In the microseismic and seismology field, determining the hypocenter of seismic events is necessary, and requires an accurate velocity model. In conventional FWI, one of the outputs is often a P- and/or S-wave velocity model obtained from (in most cases) sources on the surface. It is not difficult to imagine a framework where microseismic events could be used as additional subsurface sources that could increase the illumination in the reservoir. Furthermore, the velocity model obtained from such a scheme would be useful to re-locate said microseismic events more accurately. Therefore, this symbiotic relationship can be taken advantage of to formulate a FWI implementation where microseismic events are used to simultaneously update the velocity model, and the source position. This would involve two updates at each iteration - one for the velocity model and one for the source position. In this report, we explore in detail the source-term gradient in an elastic 2D formulation. We discuss the effect of the starting position, dominant frequency, moment tensor, and receiver geometry. Furthermore, we explore the impact of cross-talk due to having an incorrect starting velocity model. Finally, we end with preliminary results with a real dataset from the Horn River Basin, British Columbia.
View full article as PDF (1.28 Mb)