Locating seismicity on the Arctic plate boundary using multiple-event techniques and empirical signal processing

Abstract

The oceanic boundary separating the Eurasian and North American plates between 70 and 84 degrees North hosts large earthquakes which are well recorded teleseismically, and many more seismic events at far lower magnitudes that are well recorded only at regional distances. Existing seismic bulletins have considerable spread and bias resulting from limited station coverage and deficiencies in the velocity models applied. This is particularly acute for the lower magnitude events which may only be constrained by a small number of Pn and Sn arrivals. Over the past two decades there has been a significant improvement in the seismic network in the Arctic: a difficult region to instrument due to the harsh climate, a sparsity of accessible sites (particularly at significant distances from the sea), and the expense and difficult logistics of deploying and maintaining stations. New deployments and upgrades to stations on Greenland, Svalbard, Jan Mayen, Hopen, and Bjornoya have resulted in a sparse but stable regional seismic network which results in events down to magnitudes below 3 generating high quality Pn and Sn signals on multiple stations. A catalog of several hundred events in the region since 1998 has been generated using many new phase readings on stations on both sides of the spreading ridge in addition to teleseismic P phases. A Bayesian multiple event relocation has resulted in a significant reduction in the spread of hypocenter estimates for both large and small events. Whereas single event location algorithms minimize vectors of time residuals on an event-by-event basis, the Bayesloc program finds a joint probability distribution of origins, hypocenters, and corrections to travel time predictions for large numbers of events. The solutions obtained favour those event hypotheses resulting in time residuals which are most consistent over a given source region. The relocations have been performed with different 1-D velocity models applicable to the Arctic region and hypocenters obtained using Bayesloc have been shown to be relatively insensitive to the specified velocity structure in the crust and upper mantle, even for events only constrained by regional phases. The patterns of time residuals resulting from the multiple-event location procedure provide well-constrained time correction surfaces for single-event location estimates and are sufficiently stable to identify a number of picking errors and instrumental timing anomalies. This allows for subsequent quality control of the input data and further improvement in the location estimates. We use the relocated events to form narrowband empirical steering vectors for wavefronts arriving at the SPITS array on Svalbard for azimuth and apparent velocity estimation. We demonstrate that empirical matched field parameter estimation determined by source region is a viable supplement to planewave f-k analysis, mitigating bias and obviating the need for Slowness and Azimuth Station Corrections (SASCs). A database of reference events and phase arrivals is provided to facilitate further refinement of event locations and the construction of empirical signal detectors.