We have all likely experienced a quick change in the weather in a given week, or even in a single day, but many think of Earth’s climate as relatively stable.  Yet there is evidence in Earth’s history of the climate changing very rapidly, on the timescales of seasons to years, often with little to no warning.  These abrupt shifts in the climate are concerning because they may happen faster than we as a society are able to adapt.

It is argued that the greatest concentration of potential abrupt transitions in Earth’s system today is located in the Arctic, one of which is Arctic sea ice.  The current generation of global climate models produces a broad range of future sea ice projections.  Given the uncertainty of the future of Arctic sea ice, and the Arctic’s role in global climate change, diagnosing the risk of an impending shift is important.  However, there are significant challenges in forecasting these sudden changes because they often occur with little to no change in the state of the system.

Using a global climate model, a series of annual sea ice forecasts from 1961-2080 were developed using the predictor period 1862-1960.  A measure of the average skill of the forecasts was found to be a useful metric in diagnosing the risk of a rapid shift in sea ice.  Each 12-month sea ice forecast is found using an analog forecast method, which essentially searches the past climate record (1862-1960) for the five years that are most similar to the first month of the current forecast year, and the month prior (analogs).  The average of the five analogs becomes a single sea ice forecast.  This process is repeated for each year in the forecast period.

The analog forecast is compared to what’s known as a CLIPER forecast, which is a combination of CLImatology (average of all the years prior to the forecast) and PERsistence (tomorrow’s weather will be the same as today).  This provides us with a measure of the skill of the analog forecast; with a value of one meaning the analog forecast has skill.

The figure below shows the evolution of Arctic sea ice for the month of September for a particular global climate model.  In 2015, this model begins to experience a catastrophic change in sea ice, with an 83% drop in sea ice concentration by 2019, transitioning to an ice-free Arctic by the end of the record (grey line).  The blue line shows the analog forecast skill and we can see that our analog forecast exhibits skill over a CLIPER forecast until roughly 1990, when a sharp decrease in skill is observed.  This drop in predictability begins roughly 25 years before the abrupt shift in sea ice concentration, providing a forewarning of the impending sea ice collapse.

So while the reliability of future sea ice predictions would likely decrease as we near a threshold, this model analysis suggests that a forewarning of an impending rapid shift in observed sea ice data may be possible.  Although this early-warning signal may be a large step forward in our ability to say whether the probability of a abrupt climate change event is increasing, improving our understanding of the underlying causes of this behavior is crucial if we are to improve sea ice forecasts, particularly near rapid transitions.