Geoacoustic inversion based on covariance weighted Bartlett estimator

Bartlett estimator is a widely used objective function in matched field inversion (MFI) to quantify the misfit between measured acoustic field and replica field. In this paper, under the assumption of additive spatially correlated Gaussian residual errors, a normalized covariance weighted Bartlett estimator is employed in geoacoustic inversion as the objective function utilizing global optimization algorithm, differential evolution (DE). Both simulation and SW06 data are processed using this new weight-Bartlett strategy and the inversion results are compared with conventional Bartlett strategy. In the simulation, spatially colored Gaussian noise vectors are added to the generated pressure vectors. Using weight-Bartlett strategy, the estimated geoacoustic parameters including sediment thickness and sound speed of sediment and basement which are secondary sensitive to misfit function are closer to the true values than Bartlett strategy, leading to better inversion results. For source range and depth which are most sensitive to misfit function, both strategies achieve similar accurate results. While for parameters less sensitive to misfit function, offsets with respect to true values occur. For SW06 experimental data, the estimated geoacoustic parameters, e.g. sediment thickness and sound speed of sediment and basement of both strategies are consistent with other results at the same marine.