Incorporating Truck Analysis into the Highway Capacity Manual (2014)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

99 This section presents the recommended methodology for estimating truck speeds on arterial segments in between signalized intersections. 9.1 Existing Truck Treatment on Arterials in the HCM Arterial analyses in the HCM (including signalized intersections, stop-controlled inter- sections, and roundabouts) use heavy-vehicle PCE values to adjust the saturation flow rates. Unlike the freeway methods, truck PCEs are not used to adjust the vehicle flow rates. At the intersections, the saturation flow rate is adjusted by a heavy-vehicle adjustment factor, fHV, as illustrated by this equation from “Chapter 18: Signalized Arterials,” in the HCM: Equation 460s s f f f f f f f f f f fw HV g p bb a LU LT RT Lbp= where s = the adjusted saturation flow rate, s0 = the saturation flow rate under ideal conditions, fHV = the heavy-vehicle adjustment factor, and all other f.. values = other adjustment factors. This same type of adjustment is used in stop-controlled intersections and roundabouts. As with the freeway analysis methods, the heavy-vehicle adjustment factor fHV is given by 100 100 1 Equation 47f P E HV HV T( )= + − where PHV = the percentage of heavy vehicles and ET = the PCE value; the PCE value for heavy vehicles is always 2.0. The arterial is treated as a series of segments (see HCM, Chapters 16 and 17). Each segment begins and ends at a stopbar. A segment can have intermediate stop-controlled intersections and round- abouts, but no signalized intersections. A 10-step process is used to determine the “automobile LOS.” Average speed is used to assess the LOS for the vehicular traffic stream in combination with the V/C ratio. The average speed also gives an indication of delay (and travel rate). Once the steps have been completed for each segment in the arterial, the overall metrics are determined through a distance-weighted average. S e c t i o n 9 Prediction of Arterial Truck Speeds

100 incorporating truck Analysis into the Highway capacity Manual For each segment, the base free-flow speed is computed using HCM Equation 17-2: Equation 480 0S S f ff CS A= + + where Sfo = the base free-flow speed, S0 = a constant, fCS = an adjustment for cross section, and fA = an adjustment for the access points. It would be possible to add an adjustment here for the truck mix, but this is not presently our first choice. The base free-flow speed is then adjusted in HCM Equation 17-4 to account for intersection spacing through an additional adjustment factor fL: Equation 490S S ff f L= A second equation (HCM Equation 17-3) is used to compute fL. A subsequent equa- tion (HCM Equation 17-5) provides an adjustment based on vehicle proximity (effectively density): 2 1 1 52.8 Equation 500.21f v N S v m th f = + −    p where fv = the proximity adjustment factor, vm = the mid-segment demand flow rate, Nth = the number of through lanes on the segment, and Sf = the free-flow speed. Exhibit 66 shows the effect of this mid-segment lane flow rate (veh/hr/lane) on the running speed. There is no discussion about trucks, so the assumption is that the relationships are for situations where the influence of heavy vehicles is negligible. This effect of proximity is one of several adjustments that appear in HCM Equation 17-6 to compute the running time for the segment: 6.0 0.0025 3600 5280 Equation 51 1 , 1 1 t l L f L S f d dr x f v ap i other Nap ∑= − + + + = where tr = the running time, l1 = the start-up lost time, L = the segment length, fx = a control-type adjustment factor, Sf = the free-flow speed, fv = the proximity adjustment factor, dap,i = the delay due to left and right turns from the street into access point intersection i,

Prediction of Arterial Truck Speeds 101 Nap = the number of influential access point approaches along the segment, and dother = delay due to other sources along the segment. For an overall arterial, the base free-flow speed is computed via Equation 52, 1 , 1 S L L S fo F ii m i fo i i m ∑ ∑ = = = where Sfo,F = the base free-flow speed for the facility, Li = the length of segment i, m = the number of segments on the facility, and Sfo,i = the base free-flow speed for segment i. The actual travel speed for the arterial is computed in a similar manner using Equation 53, 1 , , 1 S L L S T F ii m i T seg i i m ∑ ∑ = = = where ST,F = the travel speed for the facility, Li = the length of segment i, m = the number of segments on the facility, and ST,i = the travel speed for segment i. Exhibit 66. Speed-flow relationship for urban street segments.

102 incorporating truck Analysis into the Highway capacity Manual 9.2 Approach The new methodology captures the effects of trucks on arterial speeds in two places. The first is at the intersection where the through-delay and through-stop rate are determined for point facilities like signalized intersections. Here, new PCE values have been generated that adjust the saturation flow rate. The second is at the midblock location between intersections, where the running time is determined for the section of the segment upstream of the control point (i.e., changes and/or adjustments). The intent in this latter case was to do this in a manner similar to that described previously for freeways. Development of the new methodology was accomplished in six steps: 1. Develop acceleration profiles (acceleration versus speed, one for each truck type) that can be used as inputs to VISSIM. Compare the trajectory predictions (speed versus distance) of these profiles with those from prior studies. This is identical to the freeways. 2. Use a VISSIM model of a single-lane arterial with a constant grade to see if any VISSIM param- eter values needed to be adjusted to generate speed-distance trajectories that were consistent with the findings from Step 1 above. This is identical to the freeways. 3. Conduct simulations of a wide range of truck types, weight-to-horsepower ratios, truck percentages, flow rates, and grades using a simulation model of an arterial segment that is two lanes wide with constant grades. Do this initially for one truck type (focused prin- cipally on FHWA Classes 5 and 9), and then for mixes of truck types and weight-to- horsepower ratios. This is similar to the freeways except a two-lane-wide arterial was employed. 4. Determine what predictive relationships can be used for truck PCEs and speeds based on the data from Step 3. This is similar to the procedure that was described for freeways. 5. Collect saturation flow rates for signalized intersections and see how these flow rates are affected by the truck mix. Prepare PCE values that can be used to properly adjust the satura- tion flow rates to those observed. 6. Test the resulting treatments for truck PCEs and speeds using a quasi-real case study whose setting is based on a real-world facility, but whose design and traffic mix details have been treated parametrically to allow tests of the effects of various other conditions (e.g., flow rates, truck mix percentages, and grades). 9.3 Acceleration Profiles Steps 1 and 2 involved developing the truck acceleration functions (relationships between the maximum acceleration and speed) that would be used in VISSIM to model midblock arte- rial segment speeds. This work and the results obtained were the same as was the case for the freeway analysis. 9.4 Midblock Arterial Segment Speed Model Development Steps 3 and 4 were focused on conducting the VISSIM simulations and developing the pre- dictive models for midblock arterial truck speeds. At first, the expectation was that these steps would be done in series; however, as was the case with the freeway analysis, the results from the VISSIM simulations suggested useful ways to think about the predictive equations, so the two steps were done in parallel.

Prediction of Arterial truck Speeds 103 9.4.1 Test Site Selection The investigation of the arterial segment midblock speed flow effects of trucks on extended grades was performed on a hypothetical arterial with no signals, consisting of 8 miles of level 6-lane street followed by 5 miles of 6-lane street with a grade. The grade on the 5-mile section varied from –6% to +6% in 1% increments (13 grades total). The hypothetical facility was simu- lated using the VISSIM microsimulation model (PTV Group, n.d.). 9.4.2 Simulation Model Application About 6,552 combinations of truck mix, grade, and traffic flow rate were simulated, as was the case for the freeway analysis. The parameters for each combination were as follows: • FHWA Class 5 and 9; • Weight-to-horsepower ratios: 50, 100, 150, and 200 lbs/hp; • Grades: –6% to 6% (13 grades total); • Truck percentages: 0, 10%, 20%, 30%, 40%, 50% and 100%; and • Flow rates: 180, 450, 900, 1350, 1440, 1530, 1620, 1710, and 1800 veh/hr/lane. The flow rates were intended to be equivalent to V/C ratios of 10%, 25%, 50%, 75%, 80%, 85%, 90%, 95%, and 100% for the all-automobile condition. As with the freeway analysis, scenarios were formed by grouping together the nine V/C con- ditions associated with each combination of FHWA class, weight-to-power ratio, grade, and truck percentage. This resulted in 637 scenarios: 520 mixed scenarios (24513) plus 13 all- automobile scenarios plus 104 all-truck scenarios (2413). The methodology was developed based on these scenarios. 9.5 The Predictive Procedure for Midblock Arterial Segment Speeds The predictive procedure is very similar to the one created for freeways. It makes use of the same truck speed prediction model to create the predictions of truck speeds based on grades and segment lengths. It uses a set of equations to predict what the truck and automobile speeds will be. One of the 637 scenarios can be used to illustrate the predictive procedure’s main ideas. Exhibit 67 shows a plot of 1-min flow-density data points for a mixed traffic stream on a 6% upgrade involving 30% Class 9 trucks at 150 lbs/hp. It also shows the flow-density relationship for an all-automobile traffic stream with a +6% grade. It is immediately obvious, as was in the case of the freeway analysis, that the data points for the mixed traffic stream lie well below those for the all-automobile condition. Implicitly, the speeds are very different (the slopes of the relationships). Moreover, the maximum density achieved by the mixed flow is greater than that for the all-automobile flow. In Exhibit 67, the graph shows an example flow-density relationship for an arterial segment that is two lanes wide on a +6% grade for both a mixed traffic stream involving 30% Class 9 Trucks with 150 lbs/hp and an all-automobile traffic stream. Insofar as truck speeds are concerned, Exhibit 68 shows the speed-flow plot for this same con- dition. It also shows the speed-flow relationships for the truck and automobile speeds separately (in the 30% trucks case) as well as two other cases: automobiles only and trucks only.

104 incorporating truck Analysis into the Highway capacity Manual 0 500 1000 1500 2000 2500 3000 0 10 20 30 40 50 60 70 Fl ow R at e pe r L an e (v eh /h r) Density per Lane (veh/mi) Flow-Density Relationship Auto Only Scenario Exhibit 67. Flow-density relationships for an arterial segment. 0 10 20 30 40 50 60 0 200 400 600 800 1000 1200 1400 1600 1800 Sp ee d (m ph ) Fl Note: for a +6% grade for both a mixed traffic stream involving 30% Class 9 trucks with 150 lbs/hp and an all-automobile traffic stream. ow Rate per Lane (veh/hr) Speed-Flow Relationship Auto Only Scenario Autos Trucks Exhibit 68. Speed-flow relationships on a 6% grade  30% Class 9 trucks.

Prediction of Arterial truck Speeds 105 The data points marked “Scenario” are the average mixed speeds for the 30% truck condition. At low flow, they are scattered between 40–50 mph, but they quickly drop to around 20 mph (the truck crawl speed) as the flow rate increases. The automobile speeds, marked “Autos,” follow a similar trend (as they should since 70% of the traffic stream is automobiles). This motivates a prediction model that allows the values to drop from the all-automobile condition to the all- truck condition. This is a new idea in the context of the HCM procedures. The trucks speeds, marked “Trucks,” are all at the crawl speed for the 6% grade, as they should be given the length of the segment (5 miles). The data points marked “Auto Only” are for an all-automobile condition. To make the V/C ratios match, the automobile-only flow rates are downward adjusted so that the automobile- only maximum flow rate matches that of the mixed scenario. This is effectively the reverse of the process described for Exhibit 67. The automobile-only speeds stay at or above 40 mph until capacity is reached. The data points marked “truck only” are from a simulation of a traffic stream involving 100% trucks. To make the V/C values match in this instance, the flow rates are proportionally adjusted so that the actual maximum flow rate in the all-truck circumstance (855 veh/hr/lane) maps to the maximum flow rate in the case under study (again, 1500 veh/hr/lane). 9.5.1 Truck Speeds on Arterial Segments Truck speeds (excluding intersection delays) are predicted in the same manner as they were for freeways. The truck speed is developed from the acceleration function, the deceleration function, and the length of the grade: , , Equation 54s L t L g s t o( )= ω where L = the length of the grade segment and t(L|g,so,ω) = the time required to travel the distance L given the grade involved, g, the truck’s initial speed upon entering the segment, so, and the truck’s acceleration capabilities ω. As for the freeway analysis, the same specific equations are used to predict the movement of the truck through time (see Equations 28–38). 9.5.2 Automobile Speeds on Arterial Segments Automobile speeds (excluding intersection delays) are predicted using a logistics equation, as was the case for freeways: 1 Equation 55s s s s e e a to ao to v v V v v V m m ( )= + − +       −β − ∆ −β − ∆ where sa = the automobile speed at flow rate v, sao = the automobile-only speed that would arise at flow rate v (taking into account the PCE value),

106 incorporating truck Analysis into the Highway capacity Manual sto = the truck-only speed that would arise at flow rate v (again taking into account the PCE values for the mixed flow case and the all-truck case), vm = the flow rate at which the automobile speed has accomplished half of its transition from sao to sto, DV = the range of flow rates over which the transition occurs, and b = a calibration coefficient that ensures the following holds true: 2 5 2 5 Equation 56 v V V and v V V m m −β − ∆ ∆ = − β + ∆ ∆ = − This ensures that the logit term within the large parentheses is approximately equal to 1 when v = vm – DV/2 and equal to 0 when v = vm + DV/2. Exhibit 69 shows the automobile speed function that was fitted to the automobile speeds in Exhibit 68. The smooth line represents the automobile speed estimated by Equation 4 and appropriate values of vm, DV, and b. Exhibit 69 shows the estimated automobile speed relationship for arterials using the speed- flow relationships on a +6% grade for a mixed traffic stream involving 30% Class 9 trucks with 150 lbs/hp. As with the freeway analysis, a two-step process was involved in developing a procedure to create equations that would estimate vm, DV, and b for a given situation. First, for each of the 520 scenarios, estimates of vm and DV were obtained through statistical analysis. Then, the result- ing estimates were placed in a database and curve-fitting techniques were used to develop esti- mates of the three parameters. Predicting vm proved to be most challenging. The following logic proved to be useful: 0 10 20 30 40 50 60 0 200 400 600 800 1000 1200 1400 1600 1800 Sp ee d (m ph ) Flow Rate per Lane (veh/hr) Speed-Flow Relationship Auto Only Scenario Autos Trucks Auto Model Exhibit 69. Example automobile speed relationship for arterials.

Prediction of Arterial truck Speeds 107 If (%Grade>=1%), then vm = 920 – 0.3475 * %Trk – (2.5 + 0.008 * %Trk * %Grade) * Wt/Hp + 20 * TrkType, else vm = 800. Clearly, this is not the result of a formal regression analysis; rather, it is derived from care- ful examination of the trends exhibited in vm in response to changes in the other variables involved. From a review of the results it was clear that for grades below 1% (the first seven conditions), the vm value is high if the truck percentage is 30% or less and low if it is greater. It was also clear that for grades of 1% or greater, the vm value is highest when the percent trucks is lowest, it declines as the percent trucks increases, and it falls sharply in response to increases in the weight- to-horsepower ratio. This logic is reflected in the “if-then” logic presented above, including the equation that predicts vm for grades of 1% and greater. A more detailed examination of the trends for grades of 1% or more showed that the patterns clearly matched, especially for steeper grades. For the less-steep grades, it was also clear that the stochasticity in the simulation process makes the trends less deterministic in appearance. Thus, the strength in the model presented in the “if-then” logic is that it converges to the simulation results observed as the grades increase in severity, which is a very good property for the model to have. The estimation of values for DV and b was far more straightforward. In this instance, a slight variation of Equations 28–38 was used: 1 Equation 57S S S S e e a to ao to v v v v m m ( )= + − +       − θ − θ where q reflects the combined effects of DV and b. The result was: 0.2551 0.7909 Equation 58 30.3 2v Rmθ = =p The R2 value is 0.7909 and the coefficient for vm is statistically significant given the t statistic of 30.3 (shown in Equation 58, just below the relevant parameter). As with the freeways, the predictive procedure (for both PCE and truck speed) is simple and straightforward. It appears to always correctly predict not only the density that will arise in a given situation, but also both the truck and car speeds. The procedure works whether the truck flows are of a single type or mixed. It is known to work for grades from -6% to +6% and for truck percentages up to 50%. 9.6 Arterial Case Study To illustrate application of the new arterial procedures, a real world case study was con- ducted. The setting is a 1.3-mile section of Hoosick Street in Troy, NY. The street is shown in Exhibit 70. The study section runs from 8th Street on the west to Lake Avenue on the east. (As an aside, this arterial lies immediately east of and connects directly to the Route 7 freeway section that was used as the freeway case study.) The study section has seven signalized intersections. They are, from west to east: 8th Street, 10th Street, Troy Plaza, 15th Street, Burdett Avenue, 25th Street, and Lake Avenue.

108 Incorporating Truck Analysis into the Highway Capacity Manual Hoosick Street has two through-lanes in each direction plus left-turn bays. The side streets have one through-lane on each approach plus left-turn bays. The percentage of trucks is about 6%. Periodically, NYSDOT does short counts including truck classifications. Otherwise, the street is not instrumented. The uphill direction is eastbound as the street leaves the Hudson River Valley. At 8th Street, the grade is about 6%. It then increases for about a block to 7%, declines again to about 5.25% and stays at that value until just short of the end when it decreases to 1.6%. The vertical profile is shown in Exhibit 71, including the locations of the intersections. The intersection at 8th Street is a bit complicated. On the western side, there are two entry points and two exit points on the arterial. One is the I-787 Bridge across the Hudson River. The other is the continuation of Hoosick Street (between the bridge ramps). Most of the traffic goes to and from the I-787 Bridge. A much smaller flow goes to and from Hoosick Street (from below the bridge). The entering traffic from the I-787 Bridge is traveling at 40–50 mph as it approaches the signal at 8th Street. The traffic from Hoosick Street is going much slower, having just passed through the signal at 6th Street. To conduct the analysis, a VISSIM simulation model of the arterial was created. It is based on data provided by NYSDOT. The model is capable of simulating fully actuated, semi-actuated/ coordinated, and pre-timed operation. As with the freeway case study, the actual a.m. and Exhibit 70. Case study arterial. 1 3 2 4 5 El ev at io n (fe et) Distance (miles) West (Direction) East 0 0.5 1.1 7 8t h St re et 10 th St re et Tr oy Pl az a 15 th St re et B ur de tt 25 th St re et La ke Av en ue 0.1 0.2 0.3 0.4 1.00.6 0.7 0.8 0.9 1.31.20 100 200 300 400 6 Exhibit 71. Vertical profile of the Hoosick Street case study network.

Prediction of Arterial truck Speeds 109 p.m. peaks have been studied as well as hypothetical situations involving near-capacity flows in both directions and 15% and 30% trucks. The predictions of the revised HCM procedure were checked against the performance predic- tions provided by the VISSIM simulation model. Both directions were studied in detail, but the eastbound direction is reported here because it has the uphill grades. It is very important to recognize that the HCM analyses were conducted with the signals disabled—that is, all of the signals were in constant green in both directions at the same time and for all controlled lefts. Undoubtedly, this will seem very strange to the reader, but it is important to realize that the HCM values being checked in this test are the running speeds of the vehicles given the grade, geometry, vehicle interactions, and so forth—not the overall travel times or speeds as affected by the signal timings. It turns out this can be done in a simulation model by disabling the signals. Then, as long as the model is not given any instructions about yield con- ditions, the traffic streams will pass by and through one another without interaction. Hence, interestingly, running times can be simulated directly. Exhibit 72 shows the running speeds observed by segment as well as the predictions from the proposed procedure. As can be seen, the new procedure closely predicts the observed values. The correspondence is always close between the simulated (observed) and predicted (New HCM model) running speeds. 9.7 Truck Speeds through Roundabouts The objective of this task was to develop methods to estimate truck speeds through round- abouts (as distinct from either the passenger car speeds or mixed traffic speeds). To conduct these analyses, the team had access to all of the videotapes and datasets prepared as part of the research for NCHRP Report 572: Roundabouts in the United States (Rodegerdts et al., 2007). These data encompass information related to many single lane roundabouts nationwide and a few multilane roundabouts. Excel workbooks were created for every approach that was studied. More- over, one tab in each workbook shows the sequence of vehicle events that took place including a field that indicates whether each vehicle was an automobile, motorcycle, small truck, or large truck. A small truck was considered to be a single-unit truck, a single-unit camper, or a delivery van. A large truck was a multiple-unit truck such as a tractor-trailer, a car or truck towing a boat or trailer, or a bus. 9.7.1 Analyses Three analyses have been conducted regarding trucks at roundabouts. The first examined move-up times to estimate truck PCEs. The second looked at the entry capacity equation to see how the percentage of trucks affected its calibration. The third examined the impact of facility geometry on truck speed. These studies have been conducted on the basis of data from two of the roundabouts studied in NCHRP Report 572: the single-lane roundabout in Lothian, MD, and the double-lane roundabout in Brattleboro, VT. These two were examined most intensely because they had the highest truck flow rates. Because the video recording technology in the NCHRP Report 572 project made use of a special omni-directional camera, it is possible to trace individual vehicles through the round- abouts. To learn more about truck speeds through roundabouts, these omni-directional recordings were reviewed for two roundabouts, one single lane and one double lane, to collect information about individual truck trajectories. Automobile trajectories were also collected for comparison.

110 incorporating truck Analysis into the Highway capacity Manual 8th-10th 10th-TP TP-15th 15th-BD BD-25th 25th-Lake Number of Lanes 3 3 2 2 2 2 Length (ft) 531 797 521 1312 1430 1725 Grade (%) 6.89% 5.55% 5.32% 5.25% 4.90% 4.26% Flow Rates (vph) Trucks 75.8 75.8 75.6 75.5 75.5 75.5 Autos 786.6 721.8 662.3 603.0 729.9 611.1 Total 862.4 797.5 737.9 678.5 805.4 686.6 Flow Rate (vphpl) 287.5 265.8 368.9 339.3 402.7 343.3 Running Speed Autos (mph) 44.1 44.1 44.4 44.4 45.0 44.7 Trucks (mph) 42.0 41.0 42.0 42.3 43.6 43.5 Average (mph) 43.9 43.9 44.2 44.2 44.9 44.6 New HCM Model Autos (mph) 43.6 42.4 42.6 43.1 43.4 43.4 Trucks (mph) 43.2 41.4 41.5 42.4 43.3 43.4 Average (mph) 43.5 42.3 42.5 43.1 43.4 43.4 Flow Rates (vph) Trucks 176.6 176.5 176.4 176.1 175.9 175.9 Autos 714.3 661.1 608.9 559.8 685.3 582.1 Total 890.9 837.6 785.3 735.9 861.1 758.0 Flow Rate (vphpl) 297.0 279.2 392.6 367.9 430.6 379.0 Running Speed Autos (mph) Trucks (mph) Average (mph) New HCM Model Autos (mph) Trucks (mph) Average (mph) 44.0 43.9 44.0 44.1 44.7 41.5 40.5 41.4 41.8 43.1 43.6 43.3 43.6 43.7 44.5 43.5 42 42.2 42.8 43.2 43.2 41.3 41.2 42.2 43.1 43.5 41.9 42.1 42.7 43.2 44.6 43.2 44.4 43.3 43.3 43.3 Flow Rates (vph) Trucks 364.8 364.6 364.5 364.0 363.4 363.1 Autos 583.1 546.8 509.9 472.4 599.8 535.8 Total 947.9 911.4 874.4 836.4 963.1 898.9 Flow Rate (vphpl) 316.0 303.8 437.2 418.2 481.6 449.4 Running Speed Autos (mph) 43.4 42.8 43.4 43.4 44.2 43.7 Trucks (mph) 40.9 39.7 40.8 41.2 42.5 42.4 Average (mph) 42.6 41.8 42.6 42.7 43.6 43.3 New HCM Model Autos (mph) 43.2 41.2 41.3 42.0 42.8 43.0 Trucks (mph) 43.0 40.9 40.8 41.6 42.7 43.0 Average (mph) 43.1 41.1 41.1 41.9 42.7 43.0 Data Item Eastbound Segments PM Peak, 6.1% Trucks PM Peak, 15% Trucks PM Peak, 30% Trucks Exhibit 72. The new HCm procedure versus the simulation model results for the Hoosick Street case study.

Prediction of Arterial truck Speeds 111 A set of monitoring points was superimposed on the omni-directional videotape images (Exhibit 73). There are two on each approach so that move-up times could be observed. Simi- larly, two data collection points lie on each exit. Finally, eight data collection points are on the circulating roadway: four in-between the legs of the roundabout and four at the midpoint of each splitter island. Individual vehicles were followed as they passed through the roundabouts, and timestamps were recorded when the monitoring points were passed. Hence, for example, a vehicle entering at 2 and exiting at 9 would have nine timestamps: at 2, 2, R, T, Y, U, I, 9, and 9. Distances were measured between the data collection points so that speeds (and travel rates) could be computed between all pairwise combinations of monitoring points (e.g., 2R, R7, RT, TY, Y8, YU . . .). It is important to note that the videotapes were created during time periods when the round- about was at or near capacity. The NCHRP Report 572 data collection team aimed to collect data when there was a standing queue on one or more approaches. Hence, most vehicles will be entering the roundabout from a speed near zero. A fundamental relationship related to facility design gives a sense of how the facility design relates to the speeds trucks “should” be able to travel: Equation 592v gR e f( )= + where v2 = the square of the vehicle speed, g = the gravitational acceleration rate, R = the radius of the trajectory followed by vehicles through the roundabout, e = the super-elevation (typically negative for drainage), and f = the friction coefficient. 6 6 2 2 7 7 3 3 1 1 5 5 4 4 8 8 I Q W E R T Y U Exhibit 73. Data collection points in the roundabouts.

112 incorporating truck Analysis into the Highway capacity Manual In FHWA’s roundabout design guide, Roundabouts: An Informational Guide, this relationship is employed to develop guidelines for roundabout diameters and vehicle speeds. Exhibit 74 shows the numerical guidance presented in Exhibit 6-14 of the guide (FHWA, 2000). The speed for R1 pertains to non-stop vehicles entering the roundabout; the speed for R4 is for vehicles navigating the circulating roadway. The speed for R4 is the speed that corresponds to the analyses presented here. One of the two roundabouts studied for speeds was the single lane roundabout in Lothian, MD. It has an inscribed circle diameter of 120 ft., which means the speed of vehicles on the cir- culating roadway should be about 15 mph. Exhibit 75 shows the distributions of entering and circulating speeds observed for the Lothian roundabout. Entering speeds were based on first entry movements (e.g., 2R) and right-hand exit movements (e.g., R7) while circulating speeds were based on movements between subsequent monitoring locations in the roundabout (e.g., YU, UI, IQ). The right-hand exit movements were more similar to the entering movements than to the circulating movements. The speeds are clearly different for trucks than for automobiles. In the case of entry speeds, the 80th percentile for large trucks is about 3 mph, while it is about 17 mph for automobiles. Approximate R4 Value – Radius for Conflicting Left-Turn Movement Maximum R1 Value – The Entry Path Radius Inscribed Circle Diameter (ft) Radius (ft) Speed (mph) Radius (ft) Speed (mph) Single-Lane Roundabout 100 35 13 165 25 115 45 14 185 26 130 55 15 205 27 150 65 15 225 28 Double-Lane Roundabout 150 50 15 205 27 165 60 16 225 28 180 65 16 225 28 200 75 17 250 29 215 85 18 275 30 230 90 18 275 30 Adapted from Exhibit 6-14, Roundabouts: An Informational Guide (FHWA, 2000). Exhibit 74. Recommended diameter, radius, and speed relationships for roundabouts. 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 0 5 10 15 20 25 30 Cu m ul ati ve P er ce nt ag e Vehicle Speed (mph) Entering and Circulating Speeds - Lothian Entry-Auto Entry-Light Entry-Large Circ-Auto Circ-Light Circ-Large Exhibit 75. Distributions for entering and circulating speeds— Lothian single-lane roundabout.

Prediction of Arterial truck Speeds 113 The differences in speeds on the circulating roadway are not quite so dramatic. Here the 80th percentile for large trucks is 15 mph, while for automobiles it is 20 mph. At about the 50th percentile, the speeds match the value shown in the roundabout guide, which is what should have been found. The conclusion to draw is that the speeds of trucks are clearly different from the automobiles. On entering the facility, they are strikingly different. On the circulating roadway they are less different, but still not the same. Exhibit 76 shows the same information for the two-lane Brattleboro roundabout. Again, the entry speed distributions for trucks are significantly different from the automobiles—for example, the 60th-percentile speed for the large trucks is about 3 mph, while it is 12 mph for the automobiles. The circulating speeds are more similar. The 80th-percentile speed for trucks is 16 mph, while for automobiles it is 18 mph. It seems clear that these differences in speeds should be reflected in the HCM procedures, in terms of estimating delays for trucks as they pass through isolated roundabouts, and for trucks versus automobiles as they traverse roundabouts in arterials. The field data obtained from two roundabouts (one single, one double) where the truck flows were significant, suggest that truck speeds upon entry are significantly different from and slower than automobile speeds, but the circulating speeds are fairly similar although the truck speeds are clearly lower than the automobile speeds. 9.8 Arterial Truck Travel Time Reliability Existing truck travel time reliability for one or more selected segments of an arterial street can be obtained from the National Performance Management Research Dataset (NPMRDS) for the National Highway System (NHS) (FHWA, 2013, June 26). 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 0 5 10 15 20 25 30 Cu m ul ati ve P er ce nt ag e Speed (mph) Entering and Circulating Speeds - Brattleboro Entry-Auto Entry-Light Entry-Large Circ-Auto Circ-Light Circ-Large Exhibit 76. Distributions for entering and circulating speeds—Brattleboro double-lane roundabout.

114 incorporating truck Analysis into the Highway capacity Manual Resources did not permit the development of a model for predicting truck travel time reli- ability. However, the SHRP2-L08 methodology (Kittelson and Vandehey, 2012) can be used to estimate mixed flow travel time reliability. Until such time as better methods become available, the SHRP2-L08 results might be used as a proxy for truck travel time reliability. 9.8.1 Data on Existing Truck Reliability—NPMRDS NPMRDS contains archived data on truck travel times by highway segment on the NHS, by 5-min-long time periods of the day. It is a vehicle-probe based data set. Separate travel times are reported for FHWA Vehicle Classes 7 and 8 (labeled “trucks” in the database), all other vehicle classes (labeled “passenger vehicles”), and all vehicles combined. The number of vehicles and the percent of trucks in the data are not reported. Historic data is available for non-Interstate highways on the NHS back to July 2013. A moderate amount of GIS database processing is required to make effective use of the data once downloaded. 9.8.2 Predicting Truck Reliability on Arterials As for freeways, the SHRP2-L08 methodology can be used to predict mixed flow travel time reliability for urban arterial streets. It is sensitive to recurring peak-period demands, day-to-day demand variability, the frequency and severity of bad weather, crash frequency, and the schedul- ing of work zones on the freeway facility. The methodology can be used to predict various TTIs, of which the 50th-percentile and the 95th-percentile TTIs are required. The median (50th-) and 95th-percentile TTIs predicted using the SHRP2-L08 method are entered into the following two equations, which are solved for the values of the parameters k and c: 50% 2 1 Equation 60 1 TTI kc( ) ( )= − 95% 20 1 Equation 61 1 TTI kc( ) ( )= − The agency’s target TTI threshold for on-time arrival (3.33 is recommended for arterials) is then entered into the following Burr distribution equation (along with the previously deter- mined values of k and c) to obtain the probability P of on-time arrival for mixed flow traffic on the facility: 1 1 Equation 623.33P TTITTI C k( )= − +( )= − Until a better method becomes available, the mixed flow traffic reliability (probability of on-time arrival) is assumed to be the same as for trucks. If the analyst wishes a more precise forecast, the analyst might use the SHRP2-L08 method to predict existing reliability conditions and compare that estimate with the value obtained from the NPMRDS. The ratio of the observed truck value to the estimated mixed flow value might then be used to adjust the forecasted mixed flow reliability to obtain a calibrated prediction of truck travel time reliability. However, this approach has not been tested or validated in this research.