The Labrador Sea is a climatically important region for the entire globe. As one of the main locations of deep water formation, this subpolar sea helps modulate the strength of the Atlantic Meridional Overturning Circulation. The deep water formation rate is influenced by freshwater originating from the Arctic Ocean as well as from the Greenland Ice Sheet. The freshwater entering the sea is then distributed across the Labrador Sea by boundary currents and eddy circulation. In this study we analyze a suite of simulations using a high resolution earth system model, the Regional Arctic System Model (RASM), to explore the freshwater forcing methods needed to properly represent the Labrador Sea dynamics.

The first simulation, the H-case, uses an actively coupled ocean and sea ice models only, with sea surface salinity (SSS) restored to mean climatology. The G-case removed the SSS restoring method and instead incorporated land runoff fluxes from the Coordinated Ocean-Ice Reference Experiments version 2 (COREv2) to evolve SSS. The final simulation, the R-case, is a fully coupled run that relies minimally on reanalysis forcings, but instead allows the coupled model physics to drive the simulation, with air-sea-land fluxes that develop and grow without artificial forcing.

The critical oceanographic conditions that permit deep water formation can be assessed in our model results by analyzing the strength of water column stratification within the Labrador basin. We explore three methods for calculating the mixed layer depth (MLD) methods. The MLD serves as a measure of the vertical extent of conditions amenable to deep convection and as an indicator of the deep water formation rate. We find that all three methods create a MLD signal that overpredicts, as compared to the existing observations, the spatial and temporal extent of water stratification conducive to deep convection in the Labrador Sea. The maximum density gradient (MDG) method produces on average the lowest MLD values, and hence may best represent the intermittent nature of the deep convection process in the region. Out of the three numerical experiments considered in this study, the R-case shows the most realistic depictions of the extent and depth of MLDs when compared to the G and H-cases. However, even in the R-case, the results of this study highlight the need for further improvements of the ability of the model to produce realistic levels of vertical stratification across this critical high latitude ocean basin.

Beyond the MLD analysis, we find that the R-case outperforms the G and H-cases in a number of surface ocean dynamics measures compared to observations. First, it displays a stronger West Greenland Current System (WGCS) from which Irminger Current Anticyclones are generated and buoyancy is distributed across the rest of the basin. The eddy activity in the R-case is not only more frequent, but the eddies are also more effective at supplying freshwater to the interior Labrador Sea compared to the other two simulations. We find that the salinity gradient within the WGCS and the sea surface height (SSH) gradient across the west Greenland shelf are the strongest in the R-case. These characteristics stem from freshwater entering the region through boundary currents and from local sources. The R-case also displays a realistic seasonal variation of surface hydrography, and it more closely follows the evolution of sea ice on a mean annual basis compared to the G and H-cases. This shows that the coupled modeling framework provides great improvement to the surface dynamics of the Labrador Sea.