Using self-organizing maps to investigate abrupt reductions in Arctic sea ice extent: an application to winter and summer pan-Arctic daily sea level pressure data

Abstract

Self-Organizing maps (SOMs) have been used to explore potential connections between atmospheric circulation over the Arctic (in the form of sea level pressure patterns and teleconnection indices) and sea ice loss. Extended periods of abrupt loss (rapid ice loss events, or RILEs) have been studied, along with high ice loss years (with instantaneous ice loss years >0.45 million km2). The bulk of the Thesis focuses on ice loss in 21st century projections from general circulation models (GCMs), with complimentary analysis of the observational record as represented by the NCEP/NCAR reanalysis. In addition to examining the frequency of specific weather systems (e.g. Icelandic lows, Beaufort Highs etc.), a SOM-based analogue of traditional teleconnection analysis was applied, which emphasizes impacts on the Arctic and allows for asymmetry between the positive and negative phases of a teleconnection. Results show no simple connection between sea level pressure patterns and RILEs, either in winter or summer analyses; although statistically significant circulation anomalies were identified for individual events, no common potential cause emerged. Results from composites of large ice loss years show fewer significant anomalies than found in RILE periods, but the sign of anomalies is more consistent. This suggests atmospheric drivers of ice loss are better demonstrated over short time scales (a year) rather than longer periods (years to decades). Identified circulation anomalies reflect increased winds along the transpolar drift, and these broadly resemble previously proposed ice loss mechanisms associated with the North Atlantic Oscillation, Arctic Oscillation, and Arctic Rapid Change pattern.