The impact of diabatic heating on the initialization of divergent circulations in a global forecast model

We have performed initialization and forecast experiments using the ECMWF FGGE Level IIIb analysis data and the NCAR Community Climate Model. The performances of several initialization schemes, including nonlinear normal mode initialization (NNMI) with diabatic heating effects, were examined. Global diabatic heating distributiohns were evaluated from 1) a residual of the thermodynamic energy balance using FGGE analyses (referred to as "observed" heating) and 2) a time integration of the prediction model (referred to as model generated heating). Many 24-hour forecasts were carried out with initial condistions from each initialization procedure. While the performance of NNMI with the "observed" diabatic heating was superior, the initial large-scale divergent circulations were not retained during the forecast period unless the prediction model was forced to generate a diabatic heating sufficiently similar to the "observed" heating. We conclude that not only is a realistic heating distribution necessary to produce proper initial conditions, but also physical parameterization in the prediction model must be able to generate an appropriate heating distribution. This is necessary to correctly predict the quasi-stationary, large-scale tropical circulations