Four-step Travel Demand Model Research Articles

Can Transit-Oriented Development Reduce Peak-Hour Congestion?

This paper presents a study of whether transit-oriented development (TOD) helps reduce peak-hour congestion, a topic of ongoing public debate. Applying conventional four-step travel demand modeling techniques, the study simulates traffic outcomes in three TOD scenarios in the Austin, Texas, area, where a commuter rail line is under construction and TOD proposals are being developed. With TOD, the portion of congested roadway in the Austin region is estimated to decrease by nearly 770 lane miles. Daily vehicle miles traveled (VMT) are reduced by 10 to 12 million in the region, or by 3.5 to 4.5 person miles traveled (PMT) per person. This magnitude of congestion relief for peak-hour commuting indicates a potentially significant amount of savings in highway investments through TOD practice. No major modal shifts from driving to transit or non-motorized modes occur in the TOD scenario. Still, PMT for transit is estimated to grow significantly. TOD's role as a congestion relief strategy largely lies in the concentrated development that shortens average trip length and hence generates less VMT and PMT than low-density sprawl. The study results also reveal challenges facing TOD practice: the non-TOD area benefits more than the TOD area, although TOD improves congestion regionwide. Traffic conditions in the TOD area may actually worsen due to the TOD-based concentration of people and jobs. Promoting walking or biking to minimize local driving is thus critical for TOD to succeed.

Analysis of feedback to obtain steady-state solutions in four-step modelling

The four-step travel demand model with feedback has been widely applied in developed cities; in contrast, its applications in developing countries are still very limited. When applying the model with feedback, it is important to figure out what kinds of feedback solutions should be used to obtain the uniquely converged outputs in an efficient way. Therefore, in this study, using the person-trip data collected in Jabodetabek metropolitan area of Indonesia in 2002, a newly developed four-step feedback model has been applied to compare the performances of different feedback solutions. Two major methods, i.e. the Constant Weights (CW) method and the Method of Successive Averages (MSA), are dealt with, where the CW method includes the conventionally applied Direct Feedback (DF) method as a special case. An empirical analysis result reveals that the application of the new feedback model based on the DF method converges most efficiently. It is further found that when using the CW method, the weight on the new matrix should be at least not less than 0.50 to make the feedback procedure converge quickly. Moreover, it is numerically confirmed that the trip matrices from different feedback solutions for one certain feedback process will lead to a unique convergence; in addition, the successive travel impedance matrices from any certain one of the feedback solutions could also converge finally.

ADDING A NEW STEP WITH SPATIAL AUTOCORRELATION TO IMPROVE THE FOUR-STEP TRAVEL DEMAND MODEL WITH FEEDBACK FOR A DEVELOPING CITY

In this paper, the authors develop a log-transformed aggregate multinomial logit model with spatial autocorrelation and incorporate it into a four-step travel demand model with feedback. This integrated travel demand model is named the SPA-STEP model, and using person trip data collected in Beijing, the model's performance is compared with the four-step travel demand model with feedback (called a 4-STEP model)and found to be superior in terms of accuracy.

Neural Network-Based Combined Synthetic Model of Trip Generation and Distribution

Some direct synthetic models for trip generation, distribution and model choice have been previously proposed in order to substitute the traditional four-step sequential travel demand model. However, the previous model based approach has limitations in real world applications where we have to deal with a large number of zones and various parameters effecting travel demand. In this study, we propose a neural network-based combined model for trip generation, distribution and modal spilt. The experimental results showed that the proposed direct demand model using the neural network model is an attractive proposition, particularly in areas where we have to deal with a large number of zones to determine future trips.

From Aggregate Methods to Microsimulation

Two competing approaches to travel demand modeling exist today. The more traditional four-step travel demand models rely on aggregate demographic data at a traffic analysis zone (TAZ) level. Activity-based microsimulation methods use more robust behavioral theory while focusing on individuals and households. Although the vast majority of U.S. metropolitan planning organizations continue to rely on traditional models, many modelers believe that activity-based approaches promise greater predictive capability, more accurate forecasts, and more realistic sensitivity to policy changes. Little work has examined in detail the benefits of activity-based models relative to more traditional approaches. To understand better the trade-offs between these two methodologies, results produced by both were modeled in an Austin, Texas, application. Three scenarios are examined: a base scenario, a scenario with expanded capacity along two key freeways, and a centralized-employment scenario. Results of the analysis revealed several differences in model performance and accuracy in terms of replicating travel survey and traffic count data. Such distinctions largely emerged through differing model assumptions. In general, activity-based models were more sensitive to changes in model inputs, supporting the notion that aggregate models ignore important behavioral distinctions across the population. However, they involved more effort and care in data manipulation, model calibration, and application to mimic behavioral processes better at a finer resolution. Such efforts help ensure that synthetic populations match key criteria and that activity schedules match surveyed behaviors, while being realistic and consistent across household members.

Modeling Network Traffic for Planning Applications in a Small Community

A procedure is developed to model network traffic for planning applications in small and/or medium-sized communities with limited planning resources. The proposed method is based on the theories and assumptions of conventional four-step travel demand models, but the baseline trip table is estimated from existing traffic counts using path flow estimator PFE to render a quick response approach that requires less data than the conventional approach. The proposed modeling approach is suitable for short-range, small area planning applications, such as the evaluation of alternative roadway networks that does not involve significant growth in trip generation patterns. A case study is set up with data from a small community St. Helena, Calif. to demonstrate the proposed approach. An O-D trip table is estimated with PFE from traffic counts. The estimation process is shown to be feasible as the results matched observed data with a satisfactory error bound. The traffic impacts of various scenarios of land use and network changes can be effectively evaluated with the proposed modeling approach.

Developing a Statewide Travel Demand Model from a Person-Based Time Series Household Survey

Balancing the funding needs of data collection and model development is a recurring problem for state departments of transportation and metropolitan planning organizations throughout the country. To address this problem, the Delaware Department of Transportation (DelDOT) and the University of Delaware have developed an innovative and cost-effective person-based time series household survey that has been conducted annually since 1995. The survey diverges from traditional household travel diary surveys in that it is a telephone survey that collects detailed demographic information on all household members but collects trip information from only one randomly selected household member. While the survey costs between 20% and 50% less than a typical household travel diary survey and greatly reduces nonresponse bias, this survey poses two complications. The first is that household trip generation rates must be synthesized if trip productions are to be estimated on the basis of household cross-classification tables; the second is that the data must be normalized to represent a data set for a single year. A description is provided of the procedure by which this survey was used to perform a major update of DelDOT's statewide travel demand model in 2003. The survey was relied on to provide key information in the first three steps of the four-step travel demand modeling process. This survey method has the potential for widespread application, particularly for microsimulation models, which would not require the household synthesis step.

Development of an Empirically Based Area-Type Model

Area type is often used in traditional four-step travel demand models as a predictive variable. Most often it is a factor in mode choice, but it is also commonly used in trip generation as well. Yet a reliable, fore-castable, and measurable definition of area type is generally not documented in most travel demand models. In fact, there is little literature on the subject of predicting area type in the context of travel demand models. This can be troublesome when models are applied to long-range forecasts where significant changes in population and employment result in changes in the general character of an area. With Delphi methods, geographic information system analysis, and discriminant analysis, this study documents the development of an empirically based area-type model in the Tampa Bay region. Because in many ways Tampa Bay's urban form is typical of many cities in the United States, the results should be readily transferable to other urban areas.

Propagation of Uncertainty in Transportation Land Use Models: Investigation of DRAM-EMPAL and UTPP Predictions in Austin, Texas

Propagation of uncertainty in outputs of a standard integrated model of transportation and land use has been examined. Austin-calibrated DRAM-EMPAL predictions of residence and work locations were used as inputs to a UTPP-type four-step travel demand model, and the resulting travel times were fed forward into the future period’s land use models. Covariance in inputs (including model parameters and demographic variables) was accommodated through multivariate Monte Carlo sampling of 200 scenarios. Variances in land use and travel predictions were analyzed over time and as a function of input values. Results indicate that output variations were most sensitive to the exponent of the link performance function, the split of trips between peak and off-peak, and several trip generation and attraction rates. Twenty years in the future, final uncertainty levels (as measured by coefficients of variation) solely due to input and parameter estimation errors are on the order of 38% for total regional peak-period vehicle miles traveled (VMT), 45% for peak-period flows, and 50% and 37% for residential and employment densities, respectively. This means central point estimates of key model outputs are very likely (>30%) to fall 38% to 50% below or above the mean value. In the Austin example, 15% of the region’s 200 simulated peak-period VMT estimates fell below 3.7 million mi per day and 15% exceeded 8.4 million mi. Such substantial variation is solely due to standard model parameter and input uncertainties. Other uncertainty about the future and about human behavior exists and will add more variation.

Uncertainty Propagation in an Integrated Land Use-Transportation Modeling Framework: Output Variation via UrbanSim

The effect of uncertainty in the land use component of a partially integrated land use–transportation modeling system called UrbanSim is examined. Outputs from the land use model (LUM) act as inputs for a traditional four-step travel demand model (TDM), and travel times from the traffic-assignment stage of the TDM are fed forward into the subsequent years’ LUM. This work examines the propagation of uncertainty across model stages as well as at each model stage over time. A factorized design approach is used to model uncertainty in demographic inputs (which include aggregate growth rates and mobility rates) to the land use model, as well as uncertainty in various model parameters. Results suggest that although several model inputs may affect model outputs in the short run, only those inputs that have a cumulative effect are likely to have a significant impact on outputs in the long run. Results also suggest that uncertainty in model outputs may increase for the first few years for which the model is run, as modified inputs send shocks through the urban system. However, the level of uncertainty appears to come down in later years, as households, jobs, and developers respond to changed input conditions.

Transportation Planning and TRANSIMS Microsimulation Model: Preparing for the Transition

The Transportation Analysis and Simulation System (TRANSIMS) was developed as a replacement for the four-step travel demand model. The goal was to provide a simulation system that could analyze issues facing transportation planners such as sustainable development, environmental impacts of proposed projects, and intelligent transportation systems deployment. Because TRANSIMS represents a significant shift from the current state of the practice, the transportation planning community will need to spend significant human and capital resources preparing for the transition. Some insight into these transitional issues is provided on the basis of lessons learned from research on actual calibrated transportation networks in Texas. An overview of TRANSIMS in terms of its main components along with a brief comparison with the four-step model are supplied, followed by a discussion of the data requirements with specific reference to the current data needs associated with the four-step model. The challenges inherent in converting databases that have been developed for four-step models are also discussed on the basis of the conversion of existing traffic network files. Next, the TRANSIMS traffic flow theory is described, and the challenges for transportation planners are illustrated using calibrated networks from two highway corridors in Houston, Texas, and one diamond interchange network in College Station, Texas. Last, the capabilities of TRANSIMS with respect to output analysis are described.

Modeling Framework for Estimating Emissions in Large Urban Areas

A modeling framework has been developed to produce detailed emission inventories for large urban areas. Relationships were developed between the time spent in each driving mode (cruise, acceleration, deceleration, and idle) and basic link characteristics based on simulations of selected real-world surface street networks and freeway sections using the TRAF-NETSIM and INTRAS microscopic models, supplemented by field data. These relationships were then incorporated into a specially written computer program as a postprocessor to the Urban Transportation Planning System type of four-step travel demand models. The integrated model was successfully applied to the 1,120-zone Metropolitan Transportation Commission San Francisco Bay Area network to generate vehicle activity estimates.

Land Use Impacts on Trip Generation Rates

In the conventional four-step travel demand modeling process, the number of trips made by a household is modeled in terms of household size, income, and other sociodemographic variables; any effect of location, land use, or transportation service level is discounted. This is the same as discounting any effect of household accessibility to out-of-home activities as a factor in trip generation (accessibility depending on all three: location, land use, and transportation service level). In contrast to the practice of trip generation, theory tells us that trip rates must vary with accessibility, and some (not all) empirical studies have found that they do. In light of conflicting empirical studies, and the obvious need for more precise and policy-sensitive travel forecasts, this issue is revisited. The independent effects of land use and accessibility variables on household trip rates were tested for using data from Florida travel surveys. It was found that, after controlling for sociodemographic variables, residential density, mixed use, and accessibility do not have significant, independent effects on household trip rates. Conventional trip generation models, which generate person trips by vehicle (not by all modes) and do so without regard to residential location, may not be as bad as one would imagine a priori.

Land Use Impacts on Trip Generation Rates

In the conventional four-step travel demand modeling process, the number of trips made by a household is modeled in terms of household size, income, and other sociodemographic variables; any effect of location, land use, or transportation service level is discounted. This is the same as discounting any effect of household accessibility to out-of-home activities as a factor in trip generation (accessibility depending on all three: location, land use, and transportation service level). In contrast to the practice of trip generation, theory tells us that trip rates must vary with accessibility, and some (not all) empirical studies have found that they do. In light of conflicting empirical studies, and the obvious need for more precise and policy-sensitive travel forecasts, this issue is revisited. The independent effects of land use and accessibility variables on household trip rates were tested for using data from Florida travel surveys. It was found that, after controlling for sociodemographic variables, residential density, mixed use, and accessibility do not have significant, independent effects on household trip rates. Conventional trip generation models, which generate person trips by vehicle (not by all modes) and do so without regard to residential location, may not be as bad as one would imagine a priori.