The Performance of a Medium-Range Forecast Model in Winter–Impact of Physical Parameterizations

Abstract

We present the results of a series of forecasts on seven weather situations from February 1976 using two models which differ only in their physical parameterizations. One set of parameterizations was developed at the Geophysical Fluid Dynamics Laboratory (GFDL) some years ago, the other more recently at the European Centre for Medium Range Forcasts (ECMWF). The resolution of the model (N48, 15 levels) was that which ECMWF has used in the first phase of operations, which began in August 1979. The particular aim of the experiments was to study the importance of the differences in the parameterization schemes for the model; in addition, we obtained a general view of the forecast results that might be available in the first phase of operations. Both sets of parameterizations gave similar results in terms of forecast quality. When measured by the standard objective methods, the range of predictability was 5?6 days. A study of the systematic errors in the forecasts showed that these were mainly associated with a loss of energy in the largest waves; the evolution of these systematic errors appeared to be roughly linear in time. (This is not to say that the systematic errors arise due to linear mechanisms.) A study of the energetics showed that the major part of the loss of energy in the long waves was due to a failure to maintain the stationary part of the long-wave energy. Regarding transient phenomena, the downstream intensification of baroclinic waves appeared sometimes to be predictable beyond 6 days.