A simple method for improving crustal corrections in waveform tomography

P>Accurate accounting for the effects of crustal structure on long-period seismic surface waves and overtones is difficult but indispensable for determining elastic structure in the mantle. While standard linear crustal corrections (SLC) have been shown to be inadequate on the global scale, newer non-linear correction (NLC) techniques are computationally expensive when applied to waveforms containing higher frequencies and/or overtones. We devise, implement, and verify a modified SLC approach that mimics the non-linear effects of the crust without substantially increasing the computational costs. While theoretically less accurate than the NLC approach, in practice, the reduced computational costs allow this 'modified linear correction' (MLC) technique to be applied at higher frequencies and using more detailed crustal regionalizations than is possible with NLC. In order to validate the MLC technique, we use the spectral element method to carry out a series of synthetic tests. These tests demonstrate that MLC nearly eliminates the contamination of mantle isotropic structure by unmodelled crustal effects, which can be substantial in the uppermost 150 km when using SLC. Furthermore, we show that MLC significantly reduces contamination of anisotropic structure compared to SLC, the inaccuracies of which are significant in the upper 250 km and can even obliterate the mantle anisotropic signature at depths shallower than 100 km. Finally, we apply the MLC technique to a real long period waveform data set and demonstrate the benefit of improved crustal corrections on the retrieved model.