Anomalous seismic activities in the Sichuan-Yunnan region and its adjacent areas before the Lushan M_S7.0 earthquake by the pattern informatics method

The Pattern Informatics (PI) method is a new approach for earthquake prediction based on statistical physics. This method can detect the regions where seismicity significantly deviates from the normal state (i.e., seismic quiescence or seismic activation), known as “seismic hotspots”, via strict statistical tests. Previous results show that PI is effective for medium-to long-term earthquake prediction. In this paper, after declustering and choosing completeness magnitude, the PI method was applied to earthquake catalogue from China Earthquake Networks Center since 1970, to detect the existence of seismic hotspots before the Lushan M_S7.0 earthquake on April 20, 2013. Based on suitable parameters (with grid size of 2°×2° and forecast window of 9 years) with higher prediction efficiency according to the previous researches of western China, we obtained the evolution of PI anomaly during the period from Wenchuan M_S8.0 earthquake to Sichuan Lushan M_S7.0 earthquake. The results showed that obvious seismic hotspots appeared in the convergent region (29°—31°N, 101°—105°E) of the Xianshuihe, Longmenshan and An’ninghe faults from 2009 to 2010. These seismic hotspots shrank to the range of (29°—31°N, 101°—103°E) during the years of 2011—2012 and disappeared after 2012, then the 2013 Lushan M_S7.0 earthquake occurred. In another evaluation of the evolution of PI anomaly by using a different set of calculation parameters (with grid size of 1°×1° and forecast window of 9 years), the results showed that there was a hotspot from 2009 to 2010 covering the southern Xinjin-Chengdu-Deyang fault, southern segment of Longquanshan western edge fault, Mabian-Yanjin fault, Ebian fault, and Tianquan-Yingjing fault (29°—30°N, 103°—104°E). After 2011, this hotspot disappeared and the Lushan M_S7.0 earthquake occurred in 2013. We also discussed the potential seismic risk in the studied area as well as the influence of calculated parameters on the prediction efficiency.