This page is open for editing because it is part of the Incubator. Have something to add? Please register so you can contribute. Have an option you would like to share? Please click on the 'Talk' button to enter the dialogue. The TF Resource Volunteers appreciate your feedback and interest.


Given the complex nature of travel demand forecasting, the manner in which planners and forecasters communicate transportation model results affects how those results are interpreted and utilized. The Committee for Determination of the State of the Practice in Metropolitan Area Travel Forecasting stated that, “there are many sources of error and uncertainty in travel demand forecasting, but end users of most travel forecasts would not be aware of these limitations.” (TRB, 2007). Decision makers and general audiences often lack formal training in transportation modeling and/or statistical analysis, increasing the risk that they might misinterpret modeling results presented by planners. Through clear and considerate communication with stakeholders, planners can reduce the likelihood that model results could be misinterpreted, applied inappropriately or unintentionally misrepresented, potentially avoiding unfortunate consequences from ensuing decisions and investments. To minimize the potential for these issues, transportation planners should be able to 1) fully understand and answer critical questions about their agencies’ models and model development processes, and 2) understand their models’ sensitivities to various inputs, as well as their importance and the corresponding risk of misrepresentation. Understanding a model requires understanding the assumptions, uncertainties, and limitations associated with model results, and understanding their importance when communicating those results with decision-makers.

Organizing and Presenting Spatial and Temporal Data, and Interpreting Results

The first steps in communicating model results often involve organizing and analyzing a transportation model’s available input and output data. Generally, this model data can be characterized by its spatial and temporal complexity, which depends on the level and need of the study. Spatial complexity ranges from modeling single intersections to modeling entire corridors, sub-regions, subareas or study sites. Temporal complexity ranges from modeling sub-second traffic controller logic to understanding demand shifts that may occur over a period of months or years. Considering the type of information and its complexity, an appropriate presentation method and medium can be identified to prevent communication issues with decision-makers while facilitating further analyses. The primary data outputs from Travel Demand Forecasting (TDF) models are trip tables and networks with assigned trips. Aggregating Traffic Analysis Zone (TAZ) -level data and results to larger spatial dimensions (e.g., a district or jurisdictional level), as shown in Figure 1, can assist in understanding and interpreting the model results. Temporal information associated with TDF models may also include time-of-day trip distributions and traffic state characteristics, which could be presented using charts and histograms, or using software for visualizing combined temporal and spatial data (e.g., Geographic Information Systems and/or transportation modeling software). NCHRP Report 365 discusses time-of-day characteristics in detail and presents diurnal distributions for different size urban areas.

Most Travel Demand Forecasting efforts involve multiple steps, the most common steps being trip generation, trip distribution, mode choice, and trip assignment. The following sections offer recommendations for analyzing and interpreting the results from these specific model components.

Interpreting Trip Generation Results

Trip generation data is commonly summarized by the number of trips starting and/or ending in a particular collection of TAZs which describe an area of interest, such as a central business district (CBD). Comparing the total number of trip ends in the region as a whole with expected trip-making can help identify potential issues in this step and provide context for results from the remaining modeling components.

Interpreting Trip Distribution Results

Results from the trip distribution step are also commonly described using the total number of trips starting and/or ending in an area of interest. Comparing values across scenarios can help to identify how travel patterns change with under various land use and network inputs. It may also be useful to examine the number of trips between pairs of interest areas, paying additional attention to how trips with different purposes are distributed (e.g., number of home-based work trips between a suburban area and university or business park). Desire line maps, which can be created using most commonly available GIS software, can aid in interpreting trip distribution results by displaying the results graphically.

Interpreting Mode Choice Results

In general, mode choice modeling results should be reviewed by comparing relative results across scenarios, focusing on the impact of changes between scenarios. Special attention is necessary when adding transit service, particularly in areas which currently have a low transit mode share. In this case, it can be helpful to examine the total number of transit trips and compare differences in transit use only between scenarios. This approach can also be useful when examining potential transit mode share for non-home-based work trips.

Interpreting Trip Assignment Results

Trip assignment results can be considered on a network-wide basis or for specific locations. Vehicle miles traveled (VMT) is a common network-wide performance metric which provides an overview of the service consumed from the assignment. For specific locations, macroscopic-level assignment results can be evaluated using screenlines or cordon lines. Screenlines and cordon lines are imaginary lines that are placed across all roadways covering a specific movement or subarea, and those aggregated counts can help planners quickly identify deficiencies and capacity constraints in the network. Cordon lines are especially helpful for comparing directional traffic patterns with time-of-day assignment results.

Acknowledgment of Uncertainty and Errors

There are several potential uncertainty and error sources in developing a transportation demand forecasting model. It is important to acknowledge the following error sources and communicate with the general audiences about the impact of those sources.

  • Coding Errors

Errors in coding the highway and transit networks. Errors in recording survey results.

  • Sample Errors

Errors from bias that occur in the survey sample frame. Example: A land-line telephone survey could miss households which rely on mobile phones or have no phone, which could potentially result in a demographic bias (e.g., fewer low income households and fewer young people without land lines).

  • Computation Errors

Errors which occur in developing the model programs.

  • Specification Errors

Errors from improper structure of the model where key variables or parameters are overlooked in the estimation phase. Errors from transferring model parameters from one region to another.

Sensitivity testing in traffic demand forecasting models can facilitate a sound analysis between the variation in the model outputs and the uncertainty in the model inputs. Visualization based on sensitivity testing results can also identify the possible error sources in the model, and further improve understanding of structure in the transportation demand analysis and simulation models.


NCHRP. “Evaluating and Communicating Model Results: Guidebook for Planners”. (2010) NCHRP 08-36, Task 89

Transportation Research Board, “Metropolitan Travel Forecasting, Current Practice and Future Direction,” Special Report 288, Washington, D.C. (2007).

This category currently contains no pages or media.