Objective Diagnosis of Thunderstorms and Convection Initiation Using Convection-Allowing Numerical Weather Prediction Models

Accurate forecasts of Convection/Convective Initiation (CI) and the attendant evolution of the Convection/Convective Activity (CA) are crucial to assessing the hazards posed by convective weather to a wide range of society. CI/CA forecasts play important roles in all aspects of forecasting, ranging from daily forecasts of precipitation to the creation of daily severe weather forecasts and even determining aviation routes. Unfortunately, even with advancements in convection-allowing NWP, predicting the "where" and "when" of CI remains a significant hurdle.

Recent efforts in the NOAA Hazardous Weather Testbed (HWT) in Norman, OK have centered on how best to utilize convection-allowing ensembles to glean information regarding CI/CA. Building off results from the 2011 HWT Spring Forecasting Experiment (SFE), efforts in the 2012 SFE centered on using model reflectivity values at the -10C level to determine if a grid point was convectively active (dBZ >= 35) or not (dBZ < 35). Once CA grid points were identified two separate CI identification algorithms, one object-based and the other grid-point-based, were used to identify CI grid points. Once the CI grid points were determined for each member of the ensemble, various aspects of CI (e.g., timing, probability, evolution) were calculated and used as guidance for probabilistic forecasts of CI location and timing.

This presentation will highlight a multi-facted strategy that was developed for rigorous diagnosis of CI/CA during the 2012 SFE, applied in a convection-allowing ensemble, and used to generate probabilistic guidance for experimental forecasts of the CI timing and location. This strategy was intricate, requiring implementation in the WRF model of 1) simulated reflectivity calculations that were uniquely linked to different microphysical parameterizations, and 2) run-time diagnostic code that extracted simulated reflectivity fields using a sampling interval comparable to that of observed reflectivity (5 mins). Outside of the WRF model, the ensemble output was ingested into sophisticated algorithms that determined the timing and location of CI points in all model runs. These data were further processed to generate probabilistic guidance for both the timing and location of CI (and CA). Finally, the guidance was presented to forecasters in a series of simple, easy to interpret displays that were used for preparation of probabilistic forecasts of CI and CA. The presentation will focus on this strategy while leaving a more thorough evaluation of experimental CI/CA forecasts for a separate presentation.