words, experiment B induces horizontal variability of density flux. Therefore, convection develops more rapidly in experiment B than experiment D. And, the peak of histogram is mode heavy in experiment B than experiment D.
We also carry out some experiments, in which forcing area is moving. As the moving speed increases, less heavy water is formed. The tendency is easily understood, because of an increase in a forcing area, or decrease in density flux per unit area.
5 DISCUSSION
In the numerical experiments using a three-dimensional non-hydrostatic model, in spite of the total density flux over the domain is unchanged, the density of the water made by convection may depend on the size, distribution or moving speed of an area of density forcing.
These results suggest that the dense water produced by brine rejection varies depending on the sizes or drifting speed of lead and polynia. This information is extremely useful for parameterizing dense water formation under the ice cover in a numerical model with a large (a few tens of km or larger) grid size inevitably using a hydrostatic approximation. In order to simulate the ice-covered ocean, we should include a size distribution of lead and polynia in addition to grid-averaged ice concentration. For the parameterization, we have to make clear mechanism of plume scale convection.
Acknowledgments. The work was supported by Japanese Ministry of Education, Science, Sports and Culture with great appreciations.
