Evaluation of Travel Time Methods to Support Mobility Performance Monitoring:
Otay Mesa (Page 2 of 3)
Data Collection Details
The U.S. Customs provided Border crossing statistical data. This data was evaluated for an assessment of the variability in travel conditions at the crossing. The goal of this analysis process was to obtain statistically useful data with as few data collection days as possible. In order to customize the data collection activities to the Otay Mesa crossing, the following steps were conducted:
- Define significant "seasonal" variations,
- Define significantly different days of the week,
- Identify traffic streams that experience significantly different conditions, and
- Estimate the number of days needed for the data collection survey.
Due to project constraints, data collection needed to occur between late May and early September 2001. The FY 2000 data collection site report prepared by Caltrans indicated that July and September had the highest volumes and July was selected for the FY 2001 data collection. Subsequently obtained data, shown in Table 3, indicates there is some variation in the commercial traffic by month, with the lowest volumes in the winter months and the highest volumes in the summer months. Table 3 shows that the two months with the greatest average volumes during this data collection window were June and August with July having the third highest monthly volume.
| Month | Truck Volumes |
|---|---|
| October, 1999 | 57,716 |
| November, 1999 | 54,228 |
| December, 1999 | 50,000 |
| January, 2000 | 49,378 |
| February, 2000 | 53,896 |
| March, 2000 | 58,836 |
| April, 2000 | 54,239 |
| May, 2000 | 59,955 |
| June, 2000 | 63,547 |
| July, 2000 | 60,484 |
| August, 2000 | 62,780 |
| September, 2000 | 58,640 |
| Total | 683,699 |
Source: U.S. Customs data provided by Caltrans
Tables 4 and 5 show that there is a significant difference in commercial traffic between weekdays and weekends and, further, there is a significant difference between Monday and the rest of the weekdays. Weekend traffic is 20.4 percent of typical weekday traffic and Monday traffic is 89.6 percent of typical Tuesday through Friday traffic. It was determined that collecting three days of data, from Tuesday through Thursday, would provide an adequate number of data samples to represent "typical" conditions.
| Day | Day of Week | Inbound |
|---|---|---|
| 1 | Saturday | 817 |
| 2 | Sunday | 299 |
| 3 | Monday | 2,466 |
| 4 | Tuesday | 484 |
| 5 | Wednesday | 2,614 |
| 6 | Thursday | 2,712 |
| 7 | Friday | 2,793 |
| 8 | Saturday | 795 |
| 9 | Sunday | 281 |
| 10 | Monday | 2,643 |
| 11 | Tuesday | 2,616 |
| 12 | Wednesday | 2,801 |
| 13 | Thursday | 2,721 |
| 14 | Friday | 2,457 |
| 15 | Saturday | 664 |
| 16 | Sunday | 413 |
| 17 | Monday | 2,507 |
| 18 | Tuesday | 2,663 |
| 19 | Wednesday | 2,448 |
| 20 | Thursday | 2,663 |
| 21 | Friday | 2,476 |
| 22 | Saturday | 755 |
| 23 | Sunday | 255 |
| 24 | Monday | 1,837 |
| 25 | Tuesday | 2,833 |
| 26 | Wednesday | 2,748 |
| 27 | Thursday | 2,553 |
| 28 | Friday | 2,704 |
| 29 | Saturday | 767 |
| 30 | Sunday | 265 |
| 31 | Monday | 2,443 |
| Total |  |
57,493 |
Source: U.S. Customs
| Day of Week | Week 1 | Week 2 | Week 3 | Week 4 | Week 5 | Week 6 | Average |
|---|---|---|---|---|---|---|---|
| Sunday | |
299 | 281 | 413 | 255 | 265 | 303 |
| Monday | |
2,466 | 2,643 | 2,507 | 1,837 | 2,443 | 2,379 |
| Tuesday | |
484 | 2,616 | 2,663 | 2,833 | |
2,704* |
| Wednesday | |
2,614 | 2,801 | 2,448 | 2,748 | |
2,653 |
| Thursday | |
2,712 | 2,721 | 2,663 | 2,553 | |
2,662 |
| Friday | |
2,793 | 2,457 | 2,476 | 2,704 | |
2,608 |
| Saturday | 817 | 795 | 664 | 755 | 767 | |
760 |
Source: U.S. Customs
*Note: Data from Tuesday, July 4th was omitted from this average because of the unusually low volume during the holiday. The average for all Tuesdays is 2149 in July 2000.
From discussions with U.S. Customs and Caltrans, it was learned that backups typically did not occur on the U.S. side beyond the intersection of La Media and Siempra Vivra and, when they did, they did not grow very long. Occasionally, however, the backups would reach SR-905, several long blocks north and approximately one mile farther upstream, where the San Diego Police Department would divert additional trucks so that there would not be a backup on SR-905. However, on the Mexican side, backups were said to occur on a regular basis and could stretch to the initial inbound data collection location, approximately 1.5 miles from the crossing, and beyond.
Data Collection Procedures
The data collection stations selected for the crossing were chosen because of the particular actions that occur at each site. Segments defined by the data collection stations were used to determine the commercial vehicle travel times and freight delay. As illustrated in Figures 2 through 6, the data collection sites were located at:
- An advance station located upstream of the commercial vehicle queue – IB-1 and OB-1.
- The import station (primary inspection booths before detailed, or secondary, inspection) – IB-2 and OB-2.
Data collection was conducted by recording commercial vehicle license plates as vehicles crossed fixed points within the data collection sites. Survey individuals or teams, were placed at each of the four data collection sites to record commercial vehicle license plate data. Figure 2 shows the location of the Customs facilities on both sides of the border, including station locations and major points of inspection.
Collectors at these locations would record the last five characters of the front, lower-left license plate of as many trucks as possible that passed their location. When trucking firms register many vehicles at once, they often get assigned sequential license plate numbers. Using the last five characters helps to ensure that as different trucks operated by the same firm travel across the bridge that they are uniquely identified. License plate information was entered into Handspring Visor PDAs (handheld computers) with a special application designed for this project. Each entry was time-stamped with the current date and time. Prior to each day’s collection, all PDAs were synchronized to the same time. Prior experience indicated that recording the entire license plate was too time consuming and that entering only the last four characters did not provide adequate distinction between different vehicles, so the project team chose to record the last five characters.
Typically, the queue of trucks crossing the border would extend a short distance beyond the actual crossing area. However, on occasion the queue would extend onto the local road system. When this occurred, the data collector at the #1 location would have to move further from the crossing to a point beyond the end of the queue. In this way, they could continue to record trucks before they began their wait at the end of the line. When this or any other event of interest occurred, the collectors would use an "EVENT" feature of the PDA software to record it.
For each #1 location, the supervisor would record the distance from any data collection point other than the original position. During post-processing, the data from all locations nearer to the crossing than the farthest location would be adjusted to include the additional travel time from the farthest location to the original location. The travel time would be computed at free-flow speeds, since there would have been no queue at the times that the data were collected at these closer locations. In this way, the data all would appear to be collected from the same location, the one most distant from the crossing.
The data collection team used both cell phones and hand-held, two-way radios to maintain in touch with each other. This was particularly important when the queues lengthened such that a collector had to move farther upstream. The supervisor could be kept informed without repeated trips to each data collection location. This was also useful at the end of the day when the #1 collectors would inform the #2 collectors of the last truck they recorded, so the #2 collectors would know when to stop. While interference and cell tower locations created some problems with reception, each collector was usually able to use either their radio or cell phone to reach whomever they needed to speak with.
Data Collection Sample Size
Sample sizes are typically not a concern with videotape or handheld data entry devices, because the data collection includes a large number of vehicles. However, minimum sample sizes should be verified with variability values from field data. Early research found that sample sizes from 25 to 100 license matches were necessary for a given roadway segment and time period (Turner, et. al.). In most cases, there were sufficient records to meet this requirement.
Data Collection Equipment
As outlined in the "Data Collection Procedures" section above, Handspring Visor PDAs were used as the data entry device and proved adequate to the task. Low-end models with 2MB of storage capacity were selected as the application and data size were projected to be well below this limit. The Handspring Visors use the Palm OS (operating system) and have faster processing speeds (at least in side-by-side comparison with this application) and larger screen sizes than comparable models from Palm Computing.
A custom application was developed for the Palm OS that allowed the data collectors to identify their locations (e.g., IB-1, OB-2), the number of open booths (primarily used for the customs inspection booths), special events or other comments, and license plate information. A screen shot of the application interface is shown in Figure 7.
Figure 7. Data Collection Device and Software Application
The data were downloaded via a serial cable directly from the application into a text file on the field laptop computer, which was a Dell Latitude CPx H running with a 500 MHz Pentium III processor.
Data Collection Summary
Table 6 shows the number of commercial vehicle license plates recorded for each of the stations on each of the data collection days. Table 7 shows the average daily traffic volume as recorded by U.S. Customs (inbound direction). Data from Mexican Customs have not yet been made available. Hourly volumes are used in the calculation of delay; those are shown with the delay calculations in Tables 8 through 13.
| Station | 7/17/01 | 7/18/01 | 7/19/01 |
|---|---|---|---|
| IB-1 | 1931 | 1972 | 1267 |
| IB-2 | 2150 | 1977 | 1196 |
| OB-1 | 1098 | 1140 | 1021 |
| OB-2 | 1161 | 1120 | 1078 |
| Total | 6340 | 6209 | 4562 |
| Direction | 7/17/01 | 7/18/01 | 7/19/01 |
|---|---|---|---|
| Inbound | 2847 | 2866 | 2742 |
| Outbound | not avail. | not avail. | not avail. |
| Total | 2847 | 2866 | 2742 |
Data Quality Steps
At the end of each day of data collection, the supervisor would collect the PDAs and download the data into the field laptop computer where it was stored on the hard drive. The data would be examined for any anomalies and transferred across the Internet to a secondary location for backup purposes. The IB-1 and IB-2 data would be merged together and license plates from the two locations would be "matched" using a spreadsheet developed in Microsoft Excel. As it is easy to mistake certain characters, particularly letters that looked like numbers, the license plate data was pre-processed. All 'I's were replaced with '1's; all 'O's, 'D's, and 'Q's were replaced with '0's; all 'S's were replaced with '5's; and all 'Z's were replaced with '2's. In addition, the data collectors were instructed to always use '1's for 'I's and '0's for 'O's (i.e., to use the digit, rather than the letter).
Occasionally, collectors would be unsure about a license plate and would append "QQQ" to their entry. This would typically occur when several trucks passed the collector in rapid succession or if one truck blocked the license plate of another and he or she could only manage a quick glimpse. This would allow the supervisor to search the downloaded data for a potential match by using the travel times of other trucks that were recorded in the same general time frame. During this process, the supervisor could also identify the few records in which the data collector forgot to press "ENTER" after recording a license plate before recording the next one. These ten-character entries could be split into two and the time for the first interpolated from the adjacent entries if they were less than a minute or so apart.
Data post-processing also included a step to identify any anomalies in the data, including outliers. Outliers, records that indicated travel times significantly greater than typical for that time period, were most often caused by recording the license plate of a vehicle only some of the time as it made repeated trips across the border during a single day. This is because the matching algorithm uses the most recent time at the #1 position when matching to a record from a #2 location. For example, if the vehicle was recorded as it headed from Mexico to the U.S. early in the morning, later returned to Mexico, was missed as it re-entered the U.S. later in the day, and then recorded on its subsequent return to Mexico, the #1 time from its first trip would be matched with it #1 time from the first trip (for a valid travel time) an also matched to the #2 time from its second trip (an invalid travel time). This invalid travel time would be easily identified by manual inspection of the data, aided by highlighting those travel times above a specific, but variable, threshold.
previous | next