Comparing traffic flow time-series under fine and adverse weather conditions using recurrence-based complexity measures

Abstract

Adverse weather conditions may have a significant impact on the statistical characteristics of traffic variables such as volume and speed, and consequently on their predictability. We address the problem of freeway lane speed variability under the effect of precipitation episodes and present a non-linear dynamics methodology based on recurrence quantification analysis to compare speed evolution under fine and adverse weather conditions. Findings indicate that, under fine weather conditions, section travel speed is an adequate descriptor of traffic evolution in the left and middle lanes for high demand, but fails in cases of incidents or low traffic. Moreover, when it rains, regardless of intensity and duration, a significantly dissimilar evolution of lane speed is observed; this difference should be considered in the process of short-term traffic forecasting.