Factors affecting dynamical seasonal prediction of the Arctic sea ice

Arctic sea ice variability has received increasing attention during the last decade. Seasonal prediction of the Arctic sea ice has been primarily produced with statistical methods during the past years. A few operational centers have recently implemented dynamical sea ice component in the coupled atmosphere-ocean forecast systems for seasonal climate prediction. Yet various issues remain to be resolved for an improved prediction of seasonal sea ice variations. In this study, we analyze the forecast of sea ice extent in the NCEP Climate Forecast System version 2 (CFSv2) and address factors that affect the representation of the observed sea ice variability in the forecast model. The analysis will be based on retrospective and real-time 9-month forecasts from the CFSv2 for 1982-2012. We will first assess the overall performance of the CFSv2 in capturing the observed sea ice extent climatology, long-term trend, and interannual anomalies. We will then discuss factors that affect the sea ice prediction, including: (1) consistency of the initialization of the observed sea ice concentration, (2) impacts of surface heat fluxes related to atmospheric model physics, (3) bias in sea surface temperatures, and (4) impacts of initial sea ice thickness.