ITS 2001 Proceedings, Session 7, Number 7-11 795 Complex analysis of ocean tsunami observation data for solution of the inverse problem Alexander V. Avdeev 1 , Mikhail M. Lavrentiev, Jr. 2 , Andrei G. Marchuk 1 , Elder V. Goryunov 1 , Konstantin V. Simonov 3 , and Viktor A. Okhonin 4 1 Institute of Computational Mathematics and Mathematical Geophysics, Russian Academy of the Sciences, Novosibirsk, Russia 2 Sobolev Institute of Mathematics, Novosibirsk, Russia 3 Institute of Mathematical Modeling, Krasnoyarsk, Russia 4 Institute of Biophysics, Krasnoyarsk, Russia Abstract. A special system for mareogram processing is proposed. Such a system is based on two different approaches, namely, neural network technique, and inverse problems. By using two alternative methods, it is possible to achieve better accuracy in determining space parameters of a tsunami source. The above mentioned approaches are described in the paper. Model numerical tests, processed over the realistic depth profile, are then demonstrated. 1. Introduction In this paper the inverse tsunami problem is considered, i.e., evaluation of parameters of a tsunami source based on data of observations of waves in the open ocean or on the shore. It is known that the solution of this inverse problem requires that during the analysis of mareograms the trend of the trace of wave propagation and the trend of the source of perturbation be extracted from tsunami records. For construction of the trace function and the source function based on tsunami registration data, various methods can be used, such as analytic solution of the direct and inverse problems, numerical methods of modeling of excitation, and propagation of tsunami waves. In this work a number of new approaches are proposed. They are based on complex optimization of the observation system, determination of the space distribution of tsunami source through the corresponding inverse prob- lem, and nonlinear multiparameter regression analysis (neural-network tech- nology) of tsunami wave records during which the trace function and the sought source function are reconstructed. The informational-computational technology being created here is uni- versal for the purposes of analysis of tsunami registration data. In particular, the optimization methods here applied allow one to solve the following par- ticular tasks in the tsunami related problems: to compress measured data for effective transmission and storage of 1 Institute of Computational Mathematics and Mathematical Geophysics, SB RAS, 630090, Novosibirsk, Russia (avdeev@omzg.sscc.ru, mag@omzg.sscc.ru, elder@nmsf.sscc.ru) 2 Sobolev Institute of Mathematics, Novosibirsk, Russia (mmlavr@nsu.ru) 3 Institute of Mathematical Modeling, Krasnoyarsk, Russia (lena@cc.krascience.rssi.ru) 4 Institute of Biophysics, Krasnoyarsk, Russia (noogen@krasu.ru)