Continuous Monitoring of Seismic Energy Release Associated with the 1994 Northridge Earthquake and the 1992 Landers Earthquake

We have developed a method to detect long-period precursors for large earthquakes observed in southern California, if they occur. The method allows us to continuously monitor seismic energy radiation over a wide frequency band to investigate slow deformation in the crust (e.g., slow earthquakes), especially before large earthquakes. We used the long-period records (1 sample/sec) from TERRAscope, a broadband seismic network in southern California. The method consists of dividing the record into a series of overlapping 30-min-long windows, computing the spectra over a frequency band of 0.00055 to 0.1 Hz, and plotting them in the form of a timefrequency diagram called spectrogram. This procedure is repeated daily over a daylong record. We have analyzed the 17 January 1994 Northridge earthquake (Mw = 6.7), and the 28 June 1992 Landers earthquake (Mw -7.3). No slow precursor with spectral amplitude measured over a duration of 30 min larger than that of a magnitude 3.7 was detected prior to either event. In other words, there was no precursor whose moment was larger than -0.003% of the mainshock. Introduction The initiation of an earthquake is generally considered abrupt. However, some precursory slow slip has been reported for the 1960 Chilean earthquake (Kanamori and Cipar, 1974; Kanamori and Anderson, 1975; Cifuentes and Silver, 1989), the 1944 Tonankai earthquake (Sato, 1970, 1977; Mogi, 1984), the 1983 Japan Sea earthquake (Linde et al., 1988), and the 1989 Macquarie Ridge earthquake (Ihmle et al., 1993). In particular, the leveling data before the Tonankai earthquake suggest a precursory tilt as large as 30% of that of the mainshock for a few hours prior to it. Unfortunately, data are very limited for both the Chilean earthquake and the Tonankai earthquake. Also, Kedar et al. (1994) saw no direct evidence for the precursor for the Macquarie Ridge earthquake. Hence, whether or not slow precursory deformations occur under certain circumstances is not presently resolved. In contrast, several recent studies (mainly of California earthquakes) demonstrated that if a slow precursory slip occurs, it is probably less than 1% of the mainshock in terms of seismic moment (Johnston et al., 1994, 1990, 1987; Agnew and Wyatt, 1989; Linde and Johnston, 1989). Also, modeling studies using velocity-weakening constitutive relations predict that such precursory changes on time scales of minutes are less than 1% of the mainshock moment (Lorenzetti and Tullis, 1989). An example of slow deformation detection is that of Beroza and Jordan (1990) who used spectral amplitudes of the Earth's normal modes to identify slow earthquakes otherwise undetected. We have developed a method for continuously monitoring the long-period (1 sample/sec) record from TERRAscope, the Caltech/USGS Southern California Broadband seismic network. We report here the results from two recent large earthquakes in southern California: the 1994 Northridge earthquake and the 1992 Landers earthquake. Method The long-period part of the spectrum was analyzed by means of a frequency-time diagram or "spectrogram." The spectrograms were constructed and fine-tuned particularly for analysis of TERRAscope data over a frequency band 0.00055 to 0.1 Hz. The computation is done by taking the frequency spectrum of a 30-min-long time window (hence the low-frequency end 1/1800 sec = 0.00055 Hz), advancing the window by 10 min, and repeating the computation. The cutoff amplitude for the spectrogram was determined for each station to emphasize signals that are slightly above noise level. This analysis is performed automatically every day for the Pasadena station (PAS). 17 January 1994 Northridge Earthquake Figure 1 shows the spectrogram for the mainshock of the 1994 Northridge earthquake and its aftershock sequence