The Numerical Prediction of Severe Convective Storms: Advances in Research and Applications, Remaining Challenges, and Outlook for the Future

Accurate and reliable short-range forecasts of hazardous convective weather save lives and reduce property loss. Considerable progress has been made over the past 30 years in improving high-impact severe weather forecasts, including forecasts of hail, damaging winds, flooding, and low-level mesocyclones associated with tornadoes. This progress has primarily been driven by advances in three areas: convective scale Numerical Weather Prediction (NWP) models, high-density storm-resolving observations, and data assimilation (DA) methods. In the development of convective scale NWP models within the US, many early modeling efforts focused on the simulation of convective scale phenomena in idealized environments. Then during the past few decades, the Advanced Regional Prediction System (ARPS) and Weather Research Forecast (WRF) models were developed with the express purpose of forecast applications using real observations. These two model systems inspired a wide variety of research and applications to advance short-term severe weather forecasts. The implementation and ongoing upgrades of the WSR-88D radar network and launch of GOES-R also have improved short-term severe weather forecasts by providing convective-scale detail for the model initial conditions. Research on variational and ensemble DA methods, especially for the assimilation of radar and satellite data, has further led to significant progress in improving severe weather forecasts. Looking ahead, research leading to better representation of cloud and microphysical processes in NWP models, proper hybridization of ensemble variational DA methods for convective scale phenomena, and effective use of machine learning to improve storm-scale DA and forecast models hold great promise for further improving both the accuracy and efficiency of short-term thunderstorm forecasts.