The aim of this thesis is to obtain crustal model through the inversion of deep seismic refraction data, considering lateraly homogenous horizontal plain layers over a half-space. The direct model is given by analitic expression for the travel-time curve, as a function that depends on the source-station distance and on the array of parameters, formed by velocity and thickness of each layer. The expression is obtained from the trajectory of the seismic ray by Snell's Law. The calculation of the arrivel time for seismic refraction by this method, takes into account a model with velocities increasing with depth. The occurrence of low velocity layers (LVL) are solved as a model reparametrization, taking into account the fact that top boundary of the low velocity layer is only a reflector, and not a refractor of seismic waves. The inversion method is used to solve for the possible solutions, and also to perform an analysis about the ambiguity of the problem. The search region of probable solutions is constrainted by high and lower limits of each parameter considered, and by high limits of each critical distance, calculated using the array of parameters. The inversion process used is an optimization technique for curve fitting corresponding to a direct search in the parameter space, called COMPLEX. This technique presents the advantage of using any objective function, and as being practical in obtaining diferent solutions for the problem. As the travel-time curve is a multi-function, the algorithm was adaptaded to minimize several objective funtions simultaneously, with constraints. The inversion process is formulated to obtain a representative group of solutions of the problem. Afterwards, the analysis of ambiguity is made by Q-mode factor analysis, through which is possible to find the commom properties of the group of solutions. Tests with synthetic and real data were made having as initial aproximation to the inversion process, the velocity and thickness values calculated by the straightforward visual interpretation of the seismograms. For the synthetics, it was used seismograms calculated by the refletivity method, with diferent models. For test with real data, it was used seismograms colected by the Lithospheric Seismic Profile in Britain (LISPB), in the northern region of Britain. It was verified in all tests that geometry of the model has most importance for the ambiguity of the problem, while the physical parameters present only smaller changes into the group of solutions obtained.
|
