Shallow water tomography: new trends in array processing – TOTS

The first key point in our approach is to introduce array-to-array signal processing as a way to significantly increase the number of acoustic observables extracted in a multipath oceanic environment. The nature of these observable will evolve comparing to classical ones (travel times) and we will introduce receive-launch angles as well as amplitudes.
The second key point concern the developed tomography inversion algorithms, which will be based on the use of sensitivity kernels in contrast to classical ray approach. Travel-time Sensitivity kernels, Amplitude Sensitivity Kernels adapted to array-to-array configuration have to be developed and integrated in the tomography process. We will also focus on full-wave inversion methods.
Inversion algorithms are then validated on (1) real at-sea data using two existing data set provided by the Scripps Institute of Oceanography and (2) laboratory scale data that result from the development of a research group on acoustics at LGIT.

En cours

En cours

En cours

TOTS stands for Tomographie OcÈanique en zone peu profonde : nouvelles perspectives en Traitement du Signal (Shallow water tomography: new trends in array signal processing) and addresses 3 different tomography problems:
- Classical tomography providing the spatio-temporal sound speed structure of a ocean portion,
- Detection-localization of a target in shallow water, performed using an impedance tomography
- Surface tomography.
These problems, usually considered separately, are studied jointly in the project as most of the methods developed for one of them can be used (if adapted) for the others.

Performing tomography in shallow water, which is still an almost unknown environment, is particularly interesting as its knowledge could help scientists to better understand streams, tides, human influence and pollution in coastal areas. Consequently, they need the precise knowledge of the spatial-temporal distribution of ocean temperature. Moreover, performing the other tomographies to localize an underwater source and to estimate the surface height is also very interesting as these problems have many applications (acoustic barrier in harbors for example) which are not solved yet.

As written by Munk et al, "The problem of acoustic tomography is to infer, from precise measurement of travel time, or of other properties of acoustic propagation, the state of the ocean traversed by the sound field". The first task to perform Ocean Acoustic Tomography is then to choose the best acoustic observables and to measure them as accurately as possible. Then, once the acoustic observables are extracted, the second step consists in building a forward model linking acoustic observables to ocean physical parameters. Finally, the last step of the tomographic process is to invert for ocean physical parameters such as for example, the spatial-temporal temperature distribution in the ocean.

The principal limitation is that tomography needs the separation and identification of the different arrivals in the signal which is a difficult task in a multipath recorded signal. One of the goals of our project is to overtake this limitation in shallow water configuration taking profit of source-receive array configuration. This will be done developing and using different approaches: double beamforming, adaptive double beamforming, high resolution methods.

The other main difficulty to perform shallow water tomography is to establish a link between acoustic observables and physical parameters we invert for. Classical tomography uses ray modelization of the propagation to perform this task but this model does not take into account acoustic diffraction. As a result, the way to include it in tomography inversion is to deal with Sensitivity Kernels that have been widely used in seismic tomography. Studying Sensitivity Kernels and their use in high-resolution shallow water tomography will be very useful to develop more accurate tomography methods.

Finally, all these methods will be validated on real data. We will have access to the existing FAF03 and FAF05 data set provided by the Marine Physical Laboratory (San Diego). We will also perform laboratory experiments in a more controlled environment. They will be performed in an ultrasonic tank developed by P. Roux at the Laboratoire de GÈophysique Interne et Tectonophysique (LGIT) in Grenoble.

This project is thus a multidisciplinary project, carried by researchers from different scientific fields, which deals with two main tasks: development of the methods (from the extraction of acoustic observables to inversion) and application of these methods on real data (at-sea and tank data) to perform high-resolution shallow water tomography.

Through this project, we aim at gathering, around Barbara Nicolas, a junior CNRS research scientist at Gipsa-lab, a work force in Grenoble devoted to research on shallow water tomography. The two laboratories involved in this project are GIPSA-lab and LGIT-Grenoble.