HR: 0830h
AN: S31E-0808    [PDF]
TI: A New Global Model for 3-D variations in P Wave Speed in Earth's Mantle
AU: Karason, H
EM: hrafnkellk@kaupthing.net
AF: Massachusetts Institute of Technology, 77 Mass. Av., Cambridge, MA 02139 United States
AU: * van der Hilst, R D
EM: hilst@mit.edu
AF: Massachusetts Institute of Technology, 77 Mass. Av., Cambridge, MA 02139 United States
AU: Li, C
EM: changli@mit.edu
AF: Massachusetts Institute of Technology, 77 Mass. Av., Cambridge, MA 02139 United States
AB: In an effort to improve the resolution of mantle structure we have combined complementary data sets of short- and long period (absolute and differential) travel time residuals. Our new model is based on short period P (N\~7.7x10**6), pP (N\~2.3x10**5), and PKP (N\~16x10**4) data from the catalog by Engdahl et al (BSSA, 1998), short-period PKP differential times (N\~1600) measured by McSweeney \& Creager, and long-period differential PP-P times - N\~20,000 measured by Bolton \& Masters and N\~18,000 by Ritsema - and Pdiff-PKP (N\~560) measured by Wysession. Inversion tests, spectral analysis, and comparison with geology indicate that the large-scale upper mantle structure is better constrained with the addition of PP-P, whereas the Pdiff and PKP data help constrain deep mantle structure (Karason \& Van der Hilst, JGR, 2001). The long period data were measured by cross-correlation. We solved the system of equations using 400 iterations of the iterative algorithm LSQR For the short period (1 Hz) data we use a high frequency approximation and trace rays through a fine grid of constant slowness cells to invert for mantle structure. For low frequency Pdiff and PP data we account for sensitivity to structure away from the optical ray path with 3-D Frechet derivatives (sensitivity kernels) estimated from single forward scattering and projected onto basis functions (constant slowness blocks) used for model parameterization. With such kernels the low frequency data can constrain long wavelength heterogeneity without keeping the short period data from mapping details in densely sampled regions. In addition to finite frequency sensitivity kernels we optimized the localization by using a parameterization that adapts to spatial resolution, with small cells in regions of dense sampling and larger cells in regions where sampling is more sparse (the total number of cells was \~ 350,000). Finally, we corrected all travel times and surface reflections for lateral variations in crust structure using CRUST2.0
DE: 7200 SEISMOLOGY
DE: 7203 Body wave propagation
DE: 8180 Tomography
SC: Seismology [S]
MN: 2003 Fall Meeting