Assessment of Cirrus Cloud Optical and Microphysical Data Reliability by Applying Statistical Procedures

Kurzfassung

A methodology of employing statistical procedures, specifically the principal component analysis (PCA) technique, to assess cirrus cloud data reliability is described. The approach is demonstrated by an example of a study of optical and microphysical characteristics measured during two campaigns performed at midlatitudes in the pristine Southern (SH) and polluted Northern (NH) Hemispheres within the international INCA project (Interhemispheric Differences in Cirrus Cloud Properties from Anthropogenic Emissions). The datasets were obtained by using state-of-the-art airborne instruments including the polar nephelometer and PMS particle size spectrometers for the ice-particle characterization. The approach is applied to both the measured angular scattering intensities and the ice-particle size distributions. It is shown that the PCA technique allows for impartial elimination of nonreliable channels of instruments. Furthermore, this technique is efficient in a study if the dataset is statistically homogeneous, and provides the possibility of removing specific records corresponding to distinguishing statistical ensembles. The results, expressed in terms of significant components and corresponding eigenvalues, show that the Southern and Northern Hemisphere datasets are in good agreement and they can be considered as statistically representative of the sampled cirrus. Furthermore, the frequency distributions of the cirrus cloud microphysical and optical properties can be regarded as arbitrary positive quantities, which are lognormally distributed. The validation of the measurements is provided by intercomparison of parameters estimated from different and independent techniques. The statistical relationships between quantities derived from angular scattering intensities and from ice-particle distributions as well as the similarity of the results obtained for the Southern and Northern Hemisphere cases serve as proof of the reliability of the measured cloud properties.