Long-range forecasting of Arctic sea ice

Abstract

The operational production of skillful long-range forecasts of Arctic sea ice has the potential to be very useful when integrated into the planning of Arctic operations by the U.S. Navy and other organizations. We investigated the potential for predicting October sea ice concentration (SIC) in the Beaufort Sea at lead times of one to five months. We used SIC data for 1979-2007 to statistically and dynamically analyze atmospheric and oceanic processes associated with variations of SIC in the Beaufort Sea. We also conducted correlation analyses to identify climate system variables for use as predictors of SIC. We developed linear regression models for predicting SIC based on multiple predictors. We tested these models by generating hindcasts of October SIC for 1979-2007 based on several combinations of predictors. We found two key predictors of October SIC in the Beaufort Sea at leads of one to five months--antecedent SIC in the Beaufort Sea and sea surface temperature (SST) in the Caribbean Sea in the preceding May-September period. Both of these predictors showed a consistent and statistically significant relationship with October SIC at all lead times. Both are also dynamically reasonable predictors, given the role of antecedent ice conditions, and of the Arctic Oscillation and North Atlantic Oscillation in influencing basin scale SSTs. Our hindcast verification metrics show that a linear regression model based on these two predictors produces skillful forecasts of SIC at leads of one to five months. Based on these results, we issued a forecast on 01 June 2010 for SIC in the Beaufort Sea in October 2010. We also identified and conducted mult-year, linear regression hindcasts using several other predictors (e.g., low level air temperature, low level winds, and upper ocean temperature) that proved useful at various lead times. Our results indicate a significant potential for improving long range forecasts in support of Arctic operations by the U.S. Navy and other organizations.