The ArcticWATCH project: rapid sea ice loss as an indicator of global warming

Researchers from the ArcticWATCH project study and model sea ice, an important indicator of global warming. Using complex models, which they refine through their research, they predict the future of this sea ice and its consequences for the global climate.

Globally, the planet's climate is currently experiencing human-induced global warming, which has been scientifically documented for decades. However, the global climate is not warming at the same rate everywhere on the planet.  

At the North Pole, the Arctic region is warming 3 to 4 times faster than the global average. Currently, a global warming of approximately +1.5°C (compared to the pre-industrial era) has been established. For the Arctic, this means that we are therefore at approximately +4.5°C of warming, which is enormous and extremely problematic for the ecosystem of the area.  

There are three major types of cryosphere (i.e., solid water) across the globe:  

• Snow  
• Land ice, namely glaciers and polar ice caps  
• Sea ice, which consists of frozen ocean water when temperatures reach -30°C to -40°C in the coldest areas of the Earth

This sea ice (generally referred to as “banquise” in French) is a valuable indicator for measuring global warming at the North Pole, as its quantity and extent change significantly in response to this warming. In recent years, there has already been a sharp decline in sea ice, with approximately 50% of its summer surface lost since measurements began in 1978.  

It is this sea ice that is the focus of the work of François Massonnet's team, a researcher at the Earth and Life Institute (UCLouvain). As part of the EU-funded ERC research project ArcticWATCH, François Massonnet and his team monitor, study, and model Arctic sea ice. 

While the loss of sea ice surface area is well documented and observed, it is not always uniform. In some summers, the loss is greater than in others, and scientists do not yet fully understand why. Is it related to the ocean or the atmosphere? Is it predictable or not? What are the impacts on the global climate system? These are all questions that the ArcticWATCH team is trying to answer.

To do this, they focus on specific events known as RILEs, or Rapid Ice Loss Events. RILEs are periods of sudden and significant reduction in sea ice extent, characterized by fluctuations that add to the long-term trend of sea ice decline.  

Recently, researchers involved in the project published three new scientific articles with some alarming conclusions.

A high probability of a future episode of sudden sea ice loss

In her article “Seasonality and scenario dependence of rapid Arctic sea ice loss events in CMIP6 simulations”, Annelies Sticker catalogued RILEs events, i.e., abrupt sea ice loss, in a diverse set of numerical models.

In climatology, projections are made using simulations run on numerical models. These models are like virtual laboratories, in which scientists add many different processes before running a simulation and analyzing the results for the past, present, and future of the planet Earth.

On a global scale, around 50 climate modelling groups run numerical models that are used to perform simulations and projections for the global climate. These groups are taking part in the latest phase of collaboration under the heading “CMIP-6” and studies based on these simulations feed the various IPCC reports. The advantage of using numerous models rather than one or two is that not all models necessarily take the same parameters and processes into account. Combining the simulations from each of these models therefore provides a more accurate picture of the global climate than a simulation from a single model.  

Annelies Sticker used about 26 to predict future RILEs events. Following her simulations, the researcher found that 60% of these models predicted at least one RILE event by 2030, which represents a very high probability of such an event occurring.

In addition, the doctoral researcher also found that 30% of these abrupt events even led to almost total sea ice disappearance in summer. If this were to happen, the global climate would face a significant risk of a feedback loop: less sea ice -> less reflection of solar radiation back into space -> more warming -> further reduction in sea ice -> etc.  

Beyond the climatic impacts, it should also be noted that the disappearance of the sea ice by the end of the summer would also have impacts on several other levels. In the Arctic, the ice acts as a cover for the ocean and regulates the phytoplankton cycle, which is the basis of the food chain in the oceans. If it disappears, many other animal species could also disappear, causing disruption to the entire ecosystem in this region.  

Finally, an Arctic Ocean that is seasonally ice-free would become a fully navigable ocean, leading to a total global geopolitical reconfiguration.

Studying extreme events to better understand and predict them

In his article “Ensemble design for seasonal climate predictions: Studying extreme Arctic sea ice lows with a rare event algorithm”, Jerome Sauer focused on the most extreme scenarios that could occur for sea ice, using a rare event algorithm from the field of statistical physics.

These extreme scenarios, which involve a drastic reduction in sea ice, remain poorly understood and understudied because these events occur so rarely. However, there is a possibility that they occur, and if this were to happen, we would need to be prepared. This means being able to predict what to expect and to understand the impacts on the global climate.

Jerome Sauer was able to drastically increase the simulated number of extreme events thanks to the rare event algorithm. Starting from an initial climate situation, the doctoral researcher calculated 600 possible trajectories for the evolution of sea ice. Thanks to these simulations, the return time of extreme events involving drastic reductions of sea ice can now be better estimated.  

Such an event already occurred in 2012, when a very extreme RILE was observed, with an absolute record for sea ice reduction.  

If the global climate were to remain stable, Jerome Sauer has calculated that such an event would recur on average every 1000 years. However, the climate is not stable. The probability of it recurring much sooner is therefore very high.  

Using AI to better predict the future of sea ice  

In her article “Probabilistic Forecasts of September Arctic Sea Ice Extent at the Interannual Timescale With Data-Driven Statistical Models”, Lauren Hoffman used artificial intelligence (AI) to run simulations usually performed by numerical models, to see if this tool could prove reliable and help researchers make sea ice predictions.  

AI approaches work very differently from the traditional climate models scientists have relied on for decades. Instead of solving physical equations step by step, AI learns patterns directly from data — a process that can feel like a “black box,” since we see the predictions but not every step the model took to make them. This approach is much faster and could eventually be more accurate, opening the door to new ways of forecasting sea ice change.

Through her research, Lauren Hoffman has shown that AI-based models are capable of making reliable predictions of September sea ice for the next five years.  

The postdoctoral researcher then used this model to predict future sea ice loss, and it appears that September sea ice should remain stable over the next five years.

Research to better predict tomorrow's climate

At the forefront of research on the state of sea ice in relation to global climate and climate change, the ArcticWATCH team has made major advances in the use of scientific models. The more sophisticated and refined these models are, the more accurate their predictions will be, and the better we will be able to predict and prepare for the climate of the future.  

References

Hoffman, L., Massonnet, F., & Sticker, A. (2025). Probabilistic Forecasts of September Arctic Sea Ice Extent at the Interannual Timescale With Data-Driven Statistical Models. Journal of Geophysical Research: Machine Learning and Computation, 2(3), e2025JH000669. https://doi.org/10.1029/2025JH000669 

Sticker, A., Massonnet, F., Fichefet, T., DeRepentigny, P., Jahn, A., Docquier, D., Wyburn-Powell, C., Quint, D., Shivers, E., & Ortiz, M. (2025). Seasonality and scenario dependence of rapid Arctic sea ice loss events in CMIP6 simulations. The Cryosphere, 19(8), 3259–3277. https://doi.org/10.5194/tc-19-3259-2025 

Sauer, J., Massonnet, F., Zappa, G., & Ragone, F. (2025). Ensemble design for seasonal climate predictions: Studying extreme Arctic sea ice lows with a rare event algorithm. Earth System Dynamics, 16(3), 683–702. https://doi.org/10.5194/esd-16-683-2025 

Article: Emmeline Van den Bosch, François Massonnet, Annelies Sticker, Jerome Sauer, Lauren Hoffman