Sensitivity Analysis of Wave-equation Tomography: A Multi-scale Approach
Earthquakes, viewed as passive sources, or controlled sources, like explosions, excite seismic body waves in the earth. One detects these waves at seismic stations distributed over the earth’s surface. Wave-equation tomography is derived from cross correlating, at each station, data simulated in a reference model with the observed data, for a (large) set of seismic events. The times corresponding with the maxima of these cross correlations replace the notion of residual travel times used as data in traditional tomography. Using first-order perturbation, we develop an analysis of the mapping from a wavespeed contrast (between the “true” and reference models) to these maxima. We develop a construction using curvelets, while establishing a connection with the geodesic X-ray transform. We then introduce the adjoint mapping, which defines the imaging of wavespeed variations from “finite-frequency travel time” residuals. The key underlying component is the construction of the Fréchet derivative of the solution to the seismic Cauchy initial value problem in wavespeed models of limited smoothness. The construction developed in this paper essentially clarifies how a wavespeed model is probed by the method of wave-equation tomography.
Mathematics Subject Classification (2000)35L05 35R05 35S50 35R30 86A15
Unable to display preview. Download preview PDF.
- 2.Brytik, V., de Hoop, M.V., Smith, H.F., Uhlmann, G.: Parametrix construction following polarization for the elastic wave equation with stiffness of limited smoothness: A multi-scale approach. Preprint (2009) Google Scholar
- 11.Evans, L.C.: Partial Differential Equations, vol. 19. Am. Math. Soc., Providence (2002). Google Scholar
- 14.Kingsbury, N.: Image processing with complex wavelets. Philos. Trans. R. Soc. Lond. A357, 2543–2560 (1999) Google Scholar
- 26.Stolk, C.C.: On the modeling and inversion of seismic data. Ph.D. thesis, Utrecht University (2000) Google Scholar