Spatial Modelling of Cloud

Abstract

Clouds have always been a source of scientific research as they play an important role in regulating the Earth’s weather. Much of the information we get about clouds is obtained by satellite images, which need to be processed in order to obtain interesting information. Indeed, due to the existence of different cloud types, there is a need of building a reliable cloud classification method. The aim of this thesis will be to create a spatial retrieval of cloud types, combining visible and infrared satellite images. This includes a preprocessing of the data to separate cloudy pixels from the underlying surface, a model of cloudiness using a radial basis functions neural network and a probabilistic method to classify clouds.