Abstract: Based on principal component analysis method, the authors optimized nine radon observatories in Shanghai area by using 1998 radon data. The result shows that three observatories that maintain little information can be canceled, because the rest six observatories covered 93.0% information of the total nine observatories. It means that an increase of observatory number does not always work, because the information gained from these observatories will not increase in accordance if the observatory data are correlative in the local area. The method to optimize observatory distribution is very important to direct the optimization of earthquake precursory observatories.

Key words: Principal component analysis, Eigenvector, Contribution ratio, Optimizing observatory