Because human intrusion is a potential cause of repository breaching during any 10,000-year regulatory period, one of the tenets of radwaste disposal is that a candidate environment should be free of valuable natural resources that could stimulate future explorations. Since the LWA is underlain by exploitable potash beds in the McNutt interval (Figures 2 and 4) and is surrounded by oil and gas wells, the scenarios of one or more inadvertent interceptions of the waste rooms by future drill holes are very credible. PA calculations suggest that the raising of contaminated cuttings to the surface poses the greatest (though minimal) threat to posterity. Direct flows of contaminated brine along boreholes open to the surface has been minimized in PA by assumptions of borehole plugging by future operators. Utilizing EPA modeling guidance that borehole permeability should be . . . typical of a borehole filled by the soil or gravel that would normally settle into an open hole over time, about 3.23 x 10-11 ft.2, the maximum calculated rate of flow to the Culebra was estimated to be about 2800 ft.3/yr. (Helton, 1991, p. 24), a magnitude unrealistic of the rates that could be driven along an open borehole that intercepts a saturated repository at near-lithostatic pressures, or along fractures initiated by subsidence or oil-field water-flooding (Bredehoeft., 1997). Small values are appropriate to sealed boreholes (plugged with concrete or mud) that become leached and porous as cement deteriorates over years, but axial seepage may maintain a conduit outside the seals. Thus, flow rates to the surface and the Culebra have been unconservatively discounted for human-intrusion scenarios, as well as for hydrofracture and subsidence-induced fracture conditions.
Regional Flow in the Rustler Formation
Transport modeling entails many parameters, known or approximated. In order to model the paths and rates of transport of radionuclides in groundwater, there must be defined all necessary groundwater fluxes, boundary conditions, material properties and chemical processes. In this instance, a boundary following divides encompasses about 2300 square miles to form the lateral limits, while the top and bottom of the Culebra dolomite, 23 ft. thick, form the vertical limits. Fixed hydraulic heads are assigned at the lateral boundaries, commensurate with measured or deduced heads, and the model limits are defined as either no-flow (impermeable) or discharging (permeable) boundaries. If arranged conservatively, those boundaries are so distant from the LWA that errors in selecting heads and flow properties there should have little effect at the LWA perimeter, the compliance boundary. Heads within the model, determining gradients everywhere, have been adjusted according to measured heads at wells penetrating the Culebra dolomite. If the PA model was grossly correct, that only regional flow based on current heads in the Culebra mattered, the effort might justify minimal transport via the groundwater pathways. DOE relies upon low matrix-dominated permeability of the Culebra (modeled as a single fracture) in its claim that the Rustler is thus a barrier to significant flows to the accessible environment.
The facts reveal the Rustler to be a much more complex aquifer, in terms of properties and boundaries. As Figure 5 shows, the topography slopes gently from a divide 7 miles NE of WIPP westerly towards Nash Draw, which wraps around the NW, W and SW sides of the LWA. Nash Draw is 20 miles long, bounded on the east by a ragged scarp called Livingston Ridge. Dissolution of the Salado salt at the base of the Rustler has produced a brine aquifer beneath Nash Draw in the thick zone of insoluble rubble formed by subsidence and brecciation of the overlying units (Kelly, 2000). But the Rustler mainly discharges at higher levels and to the surface in the vicinity of the salt pan, Laguna Grande, where it evaporates, at least in today's climatic setting (Phillips, 1987). The lake bottom is a deposit of at least 55 ft. of fine gypsum, attesting to a long-term anhydrite source, distinct from the sodium chloride of Salado origin discharging to the Pecos River via the brine aquifer (Phillips, 1987). Phillips documented the ephemeral nature of discharge from a lake north of Laguna Grande called Laguna Pequena (W/NW Sec. 3, T23S, R29E), recording a flow of over 100,000 gallons per minute into Laguna Grande on September 5, 1984 aft.er a record rainfall event stimulated spring flow. There was no surface inflow to Laguna Pequena, and the outflow diminished rapidly aft.er the measurement (Phillips, 2001). Phillips (1987, pp. 244-248) makes a convincing geochemical argument for a Rustler source of most water discharging to Laguna Pequena and Laguna Grande, which must be the long-term destination for any WIPP-site groundwaters.
The current southerly gradient beneath the LWA, and a fairway of high Culebra transmissibility southwards through DOE-1, directs modeled Culebra flow paths from the repository southward into regions of low transmissibilty and to a distant discharge area on the Pecos River (Malaga Bend). But had they incorporated the low heads at WIPP-29 (2968 ft.) and Laguna Grande (2950 ft.), modeled paths would have turned west to Nash Draw (Phillips and Snow, 1998, p. A-8). Gradients in the Rustler are centripetal to Laguna Grande (Kelly, 2000), consistent with the regional evidence that the Rustler drains westerly to Nash Draw. Computed travel times through the Culebra would have been much shorter if model heads had been realistically represented.
Another significant and erroneous assumption used in PA is that the Culebra dolomite is the only aquifer of concern for radionuclide transport from the repository to the limits of the LWA. Drill holes through the Rustler encountered mainly anhydrite, a dense, hard, impermeable rock in its intact form, thus the Magenta and Culebra dolomites were considered the only persistent aquifers that could be tested and characterized (See Figure 3). During shaft-sinking, Chaturvedi and Channel, (1985, Plate 1) found that there is a steep fracture with a dissolution-enlarged opening across the brittle Unnamed Lower Member anhydrite (below the Culebra), so the dolomites cannot be confined. Figure 8, illustrating that fracture, suggests not only open-conduit flow across the Culebra, but also that karst channeling must have occurred at the top of the Salado salt directly above the repository. At various levels in the 250-550 ft. thick saturated Rustler and Dewey Lake interval there may be infrequent, large-capacity fracture-dissolution conduits not usually intercepted or characterized by the vertical drillholes completed for testing the Culebra. Thus, the simplistic, two-dimensional PA model of continuous confined flow in the Culebra dolomite (23 ft. thick) fails to characterize the entire Rustler, which can capture repository discharges via boreholes or fractures, and channel more rapid open-conduit karstic flow. A tiny fraction of the formation's volume, as cave passages, may occasionally convey the preponderant fraction of the discharge, at great rates unrepresented by Doe's modeling of an intact, thin, matrix-dominated Culebra dolomite (Hill, 1999).
The Culebra dolomite has been modeled as a continuous porous medium cut by a single horizontal fracture, an erroneous conception argued to be conservative. First of all, the horizontal fractures are bedding plane breaks on shaley partings without significant hydrologic importance because they are tightly closed until unloaded or sampled by coring. As may be seen at Culebra Bluffs, there are numerous vertical and inclined fractures (Swift., 1992). Many are enlarged by dissolution near the surface, and these are probably good conductors, but DOE has never tested their individual properties in the buried Culebra. They form two orthogonal sets trending NE and NW (a tectonic pattern pervasive throughout the Delaware Basin). Cores reveal their antiquity by mineral infillings of gypsum that render the fractures locally impermeable, indeed, partitioning the matrix into isolated permeable blocks, the probable character of the Culebra east of the LWA (as at P-18). Inclined fractures are also present at Culebra Bluffs and probably wherever differential subsidence has occurred over regions of salt removal. In the subsurface, inclined fractures are also gypsum filled, but to a variable degree due to dissolution. Westward across the site, the proportion of fractures lacking gypsum infillings seen in drill cores increases (Ferrall and Gibbons, 1980), and gypsum infillings are absent at Culebra Bluffs. Neill, et. al, (1998, p.11) suggest that fracture openings formed due to dissolution of Rustler salt and consequent deformations. It may be that within the LWA, Culebra flow is not controlled by either matrix permeability nor by fracture permeability, but by elongate dissolution channels formed, perhaps at the intersections of fractures and the Culebra contacts. If DOE had fully characterized Culebra fracture flow properties by slant-hole coring, dissolution conduits would also have been found and described, and by necessity, modeled.
Tests have revealed such a great range of transmissibilities (T = hydraulic conductivity times aquifer thickness) near wells completed in the Culebra that the degree of channeling must be the main variable across the site. East of the LWA, measures are on the order of 10-3 ft.2/day. It is on the order of 0.1 ft.2 /day on site, on the order of 102 ft.2/day west of the LWA, and 103 ft.2/day in Nash Draw. Thus T increases westward by five to six orders of magnitude (Phillips and Snow, 1998, p A-3). Many individual tests give different transmissibilities, depending upon the observation well used to interpret drawdowns. Of 42 wells tested, high measures were reported at wells WIPP-13, H-6, P-14, H-11, DOE-1 and DOE-2 (Figure 5). Local and directional variability is typical of karst regions with widely distributed solution channels, reflecting chance proximity of each test well to the conduits. The observed irregular increase of T from east to west reflects a systematic increase in the dissolution of fracture fillings, coalescence of smaller channels into larger conduits and the development of cross-connecting fractures and channels to other Rustler strata (Hill, 1999, Neill, et. al., 1998). Since pumping tests in the Culebra produce greatly different responses at adjacent observation wells, reflecting Rustler anisotropy and heterogeneity, the conservative approach to modeling would be to utilize at a site the greatest interpreted transmissibility, or at least the geometric mean of directional values. Instead, DOE has arbitrarily assumed transmissibility values at well sites within 1.5 miles of the center of the repository that are one to two orders of magnitude smaller than the highest values revealed by the hydrologic tests (Phillips and Snow, 1998, p. A-3). The consequence is that computed travel times across that first 1.5 miles from the source are exaggerated by similar magnitudes.
Consistent with Doe's assumptions that Culebra flow is confined and that transmissibility is due solely to matrix or hypothetical horizontal fracture properties, PA modeling was conducted with a continuous T-field interpolated from 39 selected observations. The mathematical method assumes that T is a continuously varying point-function, an invalid procedure in the presence of occasional large, discrete solution conduits of unmeasured, thus uncertain conductivity, randomly placed with preferred orientations due to fracture controls and former hydraulic gradients. At WIPP-33, a site 0.54 miles west of the LWA, a hole drilled in the center of a sinkhole intercepted a cavern in the Dewey Lake and four caverns in the Rustler as deep as the Magenta dolomite. DOE neglected to test that well hydrologically nor subsequently to monitor it. Though only that test hole and one shaft. exposure (Figure 8) indicate the actual geometry of parts of the karst conduit system, that deficiency stems from an apparent unwillingness to explore for such features. Prudence should have demanded a conservative interpretation of the potential consequences of karst, at least until the necessary investigations were done.
Flows were computed according to the present-day fresh-water head (depending upon pressure, density and level) distribution in the Culebra, whose gradient is generally southerly over the LWA, tending to direct discharge towards Malaga Bend. The gradient in the overlying Magenta dolomite is westerly, as may also have been the gradient in the Culebra before pumping and drainage to the shaft.s altered it in the 1970's and 1980's. Culebra heads have been systematically rising for years, particularly since the sinking of the Air Intake Shaft. and the stanching of copious inflows to the repository that occurred. As the synoptic data in Figure 6 indicates (Snow, 1998), the heads at all well-recorded observation wells completed in the Culebra continued to rise during more than five subsequent years. It is possible that a primitive westerly gradient may ultimately be restored long aft.er repository closure, directing the gradient for all Rustler flows towards Nash Draw. The gradual rise indicates that on that scale and in that five year time span, the Culebra itself behaves as a continuous aquifer of modest transmissibility, at least at the points of observation where wells are completed. If large conduits are present at other levels or places, there could be transient head fluctuations that occur mainly in such a system of solution caverns, not recorded on hydrographs of the wells because none penetrate the karst conduits, nor were they instrumented for brief excursions.
Mercer (1983), Chaturvedi and Channel (1985, p. 40) and Brinster (1989, p. IV-75) discussed the fact that the Magenta has freshwater heads as great as 155 ft. higher than that of the Culebra within most of the LWA (except at WIPP-13), but that the two dolomites have coincident heads from the west boundary of the LWA to Nash Draw. Such a head difference at locations within the LWA may be wholly post-disturbance, reflecting short-term Culebra confinement and perhaps the sealing of old karst features between the two strata.