Multi-Aspect Assessment of CMIP6 Models for Arctic Sea Ice Simulation

This paper evaluates the ability of 35 models from phase 6 of the Coupled Model Intercomparison Project (CMIP6) to simulate Arctic sea ice by comparing simulated results with observation from the aspects of spatial patterns and temporal variation. The simulation ability of each model is also quantified by the Taylor score and e score from these two aspects. Results show that biases between observed and simulated Arctic sea ice concentration (SIC) are mainly located in the East Greenland, Barents, and Bering Seas and the Sea of Okhotsk. The largest difference between the observed and simulated SIC spatial patterns occurs in September. Since the beginning of the twenty-first century, the ability of most models to simulate summer SIC spatial patterns has decreased. We also find that models with the Sea Ice Simulator (SIS) sea ice component in CMIP6 show a consistent larger positive simulation biases of SIC in the East Greenland and Barents Seas. In addition, for most models, the higher the model resolution is, the better the match between the simulated and observed spatial patterns of winter Arctic SIC is. Furthermore, this paper makes a detailed assessment for temporal variation of Arctic sea ice extent (SIE) with regard to climatological average, seasonal SIE, multiyear linear trend, and detrended standard deviation of SIE. The sensitivity of September Arctic SIE to a given change of Arctic surface air temperature over 1979-2014 in each model has also been investigated. Most models simulate a smaller loss of September Arctic SIE per degree of warming than is observed (1.37 x 10(6)km(2)K(-1)).

Automated summary

This paper evaluates the performance of 35 models from CMIP6 in simulating Arctic sea ice, highlighting a decrease in the ability of most models to reproduce summer SIC spatial patterns and noting a consistent larger positive simulation bias in certain regions for models with the SIS sea ice component. Additionally, higher model resolution generally results in better matching of winter Arctic SIC spatial patterns.