Time-series Analysis of Impact of Severe Weather Events on Traffic and Mobility Patterns

Severe weather events can cause significant changes in traffic volume and pattern throughout the day, where certain routes may experience an increased traffic while others may see reduced use. Studying the impact of severe weather events on traffic volume and patterns aids in designing more resilient infrastructure and developing effective emergency response strategies. This research aims at analyzing the existing traffic data to identify disaster-induced traffic patterns in the State of Kansas. We examine the impact of historical severe-weather events obtained from the National Oceanic and Atmospheric Administration, on hourly traffic data collected by the U.S. Department of Transportation Federal Highway Administration at different traffic stations across the state. We employ long-short-term-memory (LSTM) recurrent neural networks to construct a counterfactual prediction, which is compared against the actual traffic data to analyze the causal impact of severe weather events on daily traffic patterns in different regions of the state. For proof-of-concept illustration, we report on the impact of a major snowstorm as well as a tornado event on traffic volume at nearby traffic stations.