Natural Variability of Arctic and Antarctic Sea Ice : Insight From Pre-industrial CMIP5 Runs

Abstract

Understanding natural variability of Arctic and Antarctic sea ice is important for explaining and predicting observed sea ice variability. Motivated by this, I undertook this study that analyzes sea ice variability in pre-industrial control runs for a subset of Earth System Models (ESMs) and Climate Models (CMs) from the Coupled Model Intercomparison Project Phase 5 (CMIP5), and the Alfred Wegener Institute Climate Model (AWI-CM). I compare the general characteristics of simulated mean sea ice conditions in the Arctic and Antarctic for the selected models. I find that the main characteristics of Arctic sea ice are better simulated than that of Antarctic sea ice. None of the selected models were able to simulate all of the Antarctic sea ice characteristics properly.Inter-model spread of winter sea ice volume and summer sea ice extent in both hemispheres is large, with the spread exceeding the limits of natural variability. The drivers of the Arctic sea ice minimum in September and the Antarctic sea ice minimum in February are studied using composite analysis. In the Arctic, events of low sea ice extent in September are related to (1) Sea Ice Thickness (SIT) Memory (2) a negative Sea Level Pressure (SLP) and a positive Sea Surface Temperature (SST) anomaly in the Barents Sea in March, (3) the deepening of the Aleutian Low in March with imprints of ENSO and PDO, and (4) a positive SLP anomaly in the Beaufort Sea in July. I also find that events of low September sea ice extent minima tend to cluster in time. I suggest that in addition to low frequency climate variability, clustering of low September sea ice extent events depends on the memory of the Arctic sea ice system provided by SIT and ocean surface heat content. The occurrences of the aforementioned atmospheric patterns are random, in the sense that they are not necessarily taking place in the same years. An important characteristic of the events analyzed is an anomalous thin Arctic sea ice cover in winter that preconditions these late summer events. The anticyclonic SLP anomaly in the Beaufort Sea develops in summer, so if the sea ice cover is in a fragile and weak state, atmosphere-sea ice coupling will be even more efficient in reducing sea ice extent via Ekman convergence. In the Antarctic, events of low sea ice extent minimum in February are related to (1) a positive anomalous Southern Annular Mode (SAM) in winter that enhances sea ice divergence creating widespread negative anomalies in Sea Ice Concentration (SIC), (2) a negative anomalous SAM in summer and (3) a broad positive SST anomaly in the Southern Ocean associated with the negative SAM and SIC decline. Similar to the Arctic, events of low February sea ice extent minima tend to cluster in time, however with a lower likelihood of a February sea ice extent event to be followed by another. The memory of clustering in the Antarctic is provided by low frequency variability and the memory of the sea ice system that is predominately associated with ocean surface heat content anomalies. In both hemispheres, the reduction of sea ice extent and volume in years characterized by a low sea ice extent minimum, causes a strong negative net Short Wave Radiation (net SWR) anomaly in spring. This leads to enhanced sea ice melt via the positive sea ice-albedo feedback. From the composite analysis I suggest that the mechanism for producing low sea ice extent minima events in both hemispheres are similar. Such a mechanism involves the thinning of the sea ice cover in winter, enhanced sea ice albedo feedback in spring, and the efficient role of an anomalous anticyclonic SLP pattern in summer.