Analysis of fine structures in celerity profiles for oceanography by wavelets and nonlinear filtering

A sonar system's capability can be fully determined by the knowledge of the celerity profile. Today, we are able to calculate a celerity profile based on different parameters. But in real profile, we can find faint defaults which are very difficult to predict or to be studied. We call them fine structures. Recent studies (Flatte, 1979) have examined fine structures of the sound celerity under water using statistics. The basic difficulty is in accessing these fine structures. Fine structures are integrated in a celerity profile with macroscopic structures. In order to explore fine structures, we must extract them. The key idea of this article is to consider a celerity profile c(z) as a time signal s(t). Immersion z becomes time. We use causal filters and we must reduce the dephasing and board effects. To isolate the two components, two algorithms are used: an algorithm based on the median and averaging filtering and a second one based on multiresolution analysis. Then fine structures are examined and it is shown that the main features such as the autocorrelation function are regular with z. These methods allow us to avoid making hypotheses and eventually to check hypotheses classically made. (C) 1996 Academic Press, Inc.