Office of Operations
21st Century Operations Using 21st Century Technologies
Table of Contents

Traffic Analysis Toolbox Volume XIV: Guidebook on the Utilization of Dynamic Traffic Assignment in Modeling

4.0 DTA Modeling Framework

4.1 DTA Modeling Process

Transportation modeling using DTA should follow the seven-step modeling process that has been defined in Traffic Analysis Toolbox Volume III. This process is a sound modeling workflow and applies to DTA methods very well. Figure 4.1 presents a summary of the modeling work flow starting with scoping and ending with alternatives analysis.

Figure 4.1 DTA Modeling Process

Figure 4.1 depicts the overall modeling process for a Dynamic Traffic Assignment model, including six major steps.

Source: Traffic Analysis Toolbox Volume III. Modified by Cambridge Systematics, Inc.

4.2 Scoping

Developing a scope for a DTA modeling project is the first step in the process. The initial stages of developing the scope include identification of the project objectives and identifying stakeholders and their key questions. Based on the project objectives and stakeholder issues the purpose of the model can be clarified by asking: “What questions do we want to answer with this model?”

The model scope includes the selection of the modeling tool, selection of study area, model calibration requirements, data collection plan, and the alternatives to be analyzed. Preparing a Methods and Assumptions Document including a statistical evaluation of existing available data could assist in developing an accurate level of effort for the modeling effort and is highly recommended. Finally, a level of effort and cost for the project is prepared.

Project and Stakeholder Objectives

Project objectives should be discussed with the stakeholders at the outset of a project. The objectives of a project should be converted into tangible items such as the following:

“We will study the I-62 corridor for three different Transportation System Management (TSM) alternatives. The impact of TSM alternatives will be studied on I-62 from Crossroad “X” to Crossroad “Y.” The impacts studied will include but will not be limited to queuing and weaving conflicts. The effects of TSM on adjacent arterials will be examined within one mile of I-62.”

With this information discussed and documented the appropriate modeling tool and model limits can be determined.

Model Limits

Setting model boundary limits is necessary for both the geographic area (spatial limits) and time periods (temporal limits) to be analyzed. The limits for a DTA model need to extend beyond the immediate project area and must include enough of the roadway network that travel patterns in and out of the area can be represented realistically. Some of these considerations include the following:

  • Geographic Limits:
    • Must include the significant alternative routes in the area being studied;
    • Should consider natural barriers such as water crossings to minimize the number of external stations;
    • External stations should be located where field data can be easily used to compare; and
    • The extent of queued and congested vehicles should be contained within the limits of the model.
  • Temporal Limits:
    • Duration of the model should allow for the build-up of congestion (warm up period) and the dissipation of congestion (cool down period); and
    • Time intervals should take into account data availability and the level of analysis detail required to answer the questions in the study.
  • Multi-resolution Modeling Considerations:
    • If multiple models with different resolutions are used in the same study, each model resolution may have different boundaries;
    • Different model boundaries should be based on hierarchy of the coverage and resolution associated with teach model. Generally,
    • Regional demand models cover the largest area;
    • Mesoscopic models would cover a smaller area than the regional demand models; and
    • Microscopic simulation models (most detail) would cover the smallest area.
  • Depending on software capabilities and project purposes, some or all of the different model resolutions could have the same model boundaries.

Additional resources for model limit considerations can be found in:

Figure 4.2 presents an example of how the geographic model limits can be developed. The project in this example includes the facilities labeled as segments 1 and 2. Based on the need to analyze these facilities, the modeling boundaries are two-fold. First, the regional travel demand model will be “cut” into a focused area sub model, and the focus area will be used to provide origin-destination information that will be fed into a mesoscopic DTA model. The second part is the DTA model that covers a smaller area within the regional focus area model, but one that is still large enough to include the segments that are being studied.

Figure 4.2 Sample Model Limits

Figure 4.2 is a map depicting an example modeling project where the primary subject of the project is one highway. However the model limits for the Focus Sub-Area Model are larger and a Dynamic Traffic Assignment Sub-Area Model larger still.

Source: Cambridge Systematics, Inc., ©2012 Google.

Tool Selection

The selection of the appropriate traffic analysis tool can be facilitated by the process discussed in the Traffic Analysis Toolbox Volume II. Volume II presents a detailed approach for determining the type of traffic analysis tool based on a number of criteria. Figure 4.3 presents a summary of the criteria used in tool selection. There is also a spreadsheet tool that accompanies that guidebook.

The selection of DTA methods should be based on the needs of the project and the objectives of the stakeholders. DTA methods are preferred for a project where multiple routes are being modeled and the outcomes that are being tested include traffic diversion related to either congestion and/or ITS type strategies.

Figure 4.3 Criteria for Selecting a Traffic Analysis Tool Figure

Figure 4.3 depicts the types of questions that should be considered when selecting the appropriate analysis tools. This includes analysis context and seven other areas that should be considered. The seven areas are in separate boxes with labels one through seven from left to right.

Source: FHWA – Traffic Analysis Toolbox Volume II.

Level of Effort

Based on the analysis tool, model limits, modeling requirements, and data collection needs, a work plan and corresponding level of effort (LOE) should be prepared. This LOE should anticipate the amount of time to be spent on model building and calibration.

Additional guidance on the question, “How much will it take to create a DTA model?” is difficult to provide at this time and within the context of this guidebook. The possibilities of different model boundaries, different model resolutions used in the same project, and the ranges of types of alternatives are too numerous to settle into a generic reference quantitative summary. There are resource documents for quantifying the level of effort for applying travel demand models and microsimulation models. These sources can be found at:

  • Travel Demand Modeling Association of Metropolitan Planning Organizations – Advanced Travel Modeling Study, July 2011.
  • Rossi et al., “Deciding on Moving to Activity-Based Models (or Not),” TRB 2009.
  • TMIP Online – Introduction to Travel Demand Forecasting Self Instructional CD-ROM.
  • Microsimulation Modeling, Guidance on the Level of Effort Required to Conduct Traffic Analysis Report, Number HRT-13-026.

There is no similar guidance document for mesoscopic modeling. Mesoscopic modeling is relatively new and a review of the practice and the level of effort required has not been conducted or identified as of yet.

The development of these models represents an investment by the sponsoring agency. The value of this investment should extend beyond the immediate project to include the potential re-use of the model for another phase of the project or a separate project. During the scoping phase flexibility and longevity of the model should be considered.

4.3 Data Collection

Data collection for models using DTA must include data from multiple days. The different days that data are collected should have some similarity between them primarily in terms of travel demand and spatial/temporal travel patterns. Also, if the analysis is to test impacts of an incident or inclement weather, it may be necessary to collect data from days when these occurrences happen.

Combining data from different days into one homogenous data set can cause difficulties in modeling, such as creating unrealistic congestion or not creating congestion where it typically occurs. Data should be collected over multiple days for capturing variability; however, each data day should be kept separate. Data requirements for DTA are covered in Chapter 5.

4.4 Base Model Building

Building a sound base model is necessary to create the foundation of the entire modeling process. One goal of base model development is to produce a model that is verifiable, reproducible, and accurate. Base model development tasks and data should be well documented and transparent. There are two components to base model building: supply and demand. The supply side for DTA modeling is similar to that for other types of modeling. Depending on the model type (macro, meso, or micro) the level of detail of representing geometry and traffic control may vary. The primary focus of base model building in this guidebook is the demand-side and the development of origin-destination matrices.

4.5 Error and Model Validity Checking

Checking for mistakes in coding is an essential component of the modeling framework. Errors of omission and transposition of information are common occurrences. Having an independent check of the model inputs is a step that can alleviate wasted effort further into the modeling process. There is no perfect error checking system so a analysis team must always be vigilant about the possibility that errors could still be in the model; however, attempting to render the model mistake free will save time.

Checking for model “validity” is a different activity than error checking, although the validity checking process may be useful in revealing errors as well. The validity checks discussed in this guidebook are a series of stress tests and diagnostic testing steps to help ensure that the models in fact can meet the objectives of the project. Often these stress tests are conducted as part of the model calibration effort. In this guidebook, stress testing is discussed in Chapter 7, and Error and Model Validity Checking in Chapter 8.

4.6 Model Calibration

Calibration (often, travel demand modelers refer to this definition of “model calibration” as “model validation”) is a process whereby the base model is adjusted to ensure that the model performance measures are realistic and statistically representative of observed field data. Calibration tends to be the area of a modeling that requires some effort to complete. With DTA applications there are more possibilities for route choice, temporal variability, and the complexities of the entire assignment process are introduced into the modeling steps. The calibration process is discussed in detail in Chapter 8, including the following steps:

  • Establish calibration objectives and identify the performance measures and critical locations against which the models are to be calibrated.
  • Determine the statistical methodology and criteria.
  • Determine the strategy for calibration (i.e., which model parameters are going to be adjusted and in what sequence?).
  • Conduct model calibration runs following the strategy and conduct statistical checks; when statistical analysis falls within acceptable ranges, the model is calibrated.
  • Test or compare the calibrated model with data set not used for calibration. If the model replicates the different data set, the model is validated.

4.7 Alternatives Analysis

The alternatives analysis in a DTA modeling process can vary greatly from analyzing short-term impacts such as an incident to testing long-term impacts such as an added lane. Transportation analysis software with a DTA component should have the ability to perform either type of analysis through a “one-shot” non equilibrium assignment or with an assignment using dynamic user equilibrium.

You may need the Adobe® Reader® to view the PDFs on this page.