The occurrence of the 7 May 1986 Andreanof Islands earthquake provides the opportunity to evaluate the earthquake forecast and specific prediction that had been formulated and promulgated, starting more than 2 yr before. The prediction experiment was based on the observation of seismic quiescence in the part of the Aleutian subduction zone monitored by the Central Aleutians (Adak) Seismic Network. The magnitude estimate (Ms 7 to 7.5) was based on the size of the quiescent zone and the seismic history of the Adak region since 1900. The time of occurrence was predicted as a test of the discovery by Habermann of a quiescence lasting about 3 yr prior to the 2 May 1971 Adak Canyon event, and the hypothesis that a characteristic precursor time is a valid concept. The place was originally forecast as “near Adak Island,” then specified with epicenter at the eastern margin of Adak Canyon, rupturing to the west under the canyon. The 7 May 1986 earthquake had magnitude Ms 7.7 (Mw 8.0), with epicenter 140 km east of the predicted place, rupturing east of Adak Canyon and stopping in the area predicted as the nucleation point. The time of this event was 6 months later than predicted. Thus, the specifically predicted event has not happened. The largest part of the rupture did go through the segment that had been most active before the onset of quiescence and which experienced the sharpest and most persistent quiescence. A cause-and-effect relation between preparation for this earthquake and the quiescence is still to be firmly established, but the distribution of the pattern of quiescence relative to the epicentral location and the site of the greatest moment release suggest a physical connection between earthquake and seismicity patterns. The predictions of location and time were acknowledged throughout as highly speculative. No firm basis exists for quantitative estimation of the uncertainties to be assigned to the elements of an intermediate-term prediction.
This experiment lends support to the validity of quiescence as a precursor, at least for some earthquakes, and demonstrates the need for local network data to provide the high resolution of seismicity required for prediction efforts based on seismicity patterns.