Living Earthquake Sensors
But of course animals react to other stimuli besides earthquakes and not all earthquakes have pre-shocks so scientists are trying to develop artifical sensors based on whatever methods animals use to predict earthquakes . Apparently they have not yet had a huge amount of success in their mimicry - some biological systems have evolved to do their job so extraordinarily well, it seems.
One author, the biologist and author Rupert Sheldrake (whom Salon calls 'a delightful crackpot'), has approached this problem in another way which, on the face of it, makes ideal use of something that humans have succeeded wonderfully in inventing - the internet. He proposes that a website logs records of unusual animal activity in earthquake-prone zones (like Susan and CB's ring of fire in comments below) and if there are enough reports and other activity can be ruled out then this would make an effective early-warning system. The fact that the reports come from many observers might help prevent false alarms.
A similar sort of idea has worked in China. In 1975 the behaviour of animals and various geophysical evidence was used to predict an earthquake in a province in China in 1975 which helped save thousands of lives. The trouble is the system also predicted earthquakes that didn't happen and also did not predict a dreadfully destructive one that did a year later. Prediction of earthquakes then is a complicated business. I suppose it faces the same problems as those people responsible for predicting terrorist attacks: there are always going to be false alarms and the real and vital skill is ascertaining which one (of the many) is real.