Abstract:

Two types of stochastic models have been developed to predict Arctic sea ice concentration. Linear Markov models are used to predict sea ice concentration year-round at the seasonal time scale, while an auto-regressive (VAR) model is applied to predict summer daily sea ice centration and spring melting date at intra-seasonal time scale. The linear Markov model was built to capture co-variabilities in the atmosphere-ocean-sea ice system defined by sea ice concentration, sea surface and air temperature, geopotential height and winds at the 300mb level. Multivariate empirical orthogonal functions of these variables served as building blocks of the model. The predictive skill of the model was evaluated in a cross-validated fashion. The model shows good skill in predicting the signs of sea ice anomalies within the Arctic Basin during summer and fall. Particularly in the region north of the Chukchi, Beaufort, Eastern Siberian, Kara, and Barents Seas, the skill based on anomaly correlation is above 0.6 even at a nine-month lead. Because the Arctic Basin is completely frozen in winter and spring, the predictability can be seen only in the seasonal ice zone during these seasons. The model has higher skill in the Atlantic sector of the Arctic than in the Pacific sector. The model predicts well the interannual variability of September sea ice extent (SIE) but underestimates the accelerated long-term decline of SIE, resulting in a systematic model bias. This model bias can be corrected by a constant bias correction or linear regression bias correction, leading to an improved correlation skill of 0.93 for the two-month lead SIE prediction. 

The (VAR) model is evaluated for predicting the summer time (May through September) Arctic sea ice concentration, using only the daily sea ice data and without direct information of the atmosphere and ocean. The cross-validated forecast skill of the VAR model is found to be superior to both the anomaly persistence and damped anomaly persistence at lead times of 20~60 days, especially over Northern Eurasian marginal seas and the Beaufort Sea. The daily forecast of ice concentration also leads to predictions of ice-free dates and September mean sea ice extent. While the detailed mechanism leading to the high predictability of intra-seasonal sea ice concentration needs to be further examined, the study reveals for the first time that Arctic sea ice can be predicted statistically with reasonable skills at the intra-seasonal time scales given the small signal-to-noise ratio of daily data.