Scientific Investigations Report 2006–5316
U.S. GEOLOGICAL SURVEY
Scientific Investigations Report 2006–5316
Back to Table of Contents
Idaho’s largest fresh water resource, the Eastern Snake River Plain (ESRP) aquifer, occurs in the ESRP mafic volcanic province. This aquifer is composed primarily of pahoehoe basalt with intercalated sediments and minor occurrences of andesite and rhyolite. It is characterized by transmissivities and hydraulic conductivities ranging over six orders of magnitude and linear ground-water flow velocities up to 10 ft/ d (Ackerman, 1991; Anderson and others, 1999).
A large number of well borings, logging records, and cores have been collected in the vicinity of the Idaho National Laboratory (INL) for defining the geohydrologic regime in this part of the ESRP aquifer (fig. 1). Improvements in characterization of the subsurface stratigraphy have led to increasingly sophisticated conceptual and digital models of the aquifer and of ground-water flow and transport at various spatial scales. A subregional-scale flow model (about 25 × 75 mi; see fig. 1) currently is being tested and refined at the U.S. Geological Survey-INL (USGS-INL) project office. Calibration of the model has shown that assigned hydraulic conductivity values are sensitive to the relative abundance of sediment within the aquifer and that a more realistic model of the spatial variability of sediment will be required to refine the flow simulator.
This study, conducted by the USGS in cooperation with the U.S. Department of Energy, was undertaken to develop a geostatistical model of sediment abundance in the aquifer that can be used to constrain future ground-water flow model calibrations.
The spatial model of sediment abundance is based on borehole data compiled by Anderson and others (1996) in a region of the ESRP corresponding to the geographic extent of the ground-water model (fig. 1). The sediment model was created in a Geographic Information Systems (GIS)-compliant format that can interface with the ground-water flow model on a cell-by-cell basis, in specific depth-discretized layers corresponding to the digital flow model. The relative uncertainty of the sediment abundance estimates also had to be computed on a cell-by-cell basis because the degree of borehole coverage over the model domain is relatively sparse and highly variable. Geostatistical methods were employed to analyze the borehole data to select and justify an appropriate estimation method (in this case, multiple indicator kriging) and to quantify the relative uncertainty of the resulting estimates. Finally, the sediment model was used to demonstrate an approach by which sediment abundances could be used to develop a refined calibration process for the ground-water model.
Hundreds of boreholes have been drilled in and near the INL during the past 50 years for purposes of characterizing the ESRP aquifer and monitoring chemical and radioactive wastes in the subsurface. The lithologies encountered in most of these boreholes have been characterized using geophysical logging techniques; less than 10 percent of all boreholes have been cored and selectively analyzed for properties such as geochronology, paleomagnetism, and bulk geochemistry in an ongoing attempt to characterize subsurface architecture. Systematic interpretation of these and other data sets has led to the development of a stratigraphic framework (for example, Anderson and others, 1996) and conceptual models of the aquifer (Lindholm and Vaccaro, 1988; Ackerman and others, 2006), without which the quantitative analysis undertaken in this study would not be possible.
Walker (1974) conducted one of the earliest stratigraphic studies of the INL’s subsurface using a combination of geologic and geophysical logs. His work was followed by another at the Radioactive Waste Management Complex (RWMC) in which sedimentary interbeds and individual basalt-flow groups (defined as packages of flows derived from a single shield volcano) were differentiated using natural-gamma logs (Barraclough and others, 1976). The first quantitative stratigraphic study using continuous cores was conducted at the RWMC using Potassium-Argon (K-Ar) age dating, paleomagnetic polarity and inclination measurements, and petrographic characterization (Kuntz and others, 1980). Similar studies in other areas have used 40Argon/39Argon and K-Ar ages (Champion and others, 1988; Lanphere and others, 1993, 1994; Champion and Lanphere, 1997; Champion and Herman, 2003). Detailed, small-scale stratigraphic interpretations based on basalt chemistry have been described by Reed and others (1997), Scarberry (2003), and Grimm Chadwick (2004) for selected coreholes.
The work of Anderson and coworkers (Anderson and Lewis, 1989; Anderson, 1991; Anderson and Bowers, 1995; Anderson and Bartholomay, 1995) created the first and most carefully documented and vetted compilation of subsurface lithologic variability and stratigraphic correlations across the INL. Starting at selected INL facilities, their work with natural-gamma log records that had been calibrated to available core properties, led them to eventually describe and compile a subsurface stratigraphic database of 333 boreholes across the INL and vicinity (Anderson and others, 1996; Anderson and Liszewski, 1997). This database constitutes the basis of this study; the geographic locations of its boreholes are shown in figure 1.
The database compiled by Anderson and others (1996) indicates that boreholes that penetrate the unsaturated zone of the aquifer have penetrated at least 178 identified basalt-flow groups, 103 sedimentary interbeds, 6 andesite-flow groups, and 4 rhyolite domes. Anderson and Liszewski (1997) first proposed a classification based on groups of units of similar age that was intended to show the principal stratigraphic features in the subsurface. They grouped individual basalt and sedimentary units into 14 composite stratigraphic units, each made up of 5 to 90 individual units.
Table 1 summarizes these composite stratigraphic units. Composite unit 1, the youngest, is made up of 78 basalt-flow groups and 12 sedimentary interbeds. Composite unit 14, the oldest, is made up of only four identified basalt-flow groups and one sedimentary interbed. The decreasing number of individual units in successively older composite units may be due to larger and less-frequent volcanic eruptions as well as to the decreasing availability of borehole data with depth on which to base stratigraphic correlations.
Composite units 1 through 7 represent time spans of approximately 65 to 248 thousand years each and, collectively, 800 thousand years. Composite units 1, 2, and 3 generally comprise most of the unsaturated zone and only units 4 and older influence saturated ground-water flow in the aquifer. Composite units 8 through 14 span an age range of about 800 thousand to 1.8 million years and are grouped together as one aggregated composite unit. As shown by Anderson and others (1996) and by subsequent work, these older units occur beneath most of the INL at considerable depths below the water table. In the northeastern part of the INL, however, where a stratigraphic discontinuity exposes the older units at the surface, they make up most or all of the unsaturated zone.
The approximate locus of the stratigraphic discontinuity is shown in figure 2, and its subsurface character is shown in figure 3, together with the nomenclature used to identify the layers that discretize the aquifer in the subregional ground-water flow model. The water table in figure 3 reflects hydrologic conditions in 1996, at which time composite unit 5 was generally below the water table south of the discontinuity, whereas north of it, only composite units 12 and older coincided with saturated conditions (Anderson and Liszewski, 1997). Despite recent drought conditions, a similar geohydrologic situation to that shown in figure 3 prevailed in 2004.
Inferring spatial continuity is one of the principal goals of geostatistics. Spatial autocorrelation is a statistical property that characterizes the statistical similarity of measurements separated by various distances. Various types of autocorrelation statistics known as variogram statistics can be used to describe the degree of similarity in space as well as to constrain geostatistical predictive models to make optimal estimates at unsampled locations (as well as quantify estimation uncertainty). Previous geostatistical work at the INL has demonstrated that subsurface lithologic properties in the ESRP aquifer are characterized by moderate to strong spatial autocorrelation, with greater statistical similarity over short distances that diminishes with distance. This fact can be exploited through various geostatistical modeling techniques (for example, kriging) to make best unbiased estimates of subsurface properties at unsampled locations. Previous geostatistical analysis of borehole data at the INL has been conducted with one of two general aims: (1) characterizing the nature of subsurface hydrologic variability (Welhan and Reed, 1997); and (2) modeling spatial variability to predict subsurface lithologic or hydrologic properties (Welhan and others, 2002a; Leecaster, 2002, 2004).
In previous geostatistical modeling at the INL, different approaches have been used to achieve different goals. Welhan and others (2002a) used a stochastic indicator simulator to model the spatial configuration of sediments and basalt interflow zones in the aquifer beneath the Test Area North, Leecaster (2002, 2004) applied ordinary kriging to predict sedimentary interbed thicknesses and subsurface hydraulic properties of the aquifer in an area beneath the RWMC, and Gego and others (2002) used sequential Gaussian simulation to cast particle transport and breakthrough times in a stochastic context and demonstrate the application of probabilistic (Monte Carlo) ground-water modeling methods.
Back to Table of Contents