Combining Full Waveform Inversion and Traveltime Tomography

Abstract:

By alternating full waveform inversion and traveltime tomography we attempt to merge their complementary merits and produce models of the European continent that explain both waveform and traveltime data sets.

Technological developments and advances in theoretical and numerical seismology allow us to assimilate complete waveforms for the solution of full waveform tomographic problems. On regional to continental scales, however, full waveform inversion is still and will also in the long term be limited to an intermediate period band because of computational limitations. Valuable information contained in high-frequency P and S wave traveltimes cannot be exploited. Since mainly surface waves can be observed in the period range that is technically feasible, full waveform inversion yields excellent results in the upper 300 km, where surface wave sensitivity is large. At greater depth S velocity anomalies are less well resolved and generally P velocity heterogeneities are not well constrained at all.
In contrast, classical traveltime tomography intrinsically incorporates information from high-frequency body waves. However, traveltime methods are deficient for longer periods due to the commonly used infinite frequency approximation of the wave equation.

To combine their complementary assets and extend the spectrum of exploited information, we develop an inversion scheme, where full waveform inversion and ray tomography alternate, thereby avoiding an explicit coupling of both methods and data sets. We put special emphasis on its ability to produce 3D models that explain both waveform and traveltime data sets. We start with a full waveform inversion model and machinery of Europe and Western Asia in conjunction with the traveltime tomography package FMTOMO. For the traveltime update we use the teleseismic data set that constrains the global model S40RTS, complemented with measurements distributed by the International Seismological Centre. Avoiding difficulties in modeling the complexities in the crust and upper mantle, the traveltime update primarily introduces information in deeper parts of the model. After alternating both methods the resulting model finally incorporates information from traveltime measurements and from waveform data, thus proving the compatibility of both approaches.