SOM-based circulation types in the Mediterranean basin from reanalysis and CMIP6 models

Abstract

<p><span>The aim of this work is to produce a climate atmospheric circulation classification via Self-Organizing Maps (SOM) and to identify and evaluate the respective changes in the frequency of the identified regimes for GCM projections (Coupled Model Intercomparison Project phase 6, CMIP6). The main focus is to assess the ability to accurately represent the large-scale circulation over the Mediterranean and to generate atmospheric circulation regimes that can be used as an explanatory tool in multiple research fields. The classification framework uses an unsupervised learning algorithm for a low-dimensional representation of high-dimensional datasets (SOM) to identify non-linear relationships and patterns from complex spatiotemporal climatological fields. Upon the selection of the atmospheric variables and the corresponding spatial and temporal scales, the SOM framework will be applied initially to the historical period climate series of large-scale atmospheric circulation from reanalysis datasets and following to a multi-model ensemble of GCM simulations for multiple Socioeconomic Pathways (SSPs). The resulting atmospheric circulation regime catalogs will be compared and discussed in terms of the representativeness of the large-scale circulation by the GCM climate models for the Mediterranean domain. The catalogs will assess the corresponding changes in atmospheric circulation, focusing on the regimes’ frequency of occurrence, persistence, and transition probabilities</span></p>