Understanding a promising earthquake forecasting tool by computer modelling of seismicity

Abstract

A regular feature of many real earthquake catalogues is that major earthquakes are preceded in the long term by an increase in the rate and magnitude of minor earthquakes. The largest minor earthquakes involved in this increase are usually about one unit of magnitude smaller than the major earthquake itself. Moreover, the precursor time (the time between the onset of the increase and the occurrence of the major event) and the precursory area (the area in which the increase, the major event and its aftershocks all occur) both grow in a regular statistical way with earthquake magnitude. This phenomenon is known as the precursory scale increase. It has been used to construct an earthquake forecasting tool, which works quite well in a number of earthquake prone regions, including New Zealand, California, Japan and Greece.

Why the precursory scale increase occurs is not well understood. In this study we try to understand it better by the use of computer-generated earthquake catalogues derived using simple assumptions about the physics of earthquake occurrence and the distribution of earthquake faults in the ground. If a computer-generated catalogue displays the precursory scale increase phenomenon, then perhaps the assumptions on which it is based are responsible for the phenomenon occurring in real catalogues.

It turns out that the distribution of faults in the ground is the key to the occurrence of the precursory scale increase phenomenon in the computer-generated catalogues. In a simple fault network with a small number of parallel faults, the precursory scale increase hardly occurs at all and the forecasting tool performs poorly. But in a more elaborate network involving major faults at a variety of orientations and a large number of small, randomly oriented faults, the precursory scale increase occurs before most major earthquakes and the forecasting tool performs well. Moreover the relations connecting earthquake magnitude, precursor time and precursory area are broadly similar.

The richness and variety of fault orientations therefore appear to be responsible for the precursory scale increase phenomenon. And the occurrence of the precursory scale increase in a computer-generated catalogue lends increased credence both to the forecasting tool and to the physical assumptions incorporated in the computer model. However there are some differences in detail between the patterns seen in the real and computer-generated catalogues. These could be investigated by further work on the computer model. This might include tracking changes of stress leading up to an earthquake on a major fault, and introducing more details of the physics of earthquake occurrence into the model.

---

Technical Abstract

The Every Earthquake a Precursor According to Scale (EEPAS) model has performed well as a long-range forecasting model for the larger earthquakes in a number of real seismicity catalogues. It is based on the precursory scale increase phenomenon and associated predictive scaling relations, the detailed physical basis of which is not well understood.

Synthetic earthquake catalogues generated deterministically from known fault physics and long- and short-range stress interactions on fault networks have been analysed using the EEPAS model, to better understand the physical process responsible for the precursory scale increase phenomenon. In a generic fault network with a small number of parallel faults, the performance of the EEPAS model is poor. But in a more elaborate network involving major faults at a variety of orientations and a large number of small, randomly oriented faults, the performance of the EEPAS model is similar to that in real catalogues, such as that of California and central Japan, albeit with some differences in the scaling parameters for precursor time and area. The richness and variety of fault orientations therefore appear to be responsible for conformity to the EEPAS model. Tracing the stress evolution on a set of individual cells in the synthetic seismicity model may give insights into the origin of the precursory scale increase phenomenon. It is possible that introducing visco-elastic relaxation into the synthetic seismicity program could explain some of the differences in scaling parameters.