Abstract:
Ocean acoustic tomography uses acoustic signals to infer the environmental properties of
the ocean. The procedure for tomography consists of low frequency acoustic transmissions
at mid-water depths to receivers located at hundreds of kilometer ranges. The arrival
times of the signal at the receiver are then inverted for the sound speed of the background
environment. Using this principle, experiments such as the 2004 Long Range Ocean Acoustic
Propagation EXperiment have used acoustic signals recorded across Vertical Line Arrays
(VLAs) to infer the Sound Speed Profile (SSP) across depth. The acoustic signals across the
VLAs can be represented in terms of orthonormal basis functions called modes. The lower
modes of the basis set concentrated around mid-water propagate longer distances and can
be inverted for mesoscale e ffects such as currents and eddies. In spite of these advantages,
mode tomography has received less attention. One of the important reasons for this is that
internal waves in the ocean cause significant amplitude and travel time fluctuations in the
modes. The amplitude and travel time
fluctuations cause errors in travel time estimates.
The absence of a statistical model and the lack of signal processing techniques for internal
wave e ffects have precluded the modes from being used in tomographic inversions.
This thesis estimates a statistical model for modes a ffected by internal waves and then
uses the estimated model to design appropriate signal processing methods to obtain tomographic observables for the low modes. In order to estimate a statistical model, this thesis
uses both the LOAPEX signals and also numerical simulations. The statistical model describes the amplitude and phase coherence across diff erent frequencies for modes at diff erent
ranges. The model suggests that Matched Subspace Detectors (MSDs) based on the am-
plitude statistics of the modes are the optimum detectors to make travel time estimates for
modes up to 250 km. The mean of the travel time estimates is close to the mode travel
times for the background SSP. The travel time estimates produced by the MSDs have a
smaller variance than other signal processing methods that do not take into account the
statistics of the mode signals. The MSDs are applied to the LOAPEX signals to make travel
time estimates for modes received at ranges of 50 km and 250 km. The estimated sound
speed inverse for the mid-water depths is consistent with spot measurements made during
LOAPEX. Work sponsored by Office of Naval Research Grant N00014-06-1-0223.
