Record sea ice minimum in reality and climate models

A new record minimum of Arctic sea ice cover has been observed in late August 2012. The event is a consequence of a downward trend in sea ice thickness and area. Since the start of the satellite era in 1979, it is likely about two thirds of the Arctic sea ice volume has been lost.

The underlying reason for the sea ice decline is global warming due to increased greenhouse gas emissions with Arctic warming being observed to be twice as strong as global mean values. Warming atmosphere and oceans is inevitably causing ice melt. The forcing and feedbacks involved are complex in detail. Still, climate models have shown a sea ice decline in response to future emission scenarios; however, global climate models (GCMs) have been shown to underestimate the pace of sea ice reduction. In this article, we compare observed trends and record events with climate model results. Although the efforts towards a new IPCC report are still ongoing, first reports point at simulated sea ice decline either too slow or too late. A central problem is the correct representation of sea ice thickness in coupled models, subject to various influences and feedbacks.

Already in 2007, when the previous record (4.3 million km2) was reached, scientists noted a discrepancy between the observed decline and corresponding simulations published by IPCC (AR4). The simulated sea ice decrease as an average over several global climate models ("ensemble") was clearly too conservative in terms of long-term trend and amplitude of events. However, the use of ensemble means is masking variability of individual model simulations.

It appeared that the inter-annual variability differs among the models. The stronger inter-annual variability a model shows, the more likely are events due to the highly non-linear nature of the Arctic climate system and feedback mechanisms, such as sea ice-albedo feedbacks or interaction between sea ice and atmospheric circulation changes. Different representation of processes and different sensitivity leads to a range of different amplitudes of sea ice variability. A systematic analysis of variability in AR4 models confirms that internal variability is a leading factor influencing the magnitude of the downward trend in both Arctic sea ice extent and volume (Wettstein and Deser 2011). Based on limited information on individual GCM runs, at least one of the AR4 runs was showing a sea ice cover variability of the same order as observed.

What is the situation today?

Since 2007, global and regional climate models have developed further. Most important for the polar areas, sea ice models have been improved among other variables.

Regional Arctic models have shown strong sea ice variability and ice extent trends in the same order of magnitude as the current observations. Rapid sea ice reduction events can be related largely to large scale atmospheric circulation anomalies in combination with thinning and increasingly sensitive ice conditions. (Döscher and Koenigk, 2012)

A new effort towards IPCC (AR5) is under way, organized by the Coupled Model Intercomparison Project phase 5 (CMIP5). The first scientific articles on individual global CMIP5 models have already been published.

The European community climate model EC-Earth (with significant contributions from Rossby Centre) performs scenario simulations with a realistic interannual sea ice variability, but initially too thick ice and a too weak summer trend (Koenigk et al. 2012). Not before 2040, sea ice decline accelerates to a trend similar to the recent observed one.

Another example originates from the US community model CCSM4 (Vavrus et al. 2012). Sea ice extent trend is underestimated in the ensemble average, but not for selected realizations. During certain decades, the trend accelerates to observed values.

In a first overarching effort, a study of 29 current CMIP5 models show that the trend of sea ice extent clearly relates to ice thickness (Massonet et al. 2012). It is also shown that the decadal trend towards less ice is increasing with decreasing sea ice extent up to a maximum decadal trend. Further ice reduction occurs at slower decadal trends. Although the ensemble mean of the 29 CMIP5 models still underestimates the observed sea ice reduction, a larger number of individual models, compared to CMIP3, simulates realistic ice extents and at least period wise realistic trends.

For additional robust conclusions from the ongoing CMIP and CORDEX efforts, we will have to wait for upcoming results. In the meantime, we summarize the existing studies as follows: Many GCMs are still underestimating the sea ice decline trend on the real time axis, especially in the light of the September 2012 observed sea ice minimum record. Nevertheless, rapid sea ice reduction events occur in a number of models together with realistic trends during certain periods. On the real time axis however, such events are delayed. As the sea ice thickness appears to play an important role for timing, ongoing development needs to focus on realistic thickness during the start phase of the scenarios, e.g. around year 2000. Historically, such model verification of ice thickness is a challenge as integrated Arctic-wide thickness observations does not exist. Only recently, first thickness products from the Cryosat-2 satellite appeared.


Döscher, R. and Koenigk, T.: Arctic rapid sea ice loss events in regional coupled climate scenario experiments, Ocean Sci. Discuss., 9, 2327-2373, doi:10.5194/osd-9-2327-2012, 2012.
Koenigk, T., L. Brodeau, R. G. Graversen, J. Karlsson, G. Svensson, M. Tjernström, U. Willén and K. Wyser, 2012: Arctic climate change in 21st century CMIP5 simulations with EC-Earth. Climate Dynamics 2012, DOI:10.1007/s00382-012-1505-y
Massonnet, F., Fichefet, T., Goosse, H., Bitz, C. M., Philippon-Berthier, G., Holland, M. M., and Barriat, P.-Y.: Constraining projections of summer Arctic sea ice, The Cryosphere Discuss., 6, 2931-2959, doi:10.5194/tcd-6-2931-2012, 2012.
Wettstein and Deser 2011,