| Popis: |
This paper illustrates a workflow that uses rock physics in a probabilistic framework to drive a geologically constrained classification of a seismic dataset. Rock physics is used to diagnose the reservoir sands and then build an interpretational scheme (i.e., a rock physics template) in an appropriate elastic domain to classify seismic volumes (e.g., produced by an elastic inversion); it also allows to model possible variations of the reservoir properties that might occur away from the wells, thus extending the training dataset used to drive the classification. This classification is based on a probabilistic framework which uses the Mahalanobis distance’s criterium to separate different classes; an optional step to impose a realistic vertical distribution, based on a 1D implementation of a Markov chain Monte Carlo, may be also applied. The results are finally evaluated with a simple single-value coefficient based on a statistical technique called contingency matrix. The methodology is tested on a seismic dataset around two wells targeting Pleistocene turbiditic reservoirs in the offshore Nile Delta region, and the results show that certain association of lithologies and fluids are correctly identified, in accordance with the rock physics templates previously defined. |
