Skip to contentUnited States Department of Transportation - Federal Highway Administration FHWA Home
Research Home
Report
This report is an archived publication and may contain dated technical, contact, and link information
Publication Number: FHWA-HRT-05-137
Date: July 2006

Evaluation of Safety, Design, and Operation of Shared-Use Paths

Final Report

PDF Version (1172 KB)

PDF files can be viewed with the Acrobat® Reader®

 

CHAPTER 4. OPERATIONAL DATA COLLECTION

INTRODUCTION

One of the major objectives of this research was that the LOS methodology be calibrated and validated using data collected from the United States. The challenges facing the project team included a choice of methodology to allow the efficient collection of high-quality data on the operations of shared-use paths and the selection of study paths to provide a nationwide sample of representative paths. This chapter describes how the researchers met those challenges. Chapter 5 provides the results from the operational data collection effort.

 

DATA COLLECTION METHOD

The model for estimating the number of meetings and passings experienced by a test bicyclist developed in depth in chapter 3 uses the volume, average speed, and standard deviation around the average speed of each mode on the path as inputs. The data collection to calibrate and validate this model must therefore involve all of these variables. Of course, trail characteristics must also be recorded at each site. To ensure later flexibility, it was also desirable that scenes on paths of interest be recorded from different perspectives so that additional data could be obtained later by viewing videotapes if needed.

The project team identified three possible methods of operational data collection, which included a one-camera method, a two-camera method, and a moving-bicycle method. The one-camera method involved a camera on a high perch that was able to record activity, including meetings and passings, on a long segment of path. The two-camera method involved two cameras set up several thousand feet apart along a path. From each camera, a sequence of users could be determined and, from those sequences, meetings and passings could be discerned. In the end, however, we concluded that the one-camera and two-camera methods would not provide adequate data, so we chose the moving-bicycle method (described below). Vantage points for the one-camera method would have been rare; tall buildings and hills with unobstructed views of qualifying shared-use paths are not common in the United States. The two-camera method would not have been able to identify the difference between actual passings and desired passings, because only path users would have known whether they wanted to pass and were unable to do so and why. For example, a bicyclist may not have been able to pass because of inadequate path width or congestion.

We chose the moving-bicycle method as our primary operational data collection method because it overcomes the flaws in the other methods. It is not restricted to places where special camera vantage points are available, and it can determine desired passings. The method works by collecting data from the perspective of the bicyclist, and is analogous to the moving-vehicle method of collecting volumes and travel times on the highway (such as described in chapter 15 of the HCM(4)). In the moving-bicycle method, a member of the project team rides a bicycle along a path segment of interest at a predetermined constant speed. The team member attempts to maintain that speed as closely as possible, passing when encountering slower same-direction users when there is sufficient room to do so safely while maintaining that constant speed. The team member wore a video camera on his or her helmet that recorded the number of meetings, the number of passings accomplished, and the number of passings delayed or not accomplished (i.e., reached the end of the segment before the opportunity to pass presented itself). At the same time, as the bicyclist was making his or her ride, a colleague was counting the number of users of each mode in each direction moving past the midway point of the segment. This provided the needed volume data.

A potential bias with the moving-bicycle method is that data collector judgments determine the difference between a desired pass, a following maneuver with no desire to pass, and a completed pass. To prevent this bias from affecting the results, we equipped the bicycle with a mini-computer (a Specialized™ Speed Zone P.Brain) that displayed the bicycle speed to the nearest 0.16 km/h (0.1 mi/h). In addition, the team recorded the time needed to ride the segment from start to finish to verify that the desired speed was maintained.

Moving-bicycle desired speeds were determined for each path based on the results of a prior study of the bicycle speed distribution on the path segment of interest. Typically, the data collection team conducted the prior study on the day before the moving-bicycle study was to commence. Therefore, the data collection process typically took 2 days per trail. On the first day, the team collected an adequate sample of speeds of the free-flowing bicycles (30 minimum) and of other path users using a stopwatch and a clearly marked distance. Then, the team calculated a mean and standard deviation for the bicycle speed sample. The moving-bicycle runs on the second day were typically made at three speed levels–high, medium, and low–that corresponded to the mean speed plus one standard deviation, the mean speed, and the mean speed minus one standard deviation from this prior sample.

On the second day, the team collected the moving-bicycle data. We used a stationary camera on the side of the path at the midpoint, next to the volume data collector, to provide backup for the volume count, to provide additional speed observations, and to provide other data that may have proved necessary later. To avoid fatigue, the two data collectors traded bicycle duty and stationary volume counting duty occasionally.

During the second day of data collection, the team set a goal to collect at least 20 runs at each of the three different speeds, or a total of 60 runs conducted at each path. Since we collected data for both directions along the trail, 60 runs actually provided 10 runs in each direction along the path at the three different speeds. Higher sample sizes would have been desirable, but were not usually possible, because user volumes on the paths of interest did not usually stay high for many hours of the day. We could only collect during daylight hours, and the bicycle riders became fatigued.

Equipment

The main equipment for this data collection included a stationary camera/recorder, a bicycle, a mini video camera/recorder for the bicyclist';s helmet, and a bicycle speedometer. The bicyclist's recorder was carried in a handlebar pouch, where the display was visible to the bicyclist, so that he or she could be sure that the system was recording. A microphone taped to the bicyclist's shirt was incorporated into the mini-camera system to allow the bicyclist to record comments during a run The most helpful of these comments was whether a particular event was a delayed passing or not.

We used a hybrid bicycle, which is a combination of a mountain bicycle and a road bicycle, during our data collection. Hybrid bicycles have a smoother and wider tire than mountain bicycles in order to obtain the higher speeds and increased stability that we needed. The bicycle we used was also easy to disassemble and reassemble for travel by plane, because we attempted to use the same bicycle for all of the different data collection sites to ensure more consistency during the data collection process. In the end, mechanical problems with the bicycle meant that we used a rented bicycle during one of our data collection trips (to Saint Louis, MO).

The mobile camera and recorder system we chose are generally used for surveillance operations. The mini-camera was approximately 50 millimeters (mm) (2 inches) long and 25 mm (1 inch) in diameter. The camera had 360 lines of resolution and a 3.6-mm-wide lens. The recorder was supposedly the world's smallest VCR at the time we purchased it, with an LCD monitor that was about 190.5 mm (7.5 inches) by 114.3 mm (4.5 inches) by 88.9 mm (3.5 inches), weighing about 0.68 kilograms (kg) (1.5 pounds (lb)). The rechargeable camera battery lasted about 2 h. The research team soon developed a routine of changing all batteries and cassette tapes on all cameras and recorders every 2 h or 10 runs to ensure that we kept recording when desired. Total equipment costs for the cameras, bicycle, and accessories were about $3,000.

 

SITE SELECTION

The project budget allowed for operational data collection for up to 20 trails in 10 cities across the United States. This was likely to provide a large enough sample to calibrate and validate the procedure in a credible manner. The project team sought operational data collection sites that met a strict set of criteria to ensure project success. These criteria included:

  1. Sites in most regions of the United States.
  2. Two or more sites in a city in order to reduce travel costs and meet the goal of 20 path segments in 10 cities.
  3. Sites that were well known to trail planners and designers in order to build credibility in the results.
  4. Sites that had moderate to high traffic levels for at least some portions of some days.
  5. Sites that had long segments with no intersections or turnouts (ensuring uninterrupted flow).
  6. Sites where the project team could unload equipment easily from a vehicle.
  7. Sites that had a wide variety of geometric characteristics.
  8. Sites with managers who were willing to cooperate.

Based on the knowledge of the researchers and input from FHWA staff at the February 2001 briefing, the team assembled a preliminary list of possible data collection sites that may have met some or all of these criteria. The sites included:

  1. Raleigh, NC: Shelly Lake Trail, Lake Johnson Trail, and Apex Lake Trail
  2. Hilton Head Island, SC
  3. Jekyll Island, GA
  4. Pinellas County, FL: Pinellas Trail
  5. Gainesville, FL: Gainesville-Hawthorne Trail
  6. Tallahassee, FL: St. Mark’s Trail
  7. Winter Garden and Apopka, FL: West Orange Trail
  8. Orlando, FL: Cady Way Trail
  9. Boston and Cambridge, MA: Charles River Trail
  10. Arlington, Bedford, and Lexington, MA: Minuteman Bikeway
  11. White Plains, NY: North County Trailway and South County Trailway
  12. Manhattan, NY: Hudson River Trail from Battery Park to 125th Street
  13. Brooklyn, NY: Shore Parkway and Ocean Parkway
  14. Philadelphia, PA: Schuylkill River Trail
  15. Washington, DC: Capital Crescent Trail and Washington and Old Dominion Trail
  16. Virginia Beach, VA
  17. Chicago, IL: Lakefront Path
  18. Mackinac Island, MI: M-185
  19. Madison, WI
  20. St. Louis, MO: Forest Park Bike Path and Grant’s Trail
  21. Columbia, MO
  22. Phoenix, AZ
  23. Tucson, AZ
  24. Denver, CO: Platte River Greenway and Cherry Creek Path
  25. Houston, TX: Harrisburg Rail Trail, West White Oak Bayou Trail, West Brays Bayou Trail System, and Buffalo Bayou Trail
  26. Dallas, TX
  27. Davis, CA
  28. Huntington Beach, CA: Bolsa Chica-Huntington Beach
  29. Los Angeles, CA: South Bay Trail
  30. Santa Ana, CA: Santa Ana River Trail
  31. Santa Monica, CA: Ocean Front Walk
  32. San Diego, CA: Ocean Front Walk
  33. San Francisco and Oakland, CA: San Francisco Bay Trail
  34. Portland, OR
  35. Seattle, WA: Burke-Gilman Trail
  36. Boulder, CO
  37. Portland, ME

The project team developed a questionnaire for the owners or managers of the paths listed above to determine the suitability of a particular path for data collection. The questionnaire asked:

  1. Is the trail paved?
  2. What is the length of the shortest uninterrupted segment?
  3. Which user modes use this trail?
  4. What is the predominant user mode?
  5. Are the trail users mainly recreational users or commuters?
  6. Is the trail divided by a median, berm, or pavement striping?
  7. What is the width of the trail?
  8. Would you consider the trail volumes to be high, medium, or low?
  9. What is the peak month of the year?
  10. What is the peak day of the week for the trail?
  11. What is the peak time of the day for the trail?
  12. How would you describe the trail grades (level, rolling, or mountainous)?
  13. Would we be able to park a vehicle near the trail?
  14. Do we need permission to conduct our study along your trail?

The project team sent the questionnaire to the owners or managers of the 37 trails via regular mail and e-mail and received 26 responses. From the responses, the researchers identified a list of 10 cities and an alternate city that provided the best possible opportunities to satisfy the criteria. The list included cities in all regions of the United States and cities with many of the best-known trails in the United States. The final list of sites approved by FHWA was:

  1. Seattle, WA
  2. San Francisco, CA
  3. Boston, MA
  4. Chicago, IL
  5. St. Petersburg, FL
  6. St. Louis, MO
  7. Raleigh, NC
  8. Dallas, TX
  9. Washington, DC
  10. Denver, CO
  11. Los Angeles, CA (alternate)

In the end, we used our alternate city, Los Angeles, and did not collect data in Denver because of travel and weather difficulties.

In the course of this study, the team collected data from 15 trails and 10 cities scattered across the United States. Some cities only had one usable trail. The data collection sites were:

  1. Lake Johnson Trail in Raleigh, NC
  2. Pinellas Trail and Honeymoon Island Trail near St. Petersburg, FL
  3. White Creek Trail and White Rock Lake Trail in Dallas, TX
  4. Mill Valley-Sausalito Pathway near San Francisco, CA
  5. South Bay Trail in Santa Monica, CA
  6. Sammamish River Trail near Seattle, WA
  7. Forest Park Trail and Grant’s Farm Trail in St. Louis, MO
  8. Lakefront Trail in Chicago, IL
  9. Dr. Paul Dudley Bike Path and Minuteman Bikeway in and near Boston, MA
  10. Capital Crescent Trail and Washington and Old Dominion Trail near the District of Columbia

The most restrictive criteria in terms of locating usable trail segments were the segment length and the need for moderate to high volumes of traffic. Trails with moderate to high volumes of traffic tend to be in areas with many intersections and trail connections; however, we wanted segments at least 0.8 km (0.5 mi) long between intersections to gather unbiased data using the moving-bicycle method. In the end, we compromised on segment length in a couple of places (a 0.40-km (0.25-mi) segment for the South Bay Trail and a 0.64-km (0.4-mi) segment for the Forest Park Trail). We settled for segments in other places that did not have very high volumes, as shown in chapter 5.

Tables 9 and 10 provides some details on the chosen study trails. The study trails were located in urban and suburban areas. The study trail environments included parks, lakes, beaches, highways, and downtown areas. There was a nice range of trail widths from 2.44 m to 6.1 m (8 to 20 ft). The study trails were sometimes marked with centerlines, and sometimes there were other adjacent treadways that accommodated some users. Few trails had significant horizontal or vertical curvature. Most trails had good sight distances, as judged qualitatively by the research team after riding them numerous times on a bicycle.

 

Table 9. Characteristics of operational study sites.

Location Path name Community context Trail type Study location Area landscape Width (ft) Centerline
Raleigh, NC Lake Johnson Trail Suburban Park loop 0.25 mile point to the 0.75 mile point Wooded park, lake 8–8.5 None
Redmond, WA Sammamish River Trail Suburban Linear riverside greenway In Sixty Acre Park, about 0.5 mi from NE 116th St. Grass, ballfields 10 None
Marin County, CA Mill Valley-Sausalito Pathway Suburban Rail-trail At Bothin Marsh, north of the U.S. 101 bridge Marsh, highway, bay 9.5–10.5 None
Dallas, TX White Rock Lake Trail Urban Park loop Just south of the E. Lawther/Emerald Is. park access, near Winfrey Point Grass, lake, park road 14 Solid
Chicago, IL Lakefront Trail Urban Lakefront beach trail Near trail intersection with North Avenue Grass and beach 20 Solid
Santa Monica, CA South Bay Trail Suburban Oceanfront beach trail About a mile north of the Santa Monica Pier Beach 14 Dashed
St. Louis, MO Forest Park Trail Urban Park loop On the north edge of the park, along Lindell Blvd., between mile 5.25 and 5.75 Grass and street 10 Solid
Dunedin, FL Honeymoon Is. Trail (Dunedin Causeway) Suburban Hwy. sidepath/
greenway
West of the drawbridge Beach, roadway 12 None
Arlington, MA Minuteman Bikeway Suburban Rail-trail Mile marker 7.5 near the bike shop Wooded 12 Dashed
Boston, MA Dr. Paul Dudley Bike Path Urban Linear riverside greenway South of the River, just east of the Harvard Br. River, highway 8 Dashed
Location Path name Community context Trail type Study location Area landscape Width (ft) Centerline
Vienna, VA W&OD Trail Suburban Rail-trail Near downtown Vienna Grass and trees 10 Solid
Washington, DC Capital Crescent Trail Urban Rail-trail Between K St. and Fletcher's Boathouse Wooded 10 Dashed
St. Louis Co., MO Grant's Trail Suburban Rail-trail Near I-55 and Union Road Grass and trees 12 Solid
Dunedin, FL Pinellas Trail Suburban Rail-trail North of Curlew Road Golf course and hwy. 15 Solid
Dallas, TX White Creek Trail Suburban Linear streamside greenway North of the Fair Oaks Tennis Center, on both sides of the overpass Grass, stream 8 None
1 ft =0.305 m
1 mi=1.61 km

 

Table 10. Additional characteristics of operational study sites

Location Path name Surface Shoulder Other treadways Clear zone(ft) Sight distance Horizontal curvature Vertical curvature
Raleigh, NC Lake Johnson Trail Asphalt (in poor condition) No None 1 to 4 Poor Medium Low
Redmond, WA Sammamish River Trail Asphalt No Horse trail 6 to 10 Good Low Low
Marin County, CA Mill Valley-Sausalito Pathway Asphalt Yes, 5-7 ft gravel None 5 to 7 Unlimited Low No
Dallas, TX White Rock Lake Trail 12 ft of asphalt with 1 ft concrete edges No Some bicyclists use park road 10 to 20 Unlimited Low No
Chicago, IL Lakefront Trail Concrete No None Unlimited Good None No
Santa Monica, CA South Bay Trail Concrete No None Unlimited Unlimited Low No
St. Louis, MO Forest Park Trail Asphalt Yes, 4 ft dirt on one side Joggers use 4 ft dirt shoulder 4 Good None No
Dunedin, FL Honeymoon Is. Trail (Dunedin Causeway) Asphalt No None 0 on one side, 4 on the other Unlimited Low No
Arlington, MA Minuteman Bikeway Asphalt No None 2 to 4 Fair None No
Boston, MA Dr. Paul Dudley Bike Path Asphalt No Separate ped. paths None Poor Medium No
Vienna, VA W&OD Trail Asphalt No None Unlimited Excellent Low No
Washington, DC Capital Crescent Trail Asphalt No Peds. Use adjacent towpath 2 to 4 Unlimited Low No
St. Louis Co., MO Grant's Trail Asphalt No None 2 to 4 Good Low No
Dunedin, FL Pinellas Trail Asphalt No Some ped. paths 6 to 10 Unlimited None No
Dallas, TX White Creek Trail Asphalt No None 30 to 75 Unlimited None No
1 ft =0.305 m
1 mi=1.61 km

 

Data Collection Execution

Data collection occurred from July 2001 to March 2002. Peak hours and peak times were identified for each trail location. The peak days were generally Saturdays and Sundays. Peak hours varied by location. There often appeared to be two peaks during the weekend days on most trails; one volume peak in the morning and a second volume peak in the afternoon. Trail users who appeared to be using the trail for fitness purposes appeared most often during the morning hours. Recreational and casual users, consisting of tourists and families, appeared more often during the afternoons. At data collection sites where there were commuters, the commuter peak hours were generally the weekday mornings. The data collection process generally lasted from early morning until dusk; consequently, a great variation in volume was typically collected at each site.

The data collection team attempted to collect 60 trials at each trail. Because of inclement weather and mechanical failures, it was not possible to obtain 60 trials at each trail location. In total, 771 runs were successfully completed. Table 11 shows the sample size by trail. Because it was such a high-quality site and there were no other candidate sites in the city, the team collected extra data at the Lakefront Trail in Chicago. The most disappointing data collection trip was to Washington, DC, during October 2001, when bad weather prevented all but a handful of runs at what should have been excellent data collection sites.

 

Table 11. Number of successful data collection runs by trail.

Trail Successful Runs
Lake Johnson 58
Sammamish River 58
Mill Valley-Sausalito 60
White Rock Lake 60
White Creek 60
Lakefront 90
South Bay 60
Grant's 30
Forest Park 57
Honeymoon Island 48
Pinellas 57
Minuteman 60
Dr. Paul Dudley 60
Capital Crescent 9
Washington & Old Dominion 4
Total 771

 

FHWA-HRT-05-137

Previous | Table of Contents | Next

ResearchFHWA
FHWA
United States Department of Transportation - Federal Highway Administration