Identifying coupled modes of variability through singular vector decomposition and extracting climate trends

In determining the predictability of the regional impacts of interannual modes of variability such as the ENSO cycle, decadal trends must be considered. We identify coupled ocean-atmosphere modes of seasonal climate variability using singular vector decomposition of the regression matrix of transformed ocean and atmosphere data fields. Variant time series for the modes of variability identified are then regressed on North American regional temperature and precipitation fields to determine the impacts of the identified large scale modes. The resulting projections of large scale modes on the regional data are next subtracted leaving residual variability. Once modes of seasonal variability are extracted from the regional temperature and precipitation data, a lagged regression analysis on regional climate variables is constructed to determine predictable trends. Variations in the strength and pattern of decadal trends over time are examined. It is shown that the predictability of decadal trends is enhanced by extraction of interannual modes of variability. By performing the SVD analysis of the regression matrix of large-scale to regional data, we examine if the regional impact of large-scale modes of interannual variability can be determined directly and compare to the previous results.