We have developed a double-difference algorithm to relocate earthquakes recorded at global seismic networks, using differential arrival times for first and later arriving regional and global phases to invert for the vectors connecting the hypocenters. Differential times are formed from global seismic bulletins and are accurately measured on similar seismograms by time domain waveform cross correlation. We evaluate the performance of this spherical, multiphase double-difference algorithm using three-dimensional regional-scale synthetic data and two sets of earthquake data in different tectonic settings. The first includes 3783 intermediate depth earthquakes that occurred between 1964 and 2000 in the subducting Nazca plate beneath northern Chile, where the relocated seismicity confirms a narrowly spaced double seismic zone previously imaged with temporary local seismic data. Residual statistics and comparison with accurately known locations indicate mean relative location errors at the 90% confidence level of 2.4 km laterally and 1.8 km vertically. Later events typically constrained by cross-correlation data have errors of 1.6 km laterally and 1.4 km vertically. The second data set includes 75 crustal earthquakes in the 1999 Izmit and Duzce, Turkey, aftershock sequences, where the double-difference solutions image orientation and dip of individual fault segments that are consistent with focal mechanisms and near-surface information. Fault complexity likely causes a low level of waveform similarity in this aftershock sequence and thus generates fewer correlated events compared to the Chile earthquakes. Differences between the double-difference locations and corresponding locations in global seismicity catalogs (Earthquake Data Report, EDR; International Seismological Centre, ISC; Engdahl-Hilst-Buland, EHB) are typically greater than 10 km. We evaluate the potential of cross-correlation and double-difference methods to improve hypocenter locations on a global scale.
239OOTimes Cited:0Cited References Count:38