Abstract
Rotated principal component analysis (RPCA) is a powerful tool for studying upper air height data because of its ability to distill information about the variance existing in a large number of maps to a much smaller set of physically meaningful maps which together explain a large fraction of the variance of she input dataset. However, in order to achieve this, one faces the problem of deciding how many eigenmodes to rotate. A discussion of the dangers of incorrectly choosing the rotation point and a quasi-objective technique that leads to a good compromise between over- and underrotation are presented. Finally, the use of RPCA for detecting errors and inconsistencies in upper air data along with two examples is discussed.
