The ability to distinguish foreshocks from background seismicity is very important in short-term earthquake prediction. To that end we have looked at spatial clustering (using waveform cross-correlation) and stress drops of foreshocks of two New Zealand earthquake sequences that occurred in 1990. The Tennyson sequence, located in a continental margin-type strike-slip environment, consisted of a group of foreshocks, an ML = 5.8 mainshock, and many aftershocks. A cross-correlation analysis showed five spatially close clusters of activity prior to the mainshock. Two were event pairs located within the final aftershock zone, two were clusters of four events, each located outside the aftershock zone, and the fifth was a cluster of eight immediate foreshocks loated within the aftershock zone. An analysis of two nearby control regions showed that pairs of identical events were not uncommon, but larger clusters were. Stress drops of three events in the 12 days before the mainshock, obtained by deconvolving small events as empirical Green's functions, were lower than for earlier preshocks and aftershocks. Source time functions derived from the Green's function deconvolution indicated that a unilateral rupture model was more appropriate than a circular source model.
Cross-correlation values from the ML = 5.9 Weber sequence also showed spatial clustering, but this was well removed from the mainshock in space and time. A control area also showed similar clustering, suggesting that it is a normal feature of the seismicity at a convergent margin. The Weber foreshocks, only four in all, were not highly correlated. For both sequences, foreshocks did not correlate with the aftershocks, indicating that they occurred in a region of complete coseismic stress relief.
A stress drop of 1650 bars was obtained for a 44-km-deep event that occurred within the upper part of the subducting Pacific Plate, nearby, but not related to the Tennyson sequence.