Application of Fuzzy Grade-of-Membership Clustering to Analysis of Remote Sensing Data

Abstract

A fuzzy grade-of-membership (GoM) clustering algorithm is applied to analysis of remote sensing data, in particular, the type of data used in climatic classification. The methodology is applied to a cloud product data subset derived from NASA’s International Satellite Cloud Climatology Project, which includes remotely sensed global monthly average surface temperature and precipitation data for land and coastal regions for the year 1984. GoM partitions for this case are similar to those of vector quantization and fuzzy c-means clustering algorithms, which is significant given the striking differences between the algorithms. The GoM clustering approach is shown to provide an alternative means of interpreting large heterogeneous datasets for exploratory analysis, which broadens the application base by admitting categorical data.