Predictability of Sea Ice Near Bifurcations

Abstract

There is evidence in Earth’s history of relatively stable climate regimes abruptly transitioning to alternative states. It has been argued that the greatest potential for such abrupt transitions in Earth’s system in the near future is located in the Arctic. Here we analyze the Arctic sea ice evolution of two current generation climate models that exhibit critical transitions. We demonstrate the detectability of two early warning signals: increased variance and increased autocorrelation. We introduce another metric that forewarns of abrupt changes in sea ice; a decrease in predictability before the threshold points. Observations of Arctic sea ice extent are searched for early warning signals using methods identical to the model analysis. A regional analysis is also performed for both models and observations and demonstrates that these metrics are detectable at the regional level. We show that the complexities of the Arctic sea ice system can be reduced to a simple stochastic sea ice model and determine that our warning metrics are applicable in that setting as well. Because there is often little to no change to the state of the system before an abrupt change, the robustness of these metrics make them promising indicators of the risk of upcoming regime shifts in the Arctic sea ice system. But given the decrease in predictability of sea ice near a bifurcation, an improved understanding of the physical mechanisms forcing abrupt climate change is needed if we are to improve sea ice forecasts moving forward. Understanding the reason behind the vastly different outcomes produced by the climate models may help us with this task.