Comparison of source time functions retrieved in the frequency and time domains for seismic events in a copper mine in Poland
The extraction of source time functions based on empirical Green's functions is especially convenient for source studies of seismic events in mines, where underground seismic networks are situated in the source area and are often composed of a large number of sensors. The empirical Green's function deconvolution techniques were applied to retrieve the source time functions from the records of P waves of several seismic events that occurred in 1998 at Rudna copper mine in Poland. The selected events were located within the underground network composed of 32 vertical sensors situated at a depth of about 1 km; the size of the network is about 10 by 10 km. Their moment magnitude ranged from 2.7 to 3.2. The records of smaller events from the same area and with similar source mechanism, with moment magnitude from 2.2 to 2.4, were accepted as empirical Green's functions. The results of classic deconvolution approach in the frequency domain are reported elsewhere. The results of the application of projected Landweber deconvolution in the time domain to the same set of data are described here. The projected Landweber approach is an iterative deconvolution technique, allowing introduction of physical constrains on the final source time function. This technique successfully overcomes the instability effects of the deconvolution process inherent in the frequency domain and provides stable and reliable relative source time functions retrieved at several stations. Although the results obtained in the frequency and time domains were found to be similar, the time domain approach provides more objective determination of the source time function duration, essential for a proper determination of the source dimension. The relative source time functions retrieved by both methods display directivity effects in several cases, implying unilateral rupture propagation. The rupture propagation direction and rupture velocity were estimated from the distribution of pulse widths and pulse maximum amplitudes as a function of the station azimuths.
Bibliogr. 25 poz.