Three deep Hindu Kush events were then analyzed to explore the utility of data
from Garm for locating regional earthquakes and determining source parameters.
Since the previous receiver function study indicated complex crustal structure
at Garm, attenuation caused by scattering and anelasticity was investigated
by a spectral ratio study. Attenuation is potentially important in controlling
the differential amplitudes of P and S waves which are needed in the source
mechanism study. The results, however, showed that waves from all events
traversed areas of high *Q*_{s} (which is the frequency independent S wave
quality factor) indicating that attenuation has nearly no effect
on the amplitudes. A following sensitivity study using synthetic Green's
functions showed the same result.

Observed direct P, S and SsPmp phases and times were then used to infer backazimuth, depth and distance, locating the events. The SsPmp phase appeared to be the major secondary phase that could be identified in the regional seismograms. "S" represents the direct S wave traveling from the source. The small s is the S wave after transmission at the moho. After the conversion at the free surface, the wave travels as P down to the moho where it is reflected. The small p is the upgoing P wave from the moho. Possible focal mechanisms were obtained by a grid search program that uses arrival amplitudes of synthetic Green's functions as an input. The Green's functions were generated by a wavenumber integration method developed by Barker 1984 using the determined depth and distance of each event. By integration over horizontal wavenumber or slowness, a complete seismogram is obtained. To increase the accuracy of the source parameter estimates, synthetic seismograms were computed and the waveforms were compared with the observed data. The synthetics were generated by combining the obtained Green's functions with the focal mechanisms. Using time versus depth and time versus distance synthetic moveouts, the change of the P, S and SsPmp phases in shape and timing to each other could be determined. In this way, constraints could be put on the source parameter estimates. Figure shows the observed data of the three Hindu Kush events and the synthetics generated by the wavenumber integration method.

The first event is the event of the 16th of April 1991. Three clear distinct phases, namely direct P, direct S and SsPmp, could be identified in the data. Based on this information the event could be located at a backazimuth of , a depth of 110 km and a distance of 320 km. The uncertainties were only km in depth and km in distance. The PDE (Preliminary Determination of Epicenters) location of this event gives a backazimuth of , a depth of 127 km and a distance of 296 km. Considering the fact that PDE locations can have an error of up to km, the respective location lies within the probable error of the PDE location. One possible focal mechanism was obtained for a determined depth of 110 km and a distance of 320 km. It had a strike of , a dip of and a rake of . Its T-axis was plunging to the north. The variations in this mechanism are small, being only .Due to the three different phases used in the grid search, this event is reasonably well constrained.

The second event is the event of the 4th of June 1991. It was located less accurately than the first event due to the fact that the SsPmp phase interfered with the direct S phase so that clear identification of its arrival time was not possible. This increased the uncertainties in the results. The event was located at a depth of 208 km which could vary from 190 km to 225 km. The resulting distance was 283 km and could vary from 275 km to 288 km. The backazimuth was determined to . The original PDE location gives a backazimuth of , a depth of 160 km and a distance of 283 km. Still, both locations are in good agreement. The obtained focal mechanism had a strike of , a dip of and a rake of . The deviations in the mechanism are slightly higher than those of the first event.

The third event occurred on the 11th of February 1991. Its analysis showed that the coda level is very large, dominating over small amplitude phases. This large coda level is caused by the lateral heterogeneities around Garm. Due to the fact that the coda is masking low amplitude phases, only first P and S could be identified in the data. This makes an accurate location estimate impossible. Based on only the S-P travel time, the event was located at a depth of 180 km or deeper, a distance of 275 km or less and a backazimuth of . Considering these limits, the source parameters could still agree with the PDE depth of 217 km, distance of 256 km and backazimuth of .Many different focal mechanisms could be obtained for the determined range in depth and distance. They varied essentially in the T-axes, but the P-axes were in fairly good agreement with each other plunging approximately to the south. These large variations are based on the very limited phase information in this event.

Figure 1

11/17/1997