Two relatively novel wavefield separation techniques are examined. Spectral matrix filtering (SMF) has received minimal coverage in the geophysical literature. Polarisation filtering (PF) has been more widely used, although application in exploration seismology has been limited by the general lack of multi-component recording.
SMF wavefield separation requires smoothing of the cross-spectral matrix constructed from a 2D seismic image, followed by decomposition into a number of eigenimages. This concept is introduced in the context of single-component seismic wavefield separation. Testing on synthetic vertical seismic profile (VSP) images demonstrates practical issues relating to various options of frequency and/or distance smoothing. Combined frequency/distance smoothing has potential application for single-pass separation of upgoing and downgoing VSP wavefields. This is demonstrated using real VSP data. SMF provides a useful alternative to more common VSP wavefield separation methods, although it is computationally expensive.
Application of SMF as a true multi-component wavefield separation tool is also demonstrated. In this context the desired wavefield is estimated within a time window where it is best represented. The wavefield of interest is then extracted from any other window, or component, by projection. This process is applied to the problem of separation of P and SV wavefields in VSP data, with example applications to noisy synthetics, as well as real, multi-component VSP data.
Two PF algorithms are outlined - the Principal Component method, and the Generalised Least Squares approach. Results from these two algorithms are shown to be similar. Two broad categories of polarisation filter are considered. The 'pass plane' or rotation filter is regarded as more flexible than the 'classic' or projection filter.
The first real-data application of PF concentrates on shear wave splitting (SWS) analysis of multi-component VSP data. The polarisation method proves competitive with other efficient SWS algorithms. The second application of the PF method is to the problem of phase identification in teleseismic and region earthquake recordings. PF assists in the identification of crustal conversion and reverberations from Fijian and Japanese compressional wave arrivals at the Charters Towers seismograph station, providing constraints on gross crustal thickness and velocities. In addition, PF is used to assist in the identification of crustal refraction phases from regional earthquakes recorded on microearthquake systems in North Queensland.