Constrained tomography of realistic velocity models in microseismic monitoring using calibration shots

The knowledge of the velocity model in microseismic jobs is critical to achieving statistically reliable microseismic event locations. The design of microseismic networks and the limited sources for calibration do not allow for a full tomographic inversion. We propose optimizing a priori velocity models using a few active shots and a non‐linear inversion, suitable to poorly constrained systems. The considered models can be described by several layers with different P‐ and S‐wave velocities. The velocities may be constant or have 3D gradients; the layer interfaces may be simple dipping planes or more complex 3D surfaces. In this process the P‐ and S‐ wave arrival times and polarizations measured on the seismograms constitute the observed data set. They are used to estimate two misfit functions: i) one based on the measurement residuals and ii) one based on the inaccuracy of the source relocation. These two functions are minimized thanks to a simulated annealing scheme, which decreases the risk of converging to a local solution within the velocity model.

The case study used to illustrate this methodology highlights the ability of this technique to constrain a velocity model with dipping layers. This was performed by jointly using sixteen perforation shots recorded during a multi‐stage fracturing operation from a single string of 3C‐receivers. This decreased the location inaccuracies and the residuals by a factor of six. In addition, the retrieved layer dip was consistent with the pseudo‐horizontal trajectories of the wells and the background information provided by the customer. Finally, the theoretical position of each calibration shot was contained in the uncertainty domain of the relocation of each shot. In contrast, single‐stage inversions provided different velocity models that were neither consistent between each other nor with the well trajectories. This example showed that it is essential to perform a multi‐stage inversion to derive a better updated velocity model.