A study of convective initiation failure on 22 October 2004

On the evening of 22 October 2004, a surface warm front moved northward across portions of the Northern Plains. All cycles of the operational models, including the Rapid Update Cycle (RUC), forecast convection to initiate along the front by 0000 UTC 23 October 2004, despite a large capping inversion present on observed soundings during the morning hours. Initially, the forecast models resolved the inversion, but eroded the cap throughout the day; this lead to the anticipation of severe weather by local offices in the area and by the Storm Prediction Center. However, convection did not initiate along the warm front, resulting in a null event.

This study seeks to determine why convection did not occur along the warm front, as well as the reason that all three forecast models incorrectly forecast convection. This is done by examining surface, satellite, and upper air data, and also through a simulation of the event. The WRF-ARW is initialized with the North American Regional Reanalysis dataset and run both at a resolution similar to the operational models with a convective parameterization, and at a high resolution with explicit convection. A comparison of the two WRF simulations and a comparison of the WRF simulations to the forecast models allow for an analysis of how convective parameterizations may affect the model's ability to forecast convective initiation in a capped environment, and how models handle the evolution of capping inversions in the presence of strong low level forcing for ascent. Results from this study may help to identify the necessity of running a high-resolution local model at National Weather Service offices when convection is forecast by the operational models. In addition, results will aid forecasters in understanding how models handle capping inversions, so that model failure in short term convective forecasts can be better anticipated.