TRANSPORTATON RESEARCH RECOR 1398. 31 Study of Freeay Bottlenecks in Texas ]OHN RNGERT AN THOMAS URBANK Observations of flo rates much higher than 2, passenger cars per hour per lane and the recent revision of the multilane highay chapter in the Highay Capacity Manual have led to questioning the current value of freeay capacity and the speed-flo relationship. An analysis of free-flo and queue discharge flo rates at three freeay bottlenecks in Texas found less variability in queue discharge flo rates than in free-flo flo rates. Average free-flo flo rates ranged from 2,96 to 2,21 vehicles per hour per lane (vphpl) across all lanes, hereas queue discharge flo rates averaged approximately 2,175 vphpl for the study sites. n addition, higher flos did not occur in free-flo conditions in all cases. As a result of lane interaction, some lanes are prematurely transitioned into queue discharge ithout reaching high flo rates in free-flo conditions. uring the past decade, much attention has been given to freeay capacity and the relationship of speed, flo, and density for freeay bottlenecks. Freeay capacity plays a critical role in the planning, design, and operation of freeays in general and urban freeays in particular. As a result of steadily increasing congestion in urban areas throughout the United States, traffic engineers and transportation planners have identified problems associated ith the freeay capacity numbers given in the 1985 Highay Capacity Manual (HCM) and the corresponding speed-flo relationship (J). Many studies have concluded that flos measured on freeays are frequently greater than the values given as capacity. A study by Hurdle and atta (2) in 1983 concluded that the value of 2, passenger cars per hour per lane (pcphpl) as still a good estimate of capacity. A more recent study by Agyemang uah and Hall (3) concluded that the capacity flo rate as approximately 2,3 pcphpl. A study by Urbanik et al. ( 4) found peak 15-min flo rates beteen 2, 1 and 2,3 vehicles per hour per lane (vphpl) at four sites in Texas. n addition to the measurement of high flo rates, many studies have concluded that a reduction in capacity occurs once a queue forms. Hall artd Agyemang-uah concluded that the flo rate dropped from 2,3 to 2,2 pcphpl in queue discharge conditions (3,5). f to capacities exist, a possible problem exists ith the use of the peak 15-min flo because it ould give a flo that may not be sustainable at higher demand levels here queue discharge occurs. Although many studies on freeay capacity have been done, a good understanding of the characteristics of flo during free-flo and queue discharge has not been achieved. The characteristics of flo at freeay bottlenecks, hether a reduction in flo exists once a queue forms, the shape of the speed-flo relationship, and recommendations regarding capacity are examined in this paper. Because the definition of J. Ringert, Kittelson & Associates, nc., 1455 Response Road, Suite 12, Sacramento, Calif. 95815. T. Urbanik, Texas Transportation nstitute, Texas A&M University, College Station, Tex. 77843. capacity plays a significant role in the determination of the value to use as capacity, the maximum sustainable flo ill be the focus of this paper. Maximum sustainable flo is the maximum flo rate that can be maintained for an indefinite period ith sufficient demand. For capacity to be useful, it is important that it be sustainable and repeatedly achievable. STUY STES ata ere collected at three study sites. A schematic of each study site is shon in Figures 1, 2, and 3. The primary study site that as used to evaluate the operation at freeay bottlenecks as on US-29 in Houston, Texas. This study site as chosen because of the frequent occurrence of queues during the p.m. peak period. The other to study sites ere on 1-41 in San Antonio, Texas, and at the merge of 1-35 and US-67 in allas, Texas. These to study sites ere chosen to validate the flo characteristics found at the US-29 site. These locations ere chosen because of the occurrence of congestion during a.m. peak periods. RESULTS OF ANALYSS OF US-29 ata Collection ata ere collected for 15 days at the US-29 study site. The data ere collected for 2 to 3 hr on each day using pairs of inductive loops in each traffic lane. Analysis of Flos at US-29 Study Site To be consistent ith existing convention, 15-min peak flos ere measured for each of the sample days. The average peak 15-min flo rates are given in Table 1. The flo rates in Table 1 represent the peak 15-min flo rates during the study period across all lanes for the 15 samples. ndividual lanes can have higher peak 15-min flo rates occurring at different times. Although truck traffic as present, no attempt as made to adjust for truck passenger car equivalents in the preliminary analysis. Truck percentages averaged 3. percent for all lanes combined. The highest percentages ere in the outside lane (Lane 3), hich averaged 5.3 percent trucks, hereas the inside lane (Lane 1) had the loest average truck percentage, ith. 7 percnt. These truck percentages reflect the truck traffic for the study periods. Although peak 15-min flo rates are the existing convention detailed by the 1985 HCM, it is likely that they are not
32 TRANSPORTATON RESEARCH RECOR 1398 c:s =' 1. 35m 15 m.. :; 3.3S m = :i -- 3.3S m Lane 2 1....a ; s.. -- 3.35 m Lane 1 Permanent lnducllye Loop FGURE 1 US-29 at Tidell study site..3 m nner Shoulder \ 'Cl i= Concrete Median Barrier n u sustainable. The to-capacity hypothesis asserts that to separate capacities exist: one during free-flo conditions and one hen demand exceeds capacity, creating queue discharge conditions. Therefore, distinguishing beteen regions is necessary to ensure that a measurement does not contain data from both regions at the same time. To compare the characteristics of flo before and during the period hen demand exceeds capacity, the point at hich demand exceeds the service rate of the bottleneck as identified. emand is considered to exceed the service rate hen vehicle speed is controlled by the service rate of the bottleneck, thereby making vehicles ait to resume their desired speed. The excess demand generates queues and produces queue discharge flo. The point at hich a rapid drop in 3- sec average speed occurs, hich is shon conceptually in Figure 4, as determined to be the beginning of queue dis- r 3.35 m Outer Shoulder 152 m 1 / 111-35 3.66m Lane 4 3.66m Lane 3 3.66m Lane 2 3.66m Lane 1 4.27 m nner Shoulder Concrete Median Barrier Permanent nductive Loops FGURE 2 1-35/US-67 study site.
Ringert and Urbanik 33 6.1 m,..,. 76m 61111 122 m 3 m Outer Shoulder 3.66 m Lane 3 -- 3.66m Lane 2 3.66m Lane 1 Bridge Perm nent Jnductln Loop1 3 m nner Shoulder Concrete Medl n Barrier FGURE 3 1-41 study site. charge. Vehicles are forced to a loer rate of speed than previously desired under prevailing conditions. The folloing three time periods ere used in the analysis: The 5 min immediately before the speed drop (free flo); The 5 min immediately after the speed drop (queue discharge); and The entire period after speed drop until end of queue discharge, formation of a donstream queue, or the end of data collection (queue discharge). These time periods in relation to a typical speed profile are illustrated in Figure 4. The flo rates ere calculated from the average headays during each interval. n most cases the TABLE 1 Peak 15-min Flo Rates Across All Lanes at US-29 at Tidell Highay Observation Peak 15-Minute Flo Rate (vphpl) Lane 1 Lane 2 Lane 3 Average U.S. 29 1 2336 2256 2348 2313 2 232 22 214 222 3 23 2244 2456 2333 4 238 224 26 2227 5 2288 2172 218 2213 6 2368 224 2312 2295 7 226 2224 212 221 8 2228 2196 2228 2217 9 222 2268 2132 227 1 2384 2256 2188 2276 11 2496 2244 2172 234 12 2252 222 2348 2273 13 248 2344 2156 233 14 2248 2356 228 2211 15 2312 224 268 227 Average 232 2244 2196 2253
34 TRANSPORTATON RESEARCH RECOR 1398 c.. CJ) FGURE 4 profile. FREE-FLOW TOAL QUEUE SCHARGE PERO TME Time intervals used for analysis, typical speed The F-ratio indicated that, for the most part, the to time periods have unequal variances. The variance before the speed drop as significantly higher than the variance after the speed drop for all 15 samples for Lane 1, for 9 samples for Lane 2, and for 1 samples for Lane 3. Because the variances ere significantly different in several cases, the mean headays before and during queue discharge ere compared using a t test approximation for significance of difference beteen to means ith different variances. The results of this t-test indicated that only three samples in Lane 1, to samples in Lane 2, and six samples in Lane 3 had statistically different mean headays before the speed drop compared ith after the speed drop. Therefore, most samples did not experience a statistically significant change in mean flo after the speed drop. Because of the high variance in headays and corresponding flos, a large change is required to be statistically significant. speed drop occurred at the same time in all lanes, but for consistency beteen lanes the location of the speed drop for each sample as determined on the basis of highest flo lane (Lane 1 at the US-29 site). Statistical Analysis of Speed rop A statistical analysis of the speed drop as performed in to steps. To determine hether a significant change in flo characteristics occurred after the speed drop, an F-ratio as calculated to compare the variances of headays during the 5- min intervals before and after the speed drop. A t-test as then performed to determine hether significant changes in mean headays occur after the speed drop. n using both the F-ratio and the t-test, it as assumed that the sample mean headays ere independent and the headays ithin them have a normal distribution. These assumptions are considered valid because of the large sample of headays and the central limit theorem, hich states that as sample size increases, the distribution of sample means approaches normality. Average Flos for all Samples Combined To evaluate the characteristics of flos beteen days, statistics ere calculated for the mean flo rates for the 15 sample days. ndividual flo rates for each sample ere calculated from mean headays. The statistics for the sample flos are given in Table 2. Table 2 gives the means, standard deviations, and a 95 percent confidence interval on the means for each time interval for the 15 days on hich data ere collected. The confidence interval assumes that the distribution of sample means follos a normal distribution, hich should be applicable because a relatively large sample size as used. The average flos sho some interesting trends. Whereas the mean average flo rate across all lanes does not change much for the 5-min period before the speed drop compared ith the 5-min period after the speed drop during queue discharge, the distribution of traffic across the lanes does change. n Lane 1 the mean traffic volume substantially increased in queue discharge. The opposite occurred in Lane 3, here the TABLE 2 Statistics for aily Averages for US-29 at Tidell Time Period. Statistic Lane 1 Lane 2 Lane 3 Average 5-min Before Average Flo 276 22 221 296 Speed rop (Free- (vphpl) Flo) Std ev. 136 159 187 132 95% C(+/-) '75 88 14 65 5-min After Average Flo 2266 235 1989 297 Speed rop (vphpl) Std. ev. 145 134 193 115 95% C(+/-} 8 74 17 64 Total Queue Average Flo 2246 2161 29 2166 ischarge (vphpl) Std. ev. 81 68 11 67 95% C (+/-} 45 38 56 37
Ringert and Urbanik average volume decreased after the speed drop in queue discharge. This could be explained by the fact that in free-flo conditions Lane 1 has not reached its maximum flo and operates at very high speeds hile Lane 3 is already becoming congested. When Lane 3 becomes congested and slos don even further, traffic merges into Lanes 2 and 1 and subsequently drops the speed and transitions the flo into queue discharge conditions. Therefore, if the maximum flo occurs in free-flo conditions, it ould not be seen in Lane 1 because Lane 1 transitioned from free flo ith a flo rate belo maximum flo directly to queue discharge. The analysis of means points out another problem ith using free-flo conditions to obtain a maximum sustainable flo rate. Although a maximum flo rate in free-flo conditions exists, in certain lanes it may never be reached, and therefore it ould be impossible to achieve. For example, suppose Lane 1 had a maximum flo of 2,4 vehicles per hour (vph), but because of the turbulence created by Lanes 2 and 3, the flo rate in Lane 1 transitioned directly from 2,76 to 2,266 vph in queue discharge as the data suggest. The flo rate of 2,4 vph ould never be reached and therefore could not be considered the maximum flo rate. Although it appears that for individual samples the variance decreased in queue discharge, the variance among samples (standard deviation squared) remained high in the 5-min period after the speed drop. One explanation for this occurrence is the ay in hich the boundary of the free-flo and queue discharge conditions as chosen. As shon in Figure 4, the boundary as determined on the basis of the speed drop. As can be seen in Figure 5 for Lane 1, a rapid speed drop into hat appears to be queue discharge conditions does not occur in all samples. To investigate the effects of samples that did not appear to drop directly into queue discharge, the samples that appeared to drop directly into queue discharge ere evaluated. The results of the evaluation of samples that had rapid speed drops into queue discharge indicate that the variance (and standard deviation) beteen samples decreases in the 5- min interval after the speed drop. Speed-Flo Relationship Figure 6 shos the speed-flo plots for all samples combined for each lane. Five-min averages ere used for the plots. The lines represent the estimated shape of the speed-flo relationship on the basis of the data. The ends of the lines are equal to the average total queue discharge flo. Note that some data points include donstream congestion, also shon in Figure 6. The hypothesized speed-flo relationship resulting from the analysis is plotted in Figure 7, shoing that the operational characteristics of each individual site determine the capacity. Because of lane interaction, hen Lane 3 transitions into queue discharge the other lanes quickly follo ithout reaching their theoretical maximum flo. As mentioned previously, this transition may be nearly instantaneous in some cases. The dashed lines sho the hypothesized shape of the end section of the speed-flo relationships in Lanes 1 and 2, assuming Lane 3 does not break don and prematurely transition Lanes 1 and 2 into queue discharge. 13 12 -SPEE 11 2 1... 9 c 8 3: 7 +-Jl...Jl-Hji1+JJ--:+--l+-----J.!!--=--==-tlillf---.lJF15 g a.. LL. 6 Vl 5 4 3 13 12 11 21... 9 c 8 7 +:-1Htt-t-tt-+-l-t-'--Hl-:+-:!-----<------Hr+ 15 a.. Vl 6 5 4 nstrm Queue 1-.. 5 2 4 6 8 1 12 TME (MN) 3:3-5:3 2 4 6 8 1 TME (MN) 3:3-5:3 FGURE 5 Types of speed drops during transition to queue discharge, 1 min, Lane 1: top, rapid transition (sample 2); bottom, unstable transition (sample 6). Results of US-29 Analysis From the analysis of the US-29 site the folloing observations can be made: The variance in flo decreases after the speed drop. The flo rate in Lane 3 decreased after the speed drop hile flo rates in Lanes 1 and 2 increased, indicating that hen Lane 3 transitions into queue discharge, Lanes 1 and 2 also transition to queue discharge. The transition to queue discharge can occur nearly instantaneously or over a longer period of time. Because of the variability and instability of flo during free-flo conditions and in the direct vicinity of the speed drop, use of these regions as maximum sustainable flo and capacity might not be practical. COMPARSON WTH OTHER BOTTLENECK STES Analysis of Flos at Validation Sites To verify the results of the analysis for the US-29 site, data from the 1-41and1-35/US-67 sites ere collected. ata ere collected for 4 days at the 1-41 site and 3 days at the 1-35/ US-67 site. Because one sample at the 1-41 site as taken during a rainy morning, only three samples ere used in the analysis. The peak 15-min flo rates for each of these sites are given in Table 3. 3: g LL. 35
36 TRANSPORTATON RESEARCH RECOR 1398 CJ 'C" :r: e Q.. (/) 9 8 7 6 5 4 3 6 Congestion ( +) 8 1 12 14 16 18 2 22 24 26 28 3 e ttl Q.. (/) 7o+ -------------- 5,_,_..,=o""'.:..::ns,..,tr.::;ea:::;m:.: --1 Congestion ( +) 4 --------------------------------4 CC 3+---.----.--...------------------------------ 6 8 1 12 14 16 18 2 22 24 26 28 3 11 1 e Q.. (/) 6 ------ ----o------;::::;-1' ------------------4 OaWSll:a.m.-----=----------------f Congestion ( +) 8 1 12 14 16 18 2 22 24 26 28 3 FLOW (VPH} FGURE 6 Speed-flo plots for US-29 study site, 5-min averages: top, Lane 1; middle, Lane 2; bottom, Lane 3. The peak 15-min flo rates sho some apparent differences beteen the to study sites as ell as the US-29 study site. A comparison of flo rates for individual lanes indicated that the median lane (Lane 1) is the highest flo lane for all the sites and is predominant for the 1-41 and the 1-35/US-67 sites. These to sites have very large peak 15-min flos of 2,496 and 2,492 vph. The loest flos tend to be in the outside lanes, hich are merge lanes for the US-29 and 1-41 sites but not for the 1-35/US-67 site. The average calculated flo rates for the 5 min before the speed drop, 5 min after the speed drop, and entire time during congested conditions are given in Table 4. Truck percentages averaged 1.7 percent at the 1-35/US-67 sites and 2.1 percent at the 1-41 site. The 1-41 site, hich is similar to the US-29 site, had much different flo rates and distribution of traffic across lanes. At the US-29 site the outside lane had the highest volume before the speed drop, indicating that it reached its maximum flo
13 12 11 NSE LANE <LANE 1> MLE LANE <LANE 2> 1- J: 9 J OUTSE LANE <LANE 3l... 8 J E Q.. (j) 7 J 6 51 4---; 3 j 2 1 - a 2 4 6 8 1 12 14 16 18 2 22 24 FLOW RATE <VPH> FGURE 7 Speed-flo relationship for US-29 study site. TABLE 3 Peak 15-min Flo Rates Across All Lanes Highay Observation Peak 15 Minute Flo Rate (vphpl) Lane 1 L,., T 'l T A. U.S. 29 Average 232 2244 2196-2253 1-41 1 2376 2212 1856-2148 2 2476 2196 1856-2176 3 2636 296 1688-214 Average 2496 2168 18-2155 1-35/US 67 1 248 218 2144 2124 2232 2 2588 232 2224 1856 2247 3 248 224 2324 2172 2286 Average 2492 2247 2231 251 2255 TABLE 4 Comparison of Flo Rates Before and After Speed rop at All Sites EJ 5 Minutes Before Speed rop 5 Minutes After Speed rop Entire Period After Speed rop (vphpl) (vphpl) (vphpl) Lane Lane Lane Lane Avg Lane Lane Lane Lane Avg Lane Lane Lane Lane Avg 1 2 3 4 2 3 4 1 2 3 4 U.S. 29 Average 2TT6 22 221-2()1)6 2266 235 1989-2()1)7 2246 2161 2()1) - 2166 Std ev 136 159 187-132 145 134 193-115 81 68 11-67 95% C 75 88 14-65 8 74 ltt - 64 45 38 56-37 1-41 Average 2463 2166 1856-2162 1839 179 1585-1738 1954 1864 1667-1828 Std ev 285 89 188-153 164 14 49-46 18 95 46-68 95% C 7()1) 222 466-381 4 34 121-113 448 236 115-169 1-35/US 67 Average 2679 2341 2134 1687 221 2542 2238 2231 1856 2217 2396 2154 2172 217 2185 Std ev 128 79 254 182 137 9 75 56 86 7 47 56 41 8 47 95% C 318 197 631 452 341 222 185 139 213 174 117 14 13 199 118
38 TRANSPORTATON RESEARCH RECOR 1398 before the other lanes and broke don first. After the speed drop the inside lane increased in flo hereas the outside lane decreased in volume. A much different situation occurred at the 1-41 site. All of the lanes decreased in flo after the speed drop. Further study found that the speed drop at the 1-41 site as caused by donstream congestion that resulted in the flo reduction. Because of a donstream slodon, the speed drop at the study site did not occur independently and the queue discharge flos ere controlled by the donstream bottleneck. Although the site as affected by donstream congestion, the inside lane had a very high flo before congestion set in. This shos that it is possible to reach much higher flo rates under free-flo conditions than measured at the US-29 site in the inside lanes. This agrees ith the hypothesis that the inside lane at the US-29 site prematurely dropped into queue discharge before reaching the maximum flo under free-flo conditions. The most interesting results came from the 1-35/US-67 site. At this site, traffic volumes ere substantially higher than those measured at the US-29 site. The average flos before the speed drop ere 2,679, 2,341, 2,134, and 1,687 vph for Lanes 1, 2, 3, and 4, respectively. After the speed drop the flo decreased in Lanes 1 and 2 and increased in Lanes 3 and 4, indicating that Lanes 1 and 2 broke don and transitioned Lanes 3 and 4 into queue discharge. This corresponds to the results and the model for US-29, hich determined that once a lane or lanes break don, the other lanes are also subsequently broken don. Figure 8 shos the comparison of the flo rates beteen the 1-35/US-67 site and the US-29 site for the 5 min before the speed drop and the total queue discharge period. The flo rates for the 1-35/US-67 site sho the same statistical trends as the US-29 data. The standard deviations in mean flos beteen days tended to be loer after the speed FREE-FLOW 268 QUEUE SCHARGE 234 2395 275 221 2135 2162155 217 29:: 215 1685 Lane 1 Lane 2 Lane 3 Lane 4 LANE NUMBER Lane 1 Lane 2 Lane 3 Lane 4 U.S. 29-1-35/US 67 FGURE 8 Flo-rate comparison beteen -35/US-67 and US-29 at Tidell. Lane 1 Lane2 Lane3 Lane4 LANE U.S.29 lmkhil H-35/US 67 FGURE 9 Average queue discharge flo rates.
1 -.: E Cl.. en (a) 9 8 - - -- - -- 7 ----- - - - --- q::j 5 --- - -- - ---- - - - - --- -- - --... t:ej"- - Ga - -o ---- - -- 4 - - - - ---- ------ - --- - - -- - --- -- - ------- -- - - - 3+---.----.--..---.---r---ir---...---.----.------,.---i 6 8 1 12 14 16 18 2 22 24 26 28 3 11----------------------------. -.: E Cl.. en 4 -- ------- ----- (bj Jn-+--------,------- 6 8 1 12 14 16 18 2 22 24 26 28 3 1,------- -.: E Cl.. en 9 8,-4----- 7"{\-4..------- 6,-f-------- 5 --------- - 4 3 6 8. 1 12 14 16 18 2 22 24 26 28 3 (cj 1 --- --- --- -- - --- :i: E Cl.. en (d) 9 8 -------- - ----- 7 6 ----- ------- 5 4 - - - --- 3+--.----.--------..---..-----.--- 6 8 1 12 14 16 18 2 22 24 26 28 3 FGURE 1 Speed-flo plots for 1-35/US-67 study site, 5-min averages; a, Lane 1; b, Lane 2; c, Lane 3; d, Lane 4.
4 drop than before the speed drop. This further supports the conclusion that flos in the free-flo regions contain a large amount of variability, hereas the flos during queue discharge are not nearly as variable. Even ith only three samples the confidence intervals for the average queue discharge flo rates are lo, ith 2,396 ± 117, 2,154 ± 14, 2,172 ± 13, and 2,17 ± 199 for Lanes 1, 2, 3, and 4, respectively. The average across all lanes as 2,185 ± 118 for the three samples. Although the individual lanes differ significantly, the average across all lanes during queue discharge as not statistically different than at US-29, hich had an average flo of 2,166 ± 37. Figure 9 shos the average queue discharge flo rates adjusted for heavy vehicles (assuming an E 1 of 2.) for the US-29 and -35/US-67 sites. The flos ranged from 2,2 to 2,4 pcphpl for individual lanes except in Lane 4 at the -35/US-67 site, hich as observed to not be a preferred lane and had loer flos. The averages for both sites across all lanes ere nearly identical, ith 2,23 and 2,22 pcphpl. For these reasons, queue discharge appears to be the most consistent flo for estimating the maximum sustainable flo of a facility. Although much higher flos are obviously possible, they typically do not occur across all lanes and ould be difficult to maintain. Even in queue discharge, Lane 1 at the -35/US-67 site continually had very high flos, averaging 2,396 vph. t is difficult to determine the reason for these very high queue discharge flo rates. One explanation for the high flo rates during queue discharge in Lane 1 could be the difference in the type of bottleneck at the -35/US-67 site. Speed-Flo Relationship at Validation Sites Because the -41 site as affected by donstream congestion, the relationship beteen speed and flo as not evaluated at this site. Figure 1 shos the speed-flo plots for each lane at the -35/US-67 site. Also shon in Figure 1 are TRANSPORTATON RESEARCH RECOR 1398 the projected speed-flo relationships suggested by the data. The scattered points in the center of the relationship (lo flo and higher speed) represent recovery to uncongested conditions. This phenomenon as not present in the US-29 data. Figure 11 shos the speed-flo model for the -35/US-67 site. As can be seen in this figure, Lanes 1 and 2 reach their peak flo rates during free-flo conditions hile Lanes 3 and 4 are prematurely transitioned into queue discharge. One interesting aspect of this site is the relation of Lanes 2 and 3. Both have approximately the same free-flo speeds and transition into nearly identical queue discharge flo rates, yet Lane 2 reaches it peak in free-flo conditions and Lane 3 peaks after the speed drop in queue discharge. These to lanes illustrate the effects of lane interaction. Because Lane 2 broke don, the turbulence transitioned Lane 3 into queue discharge before it reached its maximum flo rate. The extremely high flos in Lane 1 make it difficult to determine if the turbulence created by Lane 1 restricted Lane 2 from reaching its maximum free-flo flo rate. These results support the hypothesis that hen one or more lanes break don the other lanes are prematurely transitioned into queue discharge conditions. The data from the -35/US- 67 site sho that extremely high flos are possible in freeflo conditions and that relatively high flos also occur in queue discharge. Therefore, the measured flos are a function of the interactions beteen the lanes. Although lane interaction is not a ne idea, it seems to have been forgotten in many studies. One of the most significant problems is hen lanes are combined to form an overall speed-flo relationship. n free-flo conditions all lanes operate similarly, and all lanes can be combined to produce average flos. Once the facility begins to break don, average flo rates and speeds are significantly different from the individual lanes, and combining all lanes to produce a single relationship is misleading. The interaction beteen lanes may help explain the variety of results obtained in earlier studies. 13 12. 11 NSl_E LANE (LANE 1) 1 9... OUTSE LANE (LANE 4) E 8 7 6 a.. 5 Cl) 4 3 2 1 4 6 8 1 12 14 16 18 2 22 24 26 28 FLOW RATE CVPH> FGURE 11 Speed-flo relationship for 1-35/US-67 study site.
Ringert and Urbanik 41 RESULTS AN CONCLUSONS On the basis of the analysis of the US-29, -41, and -35/ US-67 study sites the folloing conclusions can be made: Variance in flo rate decreases after the speed drop to queue discharge. Peak flos for individual lanes occur in free-flo conditions before breakdon. Peak flos during free-flo conditions do not generally occur in all lanes on a facility because of an imbalance of flo rates beteen individual lanes. This prematurely transitions the flo from free-flo into queue discharge conditions. Bottleneck configuration may influence the maximum possible flo obtainable during free-flo and possibly queue discharge conditions. Queue discharge appears to be the best estimate for maximum sustainable flo and capacity. ACKNOWLEGMENTS This study as sponsored by the Texas epartment of Transportation and FHWA, U.S. epartment of Transportation. REFERENCES 1. Special Report 29: Highay Capacity Manual. TRB, National Research Council, Washington,.C., 1985. 2. V. F. Hurdle and P. K. atta. Speeds and Flos on an Urban Freeay: Some Measurements and a Hypothesis. n Transportation Research Record 95, TRB, National Research Council, Washington,.C., 1983, pp. 127-137. 3. K. Agyemang-uah and F. L. Hall. Some ssues Regarding the Numerical Value of Freeay Capacity. Proc., nternational Symposium on Highay Capacity, July 1991, pp. 1-15. 4. T. Urbanik, W. Hinsha, and K. Barnes. Evaluation of High Volume Urban Texas Freeays. n Transportation Research Record 132, TRB, National Research Council, Washington,.C., 1991, pp. 11-118. 5. F. L. Hall and K. Agyemang-uah. Freeay Capacity rop and the efinition of Capacity. n Transportation Research Record 132, TRB, National Research Council, Washington,.C., 1991, pp. 91-98. The contents of this paper reflect the vies of the authors, ho are responsible for the opinions, findings, and conclusions presented herein. The contents do not necessarily reflect the official vies of FHW A or the Texas epartment of Transportation. This paper has not been revieed by the sponsoring agencies. Publication of this paper sponsored by Committee on Highay Capacity and Quality of Service.