Maximizing the value of Warn-on-Forecast System (WoFS; Heinselman et al. 2024) output requires application of post-processing techniques that (1) predict storm hazards that are not explicitly predicted by the ensemble system, (2) calibrate ensemble forecast output based on previous system performance, and (3) provide dynamic estimates of forecast accuracy to improve usability. Machine learning (ML) methods are particularly suited to these objectives (Flora et al. 2021; McGovern et al. 2023). ML is also a promising avenue for improving storm-scale data assimilation, e.g., by replacing current cloud analysis schemes used to accelerate spin-up of thunderstorms in the ensemble. 

A particularly ambitious application of ML, specifically deep learning (DL), is the emulation of NWP models. The ability of DL models to emulate global NWP models with reasonable accuracy has recently been demonstrated by Google DeepMind's Graphcast and other DL models. The emulation of convection-allowing models (like the WoFS) is an emerging and exciting topic that is rapidly growing in priority within NOAA.

Research proposals are invited on all applications of ML/DL to the WoFS. 

Flora, M. L., C. K. Potvin, P. S. Skinner, S. Handler, and A. McGovern, 2021: Using machine learning to generate storm-scale probabilistic guidance of severe weather hazards in the Warn-on-Forecast System. Mon. Wea. Rev., 149, 1535-1557, https://doi.org/10.1175/MWR-D-20-0194.1.

Heinselman, P. L., and Coauthors, 2023: Warn-on-Forecast System: From Vision to Reality. Wea. Forecasting, 39, 75–95, https://doi.org/10.1175/WAF-D-23-0147.1.

McGovern, A., R. J. Chase, M. Flora, D. J. Gagne, R. Lagerquist, C. K. Potvin, N. Snook, and E. Loken, 2023: A Review of Machine Learning for Convective Weather. Artif. Intell. Earth Syst., 2, e220077, https://doi.org/10.1175/AIES-D-22-0077.1.