An Evaluation of the CMIP6 Historical Simulations of the Arctic Sea Ice

Abstract:

Output from a subset of models participating in the Phase 6 of Coupled Model Intercomparison Project (CMIP6) has been used to investigate the quality of their simulation of the Arctic sea ice (SI). We have used a variety of metrics in comparison with available observations and results from the Regional Arctic System Model (RASM) for the period 1979-2014. Those include the timing and rate of simulated SI decline along with multi-decadal trends, integrated ice-edge error (IIEE), and qualitative comparison of the historical simulations. In this presentation, we focus on the IIEE analyses, since those are not commonly used for the evaluation of climate model historical simulations of the Arctic SI. In addition, this metric offers a pathway to diagnosing the causes of common and specific climate model biases down to the process / parameterization level. The IIEE regional analyses expose common and individual regional model biases, which the basic SI extent analysis alone does not capture, therefore suggesting limitations in or lack of representation of some physical processes within CMIP6 models. Given the high incidence of model biases revealed by those analyses related to the marginal and shelf seas, we hypothesize that many model limitations might be related to the oceanic forcing of SI in those regions. For example, models that consistently underestimate the position of the SI edge in the Barents and Kara seas could be over/under-estimating the transport of warm Atlantic Water across or cooling over the Barents Sea. The resulting excess/shortage of oceanic heat transport could be accelerating/slowing down ice melt and delaying/accelerating freeze-up over the Barents Sea and further downstream, over the northern Kara Sea and beyond. SI biases in other regions of the Arctic Ocean and during other seasons diagnosed with this metric are also discussed.