TRAFFIC VOLUME ESTIMATION FROM SHORT-PERIOD TRAFFIC COUNTS

This paper considers the problem of estimating the annual traffic volumes at a count site when traffic counts are available for only a limited part of the year, perhaps only for a few hours or days. It presents a new method for estimating annual average daily traffic (AADT), based on regression analysis. This method uses the fact that traffic levels may vary considerably between count sites, but that variations are usually very similar between sites, especially within the same car length class. It is more precise than the traditional factor approach, and specifies the precision of the AADT estimate as a function of the sample design. This precision function can be used to optimise the sample design before counting starts. This allows a balance to be achieved between counting longer periods on roads with heavy traffic and shorter periods on roads with light traffic. Separate AADT estimates may be combined in various ways, for example to estimate annual vehicle distance travelled (AVDT) within a given region. The new method is applied to traffic data for five vehicle length classes from Oslo, Norway. It provided AADT estimates and their precisions for each length class in each direction, and precisions of weighted sums of separate AADT estimates.