Sensitivity of a regional atmospheric model to a sea state-dependent roughness and the need for ensemble calculations

Abstract

The sensitivity of an atmospheric high-resolution limited area model to a sea state-dependent roughness is examined. Two sets of Monte Carlo experiments are compared. In the first set the sea state was explicitly accounted for in the computation of the sea surface roughness. In the second set the roughness was parameterized by the standard Charnock relation. On climatic timescales of months and longer, the differences between the two sets are small. On the daily timescale large deviations between individual realizations of the two ensembles in the order of several hectopascals are occasionally found suggesting a considerable impact of the sea state-dependent roughness on the atmospheric circulation. It is shown, however, that the comparison of individual realizations, a frequently used approach in regional sensitivity studies, can be misleading. It is found here that the largest differences between the two ensembles occurred simultaneously with high inherent model variability. In these situations an eventually existing impact of the sea state-dependent roughness on the atmospheric circulation could therefore not be discriminated from the background variability and the null hypothesis that both ensembles stem from the same population could not be rejected at given risk. At times at which the internal model variability was small a statistically significant impact of the sea state-dependent roughness on the atmospheric circulation was found. However, the impact was small and it is concluded that compared with the sea state-dependent parameterization used in this study the Charnock relation represents a reasonable parameterization in regional atmospheric climate models.