A few feet below the plain is the 500,000 year-old Mescalero caliche, which has collapsed into Nash Draw. Testing the hypothesis that the caliche layer does not obstruct recharge in the vicinity of WIPP, Phillips (1987) drilled over 1000 auger holes in and around several of the depressions. Many of the topographic depressions within and around the LWA prove to be solution-subsidence dolines (sinkholes partially filled with alluvium), while others are wind-eroded deflation depressions. The caliche bed has subsided to form funnel shapes below some of the depressions, is dissolved to residual materials in many places, and is absent where solution pipes penetrate it. The deformed caliche surface proves local subsidence, and the solution pipes prove that infiltrating water has locally karstified the caliche. Phillips explored another mile-long karst valley without a surface water course that crosses the west boundary of the LWA, and a chain of four sinkholes at WIPP-33 indicating that water carves elongate E-W courses through the underlying Dewey Lake and Rustler beds, at least to the depth of the Magenta Dolomite.
Observations of storm infiltration at sinkholes prove that karst conditions exist in the subsurface. About ten inches of rain fell in the vicinity (registered at Loving, 18 miles WSW of WIPP) on September 18-19, 1985, causing the WIPP-33 sinkhole to fill like a bathtub to a depth of five feet, then disappear in a few days (Phillips, 1987, p. 86 and Greenwald, 1995). Water also ponded in the fourth depression east of WIPP-33 and rapidly infiltrated the ground. Such water can not seep away laterally from closed topographic depressions. It must follow karst conduits to the water table, joining an interconnected, perhaps locally obstructed conduit system that leads to outlets. Mature karst aquifers usually discharge at large-capacity springs along rivers or shores defining the base-level for the system, and have a strong influence on the water table everywhere upstream. Storm waters recharge such a karst system in a matter of hours, typically causing a sudden rise and a corresponding rapid increase of spring discharges, noted in days. Wells near WIPP have not been instrumented to provide hydrographs of the karst system, so none substantiate rapid water level rises, but one typical spring-flow event has been observed. On September 5, 1984, following a major storm, Phillips (1987, p. 228) witnessed transient flow into Laguna Grande amounting to at least 100,000 gallons per minute, the water welling up through the floor of nearby Laguna Pequena. Figure 7, showing flow in a channel connecting the two is thus graphic proof of karst conditions in the watershed. The chemistry of the waters was consistent with the Rustler lithology of gypsum and dolomite, dissimilar to either the brine aquifer waters flowing on Salado salt to Malaga Bend or of the effluent from the nearby potash refineries. Such hydrologic and geochemical observations should have been collected routinely by Sandia Laboratories. Years of paper studies and computational modeling have remained unsupported by basic fieldwork. Since Phillips shared his field observations with Sandia hydrologists in 1990, it seems deliberate that Phillips' thesis, one of the few data sources for surface water hydrology at WIPP, was not cited in the CCA nor its background documents.
Phillips (1997), Phillips and Snow (1998), Snow, (1998) and Hill, (1999) have speculated on the geometry of the karst system at the WIPP site. Estimates of the travel-time for radionuclides that enter the Rustler depend upon the actual karst configuration. No tracer tests have been done to determine either potential karstic flow paths or rates. The five caverns at WIPP-33 provide reliable geometry. Elsewhere, it is inferential, as are the relevant but poorly defined boundary conditions. Phillips and Snow (1998, Figure 4) analyzed the transmissibility data and head distribution among the wells, conceiving that in plan view, there are preferred channel directions from the shaft. area southwards and northwestwards, and from the LWA westwards along three paths to Nash Draw. One passes through the NW corner of the LWA and WIPP-33, one follows the karst valley near the center of the W boundary, and a third lies just south of the S boundary. Snow (1998) speculated on the channel distribution in section, envisioning a phreatic karst in the vicinity of the primitive, pre-project-disturbance water table with an extensive vadose karst above it, all influenced by the systematic joint pattern and east-dipping bedding. The water table profile is envisioned to stair-step down-gradient towards the west. Hill (1999) interpreted the evidence in favor of an intrastratal (covered) karst, mainly confined to contacts of the Rustler dolomites but fed by steep infiltration channels. The westerly-increasing transmissibility may reflect conduits at several stratigraphic levels. The water-filled WIPP-33 caverns lie at and above the Magenta. Upon examination of the many drill logs, Phillips (1997b) found residual clays, open washed-out zones or lost core intervals strongly suggestive of dissolution at every stratigraphic interval from the Dewey Lake to the top of the Salado salt, but not necessarily features that are open today. Below the seldom-observed water table, progressively rising by many feet during years of recovery aft.er drainage to the shaft.s (Figure 6), a continuous large-scale karst probably does not occur in the Culebra, otherwise the recovery would have been rapid. But if the water level does rise rapidly in karst channels following storm recharge events, and then drains rapidly to spring outlets like Laguna Pequena, it has minimal effect on the water levels recorded in Culebra matrix. Conversely, the gypsum spring deposits extending 3500 ft. along the base of Livingston Ridge, containing bones of camels and teeth of Equis, may reflect a more durable level of the water table during a cooler or wetter period, and saturation of a higher interval of the karst as much as 240 to 270 feet thick. Not only is there is a deficiency of hydrologic data on the existing karst, but also on paleokarst relevant to the 10,000 year period of interest. Lack of data does not exonerate Doe's PA modeling for neglecting it.
Effects of Faulty Assumptions
There is no doubt that there is a karst system west of the repository that is both phreatic and vadose, as recorded at WIPP-33. Sinkholes common both inside and outside the LWA attest to vadose channels eroded (due to weak clay binder) both laterally and vertically through Dewey Lake clastics. As the Rustler salt beds were being removed, anhydrite was being altered to gypsum and subsequently dissolved. This can be seen at Culebra Bluffs, where the Culebra is intact, but gypsum and residual clays represent the anhydrites and salt. We do not know the current relationship of the karst aquifer to the west-sloping water table, which bevels the east-dipping bedding of both the Dewey Lake clastics and the Rustler evaporites. The lower bounds of the karst aquifer must be limited by the truncated ends of the five Rustler salt beds, but the gradations of karstification above it are unknown. Figure 8 proves that fractures have been dissolved to depths below the Culebra in the repository area, at least locally. Below the current water table, depressed by drainage and testing, the hydrologic data reflects only fine-textured dissolution conduits, like the opened fractures in the Culebra. Only the current saturated zone can be pump-tested today. But when storms recharge the aquifer or climate becomes wetter, the water table may rise to occupy a more mature karst with larger conduits, manifested by rapid transient flow, and restore westerly gradients controlled by those conduits. In karst regions elsewhere, channelization has been found to be most active at and just below the water table, which can rise appreciably into a far more transmissive vadose zone, due to increased recharge during wetter climates. For such reasons, Doe's ignorance of the karst system, and its failure to define and understand the water table and its controls has fostered simplistic assumptions and prevented construction of a valid long-term performance model.
Dye-tracing of groundwater flows in classic karst areas such as Dalmatia, Florida and Kentucky have demonstrated transient velocities exceeding a mile per day, flows that may even now be matched west of WIPP during storm recharge-discharge events like those of September, 1984 and 1985 (Phillips, 1987). Travel-time estimates through the WIPP karst (Phillips and Snow, 1998, Snow, 1998 and Hill, 1999) have ranged between 5 years and 500 years from the center of the LWA to the accessible environment in Nash Draw, but even these may not be conservative. The evidence of phreatic karst conditions and of vadose karst features within the reach of a rising water table in the Rustler formation gives assurance that if contaminants escape through the Salado during the next 10,000 years, the Rustler will not impede rapid solute movement.
PA and supporting papers belabor the transport issues of Culebra dolomite (Snow, 1995), when that unit plays, and will play a minor role to the karst system. It treats chemical retardation for each radionuclide species using laboratory distribution (Kd) coefficients derived with powdered dolomite. Even the critics (see Appendix 8.6 of Neill, et. al, 1998) of Doe's use of such coefficients don't acknowledge that if a karst system short-circuits the Culebra, solutes cannot equilibrate with dolomite matrix. PA applies an unjustified, continuous, steady-state groundwater flow model (such as that of Corbet and Knupp, 1996) to calculate specific discharges, then arbitrarily assumes 1% effective porosity to obtain seepage velocities. It attributes effective porosity to diffusion between a (non-existent) horizontal fracture and the rock matrix, numerically consistent with six tracer tests measuring retardation in the Culebra (Jones, et. al., 1992). But if retardation is due to other factors, such as fracture pattern (Snow, 1995 and Snow and Lee, 1996) and/or a karst conduit system outside the Culebra, the assumed 1% effective porosity could be orders of magnitude too large (thus seepage velocities too small). Whereas diffusion will occur between matrix and either fractures or solution conduits, its retarding effect will be insignificant in context of the rapid transport during brief episodes of karst channel flow, especially aft.er a high water table has been restored. Likewise, ion exchange between solutes and matrix will occur, but if the major flux is via channels with relatively small surface area, chemical retardation will be minimal. The tracer tests gave direct retardation measures for surrogate species across short Culebra paths, but failed to predict the correct order of transport rate through karstic ground, because the close-spaced test wells nowhere penetrated the channels, but were completed only in the matrix and some dissolved fractures of the Culebra dolomite.
The regional flow model of Corbett and Knupp (1996) emulated in the 1996 PA discounts the effects of large, transient recharge events, such as Phillips observed (1987, pp. 84-86, 228-231), instead applying uniform vertical fluxes of 0.2 to 2.0 mm/year consistent with minimum well test interpretations. Conversely, the potential evaporation from pre-mining Laguna Grande (Phillips, 1987), the likely discharge point for the Rustler aquifers, should suggest recharge of at least 15 mm/year. Furthermore, the PA model applies the current gradients, generally southward from the repository area, instead of a more applicable westerly gradient for pre-disturbed and post-emplacement flows. Biased transmissibility data application was noted by Phillips and Snow (1998, p. A-3). Culebra transmissibilities have been extrapolated from known points of measurement to unknown areas assuming that T varies continuously. For very discontinuous media, such as karstic strata, the pilot-point interpolation concept is invalid because the extremes of the conductivity distribution are unrepresented in the data set.
Table 1 summarizes assumptions made or implied by PA. The table asserts actual conditions and describes performance implications. The preponderance of evidence supports the contention that PA modeling seriously underestimates the Cumulative Complimentary Distribution Function of releases of radioactivity to the accessible environment, reason for invalidating the certification granted by the EPA. Given these departures from rationality, the reader should be incredulous that the CCA was approved. DOE obtained the blessings of the National Research Council (1996) in spite of reservations expressed by its hydrologist (Konikow, L., 1995, and National Research Council, 1996, p. 4). The EPA was well aware that the basis for objections, then and now, has always been that there are karst conditions in the Rustler. Implications for rapid transport were not lost on the EPA (Meetings of January 14-16, 1995 at Washington, DC and of December, 1995 at Carlsbad, and Neill, et al., 1998), whose approval can best be understood in view of the unreasonable pressure placed on it by Congress (See Attachment B). In a more recent review (National Research Council, 2001, p. 28), the committee also over-rode the concerns of its hydrologist, John Sharp, that there are uncertainties of transport in the Rustler due to a lack of characterization of fractures and potential karst conduits. Careful scrutiny, especially of the karst issues, will not support the models used, nor provide reasonable expectations that Doe's predicted releases will be met. Post-emplacement monitoring will surely be ineffectual, as far as posterity will tell.
The TRU waste has to be removed from the generator sites, even if an adequate permanent repository has not been established. Forty years' study of options has failed, since geologic disposal has never satisfied everyone, but until a satisfactory method and place is established, government retains the responsibility for safeguarding all military nuclear waste. TRU waste disposal underground remains premature, leaving monitored retrievable storage as the only option. Unlike spent fuels, TRU waste offers no temptations of reprocessing, so time is an unimportant factor. A safe temporary facility could be situated near the surface of an old, stable landform, above the water table and free from flooding and infiltration hazards. For instance, at WIPP, it could probably be established on the Santa Rosa formation, but not on the Dewey Lake/Rustler karstland. The duration of institutional controls might have to be a century, a reasonable expectation for stable structures, while study of permanent disposal options is pursued. It is evident that disposal in salt at WIPP is not the answer, where travel times in the overlying aquifer will be orders of magnitude shorter than PA predicted. The EPA has erred in certifying the repository, and recertification (in 2003) should be defeated. Meanwhile, TRU waste disposal at WIPP should stop, and the waste already in place in the first panel should be retrieved before roof collapse makes it prohibitively costly and dangerous to do so.
Dr. Snow's research was supported by a grant from the Citizens' Monitoring and Technical Assessment Fund.