!C-,-

6

p

'14

-1iCo

\/3

\ F'II

Spatial Variability in Hydrologic Propertiesof a Volcanic Tuff

by J. D. Istoks, C. A. R a utnanb L. E. Flint, and A. L. Flintd

AbstractSpatial variability of hydrologic properties was quantified for a nonwelded-to-welded ash flow tuff at Yucca Mountain,

Nevada, the potential site of a high-level, nuclear waste repository. Bulk density, porosity, saturated hydraulic conductivity,and sorptivity were measured on core specimens collected from outcrops on a grid that extended vertically through the entireunit thickness and horizontally 1.3 km In the direction of ash transport from the volcanic vent. A strong, geologicallydetermined (deterministic) vertical trend in properties was apparent that correlated with visual trends In degree of weldingobserved in the outcrop. The trend was accurately described by simple regression models based on stratigraphic elevation(vertical distance from the base of the unit divided by unit thickness). No significant horizontal trends In properties weredetected along the length of the transect. The validity of the developed model was tested by comparing model predictions withmeasured porosity values from additional outcrop sections and boreholes that extended 3000 m north, 1500 m northeast, and6000 m south of the study area. The model accurately described vertical porosity variations except for locations very close tothe source caldera, where the model underpredicted porosity In the upper half of the section. The presence of deterministicgeologic trends, such as those demonstrated for an ash flow unit In this study, can simplify the collection of sitecharacterization data and the development of site-scale moL. Is.

IntroductionSite characterization and performance assessment

activities are in progress at several existing and potentialwaste storage and disposal sites in the western United States.The information provided by site characterization typicallyconsists of measurements of hydrologic properties at alimited number of data locations (e.g., core specimens fromboreholes). Performance assessment calculations (e.g., waterflow or solute transport calculations), however, typicallyrequire property values at all locations in a computationalgrid covering the site. Estimating these values from theavailable point measurements requires the development ofsite-scale models to represent the variation in geologic andhydrologic properties of subsurface materials. The develop-ment and use of models of this type has been called 'dataexpansion"(Journel and Alabert, 1989); applications to sitecharacterization and performance assessment are discussedin Freeze et al. (1990) and Rautman and Treadway (1991).

Previous studies have demonstrated that soil and rockproperties vary vertically and laterally at many sites (Nielsenet al., 1973; Gajem et al., 1981; Healy and Mills, 1991;Rautman and Flint, 1992) and that, in some situations, thisvariability may be modeled as a stochastic process (Russo

&Department of Civil Engineering, Oregon State University,Corvallis, Oregon 97331.

bGeoscience Assessment and Validation Department, SandiaNational Laboratories, Albuquerque, New Mexico 87185.

"Raytheon Services Nevada, Hydrologic Research Facility,P.O. Box 327, Mercury, Nevada 89023.

dU.S. Geological Survey, Hydrologic Research Facility, P.O.Box 327, Mercury, Nevada 89023.

Received July 1993, revised December 1993, accepted March1994.

and Bresler, 1981). Siample variograais are convenient toolsfor quantifying the spatial variability displayed by aregionalized variable and form the basis for a number ofestimation and simulation methods. Examples of the use ofgeostatistical methods to develop site-scale models of hydro-logic properties of soils are in Guma (1978), Burgess et al.(1981), and Viera et al. (1981). Geostatistical methods havealso been used to quantify the hydrologic properties ofrocks, e.g. fracture orientation and size (Smith, 1981;Young, 1987), hydraulic conductivity and unit thickness(Aboufirassi and Marino, 1984), and geochemical composi-tion (Aitchison, 1984).

Only limited information is available about the spatialvariability of hydrologic properties of volcanic rocks,although these rocks constitute the principal hydrogeologicunits at Yucca Mountain, Nevada, a potential repository forhigh-level nuclear waste. The potential repository horizon isin the unsaturated zone and water flow through thesematerials is the subject of ongoing site characterization andperformance assessment activities. Recent field investiga-tions have identified the presence of deterministic trends inhydrologic properties within individual ash flow units(Rautman, 1991; Rautman et al., 1993) and that these trendsmay have a pronounced effect on the unsaturated flowsystem controlling water flow through the potential reposi-tory horizon (Rautman and Flint, 1992). Vertically, hydro-logic properties in volcanic rocks may display deterministictrends related to degree of welding within individual coolingunits; laterally, hydrologic properties may display determin-istic trends related to distance from the source caldera;however, insufficient data are currently available to quantifythese trends.

The presence of deterministic (geologically controlled)vertical trends in degree of welding within individual cooling

Vol. 32, No. 5-GROUND WATER-September-October 1994 751

i�M

.1

Porosity Porosity I

Fig. 1. Conceptual model of cooling unit showing zonation and vertical porosity distribution.

units in tuff deposits is well-known (e.g., Ross and Smith,1961; Sparks and Wright, 1979). Figure 1 presents a simpli-fied conceptual model of an individual ash flow unit. An ashflow unit is typically underlain by a topography blanketing,Plinian air-fall ash bed (Sheridan, 1979); the thickness of theash flow unit generally decreases with increasing distancefrom the source caldera. As the ash cools, zonal features areformed. In the interior of the ash flow unit, high tempera-tures are maintained for longer periods of time than in theoverlying and underlying layers. Heat loss to the preexistingtopographic surface is initially rapid, but quickly slows asthe temperature of the preexisting surface increases. Heatloss from the upper surface of the ash flow generally con-tinues at a more uniform rate due to convective heat transferfrom this surface to the atmosphere. Depending on the rateof cooling, the initially glassy material may devitrify in situto a microcrystalline aggregate of silica polymorphs andalkali feldspar. Plastic deformation of glassy shards andreduction in porosity can occur as layers compact under theweight of overlying materials. Continued exsolution ofmagmatic gases into pores (Ross and Smith, 1961) mayallow the formation of vapor-phase alteration minerals,again typically microscopic intergrowths of silica phases andfeldspar. Vapor phase alteration is typically more advancedin the upper portions of ash flow deposits.

The hydrologic properties of the resulting tuff are sig-nificantly influenced by the initial thickness and rate ofcooling of the ash, and these vary with position within thecooling unit. For example, a high porosity zone is expectedat the base of an ash flow sheet, which cools quickly (Figure1). Porosity is expected to decrease with increasing elevationin the interior (core) of the flow unit where temperatures

remain high long enough for significant plastic deformation(welding) to occur. The minimum value of porosity isexpected near the base of the core because overburdenpressure is largest there. Above the core, porosity is expectedto first decrease gradually with increasing elevation as over-burden pressure decreases, and then to decrease rapidly withincreasing elevation through the quickly cooled, uppermostportion of the flow. These vertical trends in porosity areexpected to be most strongly expressed in the proximal endof the ash flow, nearest to the source caldera, where the unitis thickest and sufficient overburden pressure is available toproduce a densely welded core. These trends are expected tobe less apparent at the distal end of the flow, farthest fromthe source caldera where the unit is thinnest, rapid coolingcan occur through a major portion of the ash flow, andoverburden pressures are small. Although this conceptualmodel is generally accepted as valid at Yucca Mountain,variations in magma chemistry, initial temperature, or post-depositional alteration history may produce local variationsin this expected profile (Rautman and Flint, 1992).

Although the qualitative features of this conceptualmodel are well-known, only limited quantitative informa-tion is currently available on the distribution of hydrologicproperties within individual cooling units. If hydrologicproperties can be correlated with more easily observablegeologic trends (for example, trends in degree of weldingwhich can be determined visually), it may be possible tosimplify site-scale models and reduce sampling and testingrequirements for site characterization. The objective of thisstudy was to obtain a detailed description of the spatialvariability (vertical and horizontal) in hydrologic propertiesfor the nonwelded-to-welded transition shardy base micro-

752

I

The principal study area is an extensive outcrop of theshardy base microstratigraphic unit of the Tiva CanyonMember of the Miocene Paintbrush Tuff, located on thewestern flank of Yucca Mountain (Figure 2). The thicknessof the unit ranges from 5 to 15 m along the length of thesampling transect (described below) with an average thick-ness of 8.3 m. The shardy base microstratigraphic unitconsists of three subunits: the upper ash flow, lower ashflow, and pumice bed (Figure 3). The basal pumice bedsubunit ranges in thickness from 0.2 to 3 m and is overlain bya nonwelded and poorly sorted lower ash flow subunit,which ranges in thickness from 2 to 8 m. This subunit isoverlain by an upper ash flow subunit that changes fromnonwelded at the base to densely welded at the gradationalupper contact with the overlying columnar microstrati-graphic unit. The lower contact of the shardy base is grada-tional with underlying weathered pumice and ash units; theupper contact of the shardy base is identified in the field bydevitrification and vapor-phase alteration, and by thedistinctive vertical cooling joints in the overlying columnarmicrostratigraphic unit.

I4Reid Sampling

A two-dimensional outcrop sampling transect wasestablished that consisted of 26 vertical sections locatedbetween 19 and 200 m apart along a 1.3 km long exposure ofthe shardy base in Solitario Canyon (Figure 2). Thesampling transect extends vertically through the entire

MILES

0 1 2 3 NAAYuccaN KM I ounwi

e 1 2 3 NFig. 2. Location of study area on Nevada Test Site, Nyc County,Nevada. Core data were collected from a two-dimensional out-crop sampling transect In Solitarlo Canyon, vertical outcropsections at Busted Butte, Paginy Wash, and Yucca Wash, andboreholes N-li, N-37, N-53, N-54, and N-55.

stratigraphic unit of the Tiva Canyon Member of the Paint-brush Tuff at Yucca Mountain. The hydrologic properties ofthis unit are of importance because under conditions pro-ducing fracture flow in the overlying welded tuffs, it couldform a capillary barrier or a permeability barrier capable ofdiverting percolating ground water laterally (Flint et al.,1993). This study presents a quantitative description of thespatial variability of hydrologic properties for the shardybase unit based on an analysis of data for outcrop corespecimens collected from an extensive two-dimensionalsampling transect. The model developed from the two-dimensional transect is tested by comparison with addi-tional outcrop sampling transects and with core data fromrecently completed boreholes.

MethodsSite Description

Yucca Mountain, located 140 km northwest of LasVegas, Nevada (Figure 2) consists of alternating welded andnonwelded ash flow and ash fall tuffs that have been tiltedtoward the east by Basin-and-Range faulting (Carr, 1988).The upper portion of the tuff sequence is well-exposed alongthe western face of Yucca Mountain in Solitario Canyon.

Ash kw w*wfd1 WIV vIWc*Oyfl becomes dsiedwpwod: .Ik F.5 Wd MackwO

Fpimr wtme~c -oshg Wmlt

tI-t

IAWmI, lair ftW~W8*dbuss~yWnudydakIat yol-

-wwm a~wdwa $ Iem1WL tpwWpelt -PWodW oveIOWe Moky I a&WthOuIft

II

I II

LAWWhftoWWelkd.Pooflw-tc W"Amn spw, WWIlNedbrosm Ift Cluti. 11-20an1hrsmwald2anpimdw I aban.

/ball'owon e 11 my '

kee. v1 hb uwlo Wil

I \k he" t-~TfI ~~~~~~~~~~owldado i_-a Tlsh

Fig. 3. Schematic stratigraphic profile of part of the Paintbrushnonwelded Interval In Solitario Canyon showing position ofshardy base transect lithologic Interval near drill hole USWUZ-6s.

753

mm

thickness of the unit and is approximately aligned hori-zontally with the principal direction of ash transport fromthe Claim Canyon caldera, the ash source. Ten to 23 corespecimens, nominally 2.5 cm in diameter and 4 to 10 cm inlength were collected from each vertical section using aportable core drill with a diamond coring bit. Vertical spac-ing between cores ranged from 0.15 to 2.90 m with anaverage of 0.76 m. Horizontal distance between verticalsections ranged from 22 to 201 m, withan average of 52.1 m.The horizontal distance (x) from the base of each verticalsection to an origin established downslope from an existingdrill hole (USW UZ-6s) was measured with a steel tape. Thevertical distance (z) of each core above the base of the basalpumice bed subunit was measured in the field using aJacob's staff. Stratigraphic elevation, E. = z/ D, was com-puted for each core location, where D is the thickness(measured perpendicular to the bedding plane) of the unit atthe vertical section containing the core.

Core specimens were also collected from three addi-tional vertical outcrop sections, located at Yucca Wash,Pagany Wash, and Busted Butte, and from five boreholes:USW UZN-11 (N-li), USW UZN-37 (N-37), USW UZN-53(N-53), USW UZN-54 (N-54), and USW UZN-55 (N-55)(Figure 2). In an attempt to quantify the magnitude ofsampling and testing errors, five closely spaced cores werecollected within a total horizontal distance of approxi-mately I m at selected locations in the Pagany Wash andBusted Butte outcrop sections.

Laboratory AnalysesCore specimens were trimmed to a length between 2

and 5 cm using a diamond rock saw. Bulk density (Pb) andporosity (5) were determined using Archimedes' principle.Trimmed core specimens were first saturated with carbondioxide by introducing the gas into an evacuated bell jarcontaining the cores. Core specimens were then saturatedwith distilled, deaired water, reneoved from the bell jar, anddried using a damp towel to reimve surface water. Speci-mens were weighted to determine saturated weight, andweighed while submerged in water to determine saturatedvolume. The samples were dried in a controlled relativehumidity (RH) oven at 600C and 40o RH and weighed todetermine dry weight.

Saturated hydraulic conductivity (K) was measuredusing a constant-head method. Following imbibitionmeasurements, core specimens were encased inside "heatshrink" tubing and connected to a constant-head watersupply reservoir. The volume of water flowing through acore specimen for 72 hour period was collected and mea-sured; saturated hydraulic conductivity was computed usingDarcy's law. A pressure system (triaxial cell) was used tomeasure saturated hydraulic conductivity for a few of themore densely welded cores from each vertical section as wellas the cores from the Pagany Wash, Yucca Wash, andBusted Butte transects.

Sorptivity (Philip, 1957) was calculated from imbibi-tion data. Imbibition measurements were made using a'tension table' (Klute, 1986). The bottom of the tensiontable was lined with a cloth towel, a Mariotte reservoir wasconnected to the inlet port of the tension table and used to

maintain a constant head a few mm above the upper satu-rated surface of the towel. Core specimens were dried at600C and 40% RH, cooled to room temperature, weighed,and then placed on the saturated surface (time, t = 0).Periodically, the core specimens were removed, weighed,and returned to the tension table. The equivalent depth ofwater imbibed into each core as a function of time, I (t), wascomputed from the weight of water imbibed at time t, thedensity of water, and the cross-sectional area of the core.Sorptivity (S) was calculated for each core by fitting theequation I = St02 to the data (Talsma, 1969).

ResultsModel Development

The data from the two-dimensional outcrop samplingtransect in Solitario Canyon was used to develop a quantita-tive model for the spatial variability of hydrogeologic prop-erties of the shardy base microstratigraphic unit in twodirections: vertical (perpendicular to bedding) and hori-zontal (approximately aligned with the direction of ashtransport from the Claim Canyon Caldera). The utility ofthis model for predicting properties of the shardy base over alarge block encompassing the study area was tested bycomparing model predictions with measured values of asingle hydrogeologic property, porosity, for cores from theBusted Butte, Pagany Wash, and Yucca Wash outcrop sec-tions and from boreholes N-ll, N-37, N-53, N-54, and N-55.

Descriptive StatisticsThere were significant differences in properties for the

three subunits, reflecting differences in pumice and fine siltyash content, and welding. The basal pumice bed and lowerash-flow subunit are nonwelded. Porosity (), hydraulicconductivity (K), and sorptivity (S) were largest and bulkdensity (Pb) was smaliest in the basa& pumice bed due to thelarger percentage of coarse pumice grains in that subunit(Table 1). Porosity, log(K), and log(S) were smaller, and Pbwas larger in the lower ash-flow subunit than in the pumicebed, reflecting a higher content of ash. Ash contents in theupper and lower ash-flow subunits were similar, but theupper unit had smaller values of 4X, log(K), and log(S), andlarger values of Ob due to an increase in welding within theupper unit. Properties were generally more uniform (smallersample standard deviations and coefficients of variation) inthe basal pumice bed and lower ash flow subunits than in theupper ash flow subunit (Table 1).

Significant linear correlations existed between theproperties. Based on an analysis of scatter plots, linearregression models were fit for log(K) and log(S) as a func-tion of 4:

l

aV

a

b

Cc

Is

f1'5

11

aU

ti(I

tice

0

I

I

L

I

I

I

log(K) = -9.012 + 12.3130, r2 = 0.824

log(S) = -6.040 + 5.237,0, r2 = 0.826

(la)

(Ib)

These regression models are of interest because they providea means for predicting vertical variation in log(K) andlog(S) based on expected vertical variations in 4 for theconceptual model in Figure 1, and because 4' can be mea-sured at more locations (due to lower analysis costs) than Kor S.

754

Deterministic Vertical TrendsThe existence of strong vertical trends in properties is

apparent in grey-scale plots that show the approximatevalue and location of measured values for 4 and log(K)(Figure 4). Because of the similarity in trends for 4 and Pband for log(K) and log(S), only trends in 4 and log(K) willbe discussed here. In Figure 4 the contacts between the threesubunits are shown by solid lines; the irregular nature of thecontacts results in variation in subunit thickness along thelength of the transect and in some cases may be due to ashfilling of predepositional erosional topography in under-lying units. For example, ash filling of a preexisting ero-sional channel was identified in the field at a distance of 320m. Porosity and log(K) were largest in the basal pumice bedand smallest at the top of the upper ash flow subunit. In theupper ash-flow subunit, and to a lesser extent in the basalpumice bed, 4 and log(K) decreased with increasing eleva-tion above the contact with the underlying subunit. Thesetrends were observed consistently in all vertical sections andwere confirmed by field observations of degree of welding(e.g., pumice grain aspect ratios increased during the transi-tion from moderately to densely welded tuff) and by diffi-culty of drilling, which generally increased with increasingelevation within a section.

A more detailed description of the vertical trends wasobtained from an analysis of composite plots prepared by

Table 1. Descriptive Statiscs for Hydrologic Properties forSubunits and Entire Shardy Base Microstratigraphic

Unit for Cores from Two-Dimensional OutcropSampling Transect in Solitario Canyon

4' Pb log(K) log(S)

Upper ash-flow subunitMinimum 0.06 1.30 -8.8 -5.9Maximum 0.45 2.27 -3.1 -3.4Mean 0.195 1.89 -6.73 -5.08Variance 0.010 0.059 1.58 0.263CV (%) 51 13 19 10n 130 130 120 122Lower ash-flow subunitMinimum 0.20 1.13 -6.9 -5.2Maximum 0.55 1.87 -2.2 -2.9Mean 0.403 1.38 -3.83 -3.83Variance 0.002 0.012 1.00 0.230CV (%) 12 8 26 13n 140 140 133 133Basalpumnice bed subunitMinimum 0.34 0.93 -4.5 -4.1Maximum 0.60 1.54 -1.6 -3.0Mean 0.520 1.13 2.86 -3.45Variance 0.004 0.025 0.503 0.070CV(%) 12 14 25 8n 36 36 33 35Entire Shardy Base unitMinimum 0.06 0.93 -8.8 -5.9Maximum 0.60 2.27 -1.6 -2.9Mean 0.329 1.56 -4.93 -4.31Variance 0.020 0.117 3.61 0.670CV(%) 43 22 39 19n 306 306 286 290Units for K and S are mns-'and ms1,*, respectively. CV is coefficientof variation, and n is sample size.

pooling the data for all vertical sections (Figure 5). Porositydecreased with increasing stratigraphic elevation within thebasal pumice bed, remained relatively constant within thelower ash-flow subunit, and decreased with increasing strati-graphic elevation within the upper ash-flow subunit [Figure5(a)]. The rate of decrease in 4 also decreased with increas-ing stratigraphic elevation. Contacts between the subunitsappeared gradational with respect to 4. Trends for log(K)and 4 were similar, log(K) decreased with increasing eleva-tion within the basal pumice bed, remained relatively con-stant within the lower ash-flow subunit, and decreased withincreasing stratigraphic elevation in the upper ash-flow sub-unit [Figure 5(b)]. The nonlinear vertical trends in 4 andlog(K) within the upper ash-flow subunit are attributed tothe compaction and cooling history of the upper ash-flowsubunit. Compaction is at a minimum at the base of thesubunit where loss of heat to the underlying, cooler, lowerash flow caused the glass shards to lose their plasticityquickly. Compaction, and eventually welding, of the shardsincreased upward, in response to the insulating effect ofpreviously deposited materials and the increasing over-burden pressure caused by subsequently deposited ash lay-ers. Stratigraphic elevation is useful for displaying these databecause it combines data into groups with similar coolinghistory and overburden pressure.

Several empirical expressions were investigated to de-scribe the observed vertical trends in 4 and log(K). For 4, athree-part regression model was fit using a minimized least-squares procedure

4 = 0.63 - 1.54Es oS E,

I

60%11-15

tWWM

0090o

.... .-=

0%

0 200 400 600 800 1000 1200 1400HORIZONTAL DISTANCE NORTH OF DRILL HOLE USW UZ-6s (m)

-1.7

0

Uz"IC,,0

a

Fig.mi

FigSol

0

-8.8

0 200 400 600 800 1000 1200 1400

HORIZONTAL DISTANCE NORTH OF DRILL HOLE USW UZ-6s (m)Fig. 4. Horizontal and vertical variation of (a) porosity and (b) log(K) for cores from two-dimensional outcrop sampling transect InSoltarlo Canyon. Measured values are indicated using the grey-scale legend. Solid lines show location of contacts between subunits.

substantial spatial correlation in the vertical direction isfurther indication that the regression model has accuratelydescribed the observed spatial variation in these properties.

Analysis of Horizontal TrendsThe existence of horizontal trends was investigated by

plotting regression residuals vs horizontal distance along the

sampling transect (Figure 7). No systematic trends withhorizontal distance for either 4 or log(K) were apparent; ananalysis of variance for a fitted linear trend model [with j6 orlog(K) as the independent variable and horizontal distanceas the dependent variable] confirmed that no significantlinear trends existed. Sample variograms for regressionresiduals for q and log(K) were fit with pure nugget models FigSan

756

1.0 | rs _ S 1.0 ...

z O.B

Z 0.6 -a

C 0.4 -

Un

-t L +

*4. + (a)+

\++ +--- FnsED MODEL+++

+ + o

0

+0

0 0

+ UPPER ASH FLOW

0 LOWER ASH FLOW x

x BASAL PUMICE BED 0 KKo0 0

x+ a ++ xxKx

x , x+xKMQ

z 0.8 -0

w 0.6 -

$

!) 0.4 -

U3

. 4+t4 ++ ++

$4+'--- FITED 5MODEL

+ +X+ ++o0}+*4 +

7R ++ 0.4. 0+ C +

00~~~

+ UPPER ASH t

0 LOWrER ASHKll oLo

x BASAL PUMOCEBE ax

0ooC

X x xAL Mixx

0.2- 0.2 -

r0.0 - _0.0 0.2 0.4

POROSITY0.6

u.u -r_-10 -6

log (K)-2

Fig. S. Fitted regression models (solid iines) for vertical trends in (a) porosity and (b) log(K) for cores from the two-dimensional outcropsampling transect In Solitario Canyon.

0.003 n

0.002 -

(a)

JD- _ e- 4D- - -S1 _G - -a

10

I.U -

(b)

e, - -ak -ae- 4 _0,,.'0-6-

0.5 -

0.0- 1- L A A

0.001

n0.0 .- _-I--- - t. - I- - ..U.v D. Z.4 U.0 M.U UO Z.4 U.

VERTICAL SEPARATION, h VERTICAL SEPARATION, h

Fig. 6. Verticalsamplevariogramsforregressionresidualsfor(a)porosity and(b)log(K)forpooleddatafromallvertical sectionsintheSolitario Canyon outcrop sampling transect. Horizontal line represents variance of regression residuals.

2

E,,

C,,

N

CA

0

0CL

ten

to

DISTANCE (m) DISTANCE (m)Fig. 7. Regression residuals for (a) porosity and (b) log(K) for pooled data for all vertical sections in the Solitarlo Canyon outcropsampling transect.

757

L

I

0.005

0.004 -

O.OO -

0.002 -

0.001 -

(a)

9II1

III

1.2

0.6

(b)

00 it

, " 0 0 111

, 6 0 M \ P

0, 1600 200 360 460 560 860

I 11 leII

1, %0

0.000 *I. n n_.__ 6 160 200 3o0 400 5600 60 edo

HORIZONTAL SEPARATION, h (m) HORIZONTAL SEPARATION, h (m)FigS. Horizontal sample varograms for regression residuals for (a) porosity and (b) log(K) for pooled data from *ll vertical sections inthe Solitardo Canyon outcrop sampling transect. Horizontal line represents variance of regression residuals.

indicating an apparent absence of spatial correlation (Figure8). The lack of substantial spatial correlation in the hori-zontal direction suggests that properties can be consideredto be essentially constant (except for nonstructured randomvariations) in this direction.

Model ValidationEquations (2) and (3) accurately described the observed

vertical trends in 4 and log(K) within the shardy base overthe entire thickness and length of the Solitario Canyonsampling transect. The utility of equation (2) for predicting4 over a larger area encompassing the potential repositorywas tested by comparing model predictions with measuredvalues of porosity at three additional vertical outcrop sec-tions and five boreholes. The Yucca Wash section waslocated approximately 2800 m north of the northern edge ofthe Solitario Canyon transect (Figure 2) and represents adepositional environment much closer to the Claim CanyonCaldera. The Busted Butte section was located approxi-mately 3200 m southwest of the southern edge of theSolitario Canyon transect (Figure 3) and represents a deposi-tional environment further from the caldera. The PaganyWash section was located approximately 1500 m north andeast of the Solitario Canyon transect. The thickness of theshardybaseis i4matYuccaWash,4matBusted Butte, and11 m at Pagany Wash. Four of the boreholes (N-37, N-53,N-54, and N-55) are located from 1000 to 1500 m east, andborehole N-11 is located approximately 2400 mnnorth of theSolitario Canyon transect.

Predicted values of 4 were obtained for each core inthese sections and boreholes using equation (2). Compari-sons between measured and predicted values were very goodfor the Pagany Wash section but were poor for boreholeN-1l (Figure 9). Prediction errors (measured - predictedvalue) for all sections and boreholes are compared in Figure10. A prediction error of zero indicates a perfect fit; theregion within the two vertical dashed lines represents the95% confidence interval for the model predictions. AtPagany Wash and Busted Butte, prediction errors weresmall, with a mean value of zero and absolute values gener-ally within two standard errors (shown as vertical dashedlines) of the fitted equations [Figure 10(a)]. For these sec-

758

tions, prediction errors fluctuated randomly about zero,with no systematic deviation from model predictions. Themodel predicted more poorly at Yucca Wash, the sectionclosest to the source caldera. There, the model underesti-mated porosity, especially in the upper half of the section[Figure 10(a)]. The largest prediction errors were severaltimes larger than the average sampling error [shown as ahorizontal bar in Figure 10(a)] determined from an analysisof porosity data for the replicate core specimens from thePagany Wash and Busted Butte outcrop sections, suggestingthat prediction errors are not an artifact of sampling andlaboratory testing.

The model accurately predicted the vertical distribu-tion of 4 at borehole N-55, but performed more poorly inpredicting 4 for the other boreholes [Figure 10(b)]. Predic-tion errors were generally within two standard errors ofmodel prediction for cores from the lower portion of theprofile (stratigraphic elevations smaller than 0.4). At higherpositions in the section, the model systematically underpre-dicted 4; approximately one-half of prediction errorsexceeded the 95% confidence interval for the model fit. Thepoorest fit was obtained for borehole N-11, the boreholelocated closest to the caldera.

The failure of equation (2) in predicting porosity in theupper portion of the section at the northernmost outcropsections and boreholes may be due to the proximity of theselocations to the source caldera and involvement of theseregions in the transition from the intracaldera to outflow-sheet environment. Although the location of the exact mar-gin of the Claim Canyon Caldera is uncertain, the YuccaWash and N-ll sections are within 2 to 3 km of this majorstructural feature (Byers et al., 1976). Geologically, there areprofound differences in the geology of the Tiva CanyonMember in the vicinity of the Yucca Wash transect: the unitis thin and major parts of the welded portion of the unit aremissing or of completely different character. There are alsoqualitative differences in the geology and appearance of theshardy base microstratigraphic unit, although these havenot yet been systematically described or evaluated. Hildrethand Mahood (1986) discuss subtle differences within theBishop Tuff (derived from the Long Valley Caldera nearReno, Nevada) that indicate large-volume ash-flow sheets

1.0

Z 0.80

D 0.6

r- 0.4

z0

b-4

0.2

0.0

Fig. 9. Comparison of model predictions with measured values of porosity for (a) vertical outcrop section In Pagmny Wash and (b)borehole N-ll.

are emplaced as numerous individual flows of lobate formthat may bypass or overlap one another in complex fashion.Given our preliminary information, it appears reasonablethat the northern portion of the Yucca Mountain region hasbeen affected by these types of processes to a much greaterextent than the more distal portions of the outflow sheet

sampled by the remaining transects and boreholes.

SummaryDetailed field sampling and laboratory testing identi-

fied the presence of geologically determined vertical trendsin hydrologic properties for a nonwelded ash-flow tuff at the

1.0 1.0.

z . , o

z0.8 -0

- 0.6-

.,0.4 -

(a) | 1 0I ^ ! , 0

'* i 018sn80. i ~0

I A j

I a a

aA I BUSTED DU~TTI PAGANY WASH

a* &&AA&AYUCCA WASH

a 00000'a

A 0A a -

[a aj

00.8 -0

0.6 -Co.)

o.sU,

* '0~~~*(b) ' °

j X A 0

X ~~~0Al 0

!X°!

,X5jo

: x . oooNl

1 0 0 N-37A 00000 N-63

Ix * &&&A N-54

I a xxxxx N-55X ~ ~ ~ NNIN6

0.2 - 0.2 -

no0. _ . . . . .. _- .2

--- j--- , ' 0 m' * O.-0.1 0.0 0.1 D 0.2 0.3 0.4

PREDICTION ERROR

0.0 !r ~ t ~ ! a-0.2 -0.l 0. O. 0.2 0.3 o.4

PREDICTION ERRORFig. 10. Validation of vertical trend model developed in Solitarlo Canyon with measured values of porosity for (a) outcrop samplingtransects and (b) boreholes. Dashed vertical lines give approximate 95% confidence levels for fitted model. Horizontal bar In (a)represents magnitude of sampling and measurement erors.

759

site of a potential high-level nuclear waste repository. Thetrends were accurately described for a large portion of thestudy area by simple regression models based on strati-graphic elevation (vertical distance from the base of the unitdivided by unit thickness). The validity of the developedmodel was tested by comparing model predictions withmeasured porosity values from additional outcrops sectionsand boreholes. The model accurately described verticalporosity variations in two of the outcrop sections and four ofthe five boreholes but generally underpredicted porosity forthe outcrop section and boreholes located closest to thecaldera (the proximal edge of the ash flow).

Application of ResultsModels of hydrogeologic properties are needed for site

characterization and performance assessment at many exist-ing and potential waste disposal sites. The presence of geo-logically determined (deterministic) trends, such as thosedemonstrated in this study for the shardy base microstrati-graphic unit at Yucca Mountain, can simplify the develop-ment of these models by providing additional informationfor use in sampling design, property estimation betweenboreholes and core samples, and for property simulation.For example, the fitted trends presented here can be used, inconjunction with data from existing borehole and outcropsamples, to obtain preliminary estimates for the shardy baseunit over the potential waste repository block. These trendscan also be used to simulate geologically plausible sets ofrock properties for this unit for use in water-flow calcula-tions. For example, the results indicate that it should bepossible to generate porosity and log(K) fields for the poten-tial repository block using equations (2) and (3) (and ran-dom error components) with very little computationaleffort. These fields could then be used as input for MonteCarlo .ype simulation.s of waterflov' within the shardy base.To the extent that the regression equations quantify actuialgeologic processes within the cooling ash flow, incorpora-tion of those trends brings a degree of understanding intothe modeling process beyond the information containedsolely in the tabulated values of hydrogeologic properties.

AcknowledgmentsThis work was supported by the U.S. Department of

Energy, Yucca Mountain Site Characterization ProjectOffice, under contract DE-ACO04-76D00789 and Interagencyagreement DE-AI08-78ET44802.

ReferencesAboufirassi, M. and M. A. Marino. 1984. Cokriging of aquifer

transmissivities from field measurements of transmissivityand specific capacity. Mathematical Geology. v. 16, no. 1,pp. 19-36.

Aitchison, J. 1984. The statistical analysis of geochemical compo-sitions. Mathematical Geology. v. 16, no. 6, pp. 531-564.

Burgess, T. M., R. Webster, and A. B. McBratney. 1981. Optimalinterpolation and isarithmic mapping of soil properties. IV.Sampling strategy. J. Soil Science. v. 32, pp. 643-659.

Byers, F. M., W. J. Carr, R. L Christianson, P. W. Lipman, P. P.Orkild, and W. D. Quinlivan. 1976. Geologic map of theTimber Mountain Caldera area, Nye County, Nevada.U.S. Geol. Surv. Misc. Inv. Series Map 1-891. 1:48,000.

Carr, W. J. 1988. Volcano-tectonic settingof Yucca Mountain and

Crater Flat, southwestern Nevada. U.S. Geol. Survey Bul-letin 1790. p. 35.

Flint, A. L., J. A. Hevesi, and L. E. Flint. 1993. The influence oflong-term climate change on net infiltration at YuccaMountain, Nevada. Proceedings of the Fourth InternationalConference on High-Level Radioactive Waste Manage-ment, Las Vegas, NV.

Freeze, R. A., J. Massmann, L. Smith, T. Sperling, and B. James.1990. Hydrogeological decision analysis: 1. A framework.Ground Water. v. 28, no. 5, pp. 738-766.

Gajem, Y. M. A. W. Warrick, and D. E. Myers. 1981. Spatialstructure of physical properties of a Typic Torrifluvent soil.Soil Society of America Journal. v. 45, pp. 709-715.

Guma, G. S. 1978. Spatial variability of in situ available water.Ph.D. dissertation, The Univ. of Arizona, Tucson.

Healy, R. W. and P. C. Mills. 1991. Variability of an unsaturatedsand unit underlying a radioactive-waste trench. SoilScience Society of America J. v. 55, no. 4, pp. 899-907.

Hildreth, W. and G. A. Mahood. 1986. Ring-fracture eruption ofthe Bishop Tuff. Geological Society America Bulletin.v. 97, pp. 396-403.

Journel, A. G. and F. Alabert. 1989. Non-Gaussian data expan-sion in the earth sciences. Terra Nova. v. 1, pp. 123-129.

Klute, A. (ed.). 1986. Methods of soil analysis, Part 1, Physicaland mineralogical methods. 2nd ed. American Society ofAgronomy, Madison, WI.

Nielsen, D. R., J. W. Biggar, and K. T. Erh. 1973. Spatial vari-ability of field-measured soil-water properties. Hilgardia.v. 42, pp. 215-259.

Philip, J. R. 1957. Theory of infiltration, 1. The infiltration equa-tion and its solution. Soil Science. v. 83, pp. 345-357.

Rautman, C. A. 1991. Estimates of spatial correlation in volcanictuff Yucca Mountain, Nevada. Sandia National Labora-tories, Albuquerque, NM. SAND89-2270.

Rautman, C. A., J. D. Istok, A. L. Flint, and L E. Flint. 1993.Influence of deterministic geologic trends on spatial vari-ability of hydrologic properties in volcanic tuff. High-LevelRadioactive Waste Management, Proceedings of the FourthInternational Conference, Am. Nuc. Soc., LaGrange Park,IL.

Rautman, C. A. and A. L Flint. 1992. Deterministic geologicprocesses and stochastic modeling. Sandia NationalLaboratories, Albuquerque, NM. SAND91-1925C.

Rautman, C. A. and A. H. Treadway. 199; . Geologic uncertaintyin a regulatory environment: An example from the poten-tial Yucca Mountain nuclear waste repository site. Environ.Geol. Water Science. v. 18, no. 3, pp. 171-184.

Ross, C. S. and R. L. Smith. 1961. Ash-flow tuffs: their origins,geologic relations, and identification. U.S. Geol. Survey,Professional Paper 366.

Russo, D. and E. Bresler. 1981. Soil hydraulic properties asstochastic processes: 1. An analysis of field spatial vari-ability. Soil Science Society of America J. v. 45, pp.682-687.

Scott, R. B. and J. Bonk. 1984. Preliminary geologic map ofYucca Mountain, Nye County, Nevada with geologic sec-tions. U.S. Geol. Survey Open-File Rept. 84-494.

Sheridan, M. F. 1979. Emplacement of pyroclastic flows: Areview. Geological Society of America Special Paper 180.pp. 125-136.

Smith, L 1981. Spatial variability of flow parameters in a strati-fied sand. Mathematical Geology. v. 13, no. 1, pp. 1-22.

Sparks, R.SJ. and J. V. Wright. 1979. Welded air-fall tuffs.Geological SocietyAmerica Special Paper 180. pp. 155-156.

Talsma, T. 1969. In situ measurement of sorptivity. Australian J.Soil Research. v. 7, pp. 269-276.

Viera, S. R., D. R. Nielsen, and J. W. Biggar. 1981. Spatialvariability in field measured infiltration rate. Soil ScienceSociety of America Journal. v. 45, pp. 1040-1048.

Young, D. S. 1987. Indicator kriging for unit vectors: Rock jointorientations. Mathematical Geology. v. 19, no. 6, pp.481-501.

Al

tecanThsqipr4cOIdefoxtO

Ini

testiocoIwaRiupeiproanieawDamemeet abultespaiani

thaaniAncasanitesine%is r

Nei

mry

760Vo vol

, 14

I

JOURNAL * ASSOC. OF GROUND WATER SCIENTISTS & ENGINEERS * SEPTEMBER-OCTOBER 1994