Large Teleseismic P Wavefront Deflections Observed with Broadband Arrays

We measure the plane wavefront incidence azimuth for teleseismic P at large-aperture ( 50 km) broadband arrays. The incidence azimuth is determined by crosscorrelation of the P arrivals on the vertical component seismograms filtered in successive frequency bands. The periods considered range from 10 to 35 sec. At the Anza array in southern California, the plane wave direction is deflected from the great circle azimuth of the event by up to 20 . In addition, we find a surprisingly strong frequency dependence of the same magnitude and a striking antisymmetric pattern of the deflection as a function of backazimuth, whereas the curvature of the wavefront is small. Similar characteristics are found at the Gräfenberg array in Germany and the NORSAR array in Norway, however, with much weaker amplitudes of 5 . We ascribe the behavior at Anza to structure in the lower crust and uppermost mantle beneath the array, given that the observations are only a function of source backazimuth and not of source depth and source mechanism, that the wavelengths under consideration range from 50 to 270 km, and that the sign of the deviation is opposite to that predicted from shallow crustal structure and Moho topography. We are able to reproduce the magnitude and frequency dependence of the wavefront deflection using finite difference numerical modeling of plane wave propagation through simple 2D structures. Introduction Slowness and azimuthal anomalies of teleseismic P phases measured at seismic arrays have been used to infer mantle and crustal structure. An early controversy was related to whether the anomalies were caused by near-receiver structure in the upper mantle and crust, or in the lower mantle; results are summarized in more recent studies such as Steck and Prothero (1993), Lin and Roecker (1996), and Tibuleac and Herrin (1999). Most of these studies were conducted in the short-period ( 1 Hz) band. In addition, most recent applications of azimuth determination of seismic phases using arrays are related to location, detection, and discrimination problems, and are therefore more concerned with robustness in the presence of noise and multiple signals (e.g., Harris [1990], Chiou and Bolt [1993], Wagner and Owens [1996]). Once empirical corrections for systematic anomalies or so-called statics are found, they are used to calibrate individual records from the array, but for the most part, the underlying cause is left uninvestigated. In this study, we use very high signal-to-noise teleseismic P arrivals at several broadband, large-aperture arrays to study earth *Present address: CIRES and Dept. Geological Sciences, University of Colorado at Boulder, UCB 399, Boulder, Colorado, 80309-0399, [email protected]. structure. One novel aspect is the resolution of frequency dependence of azimuthal anomalies within the long-period (T 10 sec) band. Data The main focus of this study is the Anza array in southern California. In the configuration used in this study (Fig. 1), the array has nine three-component stations with Streckeisen STS-2 broadband sensors distributed irregularly over an aperture of 50 km. For part of the time under consideration in this study, four additional stations were in operation. All sensors are located on bedrock of the Peninsular Ranges batholith. Despite the large variation in topography over the area, the altitude of the stations used here varies by only 100 m. The array is bisected by the San Jacinto fault, a major branch of the San Andreas fault system. For comparison with the results from Anza, we also analyzed data from the Gräfenberg broadband array (GRF) (Fig. 2) in southeast Germany. The array consists of 13 stations with STS-1 sensors over an aperture of roughly 50 by 100 km and is located on Jurassic limestone in an area with only slight tectonic activity. A third dataset comes from the NORSAR array (Fig. 3) in southern Norway, with six broad748 V. Schulte-Pelkum, F. L. Vernon, and J. Eakins