STRUCTURE OF THE SUBDUCTING NAZCA PLATE BENEATH PERU

Arrival times from intermediate-depth (110-150 km) earthquakes within the region of flat subduction beneath the subandean zone and foreland basins of east-central Peru provide constraints on the geometry and ve locity structure of the subducting Nazca plate. Hypocentral locations and origin times for these events were determined using observations f rom a 15 station digitally recording locator array deployed in the epi central region of eastern Peru. Observed P wave arrival times for coas tal stations in Peru, some 3-6-degrees from the epicenters, are up to 4 s early relative to predicted arrival times based on the best fit ve locity-depth model used for hypocenter locations. These large negative time residuals appear to be the result of propagation paths which hav e long segments in the colder, higher-velocity subducting plate. P wav e travel times were modeled for the effects of the slab using three-di mensional (3-D) ray tracing. Computed ray paths show that travel times to coastal stations for the eastern Peru events can be satisfactorily modeled with average velocities relative to the surrounding mantle 6% lower within the uppermost slab (assumed on the basis of other studie s to be unconverted basaltic oceanic crust 6 km thick) and 8% higher w ithin the cold uppermost mantle of the slab. Ray tracing for this plat e model shows that P wave ''shadow zones'' can occur if the source-sla b-receiver geometry results in seismic rays passing through regions in which the clip angle of the slab changes significantly. Such geometri es exist for seismic waves propagating to some coastal stations from i ntermediate-depth earthquakes located east of the Andes. Observed firs t-arrival times for such cases do in fact have smaller negative residu als than those for geometries which allow for ''direct'' paths at simi lar distances. Modeling such arrivals as internally reflected waves pr opagating through the high-velocity part of the plate produces a signi ficant improvement in the travel time residuals. For the slab velociti es given above, we obtain a model thickness of approximately 36 km for the cold slab interior and a slight northwest component of dip in the region of subhorizontal subduction.