Pilot Testing of SHRP 2 Reliability Data and Analytical Products: Washington (2014)

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.

40 CHAPTER 4 Pilot Testing and Analysis on SHRP 2 L02 Product 4.1 Introduction The L02 project aims at developing tools and procedures for creating a system that monitors travel time reliability and quantifies the impact of varying conditions on the reliability. Ultimately, the L02 tools are intended to help transportation agencies answer five basic questions: 1. What is the distribution of travel times in their system? 2. How is the distribution affected by recurrent congestion and nonrecurring events? 3. How are freeways and arterials performing relative to performance targets set by the agency? 4. Are capacity investments and other improvements really necessary given the current distribution of travel times? 5. Are operational improvement actions and capacity investments improving the travel times and their reliability? The L02 project’s effectiveness at answering each of these questions was evaluated, and solutions to shortcomings are recommended. The three L02 products were also tested by applying them to a TTRMS. The three products tested include the guide and its methodology, the TTRMS and its effectiveness in monitoring reliability, and the approach to synthesizing of route- level travel times from segment-level travel times. This system helps quantify travel time reliability for a relatively large-scale network, visualize the causes of congestion, and identify segments where a performance improvement is desired. 4.2 Test Sites Test Sites A and B are selected for L02 product testing. Test Site A includes I-5 northbound and southbound from Lynnwood to Tukwila, and Test Site B covers the entirety of I-405 northbound and southbound. Test Site A runs 26.5 miles between the southern and northern termini of I-405 and experiences a peak AADT of 228,000 vehicles near milepost 163, just south of the I-5/I-90 interchange. Similarly, the I-405 route (Test Site B) is 29.4 miles long and experiences a peak AADT of 200,000 vehicles near milepost 12, just north of the I-405/I-90 interchange. These routes are illustrated in Figure 4.1. Data were not collected for the I-5 reversible express lanes, which on weekdays run southbound from approximately 5:00 a.m. to 11:00 a.m. and northbound from approximately 11:15 a.m. to 11:00 p.m. However, these time periods are often delayed or modified because of incidents and special events. These express lanes run approximately 7 miles from milepost 165 to milepost 172 and carry between two and four lanes of traffic with the number of lanes

41 increasing as the roadway approaches the downtown Seattle exits and entrances. Because of the variable nature of operation times, the limited access nature of this facility, and the integration with traffic on mainline I-5, incorporating these express lanes into the travel time calculations would likely decrease the accuracy of the results. Therefore, travel times were not calculated for the I-5 express lanes. Instead, express lane traffic is considered interacting with the mainstream traffic as on-ramp or off-ramp flows. (a) (b) Map data © 2014 Google Figure 4.1. Map of (a) Test Site A (I-5 facility) and (b) Test Site B (I-405 facility). 4.3 Data Description In this test, 5-minute loop data serve as the basis for the travel time calculations. The procedure follows the L02 Guide for travel time monitoring, in which “5-minute interval” is stated as the minimum resolution to accurately capture the effects of weather and incidents on travel time reliability. The timeframe of interest is the entire 24-hour day with data from January through

42 December 2012. Researchers chose to analyze data for weekdays Tuesday through Thursday. Some studies separated Monday and Friday from other weekdays when predicting traffic patterns, because traffic patterns during Monday and Friday may deviate from other weekdays. This way, researchers were able to capture the most and least congested periods of the day while eliminating the traffic inconsistencies that are frequently observed on Mondays, Fridays, and weekends. Data from any existing HOV lanes were also excluded. This single-loop data were then converted to speed using Athol’s method (Athol 1965) with a g-factor of 2.2. The WSDOT travel time estimation methodology specifies the minimum and maximum speeds to use for travel time calculation. Speeds higher than the maximum speed are truncated to the maximum speed value of 60 mph. Those speeds lower than the minimum speed threshold are replaced with the minimum speed of 10 mph. Segment travel times were then generated by measuring the distance between two adjacent loop locations and dividing that by the harmonic mean of the speeds measured at these locations. Finally, route-level travel times are calculated using a piecewise trajectory algorithm that sums the segment-level travel times along the route. 4.4 Regime Characterization According to the L02 Guide, a regime is defined as a pair of conditions that consists of a recurring congestion level and a nonrecurring condition. For the recurring congestion, each travel time measurement is tagged with a congestion level (free-flow, low, moderate, and high) based on the time of day and average travel time based on the entire year, as defined in Table 4.1. Table 4.1. Determination of Congestion Levels for I-5 and I-405 Congestion Level Average Annual Travel Time Free-flow <30 min Low 30–35 min Moderate 35–40 min High >40 min It is important to note that the times for congestion levels are not determined day to day but rather reflect the annual average conditions as specified in the L02 Guide. For the nonrecurring condition, data is tagged as “normal” (no nonrecurring event occurred), “weather” (a weather event is occurring that negatively affects traffic), “incident” (there is a lane-blocking incident affecting the study facility), and “overlap” (if weather and incident occur simultaneously). Incidents are tagged using data from the WITS system. Data are tagged as having an incident in progress if there is an incident blocking a lane or lanes on the route or within 2 miles downstream of the route, during the 5-minute period. Data are tagged as “weather” if there is measurable precipitation during a 1-hour period or if fog was recorded. The data are taken from local weather stations, which only report every hour. Once all data are tagged with a recurring

43 congestion level and a nonrecurring condition, the data could be plotted as a cumulative distribution function (CDF) chart, the key visual output of the L02 methodology. 4.5 Testing Results and Discussion After categorizing all the travel time data into the appropriate regimes, many useful charts can be drawn in analyzing each facility’s travel time reliability and comparing the reliability between the facilities. The travel time CDF is the key output of L02 and the most information-rich chart. Figure 4.2 shows each facility’s TTI CDF (developed following the L02 procedure). (a)

44 (b) (c)

45 (d) Figure 4.2. TTI CDFs for all test facilities: (a) I-5 North, (b) I-5 South, (c) I-405 North, and (d) I-405 South. These graphs are useful since they contain important information about the travel time reliability of each route. For example, it is easy to look at the chart for I-405 South and infer that the interquartile range for TTI under heavy congestion and adverse weather is about 1.4–1.9. It is also useful to show the relative reliability of each regime. Looking at the I-405 South chart again, travel times with adverse weather and heavy congestion are generally slower and consistently less reliable, as indicated by higher TTI above the 25th percentile and a broader distribution (less steep curve) for the “Weather Heavy” curve versus the “Normal Heavy” curve. While the CDF graphs have proven useful for quickly interpreting reliability, they were found to be less effective tools for making policy decisions and evaluating roadway improvements. The CDF graphs reveal limited information about the frequency with which a regime occurs, or its total contribution to delay. For instance, if an agency decides to improve reliability by mitigating the effects of incidents, it is crucial to quantify the impact incidents have on travel delay. Figure 4.3 and Figure 4.4 address this by showing the relative frequency with which each regime occurs and the contribution of each regime to the total travel delay. Figure 4.5 demonstrates the average travel delay for each regime on I-405 North. It can be observed that the I-405 North normally experiences the largest travel time delay under the heavy traffic conditions.

46 Figure 4.3. Relative frequency of each regime on I-405 North. The CDF graphs are useful for qualitative analysis of reliability. However, it is found that these graphs have some shortcomings in making the quantitative assessments that are desired when evaluating roadway improvements. To test the effectiveness of L02 in evaluating roadway improvements, the research team has examined the “I-405–NE 8th St. to SR 520 Braided Ramps–Interchange Improvements” project, which was completed in early 2012. Specifically, this project aimed to improve traffic flow by building new multilevel “braided” ramps to separate vehicles entering and exiting northbound I-405 between NE 8th Street and SR 520 in Bellevue. Figure 4.6 shows the layout of this improvement project.

47 Figure 4.4. Relative contribution of each regime to travel delay on I-405 North. In order to test the impact of this improvement on reliability, travel times were calculated for I-405 northbound from milepost 12.28 to milepost 15.36. For comparison, the physical extent of this project extends from milepost 13.9 to milepost 14.9. Tuesday–Thursday data were collected January–September 2011 and 2012 for before and after. The gap was created because key elements of this project began opening in early October. These data were then processed in the same method as the route-level data, and CDFs were plotted for normal, incident, and weather regimes. The CDF plots under normal and incident conditions for this analysis are shown in Figure 4.7 and Figure 4.8 and reveal significant improvements in reliability after the project. For example, in Figure 4.8 the interquartile range for TTI under heavy congestion shifted from 1.17–2.04 before the project to 1.06–1.67 after.

48 Figure 4.5. Average travel delay for each regime on I-405 North.

49 Figure 4.6. Design and layout of I-405 Braided Ramps Project.

50 Figure 4.7. Before-and-after TTI CDF for I-405 Braided Ramps Project under normal conditions.

51 Figure 4.8. Before-and-after TTI CDF for I-405 Braided Ramps Project under incident conditions.

52 Figure 4.9. TTI standard deviations for each regime before and after I-405 ramp project. However, the research team found that the CDF graph makes it somewhat difficult to extract quantitative values for reliability. In addition, graphing all regimes simultaneously would require plotting 18 curves on a single graph, which makes the charts less useful. Plotting the standard deviations by regime as a bar graph was found to be more effective for this application. The results are shown in Figure 4.9. This graph shows clear reliability improvement in 8 out of 9 regimes, with only the Normal Heavy regime getting less reliable. An examination of the CDF graph reveals TTI in this regime actually improved up to the 85th percentile, proving that the CDF is still a valuable tool for understanding the whole picture. 4.6 Practical Applications of the L02 Methodology The L02 project’s TTRMS was implemented on the Digital Roadway Interactive Visualization and Evaluation Network platform, which is currently being developed as WSDOT’s data analytics system. DRIVE Net is a framework for a regionwide web-based transportation decision system that adopts digital roadway maps as the base and provides data layers for integrating multiple data sources, including traffic sensor data, incident data, accident data, and travel time data. DRIVE Net provides a practical solution to facilitate data retrieval and integration, and enhances data usability. The system provides users with the capability to store, access, and manipulate data from anywhere as long as they have Internet connections. The goal of the platform is to remove the barriers existing in the current data sets archived by WSDOT and to achieve the integration and visualization of information needed for decision support.

53 The DRIVE Net system adopts the “thin client and fat server” architecture with three basic tiers to the web application: presentation tier, logic tier, and data tier (see Figure 4.10). Analytical tools developed include incident-induced delay forecasting using deterministic queuing theory and GPS-based truck performance measures. By implementing the reliability data generated by L02 onto DRIVE Net, transportation agencies and roadway users have access to the reliability data that have been generated from the project. Providing this easy access to the data is useful in planning future projects to improve reliability as well as in measuring their effectiveness. Regular road users may create a personal DRIVE Net account with customized travel route information to see travel time statistics on their commuting routes and explore potential alternative routes. The reliability data and analysis performed for L02 has been extended from the original study of the I-5/I-405 alternative facility to include SR 520, portions of I-90 and SR 167, and an extended segment of I-5 stretching over 100 miles. Figure 4.11 illustrates this coverage in green.

54 Figure 4.10. DRIVE Net architecture (Wang et al. 2013). DRIVENet Web Server Geospatial Data Transportation Data Data Sources Data Quality Control OpenStreetMap Server Web Mapping Service OpenStreetMap WSDOT Roadway Geometric data HPMS TMC Network … ... Data fusion Data Sources Data Sources Data fusion Loop Detectors Weather INRIX Speed WITS … ... Data fusion Dynamic Routing Incident Induced Delay Calculation Travel Time Performance Measure Real-time Traffic Freight Performance Measure Corridor Sensor Comparison Pedestrian Trajectory Reconstruction Traffic Emission Evaluation Freeway Performance Measure C lien t Sid e Server Side HTTP(S) R Server Statistical Analysis Service File Importing FTP Downloading External Database Connecting …... …...

55 © OpenStreetMap contributors Figure 4.11. Routes available on the DRIVE Net platform for L02 reliability analysis.

56 Using these new data, transportation agencies and roadway users can explore reliability anywhere along these implemented routes simply by inputting mileposts or clicking on the map. Travel time reliability information is available in two different forms: 1. Users can directly view the travel times for varying levels of reliability for a custom route by specifying a starting and ending milepost. A snapshot of this feature is shown in Figure 4.12. 2. Users can specify a starting milepost along with a given amount of travel time, and DRIVE Net can determine how far the user can travel with varying levels of reliability. A snapshot of this feature is shown in Figure 4.13. © OpenStreetMap contributors Figure 4.12. Travel times for varying levels of reliability for a custom route.

57 © OpenStreetMap contributors Figure 4.13. Travel distance with varying levels of reliability. With the depth of reliability information made available on DRIVE Net, transportation agencies can better understand the performance of their roadway networks, and drivers can make better route choices when planning their commutes. For more information, the DRIVE Net platform can be accessed at http://uwdrive.net/STARLab. 4.7 Evaluation of the L02 Objectives Overall, the L02 tools have few shortcomings and effectively help transportation agencies answer five basic questions: 1. What is the distribution of travel times in their system? 2. How is the distribution affected by recurrent congestion and nonrecurring events? 3. How are freeways and arterials performing relative to performance targets set by the agency? 4. Are capacity investments and other improvements really necessary given the current distribution of travel times? 5. Are operational improvement actions and capacity investments improving the travel times and their reliability?

58 The distribution of travel times and how it is affected by recurrent congestion and nonrecurring events is clearly and efficiently shown by creating the CDF charts using the L02 methodology. Comparing performance targets to actual freeway performance is then easily accomplished, as long as targets are expressed in a way that is compatible with the L02 output. For example, agencies should express desired performance in terms of performance at various percentiles, or as the standard deviation of travel time. The need for capacity investments and other improvements is not perfectly addressed by the L02 tools. The research team felt it was necessary to analyze the relative contribution of each regime to the overall reliability and delay. This could not be directly taken from the L02 methods; however, it did provide a strong foundation for such analysis. Finally, the L02 methodology and CDFs were helpful in determining the effectiveness of improvements and investment. However, it is important to note that L02 specifies route-level analysis, which is a much larger scale than most improvements. The research team chose to examine improvements near the segment level and found that plotting standard deviations of travel times could be more helpful for detailed analysis.