New time-lapse seismic tomographic scheme based on double-difference tomography and its application in monitoring temporal velocity variations caused by underground coal mining

Time-lapse seismic tomography aims at obtaining temporal velocity variations of the subsurface medium in different time periods, which could be very useful for volcanic monitoring, earthquake prediction and fault zone damage evaluation. The standard time-lapse velocity tomograms are generally obtained by subtracting velocity models resulting from separate seismic tomography for different time periods. However, this could introduce some artefacts in temporal velocity changes because of different data distribution and data quality at different time periods. In this study, we propose a new time-lapse seismic velocity tomography method that is based on the concept of double-difference (DD) seismic tomography. We redefine the DD equation by considering two events from two different time periods in order to evaluate the temporal velocity changes. The new method inverts data from two adjacent epochs simultaneously for the temporal velocity model changes to minimize differential arrival time residuals. As a result, it is less affected by different data distribution and quality in different time periods. We have applied the new method to a surface microseismic monitoring dataset for underground longwall coal mining. It is known that risks associated with unstable rock formation due to stress redistribution caused by coal mining could be devastating. By evaluating temporal velocity changes, we could derive stress redistribution due to mining process and then better understand potential risks. The resolution tests show that compared to standard time-lapse tomography method, our proposed method has higher resolutions in recovering temporal velocity changes. The application to the real microseismic dataset clearly shows that velocity increases by similar to 0.15 km s(-1) in the front of mining face where the stress increases and velocity decreases by similar to 0.23 km s(-1) in the gob area where the materials are loose, which is expected from numerical modelling of stress distribution and is more consistent than the standard time-lapse seismic tomography.