Determination of the Systematic Bias and Variability of the Error of the National Meteorological Center's Six-Layer Baroclinic Model

摘要

Geographical distribution of the bias and variability of the error in the 500 and 1000mb geopotential height forecasts produced by the National Mereorological Center's operational six-layer baroclinic model was investigated. Forecasts and the verifying observations for six winter seasons beginning with the 1973-74 winter were used. It was found that the model systematically damps out the amplitude of the climatological mean surface highs and lows, and 500mb troughs and ridges. Typical values of the bias in the mean 500mb standing wave troughs are on the order of 60 meters; in these regions the bias accounts for up to 30% of the rms error. Throughout the forecast period, the standard deviation of the forecast error about its own mean shows regional contrasts similiar to climatological standard deviation patterns. These non-systemtic forecast errors tend to be largest in storm tracks over the western oceans and quite small at low latitudes. Normalization of the standard deviation of the error by the climatological standard deviation of the observation results in a relatively featureless field except for a monotonic pole-to-equator increase. The normalized errors are larger at the 1000mb level than at the 500mb level over most of the hemisphere but they are of comparable magnitude over the eastern Atlantic and Europe. The vertical structure of the error field, as manifested in the geographical distributions of the correlation between the 1000 and 500mb forecast error, and the ratio of the amplitudes of error fluctuations at those levels, was found to be similiar to that of real atmospheric fluctuations.