Overcoming challenges in time reversal for passive seismic source localization

Abstract

Passive seismic source localization plays a crucial role in understanding geodynamics, monitoring geological activities, forecasting geological hazards, and managing geological fracturing processes. This doctoral thesis presents three consecutive approaches to enhance the robustness and accuracy of time-reversal source localization. The proposed methods address challenges such as source imaging artifacts, low-resolution source images, sparse and small-aperture seismic data acquisitions, and unknown seismic velocity models.

In the first approach, the focus is on refining time-reversal imaging. Conventional time-reversal source imaging methods like autocorrelation imaging or grouped crosscorrelation imaging often suffer from source imaging artifacts due to the use of low-quality seismic data or less constrained velocity models. These artifacts typically degrade the quality of the image and can lead to subsequent misinterpretation, resulting in false source location estimation. To overcome this, the Gaussian-weighted crosscorrelation imaging condition is proposed. Each time step in this method includes dividing the back-propagated wavefield, weighting seismic amplitudes using Gaussian functions, and using a zero-lag crosscorrelation. This process effectively minimizes source imaging artifacts, resulting in high-resolution, low-noise source images. Numerical examples of complex models and field examples illustrate the method’s performance, demonstrating its effectiveness in identifying sources, even within clusters and under conditions of noisy and sparse-sampled data.

The second approach addresses another challenge posed by sparse and small-aperture seismic data acquisition in time-reversal imaging techniques. Such acquisitions often lead to false wave focusing due to insufficient wave illumination. To address the issues, the maximum-amplitude path method is introduced for source localization using maximum-amplitude paths. The paths are constructed from back-projected wavefields using receiver patches selected from the acquisition. They include the maximum amplitudes of the back-projected wavefronts for each considered time step. The proposed method exploits the continuity of the maximum amplitudes of the back-projected wavefields. The point of closest proximity (or crossing point) of the paths is the source location and the corresponding time is the source time. The maximum-amplitude path method successfully overcomes the acquisition problems and provides accurate source location and source excitation time even in challenging scenarios.

In the final approach, a data-driven hybrid workflow is presented to address the challenge of the lack of velocity model in time-reversal localization methods including the proposed first and second time reversal localization approaches. The proposed workflow can simultaneously invert the source location, excitation times, and velocity model using wavefront attributes of passive seismic data. By combining wavefront tomography and time-reversal methods, the workflow eliminates the need for detailed prior information, making it particularly applicable in practical scenarios. The proposed workflow comprises the following steps. First, a set of user-defined vertical gradient velocity models is designed. Time reversal is then used to estimate the source excitation times for each model. After that, these source times and gradient models are used in wavefront tomography. The second step uses an optimization procedure to refine the velocity model and the source excitation time. Each iteration of the optimization involves a sequential application of time reversal and wavefront tomography. In the final step, the source location is refined using the optimal velocity model and the Gaussian-weighted crosscorrelation imaging condition. The proposed workflow overcomes the limitations of time reversal in the absence of a velocity model, providing good velocity models and fairly accurate source locations and excitation times.