Earthquake Prediction

A major factor in reducing the risk factor for earthquakes is predicting future earthquakes. Almost all earthquakes occur along active faults, which the United States Geological Survey defines as a "break in the Earth's crust upon which movement has occurred in the recent geologic past and future movement is expected." So, one of the first tasks of earthquake prediction is identifying and mapping active faults. Where these rupture recent geomorphological features (e.g. streams and recent glacial features) this task is made simple. However, some faults may have no geomorphological expression but have been the source of earthquakes in historic times. Unfortunately earthquake prediction is by no means an exact science. In spite of considerable effort going into prediction research, we can still only estimate probabilities of where and when. There are a number of predictive techniques which have been applied and they fall into two groups; long term earthquake modelling and precursory phenomena.

Long term earthquake modelling

Neotectonic research establishing the past earthquake activity in a region can date the intervals between major earthquakes (e.g. earthquakes which caused ground rupture). From these dates an average recurrence interval can be calculated, which allows estimation of the timing of the next major movement along a fault. Most earthquakes occur along plate boundaries which are moving past one another at an average rate of 10 cm yr^-1. If there is no movement along the plate during decades or centuries, then forces will build up and may be released in a single event. After the movement, forces will begin to accumulate at a similar rate. Around the Pacific Ocean a return period for earthquakes between 75 and 300 years is estimated, related to whether or not a fault moves episodically or by continuous creep (Bryant 1991, p 188-189).

In New Zealand, the Wellington Fault is one of the major active faults in the southern North Island. Rupture of the southern section of the fault is one of the most serious natural hazard scenarios in all of New Zealand, affecting the densely populated capital city of Wellington and smaller urban centres of Lower Hutt, Upper Hutt and Porirua. Neotectonic studies of the fault have identified that the Wellington-Hutt Valley segment ruptures as a single event; the last rupture was 340 - 490 years ago, the next oldest was 710 - 870 years ago. A relatively constant horizontal slip rate of 6 - 7.6 mm yr ^-1 along the fault has been established, with single event displacements of 3.2 - 47 m measured for the past five events. The average recurrence interval for events along the fault is calculated by dividing the single event displacement by the slip rate. A calculated average recurrence interval of 420 ± 780 years compares favourably with the time intervals (340 - 490 years and 220 - 530 years) between the last two events (Van Dissen et al. 1992).

If we plot the epicentres of past known events, we can, by extension of the observed trends, make informed deductions about possible future events. One major drawback of this method is that segments along a fault that appear aseismic, or relatively inactive, may be segments which are storing up energy for a catastrophic event. Such zones are termed seismic gaps, and they are thought to be prime sites for earthquake activity (e.g. the 1989 Loma Prieta, California, earthquake, Episodes 1989).

Precursory phenomena

Precursory phenomena which might give short term forewarning of an earthquake include (Rahn 1986; Bryant 1991; Mogi and Oyagi 1991):

• Anomalous crustal movements, or land deformation - the build up of strain in the crust will show in small lateral or vertical distortions on the surface. These can be measured using infra-red and laser survey, tiltmeters and strain gauges. In Japan a number of levelling surveys have been carried out. Up to 1955, ground level at stations close to the epicentre of the 1964 Niigata earthquake (Japan) were rising at a constant rate. However, in the years prior to the earthquake there was a sudden increase at two stations. These changes may or may not have been related to the earthquake. A similar survey series was carried out in the USSR close to, and before, the 1966 Tashkent earthquake. Here sudden changes in subsidence or uplift were recorded in 1944, with a more rapid change associated with the actual earthquake.

Tiltmeters can be used to provide a continuous measurement of ground movements. A simple water-tube tiltmeter records changes in heights of the water surfaces in separate tubes which indicates changes in ground level. Land deformation, however, may occur without any associated seismic activity and the problem then is working out just when the deformed land will fracture.

• Tide-gauge observation - the fact that land deformation precedes an earthquake implies that there must be associated changes in sea level. These changes are measured at tide-gauge stations. Two stations were used at the time of the 1964 Niigata earthquake and recorded significant changes in sea-level about 1 year before the earthquake. However, the actual shock caused a much greater change.

• Anomalous seismic activity - including seismic gaps (discussed before) and micro-earthquake swarms which indicate microfracturing along faults, affect the velocity of seismic waves through the rock. This can be measured by comparing P- and S-wave arrival time variations.

• Variations in geomagnetic and geo-electrical activity - locally anomalies in geomagnetic activity have been measured up to ten years before an earthquake and anomalies in electrical activity noted hours before events.

• Changes in level, temperature and chemical components of ground-water - ground water levels may fall, as prior to fault rupture small cracks may develop along the fault, allowing water to seep downwards or gas to seep upwards. Variations in the amount of Radon present in ground water have been observed to increase before an earthquake, possibly as fresh fracturing in crustal rocks allows more of the isotope to be absorbed into the water. Similar behaviour is also thought to occur for isotopes of Helium. Observation wells drilled in the region of the fault can be used to monitor changes in ground water level.

• Strange animal behaviour, or natural phenomena - animals may be sensitive to changes in magnetic or electrical fields that may presage an earthquake. They may be able to hear deep precursor breaking sounds not audible to the human ear. Prior to the 1775 Lisbon earthquake, for example, earthworms were described as having wriggled out of the ground.

The Japanese and Chinese have a long history of earthquake research. In China between 1960 and 1965 earthquake studies were undertaken to try and establish precursory phenomena that could be used to predict an earthquake. One example of a successful prediction was the M_s 7.5, 1975 Haicheng earthquake in China. Based on a combination of measurement of fault movements, changes in magnetic and gravitational fields, uplift of the ground and abnormal animal behaviour the earthquake was predicted and only a few lives were lost. However, the next major earthquake in China (M_s 7.5, 1976 Tangshan) was not predicted by the same methods and 650,000 people died.