Collaborative Research: From Earthquake Physics to Testable Forecasts

试图结合地震物理的模型在地震学中发挥着越来越重要的作用，在地震危险性评估中也越来越普遍。大多数情况下，为这些模型提供信息的物理概念是启发式的，并且基于专家意见。统一加州地震破裂预报（UCERF2）采用了几个这样的模型。为了提高即将到来的UCERF3的可测试性，我们建议研究特征地震假设，最大震级与断层长度的关系以及库仑应力假设。对于许多断层，地震学家确定了所谓的特征地震，并将其作为预期目标地震纳入危险评估。最突出和研究最好的例子是加利福尼亚州帕克菲尔德的6级事件序列。研究表明，这种现象可能可以用频率-震级分布的较高震级范围的低样本量来解释。最大震级与断层长度的关系在灾害评估中起着重要作用，因为它用于估计特定断层未来大事件的大小。这种关系纯粹是基于事后观察，并没有使用先验预测进行检验。所选模型中最复杂的是库仑应力模型。它显示了很好的描述能力，但也依赖于通过拟合自由参数的后验观测。它的不确定性范围仍相对未被探索。所有这些概念模型的共同点是，它们的预测能力从来没有经过严格的测试，这在某种程度上是由于将它们表述为可测试的假设很困难。我们建议探索这些假设的不确定性范围，并将其转化为可测试的假设，以便在地震可预测性研究合作实验室的框架内进行严格测试。这项研究将为UCERF3的工作增加更多的可测试性，并有助于理解和量化这些模型对地震危害评估的影响。

Models that attempt to incorporate earthquake physics play an increasingly important role in seismology and are prevalent in seismic hazard assessment. Most often, the physical ideas that inform these models are heuristically motivated and based on expert opinion. The Uniform California Earthquake Rupture Forecast (UCERF2) employs several such models. To increase the testability of the upcoming UCERF3, we propose to investigate the characteristic earthquake hypothesis, the relation of maximum magnitude to fault length, and the Coulomb Stress hypothesis. For many faults, seismologists identified so-called characteristic earthquakes and include them as expected target earthquakes in hazard assessment. The most prominent and best studied example is the sequence of magnitude 6 events at Parkfield, CA. Studies have shown that this phenomenon possibly could be explained by low sample size from the upper magnitude ranges of the frequency-magnitude distribution. The relation of maximum magnitude to fault length plays a major role in hazard assessment as it is used to estimate the size of future large events at particular faults. This relation is based purely on a posteriori observations and has not been tested using a priori predictions. The most complicated of the selected models is the Coulomb Stress model. It shows great descriptive capabilities but is also relies on a posteriori observations through fitting of the free parameters. Its uncertainty range remains relatively unexplored. Common to all of these conceptual models is the fact that they were never rigorously tested for their predictive power, owing in some part to the difficulty in formulating them as testable hypotheses. We propose to explore the uncertainty ranges of these hypotheses and to translate them into testable hypotheses to be tested rigorously in the framework of the Collaboratory for the Study of Earthquake Predictability. This study will add more testability to the UCERF3 effort and help to understand and to quantify the impact of these models on seismic hazard assessment.