China’s efforts on the CSEP and a retrospective case

China has established a Testing Center on the Collaboration for the Study of Earthquake Prediction (CSEP) for standardized evaluation of current earthquake prediction methods. Some traditional methods such as Load-Unload Response Ratio (LURR), Pattern Information (PI), Relative Intensity (RI), Epidemic Type Aftershock Sequence (ETAS), and b-value have been installed. In addition, several new methods, such as Crustal Vibration (CV), State Vector (SV), Earthquake Occurrence Rate (EOR), and Seismic Modulation Ratio (SMR) have also been involved. Their prediction performance determined by the prediction periods, magnitude of detection earthquakes, and optimal critical regions can be evaluated using the Molchan-test, R-test, N-test and ROC-test. It has a potential application in earthquake prediction by developing a Comprehensive Probability (CP) model, in which the predictive performance of each method is regarded as the weights and integrated using the Bayesian formulas. As a retrospective example, we applied the R-test to evaluate the prediction power of the LURR method by using the seismic data over the last 50 years in the north-south seismic belt of China. Results show that the predictions of LURR outperform the null hypothesis, and the optimal prediction performance for Ms > 6.0 earthquakes with a 2-year time window is derived from various model tests.