This study introduces a method for consideration of statistical distribution of traffic counts in validating the link volumes in a travel demand model (TDM). Historically the TDM’s daily link volume validation at North Central Texas Council of Governments (NCTCOG) has been based on comparisons with single-day point-counts on select links, reported as percent root mean square error (%RMSE) by functional classification. Although the utilization of acceptable measures of error indirectly acknowledges the inherent variability in the observed traffic counts but it still only provides a single-point comparison. Therefore, the comparison is deficient in informing the analyst of the expected confidence in the model results since the traffic counts are random variables and comparing them to the static model outputs requires understanding of their variability.
The major source of NCTCOG’s freeway counts is TxDOT’s saturation counts that are collected every five years mainly throughout spring and fall and therefore do not necessarily satisfy the conservation of flow. This study uses the data collected from the regional ITS systems (TransVision and DalTrans) and the automatic traffic recorders (ATR) in calculating the variance, mean, and confidence intervals for the traffic counts. This information will be used to establish acceptable deviation of model results from the counts instead of solely relying on comparisons based on calculated %RMSE. The NCTCOG region has about 1,000 center-line miles of freeways where reliable traffic count distributions can be defined for at least 100 of its segments.
Applying this methodology could inform the analyst of the statistical meaning of the popular %RMSE of the model output. The results can also be used as a guideline for defining the calibration and validation expectations on different facility types and reduce the risk of over-calibration. This presentation summarizes the methodology, data sources, results, and recommendations.