Why is Antarctic sea ice so hard to model?

Climate models are our best tools for making projections of future climate. However, over recent decades climate models have largely failed to reproduce the observed behaviour of Antarctic sea ice (the layer of frozen ocean surrounding the continent). In general, Antarctic sea-ice projections are viewed with low confidence. Modelling Antarctic sea ice is difficult because many processes affect its formation and melt. Those factors can change from day-to-day, and over decades. Gaining a better understanding of Antarctic sea ice’s evolution and dynamics is important because of the global implications¹.

Why is sea ice so difficult to model and what are we doing about it?

Global climate models (GCMs) were first created in the 1960s and 1970s, and have they have become substantially more sophisticated since then. The focus of early GCMs was to answer one main question: How will the Earth’s temperature change with increasing carbon dioxide emissions from human activities? A recent study shows that GCMs produced between 1970 and 2007 did a good job of forecasting surface warming².

Today, climate models are run on supercomputers far more powerful than those that ran earlier models. Models can now include many more of the complex interactions between the atmosphere, land, ocean and sea ice. They also resolve these processes in greater spatial detail than the original GCMs. This allows modern climate models to answer the original question in more detail, and to provide new insights into more aspects of the climate system.

Despite these advances, the behaviour of Antarctic sea ice has proved difficult to simulate. In the latest collection of climate model simulations, only a few produced a trend in Antarctic sea-ice area that was consistent with observations over the period 1979-2018. Most models tend to underestimate summer sea-ice cover and predict strong negative trends that have not been seen in observations³.

In the real world, the annual average extent of sea ice around Antarctica was increasing over those decades, reaching a record high in 2014, in stark contrast to the decline seen in the Arctic. (Check out the animation above or via YouTube showing the modelled Antarctic sea-ice concentration in 2014.)

Antarctica’s sea ice extent high was unexpectedly followed by a record low minimum extent in 2017. Since then, the area of sea ice has generally declined, with a new record low minimum in February 2022, and again in February 2023. This dramatic decrease is the subject of intense research. Proposed reasons for the change have included the combination of a strong El Niño event and coincident variability of the winds around the Antarctic continent, leading to warmer-than-average surface temperatures in the Southern Ocean^4,5.

Scientists have also been asking why models generally fail to reproduce the increase in Antarctic sea ice from 1979 to 2014. Suggestions include wind changes⁶, random variability in the climate system⁷, changes in stratospheric ozone⁸ (which was later shown to cause sea-ice loss⁹), and freshwater from Antarctic ice shelves^10-15. None of these has proven to be a definitive answer, so far.

Simulating Antarctic sea ice in a way that is consistent with observations requires more than just an understanding of the physics of the ice itself. Sea ice is tightly coupled to the atmosphere above it and the ocean below it. Changes in air temperature control sea-ice freezing and melting, and winds control how the ice is distributed around the continent. Similarly, changes in ocean temperature have a large impact on sea-ice formation and melt. Models must get all of these factors right to accurately reproduce past sea-ice behaviour and to project future behaviour, to stand a chance of informing us what might happen next in the real world.

The Antarctic Science Platform supports cross-disciplinary research that aims to improve our ability to model Antarctic Sea Ice more realistically. This includes new approaches, led by Modelling Hub researchers, to couple atmospheric and ocean processes (to the extent they affect sea-ice dynamics), and to better represent meltwater generation around Antarctica to capture the influence of melting ice shelves. The Platform is also supporting a mix of ocean and sea-ice field observations that allows specific processes to be measured, so that these can then be factored into model development.

Sea-ice evolution is critically linked with ocean circulation, global climate and marine ecosystems. Until we can understand the processes that control sea-ice evolution sufficiently well to understand its future sensitivity to a changing climate, we will be unable to fully comprehend the size, scope and speed of changes that we can expect across the globe in coming decades.

Article prepared by Andrew Pauling, Inga Smith and Max Thomas (Department of Physics, University of Otago).