CAN HEAT FLUX DATA FROM OCEAN REANALYSES PREDICT SEASONAL ARCTIC SEA ICE RETREAT?

Before the 21st century, the Arctic had been a region of the world that was mostly inaccessible to resource extraction and transportation. Sea ice decline has made the Arctic more accessible for both resource extraction and shipping routes, so, being able to accurately predict future seasonal Arctic sea ice extent is more important than ever before. Data from Bering Strait moorings have established a clear negative correlation between heat flux and sea ice where, as heat flux increases, sea ice extent decreases. This study evaluates the heat flux measurements in the Simple Ocean Data reAnalysis (SODA 3.3.1) and the Ocean ReAnalysis System 4 (ORAS4), comparing them to in situ mooring data and assessing their ability to predict sea ice retreat in the Chukchi Sea.

The SODA reanalysis was constructed at the University of Maryland. It relies on the Modular Ocean Model v5 ocean with MERRA-2 atmospheric forcing. ORAS4 was produced by the European Center for Medium-Range Weather Forecasts and relies on the NEMO V3.0 ocean model with ERA atmospheric forcing. Analysis of both reanalyses consisted of subsetting water temperature, velocity, salinity, and depth along a transect from ~189.25°E to ~193.25°E at ~66.25°N. By calculating the flux of heat through this cross-section of the Bering Strait, data from the reanalyses could be compared to heat flux data from moorings in the strait. First retreat day (FRD) for sea ice in the Chukchi Sea (i.e., the first day that sea ice concentration in the Chukchi Sea dropped below 30%) was sourced from combined passive microwave record.

While SODA did significantly correlate to mooring data for most months, the correlation between ORAS4 and the mooring data was more consistently significant. SODA only produced data with a significant correlation (p ≤ 0.05) to the FRD data when the reanalysis was assimilating mooring data (e.g 1991, 1992, 1998, 2000-2013). Therefore, SODA is insufficient for predicting sea ice retreat; it relies too heavily on data assimilation to correct the model. By contrast, ORAS4 had significant correlations with sea ice retreat even when its model output was not being nudged by mooring data. This suggests that ORAS4 and its underlying model have greater potential for predicting sea ice retreat in the Chukchi Sea.