| The objective of seismic exploration is to determine the location (imaging) and mechanical properties (inversion) of hydrocarbon resources in the earth using recorded data. The recorded data have a non-linear relationship with the property changes across a reflector. Current inversion methods either assume small property changes and solve a linear approximate form, or assume a non-linear relationship but use an indirect method to invert. The assumptions of the former methods are often violated in practice and can cause erroneous predictions: the latter category usually involves a significant computational effort (especially in multi-dimensional case) and/or has ambiguity issues in the predicted result.;In this dissertation, a more comprehensive multi-parameter multi-dimensional direct nonlinear inversion framework is developed based on the inverse scattering task-specific subseries (see, e.g., Weglein et al., 2003). The procedure is direct and non-linear without global searching and small-change assumptions; hence, it has the potential to provide more accurate and reliable earth property predictions for large contrast and complex targets.;As an initial part of the more general multi-dimensional direct non-linear inversion project, this dissertation focuses on the inversion for ID media and 2D experiments. Explicit direct non-linear inversion equations are derived for one and two parameter acoustic and three parameter elastic cases. The terms for imaging are separated from inversion-only terms.;Numerical tests show that non-linear inversion results provide improved estimates in comparison with the standard linear inversion. In this dissertation, we demonstrate that the direct non-linear elastic inversion in 2D requires all four components of data. However, we introduce an approach which only uses pressure measurements and approximately synthesizes the other three required components of data. Added value beyond the corresponding conventional linear results can still be achieved from pressure-only acquisition. This permits us to derive value from direct non-linear elastic inversion, when only pressure measurements are available. We anticipate further improvement when all four components of data are used.;Finally, the method is applied to time-lapse seismic data to distinguish pressure changes from reservoir fluid changes, a situation in which conventional methods have difficulty. Initial tests provide encouraging results. |
