A Bootstrap Technique for Testing the Relationship between Local-Scale Radar Observations of Cloud Occurrence and Large-Scale Atmospheric Fields

Abstract

A classification scheme is created to map the synoptic-scale (large scale) atmospheric state to distributions of local-scale cloud properties. This mapping is accomplished by a neural network that classifies 17 months of synoptic-scale initial conditions from the rapid update cycle forecast model into 25 different states. The corresponding data from a vertically pointing millimeter-wavelength cloud radar (from the Atmospheric Radiation Measurement Program Southern Great Plains site at Lamont, Oklahoma) are sorted into these 25 states, producing vertical profiles of cloud occurrence. The temporal stability and distinctiveness of these 25 profiles are analyzed using a bootstrap resampling technique. A stable-state-based mapping from synoptic-scale model fields to local-scale cloud properties could be useful in three ways. First, such a mapping may improve the understanding of differences in cloud properties between output from global climate models and observations by providing a physical context. Second, this mapping could be used to identify the cause of errors in the modeled distribution of clouds?whether the cause is a difference in state occurrence (the type of synoptic activity) or the misrepresentation of clouds for a particular state. Third, robust mappings could form the basis of a new statistical cloud parameterization.