No abstract

No abstract

This preliminary study employs Wolfe's cluster analysis program (NORMIX) in the bivariate normal classification of North American air masses. Once-daily observations of potential temperature and mixing ratio were classified by the NORMIX program into 'types' which represent different air masses. The cluster analysis was performed at the surface, at 850 mbar, at 700 mbar, and at 500 mbar for each of seven stations located along the 80th meridian. Individual data sets were composed of 10 years of observations taken during January or July at each level.Analyses derived from the NORMIX results showed more types at the middle latitudes than at source region latitudes. A different number of NORMTX types was found at each level above each station, except at Miami in July. Frequency tables for the Greensboro station showed that some NORMIX types at upper levels occurred in conjunction with two or more surface NORMIX types, thereby indicating the presence of overrunning. Synoptic analyses for the Greensboro station showed that each NORMIX type was associated with a specific synoptic air mass type.The NORMIX program generated discriminant functions which were used to objectively classify independent potential temperature and mixing ratio data for the Greensboro station. The results of the independent data classification compared favourably with synoptic air mass types and with the original NORMIX results.