Seasonal Prediction of Arctic Sea Ice by a Linear Markov Model

Abstract:

To assess the predictability of Arctic sea ice, a linear Markov model has been developed to forecast ice concentration in the pan Arctic region. The model was built to capture co-variability in the atmosphere-ocean-sea ice system that was defined by sea ice concentration, sea surface air and ocean temperature, geopotential height and winds at 300mb. Multivariate empirical orthogonal functions of these variables were used as building blocks. A series of model experiments were carried out to determine the dimension of the model. The predictive skill of the model was evaluated in a cross-validated fashion. The model showed considerable skill within the Arctic Basin during summer and fall. Particularly in the region north of Chukchi Sea, skills for fall predictions are above 0.6 even at 9-month lead. Because the Arctic Basin is completely frozen in winter and spring, the predictability appears in the marginal seas during these seasons. The model had higher skills in the Atlantic sector of the Arctic than in the Pacific sector. The experimental 4-month lead forecast for 2013 September pan Arctic ice extent yielded 5.16 million square kilometers, which slightly under-estimated the observation of 5.40 million square kilometer.