Publications

Two-dimensional, heterogeneous, spatially correlated models of thermal conductivity and bulk density have been created for a representative, east-west cross section of Yucca Mountain, Nevada, using geostatistical simulation. The thermal conductivity models are derived from spatially correlated, surrogate material-property models of porosity, through a multiple linear-regression equation, which expresses thermal conductivity as a function of porosity and initial temperature and saturation. Bulk-density values were obtained through a similar, linear-regression relationship with porosity. The use of a surrogate-property allows the use of spatially much-more-abundant porosity measurements to condition the simulations. Modeling was conducted in stratigraphic coordinates to represent original depositional continuity of material properties and the completed models were transformed to real-world coordinates to capture present-day tectonic tilting and faulting of the material-property units. Spatial correlation lengths required for geostatistical modeling were assumed, but are based on the results of previous transect-sampling and geostatistical-modeling work.

Indicator geostatistical techniques have been used to produce a number of fully three-dimensional stochastic simulations of large-scale lithologic categories at the Yucca Mountain site. Each realization reproduces the available drill hole data used to condition the simulation. Information is propagated away from each point of observation in accordance with a mathematical model of spatial continuity inferred through soft data taken from published geologic cross sections. Variations among the simulated models collectively represent uncertainty in the lithology at unsampled locations. These stochastic models succeed in capturing many major features of welded-nonwelded lithologic framework of Yucca Mountain. However, contacts between welded and nonwelded rock types for individual simulations appear more complex than suggested by field observation, and a number of probable numerical artifacts exist in these models. Many of the apparent discrepancies between the simulated models and the general geology of Yucca Mountain represent characterization uncertainty, and can be traced to the sparse site data used to condition the simulations. Several vertical stratigraphic columns have been extracted from the three-dimensional stochastic models for use in simplified total-system performance assessment exercises. Simple, manual adjustments are required to eliminate the more obvious simulation artifacts and to impose a secondary set of deterministic geologic features on the overall stratigraphic framework provided by the indictor models.

Exhaustive characterization of a contaminated site is a physical and practical impossibility. Descriptions of the nature, extent, and level of contamination, as well as decisions regarding proposed remediation activities, must be made in a state of uncertainty based upon limited physical sampling. The probability mapping approach illustrated in this paper appears to offer site operators a reasonable, quantitative methodology for many environmental remediation decisions and allows evaluation of the risk associated with those decisions. For example, output from this approach can be used in quantitative, cost-based decision models for evaluating possible site characterization and/or remediation plans, resulting in selection of the risk-adjusted, least-cost alternative. The methodology is completely general, and the techniques are applicable to a wide variety of environmental restoration projects. The probability-mapping approach is illustrated by application to a contaminated site at the former DOE Feed Materials Production Center near Fernald, Ohio. Soil geochemical data, collected as part of the Uranium-in-Soils Integrated Demonstration Project, have been used to construct a number of geostatistical simulations of potential contamination for parcels approximately the size of a selective remediation unit (the 3-m width of a bulldozer blade). Each such simulation accurately reflects the actual measured sample values, and reproduces the univariate statistics and spatial character of the extant data. Post-processing of a large number of these equally likely statistically similar images produces maps directly showing the probability of exceeding specified levels of contamination (potential clean-up or personnel-hazard thresholds).

A set of detailed geostatistical simulations of porosity has been produced for a layered stratigraphic sequence of welded and nonwelded volcanic tuffs at Yucca Mountain, Nevada. The simulations are produced using a composite. model of spatial continuity and they are highly conditioned to abundant drill hole (core) information. A set of derivative simulations of saturated hydraulic conductivity has been produced, in the absence of conditioning data, using a cross-variable relationship developed from similar data elsewhere. The detailed simulations reproduce both the major stratigraphic units and finer scale layering indicated by the drill hole data. These simulations have been scaled up several order of magnitude to represent block-scale effective hydrologic properties suitable for use in numerical modeling of groundwater flow and transport. The upscaling process involves the reformulation of a previously reported method that iteratively adapts an initial arbitrary grid to ``homogenize`` the detailed hydraulic properties contained within the adjusted cell limits and to minimize the size of cell in highly heterogeneous regions. Although the computation of the block-effective property involves simple numerical averaging, the blocks over which these averages are computed are relatively homogeneous, which reduces the numerical difficulties involved in averaging non-additive properties, such as permeability. The entire process of simulation and upscaling is rapid and computationally efficient compared with alterative techniques. It is thus suitable for the Monte Carlo evaluation of the uncertainty in site characterization as it affects the results of groundwater flow and transport calculations.

Hydrologic properties have been measured on outcrop samples taken from a detailed, two-dimension grid covering a 1.4 km outcrop exposure of the 10-m thick non-welded-to-welded, shardy base microstratigraphic unit of the Tiva Canyon Member of the Miocene Paintbrush Tuff at Yucca Mountain, Nevada. These data allow quantification of spatial trends in rock matrix properties that exist in this important hydrologic unit. Geologic investigation, combined with statistical and geostatistical analyses of the numerical data, indicates that spatial variability of matrix properties is related to deterministic geologic processes that operated throughout the region. Linear vertical trends in hydrologic properties are strongly developed in the shardy base microstratigraphic unit, and they are more accurately modeled using the concept of a thickness-normalized stratigraphic elevation within the unit, rather than absolute elevation. Hydrologic properties appear to be correlated over distances of 0.25 to 0.3 of the unit thickness after removing the deterministic vertical trend. The use of stratigraphic elevation allows scaling of identified trends by unit thickness which may be of particular importance in a basal, topography-blanketing unit such as this one. Horizontal changes in hydrologic properties do not appear to form obvious trends within the limited lateral geographic extent of the ash-flow environment that was examined. Matrix properties appear to be correlated horizontally over distances between 100 and 400 m. The existence and quantitative description of these trends and patterns of vertical spatial continuity should increase confidence in models of hydrologic properties and groundwater flow in this area that may be constructed to support the design of a potential high-level nuclear waste repository at Yucca Mountain.

The spatial correlation structure of volcanic tuffs at and near the site of the proposed high-level nuclear waste repository at Yucca Mountain, Nevada, is estimated using samples obtained from surface outcrops and drill holes. Data are examined for four rock properties: porosity, air permeability, saturated hydraulic conductivity, and dry bulk density. Spatial continuity patterns are identified in both lateral and vertical (stratigraphic) dimensions. The data are examined for the Calico Hills tuff stratigraphic unit and also without regard for stratigraphy. Variogram models fitted to the sample data from the tuffs of Calico Hills indicate that porosity is correlated laterally over distances of up to 3000 feet. If air permeability and saturated conductivity values are viewed as semi-interchangeable for purposes of identifying spatial structure, the data suggest a maximum range of correlation of 300 to 500 feet without any obvious horizontal to vertical anisotropy. Continuity exists over vertical distances of roughly 200 feet. Similar variogram models fitted to sample data taken from vertical drill holes without regard for stratigraphy suggest that correlation exists over distances of 500 to 800 feet for each rock property examined. Spatial correlation of rock properties violates the sample-independence assumptions of classical statistics to a degree not usually acknowledged. In effect, the existence of spatial structure reduces the ``equivalent`` number of samples below the number of physical samples. This reduction in the effective sampling density has important implications for site characterization for the Yucca Mountain Project.