A Simple Parcel Method for Prediction of Cumulus Onset and Area-Averaged Cloud Amount over Heterogeneous Land Surfaces

Abstract

The purpose of this note is to compare several methods for predicting the onset and quantitative amount of cloud cover over heterogeneous land surfaces. Among the methods tested are that of Wilde et al. (1985) and a new, simple parcel approach. Model comparison is accomplished by running each model using a series of six initial conditions from the Wangara experiment. Case days were chosen because they had relatively quiet synoptic conditions, and exhibited the formation of cumulus clouds from an initially mostly clear sky during the period of solar heating. Each model contains two or three free parameters that were systematically varied until the optimum agreement was reached between observed and predicted cloud amount. The single best run for each method was chosen based on the RMSE and coefficient of determination. The best runs are compared and plotted against the observations for the six case days. Results of these limited tests do not necessarily suggest the absolute degree of accuracy to which low cloud cover may be predicted. This is left for a future study. Rather, the focus is on the relative skill and flexibility of the various models. It is shown that parcel methods, in which surface air is lifted to its equilibrium level while being diluted by a defined amount of mixed layer air, produce substantially superior prediction of cloud amount, particularly during periods of rapid cloud onset when the mean boundary layer top is swiftly rising through a new-neutral layer. Pending, verification from independent datasets, it appears that an rms error in instantaneous cloud amount of ±10% may be achievable.