1
Daniel I Rubins
Susan Handy
Bicycle Clearance Times: A Case Study of the City of Davis
Daniel I. Rubins
University of California, Davis
Institute of Transportation Studies
Department of Civil and Environmental Engineering
One Shields Avenue
Davis, CA 95616-8762
Phone: 530-752-2570
Fax: 530-752-6572
E-mail:
Susan Handy
University of California, Davis
Institute of Transportation Studies
Department of Environmental Science and Policy
One Shields Avenue
Davis, CA 95616-8762
Phone: 530-752-5878
Fax: 530-752-3350
E-mail:
Draft Submitted to TRB for 2005 Annual Conference, Washington D.C.
Date Submitted: August 1, 2004
Word Count: 5210
Abstract
The current state of practice for traffic signal timing does not account for minimum green time or clearance interval for bicycles. Like pedestrians, bicyclists need sufficient time to cross an intersection safely. However, this need must be balanced against possible delays for motorist traffic. Accurate estimates of clearance times for bicyclists are thus essential to the safe and efficient design of traffic signals. This research project presents data on bicycle clearance times for different intersection distances near the University of California at Davis Campus and provides a methodology for measuring bicycle clearance times that other researchers can use. A total of 10 signalized intersections with varying motor and bicycle traffic volumes were videotaped for a total of approximately 11 hours. Observed clearance times and calculated speeds for standing, rolling, and quasi rolling starts are presented in this paper. The importance of physical design of intersections is also discussed. Important findings are that the clearance times vary widely for each intersection distance and that the 2nd percentile speeds for each start type are much slower than the speeds recommended for use in signal design by AASHTO.
InTRODUCTION
The design of a traffic signal for motor vehicle and bicycle traffic is a trade-off between the safety of bicyclists and convenience for motor vehicles (1). Like pedestrians, bicyclists need sufficient time to cross an intersection safely (2). The critical times for bicyclists at signalized intersections are at the beginning and end of the green periods: a properly designed signal provides sufficient time for a bicyclist entering an intersection at the end of the green phase and for a bicyclist starting-up at the beginning of the green phase (3). However, this need must be balanced against possible delays for motorist traffic: if the green phase is lengthened to accommodate bicyclists, then the delays for cross traffic may increase. Accurate estimates of clearance times for bicyclists are thus essential to the safe and efficient design of traffic signals.
Limited guidance and empirical evidence exists for the design of traffic signals to accommodate bicycles. AASHTO states that 98 percent of bicyclists should be able to clear an intersection and recommends, in the absence of data on bicyclist speeds, assumptions of 12.7 ft/sec for advanced bicyclists, 12.0 ft/sec for basic bicyclists, and 9.1 ft/sec for children (3). Several studies consider elements of bicycle clearance time, including comfortable deceleration and accelerations (4, 5). Other research has shown that the design clearance point is often well within the limit line, meaning that there is not sufficient clearance time for bicyclist to safely cross an intersection (4, 6, 7). Dean and Davis note in their review of current bicycle research that the bicyclist speed distribution studies by Forester (2), Taylor (4), Pein (5), and Opiela et. al. (8) do not relate bicyclist speed to roadway environment characteristics other than vertical grade profile. A study by Landis, et. al. uses width of outside lane, intersection distance, volume of directional traffic and total number of through lanes to determine the perceived hazard of shared-roadway environment through an intersection (9). The Manual of Uniform Traffic Control Devices (MUTCD) and Highway Capacity Manual (HCM) use motor vehicle equivalents to incorporate the impedance of bicycle traffic into signal design (10, 11). Allen et. al. highlight the need for research to quantify the effects of bicycles on the capacity of signalized intersections (12), a development that would enable the creation of a composite level of service for mixed flow of intersections and eliminate the use of motor vehicle equivalents for bicyclist (9, 13, 14).
This research project presents data on bicycle clearance times for different intersection distances near the University of California at Davis Campus and provides a methodology for measuring bicycle clearance times that other researchers can use. The City of Davis maintains an integrated system of bike lanes (48.8 miles) and bike paths (49 miles) that serves as the primary commute mode of transportation for students, faculty and staff living in Davis (15). Recent city data (1990) indicates bicycles account for approximately 20-25 percent of trips by city residence (15). The 2000 Census, indicates that 14 percent of Davis residence usually bike to work (16). The City of Davis has been nationally recognized by the League of American bicyclist as a “Bicycle Friendly Community” and earned the title of “America’s Best Cycling City” from the Bicycle Federation of America (15). The City of Davis has also been the subject of a number of transportation studies beginning in the late 1970s, though none focused on bicycle clearance times (1, 7, 12, 17, 18). In this project, a total of 10 signalized intersections with varying motor and bicycle traffic volumes were videotaped for a total of approximately 11 hours. This paper discusses the data collection procedure, results and discussion of improvements and qualitative observations.
Data Collection Procedure
In selecting intersections, we used the following selection criteria, similar to those used in a study by Opiela, Khasnabis and Datta (8):
§ Sufficient bicycle volume.
§ A variety of clearance distances.
§ An acceptable vantage point for the video recorder (e.g., set back from the intersection).
We identified 14 intersections to be videotaped and ultimately recorded approximately one hour at each of the 10 signalized intersections near the University of California at Davis Campus. The observations were made during the morning (7:30-9:30am) and evening (4-6pm) peak periods using digit video cameras with wide-angle lenses. The weather was mild for all observations and each intersection had flat approaches. The majority of the observed bicyclists were young adults because of the proximity of the intersections to the University of California at Davis Campus. Table 1 summarizes for each intersection the number of observations, range of intersection distances, range of observed speeds, and distance from campus.
The intersection distance is defined as the distance from the first crosswalk line to the first line on the other side of the intersection (third line encountered rather than the limit line on the far side). This definition was chosen because most bicyclists stop at the first crosswalk line at red lights and because bicyclists are safely out of the path of cross traffic when they cross the third line. Figure 1 shows an overhead view of the Sycamore and Russell intersection; the distances of the observed movements and camera location are show as well. Distances for each movement for each intersection were measured based on intersection diagrams provided by the City of Davis. For the 10 intersections, distances for through movements ranged from 32 feet to 117 feet while distances for left turn movements ranged from 54 feet to 130 feet. Note that not all turning movements were observed at all intersections due to obstructions based on the camera positions at several of the intersections.
To extract clearance times from the videotapes, a team of undergraduate students manually reviewed the videotapes. Time and resource constraints precluded the use of a computer-aided video capture and analysis procedure that tracks the vehicle movements and transforms the positions of vehicles to street coordinates (7). The manual review process involved manually starting and stopping the videotape for each bicyclist and recording the time that the bicyclist entered the intersection (crossed the first crosswalk line) and exited the intersection (crossed into the crosswalk on the far side of the intersection). The process did not involve dividing the clearance time into acceleration and cruise portions of a bicyclist through an intersection (5) but did differentiate between three important start types: bicyclists that entered at full speed were defined as “rolling starts”; bicyclists fully stopped at a red light with at least one foot on the ground were defined as “standing starts”; and bicyclists stopped at a red light with at least one foot on the ground several bike lengths from the first crosswalk line were defined as “quasi starts.” In addition to clearance time and start type, characteristics recorded from the videotapes included bicycles with trailers, approximate age of bicyclist, and gender of bicyclist. Special observations about interactions between bicyclists, pedestrians, automobiles and intersection design were also noted, and most illegal bicycle movements, such as diagonal travel across the intersection, were excluded from the analysis. Travel speed was calculated for each observation based on the recorded clearance time and the measured distance for the intersection movement.
As a test of the reliability of the method for extracting clearance times from the videotapes, each student extracted crossing times for the same 10-minute tape sample. This exercise produced an intraclass correlation coefficient of 0.64 for 9 students and 31 observations, showing that the clearance times recorded by different students were reasonably consistent. Probable reasons for differences in recorded clearance times include difficulty in seeing stopping and starting points and the display of times on the videotape only to the full second with 30 frames per second.
Results
For the overall sample there is a wide variation in speed from the slowest bicyclist at 2.9 ft/sec to the fastest bicyclist at 33 ft/sec. The slowest observed speed was for a bicyclist who sped up at the start of the intersection then slowed down on the far side of the intersection in response to the curb design and a queue of waiting bicycles, illustrating the importance of considering the role of the roadway environment in determining clearance times. The fastest observed speed was for an advanced bicyclist on road race bike, illustrating the importance of considering bicyclist ability. The range of speeds for the entire sample and for each of the ten intersections can be seen in Figure 2; the grouping of data points at certain speeds for each intersection is an effect of the rounding of recorded clearance times to full seconds. The descriptive statistics for all observations and for each start type are summarized in Table 2.
Overall, with outliers, the mean speed is 13.5 ft/sec and the median is 12.5 ft/sec. The mean is larger than the median because the distribution is skewed to the right (skewness > 0) and tails are larger than a normal distribution (kurtosis > 0). The relatively small standard error of the mean, 0.106, generates confidence in the central tendency of the observed speeds. The large standard deviation, 4.85 ft/sec, and variance, 23.54 ft2/sec2, suggest a broad range for the observed speeds with 95% of bicyclist falling between 3.79 ft/sec to 23.20 ft/sec. The 85th percentile for all speed observations is 18.57 ft/sec.
A comparison of mean speeds by start type shows that acceleration for bicyclists is not instantaneous as is commonly assumed for pedestrians when average speeds of 3-4 ft/sec are used in signal design (19). Standing starts have the slowest mean, 11.24 ft/sec, while rolling starts has the fastest mean speed, 16.03 ft/sec. As expected the mean of quasi starts, 11.93 ft/sec, is slightly faster than for standing starts. Figure 3 shows that the standing and quasi start types make up a greater share of low speed observations than do rolling starts. The differences between the mean values for the different start types, including the difference between standing starts and quasi starts, are statistically significant at the 1 percent significance level. As noted above, AASHTO states that 98 percent of bicyclists should be able to clear an intersection (3). For this sample, the 2nd percentile speeds are 5.3 ft/sec, 6.4 ft/sec and 6.7 ft/sec for standing, rolling, and quasi start type, respectively. These speeds are considerably slower than the assumed speeds recommended by AASHTO and suggest that longer clearance times should be used in signal design to ensure that 98 percent of bicyclists will be able to clear an intersection during the green phase.
The relationship between clearance time and intersection distance was estimated for each type of start using linear regression. The equations enable an estimation of the expected clearance time for a given intersection distance for a typical bicyclist and can be interpreted as consisting of an initial start-up time (the intercept) and an average speed once moving (the inverse of the coefficient). As the intersection distance increases, the start-up penalty becomes a smaller portion of the clearance time. Using this interpretation, the average speed for standing start is 17.7 ft/sec, for rolling start is 23.8 ft/sec, and for quasi start is 12.2 ft/sec, while the time penalties are 3.1 seconds for standing starts, 2.1 seconds for rolling starts, and 0.5 seconds for quasi starts. A smaller intercept for quasi starts than rolling starts seem counterintuitive but may be explained by the few long clearance time outliers for the rolling observations. These equations were estimated using data for movements between 32 and 139 feet and therefore the validity of these equations is questionable for smaller or larger movements. For the entire sample, the adjusted R-square is a modest 0.27; the adjusted R-squares for start types were 0.25 for rolling starts, 0.29 for standing starts, and 0.52 for quasi starts. It is important to note that there is a large amount of variation in the clearance times for each intersection distance for each start type.
discussion
Although the results presented have important limitations and should be replicated through additional studies, they begin to fill a notable gap in data on bicycle clearance times for signalized intersections. The data produced in this project can be used to develop guidelines and estimate clearance times as a function of intersection distance. Given the significant variation in clearance times, even for a relatively homogeneous sample of bicyclists, a conservative approach would be to use the 2nd percentile of speed for standing starts of 5.33 ft/sec. Alternatively, the regression equations presented in this paper can be used to estimate the average clearance time for an intersection of a specific distance, accounting for the time required for the bicyclist to accelerate.