Steven John Gibbons 

The seismic arrays of the International Monitoring System (IMS) for the Comprehensive Nuclear-Test-Ban Treaty (CTBT) are highly diverse in size and configuration, with apertures ranging from under 1 km to over 60 km. Large and medium aperture arrays with large inter-site spacings complicate the detection and estimation of high-frequency phases lacking coherence between sensors. Pipeline detection algorithms often miss such phases, since they only consider frequencies low enough to allow coherent array processing, and phases that are detected are often attributed qualitatively incorrect backazimuth and slowness estimates. This can result in missed events, due to either a lack of contributing phases or by corruption of event hypotheses by spurious detections. It has been demonstrated previously that continuous spectral estimation can both detect and estimate phases on the largest aperture arrays, with arrivals identified as local maxima on beams of transformed spectrograms. The estimation procedure in effect measures group velocity rather than phase velocity, as is the case for classical f–k analysis, and the ability to estimate slowness vectors requires sufficiently large inter-sensor distances to resolve time-delays between pulses with a period of the order 4–5 s. Spectrogram beampacking works well on five IMS arrays with apertures over 20 km (NOA, AKASG, YKA, WRA, and KURK) without additional post-processing. Seven arrays with 10–20 km aperture (MJAR, ESDC, ILAR, KSRS, CMAR, ASAR, and EKA) can provide robust parameter estimates subject to a smoothing of the resulting slowness grids, most effectively achieved by convolving the measured slowness grids with the array response function for a 4 or 5 s period signal. Even for medium aperture arrays which can provide high-quality coherent slowness estimates, a complementary spectrogram beampacking procedure could act as a quality control by providing non-aliased estimates when the coherent slowness grids display significant sidelobes. The detection part of the algorithm is applicable to all IMS arrays, with spectrogram-based processing offering a potential reduction in the false alarm rate for high-frequency signals. Significantly, the local maxima of the scalar functions derived from the transformed spectrogram beams are robust estimates of the signal onset time. High-frequency energy is of greater importance for lower event magnitudes and in the cavity decoupling detection evasion scenario. There is a need to characterize both propagation paths with low attenuation of high-frequency energy and situations in which parameter estimation on array stations fails.