Step-wise relocation of ISC earthquake hypocenters for linearized tomographic imaging of slab structure

A vast volume of seismic phase data, computed from routinely determined hypocenter locations, is presently available for seismic imaging techniques such as tomographic inversion and residual sphere analysis. Routinely reported earthquake hypocenters, however, can be in error by several tens of kilometers. These biases in hypocenter location can result in the misidentification of seismic phases and the loss of structural signal in the seismic phase data. Thus, the value of these data in constraining seismic images is significantly reduced. Furthermore, hypocenter mislocations degrade the linearization of the tomographic problem and map into the images produced by tomographic inversion. To obtain adequate reference hypocenters for linearized inversion, we used iasp91 software and reported arrival times to reprocess ISC hypocenters and phase data for northwest Pacific earthquakes. Subsequently, we inverted P- and pP-wave residuals for Earth structure and source (mis)location. We describe this step-wise relocation of ISC hypocenters, which underlies the linearization of the tomographic inversion, and compare the relocation vectors before and after inversion. The hypocenter relocations determined prior to inversion are of the order of 10 km, which is significant with regard to both the estimated standard errors and the effect on travel-time residuals, and systematic with regard to earthquake position in the subducted slab. In contrast, the spatial components of the relocation vectors determined upon inversion of P and pP residuals are up to an order of magnitude smaller, do not reveal a correlation with location in the seismic zone, and do not generally exceed the noise level. The incorporation of pP data in the relocation was essential to remove the depth bias prior to inversion, to retrieve slab signal that might have been absorbed in mislocation, and to constrain the focal depth upon seismic inversion. Non-linear inversion schemes will not be efficient in removing the depth bias or in retrieving lost slab signal unless seismic phases that contain additional information about the depth bias, like the depth phase pP, are used