A neuro-fuzzy approach for monitoring global snow and ice extent with the SSM/I

A neuro-fuzzy approach for classification of snow cover and sea ice from the Special Sensor Microwave/Imager (SSM/I) satellite data is presented. The fuzzy c-means algorithm was used as a supervised clustering method for mapping the SSM/I data of known classes into fuzzy c-partitions from which to determine the fuzzy c-means for each class. A single-hidden-layer neural network was implemented to learn the fuzzy c-means and desired responses in terms of fuzzy memberships. After training, the neural network was applied as a fuzzy membership function to fuzzify any inputs of unknown SSM/I data and defuzzify the outputs into crisp classes for mapping snow and ice extent. Weekly snow and sea ice extent was tested and compared with the one derived from the National Snow and Ice Data Center (NSIDC)