Evaluation of Phenomena Preceding Earthquakes and Earthquake Predictability

Abstract

Unusual phenomena sometimes precede a large earthquake and are considered by some as a telltale sign of that earthquake. Judging whether the phenomenon was indeed related to the earthquake is difficult for individual cases. However, the accumulation of data over time allows for statistical evaluation to determine whether there is a correlation between the occurrence of a certain type of phenomena prior to an earthquake. The focus of this study is to review such statistical evaluation. The aspects considered in this study include seismicity, crustal deformation, slow slip, crustal fluids, crustal properties, electromagnetic phenomena, and animal behaviors. The lead times range from minutes to a few decades. The magnitude of the earthquake-preceding tendency can be universally measured by the probability gainG, which is the enhancement ratio of earthquake probability suggested by the occurrence of the phenomenon. A preceding tendency is considered to exist ifGis > 1 with reasonable statistical significance. Short-term foreshock activity, that is, temporarily heightened seismicity, produces by far the highestG> 100, sometimes exceeding 10000. While this strongly contributes to empirical forecasting, a considerable part of the predictive power of foreshocks is likely to derive from the mere aftershock triggering mechanism. This enhances the probability of small and large earthquakes by the same factor. It is fundamentally different from traditional expectations that foreshock activity signifies the underlying nucleation process of the forthcoming (large) earthquake. Earthquake-preceding tendency has also been proven significant for a number of other phenomena not ascribable to the aftershock-triggering effect. Some phenomena may be indicators of physical conditions favorable for large earthquakes, while some (e.g., slow slip) may represent triggering effects other than aftershock triggering. Phenomena not ascribable to aftershock triggering have a modestGof < 20 so far. However, these phenomena, including higher-order features of foreshocks, can be combined with the highGfrom aftershock-triggering effect, sometimes yielding a fairly scaring level of forecast. For example, say ∼10% chance of an M7 earthquake in a week in a few hundred km radius.