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.

59 This section describes the development of a truck LOS model framework that is sensitive to the facility performance measures that are most important from the perspective of shippers, receivers and carriers, tempered with the need to provide actionable information to public agen- cies for the improvement of truck movements on the facility. The truck LOS model development process started with the Cambridge Systematics New York/ New Jersey Cross Harbor Freight Movement Model (Cambridge Systematics, in process). This original model, designed to analyze the entire truck trip, was adapted to the analysis of single free- way or street facilities, with the more limited goods-movement information available at that level. The model was then streamlined with the use of several default values for critical information on goods-movement characteristics in order to facilitate its application by public agencies. Finally, an alternative form was developed for the model to make its operation and results more intuitive for use by public agencies communicating their results to decisionmakers and the general public. The models were vetted with freight experts from various public agencies in two workshops to identify the model best suited for their use in evaluating highway freight improvement projects and for goods-movement planning. 6.1 Establishing a Facility’s Freight Importance Class It is desirable to be able to set different LOS thresholds according to the importance of the facility to the economic vitality of the region. Thus, an inter-regional freeway or highway critical for importing and exporting goods from the region is a vital link in the region’s freight system. Freeways and arterial streets serving a major regional intermodal terminal such as a water port, an airport, or a railroad intermodal facility may also be vital links in the region’s freight system. Roads serving a major factory complex may also be vital links for the region’s freight system. FHWA, working in conjunction with the states, has established the National Highway System (NHS) (FHWA, 2013), which consists of roadways important to the nation’s economy, defense, and mobility. The NHS consists of Interstate highways, other principal arterials, the Strategic Highway Network, major strategic highway network connectors (connectors to major military installations), and intermodal connectors. MAP-21, the Moving Ahead for Progress in the 21st Century Act (P.L. 112-141), “requires [the Department of Transportation] to establish a national freight network to assist States in strategically directing resources toward improved movement of freight on highways. The national freight network will consist of three components: • A primary freight network (PFN), • Any portions of the Interstate System not designated as part of the PFN, and • Critical rural freight corridors.” (Federal Register, 2013) S e c t i o n 6 Truck Level-of-Service Framework

60 incorporating truck Analysis into the Highway capacity Manual Since designation of the National Freight Network is not expected until after preparation of this report, a tentative three-class system (shown in Exhibit 28) employing some of the general criteria outlined in MAP-21 is recommended for classifying highway facilities by their relative importance to the region’s and national economy. The different classes of facilities are assigned different percentage thresholds for a given letter grade LOS (see Exhibit 29). The thresholds are higher for higher-class facilities and lower for the lower-class facilities. 6.2 Derivation of LOS Model 1 The Cambridge Systematics Port Authority model (described in Section 3.5) needed to be simplified and adapted for application within the single highway facility analysis environment typical of HCM analyses. The unique goods movement inputs of the Cambridge model needed to be replaced with regional defaults to enable application of the model using the data resources typically available for an HCM analysis. The derivation of Model 1 from the Cambridge model proceeds through several steps. Facility Class Description Suggested Criteria Examples I Highway facility critical to the inter-regional or within region movement of goods. • Facility carries a high volume of goods by truck (by tonnage or by value). • Trucks may account for a high volume or percentage of AADT* compared with other facilities in the region. Interstate freeway, inter- regional rural principal arterial. II Highway facility of secondary importance to goods movement within or between regions. • Facility carries lesser volumes of goods (by tonnage or value). • Trucks account for a lesser volume or percentage of AADT. Urban principal arterial, connector to major intermodal facilities (maritime port, intermodal rail terminal, airports). III Highway facility of tertiary importance to goods movement within or between regions. • Connectors to significant single origins/destinations of goods, such as major manufacturing facilities, sources of raw materials (mines, oil, etc.). • Connectors to truck service facilities and terminals. Access roads to mines, energy production facilities, factories, truck stops, truck terminals. *AADT = annual average daily traffic. Exhibit 28. Facility freight classification system. LOS Class I Primary Freight Facility Class II Secondary Facility Class III Tertiary Facility A >=90% >=85% >=80% B >=80% >=75% >=70% C >=70% >=65% >=60% D >=60% >=55% >=50% E >=50% >=45% >=40% F <50% <45% <40% *Entries are the percentage of achievement of ideal facility operating conditions for trucks. Exhibit 29. LOS Model 1 service measures and thresholds for goods movement LOS.*

truck Level-of-Service Framework 61 6.2.1 Translation of Utility to LOS The utility index output by the Cambridge model must be translated into an equivalent letter grade LOS. This is done by comparing the computed utility for actual conditions on the facility with the estimated utilities for the theoretically best- and worst-case conditions on the facility. The “closeness” of actual performance to ideal, best performance is used to assign the letter grade LOS. Conditions close to ideal, best case are assigned a letter grade of “A.” Conditions far worse are assigned a letter grade of “F.” Both the best- and the worst-case conditions for the facility would be set based on local oper- ating agency preferences. The best case would presumably represent free-flow conditions with highly reliable travel times with modest to no tolls on the facility, but this is up to the agency. The worst case would represent severe congestion, highly unreliable travel times, and a toll condition specified by the operating agency. The Truck Level of Service (TLOS) Index is then computed as follows: Equation 9TLOS Index U actual U worst U best U worst ( ) ( ) ( )( ) ( )= − − where TLOS(Index) = ratio of actual utility to utility for ideal conditions (constant free flow speed, no tolls) and U(x) = utility of trip on facility under conditions “x.” Exhibit 29 shows the recommended thresholds by LOS grade by facility class (facility classes were described in Section 6.1). 6.2.2 Translation of Facility Changes to Shipment Changes The Cambridge model is designed to be applied to the entire truck trip, while HCM analyses apply to individual facilities. In order to use the Cambridge model in an HCM analysis, it is necessary to translate facility performance changes into their equivalent effects on the entire truck trip. Translating Facility Travel Time Effects into Shipment Travel Time Changes The average shipment travel time by commodity type is obtained from the table of suggested defaults provided in Exhibit 30. The average shipment time for each commodity type was esti- mated by applying assumed typical freeway and arterial free-flow speeds to the average shipment distances obtained from FHWA’s Freight Analysis Framework (FAF). The effect of actual facility travel times for the selected analysis period on the average ship- ment times by commodity type are estimated by adding the difference between the actual facility travel time and the free-flow travel time for the facility: , , , , Equation 10AST c s a r AST c s b r T s a T s b( ) ( )( ) ( ) ( ) ( )= = = + = − = where AST(c,s,r) = Average shipment time for commodity “c,” scenario “s,” and region “r,” where commodity types and regions are as shown in Exhibit 30 and scenarios are best case (s = b), actual case (s = a), or worst case (s = w) and T(s) = End-to-end facility travel time under scenario “s.”

62 incorporating truck Analysis into the Highway capacity Manual The same approach is used to estimate the effect of worse-case conditions on average ship- ment times. The difference between the agency selected worst-case congested travel time and the free-flow travel time for the facility is added to the average shipment time obtained from FHWA’s FAF. Translating Facility Reliability into Probability of On-Time Arrival The Cambridge model uses the probability of on-time arrival to estimate utility; however, for typical HCM analyses, only facility reliability will be available. Facility reliability may be expressed in many forms such as the 85th-percentile travel time index (TTI) or the probability of automobile LOS F operation. The definition of on-time arrival is obviously specific for each shipment. For the purposes of highway planning, an agency may select an average value of on-time arrival that reflects agency goals such as the difference between free-flow speeds and congested speeds on the facility. Fol- lowing such a policy, an agency might define on-time arrival for a freeway as average truck travel times on the facility that are no more than 33% greater than free-flow travel times. Following this policy the 85th-percentile and 90th-percentile TTIs for the facility can be translated into probability of on-time arrival as follows: if TTI 85th% 1.33, - 85%, if TTI 90th% 1.33, - 90%, Else probability of on-time arrival is between 85% and 90% Equation 11 then probability of on time arrival then probability of on time arrival ( ) ( ) > < < > where TTI (P) = ratio of the “P” percentile highest travel time on facility to the free-flow travel time. Following the recommended reliability LOS thresholds suggested by SHRP2-L08 (Kittelson and Vandehey, 2012) the threshold for on-time arrival is set at the conventional HCM LOS E/F threshold for the facility. If the facility is operating at LOS F then the truck is assumed to not arrive on time. For freeways, LOS F will usually occur when TTI exceeds 1.33. For arterials, LOS F will usually occur when the TTI exceeds 3.33 (the midblock free-flow speed divided by the LOS F travel speed). Average Shipment Time by Commodity Type (hr) Pacific Rocky Mountains Southwest Midwest Northeast Southeast Alaska Hawaii Agriculture 5.1 4.4 4.1 2.9 4.0 3.8 4.7 0.6 Metal and Mining 2.9 2.9 3.2 2.9 2.6 2.8 5.4 0.6 Construction 1.4 1.7 1.4 1.5 1.1 1.4 3.9 0.6 Chemical 3.5 4.2 3.3 3.0 2.5 3.0 4.9 0.6 Wood and Paper 4.5 4.8 4.7 4.5 3.7 3.0 6.0 0.6 Electronics 11.5 9.1 10.6 8.6 8.8 8.3 21.7 0.5 Transportation and Utility 9.1 9.0 7.2 5.8 6.3 5.8 17.9 0.6 Wholesale and Retail 5.3 5.1 4.4 3.9 3.5 4.1 13.1 0.6 Regional Average 3.7 3.6 3.3 2.9 2.8 2.8 5.2 0.6 *Data extracted from FHWA Freight Analysis Framework; values computed by Kittelson and Associates. See Appendix A for derivation of average shipment times. Exhibit 30. Default average shipment times by commodity type by region.*

truck Level-of-Service Framework 63 Estimation of Effect of Tolls on Average Shipment Cost The Cambridge model requires average shipment cost by commodity type. Changes in facility tolls would change that cost upwards or downwards. A table of default average shipment costs (shown in Exhibit 31) has been derived from FHWA FAF data. Any changes in facility tolls applicable to trucks are added to the average shipment costs for the commodities: , , , , Equation 12ASC c s a r ASC c s b r Tolls s a Tolls s b( ) ( ) ( ) ( )= = = + = − = where ASC(c,s,r) = Average shipment cost for commodity “c,” scenario “s,” and region “r,” where commodity types and regions are as shown in Exhibit 30 and scenarios are best case (s = b), actual case (s = a), or worst case (s = w) (in dollars) and Toll(s) = End-to-end facility toll for trucks under scenario “s” (in dollars). The average shipment costs are related to the average shipment distances given in Exhibit 32. Average Shipment Cost ($) Pacific Rocky Mountains Southwest Midwest Northeast Southeast Alaska Hawaii Agriculture $1,200 $1,000 $1,000 $700 $900 $900 $1,200 $100 Metal and Mining $700 $700 $800 $700 $600 $700 $1,300 $100 Construction $300 $400 $300 $300 $300 $300 $1,000 $100 Chemical $800 $1,000 $800 $700 $600 $700 $1,200 $100 Wood and Paper $1,000 $1,100 $1,100 $1,000 $900 $700 $1,500 $100 Electronics $2,700 $2,100 $2,500 $2,000 $2,100 $2,000 $5,400 $100 Transportation and Utility $2,100 $2,100 $1,700 $1,400 $1,500 $1,400 $4,400 $100 Wholesale and Retail $1,200 $1,200 $1,000 $900 $800 $1,000 $3,200 $100 Regional Average $800 $800 $800 $700 $700 $700 $1,300 $100 *Data extracted from FHWA Freight Analysis Framework; values computed by Kittelson and Associates. See Appendix A for derivation of average shipment times. Exhibit 31. Average shipment costs by commodity type and region.* Average Shipment Distance (miles) Pacific Rocky Mountains Southwest Midwest Northeast Southeast Alaska Hawaii Agriculture 330 280 260 180 260 240 250 20 Metal and Mining 180 180 210 180 160 180 290 20 Construction 90 110 90 90 70 90 210 30 Chemical 220 270 210 190 160 190 260 30 Wood and Paper 280 300 300 280 230 190 320 20 Electronics 730 570 680 550 560 530 1170 20 Transportation and Utility 580 570 450 370 400 370 970 20 Wholesale and Retail 330 320 280 240 220 260 710 20 Regional Average 230 230 210 180 180 180 280 30 *Data extracted from FHWA Freight Analysis Framework; values computed by Kittelson and Associates. See Appendix A for derivation of average shipment times. Exhibit 32. Average truck shipment distance.*

64 incorporating truck Analysis into the Highway capacity Manual Computation of Composite Utility The Cambridge model computes utility for individual commodity types. A composite utility is computed for trucks on the facility by weighting the utility for each specific commodity type by its proportion of total truck commodity flows (in tons) in the region. Default proportions by commodity type (see Exhibit 33) were obtained from the FHWA FAF for the major regions of the United States. 6.3 Derivation of Model 2 Based on concerns regarding the sensitivities, computational complexities, and data require- ments of Model 1, various steps were taken to streamline the model and improve its usefulness for highway planning. 6.3.1 Graduated Effects of Reliability It was noted in the tests of Model 1 that reliability (specifically on-time arrival) was insensi- tive to changes in reliability when the probabilities of on-time arrival dropped below 85% or exceeded 90%. The model consequently showed no incremental benefits of reliability improve- ments until the 85% tipping point was reached. To provide for LOS sensitivity over the full range of possible probabilities of on-time arrival, a straight line function was created to approximate and replace the three-value on-time arrival vari- able in the original utility model. An equation with a slope of +5 and an intercept of –5 provided a reasonable fit to the original three-value OTA variable at the 85% probability of OTA without going into the positive range for probabilities of on-time arrival exceeding 95% (see Exhibit 34). The equation for estimating the contribution of reliability to utility is as follows: 5.0 5.0 Equation 13OTA POTA= ∗ − Percent Ton-Miles by Commodity Type Pacific Rocky Mountains South- West Midwest North- East South- East Alaska Hawaii Agriculture 19% 26% 16% 33% 15% 14% 13% 8% Metal and Mining 35% 35% 36% 30% 36% 32% 31% 44% Construction 16% 13% 19% 15% 16% 21% 26% 31% Chemical 10% 10% 16% 9% 15% 11% 19% 7% Wood and Paper 10% 7% 5% 5% 8% 14% 6% 4% Electronics 1% 1% 1% 1% 1% 1% 0% 0% Transportation and Utility 2% 1% 1% 2% 1% 1% 1% 1% Wholesale and Retail 7% 6% 6% 5% 8% 6% 4% 5% Total 100% 100% 100% 100% 100% 100% 100% 100% *Data extracted from FHWA Freight Analysis Framework; values computed by Kittelson and Associates. See Appendix A for derivation of average shipment times. Exhibit 33. Default table of percent of truck movements by commodity type by region.* Probability of On-Time Arrival Original OTA value Straight Line Approximation 0% to 85% –0.758 –5.000 to –0.750 85% to 90% –0.275 –0.750 to –0.500 90% to 100% 0.000 –0.500 to 0.000 Exhibit 34. Comparison of original OTA values and straight line approximation.

Truck Level-of-Service Framework 65 where OTA = on-time arrival contribution to utility equation (utils) and POTA = probability of on-time arrival expressed as a proportion (unitless). 6.3.2 Estimation of OTA Probabilities from Travel Time Indices If the cumulative distribution of TTIs for the facility is available, it is a simple matter for the analyst to read the probability of on-time arrival for any selected on-time arrival threshold—for example, the threshold might be defined as 1.33 times the free-flow travel time (see Exhibit 35). If only the median (50%) and 95th-percentile TTIs are available to the analyst, then the prob- ability of on-time arrival for a selected target TTI (e.g., 1.10) can be estimated using a fitted Burr Distribution (Burr, 1942): 1 1 Equation 14P TTI TTI c k( )( ) = − + − where P(TTI) = cumulative probability of TTI; TTI = desired target travel time index; and c, –k = distribution parameters, both greater than zero. Solving for the value of TTI that represents a certain cumulative percentile of the distribution (Taylor and Susilawati, 2012): 1 1 Equation 15 1 TTI P P kc( ) ( )= − −− where TTI(P) = percentile (P) of TTI. Exhibit 35. Probability of on-time arrival from cumulative distribution of TTIs.

66 incorporating truck Analysis into the Highway capacity Manual Thus, the median (50th-) and 95th-percentile TTIs are 50% 2 1 Equation 16 1 TTI kc( ) ( )= −− 95% 20 1 Equation 17 1 TTI kc( ) ( )= −− Equations 16 and 17 are solved for the two unknowns: k and c. One then uses these values of k and c plus the agency’s target on-time arrival threshold TTI to estimate the probability of on-time arrival. 6.3.3 Replacement of Commodity Types with Single Generic Type To increase the appeal of the procedure to transportation engineers and planners not famil- iar with goods movement and the collection of goods movement flows by commodity type, the potential of condensing the original nine commodity types in the Cambridge model into a single generic commodity type was evaluated. Examination of the sensitivity of the utilities for the nine different commodity types to time and cost noted the following: • All commodities are identically sensitive to reliability in the Cambridge model. • All commodities are identically sensitive to travel time for shipment times less than 10 hrs. • The sensitivities of the commodities to shipping cost vary significantly across types, but appear to fall into two main categories: highly price-sensitive goods (with cost coefficients ranging between –0.0086 and –0.0109) and less price-sensitive goods (with cost coefficients ranging between –0.0060 and –0.0068). Based on this examination, it was concluded that the nine original commodity types could be grouped into two classes: one class less sensitive to travel time and cost (transportation, utility, wholesale and retail goods), the other class more sensitive to travel time and cost (consisting of all other commodity types). The less cost-sensitive goods account for between 5% and 9% of all ton-miles shipped by trucks in the United States, so it was further concluded that the less cost- sensitive class could be dropped for the purposes of LOS estimation. It was also noted that the average truck shipment times and shipment costs by commodity type by region derived from the FHWA FAF are almost all under 10 hours and under $1000, so the time and cost splines in the Cambridge model were dropped. The original Cambridge model consequently can be streamlined to the following: 1 Equation 18U Reliability Cost Time( )= α ∗ − + β∗ + γ ∗ where U = perceived utility to shippers and carriers of a truck shipment of a single generic commod- ity type. All other variables (reliability, cost, time) and coefficients (alpha, beta, gamma) are as defined in Exhibit 36. 6.3.4 Replacement of Commodity Shipment Costs with Average If we replace the commodity-specific shipping costs with the average shipment cost of $750 for the Continental United States ($1,300 for Alaska and $100 for Hawaii), then the analyst no longer needs to acquire shipping cost information.

truck Level-of-Service Framework 67 6.3.5 Prorating Reliability Effects by Facility Length During testing of the streamlined model, it was noted that it significantly overestimated the value of reliability, incorrectly suggesting that shippers would be willing to pay tolls of $10 to $30 per mile for 90% probabilities of on-time arrival. Upon re-evaluation of the Cambridge model, it was noted that the reliability effect applied to the entire shipment distance rather than just the facility length. Consequently, the reliability effect within the Cambridge model incorporated a distance component. It was decided to prorate the reliability effect of the facility according to the percent of the total trip length accounted for by the facility. 6.4 Streamlined Utility Model (Model 2) The previous steps result in the following streamlined utility equation for Model 2: 5.00 1 0.01 0.32 1 Equation 19U L ASL POTA ASC Toll AST T TTIFF( ) ( ) ( )( )= ∗ ∗ − − ∗ + − ∗ + − where U = Perceived utility to shippers and carriers of a truck shipment; POTA = Probability of on-time arrival; L = Length of facility (miles); ASL = Average shipment length (200 miles Continental U.S., 280 miles Alaska, 30 miles Hawaii); ASC = Average shipment cost ($750 Continental U.S., $1,300 Alaska, $100 Hawaii); Toll = Toll paid by trucks to use facility ($); AST = Average shipment time (3 hr Continental U.S., 5 hr Alaska, 0.5 hr Hawaii); TFF = Free-flow travel time to travel length of facility (hr); TTI = Travel time index, ratio of mean truck speed for the given scenario to free-flow truck speed. The proposed LOS index and the LOS thresholds using the streamlined utility model are the same as those for Model 1 [The LOS index is (Actual–Worst)/(Best–Worst), which is used in Exhibit 29]. 6.5 Logistic Formulation with Truck Friendliness (Model 3) The streamlined model (Model 2) requires that the LOS model be applied three times for each facility: once to compute the utility for ideal conditions, once for worst-case conditions, and once for actual conditions during the selected study period (such as the weekday p.m. peak period). While the ideal condition is relatively easy to identify (100% on-time arrival, free-flow Variable Description Coefficient Reliability Probability of On-Time Arrival (0.00 – 1.00) α +5.00 Cost Shipment Cost ($) β –0.01 Time Average Shipment Time (hr) γ –0.32 Exhibit 36. Coefficients for two commodity utility model (Equation 18).

68 Incorporating Truck Analysis into the Highway Capacity Manual speeds, and no tolls), the identification of worst-case conditions is less obvious. The absolute worst case of 0% on-time arrival, zero speed, and infinite tolls is not numerically tractable, so the agency must select a “realistic” worst case with relatively little guidance as to what is a reasonable “worst” case. This can be an advantage for agencies desiring to calibrate the truck LOS results to local conditions; this can be a disadvantage for agencies not willing or able to calibrate the results to local conditions. An alternative approach was developed for estimating LOS from utility that avoids the need to explicitly identify a “worst-case condition.” It employs a logistic function that is self-limiting to values between 0% and 100% (see Exhibit 37). % 1 1 Equation 20TLOS e U x( )= + α ( )−β where %TLOS = The truck LOS index as a percentage of ideal conditions. α = Calibration parameter (determines value of %TLOS at x = 0). A value of 0.10 was selected heuristically so that model yields LOS A (>90% TLOS) under ideal reliable, free-flow, no toll conditions. b = Calibration parameter (determines rate at which function increases to 100%). A value of 200 was selected heuristically so that model yields LOS F (<50% TLOS) if any one of these conditions is present: POTA <50%, TTI > 3.25, or Toll > $1.10/mile. U(x) = Utility function. One of the desirable characteristics of a truck LOS model would be that it is not unduly influ- enced by the selection of facility length. Ideally, facilities of different lengths but with similar reliability and average operating speeds should get similar LOS ratings. In Model 2, the effects of facility length were cancelled out by incorporating the utility calcula- tions in both the numerator and the denominator of the truck LOS index. The logistic function used in Model 3 explicitly avoids the computation of multiple utilities for the same facility under Exhibit 37. The %TLOS logistic function for Model 3.

truck Level-of-Service Framework 69 best- and worst-case conditions; therefore, other methods must be explored to reduce the sen- sitivity of Model 3 to varying facility lengths. Revisiting the utility function in Model 2 (Equation 19) it can be noted that the free-flow travel time is facility length divided by the free-flow speed. The toll in that equation can be replaced with toll/mile times the facility length. These substitutions allow us to divide the equa- tion by the facility length without changing the relative effects of reliability, speed, and tolls on the utility. We obtain the following utility equation: 5 1 0.01 0.32 1 Equation 21U ASL POTA ASC L Toll mi AST L TTI FFS( )( )( ) ( )= ∗ − − ∗ + − ∗ + − where U = Perceived utility to shippers and carriers of truck shipments using the facility; POTA = Probability of on-time arrival; L = Length of facility (miles); ASL = Average shipment length (200 miles Continental U.S., 280 miles Alaska, 30 miles Hawaii); ASC = Average shipment cost ($750 Continental U.S., $1,300 Alaska, $100 Hawaii); Toll/mi = Toll rate paid by trucks to use facility ($/mile); AST = Average shipment time (3 hr Continental U.S., 5 hr Alaska, 0.5 hr Hawaii); FFS = Free-flow speed for trucks on facility (mph); and TTI = Travel time index, ratio of mean truck speed for the given scenario to free-flow truck speed. Grouping the facility length sensitive factors together, we get 5 1 0.01 0.32 1 0.01 0.32 Equation 22 U ASL POTA Toll mi TTI FFS ASC AST L( ) ( )( ) ( ) ( )= ∗ − − ∗ − ∗ − − ∗ + ∗ Sensitivity testing of Equation 22 for facilities between 1 and 10 miles in length found that the facility length dependent term (incorporating average shipping cost and average ship- ping time) caused the predicted LOS to be highly sensitive to the facility length. In addition, the average shipping cost (being in the hundreds of dollars range) tended to dominate the utility, significantly reducing its sensitivity to changes in reliability, and the travel times. The facility length dependent term of the utility equation was consequently dropped from further consideration. This last change also had the advantage of reducing the data requirements for the LOS model. Average shipping cost and average shipping time would no longer be required by the model. The result is the following utility equation for Model 3: 5 1 0.01 0.32 1 Equation 23U ASL POTA Toll mi TTI FFS( )( )( ) ( )= ∗ − − ∗ − ∗ − where U(x) = Perceived utility of truck shipments using the facility; POTA = Probability of on-time arrival; ASL = Average shipment length (200 miles Continental U.S., 280 miles Alaska, 30 miles Hawaii);

70 incorporating truck Analysis into the Highway capacity Manual Toll/mi = Toll rate paid by trucks to use facility ($/mile); FFS = Free-flow speed for trucks on facility (mph); and TTI = Travel time index, ratio of mean truck speed to free-flow truck speed. For example, for the Continental United States, the equation is 0.025 1 0.01 0.32 1 Equation 24U x POTA Toll mi TTI FFS( )( ) ( ) ( )= ∗ − − ∗ − ∗ − At one of the public agency workshops conducted to review the candidate models, several of the freight planning experts requested that the truck LOS model also include a “truck friendli- ness index” to indicate the degree to which substandard geometry, structures, or at-grade rail- road crossings hindered the ability of legal trucks with legal loads from using the facility without having to slow down for a railroad crossing or to maneuver through a geometric construction. The truck friendliness index (TFI) was consequently added to Model 3 to enable agencies to incorporate geometric limitations and at-grade rail crossing features of the facility into the truck LOS. The TFI is set at 1.00 for a facility designed and built to accommodate all federal, state, and local legal vehicles and loads with no at-grade railroad crossings. This value of 1.00 is depreciated at the agency’s discretion to account for vehicle length, width, height, turning radius, and load restrictions on truck usage of the facility. The model has been calibrated so that a TFI of 0.60 will yield LOS F in the model, regardless of the reliability, travel time, or toll on the facility. Adding the truck friendliness index to Equation 23, Model 3 can be re-specified as 1 1 1 Equation 25U x A POTA B TTI C Toll mi D TFI( )( ) ( ) ( ) ( )= ∗ − + ∗ − + ∗ + ∗ − where U(x) = Utility of facility for truck shipments. A, B, C = Calibration parameters from Model 2: A = 5/ASL, B = –0.32/FFS, C = –0.01 where ASL = average shipment length (200 miles Continental U.S., 280 miles Alaska, 30 miles Hawaii) and FFS = Free-flow speed (mph). D = Calibration parameter = 0.03 (determined heuristically so that LOS F if TFI is below 0.60). POTA = Probability of on-time arrival, with on-time being defined as a TTI of 1.33 or less for freeways, multilane highways, and two-lane highways; for urban streets, TTI ≤3.33. TTI = Travel time index for study period, ratio of free-flow speed to actual speed. Toll/mi = Truck toll charged per mile ($/mi). TFI = Truck friendliness index (1.00 = no constraints or obstacles to legal truck load and vehicle usage of facility, 0.00 = no trucks can use facility). 6.6 Reliability and Friendliness (Model 4) A fourth LOS model was created to address requests for a travel-time-reliability-only model while retaining the TFI desired by workshop participants. It was pointed out by the Philadelphia workshop participants that Model 3 and its ancestor models appeared to double count reliability by incorporating both probability of on-time arrival and the TTI, with both being measured against the same standard—the free-flow speed.

truck Level-of-Service Framework 71 This model is obtained by dropping tolls and the TTI portions of the utility equation from Model 3. The logistic function and parameters of Model 3 are retained. The utility function is reformulated to create Model 4 as follows: , 6 1 1 Equation 26U x A POTA D TFI( ) ( ) ( )= ∗ − + ∗ − where U(x,6) = Utility function for Model 4; A = Calibration parameter from Model 3 (A = 0.025); D = Calibration parameter = 0.03 (determined heuristically so that LOS F if TFI is below 0.60); POTA = Probability of on-time arrival, with on-time being defined as a TTI of 1.33 or less; and TFI = Truck friendliness index (1.00 = no constraints or obstacles to legal truck load and vehicle usage of facility, 0.00 = all trucks unable to use facility). 6.7 Results of Review by Public Agencies The four truck LOS models were reviewed by 28 public agency and university freight plan- ning experts in Philadelphia, Pennsylvania, and by 33 public agency freight planning experts in Sacramento, California. Public agency representatives came from state DOTs, MPOs, port authorities, and city and county planning agencies. The workshop participants had the following comments and conclusions about the four pro- posed LOS models: • The workshop participants agreed that truck LOS will be a useful tool to help make goods- movement projects related to trucks more competitive with other transportation improve- ment projects. It should help “getting trucks into the planning process”: – Truck LOS will be useful for MAP-21, for communicating to the general public, and for decisionmakers, – Truck LOS should be measurable, – Truck LOS should quantify different degrees of LOS F, and – Truck LOS should be calibratable to local conditions and perceptions. • The general preference of workshop participants was for Model 3 (Reliability, Speed, Cost, Friendliness) with the ability to calibrate it to local perceptions; they liked the ability to include tolls if they were an issue and felt it easy to exclude tolls if they were not an issue: “Better to have it and not use it, than to not have it.” • Model 4 (Reliability only plus Friendliness) was second favorite. • Models 1 and 2 were generally least desired; the primary objections appeared to be their greater apparent complexity. 6.8 Recommended Truck LOS Model The recommended truck LOS model is Model 3—combining speed, reliability, cost, and truck friendliness in a logistic model formulation (Equation 25).