Abstract
In the present paper, we review a new method for relating cloud observations to large-scale variables of general circulation models. The method is based on an application of the cluster analysis to synoptic analyses Of prognostic model variables provided by the National Meteorological Center. Surface cloud observations are “clustered” according to the similarity of the principal-component loading scores of the corresponding vertical soundings. The method was tested by developing a simple cloud parameterization scheme, from the cluster-stratified cloud data, and comparing it with the observations.
Parameterization results are compared qualitatively against satellite imagery and surface analysis, and quantitatively against a scheme based on one variable only, the relative humidity. Qualitative comparison shows that the new approach generates cloud parameterization consistent with observations, especially with cloud structures related to various synoptic-scale flows. Quantitative comparisons indicate a possible advantage of the present method in the areas covered by limited observations. Overall, the results are suggestive of a possible alternative for upgrading and validating cloud schemes presently used in general circulation models.
