Research Papers
Synthesis of strong motion records using slip distributions from historical New Zealand earthquakes
RE Abercrombie, S Bannister, RA Benites, JJ Mori, A Pancha, TH Webb - IGNS
Slip inversion
We have investigated four aftershocks of the 1994 Arthur's Pass earthquake (Mw 6.7, South Island, New Zealand) to determine whether the major strike-slip faults in the region, or their unmapped conjugates, were activated during the aftershock sequence. The Arthur's Pass earthquake itself was a primarily reverse event, but the majority of the aftershocks, including the largest, were strike-slip. The earthquake occurred only 25km SE of the Alpine fault, the largest of the NE-SW trending predominantly strike-slip faults which dominate this region of oblique compression. We used the empirical Green's function method to obtain source time functions for four aftershocks (ML 4.1-5.1). We then inverted for slip on each nodal plane and determined which was the fault plane. Two of these earthquakes (ML 5.1 and ML 4.2), located close to the mapped trace of the Bruce fault, occurred on fault planes striking NNW-SSE, perpendicular to the predominant strike of the regional strike-slip faults. The former ruptured northwards and the latter to the south. The largest earthquake which could have occurred on the Bruce fault in this sequence cannot, therefore, have been larger than about ML 4. A third event (ML 4.1) was located on a lineation of aftershocks parallel to the regional mapped trend. The preferred fault plane has a NE-SW strike, providing evidence of activation of right-lateral strike-slip faults parallel to the Alpine fault. The fourth earthquake studied here (ML 4.1) was located close to the mainshock fault plane and had a more oblique reverse mechanism. It showed northward directivity, but this could be matched with slip on either nodal plane and so the fault plane could not be determined. All four earthquakes had relatively high stress drops (30-100 Mpa) with the shallowest having the lowest stress drop. These results confirm that previously unknown strike-slip faults trending NNW-SSE, and perpendicular to the predominant regional trend of the mapped faults, slipped in large (ML >= 5.1) aftershocks of the Arthur's Pass earthquake.
Strong motion modelling
The 1993 Tikokino, New Zealand earthquake (ML 6.1) is modelled as a unilateral rapture, exhibiting clear source directivity to the south. The earthquake was recorded by four strong motion stations within 30km: Waipawa to the South, and three sites in Napier and Hastings to the northeast. The shorter duration and greater amplitudes (by a factor of 10) observed at Waipawa with respect to the other stations provide clear evidence for the southward rupture direction. The Tikokino earthquake occurred on a shallow dipping, oblique reverse fault, and probably represents movement at the plate interface. A high rupture velocity is required to match the distribution of observed ground shaking, and the rupture area is constrained to be c. 7 x 2km2. The moment of the preferred model is 1.1 x 1018 Nm (Mw6.0) and the stress drop about 35MPa. This high average stress drop is consistent with the rupture being confined to an isolated asperity. The model used consists of a finite, rectangular fault rupturing with prescribed velocity and direction, and with uniform slip. The fault is embedded in a planar layered seismic velocity structure. The ability of the model to match the principal features of the observed seismograms suggests that it will be a useful tool in the prediction of strong ground motion for seismic hazard studies in the region.
The ML 6.3 1987 Edgecumbe earthquake was a normal faulting event located at the northern end of the Taupo Volcanic Zone that produced clear surface faulting. There was only one close, free field, strong motion record suitable for modelling. Pure normal faulting on either a single or double fault was modelled. The relative amplitudes of vertical to horizontal motions were found to be quite sensitive to near-surface velocities, which made it difficult to get a unique solution. In both the single and double fault models we could match the amplitude, frequency content and duration of the data well, which is encouraging for future seismic hazard work. With suitable adjustment of the many free parameters, such as the velocity model, fault orientation, size and slip distribution it would be possible to fit the data with either a double or a single fault model, but the problem is too poorly constrained for the solutions so obtained to be unique. Better azimuthal coverage of recorded data is required to constrain these parameters.