论文部分内容阅读
由于危险性的潜在统计学信号特征未知,在概率地震危险性分析(PHSA)中模型的不确定性普遍存在。虽然人们很好地理解了在概率地震危险性分析中整合独特模型的参量不确定性的方法,但由于不同地震复发模型之间存在高度依赖性,实施这一方法要更为困难。我们所展示的是2002年加州地震概率工作组(WGCEP-2002)所使用的、用来把多重地震复发模型给出的概率分布进行联合的方法,其结果上有几处相反的效果。特别是,WGCEP-2002使用了一种模型的线性结合方法,该方法忽视了模型的依赖性,并在最终的危险性估计中造成了大的不确定性。并且,模型权重的选择以数据为基础,这有可能导致系统偏离最终概率分布。该工作组报告中所使用的权重选择方案亦产生出依赖于模型任意排序的结果。除了对当前的统计学问题进行分析之外,我们还为严格地把模型不确定性整合进概率地震危险性分析里展示了另一种可选方法。
Due to the unknown nature of the potential statistical signal of danger, the uncertainty of the model in the Probabilistic Seismic Hazard Analysis (PHSA) is ubiquitous. Although well-understood methods for integrating parametric uncertainties of unique models in probabilistic seismic hazard analysis are implemented, it is more difficult to implement this method because of the high degree of dependence between different models of seismic recurrence. What we show is the method used by the 2002 California Earthquake Probability Task Force (WGCEP-2002) to combine the probability distributions given by the multiple earthquake recurrence models. The results have several opposite effects. In particular, WGCEP-2002 used a linear combination of models that ignored the model’s dependence and caused major uncertainties in the final risk estimate. Also, the choice of model weights is based on data, which can lead to deviations of the system from the final probability distribution. The weight selection scheme used in the WG report also produced results that relied on arbitrary ordering of models. In addition to analyzing the current statistical issues, we show an alternative approach to rigorously integrate model uncertainty into probabilistic seismic hazard analysis.