A new method for estimating source extent parameters has been applied to the Imperial Valley earthquake of October 15, 1979, and the Victoria earthquake of June 9, 1980. From the stations of the World-Wide Standard Seismograph Network and Canadian Seismic Network, estimates of the duration of far-field long-period SH waves were made by computing the variance or second-central moment &tgr; 2 of instrument-corrected, attenuation-corrected pulses. Seventeen measurements for the Imperial Valley event and 25 for the Victoria event were made. We inverted for the source duration T, the fault length L, and the directivity parameter D by least squares. We find that for the Imperial Valley, T=9.0¿0.5 s, D=130¿43 km s, and L=46¿13 km, and for the Victoria event, T=10.9¿0.6 s, and D=251¿71 km s. L is not well constrained for this second event; the inversion gives 23 km, the lower bound obtained by constraining &tgr; 2 to be positive everywhere on the focal sphere, although a value as high as 40 km would still be within the uncertainties. The positive values of D for both events indicate that rupture propagation was predominantly to the northwest. In order to estimate the degree of ''unilateralness'' we have computed the rupture mode index defined by R=‖D‖/LT, which is 1 for a unilateral rupture and zero for a bilateral rupture. We find a value of R=0.31¿0.14 for the Imperial Valley event, significantly less than unity, which indicates that a southern component of rupture is also present. Assuming uniform slip, we estimate the northern and southern fault segments to be 26 and 20 km, respectively. The presence of a bilateral component is supported by near-field studies which have made use of stations along the southern half of the Imperial fault. For the Victoria event, due primarily to the uncertainty in L, the value of R can vary from 0.6 to 1 with the value of 1 being most probable. Assuming that the rupture is unilateral, we can obtain an estimate of fault length from the relation L=‖D‖/T, which yields L=23¿6 km.

Geodetic data and the distributions of aftershocks are consistent with a unilateral rupture mode and the fault length that we have obtained. Values of seismic moment from measurements of the zeroth moment of the displacement pulse are found to be 5.8¿0.4 for Imperial Valley and 4.8¿0.4 for Victoria, in units of 1018 N m. The presence of a southern component of rupture for the Imperial Valley event would suggest a closer and perhaps causal relationship to the Victoria event, which occurred 8 months later and about 60 km to the south. In a tectonic framework we interpret these two earthquakes as the failure of adjacent transform faults and consider the possibility that a concurrent spreading event may have occurred in the vicinity of Cerro Prieto.