Applications of neural network methods to the processing of Earth observation satellite data

The new generation of earth observation satellites carries advance sensors that gather very precise data for studying the Earth system and global climate. This paper shows that neural network methods can be successfully used for solving forward and inverse remote sensing problems, providing both accurate and fast solutions. Two examples of multi-neural network systems for the determination of cloud properties and for the retrieval of total columns of ozone using satellite data are presented. The developed algorithms based on multi-neural network are currently being used for the operational processing of European atmospheric satellite sensors and plays a key role in related satellite missions planed for the near future.