INVERSION FOR SOIL PROPERTY THROUGH DATA ASSIMILATION OF RAYLEIGH WAVE

Abstract

The characterization of geomaterials in the near-surface plays a crucial role in geotechnical and geophysical engineering. Surface wave methods, utilizing Rayleigh and Love waves, are widely employed for near-surface characterization due to their broad exploration scale. However, the inversion process in geophysics faces several challenges due to the ill-posed nature of the problems. In this paper, we propose a data assimilation approach based on the ensemble Kalman filter (EnKF) to estimate the parameters of geomaterials using active Rayleigh waves data. The EnKF is a popular ensemble data assimilation method known for its ability to handle high-dimensional and nonlinear problems. It approximates the distribution of state variables using a ensemble of a small number of samples. Unlike gradient-based generalized linear inversion methods, the EnKF-based inversion does not require the calculation of a Jacobian matrix, and the parameters are estimated based on the statistical relationship between parameters and observations. The proposed method enables joint inversion, allowing the simultaneous inversion of different types of observations to reduce non-uniqueness in the inversion problem. Numerical experiments demonstrate the effectiveness of the EnKF-based inversion method in estimating one-dimensional soil property models using dispersion curves as observation data. The method achieves accurate results with a small number of samples and provides quantitative estimates of the uncertainty. Joint inversion of dispersion curve and first arrival data was employed to estimate the parameters of a two-dimensional model. The results indicate that the accuracy of joint inversion using multiple types of observed data is higher.