Arc2-complete analysis

The analysis above has shown that much of the energy recorded at a 3-component receiver may propagate along a different course to the great circle azimuth defined by the source- receiver path. The distinction of whether this energy is scattered energy generated within the vicinity of the receiver, or whether it reflects deeper structure, can only be obtained by comparing the results found at different 3-component receivers. The arrivals which are

93

coming in on-azimuth are also tested to see whether they correlate over all 3-component stations. Strong correlation between stations will indicate that the arrivals reflect upper mantle structure, while partially coherent events may indicate local structure or suggest that arrivals on certain stations have been modified significantly by signal generated noise such that they are now unidentifiable.

A comparison is made between all four 3-component stations in the l-2Hz frequency band, and representative diagrams of 2 stations : AO & C2, displaying the waveform parameters are given in figs.5.7a,b&5.9a,b. This frequency band is chosen as on-azimuth P and Rayleigh waves are present in the coda along with phases which have propagated from different directions. The seismograms for all 3 orthogonal components correlate very well along the entire coda and distinct P, S and Rayleigh wave phases are clearly visible on the seismograms.

The values of the apparent azimuth are essentially the same in the regions where the azimuth is stable, otherwise they fluctuate with different values, which is clearly shown by the values in the P-section of the codas. The only common feature of apparent angle of incidence values is that they are low in the Rayleigh wave section of the seismogram.

In the P wave section of the coda highly rectilinear signals for each channel are in phase, but only some of these correlate between the stations. The initial, dominant P phase is observed on all channels along with several other P arrivals, however many of the P phases identified correlate only over several stations, while the other wave groups identified are distinct to respective stations. The correlation of elliptical type waves (high planarity and low rectilinearity) with out of phase motion is high between the 3-component stations in the Rayleigh wave section of the coda, which results in Rayleigh wave classification.

Utilisation of the azimuthal based parameters points out that most of the motion that correlates across the stations is on-azimuth, and therefore suggests that off-azimuthal energy is distinct to individual channels, reflecting local structure within the vicinity of the receiver.

With an increase in frequency the correlation of the seismograms, parameters and wave classification deteriorates. This is expected as the recorded coda is complex, mainly due to the numerous occurrence of interfering arrivals, and the recording of scattered energy. The

100953 41232 Amplitud0(normalisedj Apparent azimuth 360.0 _ Xxx«xx)<xx,0< « X X x * X5S a c ^ W » * XXX\ ; X

Apparent angle of incidence 90.0 _

ä PQXY

0.67 _ _raoococ*

seconds

Stability Rectilinearity Planarity Wavegroup Indet ell: P/S ell: SV/Ray rect: SH rect: P/SV = RZ J f - - ellipticaL/SH Vertical/horizontal S x x x t x * * ^ * * ^ ^ « * ^ * * * » ^ ^ S x x x x * **” S Yy ** ^ * X\ x x x*xxx**** Wavetype(on-azimuth) Indet ell: Ray ell: S V(>i) ell: S ell: S V(<i) ell: P rect: SH rect: SV(<i) rect: P Seconds

only waves that correlate above 2Hz is the initial P phase (over a broad frequency spectrum) and the SH phase in the 2-4Hz frequency band. One may think that less waves are determined and consequently there is a lower correlation of the waves between stations because the time analysis window employed here is too large. It was mentioned in chapter 4 that an increase in resolution is obtained with a reduction in window length but the stability of the parameters decreases. This behaviour is seen with this event when the data is analysed in 0.5 second intervals (instead of 1 second). More phases are determined but the extra identifications are unstable as they do not correlate across the channels. This increase in resolution is in fact only giving more information on noise local to the receiver which is generally not wanted.

This event again reconfirms that as the complexity of the coda increases, there is a rapid decrease in the number of phases identified. Also some phases are only identified over limited frequency ranges, this is clearly evident with the SH phase and Rayleigh waves. Comparing results from individual stations shows that when the coda is complex much of the coda identified is distinct to the stations, reflecting local scattered energy, and much of the well correlated energy is on-azimuth. Partially coherent energy reflecting either a local or upper mantle origin is also present at lower frequencies. The amount of coda identified decreases rapidly with frequency, probably due to the increased effect of scattered arrivals and this unidentified energy is either upper mantle and crustal phases modified by local noise, interfering arrivals or signal generated noise originating within the vicinity of the receiver.

5.1.3 Conclusion

These two events have shown that as the complexity of the coda increases, the wavetype criteria is satisfied less often. The sources of complexity can be divided into 3 distinct areas : (i) Signal generated noise : the major part of this noise is scattered energy generated within the vicinity of the receiver. In both events as the frequency increased the noise had a more dominant role, to the extent where wavetype classification was hardly feasible. Also much of the coda identified at higher frequencies was distinct to its 3-component station, reflecting

95 scattered energy with a certain wavetype. Key (1967) found that P to Rayleigh wave scattering can play a major role in the later part of the P coda, and Hendrajaya (1981) has noticed the perturbation of the initial S phase by P wave polarised arrivals generated by topography at the earth's surface.

(ii) Interference of waves : as the source-receiver distance is shortened the time interval in which the phases are recorded decreases, consequently the occurrence of interfering phases (neglecting scattered energy) becomes more numerous. This is clearly reflected by the number of phases identified in the P coda section of the two events. Many more P waves are identified for the regional event Norl compared with the local event Arc2. The time relationship of the waves in the coda allows for interference of phases of different type, especially after the P coda.

(iii) Wavetype recorded : P and Rayleigh wave phases have simple waveforms, whereas the S wave which consists of different amounts of SV and SH energy is complex. Also above the critical angle the SV component is elliptical, consequently the S wave (containing SV and SH components) is 3-dimensional and can not be distinguished from noise or other 3- dimensional interference.

Since signal estimates vary strongly with frequency, the use of a wide frequency band would result in the frequency component at the peak amplitude for a given time window dominating the polarisation estimate. This indicates that if no frequency filtering was applied to the data many of the waves actually classified would not be identified. By filtering the data a wider range of signals can be analysed separately and this also allows for a reduction in the coda's complexity. Also as the phases have band-limited frequency spectra, their signal-to- noise ratios are improved by employing narrower frequency bands.

The actual wavetype can not be obtained directly without knowledge of the signal's azimuth. The assumption that the signal is arriving in a direction which coincides with the source-receiver azimuth is not strictly true as was shown by the presence of off-azimuth energy which can correlate over several 3-component stations. By summing up the data from a 3-component array (see section 5.2) one may be able to obtain a clearer picture of the

crustal and upper mantle phases recorded in the coda, as the scattered energy will be reduced since it does not correlate over the stations.