This paper discusses what the most appropriate measures of observed traffic counts are for model validation. As the aim of validation is to compare observed counts with model outputs, one must ensure that the counts used in the comparison represent the modelled travel conditions as closely as possible. An analysis of hourly traffic counts for a calendar year in the greater Sydney region showed that the ôobserved traffic countsö may be calculated in several ways, leading to significantly different results. The following recommendations are offered. Travel models are calibrated to reproduce a typical ôaverage dayö of certain definition: all days of the year, or weekdays, or workdays, or schooldays only. Therefore, the counts used for validation must be calculated from the same set of days that the model is representing. It is more appropriate to use the median value of the counts rather than the mean, as the mean is more sensitive to a few extreme values that may occur over the year. This study has also demonstrated that there are significant differences between the model outputs and the link counts for any given time period even from the same input data. These differences are due to the static traffic assignment method used in the strategic models. (a) For the covering record of the conference, please refer to ITRD abstract no. E218380.
Abstract