The use of shape, appearance and the dynamics of clouds for satellite image interpretation

Lewis, Hugh Gifford (1998) The use of shape, appearance and the dynamics of clouds for satellite image interpretation. University of Southampton, Doctoral Thesis.

Abstract

Atmospheric processes are manifested through the formation and dispersal of clouds and cloud patterns. Consequently, clouds are of great importance to weather forecasting. The ability to study cloud patterns over wide areas and in regions where meteorological data is sparse has led to the development of subjective techniques, that use conceptual models, to analyse the cloud cover contained in satellite images. These conceptual models describe the shape, appearance and dynamics of clouds and cloud patterns and are able to provide an abundance of information about the current and future meteorological situation. Automatic interpretation techniques, however, typically rely on data sources other than satellite images to understand atmospheric processes.

In this research a new approach to satellite image interpretation, based on adaptive segmentation techniques and a neural network matching method, has been developed. The approach generates shape, appearance and dynamics characteristics of cloud features identified in satellite images at three spatial scales. These characteristics are combined in an holistic analysis to provide a detailed interpretation of the meteorological situation. Several key advantages are offered by this approach over current methods for automatic satellite image interpretation. Firstly, no data sources in addition to satellite images are required. Secondly, characteristics which quantify the shape, appearance and dynamics of small-scale cloud features are presented in the context of large-scale cloud features, and thirdly, dynamics characteristics of small-scale convective cloud features are able to model the lifecycles of these features.

Applied to real satellite images, the interpretations based on the holistic analysis of cloud shape, appearance and dynamics characteristics were shown to be in agreement with the actual meteorological situations. The use of adaptive segmentation and matching methods allowed the key cloud features to be identified, described and matched over image sequences automatically. Subsequent interpretations and forecasts based on this information can be made automatically using intelligent processing techniques.

This record has no associated files available for download.

Contributors

Download statistics