Bestimmung verschiedener Eisklassen durch statistische Analyse der Rauigkeit von Meereis

Abstract

Among the properties of sea ice, roughness is an important parameter. It affects the interactions between ice, atmosphere and ocean. The morphological properties of the top and underside surface influence the transfer of energy and momentum. In satellite remote sensing, the knowledge of surface roughness characteristics is important, because they influence the measured signal in a complex way.Based on in-depth statistical analyses of sea-ice roughness, two classification methods are investigated in this work regarding their potential to separate different ice types. These methods are discriminant and cluster analysis.In order to take different aspects of roughness into account, datasets from four different geographical locations are used. These comprise data from the Lincoln Sea north of Greenland, the Arctic Ocean near Svalbard, and the Baltic Sea. One dataset from the Arctic Ocean was obtained during summer. The available data thus enable investigations of regional as well as seasonal changes of sea-ice roughness.The statistical analyses reveal regional differences in sea-ice roughness. Surface roughness profiles are found to be nonstationary and to display fractal properties on length scales below 20~m. The distributions of height and spacing of pressure ridges are approximately exponential or lognormal, respectively. Pressure ridges are not distributed randomly over the ice surface but appear in clusters. Significant correlations exist between profiles of the sea-ice surface and draft. Spatial scales that contribute most to the surface roughness are found to be smaller than 50~m. The surface roughness is thus largely influenced by length scales comparable to observed pressure ridge widths. The statistical analyses lead to a set of parameters, consisting of mean height, RMS height, skewness, kurtosis, fractal dimension and RMS slope, which characterize the roughness and form the basis for the classification analysis. The discriminant analysis shows that the thickest ice classes can be distinguished from one another and from thinner ice using the surface roughness parameters to separate the classes. The cluster analysis reveals that different types of surface roughness cannot be distinguished clearly from one another. Synthetic roughness profiles are important for studies of the interactions between the sea-ice surface and the atmosphere. In this work a numerical model for simulations of sea-ice draft is assessed regarding its potential to generate realistic sea-ice surface profiles. It is shown that the model is capable of reproducing many of the properties of real surface roughness profiles.