Earthquakes happen when tectonic plates under the continents slip past each other violently, causing severe vibrations. Unlike eclipses, predicting earthquakes was entirely impossible prior to the 1970s. However, some headway has been made in recent decades.
In February 1975, scientists predicted an earthquake five hours in advance. The earthquake occurred in Haicheng in northeast China and it was the first time an earthquake prediction bore true. Millions of people had time to evacuate their homes and factories before the earthquake hit, saving tens of thousands of lives. Although many towns were totally destroyed, only a few hundred people died.
Another Chinese earthquake took place near T'ang-shan in August 1976. Although the earthquake was predicted a few years in advance, around 700,000 people still died from the quake. Long-term predictions seem to have limited value.
Earthquakes can be scientifically predicted by observing tiny cracks in rocks and how they widen when the rock is under stress. Other changes are also observable when a rock starts to compress, including changes in electrical resistance and the speed at which sound waves propagate through the rock. The swelling of cracks in a rock before it breaks is called dilantancy. It begins when the force on the rock is about half the force required for it to break.
In Russia and The United States, small earthquakes have been predicted up to five years in advance, but predicting large earthquakes or earthquakes around certain faults can be difficult. The rate of certainty has to be high for a warning to be issued. Earthquake prediction is still very much an imperfect science.
In 1966 at Denver, Colorado, waste liquids were injected at high pressure into a well. This loosened the friction between rocks in a fault, causing small earthquakes. Using this technique to release pressure at places like the San Andreas fault has been discussed, but not yet implemented. Clearly, smaller controlled quakes are preferable to a huge release of tectonic stress. Earthquakes disturbing the operation of nuclear power plants is a particular concern prompting research into predicting their occurrence.