Application of Fuzzy K-mean Method to the CALIOP/CALIPSO Layer Feature Classifications

Thursday, 17 December 2015
Poster Hall (Moscone South)
SHAN Zeng1, Mark A. Vaughan1, Charles R Trepte1, David M Winker1, Patricia Lucker2, Zhaoyan Liu1, Ali H Omar1, Jayanta Kar3, Sharon P Burton1, Yongxiang Hu1 and Melody A Avery1, (1)NASA Langley Research Center, Hampton, VA, United States, (2)Science Systems and Applications, Inc., Hampton, VA, United States, (3)Science Systems and Applications, Inc. Hampton, Hampton, VA, United States
Fuzzy clustering model is an essential tool for remotely sensed image classification. In our study, we applied fuzzy k-mean approach to the CALIOP/CALIPSO data and compared fuzzy clustering results with evaluated version 4 classification products. Sensitivity of data sampling and fuzzy linear discriminant analysis are discussed. Although fuzzy K-mean may not improve current retrieval algorithm, it could help us to better understand which parameters are key for atmospheric feature discriminations and which drive the ambiguous classifications. Fuzzy clustering could be further used for the classification of atmospheric features with combined A-Train observations in the future and could be a guide for the design of better new instruments.