Simultaneous use of multiple seismic arrays

Abstract

Seismic arrays provide an important means of enhancing seismic signals and determining the directional properties of the wavefield by beamforming. When multiple arrays are to be used together, the viewpoint needs to be modified from looking outwards from each array to focusing on a specific target area and so constraining the portions of the waveforms to be analysed. Beamforming for each array is supplemented by the relative time constraints for propagation from the target to each array to provide tight spatial control. Simultaneous multiple array analysis provides a powerful tool for source characterization, and for structural analysis of scatterers as virtual sources. The multiple array concept allows us to illuminate a specific point in the Earth from many different directions and thus maps detailed patterns of heterogeneity in the Earth. Furthermore, illumination of the structure from multiple directions using data from the same event minimizes source effects to provide clearer images of heterogeneity. The analysis is based on a similar concept to the backprojection technique, where a part of the seismic wave train is mapped to a specific point in space by ray tracing. In contrast to classic backprojection where the incoming energy is mapped onto a horizontal plane with limited vertical resolution, the multiarray method controls depth response by combining relative time constraints between the arrays and conventional beamforming. We illustrate this approach with application to two earthquakes at moderate depth. The results show that the use of simultaneous multiple arrays can provide improvement both in signal quality and resolution, with the additional benefit of being able to accurately locate the source of the incoming energy and map large areas with only a limited number of such arrays. Spatial analysis, Earthquake source observations, Body waves 1 INTRODUCTION

Seismic arrays play an important role in the detection and location of seismic events. The number of arrays of various types has been augmented in recent years by the new arrays in the primary seismic network established by the Preparatory Commission for the Comprehensive Nuclear-Test-Ban Treaty (CTBT) with the aim of securing more uniform detection capabilities across the globe. Other new arrays, such as Pilbara Seismic Array (PSAR) in northwestern Australia form part of enhanced tsunami warning systems.

The merit of seismic arrays is that when the signals at the various sensors are suitably phased, coherent signal is enhanced relative to incoherent background noise (e.g. Rost & Thomas 2002); the theoretical gain factor in the stacked energy for N sensors is N. For a single seismic array, the natural mode of operation is to look outwards from the array with monitoring of the full range of slowness and azimuths seeking to detect new events. Detections from multiple arrays can then be brought together to refine the location of the event, along with information from other stations.

In this paper, we examine the way in which multiple seismic arrays can be used simultaneously to look at seismic sources or the location of seismic scattering, which can be regarded as virtual sources. Such simultaneous use of multiple arrays requires concentration on those portions of the seismograms related to the same potential event, with appropriate time delays to compensate for the different epicentral distances to the arrays. Thus rather than looking outwards from a single array and sweeping through the full range of distances and azimuths via beamforming, multiarray analysis is automatically source oriented.

This difference in approach has been recognized implicitly in the use of multiple arrays to track the spatial evolution of great earthquakes. Thus Roessler et al. (2010) have exploited the coherence properties of seismograms recorded at networks of stations, treated as very large arrays, as a function of apparent source position in the neighbourhood of the hypocentre of large earthquakes. Rather than summing the semblance responses for the different large arrays they have multiplied them, which gives higher spatial resolution. Many backprojection studies for great earthquakes have employed more than one suite of stations as an array, but commonly have just compared the results from the different receiver locations (e.g. Simons et al. 2011; Ye et al. 2013). Kiser & Ishii (2012) have demonstrated the improvement that can be made when two arrays at different distances and azimuths are used together through simulations of the combined response of the central portion of the US array and the Hi-net network in Japan.

We here show how the combination of the slowness and time constraints from multiple seismic arrays can be used to achieve good spatial resolution in 3-D, when there is sufficient azimuthal separation between the arrays employed. We illustrate the simultaneous analysis by using the set of arrays within Australia to examine events in the subduction zones to the north, and also show the improvement achieved when additional azimuths are brought to bear by using North American arrays.