Analysis of Long-Period Empirical Green's Functions at GSN Stations

Abstract:

We analyze three years of ambient noise data from 144 broadband stations at the Global Seismological Network (GSN) to obtain empirical Green's functions between the station pairs based on cross-correlation calculations. Vertical-component continuous seismic recordings are downloaded from IRIS, filtered between 50 and 150 seconds and down-sampled to 1 Hz after instrument responses are removed. Daily cross-correlation functions are then calculated for each station pair in the frequency-domain with a two-step pre-whitening applied. The quality of the stacked cross-correlation functions improves as the number of stacked traces increases and the empirical Green's functions also become more symmetric. The empirical Green's functions for GSN station pairs are overall in good agreement with synthetic surface-wave Green’s functions calculated in a reference earth model PREM. The quality of the empirical Green's functions decreases with increasing distance. For example, more than 80% of the empirical Green's functions show clear surface-wave arrivals at distances less than 45 degrees while only 40% are of good quality at distances greater than 120 degree. Finally, we measure surface-wave phase delay times between the empirical Green's functions and reference seismograms computed for PREM and compare measurements with predicted delay times calculated based on global phase-velocity models.