Underwater sound propagation -- how sound waves move in the ocean and the effects of factors like temperature, ocean depth, seafloor composition and currents -- is the research focus of Eliza Michalopoulou, PhD, associate professor of mathematical sciences. She uses principles from signal processing and wave propagation, along with an approach she developed using a Gibbs sampling, an analytic tool for probabilistic inference, to solve problems such as localizing the source of a sound -- important in defense applications -- and environmental modeling, for example, studying global warming by monitoring sound speed in the ocean.
The figure above shows the source location estimate obtained using (a) the conventional narrowband Bartlett processor and (b) the Gibbs sampling localization-deconvolution approach developed by Dr. Michalopoulou. The correct source location is at 2 km in range and 34 m in depth. The conventional Bartlett processor produces a range-depth surface with the peak away from the true source location. The Gibbs sampling approach produces a clear ambiguity surface with the main mode at the correct source location.