Using Acceleration Records as Diffuse Fields for Tomography of the Valley of Mexico City: Synthetic Results

In order to improve the understanding of the seismic response of Mexico City Valley, we have proposed to perform a tomography study of the seismic wave velocities. For that purpose, we used a collection of acceleration seismograms (corresponding to earthquakes with magnitudes ranging from 4.5 to 8.1 and various epicentral distances to the City) recorded since 1985 in 83 stations distributed across the Valley. The H/V spectral ratios (obtained from average autocorrelations) strongly suggest these movements belong to a 3D generalized diffuse field. Thus, we interpret that cross-correlations between the signals of station pairs are proportional to the imaginary part of the corresponding Green function. Finally, the dispersion curves are constructed from the Green function which lead to the tomography. Other tomographies have already been made around the world using either the seismic coda or seismic noise. We used instead the ensemble of many earthquakes from distant sources that have undergone multiple scattering by the heterogeneities of the Earth and assume the wave fields are equipartitioned. The purpose of the present study is to describe the different steps of the data processing by using synthetic models. The wave propagation within an alluvial basin is simulated using the Indirect Boundary Element Method (IBEM) in 2D configuration for the propagation of P and SV waves. The theoretical Green function for a station pair is obtained by placing a unit force at one station and a receiver at the other. The valley illumination is composed by incoming waves which are simulated using distant independent sources and several diffractors. Data process is validated by the correct retrieval the theoretical Green function. We present here the in-plane Green function for the P-SV case and show the dispersion curves constructed from the cross-correlations compared with analytic results for a layer over a half-space. ACKNOWLEDGEMENTS. This study is partially supported by AXA Research Fund and by DGAPA-UNAM under Project IN104712.