Analytical Procedures for Determining the Impacts of Reliability Mitigation Strategies (2012)

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.

177 A p p e n d i x B effect of Ramp Metering on Reliability on i-285 in Atlanta, Georgia Background The Georgia Department of Transportation (GDOT) has aggressively pursued ramp metering as a control strategy on Atlanta area freeways. GDOT started very limited deployment in 1996; by 2005 there were nine operating meters. In 2006, the FastForward program was initiated by the Governor. As part of this program, 161 ramp meters were installed, most of them in 2008. All 10 of the SHRP 2 study sections in Atlanta now have ramp meters installed. Two sections, the northbound and southbound sections of I-75/I-85 (the downtown connec- tor), already had ramp meters beginning in 2005. The four sections on I-285 had meters operating as of July 7, 2008, and the four sections on I-75 had meters operating as of Octo- ber 2, 2008. It was decided to use the four I-285 sections for the analysis as there were sufficient before-and-after data when the analysis was initiated (i.e., approximately 6 months of data for each period). As shown in Chapter 4, 6 months of data provide acceptable estimates of annual reliability if winter weather is not a major factor. The locations of the ramp meters for the analysis sections on I-285 (north side) are listed below. Their operating times expand slightly beyond the beginning and ending of the peak periods, as determined in Chapter 4. • Eastbound between I-75/Cobb and Peachtree–Dunwoody Road—6:15 to 9:30 a.m. (Section 5) 44 New Northside Drive, 44 Riverside Drive, 44 Roswell Road, and 44 Peachtree–Dunwoody Road; • Westbound between Buford Highway and GA 400—6:15 to 9:45 a.m. (Section 6) 44 Buford Highway, 44 Peachtree Industrial Boulevard, 44 Chamblee–Dunwoody Road, and 44 Ashford–Dunwoody Road; • Westbound between GA 400 and I-75—3:30 to 6:30 p.m. (Section 7) 44 Glenridge Drive, 44 Roswell Road, 44 Riverside Drive, and 44 Northside Drive; • Eastbound between Roswell Road and Spaghetti Junction— 3:30 to 6:30 p.m. (Section 8) 44 Ashford–Dunwoody Road, 44 North Peachtree Road, 44 Peachtree Industrial Boulevard, and 44 Buford Highway. Methodology A before-and-after analysis was conducted. The before period was defined as January 1 through June 16, 2008. The after period was defined as July 16 through December 31, 2008. This allowed for a dead zone of 2 weeks before and after ramp meter turn-on to allow the system to stabilize. Two types of controls were used: 1. Control sections—Sections that did not have ramp meters installed. Because of the aggressiveness of Georgia DOT’s program, it was difficult to locate sections with no ramp meters for the analysis period that had similarly high-base congestion levels, although four sections were located. In addition to these sections, a reduced before-and-after period of 75 days before June 16 and 75 days after July 16 was selected for all SHRP 2 sections. As shown in the Before-and-After Analyses of Reliability Improvements

178 Phase 2 report, a period of 75 days of data is insufficient to establish reliability, but is more than adequate to estimate average congestion, as measured by the Travel Time Index (TTI) here. 2. Influencing factors—Demand (vehicle miles traveled [VMT]) and lane hours lost due to incidents were com- piled for the before-and-after periods to see to what degree they might be influencing results. The team was particularly concerned about demand changes; as shown in the Phase 2 report, gas price and availability in the summer of 2008 caused a temporarily sharp drop-off in demand. Results Figures B.1 through B.4 show the TTI for each daily peak period for the four sections receiving ramp metering. The self-imposed dead zone is readily apparent in the June 16 to Figure B.1. Section 5 peak period TTIs. Figure B.2. Section 6 peak period TTIs.

179 July 16 period. Note that Section 5 had a data outage after October 15, which means that any results will be inconclusive for this section (although they have been reported). Tables B.1 through B.4 present the results of the before-and-after analy- sis on the four study sections. Multiple reliability metrics were used to characterize the before-and-after conditions. Also included is the sustainable service rate (SSR) developed in Chapter 4. Two estimates of daily vehicle miles traveled (DVMT) were used: the DVMT in the peak period and in an extended period, including 45-minute shoulders on each side of the peak period. These estimates were used to account for queuing in the peak period, which lowers observed VMT. These results reveal • Base (average) congestion conditions as measured by TTI dropped between 7.5% and 13.4% for the period. Figure B.3. Section 7 peak period TTIs. Figure B.4. Section 8 peak period TTIs.

180 Table B.1. Section 5: I-285 Eastbound from I-75 to GA 400 (Peak Period, 7:15 to 8:45 a.m.; Section Length, 6.860 mi) Before After Change (%) Reliability Metric TTI 1.447 1.338 -7.5 Buffer Index 0.332 0.320 -3.6 Planning Time Index 1.927 1.744 -9.5 Skew statistic . . -2.1 Misery Index 2.117 1.944 -8.2 On-time at 45 mph 34.0% 48.7% 43.2 Mean SSR 1,750 1,740 -0.6 Control Statistic Peak period DVMT 1,019,705 1,010,691 -0.9 Shoulder + peak DVMT 1,694,413 1,642,311 -3.1 Peak incident lane hours lost 7.63 6.22 -18.5 Peak incident shoulder hours lost 32.25 24.83 -23.0 Peak number of incidents 61 45 -26.2 Table B.2. Section 6: I-285 Westbound from I-75 to GA 400 (Peak Period, 4:30 to 6:30 p.m.; Section Length, 6.880 mi) Before After Change (%) Reliability Metric TTI 1.814 1.571 -13.4 Buffer Index 0.708 0.766 8.2 Planning Time Index 3.099 2.774 -10.5 Skew statistic 1.231 2.559 107.9 Misery Index 3.528 3.227 -8.5 On-time at 45 mph 25.4% 44.9% 76.8 Mean SSR 1,720 1,755 2.0 Control Statistic Peak period DVMT 886,378 896,326 1.1 Shoulder + peak DVMT 2,100,381 2,105,177 0.2 Peak incident lane hours lost 26.77 15.24 -43.1 Peak incident shoulder hours lost 93.91 61.94 -34.0 Peak number of incidents 80 95 18.8

181 Table B.3. Section 7: I-285 Eastbound from GA 400 to I-85 (Peak Period, 4:00 to 6:30 p.m.; Section Length, 5.861 mi) Before After Change (%) Reliability Metric TTI 1.958 1.735 -11.4 Buffer Index 0.766 0.843 10.1 Planning Time Index 3.458 3.197 -7.5 Skew statistic 1.855 2.603 40.3 Misery Index 3.963 3.631 -8.4 On-time at 45 mph 24.4% 34.9% 43.0 Mean SSR 1,690 1,740 3.0 Control Statistic Peak period DVMT 870,265 876,770 0.7 Shoulder + peak DVMT 2,440,655 2,461,321 0.8 Peak incident lane hours lost 31.65 30.88 -2.4 Peak incident shoulder hours lost 67.50 70.35 4.2 Peak number of incidents 117 139 18.8 Table B.4. Section 8: I-285 Westbound from GA 400 to I-85 (Peak Period, 7:15 to 9:00 a.m.; Section Length, 5.595 mi) Before After Change (%) Reliability Metric TTI 1.602 1.453 -9.3 Buffer Index 0.287 0.313 9.1 Planning Time Index 2.061 1.909 -7.4 Skew statistic 0.557 0.707 26.9 Misery Index 2.470 2.267 -8.2 On-time at 45 mph 18.7% 34.7% 85.6 Mean SSR 1,920 1,910 -0.5 Control Statistic Peak period DVMT 924,065 923,262 -0.1 Shoulder + peak DVMT 1,758,676 1,713,866 -2.5 Peak incident lane hours lost 15.58 12.93 -17.0 Peak incident shoulder hours lost 31.70 45.63 43.9 Peak number of incidents 71 80 12.7

182 • As demonstrated in Chapter 4, the Buffer Index and the skew statistic generally increased as TTI dropped; the Buf- fer Index increased approximately 10%, and the skew sta- tistic increased on three of the sections as TTI dropped. Section 5 showed a 3.6% decrease in the Buffer Index, but this section also showed the lowest drop in TTI (7.5%). The skew statistic changed dramatically on some sections (up to 107% on Section 6). Further analyses may show that this fluctuation is an aberration, but it is potentially an unstable indicator of changes in reliability. • The Planning Time Index decreased on all sections; decreases ranged from 7.4% to 9.5%. The Misery Index showed the most consistent pattern for the reliability met- rics, decreasing between 8.2% and 8.5% on all sections. • SSR was relatively stable on Sections 5 and 8 (morning peaks), exhibiting slight decreases of 0.5% and 0.6%, respectively. On the two afternoon peak sections, increases of 2.0% and 3.0% were observed. Note that the afternoon peak sections also had higher base congestion levels than the morning peak sections. • Shoulder DVMT was either stable or increased slightly on the two afternoon peak sections. Shoulder DVMT on the morning peak sections decreased by 2.5% and 3.1%. • As might be expected with only 6 months of data confined to weekday periods of approximately 1.5 to 2 hours, inci- dent characteristics varied across the sections, sometimes increasing in severity, sometimes decreasing. • Incident effects were relatively stable on Sections 5, 7, and 8. Section 6 showed a significant drop-off in the time lanes and shoulders were blocked. Table B.5 shows the performance of the control sections for the entire before-and-after periods. These are sections that did not have ramp meters installed during 2008 and are not SHRP 2 study sections. The results show that congestion and reliability were relatively stable on the control sections in the before-and-after period, although their base congestion lev- els were generally lower than the SHRP 2 study sections. Table B.6 shows the change in TTI for the reduced (75-day) before-and-after period. For the nontreatment sections, a general downward trend is apparent in average congestion levels. However, the decrease on three of the four sections with ramp metering was larger than the decrease on the untreated sections. Conclusions Both average congestion and reliability (as measured by the Planning Time Index) showed improvements for the time period after ramp meters became operational, with decreases Table B.5. Performance of Control Sections Period Section TTI Buffer Index Skew Statistic Planning Time Index On-Time at 45 mph (%) Misery Index Before 23 1.081 0.364 31.948 1.474 0.916 1.689 After 23 1.101 0.417 27.518 1.560 0.893 2.098 Before 26 1.349 0.568 3.995 2.115 0.616 2.803 After 26 1.350 0.532 3.167 2.068 0.598 2.632 Before 28 1.041 0.153 11.477 1.200 0.966 1.476 After 28 1.076 0.177 4.790 1.267 0.956 1.660 Before 29 1.243 0.528 7.928 1.899 0.684 2.120 After 29 1.204 0.384 4.080 1.666 0.783 2.134 Table B.6. Change in Base Congestion for All Sections (±75 days from Ramp Meter Turn-On) TTI Section Before After Change (%) 1 1.714 1.561 -8.9 2 1.345 1.251 -7.0 3 1.313 1.288 -1.9 4 2.253 2.247 -0.3 5 1.418 1.278 -9.8 6 1.757 1.510 -14.0 7 1.862 1.693 -9.1 8 1.530 1.520 -0.6 9 1.554 1.468 -5.5 10 1.633 1.522 -6.8 23 1.031 1.026 -0.4 26 1.138 1.276 12.1 28 1.034 1.065 3.0 Note: Sections 5 to 8 are the treatment sections.

183 Table B.8. Performance Measure Comparisons for I-210 Westbound in Los Angeles Time Period Average Travel Time Buffer Index Failure Rate Planning Time Index Skew Statistic Misery Index Before After Before After Before After Before After Before After Before After Peak hour 25.7 0.1 26.5 0.1 0.246 0.006 0.245 0.004 0.051 0.004 0.046 0.003 2.358 0.016 2.410 0.010 1.062 0.048 0.932 0.038 0.391 0.009 0.389 0.010 Peak period 23.9 0.1 25.0 0.1 0.312 0.005 0.303 0.003 0.098 0.003 0.095 0.003 2.327 0.009 2.391 0.010 1.151 0.030 1.050 0.028 0.491 0.007 0.499 0.009 Counterpeak 15.9 0.0 16.8 0.0 0.152 0.003 0.210 0.004 0.039 0.002 0.052 0.002 1.399 0.013 1.530 0.009 4.374 0.223 2.895 0.099 0.466 0.014 0.491 0.015 Midday 14.8 0.0 15.1 0.0 0.040 0.001 0.064 0.002 0.002 0.000 0.016 0.001 1.114 0.003 1.186 0.004 2.989 0.084 4.350 0.121 0.147 0.005 0.286 0.011 Weekday 16.8 0.0 17.2 0.0 0.398 0.004 0.423 0.003 0.148 0.001 0.154 0.001 1.928 0.005 2.024 0.005 24.515 0.280 16.458 0.201 0.850 0.003 0.884 0.003 All year 16.1 0.0 16.5 0.0 0.294 0.004 0.331 0.004 0.124 0.001 0.130 0.001 1.810 0.006 1.899 0.007 34.160 0.442 19.620 0.216 0.844 0.004 0.886 0.004 Note: Standard errors are shown in boldface. Table B.7. Summary Demand, Weather, and Incident Characteristics on I-210 Westbound in Los Angeles Before After Demand Annual average daily traffic (AADT) 140 144 K-factor 5% 5% Weather (number Rain 8 0 of days) Fog 0 0 Snow 0 0 Wind 0 0 Incidents No collision 100 113 Collision, no injury 60 98 Collision, injury and/or fatality 9 14 of roughly 10% and 8%, respectively. The changes in demand probably explain a small amount of the decreases. Incident effects do not appear to be large enough to have a significant influence on the improvements in congestion and reliability. Therefore, the 10% and 8% decreases should be taken as an upper limit. Without a statistical model, it is difficult to know how much to adjust the decreases, but a reasonable estimate would be that ramp meters reduce average congestion by 8% to 9%, and improve reliability by 6% to 7%. Changes in year- long capacity, as measured by the SSR, are in the 2% to 3% range. effect of Adaptive Ramp Meter Control on i-210 in Los Angeles, California The results show little change to slight degradation in conges- tion and reliability due to implementing the adaptive ramp meter control (Tables B.7 and B.8 and Figures B.5 and B.6). However, these results were obtained before adjustment of the metering algorithm. Therefore, these tests will be redone with a different after period. (Note: Travel time density is the frequency percentage for the travel time measurements.) effect of implementing Rapid Clearance policy for Large- Truck Crashes on i-710 in Los Angeles, California Tables B.9 through B.12 and Figures B.7 through B.10 show the results of the analysis. effect of Ramp Meters in San Francisco Bay Area, California Tables B.13 and B.14 and Figures B.11 and B.12 present the results of the analysis. effect of Freeway Service patrol implementation in San diego, California I-8 Westbound Tables B.15 and B.16 and Figures B.13 and B.14 present the results of the analysis.

184 Figure B.6. Travel time density on I-210 westbound in Los Angeles, peak period. Figure B.5. Travel time density on I-210 westbound in Los Angeles, peak hour.

185 Table B.10. Performance Measure Comparisons for I-710 Northbound in Los Angeles Average Travel Time Buffer Index Failure Rate Planning Time Index Skew Statistic Misery Index Time Period Before After Before After Before After Before After Before After Before After Peak hour 19.9 0.1 17.0 0.1 0.337 0.011 0.238 0.007 0.098 0.005 0.065 0.003 2.330 0.040 1.874 0.031 2.401 0.108 1.940 0.081 0.754 0.028 0.682 0.027 Peak period 18.7 0.1 16.1 0.0 0.347 0.009 0.247 0.007 0.109 0.003 0.075 0.002 2.259 0.026 1.784 0.017 1.890 0.055 2.277 0.078 0.803 0.017 0.714 0.020 Counterpeak 19.1 0.1 18.9 0.1 0.358 0.005 0.319 0.005 0.126 0.003 0.100 0.003 2.185 0.017 2.082 0.011 1.721 0.060 1.476 0.043 0.625 0.012 0.538 0.007 Midday 15.3 0.0 14.7 0.0 0.169 0.009 0.124 0.005 0.063 0.003 0.040 0.002 1.657 0.019 1.444 0.013 4.006 0.198 2.305 0.082 0.645 0.018 0.497 0.014 Weekday 16.3 0.0 15.4 0.0 0.343 0.003 0.282 0.002 0.129 0.001 0.100 0.001 1.955 0.006 1.745 0.005 4.278 0.043 4.194 0.039 0.784 0.005 0.673 0.004 All year 15.5 0.0 14.8 0.0 0.325 0.002 0.257 0.002 0.129 0.001 0.097 0.001 1.855 0.007 1.659 0.004 6.328 0.078 8.684 0.131 0.797 0.006 0.671 0.004 Note: Standard errors are shown in boldface. Table B.9. Summary Demand, Weather, and Incident Characteristics on I-710 Northbound in Los Angeles Before After Demand AADT 161 159 K-factor 6% 6% Weather (number of days) Rain 11 6 Fog 0 2 Snow 0 0 Wind 0 0 Incidents No collision 196 162 Collision, no injury 139 121 Collision, injury and/or fatality 17 13 Table B.11. Summary Demand, Weather, and Incident Characteristics on I-710 Southbound in Los Angeles Before After Demand AADT 167 148 K-factor 5% 5% Weather (number of days) Rain 11 8 Fog 3 5 Snow 0 0 Wind 0 0 Incidents No collision 189 156 Collision, no injury 165 165 Collision, injury and/or fatality 12 14

186 Figure B.7. Travel time density on I-710 northbound in Los Angeles, peak hour. Table B.12. Performance Measure Comparisons for I-710 Southbound in Los Angeles Average Travel Time Buffer Index Failure Rate Planning Time Index Skew Statistic Misery Index Time Period Before After Before After Before After Before After Before After Before After Peak hour 18.0 0.1 17.0 0.0 0.167 0.006 0.177 0.004 0.037 0.004 0.019 0.002 1.792 0.016 1.641 0.008 0.529 0.028 1.053 0.042 0.366 0.009 0.311 0.008 Peak period 17.4 0.0 16.5 0.0 0.194 0.005 0.199 0.003 0.043 0.002 0.028 0.002 1.750 0.013 1.622 0.005 0.690 0.029 1.041 0.022 0.399 0.008 0.342 0.006 Counterpeak 16.4 0.1 15.1 0.0 0.262 0.007 0.218 0.007 0.076 0.003 0.061 0.003 1.802 0.021 1.596 0.011 2.573 0.099 3.657 0.138 0.610 0.016 0.604 0.018 Midday 14.8 0.0 14.1 0.0 0.190 0.007 0.103 0.003 0.049 0.002 0.031 0.001 1.527 0.015 1.349 0.011 3.314 0.116 3.513 0.103 0.566 0.022 0.402 0.010 Weekday 15.2 0.0 14.5 0.0 0.272 0.002 0.225 0.001 0.084 0.001 0.066 0.001 1.644 0.005 1.503 0.003 4.457 0.052 5.739 0.059 0.570 0.005 0.475 0.004 All year 14.7 0.0 14.1 0.0 0.280 0.001 0.212 0.002 0.096 0.001 0.070 0.001 1.590 0.004 1.461 0.002 10.589 0.231 10.381 0.101 0.575 0.004 0.473 0.003 Note: Standard errors are shown in boldface.

187 Figure B.8. Travel time density on I-710 northbound in Los Angeles, peak period. Figure B.9. Travel time density on I-710 southbound in Los Angeles, peak hour.

188 Table B.13. Summary Demand, Weather, and Incident Characteristics for I-580 Eastbound in San Francisco Bay Area Before After Demand AADT 110 111 K-factor 5% 5% Weather (number Rain 16 18 of days) Fog 6 11 Snow 0 0 Wind 0 0 Incidents No collision 182 131 Collision, no injury 28 25 Collision, injury and/ or fatality 7 8 Figure B.10. Travel time density on I-710 southbound in Los Angeles, peak period. I-8 Eastbound and SR 52 Westbound Tables B.17 through B.20 and Figures B.15 through B.18 pre- sent the results of the analysis. effect of Capacity improvements on Reliability in Minneapolis–St. paul, Minnesota Analytic Procedures for Determining the Impacts of Reliability Mitigation Strategies Preliminary Investigation of the Before-and-After Study Sections Five sites in the Minneapolis–St. Paul area were selected for the before-and-after study. A brief description of project type and duration is stated in Table B.21. The before study period for each site is approximately the year before the project start date, and the after study period is approximately the year after the project completion date. Due to lack of data for the first half of the after period of Project C, the after period was extended to ensure a full year’s data for the analysis. Table B.22 lists the before-and- after periods for all before-and-after study sections. Additional information about Projects E, G, and H includes the following: • Project E’s improvement work (adding an auxiliary lane) happened right after the operation of the HOT lane project on the same section of I-394. To isolate the effect of the HOT project from the effect of the new auxiliary lane, an additional section on I-394 was chosen to study HOT effects. In this additional section (I-394 eastbound from I-494 to Highway 169), the only improvement for the study period was the HOT project. • All study sections, except Project G, were chosen at approx- imately the locations where improvement projects were located. Due to a lack of data at or near the Project G loca- tion, a southbound section, which is about 3 miles upstream of the project site and separated by a major bottle neck (I-494), was chosen for the analysis. • For Project H, the northbound direction of the highway was selected as the study section. However, different improvement work was done on each side of the highway. The northbound section had a third lane added, and the southbound section had an auxiliary lane added. The interchange conversion from cloverleaf to a folded dia- mond affected both directions. Therefore, the treatment (as in a quasi-experimental design) for Project H was a combination of different improvement projects.

189 Table B.14. Performance Measure Comparisons for I-580 Eastbound in San Francisco Bay Area Average Travel Time Buffer Index Failure Rate Planning Time Index Skew Statistic Misery Index Time Period Before After Before After Before After Before After Before After Before After Peak hour 14.2 0.2 11.1 0.1 0.347 0.014 0.349 0.007 0.126 0.014 0.101 0.004 3.685 0.064 2.986 0.036 1.149 0.120 1.974 0.076 0.666 0.047 0.705 0.018 Peak period 13.1 0.1 10.2 0.0 0.426 0.010 0.337 0.005 0.166 0.005 0.101 0.002 3.738 0.063 2.800 0.016 1.273 0.068 1.621 0.038 0.822 0.024 0.738 0.010 Counterpeak 7.1 0.0 6.5 0.0 0.156 0.004 0.170 0.005 0.014 0.004 0.053 0.002 1.526 0.011 1.585 0.025 1.614 0.119 3.324 0.108 0.363 0.032 0.693 0.028 Midday 7.6 0.1 6.9 0.0 0.140 0.010 0.223 0.006 0.056 0.006 0.080 0.002 1.886 0.081 1.762 0.024 2.514 0.174 4.470 0.129 0.941 0.095 0.802 0.024 Weekday 8.6 0.0 7.5 0.0 0.686 0.008 0.461 0.003 0.207 0.003 0.178 0.001 3.050 0.027 2.263 0.007 5.693 0.100 5.527 0.048 1.329 0.014 0.990 0.007 All year 7.8 0.0 7.1 0.0 0.612 0.012 0.446 0.002 0.174 0.002 0.171 0.001 2.833 0.019 2.136 0.006 7.041 0.132 8.676 0.073 1.412 0.014 0.997 0.005 Note: Standard errors are shown in boldface. Figure B.11. Travel time density on I-580 eastbound in San Francisco Bay Area, peak hour.

190 Table B.15. Summary Demand, Weather, and Incident Characteristics for I-8 Westbound in San Diego Before After Demand AADT 49 46 K-factor 4% 5% Weather (number of days) Rain 9 5 Fog 0 0 Snow 0 0 Wind 0 0 Incidents No collision 6 4 Collision, no injury 6 5 Collision, injury and/ or fatality 2 0 Figure B.12. Travel time density on I-580 eastbound in San Francisco Bay Area, peak period. Peak hour travel time for both the before and after study peri- ods of each section was plotted to examine the improvement effect on average travel time and reliability. The analyzed peak hour was specific to the study section; it was identified by an algorithm designed by the research team. The peak hour travel time frequency distribution for the before and after periods also was plotted to identify the shift in distribution. By examining the plots, the team observed the following gen- eral trends: • Projects B, C, E, and E2 demonstrated reductions in average travel time and improvements in reliability of travel time; • Project G showed increases in average travel time, but it also showed improvements in reliability of travel time; and • Project H showed reductions in average travel time along with deteriorations in reliability of travel time. Due to the relative locations of the study section and the improvement project site of Project G, this section was possibly subject to other influencing factors besides the improvement project itself. For example, if more traffic took southbound Highway 169 due to improved interchange and driving con- ditions, the study section may have experienced an increase in travel time because of the I-494 bottleneck. However, even with increases in average travel time, this section showed an improvement in travel time reliability. Project H is another section that requires further inves- tigation. Multiple improvement projects implemented at the same time could have different effects than the same projects implemented separately. At this time, the effects of confounding factors (e.g., inci- dents and weather effects) have not been studied along with improvement project effects.

191 Table B.16. Performance Measure Comparisons for I-8 Westbound in San Diego Time Period Average Travel Time Buffer Index Failure Rate Planning Time Index Skew Statistic Misery Index Before After Before After Before After Before After Before After Before After Peak hour 6.1 0.0 6.1 0.0 -0.001 0.000 0.002 0.002 0.000 0.000 0.004 0.001 1.000 0.000 1.015 0.001 NA NA NA NA 0.000 0.000 0.090 0.015 Peak period 6.1 0.0 6.1 0.0 -0.001 0.000 0.006 0.000 0.000 0.000 0.002 0.000 1.000 0.000 1.018 0.000 NA NA NA NA 0.000 0.000 0.055 0.005 Counterpeak 6.6 0.0 6.6 0.0 0.106 0.010 0.103 0.010 0.061 0.002 0.056 0.002 1.505 0.025 1.440 0.018 NA NA NA NA 0.892 0.034 0.811 0.029 Midday 6.1 0.0 6.1 0.0 -0.005 0.001 0.015 0.000 0.003 0.001 0.002 0.000 1.005 0.001 1.029 0.000 NA NA NA NA 0.081 0.010 0.058 0.005 Weekday 6.2 0.0 6.2 0.0 -0.013 0.000 0.003 0.000 0.021 0.001 0.018 0.001 1.034 0.001 1.037 0.001 NA NA NA NA 0.298 0.007 0.267 0.006 All year 6.2 0.0 6.2 0.0 -0.013 0.000 0.004 0.000 0.017 0.000 0.015 0.000 1.026 0.000 1.030 0.000 NA NA NA NA 0.254 0.005 0.228 0.005 Note: Standard errors are shown in boldface. NA = Not available (skew statistics could not be computed because 10th and 50th percentile travel times were too close). Figure B.13. Travel time density on I-8 westbound in San Diego, peak hour.

192 Figure B.14. Travel time density on I-8 westbound in San Diego, peak period. Table B.17. Summary Demand, Weather, and Incident Characteristics for I-8 Eastbound in San Diego Before After Demand AADT 49 49 K-factor 6% 6% Weather (number Rain 13 8 of days) Fog 1 5 Snow 0 0 Wind 0 0 Incidents No collision 6 7 Collision, no injury 6 3 Collision, injury and/ or fatality 1 1 Peak hour travel times and peak hour travel time frequency distributions for the study segments are shown in Figures B.19 through B.32. Performance measure comparisons for the study segments are provided in Tables B.23 through B.29. effect of Large-Truck incident Rapid Clearance policy in Atlanta, Georgia Various public and private organizations in metro Atlanta work together as the Traffic Incident Management Enhance- ment (TIME) Task Force to improve the management of traffic incidents. In 2006, the TIME Task Force developed a strategic vision of initiatives to improve TIME services in metro Atlanta. One of several high-priority recommenda- tions was to quickly and safely remove large-vehicle crashes from the roadways. The Georgia Towing and Recovery Incen- tive Program (TRIP) program was developed as part of this strategic vision. TRIP is a recovery incentive program that pays heavy-duty recovery companies a monetary bonus for clearing commer- cial vehicle crashes quickly. TRIP helps to reduce the impact of major traffic incidents in metro Atlanta and to meet TIME’s aggressive clearance goal of 90 minutes or less. The program, implemented in early 2008, covers the following roadways: • I-285 (beltway) and all freeways inside its perimeter 44 I-75, 44 I-85, 44 I-20, 44 GA 400, and 44 GA 166; and • Four hot spots outside of the I-285 perimeter 44 I-85 Northside from I-285 to Pleasantdale Exit, 44 I-75 Northside from I-285 to Windy Hill Exit, 44 I-20 Westside from I-285 to Fulton Industrial Exit, and 44 I-20 Eastside from I-285 to Wesley Chapel Exit. The analysis of the TRIP program proceeded differently from the other before-and-after evaluations. Instead of tracking reliability measures directly, the team considered incident characteristics instead. This decision was due to two factors.

193 Table B.18. Performance Measure Comparisons for I-8 Eastbound in San Diego Average Travel Time Buffer Index Failure Rate Planning Time Index Skew Statistic Misery Index Time Period Before After Before After Before After Before After Before After Before After Peak hour 7.3 0.0 7.1 0.0 0.283 0.008 0.266 0.007 0.085 0.004 0.077 0.004 1.737 0.016 1.647 0.017 3.136 0.233 7.800 0.901 0.575 0.022 0.604 0.023 Peak period 6.9 0.0 6.7 0.0 0.301 0.004 0.287 0.003 0.098 0.002 0.090 0.002 1.671 0.007 1.600 0.008 13.763 0.641 47.245 3.115 0.611 0.013 0.634 0.012 Counterpeak 6.0 0.0 6.0 0.0 -0.004 0.001 -0.006 0.001 0.007 0.001 0.033 0.002 1.018 0.001 1.035 0.005 NA NA NA NA 0.187 0.024 0.279 0.015 Midday 6.0 0.0 5.9 0.0 0.003 0.001 -0.001 0.001 0.010 0.001 0.009 0.001 1.034 0.002 1.022 0.001 NA NA NA NA 0.200 0.017 0.159 0.012 Weekday 6.3 0.0 6.2 0.0 0.169 0.005 0.120 0.006 0.091 0.001 0.080 0.001 1.462 0.004 1.405 0.003 NA NA NA NA 0.579 0.005 0.535 0.004 All year 6.2 0.0 6.1 0.0 0.050 0.003 0.014 0.001 0.071 0.001 0.065 0.001 1.371 0.005 1.336 0.003 NA NA NA NA 0.530 0.004 0.494 0.004 Note: Standard errors are shown in boldface. NA = Not available (skew statistics could not be computed because 10th and 50th percentile travel times were too close). Table B.19. Summary Demand, Weather, and Incident Characteristics for SR 52 Westbound in San Diego Before After Demand AADT 65 60 K-factor 8% 8% Weather (number Rain 9 5 of days) Fog 0 0 Snow 0 0 Wind 0 0 Incidents No collision 11 8 Collision, no injury 2 3 Collision, injury and/ or fatality 2 1 First, most TRIP coverage was on highways not previously identified as SHRP 2 study sections, and sufficient traffic data were not available. Second, the after period was 2008, which was already observed to have lower congestion levels due to decreased demand. A comparison of incident statistics for the before (2007) and after (2008) periods is shown in Table B.30. Because of the TRIP incentive program, average incident duration for large-truck crashes fell almost 13%, and lane hours lost per crash dropped over 9%. During the same period, crashes not involving large trucks showed a slight decrease in average incident duration (3%), but large decreases in lane hours lost (14%). Further examination of the data reveals that in 2008 shoulder hours lost per nontruck crash increased 6% over 2007. This increase coincides with a more aggressive incident management policy instituted in 2008 to move lane-blocking vehicles to the shoulder as rapidly as possible. effect of Capacity improvements near Seattle, Washington I-405 Southbound in Kirkland, Washington Background The I-405 Kirkland Nickel Improvement Stage 1 project expanded the capacity of a bottleneck segment on a major urban north–south Interstate by adding an additional gen- eral-purpose (GP) lane. The project is the first stage of a multistage project to improve traffic conditions along a 7.6-mile segment of I-405 north of Bellevue, Washington, a major suburban city. The project location was a 2-mile southbound freeway seg- ment of I-405, located on the east side of Lake Washington (Seattle lies on the west side of Lake Washington). The seg- ment is part of a freeway commute route that experiences heavy volumes and congestion during the a.m. peak period. Traffic on that route travels toward the Bellevue central busi- ness district. Just before reaching downtown Bellevue, the route also connects to an interchange with SR 520, a major east–west freeway that provides access to downtown Seattle (westbound) and the City of Redmond (eastbound); the latter

194 Table B.20. Performance Measure Comparisons for SR 52 Westbound in San Diego Average Travel Time Buffer Index Failure Rate Planning Time Index Skew Statistic Misery Index Time Period Before After Before After Before After Before After Before After Before After Peak hour 13.2 0.1 9.7 0.0 0.369 0.012 0.206 0.006 0.135 0.005 0.045 0.004 2.329 0.053 2.162 0.021 2.275 0.142 0.834 0.041 1.018 0.038 0.441 0.017 Peak period 13.0 0.1 9.3 0.0 0.427 0.008 0.257 0.004 0.163 0.003 0.068 0.003 2.414 0.037 2.177 0.013 7.613 0.711 0.662 0.017 1.223 0.033 0.524 0.014 Counterpeak 9.3 0.0 5.8 0.0 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 1.000 0.000 1.000 0.000 NA NA NA NA 0.000 0.000 0.000 0.000 Midday 9.3 0.0 5.8 0.0 -0.002 0.000 0.001 0.000 0.000 0.000 0.000 0.000 1.009 0.001 1.008 0.000 NA NA NA NA 0.045 0.009 0.042 0.006 Weekday 10.3 0.0 6.8 0.0 0.290 0.007 0.476 0.002 0.146 0.001 0.271 0.001 1.760 0.005 1.881 0.004 NA NA NA NA 1.111 0.010 0.805 0.006 All year 10.1 0.0 6.6 0.0 0.072 0.007 0.454 0.002 0.130 0.001 0.254 0.002 1.668 0.005 1.834 0.003 NA NA NA NA 1.042 0.009 0.813 0.004 Note: Standard errors are shown in boldface. NA = Not available (skew statistics could not be computed because 10th and 50th percentile travel times were too close). Figure B.15. Travel time density on I-8 eastbound in San Diego, peak hour.

195 Figure B.16. Travel time density on I-8 eastbound in San Diego, peak period. Figure B.17. Travel time density on SR 52 westbound in San Diego, peak hour.

196 Figure B.18. Travel time density on SR 52 westbound in San Diego, peak period. Table B.21. Before-and-After Study Sites in Minneapolis–St. Paul Area Project ID Highway Location Project Description Project Cost Approximate Start Date Approximate Completion Date B I-94 Highway 100 to I-494 Add third lane in each direction. $55M September 2001 Fall 2004 C I-494 Highway 100 to Highway 5 Add third lane in each direction. $71M May 2003 Fall 2005 E I-394 Highway 100 to Highway 169 Add auxiliary lane westbound; high-occupancy toll (HOT) lane project (MnPass). $2M September 2005 October 2005 E2 I-394 I-494 to Highway 169 HOT lane project (MnPass). NA May 2005 August 2006 G Highway 169 Anderson Lakes and Pioneer Trail Convert signalized intersections to diamond interchanges. $20M Summer 2005 Fall 2006 H Highway 100 Highway 7 to Minnetonka Boulevard Add third lane northbound. Add auxiliary lane southbound. Convert Highway 7 interchange from cloverleaf to folded diamond. $7M June 2006 October 2006 Table B.22. Study Periods for Before-and-After Projects Project ID Route Directions Covered Beginning Landmark Ending Landmark Before Period After Period B I-94 Eastbound Highway 100 I-494 September 2000 to September 2001 November 2004 to November 2005 B I-94 Westbound I-494 Highway 100 September 2000 to September 2001 November 2004 to November 2005 C I-494 Eastbound Highway 5/312 Highway 100 April 2002 to April 2003 July 2006 to July 2007 E I-394 Westbound Highway 100 Highway 169 July 2004 to July 2005 November 2005 to November 2006 E2 I-394 Eastbound I-494 Highway 169 July 2004 to July 2005 November 2005 to November 2006 G Highway 169 Southbound T.H. 62 I-494 June 2005 to June 2006 November 2006 to November 2007 H Highway 100 Northbound 36th St I-394 April 2005 to April 2006 November 2006 to November 2007

197 Figure B.19. Peak hour travel time for Project B eastbound. Figure B.20. Peak hour travel time for Project B westbound.

198 Figure B.21. Peak hour travel time for Project C. Figure B.22. Peak hour travel time for Project E.

199 Figure B.23. Peak hour travel time for Project E2. Figure B.24. Peak hour travel time for Project G.

200 Figure B.25. Peak hour travel time for Project H. Figure B.26. Peak hour travel time frequency distribution for Project B eastbound.

201 Figure B.27. Peak hour travel time frequency distribution for Project B westbound. Figure B.28. Peak hour travel time frequency distribution for Project C.

202 Figure B.29. Peak hour travel time frequency distribution for Project E. Figure B.30. Peak hour travel time frequency distribution for Project E2.

203 Figure B.31. Peak hour travel time frequency distribution for Project G. Figure B.32. Peak hour travel time frequency distribution for Project H.

204 Table B.23. Performance Measure Comparisons for Project B Eastbound Travel Time Index Buffer Index Planning Time Index Skew Statistic Misery Index On-Time at 45 mph (%) Time Period Before After Before After Before After Before After Before After Before After Peak hour 1.74 1.09 0.44 0.37 2.51 1.49 0.84 144 3.08 2.11 0.23 0.91 Peak period 1.55 1.06 0.52 0.28 2.36 1.35 1.48 3,943 2.94 1.93 0.38 0.94 Counterpeak 1.23 1.01 0.63 0.02 2.00 1.03 22.73 3.13 2.55 1.20 0.79 0.99 Midday 1.09 1.01 0.12 0.05 1.22 1.06 2.08 18.89 1.80 1.15 0.96 1.00 Weekday 1.23 1.03 0.63 0.07 2.00 1.10 14.32 67.32 2.55 1.26 0.83 0.98 Table B.24. Performance Measure Comparisons for Project B Westbound TTI Buffer Index Planning Time Index Skew Statistic Misery Index On-Time at 45 mph (%) Time Period Before After Before After Before After Before After Before After Before After Peak hour 1.76 1.10 0.68 0.38 2.96 1.52 1.87 75.13 3.70 1.93 0.28 0.91 Peak period 1.86 1.12 0.65 0.36 3.07 1.52 1.55 15.56 3.71 1.90 0.25 0.89 Counterpeak 1.06 1.00 0.19 0.01 1.26 1.01 14.50 2.13 1.87 1.08 0.94 1.00 Midday 1.12 1.01 0.68 0.02 1.88 1.03 22.78 4.13 2.47 1.18 0.90 0.99 Weekday 1.28 1.04 0.97 0.17 2.52 1.22 70.60 2.14 3.31 1.41 0.83 0.98 Table B.25. Performance Measure Comparisons for Project C TTI Buffer Index Planning Time Index Skew Statistic Misery Index On-Time at 45 mph (%) Time Period Before After Before After Before After Before After Before After Before After Peak hour 3.15 2.08 0.76 1.33 5.53 4.85 0.87 4.60 6.64 6.00 0.10 0.34 Peak period 2.68 1.85 0.90 1.32 5.09 4.30 1.19 4.30 6.13 5.47 0.16 0.41 Counterpeak 1.51 1.39 0.58 0.76 2.39 2.45 1.87 9.82 3.32 3.22 0.43 0.66 Midday 1.29 1.10 0.59 0.40 2.04 1.53 44.10 4.06 2.70 1.98 0.66 0.93 Weekday 1.46 1.27 1.27 0.83 3.32 2.33 88.01 21.69 4.43 3.64 0.75 0.85 Table B.26. Performance Measure Comparisons for Project E TTI Buffer Index Planning Time Index Skew Statistic Misery Index On-Time at 45 mph (%) Time Period Before After Before After Before After Before After Before After Before After Peak hour 1.69 1.09 0.34 0.38 2.27 1.51 0.61 12.15 2.54 1.85 0.23 0.91 Peak period 1.70 1.11 0.34 0.44 2.28 1.59 0.56 12.69 2.51 1.90 0.23 0.90 Counterpeak 1.09 1.02 0.43 0.08 1.56 1.10 22.77 143.45 2.03 1.24 0.91 0.99 Midday 1.15 1.02 0.67 0.08 1.91 1.10 52.37 23.95 2.20 1.26 0.85 0.99 Weekday 1.18 1.04 0.70 0.12 2.01 1.16 78.39 43.23 2.27 1.37 0.86 0.98

205 Table B.27. Performance Measure Comparisons for Project E2 TTI Buffer Index Planning Time Index Skew Statistic Misery Index On-Time at 45 mph (%) Time Period Before After Before After Before After Before After Before After Before After Peak hour 1.23 1.11 0.72 0.32 2.12 1.47 52.27 58.17 2.70 1.87 0.81 0.91 Peak period 1.19 1.09 0.72 0.32 2.04 1.43 851.50 7.14 2.59 1.82 0.85 0.92 Counterpeak 1.01 1.01 0.00 0.00 1.01 1.00 0.50 0.50 1.14 1.12 0.99 0.99 Midday 1.00 1.00 0.01 0.00 1.01 1.00 2.13 0.50 1.07 1.08 1.00 0.99 Weekday 1.04 1.02 0.14 0.08 1.18 1.10 7.13 3.13 1.52 1.27 0.98 0.99 Table B.28. Performance Measure Comparisons for Project G TTI Buffer Index Planning Time Index Skew Statistic Misery Index On-Time at 45 mph (%) Time Period Before After Before After Before After Before After Before After Before After Peak hour 2.64 3.27 1.09 0.56 5.52 5.09 1.83 0.94 6.57 5.72 0.25 0.03 Peak period 2.38 2.78 1.25 0.72 5.37 4.77 2.30 0.89 6.43 5.47 0.35 0.09 Counterpeak 1.10 1.14 0.57 0.11 1.73 1.26 51.23 1.29 2.34 1.43 0.95 0.98 Midday 1.21 1.50 1.03 0.82 2.45 2.73 33.09 3.04 3.28 3.70 0.89 0.53 Weekday 1.40 1.63 1.47 1.22 3.45 3.62 112.22 12.55 4.85 4.43 0.84 0.67 Table B.29. Performance Measure Comparisons for Project H TTI Buffer Index Planning Time Index Skew Statistic Misery Index On-Time at 45 mph (%) Time Period Before After Before After Before After Before After Before After Before After Peak hour 1.77 1.37 0.24 0.89 2.20 2.58 0.42 10.47 2.54 4.10 0.15 0.75 Peak period 1.77 1.41 0.25 1.06 2.22 2.91 0.39 12.94 2.63 4.37 0.17 0.72 Counterpeak 1.76 1.26 0.38 0.55 2.43 1.95 0.62 9.66 0.29 2.35 0.29 0.80 Midday 1.09 1.06 0.13 0.09 1.23 1.16 3.03 4.75 1.72 1.23 0.96 0.99 Weekday 1.34 1.16 0.63 0.51 2.19 1.75 20.07 14.68 2.44 2.76 0.78 0.93 Table B.30. Crash Characteristics on Atlanta Highways Affected by TRIP Program, 2007 to 2008 Crash Type Year Number of Crashes Average Incident Duration (min) Lane Hours Lost Lane Hours Lost Per Crash Nonlarge-truck crashes 2007 3,823 48.5 2,909 0.761 2008 4,057 47.0 2,656 0.655 Difference (%) +6.1 -3.1 -8.7 -14.0 Large-truck crashes 2007 417 88.9 852 2.043 2008 420 77.4 778 1.852 Difference (%) +0.7 -12.9 -8.7 -9.3 All truck crashes 2007 4,240 52.5 3,761 0.887 2008 4,477 49.8 3,433 0.767 Difference (%) +5.6 -5.1 -8.7 -13.6

206 completion). Median trip time showed a similar reduction. In addition to a drop in average travel times, the overall reliability of travel on that freeway route improved; on days with outlier travel times at the 80th, 90th, and 95th percentile levels, the peak period travel times dropped at each of those levels. The reduction in outlier travel times and the resulting improvement in travel time reliability were reflected in the overall likelihood of encountering a congested trip on that route on any given weekday (congested trip was defined as a trip with an overall trip speed of 35 mph or less); that likeli- hood dropped significantly, from 65% before construction to 39% after the additional GP lane was opened. The TTI and Planning Index also dropped, reflecting the changes in mobil- ity and reliability, but the Buffer Index was largely unchanged. Figure B.33 illustrates that the reduction in the frequency of congestion is significant throughout much of the a.m. peak period. is a significant suburban employment center that includes, most notably, Microsoft’s headquarters. Before construc- tion, there were three GP lanes and one inside high-occupancy vehicle (HOV) lane in the segment of interest. This project added a fourth GP lane, as well as on-ramp improvements. Results After the opening of the additional southbound GP lane in the Kirkland area north of Bellevue, congestion was reduced, and travel times decreased and became more reliable. Table B.31 summarizes the change in travel time statistics for a typical 16-mile freeway commute route to downtown Bellevue that includes the construction segment. The results show a drop in the average a.m. peak period travel time from 31 minutes (before construction start) to 27 minutes (after construction Table B.31. Travel Time Data for Typical I-405 Trip Route During A.M. Peak Period A.M. Peak Period Travel Time (min) A.M. Peak Period Average Median 80th Percentile 90th Percentile 95th Percentile Skew Frequency of Congestion (<35 mph) Travel Time Index Buffer Index Planning Index 2005 (before construction) 31 30 37 40 44 0.50 65% 1.9 45% 2.8 2007 (during construction) 32 31 39 45 49 0.82 69% 2.0 52% 3.1 2008 (after construction) 27 26 32 37 41 1.53 39% 1.7 49% 2.5 Note: Travel times were based on a typical 16-mile southbound trip from Lynnwood to the Bellevue CBD that included the construction segment. Travel times were averaged over a 6:00 to 9:00 a.m. peak period. The time period each year was fixed (January to June, weekdays only) to minimize effects of seasonal variations. Figure B.33. Reductions in frequency of congestion and average travel time.

207 NE 85th (traffic is moving from bottom to top in the dia- grams). Although congestion upstream from that segment has lessened, congestion persists just downstream from the con- struction location, indicating a possible bottleneck south of the construction segment where the additional lane stops. (In Stage 2 of this project, plans call for an extension of the Stage 1 GP lane an additional mile to the south.) The congestion benefit from this project extended upstream from the construction segment for up to 6 miles along the corridor. A comparison of time–space diagrams of the average speed along the trip route before versus after the new lane was opened (Figure B.34) shows how the magnitude and duration of upstream congestion was reduced during the a.m. peak period after the addition of the GP lane between NE 124th and Figure B.34. Average speed in (top) 2007 and (bottom) 2008.

208 two GP lanes and one inside HOV lane in the segment of interest. This project added a third GP lane. Results The opening of the additional GP lane resulted in noticeably reduced travel times and higher travel time reliability, as well as reduced congestion. Figure B.35 illustrates the drop in average a.m. peak period travel times after the opening of the additional GP lane for a typical 13-mile northbound freeway commute route that included the construction segment. Table B.32 summarizes the change in travel time statistics for the 13-mile northbound trip route. Average a.m. peak period travel time dropped significantly compared with the I-405 Northbound in Bellevue, Washington Background The south segment of the I-405 South Bellevue Widening project expanded capacity at a bottleneck location on I-405 on the east side of Lake Washington by adding a new auxiliary GP lane. The project location was a 2-mile northbound urban freeway segment of I-405 that is part of a freeway commute route that experiences heavy volumes and congestion during the a.m. peak period as traffic approaches the central business district of Bellevue, a major suburban city, as well as a nearby interchange with I-90, a major east–west freeway that pro- vides access to downtown Seattle (westbound) and eastern Washington (eastbound). Before construction, there were Figure B.35. Average a.m. peak period travel time by day. Table B.32. Travel Time Data for Typical I-405 Trip Route During A.M. Peak Period A.M. Peak Period Travel Time (min) A.M. Peak Period Average Median 80th Percentile 90th Percentile 95th Percentile Skew Frequency of Congestion (<35 mph) Travel Time Index Buffer Index Planning Index 2007 (before construction) 35 35 40 43 45 -0.11 92% 2.6 31% 3.4 2008 (during construction) 32 31 40 44 46 0.41 85% 2.4 43% 3.4 2009 (after construction) 21 19 26 28 30 1.26 31% 1.5 44% 2.2 Note: Travel times were based on a typical 13-mile northbound trip from Tukwila to the Bellevue CBD that included the construction segment. Travel times were aver- aged over a 6:00 to 9:00 a.m. peak period. The time period each year was fixed (mid-January to mid-April, weekdays only) to minimize effects of seasonal variations.

209 on any given weekday dropped sharply, from 85% to 92% in previous years to 31% after construction. Figure B.36 illus- trates that the improvement in the likelihood of having a heavily congested trip is significant throughout the a.m. peak period, and continues during the shoulder of the peak period after 9:00 a.m. Interchange of I-405 Southbound and SR 167 in Renton, Washington Background This project built a grade separation ramp connecting the southbound I-405 off-ramp with the southbound SR 167 on- ramp. The project location was an interchange of two major north–south roadways (I-405 and SR 167) in Renton, Wash- ington, just south of Seattle. The interchange experiences heavy volumes and congestion during the p.m. peak period commute. The I-405/SR 167 interchange was one of the worst traffic bottlenecks in the region. This interchange, initially designed as a cloverleaf interchange, became a large bottle- neck with increasing traffic volumes and merging conflicts. In the previous lane configuration, traffic using the collector– distributor lane to exit from southbound I-405 to southbound SR 167 was forced to weave with traffic entering the collector– distributor from northbound SR 167. These merging conflicts created increased congestion on both southbound I-405 and northbound SR 167. The new grade separation ramp elimi- nated the weaving movements by providing a separate elevated lane for the I-405 southbound off-ramp to SR 167. previous 2 years, down from 32 to 35 minutes to 21 minutes after the opening of the new lane. There was a similar reduc- tion in median trip time. In addition to a drop in average travel times, the overall reliability of travel on that freeway route was significantly enhanced. A review of days with outlier travel times at the 80th, 90th, and 95th percentiles showed that average peak period travel times dropped significantly at each of those lev- els. In fact, the new 95th percentile travel time dropped below the previous average travel time. The skew factor grew, but this was more a function of the significant drop in the central tendency of the travel time distribution rather than a higher frequency of outlier travel times. Table B.32 also summarizes the change in the TTI, Buffer Index, and Planning Index values for the a.m. peak period. In this research TTI = (average a.m. peak period travel time)/ (off-peak travel time at 60 mph); Buffer Index = (95th per- centile a.m. peak period travel time - average a.m. peak period travel time)/(average travel time) * 100; and Planning Index = (95th percentile a.m. peak period travel time)/(off- peak travel time at 60 mph). The TTI and Planning Index both dropped noticeably, reflecting the reduced average and 95th percentile trip times and the resulting higher travel time reliability. The Buffer Index was essentially unchanged, but the buffer percentage value was relative to a significantly smaller average travel time. Travel time reliability also can be expressed in terms of the likelihood that a traveler will encounter heavy congestion. Table B.32 shows that the likelihood of having a heavily con- gested trip (overall trip speed 35 mph or less) on that route Figure B.36. Reductions in frequency of congestion and average travel time.

210 to December 31, 2002) in an effort to minimize the effects of seasonal variation. Figures B.37 and B.38 display before-and-after speed con- tours centered near the interchange location; southbound traffic is moving from bottom to top in each figure. The new ramp significantly reduced the bottleneck at the interchange. Results Six months of weekday data from before and after the completion of the ramp were analyzed. Study dates of July 1 to December 31, 2003, were selected and compared with the same 6-month period of the previous year (July 1 Figure B.37. Time–space speed contours of original interchange configuration in 2002. Figure B.38. Time–space speed contours of interchange after installation of grade separation ramp.

211 with the previous year. The same increase was recorded for the median speed. The overall reliability of travel on that freeway segment also improved. A review of days with outlier average peak period speeds at the 80th, 90th, and 95th percentiles showed that speeds went up by approximately 6 to 7 mph at each of those levels. Furthermore, the likelihood of encountering a congested trip (overall trip speed of 35 mph or less) on the segment on any given day dropped sharply, from 67% to 36%. Figure B.39 illustrates that the improvement in congestion was significant throughout the afternoon. The figure repre- sents travel times and frequency of congestion on an extended I-405 segment that includes the interchange area. The 11-mile In 2002, average speeds approaching the interchange stayed below 35 mph for the entire afternoon and evening (10:30 a.m. to 6:30 p.m.) and dropped below 25 mph for approximately 3 hours (1:30 p.m. to 4:30 p.m.). After the off- ramp was separated, average speeds in 2003 never dropped below 25 mph. In addition, the duration of congestion was reduced by more than one-half; average speeds do not fall below 35 mph until 1:30 p.m., and only stay at that level until 5:00 p.m. Table B.33 summarizes the change in p.m. peak period speed statistics for the 4.59-mile trip displayed in the con- tours in Figures B.37 and B.38. The route extends on I-405 from SR 900 to I-5. The results show an 8-mph increase (from 29 to 37 mph) in the average a.m. peak period speed compared Table B.33. Speed Data for Trip Segment During P.M. Peak Period P.M. Peak Period Speed (mph) Frequency of Congestion (<35 mph)Average Median 80th Percentile 90th Percentile 95th Percentile 2002 29 30 24 21 20 67% 2003 37 38 31 28 26 36% Change (%) 25 26 32 28 31 -31% Note: Speeds are based on the 4.59-mile trip from SR 900 to I-5 and were averaged over a 3:00 to 7:00 p.m. peak period. The time period each year was fixed (July 1 to December 31, weekdays only) to minimize effects of seasonal variations. Figure B.39. Reductions in frequency of congestion and average travel time.

212 were over 100% longer than free-flow speeds, but in 2003, they were only 60% longer. The grade separation ramp also significantly improved mainline throughput near the interchange. The bottleneck restricted the flow of vehicles through the segment. Table B.37 illustrates the freeway volumes just upstream of the inter- change. Freeway volumes increased significantly throughout the p.m. and by 16% to 19% during the most congested period of the evening commute (3:00 to 5:00 p.m.). Current State In the 6 years since the ramp was completed, traffic condi- tions throughout the I-405 corridor have steadily declined. Figure B.40 shows the average speeds at the I-405/SR 167 interchange using data from July 1 to December 31, 2008. Peak period speeds are approaching conditions similar to those before the ramp installation. Speeds drop below 25 mph for approximately 2 hours (versus 2 hours in 2002). However, off-peak speeds are still improved from the preramp condi- tions. In 2002, off-peak speeds dropped below 35 mph by 10:30 a.m. In 2008, speeds did not begin dropping below 35 mph until about 1:15 p.m. trip extends from the major interchange at I-90 to the inter- change at I-5. The travel time curves show that travel time improvements as a result of the new ramp were not restricted to the peak period. Travel times during the off-peak early afternoon were approximately 3 minutes faster after the ramp installation. In addition, the frequency of congestion histo- grams shows that the onset of congestion on the segment was delayed until later in the afternoon. Table B.34 displays travel time statistics for the p.m. peak period for the trip from I-90 to I-5. Mobility and reliability measures in Table B.35 were calculated based on the average, 95th percentile, and free-flow (trip time based on 60 mph) travel times. TTI compares the average travel time during the peak period to the free-flow travel time. The construction of the ramp resulted in increased mobility and a reduction in TTI. The Buffer Index and Planning Index, which measure the reliability of a trip, showed reliability improvements after the construction of the ramp. The Buffer Index measures the amount of extra time a traveler should budget to ensure an on-time arrival 95% of the time. In 2002, a traveler would have needed to budget an extra 7 minutes; with the new ramp, the traveler needed to budget less than 5 minutes extra. These improvements were more pronounced when focused specifi- cally on the segment most affected by the ramp. Table B.36 displays the mobility and reliability indices for the 4.59-mile segment near the ramp. TTI for the shorter segment dropped from 2.04 to 1.63, meaning that in 2002, average travel times Table B.34. Mobility and Reliability Measures for 11.08-Mile Trip from I-90 to I-5 P.M. Peak Period (3:00 to 7:00 p.m.) Travel Time (min) Average 95th Percentile Free Flow 2002 (old ramp configuration) 18.9 26.0 11.08 2003 (after installation of grade separation ramp) 17.8 22.5 11.08 Table B.35. Mobility and Reliability Measures for 11.08-Mile Trip from I-90 to I-5 P.M. Peak Period (3:00 to 7:00 p.m.) TTI Buffer Index Planning Index 2002 (old ramp configuration) 1.71 37% 2.35 2003 (after installation of grade separation ramp) 1.61 26% 2.03 Table B.36. Mobility and Reliability Measures for 4.59-Mile Trip Near Ramp P.M. Peak Period (3:00 to 7:00 p.m.) TTI Buffer Index Planning Index 2002 (old ramp configuration) 2.04 48% 3.01 2003 (after installation of grade separation ramp) 1.63 41% 2.31 Table B.37. I-405 Volumes Near I-405/SR 167 Interchange I-405 Southbound Throughput (vehicles/hour) Time Period 2002 2003 Change (%) 12:00 to 1:00 p.m. 3,057 3,529 15 1:00 to 2:00 p.m. 3,015 3,538 17 2:00 to 3:00 p.m. 2,952 3,466 17 3:00 to 4:00 p.m. 2,752 3,271 19 4:00 to 5:00 p.m. 2,743 3,187 16 5:00 to 6:00 p.m. 2,804 3,220 15 6:00 to 7:00 p.m. 2,869 3,236 13

213 on the east side of the lake, as well as business development on the east side (e.g., Microsoft). In August 2001, WSDOT began to use the ramp meters to try to alleviate heavy east- bound morning congestion. Results Six months of weekday data from before and after the initia- tion of morning ramp metering were analyzed. Study dates of January 1 to June 28, 2002, were compared with the same 6-month period of the previous year (January 1 to June 29, 2001) in an effort to minimize the effects of seasonal variation. Table B.38 summarizes the change in speed statistics for the 4-mile eastbound section from I-5 to just east of the Ever- green Point Floating Bridge. The results show an average 4 mph (from 32 to 36 mph) increase in the average a.m. peak period speed compared with the previous year. The same increase was recorded for the median speed. The overall reliability of travel on that freeway segment also improved. A review of days with outlier average peak period speed at the 80th, 90th, and 95th percentiles showed that speeds went up by approximately 4 mph at each of those levels. The speed benefits from the ramp metering extended through the project segment. Figure B.41 shows before-and- after speed contours of the segment affected by ramp meter- ing. In 2001, average speeds near the ramps stayed below 25 mph for over 1.5 hours. After ramp metering was initiated, Washington State DOT (WSDOT) has a current project to add a new interchange upstream of the grade separation ramp. This interchange is expected to improve access to downtown Renton and relieve some of the traffic demand at the I-405/SR 167 interchange. effect of Ramp Metering in Seattle, Washington SR 520 Eastbound Background The SR 520 Ramp Metering project managed congestion on the mainline by using ramp metering to control the frequency of vehicles entering the roadway on two on-ramps to SR 520 eastbound. SR 520 is one of two east–west roadways across Lake Wash- ington, which forms the eastern boundary of Seattle. The roadway is heavily used by commuters in both directions. The traditional eastbound evening commute from downtown Seattle to the more suburban east side of Lake Washington via the SR 520 Evergreen Point Floating Bridge has been man- aged by ramp meters on the on-ramps at Montlake and Lake Washington Boulevards (just before reaching the bridge) since 1986. Over the years, traffic conditions in what had tra- ditionally been the reverse commute direction (i.e., east- bound in the morning) worsened. This congestion was in part due to the growth of Bellevue, the major suburban city Figure B.40. Time–space speed contours in 2008 at I-405/SR 167 interchange.

214 significant throughout the a.m. peak period, and it demon- strates that ramp metering delayed the onset of congestion for the segment. Ramp metering also improved the mainline throughput of the segment. Table B.39 illustrates the freeway volumes near the Montlake Boulevard on-ramp. Freeway volumes increased by 13% to 15% during the most congested period of the morning commute (7:00 to 9:00 a.m.). Table B.40 displays travel time statistics for a typical com- mute through the area affected by ramp meters, the 14.8-mile commute from Seattle to Redmond. Mobility and reliability measures were calculated based on average, 95th percentile, and free-flow (trip time based on 60 mph) travel times. TTI compares the average travel time during the peak period to average speeds in 2002 only dropped below 25 mph for a third of that time (approximately 30 minutes). In 2001, speeds dropped below 25 mph by 7:30 a.m.; in 2002, the onset of congestion was delayed by about 15 minutes. In addition to delaying the onset, the lowest average speeds did not drop below 20 mph in 2002; in contrast, during 2001 they dropped below 20 mph for about 20 minutes during the peak conges- tion period. The main goal of ramp metering is to reduce the conges- tion on the mainline. Table B.38 shows that the likelihood of encountering a congested trip (overall trip speed of 35 mph or less) on the sample route on any given day dropped from 63% to 45% after the implementation of ramp metering. Fig- ure B.42 illustrates that the improvement in congestion was Figure B.41. Time–space speed contours. Table B.38. Speed Data for Trip Segment During A.M. Peak Period A.M. Peak Period Speed (mph) Frequency of Congestion (<35 mph)Average Median 80th Percentile 90th Percentile 95th Percentile Skew 2001 32 32 27 26 24 -0.20 63% 2002 36 36 31 29 28 0.98 45% Change (%) 12% 14% 16% 15% 13% -18% Note: Speeds were based on the approximately 4-mile trip from I-5 to just east of the Evergreen Point Floating Bridge and were averaged over a 6:00 to 9:00 a.m. peak period. The time period each year was fixed (January 1 to June 30, weekdays only) to minimize the effects of seasonal variations.

215 Table B.39. SR 520 Eastbound Throughput at Montlake Boulevard (vehicles/hour) Time Period January to June 2001 January to June 2002 Change (%) 6:00 to 7:00 a.m. 2,261 2,042 -10 7:00 to 8:00 a.m. 2,329 2,639 13 8:00 to 9:00 a.m. 2,128 2,454 15 9:00 to 10:00 a.m. 2,209 2,424 10 Table B.40. Travel Time Statistics for A.M. Commute from Seattle to Redmond A.M. Peak Period (6:00 to 9:00 a.m.) Travel Time (min) Average 95th Percentile Free Flow 2001 (before metering) 19.1 23.3 14.8 2002 (after metering) 18.6 22.2 14.8 Figure B.42. Reductions in frequency of congestion and average travel time. the free-flow travel time. The Buffer Index and Planning Index both measure the reliability of a trip based on the amount of extra time a traveler should budget to ensure an on-time arrival 95% of the time. These indices only showed small mobility and reliability improvements after the addi- tion of ramp metering (Table B.41). The travel time improve- ments produced by the metered ramps may have been dampened by the rest of the trip, since the portion of the Seattle-to-Redmond route that is east of Lake Washington usually operates at near free-flow speeds. The travel time improvements were more pronounced when focused specifi- cally on the segment most affected by the metering. Table B.42 displays the mobility and reliability indices for the 4-mile seg- ment near the ramp. TTI for the shorter segment dropped Table B.42. Mobility and Reliability Measures for 4-Mile Segment near the Ramps A.M. Peak Period (6:00 to 9:00 a.m.) Travel Time Index Buffer Index Planning Index 2001 (before metering) 1.87 32% 2.46 2002 (after metering) 1.66 31% 2.17 Table B.41. Mobility and Reliability Measures for A.M. Commute from Seattle to Redmond A.M. Peak Period (6:00 to 9:00 a.m.) Travel Time Index Buffer Index Planning Index 2001 (before metering) 1.29 22% 1.57 2002 (after metering) 1.25 20% 1.50

216 Figure B.43. Time–space speed contours in 2008 on SR 520 segment affected by ramp metering. from 1.87 to 1.66; that is, in 2001 travel times were over 87% longer than travel times at free-flow speeds, but in 2002, they were only 66% longer. Current State In the 7 years since the ramp metering was initiated, traffic conditions on SR 520 have steadily declined. Figure B.43 shows the average speeds in the segment affected by ramp metering using data from January 1 to June 30, 2008. Peak period speeds are now slower than speeds seen in 2001 before metering. Speeds remain below 35 mph from 7:15 to 10:00 a.m. and drop below 20 mph for almost 1.5 hours (versus 20 minutes in 2001). In addition, throughput vol- umes are lower in 2008 than those seen immediately after starting ramp metering (2,589 vehicles/hour from 7:00 to 8:00 a.m.; 2,308 vehicles/hour from 8:00 to 9:00 a.m.), although they are not as low as the premetering levels. WSDOT is in the beginning stages of a new project to replace the SR 520 Evergreen Point Floating Bridge. A con- tinuous HOV lane and rebuilt on- and off-ramps are planned to improve mobility and reliability on the roadway.