On improving the simulation of Atlantic Water circulation in the Arctic Ocean

[1] We performed a set of numerical experiments in order to better simulate the circulation of Atlantic Water in the Arctic Ocean by employing a coupled ice-ocean Arctic regional model. We found that the inflow of Atlantic Water via Fram Strait is weak in the case of high viscosity and diffusivity coefficients ( 1 . 10(8) cm(2)/ s and 1 . 10(7) cm(2)/s) and creates an anticyclonic circulation in the Eurasian basin. This flow increases significantly when both coefficients are scaled down ( to 1 . 10(7) cm(2)/ s and 0.5 . 10(6) cm(2)/s) but the current path is then unstable, and it becomes mostly anticyclonic after about a decade of integration. A further reduction of these coefficients leads to growing instabilities due to insufficient grid resolution. Short-term integration of the model with doubled space resolution shows better agreement with observations, but long-term calculation is still restricted by simulation time. Alternatively, a lower resolution version of this model benefits from parameterizations of subscale processes. Incorporation of the Neptune parameterization of eddy-topography interaction into the model intensifies the current steered by subsurface topography and results in cyclonic circulation of the Atlantic Water in every subbasin and in the whole Arctic region. In the absence of salinity restoring there was strong salinity drift at the surface, resulting in a salinity increase in the central Arctic by 3 units from 1960 to 2005. The Arctic freshwater content is driven basically by year-to-year variations of convective mixing involving more saline water from lower layers. After cooling and increasing in salinity through brine rejection, the surface water becomes denser, leading to even stronger convective mixing. One-dimensional model tests showed how model performance could be improved by implementing a lower vertical diffusion coefficient and higher vertical resolution.