Verification of Convection-Allowing NWP in High-Shear, Low-CAPE Environments

In the past, forecasters utilized relatively coarse numerical model guidance in generating short-range (1-3 day) forecasts of severe convective storms. They used model guidance to predict the convective environment, including vertical wind shear and CAPE, for example. However, recent improvements in model resolution now allow for explicit prediction of convective storms in NWP models, run without parameterization of deep convection (Convection-Allowing Models, or CAMs). As a result, many forecasters now use a combination of the previously established environmental “ingredients-based” approach and CAM NWP when forecasting severe convection. Many antecedent studies have evaluated CAM NWP in high-CAPE environments, and others have focused on improving discrimination between severe and non-severe low-CAPE environments. In contrast, few studies have been conducted which evaluate the performance of CAM NWP in high-shear, low-CAPE (HSLC) environments. Therefore, our goal is to both establish methods for the verification of CAM NWP in HSLC environments as well as conduct the verification on a set of HSLC severe convection cases. This will provide forecasters with information concerning the quality of CAM HSLC predictions, and will assist future studies which seek to verify CAM NWP in HSLC environments.

We focus on times and locations where most HSLC severe convection events occur, during the cool season (15 October - 15 April) in the southeastern United States. First, we compiled a database of all severe convection events within this temporal and spatial domain over two seasons (2017-2019), yielding 104 events. By applying an instability threshold of MUCAPE ≤ 1000 J/kg, the initial set of events is reduced to 70 events that could be characterized as low-CAPE during some portion of the event. If the severe convection is required to be low-CAPE throughout the entirety of the event, the initial set of events is reduced to 25 events over the two seasons. From the set of 70 events, three events are chosen as test cases with which to define the verification process; these cases exhibit varying levels of tornadic as well as non-tornadic severe convection. In the verification process, model proxies from the HRRR and HREF such as 0-3 km Updraft Helicity are compared with observations including SPC storm reports and the radar-based Mesocyclone Detection Algorithm. Verification methods from the three selected cases will then be expanded to the broader set of severe convection events in HSLC environments. Verification methods and results will be presented.