Paper No. 8
Presentation Time: 10:45 AM
USE OF OUTCROP ANALOGS TO ANALYZE LITHOFACIES AND HYDROFACIES DISTRIBUTIONS WITHIN BORDEN AQUIFER SEDIMENTS, ONTARIO, CA
Heterogeneity is commonly invoked as a dominant control on contaminant dispersion in aquifers; however our ability to reasonably model and understand the geometry and distribution of heterogeneities in the subsurface is limited. In an attempt to more accurately characterize these properties, an outcrop analog approach utilizing terrestrial lidar and high-resolution digital photography is being used to map and model facies variability in Borden aquifer sediments. In July 2010, we excavated 15 sub-vertical faces (each approximately 20 x 1.5m in size, totaling approximately 40m width and 10m height) at a sand quarry in the Borden aquifer sediments located approximately 2km from the Stanford-Waterloo experimental site. These sections have similar lithofacies as observed in core from the nearby aquifer studies. Terrestrial lidar scans and high-resolution photographs were acquired and facies distributions mapped in the field onto the photographs. In addition to field mapping, ERDAS Imagine v. 2010 was used to segment geometrically calibrated digital photos into distinct textural classes in a semi-automated approach dependent on surface heterogeneity and geometry. A zero-sum convolution filter was used to delineate bounding surfaces and enhance distinction of contorted bedding, and a standard deviation based focal analysis enhanced the texture of gravels. The first principal components of the filtered images were then combined into a new image which was classified using the Advanced RGB Clustering algorithm. Finally, the classified image was projected onto a lidar-derived 3D surface to obtain the spatial distribution of facies and textural classes. We assign hydraulic conductivity values from earlier studies to the textural/facies classes, thus allowing us to produce 3D heterogeneous groundwater models that capture realistic geometries and facies distributions. These data are also used as training images for geostatistical models.