A Quantitative Study of How Effective the Transformed Gravity Gradient Tensor is for Subway Tunnel Imaging

Abstract

We investigate the efficacy of spatial frequency filtering in improving gravity inversion when gravity gradient data are unavailable. The technique simulates gravity gradient data from gravity readings by extracting distinct frequency components. Previous research shows that the joint inversion of gravity and gradient data derives accurate density models. Given the lack of gradient instruments, the potential of frequency filtering to enhance inversions is explored. A series of inversion experiments are conducted using synthetic and field-measured gravity data. The research aims to determine if filtered gradient data, transformed from gravity data, can enhance inversion results compared to using gravity alone. Synthetic models assess the filter, showing that low-pass filtering at an optimal cutoff enhances the correlation between transformed and forward-modeled gradient components. Inversions using transformed gradients consistently outperform those with only gravity. Component combinations and cutoff wavelengths must be optimized to achieve the best reconstruction. Applying the method to Beijing subway tunnel gravity readings, gravity and transformed Uzz gradient data are combined to detail underground double tunnels. The information agrees with engineering data, validating the approach for improving inversions, especially for shallow, large targets. This work establishes the practicality of frequency filtering when direct gradients cannot be measured, contributing valuable insights into better characterizing geology through enhanced inversion. Findings imply improved interpretation; particularly where gradient acquisition proves difficult.