MULTIMODE INTERFERENCE PROBLEM AND SOLUTION IN LOVE WAVE TOMOGRAPHY

Surface waves have been widely used to image the variations of seismic velocity and anisotropy in the crust and upper mantle. However, different modes of surface waves with close group velocities interfere with each other, causing large uncertainties in phase velocity measurements and tomography models. In this study, we have computed a series of synthetic data to investigate Love wave mode interference, developed a method to separate the waveforms for different modes, and applied the method to Love wave tomography in northeastern America. We generate synthetic Love wave seismograms for different focal mechanisms and source depths and find that the energy of the first higher mode is not sensitive to source types but mainly controlled by source depths. We have developed a stacking-stripping method to separate the waveforms of the fundamental mode and the first higher mode Love waves. First, we detect different modes and estimate their phase velocities in the group-phase velocity diagram at an interested frequency band. Second, the phase-shifting and stacking approach is adopted to extract the fundamental mode waveforms. Then we extract the first higher mode waveforms by reapplying the stacking method to the residual waveforms, which are obtained by stripping the separated fundamental mode from the original data. This method has been validated using synthetic Love wave data and earthquake data recorded by the USArray Transportable Array (TA) stations. We also apply the stacking-stripping method in Love wave tomography in northeastern North America. Love wave data are from 60 large distant earthquakes recorded at 220 TA stations. Dispersion analyses show noticeable higher modes between 0.015 Hz and 0.025 Hz for 15 events. We apply the stacking-stripping method to these events to extract the fundamental mode waveforms at 0.02 Hz. Then a phase velocity map is constructed by applying the two-plane-wave inversion method to the fundamental mode data with and without mode separation analysis, respectively. The result with mode separation analysis improves the image of the slow anomaly in New England and reduces the data misfits and phase velocity uncertainties. We conclude that the stacking-stripping method helps improve Love wave data quality and tomography models.