Stochastic Conditional Simulation of Berea Sandstone Geophysical Properties with a Counterpropagation Neural Network

See related content

by Lance Besaw,
Donna M. Rizzo,

Document Type: Proceeding Paper

Part of: World Environmental and Water Resources Congress 2008: Ahupua'A

Abstract: The ability of a counterpropagation neural network to produce stochastic conditional simulations is applied and evaluated on a real dataset. This network functions as a nonparametric clustering algorithm that circumvents traditional assumptions (i.e. normal distributions) and is well suited for assessing risk and uncertainty associated with spatially auto-correlated data. Detailed geophysical measurements from a slab of Berea sandstone are used to allow comparison with sequential indicator simulation, a traditional geostatistical method of producing conditional simulations. Equiprobable simulations and ensemble estimated fields of air permeability and compressional-wave velocity are generated using an anisotropic spatial structure extracted from a subset of observation data. Results from the counterpropagation network have been proven to be statistically similar to traditional geostatistical methods and original reference fields. This neural network's simplicity and computational speed make it well suited for environmental subsurface characterization and other earth science applications with spatially autocorrelated variables.

Subject Headings: Neural networks | Stochastic processes | Spatial data | Sandstone | Data analysis | Parameters (statistics) | Algorithms | Risk management |

Services: Buy this book/Buy this article


Return to search