Applying AI to forecasting compound extreme weather events

We propose a high-impact innovation for weather forecasting that integrates global ensemble weather forecasts with an AI-driven post-processing model of extreme weather indices (XtremeCast). This innovation will provide the basis for skillful probabilistic forecasts of compound extreme weather events at extended lead times, and the ability to identify weather phenomena not resolved by the global models. Our proposed calibration methods and extreme event modules are easily adapted to the new AI-based global models, whereby AI global model outputs drive AI-based extreme event models. The project addresses specific needs for increased resilience of electric utilities in the face of compound extreme weather/climate events, including risks from the increasing penetration of renewable energy. In addition to extreme weather from severe convective systems and hurricanes, the proposed solution will incorporate forecasts of heat/cold waves, wind and solar droughts, and impacts on natural gas generation efficiency to identify compound events with the potential to push energy systems beyond capacity. An advanced user interface dashboard based on Visual Analytics will support interactive visual analysis for more rapid and insightful analysis to support complex, rapidly evolving decision making and direct integration into client-side systems. CFAN's clients in the energy and insurance sectors have communicated the need for probabilistic forecasts of compound severe weather events with greater granularity and longer lead times than is currently produced by NOAA and other market providers. CFAN’s proposed innovation responds to this need and will improve electric utility companies’ ability to manage the impacts of extreme weather and create more resilient systems through improved load planning. CFAN’s solution will further support decision making in the broader energy sector, including natural gas trading and wind farm operators and investors. There are also potential applications in insurance, emergency management, logistics, transportation, and construction sectors. Advanced online decision support tools trained with data-driven and human insights will empower algorithms and experts to continue to learn from and validate each other – this feedback will support operational adaptation to extreme weather events in a changing climate.