Comparing the empirical orthogonal function and the eigen microstate approach for surface air temperature studies

Abstract

Temperature is a climate parameter with a complex pattern and distribution. A thorough understanding of this pattern and complexity requires precise analysis. This study analyzed surface air temperature variability using the eigen microstate approach (EMA) and empirical orthogonal function (EOF) methods based on NCEP-NCAR data. The results show that the EOF variance distribution is concentrated in the initial mode (EOF 1 at 88.42%), whereas the EMA variance is more spread out, with contributions > 1% up to the sixth mode. The resulting spatial pattern is consistent with previous studies, although it exhibits phase differences and higher variance in EMA. Fourier spectral analysis shows that EOF emphasizes annual patterns, whereas EMA can separate multiscale signals across semi-annual, annual, and interannual periods. Furthermore, EMA successfully detected important global impacts linked to regional climate phenomena such as monsoons, El Niño Southern Oscillation (ENSO), ENSO Modoki, and the Indian Ocean Dipole, which are less clear in higher EOF modes due to their relatively small variance. Thus, EOF is effective for identifying the global most dominant patterns, while EMA is superior in capturing multiscale variability and global impacts linked to regional climate phenomena.