Arctic synoptic activity associated with sea ice variability using self-organizing maps

• Main Author: Mills, Catrin
• Other Authors: Walsh, John E., Cassano, John J., Nesbitt, Stephen W., Wang, Zhuo
• Format: Text
• Language: English
• Published: 2014
• Subjects: Arctic meteorology; Arctic climate; Self-organizing maps (SOMs); Arctic; Fram Strait; North Atlantic; North Atlantic oscillation; Sea ice
• Online Access: http://hdl.handle.net/2142/49755

Relationships between synoptic activity and sea ice variability in the Arctic are studied using self-organizing maps (SOMs) to categorize observed weather patterns over the 1979-2010 period. The European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-interim, or ERAI) provides the daily sea level pressures from which the SOMs are computed. Time series of frequencies and durations of synoptic weather patterns are correlated with two sea ice metrics, Fram Strait ice outflow and year-to-year changes in September pan-Arctic sea ice extent. When compared to teleconnection indices commonly associated with sea ice variability, the Arctic Oscillation (AO), North Atlantic Oscillation (NAO), and Arctic Dipole (AD), some SOM patterns correlate more strongly with sea ice metrics. For example, Beaufort High synoptic patterns are increasing in frequency in spring and summer and their spring frequencies are associated with ice loss. Icelandic Low patterns show opposing influences on sea ice from wind-forcing and thermal advection. The phase lags between the SOM occurrences and sea ice variability offer the potential for augmentation of other approaches to seasonal sea ice prediction. The ERA-interim SOM analysis is used to quantify how the Community Climate System Model, Version 4 (CCSM4) captures synoptic activity in the twentieth and twenty-first centuries. The model undersimulates patterns important for ice loss, such as broad high pressures over the continents and ice cover, and simulates strong storm track features at a higher-than-observed frequency. Large-scale teleconnection patterns, such as the AO and AD, are reasonably captured but there are spatial shifts in centers of action (which are associated with ice motion biases) and enhanced interannual variability relative to the observations. Relationships between synoptic activity and year-to-year changes in sea ice extent are not as prominent in the 20th century model experiment and further weaken in the 21st century. Accounting for seasonal SLP biases in the model enhances SOM frequency-ice correlations, suggesting that the model captures closer-to-observed atmosphere-ice linkages when SLPs are reasonably simulated. Strong low pressures over the Arctic predominate over the ice-free central Arctic during summer and fall in the 21st century.