Modeling Travel Time Reliability for Non-Interstate National Highway System Routes

About the report:

Under the Moving Ahead for Progress in the 21st Century Act, state departments of transportation are required to report travel time reliability and set performance targets for interstate and non-interstate National Highway System facilities. Although several Virginia Transportation Research Council studies have analyzed and predicted travel time reliability on interstates, a gap still exists in studies modeling travel time reliability on arterial roads. Using 4 years of INRIX and National Performance Management Research Data Set (NPMRDS) probe data, this study conducted a segment-level comparative analysis of travel time distributions and the Level of Travel Time Reliability (LOTTR). Results show that NPMRDS systematically reports higher travel times than INRIX, inflating reliability metrics. Consequently, NPMRDS classifies significantly more segments as unreliable, especially on short (less than 0.25-mile) segments. Applying sample size filters to exclude segments with low sample sizes does not substantially reduce the discrepancies between NPMRDS and INRIX travel times. Analysis reveals that a major contributing factor is the data density of NPMRDS. More than 80% of observations fall into the lowest category (density A), leading to greater volatility in the higher percentiles of travel times and inflated Level of Travel Time Reliability values.

In addition, this study developed planning-level models to predict the 50th, 80th, and 95th percentile travel times on arterial segments statewide. Linear mixed models, random forests, and Light Gradient Boosting Machine models were built using data from multiple sources, and model performances were compared. The results showed that linear mixed models perform reasonably well for segments with LOTTR below 1.5, but machine learning approaches offer superior accuracy across the full reliability spectrum. The study also found that the machine learning models trained with INRIX data underestimate the travel time percentiles calculated using NPMRDS data, with a mean absolute percentage error of more than 10 for the 50th percentile travel time.

The study recommends that the Virginia Department of Transportation further develop models trained with NPMRDS data to meet federal reporting requirements and improve accuracy. In addition, the Virginia Department of Transportation should consider improving consistency in travel time reliability analyses across divisions and programs where applicable. Researchers developed an implementation plan that includes stakeholder engagement, technical guidance development, and support for the next target-setting cycle of the system reliability performance measures set forth in 23 CFR Part 490 (May 30, 2025).

_{Disclaimer Statement:The contents of this report reflect the views of the author(s), who is responsible for the facts and the accuracy of the data presented herein. The contents do not necessarily reflect the official views or policies of the Virginia Department of Transportation, the Commonwealth Transportation Board, or the Federal Highway Administration. This report does not constitute a standard, specification, or regulation. Any inclusion of manufacturer names, trade names, or trademarks is for identification purposes only and is not to be considered an endorsement.}

Authors