Consideration of Weather Conditions to Estimate Missing Traffic Data

Abstract

Estimating missing traffic data is an essential task for transportation agencies. Several imputation methods have been developed in the literature to estimate the missing traffic volumes. However, none have considered the variation in traffic patterns caused by severe winter weather conditions. Highway traffic volumes are highly influenced by weather conditions; therefore, a detailed investigation was carried out to develop relationships between weather and highway traffic volumes and to use them for reliable estimation of missing traffic volumes. The study was based on hourly traffic data from permanent traffic counter sites located on provincial highways of Alberta, Canada, using 11 years of data from 1995 to 2005. Weather data were obtained from Environment Canada weather stations located within 10 mi of the chosen permanent traffic counter sites. Cold and snowfall represented the winter conditions. Multiple regression analysis was used to develop relationships between hourly traffic volumes, categorized cold, and total snowfall. The study models showed a strong association between traffic volumes and weather conditions. Weekend traffic was more susceptible to weather than weekday traffic. In cases of extreme cold (≤25°C), the peak hours experienced fewer reductions in traffic (6% to 13%) than off-peak hours (10% to 17%). The amount of reduction in traffic volume caused by each centimeter of snowfall varied from 0.5% to 2.0%. The traffic–weather relationships developed were used to estimate missing hourly volumes. The errors were 30% to 75% less than the traditional methods used by highway agencies.