Operational Evaluation of Part-Time Shoulder Use for Interstate 476 in the State of Pennsylvania

,


Introduction
e use of the shoulder as a travel lane is applied mainly in European countries such as the Netherlands, Germany, and the United Kingdom [1]. e extent of PTSU in Europe is expansive with 1000 kilometers in the Netherlands, over 200 kilometers in Germany, and an initial evaluation of 18 kilometers in the United Kingdom, as of 2010 [1]. In the USA, implementations began as bus-on-shoulder (BOS) lanes but have expanded to all types of PTSU in sixteen states with nine of them being at least a general purpose lane of PTSU [2]. Each of these sixteen states implemented PTSU in their unique design to fit the requirements for the highway. Some use simple signage to indicate operating times, and others combine PTSU with other roadway improvement methods such as high occupancy vehicle (HOV) lanes. Overall PTSU in the USA is predominantly transit based [3]. With the variety of circumstances of transportation infrastructure in the USA, there is a need to further expand the understanding of PTSU in USA applications, especially in localized research. ough the form of PTSU presented in this study is a simple implementation that utilizes existing technologies, results indicated significant benefits. With further improvements in smart cities and connected vehicles, the current literature suggests the dynamic lane management strategies will become increasingly more relevant and effective, making understanding PTSU imperative. [4].
I-476 in Philadelphia, Pennsylvania, is a prime candidate for PTSU implementation. e roadway decreases from a six-lane highway to a four-lane highway for the southern 14.5 kilometers of the roadway before it splits into Interstate 95 (I-95), shown in Figure 1 in Methodology and Calibration Results. Within the four-lane highway segment, the northbound (NB) hourly volumes range 3,600 to 4,700 vehicles per hour (veh/h) and the southbound (SB) hourly volumes range 2,400 to 4,200 veh/h. Due to these high volumes and the reduced geometric cross section, the Delaware Valley Regional Planning Commission (DVRPC), the regional Metropolitan Planning Organization (MPO) of Greater Philadelphia, has included the implementation of PTSU on this section of highway in the long-range plan [5][6][7]. e nearest example of a significant implementation of PTSU is in the Washington, District of Columbia area, which is two to three hours south of I-476. is research aims to evaluate possible operational effects of PTSU on I-476 with a Vissim-based model case study [8].

Literature Review
e major implementations of PTSU, or known internationally as hard shoulder running (HSR), can be broken down into three types, BOS, static, and dynamic PTSU [2]. BOS PTSU is common around major metropolitan areas where there is significant transit traffic. BOS allows the use of the shoulder for bus travel during specific periods throughout the day to ensure transit reliability. Static PTSU is the use of the shoulder as a travel lane at specific times of the day, typically the primary commuting hours. Static PTSU could include various vehicle type restrictions such as passenger car only lanes, heavy truck only lanes, or HOV lanes [2]. Dynamic PTSU is similar to static PTSU, but the hours of operation are adjustable. Instead of specific hours of operations, the shoulder lanes open when the density of the roadway increases to a site-specific level that would benefit significantly from PTSU.
is method is ideal for areas where traffic volumes are highly variable throughout the day such as roadways around major tourist areas where the arrival times to the area are sporadically throughout the day.
As PTSU becomes more prominent in the USA, the operational effects will be further evaluated and understood. Even within the limited applications in the USA, a report by the Federal Highway Administration (FHWA) summarized the capacity potential of PTSU in the USA to range from 1,000 to 2,000 veh/h, which agreed with another USA-based study out of the state of Virginia [2,9]. e one simulationbased study on a roadway in Buffalo, New York region, evaluated the effects of the length of the PTSU areas (length of the bottleneck versus the length of the bottleneck along with the upstream queue) and the width of the shoulder lane (3.05 meters versus 3.66 meters) [10]. e shorter and narrower PTSU implementation had a capacity of 1,262 veh/h, while the longer and wider PTSU implementation had a capacity of 1,687 veh/h. Others have found the PTSU lane to be able to handle 1/3 to1/2 of the mainline traffic volume [2,11]. Other studies have found the capacity to be around 15 to 20% of the mainline volume, depending on the number of lanes on the highway [11,12]. e total capacity of a roadway has been found to increase by 7 to 35% by European-based projects [13][14][15]. e additional capacity created by a third lane on a two-lane segment of a United Kingdom (UK) highway was found to offset the significant negative effects of higher lane volumes on the average vehicle speed until the total segment volume increased by about 75% [16]. A theoretical study based on Interstate 70 (I-70) in the state of Colorado, USA, determined that an additional lane for a two-lane segment of I-70 could reduce the lane density by 33%, assuming no induced volume caused by the improved operational performance [17]. However, a study based in the state of Texas, USA, found that the overall roadway density could reduce up to 70% [18]. e operational performance effect indicators vary depending on the location. An evaluation of PTSU in the state of Washington found that the travel time reduced from nine minutes to one and a half minutes for a 2.49-kilometer segment of PTSU [2]. A United Kingdom study for the M42 motorway found that the variability of the travel time decreased by 27 to 34%, and a France-based study found an over improvement in travel time reliability [19,20]. Congestion was found to reduce by up to 35% in a German study and congestion frequency to reduce by up to 82% [21,22]. A study based in Alabama, USA, found that left shoulder PTSU could reduce the total travel time by 0.54 minutes per kilometer and 0.05 minutes per kilometer for right shoulder PTSU [23]. e overall reduction in delay was determined to be 0.25 to 0.62 minutes per kilometer of PTSU implementation, based on two studies [2,23]. e design can be optimized to minimize total time delay during traffic incidents [24]. An analysis about the density of a highway with shoulder use has found to improve safety due to the overall reduction in the density of the highway [25]. ese results highlight the variability of the operational effects of PTSU.
is study aims to improve the knowledge of PTSU for the USA implementations. ere are four total interchanges, Interchanges 1, 3, 5, and 9. e analysis segments are from the center of each interchange to the next interchange. For the NB direction, the analysis segments are between Interchange 1 to 3, Interchange 3 to 5, and Interchange 5 to 9. e SB direction analysis segments are the same but with the interchanges in the decreasing number. Figure 1 includes a network map of the modeled interstate with the hourly volumes used within the analysis. e analysis period is the two-hour morning commuting period from 7 AM to 9 AM. e NB direction has higher volumes overall though the SB direction has volumes that are comparable in specific segments. e percentage of trucks on I-476 is approximately 10% throughout, and 10% will be used as the network average for the Vissim model. e posted speed limit for I-476 is 55 miles per hour (mph; 88.5 kilometers per hour (km/h)). e model was calibrated using travel times that were collected in autumn 2015 by researchers at Villanova University, shown in Table 1. e model was calibrated using a single seed number where the simulated travel time was found to be not statistically different from the real world measured travel times of the analysis segments. Once completed, the variability was assessed using the speed between the interchanges to validate the model by determining if the variability of the model to be less than 10% from the average values measured within the microsimulation based on ten simulation runs using ten random seed values. e speed data were from INRIX data provided by the Regional Integrated Transportation Information System online database, Probe Data Analytics Suite [26]. e model was run for ten simulation runs which were determined to be sufficient to capture the variability of the microsimulation model with 95% confidence and completed the calibration process. e average travel times of the ten randomly seeded simulation runs for each analysis segment is included below in Table 1.

Results and Analysis
e benefit of PTSU on the operational performance is evident in the NB performance change from the base I-476 conditions. Figure 2 includes the average travel time along with the 95% confidence interval of the base scenario and the aforementioned PTSU scenarios. e variability of the scenarios is relatively minimal. e highest variability is in the SB segment, Interchange 3 to 1, where the segment ends in a split of I-476 with one direction exiting onto I-95 SB and the other onto I-95 NB. Interchange 1 SB has high levels of lane changing that can cause inconsistent travel times at the end of this analysis segment. e SB direction only saw a significant reduction between Interchanges 5 and 3 where the reduction is around 43%. e other SB segments saw a reduction of zero to 15% when compared to the base scenario. Due to geometric limitations of one of the I-95 ramps, with the induced PTSU scenario, there is a significant increase in travel time between Interchanges 3 and 1 and a minimal increase between Interchanges 5 and 3. Without this geometric limitation, the performance for the SB direction with the induced PTSU scenario is expected to be similar to the induced PTSU in the NB direction, lower or similar to the base scenario travel time values. For the other NB PTSU scenarios, the reduction is over 60% between Interchanges 1 and 5. From Interchange 5 to 9, the speed of the traffic was near the posted speed limit in the base scenario, and with PTSU, the travel time reduced by about 23% from the base scenario. Between all of the analysis segments, there is minimal difference between the noninduced PTSU scenarios in average travel time in both directions. e most variable is between Interchanges 3 and 1 SB where the need for lane changing is higher for trucks in the PTSU lane which caused an increase in travel time. However, the other noninduced PTSU scenarios were statistically similar to each other in the SB Interchange 3 to 1 segment. e density of the NB segments depicted in Figure 3 illustrates how the overall congestion forms within the network. In the base conditions, the congestion begins between Interchanges 3 and 5 before it extends into the Interchange 1 to 3 segment. e highest volumes are between Interchanges 5 and 9, but the end of that segment introduces a new lane, and it is the longest segment of I-476 in the analysis network without a ramp. ese two factors allow for better operations between Interchanges 5 and 9 where the density stabilizes at the border between LOS D and LOS E. is allows for sporadic areas of congestion to form due to unstable flow but not a complete breakdown of the traffic flow like the other NB segments. e other two NB base scenario segments achieve a LOS of F within the first 30 minutes of the two-hour congestion period. With PTSU, the density drops to around sixteen vehicles per kilometer per lane where the LOS of I-476 is between LOS C and LOS D. For the induced PTSU scenario, the density between Interchanges 3 and 5 increases similarly to the base scenario, but it levels off for the rest of the two-hour commuting period. e level off in density for the Interchange 3 to 5 segment allows for the density for the Interchange 1 to 3 segment to peak and decrease in the second hour of the analysis periods where the traffic volumes decrease. While the Interchange 3 to 5 segment operates at a LOS F, the PTSU minimized the increased LOS F density of the Interchange 1 to 3 segment to be maintained for only an hour instead of the base scenario of an hour and a half.
e SB direction has less congestion which results in the density over time profile for the SB direction to be simpler Advances in Civil Engineering than the NB direction, shown in Figure 4. In the base conditions, the density increases to just above a LOS C in the second hour of the analysis period. In the induced PTSU scenario, due to the geometric issue mentioned, the density begins to increase around 30 minutes into the two-hour analysis period for the Interchange 3 to 1 segment. e Interchange 3 to 1 segment eventually levels o at around 80 minutes into the analysis period when Interchange 5 to 3 segment begins to increase as the congestion extends into that segment. e noninduced PTSU scenarios operate around LOS B except for the last half hour of the analysis period. Due to the higher volumes in the second hour for the SB direction, the density between Interchanges 3 and 1 begins to increase, but in the worst case scenario, the truck PTSU scenario, only reached LOS F at the end of the twohour analysis period. e general PTSU and passenger car PTSU only reach a LOS of D by the end of the analysis period.   is expanded the overall range determined in the literature, 0.05 to 0.62 minutes per kilometer [2,23]. e two most congested segments have a travel time reduction range of 1.08 to 1.86 minutes per kilometer. e higher initial levels of congestion presented an ideal scenario for increased travel time reductions. e travel time reductions in the other analysis segments fell within the range determined in the literature review section [2,23]. Similar to the travel time reductions, the two most congestion segments of I-476, between Interchange 1 and Interchange 5 NB, have lane density reductions higher than the range determined in the literature [17,18]. However, the increased density reductions found in this study are only slightly higher than the literature upper limit for roadway density reduction. e less congested segments had lane density reductions similar to the lower limits of the literature review range. e results of this study fit within the current body of knowledge of PTSU. However, it does provide additional detail on the variability of the operational benefit of PTSU over time and location.

Conclusions
e objective of this study was to understand the operational effects of PTSU. e addition of a lane during the two-hour morning commuting period provided significant improvement to the travel time. e NB direction had higher levels of congestion, so the benefit was predominantly in the NB direction. However, the SB direction did have significant travel time reduction in the one analysis segment with higher levels of congestion in the base conditions. e use of PTSU had a significant effect on the lane density throughout both directions. Instead of segments of I-476 becoming unstable during the morning commuting period, the traffic flow remained relatively stable. With the higher volumes in the NB direction, the traffic flow is not going to be perfectly stable as it was on the border between LOS C and D. However, the traffic density will be less variable over time than the base scenarios where the density is increasing throughout the two-hour analysis period. Based on the conditions in this model, the optimal PTSU scenario was the passenger cars only PTSU scenario. Trucks in the PTSU lane adds merging issues at interchanges and lane changing issues at interchanges such as Interchange 1 SB where there is a major roadway split. Depending on the truck volume on a roadway, the effects of the trucks will vary. PTSU is a unique solution that utilizes the existing roadway more efficiently and provides a stable roadway that operates at a faster speed.
From this study, the following key points can be applied to other PTSU scenarios. e improved operational performance, created by PTSU, may induce traffic. Understanding the geometric limitations of the roadway segment can be a significant impact on the effectiveness of the implementation, as any current limitations will likely be exacerbated with PTSU. If possible, it is recommended to use the left shoulder as the literature suggests that the left   shoulder PTSU can be more beneficial than right shoulder PTSU [22]. Also, the vehicle composition of the roadway could be a determining factor if vehicle type restrictions are necessary for any future PTSU implementation such as the quantity of heavy vehicles impacting the shoulder pavement performance. One last key point that could influence the PTSU implementation is the extent of the area that could benefit from PTSU. For example, if PTSU is only needed between two interchanges, extending PTSU further could reduce the cost-effectiveness of the implementation. However, in cases where PTSU is not extended far enough along the roadway segment, the operational benefit may be minimized by significant pockets of congestion still existing along the roadway segment. ese key points can aid in increasing the overall effectiveness of PTSU in other scenario applications.

Future Work
is study focused on ideal conditions where no incidents are present. Future work will evaluate the effects of PTSU in crash-based scenarios that obstruct one of the lanes during the entire commuting period. Blocking the lane for the entire commuting period will provide a worst case scenario for a crash-based scenario but will add insight to the effects of PTSU in crash mitigation during the peak commuting period. e inclusion of crash-based scenarios will strengthen the understanding of PTSU operation effects by evaluating PTSU functionality in nonideal operating scenarios.

Data Availability
e data used to support the findings of this study are available from the corresponding author upon request.