Broadband nonlinear inversion for geoacoustic parameters in shallow water

The properties of the bottom play an important role in the acoustic propagation in shallow water. Direct estimation of bottom properties is costly, time-consuming and spatially limited. Indirect methods based on acoustic measurements may offer a suitable alternative approach. Linear inversions are limited by the fact that they often require detailed background information. The relationship between the measurement and the unknown parameters is non-linear and the linear schemes rely on perturbation around this background. Non-linear inversion methods based on global optimization provide a viable option for the estimation of bottom properties such as compressional sound speed, attenuation and potentially density. In this work, a non-linear inversion technique, using travel time of broadband signals generated by explosive sources, is presented. A Genetic Algorithm (GA) is used as the basic tool for global optimization. The compressional wave speed of the sediment is mapped using this inversion scheme. The inversion for compressional wave speed is based on the travel time dispersion characteristics of the broadband signals. Another inversion technique, based on mode amplitude ratios, is used to estimate the compressional wave attenuation. The technique can be applied to shallow water ocean waveguides in depths up to 500 m and over ranges of up to 50 km. In addition to these sediment parameters, source-receiver range and sound speeds in the water column are also estimated along with the compressional wave speeds and attenuation.; The inversions are carried out using synthetic as well as field data. Field data was collected during the Shelf Break Primer Experiment, which was conducted in the summer of 1996. Compressional wave speed inversions corresponding to various types of environmental assumptions were carried out. They are (1) Range independent waveguide. (2) Range dependent bathymetry and water column parameters and range independent sediment parameters. (3) Range dependent bathymetry, water column properties and sediment parameters.; The inversions are compared with data from a deep core and a number of shallow gravity cores. All the inversions were carried out in the shelf region of the experimental area. In this area the variations in water column and bathymetry were mild and gradual. Hence adiabatic theory was used to model the acoustic propagation in range dependent inversions. The error estimates were based on an a posteriori analysis and Hessians. Resolution lengths were also calculated by applying a linear perturbation approach and using a posteriori analysis.; The compressional wave attenuation is estimated using only the range independence assumption. This inversion technique is applied to synthetic as well as field data. The estimated values are compared with historic data and attenuation calculated using the gravity core data.; The increased effectiveness of the non-linear optimization schemes in dealing with multi-modal objective functions comes at a high cost of computation. In order to increase the efficiency of the inversion schemes without sacrificing the robustness, various hybrid schemes are implemented. These hybrid approaches consist of combining the Genetic Algorithm with local search methods (Levenberg-Marquardt scheme, Differential Evolution and Neighborhood Approximation).