AmericanChemicalSociety.com

The goal of this project is to describe the distributions and possible causes of apparent groupings of channel-belt sand bodies in natural and experimental alluvial basins. Clustering of sand bodies is a prominent feature of latest Cretaceous deposits in much of Wyoming, and has been observed elsewhere as well. Spatial groupings of similar character have been observed in fluvial experiments performed at the St. Anthony Fall Laboratory (University of Minnesota) with no changes in boundary conditions. These results suggest that extrinsic forcing is not a requirement for clustering and that such features may be an example of natural self-organization.

Much of the effort over the past year has involved developing statistical approaches to analyzing clustering as seen in one-dimensional lithologic logs as proxied by subsurface gamma-ray well log data. This analysis is forming part of the basis of a Ph.D. study by Mr. Peter Marcy of the Department of Statistics at the University of Wyoming. In a preliminary study two sites within in the Lance Formation (Late Cretaceous) from the Wind River Basin were analyzed in order to evaluate the degree of local versus regional consistency of sand body thickness and spacing as seen in vertical successions. Each site included four to six wells located within a short distance of each other.

Most early effort was spent in finding a suitable way to convert continuous data from well wire lines to binary data (i.e. sandstone vs. shale) for statistical analysis. Long sections (c. 900 m) were broken into 150 m intervals, each of which summary and descriptive statistics were calculated. These results were than compared within and between wells in order to assess similarity spatially and with depth. We found that in no case were sand body distributions random. However, spatial proximity did not consistently explain trends observed. In some cases wells far apart had similar summary statistics (median/max thicknesses for both shale and sandstone). Wells closer together, although in many cases having similar characteristics, at times possessed drastically different summary statistics.  

These surprising results have inspired us to delve more deeply into a subfield of statistics called Analysis of Object Data. In addition, we have begun a sensitivity analysis whereby we assess the statistical impact of setting different cut-off values for autopicking of sandstone and shale units from well logs. This work will form part of Mr. Marcy’s Ph.D. study.

Additionally work at the St. Anthony Falls Laboratory (University of Minnesota) focused on mapping channels in strike sections of previous experiments as well as work modeling the degree to which channelization and clustering influences flow-path variability in reservoirs. Channel deposits in the experimental run were mapped by undergraduate researcher Shawn Wieland. Preliminary analysis on these deposits shows relationships between channel thickness, number of stories (avulsion reoccupations), and story thickness. Work evaluating the spatial distributions of channels and channel properties is ongoing.

We have begun to evaluate the effect of channel clustering on fluid flow paths in collaboration with Dr. Martin Saar (University of Minnesota) and Dr. Stuart Walsh (Cornell University). Flow simulations were conducted on cross-sectional panels from experimental deposits. Panels ranged in percentage of sand (permeable) and coal (impermeable) and represented different degrees of channelization and clustering (ranging from completely channelized to no channels). Preliminary results suggest that highly channelized intervals (e.g., panels from within clustered sections) show many flow paths across the panels, and a wide distribution of flow velocities, with many slow flow paths and few fast flow paths. In contrast, panels containing no channelization or a mix of channelization and planar bedding (e.g., intervals between clusters and intervals on the margins of clusters, respectively) tend to be dominated by relatively few, fast flow paths. This work suggests that there may be consistent relationships between flow path speed and variability in highly channelized reservoir intervals.