Estimating time-lapse velocity changes in the earth by full waveform inversion of repeating seismic events

Abstract:

Passive seismic monitoring of natural or induced seismicity can provide important information about time-variant subsurface phenomena, especially concerning fluid flow and stress-strain conditions. Traditional monitoring focuses on locating seismic events, and interpreting the results in terms of changes to subsurface properties. Instead, we directly estimate the subsurface velocity changes, which can be more quantitatively related to changes in fluid and stress states via rock physics and geomechanical relationships. The expected velocity changes are often so small that conventional traveltime tomography has difficulty in providing reliable detailed estimates. Our method is based on full waveform inversion (FWI) of the seismic data. FWI estimates a high-resolution subsurface model by fitting waveforms in the data domain based on a numerical solution to the wave equation. We evaluate the performance of our FWI by using a synthetic set of a limited number (up to three) repeating events recorded at a surface array of sensors. We use a realistic complex velocity model which includes fine-scale heterogeneity, and consider the case where fluid pore pressure has changed the stress state leading to a small velocity increase in two distinct thin layers. We assume that we have reasonable estimates of the event locations and the background velocity model before starting the inversion. We compare parallel, double-difference and bootstrapping techniques, and analyse effects of accuracy in the background velocity model to the inversion. FWI retrieves the correct velocity change with excellent accuracy at a high spatial resolution less than one wavelength. This is due to the ability of FWI to invert the full waveform data, rather than traveltime arrivals only. We analyse the inversion sensitivity kernels and show that the spatial resolution improves, and the required density of the sensor array reduces, for velocity changes nearer to the seismic source location.