Development of stochastic model for the Japanese Islands by combining crust deformation and seismic events

研究背景：由于地震是由地壳应力引发的，因此估算日本列岛地壳应力的三维分布对地震危险性和灾害预测至关重要。然而，测量地下10公里深处的应力是不可能的。另一种方法是建立随机模型，利用概率分布函数来解释应力分布的不确定性；该函数是通过对地壳变形和地震活动的有限测量来定义的。通过对地壳形变和地震活动测量积累的数据进行综合，得到了一套可接受的随机模型。研究目的：本研究的主要目的是建立一个储存在日本岛地壳中的应力分布的随机模型。通过对GPS测量的地壳形变和地震仪台网测量的地震活动的再现，找到了一套可接受的模型。这一套将提供一个能够对未来事件做出预测的模型。提出了一种新的随机模型分析方法，使分析所需的数值计算最小化。研究方法：首先，建立一个与实际板相匹配的物理模型。将有限的应力信息用概率函数表示，建立了随机模型。研究了在物理模型的基础上建立随机模型的合理方法。采用蒙特卡罗模拟的方法，采用大规模的计算方法，确定了一套可接受的模型，并选择了一个最合适的模型，使其成为预测地壳变形和地震活动的原型。主要成果：建立了日本列岛地壳的物理模型，该模型具有最高的空间分辨率，可以通过大规模并行计算进行分析。该物理模型被扩展为随机模型，其中地壳应力的三维分布以概率函数的形式给出。提出了一种采用新的函数离散化格式的分析方法，以便有效地分析随机模型的概率行为。利用地壳形变和地震活动的实测资料，对随机模型进行了分析。研究了用大规模数值计算分析这种随机模型的潜在用处。少

Background of research: The estimation of three-dimensional distribution of stress in the crust of the Japanese Islands is essential for earthquake hazard and disaster prediction, since an earthquake is triggered by crust stress. However, it is impossible to measure stress as deep as 10 kilometer below the ground. An alternative is the construction of a stochastic model which accounts for the uncertainty of the stress distribution by using a probabilistic distribution function; the function is defined by using limited measurement for the curst deformation and the seismic activities. A set of admissible stochastic models are found by integrating data which will be accumulated through measurement of the crust deformation and the seismic activities.Objective of research: The primary of objective of this research is the construction of a stochastic model for the stress distribution which is stored in the crust of the Japanese Island. Aset of admissible models are found by reproducing data … More for the curst deformation that is measured by GPS and for the seismic activities which are measured by seismograph network. This set will provide a certain model which is able to make prediction for future events. Anew analysis method for the stochastic model is developed, so that numerical computation that is required for the analysis is minimized.Methodology of research: First, a physical model which has the configuration that matches the actual plates is constructed. A stochastic model is constructed by using limited information for stress which are expressed in terms of a probabilistic function. Arational method for constructing the stochastic model based on the physical model is studied. A set of admissible models are determined by means of the Monte-Carlo simulation which uses large-scale computation, and a most suitable model is chosen so that it will be a prototype which produces prediction of crust deformation and seismic activities.Major results: A physical model is constructed for the crust of the Japanese Islands, and this model is of highest spatial resolution so that it is analyzed only by a large-scale parallel computation. This physical model is extended to a stochastic model in which three-dimensional distribution of stress in the crust is given as a probabilistic function. An analysis method that uses a new discretization scheme for functions is developed, so that the probabilistic behavior of the stochastic model is efficiently analyzed. The data which are measured for the crust deformation and the seismic activities are used in analyzing the stochastic model. The potential usefulness of analyzing this stochastic model with large-scale numerical computation is studied. Less