A Comparison of Fully-Coupled and Reanalysis-Forced Simulations of the Effect of Arctic Cyclones on Short-Term Sea-Ice Evolution

We showed in a recent study that in the Navy-ESPC coupled (ice-ocean-atmosphere) model, sea-ice melting was locally enhanced in association with a strong cyclone in August 2012. Bottom melting was enhanced by storm-induced turbulent vertical mixing of a sub-surface oceanic warm layer, and surface melting was enhanced by strong surface winds advecting warm air over the ice from Alaska and open waters. The increase in melting resulting from turbulent mixing of the ocean supports a previous study (Zhang et al. 2013) that used a different ice-ocean model that was forced by a reanalysis atmosphere, but there were some discrepancies as well. Most notably, in our study, there was no apparent basin-wide signature of the cyclone evident in the evolution of the sea ice, whereas in the previous study, there was a clear peak in the Arctic-mean melting resulting from the cyclone.

Here, we examine additional simulations of the August 2012 cyclone, where the ice-ocean components of the Navy-ESPC model are forced by either NCEP/NCAR or ERA5 atmospheric reanalysis, and we compare these to the fully-coupled simulations. The overall rate of melting is much greater in the reanalysis-forced simulations, and this results in a substantially worse “forecast” of sea-ice extent than in the coupled model, despite having largely removed the effect of forecast errors in the atmosphere. This is in contrast to the successful simulation of sea-ice extent in the forced simulation of Zhang et al. (2013). We investigate the possible reasons for these differences, including biases in reanalysis forcing (for which the forced model of Zhang et al. is tuned, but for which Navy-ESPC is not), biases in the Navy-ESPC model, and differences in the initial distribution of sea ice and ocean temperature. We also examine additional cases of strong cyclones in other years and other months within the melt season, in order to assess the robustness of the simulated sea-ice response, and the degree to which the response is seasonally-dependent.