While scientists have observed and studied the rapid decline in Arctic sea ice over the past few decades, a high level of uncertainty remains about how the rate of ice loss will change over the coming years. Machine learning techniques have the potential to help us untangle the underlying mechanisms in order to reduce these uncertainties. Providing reliable forecasts of likely Arctic sea ice cover would help improve regional and global climate model predictions.
Explaining the science
The decrease of Arctic sea ice cover is concerning as its white surface reflects up to 80 percent of incoming sunlight, deflecting additional energy away from the Earth's lower atmosphere and back towards space. With a declining spatial extent of sea ice, the dark ocean surface which replaces it absorbs considerably more sunlight energy. This leads to a self-perpetuating cycle with further warming of the atmosphere and surface waters and more melting of ice. This process is known as ‘ice-albedo feedback’ and is most prevalent during the summer months. Conversely, in the darker months other climate drivers, such as variability in oceanic currents and atmospheric weather patterns, are more likely to dominate the variation in sea ice.
This research is being undertaken by the British Antarctic Survey (BAS) AI Lab.
The aim of this research is to further our understanding of the complex inter-connectedness of large-scale climate phenomena and sea ice variability in order to reduce uncertainties in future climate predictions.
Applying machine learning techniques in new ways to this area will enable us to work more efficiently and effectively than using current climate model simulators alone, not least as these simulators are highly computationally expensive. The machine learning predictions generated within this project will be benchmarked against those from state-of-the art climate simulators to track ongoing improvements.
The vast areas of white ice and snow in the Arctic region is responsible for reflecting sunlight back to space, keeping our planet cool and regulating regional and global weather patterns. It is also an extreme living environment which supports a wide range of indigenous people and wildlife.
Being able to better predict the likely fluctuations in sea ice extent across the Arctic region, and the impact this can have on the global climate, is fundamentally important as these changes will effect weather patterns which in turn will have impacts on food, water, and economic security across the globe. Furthermore, improving our understanding of the potential changes facing this irreplaceable environment can also be used to support wildlife conservation efforts and indigenous peoples around the polar regions, whose way of life is under threat through environmental change.