National Academies Press: OpenBook

Multimodal Level of Service Analysis for Urban Streets (2008)

Chapter: Chapter 6 - Transit LOS Model

« Previous: Chapter 5 - Auto LOS Model
Page 72
Suggested Citation:"Chapter 6 - Transit LOS Model." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 72
Page 73
Suggested Citation:"Chapter 6 - Transit LOS Model." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 73
Page 74
Suggested Citation:"Chapter 6 - Transit LOS Model." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 74
Page 75
Suggested Citation:"Chapter 6 - Transit LOS Model." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 75
Page 76
Suggested Citation:"Chapter 6 - Transit LOS Model." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 76
Page 77
Suggested Citation:"Chapter 6 - Transit LOS Model." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 77
Page 78
Suggested Citation:"Chapter 6 - Transit LOS Model." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 78
Page 79
Suggested Citation:"Chapter 6 - Transit LOS Model." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 79
Page 80
Suggested Citation:"Chapter 6 - Transit LOS Model." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 80
Page 81
Suggested Citation:"Chapter 6 - Transit LOS Model." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 81

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

72 6.1 Model Development The Transit Capacity and Quality of Service Manual (TCQSM) provides a family of LOS models for dealing with several dimensions of transit service at different levels of ge- ographic aggregation. The TCQSM is oriented to the entire service area, the entire route, or the bus stop. It was necessary to extract a subset of these quality-of-service measures that were most appropriate for a single urban street. The urban street is at a level of aggregation that is greater than the bus stop level and incorporates multiple routes using the street, but it covers just the portion of the routes that actually use the street. Thus a different geographic focus was necessary in the development of the Urban Street transit level of service model. Transit riders were surveyed on portions of routes using a specific urban street to determine what factors most signifi- cantly influenced their perceived quality of service. It was quickly discovered that passengers were basing their LOS rat- ings on their entire trip experience up to that point and not just the portion of their trip on a specific urban street. In ad- dition, an on-board survey can survey only those that even- tually chose to ride transit; it cannot take into account the opinions of those who chose not to ride that bus or selected a different route. Consequently, the surveys were used to iden- tify the key factors influencing perceptions of quality of ser- vice, but LOS models were not fitted to the on-board survey levels of service. An alternative source of data on traveler preferences was necessary to construct an urban street level of service model for transit. The working hypothesis of the research team was that “people vote with their feet.” When confronted with a choice, people will pick the service that gives them more of what they value, in our case, quality of service. Thus, standard models of transit mode choice were consulted to identify the relationships between various service characteristics and the likely proportional increase in ridership. LOS E was set for a hypothetical, base transit service on an urban street. A mode choice model would then be used to compare the ridership for the actual transit service to that for the hypothetical base case. An increase in ridership over the hypothetical base case would be interpreted as an indication of a preference for the actual service over the base case. The actual service would be assigned a level of service superior to E. Similarly, lesser ridership would be interpreted as an indi- cation of poorer quality of service and would be assigned a level of service inferior to E. The application of mode choice models at the urban street level was considered impractical, so mode choice models were replaced with elasticities derived from typical mode choice models. The elasticities predict the percent increase in ridership as a function of percent change in the transit service characteristics. Selection of Explanatory Variables for LOS The Phase 1 surveys asked passengers to rate their satisfac- tion with 17 specific aspects of their trip. A multiple linear re- gression model was developed that related individual factor ratings to the overall satisfaction rating. The factors that added significance to the model were • Close to home rating; • Close to destination rating; • Frequency rating; • Reliability rating; • Driver friendliness rating; • Seat availability rating; and • Travel time rating. Of these factors, “close to home” and “close to destination” relate to getting to the stop, “frequency” and “reliability” re- late to waiting at the stop, and “driver friendliness,” “seat availability,” and “travel time” relate to the ride on the bus. C H A P T E R 6 Transit LOS Model

73 Other considerations also had to be taken into account during this factor selection process: 1. The factors included in the model should be under the con- trol of either the transit operator or the roadway owner; 2. To the extent possible and warranted, the factors as a whole should reflect the influence of other modes on tran- sit quality of service; 3. The factors should be readily measurable in the field; 4. The factors should reflect conditions existing within the urban street right-of-way; and 5. The factors should have a documented impact on some as- pect of customer satisfaction. Based on these criteria, “driver friendliness” was dropped from consideration. Although partially under the control of the transit operator, this factor can only be measured through a customer satisfaction survey, which we felt made it imprac- tical to include. In addition, we are not aware of any research relating different levels of driver friendliness to some measur- able aspect of satisfaction (for example, increased ridership). The factors “close to home” and “close to destination” gen- erated considerable discussion among the project team. Walking distance to the stop depends on a number of factors beyond the urban street right-of-way, including land use pat- terns, street connectivity, transit route structure, stop loca- tions, and sidewalk provision on connecting streets, which would tend to suggest not including these factors. At the same time, there are known relationships that describe how bus pa- tronage declines the farther one has to walk to a stop. One potential surrogate measure identified through initial statistical modeling is “number of stops per mile”—the more stops per mile, the shorter the distance passengers may have to walk to get to a stop once they reach the street with transit service. However, there are two potential difficulties with this measure. First, the more stops per mile, the slower the bus travel time. Travel time is already identified as a potential factor, so adding stops per mile to the model would be redundant. Second, long stop spacing may or may not be in- convenient to passengers, depending on how convenient the stops are to where passengers actually want to go. Without knowing something about adjacent land development pat- terns (which takes the analyst beyond the urban street right- of-way), it is hard to make a judgment about the impact of stop spacing on customer access. Another potential surrogate measure would be the dis- tance of the bus stop from the nearest intersection. This is something that may be influenced by the auto mode—for ex- ample, traffic engineers frequently do not want far-side bus stops located adjacent to intersections, in situations where buses must stop in the travel lane, because of the potential for cars to stop behind the bus and block the intersection. Moving the stop farther from the intersection increases walk- ing distances for passengers arriving from three of the four di- rections at the intersection, which can be related to walking time. On the other hand, near-side/far-side stop location trade-offs can be evaluated through changes in travel speed, using methodologies found in the TCQSM. A third potential surrogate, and the one recommended by the project team, is pedestrian LOS. Pedestrian LOS relates to the ease of access to and from destinations along the urban street, the quality of pedestrian facilities serving the bus stop, and the difficulty of crossing the street. It will be a part of the multimodal urban street LOS methodology; therefore, no ad- ditional data collection will be required. It is a measure of the impact of another mode on the transit mode and can be impacted by roadway agency actions. In short, it meets all of the criteria set out above. The TCQSM provides an areawide measure, “service coverage,” that addresses the “close to home” and “close to destination” factors. This measure accounts for land use pat- terns, street connectivity, and street-crossing difficulty, at the cost of requiring more data than is desirable for an urban street analysis. The four remaining candidate factors are travel time, re- liability, seat availability, and frequency. All are impacted by conditions on the urban street, or by transit or roadway agency actions. All are related to TCQSM measures, which is important from a consistency standpoint. The first three factors can be related to travel time, which addresses a panel request to consider travel speed in the transit LOS model. The key remaining question is: Do relationships exist between passenger satisfaction and different values of these factors? The answer to this question appears to be “yes.” Consid- erable research has been conducted on traveler ridership re- sponses to changes in service frequency and travel time. (Both TCRP Report 95: Traveler Response to System Changes, and the Victoria Transport Policy Institute’s Online TDM Encyclopedia provide extensive summaries of the literature pertaining to ridership responses to transit system changes.) For example, as bus headways decrease from 60 minutes to 30 minutes, from 30 minutes to 15 minutes, and so on, rid- ership increases, although in an ever-decreasing proportion to the amount of added service. All other things being equal, the relative amount of ridership one would expect at a given headway, compared to a 60-minute headway, is reflective of the difference in customer satisfaction between the two headways. There is comparatively little research on the impacts of re- liability and crowding on ridership. However, reliability can be converted to an “excess wait time”—the average addi- tional amount of time one would wait for a bus as a result of non-uniform headways. The excess wait time can, in turn, be

74 converted into a perceived wait time, with an impact greater than the actual wait time [98]. Time spent standing or even seated in a crowded transit vehicle is also perceived by pas- sengers as being more onerous than the actual travel time (See, for example, Balcombe [99]). Thus, the level of crowd- ing on a bus can be used to convert an actual in-vehicle travel time to a perceived in-vehicle travel time. Furthermore, vari- able headways result in uneven loadings on buses, with the re- sult that late buses are more crowded than would be sug- gested by an average peak-hour or peak-15-minute load. Finally, research exists to document how certain kinds of stop amenities can help reduce the perceived waiting time at bus stops. Therefore, the recommended factors to include in the tran- sit LOS model are the following: • Service frequency (headways); • Travel time (speed); • Crowding; • Reliability (headway variability); • Presence of stop amenities documented to reduce per- ceived wait time; and • Pedestrian LOS. Proposed General Model Form for Transit LOS The proposed general form for the transit LOS model is a linear combination of the quality of service accessing the bus stop on foot and the quality of service involved in waiting for and riding the bus. It is similar to many transit mode choice models incorporating the factors of accessibility, wait time, and travel time to predict the probability of choosing transit. This model form varies slightly from traditional mode choice models in that it blends wait time and travel time into a sin- gle factor before adding the result to the accessibility. Only pedestrian accessibility is considered (as opposed to auto ac- cessibility) because this model is designed for application in an urban street environment where park and ride is less likely a phenomenon. Transit LOS = a1 * Pedestrian Access + a2 * Transit Wait/Ride (Eq. 20) Where: a1, a2 = calibration parameters Pedestrian Access Score = A measure of the pedestrian level of service for the street. Transit Wait/Ride Score = A measure of the quality of transit ride and waiting time. The quality of the pedestrian access can be conveniently obtained by employing the pedestrian level of service score for the street. The quality of the transit wait/ride experience would be measured based on the average wait time for a bus and the perceived travel time on the bus. The ratio of transit patronage for the actual wait time di- vided by the patronage for a base wait time gives an indica- tion of the perceived quality of the service provided relevant to the wait time. The ratio of transit patronage for the perceived travel time rate (minutes per mile, the inverse of the speed) divided by the patronage for a base travel time rate gives an indication of the perceived quality of the service provided relevant to the travel time rate of service. The patronage ratios are estimated based on patronage elasticities obtained from various research, as explained in the following sections. Elasticity Concept For practical purposes it is not feasible to apply full mode choice models to an isolated urban street (because the mode choice models are designed to consider the entire trip while the street is limited to portions of the trip). Elasticities were adopted instead since they can be used to predict changes in ridership without having to consider the full length of the trip. A basic, hypothetical transit service for the urban street is assumed which would provide LOS E as far as transit patrons are concerned. The difference between the actual service and the hypothetical service is converted into an estimated per- centage change in ridership to determine by how much the actual service LOS exceeds (or falls below) LOS E. The two key components of the TransitWaitRideScore are the headway factor and the perceived travel time factor. These in turn, are related to documented traveler responses to changes in headway and changes in travel time. These re- sponses are quantified in terms of elasticities. Transit elasticities reflect the percent change in transit rid- ership resulting from a 1% change in an attribute of the ser- vice (e.g., fare, frequency, travel time, service hours, etc.). The relationship between demand and service attribute need not be linear. This is the case with service frequency: dou- bling the frequency from one bus to two buses an hour on a route has a much greater percentage impact on ridership than doubling the frequency from six buses to twelve buses an hour. Thus, the value of elasticity, E, may be different de- pending on where one starts from and where one ends up. TCRP Report 95, the source of many of the elasticity values used for the transit model, uses the concept of mid-point arc elasticity to approximate this relationship, based on average before-and-after values of the two variables, ridership and service attribute [100]. These relationships are illustrated in Exhibit 82 and expressed mathematically in the equation below.

75 P Q 0 Q1Q2 P2 P1 Ridership Se rv ic e At tri bu te Q1Q2 (Q1+Q2)/2 (P1+P2)/2 P2 P1 Q2-Q1 Exhibit 82. Mid-Point Arc Elasticity. (Eq. 21) Where E = mid-point arc elasticity; P = the before (P1) and after (P2) prices (e.g., fare, headway, travel time); and Q = the before (Q1) and after (Q2) ridership demand. Elasticity is relative—the actual ridership of a route is de- termined by many factors, including the type and density of adjacent land uses, the demographic characteristics of per- sons living near the route (e.g., age and vehicle ownership), and the ease of access to bus stops. However, given no changes in these external factors, one can estimate the change in a route’s ridership resulting from a change in a single ser- vice attribute under the control of a transit or roadway agency. Whether the resulting change in ridership is from 100 to 150 riders, or from 1,000 to 1,500 riders makes no difference—one can still estimate how much more attractive (satisfactory) one level of a particular service attribute to passengers is compared to another. Bus Frequency Elasticity Elasticities related to how often bus service is provided can be expressed as frequency elasticities (using positive numbers—increased frequencies result in increased rider- ship) or as headway elasticities (using negative numbers— decreased headways result in increased ridership). Trends identified in TCRP Report 95 suggest that frequency elastici- ties can be +1.0 or greater, in situations where the original service was very infrequent (60-minute headways or longer), that is, doubling the frequency may more than double the rid- ership in those situations. With more frequent service as a starting point, typical frequency elasticities are in the +0.3 to E Q Q Q P P P Q P P P Q Q = + ÷ + = + + Δ Δ Δ Δ( )/ ( )/ ( ) (1 2 1 2 1 2 1 22 2 ) ( )( ) ( )( ) = − + + − Q Q P P Q Q P P 2 1 1 2 1 2 2 1 +0.5 range, dropping to as low as +0.2 with very frequent service (i.e., service every 10 minutes or better). The recommended transit LOS model uses the following frequency elasticity values, based on typical values reported in TCRP Report 95: +1.0 for 1–2 buses/hour, +0.5 for 2–4 buses/hour, +0.3 for 4–6 buses/hour, and +0.2 for 6 or more buses/hour. Solving for Q2 in Equation 1, one can estimate fu- ture ridership demand based on a given starting demand and an assumed elasticity, as shown Equation 22 below. (Eq. 22) Thus, with an elasticity of +1.0, a route with a ridership of 100 passengers at 60-minute headways (P1 = 1 bus/hour) would be expected on average to have a ridership of 133 pas- sengers if headways improved to 45 minutes (P2 = 1.33 bus/hour), and a ridership of 200 passengers if headways improved to 30 minutes. With the decreased response to frequency changes assumed to begin at 30-minute head- ways, ridership would increase to 244 passengers at 20-minute headways (E = +0.5, Q1 = 200 passengers, P1 = 2 buses/hour, and P2 = 3 buses/hour), 280 passengers at 15-minute headways, 316 passengers at 10-minute headways, and 379 passengers at 5-minute headways. For any given frequency or headway, one can estimate the ridership relative to a 60-minute head- way and, thus, the relative attractiveness of the service. In this example, 10-minute headways produce 3.16 times the num- ber of passengers compared to 60-minute headways, all other things being equal; therefore, the value of fh that would be used for 10-minute headways would be 3.16. If local data were available, local elasticities could be sub- stituted for the typical national values used in the model. Travel Time Elasticity A review of transit travel time elasticities in the literature, conducted by TCRP Project A-23A (Cost and Effectiveness of Selected Bus Rapid Transit Components), found a typical range of -0.3 to -0.5 (that is, for every 1% decrease in travel time, ridership increases by approximately 0.3 to 0.5%) [101]. The TCRP Report 95 chapter on Bus Rapid Transit, where travel time elasticities will probably be covered, has not yet been published. Assuming some baseline travel time that passengers would be satisfied with, additional travel time above this baseline value would be less satisfactory, while a reduction in travel time would be more satisfactory. The relative sat- isfaction of passengers associated with a given travel time can be expressed in terms of the ridership expected at the actual travel time, relative to the ridership that would occur at the baseline travel time. For example, a route with Q E PQ E P Q E P E P 2 1 1 2 1 2 1 1 1 1 1 = −( )( ) − +( )( ) −( )( ) − +( )( )

76 0 2 4 6 8 10 12 14 16 18 20 100M+ 50-100M 25-50M 10-25M 5-10M 2.5-5M 1-2.5M 0.5-1M 0.25-0.5M <0.25M Annual Boardings A ve ra ge B us S pe ed (m ph ) 0 10 20 30 40 50 60 70 80 90 100 N um be r o f A ge nc ie s Exhibit 83. Average System Bus Speed by Number of Annual Boardings. an average passenger travel time of 25 minutes would be ex- pected to have 10% higher ridership than a route with an average passenger travel time of 30 minutes, all other things being equal, assuming an elasticity of -0.5. (This value is calculated using Equation 2, rather than by taking a 16.7% difference in travel times and multiplying by 0.5). There- fore, if a 30-minute travel time was set as the baseline, the value of fptt would be 1.10 for 25-minute travel times, as- suming for the moment no other influences on perceived travel time. Because urban street LOS focuses on the quality of urban street segments, rather than the bus trip as a whole, the alter- native transit LOS model works with travel time rates (e.g., 6 minutes per mile) instead of travel times (e.g., 30 minutes) or travel speeds (e.g., 10 mph). Travel time rates are the inverse of speeds and, over a given distance, change at the same rate that travel times do. For example, if a bus’ travel time to cover 2 miles decreases from 12 to 11 minutes, the travel time de- creases by 8.3% and so does the travel time rate (from 6 min- utes per mile to 5.5 minutes per mile). Because the rate of change is the same, travel time elasticities should also apply to changes in travel time rates. The fptt factor serves to increase or decrease LOS when tran- sit service is particularly fast or slow, compared with some neutral, baseline value. The fptt and fh factors, in combination, produce a TransitWaitRideScore that represents the percent increase in ridership for a particular headway and perceived travel time rate, compared with a baseline of 60-minute ser- vice at a baseline speed. Selection of a Baseline Travel Time Rate Originally 6 minutes/mile (10 mph) was proposed as a preliminary value for the baseline travel time rate, based on LOS ranges given in TCRP Report 26 [102]. Testing of the preliminary model using real-world data for the entire TriMet bus system in Portland, Oregon, found that the travel time rate of 6 minutes/mile resulted in LOS ratings for Port- land that were too high (62-69% of all street segments ended up as LOS A). The National Transit Database (NTD) can be used to cal- culate a systemwide average bus speed, in terms of revenue miles operated divided by revenue hours operated. “Revenue service” consists of the time a bus is in passenger operation— it does not include travel to or from the bus garage, but does include driver layover time at the end of the route. To elimi- nate the effect of layover time on average speed, we assumed that layover time was 10% of total revenue hours, which is a typical transit industry standard, although local contracts with the bus drivers’ union may specify a different value. For all bus systems reporting to the NTD, the median speed is 15.2 mph. As shown in Exhibit 83, average speed is a function of city size: the larger the city, the lower the speed. For the seven largest bus operators (serving 100 million or more annual boardings), the mean speed was 12.3 mph; for the 60 smallest bus operators reporting to the NTD (10 or more buses in service and fewer than 250,000 annual board- ings), the mean speed was 18.3 mph. The 33 bus operators in the 10 to 25 million boardings category (Montgomery

77 County, Maryland, and Cincinnati, Ohio are toward the top of this group) had a mean speed of 14.8 mph, while the 91 bus operators in the 1 to 2.5 million boardings category (Appleton, Wisconsin, and Pueblo, Colorado are toward the bottom of this group) had a mean speed of 16.0 mph. In total, 247 bus agencies out of 468 reporting (53%) are represented by groups with mean speeds within 1 mph of the median speed of 15.2 mph; thus, this median value is representative of most U.S. bus agencies. When a baseline travel time rate of 4 minutes/mile (15 mph) was tested against the Portland data, a much better distribution of LOS grades was obtained, with only 30 to 39% of route segments receiving LOS A grades, depending on whether route-average or segment-specific speeds were used in the calculation. However, a baseline travel time rate of 4 minutes/mile, when applied to the San Francisco surveys, results in LOS grades that were too low relative to the frequency of service provided. This suggests that a different baseline travel time rate may be ap- propriate for dense urban areas such as San Francisco or down- town Washington, DC. Further testing of the speed elasticity used in the model would also be appropriate. Reliability One way that the TCQSM measures transit reliability is through the coefficient of variation of headway deviations—the standard deviation of headway deviations divided by the mean scheduled headway. (A headway deviation of a given bus is the actual headway minus the scheduled headway. When buses arrive exactly on schedule every time, cvh = 0; when two buses consistently arrive together, cvh = 1.) Some believe that a better reflection of headway reliability from a passenger point-of-view is given by excess wait time (e.g., Furth and Miller (previously cited) and Transport for London’s transit performance standards), which is the aver- age additional time a passenger must wait for a bus to arrive because of non-uniform headways. When passengers arrive randomly at a stop—the case when service is relatively fre- quent (the TCQSM suggests this occurs at headways of 10 to 12 minutes or less)—the average passenger will wait half a headway for a bus to arrive. When a bus is late, passengers will wait longer than half a headway on average. The difference in these two times is the excess wait time. For random passenger arrivals, excess wait time is calcu- lated as half the scheduled headway multiplied by the square of the coefficient of variation of headways (or headway devia- tions) [103, 104]. For non-random arrivals (i.e., for longer headways, when passengers would be expected to be familiar with the schedule and arrive a few minutes before the sched- uled departure time), excess wait time is the average number of minutes that buses are behind schedule. Buses more than a minute early can be treated as being one headway behind schedule, because passengers arriving near the scheduled de- parture time would have to wait for the next bus. Excess wait time adds to a passenger’s overall wait time; it also affects a passenger’s perceived wait time. A common value in the liter- ature is that passengers perceive or value wait time approxi- mately twice as much as in-vehicle time; Furth and Muller suggest a value of 1.5 and suggest accounting for “potential wait time,” an allowance a rider makes to show up earlier for service known to be unreliable, with a value of 0.75 of in- vehicle time [98]. Excess wait time makes a passenger’s trip take longer than intended (i.e., the perceived speed or travel time rate for the trip is slower). However, the effect of excess wait time on the travel time rate varies depending on the length of the trip: a 2-minute excess wait has a bigger proportional effect on a 10-minute trip than a 2-minute wait for a 20- or 30-minute trip. The difficulty is in determining what an appropriate trip length should be. The recommended solution is to compare the excess wait time with the average trip time. The NTD provides information on weekday boardings and passenger miles by mode each year for most transit systems; an average trip length (miles/ boarding) can be computed from these two variables. (This calculation assumes that trip lengths are consistent throughout the day, which may or may not be the case. Passenger miles are only available as daily values. Although the NTD allows agen- cies to report boardings in smaller time increments than a day (e.g., AM peak, midday, etc.), most choose not to.) Dividing the excess wait time (minutes) by the computed average trip length (miles) provides the average effect on the overall travel time rate (minutes/mile), which can then be converted to a perceived travel time rate. For example, if the analysis were being performed in Portland, the average week- day passenger miles in 2003 were 765,100, while the average weekday boardings were 214,158, resulting in an average trip length of 3.57 miles. If the average excess wait time was 2 min- utes, the additional travel time rate would be (2 / 3.57) = 0.56 minutes/mile. The perceived additional travel time rate could be up to twice this value, or 1.12 minutes/mile. Effect of Stop Amenities on Perceived Waiting Time Research presented in TRL Report 593 suggests that certain stop amenities, including shelters, lighting, and seating, can reduce perceived journey time by providing a more comfort- able waiting environment. Exhibit 84 presents these values, converted from pence to in-vehicle time, using an in-vehicle time value of 4.2 pence per minute. Some authors have suggested that real-time displays at bus stops showing the number of minutes until the next bus arrival

78 Crowding Penalty (pence/min) Load Factor (p/seat) Seated Passengers Standing Passengers 0.80 0.0 -- 0.90 0.4 -- 1.00 0.8 6.5 1.10 1.2 7.0 1.20 1.6 7.5 1.30 2.0 8.0 1.40 2.4 8.5 1.50 -- 9.0 1.60 -- 9.5 NOTE: The baseline value of in-vehicle time for rail passengers is 7.2 pence/minute, in 2000 prices. Intermediate values are obtained through linear interpolation. The baseline value of in-vehicle time for bus passengers is 4.2 pence/minute; no corresponding crowding penalties are available. SOURCE: Derived from Balcombe [106]. Exhibit 85. British Crowding Penalties for Rail Passengers. Amenity In-Vehicle Time Value (min) Shelter with roof and end panel Shelter with roof 1.3 1.1 Lighting at bus stop 0.7 Molded seats Flip seats Bench seats 0.8 0.5 0.2 SOURCE: Calculated from Steer Davies Gleave, Bus passenger preferences. For London Transport buses. (1996) in Balcombe, R. (editor), [105]. Exhibit 84. In-Vehicle Time Value of Stop Amenities. should reduce, if not eliminate, the perceived travel time penalty. (In other words, perceived excess wait time would be the same as actual excess wait time when real-time bus-arrival information is provided.) Although this seems to be a reason- able hypothesis, to date, the project team has not seen any literature documenting such an effect. Crowding Perceived Travel Time Effects In the same way that passengers perceive or value wait time more than in-vehicle time, passengers also perceive or value time spent in crowded conditions more than time spent in uncrowded conditions. Exhibit 85 presents values of train crowding used in Great Britain. To demonstrate how the Exhibit 85 information could be applied, consider a situation where a train is operating with a passenger load that is 120% of its seating capacity (i.e., a load factor of 1.20). In the absence of crowding, rail commuters value in-vehicle time at 7.2 pence/minute. At a load factor of 1.20, seated commuters experience a penalty of 1.6 pence/minute due to the more crowded conditions, while standing commuters experience a penalty of 7.5 pence/minute (i.e., standees perceive their time spent standing as being more than twice as onerous as being seated in an uncrowded car- riage). The weighted average penalty for all passengers in the train would be 2.58 pence/minute, corresponding to a value of time 36% higher than in uncrowded conditions. This value- of-time factor, 1.36, corresponds to factor a1 in the perceived travel time rate equation. In application, the transit LOS model would provide a lookup table based on a similar calcu- lation using bus passenger value-of-time to directly provide a1. Section 6.10 provides an example of such a lookup table. No corresponding British values exist for bus crowding penalties and no American work was found that could be used to compare U.S. rail crowding perceptions to U.K. per- ceptions. Additional research would be needed to establish U.S. crowding penalty values. In the absence of other data, the recommended transit model uses a combination of U.K. bus value-of-time data and rail crowding penalties. Load Variability Effects Unreliable operations tend to result in higher levels of crowding on buses that are running late, because these buses pick up not only their own passengers, but passengers who have arrived early for the following bus. For existing- condition analyses, this crowding can be measured directly. For future-condition analyses, for frequent service (i.e., head- ways approximately 10 minutes or less), this additional crowding can be estimated as the mean load multiplied by (1 + cvh) (Derived from the TCQSM, Part 4, Appendix E, Equa- tions 4-22 and 4-23). At long headways, a late bus will pick up its normal load, because passengers will have timed their ar- rival at the bus stop to the expected departure time (Furth and Miller, previously cited). There is also an intermediate range of headways with a mix of randomly and non-randomly ar- riving passengers [107]. This late-bus load can be used in the perceived travel time calculation described in the previous section, thus incorporating the effect of unreliable service’s crowding into the LOS measure. The literature review uncovered no information directly relating overcrowding and/or reliability to transit demand. In the United States, the San Francisco County Transporta- tion Authority appears to be the only agency to have docu- mented a test of reliability and crowding factors for use in a travel demand model. The tested values were based on a stated-preference telephone survey, but were not incorpo- rated in the final model because the predicted number of boardings did not reasonably match the observed number of boardings [108]. 6.2 Recommended Transit LOS Model The recommended transit LOS model predicts the average quality of service rating that transit riders would give the bus service on an urban street. The model is as follows:

79 LOS Numerical Score A ≤ 2.00 B >2.00 and ≤ 2.75 C >2.75 and ≤ 3.50 D >3.50 and ≤ 4.25 E >4.25 and ≤ 5.00 F > 5.00 Headway (min) Frequency (bus/h) fh 60 1 1.00 45 1.33 1.33 40 1.5 1.50 30 2 2.00 20 3 2.44 15 4 2.80 12 5 2.99 10 6 3.16 7.5 8 3.37 6 10 3.58 5 12 3.79 NOTE: The following frequency elasticities are assumed: +1.0 for 1-2 buses/hour, +0.5 for 2-4 buses/hour, +0.3 for 4-6 buses/hour, and +0.2 for 6 or more buses/hour. Elasticities derived from data reported in TCRP Report 95, Chapter 9. Exhibit 87. Headway Factor Values. Exhibit 86. Transit LOS Thresholds. Transit LOS Score = 6.0 − 1.50 * TransitWaitRideScore + 0.15 * PedLOS (Eq. 23) Where PedLOS = The pedestrian LOS numerical value for the facility (A=1, F=6). TransitWaitRideScore = The transit ride and waiting time score, a function of the average headway between buses and the perceived travel time via bus. The computed transit LOS score is converted to a letter LOS grade using the equivalencies given in Exhibit 86. These are the same thresholds as used for auto. Estimation of the Pedestrian LOS The pedestrian LOS for the urban street is estimated using the pedestrian LOS model described in a later chapter. Estimation of the Transit Wait Ride Score The transit wait and ride score is a function of the headway between buses and the perceived travel time via bus for the urban street. TransitWaitRideScore = fh * fptt (Eq. 24) Where fh = headway factor = the multiplicative change in ridership expected on a route at a headway h, relative to the rider- ship at 60-minute headways; fptt = perceived travel time factor = the multiplicative change in ridership expected at a perceived travel time rate PTTR, relative to the ridership expected at a baseline travel time rate. The baseline travel time rate is 4 minutes/mile except for central business districts of metropolitan areas with over 5 million population, in which case it is 6 min/mile. Exhibit 87 provides fh values for typical bus headways. The perceived travel time factor is estimated based on the perceived travel time rate and the expected demand elasticity for a change in the perceived travel time rate. (Eq. 25)F e BTTR e TTR e TTR e BTT PTTR = −( ) − +( )[ ] −( ) − +( ) 1 1 1 1 R[ ] Where F(PTTR) = Perceived Travel Time Factor PTTR = Perceived Travel Time Rate (min/mi) BTTR = Base Travel Time Rate (min/mi). Use 6 minutes per mile for the main central business district of metropolitan areas with population greater than or equal to 5 million. Use 4 minutes per mile for all other areas. e = ridership elasticity with respect to changes in the travel time rate. The suggested default value is −0.40, but local values may be substituted. Exhibit 88 below illustrates the application of this equation for selected perceived travel time rates and a selected elasticity. The perceived travel time rate (PTTR) is estimated based on the mean speed of the bus service, the average excess wait time for the bus (due to late arrivals), the average trip length, the average load factor for the bus service, and the amenities at the bus stops. F(PTTR) BTTR: 4 min/mi 6 min/mi PTTR (min/mi) 2 1.31 1.50 2.4 1.22 1.41 3 1.12 1.31 4 1.00 1.17 6 0.85 1.00 12 0.67 0.76 30 0.53 0.58 Notes: • F(PTTR) = Perceived Travel Time Factor • PTTR = Perceived Travel Time Rate. • BTTR = Base Travel Time Rate (default is 4 minutes per mile. 6 minutes per mile BTTR is used for the central business districts (CBDs) of metropolitan areas with 5 million or greater population). • Based on default value of –0.40 for elasticity. Exhibit 88. Example Perceived Travel Time Factors (F(PTTR)).

80 Load Factor (pass/seat) a1 0.80 1.00 default 1.00 1.19 1.10 1.41 1.20 1.62 1.30 1.81 1.40 1.99 1.50 2.16 1.60 2.32 Notes: Load factor is the average ratio of passengers to seats for buses at the peak load point within the study section of the street.If bus load factor is not known, a default value of 1.00 can be assumed for the load weighting factor (a1). Exhibit 89. Passenger Load Weighting Factor (a1). PTTR= a1 * IVTTR + a2 * EWTR − ATR (Eq. 26) Where PTTR = perceived travel time rate. IVTTR = actual in-vehicle travel time rate, in minutes per mile. EWTR = excess wait time rate due to late arrivals = excess wait time (minutes) / average trip length (miles). a1 = passenger load weighting factor (a function of the average load on buses in the analysis segment dur- ing the peak 15 minutes). a2 = 2 = wait time factor, converting actual wait times into perceived wait times. ATR = amenity time rate = perceived travel time rate re- duction due to the provision of certain bus stop amenities = in-vehicle travel time value of stop amenities (minutes) / average trip length (miles). In-Vehicle Travel Time Rate The in-vehicle travel time rate is equal to the inverse of the mean bus speed converted to minutes per mile. (Eq. 27) Where IVTTR = In-Vehicle Travel Time Rate (min/mi)r. Speed = Average speed of bus over study section of street (mph). When field measurement of mean bus speed is not feasible, the mean schedule speed can be used. Identify two schedule points on the published schedule for the bus route(s). Mea- sure the distance covered by the bus route(s) between the two points. Divide the measured distance by the scheduled travel time between the two schedule points. The bus speed estima- tion procedure given in Chapter 27 (Transit) may be used to estimate future bus speeds. The in-vehicle travel time rate is multiplied by a passenger load weighting factor (a1) to account for the increased dis- comfort when buses are crowded. Values of the passenger load weighting factor (a1) are given in Exhibit 89. Excess Wait Time Rate The excess wait time is the sum of the differences between the scheduled and actual arrival times for buses within the study section of the street divided by the number of observa- tions. Early arrival without a corresponding early departure is counted as being on-time. However, early arrival with an early departure is counted as being “one headway” late for the pur- poses of computing the average excess wait time for the street. IVTTR Speed = 60 The excess wait time rate is the excess wait time (in min- utes) divided by the mean passenger trip length for the bus route(s) within the study section of the street. For average passenger trip length, a default value can be taken from national average data reported by the American Public Transit Association (APTA) http://www.apta.com/ research/stats/ridership/trlength.cfm ). In 2004, the mean trip length for bus passenger-trips nationwide was 3.7 miles. More locally specific values of average trip length can be obtained from the NTD. Look up the annual passenger miles and annual unlinked trips in the transit agency profiles stored under NTD Annual Data Publications at http://www.ntd program.gov/ntdprogram/pubs.htm#profiles. The mean trip length is the annual passenger-miles divided by the annual unlinked trips. Amenity Time Rate The amenity time rate is the time value of various bus stop improvements divided by the mean passenger trip length. The mean passenger trip length is the same distance used to compute the Excess Wait Time Rate (described above). (Eq. 28) Where ATR = Amenity Time Rate (min/mi) Shelter = Proportion of bus stops in study section direction with shelters Bench = Proportion of bus stops in study section direction with benches ATL = Average passenger trip length (miles) ATR Shelter Bench ATL = +1 3 0 2. * . *

81 Operator Rte Freq . (bus/h) Spd (mph) OTP % Shelte r (% ) Bench (% ) LF (p/seat) Ped LOS CBD Surv ey LOS FDOT LOS TCQSM LOS Mode l LOS TriMet 14 8 11.8 75% 34% 47% 0.55 C No A A C A TriMet 44 4 14.8 76% 30% 41% 0.83 C No A B D A AC Transit 72R 5 15.7 66% 74% 75% 1.10 D No A B D B AC Transit 72 4 12.1 53% 39% 46% 1.10 D No A B D B WMATA 38B 4 10.1 46% 29% 26% 0.38 D No A C D B WMATA 2B 2 14.0 67% 13% 15% 1.10 D No A E D D AC Transit 218 1 15.1 72% 11% 15% 1.10 C No A E E E AC Transit 51 8 11.8 54% 28% 51% 1.10 D No B A C A SF Muni 14 10 9.2 57% 54% 56% 1.30 E Ye s B A C C SF Muni 30 7 7.4 59% 44% 44% 1.30 E Ye s B A C D SF Muni 1 20 8.8 63% 44% 44% 1.30 C Ye s B A C A SF Muni 38 8 9.8 59% 68% 69% 1.30 F Ye s B B C D SF Muni 38L 9 12.1 48% 84% 86% 1.10 F No B B C A Broward 18 4 13.6 65% 23% 75% 1.10 E No B C D B % Exact Match 100% 21% 0% 21% % Within 1 LOS 100% 86% 43% 71% Notes: 1. OTP = on time performance with 5 minutes late considered on-time. 2. LF = load factor 3. Shelter = percent of bus stops with shelters. 4. Bench = percent of bus stops with benches. 5. Survey = the mean level of service reported in the field survey. 6. FDOT = Florida Quality/Level of Service Handbook method. 7. TCQSM = Transit Capacity and Quality of Service Manual. The TCQSM does not produce a single letter grade LOS for transit routes. The letter grade reported here is an average, a grade point average (GPA) of the numerous LOS ratings that the TCQSM reports for any given transit route. 8. Model LOS = the letter grade predicted by the recommended transit LOS model. Exhibit 90. Evaluation of Proposed Transit Model and TCQSM against Field Survey Results. Notes: 1. Shelters with benches are counted twice—once as shelters, once as benches. 2. Coefficients adapted from Steer Davies Gleave, Bus pas- senger preferences. For London Transport buses. (1996) in Balcombe, R. (editor) [109]. 6.3 Performance of Transit LOS Model Exhibit 90 compares the ability of the existing TCQSM LOS models and the proposed transit LOS model to predict the mean LOS response for each bus route obtained from the field surveys. None of the models reproduce the mean levels of service reported by passengers in the on-board surveys very well. Both the FDOT LOS and proposed LOS model match the passen- ger surveys about 21% of the time. Although a better match might have been desirable, the on-board survey results indi- cate a high degree of acceptance for a wide range of conditions. It is thought that passengers not satisfied with the service are less likely to ride the buses and thus were undersampled in the survey. Consequently, it was considered acceptable that the proposed transit LOS model should predict poorer levels of service than obtained in the on-board surveys. The scope of the TCQSM LOS model is quite a bit differ- ent than the urban street. The TCQSM is designed to repre- sent the entire trip, while this research is limited to transit service on a given street. Also, the TCQSM provides six dif- ferent letter grade levels of service, depending on the geo- graphic scope and aggregation of the analysis. Only the worst result is shown in the table. Both the FDOT LOS model and the proposed transit LOS model predict a range of LOS A to E for the transit routes sur- veyed. All three LOS models, FDOT, TCQSM, and the pro- posed transit LOS model, tend to agree that WMATA Route 2B and AC Transit Route 218 are LOS D/E, which passengers rated as LOS A.

Next: Chapter 7 - Bicycle LOS Model »
Multimodal Level of Service Analysis for Urban Streets Get This Book
×
MyNAP members save 10% online.
Login or Register to save!
Download Free PDF

TRB’s National Cooperative Highway Research Program (NCHRP) Report 616: Multimodal Level of Service Analysis for Urban Streets explores a method for assessing how well an urban street serves the needs of all of its users. The method for evaluating the multimodal level of service (MMLOS) estimates the auto, bus, bicycle, and pedestrian level of service on an urban street using a combination of readily available data and data normally gathered by an agency to assess auto and transit level of service. The MMLOS user’s guide was published as NCHRP Web-Only Document 128.

Errata

In the printed version of the report, equations 36 (pedestrian segment LOS) and 37 (pedestrian LOS for signalized intersections) on page 88 have been revised and are available online. The equations in the electronic (dpf) version of the report are correct.

In June 2010, TRB released NCHRP Web-Only Document 158: Field Test Results of the Multimodal Level of Service Analysis for Urban Streets (MMLOS) that explores the result of a field test of the MMLOS in 10 metropolitan areas in the United States.

  1. ×

    Welcome to OpenBook!

    You're looking at OpenBook, NAP.edu's online reading room since 1999. Based on feedback from you, our users, we've made some improvements that make it easier than ever to read thousands of publications on our website.

    Do you want to take a quick tour of the OpenBook's features?

    No Thanks Take a Tour »
  2. ×

    Show this book's table of contents, where you can jump to any chapter by name.

    « Back Next »
  3. ×

    ...or use these buttons to go back to the previous chapter or skip to the next one.

    « Back Next »
  4. ×

    Jump up to the previous page or down to the next one. Also, you can type in a page number and press Enter to go directly to that page in the book.

    « Back Next »
  5. ×

    To search the entire text of this book, type in your search term here and press Enter.

    « Back Next »
  6. ×

    Share a link to this book page on your preferred social network or via email.

    « Back Next »
  7. ×

    View our suggested citation for this chapter.

    « Back Next »
  8. ×

    Ready to take your reading offline? Click here to buy this book in print or download it as a free PDF, if available.

    « Back Next »
Stay Connected!