A New Hurricane Impact Level Ranking System: A Multivariable Approach to Forecasting Loss Using Artificial Neural Networks for Communicating Risk to the Public

Hurricanes consist of multiple meteorological hazards, including winds, storm surges, and precipitation, which can have varying effects, depending on location. However, traditional warnings targeted towards the public categorize the intensity of these storms purely by wind speed. The severity of these multiple hazards, along with the population affected and locational characteristics, such as the presence of wetlands and/or bays, interact in a complex and intricate manner, making it increasingly difficult to connect these variables to a clear and concise socio-economic outcome that could be used to communicate risk to the public.

This presentation will discuss a newly proposed Hurricane Impact Level Ranking System and the artificial neural network model developed for its forecasting use. The Hurricane Impact Level Ranking System uses thresholds of economic damage to categorize historical events in order to provide a comparative level for a new oncoming hurricane event within the United States. This approach proposes a new way to synthesize these multiple hazards from one hurricane event into a simpler form the public more easily comprehends: cost. In real time use, an artificial neural network relies on established patterns from historical events as a comparative algorithm to forecast results for a new event. The results of the research conducted in building this model have led to additional understanding of how best to approach a multivariable assessment of a natural hazard. Use of this framework and ranking system could improve how the scientific community conveys risk and vulnerability to the public during natural hazard events in order to minimize cases of crying wolf and subsequent disregard to later warnings.