10. FINDINGS AND LESSONS LEARNED

This report has presented the results of an effort to pilot test the feasibility of several mobility-based work zone performance measures. The measures encompassed various dimensions of:

Work zone exposure
Traffic queuing
Motorist delays
Travel time reliability

A discussion of useful safety-related measures was also presented. However, because of anticipated lag times in obtaining crash data from the various pilot test locations (and the straightforward nature with which safety-related performance measures are typically developed), exposure and mobility-related performance measures only were targeted at each of the pilot test locations.

The pilot test also incorporated different methods of data collection to support performance measurement computations. The first method was to rely on field personnel at the project to manually document the occurrence of traffic queues that develop because of temporary lane closures or other work activities that constrain roadway capacity. Use of this method assumes that traffic conditions in the absence of the work zone would have been uncongested, such that any queuing or congestion that observed could be attributed solely to the work zone. The second method was to use electronic traffic surveillance data to monitor traffic conditions and compute desired performance measures. This data may come from a regional TMC already in place, or from sensors specifically deployed as part of a portable work zone ITS. The third method relied on truck speed data collected by third-party vendors willing to provide limited access for transportation monitoring purposes. A screening process was undertaken to access available speed observations at each project on specific times and dates when traffic impacts were expected to exist. Organizing these data by location and time of day was expected to provide another method of tracking the effect of work zones on traffic mobility on a facility.

In the following section, a synopsis of the key lessons learned with respect to data collection and analyses of work zone exposure and mobility data are presented. Next, the various performance measures are compared across projects to discuss how they might be used to assess for critiquing project-level decisions and actions, as well as agency policies and processes.

DATA COLLECTION AND ANALYSIS FINDINGS AND LESSONS LEARNED

Manual Documentation of Queues by Field Personnel

Overall, the results of the pilot test suggest that it is possible to obtain fairly good quality queue data from field personnel upon which to base work zone mobility performance measure computations. The impact of data collection demands on field personnel workload was a potential concern initially in this pilot test, but workload-related problems did not appear to materialize. Comparisons of the queue length and duration data to data from traffic surveillance sensors and ground-truth measurements (when available) suggest that reasonable estimates can be obtained. Very few instances were identified where field personnel failed to document either the work and lane closure activity (this type of information is normally needed for project diary entries anyway) or the development of queues. The fact that this was a special request as part of a research project may be part of the reason the effort was so successful. It is likely that regular reminders (possibly as a special note in the project diaries) would be beneficial to agencies striving to adopt this process as part of their monitoring efforts. It is also likely that an indication of the importance of this documentation effort by upper management in an agency would further ensure personnel consistency in reporting.

These successes notwithstanding, a few key lessons were also learned through this pilot test effort.

It will be important for agencies to establish simple operational definitions for field personnel as to what exactly constitutes a traffic queue – In most cases, agencies will want to use a reasonable and consistent indicator, such as when vehicle speeds drop below 35 mph. Other agencies may choose to use a much lower value to define what constitutes a queue (some agencies use values as low as 10 mph to define queuing). Unfortunately, it may not be possible for field personnel to accurately gauge vehicle speeds and make distinctions as to whether a backup truly meets the agency definition of a queue. All that an agency may be able to expect is for field crews to be able to define when and how far upstream traffic has “backed up.” Estimates of what the speed in queue likely was at that work zone may then need to be computed after the fact using traffic flow relationships.
Estimates of queue lengths need to include a description of the location of the queue relative to the lane closure (upstream, beyond taper and into work zone, partially upstream and partially within the work zone, etc.) – Related to the previous bullet, the comparison of queues documented by field personnel and the electronic traffic surveillance sensor data at several sites indicated that some queues occurred at the point of the lane closure transition, whereas others occurred within the work zone next to the location of work activity or other disturbance. When queues occur within the work zone where the number of lanes available remains constant, traffic flow relationships suggest that speeds in those queues will be much higher than in queues at and upstream of a lane closure transition. Equation 1 identified the queue speed estimated for the lane closure transition queues. For queues within the work zone, the speed at capacity flow (approximately one-half of the free-flow speed) is a more realistic value to use.
Manual methods will be less effective in capturing short (less than one hour) disturbances that result in queues – Realistically, neither manual methods or electronic surveillance data are likely to be all that effective in capturing short-duration queues that may form because of small fluctuations in traffic flow behavior, a temporary disruption in flow by work vehicles entering or exiting the work area, or other disturbance. Given the workload that field personnel typically carry during work zone activities, only the more significant long-lasting queues that develop are likely to be documented.

Use of Electronic Traffic Surveillance Data

The results of this pilot test also indicate that electronic traffic surveillance data, when available, can be utilized to measure work zone exposure and to compute the effects of work zones on traffic mobility. However, some challenges do exist in obtaining and applying these types of data to work zone exposure and mobility-related performance measurement. The major lessons learned relative to the use of electronic surveillance data are as follows:

Field personnel documentation of when and where lane closures were placed and hours of work activity will still be needed to compute mobility performance measures using electronic traffic surveillance data – Although it may be desirable to be able to rely on electronic traffic surveillance data alone to determine when lanes were closed, which lanes were closed, when the lane closures were removed, etc. (and thereby eliminate the need to extract these data from field personnel), the realities of electronic surveillance data is that it is often not easy to distinguish between changes in conditions that are due to the work zone actions listed, those which are due to incidents, and those due to normal “noise” in the traffic stream.
It is important to make sure that the traffic sensors themselves will remain operational when work begins – At some of the pilot test sites, it appears that power was disrupted temporarily to the traffic sensors when work was occurring. As a result, data were not available during the exact times they would be needed for monitoring and performance measurement computation purposes. If temporary power disruptions in the vicinity are anticipated, it will be necessary to take steps to fill in gaps in coverage (by manual documentation of conditions by field personnel or by deploying portable traffic sensors that can provide the data during the power outage).
The level of accuracy in work zone mobility performance measurement achievable with electronic traffic surveillance data depends heavily on the design of the system (particularly traffic sensor spacing) – Some of the key factors that will affect the extent to which accurate performance measures can be recorded and computed include 1) the availability of speed data from each sensor (critical to the estimation of delays and queue lengths), 2) the spacing between sensors (closer sensor spacing allow for finer resolution of queue lengths and queue length growth/dissipation over time), and 3) the limits of surveillance relative to the length of impacts that develop (if queues grow beyond the limits of surveillance, the measures computed from that data may be counterintuitive to what is actually occurring at the site).
Aggregation of traffic sensor data to hourly averages is a reasonable compromise between accuracy and practicality for monitoring work zone mobility impacts – Data at this level of detail appears to be still sensitive enough to detect the onset and dissipation of congestion, but not so detailed so as to overburden the analyst.

Truck Probe Spot Speed Data

At each of the pilot test locations, analysis and comparison of truck probe speed data for work zone mobility performance monitoring was fairly limited because of a lack of adequate sample size across time and space. In some cases, work activity occurred during nighttime hours when traffic volumes are lower, and this contributed to the low sample size. However, even at the Las Vegas pilot test locations, the amount of truck transponder data obtained was not enough to allow a full comparison to other data sources. There were instances at the three nighttime project pilot test sites in which the knowledge that a work zone was present and that congestion had developed (via other data sources) could be correlated with a reduced speed measurement obtained from the truck transponder data. Unfortunately, there were more instances in which truck speed data were not available at the location and time that work zone queuing was known to have developed. Consequently, its value as a primary data source for mobility monitoring and performance measurement is currently rather limited.

The limitations of this current pilot test notwithstanding, the potential use of vehicle probe speed data for work zone performance measurement is still attractive to highway agencies. As private sector vendors of speed and travel time data continue to evolve, an adequate supply of this type of data may someday become available. Although there were few actual lessons learned through the review of truck transponder speed data in this pilot test effort, the following are three issues to be considered in future efforts to use this type of data:

The choice of segment length must balance the trade-offs between amount of data that is available and level of accuracy of queue and delay estimates – For this pilot test, the decision was made to use one-mile segments and one-hour time periods in the analysis. It was found that the amount of data available was insufficient to support the analysis at this level of detail. Unfortunately, even a longer segment length would not have proven beneficial for the pilot test (researchers did examine the possibility of combining mile segments each hour, without much improvement in coverage). For other vehicle probe data sources, the goal should be to dissect the affected roadway segment into intervals that allow for multiple readings (at least three would be preferable in order to establish the degree of variability of each segment estimate) in each segment in each hour. If a reasonable speed estimate can be established in each segment each hour, the computational procedures for determining queue lengths and delays from spot sensor data could be applied.
A common reference system must be used to define the segment endpoints and work zone location – In many work zones, the location of work activity and temporary lane closures will change from work period to work period. Documentation of where the work zone is located each period is essential to matching the vehicle probe speed data in each segment to that location so that queue lengths can be approximated.
The potential reaction of the vehicle fleet to the presence of a work zone and resulting congestion should be considered when analyzing the data – In the case of truck transponder data, communication between drivers of the presence of severe congestion at a work zone could lead to many drivers leaving the roadway to take a break and wait for congestion to dissipate. Although it was not possible in this pilot test to verify or refute whether this occurred at any of the sites, the potential does exist for such behavior to occur.
It is important to remember that speeds within each road segment are not true “spot” speeds – The aggregation of speeds obtained from vehicle probes will either be a compilation of speeds all along the length of the roadway segment, or an estimate of speeds using elapsed travel time on that segment or portion thereof. Significant changes in roadway characteristics within the segment length can cause much different speeds at different locations within the segment. This increased variability makes it more difficult to assess the impacts of a work zone, and implies that a greater sample size will be needed.

Comparison across Possible Data Collection Methods

Table 47 compares and contrasts the various data collection methods currently available for monitoring work zone impacts and assessing performance. The table includes point-to-point travel time measurement systems mentioned previously, but which were not available for inclusion in the pilot test effort.

Table 47. Comparison of Data Collection Methods for Work Zone Monitoring and Performance Measurement
Data Source	Advantages	Disadvantages	Other Considerations
Manual Measurement of Queue by Field Personnel	Direct control of data by agency Easy to implement Minimal additional cost to agency	Increases work load of field personnel (inspector, TTC supervisor, etc.) Limited to locations where recurrent congestion not present	Important to note location of start and end of queue relative to work zone lane closure each work period
Electronic Spot Speed Data:
Existing TMC already in place	Minimal additional cost to agency Availability of “before” data (allows assessment of incremental effects of work zone)	Location of devices may not be optimum for work zone assessment purposes Can require significant effort to extract and process desired data from entire system	Important to ensure that sensors will remain operational during work activities
Portable ITS devices (sensors on trailers, portable sensors in channelizing devices, etc.)	Allows for optimum placement of sensors for work zone monitoring purposes	Work zone ITS can be costly to the project	Portable devices must be placed consistently each work period (or documented if changed each day) Sensors must extend beyond the limits of anticipated congestion
Vehicle probe data (i.e. truck transponders)	Does not require agency to purchase technology to deploy Does not require technology to be moved or maintained	Sample size is an issue for truck transponder data May require purchasing from third-party vendors	Important to remember that speeds are not true “spot” speeds, but are distributed across the segment in which they are included
Electronic Point-to-Point Travel Time Data::
AVL systems	Very accurate tracking of speed profiles possible	Vehicle fleets to draw data from are usually very limited Will require agreements with agencies or vendors who collect these data	The potential exists for obtaining data at a finer level of detail than is needed, which could create data management and analysis challenges
AVI toll tags	Available sample size can be fairly high	Queues difficult to measure without multiple, very closely spaced sensors May require agreement with toll agency to gather data Deployment of additional transponder readers can be costly	Generally limited to regions where a significant proportion of the driving population has toll tags in their vehicle
License plate recognition systems	Available sample size can be fairly high	Costly to implement Sample size availability depends on ability of system to match license plates Queues difficult to measure without multiple, very closely spaced sensors	May create concerns about privacy with local citizens
From bluetooth readers	Data can be obtained unobtrusively from roadside devices	Requires purchase and deployment of Bluetooth readers Queues difficult to measure without multiple, very closely spaced sensors	Dependent on the volume of traffic present
Cell phone tracking	Large potential sample size within traffic stream	Requires agreements with 3rd party vendors to obtain data	Dependent on the volume of traffic present

FINDINGS AND LESSONS LEARNED REGARDING PERFORMANCE MEASURE COMPUTATION AND INTERPRETATION

The proposed measures themselves offer insights into their potential usefulness to highway agencies for both current and future project-level decision-making, and for process-level reviews.

Exposure Measures

The importance of vehicle-based measures of exposure (volumes through the work zone, vehicle-miles-traveled through the work zone) for computing and normalizing queue, delay, and crash-based performance measures was discussed earlier in this report. The importance and lessons learned regarding some of the additional exposure measures proposed are discussed below.

Work activity measures (percent of days worked, average hours per day of work) will be useful in tracking and comparing contractor level of effort – However, many factors affect contractor ability to work on a given project. Consequently, these data are likely to be most useful when averaged across multiple projects, or in some form of compliance assessment to a threshold target (e.g., percent of projects with less than 40 percent of work days with activity). These values would also be useful for monitoring efforts of individual contractors (again, either in terms of averages or as a percentage of projects with work activity percentages below a preset threshold). Finally, these measures could also be extrapolated across all projects in a region as a way to quantify total exposure for use by public information offices or other needs.
Capturing lane closure hours (percent of hours involving 1, 2, etc. lanes closed) is relevant only if total lanes remaining open is also captured– Agencies will need to capture and categorize these data based on lanes open/lanes closed configurations. One lane closed on a five-lane section is much different than on a two-lane section. Given how the number of lanes available for traffic can vary from location to location along a section of multi-lane roadway, establishing field documentation procedures for lane closures that require the number of lanes closed, number of lanes open, and location of the lane closure(s) will be important. Proper documentation of lane closure parameters will also facilitate collection and analysis of delay and queue measures (especially when relying on electronic traffic surveillance sensor data).
It will be important to stratify projects based on the type of lane closure being utilized when examining trends or evaluating compliance to thresholds – For example, projects involving short-term lane closures each day or night should be examined differently than those involving long-term lane closures. Information regarding percent of lane closure hours with inactivity will be much more valuable for the short-term lane closure projects than for long-term lane closure projects, since the traffic control designer should already have acknowledged that there will be many inactive hours with lane closures for a long-term lane closure traffic control option. While it still may be worthwhile to quantify this information, mixing the results with short-term lane closure projects will tend to skew the values and make it difficult to assess agency decisions and policies for either type of lane closure being used.
Computation of the total lane-mile-hours of lane closures during the project can provide another way to normalize delay measures – Depending on the number of lanes available in a given direction of travel, contractors have options in term of how many lanes are closed for a given work shift. However, the number of lanes closed at a time also affects how many work shifts are needed to complete a task. Normalizing vehicle-hours of delay on a per-lane-mile-hour of closure basis can help connect traffic impacts to both lane closure and work productivity measures, if measured across the total duration of the work task being completed.

Queue Measures

The argument was made earlier in this report that queuing measures themselves were valuable for work zone mobility monitoring and impact assessment. Furthermore, an approach was proposed for estimating delay impacts from queue length and duration documentation by field personnel during work zone activities. The pilot test results did indicate the need to include additional details during queue length documentation to improve delay estimates. Additional insights and lessons learned regarding queue measures are provided below.

The percent of time when queues exceed selected threshold values should be useful as a potential performance specification for work zone traffic control and impact mitigation – The results of the pilot tests illustrate that unexpected queues can occur from time to time due to fluctuations in traffic flows and small, temporary disturbances to that flow near areas of work activity. These can occur even if analysis during traffic control plan design indicates that traffic demands do not exceed the expected traffic capacity through the work zone (although the likelihood increases the closer demand is to expected capacity value). It appears that many of these queues are fairly short in duration and length. This performance measure provides a simple way to distinguish between those occasional short-duration queues that are out of the control of the contractor/highway agency, and those that are systemic in exceeding agency threshold targets.
The measure “percent of traffic encountering a traffic queue” provides a direct indication of the breadth of impacts of the work zone to the motoring public – However, it may be challenging for agencies to use this measure where recurrent congestion is present prior to the start of the project, and where long-term lane closures have been deployed. For these situations, it will be necessary to first compute this measure for the before-project condition, and then measure the change in the measure once the work zone has been installed.
The average duration of queue per work shift or day provides valuable information at projects where recurrent congestion was already present prior to the start of the work zone, or where electronic surveillance may not cover the entire length of work zone impacts on a roadway– The measure also provides useful insights about scheduling decisions for short-term lane closures each work period.
The average queue length and maximum queue length measures have value for assessing both project-level and process-level traffic management decisions – As stated earlier in this report, the maximum queue length is a useful measure for evaluating how well planning and impact analysis tools and procedures worked for the project, and to assess whether advance warning signs and other traffic control provided adequate warning for the project. In addition, it is expected that tracking these measures across multiple projects in a region (appropriately categorized by roadway type) to determine whether policies and procedures should be modified in terms of capacity values assumed, analysis tools and procedures allowed, impact mitigation strategy emphases, etc.

Delay Measures

Vehicle-hours of delay are generally the most significant contributor to the additional road user costs created by work zone activities, and so can have significant implications in terms of contracting strategies used and penalties and incentives incorporated into contract language. Relating these total costs in terms of impacts to individual motorists is also important from a customer service perspective that many agencies emphasize. Since queues and delays are correlated, similar trends and insights can be gained through examination of both types of measures. The delay measures pilot tested in this effort do appear to provide information that could be valuable to a highway agency or a contractor, except for the average delay per entering vehicle measure. Initially, this measure was proposed as a way to account for the fact that a work zone can affect some motorists very significantly during those times when congestion and queues are present (and be perceived somewhat negatively by those motorists), but have no effect and so be perceived much more positively by motorists during other times of the work activity when queues have not developed. In reality, attempting to average the impacts across all vehicles only appears to mute the overall effect of the project and give a somewhat unrealistic perception that the impacts to motorists were not all that significant. The average delay per queued vehicle, along with the percent of traffic encountering a queue, appears to be a more straightforward way to account for both the intensity and breadth of work zone impacts on motorists.

Travel Time Reliability Measure

The use of the buffer index as a travel time reliability measure for work zone performance monitoring was demonstrated for one of the pilot test project locations. For projects where recurrent congestion already exists and where reliability is already affected, computation of buffer indices and comparison to the pre-work zone condition appears to offer useful insights into another dimension of user impacts due to the work zone. Two different reliability thresholds were computed (95th percentile and 80th percentile travel times) for the pilot test, and both indicated impacts due to the work zone. However, the magnitude of the impacts differed. Given that travel time reliability measurement research in general has not provided recommendations about appropriate thresholds to use, it would seem appropriate to compute and monitor both as part of work zone mobility performance measurement at the present time.

Safety Measures

Although data were not available to allow computation of safety performance measures for the pilot test locations, the process by which such measures are computed is fairly straightforward. Traditionally, agencies do (eventually) have access to crash data through their statewide crash records system. What is usually missing is the exposure data that is needed to convert the crash data to meaningful rates that can be compared to pre-work zone conditions or across projects, regions, or the state. In addition, the crash data can also be combined with work activity exposure data to assess the impacts of the project or project task upon overall crash numbers or crash costs.

It is also important to utilize and interpret the safety measures (as well as the exposure and mobility-related measures) relative to the goals of a particular project when making decisions regarding the effectiveness of a particular approach used to complete the work. For example, performing a particular task using only nighttime lane closures may result in a higher crash rate (or increase in crash rate) on per vehicle-mile-traveled basis than doing the work doing the day. However, because of the much lower traffic volumes present at night, the total impact upon traffic crashes over the duration of completing that work task might still have been much lower than if it had occurred during the day. Ensuring that the appropriate measure was computed (total crash costs for the project, in this example) will allow agencies to continue to improve their overall processes and procedures for delivering a functional, safe, and high-quality transportation product to the public.

NEXT STEPS

To assist practitioners in applying the findings and lessons learned from this pilot test effort, a primer on selecting and computing work zone performance measures is being developed to accompany this report.

previous | next