Eigenvector Filtering of Three-Dimensional Pressure Field Data

Abstract

Grid-point data on the MSL pressure and 500 mb height fields over western North America are filtered with the eigenvectors of their covariance matrix. The filtering allows virtually all the information in 134 grid-point data for each day of the 8-year sample to be represented by 15 eigenvector coefficients. The three-dimensional eigenvector patterns are meteorologically coherent, and the filtering process can be used as a screen for bad data or for anomalous circulation patterns.