SIGNIFICANCE TEST FOR EMPIRICAL ORTHOGONAL FUNCTION (EOF) ANALYSIS OF METEOROLOGICAL AND OCEANIC DATA

摘要

A Monte Carlo experiment was done to create a statistically significant test table for selecting EOF modes whose signals are above the noise level. Numerical analysis showed that it is feasible to choose a shorter data set between temporal and spatial dimensions for conducting an EOF analysis, and thus simplify calculation of the covariance matrix and solution of its eigenvalues and eigenvectors. The ex- periment showed that the dominant EOF modes remained the same. This approach is more computationally efficient, reducing both computing time and computer memory requirement, while still retaining the domi- nant signal.