These variables are service frequency, service coverage, percent of MSA households that do not own cars, and unemployment rate. All four variables behaved as hypothesized. Two variables (service frequency and coverage) are under the control of transit managers. As service frequency and coverage increase, so does the transit commute mode share. The elasticities indicate that service frequency has a stronger effect on commute mode share than service coverage (elasticities of 0.906 and 0.635, respectively). This finding is consistent with other literature. The other two variables are beyond the control of transit managers. Perhaps not surprisingly, the larger the share of carless households in the MSA, the higher the transit commute mode share. In fact, this variable has the strongest effect on transit commute mode share (elasticity of 0.949). In addition, and also not surprisingly, the economic health of the metropolitan area has an effect on the transit commute mode share. As unemployment rates increase, transit commute mode share falls. The second model, shown in Table 5, focuses on the relationship between transit commute mode share and our set of explanatory variables in the medium sized MSAs (population of 1 million to 5 million). The model has high R squared and F statistics, indicating that it is a strong explanatory model. As with our first model, we found no statistical relationship between the strength of the CBD and transit ridership. Three of the four explanatory variables that were significant in the first model are also significant in this model. These variables are: service frequency, service coverage, and the percent carless households. All three variables behaved as hypothesized. As in the first model, MSAs whose transit agencies offered more frequent service and/or better service coverage had higher transit commute mode shares. As in the first model, MSAs with a higher percent of carless households had higher transit commute mode shares. These variables are inelastic with respect to transit commute mode share, with a similar rank order pattern as the model for all MSAs. The third model, shown in Table 6, focuses on the relationship between transit commute mode share and our set of explanatory variables in the small MSAs (population 500,000 to 1,000,000). Again, the R squared and F statistics indicate that this is a powerful model. This is the only one of the three models where our multicollinearity test statistics are not comfortably within widely acceptable ranges. One variable, percent of MSA households that do not own a car, has collinearity statistics that are just barely beyond this range, although the statistics are negligible. M l l A ( e r a h S ods for transporting wheelchair-seated travelers. International Truck & Bus Safety & Security Symposium, November 14–16, 2005, Itasca, IL. ISO. 2010. ISO/DIS 10865 Part 1: Assistive products for persons with disability— Wheelchair containment and occupant retention systems for motor vehicles designed for use by both sitting and standing passengers—Part 1: Systems for rearward facing wheelchair-seated passengers. Geneva, Switzerland: International Standards Organization. Nelson/Nygaard Consulting Associates. 2008. Status report on the use of wheelchairs and other mobility devices on public and private transportation. Washington, D.C: Easter Seals Project ACTION. Project Action. 2008. Status report on the use of wheelchairs and other mobility devices on public and private transportation. San Francisco, CA. Shaw, G. 2008. Investigation of large transit vehicle accidents and establishing appropriate protection for wheelchair riders. Journal of Rehabilitation Research and Development 45(1): 85–108. Society of Automotive Engineers (SAE). 1999. SAE J2249: Wheelchair tiedowns and occupant restraint systems: Surface vehicle recommended practice. Warrendale, PA: SAE. Songer, T., S. G. Fitzgerald, and K. Rotko. 2004. The injury risk to wheelchair occupants using motor vehicle transportation. 48th Annual Proceedings of the Association for the Advancement of Automotive Medicine. About the Authors Linda van Roosmalen , Ph.D. () is Principal at LINC Design LLC, Verona, PA (www.LINC-Design.com) and lead investigator on the RERC on Wheelchair Transportation Safety. Douglas A. Hobson , Ph.D. () is Emeritus Associate Professor at the University of Pittsburgh in the Department of Rehabilitation Science and Technology. Patricia E. Karg , M.S. () is Assistant Professor at the University of Pittsburgh in the Department of Rehabilitation Science and Technology and CoDirector of the RERC on Wheelchair Transportation Safety. Emily DeLeo , B.S. () is a graduate student at the School of Physical Therapy, Duke University, Durham, NC. Erik Porach , B.S. () is a Research Specialist at the University of Pittsburgh Department of Rehabilitation Science and Technology. He supports projects in the RERC on Telerehabilitation, Spinal Cord Injury, and on Wheelchair Transportation Safety. Simulated Analysis of Exclusive Bus Lanes on Expressways: Case Study in Beijing, China Lin Zhu, Beijing Jiaotong University Lei Yu, Ph.D., P.E., Texas Southern University Xu-Mei Chen, Ph.D., Beijing Jiaotong University Ji-Fu Guo, Beijing Transportation Research Center Abstract Deploying exclusive bus lanes is considered an important strategy for supporting public transit priority policy. This paper uses a simulation approach to evaluate planned exclusive bus lanes on expressways in Beijing, China. Two scenarios for deploying exclusive bus lanes—a curbside bus lane scenario and a median bus lane scenario—were designed. Then, a micro-simulation network platform using VISSIM was established and calibrated, with all relative errors between the simulated timevarying speeds and the field speeds less than 15 percent. Afterwards, the two bus lane scenarios were simulated, evaluated, and subsequently compared with current traffic conditions without bus lanes. It was found that for both the mainline and the whole network, the operational efficiencies of buses, general traffic, and all mixed traffic are improved with the deployment of exclusive bus lanes. Further, the median bus lane scenario slightly outperforms the curbside bus lane scenario in this case. Introduction As of April 2012, the number of motor vehicles in Beijing, China, was about 5.06 million. The increasing number of vehicles has resulted in many problems, such as traffic congestion, increased emissions, and noise. Improving public traffic is a key strategy for solving traffic problems and has received increased attention from various government agencies in Beijing. Employing exclusive bus lanes is also a basic public transit priority. The basic idea of deploying exclusive bus lanes is to accommodate large travel demands and improve urban traffic operational efficiency by implementing the proper allocation of space and time resources between buses and general traffic (Yang and Ma 1997). In 1997, Beijing installed its first exclusive bus lane on the right curbside lane of Chang’an Street, which is used only by buses from 6 am to 8 pm. Construction of bus lanes in Beijing is quite slow, and only about 20 bus lanes are built each year. Until now, the total length of exclusive bus lanes was about 303 kilometers in Beijing, which is far from the requests of public transit planning and management. Therefore, in the 12th five-year plan of Beijing, more than 150 exclusive bus lanes will be built. Expressways are the major arterials in the urban traffic network of Beijing, carrying more than 50 percent of the total daily traffic of the city. Traffic conditions on expressways in Beijing indicate that there are extremely high traffic volumes, a high density of public transit lines, large bus flows at bus stops, and high densities at on- and off-ramps. A commonly-observed phenomenon is that, to get through the already-congested roads, automobiles and buses must compete for the rightof-way without concessions, resulting in even worse traffic conditions. Therefore, relevant agencies are proposing to deploy exclusive bus lanes on expressways to reduce conflicts between vehicles by physically separating automobiles and buses. However, exclusive bus lanes are usually constructed on urban arterials and other key roads of general grades, and employing exclusive bus lanes on expressways is less common. There is no doubt that exclusive bus lanes will have some degree of impact on road traffic, which has been studied by a number of researchers. Based on simulations and field surveys, St. Jacques and Levinson (1997) developed an analysis procedure for estimating capacities and speeds on arterials with at least one exclusive bus lane with either no, partial, or exclusive use of the adjacent lane. Siddique and Khan (2006) used NETSIM to model and forecast traffic conditions along BRT corridors in Ottawa for 2021, which were compared with traffic conditions in 2001. The study focused on the capacity analysis of BRT operation on exclusive bus lanes. Although the deployment of exclusive bus lanes on expressways has been planned, relevant studies are still rare. Chen et al. (2009) analyzed the impacts of exclusive bus lanes on the capacity of the ring-road expressway using the VISSIM model. The analyzed parameters included the styles and distances of ramps, length of weaving sections, bus headway, and others. The simulation results showed that weaving section length and bus headway are more sensitive, especially for on- and off-ramps for curbside bus lanes. In light of the above, the research in this paper simulated the impact of deploying exclusive bus lanes on an expressway. To this end, it first carried out a series of comprehensive traffic surveys and data collections along the Western 3rd Ring-Road Expressway in Beijing. Second, it explained the conditions of setting an exclusive bus lane and designed two bus lane scenarios, including a median bus lane and a curbside bus lane. Then, it established a simulation platform using VISSIM for the Western 3rd Ring-Road Expressway and calibrated the model parameters. Finally, it comparatively simulated and evaluated the two designed scenarios of an exclusive bus lane. Study Area and Data Collection Description of Western 3rd Ring-Road Expressway Existing ring-road expressways in Beijing include the 2nd, 3rd, 4th, 5th, and 6th ringroad expressways. The selected network in this study is the main portion of the Western 3rd Ring-Road Expressway Network, which is about 8.5 kilometers long in the south-north direction, including 7 interchange bridges, as shown in Figure 1. The Western 3rd Ring-Road Expressway provides 3 lanes in each direction with widths of 3.5 meters, 3.25 meters, and 3.5 meters for median, center, and shoulder lanes, respectively, as well as an emergency lane that is 4.75 meters wide. There are frontage roads present with two lanes in each direction along the expressway, which are connected to the mainline through ramps. A green zone with a width of 2 meters is reserved in the middle of two directions as well as between the mainline and frontage roads. One of the busiest traffic corridors in Beijing, the Western 3rd Ring-Road Expressway is crossed by four urban expressways and four major arterials. Along and near the expressway, there exist Lize Bridge Coach Station, Liuli Bridge Coach Station, Lianghuachi Coach Station, and Beijing Western Railway Station, the largest railway station in Beijing. Public transit demand is quite high in this area, with a total of more than 40,000 passengers per day getting on and off buses at each bus stop. The bus cross-sectional volume in the peak hours reaches 300 vehicles per hour, including 12- or 14-meter single buses; articulated buses 14, 16 or 18 meters in length; and 10- or 12-meter double-deck buses. Thus, many large vehicles run on the expressway simultaneously, which considerably affects the traffic conditions on the expressway. Data Collection and Preparation Using data sources, collection methods, and data characteristics and usage, this study carried out four tasks of data collection to support the research. The first task was collecting geographical data, such as the latest version of the Beijing E-map and aerial map, the regional road GIS map, and the transit route GIS map, which provided the geographic and structure information about the network. The second task was collecting information about network facilities and traffic control measures, including road geometric information (length, width, and number of lanes), locations of on- and off-ramps, traffic paths at intersections and overpasses, intersection signal timings, and information about bus stops (location, form, length of platform, and number of berths) and distribution of transit routes. The third task was collecting traffic flow data at network entrances and diversion points, which are required by the simulation model. Specifically, the data contain flows at 30 network entrances, diversion flow ratios at 45 ramps, traffic volumes at 8 approaches of 2 signalized intersections and 78 diversion points of 7 overpasses, and bus headways of each bus line at the entrances. The final part task collecting data from Remote Traffic Microwave Sensors (RTMS) and the transit data from Global Position System (GPS). RTMS data can provide flow and speed information at 20 sections along the Western 3rd Ring-Road Expressway. The original data collected at 2-minute intervals by RTMS were aggregated into data at time intervals of 10 minutes, 1 hour, or 2 hours. In the study, hand-carried GPS units were used to collect bus speed data at 2-second intervals for selected bus lines by boarding on buses. These data were used in the calibration and validation of the simulation model. Design of Exclusive Bus Lanes on Expressways Conditions of Setting an Exclusive Bus Lane This section explains the conditions of setting an exclusive bus lane, as follows: (1) Geometric conditions on the road: There should be at least 2 lanes in each direction on the road, and it is better if there are 3 or 4 lanes (Lu 2003). Considering the needed space for bus vehicles, the width of a bus lane usually equals 3.5 meters, which can be appropriately reduced but should be at least 3 meters (Yang 2003). The Western 3rd Ring-Road Expressway has 3 lanes in each direction with widths of 3.5 meters, 3.25 meters, and 3.5 meter, respectively, and an emergency lane that is 4.75 meters wide. Accordingly, the geometric structure of the Western 3rd Ring-Road Expressway meets the physical requirements of deploying exclusive bus lanes. (2) Traffic saturation level on the road: It is necessary to deploy an exclusive bus lane when the volume-to-capacity ratio on a road arrives at or exceeds the value of 0.8 (Zhang et al. 2000). According to the surveyed flow data, the average volume-to-capacity ratio on the Western 3rd Ring-Road Expressway is 0.94, and the values of several sections are higher than 1. (3) Bus volume on the road section: It is suggested to build an exclusive bus lane if bus volume on a road section in peak hours is higher than 150 vehicles per hour (Yang et al. 2000) . The field surveyed data indicate that the bus volume in the peak hours on the mainline of the Western 3rd Ring-Road Expressway is more than 225 vehicles per hour. (4) Public transit passenger volume on the road section: The Highway Capacity Manual (National Research Council 2000) suggests that passenger volume on a bus lane should be 50 percent higher than that on other lanes, and this value should be more than 3,000 person-trips per hour on the planned bus lane in Shanghai (Lin et al. 2007). According to the surveyed data, the average passenger volume of public transit on the Western 3rd Ring-Road Expressway is about 17,750 person-trips per hour in the peak hours and occupies about 70 percent of total service passenger volume on the section. Consequently, it is qualified and necessary to deploy an exclusive bus lane on the Western 3rd Ring-Road Expressway in Beijing. Scenario Designs of Exclusive Bus Lane The key elements in the design of exclusive bus lanes on expressways include physical location of the bus lane, structure of the bus stops, ramp control towards buses, and corresponding adjustment of bus lines, all of which have been considered in the designs of the two exclusive bus lane scenarios on the Western 3rd Ring-Road Expressway. Basic on the current structure of the roads, two exclusive bus lane scenarios were designed (as shown in Figure 2), which were modeled and evaluated with the established VISSIM simulation model. Basic Scenario Current traffic conditions without the exclusive bus lane were simulated based on the field data collected. Scenario 1: Curbside Bus Lane Scenario With the road structure of the mainline unchanged, a curbside lane was used as the bus lane in this scenario. There were no major structural and positional changes on bus stops, i.e., bus bays remained the same, passenger waiting areas continued to occupy the green zone, and buses parked in the emergency lane, as shown in Figure 2. Buses ran on the curbside lane and could enter or exit the bus lane conveniently. All buses followed the current routes. Scenario 2: Median Bus Lane Scenario In this scenario, the median lane was used as the bus lane. Therefore, bus stops were moved from the curbside to the center of the expressway. The widths of the general traffic lanes and the emergency lane were slightly narrowed to ensure adequate space required for parking the buses. Figure 2 shows the configuration of the bus stop area in the median bus lane scenario. Buses had to enter or exit the median bus lane by crossing two general traffic lanes, which caused a serious interference with traffic. Therefore, a bus ramp control strategy was proposed, which meant that when running in the median bus lane, buses could enter the mainline only from one on-ramp and exit the mainline only from one off-ramp. In this design, bus access ramps in both directions were placed at the upstream and downstream links of Lianhua Bridge. To accommodate this design, some bus lines were also adjusted correspondingly. Establishment of Simulation Model Description of Simulation Approach In this study, the traffic simulation technique was used to model, evaluate, and analyze the scenarios of an exclusive bus lane on the Western 3rd Ring-Road Expressway. VISSIM, a widely used micro-simulation model, was employed. A simulation framework was developed using VISSIM for this study, based on the network information and traffic data of the Western 3rd Ring-Road Expressway, as shown in Figure 3. (1) Develop a simulation platform of the Western 3rd Ring-Road Expressway Network using VISSIM based on the surveyed data. (2) Calibrate the physical attributes of the network, the vehicle desired speed distribution using the frequency analysis, and the driving behavior parameters using a combined calibration algorithm (introduced in the next subsection). (3) Design and run simulation scenarios, including the simulation of current traffic conditions and simulations of two different designs of the exclusive bus lane. (4) Select a set of performance measures to analyze the simulation results and evaluate the effectiveness of different designed scenarios. Calibration of Simulation Model Roads and vehicles are the basic elements of urban traffic systems; therefore, a traffic simulation model usually consists of a network element and a traffic element. The former describes the geometric structures of roads and the connective relations of links, while the latter describes the moving characteristics of vehicles. Consequently, the calibrations of the two aspects underlie the accuracy and reliability of scenario experiments and evaluations. Calibration of the network model is completed by adjusting the static traffic parameters, including the connections of links, flow paths and ratios at key nodes, and locations of functional change of lanes. The selected precision indicator for this calibration was the Relative Errors (RE) of two-hour accumulated flows from 7–9 am between the simulated results and the collected RTMS data. The locations of detectors in the simulation model were made consistent with RTMS detectors in the real network, as shown in Figure 4. After calibrating the network model, the maximum RE of the two-hour accumulated flows was 9.00 percent. This result satisfied the requirements of the study. Calibration of the traffic model is conducted to adjust the default model parameters to capture the actual traffic behaviors in the real network. In VISSIM, the key parameters that needed to be calibrated included the desired speed distribution and driving behavior parameters. In VISSIM, desired speed distribution is defined to describe the fact that a driver will travel at a desired speed (with a stochastic variation) when not hindered by other vehicles. The maximum and minimum values for the desired speed, as well as the intermediate points, are determined by a frequency analysis of the vehicle speed data collected during free-flow periods. The desired speed distribution of general traffic was obtained by analyzing the speed data from RTMS on the Western 3rd Ring-Road Expressway from 12–6 am on October 7 and 8, 2008; the desired speed distribution of buses on general lanes was obtained by analyzing the GPS speed data of buses on the Western 3rd Ring-Road Expressway from 2–4 pm(the lowest bus-flow period). It was noted that there are no deployed bus lanes on the Western 3rd Ring-Road Expressway at present. Therefore, the GPS speed data of buses on the section from Xizhimen Bridge to Jishuitan Bridge of the 2nd ring-road from 7–9 am were collected and analyzed to obtain the desired speed distribution of buses on the bus lane and represent the bus running condition on the bus lane on the Western 3rd Ring-Road Expressway. In VISSIM, driving behavior parameters describe the vehicle-following and lanechanging behaviors, lateral behavior, and reaction behavior to signals. In a sensitivity analysis of parameters using a simple network, 10 driving behavior parameters were screened, including the maximum look-ahead distance, average standstill distance, additive part of safety distance, multiplicable part of safety distance, maximum deceleration for lane changes, accepted deceleration for lane changes, waiting time before diffusion, minimum headway for lane changes, reduction rate (as meters per 1 m/s²), and minimum lateral distance for 50 km/h. Combining the Generic Algorithm (GA) with the Simultaneous Perturbation Stochastic Approximation Algorithm (SPSA), a calibration program for driving behavior parameters was developed using Visual C++ and MATLAB languages. First, the parameters were locked in a relatively small area using the GA, overcoming the SPSA’s shortcoming of inefficient global optimization; then, the SPSA was used to solve the problem in the locked area, overcoming the slow convergence of GA. In the calibration process, the Sum of Squared Error (SSE) between the simulated time-varying speeds and the actual speeds was selected as the measure to determine the best combination of the 10 parameters for the Western 3rd Ring-Road Expressway simulation model. The calibration algorithm is explained in detail by Chen et al. (2011). After completing the calibrations of network model and traffic model, the timevarying speeds at 10-minute intervals were output from 20 detectors, and then the simulation results were compared with the RTMS data. The simulated time-varying speeds and the actual RTMS speeds were plotted around the 45-degree line, as shown in Figure 5. The relative errors of time-varying speeds at all 20 detectors in the network was less than 15 percent. Measures of Effectiveness This study evaluated the impacts of exclusive bus lane schemes on traffic conditions on expressways using a traffic simulation model. The evaluation targets included buses, general cars, and all traffic. The whole network is composed of the Western 3rd Ring-Road Expressway, frontage roads, and crossing roads. According to different parts of the network and targets for traffic evaluation, the Measures of Effectiveness (MOEs) were selected, as listed in Table 1. All the selected MOEs focus on the efficiencies of buses, general traffic, and all traffic for evaluating bus operational impacts, economic benefits, environment effects, etc. For the mainline of the Western 3rd Ring-Road Expressway, the average crosssectional speeds were obtained from 20 detectors at 10-minute intervals, and the average travel time was the mean travel time of all vehicles that completed travel on the mainline. For the whole network, average travel speed was equal to total distance traveled by all vehicles divided by total travel time, while average delay per unit distance was calculated by total travel delays of all vehicles divided by total travel distance. In addition, to evaluate the performance of all mixed traffic in the entire network, the numbers of passengers in buses and general vehicles were used as the weights in calculating measures. All traffic Bus All traffic Measures of Effectiveness Average section speed (km/h) Average travel time (s/veh) Average section speed (km/h) Average travel time (s/veh) Average travel time (s/person) Average travel speed (km/h) Average delay per unit distance (s/km) Average travel speed (km/h) Average delay per distance (s/km) Average delay per unit distance (s/km/person) Simulations and Analyses Simulation runs were implemented for the designed scenarios using the established VISSIM platform for the Western 3rd Ring-Road Expressway. Traffic flows input in the basic scenario came from the field data, while those in the curbside and median bus lane scenarios were obtained from the outputs of the mesoscopic INTEGRATION model, which covers a larger area network (BJTU and BTRC 2008). The simulation period was set to three hours. The first half-hour is the warm-up time used to load the network with traffic, the last half-hour was the clear-up time used to empty the network, and the middle two hours were used to simulate the actual period of 7–9 am. Spatial-temporal speed distributions were generated to illustrate the impacts of exclusive bus lanes on buses and general traffic, as shown in Figures 6 and 7. In the morning peak hours, the current cross-sectional speeds of buses and general traffic in the outer-ring direction were higher than those in the inner-ring direction. Figure 6 shows that the detected bus speeds from most of detectors increased visibly after deploying the exclusive bus lanes. However, within the curbside bus lane, the speeds at Sensor 3 in the outer-ring direction were reduced as buses are interfered with by cars entering or exiting the mainline, which generated a bottleneck in the outer-ring. As shown in Figure 7, the traffic conditions on the general lanes in the two exclusive bus lane scenarios were much better than current conditions, especially in the outer-ring direction. In the median bus lane scenario, because bus access ramps are set near Lianhua Bridge, plus more loading traffic in the inner-ring, the traffic in the south of Lianhua Bridge remained terrible. Figure 8 illustrates the average travel times of different vehicles completing travel on the mainline of the Western 3rd Ring-Road Expressway under the three scenarios. From Figure 8, the following conclusions can be derived: (1) Currently, the travel time in the inner-ring direction is higher than that in the outer-ring direction, and the bus travel time is higher than the general traffic travel time. (2) In the median bus lane scenario, the travel times of both buses and general traffic decreased significantly. (3) In the curbside bus lane scenario, although the right-of-way of traffic was well defined, the freedom of general traffic for entering and exiting the mainline was compromised, which increased the travel time of general traffic by 3.4 percent. (4) The average travel times of all mixed traffic decreased in both directions with the exclusive bus lanes. Apparently, the magnitude of decrease in the median bus lane scenario is much bigger. Impacts of the exclusive bus lane on traffic conditions of the entire studied network were analyzed. Results of MOEs are shown in Table 2. From the values of average travel speed and average delay per unit distance for all traffic in the simulation network, the median bus lane scenario outperformed the curbside bus lane scenario in this case. (1) Apparently, in the morning peak hours, the traffic on the inner-ring of the Western 3rd Ring-Road is more congested than that on the outer-ring. The former carries more traffic volumes and experiences lower speed. (2) For the mainline of the expressway, the average speeds of buses improve with the exclusive bus lanes, and the average travel time decreases. (3) The spatial-temporal speeds of general traffic on the expressway have more noticeable characteristics with the deployment of exclusive bus lanes. Congestion appears mainly north of Huayuan Bridge on the outer-ring and south of Lianhua Bridge on the inner-ring. (4) For the case network in this paper, the traffic operational efficiency of traffic in the bus lane scenarios was improved. Further, the median bus lane scenario slightly outperformed the curbside bus lane scenario. This is the first relatively complete case study on the design and evaluation of exclusive bus lanes on urban expressways. The entire study was carried out on the basis of comprehensive and extensive field traffic data. Therefore, the study results are of practically significance. Evaluations in the paper, which focus mainly on traffic impacts, are still quite general. It is recommended that further studies be conducted on the special traffic operational problems associated with the deployment of exclusive bus lanes, such as impacts on the traffic near bus stops, traffic conditions near on- and off-ramps, etc. This paper focuses only on the corridor network of the Western 3rd Ring-Road Expressway. It is, therefore, recommended that the study be expanded in the future to other networks with different locations and scopes, such as the other ring-road expressways or radial roads. Acknowledgment This paper was prepared based on support from Major Projects of the Beijing Transportation Research Center #T08L04404 and #CZ200704 and from Fundamental Research Funds for the Central Universities #2009YJS045. Beijing Jiaotong University (BJTU) and Beijing Transportation Research Center (BTRC). 2008. The Research Report of Dynamic Traffic Model for Beijing. Beijing, China. Chen X. M., L. Yu, L. Zhu, J. F. Guo, and M. Z. Sun. 2009. Capacity analysis of weaving sections on an urban expressway with exclusive bus lanes using microscopic traffic simulation. 88th Transportation Research Board Annual Meeting CDROM, #09-0535, Washington, D.C. Chen X. M., L. Yu, L. Y. Zhu, Y. Zhang, and Z. Lin. 2011. Calibrating and validating a micro-simulation model of a bus rapid transit corridor with heuristic optimization methods. 90th Transportation Research Board Annual Meeting CD-ROM, #11-2641, Washington, D.C. Lin T., K. F. Yang, and J. X. Zheng. 2007. Study on bus lane system planning method: Case study of Shenzhen. Traffic & Transportation 7: 1–4. Lu J. 2003. Study of installation conditions of bus lanes. Communications Standardization 1: 59–61. National Research Council. 2000. Highway Capacity Manual 2000. Washington, D.C.: Transportation Research Board. Siddique, A. J., and A. M. Khan. 2006. Microscopic simulation approach to capacity analysis of bus rapid transit corridors. Journal of Public Transportation 9: 181–200. St. Jacques, K., and H. S. Levinson. 1997. Operational analysis of bus lanes on arterials. TCRP A-7. Washington, D.C.: Transportation Research Board, National Research Council. Yang, X. G. 2003. Manual of Urban Traffic Design. Beijing: China Communications Press. About the Authors Lin Zhu () is a Ph.D. Candidate at School of Traffic and Transportation, Beijing Jiaotong University, Beijing, China. Her research interests involve public transit planning and management, transportation simulation, and Intelligent Transportation Systems. Lei Yu, Ph.D., P.E. () is Professor and Dean of the College of Science and Technology, Texas Southern University, and a Yangtze River Scholar of Beijing Jiaotong University. He has managed more than 80 research projects and published more than 200 scientific papers. Xu-Mei Chen , Ph.D. () is Associate Professor at the School of Traffic and Transportation, Beijing Jiaotong University, Beijing, China. Her researches interests involve ITS technologies, urban traffic planning and management, public transit operation, and industry policy of transportation. Yang , X. G. , G. W. Zhou , M. S. Hang , and C. H. Shi . 2000 . Technologies and methods of transit priority . Urban Transport of China 2 : 1 - 12 . Zhang , W. H., Y. J. Huang , and G. Hu . 2003 . Study on design standards for urban bus lanes . Communications Standardization 7 : 33 - 36 .


This is a preview of a remote PDF: https://scholarcommons.usf.edu/cgi/viewcontent.cgi?article=1084&context=jpt

Full Issue 15(4), Journal of Public Transportation, 2012,