Development and Use of an Agent-based Model to Assess the Effect of Forecast Credibility on Urban Traffic During Snow Events

Abstract

With the difficulties in snow accumulation prediction, the potential for false alarms and forecast misses arise. These forecast errors can lead to a lack of public trust and poor decisions in responding to future weather hazards. There has been little research on how individuals respond in the future to false alarms and forecast inconsistencies. We developed an agent-based traffic model to demonstrate how snow forecasts and public response interplay. This model factors receptiveness to expertise, forecast severity, and forecast credibility into the agents’ work-related travel decisions. Agents are grouped into three categories: firm workers, service workers, and household workers, where firm workers can work from home, service workers must go into work, and household workers always work from home. It was found that forecast severity has the most effect on the number of agents traveling, while credibility factors into agents’ decisions if they have the option to work from home. Owing to uncertainties in actual accident rates during snowfall, no firm conclusions were made in terms of how such events might interact with forecast severity and credibility, although there does appear to be potential for significant regional differences in these effects. This model is a first attempt at simulating the role that these factors play in work-related travel decisions and outcomes, but it is deliberately simple. Recommendations are made regarding useful enhancements to the model framework.