SEISMOLOGICAL ASPECTS OF THE 1989-1990 ERUPTION AT REDOUBT VOLCANO, ALASKA - THE MATERIALS FAILURE FORECAST METHOD (FFM) WITH RSAM AND SSAM SEISMIC DATA

Seismic activity during the December 1989 to April 1990 eruption of Redoubt Volcano, Alaska, has been tracked by the Alaska Volcano Observatory using Real-time Seismic Amplitude Measurement (RSAM) and Seismic Spectral Amplitude Measurement (SSAM) systems with up to five stations. Data consist of 10-minute averages of the absolute amplitudes of seismometer output. These data have been used to test in hindsight the Materials Failure Forecast Method (FFM), which attempts to define the time of eruption with a time series for precursory phenomena whose rate of change accelerates measurably before events. Practical application of the method emphasizes inverse-rate plots, both for early detection of signal emergent from background, and for event forecasting. Eruption windows are determined by graphical or numerical extrapolation of the inverse-rate trend and by intersection of a data envelope, reflecting data scatter and consistency, with an empirical critical rate near the time of eruption. Prior to the dome-destroying eruption of January 2, rate changes were of sufficient consistency, duration, and intensity for a qualitative or quantitative FFM predictive analysis, using either RSAM or SSAM. Signal-to-noise ratio was high for both RSAM or SSAM data sets, and FFM analyses could have provided useful support for decision making. Following January 2, in association with a rapid succession of dome-collapse eruptions, signal strength diminished and RSAM signal-to-noise ratio declined. The most distinct patterns on RSAM reflected noise rather than signal, and forecasting exclusively based for RSAM would have been misleading. By eliminating the noise in excluded frequencies, an enhanced signal-to-noise ratio was generally produced by SSAM for banded frequencies near 2 Hz. After January 2, only SSAM exhibited signal-to-noise ratios suitable for FFM analysis. Inverse-SSAM plots could have been informative for early detection of precursory long-period seismicity. The prospect of combining the FFM approach with spectral amplitudes in selected frequencies offers considerable promise.