Traffic volume estimation from short period traffic counts

This paper considers the problem of estimating the yearly traffic volume at a count site, when traffic counts are available for only a limited part of the year, perhaps only a few hours or days. A new method for estimating annual average daily traffic (AADT) based on regression is presented. In addition to being more precise than the traditional factor approach, the new method supplies the precision of the AADT estimate as a function of the sample design. This precision function may be used to optimize the sampling design before the actual counting is performed. Separate AADT estimates may be combined in various ways, for instance to an estimate of annual vehicle distance travelled (AVDT) within a specific region. The new method is applied to traffic data from Oslo, Norway. For each count site, hourly counts of number of vehicles within five length classes are available in both directions of the road. The method provides AADT estimates and their precision for each length class within each direction, as well as for weighted sums of separate AADT estimates.