Comprehensive Earthquake Prediction Research Based On Probability Gain Model

The comprehensive probability gain earthquake prediction method, in this thesis, can accommodate a variety of qualitative and quantitative individual earthquake prediction methods, which are deterministic or probabilistic. The relationship between the probability gain, the forecast performance and the probability of the predictor helps to make the process of earthquake preparation and occurrence, and precursor expression closely linked,which is important to explore or interpret the physical nature of earthquake preparation and occurrence , and improve the prediction method. However, the comprehensive probability gain prediction model is based on the assumption that different predictors are independent, which is not fully consistent with the actual situation. Therefore, we need to explore two issues. The first issue is the correlation between different predictors; the second issue is the influence of the predictors with a certain correlation on the results of comprehensive probability gain prediction. For the first issue, there are many methods to estimate its correlations in practice, mainly including physical nature and correlation coefficient. The second issue is the focus of research of the thesis.The influence of two predictors with some relevance on the results of comprehensive probability gain prediction is analyzed and discussed in detail by using Monte Carlo simulation. Its results show that the difference between joint probability gain and apparent probability gain is bigger with the correlation between two predictors increasing. So it is necessary to correct its apparent probability gain by depending on its joint probability gain.On the basis of research of the comprehensive probability gain prediction method and the correlation between its predictors, the approach how to integrate different individual predictors efficiently was explored by using two concrete predictors, which are PI(Pattern informatics) algorithm and the normalized heterogeneity parameter Kcv of spatial distribution of earthquake。PI algorithm, based on statistical mechanics of complex systems, is an earthquake forecasting model, which was researched as one individual prediction method of comprehensive probability gain prediction, and was further applied to the next decade large-scale earthquake prediction in Asian area. The effectiveness of PI algorithm was evaluated by analyzing its actual test of retrospective prediction from 2000/1/1 to 2009/12/31. According to the evaluated result, I continue to predict the Mw6.9 probability of long-term seismic hazard forecasting from 20101/1/ to 2019/12/31 in Asian area.The normalized heterogeneity parameter Kcv of spatial distribution of earthquake was redefined by depending on original non-uniformity index Cv of spatial distribution of earthquake. On this basis, according to the spatial-temporal probability gain model and the assessment method of efficiency of earthquake prediction methods which was designed by using different spatial deviations and time delays between target earthquake and predictor grid, the empirical earthquake probability gain spectrum and efficiency spectrum of this novel method were calculated using the heterogeneity research and earthquake cases statistics over past 20 years. The robustness and effectiveness of Kcv-value method were further evaluated by analyzing its probability gain values, the probability of Mw6.9 earthquake occurrence from 2000 to 2009 and its actual test of retrospective prediction. The approach provides a way of predicting the Mw6.9 earthquake probability and probability gain from 2010 to 2019, and provides a reference to determine the Asian earthquake risk zone in the next decade and another individual prediction method of comprehensive probability gain prediction.Based on the above research of PI algorithm and Kcv-value, the correlation between the two individual methods was studied in detail. According to comprehensive probability gain prediction method considering relevance between predictors, the research of retrospective prediction and forecasting efficiency of comprehensive earthquake prediction were conducted by predicting long-term seismic hazard of the Mw6.9 earthquake from 2000/1/1 to 2009/12/31 in Asian area. Its results show that the overall forecast performance was improved by integrating different individual prediction methods with comprehensive probability gain prediction method considering relevance between predictors, which is justly the significance of comprehensive probability prediction method.