dc.description.abstract | Earthquakes originate from a driven nonlinear threshold system. It’s impossible to fully understand the dynamical processes and measure the internal state variables, but we can know the system by pattern state. Rundle et al. purposed the Pattern Informatics (PI) method to analyze the changes of seismicity before and around a large earthquake[Rundle et al., 2003; Tiampo et al., 2002a; Chen et al., 2005; 2006; Wu et al., 2008a]. In this study, we calculated the anomaly area which associated with large earthquakes in Taiwan region by PI method from Taiwan CWB earthquakes catalog. Chenong et al. [2014] well applied the Soup-of–group (SOG) model, a mathematics model, to earthquakes system. In SOG earthquake model, the numbers of small events had expectation decreased (seismic quiescence) in order of magnitude before the large earthquake occur. Therefore, in this study, we first time ever improved the PI method, inspired from SOG model, by calculating the change of seismicity rate by dividing the magnitude range into several segments and multiple them to get a new PI relative probability. We retrospectively tested the target earthquakes with magnitude larger than 5.5 from 2000 to 2016 by new PI method, and objectively evaluated the performances of the new method by the Relative Operating Characteristic (ROC) method which were significantly better than original results. Finally, we obtained absolute PI values by comparing the PI values of all the grids in space at 204 timing, and it was possible to define a true absolute high value region for all of time and space in the future calculations. The most important is that the future large earthquakes will occur with high probability in these anomaly areas determined by absolute PI high value from this study, so there is a high probability that absolute PI values can be delicate converted into the probability of earthquake occurrence. | en_US |