Comparison of Fluid Prediction Success between AVO and Bright Spot Techniques in the Marco Polo Field, the Gulf of Mexico

摘要

Bright spot amplitude anomalies in seismic data are common indicators of natural gas; however, an interpretation based purely on these amplitude anomalies often yields a false indication of gas-saturated sands. A data set from the Marco Polo field, the Gulf of Mexico, demonstrates this problem. A discovery well was drilled into a sequence of bright spot anomalies that were indeed gas-saturated sands. This suggested that other bright spots in the seismic section also corresponded to gas sands and that non-bright spots were to brine-saturated sands. Nine development wells were later drilled into those bright spots, but not all of them were gas sands and not all of non-bright spots were brine-saturated sands. This study utilized Gassmann fluid substitution and three seismic amplitude versus offset (AVO) techniques (intercept and gradient, elastic impedance, and Lambda-Mu-Rho) as a comparison to purely using bright spots technique for fluid-type prediction at the location around the discovery well. This study used borehole information only from the discovery well in the purpose of AVO calibration. Forward models for the three AVO techniques were created from the well-log information in order to predict differences in the modeled attributes between gas- and brine-saturated scenarios. Pre-stack seismic data were inverted for intercept and gradient attributes, elastic impedance (EI) volumes, and Lambda-Mu-Rho (LMR) volumes. These volumes were compared to the forward models to predict gas- and brine-saturated locations. The prediction results were evaluated with information from the nine development wells. The intercept and gradient, elastic impedance, and LMR techniques yielded correct predictions of 52%, 61%, and 70%, respectively, of the observed sands. The traditional bright spot method yielded only 45% of correct fluid prediction. In conclusion, the pre-stack AVO techniques provided a better fluid prediction than relying solely on the post-stack bright spots alone. Furthermore, the prediction results improved as the computational intensity of the inversion increased from the intercept and gradient, to the elastic impedance, and to the LMR technique.