Abstract:The seismic correlation length prior to the Lushan earthquake(April.20,2013) has been studied by use of single-link cluster(SLC) analysis.The catalogs used were downloaded from China Earthquake Networks Center(CENC,http://www.csndmc.ac.cn).The earthquakes were located in an area centered in the mainshock(30.3N°,103.0°E) and ranging 27°~33° to 100°~106° from 5 years before the mainshock to its origin time.We selected 3.5 as the minimum magnitude and checked the completeness of the catalogs used prior to the mainshock using the Gudengberg formula,which shows good completeness above magnitude 3.5.The results showed that the growing correlation length of earthquakes for moderate-size earthquakes,which had been observed prior to the mainshock nearly two years ago to the origin time,can be fitted well by the power law.At the same time,we have developed a method to test whether or not the result obtained is simply caused by data selection to produce the desired pattern.We selected a time interval before the mainshock in which no great earthquake occurred and several years later when the greater earthquake occurred.The catalogs in the time interval will be analyzed using the same method as the catalogs prior to the mainshock study to check whether the correlation length shows an increase with the power law.We cannot find an obvious increasing process by the power law.For the preparation and occurrence stages of the earthquake process,characterized as a selforganized critical phenomenon,to test directly for growing spatial correlation lengths prior to large earthquakes is an independent approach to detect critical point behavior in observed seismicity.The results from this paper showed that the self-organized critical point characteristic was observed prior to the Lushan 7.0earthquake.The studies in this paper illuminate the notion that the growing seismic correlation length by the power law represents a physical process of criticalpoint characteristics in the source area preparing for a great earthquake.In future work,the predictive power of growing correlation length should be tested.In an area,the grid search technique can be applied to extract regions of growing correlation length systematically from a data set.The estimates for occurrence time,epicenter,and magnitude of expected mainshock may then be obtained from the fitting parameter,critical region,and value of correlation length.For a further reduction of uncertainties in a predictive approach,it also appears promising to combine the concepts of growing correlation length and other predictive methods