A Synoptic Evaluation of Normal Mode Initialization Experiments with the NMC Nested Grid Model

Abstract

The Regional Analysis and Forecast System at the National Meteorological Center consists of an optimum interpolation objective analysis scheme, an adiabatic nonlinear normal model initialization (NNMI) and a hemispheric Nested Grid Model (NGM) to provide 48 h forecasts. We investigate here the effect NNMI has on the analyses and forecasts produced by this system. An eight vertical mode NNMI procedure led to significant reductions of the divergent component of the analyzed wind field in regions of heavy precipitation. This is shown to contribute to a systematic spinup error in NGM 0?12 h precipitation forecasts, especially from the 0000 UTC runs. Forecasts starting with no initialization had unacceptable noise levels. NNMI using two vertical modes yielded the best combination of noise-free forecasts and unsuppressed initial precipitation rates. A physical interpretation of this result is provided using the vertical structure functions of the normal modes. Tests of the two-mode NNMI in an operational setting confirmed that the 0?12 h NGM precipitation amounts increased along with a reduction of an excessive precipitation bias in the 12?36 h period. The two-mode procedure was implemented operationally in August 1987. It is suggested that if one has a reliable data assimilation system, especially if the 6 h forecasts used as background fields for the analysis have realistic precipitation rates and attendant divergent circulations, then a two-vertical-mode adiabatic NNMI provides many of the benefits of a traditional diabatic normal mode initialization procedure.