Abstract

Lawrence Livermore National Laboratory (LLNL) (Fig. 1) is in the final stages of the Superfund decisionmaking process for site remediation and restoration. In the process of characterizing the subsurface of the LLNL site, we have developed unique methods of collecting, storing, retrieving, and imaging geologic and chemical data from more than 350 drill holes. The lateral and vertical continuity of subsurface paleostream channels were mapped for the entire LLNL site using geologic descriptions from core samples, cuttings, and interpretations from geophysical logs. A computer-aided design and drafting program, SLICE, written at LLNL, was used to create two-dimensional maps of subsurface sediments, and state-of-the art software produced three- dimensional images of the volatile organic compound (VOC) plumes using data from water and core fluid analyses.

Introduction

The Lawrence Livermore National Laboratory (LLNL) is a multiprogram national laboratory. Its primary mission is nuclear weapons research and development. However, over the past 20 years LLNL has diversified into other fields such as magnetic and laser fusion, biomedical and environmental sciences, and, in the past decade, environmental remediation and restoration. LLNL has drilled over 350 monitor wells in order to characterize the subsurface. Characterization data from these wells, which consist of water and soil chemical and geochemical analyses, water levels, and geologic data, are stored on a main frame computer and are accessed via a data retrieval system. The Livermore Valley is relatively flat, underlain by a complex alluvial sedimentary basin drained by two intermittent streams. The subsurface consists of unconsolidated sand, silt, clay, and gravel with multiple, poorly connected zones of high permeability. The hydrogeologic system is characterized as leaky, with horizontal hydraulic communication of up to 80 ft and a vertical connection of nearly 50 ft. Contaminants at LLNL were initially discovered in the ground water by LLNL in 1983, and LLNL was added to the U.S. Environmental Protection AgencyÕs (EPA) National Priority List (Superfund) in 1987. The contaminants at LLNL are primarily in the ground water and are predominantly composed of the chlorinated volatile organic compounds (VOCs) trichloroethylene and perchloroethylene. In addition, LLNL has identified isolated areas where the subsurface sediments were contaminated with low levels of mixed waste (VOCs and tritium). over the last 40 years, the VOCs have migrated naturally in the subsurface, dispersing over an area of approximately 1.3 mi2 and to a depth of approximately 200 ft. State-of-the-art imaging tools were necessary to visualize the lateral and vertical continuity of the paleostream channels. A computer-aided design and drafting (CADD) program, SLICE, was used to create two-dimensional maps of channels in the subsurface, and advanced, commercial, three- dimensional modeling packages were used to model and calculate volumes of the contaminant plumes. Future plans include producing animated programs of flow and transport of VOCs.

Data Collection and Retrieval

Information Model

Most of the data collected as part of the environmental cleanup and monitoring activities at LLNL are stored in a central INGRES relational database. A highly simplified version of the information model used to structure this database is given in Fig. 2. The starting point of the model is the sampling location (of which there are currently almost 5,000), which can be anything from an air-monitoring device, to a location where storm-drain runoff is tested, to the more common subsurface boreholes and ground water monitoring wells. The minimum information stored about a location are its well number, type, and x and y coordinates. Subsurface locations, such as boreholes and monitoring wells, usually have one or more depth-below-surface z coordinates. Boreholes have several depths associated with them because repeated soil samples are taken as the holes are drilled. Since 1983, 60,000 air, soil, and water samples have been taken from the various sampling locations. Some locations, particularly the ground water monitoring wells, are sampled on a quarterly basis, introducing a fourth dimension of time. Each sample, in turn, has been analyzed either in the field or by certified laboratories for several distinct chemical and physical properties, leading to a total of over 1 million measurements of everything from concentrations of VOCs to temperature and hydraulic conductivity. The last type of information in Fig. 2 is the lithology (i.e., sediment type) encountered as a borehole is drilled. Because these data are drawn from observations made by geologists in the field rather than laboratory analytical procedures, they are kept separate from the air/vegetation/soil/ water sample data.

Retrieval for Imaging

The data most commonly retrieved for imaging purposes include soil and water chemistry and lithology. Because the model was designed to economically store information rather than to suit the needs of any single application of the data, all three types of retrieval require some degree of processing and reformatting. Until recently, this process was handled in two steps. First, a Structured Query Language (SQL) query was generated to retrieve the required coordinates for the locations of interest and the desired attribute (lithology, total VOCs in soil, tritium in water, etc.) and to generate an ASCII file. Second, a data reformatting program was written to prepare the data for input to the imaging tool. The preparation could be as simple as averaging over laboratory duplicates or as complex as merging soil chemistry, water chemistry, and lithology data from the same borehole. The result was a second ASCII file suitable for loading into the imaging tool. The drawback to this process was its dependence on database and programming specialists to effect the transfer of data from the database to the tools.

As part of an effort to give consumers more direct access to data and control over their uses of it, an information flow manager program, known as the INTEGRATOR, has been put into service to replace the steps just described. This interactive system, developed at LLNL under the auspices of the Treaty Verification Program, is designed to: 1) provide a graphical interface to the database so that queries can be generated without knowledge of SQL; 2) route information between the database and external applications, such as statistical packages, imaging tools, or any other software of interest, without the user having to be familiar with the required formatting conventions; and 3) store and manage unconventional forms of data, such as documents, images, and personal notes, so that they appear to be part of the original database. A simple schematic representation of the INTEGRATOR is given in Fig. 3. It can best be thought of as an octopus, which extends tentacles to (i.e., interfaces with) other packages. Once an interface is built, customized procedures for data transfer and processing can be quickly created and turned over to users for routine operation and minor adjustments.

Imaging Tools

SLICE Program

The SLICE program, developed by Robert McPherrin (LLNL), constructs pseudo three-dimensional images from information stored in the geologic data base. SLICE was originally developed in ComputervisionÕs CVMACTM language and runs on a main frame computer. The program was recently ported to AutoCADÕsTM Autolisp language. Thus, it can be used on a variety of platforms.

The SLICE program is made up of three modules: 1) SLICE (cutting plane and statistical), 2) cross section, and 3) fence diagram. The SLICE cutting plane module first calculates a planer slice through the subsurface at a specified angle (dip) from any depth in a drill hole. Then, the program draws the planer surface as it intersects the drill hole, color codes and numbers the soil types, and gives the depth in the well and thickness of the geologic unit that was sliced through. Finally, SLICE stacks the series of two-dimensional depth slices, creating a pseudo three-dimensional planer view of the subsurface. The geologist can then correlate the well intersects to develop a conceptual model of the channel structure.

The SLICE programÕs statistical analysis section performs the same function as the cutting plan section; however, a pie chart or histogram can show the percentage of sand, silt, clay, or gravel for any depth interval.

Geologists commonly use the cross section and fence diagram module to correlate the soil types between wells. The cross section shows a vertical view of a series of wells. The different soil types are shown as various colors: the blues are course-grained sediments and the yellows and reds are finer grained. Because color images were not used in this paper, the gray scale figures may not show the kind of detail normally shown in color images. Fig. 4 shows an example of a geologic cross section in gray scale.

Three-Dimensional Imaging

A state-of-the-art, three-dimensional software package created by Dynamic Graphics, Inc., is used to produce images of the contaminant plumes and the subsurface geologic structure (channels). The software is issued in two modules: Interactive Surface ModelingTM (ISM) and Interactive Volume modelingTM (IVM).

The ISM program provides tools for automated surface modeling, mapping, and analysis. The program uses data files entered as either columns and rows or as irregularly spaced scattered data to calculate a two-dimensional grid file. From these two-dimensional grids, surface contour maps, mesh perspective views, cross sections, and fence diagrams are produced.

IVM models, displays, and analyzes properties or characteristics that vary continuously in three-dimensional space. Computer correlationÕs in the three- dimensional model are made using an iterative three-dimensional, minimum- tension spline algorithm designed to honor data closely while providing a smooth, reasonable interpolation away from input data points. The resulting grids can be manipulated using various mathematical grid operations to get the obtain the desired grid model. A three-dimensional grid display model is generated from the three-dimensional grid file. Examples of these models are shown in Figs. 5 and 6. To change the display perspective, the display models can be interactively manipulated on screen by changing such operations as the azimuth and the view point. The display models are usually shown in a variety of colors to show the distinctive property values. For this paper, the images are shown in shades of gray, which may be difficult to interpret.

The ISM software package can be run on a variety of mainframe platforms: VAXTM, Sun Sparc station, etc.; however, IVM can only be run on the Silicon Graphics, Inc., series of computers.

Geologic Imaging

The geologic imaging was accomplished by assigning numerical values to 16 sediment types. The soil (sediment) types identified in our study area ranged from clean gravel to clay. The following values were given to soils: 1) gravel, 2) sandy gravel, 3) silty gravel, 4) clayey gravel, 5) gravelly sand, 6) gravelly silt, 7) gravelly clay, 8) sand, 9) silty sand, 10) clayey sand, 11) sandy silt, 12) silt, 13) clayey silt, 14) sandy clay, 15) silty clay, and 16) clay.

Using IVM, we gridded the data according to the numerical sediment types. Fig. 5 is an example of a three- dimensional model of the LLNL site geology. The coarse-grained sediments, represented by gravel through silty sand (Nos. 1Š9), were combined into the blue-through-gray scales, and the clayey sands through clay (Nos. 10Š16) were given yellow through red. For this paper (shades of gray), the darker colored grays are the coarser grained sediments.

Contaminant Plume Imaging

The contaminant plume models use IVM to show the extent of subsurface plumes in the third dimension. We have produced three-dimensional models of various VOCs and tritium over the entire sight and in specific source areas. Fig. 6 is an example of VOCs found in ground water for the LLNL Site. The darker grays are source areas for VOC contamination. The lighter shades of gray indicate lower levels of contamination of VOCs measured in parts per billion.

Animation and Presentation

Three-dimensional renderings have an innate power to convey the structure of the subsurface, yet static views of the models may allow misinterpretation of lithologic or chemical distributions. When working at the Silicon Graphics workstation, the user may freely rotate, tilt, and zoom on the model to gain better perspective on individual details or areas of special interest. However, the relatively high cost of the workstation makes wide use of this interactivity prohibitive. As a means of expanding the potential audience for the images, we are developing a system for creating sequences of images on the Silicon Graphics, which can be animated on Macintosh desktop computers and included in presentations, training materials, and communications between collaborating scientists at other facilities.

The IVM module can create a directory of Silicon Graphics screen shots in SGI Image file format (a compressed 24-bit file structure) under script control. The script can dictate the starting and ending viewpoint altitude, azimuth, and distance, and interpolate between these points, generating a requested number of intermediate views. A similar approach, under development, would extend the concept across multiple models, e.g., renderings of a remediation site at different simulated time steps. The Image files are sent to the Macintosh via EtherNet and TCP/IP protocols, where they may be expanded to the native Macintosh PICT or other formats. The Macintosh user has the option to select the final pixel dimensions of the images. Owing to the ability of the Macintosh to share data among application programs, the user can add a wide variety of image manipulations, labels, and other annotations. Recently, Apple Computer has released a new standardized architecture for handling complex, time-based data, called QuickTimeTM. QuickTimeTM provides a foundation for image compression and decompression, synchronization of multiple data types (e.g., animated sequences, digitized video, sound, etc.) for production of sophisticated presentations, referred to in the Apple jargon as movies. These, in turn, can be incorporated into virtually any other document type. To communicate our characterization investigations and remedial alternatives to the public, we use commercial applications to animate our three-dimensional renderings and combine them with explanatory narratives. We are also planning an exploration of accelerated use of the INGRES-Integrator system described above, with short QuickTimeTM movies as the interim reporting medium. With these tools, we hope to greatly expand the utility of the SGI/DGI system as an engine for delivering multidimentional renderings to a wider audience, both inside and outside our environmental restoration community.

Conclusions

Imaging tools are invaluable for visualizing subsurface geology as well as the shapes and sizes of contaminant plumes. Such visualizations are effective in presenting complex and detailed, three-dimensional data to scientists working with the data and the various regulatory agencies.

Work performed under the auspices of the U.S. Department of Energy by the Lawrence Livermore National Laboratory under Contract W-7405-Eng-48.