Tackling the challenge of nowcasting elevated convection

Observational and modeling studies suggest that elevated convection accounts for approximately 50% of all convective events. Elevated convection has its updraft roots based above the convective boundary layer and occurs primarily during nighttime. Both human forecasters and numerical weather prediction models experience difficulties in successfully forecasting elevated convection. For example, the forecast skill scores for the Rapid Update Cycle (RUC) model exhibit a diurnal modulation with lowest scores during nighttime, which is an indication of poor forecast skills for elevated convection.

We have been exploring a new approach to forecasting elevated convection based on a conceptual understanding of the associated processes. The procedure builds upon a fuzzy logic technique using a variety of predictor fields based on both direct observations and derived fields from numerical weather prediction models. The fuzzy logic algorithm has been evaluated on many cases. Results so far are encouraging, which prompted us to initiate integration of the elevated convection into the NCAR AutoNowCaster (ANC) system to complement its existing surface-based convection forecast capability.