Earth 1 D , 2 D and 3 D Monte Carlo ambient noise tomography 1 using a dense passive seismic array installed on the 2 North Sea seabed

Xin Zhang,F. Hansteen,A. Curtis,S. Ridder

2019-01-01

Abstract:16 In a variety of geoscientific applications we require 3D maps of properties of the Earth’s 17 interior, and the corresponding map of uncertainties to assess their reliability. On the 18 seabed it is common to use Scholte wave dispersion data to infer these maps using inversion19 based imaging theory. Previously we introduced a 3D fully nonlinear Monte Carlo to20 mography method that inverts for shear velocities directly from frequency-dependent travel 21 time measurements, and which improves accuracy of the results and better estimates un22 certainties. Here for the first time we apply that method to real data and compare it to 23 two of those previous methods. We cross correlated 6.5 hours of ambient noise data recorded 24 on a dense seismic array over Grane, North Sea, and observed two Scholte wave modes. 25 For each mode, phase velocity maps are estimated using Eikonal tomography, which are 26 in turn used to study the shear-wave velocity structure of the subsurface. We applied 27 three nonlinear inversion methods to the Grane data: standard 1D depth inversions, a 28 2D joint inversion along a vertical cross-section, and a fully 3D inversion. We compare 29 the shear-velocity and uncertainty structures estimated along the same 2D cross-section. 30 Thus we show that the standard 1D inversion method causes errors in the results due 31 to independence of those 1D inversions, whereas the 2D and 3D inversions improve re32 sults by accounting for lateral spatial correlations. The 3D inversion bypasses the ini33 tial seabed Eikonal tomography step, and therefore avoids the errors it introduces into 34 subsequent 1D and 2D inversions. 35