GRAPHICAL AND PC-SOFTWARE ANALYSIS OF VOLCANO ERUPTION PRECURSORS ACCORDING TO THE MATERIALS FAILURE FORECAST METHOD (FFM)

The Materials Failure Forecasting Method for volcanic eruptions (FFM) analyses the rate of precursory phenomena. Time of eruption onset is derived from the time of ''failure'' implied by accelerating rate of deformation. The approach attempts to fit data, OMEGA, to the differential relationship OMEGA = AOMEGA, where the dot superscript represents the time derivative, and the data OMEGA may be any of several parameters describing the accelerating deformation or energy release of the volcanic system. Rate coefficients, A and alpha, may be derived from appropriate data sets to provide an estimate of time to ''failure''. As the method is still an experimental technique, it should be used with appropriate judgment during times of volcanic crisis. Limitations of the approach are identified and discussed. Several kinds of eruption precursory phenomena, all simulating accelerating creep during the mechanical deformation of the system, can be used with FFM. Among these are tilt data, slope-distance measurements, crater fault movements and seismicity. The use of seismic coda, seismic amplitude-derived energy release and time-integrated amplitudes or coda lengths are examined. Usage of cumulative coda length directly has some practical advantages to more rigorously derived parameters, and RSAM and SSAM technologies appear to be well suited to real-time applications. One graphical and four numerical techniques of applying FFM are discussed. The graphical technique is based on an inverse representation of rate versus time. For alpha = 2, the inverse rate plot is linear; it is concave upward for alpha < 2 and concave downward for alpha > 2. The eruption time is found by simple extrapolation of the data set toward the time axis. Three numerical techniques are based on linear least-squares fits to linearized data sets. The ''linearized least-squares technique'' is most robust and is expected to be the most practical numerical technique. This technique is based on an iterative linearization of the given rate-time series. The hindsight technique is disadvantaged by a bias favouring a too early eruption time in foresight applications. The ''log rate versus log acceleration technique'', utilizing a logarithmic representation of the fundamental differential equation, is disadvantaged by large data scatter after interpolation of accelerations. One further numerical technique, a nonlinear least-squares fit to rate data, requires special and more complex software. PC-oriented computer codes were developed for data manipulation, application of the three linearizing numerical methods, and curve fitting. Separate software is required for graphing purposes. All three linearizing techniques facilitate an eruption window based on a data envelope according to the linear least-squares fit, at a specific level of confidence, and an estimated rate at time of failure.