Clustering of Sea Surface Temperature Variability Across the Large Ensemble

Co-Author: Naomi Goldenson
Co-Author: Andrew Rhines
A consistent picture of modes of variance across the large ensemble is helpful for interpreting the model climate, as well as applications like the selection of ensemble members for downscaling. Calculating the empirical orthogonal functions (EOFs) for each ensemble member independently yields similar but non-identical patterns for the principal modes of variance. We show the results of calculating one consistent set of EOFs using all of the ensemble members at once. Then we use the results to conduct a clustering exercise on the (consistently defined) principal components (PCs). This can be done in the time as well as frequency domains. Finally we show the application of the approach to select ensemble members to force prescribed sea surface temperature simulations to study Northwest regional climate. The method selects ensemble members that span a range of regional trends, while preserving information about the relative frequency of similarly clustered ensemble members.