Estimating oceanographic properties from ambient noise

One of the most significant factors that affect the accuracy with which a target can be detected and localized with passive sonar in the ocean is the a priori knowledge, or lack thereof, of the local oceanographic environment. Acoustic propagation is strongly affected by water sound velocity variations, as well as the physical properties of the surface and bottom that form the boundaries of the ocean waveguide. The aim of this work is to make use of the ambient noise field to estimate these environmental properties. Wind and breaking waves at the surface cause Gaussian noise to radiate into the water. These stochastic signals propagate through the water and reflect off the surface and bottom over a wide range of angles and frequencies where they can be observed by a passive sonar array. The noise field itself thus contains information about the properties of the environment. Spectral snapshots computed from short-time Fourier transforms of observed noise data are circularly-symmetric complex-normally distributed. Using physics models for the noise sensor-sensor covariance matrix, the theoretical bounds on estimator performance can be obtained, and an asymptotically-optimal estimator can be implemented. This passive environmental estimation technique is explored through simulation and measured experimental data.