Improving Sea Ice Prediction in the NCEP Climate Forecast System Model

Abstract:

Skillful prediction of Arctic sea ice is important for the wide variety of interests focused in that region. However, the current operational system used by the NOAA Climate Prediction Center does not adequately predict the seasonal climatology of sea ice extent and maintains too high sea ice coverage across the Arctic. It is thought that the primary reasoning for this lies in the initialization of sea ice thickness. Experiments are carried out using the Climate Forecast System (CFSv2) model with an improved sea ice thickness initialization from the Pan-Arctic Ice Ocean Analysis and Assimilation System (PIOMAS) rather than the default Climate Forecast System Reanalysis (CFSR) sea ice thickness data. All other variables are initialized from CFSR. In addition, physics parameterizations are adjusted to better simulate real world conditions. Here we focus on hindcasts initialized from 2005-2014. Although the seasonal cycle of sea ice is generally better captured in runs that use PIOMAS sea ice thickness initialization, local sea ice freeze in early winter in the Bering Strait and Chukchi Sea is delayed when both sea ice thickness configurations are used. In addition ice freeze in the North Atlantic is more pronounced than in the observations. This shows that simply changing initial sea ice thickness is not enough to improve forecasts for all locations. Modeled atmospheric and oceanic parameters are investigated including the radiation budget, land surface temperature advection, and sub-surface oceanic heat flow to diagnose possible reasons for the modeling deficiencies, and further modifications to the model will be discussed.