In the current practice at the Florida Department of Transportation (FDOT), seasonal factors (SFs) are used in the calculation of annual average daily traffic (AADT) at portable traffic monitoring sites (PTMS). The permanent traffic monitoring sites (TTMSs) are first manually classified into different groups (known as seasonal categories) based on similarities in traffic characteristics of roads and on engineering judgment.
FDOT districts then assign a seasonal factor category to each PTMS according to the site's geographical locations, assuming that seasonal variability and traffic characteristics at the short-term and permanent count sites are similar in the same geographic area. The goal of this study was to develop a more objective and data-oriented approach by seeking to explain the underlying causes of seasonal fluctuation patterns in traffic data and to develop a methodology rooted in statistics for assigning seasonal factors to short period counts. Various clustering methods from the nonparametric hierarchical cluster analysis and parametric model-based analysis were evaluated in this study and those that better described the truth-in-data in the seasonal factor grouping process were identified. Conventional multiple linear regression analyses were conducted to identify possible factors contributing to the seasonal fluctuations in traffic volumes. Based on these factors, a fuzzy decision tree was developed to determine the seasonal factor group assignment for short period counts in the southeast Florida urban area. The methodologies developed in this study for seasonal factor grouping and assignment were implemented in a prototype GIS program, which also supported visualization of data on transportation system, traffic data, land use, and statistics about seasonal factors and groupings.