National Academies Press: OpenBook

Statewide Travel Forecasting Models (2006)

Chapter: Chapter Three - Case Studies

« Previous: Chapter Two - Survey of Statewide Travel Forecasting Practice
Page 32
Suggested Citation:"Chapter Three - Case Studies." National Academies of Sciences, Engineering, and Medicine. 2006. Statewide Travel Forecasting Models. Washington, DC: The National Academies Press. doi: 10.17226/13958.
×
Page 32
Page 33
Suggested Citation:"Chapter Three - Case Studies." National Academies of Sciences, Engineering, and Medicine. 2006. Statewide Travel Forecasting Models. Washington, DC: The National Academies Press. doi: 10.17226/13958.
×
Page 33
Page 34
Suggested Citation:"Chapter Three - Case Studies." National Academies of Sciences, Engineering, and Medicine. 2006. Statewide Travel Forecasting Models. Washington, DC: The National Academies Press. doi: 10.17226/13958.
×
Page 34
Page 35
Suggested Citation:"Chapter Three - Case Studies." National Academies of Sciences, Engineering, and Medicine. 2006. Statewide Travel Forecasting Models. Washington, DC: The National Academies Press. doi: 10.17226/13958.
×
Page 35
Page 36
Suggested Citation:"Chapter Three - Case Studies." National Academies of Sciences, Engineering, and Medicine. 2006. Statewide Travel Forecasting Models. Washington, DC: The National Academies Press. doi: 10.17226/13958.
×
Page 36
Page 37
Suggested Citation:"Chapter Three - Case Studies." National Academies of Sciences, Engineering, and Medicine. 2006. Statewide Travel Forecasting Models. Washington, DC: The National Academies Press. doi: 10.17226/13958.
×
Page 37
Page 38
Suggested Citation:"Chapter Three - Case Studies." National Academies of Sciences, Engineering, and Medicine. 2006. Statewide Travel Forecasting Models. Washington, DC: The National Academies Press. doi: 10.17226/13958.
×
Page 38
Page 39
Suggested Citation:"Chapter Three - Case Studies." National Academies of Sciences, Engineering, and Medicine. 2006. Statewide Travel Forecasting Models. Washington, DC: The National Academies Press. doi: 10.17226/13958.
×
Page 39
Page 40
Suggested Citation:"Chapter Three - Case Studies." National Academies of Sciences, Engineering, and Medicine. 2006. Statewide Travel Forecasting Models. Washington, DC: The National Academies Press. doi: 10.17226/13958.
×
Page 40
Page 41
Suggested Citation:"Chapter Three - Case Studies." National Academies of Sciences, Engineering, and Medicine. 2006. Statewide Travel Forecasting Models. Washington, DC: The National Academies Press. doi: 10.17226/13958.
×
Page 41
Page 42
Suggested Citation:"Chapter Three - Case Studies." National Academies of Sciences, Engineering, and Medicine. 2006. Statewide Travel Forecasting Models. Washington, DC: The National Academies Press. doi: 10.17226/13958.
×
Page 42
Page 43
Suggested Citation:"Chapter Three - Case Studies." National Academies of Sciences, Engineering, and Medicine. 2006. Statewide Travel Forecasting Models. Washington, DC: The National Academies Press. doi: 10.17226/13958.
×
Page 43
Page 44
Suggested Citation:"Chapter Three - Case Studies." National Academies of Sciences, Engineering, and Medicine. 2006. Statewide Travel Forecasting Models. Washington, DC: The National Academies Press. doi: 10.17226/13958.
×
Page 44
Page 45
Suggested Citation:"Chapter Three - Case Studies." National Academies of Sciences, Engineering, and Medicine. 2006. Statewide Travel Forecasting Models. Washington, DC: The National Academies Press. doi: 10.17226/13958.
×
Page 45
Page 46
Suggested Citation:"Chapter Three - Case Studies." National Academies of Sciences, Engineering, and Medicine. 2006. Statewide Travel Forecasting Models. Washington, DC: The National Academies Press. doi: 10.17226/13958.
×
Page 46
Page 47
Suggested Citation:"Chapter Three - Case Studies." National Academies of Sciences, Engineering, and Medicine. 2006. Statewide Travel Forecasting Models. Washington, DC: The National Academies Press. doi: 10.17226/13958.
×
Page 47
Page 48
Suggested Citation:"Chapter Three - Case Studies." National Academies of Sciences, Engineering, and Medicine. 2006. Statewide Travel Forecasting Models. Washington, DC: The National Academies Press. doi: 10.17226/13958.
×
Page 48
Page 49
Suggested Citation:"Chapter Three - Case Studies." National Academies of Sciences, Engineering, and Medicine. 2006. Statewide Travel Forecasting Models. Washington, DC: The National Academies Press. doi: 10.17226/13958.
×
Page 49
Page 50
Suggested Citation:"Chapter Three - Case Studies." National Academies of Sciences, Engineering, and Medicine. 2006. Statewide Travel Forecasting Models. Washington, DC: The National Academies Press. doi: 10.17226/13958.
×
Page 50

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.

33 This chapter presents five case studies of statewide models. The case studies emphasize the differences between statewide and urban models. Two case studies, Kentucky and Indiana, focus on the passenger component and two other case studies, Virginia and Wisconsin, focus on the freight component. The fifth case study, Ohio, presents a comprehensive framework for dealing with both passenger travel and freight while ac- counting for changing locations of economic activity. These particular case study states were selected because their mod- els integrate a number of features described in chapter two, and they do not duplicate material found in the report from NCHRP Project 8-43. CASE STUDY 1: KENTUCKY PASSENGER COMPONENT Kentucky has had one of the longest involvements with statewide travel forecasting, starting in 1971. The Kentucky statewide model (KYSTM) has just recently been updated (a previous version was summarized in the Guidebook). Kentucky’s model has always had a goal of being efficient in its expenditure of resources, achieving a very useful model on a small budget by piggybacking on data obtained from existing sources. The only data collection specifically for the model was the purchase of additional NHTS samples. The model also has a truck component. Kentucky’s model is presented here as an example for states with modest fore- casting needs and states now considering models for the first time or that are in the process of reactivating a dormant model. The model is well integrated with agency decision making. “The main purposes of this model are to support highway plan- ning and investment decisions, to permit a consistent method- ology in project evaluations, and to allow testing alternative land use strategies.” Previous versions of the model have been used for corridor planning, project-level traffic forecasts, re- gional planning and weigh station location. A stakeholder meeting was held early in the process to ensure that the model development met agency needs. The revision took approxi- mately 2 years and cost about $370,000. The overall structure of the passenger component consists of three steps: trip generation, trip distribution, and traffic as- signment. Only highway vehicles are assigned to a network; therefore, a mode split step is not necessary. The application of the KYSTM’s steps is similar to traditional urban models, except that there are three additional trip purposes to account for long distance travel. The KYSTM highway network is very large, spanning all of the contiguous 48 states. The network, as shown in Figure 6, is focused on the state, with considerable detail extending approximately halfway into its neighboring states. This net- work has more than 77,000 links, more than 3,600 zones within the state, and more than 1,100 zones outside the state. Figures 7 and 8 illustrate the Kentucky zone system. The ex- pansive nature of the network has allowed analysis of diver- sion from out-of-state highways. Kentucky Statewide Passenger Component Summary State population: 4.1 million State area: 40,411 square miles Gross state product: $129 billion No. of internal zones: 3,644 External zone structure: Halo + BEA regions Internal zone structure: TAZs, aggregations of TAZs No. of external zones: 1,109 No. of links: 77,272 Passenger modes: Automobile Trip purposes: Home-based work Home-based nonwork Nonhome-based Long distance business Long distance—Recreation/vacation Long distance—Other Special generators: Military bases Trip productions: Rates per household based on MSA size, area type Trip attractions: Rates per level of activity Trip distribution: Gravity expression, Fratar Mode split: None Assignment: Static equilibrium with subzones Delay estimation: BPR curves Major data: NHTS, HPMS, ATS, vendors Time frame: Two years of development time Computation time: 1 h In-house staff: 1 FTE The zone structure was built for compatibility with other databases. It is readily seen that zones well outside of Kentucky are based on BEA Economic Areas. Zones within Kentucky were created as part of census TAZ-UP participation. A level CHAPTER THREE CASE STUDIES

34 of detail was selected so that the model could evaluate projects such as I-66 and I-69 and still be reasonably accurate within ur- ban areas. Zones were custom aggregated from census TAZs in dense urban areas; however, census TAZs were adopted in fringe urban areas. TAZs in 72 rural counties were built from census block groups or census places. The halo of zones around Kentucky was represented by 660 census tracts and 296 coun- ties. The model was built from secondary data. Data sources included workplace employment data from Woods & Poole, D&B, and Claritas. Trip rate information was derived from the 2001 NHTS. Because there were only 390 samples in the standard NHTS, Kentucky contracted for an additional 1,154 samples to provide better geographic coverage. Household socioeconomic data were mostly derived from the 2000 Census. The census also provided journey-to-work data. Information on long distance travel came from the 1995 ATS. Network data within Kentucky were obtained from the Kentucky Transportation Cabinet Highway Information Sys- tem. Network data outside Kentucky were developed from the NHPN for roadway geography and the HPMS for road- way characteristics. A single trip production rate was applied for any zone for any trip purpose. Trip production rates per household were separately calculated for different MSA sizes and different Claritas area types (rural, town, suburban, second city, and urban), which were included in the NHTS. Trip attraction equations were taken directly from NCHRP Report 187, hav- ing demographic variables of households, retail employees, and nonretail employees. The NHTS also yielded automobile occupancy rates, one for each short distance purpose. The short distance occupancy rates were similar to those seen within urban areas. Long distance automobile occupancy rates were derived from the ATS, as found on Table 4. Kentucky adopted a philosophy of using actual OD tables wherever possible. Thus, trip distribution was accomplished with a gravity expression only for home-based nonwork and nonhome-based purposes. Friction factors were chosen to match the trip length frequency distributions from the na- tional and Kentucky NHTS. For home-based work, a zone- to-zone production-to-attraction table from the 2000 Census journey-to-work data was Fratar factored. Fratar factored trip tables from the ATS were used for long distance trip pur- poses. Table 5 shows how each long distance trip purpose FIGURE 6 Kentucky’s highway network. (Source: Wilbur Smith Associates 2005a.) FIGURE 7 Kentucky’s zone system, in state. (Source: Kentucky response to Peer Exchange questionnaire 2004.) FIGURE 8 Kentucky’s zone system, out-of-state. (Source: Kentucky response to synthesis questionnaire February 2005.) Long Distance Trip Purpose National Sample Kentucky Sample Business 1.82 1.80 Tourist 3.23 3.31 Other 2.48 2.43 Source: 1995 ATS. TABLE 4 LONG DISTANCE AUTOMOBILE OCCUPANCY RATES

35 was handled. OD data from the ATS were available only for county-to-county trips. The OD table was expanded to TAZs by apportioning the trips by zonal households, zonal em- ployment, or both depending on the trip purpose and trip end. Traffic assignment was accomplished with a static user- equilibrium technique, with trucks preloaded to the network and weighted by passenger car equivalent factors that de- pended on terrain. Delay came from BPR curves as a func- tion of free flow speed and capacity. Free flow speeds were drawn from a table, and these speeds varied by functional classification, terrain type, number of lanes, and posted speed limit. Capacity per lane was determined from number of lanes, terrain, and functional class. Forecasts can be made for a full day or for shorter periods within a day. It is well known that large zones can lead to lumpy traffic assignments. Kentucky’s traffic assignment method divided TAZs into smaller subzones in order to improve the smooth- ness of the results. Subzones were built around highway routes within zones with the number of trips allocated to a subzone being in proportion to the mileage of each route. For some trip purposes the mileage was weighted such that routes of higher functional classes got more trips. Validation results were not available at the time of this writing. Sources for this case study were: Kentucky response to Peer Exchange questionnaire (2004), Kentucky response to Synthesis questionnaire (Feb. 2005), and Wilbur Smith Associates (2005a). CASE STUDY 2: INDIANA PASSENGER COMPONENT The Indiana Statewide Travel Demand Model (ISTDM) (Bernardin, Lochmueller & Associates, Inc. and Cambridge Systematics, Inc. 2004) was developed principally to assist corridor-level economic development studies. ISTDM was re- cently expanded from a more localized model for the 26-county I-69 study area in southwestern Indiana. The local network was broadened to include the entire state, the TAZ structure was re- fined, traffic signals were integrated into the network, and new procedures for estimating free-flow speed and roadway capac- ities were developed. The model structure for the passenger component was similar to that of a four-step UTP model. Indiana Statewide Passenger Component Summary State population: 6.2 million State area: 36,420 square miles Gross state product: $214 billion No. of zones: 4,720 External zone structure: Halo Internal zone structure: TAZs No. of links: 34,500 No. of signals: 3,900 Travel modes: Automobile, truck, intercity bus/rail Trip purposes: Home-based work Home-based nonwork Nonhome-based Long trip Trip productions: Rates per household based on household size, automobile ownership, and area type Trip attractions: Rates per employment categories and households Trip distribution: Gravity expression Mode split: Fixed shares for short trip purposes Multinomial logit for long trip purpose Assignment: Static equilibrium with feedback to distribution Delay estimation: BPR travel time volume curves Truck models: Commodity based for freight trucks; empirical for non-freight trucks Major data: Census, NHTS, CTPP, own surveys Time frame: Seven years of continuous improvement following 3 years of initial development Computation time: 2 h In-house staff: 0.5 FTE The ISTDM covers all 92 counties in Indiana and parts of adjacent states. A detailed network was developed for areas within the state of Indiana, including all state jurisdictional highways (more than 19,500 links) and additional local streets (more than 11,500 links). A less detailed network was used for areas outside Indiana, as shown in Figure 9. Data from INDOT’s updated Road Inventory Data (RID 2000) were incorporated into the network including number of lanes, shoulders, medians, access control types, traffic and truck count data, and functional classifications. A total of 4,720 TAZs were created with external stations representing the areas in neighboring states (Figure 10). The TAZ structure was developed to generally conform to the roadway network and previously developed TAZs from the CTPP. New zones were created by subdividing CTPP zones. More than 10,000 centroid connectors (a maximum of three per zone) were added to the network using a fully automated process. Long Distance Trip Purpose Production Attraction Balance To Business Household Total employment Production Tourist Household Retail/service employment Production Other Household Households and total employment Production TABLE 5 METHOD OF FRATAR FACTORING LONG DISTANCE OD TABLES

36 locate state jurisdictional highway signals (gray dots in Fig- ure 11), and the INDOT’s crash database for 1997 through 1999 was used to locate signals on local streets (black dots in Figure 11). Therefore, signals on local roads without a crash were missing from the ISTDM network. A new procedure was developed to estimate free-flow speed based on detailed geometric features and functional types of the roadway. The data were obtained from the RID 2000 and the original I-69 speed survey database. Nonlinear regression analysis was conducted to define free-flow speed based on posted speed for each unique facility type (number of lanes, divided/undivided, area type, and access control type). Figure 12 gives the formulas developed for major fa- cility types. Highway Capacity Manual 2000 (HCM 2000) procedures were followed to calculate speed reduction factors based on the limiting factors from HCM 2000. The speed reduction factors were applied to estimate peak-hour roadway capaci- ties. Daily capacities were then obtained by factoring the hourly capacities with the inverse of time-of-day factors (i.e., the percentages of daily traffic in the peak hour). Figure 13 gives an example of curve-fitted capacity adjustment factors for lateral clearance. A similar procedure was used for all capacity-reduction factors. FIGURE 9 Indiana Statewide Travel Demand Model network. FIGURE 10 Indiana Statewide Travel Demand Model ISTDM TAZ structure. ISTDMnet INDOT Inventory New Signals from Crash Data FIGURE 11 Traffic signals in Indiana Statewide Travel Demand Model network. Traffic signals in the entire state were located on the net- work. Signal information integrated to the network includes signal location, approach priority, and number of upstream signals. Almost 3,900 traffic signals were located on the net- work. INDOT’s traffic signal data from 1997 was used to

37 Subsequently, the free-flow speed and roadway capacities were adjusted to account for signal delays by a process that first estimates control delays, d, at signals using a simplified version of the HCM 2000 uniform delay term: where C is the cycle length, g is the green time, and PF is the progression factor. The delay is then used in an empirical for- mula to create capacity-reduction factors for links with signals. ISTDM trip generation models were developed for four trip purposes (home-based work, home-based other, non- home-based, and long purpose) and for three area types (ur- ban, suburban, and rural). Cross classification of household d C g C PF= −⎛⎝ ⎞⎠2 1 2 * size and automobile ownership was used for trip production estimation. Trip attractions were related to employment cate- gories and number of households. Attraction trip rates as de- rived from linear regression are shown in Table 6. Year 2000 Census household data, the 1995 Indiana Travel Survey, and 2001 NHTS data were used for model development. The Cor- ridor 18 Model dataset was adopted for external long purpose trips. Stratification curves were developed to breakout the households into categorical groupings to apply the cross-clas- sification trip rates. The curves were calibrated using the CTPP TAZ level data. Figure 14 presents an example of the stratification curves. Gravity expressions were used for ISTDM trip distribu- tion. The friction factors were calibrated by trip purposes us- ing the 1995 Indiana Household Survey and the 2001–2 Area Type Free-Flow Speed 1,2 Condition Note 2-lane 2-way undivided highways 2-lane 2-way divided highways Multilane undivided highways Multilane divided highways Full acess controlled highways Rural Suburban Urban Rural Suburban Urban Rural Suburban Urban Rural Suburban Urban 0.009751 · PSPD2 + 30.03397 (0.000017 · (PSPD – 72.323105)2 + 0.019702)–1 + 19.835323 (0.119687 – 0.023365 · ln(PSPD))–1 + 0.373821 · PSPD (0.119687 – 0.023365 · ln(PSPD))–1 + 0.373821 · PSPD (0.081714 – 0.016217 · ln(PSPD))–1 6.189 + 0.9437 · PSPD 117.640917 · PSPD0.0015+0.001279·PSPD – 98.065483 3.180682·PSPD0.857638 – 84.105587 · e–41.803252 / PSPD 3.180682 · PSPD0.857638 – 84.105587 · e–41.803252 / PSPD (0.000017·(PSPD – 72.323105)2+ 0.019702)–1 + 19.835323 (0.000071 · (PSPD – 64.166165)2+ 0.035258)–1 + 9.061039 · ln(PSPD) 2.836165 · PSPD – 0.071256 · PSPD2+ 0.000744 · PSPD3 16.0359 + 0.8223 · PSPD 25 25 25 25 25 25 25 25 25 25 25 25 25 ≤ PSPD ≤ 55 25 ≤ PSPD ≤ 55 25 ≤ PSPD ≤ 55 PSPD < 25 PSPD < 25 PSPD < 25 25 ≤ PSPD ≤ 55 25 ≤ PSPD ≤ 55 25 ≤ PSPD ≤ 55 PSPD < 25 PSPD < 25 PSPD < 25 25 ≤ PSPD ≤ 65 25 ≤ PSPD ≤ 50 25 ≤ PSPD ≤ 55 25 ≤ PSPD ≤ 55 50 ≤ PSPD ≤ 65 25 ≤ PSPD ≤ 55 25 ≤ PSPD ≤ 55 PSPD < 25 PSPD < 25 PSPD < 25 PSPD < 25 PSPD < 25 PSPD < 25 PSPD = 55 PSPD = 60 PSPD = 65 PSPD = 70 No or Partial Access Control No or Partial Access Control No Access Control 64.00 67.06 70.21 73.30 Note: 1 Free-flow speeds in mph. 2 PSPD: Posted speeds in mph FIGURE 12 Estimation formulas for free-flow speed.

38 NHTS dataset (see Figures 15 and 16). Socioeconomic ad- justment factors (k-factors) were also validated to adjust trip distributions not explained by friction factors. ISTDM im- plemented a single feedback loop of congested times to the gravity expressions. Fixed-mode shares for home-based work, home-based other, and nonhome-based trips by area types (urban, subur- ban, and rural) were calculated from the 1995 Indiana House- hold Survey and the 2001 NHTS data. Automobile occupancy rates were also obtained from the 1995 survey. For the long trip purpose, a multinomial logit expression was adapted from the California High Speed Rail Study Model and then recali- brated for the ISTDM for a division of trips between automo- bile and intercity bus/rail hybrid. Table 7 shows the calibrated model parameters. “Freight and non-freight trucks were estimated separately. For freight trucks, base year 1993 truck trip tables from the Indiana University study were factored up to year 2000 lev- els by commodity group.” Non-freight truck trip tables were estimated from truck ground counts after first removing freight trucks. The ISTDM used a multiclass assignment approach for traffic assignment, with truck trips and automobile trips loaded to the network at the same time. Two trip tables were developed for truck trips: freight truck trips and non-freight truck trips. The traffic assignment procedure was run twice by including a feedback loop to trip distribution so that the gravity expression could use travel times based on the ini- tially assigned roadway volumes. BPR travel time and vol- ume curves were specified by functional classification. 1.0000 0.9900 0.9800 0.9700 0.9600 0.9500 0.9400 0.9300 0.9200 0.9100 0.9000 A dju stm en t F act or 75 72.5 70 67.5 65 62.5 60 57.5 55 6 ft 4 ft 2 ft 0 ft Lateral Clearance Free-Flow Speed FIGURE 13 Capacity-reduction factors for lateral clearance for two-lane freeways. Trip Purpose Demographic Category Rate Home-Based Work Employment in retail, FIRE, education, services, and government sectors 1.400 Employment in non-retail; construction; manufacturing; agriculture, forestry, and fisheries; and transportation sectors 1.120 Home-Based Other Employment in retail sector 4.850 Employment in FIRE, education, services, and retail sectors 3.200 Employment in education sector 1.750 Households 1.650 Nonhome-Based Employment in retail sector 4.490 Employment in FIRE, education, services, and government sectors 1.130 Employment in non-retail, construction, manufacturing, and transportation sectors 0.380 Households 0.590 Long Total employment 0.023 Employment in FIRE, education, services, and government sectors 0.090 Employment in agriculture, forestry, and fisheries; mining; construction; manufacturing; non-retail; and FIRE sectors 0.030 Employment in retail and services sectors 0.020 Notes: FIRE = finance, insurance, and real estate TABLE 6 TRIP ATTRACTION RATES BY TRIP PURPOSE

39 The ISTDM model was validated by comparing the base year observed daily traffic counts to the model estimates. Statistics used for validation included: percent RMSE, systemwide average error, mean loading errors, and total VMT errors. Once possible sources of model errors were identified, the components were revaluated and corrected. Adaptations included modifying trip production rates, ad- justing friction factors or k-factors in the gravity expression, adjusting volume–delay functions, and modifying centroid connectors. Overall, the ISTDM shows base–case forecasted volume as being close to actual volumes, as shown in Figure 17. The RMSEs in Figure 17 are similar to what might be seen in an urban model. The systemwide RMSE is 39.45%. The ISTDM also includes a post-processor that uses the output of the travel model to estimate speeds, levels of ser- vice, crashes, and other measures of effectiveness. The ISTDM paid particular attention to its socioeconomic forecasts, which underlie the traffic forecasts. Zonal popula- tion forecasts were developed by first establishing county control totals and then distributing the totals to TAZs using an accessibility-based regression model. Historical data from Woods & Poole economics forecasts (April 2004), Indiana State Data Center forecasts by county, and the Regional Eco- nomics Model, Inc. (REMI) forecast for the state of Indiana were examined to produce county-level population. Inde- pendent variables in the regression model included: • Total population, • Total households, • Population density, • Population under age 17, • Percent of households with head of household over age 65, • Household workers, • Average household income, • Accessibility to wealth (by place of residence), • Accessibility to unoccupied housing units, • Accessibility to schools, • Accessibility to university enrollment, 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 Pe rc e n t D is tri bu tio n 1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 4 Average Persons per Household H1 HH3 HH2 HH4 FIGURE 14 Household size stratification curves. (Source: Bernardin, Lochmueller & Associates, Inc. and Cambridge Systematics, Inc., 2004 and Indiana response to Peer Exchange questionnaire, Longboat Key, Florida, September 2004.) H1 = one-person household; HH2 = two-person household; HH3 = three-person household; HH4 = four-person household. 50000 45000 40000 35000 30000 25000 20000 15000 10000 5000 0 Fr ic tio n Fa ct or 3 7 11 15 19 23 27 31 35 39 43 Trip Length in Minutes HBW HBO NHB FIGURE 15 Short trip friction factors. (Source: Indiana response to synthesis questionnaire February 2005.) HBW = home-based work; HBO = home-based other; NHB = nonhome-based. 700000000 600000000 500000000 400000000 300000000 200000000 100000000 0 Fr ic tio n Fa ct or 50 62 74 86 98 11 0 12 2 13 4 14 6 15 8 17 0 18 2 19 4 Trip Length in Miles FIGURE 16 Long trip friction factors. (Source: HBA Specto Incorporated and Parsons Brinckerhoff Ohio 2005.). 120 100 80 60 40 20 0 Pe rc e n t R M S Er ro r 0 20,000 40,000 60,000 80,000 100,000 120,000 Average Volume in Range FIGURE 17 Validation accuracy for the Indiana model. (Source: Hunt and Abraham 2003.) Variable Original Values Adjusted Values Cost ($) –0.0276 –0.0276 IVTT—Line Haul Travel Time (min) –0.0069 –0.0069 OVTT—Access/Egress Time (min) –0.0083 –0.0083 Bias Constant –0.87 –1.15 TABLE 7 REVALIDATED MULTINOMIAL LOGIT EXPRESSION PARAMETERS (long trip purpose)

40 • Travel time to nearest city center, • Travel time to nearest airport, and • Travel time to nearest major arterial. The regression model was calibrated by comparing the re- gression of year 2000 population against 1990 socioeco- nomic data with actual 2000 data. Then the model was used to produce population changes from year 2000 to 2030 in terms of changes in zonal shares of county totals. “Only half the modeled shift in zonal share of county population pre- dicted by the regression model was applied to bias the final allocation towards the existing distribution of population given the inherent uncertainty in land use forecasting.” The same approach for forecasting population was used to forecast zonal employment. The independent variables in- cluded in the accessibility regression model were: • Total population, • Total households, • Population density, • Aggregate personal income, • Presence of airport, • Presence of hospital, • University enrollment, • Travel time to nearest city center, • Travel time to nearest major arterial, • Travel time to nearest freeway, • Accessibility to intermodal freight facilities, • Accessibility to households, • Accessibility to population, • Accessibility to university enrollment, and • Accessibility to wealth (by place of residence). “Only one-third of the modeled shift in zonal share of county employment predicted by the regression model was applied to bias the final allocation towards the existing distribution of employment given the inherent uncertainty in land use forecasting and the r-squared for the regression model.” ISTDM has been used for statewide system planning, cor- ridor planning, bypass studies, economic development stud- ies, air quality analysis, project prioritization, inputs to eco- nomic modeling, and long-term investment studies. Sources for this case study were: Bernardin, Lochmueller & Associates, Inc., and Cambridge Systematics, Inc. (2004), Indiana response to Peer Exchange questionnaire, Longboat Key, Florida (Sep. 2004), and Indiana response to Synthesis questionnaire (Feb. 2005). CASE STUDY 3: OHIO COMBINED PASSENGER AND FREIGHT COMPONENTS Both Ohio and Oregon have statewide models that differ sig- nificantly from the typical four-step UTP model seen else- where. These two statewide models share many similarities, particularly their emphasis on forecasting the spatial distri- bution of economic activity and land use. The Oregon statewide model was recently described in the draft report for NCHRP Project 8-43. This section will emphasize the economic activity portions of Ohio’s model, how the eco- nomic activity portions integrate with other components, and the microsimulation of activity-based trip patterns. Both the Ohio and Oregon models have the philosophy that travel is a consequence of human and economic activities; therefore, the spatial organization of the state’s economy is first mod- eled comprehensively and aggregately. Activities result in trip making, which is then modeled in a disaggregated fash- ion, both in space and in time. The scope of the Ohio model was decided on after a study of stakeholder needs. The model was designed to address three principal issues: economic development, congestion mitigation, and truck flows. Ohio deliberately staged its model development by first creating an “interim model,” which is currently operational. Of greater interest here is the “advanced” model, which is sched- uled to be operational soon. The overall structure of Ohio’s model may be seen in Figure 18 as being made up of several submodels. The submodels that seem most unusual in a statewide context are the Land Development submodel and the Activity Allocation submodel. These submodels are similar to aggregate land use models that have been implemented in some metropolitan areas. Because these submodels deal with both household and industry location simultaneously, there is an intrinsic linkage between the passenger and freight compo- nents. The other submodels, some nontraditional, replace sim- ilar functions of a four-step model or are post-processors. Ohio Statewide Model Summary State population: 11.4 million State area: 44,828 square miles Gross state product: $403 billion No. of zones: 5,103 External zone structure: Halo, states Internal zone structure: TAZs, grid cells No. of highway links: 250,000 Freight modes: Truck No. of commodities: 28 categories No. of industries: 15 categories Household composition: Microsimulation Tour formation: Microsimulation Passenger mode split: Microsimulation Truck vehicle split: Microsimulation Assignment: Static equilibrium, multiclass Delay estimation: BPR curves Major data: Household and business surveys, TRANSEARCH, CTPP, ES-202, County Business Patterns, assessor land values Time frame: Eight years of development time Computation time: Not determined In-house staff: 1 FTE

41 To keep computations reasonable, Ohio adopted three nested-zone structures. The economic activity portions of the model use approximately 700 “activity model zones,” which are each made up of whole TAZs. The 5,103 TAZs are com- posed of many grid cells for (1) maintenance of land use and demographic data and (2) disaggregation of traffic assign- ment. Ohio’s grid cells are also used for providing locations of origins and destinations for those steps that microsimulate freight and person travel. Small TAZs cover all of Ohio and a halo of approximately 50 mi into surrounding states. Larger zones extend to the rest of the 48 contiguous states. Ohio’s TAZs are shown in Figure 19 and the network within Ohio is shown in Figure 20. An extensive data collection effort was needed to support the goals of the model. The major data sources were: • Household travel surveys, • Household long distance travel survey, • GPS-based travel survey, • Business establishment survey, • National Transport Networks, • Ohio DOT Roadway Information Database, • U.S. Census, • ES-202, • TRANSEARCH, Interregional Economic Model Aggregate Demographic Model Land Development Model Activity Allocation Model Employment Spatial Disaggregation Model Disaggregate Household Synthesis Model Personal Travel Tour Models Commercial Travel Tour Models Network Assignment Models Air Quality & Accident Models Sub-Area Traffic Micro-Simulation Model Travel Demand Models Times & Costs FIGURE 18 Overall structure of Ohio’s statewide travel forecasting model. (Source: Hunt et al. 2004a.) FIGURE 19 Ohio’s traffic analysis zone structure. (Source: Ohio’s response to the synthesis questionnaire 2005.)

42 • Department of Natural Resources land use data, • County assessor land value data, • Ohio DOT traffic counts, • IMPLAN (IO model), • Roadside surveys, • Travel time studies, • CTPP outside Ohio, • County Business Patterns, • BEA Regional Economic Information System program, • College and university enrollments, and • County auditor data. The household travel surveys were composed of new sur- veys in small and medium MPOs, in addition to existing sur- veys in larger MPOs. These household surveys combined to yield approximately 25,000 responses. A GPS survey was si- multaneously conducted to monitor underreporting of trips. The household long distance survey elicited information about trips of greater than 50 mi from 2,000 households. Roadside surveys were taken at approximately 700 locations. Approxi- mately 800 business establishments were surveyed to provide information about services and commodities that are not in- cluded in the TRANSEARCH database. NHTS was not used. The activity allocation and land use submodels were based on PECAS (Production Exchange and Consumption Alloca- tion System) (Hunt and Abraham 2003), a land use model de- veloped at the University of Calgary. In a manner similar to a compact IO table, PECAS tracks the flow of goods and ser- vices between industries and final demand (households), but does so spatially as well as monetarily. The model locates pro- ducers and consumers within zones in such a manner as to cre- ate a supply/demand equilibrium throughout the state. The supply/demand equilibrium is maintained by adjusting prices of commodities, services, labor, and land (or floor space). The allocations of goods, services, and labor are undertaken using logit and nested-logit expressions, where utility functions con- tain (1) the cost of travel or transport, (2) the size of zone, and (3) the price of the commodity. The allocations depend on what is already present or has been allocated in a previous time period. Industry is organized into the following categories: • Agriculture, forestry, and fishing • Primary metals • Light industry • Heavy industry • Transportation equipment • Wholesale • Retail • Hotel and accommodation • Construction • Health care • Transportation handling • Other services • K-12 education • Higher education • Government and other. Households are divided into six categories by income. The model is stepped through a sequence of 5-year time pe- riods until the planning year has been reached. The Land De- velopment submodel determines how categories of land are developed using a series of logit expressions. Land uses are: • Residential, • Commercial, • Light industrial, • Heavy industrial, • Grade school, • Post-secondary institutional, • Health institutional, • Agricultural, • Forest and protected resource, and • Vacant. Ohio is also using the capability of PECAS to separate land uses in serviced and unserviced categories. The traditional generation, distribution, and mode split steps for personal travel are replaced by microsimulation of household travel decisions. Separate submodels are provided for household synthesis; short-distance, home-based person tours; long distance, home-based person tours; commercial, work-based person tours; and visitor person tours. • Household synthesis—This submodel uses a Monte Carlo process to create a list of households by TAZ. Each household has attributes that are required by other submodels. The Monte Carlo probabilities are based on FIGURE 20 Ohio’s network within state. (Source: Ohio’s response to the Peer Exchange questionnaire 2004.)

43 the existing composition of the zone and the quantities of newly developed land. • Person tours—The four tour submodels are conceptually similar. They use microsimulation to create a list of tours and then a list of trips within tours. Selection probabili- ties come from logit expressions. Trips have attributes of origin zone, destination zone, start time, and mode. Transport of large commodities is handled somewhat tra- ditionally, once the flows of goods have been established by the economic activity modules. Flows between activity model zones are converted to flows between TAZs by ap- portioning flows according to employment levels. OD flows of goods are converted to a whole number of vehicles grouped by vehicle types and departure times, using a Monte Carlo process. The list of vehicle trips, so obtained, can be post-processed in a traffic microsimulation or aggregated for a traditional traffic assignment. The 28 commodity cate- gories are consistent with two-digit STCC. Service and delivery commercial tours are created with microsimulation. As with person tours, logit expressions are used to obtain selection probabilities. The overall number of tours relates to the amount and types of employment in the activity model zone. The attributes of each trip are deter- mined in the following order: stop purpose, stop TAZ, de- parture time (accounting for earlier stops on the tour), stop subzone, and vehicle type (light, medium, and heavy). This method is described in an article about Calgary’s urban model (Hunt et al. 2004b). This method has these processes: tour generation, tour stop time, tour purpose and vehicle type, next stop purpose, next stop location, and stop duration. The last three processes are performed iteratively with ear- lier stops in the tour influencing the nature of later stops. Traffic assignment is stochastic, multiclass, and user- optimal equilibrium. Capacities are coded for 24-h. Delay for the equilibrium assignment is calculated with BPR curves. Transit assignment is also done. Post-processors have been provided for air pollution emis- sions and accident calculations and for traffic microsimula- tion of small portions of the network. Sources for this case study were: Hunt and Abraham (2003), Hunt et al. (2004a), HBA Specto Incorporated and Parsons Brinckerhoff Ohio (2005), Ohio’s response to the Peer Exchange questionnaire (2004), and Ohio’s response to the Synthesis questionnaire (2005). CASE STUDY 4: VIRGINIA FREIGHT COMPONENT The Virginia freight component is designed to properly ac- count for trucks on highways when loading passenger auto- mobiles. The model combines trucks and automobiles within an equilibrium multiclass traffic assignment step that preloads trucks using all-or-nothing assignment. Truck OD tables are derived from Reebie’s TRANSEARCH database and from systematic adjustments based on truck counts. The TRANSEARCH data for Virginia gave commodity flows in tons from, to, and within Virginia. Data were orga- nized geographically by state, BEA region, and Virginia county. Separate tables were given for each two-digit commodity group from STCC for truck, railroad, water, and air. Trucks were further divided into truck-load, less-than- truckload, and private. Eventually the model was organized into 28 commodity groups, as listed in Table 8. The TRANSEARCH database omits many agricultural products and local service and delivery trucks, which particularly affect estimates of truck movements within the state. The freight component uses the same highway network as the passenger component. This network has nearly 247,000 links and almost 1,600 TAZs. The network is illustrated in Fig- ures 21 and 22, although it is difficult to get a sense of the highly detailed network within Virginia from these figures. The zone system is illustrated in Figures 23 and 24. It can readily be seen that the network and zone system span the full contiguous 48 states, but is sharply focused on Virginia. A moderately de- tailed network and set of zones extend well into adjacent states and beyond. Virginia implements subzoning for traffic assign- ment that helps eliminate lumpy vehicle loadings to links. Virginia Statewide Freight Component Summary State population: 7.1 million State area: 42,769 square miles Gross state product: $304 billion No. of zones: 1,584 External zone structure: Halo, aggregations of states Internal zone structure: Micro/macro No. of links: 246,935 Freight modes: Truck No. of commodity categories: 28 Production: Employment by industry group Consumption: IO, employment by industry group, population Distribution: Fratar factoring freight flow database, OD table estimation to truck ground counts Mode split: Fixed shares Truck-type split: Fixed shares Assignment: Static equilibrium, multiclass Delay estimation: BPR curves Major data: TRANSEARCH, IO tables Time frame: Three years of development time Computation time: 2.5 h In-house staff: 1 FTE Virginia’s freight component concept is illustrated in Fig- ure 25. OD tonnages by trucks from the TRANSEARCH data- base are converted to truck loads by the payload factors listed in Table 8, adopted from Texas. Daily tonnage was taken to be 1/365th of yearly tonnage. An initial traffic assignment was made. The truck OD table from the TRANSEARCH database

44 was found to substantially underestimate truck volumes be- cause of the missing commodities. Instead of attempting to model these missing commodities directly, Virginia adopted a method of correcting the TRANSEARCH data by comparing the assigned volumes to ground counts. Virginia used a maximum likelihood method of OD table estimation from ground counts that was contained within their travel forecasting software package. This method required a “seed” OD table, as well as numerous truck ground counts. The seed OD table was created by a gravity expression, where total employment by zone was taken to be the measures of both trip productions and trip attractions. The TRANSEARCH commodities were assigned to the network and the differences from ground counts were found. These differences were as- sumed to consist of trucks carrying the missing commodities in the TRANSEARCH database. The resulting OD table form of the gravity expression was scaled so that, on average, the to- tal number of trucks was correct when assigned to the network. This scaled table was adjusted to the difference between the assignment and the counts. Each commodity was forecasted individually by Fratar factoring its OD table. Each of the 28 commodity groups has been matched to a similar industry group for calculating changes in commodity production. Changes in production are directly proportional to changes in industrial employ- ment. For commodity consumption, a weighted combination of industry employment and final demand is used. The weights are derived from analysis of sales from the National Input–Output Tables, Direct Requirements Table. Final demand was forecasted in proportion to a weighted combi- nation of population and employment. Forecasts in employ- ment were provided for counties by Woods & Poole and modified by national productivity coefficients. County-level data were apportioned to TAZs according to employment totals. There were no special generators. Movement Type STCC Commodity Type Intrastate Interstate Through 1 Farm products 16.1 9 Fresh fish or marine products 12.6 10 Metallic ores 11.5 11 Coals 16.1 14 Nonmetallic ores 16.1 19 Ordinance or accessories 3.1 20 Food products 17.9 21 Tobacco products 9.7 22 Textile mill products 15.2 23 Apparel or related products 12.4 24 Lumber or wood products 21.1 25 Furniture or fixtures 11.3 26 Pulp, paper, allied products 18.6 27 Printed matter 13.8 28 Chemicals or allied products 16.9 29 Petroleum or coal products 21.6 30 Rubber or miscellaneous plastics 9.1 31 Leather or leather products 10.8 32 Clay, concrete, glass, or stone 14.4 33 Primary metal products 19.9 34 Fabricated metal products 14.3 35 Machinery 10.8 36 Electrical equipment 12.7 37 Transportation equipment 11.3 38 Instruments, photo, optical equip. 9.4 39 Misc. manufacturing products 14.2 40 Waste or scrap metals 16.0 50 Secondary traffic 16.1 16.1 12.6 11.5 16.1 16.1 3.1 17.9 16.4 16.1 12.4 21.0 11.3 18.5 13.6 16.9 21.6 9.2 11.0 14.3 19.9 14.3 10.8 12.8 11.3 9.4 14.4 16.0 16.1 16.1 12.6 11.5 16.1 16.1 3.1 17.9 16.8 16.5 12.5 21.1 11.4 18.6 13.9 16.9 21.6 9.3 11.3 14.4 20.0 14.3 10.9 12.9 11.3 9.7 14.8 16.0 16.1 Note: STCC = Standard Transportation Commodity Code. TABLE 8 VIRGINIA PAYLOAD FACTORS FOR COMMODITIES FIGURE 21 Virginia’s zone system, full extent. (text continues on page 47)

45 FIGURE 22 Virginia’s zone system, in and near state. FIGURE 23 Virginia’s highway network, full extent.

46 FIGURE 24 Virginia’s highway network within state. Reebie2001 Data Rail Commodity Flows Truck Commodity Flows Air Commodity Flows Water Commodity Flows Truck Loading Factors by Commodity Type Reebie Truck Trips Truck OD Seeds (short distance) Initial Truck Network Assignment Local Truck Matrix Estimation Local Truck Trips Overall Truck Trips Network Assignment FIGURE 25 Major steps in Virginia’s truck model.

47 The capacity restraint assignment involved estimating de- lays with the BPR curve, which requires free speed and ca- pacity for a link. Free speeds and 24-h lane capacities were set separately by functional class. After an initial traffic as- signment, capacities were adjusted upward within urban ar- eas to account for the sparse network there. Because of the rural orientation for the model, the passenger car equivalent factor for trucks was one. Both Virginia and Louisiana (Wilbur Smith Associates 2004) implemented essentially two distinct travel forecasting models, referred to as the “micro” model and the “macro” model. Together these two models create a way to consider long trips across states while still working at a sufficiently detailed scale for trips within the state. The purpose of the macro model is to provide information on trips passing through Virginia or having one end within Virginia. The macro model spans the entire United States and works at the county level within the state. The macro model has just 204 TAZs, of which 135 are within Virginia. The macro network has almost 59,000 links. The micro model operates within Virginia at the level of cen- sus tracts and places. The micro model provides the necessary forecasts of travel to satisfy statewide planning needs. (Sources: Wilbur Smith Associates 2003, Virginia’s response to the Peer Exchange questionnaire 2004, Virginia’s response to the Syn- thesis questionnaire 2005, Wilbur Smith Associates 2005b.) CASE STUDY 5: WISCONSIN FREIGHT COMPONENT At the time of this writing, Wisconsin had just finished the third generation of its travel forecasting model. However, documentation of model details had not been completed. This case study is based on a series of interim memoranda, the consultant’s scope of work, questionnaire responses, and interviews with the modeling team. Wisconsin’s overall statewide modeling effort was designed to meet these needs: • Long-range plan development (statewide and urban) • Air quality conformity analysis • Corridor planning for capacity investment, program- ming, and design • Modal investments (e.g., introduction of new intercity bus service) • Traffic forecasting for project design • Traffic Impact Analysis • Traffic diversion impacts • Modal diversion impacts • Congestion mitigation planning—Wisconsin DOT Intelligent Transportation System “blue route” corridor planning efforts • Detour simulation analysis • Bypass feasibility studies • EIS traffic data input. A major motivation for building the statewide model was the need to determine the impacts truck traffic has on major highways. The freight component addresses these needs by forecasting commodity-carrying truck volumes. Wisconsin Statewide Freight Component Summary State population: 5.5 million State area: 65,503 square miles Gross state product: $200 billion No. of zones: 1,875 External zone structure: Halo, states, aggregations of states, BEA regions Internal zone structure: Aggregations of TAZs No. of links: 200,000 No. of commodities: 25 categories Freight modes: Truck, rail, water (deep and inland), air cargo Production: Employment by industry group Consumption: IO table, employment by industry group, population Distribution: Gravity expression Mode split: Fixed shares Truck type split: Fixed shares Assignment: Static equilibrium, multiclass Delay estimation: BPR curves Major data: TRANSEARCH Time frame: 2.5 years of development time Computation time: 2 h In-house staff: 3 FTEs Wisconsin’s freight component is multimodal and commodity-based. The key database for the model was Ree- bie’s TRANSEARCH from 2001 aggregated to BEA regions. This database was factored into counties using com- modity flow information for Wisconsin that was assembled by Reebie in 1996. The following is a list of the commodity groups, each of which consist of whole two-digit STCC groups or represent intermodal shipments. • Farm and fish; • Forest products; • Metallic ores; • Coal; • Nonmetallic minerals; • Food; • Lumber; • Pulp, paper, allied products; • Chemicals; • Petroleum or coal products; • Clay, concrete, glass, and stone; • Primary metal products; • Fabricated metal products; • Transportation equipment; • Waste or scrap equipment;

48 • Secondary warehousing; • Rail drayage; • Other minerals; • Furniture or fixtures; • Printed matter; • Other nondurable manufacturing products; • Other durable manufacturing products; • Miscellaneous freight; • Hazardous materials; and • Air drayage. These commodity groups were selected to emphasize those commodities that were of the greatest economic importance to Wisconsin and to allow a direct match to industrial categories. Wisconsin’s freight component essentially contains the major UTP four-steps, as illustrated in Figure 26. Wisconsin’s zone system for freight differs somewhat from the passenger component. The zone system consists of (1) 1,642 small TAZs within Wisconsin, (2) counties within a thin halo around Wisconsin, (3) a few states or BEA re- gions near Wisconsin, (4) multistate regions for the rest of the contiguous United States, and (5) four huge zones for the rest of North America (see Figure 27). TAZs within Wis- consin and its halo match the passenger component exactly. The truck network is nearly identical to the passenger car network within Wisconsin and its halo, as seen in Figure 28. This network is very detailed within and near Wisconsin and it spans most of the contiguous United States, except for the Southeast, at a coarser level of detail (owing to the aggre- gated southeast freight zone using Atlanta as a loading point). This contrasts with the passenger component whose network extends only into the halo. Wisconsin’s truck net- work is nationwide to “account for global market impacts on FIGURE 27 Wisconsin’s freight component zone system. Generation Distribution Truck Shares & Payloads Truck Assignment Total Tons Tons by O-D O-D Trucks Trucks by Route FIGURE 26 Structure of Wisconsin’s freight component. O-D = origin–destination.

49 freight movements.” Rural portions of Wisconsin contain all functional classes that are major collector or higher. Urban portions of Wisconsin contain all functional classes that are collector or higher, except for the counties covered by the Southeastern Wisconsin Regional Planning Commission. Network attributes for links within Wisconsin come from ei- ther the Wisconsin Information System for Local Roads or the Wisconsin DOT’s State Trunk Network inventory. The network outside of Wisconsin was obtained from FAF, NHPN, and TIGER line files. Commodity generation equations were developed by linear regression between commodity production and industrial em- ployment for each commodity group based on county-level data. In a manner similar to Virginia, Wisconsin identified consuming industries and final demand for a given commod- ity group by using a national IO table. Regression analysis was then performed to ascertain the relationships between con- sumption totals in the TRANSEARCH database and zonal em- ployment and population. The independent variables used in the regression are shown in Table 9. Employment data were FIGURE 28 Wisconsin’s freight component network. Commodity Production Consumption Farm and Fish SIC01 + SIC02 + SIC07 + SIC09 SIC20 + SIC54 Nonmetallic Minerals SIC14 + SIC15 + SIC16 + SIC17 SIC14 + SIC15 + SIC16 + SIC17 Food SIC20 Population Lumber SIC24 SIC24 + SIC25 + SIC50 Pulp, Paper, Allied Products SIC26 SIC26 + SIC27 Chemicals SIC28 Total employment Clay, Concrete, Glass, and Stone SIC32 Population Primary Metal Products SIC33 SIC33 + SIC34 Fabricated Metal Products SIC34 Population Transportation Equipment SIC37 SIC42 Secondary Warehousing SIC42 Population Furniture or Fixtures SIC25 Population Printed Matter SIC27 Total employment Other Nondurable Manufacturing Products SIC21 + SIC22 + SIC23 Population Other Durable Manufacturing Products SIC30 + SIC31 + SIC35 + SIC36 + SIC38 + SIC39 SIC50 TABLE 9 INDEPENDENT VARIABLES FOR TONNAGE GENERATION FOR SELECTED COMMODITY GROUPS

50 obtained from Wisconsin’s Department of Workforce Devel- opment. Employment and demographic forecasts came from Woods & Poole growth rates applied to the Department of Workforce Development base data. Production or consumption of certain commodity groups did not correlate well with demographic variables. These com- modity groups were handled by factoring base year production and consumption data from the TRANSEARCH database. Wisconsin has 27 special generators for freight, which were county and commodity combinations. These special generators consist of retail distribution centers, truck–rail in- termodal terminals, ports, airports, and obvious outliers from the trip generation calibration, such as a highly automated General Motors assembly plant. The only primary data col- lection specifically for the freight component was a pilot truck survey at the Union Pacific intermodal terminal in Rochelle, Illinois. Another survey at this location is planned. When forecasting the relationship between employment and commodity production it is important to account for changes in worker productivity. Wisconsin obtained worker productivity factors for future years from a regional economic model. Trip distribution is handled by a gravity expression, where the friction factor for each commodity has been calibrated such that the model replicates average trip lengths from the TRANSEARCH data applied to the FAF highway network. The metric for spatial separation was distance in miles, dij. Therefore, friction factors were determined by this formula f (dij) = exp(dij /γ) where γ is a constant that varies by commodity group. Val- ues of γ range from approximately 100 to 2,800, depending on the commodity. Wisconsin’s freight component has four principal modes: truck, air cargo, railroad, and water shipping (both deep and inland). The model also explicitly considers three intermodal combinations (truck–air, truck–rail, and truck–water) by including drayage links on the highway network between Wisconsin counties and major intermodal terminals, some of which are located in Illinois and Minnesota. Mode split was accomplished by fixed shares as derived from the TRANSEARCH database. Air, rail, and water modes are not assigned to a network. Wisconsin’s highway traffic assignment is 24-h, multi- class, and user-optimal equilibrium. Trucks are loaded to the network at the same time as passenger cars; therefore, the route choice of trucks is influenced by congestion. Trucks re- ceive a constant passenger car equivalent factor of 1.9. De- lay was estimated with BPR curves. Annual tonnages of commodities were converted to daily trucks by using the payload factors from Table 10 and an assumed 306 trucking days per year. Table 10 was derived principally from Wisconsin records within VIUS. The only validation for the freight component that was distinct from passenger traffic was a comparison of com- modity tonnages between the model and TRANSEARCH. Assigned trucks were also compared with truck counts at ap- proximately 300 stations for reasonableness—a direct com- parison is not possible because the model forecasts com- modity carrying trucks only, not total trucks. Total truck VMT was checked against available data sources. Outputs from the freight component aid other planning ef- forts. An important feature of Wisconsin’s model is its interface with MPO models in the state. Internal truck travel in the MPO models is handled with procedures taken from the QRFM, but external traffic patterns come from the statewide model. In ad- STCC Description Tons per Truck 1 Farm products 8 Forest products 9 Fresh fish or other marine products 10 Metallic ores 11 Coal 13 Crude petroleum, natural gas, or gasoline 14 Nonmetallic minerals, excluding fuels 19 Ordnance or accessories 20 Food or kindred products 21 Tobacco products 22 Textile mill products 23 Apparel or other finished textile products 24 Lumber or wood products 25 Furniture or fixtures 26 Pulp, paper, or allied products 27 Printed matter 28 Chemicals 29 Petroleum or coal products 30 Rubber or miscellaneous plastics products 31 Leather or leather products 32 Clay, concrete, glass, or stone products 33 Primary metal products 34 Fabricated metal products 35 Machinery—Other than electrical 36 Electrical machinery, equipment, or supplies 37 Transportation equipment 38 Instruments—Photographic or optical goods 39 Miscellaneous manufacturing products 40 Waste or scrap materials 41 Miscellaneous freight shipments 42 Shipping devices returned empty 43 Mail and express traffic 44 Freight forwarder traffic 45 Shipper association or similar traffic 46 Miscellaneous mixed shipments 47 Small packaged freight shipments 48 Hazardous waste 49 Hazardous materials 99 Unknown 24 13 6 24 24 14 19 24 18 5 5 3 15 3 16 9 22 19 4 3 19 24 9 8 12 5 2 16 23 4 3 4 3 7 4 16 18 12 Note: STCC = Standard Transportation Community Codes. 23 TABLE 10 WISCONSIN PAYLOAD FACTORS BY TWO-DIGIT COMMODITY CODES

51 dition, forecasts from the statewide model are used to validate or supersede forecasts made from historical data using Box–Cox regression analysis. Outputs are also processed through STEAM (Surface Transportation Efficiency Analysis Model) from FHWA to obtain systemwide benefits. Major updates of Wisconsin’s model are planned to occur on a 6-year cycle to coincide with Wisconsin DOT’s Six- Year Highway Improvement Program. DISCUSSION The five case studies are representative of the newer gener- ation of statewide travel forecasting models. Except for their philosophy in following a three- or four-step forecasting process, these case studies differ remarkably in both their details and execution. Each state has customized the model steps to match its own planning objectives. This chapter shows three distinct methods of modeling statewide passen- ger travel. However, there is more similarity in the freight models, particularly in basing the forecasts on commodity movements. Ohio’s model emphasizes how non-freight commercial vehicles can be important to a forecast and might need special treatment apart from freight-carrying vehicles. Furthermore, the five case studies show that statewide models are becoming large and complex. The models are in- creasing the demand for high-quality secondary data, faster hardware and algorithms, better data visualization methods, and greater expertise.

Next: Chapter Four - Findings and Suggestions for Research »
Statewide Travel Forecasting Models Get This Book
×
MyNAP members save 10% online.
Login or Register to save!
Download Free PDF

TRB’s National Cooperative Highway Research Program (NCHRP) Synthesis 358: Statewide Travel Forecasting Models examines statewide travel forecasting models designed to address planning needs and provide forecasts for statewide transportation, including passenger vehicle and freight movements. The report explores the types and purposes of models being used, integration of state and urban models, data requirements, computer needs, resources (including time, funding, training, and staff), limitations, and overall benefits. The report includes five case studies, two that focus on passenger components, two on freight components, and one on both passenger and freight.

  1. ×

    Welcome to OpenBook!

    You're looking at OpenBook, NAP.edu's online reading room since 1999. Based on feedback from you, our users, we've made some improvements that make it easier than ever to read thousands of publications on our website.

    Do you want to take a quick tour of the OpenBook's features?

    No Thanks Take a Tour »
  2. ×

    Show this book's table of contents, where you can jump to any chapter by name.

    « Back Next »
  3. ×

    ...or use these buttons to go back to the previous chapter or skip to the next one.

    « Back Next »
  4. ×

    Jump up to the previous page or down to the next one. Also, you can type in a page number and press Enter to go directly to that page in the book.

    « Back Next »
  5. ×

    To search the entire text of this book, type in your search term here and press Enter.

    « Back Next »
  6. ×

    Share a link to this book page on your preferred social network or via email.

    « Back Next »
  7. ×

    View our suggested citation for this chapter.

    « Back Next »
  8. ×

    Ready to take your reading offline? Click here to buy this book in print or download it as a free PDF, if available.

    « Back Next »
Stay Connected!