4
Current State of the Practice

The committee obtained the information needed to categorize the current state of the practice in travel model development and forecasting from three sources:

  • A review of the literature;

  • A web-based survey of 381 metropolitan planning organizations (MPOs), 228 of which were represented by the responses received; and

  • Interviews with staff at a sample of 16 agencies [MPOs or state transportation agencies (STAs) that provide travel forecasting services for multiple MPOs], designed to obtain more detailed information.

The literature review provided insights into the state of the practice as perceived by knowledgeable authors engaged in research on or the application of travel forecasting methods. The literature also notes many of the perceived shortcomings of current practice and suggests approaches for improvement. Such critiques tend to be of two types: those that question the basic paradigm on which current practice is founded and those that question specific aspects of implementation. The noted shortcomings are discussed more fully in Chapter 5.

The web-based survey provided a broad view of travel forecasting as it is practiced by MPOs of various sizes across the nation that deal with a wide variety of planning issues. Even with an extensive questionnaire, however, the survey could address only the general methods used by each agency. Follow-up interviews were therefore conducted with MPOs represented by committee members and with several agencies known to have implemented new procedures or to be active in relevant professional organizations. The survey and interview findings are summarized below. Additional detail is provided in the electronic annex to this report [available at onlinepubs.trb.org/onlinepubs/reports/VHB-2007-Final.pdf (VHB 2007)]. Information derived from the



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 46
4 Current State of the Practice T he committee obtained the information needed to categorize the cur- rent state of the practice in travel model development and forecasting from three sources: • A review of the literature; • A web-based survey of 381 metropolitan planning organizations (MPOs), 228 of which were represented by the responses received; and • Interviews with staff at a sample of 16 agencies [MPOs or state trans- portation agencies (STAs) that provide travel forecasting services for multiple MPOs], designed to obtain more detailed information. The literature review provided insights into the state of the practice as per- ceived by knowledgeable authors engaged in research on or the application of travel forecasting methods. The literature also notes many of the perceived short- comings of current practice and suggests approaches for improvement. Such cri- tiques tend to be of two types: those that question the basic paradigm on which current practice is founded and those that question specific aspects of imple- mentation. The noted shortcomings are discussed more fully in Chapter 5. The web-based survey provided a broad view of travel forecasting as it is practiced by MPOs of various sizes across the nation that deal with a wide vari- ety of planning issues. Even with an extensive questionnaire, however, the sur- vey could address only the general methods used by each agency. Follow-up interviews were therefore conducted with MPOs represented by committee members and with several agencies known to have implemented new proce- dures or to be active in relevant professional organizations. The survey and interview findings are summarized below. Additional detail is provided in the electronic annex to this report [available at onlinepubs.trb.org/onlinepubs/ reports/VHB-2007-Final.pdf (VHB 2007)]. Information derived from the 46

OCR for page 46
Current State of the Practice 47 survey is descriptive of the methodology used and many of the details of its application. While this information documents the state of the practice, it does not reveal whether the models used produce accurate forecasts. For both the web-based survey of all MPOs and the targeted interviews with selected MPOs, respondents were guaranteed confidentiality. Thus the information presented here is either in summary form (most of the web-based survey findings) or linked to an agency identified by number rather than name. Only when the information is generally available through a published source is reference made to a specific MPO. WEB-BASED SURVEY The web-based survey was structured to obtain information that would quantifiably describe the travel forecasting procedures of a broad sample of MPOs. The express purpose was to identify the state of the practice in travel demand modeling on the basis of the current practices of regional MPOs. The survey was designed to incorporate specific questions raised by the com- mittee with regard to travel demand forecasting, as well as to provide an assessment and categorization of common modeling methods. Initial versions of the survey were developed and pretested by two MPOs in May 2005. The final surveys were originally distributed to all MPOs in June 2005, and responses were received through December 2005. The committee made a special effort, with assistance from the Association for Metropolitan Planning Organizations (AMPO) and others, to obtain information from those MPOs classified as large (i.e., in areas whose population exceeds 1 million). The survey was sent to 381 MPOs identified in databases obtained from the Federal Highway Administration (FHWA) and AMPO. In addition, each STA received an e-mail with a link to the survey and notification that a survey request had been sent to each of the MPOs in the state. STAs and regional MPOs were asked to coordinate and cooperate in responding to the survey. This was of particular importance for those states in which most of the travel demand forecasting work, including model development or application, is done by the STA. In these states, the STA completed and submitted the sur- vey for each MPO. When the analysis data set was closed, responses reporting data for 228 MPOs had been received. These 228 represent 60 percent of the 381 MPOs to which the survey was distributed—84 percent of the 43 MPO areas with a population of more than 1 million (large), 57 percent of those with a population of 200,000 to 1 million (medium), and 57 percent of those with

OCR for page 46
METROPOLITAN TRAVEL FORECASTING Current Practice and Future Direction 48 MPO Size Large Medium Small FIGURE 4-1 MPOs from which responses were received. a population of less than 200,000 (small). Since not all questions were answered for each MPO, the number of responses was not the same for all questions. Figure 4-1 shows the locations of the MPOs for which survey responses were received. All states except Hawaii are represented by the responses. Following is a summary of the basic steps of the travel modeling and forecasting process as it is currently practiced at most MPOs, based on the survey results. Input Data Agencies make use of extensive input data in developing travel models and preparing travel forecasts. These data include the following: • Traffic and vehicle classification counts, highway travel times and speeds, and results of traffic origin–destination surveys;

OCR for page 46
Current State of the Practice 49 • Transit ridership and boarding counts; • Roadway characteristics, such as functional classification, number of lanes, link distances, and intersection characteristics; • Transit routes and schedules; • Results of home interview surveys, including household characteristics and individual trips made by purpose, origin–destination, time of day, and mode; and • Current and future estimates of small-area employment, population, and households, along with other socioeconomic characteristics, such as household income and vehicle ownership. Some of these data are current, while some are forecast for future years. The committee’s web-based survey found that almost all MPOs require forecasts of population, households, and employment as input to their travel forecast- ing process. About half also forecast one or more of the following: household size, automobile ownership, and income. In general, MPOs are responsible for preparing these forecasts, although they often obtain assistance from other state or local agencies and consultants. Area System The entire region is divided into travel analysis zones (TAZs) and sometimes larger districts, which usually can be related to U.S. census tracts. The number of TAZs in a region varies from several hundred to several thousand, depend- ing on the region’s size. All travel is assumed to be to or from these zones. Each zone has a “centroid,” from which all traffic is assumed to start. The zone system is often mapped in a geographic information system (GIS) database. The number of TAZs for MPOs that responded to the committee’s sur- vey is, on average, 280 for small MPOs, 870 for medium MPOs, and 1,760 for large MPOs. The average TAZ density is 0.9, 0.8, and 0.5 TAZs per square mile for small, medium, and large MPOs, respectively. Networks Highway and transit networks are a principal means by which the supply side of transportation is represented. It is in the “assignment” process (discussed below) that demand and supply are brought together.

OCR for page 46
METROPOLITAN TRAVEL FORECASTING Current Practice and Future Direction 50 The highway network is represented as individual, connected links between intersections. Usually all freeways, expressways, principal arterials, minor arterials, and feeder/collector routes are included. Data on roadway characteristics are associated with each link. Current highway networks range in size from 4,200 links for small MPOs to more than 20,000 for large MPOs. The transit network (if there is one) is represented as routes for the vari- ous transit systems in the metropolitan area. Some of these routes run on the highway network and share highway links, while others are on their own right-of-way. Transit networks are typically more complex than highway net- works because of the multiple modes involved and the need to consider oper- ating frequencies and schedules. The vast majority of MPOs that have rail transit within their area include the entire rail network in their transit model. More than 80 percent of all MPOs and 90 percent of large MPOs include at least 75 percent of available express bus miles in their transit network. All of the large MPOs that reported having local bus service include at least three- quarters of the local routes in their network. In contrast, more than 60 per- cent of the small MPOs and 20 percent of the medium MPOs that reported having local bus service include less than three-quarters of local service miles in their network. The networks are connected to the TAZs in the area system through “centroid connectors,” which attach to the centroid at or near the center of each zone. Most networks are mapped and edited by using GIS software. Trip Generation The trip generation step involves estimating how many trips are expected to be made to and from each TAZ for various purposes, such as work, school, shopping, and commercial transport. As many as nine trip purposes are cur- rently used in MPO models; smaller MPOs are more likely to use fewer pur- poses. The estimation procedure employs mathematical models that associate each purpose with demographic characteristics of the TAZ, such as popula- tion, households, employment, vehicle ownership, and income. Current information on these variables may be obtained from special household sur- veys or census reports; future information is derived from forecasts, as noted above in the discussion of input data.

OCR for page 46
Current State of the Practice 51 Trip Distribution The trip distribution process is used to determine the number of trips between each pair of zones. Most MPOs accomplish this with a “gravity model” that assumes the number of trips between zones is (a) directly related to the number of trips generated from each zone and (b) inversely related to the difficulty of travel between two zones, which is usually a function of travel time and cost. Gravity models may be insensitive to socioeconomic or geo- graphic variables that influence travel behavior and consequently produce results that do not correspond to actual travel patterns. In this case, the inter- changes between zones may be adjusted by using so-called K-factors. The extensive use of K-factors is not recommended because they interfere with a model’s ability to predict future travel (Ismart 1990). Slightly fewer than 50 per- cent of all MPOs responding to the web-based survey reported using K-factors or a similar type of adjustment factor in their trip distribution model. Another model used for trip distribution is called “destination choice.” This type of model includes traveler characteristics (e.g., income, automobile ownership), travel conditions, and variables that influence the attractiveness of each destination (e.g., employment by job category, land use categories by square foot). The model can thus take into account differences in circum- stances that influence travelers’ destination choices and are poorly accounted for in a gravity model. Some believe that destination-choice models are supe- rior to gravity models for determining trip distribution, provide more infor- mation for use in policy analysis, and may require the use of fewer K-factors to adjust trip flows (e.g., Deakin and Harvey 1994, 43). The committee’s survey found that 11 MPOs are currently using destination-choice models. Mode Choice Mode choice is the allocation of trips between automobiles and public tran- sit. Within automobile travel, there is further allocation between drivers and passengers; within public transit, there may be allocation among local bus, express bus, and various rail options. Some MPOs include bicycle and walk- ing trips in their mode choice model. This modal determination is made on the basis of the trip’s purpose, origin, and destination; characteristics of the traveler; and characteristics of the modes available to the traveler. More than

OCR for page 46
METROPOLITAN TRAVEL FORECASTING Current Practice and Future Direction 52 90 percent of large MPOs and 25 percent of small MPOs reported using a mode-choice model. More than half of large MPOs reported that represen- tation of nonmotorized trips is part of their model set; few medium MPOs and almost no small MPOs model nonmotorized trips. Assignment Assignment is the allocation of trips to actual routes in the transportation net- work described above. The committee’s survey showed that a number of small (8 percent) and medium (4 percent) MPOs use the “all-or-nothing” assign- ment method, which allows travel between zones to be assigned according to the least-time route without regard to congestion. Most MPOs (73 percent of small, 74 percent of medium, and 91 percent of large MPOs) use the more sophisticated “equilibrium” method, which accounts for congestion and delay in assigning travel to specific routes. This method may require a number of iterations to achieve stability. In many smaller MPO regions, there is little traffic congestion, and transit service is minimal. For such regions, it is reasonable for MPOs to assign aver- age daily (24-hour) travel, a method that requires the use of factors to represent probable morning and afternoon peak period demand and resulting conges- tion. More complex regions with traffic congestion and more extensive transit operations model travel by time periods within the day and account more explicitly for congestion effects on route choices. Among large MPOs, 75 per- cent assign travel for at least two and as many as five time periods, including a.m. peak period, p.m. peak period, midday, evening, and nighttime. Feedback Travel times are typically required to estimate trip distribution and mode choice; however, travel times depend on the level of congestion on routes in the network, which is determined only after trip assignment has been com- pleted. Once congested travel times have been determined by the assignment process, these adjusted travel times should ideally then be fed back through the distribution, mode-choice, and assignment processes to produce more realistic estimates of travel. Feedback is a model feature required for metro- politan areas that are not in attainment of federal clean air standards.

OCR for page 46
Current State of the Practice 53 The use of feedback has become more common as advances in comput- ing power have enhanced the ability to iterate at reasonable time and cost. More than 80 percent of large MPOs feed back times to distribution and mode choice; 40 percent feed back congestion effects to forecasts of land use and automobile ownership. Postprocessing for Emissions Calculations Hourly link-specific traffic volumes and speeds must be calculated for use as inputs to the Environmental Protection Agency’s (EPA’s) MOBILE emissions model or California’s EMFAC model. These detailed emission model inputs are not usually travel model outputs and so must be postprocessed after the model has been run. Commercial and Freight Travel The treatment of commercial and freight travel is one area in which most travel forecasting models need substantial improvement. The development of better models is hampered by a lack of data on truck and commercial vehicle travel both within and beyond the metropolitan area. Truck trips are modeled in some fashion by about half of small and medium MPOs and almost 80 percent of large MPOs; few MPOs have the ability to model all freight movement. Movement Toward Advanced Models About 20 percent of small and medium MPOs and almost 40 percent of large MPOs reported that they are exploring replacing their existing model with an activity- or tour-based model (see Figure 4-2). Three U.S. cities have imple- mented such advanced models, and eight others are in the design process (see Chapter 6). The committee’s in-depth interviews with selected MPOs, how- ever, revealed that many of them are satisfied with their current model and believe it is adequate for most planning purposes. In the web-based survey, 70 percent of large and medium MPOs identi- fied features of their models needing improvement. The most commonly cited improvement was developing a tour- or activity-based model.

OCR for page 46
METROPOLITAN TRAVEL FORECASTING Current Practice and Future Direction 54 All MPOs (n = 198) 24% Large MPOs (n = 28) 38% Medium MPOs (n = 67) 21% Small MPOs (n = 103) 23% FIGURE 4-2 MPOs considering activity- or tour-based models. MPO INTERVIEWS The committee’s in-depth data gathering, including interviews of key MPO staff and supplemental written documentation provided by selected MPOs, offers insights beyond those obtainable from the mere tabulation of survey data. While these efforts could not be of sufficient depth or detail to allow assessment of the degree to which the procedures used by any agency produce accurate or valid forecasts, they do offer a view of specific practices used or contemplated by at least some of the more active MPOs. Given the small number of in-depth interviews, the methods and procedures of these agencies cannot be viewed as average or representative of the practice of most MPOs; rather, they are a snapshot of what at least a few active agencies have undertaken. After reviewing the web-based survey findings, the committee identified several topics on which it would be desirable to obtain further information through discussions with a number of MPOs. These topics included the following: • Validation, • Sensitivity analysis, • Staffing and budget, • Advanced practices, • Barriers to improvement, and • Perceived shortcomings of current methods. The committee identified 16 MPOs or STAs as candidates for these discussions (see Table 4-1). These agencies were selected because there was

OCR for page 46
Current State of the Practice 55 TABLE 4-1 Interviewed Agencies In-Person Interview Phone Interview Agency Area Agency Area East-West Gateway St. Louis, Missouri Atlanta Regional Atlanta, Georgia Commissiona Council of Governments Chicago Area Chicago, Illinois Mid-Ohio Regional Columbus, Ohio Transportation Studya Planning Commission Community Planning Boise, Idaho North Carolina North Carolina Association of Department of Southwest Idaho Transportation MetroPlan Little Rock, Ohio Department of Ohio Arkansas Transportation MetroPlan Orlando Orlando, Florida Sacramento Council Sacramento, Metropolitan San Francisco, of Governments California Transportation California Commissiona Virginia Department Virginia of Transportation Metro Portland, Oregon North Central Texas Dallas–Ft. Worth, Council of Texas Governmentsa Pikes Peak Area Colorado Springs, Council of Colorado Governments Regional Las Vegas, Nevada Transportation Commission of Southern Nevada a Agency was not interviewed in person but provided answers to the interview questions in written form. some indication that they were or had been engaged in developing or applying procedures that might be considered as advancing the state of the practice, were active in organizations such as AMPO, or had developed or applied travel forecasting models for multiple MPOs within a state. The committee visited six of these agencies; the rest either were interviewed via phone or provided responses to a detailed questionnaire. Practices found by these agencies to be useful and to lead to better forecasts are likely to become more widely adopted and over time to be incorporated into the state of the practice. As noted above, to protect the identity of the

OCR for page 46
METROPOLITAN TRAVEL FORECASTING Current Practice and Future Direction 56 responding agencies, the committee excluded specific agency names from the following discussion. As is the case with the web-based survey information, more detailed information from the in-depth data gathering may be found in the electronic annex to this report. Estimation, Calibration, and Validation Model validation must be understood as one of four closely related processes— estimation, calibration, validation, and application. The correct conduct of these processes is crucial to the quality of model results. • Model estimation: Information on actual travel is gathered by such means as household travel surveys and transit on-board surveys. Statistical estima- tion procedures are then used to create a model that can replicate the actual travel data. • Model calibration: After the model has been estimated, it is calibrated so that predicted travel accords with observed travel on highway and transit networks. • Model validation: After the model has been estimated and calibrated, it is validated to test its ability to predict future behavior. Validation requires comparing the model output with information other than that used in esti- mating or calibrating the model. The model output is compared with observed travel data, and parameters are adjusted until the output falls within an acceptable range of error. There are two superior (but not often performed) ways of checking model performance: (a) the historical method, in which a prior-year model is used to forecast current travel, which is then compared with actual current travel; and (b) backcasting, in which a current- year model is used to estimate travel for a prior year, which is then compared with actual travel in the prior year. Backcasting is used by 5 percent of all and 13 percent of large MPOs. (An example of the historical method is given in Chapter 5.) • Model application: Although a model may replicate base-year condi- tions, the model forecasts for future-year conditions should be checked for reasonableness. The sensitivity of the models in response to system or policy changes may be used as part of the reasonableness check (FHWA 1997). The committee’s survey and interviews revealed that true validation is often hampered by a lack of independent data sources. Even the more active MPOs

OCR for page 46
Current State of the Practice 57 validate against much of the same data (for example, nonwork trip generation, trip distribution, and mode choice) used to develop their models. Moreover, there are no commonly agreed-upon standards for an acceptable range of error other than thresholds suggested by FHWA and STA guidance such as the Ohio Department of Transportation’s Traffic Assignment Procedures (Giaimo 2001). Sensitivity Analysis As noted above, sensitivity testing is key to checking the reasonableness of travel forecasts. Formal procedures used for sensitivity analysis are described in the literature (Barton-Aschman Associates and Cambridge Systematics 1997). Two agencies interviewed for this study have begun changing some aspect of the system (e.g., inserting or removing employment and residential units in several zones, changing travel times) and then analyzing the forecast changes in trip making, trip distribution, mode shares, and network conges- tion. These agencies also remove links from the highway network to determine the impact on traffic volume on other highways in the network. In addition, specific aspects of the model may be tested, such as the sensitivity of mode choice to transit fares. Such sensitivity testing is done in a small number of agencies, but the prac- tice is not widespread. Agencies that do perform sensitivity analysis appear to do so on an ad hoc basis. The Federal Transit Administration’s (FTA’s) Summit tool has also been used for model checking. Staffing and Budget MPOs vary significantly in the number of staff devoted to travel forecasting. While the committee’s web-based survey of MPOs did not request informa- tion on the size of travel forecasting staff, this was a topic of the in-depth interviews. Among the MPOs interviewed, the staff reported as working on travel forecasting ranged from one person part-time with support from the state agency to as many as seven working at least part-time on some aspect of the process. The agency typical of this group of MPOs has two or three staff involved in travel forecasting and spends $150,000 to $200,000 annually on model application. Another study found that MPOs with a population of less than 500,000 have an average of one full-time travel modeler on staff, while larger agencies average three full-time modelers. The same study found that

OCR for page 46
METROPOLITAN TRAVEL FORECASTING Current Practice and Future Direction 58 virtually all MPOs believe it is either difficult or very difficult to hire experi- enced travel modelers (Urban Transportation Monitor 2006). Typically, model development is specifically budgeted for when a major upgrade is undertaken. Most of the interviewed agencies reported using con- sultants for model development, but a few have budgets large enough to sup- port staff that can devote at least some time to consideration or development of model improvements. Most of the agencies reported an increase in both staff and budget over the past 3 years. Advanced Practices Advanced practices include not only a major shift in the modeling paradigm from trip- to tour- or activity-based, but also incremental improvements to the four-step trip-based process. The interviews revealed several practices in use by the MPOs that have the potential for more widespread application. One of the agencies interviewed has an operational advanced model and five more are actively developing such models, while several others expressed an inter- est in doing so. Other agencies are less interested in the near-term imple- mentation of advanced modeling practices and appear to be satisfied with their current models. In addition, some agencies appear to be interested in developing more effective truck models and special generator models.1 Obstacles to Improvement Agencies interviewed cited a desire for tangible evidence that new procedures perceived as more complex or requiring significantly greater effort for devel- opment and application would yield forecasts notably better than those pro- duced with currently accepted procedures. Other factors cited as impeding the adoption of advanced techniques were the unavailability of vendor-supplied software needed for implementation, a lack of sufficient staff to apply the new techniques, the difficulty of finding staff versed in the development and appli- cation of the techniques, and insufficient funds for the purpose. As noted, 1 “Special generators” are developments such as airports, universities, shopping centers, and hospitals that place special demands on the transportation system.

OCR for page 46
Current State of the Practice 59 some of the MPOs interviewed believe their current models are doing an ade- quate job for the issues they are asked to address. Perceived Shortcomings of Current Methods Many MPOs would like to have improved procedures for studying policy and land development issues and for addressing truck trips and freight move- ment. Agencies also recognize that current regional travel forecasting proce- dures are not capable of addressing some policy issues and fail to provide the detail often requested for design studies or impact analyses. MATCHING THE MODEL TO THE CONTEXT The committee finds that no single approach to travel forecasting or set of travel forecasting procedures is “correct” for all applications or all MPOs. Rather, travel forecasting tools developed and used by an MPO should be appropriate to the nature of the questions being posed by the constituent jurisdictions and the types of analyses being conducted. Using a simplistic model to analyze complex issues can lead to findings that do not properly reflect the likely traveler response patterns. Similarly, applying an overly com- plex method to more straightforward issues not only diverts resources that might have better uses but also creates an opportunity to introduce errors related to factors not directly applicable to the problem at hand. Figure 4-3 illustrates how the modeling approach employed can be tailored to the issues being addressed. As the detail required to address a transportation issue increases, so, too, should the complexity of the analysis techniques. In a smaller metropolitan area experiencing little or no growth, with little transit, and having no air quality problems, a three-step model will likely be suffi- cient to determine the proper number of lanes for a new roadway. At the other end of the spectrum is a rapidly growing metropolitan area that is not in attainment of air quality standards, has severe congestion, and is planning to apply dynamic tolling for high-occupancy travel lanes on which there will also be bus rapid transit. In such an area, it will be desirable to have a travel forecasting process that (a) is sensitive to prices; (b) allows analysis of mode choice, time-of-day choice, and trip chaining; (c) permits detailed assess- ment of travel speeds by segment and time; and (d ) incorporates sufficient

OCR for page 46
Level of detail required for analysis Aggregate Disaggregate Land use High-occupancy Corridor studies, Roadway Transit New effects on mode Air quality travel lanes, peak spreading, Typical transportation sizing Starts choice analysis/tolls variable tolls saturated networks issues Transportation analysis methods Five-step (automobile availability) with Typical land use land use Population Household Traffic Three-step Four-step issues variables synthesis activity-based microsimulation Slow to moderate Spreadsheets, growth geographic information systems Fast growth, Lowry-type growth impact accessibility- analysis based modelsa Growth, housing Real estate costs, market models environmental justice Economic Markets and development input/output Land use analysis methods models Economic Disaggregate development and business and environmental residential justice location models = Reasonable combination of models a The Lowry-type accessibility-based model was first developed by Ira S. Lowry for Pittsburgh. Such models estimate the location and scale of (a) employ- ment for basic industries and services whose clients are outside the region, (b) employment for retail activities serving the region, and (c) the resident population of the region (Chapin and Kaiser 1979). FIGURE 4-3 Matching the model to the context.

OCR for page 46
Current State of the Practice 61 information about travelers to support an analysis of disproportionate impacts on minority and low-income populations. SUMMARY FINDINGS AND RECOMMENDATIONS The information used by the committee to describe the current state of the practice in metropolitan travel forecasting was obtained from three sources: a review of the literature, a web-based survey that yielded responses representing 228 MPOs, and interviews of staff at a sample of 16 agencies (MPOs or STAs). The basic modeling approach at most MPOs remains a sequential four- step process in which the number of daily trips is estimated, distributed among origin and destination zones, divided according to mode of travel, and finally assigned to highway and transit networks. Certain practices are common to most MPOs, while others differ according to local circumstances: • Common practice: Forecasts of population, households, and employment are required as input to the travel forecasting process. • Differing practice: About half of MPOs also forecast one or more of the following: household size, automobile ownership, and income. • Common practice: The modeled region is divided into TAZs. The zone system is mapped in a GIS database. • Differing practice: The number of TAZs in a region varies from several hundred to several thousand, depending on the region’s size. • Common practice: Transportation supply is represented through highway and transit networks mapped in a GIS database. • Differing practice: Highway networks range in size from 4,200 links for small MPOs to more than 20,000 for large MPOs. The larger the MPO, the more likely it is to have complete representation of transit routes and service on the transit network. • Common practice: Trip generation is used to estimate how many trips are expected to be made to and from each TAZ. • Differing practice: Trips for different purposes, such as work, school, shopping, and commercial transport, are estimated. As many as nine trip pur- poses are currently used in MPO models; smaller MPOs are more likely to use fewer purposes. • Common practice: Trip distribution—the process of determining the number of trips between each pair of zones—is accomplished primarily with a gravity model.

OCR for page 46
METROPOLITAN TRAVEL FORECASTING Current Practice and Future Direction 62 • Differing practice: Destination-choice models are used by 11 MPOs for trip distribution. Such a model can take into account differences in circum- stances that influence travelers’ destination choices, which are poorly accounted for in a gravity model. • Common practice: Mode choice is the allocation of trips between auto- mobiles and public transit. Within automobile travel, there is allocation between drivers and passengers; within public transit, there may be allocation among local bus, express bus, and various rail options. • Differing practice: Some MPOs include bicycle and walking trips in their mode-choice model. More than 90 percent of large MPOs reported using a mode-choice model, while 25 percent of small MPOs reported using such a model. • Common practice: Assignment is used to allocate trips to actual routes in the transportation network. • Differing practice: Many smaller MPO regions have little traffic conges- tion and minimal transit service, and MPOs may assign average daily (24-hour) travel. More complex regions with traffic congestion and extensive transit operations model travel by time periods within the day to better account for the effects of congestion on route choice. Among large MPOs, 75 percent assign travel for at least two and as many as five time periods. The committee’s web-based survey and MPO interviews revealed a num- ber of areas for improvement in metropolitan travel forecasting. First, about 50 percent of all MPOs use K-factors or a similar type of adjustment factor in their trip distribution models. Extensive use of K-factors is not recommended because they interfere with a model’s ability to predict future travel. Second, most travel forecasting models are in need of substantial improve- ment to address commercial and freight travel. A lack of data on truck and commercial vehicle travel both within and beyond the metropolitan area is a major issue. Truck trips are modeled in some fashion by about 50 percent of small and medium MPOs and almost 80 percent of large MPOs. Few MPOs have the ability to model all freight movement. Third, models are validated to test their ability to predict future behavior. Validation requires comparing the model output with information other than that used in estimating or calibrating the model. The model output is com- pared with observed travel data, and parameters are adjusted until the out- put falls within an acceptable range of error. Validation is often hampered by

OCR for page 46
Current State of the Practice 63 a lack of independent data sources, and many MPOs validate against much of the same data used to develop the models. Fourth, sensitivity testing is a key to checking the reasonableness of travel forecasts. Such testing is currently done by only a small number of agencies. The committee recommends use of these tests, which vary model inputs and assumptions to determine whether the changes in modeled results are realistic. FTA’s Summit tool can also be used for model checking. Finally, in their responses, 70 percent of MPOs mentioned the most- needed improvements to their modeling processes. The most commonly cited improvement was a tour- or activity-based model. About 20 percent of small and medium MPOs and almost 40 percent of large MPOs reported that they are exploring the idea of replacing their existing model with a tour- or activity-based model. MPO staffs recognize the limitations of their current forecasting pro- cedures. Yet the agencies interviewed reported the following barriers to implementing advanced modeling practices: • A lack of tangible evidence that new procedures would yield forecasts notably better than those produced with currently accepted procedures. • The unavailability of vendor-supplied software needed to implement new techniques. • Resource and staff limitations. Among the agencies surveyed, staff reported as working on travel forecasting ranged from one person part-time with support from the state agency to as many as seven working at least part- time on some aspect of travel forecasting. Another study found that MPOs with a population of less than 500,000 have an average of one full-time travel modeler on staff, while larger agencies average three full-time modelers. The same study found that virtually all MPOs believe it is either difficult or very difficult to hire experienced travel modelers. • Some of the MPOs interviewed believe that their current models are doing an adequate job given the issues MPOs are asked to address. The committee finds that no single approach to travel forecasting or set of travel forecasting procedures is “correct” for all applications or all MPOs. Travel forecasting tools developed and used by an MPO should be appro- priate to the nature of the questions being posed by the constituent jurisdic- tions and the types of analyses being conducted. As the detail required to address a transportation issue increases, the complexity of the analysis tech- niques should also increase. The committee recommends that in their

OCR for page 46
METROPOLITAN TRAVEL FORECASTING Current Practice and Future Direction 64 planning guidance and planning regulations, the U.S. Department of Transportation, FHWA, FTA, and EPA allow MPOs substantial flexibil- ity in their travel demand modeling practices, recognizing that one size does not fit all and that unnecessary technical planning requirements could inhibit innovation and advanced practice. This chapter has presented information on the current state of the prac- tice in metropolitan travel forecasting, including common practice, variations in practice, areas needing improvement, and reported barriers to improve- ment. The next chapter reviews in greater detail the shortcomings of current forecasting models. REFERENCES Abbreviation FHWA Federal Highway Administration Barton-Aschman Associates and Cambridge Systematics. 1997. Model Validation and Reasonableness Checking Manual. Prepared for Travel Model Improvement Program, FHWA. tmip.fhwa.dot.gov/clearinghouse/docs/mvrcm. Chapin, F. S., Jr., and E. J. Kaiser. 1979. Urban Land Use Planning, 3rd ed. University of Illinois Press, Urbana. Deakin, E., and G. Harvey. 1994. A Manual of Regional Transportation Modeling Practice for Air Quality Analysis, Chapter 3. National Association of Regional Councils, Washington, D.C. tmip.fhwa.dot.gov/clearinghouse/docs/airquality/mrtm/ch3.stm. FHWA. 1997. Model Validation and Reasonableness Checking Manual. tmip.fhwa.dot.gov/ clearinghouse/docs/mvrcm/. Giaimo, G. 2001. Traffic Assignment Procedures, Ohio Department of Transportation. www. dot.state.oh.us/urban/menu.htm. Ismart, D. 1990. Calibrating and Adjustment of System Planning Models. FHWA, U.S. Department of Transportation. ntl.bts.gov/DOCS/377CAS.html. Urban Transportation Monitor. 2006. Transportation Demand Modeling and Planning Issues. Vol. 20, No. 20, Nov. 10. VHB. 2007. Determination of the State of the Practice in Metropolitan Travel Forecasting: Findings of the Surveys of Metropolitan Planning Organizations. Transportation Research Board of the National Academies, Washington, D.C. onlinepubs.trb.org/onlinepubs/ reports/VHB-2007-Final.pdf.