Function-Based Training Image Construction of Fluvial Point Bars: A Modern Analog Example From the Brazos River, Texas

Abstract

Point bars are difficult to model due to complex lateral and vertical heterogeneities at multiple scales. The Multi-Point Statistics (MPS) algorithm can create complex facies geometries while honoring seismic and well conditioning, but MPS requires the user to first provide a Training Image (TI), a three-dimensional conceptual model of the depositional facies that captures the complex lateral and vertical relationships between facies This paper discusses a technique to build a TI for a fluvial point bar that combines a conceptual model of point bars with data from closely spaced wells in a modern analog. This TI was then used in MPS to populate point bar facies in a 3D framework built with layers inclined parallel to bedding. The fluvial point bar TI was constructed in a horizontally-layered “sugar cube” framework where each TI layer represents a single time slice. To populate the TI, we used mathematical functions to describe a fining upward character, an upstream to downstream fining component, and vertical cyclicity with the thickest beds at the base and thinner beds at the top. Cores and LPSA data were used to define majority facies codes at different positions within the system from upstream to downstream, from bar top to bar base. To capture the trends observed in the system, these majority facies codes were used as targets for optimizing a linear equation with cell indices I, J, and K as independent variables. These equations were used to assign facies code trends across all grid cells of the TI. Four vertical fining upward cycles observed in the cores were placed into the TI using look-up functions with cell index K as the independent variable. This piece-wise linear function was developed to describe the locations of discontinuities at the boundaries of each major cycle. A correlated Gaussian noise function was used to add realistic variability to the TI. These trend functions were combined to create the lateral and vertical trends and cycle discontinuities in the TI. A rank transform function was used to impose the correct facies proportions as measured from core and wireline log data. The function-based TI was used in MPS for facies modeling in an inclined 3D layer framework. Bed and lateral accretion package geometries were informed by closely spaced lines of electrical resistivity tomography (ERT) and Ground Penetrating Radar (GPR) data. With this TI, MPS captured the trends and cyclicity accurately, although it was computationally expensive.

AAPG Datapages/Search and Discovery Article #90216 ©2015 AAPG Annual Convention and Exhibition, Denver, CO., May 31 - June 3, 2015