Finite-Frequency Sensitivity Kernels for Seismic Noise Interferometry Based on Differential Time Measurements

Abstract

Full waveform adjoint tomography has achieved great success in applications from global Earth structure using earthquakes to exploration seismology using active sources. When combined with ambient seismic noise data, however, the ambient noise cross correlations are subject to strong variability and bias due to nonstationary and unevenly distributed noise sources. As a result, the shape of the structure sensitivity kernel can deviate significantly from the classic banana-doughnut kernel of a point source, which has been used in previous studies as an inference from empirical Green's function and contains bias. In this study I calculate the sensitivity kernel for ambient noise cross correlation by introducing an additional station to the classic two-station setting. I compute sensitivity kernels for differential traveltime measurements. These differential sensitivity kernels show promise for canceling the overlapping part of the original source and structure kernels for pairs of stations in interferometry, thus significantly reducing the effect of nonisotropically distributed and nonstationary noise sources. I apply the ambient noise differential sensitivity kernel to synthetic data examples based on 2-D membrane waves, though the approach will apply for 3-D as well. Our results for multiple station pairs show promise for velocity tomography based on seismic noise interferometry when perfect information on noise source distribution is not available.