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.

This page is part of the Category Project-level traffic forecasting.


Project-level traffic forecasting is a specialized set of techniques intended to refine and improve the output of generalized demand forecasting methods, typically by using empirical data along with other validation methodologies. Project-level traffic forecasting became well established with the publication of NCHRP Report 255 (1982), which was later superseded by NCHRP Report 765 (2014). The "Hawaii Guidelines" are a rewritten version of much of the material in NCHRP Report 765, and these guidelines form the basis for this article.

The Practice of Project-Level Travel Forecasting

Project-level traffic forecasting describes the techniques employed to produce precise and accurate quantitative estimates of future traffic on a single (or small set of adjacent) road segment(s). As distinct from area-wide travel demand modeling, project-level forecasting typically employs additional statistical methods and empirical controls that allow the result to satisfy rigorous validation criteria. The purpose of these guidelines is to describe both best and acceptable practice for performing project-level traffic forecasts for state DOT's and similar agencies. The guidelines describe a number of techniques and options that are all acceptable within their intended scope, specific to the technique.

The forecasting process itself varies depending on the location and context of the project. Project sponsors (most typically states, counties or municipalities) that are accountable for a project's success will often require that the underlying design assumptions (in this case, forecast traffic volumes) be rigorously developed using thoroughly vetted techniques that can withstand expert scrutiny. To this end, the National Cooperative Highway Research Program (NCHRP) periodically sponsors research to compile, explain and synthesize best practices. By extension, satisfying area-wide model validation standards are prerequisite to applying subsequent refinements at the project level.

A variety of forecasting techniques should be considered with respect to the project context; oftentimes with techniques used in combination. While there is some limited role for professional judgment in applying these techniques, project-level forecasts must always be accompanied with a document that describes how the forecast was accomplished. In some cases there may be special reporting requirements based on the nature of the forecast request.

Using Regional Travel Demand Models

The logical starting point in preparing a project-level traffic forecast is with the standard raw data output from a regional travel demand model. It is important to understand that it is virtually impossible to calibrate a regional model's performance to satisfy the validation criteria for every conceivable project level forecast that might occur. In fact, if a regional model is fortunate enough to produce a perfect match to observed conditions at a project location, it is very likely a simple lucky shot. Ideally, the starting point for a project-level forecast would conform to a best-practice modeling standard. Practically speaking, however, an acceptable standard may be found in MPO models used for regional travel forecasting in the United States by virtue of their regular application in evaluating large capital investments that receive close scrutiny. In either case, forecasting assumptions that deviate from or correct the source material should be documented.

One important distinction between regional travel modeling methods and their project-level forecasting counterparts is in how the concepts of "highway capacity" and "travel delay" are defined and calculated. Additionally, project-level forecasts very often must be refined both temporally and geographically by interpolating between forecast years and by pivoting with select link analysis for small developments.

In preparing a project-level traffic forecast for single location, the analyst should recognize that observed traffic volumes, in fact, vary from day-to-day. Because travel demand models typically provide only an average annual daily traffic volume (AADT), it is important to verify that this average reflects an acceptable margin of statistical variation in observed counts .

OD table refinements

Temporal refinements and directional split refinements

Vehicle mix refinements

Turning movement refinements

Screenline refinements

Speed and travel time refinements


Refinement for Evaluation

Refining vehicle class forecasts for evaluation

Refining speeds for evaluation

Conventional Post-Processing

Highway noise analysis

Safety analysis

User benefits

Pavement design

Other Evaluation Issues

Measures of effectiveness and performance measures

Air quality estimation

Traffic microsimulation

Land use models

Custom Project-Level Models

Scenario Comparisons

A scenario might involve the economic, demographic or land-development environment of the project....Scenario/sensitivity testing

Within a single scenario, there is still the possibility of uncertainty in a forecast....Reporting of reasonable bounds on forecast values

Techniques for Increasing Spatial Resolution

Windowing with OD table estimation from traffic counts

Working with vehicle re-identification data

Subarea focusing

Blended Models

Hybrid models

Multi-resolution models

Improving Temporal Detail

Temporal resolution

Traffic dynamics

Guidelines for Specific Project Types

Bypasses of regional scope

Bypasses of local scope

Time Series Methods

Linear Regression Techniques

Trend models

Linear models with explanatory variables


Box-Jenkins/ARIMA Methods

Autoregressive (AR) models

Autoregressive with explanatory variables (ARX or SAR) models

Box-Cox transformations

Time Series Examples

Example of an autoregression model with Box-Cox transformation

Case Studies