Institute of Computational Mathematics and Mathematical Geophysics


Международная Конференция по Математическим Методам в Геофизике «ММГ-2003»

Россия, Новосибирск, Академгородок 8-12 октября 2003 г.

Тезисы докладов


Стендовые доклады (секция 1)

Трехмерная нелинейная сейсмическая томография без трассирования лучей: одновременное восстановление скоростной структуры и гипоцентров землетрясений.

Кабанник А.В., Чеверда, В.А.

Геофизическая служба СО РАН (Новосибирск)

The paper deals with the classical statement of seismic tomography -- one should refine the background model of P(S)-wave propagation by the travel-time residuals for the given set of earthquakes observed from some irregular set of points on the Earth surface (station positions), and determine earthquake hypocenters simultaneously. The formal splitting procedure was used to solve velocity and hypocenter parts of the system of linear equations separately at each iteration of nonlinear process (Pavlis & Booker, 1980). As we are going to deal with 3D Earth model the routine approach to the problem claims a huge amount of RAM in order to provide storage of the matrices. In order to avoid this the LSQR approach to resolve the problem is proposed and implemented.

The advantage is that now one should compute action of the tomographic operator and its adjoint onto the given vector only on each step of the iterative process. In order to do this one should resolve the system of linear ordinary differential equations of the first order and does not claim ray-tracing at all. The operators are calculated using the timefield in reference model, which was calculated by finite-difference eikonal equation solving technique (Podvin & Lecompte, 1991).

In spite of the fact that our method requires much less computational resources than traditional inverse approaches the real data inversion can take a long time. The computation time is strongly decreased using parallel computations. Indeed, the computation of the operator′s action for each seismic source is independent from other sources. Therefore we can distribute the independent tasks of the operators′ calculation between CPU′s in parallel computer. Synthetic data and local earthquake data from Altai-Sayan seismological stations network inversions were performed on MVS1000/M parallel supercomputer in Siberian Supercomputer Center.

References

Podvin, P, and I. Lecompte, Finite difference computation of traveltimes in verycontrasted velocity models: A massively parallel approach and its associatedtools, Geophys. J. Int., 106, 217-284, 1991 Pavlis, G.L., and Booker J.R., The mixed discrete-continuous inverse problem: Application to the simultaneous determination of earthquake hypocenters and velocity structure, J. Geophys. Res., 85(B9), 4801-4810, 1980



Ваши комментарии
Обратная связь
[ICT SBRAS]
[Головная страница]
[Конференции]

© 1996-2000, Институт вычислительных технологий СО РАН, Новосибирск
© 1996-2000, Сибирское отделение Российской академии наук, Новосибирск
    Дата последней модификации: 06-Jul-2012 (11:52:46)