Erratic animal behavior has always been a mainstay in the earthquake-predicting folk-lore in China. This one is no exception. But it is even more than that. There is a claim that a paper, published in a scientific journal in 2006, had actually predicted an earthquake in Sichuan area in 2008!
The paper, titled Study on Earthquake Tendency in Sichuan-Yunnan Region Based on Commensurability, was published in the September, 2006, issue of the Journal of Catastrophology, can be downloaded here (in Chinese, with English translation of title and abstract at the end). It listed four authors, Long Xiaoxia(龙小霞), Yan Junping(延军平), Sun Hu(孙虎), and Wang Zuzheng(王祖正), all from the Shaanxi Normal University (陕西师范大学). Its abstract claimed that
Based on the data analysis of earthquake disasters in Sichuan-Yunnan region, the tendency of the next strong earthquake(s) is predicted by commensurability of ternary, quaternion and quintuple, for the purpose of disaster prevention and reduction.It is a very short paper, consisted mainly the tabulation and listing of historical data. With much parameter fitting, it reached its conclusion that the year 2008 will be the most likely time an earthquake of 6.7 or higher would happen in the Sichuan-Yunnan region.
After the publicity, New Threads published a series of articles debunking the methodology of this prediction at here, here, here, and here. Most point out that the paper had engaged in simple regression and extrapolation with a selective use of data. For example, the paper only used the years in which earthquakes of ranking 6.7 or higher. There was no explanation why 6.7 was chosen. It also treated the calendar year as an atomic unit of datum, ignoring the fact that there might be multiple earthquakes within one year.
Most damning of all is that, one author found, should the authors of the paper did their research before 1993, their method would have predicted a strong earthquake there in the year 1993. There was none.
So much for this prediction of earthquake.