Waveform inversion for the geoacoustic parameters of the ocean bottom

Abstract

The geoacoustic properties of ocean sediments in deep water environments are important parameters necessary to predict low-frequency acoustic fields in the water column. A full wave method for obtaining the geoacoustic parameters from the acoustic field measured as a function of range with a cw source is presented. By assuming horizontal stratification, the unknown geoacoustic parameters are reduced to functions of one variable, i.e., depth. The problem of estimating the geoacoustic properties of the ocean floor is then cast as a parameter estimation problem in which a cost function φ(m), where m is a vector containing the unknown parameters, is minimized. This problem is then solved using a nonlinear optimization algorithm. This algorithm requires the determination of the derivative ∂φ/∂m. For a fluid bottom model an efficient algorithm for obtaining these partial derivatives is presented. The performance of the inversion algorithm is studied using noise free and noisy synthetic data. These inversions are carried out using the complex pressure field and the magnitude of the field as data. For the noise-free case, both approaches yield estimates close to the true value. In the case of noisy data, inversions carried out using the magnitude of the pressure field as data do not perform as well as inversions where the data are the complex pressure field. However, in both cases the algorithm is stable. The effect of modeling errors on the estimates is studied and it is shown that even small errors in source/receiver location lead to significant errors in the estimates. The effect of modeling the sediment as a fluid on the estimation of its geoacoustic properties is studied. In the case where the sediment shear speed is much smaller than compressional wave speed, the fluid approximation has no significant effect on the estimate of the compressional wave speeds. On the other hand, if the shear speed is such that considerable conversion exists, the fluid bottom model leads to a poor estimate of the compressional wave speed. In both cases the estimates of compressional wave attenuation and density are significantly affected by the fluid approximation. Finally this method is applied to data obtained in a field experiment and an estimate of the compressional wave speed profile in the sediment layers is obtained. This result is compared with the model obtained by iteration of forward models [G. V. Frisk et al., J. Acoust. Soc. Am. 80, 591–600 (1986)].