People commonly use cross-correlation based seismic interferometry to retrieve Green's functions from ambient seismic noise recordings. This approach requires that seismic sources are isotropically distributed in all directions around two receivers. However, this assumption is rarely valid in practice. Thus people have begun to apply full-waveform inversion theory to seismic noise crosscorrelation functions, functions that include both source and structure information. Source information (like locations and strengths) are essential for accurate structure information estimation. In this paper, we explain physically two types of source sensitivity kernels: one derived from traveltime mists and the other derived from waveform mists. We use these kernels for source inversion, and demonstrate the benets of using multicomponent crosscorrelations in this source estimation process.
