Cover Image

Not for Sale



View/Hide Left Panel
Click for next page ( 167


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 166
vehicles across the transportation network using multiple modes. The system can forecast how changes in transportation policy or infrastructure might affect individual trips by time of day. In addition, the model is capable of evaluating impacts on different subpopulations, such as minorities and low-income groups, because it simulates individual travelers, taking into account their demographic characteristics. Highway economic requirements system (HERS) model The Highway Economic Requirements Systems (HERS) model allows you to examine the issue of accessibility from a different perspective than that afforded by the exclusive use of travel demand modeling systems. It gives you the opportunity to assess environmental justice concerns based on the actual and forecasted performance of the road segments used most frequently by protected populations. Performance can be measured in terms of average vehicle speed. The recent innovation of making an interface between HERS-ST (State) and TransCAD makes it possible to analyze the travel experience of members of protected populations as they move between traffic analysis zones (TAZs), particularly for trips to work, school, child care facilities, and other social services, and recreation. The focus here is on the use of HERS at the state level because many MPOs will be using it increasingly for routine analysis of travel behavior. METHODS FOR STUDYING ACCESSIBILITY Table 7-1 summarizes the methods for studying accessibility that we present in this chapter. Before conducting an in-depth analysis of how a transportation project might affect accessibility for protected populations, it makes good sense to conduct a preliminary assessment. This assessment should be simple and should use an off-the-shelf method of analysis. The most efficient approach would be to apply the travel demand model already in use within the agency. Method 1. Unmodified transportation demand models As discussed earlier, transportation planning agencies commonly use four-step travel demand (TD) models, which are capable of measuring travel time between TAZs under varying traffic conditions. By comparing travel time estimates before and after modeling a project's characteristics, changes in travel time can be assessed. This method is a useful indicator of a project's impact on trip costs, level of accessibility, and transportation choice. When to use. Standard TD models can be used to obtain a preliminary assessment of changes in travel time or V/C ratios affecting TAZs with relative concentrations of protected populations. This assessment allows you to determine the likelihood that a project would improve or worsen environmental justice within the community by changing the relative accessibility of areas within the activity space of protected populations. Analysis. The starting point for determining the existence of an environmental justice problem lies in identifying those TAZs in which a high proportion of members of protected populations reside. In the analysis, these TAZs will be considered as the origin of travel. Likewise, TAZs that are the common or primary destinations for these special populations are identified, including job 171

OCR for page 166
and shopping centers, social service agencies and providers (including daycare centers), and schools. Table 7-1. Summary of methods for studying accessibility Assessment Appropriate Use Data Expertise Method level uses when needs required 1. Unmodified Screening Estimate travel The project will impact travel Medium Standard travel transportation demand (TD) demand patterns demand modeling; census data demand between TAZs analysis models 2. Adapted Detailed Estimate travel The project will impact travel Medium/ Standard travel transportation demand (TD) demand patterns and High demand modeling; demand between census protected population census data models tracts distribution is uniform within analysis census tracts 3. Advanced Detailed Estimate travel The project will impact travel High Standard travel adapted demand (TD) demand patterns and demand modeling; transportation between census protected population census data models blocks distribution is not uniform analysis within census tracts 4. HERS-ST Screening/ Estimate traffic The project will impact travel Medium HERS-ST model detailed congestion cost for protected populations application; and/or travel TransCAD cost 5. Activity- Detailed Estimate traffic Detailed, dynamic analysis of High Advanced based travel congestion traffic patterns is required or modeling tools simulation and/or travel for large or high-impact and techniques cost projects 6. Transportation Detailed Estimate traffic Detailed, dynamic analysis of High Advanced analysis and congestion traffic patterns is required or modeling tools simulation and/or travel for large or high-impact and techniques system cost projects (TRANSIMS) The transportation demand model is first run with the data that characterize the current transportation system. The results, either travel time or V/C ratios of road links between TAZs, are recorded. The next phase of analysis involves running the model again, but this time with the data that embody the intended transportation project. The focus, as before, is on the times or V/C ratios for travel between principal origin-destination (O-D) pairs by protected and other groups. Data needs, assumptions, and limitations. The data requirements for analyzing differences and changes in travel time and V/C ratios consist of demographic data such as: Household size; Number of persons in household of working age; Household income and availability of vehicles; 172

OCR for page 166
Nonresidential land use data that include number of employees, floor area, and retail sales; Zone data such as population density and distance from central business district or other business centers; and Data about designed highway capacity. Departments of transportation collect most of these data in the course of building TD models. As mentioned previously, TD models are based on simplifying assumptions that do not accurately depict factors such as trip-chaining, and so are limited in their ability to account for the relationships at work in human travel behavior patterns. Moreover, the results may be significantly skewed by estimates of economic activity, land use, and people's propensity to travel, all of which are approximated in the model. Results and their presentation. Whether using travel time or V/C ratios, a comparison is made of the results obtained on trips between origin and destination TAZs for protected and other populations under existing conditions of the network. If the comparison reveals that travel times or V/C ratios related to protected populations are typically greater than for other groups, it may be concluded that low-income and minority groups are most likely carrying a disproportionate burden of transportation-related costs; and therefore an environmental justice problem may exist. Of course, if there is no significant difference, there is probably no environmental justice problem. One note of caution in presenting the results: because of the aggregate level at which the analysis is conducted and the difficulty of definitively knowing which road segment(s) were used, you would only be able to say, for example, "About 95 percent of trips from zone 1 to zone 2, representing protected population groups, experienced a reduced/increased travel time or V/C ratio." This percentage is based on the proportion of the population in zone 1 who are members of protected groups. Assessment. Caution should always be used in drawing conclusions from the results of this type of analysis because of the underlying simplifying assumptions with respect to the factors that influence the choice of mode, the impact of various public policies on people's travel patterns, and the relationship between land use and mode choice, among others. In short, the results should be viewed as crude and should be interpreted as indicating only the likelihood of an environmental justice problem even when the magnitude of the changes in travel time or V/C ratios is significant. In such a case, a more detailed analysis is required. Method 2. Adaptation of transportation demand models By making use of TAZs, TD models allow you to take advantage of the demographic data contained within them to enhance the analysis. TAZs typically are aggregations of census tracts and may be redefined based on the presence of protected populations within zones before the model is run. When to use. These models are appropriate when the preliminary analysis indicates that a more accurate method of estimating changes in travel-time costs is needed. Though more costly in terms of time necessary to redefine TAZs, this remains a relatively inexpensive method because 173

OCR for page 166
it does not require new computer software. This method is suitable for small- to medium-sized projects. Analysis. The first step in redefining TAZs is to identify which zones contain the residences of members of protected groups. The second step is to dissect TAZs into units that are smaller groups of census tracts based on the relative presence or absence of protected and other populations. TAZs should be configured so as to be as homogenous as possible in terms of income and/or race. The third step is to identify the destination TAZs to which households of protected populations are routinely attracted, such as job and shopping centers, social service agencies and providers (including daycare centers), and schools.1 The transportation demand model is first run with the data that characterize the current transportation system and the travel times on road links between origin and destination TAZs. The next phase of analysis involves running the model again, but this time with the data that characterize the intended transportation project. As before, a record is made of travel times between principal origins and destinations by the respective kinds of groups. Data needs, assumptions, and limitations. The data required for this analysis are the same as for other routine analyses using the TD model. This includes demographic, nonresidential land use, zone, and road data as mentioned in the description of the preliminary assessment method. Again, departments of transportation routinely collect most of these data in the course of building TD models. The limitations are the same as with the unmodified TD models; the simplifying assumptions that the models are based on leave out factors such as trip-chaining and so are limited in their ability to account for relationships among travel behavior patterns. As with unmodified models, results may be skewed by estimates of economic activity, land use, and people's propensity to travel, all of which influence the model. Estimating travel-time savings is a challenge because of the significant issues involved in attaching economic value to travel time. Researchers have yet to agree on the following: What fraction of the wage rate should be used for work-related travel. What fraction of the work-related travel rate should be used for personal or nonwork- related travel. What fraction of a driver's hourly time value should be assigned to passengers in the vehicle. Whether a lower time value should be used for commuting trips that are shorter than the common travel-time budget (i.e., the amount of time people are willing to spend journeying to and from work) and a higher value for the time increments that exceed this budget. 1 TIGER/Line data based on the 2000 census are currently available from the U.S. Census Bureau regarding the location of employment centers (including shopping and major retail centers; industrial buildings/parks; office complexes/parks; government centers; and major amusement centers), educational and religious institutions, and transportation terminals. 174

OCR for page 166
Whether the same time value should be applied for very short periods of time saved (e.g., 30 seconds) as for longer periods (e.g., over 5 minutes). How to take into account variation in time en route and hence unreliability of arrival time. How to include changes in travel time for pedestrians, cyclists, and others using nonmotorized transportation modes. Completely overcoming these challenges is beyond the scope of this guidebook. The choice of method used to estimate travel-time savings is influenced primarily by how much detail is perceived to be necessary to make a decision about a project alternative. For many small projects, the preliminary assessment described earlier using travel demand models will prove adequate. For more extensive projects, or those where a reasonably strong likelihood exists that the benefits and costs of the project may raise questions of equity, a more accurate and detailed estimation of travel-time savings is needed. Results and their presentation. The travel times obtained on trips between origin and destination TAZs for protected populations are compared with those for other populations under existing conditions of the network. The analysis is then repeated with the transportation improvement in place. As in the preliminary assessment, if the comparison reveals that travel times of protected populations tend to be significantly higher than those of other groups and that the project would do little to reduce the disparity or even worsen it, the conclusion may be drawn that an environmental justice problem currently exists. As in the case of the preliminary assessment, the results should be presented in terms of trips originating in each applicable TAZ in comparison with all TAZs. Assessment. Although the results acquired using this method reflect a greater degree of accuracy than the preliminary analysis, it too is quite aggregate. As such, it can provide a general sense of the extent to which travel times to important destinations would improve or worsen for protected populations. It also can be used to compare such changes with those of travelers in general. If unfavorable results emerge relative to environmental justice, more detailed analyses will be required. Method 3. More advanced adaptation of transportation demand models This method is an advance on the previous one and disaggregates the applicable TAZs using census-block-group data instead of tract data. As before, the TAZs are redefined based on the presence of protected populations within zones before the model is used to determine travel times between analysis zones. When to use. This method is appropriate when a more accurate assessment of changes in travel time is needed than that afforded by the preliminary analysis or tract-level analysis. This method will be more costly because the tract data have to be replaced by block-group data. The method is suitable for small- to large-scale projects and is particularly useful for achieving relatively high accuracy in determining the probability of an environmental justice problem using TD models. 175

OCR for page 166
Analysis. To redefine and prepare the TAZs for use, the four steps outlined in the previous method are followed, but with block group data as the basis. Care must be taken to ensure that the TAZ geometry matches the network geometry. As before, the TD model is first run with the data that characterize the current transportation system and the travel times on road links between origin and destination TAZs, then run again with the data that characterize the intended transportation project. Data needs, assumptions, and limitations. The data required for this analysis are essentially the same as for other routine analyses using the TD model with the exception that the demographic and nonresidential land use data are at the census block group level instead of the census tract level. Census block group data, including those on population density, can be downloaded from the U.S. Census Bureau Web site; zone and road data are the same as mentioned in the description of the previous assessment method. Departments of transportation routinely collect data for modeling in the course of building TD models. The same limitations described previously affect the reliability of results for this approach. Results and their presentation. The travel times obtained for the trips between origin and destination TAZs for protected populations are compared with those from other populations under the existing network conditions. As in the preliminary assessment, if the comparison reveals that travel times of protected populations are consistently significantly higher than those for other groups, then low-income and minority groups are likely carrying a disproportionate burden of travel time costs; and therefore an environmental justice problem exists. Assessment. These results are about the most accurate that can be obtained using a traditional TD model. Nevertheless, they still represent a probability of occurrence, even though we may express it with greater confidence. This level of confidence may be adequate for most small- to medium-sized projects, but larger projects and highly sensitive projects may require the use of a of model that provides an even higher level of accuracy. Method 4. HERS-ST model Aggregate models, such as the HERS model, often are less expensive to use than more disaggregate models. Much of the necessary data for these models is routinely collected and updated by states and maintained by the FHWA in the Highway Performance Monitoring System (HPMS) database.2 The HPMS database does not, however, routinely include data on urban and rural local roads, nor on rural minor collectors, as indicated in Figure 7-1. These categories of roads are important in assessing environmental justice concerns because low-income populations and minority populations are likely to use them, and their performance would have an impact on travel costs for these protected populations. The data problem is partially resolved by HERS-ST, which treats the performance of these roads in terms of changes 2 The FHWA and the states, beginning in 1978, jointly developed and implemented a continuous data collection system called the Highway Performance Monitoring System (HPMS). Currently, the HPMS contains more than 110,000 sample sections, the most comprehensive nationwide data system available regarding the physical condition and usage of the nation's transportation infrastructure. 176

OCR for page 166
in vehicle speed or level of congestion in one of the submodels. Figure 7-2 is a graphic representation of the HERS model. Rural roads Arterials Collectors Local roads Rural interstates Rural major collectors Rural local roads Rural other Rural minor collectors principal arterials Rural minor arterials Urban roads Arterials Collectors Local roads Urban interstates Urban collectors Urban local roads KEY Urban freeways and FHWA uses HERS expressways to project needs for these highways Urban other principal arterials FHWA uses non-HERS methods to project needs for these highways Urban minor arterials Figure 7-1. FHWA's road classification system Source: U.S. General Accounting Office 2000. It is important to note that an interface can be created between HERS-ST and TransCAD. This enables you to identify those segments of the road network, including the urban and rural roads that are most likely to be used by members of protected populations. These segments often represent the probable routes between the origin TAZs and destination TAZs for the most essential trips made by such populations. The TAZs may be defined using either census tract or block-group data, but using the latter provides more detailed and accurate information, as noted earlier. When to use. The HERS-ST model is best suited to estimating changes in average vehicle speed or levels of congestion, as it takes into account factors such as traffic volume, pavement condition, and lane width. When this capability is coupled with the GIS-based TransCAD, the model becomes a reasonably accurate measure of how the existing road network affects protected populations and how the intended improvements will alter that. Of course, additional costs are incurred with this increased proficiency and must be weighed against the size, cost allocation and social and political significance of the project. 177

OCR for page 166
Travel forecast Estimate costs Evaluate Forecast (travel time, and select pavement operating, safety, improvements for condition emissions, etc.) implementation Forecast vehicle speed Estimate improvement costs Figure 7-2. Simplified representation of the HERS model Source: U.S. General Accounting Office 2000. Analysis. In the data analysis of urban and rural local roads and rural minor collectors, HERS- ST clumps together lower rural classified roads with rural major collectors and lower classified urban roads with urban minor collectors. In this approach, the derived output information on these roads is separated from the rest prior to reporting. Another approach would be to analyze the lower classified roads separately from the rest of the system once the induced deficiency and cost data are appropriately adjusted prior to conducting the analysis. As in the methods using travel demand models, defining the TAZs at the block-group level is an option, but if the choice is made to redefine them, every effort must also be made to ensure that each TAZ's geometry matches that of the road network. In addition, the interface with TransCAD provides data on the volume of traffic traveling on each road segment, which it tags with a unique identifier. This identifier is what makes it possible for TransCAD to interface with HERS-ST, and the traffic volumes relayed to the HERS model via this mechanism are incorporated into its computations. HERS-ST consists of a number of submodels with the output of one becoming the input for another (U.S. General Accounting Office 2000). The travel forecast submodel projects traffic growth, and it utilizes current traffic volume data, along with data related to the cost of travel--travel time, safety, and vehicle operating costs--and combines them with the state's projection of traffic growth and with a measure of the price elasticity of travel demand. The 178

OCR for page 166
output of this sub-model is the input to the pavement condition submodel and, subsequently, the vehicle speed sub-model. Thus, the change in average vehicle speed is an important measure because it not only reflects changes in the volume of traffic and associated congestion, but also the quality of the road surface, which is subject to wear-and-tear effects. Furthermore, changes in average speed, which can be measured for each road segment, including those most frequently used by protected populations, are key to assessing whether an environmental justice concern exists. The model begins by assessing the current condition of the highway segments in the data sample. Average speeds under existing conditions may be first compared with those for segments used by members of other groups, followed by a similar comparison of forecasted values that reflect expected changes that would result from the intended project. Data needs, assumptions, and limitations. HERS-ST requires data on average annual daily traffic (AADT), highway capacity, pavement condition, and lane width. Most of these data are also readily available from the HPMS database. Moreover, HERS-ST provides the option to substitute more accurate local data, where feasible. Additional effort will be needed to gather more specific data on urban and rural local roads, as well as on rural minor collectors. If the decision is made to redefine the TAZs in TransCAD using data at the block group level, the applicable data must be obtained from census files. This model assumes that the forecast for each road segment represents the level of use that will occur if a constant level of service is maintained on the segment. There is also an implicit assumption that the model captures the net effect of all changes in the transportation network and the economy through its assumed price elasticity of travel demand. In addition, the model assumes that all roads face the same weather conditions. Some of these assumptions give rise to the model's limitations, which are stated below (U.S. General Accounting Office 2001). Because it analyzes each road segment independently rather than the entire network as a whole, it does not completely account for the interrelationships between all segments and different transportation modes (e.g., how traffic is redistributed as improvements are made). It does not fully account for the uncertainties associated with its methods, data, and assumptions. For example, the model uses the price elasticity of demand for travel to incorporate information on how changes in vehicle user costs affect travel; there is thus an implicit assumption that the model captures the net effect of all cost changes in the transportation network. The overall economy is thus assumed to remain constant because its condition affects consumer choice. The accuracy of estimates generated by HERS-ST is uncertain because the model uses data that vary in quality. For example, the state-supplied data on pavement roughness vary significantly in quality because different states use different devices and approaches to measure it. In addition, some data used in the model, such as pavement resurfacing costs, are usually outdated. Users may exercise the option of using more accurate, local construction data. The model uses information to project the future condition of the road pavement, which does not take full account of environmental conditions that affect highways. For example, 179

OCR for page 166
the assumption is made that all road segments experience freezing and thawing conditions, while this is not the case in the warmest parts of the country. Results and their presentation. Changes in average speed on the various road segments are generated by the model and may be displayed in tabular form or reflected in maps generated by TransCAD, which receives the output of HERS-ST via a routing system and dynamic segmentation process. Maps can also be used to display the location of the intended improvements to in-house staff or to policymakers. Assessment. The HERS-ST model differs from the national-level HERS model in significant ways that can be an asset to the analyst. First, it allows the user to override some or all of the improvement decisions generated by the model. For example, users can specify the type of improvement they see fit for any segment of the highway in any funding period, whether or not the specified improvement is economical. FWHA contends that this capability gives users the opportunity to apply specific knowledge of a particular condition. Second, as implied earlier, the HERS-ST can analyze more classes of roads and provide a higher level of detail in its results with respect to every segment analyzed. Third, this model permits the substitution of more relevant state data for national-level data so that local conditions may be modeled more accurately. Finally, this model provides the user with the option of analyzing a statistical sampling of highways drawn from the HPMS database or analyzing all segments of the state's road network. Method 5. Activity-based travel simulation A new set of travel forecasting and analysis procedures based on travelers' daily activity patterns is being encouraged under the TMIP. TMIP is an attempt to satisfy the need for more accurate and sensitive travel forecasts and to facilitate better-informed decision making on transportation matters. Activity-based simulation models of human activity and travel behavior contain several modules. These modules enable the researcher to combine stated and revealed preference data along with baseline activity patterns, network and land-use data, and socio-economic and demographic data. Not only does this type of model check the network data for logical consistency and missing information, it also assesses whether a modified travel pattern is feasible, based on a human adaptation and learning module. Behavioral responses are captured by the statistics accumulator within the evaluation module, which provides descriptive and frequency statistics about vehicle miles traveled, number of trips by mode and time of day, number of stops by purpose, trip chains, vehicle occupancy, and travel times by trip purpose, among other classifications. Because this micro-simulation approach does not rely upon over-simplifying assumptions, it does not reduce the complexity and realism of the response and adaptation patterns of the travelers being modeled. As a result, the model is capable of providing highly accurate analysis of travel-time savings compared to most currently available models. As might be expected, this improved accuracy comes at a relatively higher cost. Further detail on this sort of model may be found in RDC, Inc. (1995). When to use. Activity-based simulation is most appropriate when the project to be implemented is costly. It also is suitable when a relatively high level of precision is needed to determine the 180

OCR for page 166
travel-time savings that would occur in various areas of the community if the project were implemented. Analysis. One significant advantage of this type of model is that it permits a dynamic, longitudinal analysis of travel behavior, as opposed to the static, cross-sectional analysis afforded by the traditional four-step demand models. This means not only that behavior is examined over a continuous time frame but also that impacts originating within and outside of the transportation system can all be evaluated together. As a consequence, people's entire daily itinerary is the focus of analysis, rather than individual trips. In addition, whereas evaluation has traditionally been based on capacity and level of service, this approach evaluates the impacts of transportation policy measures and projects based on time-use utility, which is represented by the daily time- use patterns of the target population. Data needs, assumptions, and limitations. Because this type of model focuses on the entire daily itinerary of travelers, it requires considerably more data than traditional models, a factor that contributes significantly to its running costs. In addition, it uses response data that must be gathered by means of a survey. Thus, the magnitude of potential benefits from its use should be carefully weighed against the costs. Nonetheless, many of the data requirements are similar to those of four-step models and may be obtained from most MPOs. These include data on TAZs, including network system and travel time, mode choice, trip distribution, and land use inventory. Demographic and socio-economic data by TAZ, such as household size, vehicle ownership, income, and race (white and nonwhite categories) are also needed and may be obtained from the Census Bureau. Original data needed include information from trip diaries for the revealed preference analysis. If the project being evaluated requires a change in TDM strategies, the type and characteristics of these strategies can also be input. To do this, however, a survey must be designed to collect stated preference data in the form of potential responses to the anticipated impacts or policy changes. The same survey can also be used to gather information to complement that received from trip diaries, such as tradeoffs between parking costs and walking distance. Activity-based simulation models are based upon the assumption that travelers engage in "satisficing" behavior (making appropriate choices with limited information), as opposed to always making optimal decisions or decisions that always maximize their utility in the purest understanding of the concept, as is typically assumed in traditional models. This satisficing assumption more appropriately reflects the reality of day-to-day living in a world where individual travelers do not have perfect information of events and concerns that affect their decision making. In other words, most travelers often make decisions with the intention of "making do" with the current circumstance, and this will be reflected in random or stochastic travel behavior because factors and constraints will affect persons differently. Moreover, the model assumes that the marginal utilities of travel vary across people, modes, and environmental conditions encountered, and that route choice preferences vary according to socio-economic characteristics and perceptions of individuals. Both of the latter assumptions impact the individual's valuation of time and allow for the differential analysis of travel-time savings across income and racial groups. The model's limitations derive from the fact that it is still in development. 181

OCR for page 166
Results and their presentation. Activity-based models can be configured to generate descriptive statistics, and they are also capable of carrying out statistical tests and providing statistical analyses in the form of response distributions. They can also cross-classify these response distributions against socio-economic and demographic variables, which allows a level of disaggregation that facilitates the application and assessment of environmental justice criteria. Detailed results can be presented in tabular and graphical formats that are easy to comprehend. Assessment. This form of model has many capabilities and has the potential to provide accurate analysis of investment and demand management policies. The level of financial and technical support that the FHWA has devoted to this process reflects its commitment to fully developing activity-based simulation models and making them widely available. A major benefit of this form of model is its potential for giving relatively accurate valuations of travel time savings. A certain expertise will be required to design a survey instrument capable of eliciting the necessary information. The goal should be to make questions as simple as possible and yet clear enough to obtain the required data. One approach to such a survey design is that used in the Adaptive Stated Preference survey instrument (Richardson 2001, p. 13). Of course, all survey questions should be pilot-tested to ascertain whether the target population is able to comprehend them fully. Realistically, it is highly unlikely that an agency would develop so ambitious a model solely for assessing environmental justice implications of a project. Rather, this type of model is most likely to be developed to meet an agency's general needs for travel demand analysis. In such a case, it is feasible to enhance the modeling effort to provide a first-rate capability to evaluate the environmental justice effects of almost any significant transportation project. Method 6. The Transportation Analysis and Simulation System The TRANSIMS is an activity-based travel demand model that functions as six integrated modules, along with a feedback selector/iteration database. The feedback mechanism is the primary modeling tool as it functions to achieve consistency among the various computational modules (Los Alamos National Laboratory and Price Waterhouse Coopers 2002, p.3). This mechanism is critical to simulating decision/choice responses of individuals to events such as accidents, closure of a segment of highway, or interruption of transit service that occur directly within the transportation system; it is also helpful in evaluating policy alternatives that affect the use of an entire transportation system, even though the policy may be targeting a particular segment of the system. TRANSIMS simulates the movement of individuals and vehicles across the transportation network and can also forecast how changes in transportation policy or infrastructure might affect individual trips by time of day. The results of the simulation are aggregated only after the activities have been set, the trips routed, and the entire set of individual trips simulated in the presence of all other travelers. Because of this capability, TRANSIMS promises a substantially expanded scope of analysis along with improved analytical ability, particularly when evaluating the impacts of potential transportation projects on different populations. One of the first requirements to make this model operational is the creation of a detailed network that represents the future transportation infrastructure. Infrastructure includes signs, signals, streets, highways, and transit information, along with information about where activities (e.g., residential, commercial, and recreational) will occur and where parking lots will be located. This 182

OCR for page 166
network supplies data to all the modules. Figure 7-3 depicts the primary modules in the center row; each is dependent on external data inflows, which are shown on the top line. The data produced by each module, indicated in the bottom row, becomes the input for other modules. The population synthesizer creates a synthetic population of households and individuals that are distributed both geographically and demographically according to the input data related to the metropolitan area under study. Vehicles are also assigned to households and individuals according to the input data. This synthetic population then interacts with the other modules. The first of these is the activity generator, where an activity list is constructed for each individual in a household by matching his or her demographics against information gathered from household travel and activity surveys. At this point, the synthetic population has places to go, and the means of going to those places are supplied by the route planner module. This module computes the fastest route to each activity by each individual based on the activity information and trip plans supplied from travel diaries and stated choices of transportation mode. Mode choice is also accomplished within the route planner module using external functions, such as logit and travel cost functions. In addition, shared rides, in which the passenger and the driver are from different households, are accounted for by this module as long as information related to the dependency is recorded with other household information in the population synthesizer module. Input files Census Population Traveler Transit Network Air quality Arbitrary forecast survey surveys box data Modules Population Activity Route Traffic micro- Emissions Output synthesizer generator planner simulator estimator visualizer Input & output files Synthetic Activity Vehicle Traveler Simulation Emissions MODELS3 population plans output inventory database Figure 7-3. The TRANSIMS architecture from the perspective of data flow Source: Los Alamos National Laboratory and Price Waterhouse Coopers 2002. The traffic microsimulator module processes the output of the aforementioned modules, causing the synthetic individuals to interact with one another and realistic features of the traveling environment. This module simulates the movement of individuals throughout the network, including their use of private vehicles and public transportation, and the high level of realism in the simulation is directed by the selector/iteration database, which utilizes an iterative process and feedback mechanism. The next step in the process, calculating vehicle emissions, is not 183

OCR for page 166
pertinent to our analysis of transportation user effects; but the one following, the creation of a visual representation of the model's output, is of particular interest. When to use. This model is most useful when the situation requires a high degree of accurate information regarding the impact of a proposed project on protected populations. Obtaining such accuracy is relatively more expensive compared to other methods, but because this type and standard of modeling is in demand due to federal requirements for other information and decision making, the use of TRANSIMS may become widespread in the not-too-distant future. Analysis. The special ability of this model to simulate the travel of an individual over an extended time period, beyond peak periods and for travel other than commutes to work and other basic activities, means that the potential for more thorough comparative analysis is greatly enhanced. For the purpose of assessing environmental justice, the most critical stage is gathering and inputting accurate demographic data in the population synthesizer. Such information determines how the individual is going to travel across the simulated transportation network and, hence, the quality of the model's output. While current use of TRANSIMS does not require demographic data about race, this must be included to facilitate environmental justice assessment. Furthermore, the matter of trip-chaining, which often is an important part of low-income people's travel itinerary, is dealt with explicitly by the model within the activity generation module--a discrete choice-based model that generates trip chains along with activity locations using the data related to the synthesized householder's travel itinerary and domicile location. The operation of the traffic microsimulator module gives this model an important advantage over traditional demand models because it is capable of simulating multiple travelers per vehicle and multiple trips per traveler, both factors that are fairly common to low-income and minority travel routines. Another key feature that facilitates determination as to whether an environmental justice problem will exist is the output visualizer module. The module allows the user to select for display any data value of interest that can be drawn on any link of any size on a given network. Because TRANSIMS is a completely disaggregate system, much care is required in calibrating and applying mode choice. Data needs, assumptions, and limitations. Much detailed information related to individual travel is required by this model, so building the database can take considerable time. In addition, the data need to be location-specific for the model to be most useful, so considerable data have to be gathered at the local and regional level. The bulk of the data is keyed into the population synthesizer and consists of geographic and demographic information at both census tract and block-group levels. TIGER/Line layouts of census tracts and block groups make up the geographic data, while summary tables (STF-3A), and public use microdata area (PUMA) samples are obtained from the U.S. Census Bureau. TIGER/Line data are used to build the transportation network, which must be able to reflect the location of workplaces, shops, stores, schools, daycare and recreation centers, hospitals, and other areas identified from household activity surveys, along with parking lots. Other network data include number of lanes, streets, freeways, highways, ramps, turn pocket lanes, and intersections (with and without traffic signals). It is important to stress that in order for the model to provide a predictive output, a 184

OCR for page 166
forecast marginal demographic file consisting of race, household size, income and age data based on census tract and block-group data must first be generated before it is keyed into the synthesizer. The master area block level equivalency/geographic correspondence engine (MABLE/Geocorr)3 is also utilized to generate a link between the PUMA samples and census blocks. Because the population synthesizer assigns individuals to activity locations, household travel and activity surveys (including travel diaries) are important sources of information about the types of activities individuals engage in (e.g., work, school, and shopping) as well as the start, stop, and travel times associated with them. Trip-chaining activities, including stop and start times, must also be included for origin-destination travel because the traffic microsimulator chains together several legs to form a trip. In fact, data on network travel times and activity locations are essential elements that allow the model to select a likely location for each activity, and each location's relative attractiveness is computed using criteria such as the number of retail employees or the amount of retail-store floor space. Unlike conventional travel demand models, TRANSIMS is not merely concerned about peak- hour activity, so travel diaries should cover an extended period of about a month to ensure that a reliable trend can be established. Information on whether individuals walk, use private vehicles or transit, or use any other mode of transportation is also gathered from the surveys. Transit data, such as route paths, terminals, and schedule of stops, are part of the required network data. This model assumes that the traveler always makes rational choices and so takes the route and uses the mode of transportation, including walking, that yields the shortest time between two points, while taking into account any situation or obstacle that may cause delay on any of the possible routes. For example, if the input information is that the traveler walks to and from work, the model will subsequently compute all work-related activity as accomplished by walking unless programmed otherwise. In other words, the travel cost function of the synthetic traveler is based on a predetermined, user-defined cost structure obtained from the survey data. On the other hand, if the information reflects that the individual drives his own vehicle and sometimes walks part of the distance, or takes transit, the model assumes that if the delay using motorized transportation would cause the time traveled between two points to be longer than it would take to walk, then that individual would, in fact, decide to walk. Though the model structure may not reflect all decisions made in reality, this should not be regarded as a serious limitation because it is applied without bias. Like other methods of forecasting, the assumption is made that once the characteristics of the surrounding infrastructure are recorded and there is no subsequent change, the forecast year behavior of the traveler is the same as that captured in the base year. Any change in infrastructure, such as changes to a roadway or to the level of transit service, and the area's population are assumed to be reflected in the base-year calibration function. Changes in such things as travel time, transit fares, and parking costs are assumed to affect modal choice. The intensity of the impact on an individual traveler depends on the nature of the demographic data input obtained from the Census Bureau 3 MABLE/Geocorr supports data maintained by the Center for International Earth Science Information Network (CIESIN) at Columbia University. See http://plue/sedac.ciesin.org/plue/geocorr. 185