Modeling surface interactions in the Arctic Ocean: accuracy vs. computational cost

This paper compares various methods that are used in numerical models to determine heat and momentum fluxes over a sea-ice surface and the adjacent temperature profiles in the snow/ice and lower atmosphere. During the Surface Heat Budget of the Arctic (SHEBA) experiment, the authors and others obtained a near-continuous record of detailed snow/ice, surface and atmospheric surface layer profile measurements of temperature, humidity, wind speed, radiation, heat fluxes and momentum fluxes. These results allowed us to examine and test various parameterizations of surface fluxes. In particular, the SHEBA data allow us to address the problems of estimating fluxes in very stable conditions when standard Monin-Obukhov surface layer theory does not apply; these results are applicable to stable surface layers everywhere on Earth.

For some processes, we found that using very simple assumptions (such as assuming that near surface humidity is usually very close to the ice saturation value) can produce results that are as good or more accurate than methods in common used today. But the SHEBA results also highlighted the complexity caused by factors such as extreme stability, horizontal variations in ice and cloud characteristics, open water areas and other factors. The authors have been working to more fully understand these factors and ways to model them in the most accurate way possible.

However, many of the more sophisticated, and presumably more accurate, parameterizations are computationally too expensive to include in global climate models or even regional models. Therefore there is a need to develop atmospheric surface layer and ice sub-models that are computationally efficient. Using the SHEBA data for verification, the authors test various parameterization schemes for accuracy. We examine simplifications such as using a geostrophic drag coefficient for momentum, radiation parameterizations based on surface temperature and cloud cover, and non-iterative turbulent surface flux schemes. We also examine, in a one-dimensional context, the effect of using various vertical grid point spacings in the lower atmosphere and snow/ice medium, even simpler assumptions that treat the upper snow/ice and atmospheric boundary layer as one layer, and other simplifications that are under development The result will be a cost-benefit analysis that model developers can use to compare the computational cost vs. the estimated accuracy of a particular algorithm or sub-model.