Towards Unified Weather-to-Seasonal Prediction Modeling with GFDL SHiELD

Abstract:

We present GFDL SHiELD, the System for High-Resolution Prediction on Earth-to-Local Domains. This system couples the nonhydrostatic FV3 Dynamical Core to a heavily modified form of the NCEP GFS Physics. It incorporates a wide range of modeling applications, including traditional medium-range global and short-range regional prediction, and new capabilities such as subseasonal-to-seasonal prediction, global-to-regional convective-scale modeling, and global-cloud resolving modeling. SHiELD forms part of the fourth-generation GFDL modeling suite, providing a link between weather and climate modeling, and permits advances to be seamlessly integrated within other FV3-based and Unified Forecast System (UFS) models.

We show several improvements in model simulations and forecast skill, starting from earlier years' versions of SHiELD. We show progressive improvements in hemispheric prediction skill, precipitation forecasting, tropical cyclone track and intensity prediction, and in the prediction of severe convective storms. The most recent versions of SHiELD also show an excellent diurnal cycle of precipitation, skillful prediction of the Madden-Julian Oscillation out to 28 days, and much improved tropical cyclone structure. We also discuss updates to the global cloud-resolving configuration. We close with some lessons learned about unified modeling, and the challenges and opportunities it poses.