Hybrid Numerical Weather Prediction: Downscaling GraphCast AI Forecasts for Downslope Windstorms

Abstract Preemptively managing electrical circuits during periods of elevated wildfire risk, such as downslope windstorms in complex terrain, necessitates accurate, high-resolution forecasts days in advance. Currently available high-resolution operational guidance is limited with respect to the forecasting period and/or spatial detail it can provide. As a consequence, many public utilities support their own NWP systems, primarily WRF-based, with initial conditions and boundary forcings supplied by operational global models such as the GFS. This study demonstrates pairing Google’s artificial intelligence (AI)-based GraphCast (GC) model, which operates on a 0.25° grid and produces global forecasts every 6 h, with a “last mile” WRF (WRF-LM) downscaling framework to refine GC’s spatial and temporal resolution for utility forecasting needs. We illustrate the potential of this system through a case study involving the December 2021 Marshall Fire windstorm in the Boulder, Colorado, area. Our results suggest that GC-initialized WRF forecasts are competitive with those guided by NCEP and ECMWF operational products, offering a promising hybrid approach for utility weather forecasting. Significance Statement Downslope windstorms are difficult to forecast with small errors potentially affecting the timing, magnitude, and location of concerning winds. Due to the elevated wildfire danger caused by these windstorms, accurate forecasts with lead times greater than 48 h are often needed to allow utilities time to prepare for the de-energization of transmission lines to reduce ignition risk. This paper examines methods of downscaling coarse synoptic-scale forecasts using the downslope windstorm associated with the 2021 Marshall Fire as our case study. A novel aspect of this work is the incorporation of machine learning weather prediction models and downscaling their forecasts using a high-resolution numerical model.