Abstract

A method is presented for the automatic construction of a cloud-type and cloud-amount nephanalysis. The uses to which such a nephanalysis could be put are first considered, and then a combination of brightness and texture measurements is used to characterize cloud types on DMSP satellite imagery. Texture is evaluated by measuring the standard deviation and vector dispersion of the density values within small window areas of the whole satellite picture, and then by a discriminant analysis scheme window areas are allocated to one of three cloud categories and a no-cloud category. The automatic nephanalysis is constructed by analyzing all the adjacent window areas on the image. The results of the analysis of visible and infrared images show accuracies of greater than 72% for each automatic nephanalysis. Finally, the use of multispectral cloud image analysis is considered in contrast to the unispectral brightness and texture approach.

This content is only available as a PDF.