中长期地震危险性概率预测中的统计检验方法Ⅰ:Molchan图表法
|
摘要
对中长期地震危险性概率预测结果的统计检验是地震预测预报的重要研究内容,采用通用的统计检验方法是促进地震预测理论、模型"无障碍"交流和发展的必要条件。本文通过对青藏高原东北缘地区PI算法和RI算法回溯性预测结果统计检验的实例,介绍了如何运用CSEP计划中已常规采用的Molchan图表法进行地震概率预测统计检验和分析,并分别考虑了网格权重和地震活动权重两种算法计算"时空占有率"的情况。结果表明,Molchan图表法可通过显著性水平α和概率增益Gain有效地评价概率预测模型的预测能力。此外,对于同样的预测结果,在考察不同的预测策略时得到的统计检验结果也可能不同。其中地震活动权重相对网格权重采用更为严格的统计检验。
The statistical test of long-term and intermediate-term seismic hazard probabilistic forecasting is an important aspect of earthquake forecast/prediction.Using general statistical test method is one of the necessary conditions for promoting earthquake forecast/prediction theory and model accessibility exchange and development.Based on the retrospective forecast results of PI and RI algorithm in the northeastern marginal region of the Qingzang plateau,we try to introduce how to apply the Molchan error diagram to do the statistical test of earthquake probabilistic forecast,and the situation of spatial-box number weighted and seismic rate weighted are considered respectively to calculate the fraction of space-time occupied by alarm.The result shows that the Molchan error diagram can evaluate the probabilistic earthquake forecast model effectively by using the significance level α and the probability gain.Furthermore,the significant different testing results also can be found when different forecast/prediction strategies are employed even if for the same forecast/prediction results.
The statistical test of long-term and intermediate-term seismic hazard probabilistic forecasting is an important aspect of earthquake forecast/prediction.Using general statistical test method is one of the necessary conditions for promoting earthquake forecast/prediction theory and model accessibility exchange and development.Based on the retrospective forecast results of PI and RI algorithm in the northeastern marginal region of the Qingzang plateau,we try to introduce how to apply the Molchan error diagram to do the statistical test of earthquake probabilistic forecast,and the situation of spatial-box number weighted and seismic rate weighted are considered respectively to calculate the fraction of space-time occupied by alarm.The result shows that the Molchan error diagram can evaluate the probabilistic earthquake forecast model effectively by using the significance level α and the probability gain.Furthermore,the significant different testing results also can be found when different forecast/prediction strategies are employed even if for the same forecast/prediction results.
引文
[1]Yamashina K.Trial of earthquake prediction in Japan and a statistical test of time-shift[J].Tectono-physics,2006,417(1-2):169-182.
[2]Schorlemmer D,Zechar J D,Werner M J,et al.First results of the regional earthquake likelihoodmodels experiment[J].Pure and Appl Geophys,2010,167(8):859-876.
[3]Schorlemmer D,Gerstenberger M C.RELM Testing Center[J].Seismol Res Lett,2007,78:30-36.
[4]Molchan G M.Strategies in strong earthquake prediction[J].Phys Earth Plane Inter,1990,61(1-2):84-98.
[5]Molchan G M.Structure of optimal strategies of earthquake prediction[J].Tectonophysics,1991,193:267-276.
[6]Molchan G M.Earthquake prediction as a decision making problem[J].Pure and Appl Geophys,1997,149:233-247.
[7]Molchan G M.Space-time earthquake prediction:the error diagrams[J].Pure and Appl Geophys,2010,167:907-917.
[8]Tiampo K F,Rundle J B,McGinnis S,et al.Mean field threshold systems and phase dynamics:Anapplication to earthquake fault systems[J].Europhys Lett,2002,60(3):481-487.
[9]Shen Z K,Jackson D D,kagan Y Y.Implications of geodetic strain rate for future earthquakes,witha five-year forecasts ofM5 earthquakes in Southern California[J].Seismol Res Lett,2007,78(1):116-120.
[10]Zechar J D,Jordan T H.Testing alarm-based earthquake predictions[J].Geophys J Int,2008,172:715-724.
[11]Keilis-Borok V I,Soloviev A A(eds).Nonlinear dynamics of the lithosphere and earthquake predic-tion.Springer-Verlag[M].In:Berlin-Heidelberg,2003.
[12]Kossobokov V G.Earthquake prediction:principles,implementation[J].Perspect Comput Seismol,2005,36-1:3-175.
[13]Rundle J B,Klein W,Turcotte D L,et al.Precursory seismic activation and critical-point phenome-na[J].Pure Appl Geophys,2000,157:2 165-2 182.
[14]Rundle J B,Turcotte D L,Shcherbakov R,et al.Statistical physics approach to understanding themultiscale dynamics of earthquake fault systems[J].Rev Geophys,2003,41(4):1 019,doi:10.1029/2003RG000135.
[15]Chen C C,Rundle J B,Holliday J R,et al.The 1999 Chi-Chi,Taiwan,earthquake as a typical ex-ample of seismic activation and quiescence[J].Geophys Res Lett,2005,32,L22315,doi:10.1029/2005GL023991.
[16]Holliday J R,Chen C C,Tiampo K F,et al.A RELM earthquake forecast based on pattern infor-matics[J].Seism Res Lett,2007,78:87-93.
[17]蒋长胜,吴忠良.对地震预测的一个统计物理算法在川滇地区的回溯性预测检验[J].中国科学:D辑,2008,38(7):852-861.
①http://www.cscptesting.org
[2]Schorlemmer D,Zechar J D,Werner M J,et al.First results of the regional earthquake likelihoodmodels experiment[J].Pure and Appl Geophys,2010,167(8):859-876.
[3]Schorlemmer D,Gerstenberger M C.RELM Testing Center[J].Seismol Res Lett,2007,78:30-36.
[4]Molchan G M.Strategies in strong earthquake prediction[J].Phys Earth Plane Inter,1990,61(1-2):84-98.
[5]Molchan G M.Structure of optimal strategies of earthquake prediction[J].Tectonophysics,1991,193:267-276.
[6]Molchan G M.Earthquake prediction as a decision making problem[J].Pure and Appl Geophys,1997,149:233-247.
[7]Molchan G M.Space-time earthquake prediction:the error diagrams[J].Pure and Appl Geophys,2010,167:907-917.
[8]Tiampo K F,Rundle J B,McGinnis S,et al.Mean field threshold systems and phase dynamics:Anapplication to earthquake fault systems[J].Europhys Lett,2002,60(3):481-487.
[9]Shen Z K,Jackson D D,kagan Y Y.Implications of geodetic strain rate for future earthquakes,witha five-year forecasts ofM5 earthquakes in Southern California[J].Seismol Res Lett,2007,78(1):116-120.
[10]Zechar J D,Jordan T H.Testing alarm-based earthquake predictions[J].Geophys J Int,2008,172:715-724.
[11]Keilis-Borok V I,Soloviev A A(eds).Nonlinear dynamics of the lithosphere and earthquake predic-tion.Springer-Verlag[M].In:Berlin-Heidelberg,2003.
[12]Kossobokov V G.Earthquake prediction:principles,implementation[J].Perspect Comput Seismol,2005,36-1:3-175.
[13]Rundle J B,Klein W,Turcotte D L,et al.Precursory seismic activation and critical-point phenome-na[J].Pure Appl Geophys,2000,157:2 165-2 182.
[14]Rundle J B,Turcotte D L,Shcherbakov R,et al.Statistical physics approach to understanding themultiscale dynamics of earthquake fault systems[J].Rev Geophys,2003,41(4):1 019,doi:10.1029/2003RG000135.
[15]Chen C C,Rundle J B,Holliday J R,et al.The 1999 Chi-Chi,Taiwan,earthquake as a typical ex-ample of seismic activation and quiescence[J].Geophys Res Lett,2005,32,L22315,doi:10.1029/2005GL023991.
[16]Holliday J R,Chen C C,Tiampo K F,et al.A RELM earthquake forecast based on pattern infor-matics[J].Seism Res Lett,2007,78:87-93.
[17]蒋长胜,吴忠良.对地震预测的一个统计物理算法在川滇地区的回溯性预测检验[J].中国科学:D辑,2008,38(7):852-861.
①http://www.cscptesting.org
版权所有：© 2023 中国地质图书馆 中国地质调查局地学文献中心