Landslide Versus Splay: The Source of the Near-Field Tsunami of 1 April 1946

The large runup in Sumatra in 2004 showed that it is important to understand contributions to a tsunami other than that from interface slip on the subduction zone. More extreme even than 2004, the Aleutian tsunami of 1946 had a runup of 42m at Scotch Cap on Unimak Island, 160km from the epicenter. We previously proposed that a large debris avalanche would explain both the near and the far-field tsunamis of 1946. Okal, et al., however, showed that the triggering earthquake was large enough (at least M8.5) to produce the observed far-field tsunami, so they instead invoked a smaller rotational slump to produce the excess local runup. Neither explanation is valid: a multibeam survey with R/V Roger Revelle in 2004 showed that there is no single landslide or slump capable of producing significant runup at Scotch Cap. A third type of slope failure can produce a tsunami, however: widespread liquefaction can trigger a multiplicity of thin landslides which will generate a tsunami as they engulf water to produce a turbidity current (this is the probable generation mechanism for the Grand Banks tsunami of 1929). Unfortunately, the multibeam system on the Revelle was in poor adjustment during the 2004 survey, limiting resolution to about 25m. The flutes, furrows, and low scarps of thin-skinned failure were therefore unobservable. During the same cruise, however, ROV Jason was used for sea floor exploration and sampling. Push cores were obtained at nine different sites during seven dives on the forearc between 1900 and 4900m depth. At each of those sites, we performed 210Pb dating on samples taken at 1-cm intervals down the core. Apart from an occasional surface mixed layer, cores showed steady sediment accumulation at rates varying from 0.65 mm/y on a basement high on the Aleutian Terrace (3190m depth) to 2.93 mm/y in a valley on the upper slope at 1965m. Two cores, one from the basement high, the other from the mouth of a canyon opening onto the Aleutian Terrace, exhibit sediment disruption followed by uniform sedimentation; the disruption dates to 1946. Because only those two cores show any effects of the 1946 earthquake, we infer that the earthquake did not induce widespread slope failure on the forearc. In the absence of such failure, the likeliest source of the excess local runup is motion on splay faults. Multiple splays outcrop on the terrace and lower trench slope. The most conspicuous are seaward-verging faults on the lower slope, as well as the deformation front itself. Motion on any of these slip surfaces would have made a large amplitude, short wavelength contribution to the tsunami source, consistent with the extreme runup at Scotch Cap. In 1946, as in 2004, it appears that motion on splays contributed significantly to the local runup. Similar large local runup should be expected along all subduction zones (e.g., Cascadia) with well-developed splay faulting.