Submesoscale Atmospheric Boundary Layer Processes over Fragmented Sea Ice

In mesoscale numerical weather prediction (NWP) models, sea ice cover is typically represented by grid-cell-average ice concentration and thickness. The relevant variables – surface heat and moisture flux, roughness, albedo and so on – are calculated as a weighted average of the respective values over sea ice and open water. With typical model resolutions of a few kilometers, all smaller-scale variability related to nonuniform spatial distribution of sea ice within model grid cells cannot be taken into account. As the larger-scale effects of these submesoscale processes are largely unknown, no parameterizations suitable for NWP models are available. Only very recently, significance of the size distribution and spatial arrangement of sea ice floes on the sea surface has attracted attention. There is growing observational and theoretical evidence that floe-level processes have significant influence on the dynamics and thermodynamics of the lower atmosphere and upper ocean, as well as the ice cover itself.  In a very recent numerical study of Horvat et al. (Geophys. Res. Lett., accepted for publication), the authors demonstrated that purely thermodynamically generated effects related to heat flux gradients at floes’ boundaries are responsible for formation of eddies and, through a number of feedbacks, for faster ice melting, with melting rates dependent on the size of ice floes within the model area. These  results, based on idealized model simulations, clearly demonstrate our limited knowledge of atmosphere–sea ice–ocean  interactions and boundary layer processes over fragmented, strongly nonuniform sea ice.

The goal of this work is to analyze three-dimensional air circulation within the atmospheric boundary layer over fragmented sea ice, and to obtain a better understanding of area-averaged effects of processes taking place at the level of individual ice floes.

To this end, we perform a series of high-resolution numerical simulations with the Weather Research and Forecasting (WRF) model. The model domain is rectangular, and periodic boundaries are used in both horizontal directions. A simple one-dimensional  ocean mixed layer model is used, initialized with a mixed layer depth and a temperature lapse rate below the mixed layer typical of Arctic conditions (coupling with a 3D upper ocean model is planned for the future). The atmospheric model is initialized with air temperature and moisture profiles representative for the Arctic Ocean. In a reference model run, constant ice concentration over the whole model domain is prescribed, which ensures horizontal homogeneity of conditions. The results of the reference run are then used as initial conditions for a series of simulations with the same total sea ice area and volume, but different spatial distribution of ice floes, including: (i) a single, elongated lead in a compact ice cover, (ii) randomly distributed floes of equal sizes, (iii) clustered floes of equal sizes, (iv) randomly distributed floes with a power-law size distribution, and (v) clustered floes with a power-law size distribution. Additionally, the simulations are repeated with a number of different wind speeds, as well as without wind, in order to analyze the relative role of advective and convective processes in the ocean–atmosphere  heat and moisture exchange.

We analyze in detail the three-dimensional structure of the atmospheric circulation, heat and moisture transport, surface heat flux, as well as domain-averaged properties of the atmospheric boundary layer in different model runs. We demonstrate that – especially at medium ice concentrations – the domain-averaged values are sensitive to the subgrid-scale spatial distribution of sea ice. This suggests that parameterizing these effects should lead to the improvement of the performance of NWP models in regions with medium ice concentration.