Statistical-seismological Features Of Seismicity In The CSEP Testing Region Of The Central China South-North Seismic Zone With Implications For Earthquake Predictability

As a preparation work facilitating the testing region of the central China north-south seismic belt of the Collaboratory for the Study of Earthquake Predictability (CSEP), this thesis focuses on the earthquake catalogue of the study region (21～41.5°N,97.5～107.5°E, from the south to the north, including Yunnan, Sichuan, Gansu and their surrounding provinces), exploring the statistical seismological features of seismicity. Time-dependent completeness of earthquake catalogue was given as a background of the instrumentally recorded earthquakes in the testing region, which shows that since 1970 the overall completeness magnitude had been 3.0. Ergodicity exhibits in the whole testing region from 1990 to the present time, with region dependence, and was apparently interrupted by the 2008 Wenchuan earthquake sequence. As a development of the baseline study using the Triple-S (Simple Smoothing Seismicity model) approach, considering both activation and quiescence of seismicity in a unified framework of’fluctuation’, the Pattern Informatics (PI) algorithm was applied to the analysis of the five-year-scale predictability of the strong earthquakes with magnitude above 6.0.In comparison to the previous work, which uses the PI algorithm in the retrospective analysis, this thesis tries to put the algorithm in forward forecast test. The test uses the parameters as follow:cutoff magnitude 3.0, target magnitude 6.0+, grid size 0.2°×0.2°,’background window’10 years,’anomaly identification window’ 5 years, and’forecast window’5 years. We use the ROC (Receiver Operating Characteristic) diagram to evaluate the performance of the PI forecast. The result shows that the PI forecast outperforms not only random guess but also the relative intensity (RI) forecast.As an extended application of the PI algorithm in the theoretical discussion on the predictability of model seismic activity in a systerm with self-organised criticality (SOC), a simple cellular automaton (CA) model was developed, with a degenerated PI algorithm applied to the model seismicity. It was shown that the performance of the PI forecast outperforms random forecast.In the practice oriented discussion, to tackle the false-alarms in the Annual Consultation, we try to use the concept of reverse tracing of precursors (RTP) considering the long-term seismic activities in the alarm regions, or the regions with increased probability of earthquakes. We apply the PI and RI algorithms as the long-term estimation of sismic hazard in the RTP analysis. We select the central China north-south seismic belt as the study region and the period 2004-2012, which is a continuation of the previous work. A retrospective test shows that the RTP analysis using the PI/RI results can correctly remove some of the false-alarm regions and thus reduce the false-alarm rate of the Annual Consultation, without reducing the hit rate. It turns out that the RTP approach originally proposed for short-term earthquake predictions seems also able to contribute to the Annual Consultation.