Real data assimilation for optimization of frictional parameters and prediction of afterslip in the 2003 Tokachi-oki earthquake inferred from slip velocity by an adjoint method
Real data assimilation for optimization of frictional parameters and prediction of afterslip in the 2003 Tokachi-oki earthquake inferred from slip velocity by an adjoint method
复制标题
2003年十胜冲地震摩擦参数优化和后滑预测的真实数据同化，通过伴随方法从滑移速度推断
DOI：
10.1093/gji/ggv289
复制
发表时间：
2015
影响因子：
2.8
通讯作者：
Kazuro Hirahara
中科院分区：
文献类型：
--
作者：
Masayuki Kano;Shin’ichi Miyazaki;Yoichi Ishikawa;Yoshihisa Hiyoshi;Kosuke Ito; Kazuro Hirahara
数据同化是在模型动力学约束下优化数值模型参数的一种技术，它能更好地拟合观测值。优化后的参数可以用于数值模型的后续预测，与没有经过数据同化优化的结果相比，预测的物理变量至少更接近于未来的观测值。在这项工作中，开发了一个伴随数据同化系统，用于优化在滑后期相对大量的空间非均匀摩擦参数，其中物理约束是准动态运动方程和实验室推导的速率和状态相关的摩擦定律，该定律描述了俯冲带滑动速度的时间演变。观测到的变量是在板块界面上估计的滑移速度。在将该方法应用于2003年多立树地震余震的实际资料同化之前，进行了综合资料同化实验，验证了优化余震区摩擦参数的可行性。如果观测结果能够捕捉到板界面滑移的加速度和衰减相位，则当前系统能够采用基于数值模型的物理约束来优化摩擦参数sa - b， a&l。另一方面，在余震振幅小于1.0 cm d−1的区域，不太可能约束摩擦参数。其次，对2003年德立树地震的实际资料进行同化，将全球导航卫星系统时间序列的时变反演得到的滑动速度数据纳入其中。a - b、aandl0的优化值分别为0 （10kpa）、0 （102kPa）和0 （10mm）。优化后的摩擦参数与观测值拟合较好，滑移速度预测能力较好。进一步的实验也表明了采用细网格模型的重要性。这将有助于进一步了解板块界面上的摩擦特性，并有助于建立预报系统，为后续地震的可能性提供有用的信息。
Data assimilation is a technique that optimizes the parameters used in a numerical model with a constraint of model dynamics achieving the better fit to observations. Optimized parameters can be utilized for the subsequent prediction with a numerical model and predicted physical variables are presumably closer to observations that will be available in the future, at least, comparing to those obtained without the optimization through data assimilation. In this work, an adjoint data assimilation system is developed for optimizing a relatively large number of spatially inhomogeneous frictional parameters during the afterslip period in which the physical constraints are a quasi-dynamic equation of motion and a laboratory derived rate and state dependent friction law that describe the temporal evolution of slip velocity at subduction zones. The observed variable is estimated slip velocity on the plate interface. Before applying this method to the real data assimilation for the afterslip of the 2003 Tokachi-oki earthquake, a synthetic data assimilation experiment is conducted to examine the feasibility of optimizing the frictional parameters in the afterslip area. It is confirmed that the current system is capable of optimizing the frictional parametersA–B,AandLby adopting the physical constraint based on a numerical model if observations capture the acceleration and decaying phases of slip on the plate interface. On the other hand, it is unlikely to constrain the frictional parameters in the region where the amplitude of afterslip is less than 1.0 cm d−1. Next, real data assimilation for the 2003 Tokachi-oki earthquake is conducted to incorporate slip velocity data inferred from time dependent inversion of Global Navigation Satellite System time-series. The optimized values ofA–B,AandLare O(10 kPa), O(102kPa) and O(10 mm), respectively. The optimized frictional parameters yield the better fit to the observations and the better prediction skill of slip velocity afterwards. Also, further experiment shows the importance of employing a fine-mesh model. It will contribute to the further understanding of the frictional properties on plate interfaces and lead to the forecasting system that provides useful information on the possibility of consequent earthquakes.