Abstract: With the apparent increase in the frequency and intensity of hurricanes, blizzards, and thunderstorms, predicting the impacts of these natural hazards is becoming more important. For weather-related hazards, numerical weather prediction (NWP) models are perhaps the best way to describe these events in both time and space. However, these physical models currently have distinct technical shortcomings, especially when applied to particular hazards. And impact models forced with NWP forecasts become inextricably linked, where the uncertainties and biases of the weather forecasts are propagated into the impact predictions. The strengths and weaknesses of NWP in terms of what it can accurately predict and how these aspects can be exploited to create useful, empirical, natural hazard impact models will be discussed. Specifically, lessons learned from creating the Outage Predictions Model (OPM), an operational, machine-learning-based, power outage prediction system forced with NWP forecasts, will be presented.
Bio: Peter Watson is a Ph.D. student in environmental engineering at the University of Connecticut and a graduate of the University of Chicago. His research interests are in the application of machine learning and other modern modeling techniques to predict the impacts of natural hazards to inform the response, adaptation, and mitigation of these hazards.