Broadband Time Domain Modeling of Earthquakes from Friuli , Italy

Short-period (SP) and long-period (LP) seismograms written by the main shock and two principal aftershocks of the 1976 Friuli, Italy, earthquake sequence are modeled in the t ime domain using synthetic seismograms. The main shock occurred on 6 May 1976 (20h OOm, Ms = 6.5) and both aftershocks on 15 September 1976 (03h 15m, Ms -6.0 and 09h 21m, Ms = 5 .9 ) , Source models were determined init ial ly by trial and error and then refined using a waveform inversion program. Two point sources of radiation are required to adequately model the aftershock short-period records. For the 09h 21m aftershock, the model derived from short-period records also produces good fits to the longperiod data. The seismic moment of this earthquake is found to be 0.8 to 1.0 x 1025 dyne-cm. The SP model for the 03h 15m aftershock, on the other hand, predicts long-period synthetics which do not agree with the observations. In particular, the SP moment ( 0 . 3 7 x 102s dyne-cm) is about 21⁄2 times smaller than the LP moment (1 x 102s dyne-cm). Adding a long-period component to the SP model considerably improves LP waveform and moment agreement. In the case of the main shock, a reasonable f i t to the observed SP data is obtained using three point sources of radiation. However, LP synthetics computed using this model do not agree with the observations, and the SP moment (0.65 x 102s dyne-cm) is a small fraction of the LP moment (3 to 5 x 102s dyne-cm). Time funct ion durations indicate that the individual events inferred from the SP records are radiated from patches of the fault having radii of 2 to 4 km and stress drops in the range 35 to 276 bars. In comparison, stress drops estimated from LP data are found to be 12 bars (main shock) and 24 bars (09h 21m aftershock). These observations suggest that the short-period instruments are sensitive to the high-frequency radiation emitted from small, high-stress drop areas on the fault plane whereas the long-period instruments record the overall motion during the earthquake. INTRODUCTION Recent years have witnessed considerable advances in understanding earthquake source processes by detailed investigation of body waves. Most studies done to date have concentrated on P and S waves recorded by long-period seismographs. A partial list of contributors includes Wyss and Hanks (1972), Burdick and Mellman (1976), Langston and Butler (1976), Langston (1976, 1978), Ebel et al. (1978), and Cipar (1980). Less attention has been paid to teleseismic short-period records. This is due, no doubt, to the rather complex appearance of these seismograms resulting from the sensitivity of short-period radiation to the vagaries of the seismic source and propagation medium. Burdick and Mellman (1976) used a long-period-short-period deconvolution technique to obtain a more detailed time function for the Borrego Mountain, California, earthquake. Somerville et al. (1976) used the Savage (1966) source model and an inversion scheme to obtain fault parameters for two shortduration earthquakes. Langston (1976) found that crustal effects were important in time domain modeling of the short-period records of the Koyna, India, earthquake. 1215