A probabilistic estimation of earthquake occurrence on the basis of the appearance times of multiple precursory phenomena.

Abstract

Appearance times of earthquake precursory phenomena are analyzed to calculate the synthesized probability of earthquake occurrence when multiple disciplines of precursors are observed. Various disciplines of precursors are classified into three groups: long, middle and short terms. We assume they are independent mutually. By analyzing the precursor time (Tp) data for each term using Weibull distribution function, it is found that distributions of Tp data for long, middle, and short terms have the properties of increasing, decreasing, and nearly constant hazard rates, respectively. Using Weibull parameters estimated for each term, we calculate the conditional probability of earthquake occurrence on the basis of the appearance times of precursors; where "conditional" means calculating probability on the condition that no earthquake has happened yet by the current time. Reliability parameter r is introduced for each term to take account of such a possibility that a part of the "precursors" are spurious ones. Introducing of the parameter r (<1) has an effect to make the conditional probability for every term decrease monotonously long after the appearance of each precursor. The synthesized probability is obtained by the formula of Utsu (1977, 1983). A probability change is calculated for the case of the Izu-Oshima-Kinkai earthquake of 1978 by applying this method to the observed precursor data.