Abstract

The purpose of this note is to encourage the use of field-weighting over field-interpolation in achieving grid-area parity in principal component analysis of geophysical data. While field-weighting has been shown to achieve exact parity in unrotated analysis, we show that it is highly effective in rotated analysis too, by comparing the structures of the three leading modes of 200mb geopotential height variability in the 1958-98 NCEP reanalysis winters obtained from both methods.



Table-I: Percentage variance explained by the three leading modes of 200mb geopotential interannual variability in the 30S-87.5N domain in the 1958-98 NCEP reanalysis winters (D,J,F). The analysis is conducted using the [cos(Lat.)]½ and [cos(Lat.)] weightings, and also with field interpolation onto the equal-area grid. Note the closeness between the [cos(Lat.)]½ weighting and interpolation based results, and the change in the order of the modes with [cos(Lat.)] weighting.

Pattern [cos(Lat.)]½ weighting Interpolation [cos(Lat.)] weighting
NAO 15.14% 16.39% 9.88%
TNH 13.38% 13.16% 19.57%
PNA 11.75% 11.18% 12.55%