Accurate earthquake prediction, which could save many thousands of lives and minimize earthquake damage, has so far eluded seismologists. One attempt at predicting an earthquake took place during the latter part of the twentieth century in the town of Parkfield in California. Based upon a known series of earthquakes, the experiment attempted to predict a ‘window’ during which the next tremor of significant size would occur.
The Parkfield Earthquake Sequence
Parkfield lies on the San Andreas Fault in California. The fault is part of an active transform fault zone which marks the boundary between the North American tectonic plate, which is moving roughly south-westwards at a rate of around 23mm each year, and the Pacific plate, which is moving approximately north-eastwards at around 79mm each year (This Dynamic Planet map). The relative motions of these plates generates regular earthquakes, many of significant size.
The earthquake history of Parkfield is well-documented but the sequence of relevance to the experiment is best summarized in the article which originally proposed the prediction model. Looking at earthquake events where the mainshock (the largest tremor) was M6.0 or greater, WH Bakun and AG Lindh described a sequence of seismic events.
Establishing the true nature of the sequence of earthquakes in Parkfield, especially in its early part, is problematic due to unsophisticated methods of measuring and recording historic earthquakes. Bakun and Lindh looked at shocks which occurred in 1857, 1881 1901, 1922, 1934 and 1966 and which had magnitudes varying between approximately 5.0 and 6.4. These events were shown to have a mean interval of just under 22 years and their occurrence was described as ‘remarkably uniform’.
Parkfield and the Seismic Gap Theory
The seismic gap theory is, in essence, based upon the idea that strain generated by earth movement accumulates along a given section of a fault at a steady rate and that, once a certain threshold is reached the strain will be released and an earthquake will occur. The strain will then re-accumulate until the threshold is exceeded again and another tremor takes place (Yufang Rong et al).
On this basis, the risk of an earthquake is lowest immediately after an event and increases with time. In theory (and many assumptions have to be made about the rate of strain accumulation, among other things), such a pattern would be expected to produce sequences of earthquakes along a fault segment which are broadly the same in magnitude and in other characteristics and which occur with a relatively even frequency – as at Parkfield.