Improvements to Seismic Monitoring of the European Arctic Using Three-Component Array Processing at SPITS Steven J. Gibbons Johannes Schweitzer Frode Ringdal Tormod Kværna Svein Mykkeltveit
Seismicity at regional distances. High frequency processing essential. The SPITS array is an auxiliary IMS seismic station (AS72) a nine element array with an aperture of approximately 1 km (among the smallest arrays in the IMS). Its small size is partly due to logistics, but most significantly to ensure that the array can detect and characterize phases from seismic events at regional distances. There is significant seismicity at regional distances from SPITS! All such signals need to be detected, located and classified. Ocean-generated microseisms dominate at frequencies below 2 Hz. This means that, for low magnitude events, we need coherent signals at high frequencies.
Array configuration optimal for detection and classification of HF regional seismic signals. Configuration of SPITS drawn to scale with the (also relatively small aperture) ARCES array (PS28). Both arrays have the classic design with sensors in concentric rings of log-periodic radii. This design provides both optimal azimuthal symmetry and an optimal range of intersite distances. The array response functions (for signals with dominant frequency 4 Hz) are displayed below. The array configurations provide minimal sidelobes. The smaller aperture of SPITS provides a poorer theoretical slowness resolution than ARCES. Prior to 2004, SPITS had only a single 3-component seismometer (SPB4).
Secondary phases are essential for regional events with few observing stations. The level of seismicity in the eastern Barents Sea (including Novaya Zemlya) is low. The figure to the right displays all seismic events from the eastern Barents Sea detected between 1992 and 2004. Secondary phase detection is crucial for phase association given the small number of observing stations and large azimuthal gap. The high amplitude Lg phases often seen on continental propagation paths are absent. P- and S- beams on the vertical sensors of the array are displayed for each event. S-phase arrivals are frequently poor: emergent signals low SNR
Secondary phase amplitudes usually larger on horizontal traces than on vertical traces. It is clear on the signals from the Kara Sea event of August 16, 1997, that the Sn arrival has a far greater amplitude and SNR on the single 3-component sensor horizontal component (transverse rotation) than on the Sn beam of vertical sensors. It is difficult to incorporate a single 3-C sensor into the detection pipeline for secondary phases. (for instance, it is difficult to measure the slowness and backazimuth) A 3-component (sub)array would be able to improve the SNR for secondary phases through beamforming. allow direction estimates from the horizontal components using f-k analysis.
SPITS upgrade in 2004. HF energy from events at far-regional distances. Improved S phases on 3C beams. The refurbished array was installed in August 2004. The main features of the upgrade were: sampling rate from 40 Hz 80 Hz short period to broadband sensors 3-C sensors at 6 out of 9 sites. On March 5, 2006, a small seismic event close to the north coast of Novaya Zemlya was detected. We see: significant energy for both P- and S- phases well above 20 Hz: the Nyquist frequency of the old array (at a distance of over 1100 km) far greater SNR for Sn phase on horizontal beam a particular improvement to the SNR at frequencies close to 2 Hz
SNR for S-phases always higher on transverse beams than vertical beams. Better S arrival times. Since the upgrade in 2004, six seismic events have been detected in the eastern Barents Sea or on Novaya Zemlya. In each case a significantly greater SNR is measured for the Sn phase arrival on the transverse beam than on the vertical beam (despite the vertical beam containing more channels than the horizontal beam). The detection capability is improved given the significantly higher SNR. The Pn coda is barely above the background noise level for some events. Sn arrival times can be picked with less uncertainty on the horizontal beams.
Accurate phase identification essential for automatic association. Use apparent velocity estimates. While a phase detection and a robust arrival time pick are important; a backazimuth (indicating direction to the source) and correct phase identification are also necessary for the phase association algorithm. Normally, a regional arrival can be estimated from the apparent velocity estimate from an array. The figure to the right shows apparent velocity measurements for analystconfirmed phases at the ARCES array as a function of backazimuth. A slight sinusoidal variation is observed but, in general, an apparent velocity of 6.1 km/s will separate regional P phases from regional S phases.
Phase identification with app. vel. difficult at SPITS. P and S phases often overlap in parameter space. If we examine the apparent velocities of phases observed at SPITS a clear problem emerges. The velocities of regional P- and S- phases overlap. No single value of apparent velocity will separate them. The curves indicate that an azimuthdependent border could be defined which would work for most source regions. HOWEVER This breaks down in the region to the north-west of the array. Clear regional P and S phases approach the station with apparent velocities usually associated with teleseismic P-phases. Can 3-component processing help?
Quality of f-k analysis for this S phase far better on horizontal traces than on vertical. In this first example (an earthquake on Mohns Ridge) the first thing to notice is that the Sn arrival is almost invisible on the vertical component beam. On the horizontal beam, the S-arrival is weak but visible. Attempting to measure the slowness vector for the Sn arrival on the vertical component traces gives a qualitatively incorrect estimate: very little coherency between channels. Attempting to measure the slowness vector on the horizontal traces in exactly the same time window gives an apparent velocity consistent with an S-phase. The horizontal component channels are far more coherent.
Quality of f-k analysis for this S phase far better on horizontal traces than on vertical. In this second example to the north west of Spitsbergen, both P and S phases are observed with good SNR on the respective vertical beams. The SNR for the Sn arrival is improved by forming a beam on the horizontal traces. Performing f-k analysis at the time of the Sn arrival on the vertical component traces gives a qualitatively misleading estimate for an S phase. Performing f-k analysis in the same time window on the horizontal channels provides a P-type apparent velocity (albeit less than for the Pn phase). Most significant is the difference in coherence between vertical and horizontal channels.
Comparing the coherence between horizontal and vertical traces may be the most robust P/S discriminant. It appears to be a generic result that for regional P-phases, the coherence between the vertical channels is greater than the coherence between the horizontal channels... and that for regional S-phases, the coherence between the horizontal channels is greater than the coherence between the vertical channels. This property provides an almost perfect separation between P and S phases even for the situations where the apparent velocities obtained are somewhat different from those expected.
3-Component Array Processing - SUMMARY The ability to perform beamforming on rotated horizontal components on a small aperture regional array can significantly improve the signal-to-noise ratio for secondary phases. Reduce detection threshold Improve arrival time estimates Reduce the risk of non-detection of events The ability to perform f-k analysis for S-phases on rotated horizontal components can result in qualitatively correct slowness estimates and good event hypotheses in cases where this fails for the vertical-channels. Separation between regional P and S phases can be made far more reliable by comparing the coherence between horizontal traces with the coherence between vertical traces. We advocate strongly routine 3-C array processing for all small to medium aperture arrays with multiple 3-C sensors - and advocate the deployment of multiple 3-C sensors at all small to medium aperture arrays with only a single three component seismometer.