Analysis of 9 years of data from the Humboldt Bay seismic network sheds new light on the structure and evolution of the Gorda plate and Mendocino triple junction. Significant findings include the buttressing effect of the Pacific plate which demonstrates that there is no underthrusting of the Gorda plate along the Mendocino fault, and the pattern of left-lateral northeast-trending faults, which demonstrates how the Gorda plate accommodates N--S shortening without such underthrusting. Focal mechanisms are consistent with northward compression of the Gorda plate by the Pacific plate until the Gorda plate passes the triple junction, beyond which the N--S compressive stress is effectively removed, and the focal mechanisms show a change from strike slip to normal faulting with downslab tension. Another significant feature is the shallow Benioff zone (10--40 km) which shows a double seismogenic layer. Unlike most other double seismic layers in deep subduction zones, this one appears to be due to underplating, with a new subduction zone being developed east of the former one.

A corner of continental margin material has been partially subducted. We attribute this double seismic zone to reactivation of both the old and new subduction boundaries under N--S compression produced by Pacific--Gorda plate interaction at the triple junction. Because the geometry of this triple junction indicates that it is unstable and because there is evidence that as part of its evolution the San Andreas fault has migrated eastward, we have constructed a model that accounts for both the eastward migration of the San Andreas fault and the doubling of the seismic zone in the Gorda plate. This can be done by assuming that at the time of the last eastward jump of the fault, the overriding continental margin north of the triple junction was broken and underthrust producing a new subduction zone that is collinear with the San Andreas fault. Other scenarios for the evolution of the Mendocino triple junction do not require that the subduction zone migrate eastward. However, the position and dip of the double zone indicate that an eastward jump of 100 km some 5 my ago could have taken place. This is consistent with other evidence for an eastward jump of the San Andreas fault. ¿ American Geophysical Union 1993