Sandia National Laboratories - Spatial Statstics

		Spatial Statistics Projects Site Characterization for Nuclear Waste Repositories, Japanese Nuclear Cycle Development Institute

The Japanese Nuclear Cycle Development Institute (JNC) is in the process of developing site characterization techniques at two sites in Japan. One of these is the Mizunami site in central Japan (about 45 miles NE of Nagoya) in an area of granitic rock. The Mizunami site will be developed into a 1000 meter deep underground research laboratory in the next few years. Sandia National Laboratories is assisting JNC with the development and testing of site characterization techniques in support of this development.

Specifically, Sandia is working with JNC on the following site characterization activities:

Application of geostatistical techniques and stochastic modeling approaches to site characterization.
In the initial stages of site characterization, the conceptual models of the subsurface are extremely uncertain. As additional data are collected, these conceptual models can be refined and some may be eliminated. Geostatistical approaches to site characterization can be used to determine optimal locations for collection of additional data based on the application of different performance measures. For example, additional data collected to reduce uncertainty in fracture frequency may not be obtained in the same location as additional data collected to reduce uncertainty in permeability.

This figure shows the areas of maximum uncertainty (blue) in the value of the fracture permeability as determined through geostatistical modeling. The blue area near the bottom of the model corresponds to a fault that was not included in the original conceptual model and the upper region of high uncertainty corresponds to the granitic-sedimentary contact.

Other work in this area includes probabilistic predictive modeling of a series of cross-well interference tests using a non-radial flow dimension conceptual model (McKenna and Roberts, 2000).

Modeling of large-scale, fractured rock, groundwater flow systems.
When construction of the underground facility at the Mizunami site begins, pumping of large amounts of groundwater will be necessary to maintain the water table below the level of construction. Current efforts are aimed at predicting the amount of drawdown that will occur for different pumping scenarios. These predictions will be checked against a series of long-term pump tests scheduled for the end of 2001.

Sandia National Laboratories has completed the preliminary, pre-pumping, ground water flow model (McKenna et al., 2001).

This model contains over nearly 1,000,000 nodes and was solved on a cluster of 36 PC's linked into a single supercomputer. In the construction of this model, special attention was put on examining different conceptual models of the fault network in the vicinity of the site. See an animation of the 3-D fault network.

Estimation of Permeability from Geophysical and Borehole Televiewer Logs.
This is a new task (Fall 2001). JNC and Sandia will be working together to see if neural networks can be used to predict values of measured hydraulic conductivity, K, using only information collected in geophysical logs and borehole televiewer logs in the same interval as the K measurement. The relationship between K and many different geophysical measurements may be highly non-linear, but neural networks may be able to define this relationship well enough to allow for relatively accurate estimation of K prior to conducting expensive downhole measurements.

Papers:

McKenna, S.A. and C.A. Rautman, (in review), Stochastic Modeling of Fracture Frequency along a Cross-Section at the MIU Site, Tono Region, Japan, Sandia National Laboratories, Albuquerque, New Mexico, 122 pp.
McKenna, S.A. and P.C. Reeves, (in press) Fractured Continuum Approach to Stochastic Permeability Modeling, accepted for publication in: Stochastic Modeling and Geostatistics: Principles, Methods and Case Studies, 2nd ed., AAPG, Tulsa, Oklahoma, December 1999.
McKenna, S.A., M. Eliassi, K. Inaba and H. Saegusa, 2001, Steady-state Groundwater Flow Modeling of the MIU Site Area, Groundwater Flow in Discrete Fractures Symposium, Japanese Geotechnical Socitey, Tokyo, September 10-11, 14 pp.
McKenna, S.A., 2001, Probabilistic Approach to Site Characterization: MIU Site, Tono Region, Japan, SAND2000-3200, Sandia National Laboratories, Albuquerque, New Mexico, 32 pp.
McKenna, S.A. and R.M. Roberts, 2001, Predictive Modeling of MIU3-MIU2 Interference Tests: MIU Site, Toki, Japan, SAND2000-0771. Sandia National Laboratories, Albuquerque, New Mexico, 18 pp.
Makino, H., S.A. McKenna and K. Wakasugi, 2001, Sensitivity Analysis of Monte Carlo Simulation Results Using the Kolomogorov-Smirnov d Statistic, presented at: IAMG annual meeting, September, 9-12, 2001, Cancun, Mexico.