Interannual variability of patterns of atmospheric mass distribution

Using the 40-yr European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-40) for 1958 to 2001, adjusted for bias over the southern oceans prior to 1.979, an analysis is made of global patterns of monthly mean anomalies of atmospheric mass, which is approximately conserved globally. It differs from previous analyses of atmospheric circulation by effectively area weighting surface or sea level pressure that diminishes the role of high latitudes. To examine whether global patterns of behavior exist requires analysis of all seasons together (as opposite seasons occur in each hemisphere). Empirical orthogonal function (EOF) analysis, R-mode varimax-rotated EOF analysis, and cyclostationary EOF (CSEOF) analysis tools are used to explore patterns and variability on interannual and longer time scales. Clarification is given of varimax terminology and procedures that have been previously misinterpreted. The dominant global monthly variability overall is associated with the Southern Hemisphere annular mode (SAM), which is active in all months of the year. However, it is not very coherent from month to month and exhibits a great deal of natural unforced variability. The third most important pattern is the Northern Hemisphere annular mode (NAM) and associated North Atlantic Oscillation (NAO), which is the equivalent Northern Hemisphere expression. Neither of these is really a global mode, although they covary on long time scales in association with tropical or external forcing. For monthly data, the second mode is coherent with Nino-3.4 sea surface temperatures and thus corresponds to El Nino-Southern Oscillation (ENSO), which is truly global in extent. It exhibits more coherent evolution with time and projects strongest onto the interannual variability, where it stands out by far as the dominant mode in the CSEOF analysis. The CSEOF analysis extracts the patterns phase locked with annual cycle and reveals their evolution throughout the year. Standard EOF and varimax analyses are not able to evolve with time of year unless the analysis is stratified by season. Varimax analysis is able to extract the SAM, NAM, and ENSO modes very well, however.