Pilot Testing of SHRP 2 Reliability Data and Analytical Products: Southern California (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.

The inconsistency in these definitions of recurring delay presents two problems. First, the user cannot multiply the vehicle-hours of delay shown in the Results Summary (see Figure 5.2 for an example) by the appropriate value of time and reliability ratio to estimate the congestion costs. Second, the value of recurring delay reported by the C11 tool includes incident-related delay and does not correspond to the delay benefits reported in traditional benefit-cost models. As a result, adding the cost of unreliability (as reported by the C11 tool) to the delay benefits in a standard benefit-cost model would underestimate the benefits by ignoring the incident-related delay savings. Difficult to Use Reliability Ratios from Other Sources As described above, the C11 tool estimates the cost of unreliability as the difference in the 50th and 80th percentile TTI figures. This can be seen in the way that the model estimates travel time equivalents. Equation 3 shows the relationship provided in the C11 technical documentation (Cambridge Systematics et al. 2013b): Equation 3: TTI𝐹𝐹(VT) = TTI50 + 𝛼𝛼 × (TTI80 − TTI50), where TTIe(VT) is the TTI equivalent on the segment 𝛼𝛼 is the Reliability Ratio (value of reliability/value of time) In this equation, the reliability ratio is defined as the difference in the 50th percentile and 80th percentile TTI. Other recent studies of the value of reliability (especially those in Europe) define the reliability ratio in terms of a single standard deviation in travel time. This is roughly equivalent to the difference in the 50th and 84th percentile TTI (assuming a one-tailed normal distribution). Through discussions with the C11 development team, the study team determined that the C11 tool uses a value for the reliability ratio from a U.S. study that defined reliability in terms of the 50th and 80th percentiles. So the tool estimates reliability benefits correctly, using this estimation of the reliability ratio. However, if a user were to change the reliability ratio to use a locally adopted value, the user should make sure it is defined in terms of the 50th and 80th percentile. By way of comparison, the L07 tool (described next in this report) uses a reliability ratio defined in terms of the standard deviation. This definition is consistent with more recent valuation studies. 5.3 Baseline Condition Estimation of the C11 Reliability Analysis Tool The first step in testing the C11 Reliability Analysis Tool was to make sure that the tool estimated reliability measures consistent with real-world conditions on the facilities. As indicated previously, the study team was able to calibrate the C11 tool to baseline conditions by adjusting the peak capacity and the hourly distribution of demand. 76

The calibration process consisted of running the tool with no adjustments, and then adjusting the peak capacity and hourly distribution of demand until the tool’s TTI results were as close as possible to real-world conditions. Real-world conditions were measured using Caltrans PeMS. The PeMS data consists of infrastructure-based sensor recordings along each of the freeway facilities over multiple years. The study team used data for 2010 as the baseline year for consistency with the CSMPs developed for the facilities. As seen in Figure 5.4, the C11 tool allows the user to utilize one of two options to input capacity: 1. Enter the peak capacity manually, or 2. Select the terrain type, allowing the tool to automatically calculate peak capacity. The pilot team initially estimated reliability using the terrain type method. These results were later adjusted by using the manual peak capacity method. Figure 5.4. Capacity data input in the C11 Reliability Analysis Tool. Initial Run In the initial run of the tool for the I-210 facility, the team allowed the tool to determine the peak capacity based on terrain and made no adjustments to the tool’s default hourly distribution. 77

When Flat was entered in the Terrain field, the tool calculated a peak capacity of 2,233 vehicles per hour per lane (vphpl). The resulting TTI curve (represented by the nearly level blue line in Figure 5.5) showed values near 1.0 for the entire day. This result is not representative of the actual baseline TTI curve, which is calculated using PeMS data and represented by the red line. The PeMS data indicate an increase in TTI in the p.m. peak period until 5:00 p.m., after which the TTI gradually decreases back to a value of 1.2 by 7:00 p.m. Figure 5.5. Mean TTI on the I-210 with no adjustments. Adjusting Peak Capacity Instead of utilizing the tool’s automatic peak capacity calculation based on terrain type, the team then attempted to manually enter in a lower peak capacity value to force the TTI values upward to match the PeMS TTI curve. As seen in Figure 5.6, this strategy was effective in bringing the p.m. peak values up to match PeMS values, but it also caused the tool’s a.m. TTI values to peak in a manner that does not correspond to the actual baseline TTI values, which are at or near 1.0. 78

Figure 5.6. Mean TTI on the I-210 with adjustment to peak capacity only. Adjusting Peak Capacity and Hourly Distribution of Demand After realizing that the TTI hump in the a.m. peak was likely the result of the C11 tool applying a standard hourly distribution of demand to all runs, the study team discovered a default hourly distribution of demand in a hidden password-protected tab. Once these figures were modified to reflect the hourly distribution of traffic volumes along the facility (as a proxy for demand) and the peak capacity was readjusted, the tool was able to produce a curve that, while far from perfect, more closely resembled PeMS data (Figure 5.7). However, the tool still produced a small TTI hump in the a.m. peak period that cannot be found in the baseline TTI data from PeMS. In addition, the TTI peak in the p.m. peak period occurs approximately 2 hours earlier than the actual hump that occurs in the baseline TTI data reported in PeMS. The observations do not render the C11 tool unfit for practical application, but they do suggest that the algorithms for estimating reliability are inadequate to closely capture reliability along the I-210 facility. The tool does a better job of estimating reliability for the I-5 facility, as described in the next section. 79