S U M M A R YWe systematically analyse temporal changes of fault zone (FZ) site response along the Karadere-Düzce branch of the North Anatolian fault that ruptured during the 1999İzmit and Düzce earthquake sequences. The study is based primarily on spectral ratios of strong motion seismic data recorded by a FZ station and a station ∼400 m away from the fault and augmented by analysis of weak motion records. The observations are used to track non-linear behaviour and temporal changes of the FZ site response. The peak spectral ratio increases 80-150 per cent and the peak frequency drops 20-40 per cent at the time of the Düzce main shock. These co-main shock changes are followed by a logarithmic recovery over an apparent timescale of ∼1 d. However, analysis of temporal changes at each individual station using weak motion waveforms generated by repeating earthquakes show lower-amplitude longer-duration logarithmic recoveries that are not detected by the spectral ratio analysis. The results are consistent with a reduction of S-wave velocities in the top 100-300 m during the Düzce main shock of 20-50 per cent or more and logarithmic post-main shock recovery on a timescale of 3 months or more. The observations support previous suggestions that non-linear wave propagation effects and temporal changes of seismic properties are generated in the shallow material by strong ground motion of nearby major earthquakes.

S U M M A R YWe perform a systematic survey of triggered deep 'non-volcanic' tremor beneath the Central Range (CR) in Taiwan for 45 teleseismic earthquakes from 1998 to 2009 with M w ≥ 7.5 and epicentral distance ≥1000 km to the broad-band station TPUB. Triggered tremors are visually identified as bursts of high-frequency (2-8 Hz), non-impulsive and long-duration seismic energy that are coherent among many seismic stations and modulated by the teleseismic surface waves. Out of the 45 earthquakes, we identified nine teleseismic events associated with nine tremor sources in the southern and five in the northern CR. Most of the tremor sources are located within the depth range of 15-25 km in the lower crust above the Moho. We find that the amplitudes of the surface waves play an important role in determining the triggering potential, and the apparent triggering threshold is ∼0.1 cm s −1 , or 7-8 KPa. However, such threshold is partially controlled by the background noise level, which could prevent weaker tremor triggered by surface waves with smaller amplitudes from being identified. The amplitudes of the triggered tremor show a positive correlation with the amplitudes of the triggering surface waves, consistent with the predictions by the 'clock-advance' model. In addition to amplitudes, other factors, such as frequency contents and incidence angles, also affect the triggering potential. We find that intermediate-period (30-10 s) surface waves could trigger/modulate tremors, suggesting that long-period (>30 s) surface waves are not always required in long-range triggering. Tremors appear to be triggered by both Love and Rayleigh waves. When the incidence angles are parallel to the strike of the CR, all six events triggered tremor primarily during the Rayleigh waves. For strike normal incidence, only the 2001 M w 7.8 Kunlun earthquake showed predominant Love-wave triggering. This observation can be qualitatively explained by a simple Coulomb failure for a left-lateral shear on the low-angle detachment fault beneath the southern CR.

We use noise correlation and surface wave inversion to measure the S wave velocity changes at different depths near Parkfield, California, after the 2003 San Simeon and 2004 Parkfield earthquakes. We process continuous seismic recordings from 13 stations to obtain the noise cross‐correlation functions and measure the Rayleigh wave phase velocity changes over six frequency bands. We then invert the Rayleigh wave phase velocity changes using a series of sensitivity kernels to obtain the S wave velocity changes at different depths. Our results indicate that the S wave velocity decreases caused by the San Simeon earthquake are relatively small (~0.02%) and access depths of at least 2.3 km. The S wave velocity decreases caused by the Parkfield earthquake are larger (~0.2%), and access depths of at least 1.2 km. Our observations can be best explained by material damage and healing resulting mainly from the dynamic stress perturbations of the two large earthquakes.

We analyzed temporal changes in site response associated with the strong ground motion of the 2004 M w 6.6 Mid-Niigata earthquake sequence in Japan. The seismic data were recorded at a site with accelerometers at the surface and a 100-m-deep borehole. We computed the empirical surface-to-borehole spectral ratios and used them to track temporal changes in the top 100 m of the crust. We observed that the peak spectral ratio decreases by 40%-60% and the peak frequency drops by 30%-70% immediately after large earthquakes. The coseismic changes are followed by apparent recoveries, with the time scale ranging from several tens to more than 100 sec. The coseismic peak frequency drop, peak spectral ratio drop, and the postseismic recovery time roughly scale with the input ground motions when the peak ground velocity is larger than ∼5 cm=sec (or the peak ground acceleration is larger than ∼100 Gal). Our results suggest that at a given site the input ground motion plays an important role in controlling both the coseismic change and the postseismic recovery in site response.

In summer 1995, in collaboration with the deep multi-channel seismic profiling project around the island of Taiwan, an onshore-offshore wide angle deep seismic profiling experiment was conducted in Taiwan. The re sults are expected to provide the first complete seismic images of the deep crustal structure for a better understanding of the Taiwan orogeny and subduction-collision system. The experiment consists of three profiles, one along each of the central and southern cross-island highways and another on the south-link highway of the island. For the first two lines, 35 three component portable seismographs were deployed along each of the 116 and 135 km-long profiles onshore, with airgun shots being fired at distances