Keeping an advanced sea-ice forecast on track with advanced data assimilation

A new study shows that including satellite-based data for sea ice concentration and thickness through data assimilation can make our sea-ice forecasts with neXtSIM more accurate.

Sea ice in the Arctic. Photo: Espen Storheim / NERSCSea ice in the Arctic. Photo: Espen Storheim / NERSCSea ice is an extremely important factor in Earth's climate. Its white colour keeps the Arctic Ocean from warming, and with decreasing sea ice, the ocean warms up more. To understand its impact on the climate system, we need to know as much details as possible about it and its behaviour.

One available source are the satellites that monitor ice concentration, thickness, and motions. They give us an idea of the current situation but say nothing of the future. To produce forecasts for the near and further future (days to decades), computer models are necessary. Forecasts are relevant in the climate context but also in more immediate terms to keep human activity like shipping in the Arctic regions as safe as possible. But computer models tend to drift away from reality in the long run, so we use mathematical methods called “data assimilation” to recall them to reality and produce the best possible forecasts. When data is assimilated, a model is fed with types of data relevant to the forecast it is supposed to produced. This way the model can adapt to real-world conditions instead of “just” making an educated guess: The data assimilation process therefore adjusts and corrects the model, leading to more reliable forecasts. 

The neXtSIM model developed at NERSC is the most advanced sea-ice model today in terms of the sea-ice mechanics. It is very efficient at simulating how the ice cracks, converges, or diverges and that makes its simulations of sea-ice motions more accurate than the traditional models. But sea-ice behaviour is very complex beyond its mechanics, so knowing better the movements of the ice does not necessarily mean that the –model will forecast precisely where the ice is freezing, melting and how thick it will grow. So, we are left with the question: Can we assimilate the satellite observations of sea ice concentration and thickness without spoiling the soup? 

But one thing that we have learned in the past with the traditional sea-ice models in the Arctic is that we need advanced data assimilation methods like the Ensemble Kalman Filter, the kind of method that modulates the effect of the data depending on the "errors of the day”: The method will make a different correction whether the ice is melting, freezing, or being blown away and that generally saves the soup.

The new study under the lead of Sukun Cheng (previously at the Nansen Center, now at Dalian Maritime University, China), shows that it is possible to make the forecasts produced by our sea-ice model neXtSIM more accurate by using data assimilation. Cheng, together with colleagues from the Nansen Center, the UK, Italy, France, and the US, used satellite data of sea-ice concentration and thickness for the winter of 2019-2020. This is just a short period for testing, but their results show that it would be worth continuing with longer time periods. They also show that the forecasts by the model have become more accurate for both concentration and thickness of sea ice without deteriorating the sea-ice motions that we were so satisfied with.

The authors, along with other colleagues working on improving neXtSIM, hope to provide a fully operational sea-ice forecasting product over time, which will be beneficial for decision-making, planning shipping routes, and managing resources in the Arctic. Co-author Laurent Bertino (NERSC) on the versatility of their findings: “Several systems in other scientific and engineering disciplines such as explosions, forest fires, wind flow on turbines share analogous model features with an adapted mesh and could adopt the same solution.”



Cheng, S., Chen, Y., Aydoğdu, A., Bertino, L., Carrassi, A., Rampal, P., and Jones, C. K. R. T.: Arctic sea ice data assimilation combining an ensemble Kalman filter with a novel Lagrangian sea ice model for the winter 2019–2020, The Cryosphere, 17, 1735–1754,, 2023.


Add comment

Login to post comments