By: Adam Bateson
One of the most frequently used visual devices to illustrate climate change is that of a polar bear on sea ice surrounded by open ocean (Fig. 1). Polar bears are today identified as a vulnerable species, with sea ice decline the primary reason for this assessment. This is one of several reasons why we need to understand how the Arctic sea ice cover is likely to change in future. Sea ice is also of relevance to meteorology and climatology since the sea ice cover acts as a barrier to the exchange of heat, moisture, and momentum at high latitudes. Furthermore, the retreat of sea ice is a positive feedback in our climate system. Sea ice reflects over 50% of the incident solar radiation (snow-covered sea ice can reflect 90% of this radiation) whereas open ocean reflects only about 6%. The replacement of sea ice cover with open ocean therefore enhances any warming effect. This effect is referred to as the ice-ocean albedo feedback (or the ice-albedo feedback, a more general term).
Figure 1: Polar bears have been described as the ‘poster animal’ of climate change. Photo credit: Hans-Jurgen Mager at Unsplash.com.
It is therefore important to ensure that we have reasonable confidence in how the sea ice is likely to change in the future. However, a recent analysis of climate model performance in simulating the Arctic sea ice found that models generally underestimated the sensitivity of September sea ice area to a given amount of global warming (Notz et al., 2020). Why are climate models failing to adequately capture the sensitivity of sea ice area to warming? This could be due to an inadequate representation of the model forcing that drives changes in the sea ice cover e.g. wind speed, surface temperature. Alternatively, or in addition, it could be due to missing or poorly captured sea ice physics.
My own research focuses on improving the representation of sea ice physics within models of sea ice. Sea ice and its interaction with the ocean and atmosphere is complicated (Fig. 2). It is not possible to accurately represent all processes whilst maintaining computational efficiency within sea ice and climate models. Instead, we must identify and develop aspects of the model physics that will best enable us to answer research questions of interest.
Figure 2: A cartoon schematic to illustrate several important processes that determine the evolution of sea ice and its interactions with the atmosphere and ocean. Note this figure just gives an example of the complexity of the sea ice-ocean-atmosphere system and is far from exclusive in highlighting important physical processes. Figure is a reproduction of Fig. 2 from Lee et al. (2012).
Sea ice is made up of individual pieces of ice that we call floes. Observations of these floes show that they can range in size from scales of just metres to tens of kilometres (Fig. 3). However, sea ice models have historically assumed floes adopt a constant size if they explicitly consider the size of individual floes at all. The size distribution of a set of floes is generally referred to as the Floe Size Distribution, or FSD. Floe size can impact several processes within sea ice including the volume of melt from the side of floes (lateral melt) and momentum exchange between the sea ice and the atmosphere and ocean. A recent study found that sea ice extent can reduce by about 20% for a distribution of floes with a diameter of 3 m compared to floes with a diameter of 300 m (the standard value used in the Los Alamos Sea Ice model, more commonly referred to as CICE) just by accounting for the impact of floe size on lateral melt rate (Smith et al., 2022). This therefore makes the improved representation of floe size a potential target to improve sea ice model performance.
Figure 3: Sea ice is made up of individual pieces of ice called floes. These floes can vary in size from just metres to tens of kilometres. Figure is a reproduction of Fig. 3 from Williams et al. (2016) under CC BY 3.0 (https://creativecommons.org/licenses/by/3.0/us/).
There are several processes thought to influence floe size including lateral melting, wave break-up, brittle fracture, and floes welding together. Several recent studies have proposed models of the FSD: some prioritise physical fidelity by allowing the shape of the FSD to emerge from the model (e.g. Roach et al., 2018, 2019); whereas others prioritise computational efficiency and make assumptions about FSD shape (e.g. Bateson et al., 2020). A recently submitted study compared these different approaches to modelling the FSD and considered the advantages of disadvantages of each (Bateson et al., 2021). In practice, the better approach depends on both the research question at hand and the resources available. To return to the research question under consideration here, the study found that, when optimised against observations of floe size, neither FSD model produced a significant improvement in simulating sea ice area. However, this study also highlighted several limitations in this conclusion such as impacts of floe size not considered in the study e.g. on the rheology of sea ice.
It is unlikely that there is going to be a single ‘silver bullet’ to improve the representation of Arctic sea ice in climate models. Progress will be incremental as we improve our understanding of relevant physical processes in sea ice and better capture atmospheric and oceanic forcing of sea ice.