Statistical Inversion of Acoustic Backscatter Profile Data with Sediment-Induced Attenuation

Abstract:

Multi-frequency acoustic backscatter inversion is a proven technique for measuring profiles of suspended sediment size and concentration. However, current inversion techniques tend to become inaccurate when sediment-induced acoustic attenuation is high (e.g., large suspended concentrations). In such cases, statistical fluctuations cause an accumulation of errors along the observed profile, which leads to an ill-posed inverse problem, even if the intrinsic properties of the scatterers are exactly known. Here, a statistical inversion methodology is introduced in which the ill-posed problem is regularized by explicitly considering the uncertainty of the observational data. The method uses the Extended Kalman Filter to estimate the propagation of uncertainty through the profile, followed by a variational smoothing step. This approach is shown to successfully estimate concentration and size profiles in cases where direct inversion fails. The method is also applied to laboratory experiments involving a multi-frequency megahertz acoustic profiler (MFDop) measuring a sediment-laden jet.