We have performed a number of assimilation and forecasting experiments on long lived supercellular convection. Both simulated data and observed radar data experiments have been performed. The real data experiments are described in Part II of this two-part presentation, in which a tornadic supercell that occurred during the CASES experiment near Wichita, Kansas in May of 1997 is studied. The control simulation in the simulated data experiments uses a sounding obtained a few hours before the tornadic supercell was observed. A warm and moist bubble is placed at the low levels to initiate the convection. The goal of this study is twofold: 1. to examine the sensitivity of the forecast to initial conditions that are obtained by assimilating Doppler radar observations using a four-dimensional variational (4DVAR) technique. 2. to investigate what are the essential features that must be retrieved in order to produce a good forecast.

The control simulation is used to generate series of radar radial velocity and reflectivity data in a frequency of five minutes, similar to the WRS-88D data. These data are assimilated by the 4DVAR system to provide initial conditions for a subsequent control forecast. The data coverage and quality are then varied to examine the sensitivity of the forecast to the retrieved initial conditions. We also evaluate the sensitivity of the forecast with respect to the ambient flow and to model error associated with the microphysical parameterization. Through these sensitivity experiments, we attempt to determine the mechanisms that control the motion and evolution of this supercell storm and the essential aspects in the initial fields and in the ambient flow for a successful forecast. Experiments are also conducted to examine the data requirement for estimating parameters in the microphysical schemes in addition to retrieving the initial conditions.