Datapages, Inc.Print this page

Abstract: Estimating Accuracies and Uncertainties of Quantitative Stratigraphic Predictions

Margaret A. Lessenger, Timothy A. Cross

Inverse modeling experiments with a synthetic stratigraphic cross section that resembles the Mesa Verde Group of the San Juan basin, Colorado, reveal strategies for predicting stratigraphic architecture and facies distributions from limited data. Values of stratigraphic process parameters and responses are completely known for the synthetic cross section. By sampling the cross section with variable numbers and positions of "pseudowells" and by varying the accuracy and resolution of stratigraphic correlations, it is possible to evaluate how well inverse stratigraphic modeling techniques will estimate process parameters of real sedimentary basins.

The inverse model accurately recognizes the relative contributions of different stratigraphic processes to observed stratigraphy. Short-term eustasy strongly affects sediment volume partitioning, cycle symmetry and frequency of unconformity and hiatal surfaces more than other process parameters. Tectonic movement and lithosphere strength affect stacking patterns of progradational/aggradational units, large-scale symmetry of stratigraphic sequences and basin shape. Sediment supply changes aspect ratios (thickness:width) and volumes of sediment within facies tracts, but does not cause sediment volume partitioning or affect stacking patterns. These experiments refute the common assumption that a change in the value of one process parameter can be fully compensated by an opposite change i value of another parameter to produce an identical stratigraphic response.

Accuracy and resolution of stratigraphic correlations affect the accuracy of the inversion, but in different ways. Inversions based on low resolution correlations tend to underestimate tectonic subsidence, and are poor estimators of eustasy. If high-resolution correlations are less certain than low-resolution correlations, it is better to reduce the resolution of correlations, than to entirely miscorrelate. For example, an inversion using a twofold division of the Mesa Verde across all facies tracts produces a cross section very close to the truth, whereas an inversion using three, incorrectly correlated divisions of the Mesa Verde produces a very poor result.

Paleogeographic and spatial positions of wells affect the ability of the inversion to accurately estimate tectonic subsidence and stratigraphic architecture. Widely spaced wells give more information about tectonic subsidence than closely spaced wells. Inversion of a single well is nearly equivalent to inversion of multiple, closely-space wells in estimating the geometry, thicknesses and lateral extents of facies tracts.

These experiments are teaching us how variations in information quality and data types affect the accuracy and uncertainty of quantitative stratigraphic prediction. This knowledge will assist decision-making by framing questions such as: "How much and what type of additional data must I collect to improve the accuracy of stratigraphic prediction by X%?"; "Would the addition of seismic data provide more or less value than additional well information in select locations?"; "Is high-resolution correlation as important at the scale of exploration as it is for reservoir characterization?"

AAPG Search and Discovery Article #90956©1995 AAPG International Convention and Exposition Meeting, Nice, France