Automatic Interpretation of Downhole Seismic Geophysical Data Accounting for Uncertainty and Measurement Error

Abstract

Accurate quantification of the shear-wave velocity, Vs, and thus the small-strain shear modulus, G0, is a critical geotechnical design task. Recent research has demonstrated the high level of complexity and uncertainty associated with acquisition and interpretation of relevant in situ test data. However, traditional interpretation approaches deliver individual, deterministic values to the end user, and existing methods to communicate errors and uncertainties are limited in scope and/or insufficiently robust. This study builds on an existing method for uncertainty quantification to propose a more comprehensive, automatic approach to evaluate uncertainty in the Vs profile; the approach also estimates magnitudes of typical measurement errors during acquisition. A Bayesian framework is used to derive probabilistic Vs profiles with meaningful uncertainty ranges, and measurement errors are modeled explicitly, enabling their direct estimation and ensuring the evaluation of Vs is robust against real, imperfect data. Existing methods tend to neglect such errors, which can lead to incoherent or misleading Vs profiles. The robustness and value of the method is demonstrated via application to a wide range of publicly available and purposefully collected offshore and onshore data. The resulting measurement errors are realistic when compared with auxiliary measurements made during acquisition, supporting the model assumptions. The proposed approach can be incorporated into existing workflows to provide designers with valuable probabilistic profiles of Vs and G0.