| Popis: |
In this work we extend a previously developed (Mazzotti & Zamboni, 2003) non-linear method of petrophysical AVA inversion to the linear case to reduce the computation time and to make feasible the inversion of large data volumes. At first a linear regression analysis of the well log data is performed and a set of petrophysical equations are calculated; a new parameterization of reflection coefficients is then obtained through the insertion of the petrophysical equations into the linear expression of Aki's reflection coefficient. Afterwards, in order to take into account the a priori well log information, a linear kernel is implemented in a least square inversion algorithm with probabilistic boundaries. In fact, the probabilistic boundaries limit the variability range of petrophysical parameters and they reduce the uncertainty of the inverse problem solution. The results of this approach provide the same accuracy as the non-linear one either on synthetic data and on real data, relative to a case where porosity and saturation of gassands are the petrophysical properties that produce a noticeable influence on the AVA response. The AVA inversion of synthetic data demonstrates that the porosity of gas-sand is the best solved parameter, followed by its saturation and that its resolution is strictly influenced by the level and the type of noise. The AVA inversion of real data confirms the systematic studies made on synthetic data and it underlines that gassand porosity is the best solved parameter, and that, in presence of high amplitude noise and limited angle range, sand saturation in the range 0% to 97% is an unresolved parameter. |
