Spatial clustering of sound speed distributions

It is increasingly important to our understanding of the effects of climate change on our oceans that we can model sound propagation in the future environments that are being predicted. This brings conflicting demands to increase the fidelity of sound propagation models, to account for a greater number of dynamic ocean properties, and their efficiency, in particular, to account for the spatial and temporal distribution of sound speed data. In this paper, different spatial clustering methods are developed to group similar sound speed data into a small number of clusters based on acoustically relevant properties of the sound speed structure. The clusters then represent the sound speed distribution, derived from thousands of individual sound speed profiles, with a small number of effective sound speed profiles which have similar properties. This is illustrated on data for the Norwegian Sea where the performance of each method is evaluated, using measures of the separation between clusters, and by comparing sound speed predictions using profiles sampled from each cluster and using the effective sound speed profile. We show that spatial clustering is an effective tool for reducing the number of sound speed profiles that are needed to represent their spatial distribution.