Computational methods for detection and imaging of oil in sea water

The oil accident of Deepwater Horizon in the Gulf of Mexico has indicated the need for developing systems for detection of oil submerged in the water column and on the seafloor. There are currently no methods to efficiently map submerged oil in the sea. This leads to a poor understanding of the extent of contamination in case of an oil spill form a deep water drill, which reduces the efficiency of the clean-up process dramatically.

The purpose of this project is to develop a numerical method for imaging of oil in sea water, and for determination of the physical properties of the medium (speed of sound, density, viscosity and inhomogeneity).
As a novelty, an approach will be developed in order to allow inversion of acoustic waveform data, constrained by complex prior information which can be extracted from training images. The goal of this approach is to efficiently produce solutions to the inverse problem that not only fit the data, but are also statistically similar to the training images.

In this work it will be assumed that the main features of the sea water velocity field can be modelled as a smooth background velocity field, and that the polluted areas are concentrated in oil/water layers, characterized by an impedance contrast and a certain inhomogeneity. In this way the problem is seperated into two: (1) The travel time problem where ray tracing is used to optimize the background velocity to match the arrival times of reflected events, and (2) a reflectivity problem where a model of impedance contrasts are used to match reflected amplitudes. The a priori constraints on the solution will be supplied by (1) a thermodynamic model (provided by project partners), and (2) a geostatistical training-image model expressing constraints on the geometry (thicknesses, areal extent and shapes) of bodies of mixed oil/water, surrounded by near- homogeneous sea water.

Supervisor: Prof. Klaus Mosegaard,