Community Research and Development Information Service - CORDIS


In order to investigate whether correlations exist between cloud cover/cloud types over an up-welling area off Northwest Africa and marine biological activity induced by nutrient rich waters brought to the surface by this process, it was decided to study the suitability of using neural networks (NN) as a tool for cloud classification.
A Kohonen's self organizing map (SOM) was implemented and employed for several experiments. The network was made up of a variable number of input channels and 28 output clusters arranged in a mono-dimensional array. In the experiments, an image acquired by advanced very high resolution radiometer (AVHRR) over an area of the Atlantic Ocean (0-40N, 5-35W) was employed; the image was acquired from the Joint Research Centre database. The results reported show a first unsupervised classification of the scene and a derived 'hand' classification based on the unsupervised clustering, brightness, temperatures and albedos.

Additional information

Authors: CHIUDERI A, JRC Ispra (IT);MURRAY N, JRC Ispra (IT)
Bibliographic Reference: Paper presented: Data User's Conference, Winchester (GB), September 4-8, 1996
Availability: Available from (1) as Paper EN 39841ORA
Record Number: 199610738 / Last updated on: 1996-08-06
Original language: en
Available languages: en