2. Traffic Delay Comparisons
The Study Team evaluated PM Peak traffic volume and speed data at selected locations along I-95, I-270 and I-495, derived from operational data from the CHART advanced transportation management system of the Maryland DOT, and archived by the University of Maryland. The first two of these corridors are radial with respect to regional travel, while the third one is part of the main circumferential route within the region. A generally similar analysis approach was used for each corridor. First, graphical summaries of the variation of average speed by time-of-day at each detector, and cumulative delay along the analysis segment were developed to identify the best potential days for analysis: (a) "August lite" volumes and delay, (b) September and October typical volumes and delay, and (c) special days with unusual traffic demand and/or use patterns. Second, the consistency in average speed and cumulative volumes at sequential detectors were studied to identify peaking patterns along the corridor as they relate to roadway features, such as bottlenecks or lane drops. Third, the analysis examined the paired-relationship between concurrent observations of volume and average speed to better understand how the relationship varies under various circumstances. The analysis demonstrates the following:
The summary analysis for each of the three corridors is presented separately below. Additional detail on I-270 is provided in Appendix A.
A separate analysis of AM Peak traffic volumes and speeds was also performed. AM Peak volumes are typically lower than PM Peak volumes, and the locations of the traffic detectors are not as conducive to a delay analysis. This analysis is located subsequent to the I-495 analysis.
I-95 Northbound for PM Peak Period
I-95 Corridor: General Locations of Detectors
The CHART system of the Maryland DOT operates a set of traffic flow detectors along the I-95 Corridor between the Baltimore and Washington, D.C. beltways. The five-minute summaries of the volume and average speed are archived by the Center for Advanced Transportation Technology of the University of Maryland (UMD-CATT). This analysis focuses on data from the six detectors of Brooklyn Bridge Road to Montgomery Road on I-95. Two other working detectors that were not included in the detailed analysis are located on I-95 near I-495, the Capital Beltway. The length of the I-95 analysis section is approximately 10.5 miles with a consistent directional cross-section of four lanes in each direction. The evaluation only included the six sites identified on the display below and does not include the two sites shown on MD 32.
As part of our analysis, the Study Team examined various traffic characteristics obtained from measured variations in five-minute volumes and five-minute average speeds at each of the six detectors of the analysis section. The data is nominally collected "24 x 7 x 365", or twenty-four hours a day, 7 days per week, for 365 days per year. The operational requirements of CHART, the program for which the data is collected does not need to have data from each and every detector for each and every time increment. Some detectors and/or parts of the communication system are more reliable in the collection of the traffic flow data than others. That typically results in small gaps of one or a few observation time periods and sometimes for hours and even days on end in the archived data. Nevertheless, generally speaking there is an adequate amount of reliable data to identify various traffic characteristics including: (a) variation in volume by time of day that is used to define the "peak traffic volume periods", (b) comparative speed range that can be used to define different degrees of congestion, such as free-flow/uncongested, slowing, slow, or jammed or stop-and-go conditions, and (c) section delay, the difference between the time it would take to travel the section at the observation time compared to the time it would take to travel that same distance at the speed-limit (Speed-Limit-Travel-Time, SLTT) or when traveling at free-flow. This characteristic can have its own peaking patterns that may be similar to or differ from volume-peaks.
The traffic characteristic of cumulative "section delay" for a typical Thursday and Friday in October are plotted and are shown in Figure 2-1 below. At the speed limit of 65 mph in this section it takes about 9 minutes and 30 seconds (570 seconds) to travel the approximate 10.5 miles. However, for a traveler beginning this section about 10 minutes before 5 PM on Friday 10-12-07, the amount of calculated section delay was about 540 seconds, or about 9 minutes more than the expected SLTT of 9 minutes, 30 seconds.
A good understanding of the patterns of variation in delay may assist in identifying and analyzing the effectiveness of strategies aimed at managing the delay. For example, as shown in Figure 2-1, congestion on Thursday, 10-11-07 in this section was more prolonged and occurred later in the afternoon relative to congestion shown for Friday 10-12-07, that was more peaked and started earlier in the afternoon. Examination of archived data for other pairs of Thursdays and Fridays would show a similar set of relative peaking patterns. Thus, strategies that address differences between days of the week will help address the effectiveness of strategies trying to restore freer flowing traffic.
The second characteristic considered was the length or duration of the peak period. During the light traffic days of early August, the duration of the peak within the peak period is about 90 minutes (about 4:30 to 6:00 PM) as shown in Figure 2-2 below. For Wednesday 8-1-07, even though section delay was very light, for most of the 5 to 6 PM period, it was somewhat slower than the other times of day. However, for a typical October day of 10-11-07, the traffic data shows that the section delay was much slower and longer in duration extending to about 7:00 PM. During the summer it seems that relatively more people leave work earlier and do not stay as late as they do in October. Thus, strategies that account for seasonal as well as time-of-day patterns of travel behavior will be more effective than those that do not do so.
A third traffic characteristic that affects traffic volume, speed, and delay is that of trip purpose and the effect trip purpose has on the temporal and spatial distribution of travel demand. An interesting illustration of this can be seen by examining the section delay characteristics for another day that same week in October, that of the Columbus Day Holiday of Monday 10-8-07. Even though Federal offices and some schools in the region and vicinity of this section were closed, section delay is seen in Figure 2-3 in the 3:30 to 5 PM time period. It is inferred that enough people did not have their usual trip purpose of a work trip heading home that afternoon. Instead, many people attended mostly to personal business and shopping, that is typical for that holiday. However, during the usual commute peak time of 5 to 6 PM, the section delays were as light as they are during the light demand days of August. Strategies for restoring freer flowing traffic need to address trip purpose too.
The overall data set for this section of I-95 was also analyzed in more depth to address the first research question of, how much traffic needs to be taken off freeways operating at various levels of congestion to restore free flow during rush hours?
To address that question the Study Team has taken the approach of first needing to "boil-down" the time-of-day variation in section delay to one indicator of delay. The measure of average delay per 5-minute increment over the 4-hour peak period was selected as the representative indicator of delay. Thus, the patterns of section delay from the prior three figures were reduced to one four-hour average. For example, the day with the most peak section delay of those shown in the prior figures was that of Thursday,10-11-07, and averaging over the entire four hours results in a value of about 156 seconds of average delay over the four-hour peak period in excess of the SLTT. That is an indicator measure for the entire analysis section of 10.5 miles. In addition, we have a measure of the variation in demand along the section for that four-hour period by adding up the cumulative volumes observed at each detector. For example, nearly 30,000 vehicles passed the detector at Brooklyn Bridge Road as the traffic the entered the section from the south heading northbound in the four-hour peak period on Thursday 10-11-07, while about 24,000 vehicles left the section at the north end at Montgomery Road during the four-hour peak period.
In Figure 2-4a the Study Team graphed the average section delay versus the total four-hour volume for each detector for the several example days being considered. The analysis results shown here have the three "August-lite" samples with less peak-period volume and less peak-period section delay than the two selected "Mid-October" sample days. Figure 2-4a also shows that there is a generally consistent relationship among the different detectors across the sample days—generally speaking as the volumes decline or increase there appears to be a linear relationship to the values of average delay.
While Figure 2-4a tends to show the "traditional relationship" of volume versus delay – more volume – more delay, however, examination of other examples of "special days" tends to indicate that the nature of the relationship between volume and average delay is more complex.
Figure 2-4a, for 3 of the 5 detectors for the two typical October days the volumes and average delay show an "inverse relationship"—when section delay increases, the total volume throughput declines somewhat. On examining the similar results for two special days—that of incident conditions on Thursday 8-9-07 and of Columbus Day on Monday 10-8-07, in Figure 2-4b we see that for the former, significantly different patterns of regional demand can result in volumes and section delay that are different than usual, while for the latter, the section delays increase significantly with little change in volume—in fact the volumes appear to be somewhat lower relative to the typical days of October, but significantly greater than the August-lite days. While the "connecting-of-the-dots" as illustrated in Figure 2-4b does nominally provide for what seems to be a larger continuum of conditions, it is not clear whether a more extensive analysis of many more samples would result in the relationship that is depicted.
Returning to comparing the October to August samples, a reduction of about 10% of the volume (about 3,000 vehicles in the four hours) at the Brooklyn Bridge Rd detector may be sufficient to return traffic there to freer flowing conditions, while at the next detector of the Welcome Center, a reduction of traffic of about 4%, or about 1,150 vehicles may be sufficient. At the northern-most detector at Montgomery Road about an 8% volume reduction or about 2,000 vehicles appears to be needed. Being able to achieve significant reductions in section delay based upon volume reductions at MD 32 may require a reduction of about 20% or 4,500 vehicles. Perhaps the significant amounts of ramp-to-ramp turning movements and weaving of traffic are creating more conflicting movements at the MD 32 interchange area.
Comparison of Paired-Values:The next part of the analysis examined the paired-relationships between concurrent observations of volume and average speed. It has been found that the data analyzed both for the system of detectors along the analysis section as well as for each detector generally replicate the types of patterns associated with the volume-speed-density curves of the Highway Capacity and Quality of Service Manual (HCQSM). However, the Study Team thinks that this analysis has a few aspects that distinguish it from the work that underlies the HCQSM. In particular, unique aspects of this analysis have included: (a) tracing this phenomenon along a series of detectors, along this 10.5 mile section, as well as (b) tying the results back to specific dates and times of generally known variations in travel demand.
In Figure 2-5 and Figure 2-6 for each detector the paired-values of volumes (x-axis) and average speed (y-axis) per each 5-minute summary interval are plotted, respectively first for a typical day in August (8-1-07) and then for a typically congested day at another time of the year, shown here as October (10-11-07). Four hours of such data for each day are shown in this case for each of the six detectors. The following observations are noted based upon Figure 2-5 and Figure 2-6:
In Figures 2-5 and 2-6 there are 576 independent sets of paired-values of 5-minute volumes and average speed (4 hours, times 12 observations per hour, times 6 detectors, times 2 days). If all of those points were put on one graph and the different marker shapes and color-codes for the detectors were not shown, then the reader would find it just about impossible to discern the relationships among the volumes and average speed that are there. In the remainder of the analysis we do the opposite of that and highlight similarities and differences for paired-graphs for the pair-values of volume and average speed for 3 of the 6 detectors. Those paired-graphs demonstrate a general consistency with HCQSM findings and they are used to estimate the "tip capacity" that breaks down the flow of traffic and creates significant delays.
Paired Values of Volume - Average Speed by Sample Day
The questions addressed in the remaining three sets of paired-graphs are: (1) does temporally increasing travel demands, reflected by marginally higher and higher 5-minute volumes, result in traffic congestion?, and (2) if it does, what are the specific values of demand (high volumes) that we should manage to avoid through application of various strategies, so as to maintain the freer-flowing traffic conditions?
On the following page, just the 5-minute paired-values of volume and average speed of I-95 at the Welcome Center are plotted paying attention to the temporal sequence of the observations, which is graphically illustrated by the line that is "connecting-the-dots". In Figures 2-7 and 2-8 it is demonstrated graphically that a small degree of marginally increasing demand (volume) is a main ingredient that results in the precipitation of congested and slow average speeds for one or more of the subsequent 5-minute time intervals. A key conceptual consideration is to examine the relationship among the different paired-values of the quantity of the traffic flow (volume) relative to the quality of the traffic flow (average speed) sequentially through time.
It has long been observed and reported in the HCQSM and elsewhere that when the quantity of flow is very high approaching the capacity of the roadway that further small increases in the quantity of flow can result in a quick and sustained reduction in the quality of the traffic flow, as measured by the average speed. That in turn has a negative feedback on subsequent time intervals when the same quantity of flow can no longer be served and queues form back upstream from the point of constriction. A series of relatively closely spaced detectors summarizing flow characteristics at relatively short time intervals can generally discern such traffic flow behavior, as illustrated in Figure 2-7 and 2-8 although the fixed location of the detector may not be exactly where the traffic flow initially broke down.
Another factor in precipitating a sudden decrease in traffic speed is usually some random event such as very short-term driver inattentiveness. Drivers taking their eyes off the road ahead for a very small interval may result in drivers braking extra hard when their eyes are back on the road and perceive that the vehicles ahead are now too close. That hard braking can cause the driver behind to react in a similar fashion, and the driver behind that one to do the same, and so on. A "shock-wave" of brake lights ahead can quickly propagate upstream at speeds that approach that of the forward movement of the traffic flow. This phenomenon of a "reverse compression shock wave" moving back upstream in the flow, at speeds of 100 mph or more relative to the forward speed of the vehicles can result in minor and even severe crashes taking place.
In Figure 2-7 for 8-1-07 it shows that at the "Welcome Center" detector on I-95, the maximum 5 minute volume approaches but does not exceed about 650 vehicles per 5-minute interval. However, in Figure 2-8 the 5-minute volumes initially start to exceed the capacity, which is somewhat elastic. However, for a short period of time the flow is not a sustained enough flow to cause an immediate break down. Shortly afterward the volume increases again and there is enough sustained volume and within a twenty minute period to cause the average speed decreases to about 25 mph and then 10 minutes later it decreases again to about 15 mph. It then took over an hour and a half more before the average speeds returned to free flow levels at about 7 PM. The maximum observed flow at this location was about 690 vehicles, enabling the maximum flow per lane to be calculated. In the I-95 section there are 4 through lanes. Expanding the 5-minute volume of 690 vehicles by 12 (12 – 5 minute time periods per hour), the hourly maximum volume is about 8,300 vehicles per hour. Dividing by the 4 lanes gives a maximum volume of about 2,075 vehicles per lane per hour. This corresponds well with the HCQSM's maximum volume of 2,400 passenger cars per lane under ideal condition. If the I-95 data is adjusted for trucks and the peak hour factor, the recorded maximum traffic volume is close to the theoretical HCQSM maximum flow.
In Figure 2-8 it should also be noted that during congested flow the traffic flow is much lower at the slower speeds. As traffic becomes congested there is a drop in the maximum flow rate of 10 to 20% versus the flow rate just prior to a traffic breakdown. By avoiding traffic breakdown not only does the traffic travel at higher speeds, but under the non-traffic breakdown conditions a higher throughput volume may be maintained, serving more travelers in less time.
Marginally Higher Volumes Cause Congestion?
Each Freeway Segment has its Own Capacity
In Figures 2-11 and 2-12 we again are able to show and estimate the maximum flow for the roadway section just prior to congestion breakdown. Figure 2-12 also demonstrates the 10 to 20% reduction in traffic flow due to being in a congested flow versus a non-congested flow. The relationships shown in this series of figures are important to understanding and estimating the reductions in delay with very modest reductions in traffic volumes.
Next Freeway Segment also has its Own Capacity
I-270 Corridor: General Locations of Detectors
Parts of the I-270 Corridor were also analyzed to evaluate and confirm many of the traffic characteristics and speed-volume relationships discussed in the previous section for parts of I-95. In a similar fashion this analysis relied on the CHART system of the Maryland DOT that operates a set of traffic flow detectors along the I-270 Corridor in Frederick and Montgomery Counties. The five-minute summaries of the volume and average speed are also archived by the Center for Advanced Transportation Technology of the University of Maryland (UMD-CATT). This analysis focuses on data from the six detectors on I-270 from I-370 north to MD 109, in Montgomery County, a distance of about 12.9 miles. The general location and names for those detectors are shown in Figure 2-13 below. Seven of the I-270 detectors located in Frederick County are shown here for informational purposes. It should be noted that the number and use of lanes for this section of I-270 changes a few times and differs for southbound and northbound travel. One of those lane changes highlighted in the analysis is the lane-drop from 3 to 2 lanes in the Northbound direction between MD 121 and Comus Road, where the concurrent-flow High Occupancy Vehicle lane also ends, which operates weekdays between 3:30 and 6:30 PM. Congested conditions are often found at and near this lane drop.
As part of our analysis, the Study Team examined various traffic characteristics obtained from measured variations in five-minute volumes and five-minute average speeds at each of the six detectors of the analysis section of I-270. Similar to I-95 discussed previously, this operational data is nominally collected "24 x 7 x 365" and the other features previously discussed also apply. Generally speaking there is an adequate amount of reliable data to identify various traffic characteristics including: (a) variation in volume by time of day used to define the "peak traffic volume periods", (b) comparative speed range that can be used to define different degrees of congestion, such as free-flow/uncongested, slowing, slow, or jammed or stop-and-go conditions, and (c) section delay, the difference between the time it would take to travel the section at the observation time compared to the time it would take to travel that same distance at the speed-limit (Speed-Limit-Travel-Time, SLTT) or when traveling at free-flow. This last characteristic can have its own peaking patterns that may be similar to or differ from volume-peaks. This latter traffic characteristic of cumulative "section delay" for a typical Thursday and Friday in October, as well as a Tuesday in April are plotted and are shown in Figure 2-14 below. For this corridor the analysis uses a 5-hour peak period of 2 to 7 PM as congested section delays seems to be more prolonged in this corridor than in the I-95 Corridor. As noted above, this analysis section of I-270 is about 12.9 miles long. The speed limit there is mostly 65 mph and the travel time to travel at that speed is just under 12 minutes (11 minutes, 56 seconds, or 716 seconds). For a traveler beginning this section about before 4 PM on Friday 10-12-07, the amount of calculated section delay was about 720 seconds, or about 12 minutes more than the expected SLTT of coincidently also about 12 minutes.
A good understanding of the patterns of variation in delay can better inform strategies aimed at managing that delay. For example, as shown in Figure 2-14 above, congestion on Thursdays is usually more prolonged and later in the afternoon relative to congestion on Fridays, which is usually more peaked and starts earlier. This pattern was also found in the I-95 Corridor analysis. For the Tuesday example of 4-29-08, the peak is even later in the PM. Strategies that address differences between days of the week will help the effectiveness of other strategies trying to restore freer flowing traffic.
Another characteristic considered was the length or duration of the peak period. Even during the lighter traffic days of August, the duration of the peak within the peak period is about 3 hours shown in Figure 2-15 below. Even one of the least congested summer days examined of 7-30-07 shows a peak of about 30 minutes duration at about 5:30 PM. For the typical October day of 10-11-07 Figure 2-15 shows that the duration of the peak section delay is even longer and starting before 3 PM and extending to nearly 7 PM. Thus strategies that account for seasonal as well as time-of-day patterns of travel behavior will be more effective than those that do not do so.
A third traffic characteristic that affects traffic volume, speed, and delay is that of trip purpose and the effects that systemic changes in trip purpose have on the temporal and spatial distribution of travel demand. An interesting illustration of this can be seen by examining the I-270 section delay characteristics for two "special days". The first of these special days is the Columbus Day Holiday of Monday 10-8-07, which was also examined in the I-95 analysis. For the I-270 analysis even though Federal offices are closed as are some schools in the region, the schools in Montgomery County remain open for parent visitations. The section delay for Columbus Day is seen in Figure 2-16 shows that I-270 was still congested for about 2 hours between about 4:30 and about 6:30 PM. This suggests that many people still traveled based on their usual trip purpose of a work trip heading home that afternoon, even though a significant proportion of workers (Federal employees) had the day off. Perhaps that is why during the first half of the PM peak period from about 2 to 4:30 PM the section delay was much lighter than usual for October and even lighter than an August-lite day of Monday 8-6-07. In addition, many people who were off work probably attended to personal business and shopping, that is typical for that holiday. Strategies for restoring freer flowing traffic need to anticipate and address trip purpose as well.
The second of these "special days" systemically affecting travel demand is somewhat unique to Montgomery County and the analysis for this section of I-270 appears to be sensitive to those systemic changes. A relatively significant portion of residents (perhaps 10 to 15%) observe the religious holiday of Rosh Hashanah and do not go to work on that day, which was Thursday, 9-13-07 in 2007. In addition, for the past few decades if that Holiday falls on a weekday then the schools are closed county-wide. About 25 % of the households in the County have school age children and a significant proportion of the adults from those households chose not to work and remain home with their children. In addition, for many of the observers of the religious Holiday, there is a custom of visiting relatives and friends throughout the afternoon after the religious services are over resulting in higher than normal social trips and travel than normally occurs on a weekday. Figure 2-17 below shows the effects of the combination of these relatively large and systemic changes in travel demand on the section delay along I-270. The peak delay occurred between 2:30 and 4:30 PM, which probably corresponded to the extra amount of social travel and trip making. On the other hand, during the usual commuter peak-times, there in essence was no section delay and the traffic conditions were more like those associated with weekend travel.
Because of the relatively uniformly light traffic congestion along I-270 on that religious Holiday, and the temporal and spatial changes in travel due to the shifts in trip purpose and destination, the throughput of traffic was high, which will be reviewed later in the analysis summary. Due to those high volumes of throughput and the low congestion, and because data was available at all detector locations, September 13, 2007 was selected as a "benchmark" for non-congested time periods for I-270 instead of one of the August-lite days previously examined. Appendix A in this document provides the detailed speed/volume plots for each detector site for the "August-lite" days that were considered plus discussion as to the relative merits of each day and the rationale for the September 13 selection.
The overall data set for this section of I-270 was also analyzed in more depth to address the first research question of, how much traffic needs to be taken off freeways operating at various levels of congestion to restore free flow during rush hours?
To address that question the Study Team has taken the approach of first needing to "boil-down" the time-of-day variation in section delay to one indicator of delay. The measure of average delay per 5-minute increment over the 5-hour peak period was selected as the representative indicator of delay. Thus, the patterns of section delay from the prior four figures were reduced to one five-hour average value. For example, the day with the most peak section delay of those shown in the prior figures was that of Friday,10-12-07, and averaging over the entire five hours results in a value of about 280 seconds of average delay over the five-hour peak period in excess of the SLTT. That is an indicator measure for the entire analysis section of 12.9 miles.
In addition, we can measure the variation in demand along this analysis section of I-270 for that five-hour period by adding up the cumulative volumes observed at each of the six detectors. For example, about 27,500 vehicles passed the detector at MD 118 (Germantown Road) and that appears to be the location of the maximum quantity of flow along this analysis section of I-270 based upon the data from the detectors. However, since there is a long spatial gap of about 5.25 miles between the prior detector at the Express Lanes at I-370 and the one at MD 118, there may actually be a higher volume location based upon other data sources.
In Figure 2-18a the Study Team graphed the average section delay versus the total five-hour volume for each detector for the several example days being considered. The analysis results shown there have the three "August-lite" samples (7-30-07, 8-6-07, and 8-13-07) with less peak-period volume and less peak-period section delay than the three selected sample days for other times of the year (10-11-07, 10-12-07, and 4-29-08). Figure 2-18a also shows that there is a generally consistent relationship among the different detectors across the sample days—generally speaking as the volumes decline or increase there appears to be a linear relationship to the values of average delay. However, for two of the detector locations the relationship seems less strong, as discussed more next.
In Figure 2-18a above, for the detectors at "I-370 – Express Lanes Only", and at the "Truck Weigh Station" there appears to be a less firm relationship between total volume average section delay in that the apparent linearity for the other four seems more independent at these two locations. For the detector at I-370, the archived data from UMD-CATT also had a set of values of 5-minute volumes and average speed for the "Local Lanes" at I-370. The study team chose to ignore that data set as we reasoned that we were interested in the variation in "mainline" volumes, speed, and delay and thus traffic flow of the local lanes represented a different set of traffic flows for our study purposes. At the truck weigh station the detector is positioned after the off ramp and before the on ramp from the weigh station. Thus traffic that turns into that area does not get counted in either the total volumes or the average speed.
For purposes of this part of the analysis, we feel comfortable with excluding the information for those two detectors in a revised version of the graph, which is termed Figure 18b given below. In addition, while Figure 2-18a tends to show the "traditional relationship" of volume versus delay – more volume – more delay, however, examination of other examples of three distinct "special days" in Figure 18b tends to indicate that when we "connect-the-dots" that the nature of the relationship between volume and average delay is perhaps richer in information and more complex.
Before returning to comparing the October to August samples, we need to put the three "special days" in context. We have already identified and partially discussed two of them: Monday 10-08-07, Columbus Day, and Thursday, 9-13-07, which was a religious holiday. Both of those days had systemic changes in significant components of the overall regional travel demand so it is not surprising that the values for total volume and average section delay do not "fall in line" with the more normal traffic days with consistent demand patterns. The third "special day" that is shown in Figure 2-18b is the day after the last one mentioned—that of Friday 9-14-07. For that day, this analysis section of I-270 experienced the most congested conditions of the samples examined for this analysis—and as shown in the Figure it had an average section delay of about 480 seconds (about 8 minutes) for over the whole 5-hour time period studied. That day is the one most to the right in Figure 2-18b. That sample tends to illustrate that more congestion can have a negative feedback and result in less throughput within a fixed time period than other less congested sample days. That pattern appears to be the case for each of the four detector locations that are shown in Figure 2-18b.
Yet, for the left side of that Figure, the relationships seen indicate that by better managing and reducing demand, there can be significant reductions in the amount of congestion and delay associated with that corridor. The particular amounts of demand reductions needed for the four locations associated with these detectors is the following going northbound in the flow direction in the PM:
Perhaps the larger and more difficult reduction would be needed in the vicinity of the MD 121 detector because: (a) that is where the number of lanes drops from 3 to 2 which results in more traffic friction, and (b) during the 3:30 to 6:30 PM period the HOV lane ends there as well. The users of the HOV lane, who are in the left lane, do not need to merge to the right. Rather the right most traffic lane is the one that is dropped and the users of that lane need to merge with the traffic in the center lane.
Comparison of Paired-Values: The next part of the analysis examined the relationships between concurrent observations of paired-values of five-minute volume and average speed. It has been found that the data analyzed both for the system of detectors along this analysis section of I-270, as well as for each detector, generally replicate the types of patterns associated with the volume-speed-density curves of the HCQSM. However, the Study Team thinks that this analysis has a few aspects that distinguish it from the work that underlies the HCQSM. In particular, unique aspects of this analysis have included: (a) tracing this phenomenon along a series of detectors, along this 12.9 mile section, as well as (b) tying the results back to specific dates and times of generally known variations in travel demand.
In Figure 2-19 and Figure 2-20 for each detector the paired-values of volumes (x-axis) and average speed (y-axis) per each five-minute summary interval are plotted, respectively first for an identified day of little congestion of Thursday 9-13-07, given in Figure 2-19. As noted and discussed above, that day was a religious Holiday for many residents in the corridor, and many others had changed travel patterns due to the schools being closed for the day. Figure 2-20 shows the similar paired-values for a typical congested day of Thursday 10-11-07. Five hours of such data for each day are shown in this case for each of the six detectors. The following observations are noted based upon Figure 2-19 and Figure 2-20:
In Figures 2-19 and 2-20 taken together there are 720 independent sets of paired-values of 5-minute volumes and average speed (5 hours, times 12 observations per hour, times 6 detectors, times 2 days). If all of those points were put on one graph and the different marker shapes and color-codes for the detectors were not shown, then the reader would find it very difficult to discern the relationships among the volumes and average speed that are there when such distinctions are accounted for. In the remainder of the analysis we do the opposite of that and highlight the similarities and differences for "paired-graphs" for the pair-values of volume and average speed for each of the six of the detectors taken one at a time and in sequence. For those paired-graphs we also "connect-the-points" sequentially in time so as to focus on the temporal as well as the spatial variations in how the paired-values of volume and average speed change over the five-hour PM peak period. Those paired-graphs demonstrate a general consistency with HCQSM findings and they are used to estimate the "tip capacity" that breaks down the flow of traffic and creates significant delays.
The following series of figures (Figures 21a and b through Figures 26a and b) compare the paired-values for each detector in sequence heading northbound from I-370. Text annotations are added to the figures in order to have them appear as large graphically as feasible for clarity of viewing their content. However, the pairs are kept on the same page to better enable the readers to more easily do their own comparisons between the paired-graphs and among the different paired sets, which may also help readers who may have difficulty seeing different colors. After this series of figures, Figure 27 shows six similar graphs of paired-values of five-minute volumes and average speed for Friday, 9-14-07. On that day in the afternoon and evening there was intensive "get-away congestion" along most of I-270 for many hours and each of the 6 detectors in this analysis section of I-270 operated at or near the capacity of the roadway at those locations.
We conclude this analysis of this section of the I-270 Corridor with a check of some of the derived capacities shown in Figure 27 (a thru f) compared to information from the HCQSM. The maximum observed flow at the MD 118 detector was about 580 vehicles, enabling the maximum flow per lane to be calculated. In that part of I-270 section there are 3 through lanes. Expanding the 5-minute volume of 580 vehicles by 12 (12 – 5 minute time periods per hour), the hourly maximum volume is about 6,960 vehicles per hour. Dividing by the 3 lanes gives a maximum volume of about 2,320 vehicles per lane per hour. This corresponds well with the HCQSM's maximum volume of 2,400 passenger cars per lane under ideal condition. Compared to the I-95 Corridor analyzed above, the part of the I-270 data has many fewer trucks particularly during the peak period. Thus, it would be expected that the recorded maximum traffic volume is close to the theoretical HCQSM maximum flow.
The maximum observed flow at the three more northern detectors from MD 121 to MD 109 had derived capacities of about 420 to 430 vehicles per 5-minutes, enabling the maximum flow per lane to be calculated. In that part of I-270 section there are 2 through lanes. Expanding the 5-minute volume of 420 vehicles by 12 (12 – 5 minute time periods per hour), the hourly maximum volume is about 5,040 vehicles per hour. Dividing by the 2 lanes gives a maximum volume of about 2,520 vehicles per lane per hour. This is somewhat higher in value that the HCQSM's maximum volume of 2,400 passenger cars per lane under ideal condition. However, given the relatively low percent of trucks compared to the I-95 Corridor analyzed above, particularly during the peak period and the experienced set of commuters using the roadway it may be possible that a capacity value that is about 5% higher than the theoretical HCQSM maximum flow could be found. In fact the HCQSM has reported capacities exceeding 2700 passenger cars per hour for individual lanes for short periods of time.
I-495 Corridor: General Locations of Detectors
In addition to the I-95 and I-270 Corridors, a small part of I-495, the Capital Beltway, was also evaluated to confirm many of the traffic characteristics and speed-volume relationships discussed in the previous sections. The two previous corridors are "radials" with respect to their traffic orientation and function within the region, while I-495 serves "circumferential" traffic patterns. One question for this research is whether the radial versus circumferential orientation would result in findings and conclusion that differ from those found for the two radial corridors.
Our analysis shows that the same basic results have been found. However, the traffic demand patterns around the Beltway frequently are less continuous than those of the radial corridors, with "crisscross-flows" quickly adding or subtracting volumes. Sequential continuity of flows between adjacent detectors is not always found and as a result the detectors may not be measuring the same traffic demand composition. We think that was the case with the three detectors used here. Further, data quality issues with the volume summation for the detector at Persimmon Tree Road, resulted in it not being used as part of the analysis. Thus the analysis summarized here does differ somewhat in form and content from that presented for the prior two corridors.
The general location of the detectors used in the analysis is shown in Figure 2-28. CHART operates a limited set of traffic flow detectors along I-495 in Montgomery and Prince George's Counties. Traffic flow detectors along the Virginia parts of the Capital Beltway have been out of service for a while and data was not available. The analysis used 5-minute summaries for total volume and average speed archived by UMD-CATT as derived from the CHART system, but only data for the flows on the "Outerloop" of I-495 are available. The southbound HOV lane of I-270 ends just prior to the merge of the I-270 Spur with the I495 Outerloop.
As part of the analysis, the Study Team looked at potential differences in the peaking of volume and average speed traffic flow characteristics of two detectors on the Outerloop near Greentree Rd and MD 190, River Road. Estimates of delay in the vicinity of those detectors were plotted and are shown in Figure 2-29 and 2-30, at Greentree Rd and MD 190, respectively. It should be noted that these delays were estimated in a different but similar manner to those given for the two prior corridors. Rather than estimating and using the cumulative delay for the section of the corridor as was done for the other two cases, delay for each of the two detectors was separately estimated relative to the observed free-flow average speed at each detector and assumed that speed was representative of the speed about one-half mile on either side of the detector. Thus the inverse of the average speed applied over one mile yielded an estimate of the travel time per 5-minute interval, and subtracting the free-flow travel time resulted in the delay for each 5-minute period.
One notable traffic characteristic that is different at the I-270 and I-95 locations is that of the length of the peak period. During the heavier and more normal traffic days of September and October, the duration of the peak within the peak period is about 90 minutes between about 5:30 and 7 PM, as shown in Figure 2-29. The duration of the peak within the peak period shown in Figure 2-30 is substantially longer starting about 3:30 PM and continuing until 6:30 to 7 PM. Thus strategies that account for time-of-day patterns of travel behavior will need to be prepared to function over prolonged time periods and not just be peak-hour oriented.
An important aspect of these two Figures is to show that delay during the lighter traffic volume days of summer can result in the near absence of congestion and delay. The example the delay shown for Wednesday 8-8-07 given in Figure 2-30 shows that to be the case. However, for the Greentree Road detector shown in Figure 2-29, the traffic volume data at that location for the week of August 6, 2007 were questionably low and were not used. Instead one of the traffic days of mid-September, that of Thursday 9-13-07 was used as that was a day that the entire Montgomery County school system was closed due to a religious holiday resulting in a significant and systemic shift from normal traffic demands.
Indeed the two other days shown in Figure 2.29 that straddle that low-demand day, those of Tuesday 9-11-07 and Friday 9-14-07 had more normal traffic characteristics typical for that time of year. Of the days sampled for this detector in the analysis, those two other days are the ones that had the most delay at this detector. Returning to Figure 2-30, it can be seen that the delay by time-of-day for Thursday 9-13-07 was very light, being very similar to that found for Wednesday 8-8-07. Another distinguishing traffic characteristic shown in Figure 2-30 for the detector at MD 190 is that the high traffic delays for Friday 9-14-07 were not only long in duration, they were also very erratic and volatile within this time period, probably indicating that surges of different traffic flow demands were passing this location. The shape of the highest delay patterns shown in Figure 2-29 are more similar to that usually associated with a "normal distribution", indicating that more consistent travel demands were being observed.
The data set was also analyzed in detail to address the research question of:
To address that question in this case the particular and similar data for other days shown in the prior two figures needed first to be "boiled-down" to one indicator of delay. The measure of average delay per 5-minute increment over the 4-hour peak period was selected as the representative indicator of delay. Thus all of the erratic and volatile ups and downs of delay for Friday 9-14-07 from Figure 2-30 above were reduced to a value of about 35 seconds of average delay per 5 minutes over the 4-hour period, as shown below in Figure 2-31. Similarly, the more normal variation of delay shown in Figure 2-29 for Friday 9-14-07 was reduced to a value of about 13 seconds of average delay over the four-hour peak period, as shown below in Figure 2-31. The other dimension of Figure 2-31 is the cumulative volume for the corresponding 4-hour peak period at each detector on the day shown.
The derived information shown in Figure 2-31 indicates a similar pattern for the two detectors analyzed. Maximum flows seem to occur when the average delay is minimal—on the order of 5 to 10 seconds per 5-minute period. As the average delay increases above that range, there appears to be a reduction in total volume that is served by that part of the roadway, with reductions of about ten percent being found. However, going to the left in Figure 2-31 away from the maximum flow, as the average delay approaches zero, there also appears to be a relatively sharp drop in total volume.
An important observation to note goes to the main premise of the research question, which assumes that the traffic volume or demand is the independent variable, while the average speed or delay is the dependent variable. That may not necessarily be the case. The findings in this and the prior two cases show that strategies that reduce demand by small, marginal amounts can result in significantly less amounts of delay or congestion. However, trying to have a goal of restoring free-flow at all times and completely reduce congestion may not be the optimum situation. Trying to manage demand so that the traffic flows are located and timed everywhere at the same time to have free-flow conditions will be extremely difficult if not near impossible from a practical sense. That is because there seems to be a feedback in the relationship between volume and delay—perhaps interdependencies.
To keep flows at or near maximum during the peak of the peak requires the flow to be just upstream and ready to move onto the next part of the roadway, but without queuing or delay. However, if the managers try to avoid any delay, then being able to sustain a maximum traffic flow to pass through the next roadway segment becomes nearly impossible too—so avoidance of some minor delay will likely result in less than maximum flow or throughput of traffic. There appears to be a trade-off between minimizing delay and maximizing volume. Achieving both at the same time everywhere throughout a roadway system is truly a significant challenge. This is how we interpret the results shown in Figure 2-31.
On sample days in early August when demand is low there still tends to be some minor amount of congestion or delay, as shown in Figure 2-31 for 8-6-08 and 8-8-08 for the detector near MD 190. Even weekdays with large systemic demand reductions, the holidays of 9-13-07 and 10-08-07, still have some minor amounts of average delay. Trying to manage the demand so that the flows have no delay will likely result in under-utilized capacity—there would still be some minor amounts of delay but significant drops in throughput would likely result.
To better understand this apparent trade-off in management objectives, the next part of the analysis examined the paired-relationship between concurrent observations of volume and average speed.
In Figure 2-32 and Figure 2-33 for each Outerloop detector the paired-values of volumes (x-axis) and average speed (y-axis) per each 5-minute summary interval are plotted, respectively first for a typical day in August (8-8-07) and then for a typically congested day at another time of the year, such as in September (9-14-07). Eight hours of such data are shown in this case with the clusters of points on the left-side being for the 4-hour AM peak period of 6 to 10 AM, and the clusters in the mid and right-sides being for the 4-hour PM peak period of 3 to 7 PM. The following observations are noted based upon Figures 2-32 and 2-33:
So the question we need to address is: Do marginally higher volumes cause congestion? And if they do, what are the values of the higher volumes when that starts to happen?
As noted in the previous Figures, higher volumes are associated with speeds beginning to slow down; and if the flow rate volumes become high enough, and begin to exceed the roadway capacity, then: (a) significantly slower speeds, congestion and delay rapidly increase at the point of observation-detection; (b) the flow rate declines significantly; and (c) queues begin to form upstream from the point of observation-detection.
In Figure 2-34 the same sets of the 5-minute volumes of I-495 for the detector near Greentree Road are plotted versus average speed with lines connecting sequential points. The low-demand weekday of 9-13-07 was used here instead of the early August sample because the PM peak period volumes for 8-8-07 were exceedingly low even for a lower volume day. (There may be some data quality issues with part of the sample for that day.) Even with using the higher demand day of 9-13-07, as shown in Figure 2-38 almost all of the paired-values had average speeds over 60 mph, five had speeds between 55 and 60 mph, and just one had a speed of about 53 mph. In essence almost the whole peak period was free-flowing.
Examining the volumes in Figure 2-34, the 5-minute volumes barely exceeded a value of 400 vehicles, which is a good deal away from the estimated 5-minute flow rate capacity of about 525 vehicles. The volumes for 9-14-07 given in Figure 2-35 have perhaps two-thirds to three-quarters of the paired-observations over the value of 400 vehicles per 5-minute period—and perhaps half of the observations having moderately congested speeds of less than 50 mph, but over 35 mph.
It is noted that in both Figures the onset of the flow breakdown seems to have begun at volumes that are closer to the value of 400 vehicles per 5-minutes. Perhaps an explanation for this apparent discrepancy stems from the roadway geometrics just downstream from the location of this detector. The roadway in the vicinity before and after the detector is on a long down-grade that about 2,000 feet past the detector enters a relatively sharp turn (for Interstate standards) after which I-495 merges with the southbound I-270 Spur. There are warning signs indicating a sharp turn ahead that can be seen from the general vicinity of the detector near Greentree Rd. Unfamiliar or overly cautious drivers may slow down even if the volumes are lighter than capacity flow rates. That could explain why at lower volumes slower speeds are seen.
Another different finding seen in Figures 2-34 and 2-35 is in both AM clusters of points where lighter volumes have slower speeds—the opposite of one of the premises of the analysis. Perhaps here too the roadway geometrics can help explain the observation if the drivers widely separated from others may actually be more observant of speed limits and warning signs.
If the maximum 5-minute flow rate in the vicinity of this detector is about 525 vehicles then the maximum hourly flow per lane may be calculated. At this detector near Greentree Rd the I-495 Outerloop has 3 through lanes. Expanding the 5 minute volume of 525 vehicles by 12 (12 – 5 minute time periods per hour), the hourly maximum volume is 6,300 vehicles per hour. Dividing by 3 since there are three lanes, yields a per-lane maximum volume of 2,100 vehicles per hour. This corresponds well with the HCQSM Manual's maximum volume of 2,400 passenger cars per lane under ideal condition.
Turning the remainder of the analysis towards the other detector near MD 190, River Road, it is located just before the interchange with MD 190. Figure 2-36 shows the paired 5-minute volume and speed data for the early August light demand conditions of the sample day of 8-8-07. Figure 2-37 shows the heavier demand sample day from other times of the year of 9-14-07. The following general observations about the volume-speed relationships at that general location are made:
Regarding the capacity of the I-495 Outerloop near the MD 190 Detector, if the maximum 5-minute flow rate in the vicinity of this detector is about 575 vehicles then the maximum hourly flow per lane may be calculated. At this detector near the I-495 Outerloop effectively has three through lanes and two auxiliary lanes for traffic exiting at MD 190 and immediately after for I-495X, Clara Barton Parkway. Expanding the 5 minute volume of 575 vehicles by 12 (12 – 5 minute time periods per hour), the hourly maximum volume is 6,900 vehicles per hour. Dividing by 3 since there are effectively three lanes, yields a per-lane maximum volume of 2,300 vehicles per hour. This corresponds very well with the HCQSM Manual's maximum volume of 2,400 passenger cars per lane under ideal condition.
AM Peak Volume Analysis
Comparison of AM and PM Peak Period Volumes
The analyses presented above in the Task 2 discussion focused on evening PM peak period traffic conditions. During the review of the analysis a question was raised as to whether the seasonal reductions in volumes observed in comparing PM peak period summer-lite traffic to PM peak period traffic at other times of the year would also be the case for the AM peak period. That question was based upon the premise that is a long-standing generalization in transportation planning that, "the PM peak period tends to be much longer and more uniform in shape, while the AM peak is shorter and much more sharply peaked in its shape".
To address this question some additional analysis, using archived detector data from RITIS at the UMD CATT Lab, was carried out that examined the volume data for one detector on I-95 at Cherry Hill Rd and one detector on I-270 at MD 118, Germantown Road. Peak direction volumes for a four-hour AM peak period of 6 AM to 10 AM and a four-hour PM peak period of 3 PM to 7 PM in the opposing direction were summarized for selected days that were used in the prior analyses. The tabulation of that information for the Cherry Hill Road detector and the MD 118, Germantown Road detector is given below in Figures 2-38a and 2-38b, respectively. The tables identify the particular days that were selected for this analysis including: their date, day-of-the-week, and whether there were special conditions.
Figure 2-39 is a graph that plots and compares the AM versus the PM data from the two previous Figures. The AM peak volume scale is on the vertical axis, with the PM peak volume scale on the horizontal axis. The blue diagonal indicates the center point where data points would lie if there were equal volumes in the AM and the PM peak. This shows that in all cases the AM peak period peak direction volumes are less than the corresponding PM peak period peak opposing direction volumes for that same day. In the graph the days that were summer-lite days are marked with a circle around their data points, and the days that had special conditions are marked with a square around their data points. A generalization from the pattern shown in the graph for the limited number of days analyzed for these two locations is that there seems to be a relatively consistent relationship between the AM peak period peak direction volumes and the PM peak period opposing direction volumes for the same day that do not appear to differ much for: (a) the summer-lite days compared to the more normal traffic days at other times of the year, nor (b) the special days with light traffic compared to the more normal traffic days at other times of the year.
We thus conclude this additional analysis with the finding that the AM peak period peak direction volumes are about 80 to 85 percent of the PM peak period opposing direction volumes as a first approximation, although on any given day that range could be a somewhat smaller or larger percentage. That implies that the AM-oriented strategies of travel demand reductions may need shorter durations than during the afternoon peak period in order to have reduced delays. It is important to note that during the morning peak period significant congestion is experienced. The congestion however does not have as long a duration as the afternoon peak. But just as it is the case in the afternoon, a 10 to 14% drop in the morning peak volume will reduce delay. However, the reduction needed will only apply to that portion of the morning peak period that is congested.
What Does This All Mean?
The analysis for these sections of I-95, I-270 and I-495 has demonstrated that there are clearly many places where the capacity of those parts of the roadways can be reached and exceeded. When that starts to happen there are relatively quick, sharp and prolonged drops in speed and a decrease in throughput of the traffic flow. It can sometimes take a considerable amount of time until the upstream volumes and demand decline enough to enable the roadway operating at reduced performance to once again have flows that can get back to freer flowing speed.
We also have seen examples in the analysis where the reduced roadway performance propagates or flows downstream from the constriction locations. We have seen that the turbulence in flow that occurs at the bottlenecks does not always clear up just after vehicles pass that point. Many drivers may still be driving cautiously while others are impatient so that in the mix of traffic there can be speed-ups and slow-downs taking place somewhat simultaneously until the vehicle density spreads out enough to absorb the fluctuations in flow. Then the cautious drivers begin to drive at freer flowing speeds and the whole flow itself tends to operate without congested conditions.
In conclusion, a main finding of this Study has been that if relatively small changes can be made in peak demand (volume) through various programs and strategies, such as congestion pricing, then two beneficial things can happen: (1) there can be relatively large decreases in congestion and delay at key choke points, and (2) there can be increased through-put along those roads during peak times of travel—thus by effectively managing the demand more travelers can be served per time period with the available fixed-supply of roadway capacity.
The analysis has shown that in many instances the amount of needed demand reduction can be on the order of five to ten percent of the peak period flow. However, there still may be a particularly difficult bottleneck in a corridor that would need reductions on the order of 15 to 20 percent. While demand reduction and operational strategies may be able go a long way in improving the flow at those locations so that capacity is exceeded less often and/or recovery is quicker, there still may need to be localized geometric or lane use changes at those locations to routinely have freer flowing traffic at those troublesome locations.
Based on our analysis we now have the information and data available to estimate the delay reduction associated with traffic reduction.
Summary Traffic Analysis
The review and evaluation of traffic data on the Washington DC Interstate System has provided a clear picture of the relationship between traffic flow, speed, and delay. All three highway sections on I-95, I-270, and I-495 showed similar traffic flow characteristics. In fact, there were no differences in the relationship between flow, speed, and delay whether the facility was a radial or circumferential freeway. And the traffic flow relationships from the field data validated both the maximum per lane capacity and the operational flow during congestion.
The recorded maximum capacities in all three sections ranged approximately between 2100 and 2400 vehicles per hour. These values compare quite well with the Highway Capacity and Quality of Service Manual (HCQSM) speed-flow curves as shown in Figure 2-40.
In the Speed Flow Density Curve, the maximum flow rates are in passenger cars per lane per hour. The Congestion Delay Study results are in vehicles per lane per hour. If the Study's capacity results were adjusted for heavy vehicles, and geometric conditions such as interchange spacing and traffic friction at the ramps and weaving sections, the maximum capacities would be even closer to the HCQSM Speed Flow Density Curves.
It should also be noted that Figure 2-40 (HCQSM Exhibit 23-3 – Speed Flow Density Curve) is a discontinuous curve. In other words, if the demand on the freeway system exceeds the maximum capacity (in this case 2400 passenger per hour), then the flow breaks down, speeds drop to the 20 to 40 mile per hour range and the capacities are reduced to approximately 2000 passenger cars per lane per hour. Based on the field data this is exactly what occurred in all three sections that were analyzed in the study.
Having verified the speed-flow-delay relationship, the reduction in delay due to a reduction in traffic flow or the shifting of traffic from the peak period may be estimated. As part of the study, traffic volume, speed, and delays in August and during holidays was evaluated and compared to the volume, speed, and delays in October when the traffic was heavier. Therefore the reduction in delay if modest reductions in volume occur may be estimated. What follows is an estimate of the potential delay reduction that may result if volumes are reduced.
I-95 Section Delay Analysis - I-95 Northbound from Brooklyn Bridge to Montgomery Road
Based on Figure 2-2 and Figure 2-3 for the 4 hour afternoon peak period that was analyzed the delay for the comparatively light travel days in August was estimated to be an average of 40 seconds per vehicle. During October, with heavier traffic flows the delay was estimated to be about 150 seconds per vehicle. As noted earlier in the report delay is calculated by taking the time difference between driving the 10.5 mile section of roadway at 65 miles an hour versus the time it would take to drive the section at the congested speed. Therefore the travel time would be:
The estimated congested travel times are the average over the 4 hour afternoon peak period. Individual afternoon peak hours will have significantly higher and lower delays.
The next step in the analysis is to estimate the vehicle hours of delay for the typical days selected in August and October.
On a typical day approximately 25,000 vehicles traverse the I-95 study section. The difference between average delay in August versus October is 110 seconds (150 – 40).
If the average vehicle occupancy rate for all trip purposes in the Washington, DC area is 1.2 then the person hours of delay is 920 hours. A typical driver values their time at 1/3 to 1/2 of their hourly wages. If each person values their time at $15 per hour then the delay on this one 10.5 mile section of I-95 is costing $13,800 per afternoon peak period (920 X $15/hour). If converted to a yearly cost, the delay cost would be $3,450,000 ($13,800 X 250 working days).
This can also be expressed as a delay cost / vehicle mile which would be:
Percent Reductions in Traffic and Delay
In October it takes an extra 150 seconds to traverse the I-95 10.5 mile study section. In August during light traffic it takes an extra 40 seconds. Therefore if we reduce our traffic below the tipping point of congestion then the percentage reduction of delay would be:
In order to get this reduction in the I-95 corridor we would need a drop in the traffic from 25,000 vehicles during the afternoon peak to 22,000. This is a traffic reduction of 12% and corresponds to the average traffic difference between August and October. A 12% traffic reduction would result in a 73% reduction in delay.
Analysis of the Peak Hour of the Afternoon Peak Period
During the peak hour in the afternoon peak period, the traffic flow is approximately 8,000 vehicles versus 25,000 vehicles during the entire afternoon peak period. From Figure 2.1 the average delay on the I-95 study section during the peak hour is 350 seconds. With no congestion the sections travel time is 570 seconds and with peak hour congestion it is 920 seconds. By reducing the demand by 12% (typical traffic reduction for August versus October) to 7400 vehicles during the peak hour there would be a significant percent delay reduction as the following calculations will show:
By reducing demand 12% we can achieve a delay reduction of 89%. The cost savings to the individual vehicle on a per mile basis would be:
Note: VOT = Hourly Value of Time
The savings in delay during the peak hour if traffic is reduced during the peak hour by 12% would be 14.76 cents/ vehicle mile.
I-270 Delay Analysis: I-270 Northbound – Section from I-370 to MD 109
Based on Figure 2-14 and Figure 2-15 for the 5 hour afternoon peak period that was analyzed the delay for the comparatively light travel days in August was estimated to be an average of 80 seconds per vehicle. During October, with heavier traffic flows the delay was estimated to about 300 seconds per vehicle. As noted earlier in the report delay is calculated by taking the time difference between driving the 12.9 mile section of roadway at 65 miles an hour versus the time it would take to drive the section at the congested speed. Therefore the travel time would be:
The estimated congested travel times are the average over the 5 hour afternoon peak period. Individual afternoon peak hours will have significantly higher and lower delays.
The next step in the analysis is to estimate the vehicle hours of delay for the typical days selected in August and October.
On a typical day approximately 22,000 vehicles traverse the I-270 study section. The difference between average delay in August versus October is 220 seconds (300 – 80).
If the average vehicle occupancy rate for all trip purposes in the Washington, DC area is 1.2 then the person hours of delay is 1613 hours. A typical driver values their time at 1/3 to ½ of their hourly wages. If each person values their time at $15 per hour then the delay on this one 10.5 mile section of I-270 is costing $24,195 per afternoon peak period (1613 X $15/hour). If converted to a yearly cost, the delay cost would be $6,048,750 ($24,195 X 250 working days).
This can also be expressed as a delay cost / vehicle mile which would be:
Percent Reductions in Traffic and Delay
In October it takes an extra 300 seconds to traverse the I-270 12.9 mile study section. In August during light traffic it takes an extra 80 seconds. Therefore if we reduce our traffic below the tipping point of congestion then the percentage reduction of delay would be:
In order to get this reduction in the I-270 corridor we would need a drop in the traffic from 22,000 vehicles during the afternoon peak to 19,000. This is a traffic reduction of 14% and corresponds to the average traffic difference between August and October. A 14% traffic reduction would result in a 73% reduction in delay.
Analysis of the Peak Hour of the Afternoon Peak Period
During the peak hour in the afternoon peak period, the traffic flow is approximately 6,800 vehicles versus 22,000 vehicles during the entire afternoon peak period. From Figure 2.14 the average delay on the I-270 study section during the peak hour is 420 seconds. With no congestion the sections travel time is 716 seconds and with peak hour congestion it is 1136 seconds. By reducing the demand by 14% (typical traffic reduction for August versus October) to 6000 vehicles during the peak hour there would be a significant percent delay reduction as the following calculations will show:
By reducing demand 14% we can achieve a delay reduction of 81%. This demand to delay percentage ratio is similar in scale to the delay reduction found on I-95.
The cost savings to the individual vehicle on a per mile basis would be:
Note: VOT = Hourly Value of Time
The savings in delay during the peak hour if traffic is reduced during the peak hour by 14% would be 13.18 cents/ vehicle mile.
In Sections 3 and 4 that follow, we address how many peak period trips can be truly deemed discretionary, and therefore may be able to shift times to reduce congestion (Section 3), and how effective congestion pricing and similar strategies have been in encouraging travel mode and travel time pattern changes (Section 4).
 Although AM Peak 4-hour volumes are 15 to 20 percent lower than PM Peak volumes, drivers do experience congestion and delay in the AM Peak, although it may be more peaked or of shorter duration than congestion in the PM Peak. A full analysis of bottlenecks and series of data, comparable to that performed for the PM Peak, would be required to establish "why" and "where" there is AM Peak delay, and "how much" is there, in order to define comparable volume reduction targets for the AM Peak as for the PM Peak.
United States Department of Transportation - Federal Highway Administration