Abstract. We used four linear seismic arrays of portable seismometers at the northern AwajiIsland, Japan, to record fault zone trapped waves from aftershocks of the 1995 M7.2 Hyogoken Nanbu (Kobe) earthquake from April to June 1996. Three arrays were deployed across the Nojima fault, which ruptured during the mainshock, while one array was deployed across the Higashiura fault, which did not break recently. We observed significant fault zone trapped waves with relatively large amplitudes and the long wave train following S waves only when both the stations and aftershocks were located close to the Nojima fault. The coda-normalized spectral amplitudes of trapped waves show a maximum peak at 4-7 Hz, which decreases rapidly with distance from the fault trace. The normalized amplitudes of trapped waves also show a decrease with hypocentral distance along the fault, giving an apparent Q of approximately -25 at 4-7 Hz. In comparison, the array across the Higashiura fault recorded much shorter wave trains with higher frequencies after S arrivals for the same events. We simulate these trapped waves as S waves guided in a low-velocity waveguide sandwiched between high-velocity wall rocks. We find an adequate fit by using a waveguide 60 m wide at the northern site and 30-40 m wide elsewhere along the Nojima fault, a waveguide S velocity of 1.5-1.7 km/s, and a Q value of 25. For the Higashiura fault, the S velocity is 2.5 km/s, and the Q value is 80. The locations of aftershocks for which we observed fault zone trapped waves show that the Nojima waveguide is 9 km long and dips southeastward at 80%85 ø to a depth of ~ 16 km. It extends 6 km farther southwestward along the Asano fault, though there are no obvious surface breaks along it. However, the waveguide is disconnected from the Suma fault on the main island, which was also ruptured during the Kobe earthquake, possibly because of the existence of an offset between the Nojima fault and the Suma fault.

Microearthquake data from a downhole seismometer network on the San Andreas fault appear to outline two aseismic asperities that may correspond to the locations of the foreshocks and main shocks of the Parkfield characteristic earthquakes. The source parameters of the microearthquakes show that a few of the earthquakes have significantly higher stress drops than most. Furthermore, the magnitude-frequency statistics suggest that at local magnitude 0.6, the cumulative number of small events begins to fall off the usual Gutenberg-Richter (b = -1) relation, in which the number of events increases exponentially with decreasing magnitude. The downhole seismometer data establish a baseline from which the evolution of the earthquake process at Parkfield can be monitored and suggest that different mechanical conditions than those that lead to the typical Gutenberg-Richter relation may be operating for the smallest of Parkfield microearthquakes.

No abstract

Appendix 1. Abbreviated list of airgun shotpoint locations, FFID numbers, and shot times Appendix 2. Reftek and OBS station locations and elevations Appendix 3. List of timing used for Reftek stations 56 Appendix 4. List of stations having problems with spurious 12-second shifts Appendix 5. List of location of Reftek station data on archival tapes

We have developed a technique to measure seismic attenuation within an active fault‐zone at seismogenic depths. Utilizing a pair of stations and pairs of earthquakes, spectral ratios are performed to isolate attenuation produced by wave‐propagation within the fault‐zone. This empirical approach eliminates common source, propagation, instrument and near‐surface site effects. The technique was applied to a cluster of 19 earthquakes recorded by a pair of downhole instruments located within the San Andreas fault‐zone, at Parkfield California. Over the 1‐40 Hz bandwidth used in this analysis, amplitudes are found to decrease exponentially with frequency. Furthermore, the fault‐zone propagation distance correlates with the severity of attenuation. Assuming a constant Q attenuation operator, the S‐wave quality factor within the fault‐zone at a depth of 5‐6 kilometers is 31 (+7,−5). If fault‐zones are low‐Q environments, then near‐source attenuation of high‐frequency seismic waves may help to explain phenomenon such as fmax. Fault‐zone Q may prove to be a valuable indicator of the mechanical behavior and rheology of fault‐zones. Specific asperities can be monitored for precursory changes associated with the evolving stress‐field within the fault‐zone. The spatial and temporal resolution of the technique is fundamentally limited by the uncertainty in earthquake location and the interval time between earthquakes.