A data-model intercomparison study of Arctic sea-ice variability

Abstract

The dynamic-thermodynamic granular rheology sea-ice model of Tremblay and Mysak is validated against 40 years of observed sea-ice concentration (SIC) data. Subsequently, the mechanisms responsible for producing SIC anomalies in the model are evaluated by studying the coupled variance (using the singular value decomposition method, SVD) between the simulated SIC anomalies and the ice speed and air temperature anomalies. To execute this validation, a 49-year (1949–97) simulation (including a 9-year spin-up period) of the Arctic and peripheral sea-ice cover using daily varying winds and monthly mean air temperatures is produced. In general, the simulated SIC variations for 1958–97 in the East Siberian, Chukchi and Beaufort seas are in agreement with observations, while larger discrepancies occur in the Laptev and Kara seas. Moreover, the sensitivity of the model to southerly wind anomalies in creating summer SIC anomalies compares well with the observed sensitivity; however, the model's sensitivity to summer air temperature anomalies is weaker than observed. The summer SIC anomalies over an entire sea are not influenced by variations in the level of river runoff. Results from the SVD analysis show that the main source of variability in the peripheral seas is associated with the variation in the strength of the Arctic High; in the East Siberian and Laptev seas, the strengthening and weakening of the Transpolar Drift Stream also play an important role. Over the entire Arctic domain, surface air temperature anomalies are negatively correlated with sea-ice anomalies. Finally, the observed downward trend in total sea-ice cover in the last two decades as well as record minima in the East Siberian Sea are well reproduced in the simulation.