New and refined constraints on three‐dimensional Earth structure from normal modes below 3 mHz

Abstract

Abstract. We present the results of generalized spectral fitting (GSF) regressions which estimate normal mode structure coefficients for the observable spheroidal and toroidal free oscillation multiplets below 3 mHz. The size, accuracy, and precision of our new catalogue of modal constraints make it a powerful new tool for assessing and refining three-dimensional Earth models. The estimates include more than 3100 coefficients for 90 multiplets and 25 pairs of coupled multiplets, including several deep mantle overtones previously obscured by fundamental modes. The coefficients constrain mantle structures of both even and odd spherical harmonic degrees, through degree 12 in some cases. Improvements in accuracy and precision have been achieved with three innovations: the development of GSF, an enhancement of the established spectral fitting technique which incorporates both Coriolis and structural coupling between multiplets; the application of G SF to an edited, high signal-to-noise and geographically diverse data set of more than 4500 seismograms from 33 high moment earthquakes; and the assignment of coefficient uncertainties using a Monte Carlo method to simulate the effects of seismic noise, theoretical errors, and coefficient covariances. The results of G SF are assessed by examining the internal consistency of estimated coefficients and through comparisons with recent mantle models. The new catalogue of structure coefficients and uncertainties is available as an electronic supplement to this paper and through the University of Colorado internet site.

Regularization uncertainty in density models estimated from normal mode data

“…Recent analyses have produced new catalogues of normal mode structure coefficients [e.g., Tromp and Zanzerkia, 1995;He and Tromp, 1996;Resovsky and Ritzwoller, 1998]. At long wavelengths, and particularly in the upper-mantle transition zone and the mid-mantle (∼ 700−2000 km), these measurements complement the sensitivity of body-wave and surface-wave data.…”

“…At long wavelengths, and particularly in the upper-mantle transition zone and the mid-mantle (∼ 700−2000 km), these measurements complement the sensitivity of body-wave and surface-wave data. Misfit statistics to normal mode data help to discriminate among recent models [Resovsky and Ritzwoller, 1998;hereafter R&R98], and inversions of these data have produced a new long-wavelength scaling-relation mantle model [Resovsky and Ritzwoller, 1999;hereafter R&R99]. There have also emerged models that exploit normal mode sensitivity to vp, ρ, and topography on major radial discontinuities (hereafter called boundary topography) throughout the mantle [Ishii and Tromp, 1997;Tromp and Ishii, 1998;Kuo et al, 1998].…”

Error bars for the global seismic Q profile

“…In this way, we make the 150-300 s spheroidal fundamental modes less important relative to data from overtones and from the high Q fundamentals, with periods longer than 300 s that are less subject to noise biasing. The latter kind of data has been greatly improved with generalized spectral fitting analysis of the Coriolis-coupled fundamentals in the 300-600 s band [28]. We also test our results for sensitivity to this approximation by performing breducedQ data searches in which only one of the two controversial data sets is included, with its originally stated error.…”

“…The unpublished data attributed to bUmQ on the REM pages have anomalously low uncertainties, which, given that there is no documentation to suggest that a superior technique was used, have been multiplied by a factor of 5.0 to provide consistency with the other data. Similarly, uncertainties for the He and Tromp [27] measurements were multiplied by 2.0 (because of errors known to be caused by the use of too few events [28]), those for Resovsky and Pestana (2002) by 3.0 (based on our familiarity with those data), and those for the inconsistent 150-300 s SW and FO measurements of spheroidal fundamentals by 3.0. Except for the latter adjustment, which is discussed below, these changes improve model misfits without altering the other characteristics of the model space.…”

“…These include SW fundamental mode measurements [7,[22][23][24], FO fundamental data [25], FO overtone and longer period fundamental measurements [26][27][28][29] and some unpublished updates of these data. The unpublished data attributed to bUmQ on the REM pages have anomalously low uncertainties, which, given that there is no documentation to suggest that a superior technique was used, have been multiplied by a factor of 5.0 to provide consistency with the other data.…”