A surprising, but not unexpected, multi-decadal hiatus in Arctic sea ice decline

Since the beginning of the satellite era in 1979, September sea ice area over the Arctic has nearly halved, and the rapid sea ice decline in the early 2000s fueled concerns that the first ice-free Arctic summer could occur before 2020.  In light of these dire forecasts, therefore, it is remarkable that no statistically significant decline in September Arctic sea ice coverage has occurred over past two decades, as we demonstrate here.

This 20-year-long hiatus in pan-Arctic sea ice decline is robust across observational datasets, and also robust to the choice metric (sea ice area or extent).  In fact, the present hiatus is seen in all months of the year, not only in September.

One is immediately led to ask whether climate models are able to capture multi-decadal-long periods with no Arctic sea ice decline in the presence of strong anthropogenic radiative forcing.  To answer this, we analyze multiple large ensembles of CMIP5 and CMIP6 simulations over the first half of the 21st century.  We find that 20-year-long periods with no significant Arctic in sea ice loss occur rather frequently in climate models, even under high emissions scenarios.  Roughly 20-30% of the hundreds of model runs analyzed here show 2005-2024 trends smaller than the small observed trend, although the intermodel spread is considerable.  Sub-selecting models based on their ability of simulating observed climatological sea ice, or on their equilibrium climate sensitivity, or on the specific scenario forcings, does not alter the key result.

Our analysis also reveals that the simulated forced response, in most models, is considerably larger than the observed trends.  This highlights the important role of internal variability in enhancing and, more recently, in reducing Arctic sea ice decline.  In fact, our analysis of model output reveals that a different configuration of internal climate variability could have caused an overall growth in Arctic sea ice over the past 20 years, surprisingly enough.

Looking at the coming decades, and focusing on those climate simulation which are consistent with the present observed trends, we find the hiatus is 50% likely to persist for another five years, and could plausibly last another full decade (with roughly 30% chance).

Taken together, our findings lead us to conclude that the current hiatus in Arctic sea ice decline, even as emissions of greenhouse gases continue to rise, is not an unexpected event that questions our basic understanding of the Arctic climate system.  Climate models show that it can occur rather frequently, and that it could plausibly persist for years to come.