SlideShare a Scribd company logo
1 of 13
Download to read offline
WATER RESOURCES RESEARCH, VOL. 30, NO. 2, PAGES 211-223,FEBRUARY 1994
Hydraulicconductivityestimationfor soilswith
heterogeneouspore structure
WolfgangDurner
InstituteofTerrestrialEcology,SoilPhysics,FederalInstituteofTechnology,Z•irich,Switzerland
Abstract.The hydraulicconductivityfunction,whichis requiredto solvetheRichards
equation,isdifficulttomeasure.Thereforepredictionmethodsarefrequentlyused
wheretheshapeof the conductivityfunctionis estimatedfromthe moreeasily
measuredwaterretentioncharacteristic.Errorsin conductivityestimationscanarise
eitherfroman invalidityof the predictionmodelfor a givensoil,or fromanincorrect
descriptionoftheretentiondata.Thisseconderrorsourceisparticularlyimportantfor
soilswithheterogeneousporesystemsthat cannotbe adequatelydescribedby the
usuallyusedretentionfunctions.To describetheretentioncharacteristicsofsuchsoils,
aflexible0(½)functionwasformedby superimposingunimodalretentioncurvesofthe
vanGenuchten(1980)type. By combiningthisretentionmodelwith the conductivity
predictionmodelofMualem(1976),conductivityestimationsforsoilswith
heterogeneousporesystemsareobtained.Estimatedconductivitiesbythismodeland
theclassicalvan Genuchten-Mualemmethodcandiffer by ordersof magnitude.Thus
reporteddisagreementsbetweenmeasuredandestimatedconductivitiesmayinsome
casesbedueto aninadequatedescriptionof theretentiondataratherthandueto a
failureof the predictionmodel.
Introduction
Theincreasingconcernwith groundwaterpollutionand
contaminationof soilsby hazardoussubstanceshasstimu-
latedthedevelopmentof numerousmathematicalmodelsof
pollutanttransportin soils.Today,thenumericalsimulation
ofwaterand pollutanttransportin unsaturatedsoilshas
becomea standardtool for assessingenvironmentalpoilu-
tionproblems.Despitetheexistenceandtheuseof simplis-
ticmodelsfor managementpurposes[e.g., Carsel et al.,
1985;Barraclough,1989](seeoverviewsby Addiscottand
Wagenet[1985]andLoagueandGreen[1991]),aswellasthe
recentintensivedevelopmentof specificmacroporemodels
[e.g.,Jarviset al., 1991;Chen and Wagenet, 1991;Gerke
andvanGenuchten,1993],the mostimportantapproaches
tomodelingtransientwater and solutetransportin the
vadosezoneare basedon the Richardsequation.To solve
thisequation,theknowledgeof the effectivesoilhydraulic
properties,namely,thesoilmoisturecharacteristic0(½)and
theK(O) or K(½) relationshipoverthe wholerangeof
moistureconditionsis required.Here, 0 (cubiccentimeters
percubiccentimeter)is the volumetric water content, ½
(centimeters)is the matricpressurehead,andK (centime-
tersperday)is thehydraulicconductivity(a scalarproperty
fortheone-dimensionalcase).For usein models,thehy-
draulicpropertiesare commonlyexpressedby analytical
functions,oftencalledparametericmodels,as,forexample,
listedby van Genuchtenand Nielsen [1985],Bruceand
Luxmoore[1986],andMualem[1986].The parametersof
thesemodelscanbe obtainedby fittingthe functionsto
experimentalwater retention and conductivity data. Alter-
natively,theycanbeestimatedfrommoreeasilymeasured
Copyright1994bytheAmericanGeophysicalUnion.
Papernumber93WR02676.
0043-1397/94/93WR_02676505.00
soilproperties[Bloemen,1980;Saxtonet al., 1986;WOsten
and vanGenuchten,1988;Vereeckenet al., 1989],or deter-
minedbyparameterestimationtechniqueswhentheinverse
problemis solved[Zachmannet al., 1981,1982;Hornung,
1983;Kool et al., 1987;Kool andParker, 1988].
Of all hydraulicproperties,the unsaturatedhydraulic
conductivityK is mostdifficultto measure.Thereforethe
use of indirect methodswhere K(0) is estimatedfrom more
easilymeasuredsoilpropertieshasbecomemoreandmore
common[van Genuchtenand Leij, 1992].A potentially
powerfulclassofmethodsresultsfrompore-sizedistribution
models,wherethewaterretentioncurveofaporousmedium
is interpretedasstatisticalmeasureof its equivalentpore-
sizedistribution[Corey,1992].In thisapproach,theconduc-
tivityisestimatedbyapplyingtheconceptof viscousfluid
flowthroughcapillariesandbyusingaconceptualmodelto
describethe pore interactionand pore connectivity
[Mualem,1986].Threebasicgroupsofstatisticalprediction
modelsmaybedistinguished:(1)thecapillary-bundlemodel
ofPurcell[1949]withitssubsequenttortuositycorrections
byYuster[1951],Burdine[1953],WyllieandGardner[1958],
andAlexanderand Skaggs[1986];(2) the cut andrandom
rejoinmodelof ChildsandCoilis-George[1950]withits
subsequentmodificationsbyMarshall[1958],Millingtonand
Quirk[1961],Campbell[1974]andothers;and,mostre-
cently,(3)theslabmodelofMualem[1976].Mualemand
Dagan[1978]andMualem[1986,1992]providecomprehen-
sivereviewsfor thesegroupsof conductivityestimation
models.
Closed-formexpressionsfortheretentionfunctionandthe
conductivityfunctionareoftenusedinparameterestimation
techniqueswheretheinverseproblemissolved[Kooland
Parker,1987a,b].Applyingapredictivemodelthatcouples
theconductivityfunctionwiththeretentionfunctionisthen
ofparticularadvantagebecauseitminimizesthenumberof
211
212 DURNER: HYDRAULIC CONDUCTIVITY ESTIMATION FOR SOILS
parameters needed to express the hydraulic relationships.
For the validity of this approachit is essentialthat the
prechosenfunctional representationof both the retention
curve and the conductivity curve as well as the model that
couplesthe two functions are appropriateover the whole
moisture range.
In principle, the statisticalmodelscanbe usedto estimate
absoluteconductivityvalues,but in practicethey are used
only to derive the shapeof the conductivityfunctionfrom
the shapeof the retentioncurve.To obtainabsolutevalues,
theestimatedconductivitycurvemustbe scaledwithatleast
onemeasuredconductivity,the "matchingvalue" [Nielsen
et al., 1960].Most frequently,the saturatedconductivityis
taken as matchingvalue, althoughseveral authors[van
GenuchtenandNielsen, 1985;Mualem, 1986;Luckneretal.,
1989;Nielsen and Luckher, 1992]have warnedagainstthis
practice.
Errors of conductivity estimates may arise from two
sources:(1) errorsdueto aninappropriatedescriptionofthe
soil moisture characteristic and (2) errors stemmingfrom
failureof thepredictionmodel.Beginningwiththeworkof
Nielsen et al. [1960], numerousstudies have compared
estimated with measured conductivities and have discussed
the validity of predictionmethodsfrom an empiricalor
semiempiricalpoint of view [e.g., KabIan et al., 1989;
Michiels et al., 1989;Stephens,1992].As a generalresult
from theseinvestigations,conductivityestimationmethods
appearto berelativelyreliablefor sandysoilswithnarrow
particlesize distributions,whereasfor loamyandclayey
soils they often fail. This failure is mostly attributedto
incorrect values of the tortuosity coefficient (or pore-
interactionfactor) r (dimensionless),whichis an empirical
parameterin thepredictingequations(see(6)inthetheory
section).SchuhandCline [1990]haveshownthatßcanvary
overa widerange.In additionto theseexperimentalinves-
tigations,theoreticaldiscussionsand reviews[Brutsaert,
1968;Gardner,1974;Mualem, 1976;Parkesand Waters,
1980;van Genuchtenand Nielsen,1985;Mualem, 1986;
AlexanderandSkaggs,1986]aswellasmathematicalsensi-
tivity analyses[Sakellariou-Makrar•tonakiet al., 1987;Vo-
gel and Cislerova,1988;Sidiropoulosand Yannopoulos,
1988]havebeenpublished.Despitetheextensivetreatment
ofthepredictionproblemanddueto thelackof a reliable
data base of measurementsin undisturbedsoils,the results
appearambiguousandinconclusive.A statementwhichwas
publisheda decadeagoappearsto be stillvalid:"The
problemofwhich[conductivity]measurementmethodand
whichpredictionmethodaremostvalidisundoubtedlyone
of the most vexingonesin contemporarysoil physics"
[Ragabet al., 1981,p. 384].
When discussingvalidity problems,mostinvestigators
have concentratedon the failure of the predictivemodel,
moreor lessdisregardingtheproblemofinterferenceswith
thefirst errorsource.No investigationhasbeenpublishedso
far on the systematicerrorsof conductivityestimation
methodsfor soilswith heterogeneouspore systemswhich
arisefromtheuseof nonoptimalwaterretentionfunctions.
Theobjectiveofthispaperistoanalyzeconductivityesti-
mationsfor suchsoils.After reviewingthecharacteristicsof
heterogeneouspore systems,conductivityestimatesare
obtainedby combininga flexibleretentionfunctionwith
Mualem's[1976]model.The resultsof thismethodare
comparedwithresultsobtainedbyclassicalunimodalmod-
els.Thedifferencesareanalyzedandthegenerallimitations
ofstatisticalconductivityestimationmethods,particularly
for soilswith a nonnegligiblefraction of largepores,are
discussed.
Theory
Water Retention Curves and Equivalent
Pore Size Distributions
Theporesystemofa rigidsoilcanbe characterizedbyits
equivalentpore-sizedistribution,which is commonlycalcu.
latedfrom the specificmoisturecapacityC* = dO/de.
Accordingto Laplace's law, the matric pressureat whicha
water-filled pore startsto drain is inverselyproportionalto
theequivalentradiusr of theporenecks.If C* asameasure
of pore densityis plottedversus• on a linear scale,pores
withlargeradiiofdifferentordersof magnitudearesqueezed
together,whichmaskstheshapeofthepore-sizedistribution
at the hydrologically important range close to saturation.
Sincethe matricpressureheadvariesover ordersofmagni.
tude,theretentioncharacteristicsarefrequentlyplotted0na
logarithmicpressureheadscale,whichis convenientlyex-
pressedinpF units,definedfor theunsaturatedrangebypœ
-- log•0(-½) [Schofield,1935],where ½ is expressedin
centimeters.The pF scale is equivalent to a logarithmic
scalefor the pore radius,and the derivativeof thewater
contentversuspF thereforeappropriatefor visualizingthe
pore-sizedistributionover differentordersof magnitude.
The so-definedpore-sizedensityis relatedto C* by
dO(r) dO(pF) dO(½t) de dO()
d logxor dpF d logo I,l d 1ogo
= [loge (10)]I½1C*.
Thepore-sizedistributionsthatare depictedinFigures
1-11 are calculatedby (1). Accordingly,the shadedarea
betweentwopF valuesindicatesthefractionofporespace
thatdrainswhenthepressureheadischangedbyoneorder
of magnitude.
Numerousmathematicalfunctionshavebeenusedinthe
literatureto express0(½)in analyticalform[Brooksand
Corey,1964;King,1965;Brutsaert,1966;Visser,1968:
Laliberte,1969;FarrelandLarson,1972;Rogowski,1972:
Su andBrooks,1975;Haverkampet al., 1977;Gillham,
1976;Simmonsetal., 1979;D'Hollander,1979;vanGenu.
chten,1980].Mostpopularamongthesefunctionsarethe
equationsofBrooksandCorey[1964](hereinanotationthat
isslightlymodifiedfromtheoriginalversion)'
= >
s, = I 1
and,morerecently,theequationofvanGenuchten[1980]:
Se= [1+ -m,
whereSeistheeffectivesaturation(dimensionless),defined
bySe= (0- 0•)/(0•- 0•),withOsand0•indicatingthe
saturatedandresidualvolumetricwatercontents,respec-
tively,a> 0(centimeters)isascalingfactorthatdetermines
thepositionoftheporesizemaximum,andX,n,andmare
dimensionlesscurve-shapeparameters••bjecttoA>0,m
>0, andn > 1.
DURNER: HYDRAULIC CONDUCTIVITY ESTIMATION FOR SOILS 213
-103
• 8
0.25 ......................................................................................1 •
_10-2 _104 _10ø _10• _102
reducedpressureheadW*
Figure1. Retentioncurve(solidline), first derivative
dSeldlog•* (shadedarea),andsecondderivatived2Se/d
(log•,)2 (dashedline)ofthevanGenuchtenfunctionwith
n = 1.5 and rn = 1 - 1/n. The reducedpressurehead
(dimensionless)is definedby •/a. The shadedarearepre-
sentsthepore-sizedistribution.The positionof maximum
poredensityisat½*= -(I/m)
Equation(2)and(3)caneasilybecombinedwithconduc-
tivitypredictionmodelsto obtainclosed~formexpressions
forK(O) andK(•) [Mualem, 1992].By this, both 0(•) and
K(½)are determinedby the parametersof the retention
curveplus one additionalmatchingfactor to scale the
predictedhydraulicconductivityfunction.BrooksandCorey
[1964]combinedtheirfunctionwith thepredictivemodelof
Burdine[1953]; Campbell [1974] combinedthe samereten-
tionfunctionwith a simplifiedversion of the Childs and
Coilis-George[1950]predictionmodel.The retentionmodel
ofvanGenuchten[1980]is mostfrequentlycoupledwith the
predictivemodelof Mualem [1976].The resultingset of
hydraulicfunctionshas been thoroughlyreviewed by van
Genuchtenand Nielsen [1985], Luckner et al. [1989], and
Nielsenand Luckner [1992]. Kool and Parker [1987a] ex-
tendedit to include hysteresisand air entrapment.
All the above-listed retention functions reflect unimodal
and,if continuouslydifferentiable,smoothnormalto lognor-
malshapedpore-sizedistributions.This is illustratedby
Figure1,wheretheretentionmodelof van Genuchten[1980]
isdepictedtogetherwithitsunderlyingpore-sizedistribution
(shadedarea). As comparedto the Brooks and Corey
thnction,the van Genuchten function has the favorable
propertythat its derivative dO/dpF is continuousand be-
comesasymptoticallyzero toward the fine and largepores.
Theextensionof the pore-sizedistributiontowardthefine
poresisdeterminedbytheproductmn(whichfor a½>> I is
equivalentto • in the Brooks and Corey function), whereas
theextensiontowardthe largeporesis determinedby the
ratiom/n.Thelargerthisratiois,thesteeperisthedecrease
inporedensityfromitsmaximumtowardthelargepores.As
•ill beshownlater, the width of the pore-sizedistribution
towardthelargeporesis of particularimportancefor esti-
matingthe conductivityfunction. Summarizing,the van
Genuchtenretentionmodelinitsgeneralform(3)represents
acontinuous,smooth,unimodal,and bell-shapedpore-size
distributionwiththeparametera primarilydeterminingthe
positionoftheporedensitymaximumandtheparametersm
andndeterminingthewidthtowardthefineandlargepore
sizes.VanGenuchtenandNielsen[1985]comparedvarious
•ater retention models and concluded that (3) fits the water
retentiondataof mostsoilsalmostperfectly.
However, to obtain the conductivity estimates as closed-
form functions,the parametersof the van Genuchtenequa-
tion mustbe subjectedto the additionalconstraintm + l/n
- i, where i is an integer value, for the van Genuchten-
Mualem model generally taken as unity [van Genuchten,
1980].This constraint eliminates someof the flexibility of the
function, since a pore-size distribution that extends far
toward fine pores is then always coupledwith a relatively
broad distribution towards the large pores, and vice versa.
Thisis in fairly goodaccordancewith experimentaldatafor
many soils.There are casesof fine-textured soils, however,
where this constraint attributes a part of the pore spaceto
unrealistically large pores.
There is some discussion whether the van Genuchten
functionis a physicallyrealisticdescriptionof 0(•) overthe
whole moisturerangewhen its parametern lies in the range
1 < n < 2 [Nielsen and Lucknet, 1992]. In this case, the
secondderivative,d20/d•2, becomesinfinitetowardsatu-
0
ration,andtheintegral.f0,• d0forlim•0 isunbounded.
This means that an infinite amount of work were necessary
to extract all water from the soil [Mualern, 1978;Sakellario-
Makrantonaki et al., 1987]. According to our view, for
practicalpurposesthisdoesnotrestricttheusefulnessofthe
van Genuchten function as an empirical model for the
retentioncurve. Nevertheless,it appearsimportantto real-
izethatit expressestheretentioncharacteristicaswellasthe
derivedpropertyC* andtherelatedconductivityestimation
in a purelydescriptivefashion.The coefficientsshouldnot
be interpretedasparametersin a physicalsense.
Water Retention Characteristics of Soils
With HeterogeneousPore Systems
Undisturbed soilsfrequently have pore systemsthat are
differentfrom the unimodal,approximatelynormaldistrib-
utedtypewhichis representedby Figure1. Consequently,
attemptsto fit theirretentiondatawitha simplesigmoidal
curveleadto unsatisfyingresults.A schemeof typicalfitting
errorsisdepictedinFigure2. It appearsthatsuchdeviations
aretacitlyassumedto bewithintheconfidenceintervalof
thecorrespondingmeasuredvalues.If theyare not, such
deviationswould be relevant as will be shown later. This
could be used to define the term "heterogeneous" with
respectto hydrologicalproperties.Wecandesignatea pore
systemasheterogeneousif thepore-sizedistributionof a
representativeelementaryarea[BearandBachmat,1991]
cannotbe correctlydescribedby the van Genuchtenfunc-
tion(oranyotheroftheabovementionedunimodalretention
functions).
Poresystemsthatarenonconformwithasimplesigmoidal
retentioncharacteristicmay be a resultof specificparticle-
size distributionsor be due to the formation of secondary
poresystemsbyvarioussoilgeneticprocesses.Aggregation
processesinloamysoilsmayleadtopore-sizedistributions
thatcausetypicalfittingproblemsin the midporerange,as
illustratedin Figure 2 I,bottomleft) [e.g., Sharma and
Uehara, 1968;Smettemand Kirkby, 1990].Besidesaggre-
gation,biologicalsoil-formingprocessesmayleadtosecond-
aryporesystemsin thelarge-porerangein soilsof any
texture{Figure2 (topleft))[e.g.,Othmeret al., 1991;De
Jongetal., 1992].Morainicsoilsandsolifluctionsoilstend
tohavea significantfractionofporesovera widerangeof
tensions,whichleadsto fittingproblemsas illustratedin
Figure2 (topright)[e.g.,ParkesandWaters,1980;Jacob-
214 DURNER: HYDRAULIC CONDUCTIVITY ESTIMATION FOR SOILS
pressure head
Figure 2. Schematictypes of deviationsbetween retention
data of structured soils and fitted unimodal retention curves.
The circles indicate typical trends, which help to identify
structured pore systems even in cases where only few
supporting data are available. (Top left) Increasing devia-
tions between data and fitted curve when approachingsatu-
ration, observed in undisturbed soils of any texture [e.g.,
Othrner et al., 1991;De Jong et al., 1992]. (Top fight) More
or less straight retention characteristic in the wet rangeup to
pF3-4, where the main pore system is located. Typical for
morainic and solifiuction soils [e.g., Schj•nning, 1985; Ja-
cobsen, 1989]. (Bottom left) Different slopes of fitted curve
and data in the midpore range. Typical for aggregatedloams
[e.g., Sharma and Uehara, 1968; Smettem and Kirkby,
1990]. (Bottom right) Considerable change in water content
very close to saturation,experimentally observedin uncon-
solidatedsands(see,for example, King [1965](liquid: soltrol
C), Ragab et al. [1981],and Stephensand Rehfeldt[1985]).
mi) aresubjectto theconditionso•i > 0, mi > O,ni > 1.
Weexplicitlydonotimposetheadditionalconstraintmi +
1/ni= 1. Figure3 showsthesuperpositionoftwohypothet.
ical pore-sizedistributionswhich definea bimodalpore
systemaccordingto (4). The dark-shadedarea canbe
interpretedasfine-texturedpore system,with a maximum
pore-sizedensityaroundpF4 (& r = 0.2/am), whereasthe
light-shadedareacan be seenas a secondaryporesystem
with its maximum density in the range of r = 0.1 mm.
The superpositionof two unimodalporesystemshasbeen
previouslyusedby PetersandKlavetter[1986]to describe
the hydraulicpropertiesof fracturedrock. Othmeretal.
[1991]andRossand Smettem [1993]usedthis approachto
describe the retention characteristics of soils with distinct
secondarypore systems.Pruesset al. [1990]modeledthe
coupledtransportof water, heat, and air in partiallysatu-
ratedporousrockby explicitlyconsideringfractureeffects.
They comparedthe resultswith those obtainedby treating
theporousmediumasa singlecontinuumandderivedsimple
criteriafor the applicabilityof the effective continuumap-
proximation.WangandNarasimhan[1985,1990]addressed
theproblemof findingeffectiveparametervaluesandcom-
paredfracturesin tuff with macroporesin soil.
The multimodal retention model, (4), keeps the functional
propertiesof thebasicvan Genuchtenmodel,but is,dueto
the increased number of coefficients, able to account forthe
deviations discussedabove. It is continuously differentiable,
asymptoticto a zero slopetowardsthe fineandlargepores,
andstronglymonotonicover the whole moisturerange.By
keepingk small,the functionis well-behavedfor interpola-
tion purposes,reducingthe noisein measureddatawhile
followingtheshapeof the measuredcurve.Hampton[1990]
showedthat the useof unconstrainedsplinesfor describing
sen, 1989;Lafolie et al., 1989].With unconsolidatedsands,
retention characteristics of the type in Figure 2 (bottom
right)havebeenobserved,evenin caseswherethiscannot
bedueto boundaryeffectsat thebeginningofdrainage[e.g.,
King, 1965;Ragab et al., 1981;Stephensand Rehfeldt,
1985].In naturalsoils,the schematictypesof Figure2 will
always appearto some degreein combination(compare
examplesof Boelset al. [1978],CurrieandRose[1985],
Dixon [1972],Hantschelet al. [1987],Jeppsonet al. [1975],
Parker and van Genuchten[1985],Puckett et al. [1985],and
Tsuji et al. [1975]).Note that theseheterogeneouspore
systemscanbefoundforfieldaveragesaswellasforsingle
soilsamples[e.g.,AndersonandCassel,1986].
Multimodal Retention Function
Toproperlydescribetheretentioncharacteristicsofsoils
withheterogeneousporesystems,we introducea multimo-
dal retention function which is constructedby a linear
superpositionof subcurvesof the van Genuchtentype
[Durner, 1991, 1992]:
k
Se--Z wi
i=!
1+ (ai!½!)n" '
(4)
wherek is the numberof "subsystems"thatformthetotal
pore-sizedistribution,andwi areweightingfactorsforthe
subcurves,subjectto0 < wi < 1and5'.wi = 1.Asforthe
unimodalcurve,the parametersof the subcurves(ori, ni,
._o 0.8
'• 0.6
ß.= 0.4
• 0.2
0 2 4
pF
0.2 ...........
1500 15 .15
poreradius(I,Lm)
.0015
Figure3. Constructionofamultimodalretentionfunction,
(4).(Top)Bimodalretentioncurve(solidline),unimodal
subcurveforthetexturalporesystem(dottedline,wi =0.6,
al =0.005cm-l nl = 14 mi= 1- 1/ni)andunimodal
subcurvefor a secondaryporesystem(dashedline.w2=
0.4,a2= 0.1cm-1 n2= 2 2 m2= 1- 1/n,).(Bottom)
Pore-sizedistributionsof bimodalfunction(s•lidline),of
texturalporesystem(darkshaded),andof secondarypore
system (light shaded).
DURNER: HYDRAULIC CONDUCTIVITY ESTIMATION FOR SOILS 215
pF pF
retentiondatacanleadtoconsiderableproblemsinnumer- 0
icalsimulationsofwatertransportduetotheroughshapeof 1.2 10ø
thewatercapacityfunction.Incontrasttoconstrainedspline
interpolations,whichappearstobethebestalternative•,, 10scheme,(4)requireslesscoefficients.Finallyitscoefficients• 0.8
can,toacertaindegree,beinterpretedinthesamemannerasthebasicvanGenuchtencoefficients,namely,•i=l/oti, 0.4
indicatingthepositionsofporedensitymaxima,andhimi ß •ø•
andni/mi, determiningthewidthoftheunderlyingpore-size 0.2
distributions.However,if theporesystemisnotdistinctly 0
bimodalor multimodal,thecoefficientsof the multimodal
retentionfunctiontendtobehighlycorrelated.It istherefore
importantthattheynotbeconsideredas"parameters"with
aphysicalmeaning;rathertheymustbeseenascurveshape
coefficientslikethoseof anyalternativeinterpolationfunc-
tion.
Themultimodalretentionfunctionis fitted to data by
minimizingtheobjectivefunctionZ:
N
g(P)= • •oill0i- • (½i,P)II (5)
i=1
whereP= {0s,Or,ai,hi,mi,wi}Tistheparametervector,
Nisthenumberof measureddata, ooiaretheweightsfor the
leastsquaresminimization,p.determinestheminimization
norm,andOiandb(½i,P) arethemeasuredandcalculated
watercontentsat pressure½i, respectively.If p. = 2, (5)
resultsin a weightedleast'squaresscheme,which corre-
spondstoamaximumlikelihoodfitundertheassumptionof
independentandnormallydistributeddataerrors.If outliers
aremorefrequent,p.canbe setto unity.In thiscase,(5)
minimizesthe sumof absolutedeviationsbetween measured
watercontentsandthe retention curve. The fitting procedure
weuseis basedon a multidimensionalbracketing method.
Thenumericalprocedureisrobustandunconditionallycon-
vergent,butcomputationallynot very efficient.For each
coefficientPi, thefollowingpropertiesmaybetailoredfor
specificneeds'(1) initial guessof Pi, (2) Pi constantor
optimized,and(3)allowedminimumandmaximumvalueof
Pi.Theoptimizationisterminatedwhentheimprovementof
theobjectivefunctionorthechangesofallparametervalues
arebelowtheir tolerance values between two iteration steps,
(•Z< Sz,Api< epi),orwhenthemaximumnumberof
iterationsisexceeded.(The programis availablefromthe
authoronrequest.)
ConductivityEstimation
Therelativehydraulicconductivityfunctioniscomputed
by numericalevaluation of Mualem's [1976] predictive
modelonbaseof the unimodalor multimodalrepresentation
ofq•(Se):
g--SeLf()]'
(6)
x•heref isgivenby
Se1,is;.
f(Se)-- ½(5;)
(7)
In[6),SJisanempiricalcorrectionfunction,whichallows
foraprioriunknowneffectsofporeconnectivityandtortu-
1õ 1.5 0.01õ 0 0.2 0.4 0.6 0.8
poreradius(I.U'n) effectivesaturation
Figure4. Conductivityestimationfor a structuredpore
system.(Left) Bimodalretentioncurve (solid line) and
underlyingpore-sizedistribution(shadedarea).(Right)Pre-
dictedrelativeconductivityfunctionsK(½) (fromtopleftto
bottomright,relatedtotopaxis)andK(Se) (fromtopfight
to bottomleft, relatedto bottomaxis).The coefficientsare
}t,•= 0.1, a• = 0.2cm-1, - ,
= 0.1, a2 = 0.001cm-l n2 = 22 andm2 = 1 - 1/n,.
osityin Mualem'smodel.Mualem[1976]foundanaverage
valueofr = 0.5tobeoptimalfora setof45samplesoils.The
absoluteconductivityKabs(Se)canbecalculatedbymatch-
ingthepredictedfunctionK(S•) to a matchingvalueKref,
measuredat somereferencesaturationSref, accordingto
Kabs(Se)= (Se/Sref)•'KrefK(Se)' (8)
Results and Discussion
ConductivityEstimationfor HeterogeneousPoreSystems
Forthepurposeofcomparingconductivityestimatesfor
bimodalandmultimodalsoils,we assumeat this pointthat
boththe retentionmeasurementsof the examplesoilsand
thepredictivemodelitselfarefreeoferror.Ofcourse,these
assumptionsare notgenerallyjustifiedandhaveto be
discussedlater in their own right.
InFigure4,thehydraulicpropertiesofahypotheticalsoil
with a narrowtexturalpore-sizedistributionand distinct
secondaryporesystemareshown.Twofindings,whichare
ofgeneralnature,areclearlyillustrated:(1)theshapeofthe
retentioncurveisdirectlyreflectedintheshapeof K(½)and
(2)thesecondaryporesystemincreasestheconductivityby
someordersofmagnitudewhiletakingonlyfewpercentof
porespace.TheparticularshapeofK(0)cannotbereflected
byunimodalmodels.The"similarity"betweentheshapes
of0(½)andK(½)isphysicallymeaningful:atpressureranges
wherethe specificwatercapacityis high,indicatedby a
steepslopeoftheretentionfunction,theconductivitymust
dropsteeplyaswell,sinceduringadrainageprocessalarge
portionoftheporespaceemptiesin thatrange.A steep
decreaseoftheconductivityclosetosaturationistypicalfor
soilswith a secondaryporesystemin the large-porerange.
Compositeconductivityfunctionswithsimilarcharacteris-
ticshavebeenderivedfromexplicitconsiderationof two
poroussystems,e.g.,for fracturedrock[Wangand
Narasimhan,1985;PetersandKlavetter,1988]orforsoils
withmacropores[Othmeretal., 1991;ChenandWagenet,
1991;Chenet al., 1993].
If oneattemptstodescribeabimodalsoilwithsimplified
unimodalhydraulicfunctions,theerrorinvolvedismuch
216 DURNER: HYDRAULIC CONDUCTIVITY ESTIMATION FOR SOILS
100
......"' 0 2 pF 4 6...... 0 2 pF4 6
..-..........,. c .... 11.1.:::::::...%I,,,"" //' 0.4 ............ ),
10'••or ?? function
,•o-•/ ,/ A'tr•' • o.•t-........ "•-_I •
.......:....,¾,.....i ,::..,i...............h• 10•
• o '150 1.5 0.015 0.1 0.2 0.3 0.4 0.5
10'4 ix)reradius(!.u'n) watercontent
Figure7. Hydraulicpropertiesof RideauClay Loam[De
Jonget al., 1992].See Figure6 for explanations.
0 0.1 0.2 0.3 0.4 0.5
water content
Figure 5. Schematic representation of the conductivity
calibration problem for bimodal pore system. Line A, con-
ductivity curve of a bimodal pore system; line B, unimodal
predicted curve matched to unsaturated conductivity; line C,
unimodal predicted curve matched to saturated conductiv-
ity; and line D, unimodal predicted curve with optimized
pore interaction factor r.
large[for theconductivityestimationsthanfortheretention
function. Matching the unimodal conductivity function with
the measuredsaturatedconductivity value Ks, asis common
practice,will leadto an overestimationof the unsaturat6d
conductivity in the whole unsaturatedmoisturerange(Fig-
ure 5, curve C). If, on the other hand, the unimodalpredicted
function is matched to an unsaturated conductivity value,
the increaseof hydraulicconductivitynear saturationis not
reflected, thus leading to a gross underestimationof the
soil's conductivity near saturation (Figure 5, curve B).
However, the useof this functionin simulationsof unsatur-
ated water transportleads to a valid descriptionof the
hydraulicsystemas longas saturationis notapproached
[WangandNarashimhan,1985].The double-porositychar-
0 1000.6
0.4
0.2
pF pF
2 4 6 0 2 4 6....
i i 1
'"•,'....-.'[•-unimodal-",, • trimodal
:•-'.,-.,.c:-'x
15 0.15 0.0015 0 0.2 0.4 0.6
poreradius(I.u'n) watercontent
Figure6. Hydraulicpropertiesof SolarVillageClay
[Hampton,1989].(Left) Measuredretentiondata(solid
circles),fittedbimodal(solidline),andunimodal(dashed
line)retentioncurve,andcorrespondingbimodal(black
shaded)andunimodal(lightshaded)pore-sizedistributions.
(Right)Predictedbimodal(solidlines)andunimodal(dashed
lines)conductivitycurves.TheK(½)curves,fromtopleftto
bottomright,arerelatedto thetopaxis.TheK(0) curves,
fromtopfighttobottomleft,arerelatedtothebottomaxis.
The coefficientsof thefunctionsarelistedinTable1.
acteristic of the conductivity function cannot be reflectedby
the unimodal estimation, even if the tortuosity factor• is
manipulated(Figure 5, curve D).
In Figures6, 7, and 8, measuredretention data of three
soils are depicted, representing structured pore systems
which can be frequently found. The chosen examplesare
similar to the schematic soil types described in Figures2
(bottomleft), 2 (top left), and 2 (top right), respectively.In
additionto the experimentaldata, Figures6-8 showtheleast
squaresfittedunimodalandmultimodalretentioncurves,the
underlying pore-size distributions, and the estimatedcon-
ductivity curves. Some physical propertiesof the soilsand
the coefficients of the fitted functions are listed in Table 1.
Unlike the unimodal function, the multimodal functionde-
scribesthemeasureddataof eachsoilaccurately,anditis
obvious that in each case the two functions representfun-
damentally different pore systems.
Figure6 (left)depictsthehydraulicpropertiesofacolumn
packedwith smallclay aggregates[Hampton,1989].The
pore-sizedistributionisdistinctlybimodal,witha maximum
poredensityoftheintra-aggregateporesatr ---0.01/xm,and
a secondarymaximumfortheinteraggregateporesatr --20
/xm.Whereastheprimaryporesystemisdeterminedbythe
particle-sizedistributionofthatsoil,theporespacebetween
theaggregatesdependsonlyonthepackingandistherefore
essentiallyindependentof the soiltexture.Notethatthe
existenceandextentof secondaryporesystemsinundb
turbedsoilsposesa limitationto thesuccesspotential0f
regressionmethodsthatseekto predicthydraulicrelation-
shipssolelyfromparticle-sizedistributions.Figure7(left}
showsdataof anundisturbedcoreof Rideauclayloam[D}
pF
0 2 4
0.5 i i i i
0.4
0.:3
0
15 1.5 0.015
poreradius(llm)
pF
6 0 2 4 610
10•li10•
0 0.1 0.2 0.3 0.4
water content
Figure8. HydraulicpropertiesofROnhaveSandyL0an•
[SchjOnning,1985].SeeFigure6 for explanations.
DURNER: HYDRAULIC CONDUCTIVITY ESTIMATION FOR SOILS 217
Table 1. Parametersof Example Soils
Classification
Rideau R0nhave
Solar Vii!age Clay Sandy
Clay* Loam? Loams AggregatedLoamõ
Texture
Type -3
Density,gcm
Porosity,%
Samplevolume,cm3
log(Ks),ms-•
Unimodalvan Genuchten
Os,cm3cm-3
Or,cm3cm-3-1
a, cm
m
clay loam heavy clay coarsesandy
aggregates loam
repacked undisturbed undisturbed
1.3 1.55
52.7 41.3
12.3 345 100/250
-5.7 -4.6 -5.3
0.556 0.439 0.403
0.000 0.000 0.000
0.0328 0.0027 0.0212
1.160 1.134 1.181
0.138 0.!18 0.153
loam aggregates
repacked
0.98
0.63
750-1500
-4.4
Bimodal Bimodal
Multimodal van Genuchten
Trimodal Bimodalll Bimodalô Bimodal**
Os,cm3cm-3 0.556 0.473
Or' cm3cm-3 0.044 0.132
wl 0.54 0.14
-• 0.0252 0.6634
al, cm
nl 2.482 2.060
rnl 0.243 0.515
w2 0.46 0.86
- • O.00006 O.O011Ot2, cm
n2 2.272 1.286
rn2 0.825 0.223
W3
-1
Ot3, ½m
rt3
m2
0.403 0.633 0.633 0.633
0.000 0.160 0.160 0.154
0.21 0.68 0.65 0.62
0.715 0.4422 0.3320 0.2799
1.349 4.675 3.123 2.849
0.210 0.304 0.680 1.0130
0.16 0.32 0.35 0.38
0.0093 0.0119 0.0112 0.0084
1.550 3.471 2.527 1.741
0.355 0.422 0.604 1.000
0.62
0.00017
1.818
0.450
*Source:D. R. Hampton(personalcommunication,1989).
?Source:DeJonget al. [1992].
$Source:SchjOnning[1985].
õSource:Smettemand Kirby [1990].
IlSoconstraint.
ôHere,rni = 1 -- 1/ni.
**Here mi = 1.
Jonget al., 1992].Thepore-sizedistributionis againbi-
modal,but the secondarypore systemis now lesspro-
nouncedandshiftedtowardlargerporesizes,ascompared
tothepreviousqxample.Forthissoilwecandeducethe
existence,but notthe actualshapeof the secondarypore
system,becausethe measurementsare insufficientin the
importantrangenearsaturation.Forthepurposeofanalyz-
ingbimodalconductivityestimationsof suchporesystems
weassumeherethat the fittedbimodalfunctionis a valid
modelofthesoil'strueretentionpropertiesevenintherange
whereno data are available. As a third example, the reten-
tionpropertiesofanundisturbedsampleofR0nhavesandy
10am[SchjCnning,1985]aredepictedinFigure8(left).Ina
strictsense,thisporesystemisunimodal.Nevertheless,it
cannotbeadequatelydescribedbytheclassicalsigmoidal
retentioncurves.Thecontinuous,almostlineardecreaseof
thewatercontentfrompFOtopF3indicatesaconsiderable
poredensityovera widerangeofporesizes,whereasthe
mainporesystemislocatedin therangeof finepores.
Acomparisonoftheconductivityestimatesfromtheuni-
andthemultimodalretentionfunctions(Figures6(right)to8
(right))showsthefollowing.
1. Conductivityestimatesfromuni-andmultimodalde-
scriptionsoftheretentioncurvecandifferbyordersof
magnitude.ThesedifferencesarevisibleinboththeK(½)
and the K(0) plane.
2. Thelargestdifferencesoccurin thewet moisture
range,whichis veryimportantfor solutetransportpro-
cesses.
3. ThedifferencesintheestimatedK(½)curvesarisein
mostcasesfromdifferentslopesnearsaturation.Asthesoil
dries,thecurvesremainapproximatelyparallel(Figure7)or
approacheachotheragain(Figures6and8),providedthe
sameresidualwatercontentisapproachedbytheretention
functions.
Simplifiedsigmoidalretentionfunctionscannotreflectthe
smallbutimportantchangesinwatercontentthattakeplace
inporerangesdistantfromthemainporesystem.The
limitingcaseforthisinabilityisretentionfunctionsofthe
typeof(2),wherebothwatercontentandconductivityare
218 DURNER: HYDRAULIC CONDUCTIVITY ESTIMATION FOR SOILS
constant between saturation and the air entry point. Func-
tions of thistype representdiscontinuouspore-sizedistribu-
tions and are by definition not adequatefor conductivity
estimations in loamy and clayey soils.
Whether the conductivity estimation by a nonoptimal
retention function will tend to lead to an overestimation or
an underestimationcan be simply evaluatedby comparing
the slope of the fitted retention curve with the slopeof the
measured data. If the fitted retention curve underestimates
the true water release near saturation (as in Figures 2 (top
left), 2 (top right), 2 (bottomright), 7 (left) and8 (left)), the
predictedchangeofK(½) will aswell be smallerthanthetrue
change.Consequently,if matchedto Ks, thepredictedcurve
results in an overestimation. In studies where conductivity
predictionshave been comparedwith measurements,this
type of predictionerrorhas,indeed,frequentlybeenfound
[e.g., Parkesand Waters, 1980].If, on the otherhand,the
fitted retention curve indicatesa steeperslope than the
measuredwater content data (as in the wet rangeof Figures
2 (bottomleft) and6 (left)),thepredictedchangeofK willbe
larger than the change that were predictedby a more
accurateretentionfunction. Accordingly,thetwo estimated
K(•) curvesof the morainicsoil(Figure8) approacheach
otheragainatdrierconditions,sincetheunimodal0(½)curve
hasa steeperslopethanthemultimodaloneat intermediate
pore sizes.
Usingconductivityestimationsfrommultimodalretention
curvesas an investigationtool, someof the classicalprob-
lems of the predictionmethodswill be discussedin the
followingsections,namely(1) theroleof theparameterOr,
(2)thequestionwhethertheparticularmathematicalformof
the retention function is of relevance, and (3) the sensitivity
of conductivitypredictionson the slopeof the retention
curve near saturation, which will be found to be most
important.
Role of 0r and Os
Throughoutthehistoryof conductivityestimations,the
searchfor a "correct" residualwater content0r hasbeena
continuoussourceof vexation.Statementsonthe roleof Or
for conductivitypredictionsrangefrom"veryimportant"
[Sidiropoulosand Yannopoulos,1988]to "unimportant"
[VogelandCislerova,1988].Partly,thismaybeduetothe
factthatthereis noagreementonthedefinitionand,conse-
quently,thecorrectdeterminationofOr[Sidiroupoulosand
Yannopoulos,1988;Lucknetet al., 1989;Nimmo,1991;
NielsenandLuckner,1992].Mualem[1976],whoconsidered
the correctvalueof the residualwatercontentto be impor-
tantfortheprediction,devoteda considerablepartofhis
classicalpaperto thedeterminationof Or-VanGenuchten
[1980]proposedtosetOrequaltoameasuredwatercontent
at a highsuction.Recentreviews[vanGenuchtenand
Nielsen,1985;NielsenandLuckher,1992]recommendtreat-
ing Orasa purefittingparameter.
From(6)we canseethatin principletheconductivity
predictioninthewetrangeisnotaffectedbythevalueofOr,
sincethehydraulicconductivityofa soilisdeterminedat
anysaturationbyonlyafewpercentofthelargestwater-
filledpores.Durner[1992]confirmedthisindependency
usingwaterretentiondataofa loamfromRoskilde,Den-
mark(similartoFigure8).Contrarytoaunimodalmodel,a
changeof Orfrom0.0to 0.19hadnoinfluenceonthe
conductivitypredictionintherangeoflargepores,sinceit
didnotaltertheshapeofa fittedbimodalretentioncurvein
thatrange.Anyinfluenceof Orinunimodalhydraulicmodels
is thereforecausedby interferencesof the curveshape
parameters.For the vanGenuchtenretentionmodel,asan
example,a decreaseof thevalueof Orcausesa shiftofthe
parametern toward a smallervalue. A similareffectis
causedby an increaseof the parameterOs.Thereforeifthe
interest lies purely in the wet moisture range, we can
disregarddatapointsin thedry rangeif thisleadstoabetter
descriptionin the rangeof interest.
Effectsof parametercorrelationscanbe shownbydataof
Kablan et al. [1989], who comparedconductivitymeasure.
mentsfor a fine sandwith conductivity estimatesbythevan
Genuchten-Mualem model. In the fitting procedureforthe
retentioncurve,the parameterOswas setequalto thelargest
singlemeasuredwater content.Sincefield datawitha
considerablescatter were used, this value was far higher
thanan optimallyfitted value.As a consequence,the0pti.
mizedvaluesof Orand n bothbecametoo small,andthe
predictedconductivityfunctioncorrespondinglyunderesti.
mated the decrease in conductivity at low saturation,thus
leadingto the conclusionthat the predictivemodelworked
well at highsaturations,but failed at low saturations.The
exampleexhibitsa dilemmain fittingproceduresforthe
retention curve. To get a best overall fit, all parameters
shouldbe simultaneouslyoptimized [compareNielsenand
Luckner,1992].An independentdetermination,asisdesir-
ablefor parametersof physicalsignificance,e.g., intrans-
portmodels,ishereinappropriate.Ontheotherhand,ifan
optimizedvalueof Osislowerthatactuallyobservedvalues,
informationrelevantfor K(0) at largeporesmaybemasked
bytheoverallfittingprocedure[Mualem,1992].
Mathematical Formulation of the Retention Function
Asisexpressedby(6),theconductivitypredictionmodels
estimatetherelativeconductivityfunctionpurelyfromthe
shapeoftheretentionfunction.Thuswecanpostulatethat
thespecificmathematicalformofa retentionmodelhasno
influenceon the conductivitypredictionas longasit pro-
videsaproperdescriptionofthemeasureddata.Weverify
thisbydescribingastronglyaggregatedloamwithabimodal
poresystem(databySmettemandKirkby[1990])usingthree
differentretentionmodels,basedonthebimodalformof(4):
(1)theunconstrainedmodelwithvaryingmiandhi,(2)the
constrainedmodelwithmi" 1- l/hi, and(3)asimplified
formwithmi = 1.Forcomparison,wealso•interpolatethe
retentiondatalinearly,withextrapolationsin thedryrange
to watercontentzeroat pF7. To definethepore-size
distributiontowardthelargepores,themeasuredvalueof0
atpFOistakenasOs.Figure9showstheretentioncharac-
teristic,theunderlyingpore-sizedistribution,andthepre-
dictedK(½)curve,asrepresentedbythislinearinterpola-
tion.Thepore-sizedistributionhasaveryruggedshapeand
isnoncontinuous.Suchadescriptionofthehydraulicprop-
ertieswouldcauseproblemsin numericalsimulations
[Hampton, 1990].
Acomparisonofallfourretentionmodelsandtherelated
estimatedconductivityfunctionsisshowninFigure10.The
agreementamongthefourmodelsaswellastheagreement
betweentheestimatedandmeasuredconductivitiesisal-
mostperfect(Figure10(right)).We concludethatany
smoothfunctionwhich describesthe retentiondataaccu-
ratelyappearstobesuitableinthiscase,regardlessofits
DURNER:HYDRAULICCONDUCTIVITYESTIMATIONFORSOILS 219
0
o.4
0.2
pF
,,
i i i i
15oo 150 15 3
poreradius
!.5 0 I 2
pF
1
1oø •_
10'2'• >
I0"•• 0,2
Figure9. Hydraulic properties of AggregatedLoam
[SmettemandKirkby, 1990].(Left) Experimentalretention
data(solid circles), linear interpolation (solid line), and
pore-sizedistribution(shadedarea),asrepresentedby the
linearinterpolated retention curve. (Right) Estimated con-
ductivityfunction.
0 ! 2 8 4 5 0 5
pF
I 2 3 4
pF
10ø
10•-•
10.48
10.• '•
Figure11. Dependenceof conductivityestimationsonthe
asymptotic slope of the retention curve near saturation.
(Left) Retentioncurves-dottedline, Brooksand Corey
function(2) (a-t = 10 cm, A = 0.15);solidline, van
Genuchtenfunction(3),withconstraintrn= 1- 1/n(a- t =
10.5cm, n = 1.15, nm = 0.15); dashedline, vanGenuchten
functionwithoutconstraint(a-i = 10.5cm,n = 1.0, nrn=
0.15). (Right)Predictedconductivitycurves.
mathematicaltype. As will be shownnext, thisis, however,
onlytrueaslongasthe functionsdo really correspond,and
donotdifferin their asymptoticbehaviortoward saturation.
Slopeof the RetentionCurve Near Saturation
Theasymptotic slope of the retention function near satu-
rationis the most sensitive determinant for conductivity
estimationsif smoothretention functions (i.e., functions that
representcontinuouspore-size distributions) are used. This
originatesfrom the use of the Hagen-Poiseuillelaw in the
predictionmodels,which statesthatthe flux densityin pores
ofacertainsizeis inverselyproportionalto thesquareof the
poreradius.Consequently,all statisticalconductivitypre-
dictionmethodsintegrateover the inverseof • [Mualern,
1992].The smallerthe suctioniswhere changesin the water
contenttake place, the greater is its contribution to the
conductivityintegral(e.g., (7)). If retentionfunctionswith a
distinctair entrypointare usedfor the conductivityestima-
tion[Brooksand Corey, 1964;Campbell, 1974],the inte-
pF pF
o 3 o 1 2ß
0.6'
1 2
i i' i i
,
1500 150 15
porerad:us(!ran)
ß5 0 0.2 0.4 0,6
water content
10ø
10'2
10'4
Figure10. Conductivityestimationfor AggregatedLoam
[SmettemandKirkby,1990]fromfourmathematicallydif-
ferentretentionfunctions.(Left)Fittedretentionmodels:
linearinterpolated(solidline),bimodalretentionfunction
(equation(4))withrni allowedtovary(dottedline),mi = 1
-' 1/ni(dashedline),andmi = 1(dashed-dottedline).See
Table1 forcoefficientsof themodels.(Right)Estimated
conductivityfunctionsandmeasuredconductivitydata.The
fourretentionmodelsyieldalmostidenticalestimates.The
agreementbetween estimates and measurements is excel-
lent.
grandin (7), 1/½dS•, is bounded,sincedS[, = 0 between
saturationandtheair entrypressure.Thisisdifferentfor any
retentionfunctionwhich approachessaturationasymptoti-
cally. Now the valueof the integranddependson how fast
dS•, approacheszero as 1/0 approachesinfinity, i.e., on the
slopeof theretentionfunction.The absolutesensitivityof K
to Se, definedby Sabs= dK/dSe, becomesinfinitenear
saturation.For the closed-formconductivityfunction of the
van Genuchten-Mualem model,
g = - - s/m)m]2 (9)
this absolutesensitivityis givenby
2Se
'
Sabs-'rSe ~q- l/m)1-m' (10)(1 --Se
whereg(Se) = [1 - (1 - S•I/")m].Whentheeffective
saturationapproachesunity, the first right-handterm andthe
numeratorof the secondterm in (10) approachfinite values,
whereas the denominator of the second term becomes zero,
becausethe value of the parameter m is boundedby 0 < m
< 1. This causes[$abs[to approachinfinity.Due to this
sensitivity characteristic,even small changesin water con-
tent near saturation(indicatingthe existence of large pores)
are very importantfor the shapeof the estimatedconductiv-
ity function. Measurementsclose to Osmust therefore be
taken at small pressureintervals and as accurately as possi-
ble.
Reliable conductivity estimations become impossible if
measurements near saturation are missing, regardless
whether the pore systemis heterogeneousor not. To illus-
trate this, we have fitted three different unimodal functions
to hypotheticalretentiondata(Figure 11 (left)). The reten-
tion models are the Brooks and Corey function, the van
Genuchten function with m = 1 - l/n, and the van
Genuchten function with n = 1.0. Since there are no
measureddata between saturationand Se = 0.8 at pF2,
eachfunctiondescribesthe dataequally well. However, the
correspondingpore-sizedistributions,whichareidenticalin
the fine-porerange, are very differentin their asymptotic
behaviorin the large-porerange.As was discussedabove,
the Brooks and Corey function(dashedline) restrictsthe
220 DURNER:HYDRAULICCONDUCTIVITYESTIMATIONFORSOILS
sizeof thelargestporesconsideredby theconductivity
predictionmodelto theequivalentradiusoftheairentry
pressure(r --- 0.15 mm in the givenexample).The two
smoothfunctionsrepresentacertainamountofporesinthe
large-porerangeandcanevenconsiderporesizesthatare
physicallyabsurd.Theconductivityestimatesbasedonthe
threefunctionsareshiftedagainsteachotherby ordersof
magnitude, and without additional measurementsbetween
saturationandpF2 thereis noway to decidewhichmodel
represents the soil best.
The accuracyof watercontentmeasurementsisgenerally
limited by the error of the measurementmethod. Unfortu-
nately,it becomesexperimentallymoreandmoredifficultto
measuretheretentioncharacteristicof a soilaccuratelyand
reproduciblyat very low suctions.It isfurtherquestionable
whetherretentioncurves,indeed,representthe pore-size
distributionof a porousmediumat high saturation,where
the nonwettingphaseis notcontinuous[Corey,1992].This
meansthat the conductivityestimationbecomesby defini-
tion inaccuratewhen approachingsaturation,even if we
assumethat the predictive model itself is free of error and its
applicability does not depend on the pore sizes. As a
consequence, conductivity estimationsfrom retention func-
tionsthat representa certainporespaceintherangeof very
smallsuctions(say, belowpFO) are unreliable,regardlessof
whether unimodal or multimodal functions are used. This is,
for example, generally the case if the parametern in the
constrainedvan Genuchten equation(m = 1 - l/n) is close
to unity.
ComparisonsBetweenMeasured and
Estimated Conductivities
To assess whether conductivity estimations from more
accurate retention functions are of advantage in practice,
comparisonsof unimodal and multimodal predictions with
experimental data are necessary.To be of statisticalsignif-
icance, suchcomparisonsmustrest on a broaddatabaseof
well-documented measurements of high quality. Existing
data have frequently been measuredon homogenizedand
packedsoilsamples,anda varietyof measurementmethods
have been used. To date, most of the publishedempirical
validation studies on conductivity estimations used rela-
tively small databasesof the authors' own measurements.
This mightbe onereasonwhy manyof thesestudiesresulted
in a new proposed"optimal" value of the tortuosityor
pore-interactionfactorr [e.g.,Kunzeetal., 1968;Greenand
Corey, 1971;Jackson, 1972; Ghosh, 1977;Ragab et al.,
1981;Talsma, 1985;Alexanderand Skaggs, 1986].
In caseswhere measureddata appear basically reliable,
often the retention measurements near saturation are insuf-
ficient. Then, the hydrologicallyimportantsecondarypore
systemis not strictlydetermined,aswasdiscussedabove.
Therefore estimationsby a flexible retention function are
uncertain, and a better or worse agreementwith measure-
mentsmightbe achievedat random.Durner[1992]showed
for Beit Netofa clay (whichservedvan Genuchten[1980]as
an examplefor possiblefailuresof the van Genuchten-
Mualem model) that a conductivityestimationbased on a
bimodal retentionfunction led to a very goodagreementwith
measureddata. However, since for this soil no retention
meaœurementsare available in the pressurerangebetween
saturationandpF2, the saturatedwatercontentwasarbi-
trarilyset2%largerthanthelargestmeasuredwatercontent.
A differentvaluefor 0• mighthavechangedtheestimation
considerably,leadingtoa badagreement.Note,however,
thatit wasnotpossibleto cometo a satisfactoryagreement
betweenmeasuredandestimatedconductivitiesusingthe
unimodal van Genuchten-Mualem model.
Sinceparticularlyforheterogeneoussoilswithsecondary
poresystemsthe problemof measurementsnearsaturation
is furthercomplicatedby macropores,air encapsulations,
and hysteresiseffects,we do not attempt to compareinthis
paper unimodaland multimodalconductivitypredictions
with measurements.It appearsdoubtfulwhethera validation
attemptin this classicalway is viable with presentlyavail-
able data.
Conclusions
The existenceof a secondarypore systemgreatly affects
the shapeof the estimatedhydraulic conductivitycurvein
the wet range. Estimation differencesdue to differentreten-
tion modelscan be larger than differencesdue to the useof
alternative prediction models [Childs and Coilis-George,
1950; Burdine, 1953; Mualem, 1976]. Consequently,the
ignoranceof a secondarypore system may lead to large
errors. We have shown for soils with heterogeneouspore
systemsthat the commonlyappliedunimodalretentionfunc-
tions are inadequate for conductivity estimations, because
they lack the necessaryflexibility in the rangeof largepores.
Their use may lead to wrong results in flow and transport
simulationsor in parameterestimationprocedureswherethe
inverseproblemis solved[Durner, 1991].Validation studies
of hydraulic conductivity prediction methods shouldbe
reevaluatedif the appliedretention modeldid not accurately
describe the measured data. Particularly for loamy and
clayey soilsthe reasonsfor the frequent failure of conduc-
tivity prediction methods should be reconsidered.In some
casesit may not be the failure of the predictive model,but
the use of inadequate retention functions which leadstoa
discrepancybetweenestimatesandmeasurements.Forsoils
with very wide pore-sizedistributions,unimodalretention
modelsare sometimesforcedto representa nonnegligible
part of the pore spacein the range of unrealisticallylarge
poresizes.In suchcases(indicated,e.g., by thevanGenu-
chtenparametern beingcloseto unity) the conductivity
predictionnear saturationis by definitionunreliable.
The almostperfectagreementbetweenmeasuredand
estimatedconductivitiesfor theloamysoilof Smettemand
Kirkby[1990,Figure10b]indicatesthatconductivitypredic-
tionmethodsbasingonstatisticalpore-domainmodelshave
a high successpotential not only for soils with narrow
pore-sizedistributions,but as well for soilswith heteroge-
neousporesystems.As comparedto unimodalretention
functions,theuseofthemultimodalretentionfunction(orof
anyotherflexibleretentionfunction)leadsto morereliable
conductivityestimatesifit improvesthefitofthemeasured
retentiondataandif theshapeofthefittedretentionfunction
in the rangeoflargeporesis strictlydeterminedby measure-
ments.Particularlyfor validationsof flow and transport
modelsonlaboratoryscale,thehydrauliccharacteristicof
thesoilsampleshouldbedescribedasaccuratelyaspossi-
ble.
Theslopeofthewaterretentioncurvenearsaturationis
themostimportantpropertyindeterminingtheshapeofthe
estimatedconductivityfunction.Evensmalluncertaintiesin
DURNER: HYDRAULIC CONDUCTIVITY ESTIMATION FORSOILS 221
measuredwatercontentsnear saturationlead to largeesti-
mationuncertainties.Whena retentioncharacteristichasno
distinctairentrypoint,the measuredsaturatedconductivity
isgenerallynot an appropriatematchingvaluefor the
predictedconductivityfunction.Thisconclusion,whichwas
alreadystatedfor unimodalporesystems[e.g., van Genu-
chtenandNielsen, 1985;Lucknet et al., 1989;Nielsenand
Lucknet,1992],is even more importantfor heterogeneous
poresystems.Theuseofmultimodalfunctionscanhelpto
explainthisproblem;however,it cannotsolveit.
Wemayexpectthat the basicconceptsthat underliethe
statisticalconductivitypredictionmodelsgenerallyholdless
forsecondarypore systems as compared to textural ones.
Theshapeandorganizationof voidsin aggregatedsoilscan
beverydifferent,dependingon soilgenesisand aggregation
state[WangandNarasimhan, 1990].As an example,a net of
planarvoidswill behavehydrologicallyquite differently
fromthe pore spacebetween sphericalaggregates,or the
porespaceduetomicrorootchannels.Nevertheless,eachof
theseporesystemsmay have a similarly shapedretention
characteristic.Therefore we shouldregard •-in the correc-
tionfunction SJ as a purely empirical fitting parameter
[Roulieret al., 1972; Russo, 1988]. For use in parameter
estimationprocedures, Nielsen and Lucknet [1992] even
suggestto decouple the retention and the conductivity
function,and to use the closed-form conductivity equation
of the van Genuchten-Mualem model (9) as an empirical
expressionfor K(Se). For soils with heterogeneouspore
systems,we expectthe bestresultsby stillusingthe shapeof
theestimatedconductivity function while treating the two
parametersKs and ß as unknowns. By this strategy the
informationon the pore system containedin the retention
functionwill be usedto fullest advantage.Close to saturation
theconductivitiesof soilscontaininglarge pores shouldbe
directlymeasured.This means, however, that the attractive-
nessofthe predictionmethodsis considerablyreduced.
Propertiesof heterogeneouspore systemscannotbe iden-
tifiedbyparameterestimationtechniqueswheretheinverse
problemis solved,if a priori unimodalhydraulicproperties
areassumed.Due to the ill-posednessof the mathematical
problem,it appearsontheotherhandimpossibleto usemore
flexiblehydraulicfunctionsand to identify the increased
numberof coefficientsby the inversemethod.Thereforethe
retentioncurvesof soilswith heterogeneouspore systems
shouldbedirectlyfittedto measureddata.Preferably,these
measurementsare obtainedfrom a transient flow experiment
bysimultaneouslymeasuring½and0inhighresolution[e.g.,
Sobczuketal., 1992].Theflowcharacteristicsof theexper-
imentcanthenbeusedto optimizethetwoconductivity-
relatedunknowns,Ksandr, bysolvingtheinverseproblem.
Finally,wewanttorecallthatthereliabilityofconductiv-
itypredictionmethodsisgenerallyrestrictedbysomefun-
damentalproblems,e.g.,theassumptionof isotropicorga-
nizationofporesindependentoftheirsize,thevalidityofthe
Hagen-Poiseuillelawforflowinlargepores,theproblemof
airentrapment,andtheproblemofhysteresis.Thespecific
findingsinthispaperforheterogeneousporesystemsare
neverthelessapplicable.Theresultsarefurtherindependent
ofthechoiceoftheparticularstatisticalconductivitypredic-
tionmodel.Theextensionto thefieldscalerequiresthe
considerationofspatialandtemporalvariabilityaswellas
theconsiderationofscaleproblemsanderrorsinducedby
themeasurementmethod.Thefurtherdevelopmentofsuit-
ablemeasurementmethodsto identifyheterogeneouspore
systemsappearsthereforeto be of primary importance.
Acknowledgments.The authorthanksDuaneHampton,Kalama-
zoo,Michigan,OleJacobsen,Tinglev,Denmark,andRichardDe
JongandG. ClarkeTopp,Ottawa,Ontario,Canada,whoprovided
thesoilphysicaldataofsomeoftheexamplesoilsinthispaper.The
authorgratefullyacknowledgesthegrantof a long-termfellowship
oftheEuropeanEnvironmentalResearchOrganization(EERO).
References
Addiscott,T. M., andR. J.Wagenet,Conceptsof soluteleachingin
soils:A reviewofmodellingapproaches,J. SoilSci.,36,411-424,
1985.
Alexander,L. R., andR. W. Skaggs,Predictingunsaturatedhydrau-
lic conductivityfrom the soil water characteristic, D'ans. Am.
Soc. Agric. Eng., 29, 176-184, 1986.
Anderson,S. H., andD. K. Cassel,Statisticaland autoregressive
analysisof soil physicalpropertiesof Portsmouthsandyloam,
Soil Sci. Soc. Am. J., 50, 1096-1104, 1986.
Barraclough,D., A usablemechanisticmodelof nitrate leaching,I,
The model, J. Soil Sci., 40, 543-554, 1989.
Bear,J., andY. Bachmat,Introductionto TransportPhenomenain
PorousMedia, Kluwer Academic,Hingham, Mass., 1991.
Bloemen, G. W., Calculation of hydraulic conductivitiesof soils
from texture and organic matter content, Z. Pfianzenerniihr.
Bodenkd., !43,581--605, 1980.
Boels, D., J. B. H. M. Van Gils, G. J. Veerrnan, and K. E. Wit,
Theory and system of automatic determination of soil moisture
characteristics and unsaturated hydraulic conductivities, Soil
Sci., 126, 191-199, 1978.
Brooks, R. H., and A. T. Corey, Hydraulic propertiesof porous
media,Hydrol. Pap., 3, pp. 1-27, Colo. State Univ., Fort Collins,
1964.
Bruce, R. R., and R. J. Luxmoore, Water retention: Field methods,
in Methods of Soil Analysis, part 1, Physical and Mineralogical
Methods, 2nd ed., edited by A. Klute, pp. 799-824, ASA and
SSSA, Madison, Wisc., t986.
Brutsaert, W., Probabilitylawsfor pore size distributions,Soil Sci.,
IOI, 85-92, 1966.
Brutsaert, W., The permeabilityof a porous mediumdetermined
from certain probability laws for pore-size distribution, Water
Resour. Res., 4, 425-434, 1968.
Burdine, N. T., Relative permeability calculationsfrom pore size
distribution data, Trans. Am. Inst. Min. Metall. Pet. Eng., 198,
71-78, I953.
Campbell, G. S., A simple methodfor determiningunsaturated
conductivityfrom moistureretentiondata, Soil Sci., I 17, 311-314,
1974.
Carsel,R. F., L. A. Mulkey, M. N. Lorber, and L. B. Baskin,The
pesticideroot zonemodel:A procedurefor evaluatingpesticide
leachingthreatsto groundwater,Ecol. Modell., 30, 49-69, 1985.
Chen,C., andR. J. Wagenet,Simulationof water andchemicalsin
macroporesoils,I, Representationof the equivalentmacropore
influence and its effect on soil-water flow, J. Hydrol., 130,
105-126, 1991.
Chen, C., D. M. Thomas, R. E. Green, and R. J. Wagenet,
Two-domain estimation of hydraulic properties in macropore
soils, Soil Sci. Soc. Am. J., 57, 680-686, 1993.
Childs, E. C., and N. Coilis-George,The permeabilityof porous
materials,Proc. R. Soc. London, Ser. A, 201,392-405, 1950.
Corey,A. T., Pore-sizedistribution,in Proceedingsof theInterna-
tional Workshop,IndirectMethodsfor EstimatingtheHydraulic
Propertiesof UnsaturatedSoils,editedbyM. T. vanGenuchten,
F. J. Leij, andL. J. Lund, pp. 37-44, Universityof California,
Riverside, 1992.
Currie,J. A., andD. A. Rose,Gasdiffusionin structuredmaterials:
The effect of tri-modal pore-size distribution, J. Soil Sci., 36,
487-493, 1985.
DeJong,R., G. C. Topp,andW. D. Reynolds,Theuseofmeasured
andestimatedhydraulicpropertiesin the simulationof soilwater
contentprofiles:A casestudy,inProceedingsof theInternational
Workshop,IndirectMethodsfor EstimatingtheHydraulicProp-
ertiesof UnsaturatedSoils, pp. 569-584, editedby M. T. van
222 DURNER:HYDRAULICCONDUCTIVITYESTIMATIONFORSOILS
Genuchten,F.J.Leij,andL.J.Lund,UniversityofCalifornia,Riverside, 1992.
D'Hollander,E.H., Estimationoftheporesizedistributionfrom
themoisturecharacteristic,WaterResour.Res.15 107-112,I979. ' '
Dixon,R. M., Controllinginfiltrationinbimodalporoussolids:
Air-earthinterfaceconcepts,FundamentalsofTransportPhe-
nomenainPorousMedia,vol.2, pp. 107-117,Universityof
Guelph,Ontario,1972.
Durner,W.,VorhersagederhydraulischenLeiffiihigkeitstrukturi-
erterB6den,BayreutherBodenkundlicheBerichte,20, 1-180,1991.
Durner,W.,Predictingtheunsaturatedhydraulicconductivityusing
multi-porositywaterretentioncurves,inProceedingsof the
InternationalWorkshop,IndirectMethodsforEstimatingthe
HydraulicPropertiesof UnsaturatedSoils,editedbyM. T. van
Genuchten,F.J.Leij,andL.J.Lund,pp.185-202,Universityof
California,Riverside,1992.
Farrel,D. A., andW. E. Larson,Modelingtheporestructureof
porousmedia,WaterResour.Res.,8, 699-706,1972.
Gardner,W.R.,Thepermeabilityproblem,Soil$ci.,117,243-249,
1974.
Gerke,H.H.,andM.T. vanGenuchten,Adual-porositymodelfor
simulatingthe preferentialmovementof waterandsolutesin
structuredporousmedia,WaterResour.Res.,29,305-319,1993.
Ghosh,R.K., Determinationofhydraulicconductivityfrommois-
tureretentiondata,SoilSci., 124,122-124,1977.
Gillham,R. W., Hydraulicpropertiesofa porousmedium:Mea-
surementandempiricalrepresentation,SoilSci.Soc.Am.J.,40,
203-207, 1976.
Green,R.E., andJ. C. Corey,Calculationofhydraulicconductiv-
ity: A furtherevaluationof somepredictivemethods,SoilSoc.
Am. Proc., 35, 3-8, 1971.
Hampton, D. R., Coupled heat and fluid flow in saturated-
unsaturatedcompressibleporousmedia,Ph.D. thesis,Colo.State
Univ., Fort Collins, 1989.
Hampton,D. R., A methodtofit soilhydrauliccurvesin modelsof
flow in unsaturatedsoils,in ComputationalMethodsin Subsur-
face Hydrology,editedby G. Gambolatiet al., Computational
MechanicsPublishers,Boston,Mass., 1990.
Hantschel,R., W. Durner,and R. Horn, Die Bedeutungyon
Porenheterogenit•itenffr dieErstellungderpF-WG-undK-0-
Kurven, Desch.BodenkdI.Mitt., 53, 403-409,1987.
Haverkamp,R., M. Vauclin,J. Touma,P. J. Wierenga,andG.
Vachaud, A comparison of numerical simulationmodels for
one-dimensionalinfiltration,SoilSci. Soc.Am. J., 41,285-294,
1977.
Hornung,U., Identificationof nonlinearsoilphysicalparameters
from an input-outputexperiment,in Workshopon Numerical
Treatmentsof Inverse Problemsin Differentialand Integral
Equations,editedby P. Deyuflhardand E. Haier, pp. 227-237,
Birkhauser, Boston, Mass., 1983.
Jackson,R. D., On the calculationof hydraulicconductivity,Soil
$ci. Soc. Am. Proc., 36, 380-382, 1972.
Jacobsen,O. H., Unsaturatedhydraulic conductivityfor some
Danishsoils,Methodsandcharacterizationof soils(in Danish),
Tidsskr.Planteavl, 93(S2030), 1-60, 1989.
Jarvis,N.J., P.-E. Jansson,P. E. Dik, andI. Messing,Modelling
water andsolutetransportinmacroporoussoil,I, Modeldescrip-
tion and sensitivity analysis,J. Soil Sci., 42, 59-70, 1991.
Jeppson,R. W., W. J. RawIs, W. R. Hamon, andD. L. Schreiber,
Use of axisymmetricinfiltration modeland fielddatato determine
hydraulic propertiesof soils, Water Resour. Res., I1, 127-138,
1975.
Kablan, R. A. T., R. S. Mansell, S. A. Bloom,andL. C. Hammond,
Determinationsof unsaturatedhydraulicconductivityfor Candler
Sand, Soil Sci., 148, 155-164, 1989.
King, L. G., Descriptionof soil characteristicsfor partially satu-
rated flow, Soil Sci. $oc. Proc., 359-362, 1965.
Kool, J. B., and J. C. Parker, Development and evaluation of
closed-formexpressionsfor hysteretic soil hydraulicproperties,
Water Resour. Res., 23, 105-114, 1987a.
Kool, J. B., and J. C. Parker, Estimating soil hydraulicproperties
from transient flow experiments: SFIT User's Guide, Electric
Power Research Institute, Palo Alto, Calif., 1987b.
Kool, 5•. B., and J. C. Parker, Analysis of the inverse problem for
transientunsaturatedflow,WaterResour.Res.,24,817--830,1988.
Kool,J.B.,J.C.Parker,andM. T. vanGenuchten,Parameter
estimationforunsaturatedflowandtransportmodels:Areview,J. Hyclrol., 91,255-294, 1987.
Kunze,R.J.,G.Uehara,andK.Graham,Factorsimportantinthe
calculationofhydraulicconductivity,SoilSci.Soc.Am.Proc.,32,760-765, 1968.
Lafolie,F.,R.Guennelon,andM. T. vanGenuchten,Analysisof
waterflowundertrickleirrigation,II, Experimentalevaluation,
Soil Sci. Soc. Am. J., 53, 1318-1323, 1989.
Laliberte,G.E.,Amathematicalfunctionfordescribingcapill•
pressure--desaturationdata,Bull.Int. Ass.Sci.HydroI.,I4,
131-149, 1969.
Loague,K., andR.E. Green,Statisticalandgraphicalmethodsfor
evaluatingsolutetransportmodels,Overviewandapplications,J.
Contam. Hydrol., 7, 51-74, 1991.
Luckner,L., M.T. vanGenuchten,andD.R.Nielsen,Aconsistent
setof parametricmodelsfor thetwo-phaseflowof immiscible
fluidsinthesubsurface,WaterResour.Res.,25,2187-2193,1989.
Marshall,T. J., A relationbetweenpermeabilityandsizedistribu.
tion of pores,J. Soil Sci., 9, I-8, 1958.
Michiels,P.,R. Hartmann,andE. DeStrooper,Comparisonsoflhe
unsaturatedhydraulicconductivityof a coarse-texturedsoilas
determinedin thefield,in thelaboratory,andwithmathematical
models,Soil Sci., 147, 299-304, 1989.
Millington,R. J., andJ.P. Quirk,Permeabilityof poroussolids,
Trans. Faraday Soc., 57, 1200-1207, 1961.
Mualem,Y., Anewmodelforpredictingthehydraulicconductivity
of unsaturatedporousmedia,WaterResour.Res.,12,5!3-522,
1976.
Mualem,Y., Hydraulicconductivityof unsaturatedporousmedia:
Generalizedmacroscopicapproach,Water Resour.Res.,14,
325-334, 1978.
Mualem,Y., Hydraulicconductivityof unsaturatedsoils:Prediction
andformulas,in Methodsof SoilAnalysis,part 1, Physicaland
MineralogicalMethods,2nded.,editedbyA. Klute,pp.799-824,
ASA and SSSA, Madison, Wisc., 1986.
Mualem,Y., Modelingthe hydraulicconductivityof unsaturated
porousmedia,in Proceedingsof the InternationalWorkshop,
IndirectMethodsfor Estimatingthe HydraulicPropertiesof
UnsaturatedSoils,editedbyM. T. vanGenuchten,F. J.Leij,and
L. J. Lund,pp. 15-36, Universityof California,Riverside,1992.
Mualem,Y., andG. Dagan,Hydraulicconductivityof soils:unified
approachto the statistical models, Soil Sci. Soc. Am. J., 42,
392-395, 1978.
Nielsen,D. R., and L. Luckner,Theoreticalaspectsto estimate
reasonableinitial parametersand rangelimits in identification
proceduresfor soil hydraulicproperties,in Proceedingsofthe
International Workshop,Indirect Methodsfor Estimatingthe
HydraulicPropertiesof UnsaturatedSoils,editedby M. T. van
Genuchten,F. J. Leij, andL. J. Lund,pp. 147-160,Universityof
California,Riverside, 1992.
Nielsen, D. R., D. Kirkham, and E. R. Perrier, Soil capillary
conductivity:Comparisonofmeasuredandcalculatedvalues,Soil
Sci. Soc. Am. Proc., 24, 157-160, 1960.
Nimmo, J. R., Comment on the treatment of residual water content
in "A consistentsetof parametricmodelsfor thetwo-phasett0•
of immisciblefluidsin the subsurface"by L. Luckheretal.,
WaterResour.Res., 27, 661-662, 1991.
Othmer,H., B. Diekk•ger, andM. Kutilek,Bimodalporosityand
unsaturatedhydraulicconductivity,SoilSci., 152,139-149,1991.
Parker,J. C., andM. T. vanGenuchten,Determiningsoilhydraulic
propertiesfromone-stepoutflowexperimentsbyparameteresti-
mation,II, Experimentalstudies,Soil Sci. Soc. Am. J., 49,
1348-1359, 1985.
Parkes,M. E., andP. A. Waters,Comparisonof measuredand
estimatedunsaturatedhydraulicconductivity,Water Resour,
Res., 16, 749-754, 1980.
Peters,R. R., and E. A. Klavetter, A continuummodelforwater
movementin an unsaturatedfractured rock mass, WaterResour.
Res., 24, 416-430, 1988.
Pruess,K., J.S.Y. Wang,andY. W.Tsang,Onthermohydrologi½
conditionsnearhighlevelnuclearwastesemplacedinpartially
saturatedfracturedtuff,2, Effectivecontinuumapproximation,
WaterResour.Res., 26, 1249-!261, 1990.
DURNER:HYDRAULICCONDUCTIVITYESTIMATIONFORSOILS 223
Puckett,W. E., J.H. Dane,B. F. Hajek,Physicalandmineralogical
datatodeterminesoilhydraulicproperties,SoilSci.Soc.Am. J.,
49,831-836,1985.
Purcell,W. R., Capillarypressures--theirmeasurementusingmer-
curyandthecalculationof permeabilitytherefrom,Trans.Am.
Inst.Min. Metall. Pet. Eng., 186, 39-48, 1949.
Ragab,R.,J.Feven,andD. Hillel,Comparativestudyofnumerical
andlaboratorymethodsfor determiningthe hydraulicconductiv-
ityfunctionof a sand,SoilSci.,131,375-388,1981.
Rogowski,A. S., Estimationofthesoilmoisturecharacteristicand
hydraulicconductivity:Comparisonof models,Soil Sci., 114,
423-429, 1972.
Ross,P. J., and K. J. R. Smettem, Describingsoil hydraulic
propertieswithsumsof simplefunctions,SoilSci.Soc.Am. J.,
57,26-29, 1993.
Roulier,M. H., L. H. Stolzy,J. Letey, andL. V. Weeks,Approx-
imationof fieldhydraulicconductivityby laboratoryprocedures
onintactcores,Soil Sci. Soc. Am. Proc., 36, 387-393, 1972.
Russo,D., Determiningsoil hydraulicpropertiesby parameter
estimation:On the selectionof a modelfor the hydraulicproper-
ties,WaterResour. Res., 24, 453-459, 1988.
Sakellariou-Makrantonaki,M., C. TzimopouIos,andD. Gouliaras,
Analysisof a closed-formanalyticalmodelto predictthehydrau-
lic conductivityfunction, J. Hydrol., 92, 289-300, 1987.
Saxton,K. E., W. J. RawIs,J. S. Romberger,andR. I. Papendick,
Estimatinggeneralizedsoil-water characteristicsfrom texture,
SoilSci. Soc. Am. J., 50, 1031-1036, 1986.
SchjCnning,P., Por½sitetsforholdi landbrugsjord,I, Modeller og
jordtypeforskelle,(in Danish), Tidsskrift.Plantearl., 89, 411-423,
1985.
Schofield,R. K., ThepF of thewaterin soil, Trans.3rd Int. Congr.
SoilSci., 2, 38-48, 1935.
Schuh,W, M., and R. L. Cline, Effect of soil properties on
unsaturatedhydraulic conductivity pore-interaction factors, Soil
Sci.Soc. Am. J., 54, 1509-1518, 1990.
Sharma,M. L., and G. Uehara, Influence of soil structure on water
relationsin low humic Iatosols, I, Water retention, Soil Sci. Soc.
Am. Proc., 32,765-771, 1968.
Sidiropoulos,E., and S. Yannopoulos, Sensitivity analysis of
closed-formanalyticalhydraulicconductivitymodels,J. Hydrol.,
IOI, 159-172, 1988.
Simmons,C. S., D. R. Nielsen, and J. W. Biggat, Scalingof
field-measuredsoil-waterproperties,I, Methodology,ttilgardia,
50, 1-25, 1979.
Smettem,K. R. J., andC. Kirkby, Measuringthe hydraulicprop-
ertiesof a stableaggregatedsoil,J. Hydrol., 117, 1-13, 1990.
Sobczuk,H. A., R. Plagge,R. T. Walczak,andC. Roth,Laboratory
equipmentandcalculationprocedureto rapidlydeterminehyster-
esisof somehydrophysicalpropertiesunder nonsteadyflow
conditions,Z. Pflanzenern•ihr.Bodenkd., 155, 157-163, 1992.
Stephens,D. B., A comparisonof calculatedandmeasuredunsat-
uratedhydraulicconductivityof two uniform soilsin New Mex-
ico, in Proceedingsof the International Workshop,Indirect
Methodxfor EstimatingtheHydraulicPropertiesof Unsaturated
Soils,editedbyM. T. vanGenuchten,F. J. Leij, andL. J. Lund,
pp.249-262,Universityof California,Riverside,1992.
Stephens,D. B., andK. R. Rehfeldt,Evaluationof closed-form
analyticalmodelstocalculateconductivityinafinesand,SoilSci.
Soc.Am.•., 49, 12-19,1985.
Su,C.,andR.H. Brooks,Soilhydraulicpropertiesfrominfiltration
tests,paperpresentedat WatershedManagementProceedings,
Irrig.andDrain.Div.,Am.Soc.Civ.Eng.,Logan,Utah,Aug.
11-13, 1975.
Talsma,T., Predictionof hydraulicconductivityfromsoilwater
retentiondata,SoilSci., 140, 184-188,1985.
Tsuji,G. Y., R. T. Watanabe,andW. S. Sakai,Influenceof soil
microstructure on water characteristics of selected Hawaiian
soils,SoilSci. Soc.Am. Proc.,39, 28-33, 1975.
vanGenuchten,M. T., A closed-formequationfor predictingthe
hydraulicconductivityofunsaturatedsoils,SoilSci.Soc.Am. J.,
44, 892-898, 1980.
vanGenuchten,M. T., andF. J. Leij, Onestimatingthehydraulic
propertiesof unsaturatedsoils,in Proceedingsof theInterna-
tionalWorkshop,IndirectMethodsfor EstimatingtheHydraulic
Propertiesof UnsaturatedSoils,editedbyM. T. vanGenuchten,
F. J. Leij, andL. J. Lund, pp. 1-14, Universityof California,
Riverside, 1992.
van Genuchten,M. T., and D. R. Nielsen, On describingand
predictingthe hydraulicpropertiesof unsaturatedsoils, Ann.
Geophys.,3, 615-628, 1985.
Vereecken,H., J.Maes,J. Feyen,andP. Darius,Estimatingthesoil
moistureretentioncurvefromtexture,bulk density,andcarbon
content, Soil Sci., 148, 389-403, 1989.
Visser,W. C., An empiricalexpressionfor thedesorptioncurve,in
Waterin the UnsaturatedZone, Proceedingsof the Wageningen
Symposium,vol. 1, editedby P. E. RijtemaandH. Wassink,pp.
329-335, IASH/AIHS, Unesco, Paris, 1968.
Vogel, T., and M. Cislerova, On the reliability of unsaturated
hydraulic conductivity calculated from the moisture retention
curve, Transp. PorousMedia, 3, 1-15, 1988.
Wang, J. S. Y., and T. N. Narasimhan,Hydrologic mechanisms
governingfluid flow in a partially saturated fractured porous
medium, Water Resour. Res., 21, 1861-1874, 1985.
Wang, J. S. Y., and T. N. Narasimhan, Fluid flow in partially
saturated, welded-nonmeldedtuff units, Geoderma, 46, 155-168,
1990.
W6sten, J. H. M., and M. T. van Genuchten, Using texture and
other soil properties to predict the unsaturated soil hydraulic
functions, Soil Sci. Soc. Am. J., 52, I762-1770, 1988.
Wyllie, M. R. J., and G. H. F. Gardner, The generalizedKozeny-
Carman equation, World Oil, 146, 210-228, 1958.
Yuster, S. T., Theoretical considerationsof multiphaseflow in
idealizedcapillarysystems,WorldPetr. Congr.Proc., 3rd, 2,
437-445, 1951.
Zachmann, D. W., P. C. Duchateau, and A. Klute, The calibration
of the Richardsflow equationfor a draining column by parameter
identification, Soil Sci. Soc. Am. J., 45, 1012-1015, 1981.
Zachmann, D. W., P. C. Duchateau, and A. Klute, Simultaneous
approximationof water capacityand soil hydraulicconductivity
by parameteridentification,Soil Sci., 134, 157-163, 1982.
W. Durner, Institute of Terrestrial Ecology, Soil Physics,Federal
Institute of Technology,Grabenstr. 3, Zfirich, CH-8952 Schlieren,
Switzerland.
(ReceivedDecember8, 1992;revisedSeptember3, 1993;
acceptedSeptember16, 1993.)

More Related Content

What's hot

Oil and gas reserves sensitivity to log evaluation
Oil and gas reserves sensitivity to log evaluationOil and gas reserves sensitivity to log evaluation
Oil and gas reserves sensitivity to log evaluationPeter Cockcroft
 
Report - Forschungszentrum Jülich
Report - Forschungszentrum JülichReport - Forschungszentrum Jülich
Report - Forschungszentrum JülichArjun Narayanan
 
Ess 16-3959-2012
Ess 16-3959-2012Ess 16-3959-2012
Ess 16-3959-2012Simone88888
 
Time-lapse analysis with earth resistance and electrical resistivity imaging
Time-lapse analysis with earth resistance and electrical resistivity imagingTime-lapse analysis with earth resistance and electrical resistivity imaging
Time-lapse analysis with earth resistance and electrical resistivity imagingDART Project
 
Birr - Identifying Critical Portions of the Landscape
Birr - Identifying Critical Portions of the LandscapeBirr - Identifying Critical Portions of the Landscape
Birr - Identifying Critical Portions of the LandscapeJose A. Hernandez
 
Limitations of vs30 etc.
Limitations of vs30 etc. Limitations of vs30 etc.
Limitations of vs30 etc. Robert Pyke
 
Limitations of vs30 etc.
Limitations of vs30 etc. Limitations of vs30 etc.
Limitations of vs30 etc. Robert Pyke
 
Ellsworthetal1996SSSAJpaper
Ellsworthetal1996SSSAJpaperEllsworthetal1996SSSAJpaper
Ellsworthetal1996SSSAJpaperellswort
 
Improved analyses of liquefaction and settlement
Improved analyses of liquefaction and settlementImproved analyses of liquefaction and settlement
Improved analyses of liquefaction and settlementRobert Pyke
 
Limitations of simplified methods for estimating seismic settlements linked i...
Limitations of simplified methods for estimating seismic settlements linked i...Limitations of simplified methods for estimating seismic settlements linked i...
Limitations of simplified methods for estimating seismic settlements linked i...Robert Pyke
 
Hydrological Modelling of Shallow Landslides
Hydrological Modelling of Shallow LandslidesHydrological Modelling of Shallow Landslides
Hydrological Modelling of Shallow LandslidesGrigoris Anagnostopoulos
 
IGARSS2011_DInSAR_MORRISON.ppt
IGARSS2011_DInSAR_MORRISON.pptIGARSS2011_DInSAR_MORRISON.ppt
IGARSS2011_DInSAR_MORRISON.pptgrssieee
 

What's hot (20)

Oil and gas reserves sensitivity to log evaluation
Oil and gas reserves sensitivity to log evaluationOil and gas reserves sensitivity to log evaluation
Oil and gas reserves sensitivity to log evaluation
 
Report - Forschungszentrum Jülich
Report - Forschungszentrum JülichReport - Forschungszentrum Jülich
Report - Forschungszentrum Jülich
 
WCEE2012_1941
WCEE2012_1941WCEE2012_1941
WCEE2012_1941
 
Ess 16-3959-2012
Ess 16-3959-2012Ess 16-3959-2012
Ess 16-3959-2012
 
Alfonso Senatore
Alfonso SenatoreAlfonso Senatore
Alfonso Senatore
 
Time-lapse analysis with earth resistance and electrical resistivity imaging
Time-lapse analysis with earth resistance and electrical resistivity imagingTime-lapse analysis with earth resistance and electrical resistivity imaging
Time-lapse analysis with earth resistance and electrical resistivity imaging
 
Birr - Identifying Critical Portions of the Landscape
Birr - Identifying Critical Portions of the LandscapeBirr - Identifying Critical Portions of the Landscape
Birr - Identifying Critical Portions of the Landscape
 
Limitations of vs30 etc.
Limitations of vs30 etc. Limitations of vs30 etc.
Limitations of vs30 etc.
 
Limitations of vs30 etc.
Limitations of vs30 etc. Limitations of vs30 etc.
Limitations of vs30 etc.
 
Poster AGU 2013
Poster AGU 2013Poster AGU 2013
Poster AGU 2013
 
Ellsworthetal1996SSSAJpaper
Ellsworthetal1996SSSAJpaperEllsworthetal1996SSSAJpaper
Ellsworthetal1996SSSAJpaper
 
Spg nitin
Spg nitinSpg nitin
Spg nitin
 
Advances in Rock Physics Modelling and Improved Estimation of CO2 Saturation,...
Advances in Rock Physics Modelling and Improved Estimation of CO2 Saturation,...Advances in Rock Physics Modelling and Improved Estimation of CO2 Saturation,...
Advances in Rock Physics Modelling and Improved Estimation of CO2 Saturation,...
 
Improved analyses of liquefaction and settlement
Improved analyses of liquefaction and settlementImproved analyses of liquefaction and settlement
Improved analyses of liquefaction and settlement
 
Limitations of simplified methods for estimating seismic settlements linked i...
Limitations of simplified methods for estimating seismic settlements linked i...Limitations of simplified methods for estimating seismic settlements linked i...
Limitations of simplified methods for estimating seismic settlements linked i...
 
Hydrological Modelling of Shallow Landslides
Hydrological Modelling of Shallow LandslidesHydrological Modelling of Shallow Landslides
Hydrological Modelling of Shallow Landslides
 
Combined experimental and modeling procedure
Combined experimental and modeling procedureCombined experimental and modeling procedure
Combined experimental and modeling procedure
 
Assessing Uncertainty of Time Lapse Seismic Response Due to Geomechanical Def...
Assessing Uncertainty of Time Lapse Seismic Response Due to Geomechanical Def...Assessing Uncertainty of Time Lapse Seismic Response Due to Geomechanical Def...
Assessing Uncertainty of Time Lapse Seismic Response Due to Geomechanical Def...
 
Pore scale dynamics and the interpretation of flow processes - Martin Blunt, ...
Pore scale dynamics and the interpretation of flow processes - Martin Blunt, ...Pore scale dynamics and the interpretation of flow processes - Martin Blunt, ...
Pore scale dynamics and the interpretation of flow processes - Martin Blunt, ...
 
IGARSS2011_DInSAR_MORRISON.ppt
IGARSS2011_DInSAR_MORRISON.pptIGARSS2011_DInSAR_MORRISON.ppt
IGARSS2011_DInSAR_MORRISON.ppt
 

Similar to Durner 1994

rophecy,_reality_and_uncertainty_in_distributed_hydrological_modelling.pdf
rophecy,_reality_and_uncertainty_in_distributed_hydrological_modelling.pdfrophecy,_reality_and_uncertainty_in_distributed_hydrological_modelling.pdf
rophecy,_reality_and_uncertainty_in_distributed_hydrological_modelling.pdfDiego Lopez
 
Jwrhe10065 20150204-144509-1366-46826
Jwrhe10065 20150204-144509-1366-46826Jwrhe10065 20150204-144509-1366-46826
Jwrhe10065 20150204-144509-1366-46826yimer mulate
 
dahlstrom_doherty_MODFLOW98
dahlstrom_doherty_MODFLOW98dahlstrom_doherty_MODFLOW98
dahlstrom_doherty_MODFLOW98Dave Dahlstrom
 
Permeability prediction
Permeability predictionPermeability prediction
Permeability predictionsajjad1984
 
APPLICATION OF KRIGING IN GROUND WATER STUDIES
APPLICATION OF KRIGING IN GROUND WATER STUDIESAPPLICATION OF KRIGING IN GROUND WATER STUDIES
APPLICATION OF KRIGING IN GROUND WATER STUDIESAbhiram Kanigolla
 
25. Centrifuge validation of a numerical model for.pdf
25. Centrifuge validation of a numerical model for.pdf25. Centrifuge validation of a numerical model for.pdf
25. Centrifuge validation of a numerical model for.pdfPinakRay2
 
An Application Of Kriging To Rainfall Network Design
An Application Of Kriging To Rainfall Network DesignAn Application Of Kriging To Rainfall Network Design
An Application Of Kriging To Rainfall Network DesignScott Bou
 
Measuring Hydrualic Conductivity Using Petro Physical Measurements 2013 compa...
Measuring Hydrualic Conductivity Using Petro Physical Measurements 2013 compa...Measuring Hydrualic Conductivity Using Petro Physical Measurements 2013 compa...
Measuring Hydrualic Conductivity Using Petro Physical Measurements 2013 compa...Christopher Estevez
 
2005 emerging insights into the dynamics of submarine debris flows
2005 emerging insights into the dynamics of submarine debris flows2005 emerging insights into the dynamics of submarine debris flows
2005 emerging insights into the dynamics of submarine debris flowsshh315
 

Similar to Durner 1994 (20)

poster_7.4_Sulskis
poster_7.4_Sulskisposter_7.4_Sulskis
poster_7.4_Sulskis
 
rophecy,_reality_and_uncertainty_in_distributed_hydrological_modelling.pdf
rophecy,_reality_and_uncertainty_in_distributed_hydrological_modelling.pdfrophecy,_reality_and_uncertainty_in_distributed_hydrological_modelling.pdf
rophecy,_reality_and_uncertainty_in_distributed_hydrological_modelling.pdf
 
79e4150a1090eca7a0
79e4150a1090eca7a079e4150a1090eca7a0
79e4150a1090eca7a0
 
Jwrhe10065 20150204-144509-1366-46826
Jwrhe10065 20150204-144509-1366-46826Jwrhe10065 20150204-144509-1366-46826
Jwrhe10065 20150204-144509-1366-46826
 
Kowalsky_etal_WRR_2001
Kowalsky_etal_WRR_2001Kowalsky_etal_WRR_2001
Kowalsky_etal_WRR_2001
 
dahlstrom_doherty_MODFLOW98
dahlstrom_doherty_MODFLOW98dahlstrom_doherty_MODFLOW98
dahlstrom_doherty_MODFLOW98
 
Permeability prediction
Permeability predictionPermeability prediction
Permeability prediction
 
APPLICATION OF KRIGING IN GROUND WATER STUDIES
APPLICATION OF KRIGING IN GROUND WATER STUDIESAPPLICATION OF KRIGING IN GROUND WATER STUDIES
APPLICATION OF KRIGING IN GROUND WATER STUDIES
 
25. Centrifuge validation of a numerical model for.pdf
25. Centrifuge validation of a numerical model for.pdf25. Centrifuge validation of a numerical model for.pdf
25. Centrifuge validation of a numerical model for.pdf
 
Iv3615381548
Iv3615381548Iv3615381548
Iv3615381548
 
An Application Of Kriging To Rainfall Network Design
An Application Of Kriging To Rainfall Network DesignAn Application Of Kriging To Rainfall Network Design
An Application Of Kriging To Rainfall Network Design
 
Hydrological Modelling of Slope Stability
Hydrological Modelling of Slope StabilityHydrological Modelling of Slope Stability
Hydrological Modelling of Slope Stability
 
Measuring Hydrualic Conductivity Using Petro Physical Measurements 2013 compa...
Measuring Hydrualic Conductivity Using Petro Physical Measurements 2013 compa...Measuring Hydrualic Conductivity Using Petro Physical Measurements 2013 compa...
Measuring Hydrualic Conductivity Using Petro Physical Measurements 2013 compa...
 
Research Plan
Research PlanResearch Plan
Research Plan
 
2005 emerging insights into the dynamics of submarine debris flows
2005 emerging insights into the dynamics of submarine debris flows2005 emerging insights into the dynamics of submarine debris flows
2005 emerging insights into the dynamics of submarine debris flows
 
Icmf2016 ghodke apte
Icmf2016 ghodke apteIcmf2016 ghodke apte
Icmf2016 ghodke apte
 
CSU_Poster
CSU_PosterCSU_Poster
CSU_Poster
 
Kowalsky_etal_AWR_2004
Kowalsky_etal_AWR_2004Kowalsky_etal_AWR_2004
Kowalsky_etal_AWR_2004
 
Scheel et al_2011_trmm_andes
Scheel et al_2011_trmm_andesScheel et al_2011_trmm_andes
Scheel et al_2011_trmm_andes
 
Espino et al. 1995
Espino et al. 1995Espino et al. 1995
Espino et al. 1995
 

Recently uploaded

Java Programming :Event Handling(Types of Events)
Java Programming :Event Handling(Types of Events)Java Programming :Event Handling(Types of Events)
Java Programming :Event Handling(Types of Events)simmis5
 
Glass Ceramics: Processing and Properties
Glass Ceramics: Processing and PropertiesGlass Ceramics: Processing and Properties
Glass Ceramics: Processing and PropertiesPrabhanshu Chaturvedi
 
High Profile Call Girls Nagpur Meera Call 7001035870 Meet With Nagpur Escorts
High Profile Call Girls Nagpur Meera Call 7001035870 Meet With Nagpur EscortsHigh Profile Call Girls Nagpur Meera Call 7001035870 Meet With Nagpur Escorts
High Profile Call Girls Nagpur Meera Call 7001035870 Meet With Nagpur EscortsCall Girls in Nagpur High Profile
 
(ANJALI) Dange Chowk Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...
(ANJALI) Dange Chowk Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...(ANJALI) Dange Chowk Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...
(ANJALI) Dange Chowk Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...ranjana rawat
 
Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...
Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...
Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...Dr.Costas Sachpazis
 
UNIT - IV - Air Compressors and its Performance
UNIT - IV - Air Compressors and its PerformanceUNIT - IV - Air Compressors and its Performance
UNIT - IV - Air Compressors and its Performancesivaprakash250
 
MANUFACTURING PROCESS-II UNIT-5 NC MACHINE TOOLS
MANUFACTURING PROCESS-II UNIT-5 NC MACHINE TOOLSMANUFACTURING PROCESS-II UNIT-5 NC MACHINE TOOLS
MANUFACTURING PROCESS-II UNIT-5 NC MACHINE TOOLSSIVASHANKAR N
 
Extrusion Processes and Their Limitations
Extrusion Processes and Their LimitationsExtrusion Processes and Their Limitations
Extrusion Processes and Their Limitations120cr0395
 
Sheet Pile Wall Design and Construction: A Practical Guide for Civil Engineer...
Sheet Pile Wall Design and Construction: A Practical Guide for Civil Engineer...Sheet Pile Wall Design and Construction: A Practical Guide for Civil Engineer...
Sheet Pile Wall Design and Construction: A Practical Guide for Civil Engineer...Dr.Costas Sachpazis
 
Booking open Available Pune Call Girls Pargaon 6297143586 Call Hot Indian Gi...
Booking open Available Pune Call Girls Pargaon  6297143586 Call Hot Indian Gi...Booking open Available Pune Call Girls Pargaon  6297143586 Call Hot Indian Gi...
Booking open Available Pune Call Girls Pargaon 6297143586 Call Hot Indian Gi...Call Girls in Nagpur High Profile
 
UNIT-II FMM-Flow Through Circular Conduits
UNIT-II FMM-Flow Through Circular ConduitsUNIT-II FMM-Flow Through Circular Conduits
UNIT-II FMM-Flow Through Circular Conduitsrknatarajan
 
Booking open Available Pune Call Girls Koregaon Park 6297143586 Call Hot Ind...
Booking open Available Pune Call Girls Koregaon Park  6297143586 Call Hot Ind...Booking open Available Pune Call Girls Koregaon Park  6297143586 Call Hot Ind...
Booking open Available Pune Call Girls Koregaon Park 6297143586 Call Hot Ind...Call Girls in Nagpur High Profile
 
MANUFACTURING PROCESS-II UNIT-2 LATHE MACHINE
MANUFACTURING PROCESS-II UNIT-2 LATHE MACHINEMANUFACTURING PROCESS-II UNIT-2 LATHE MACHINE
MANUFACTURING PROCESS-II UNIT-2 LATHE MACHINESIVASHANKAR N
 
Call for Papers - Educational Administration: Theory and Practice, E-ISSN: 21...
Call for Papers - Educational Administration: Theory and Practice, E-ISSN: 21...Call for Papers - Educational Administration: Theory and Practice, E-ISSN: 21...
Call for Papers - Educational Administration: Theory and Practice, E-ISSN: 21...Christo Ananth
 
University management System project report..pdf
University management System project report..pdfUniversity management System project report..pdf
University management System project report..pdfKamal Acharya
 
KubeKraft presentation @CloudNativeHooghly
KubeKraft presentation @CloudNativeHooghlyKubeKraft presentation @CloudNativeHooghly
KubeKraft presentation @CloudNativeHooghlysanyuktamishra911
 
College Call Girls Nashik Nehal 7001305949 Independent Escort Service Nashik
College Call Girls Nashik Nehal 7001305949 Independent Escort Service NashikCollege Call Girls Nashik Nehal 7001305949 Independent Escort Service Nashik
College Call Girls Nashik Nehal 7001305949 Independent Escort Service NashikCall Girls in Nagpur High Profile
 

Recently uploaded (20)

Java Programming :Event Handling(Types of Events)
Java Programming :Event Handling(Types of Events)Java Programming :Event Handling(Types of Events)
Java Programming :Event Handling(Types of Events)
 
DJARUM4D - SLOT GACOR ONLINE | SLOT DEMO ONLINE
DJARUM4D - SLOT GACOR ONLINE | SLOT DEMO ONLINEDJARUM4D - SLOT GACOR ONLINE | SLOT DEMO ONLINE
DJARUM4D - SLOT GACOR ONLINE | SLOT DEMO ONLINE
 
Glass Ceramics: Processing and Properties
Glass Ceramics: Processing and PropertiesGlass Ceramics: Processing and Properties
Glass Ceramics: Processing and Properties
 
High Profile Call Girls Nagpur Meera Call 7001035870 Meet With Nagpur Escorts
High Profile Call Girls Nagpur Meera Call 7001035870 Meet With Nagpur EscortsHigh Profile Call Girls Nagpur Meera Call 7001035870 Meet With Nagpur Escorts
High Profile Call Girls Nagpur Meera Call 7001035870 Meet With Nagpur Escorts
 
(ANJALI) Dange Chowk Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...
(ANJALI) Dange Chowk Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...(ANJALI) Dange Chowk Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...
(ANJALI) Dange Chowk Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...
 
Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...
Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...
Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...
 
UNIT - IV - Air Compressors and its Performance
UNIT - IV - Air Compressors and its PerformanceUNIT - IV - Air Compressors and its Performance
UNIT - IV - Air Compressors and its Performance
 
MANUFACTURING PROCESS-II UNIT-5 NC MACHINE TOOLS
MANUFACTURING PROCESS-II UNIT-5 NC MACHINE TOOLSMANUFACTURING PROCESS-II UNIT-5 NC MACHINE TOOLS
MANUFACTURING PROCESS-II UNIT-5 NC MACHINE TOOLS
 
Extrusion Processes and Their Limitations
Extrusion Processes and Their LimitationsExtrusion Processes and Their Limitations
Extrusion Processes and Their Limitations
 
Sheet Pile Wall Design and Construction: A Practical Guide for Civil Engineer...
Sheet Pile Wall Design and Construction: A Practical Guide for Civil Engineer...Sheet Pile Wall Design and Construction: A Practical Guide for Civil Engineer...
Sheet Pile Wall Design and Construction: A Practical Guide for Civil Engineer...
 
Booking open Available Pune Call Girls Pargaon 6297143586 Call Hot Indian Gi...
Booking open Available Pune Call Girls Pargaon  6297143586 Call Hot Indian Gi...Booking open Available Pune Call Girls Pargaon  6297143586 Call Hot Indian Gi...
Booking open Available Pune Call Girls Pargaon 6297143586 Call Hot Indian Gi...
 
UNIT-II FMM-Flow Through Circular Conduits
UNIT-II FMM-Flow Through Circular ConduitsUNIT-II FMM-Flow Through Circular Conduits
UNIT-II FMM-Flow Through Circular Conduits
 
Booking open Available Pune Call Girls Koregaon Park 6297143586 Call Hot Ind...
Booking open Available Pune Call Girls Koregaon Park  6297143586 Call Hot Ind...Booking open Available Pune Call Girls Koregaon Park  6297143586 Call Hot Ind...
Booking open Available Pune Call Girls Koregaon Park 6297143586 Call Hot Ind...
 
MANUFACTURING PROCESS-II UNIT-2 LATHE MACHINE
MANUFACTURING PROCESS-II UNIT-2 LATHE MACHINEMANUFACTURING PROCESS-II UNIT-2 LATHE MACHINE
MANUFACTURING PROCESS-II UNIT-2 LATHE MACHINE
 
Call for Papers - Educational Administration: Theory and Practice, E-ISSN: 21...
Call for Papers - Educational Administration: Theory and Practice, E-ISSN: 21...Call for Papers - Educational Administration: Theory and Practice, E-ISSN: 21...
Call for Papers - Educational Administration: Theory and Practice, E-ISSN: 21...
 
University management System project report..pdf
University management System project report..pdfUniversity management System project report..pdf
University management System project report..pdf
 
Roadmap to Membership of RICS - Pathways and Routes
Roadmap to Membership of RICS - Pathways and RoutesRoadmap to Membership of RICS - Pathways and Routes
Roadmap to Membership of RICS - Pathways and Routes
 
KubeKraft presentation @CloudNativeHooghly
KubeKraft presentation @CloudNativeHooghlyKubeKraft presentation @CloudNativeHooghly
KubeKraft presentation @CloudNativeHooghly
 
College Call Girls Nashik Nehal 7001305949 Independent Escort Service Nashik
College Call Girls Nashik Nehal 7001305949 Independent Escort Service NashikCollege Call Girls Nashik Nehal 7001305949 Independent Escort Service Nashik
College Call Girls Nashik Nehal 7001305949 Independent Escort Service Nashik
 
Water Industry Process Automation & Control Monthly - April 2024
Water Industry Process Automation & Control Monthly - April 2024Water Industry Process Automation & Control Monthly - April 2024
Water Industry Process Automation & Control Monthly - April 2024
 

Durner 1994

  • 1. WATER RESOURCES RESEARCH, VOL. 30, NO. 2, PAGES 211-223,FEBRUARY 1994 Hydraulicconductivityestimationfor soilswith heterogeneouspore structure WolfgangDurner InstituteofTerrestrialEcology,SoilPhysics,FederalInstituteofTechnology,Z•irich,Switzerland Abstract.The hydraulicconductivityfunction,whichis requiredto solvetheRichards equation,isdifficulttomeasure.Thereforepredictionmethodsarefrequentlyused wheretheshapeof the conductivityfunctionis estimatedfromthe moreeasily measuredwaterretentioncharacteristic.Errorsin conductivityestimationscanarise eitherfroman invalidityof the predictionmodelfor a givensoil,or fromanincorrect descriptionoftheretentiondata.Thisseconderrorsourceisparticularlyimportantfor soilswithheterogeneousporesystemsthat cannotbe adequatelydescribedby the usuallyusedretentionfunctions.To describetheretentioncharacteristicsofsuchsoils, aflexible0(½)functionwasformedby superimposingunimodalretentioncurvesofthe vanGenuchten(1980)type. By combiningthisretentionmodelwith the conductivity predictionmodelofMualem(1976),conductivityestimationsforsoilswith heterogeneousporesystemsareobtained.Estimatedconductivitiesbythismodeland theclassicalvan Genuchten-Mualemmethodcandiffer by ordersof magnitude.Thus reporteddisagreementsbetweenmeasuredandestimatedconductivitiesmayinsome casesbedueto aninadequatedescriptionof theretentiondataratherthandueto a failureof the predictionmodel. Introduction Theincreasingconcernwith groundwaterpollutionand contaminationof soilsby hazardoussubstanceshasstimu- latedthedevelopmentof numerousmathematicalmodelsof pollutanttransportin soils.Today,thenumericalsimulation ofwaterand pollutanttransportin unsaturatedsoilshas becomea standardtool for assessingenvironmentalpoilu- tionproblems.Despitetheexistenceandtheuseof simplis- ticmodelsfor managementpurposes[e.g., Carsel et al., 1985;Barraclough,1989](seeoverviewsby Addiscottand Wagenet[1985]andLoagueandGreen[1991]),aswellasthe recentintensivedevelopmentof specificmacroporemodels [e.g.,Jarviset al., 1991;Chen and Wagenet, 1991;Gerke andvanGenuchten,1993],the mostimportantapproaches tomodelingtransientwater and solutetransportin the vadosezoneare basedon the Richardsequation.To solve thisequation,theknowledgeof the effectivesoilhydraulic properties,namely,thesoilmoisturecharacteristic0(½)and theK(O) or K(½) relationshipoverthe wholerangeof moistureconditionsis required.Here, 0 (cubiccentimeters percubiccentimeter)is the volumetric water content, ½ (centimeters)is the matricpressurehead,andK (centime- tersperday)is thehydraulicconductivity(a scalarproperty fortheone-dimensionalcase).For usein models,thehy- draulicpropertiesare commonlyexpressedby analytical functions,oftencalledparametericmodels,as,forexample, listedby van Genuchtenand Nielsen [1985],Bruceand Luxmoore[1986],andMualem[1986].The parametersof thesemodelscanbe obtainedby fittingthe functionsto experimentalwater retention and conductivity data. Alter- natively,theycanbeestimatedfrommoreeasilymeasured Copyright1994bytheAmericanGeophysicalUnion. Papernumber93WR02676. 0043-1397/94/93WR_02676505.00 soilproperties[Bloemen,1980;Saxtonet al., 1986;WOsten and vanGenuchten,1988;Vereeckenet al., 1989],or deter- minedbyparameterestimationtechniqueswhentheinverse problemis solved[Zachmannet al., 1981,1982;Hornung, 1983;Kool et al., 1987;Kool andParker, 1988]. Of all hydraulicproperties,the unsaturatedhydraulic conductivityK is mostdifficultto measure.Thereforethe use of indirect methodswhere K(0) is estimatedfrom more easilymeasuredsoilpropertieshasbecomemoreandmore common[van Genuchtenand Leij, 1992].A potentially powerfulclassofmethodsresultsfrompore-sizedistribution models,wherethewaterretentioncurveofaporousmedium is interpretedasstatisticalmeasureof its equivalentpore- sizedistribution[Corey,1992].In thisapproach,theconduc- tivityisestimatedbyapplyingtheconceptof viscousfluid flowthroughcapillariesandbyusingaconceptualmodelto describethe pore interactionand pore connectivity [Mualem,1986].Threebasicgroupsofstatisticalprediction modelsmaybedistinguished:(1)thecapillary-bundlemodel ofPurcell[1949]withitssubsequenttortuositycorrections byYuster[1951],Burdine[1953],WyllieandGardner[1958], andAlexanderand Skaggs[1986];(2) the cut andrandom rejoinmodelof ChildsandCoilis-George[1950]withits subsequentmodificationsbyMarshall[1958],Millingtonand Quirk[1961],Campbell[1974]andothers;and,mostre- cently,(3)theslabmodelofMualem[1976].Mualemand Dagan[1978]andMualem[1986,1992]providecomprehen- sivereviewsfor thesegroupsof conductivityestimation models. Closed-formexpressionsfortheretentionfunctionandthe conductivityfunctionareoftenusedinparameterestimation techniqueswheretheinverseproblemissolved[Kooland Parker,1987a,b].Applyingapredictivemodelthatcouples theconductivityfunctionwiththeretentionfunctionisthen ofparticularadvantagebecauseitminimizesthenumberof 211
  • 2. 212 DURNER: HYDRAULIC CONDUCTIVITY ESTIMATION FOR SOILS parameters needed to express the hydraulic relationships. For the validity of this approachit is essentialthat the prechosenfunctional representationof both the retention curve and the conductivity curve as well as the model that couplesthe two functions are appropriateover the whole moisture range. In principle, the statisticalmodelscanbe usedto estimate absoluteconductivityvalues,but in practicethey are used only to derive the shapeof the conductivityfunctionfrom the shapeof the retentioncurve.To obtainabsolutevalues, theestimatedconductivitycurvemustbe scaledwithatleast onemeasuredconductivity,the "matchingvalue" [Nielsen et al., 1960].Most frequently,the saturatedconductivityis taken as matchingvalue, althoughseveral authors[van GenuchtenandNielsen, 1985;Mualem, 1986;Luckneretal., 1989;Nielsen and Luckher, 1992]have warnedagainstthis practice. Errors of conductivity estimates may arise from two sources:(1) errorsdueto aninappropriatedescriptionofthe soil moisture characteristic and (2) errors stemmingfrom failureof thepredictionmodel.Beginningwiththeworkof Nielsen et al. [1960], numerousstudies have compared estimated with measured conductivities and have discussed the validity of predictionmethodsfrom an empiricalor semiempiricalpoint of view [e.g., KabIan et al., 1989; Michiels et al., 1989;Stephens,1992].As a generalresult from theseinvestigations,conductivityestimationmethods appearto berelativelyreliablefor sandysoilswithnarrow particlesize distributions,whereasfor loamyandclayey soils they often fail. This failure is mostly attributedto incorrect values of the tortuosity coefficient (or pore- interactionfactor) r (dimensionless),whichis an empirical parameterin thepredictingequations(see(6)inthetheory section).SchuhandCline [1990]haveshownthatßcanvary overa widerange.In additionto theseexperimentalinves- tigations,theoreticaldiscussionsand reviews[Brutsaert, 1968;Gardner,1974;Mualem, 1976;Parkesand Waters, 1980;van Genuchtenand Nielsen,1985;Mualem, 1986; AlexanderandSkaggs,1986]aswellasmathematicalsensi- tivity analyses[Sakellariou-Makrar•tonakiet al., 1987;Vo- gel and Cislerova,1988;Sidiropoulosand Yannopoulos, 1988]havebeenpublished.Despitetheextensivetreatment ofthepredictionproblemanddueto thelackof a reliable data base of measurementsin undisturbedsoils,the results appearambiguousandinconclusive.A statementwhichwas publisheda decadeagoappearsto be stillvalid:"The problemofwhich[conductivity]measurementmethodand whichpredictionmethodaremostvalidisundoubtedlyone of the most vexingonesin contemporarysoil physics" [Ragabet al., 1981,p. 384]. When discussingvalidity problems,mostinvestigators have concentratedon the failure of the predictivemodel, moreor lessdisregardingtheproblemofinterferenceswith thefirst errorsource.No investigationhasbeenpublishedso far on the systematicerrorsof conductivityestimation methodsfor soilswith heterogeneouspore systemswhich arisefromtheuseof nonoptimalwaterretentionfunctions. Theobjectiveofthispaperistoanalyzeconductivityesti- mationsfor suchsoils.After reviewingthecharacteristicsof heterogeneouspore systems,conductivityestimatesare obtainedby combininga flexibleretentionfunctionwith Mualem's[1976]model.The resultsof thismethodare comparedwithresultsobtainedbyclassicalunimodalmod- els.Thedifferencesareanalyzedandthegenerallimitations ofstatisticalconductivityestimationmethods,particularly for soilswith a nonnegligiblefraction of largepores,are discussed. Theory Water Retention Curves and Equivalent Pore Size Distributions Theporesystemofa rigidsoilcanbe characterizedbyits equivalentpore-sizedistribution,which is commonlycalcu. latedfrom the specificmoisturecapacityC* = dO/de. Accordingto Laplace's law, the matric pressureat whicha water-filled pore startsto drain is inverselyproportionalto theequivalentradiusr of theporenecks.If C* asameasure of pore densityis plottedversus• on a linear scale,pores withlargeradiiofdifferentordersof magnitudearesqueezed together,whichmaskstheshapeofthepore-sizedistribution at the hydrologically important range close to saturation. Sincethe matricpressureheadvariesover ordersofmagni. tude,theretentioncharacteristicsarefrequentlyplotted0na logarithmicpressureheadscale,whichis convenientlyex- pressedinpF units,definedfor theunsaturatedrangebypœ -- log•0(-½) [Schofield,1935],where ½ is expressedin centimeters.The pF scale is equivalent to a logarithmic scalefor the pore radius,and the derivativeof thewater contentversuspF thereforeappropriatefor visualizingthe pore-sizedistributionover differentordersof magnitude. The so-definedpore-sizedensityis relatedto C* by dO(r) dO(pF) dO(½t) de dO() d logxor dpF d logo I,l d 1ogo = [loge (10)]I½1C*. Thepore-sizedistributionsthatare depictedinFigures 1-11 are calculatedby (1). Accordingly,the shadedarea betweentwopF valuesindicatesthefractionofporespace thatdrainswhenthepressureheadischangedbyoneorder of magnitude. Numerousmathematicalfunctionshavebeenusedinthe literatureto express0(½)in analyticalform[Brooksand Corey,1964;King,1965;Brutsaert,1966;Visser,1968: Laliberte,1969;FarrelandLarson,1972;Rogowski,1972: Su andBrooks,1975;Haverkampet al., 1977;Gillham, 1976;Simmonsetal., 1979;D'Hollander,1979;vanGenu. chten,1980].Mostpopularamongthesefunctionsarethe equationsofBrooksandCorey[1964](hereinanotationthat isslightlymodifiedfromtheoriginalversion)' = > s, = I 1 and,morerecently,theequationofvanGenuchten[1980]: Se= [1+ -m, whereSeistheeffectivesaturation(dimensionless),defined bySe= (0- 0•)/(0•- 0•),withOsand0•indicatingthe saturatedandresidualvolumetricwatercontents,respec- tively,a> 0(centimeters)isascalingfactorthatdetermines thepositionoftheporesizemaximum,andX,n,andmare dimensionlesscurve-shapeparameters••bjecttoA>0,m >0, andn > 1.
  • 3. DURNER: HYDRAULIC CONDUCTIVITY ESTIMATION FOR SOILS 213 -103 • 8 0.25 ......................................................................................1 • _10-2 _104 _10ø _10• _102 reducedpressureheadW* Figure1. Retentioncurve(solidline), first derivative dSeldlog•* (shadedarea),andsecondderivatived2Se/d (log•,)2 (dashedline)ofthevanGenuchtenfunctionwith n = 1.5 and rn = 1 - 1/n. The reducedpressurehead (dimensionless)is definedby •/a. The shadedarearepre- sentsthepore-sizedistribution.The positionof maximum poredensityisat½*= -(I/m) Equation(2)and(3)caneasilybecombinedwithconduc- tivitypredictionmodelsto obtainclosed~formexpressions forK(O) andK(•) [Mualem, 1992].By this, both 0(•) and K(½)are determinedby the parametersof the retention curveplus one additionalmatchingfactor to scale the predictedhydraulicconductivityfunction.BrooksandCorey [1964]combinedtheirfunctionwith thepredictivemodelof Burdine[1953]; Campbell [1974] combinedthe samereten- tionfunctionwith a simplifiedversion of the Childs and Coilis-George[1950]predictionmodel.The retentionmodel ofvanGenuchten[1980]is mostfrequentlycoupledwith the predictivemodelof Mualem [1976].The resultingset of hydraulicfunctionshas been thoroughlyreviewed by van Genuchtenand Nielsen [1985], Luckner et al. [1989], and Nielsenand Luckner [1992]. Kool and Parker [1987a] ex- tendedit to include hysteresisand air entrapment. All the above-listed retention functions reflect unimodal and,if continuouslydifferentiable,smoothnormalto lognor- malshapedpore-sizedistributions.This is illustratedby Figure1,wheretheretentionmodelof van Genuchten[1980] isdepictedtogetherwithitsunderlyingpore-sizedistribution (shadedarea). As comparedto the Brooks and Corey thnction,the van Genuchten function has the favorable propertythat its derivative dO/dpF is continuousand be- comesasymptoticallyzero toward the fine and largepores. Theextensionof the pore-sizedistributiontowardthefine poresisdeterminedbytheproductmn(whichfor a½>> I is equivalentto • in the Brooks and Corey function), whereas theextensiontowardthe largeporesis determinedby the ratiom/n.Thelargerthisratiois,thesteeperisthedecrease inporedensityfromitsmaximumtowardthelargepores.As •ill beshownlater, the width of the pore-sizedistribution towardthelargeporesis of particularimportancefor esti- matingthe conductivityfunction. Summarizing,the van Genuchtenretentionmodelinitsgeneralform(3)represents acontinuous,smooth,unimodal,and bell-shapedpore-size distributionwiththeparametera primarilydeterminingthe positionoftheporedensitymaximumandtheparametersm andndeterminingthewidthtowardthefineandlargepore sizes.VanGenuchtenandNielsen[1985]comparedvarious •ater retention models and concluded that (3) fits the water retentiondataof mostsoilsalmostperfectly. However, to obtain the conductivity estimates as closed- form functions,the parametersof the van Genuchtenequa- tion mustbe subjectedto the additionalconstraintm + l/n - i, where i is an integer value, for the van Genuchten- Mualem model generally taken as unity [van Genuchten, 1980].This constraint eliminates someof the flexibility of the function, since a pore-size distribution that extends far toward fine pores is then always coupledwith a relatively broad distribution towards the large pores, and vice versa. Thisis in fairly goodaccordancewith experimentaldatafor many soils.There are casesof fine-textured soils, however, where this constraint attributes a part of the pore spaceto unrealistically large pores. There is some discussion whether the van Genuchten functionis a physicallyrealisticdescriptionof 0(•) overthe whole moisturerangewhen its parametern lies in the range 1 < n < 2 [Nielsen and Lucknet, 1992]. In this case, the secondderivative,d20/d•2, becomesinfinitetowardsatu- 0 ration,andtheintegral.f0,• d0forlim•0 isunbounded. This means that an infinite amount of work were necessary to extract all water from the soil [Mualern, 1978;Sakellario- Makrantonaki et al., 1987]. According to our view, for practicalpurposesthisdoesnotrestricttheusefulnessofthe van Genuchten function as an empirical model for the retentioncurve. Nevertheless,it appearsimportantto real- izethatit expressestheretentioncharacteristicaswellasthe derivedpropertyC* andtherelatedconductivityestimation in a purelydescriptivefashion.The coefficientsshouldnot be interpretedasparametersin a physicalsense. Water Retention Characteristics of Soils With HeterogeneousPore Systems Undisturbed soilsfrequently have pore systemsthat are differentfrom the unimodal,approximatelynormaldistrib- utedtypewhichis representedby Figure1. Consequently, attemptsto fit theirretentiondatawitha simplesigmoidal curveleadto unsatisfyingresults.A schemeof typicalfitting errorsisdepictedinFigure2. It appearsthatsuchdeviations aretacitlyassumedto bewithintheconfidenceintervalof thecorrespondingmeasuredvalues.If theyare not, such deviationswould be relevant as will be shown later. This could be used to define the term "heterogeneous" with respectto hydrologicalproperties.Wecandesignatea pore systemasheterogeneousif thepore-sizedistributionof a representativeelementaryarea[BearandBachmat,1991] cannotbe correctlydescribedby the van Genuchtenfunc- tion(oranyotheroftheabovementionedunimodalretention functions). Poresystemsthatarenonconformwithasimplesigmoidal retentioncharacteristicmay be a resultof specificparticle- size distributionsor be due to the formation of secondary poresystemsbyvarioussoilgeneticprocesses.Aggregation processesinloamysoilsmayleadtopore-sizedistributions thatcausetypicalfittingproblemsin the midporerange,as illustratedin Figure 2 I,bottomleft) [e.g., Sharma and Uehara, 1968;Smettemand Kirkby, 1990].Besidesaggre- gation,biologicalsoil-formingprocessesmayleadtosecond- aryporesystemsin thelarge-porerangein soilsof any texture{Figure2 (topleft))[e.g.,Othmeret al., 1991;De Jongetal., 1992].Morainicsoilsandsolifluctionsoilstend tohavea significantfractionofporesovera widerangeof tensions,whichleadsto fittingproblemsas illustratedin Figure2 (topright)[e.g.,ParkesandWaters,1980;Jacob-
  • 4. 214 DURNER: HYDRAULIC CONDUCTIVITY ESTIMATION FOR SOILS pressure head Figure 2. Schematictypes of deviationsbetween retention data of structured soils and fitted unimodal retention curves. The circles indicate typical trends, which help to identify structured pore systems even in cases where only few supporting data are available. (Top left) Increasing devia- tions between data and fitted curve when approachingsatu- ration, observed in undisturbed soils of any texture [e.g., Othrner et al., 1991;De Jong et al., 1992]. (Top fight) More or less straight retention characteristic in the wet rangeup to pF3-4, where the main pore system is located. Typical for morainic and solifiuction soils [e.g., Schj•nning, 1985; Ja- cobsen, 1989]. (Bottom left) Different slopes of fitted curve and data in the midpore range. Typical for aggregatedloams [e.g., Sharma and Uehara, 1968; Smettem and Kirkby, 1990]. (Bottom right) Considerable change in water content very close to saturation,experimentally observedin uncon- solidatedsands(see,for example, King [1965](liquid: soltrol C), Ragab et al. [1981],and Stephensand Rehfeldt[1985]). mi) aresubjectto theconditionso•i > 0, mi > O,ni > 1. Weexplicitlydonotimposetheadditionalconstraintmi + 1/ni= 1. Figure3 showsthesuperpositionoftwohypothet. ical pore-sizedistributionswhich definea bimodalpore systemaccordingto (4). The dark-shadedarea canbe interpretedasfine-texturedpore system,with a maximum pore-sizedensityaroundpF4 (& r = 0.2/am), whereasthe light-shadedareacan be seenas a secondaryporesystem with its maximum density in the range of r = 0.1 mm. The superpositionof two unimodalporesystemshasbeen previouslyusedby PetersandKlavetter[1986]to describe the hydraulicpropertiesof fracturedrock. Othmeretal. [1991]andRossand Smettem [1993]usedthis approachto describe the retention characteristics of soils with distinct secondarypore systems.Pruesset al. [1990]modeledthe coupledtransportof water, heat, and air in partiallysatu- ratedporousrockby explicitlyconsideringfractureeffects. They comparedthe resultswith those obtainedby treating theporousmediumasa singlecontinuumandderivedsimple criteriafor the applicabilityof the effective continuumap- proximation.WangandNarasimhan[1985,1990]addressed theproblemof findingeffectiveparametervaluesandcom- paredfracturesin tuff with macroporesin soil. The multimodal retention model, (4), keeps the functional propertiesof thebasicvan Genuchtenmodel,but is,dueto the increased number of coefficients, able to account forthe deviations discussedabove. It is continuously differentiable, asymptoticto a zero slopetowardsthe fineandlargepores, andstronglymonotonicover the whole moisturerange.By keepingk small,the functionis well-behavedfor interpola- tion purposes,reducingthe noisein measureddatawhile followingtheshapeof the measuredcurve.Hampton[1990] showedthat the useof unconstrainedsplinesfor describing sen, 1989;Lafolie et al., 1989].With unconsolidatedsands, retention characteristics of the type in Figure 2 (bottom right)havebeenobserved,evenin caseswherethiscannot bedueto boundaryeffectsat thebeginningofdrainage[e.g., King, 1965;Ragab et al., 1981;Stephensand Rehfeldt, 1985].In naturalsoils,the schematictypesof Figure2 will always appearto some degreein combination(compare examplesof Boelset al. [1978],CurrieandRose[1985], Dixon [1972],Hantschelet al. [1987],Jeppsonet al. [1975], Parker and van Genuchten[1985],Puckett et al. [1985],and Tsuji et al. [1975]).Note that theseheterogeneouspore systemscanbefoundforfieldaveragesaswellasforsingle soilsamples[e.g.,AndersonandCassel,1986]. Multimodal Retention Function Toproperlydescribetheretentioncharacteristicsofsoils withheterogeneousporesystems,we introducea multimo- dal retention function which is constructedby a linear superpositionof subcurvesof the van Genuchtentype [Durner, 1991, 1992]: k Se--Z wi i=! 1+ (ai!½!)n" ' (4) wherek is the numberof "subsystems"thatformthetotal pore-sizedistribution,andwi areweightingfactorsforthe subcurves,subjectto0 < wi < 1and5'.wi = 1.Asforthe unimodalcurve,the parametersof the subcurves(ori, ni, ._o 0.8 '• 0.6 ß.= 0.4 • 0.2 0 2 4 pF 0.2 ........... 1500 15 .15 poreradius(I,Lm) .0015 Figure3. Constructionofamultimodalretentionfunction, (4).(Top)Bimodalretentioncurve(solidline),unimodal subcurveforthetexturalporesystem(dottedline,wi =0.6, al =0.005cm-l nl = 14 mi= 1- 1/ni)andunimodal subcurvefor a secondaryporesystem(dashedline.w2= 0.4,a2= 0.1cm-1 n2= 2 2 m2= 1- 1/n,).(Bottom) Pore-sizedistributionsof bimodalfunction(s•lidline),of texturalporesystem(darkshaded),andof secondarypore system (light shaded).
  • 5. DURNER: HYDRAULIC CONDUCTIVITY ESTIMATION FOR SOILS 215 pF pF retentiondatacanleadtoconsiderableproblemsinnumer- 0 icalsimulationsofwatertransportduetotheroughshapeof 1.2 10ø thewatercapacityfunction.Incontrasttoconstrainedspline interpolations,whichappearstobethebestalternative•,, 10scheme,(4)requireslesscoefficients.Finallyitscoefficients• 0.8 can,toacertaindegree,beinterpretedinthesamemannerasthebasicvanGenuchtencoefficients,namely,•i=l/oti, 0.4 indicatingthepositionsofporedensitymaxima,andhimi ß •ø• andni/mi, determiningthewidthoftheunderlyingpore-size 0.2 distributions.However,if theporesystemisnotdistinctly 0 bimodalor multimodal,thecoefficientsof the multimodal retentionfunctiontendtobehighlycorrelated.It istherefore importantthattheynotbeconsideredas"parameters"with aphysicalmeaning;rathertheymustbeseenascurveshape coefficientslikethoseof anyalternativeinterpolationfunc- tion. Themultimodalretentionfunctionis fitted to data by minimizingtheobjectivefunctionZ: N g(P)= • •oill0i- • (½i,P)II (5) i=1 whereP= {0s,Or,ai,hi,mi,wi}Tistheparametervector, Nisthenumberof measureddata, ooiaretheweightsfor the leastsquaresminimization,p.determinestheminimization norm,andOiandb(½i,P) arethemeasuredandcalculated watercontentsat pressure½i, respectively.If p. = 2, (5) resultsin a weightedleast'squaresscheme,which corre- spondstoamaximumlikelihoodfitundertheassumptionof independentandnormallydistributeddataerrors.If outliers aremorefrequent,p.canbe setto unity.In thiscase,(5) minimizesthe sumof absolutedeviationsbetween measured watercontentsandthe retention curve. The fitting procedure weuseis basedon a multidimensionalbracketing method. Thenumericalprocedureisrobustandunconditionallycon- vergent,butcomputationallynot very efficient.For each coefficientPi, thefollowingpropertiesmaybetailoredfor specificneeds'(1) initial guessof Pi, (2) Pi constantor optimized,and(3)allowedminimumandmaximumvalueof Pi.Theoptimizationisterminatedwhentheimprovementof theobjectivefunctionorthechangesofallparametervalues arebelowtheir tolerance values between two iteration steps, (•Z< Sz,Api< epi),orwhenthemaximumnumberof iterationsisexceeded.(The programis availablefromthe authoronrequest.) ConductivityEstimation Therelativehydraulicconductivityfunctioniscomputed by numericalevaluation of Mualem's [1976] predictive modelonbaseof the unimodalor multimodalrepresentation ofq•(Se): g--SeLf()]' (6) x•heref isgivenby Se1,is;. f(Se)-- ½(5;) (7) In[6),SJisanempiricalcorrectionfunction,whichallows foraprioriunknowneffectsofporeconnectivityandtortu- 1õ 1.5 0.01õ 0 0.2 0.4 0.6 0.8 poreradius(I.U'n) effectivesaturation Figure4. Conductivityestimationfor a structuredpore system.(Left) Bimodalretentioncurve (solid line) and underlyingpore-sizedistribution(shadedarea).(Right)Pre- dictedrelativeconductivityfunctionsK(½) (fromtopleftto bottomright,relatedtotopaxis)andK(Se) (fromtopfight to bottomleft, relatedto bottomaxis).The coefficientsare }t,•= 0.1, a• = 0.2cm-1, - , = 0.1, a2 = 0.001cm-l n2 = 22 andm2 = 1 - 1/n,. osityin Mualem'smodel.Mualem[1976]foundanaverage valueofr = 0.5tobeoptimalfora setof45samplesoils.The absoluteconductivityKabs(Se)canbecalculatedbymatch- ingthepredictedfunctionK(S•) to a matchingvalueKref, measuredat somereferencesaturationSref, accordingto Kabs(Se)= (Se/Sref)•'KrefK(Se)' (8) Results and Discussion ConductivityEstimationfor HeterogeneousPoreSystems Forthepurposeofcomparingconductivityestimatesfor bimodalandmultimodalsoils,we assumeat this pointthat boththe retentionmeasurementsof the examplesoilsand thepredictivemodelitselfarefreeoferror.Ofcourse,these assumptionsare notgenerallyjustifiedandhaveto be discussedlater in their own right. InFigure4,thehydraulicpropertiesofahypotheticalsoil with a narrowtexturalpore-sizedistributionand distinct secondaryporesystemareshown.Twofindings,whichare ofgeneralnature,areclearlyillustrated:(1)theshapeofthe retentioncurveisdirectlyreflectedintheshapeof K(½)and (2)thesecondaryporesystemincreasestheconductivityby someordersofmagnitudewhiletakingonlyfewpercentof porespace.TheparticularshapeofK(0)cannotbereflected byunimodalmodels.The"similarity"betweentheshapes of0(½)andK(½)isphysicallymeaningful:atpressureranges wherethe specificwatercapacityis high,indicatedby a steepslopeoftheretentionfunction,theconductivitymust dropsteeplyaswell,sinceduringadrainageprocessalarge portionoftheporespaceemptiesin thatrange.A steep decreaseoftheconductivityclosetosaturationistypicalfor soilswith a secondaryporesystemin the large-porerange. Compositeconductivityfunctionswithsimilarcharacteris- ticshavebeenderivedfromexplicitconsiderationof two poroussystems,e.g.,for fracturedrock[Wangand Narasimhan,1985;PetersandKlavetter,1988]orforsoils withmacropores[Othmeretal., 1991;ChenandWagenet, 1991;Chenet al., 1993]. If oneattemptstodescribeabimodalsoilwithsimplified unimodalhydraulicfunctions,theerrorinvolvedismuch
  • 6. 216 DURNER: HYDRAULIC CONDUCTIVITY ESTIMATION FOR SOILS 100 ......"' 0 2 pF 4 6...... 0 2 pF4 6 ..-..........,. c .... 11.1.:::::::...%I,,,"" //' 0.4 ............ ), 10'••or ?? function ,•o-•/ ,/ A'tr•' • o.•t-........ "•-_I • .......:....,¾,.....i ,::..,i...............h• 10• • o '150 1.5 0.015 0.1 0.2 0.3 0.4 0.5 10'4 ix)reradius(!.u'n) watercontent Figure7. Hydraulicpropertiesof RideauClay Loam[De Jonget al., 1992].See Figure6 for explanations. 0 0.1 0.2 0.3 0.4 0.5 water content Figure 5. Schematic representation of the conductivity calibration problem for bimodal pore system. Line A, con- ductivity curve of a bimodal pore system; line B, unimodal predicted curve matched to unsaturated conductivity; line C, unimodal predicted curve matched to saturated conductiv- ity; and line D, unimodal predicted curve with optimized pore interaction factor r. large[for theconductivityestimationsthanfortheretention function. Matching the unimodal conductivity function with the measuredsaturatedconductivity value Ks, asis common practice,will leadto an overestimationof the unsaturat6d conductivity in the whole unsaturatedmoisturerange(Fig- ure 5, curve C). If, on the other hand, the unimodalpredicted function is matched to an unsaturated conductivity value, the increaseof hydraulicconductivitynear saturationis not reflected, thus leading to a gross underestimationof the soil's conductivity near saturation (Figure 5, curve B). However, the useof this functionin simulationsof unsatur- ated water transportleads to a valid descriptionof the hydraulicsystemas longas saturationis notapproached [WangandNarashimhan,1985].The double-porositychar- 0 1000.6 0.4 0.2 pF pF 2 4 6 0 2 4 6.... i i 1 '"•,'....-.'[•-unimodal-",, • trimodal :•-'.,-.,.c:-'x 15 0.15 0.0015 0 0.2 0.4 0.6 poreradius(I.u'n) watercontent Figure6. Hydraulicpropertiesof SolarVillageClay [Hampton,1989].(Left) Measuredretentiondata(solid circles),fittedbimodal(solidline),andunimodal(dashed line)retentioncurve,andcorrespondingbimodal(black shaded)andunimodal(lightshaded)pore-sizedistributions. (Right)Predictedbimodal(solidlines)andunimodal(dashed lines)conductivitycurves.TheK(½)curves,fromtopleftto bottomright,arerelatedto thetopaxis.TheK(0) curves, fromtopfighttobottomleft,arerelatedtothebottomaxis. The coefficientsof thefunctionsarelistedinTable1. acteristic of the conductivity function cannot be reflectedby the unimodal estimation, even if the tortuosity factor• is manipulated(Figure 5, curve D). In Figures6, 7, and 8, measuredretention data of three soils are depicted, representing structured pore systems which can be frequently found. The chosen examplesare similar to the schematic soil types described in Figures2 (bottomleft), 2 (top left), and 2 (top right), respectively.In additionto the experimentaldata, Figures6-8 showtheleast squaresfittedunimodalandmultimodalretentioncurves,the underlying pore-size distributions, and the estimatedcon- ductivity curves. Some physical propertiesof the soilsand the coefficients of the fitted functions are listed in Table 1. Unlike the unimodal function, the multimodal functionde- scribesthemeasureddataof eachsoilaccurately,anditis obvious that in each case the two functions representfun- damentally different pore systems. Figure6 (left)depictsthehydraulicpropertiesofacolumn packedwith smallclay aggregates[Hampton,1989].The pore-sizedistributionisdistinctlybimodal,witha maximum poredensityoftheintra-aggregateporesatr ---0.01/xm,and a secondarymaximumfortheinteraggregateporesatr --20 /xm.Whereastheprimaryporesystemisdeterminedbythe particle-sizedistributionofthatsoil,theporespacebetween theaggregatesdependsonlyonthepackingandistherefore essentiallyindependentof the soiltexture.Notethatthe existenceandextentof secondaryporesystemsinundb turbedsoilsposesa limitationto thesuccesspotential0f regressionmethodsthatseekto predicthydraulicrelation- shipssolelyfromparticle-sizedistributions.Figure7(left} showsdataof anundisturbedcoreof Rideauclayloam[D} pF 0 2 4 0.5 i i i i 0.4 0.:3 0 15 1.5 0.015 poreradius(llm) pF 6 0 2 4 610 10•li10• 0 0.1 0.2 0.3 0.4 water content Figure8. HydraulicpropertiesofROnhaveSandyL0an• [SchjOnning,1985].SeeFigure6 for explanations.
  • 7. DURNER: HYDRAULIC CONDUCTIVITY ESTIMATION FOR SOILS 217 Table 1. Parametersof Example Soils Classification Rideau R0nhave Solar Vii!age Clay Sandy Clay* Loam? Loams AggregatedLoamõ Texture Type -3 Density,gcm Porosity,% Samplevolume,cm3 log(Ks),ms-• Unimodalvan Genuchten Os,cm3cm-3 Or,cm3cm-3-1 a, cm m clay loam heavy clay coarsesandy aggregates loam repacked undisturbed undisturbed 1.3 1.55 52.7 41.3 12.3 345 100/250 -5.7 -4.6 -5.3 0.556 0.439 0.403 0.000 0.000 0.000 0.0328 0.0027 0.0212 1.160 1.134 1.181 0.138 0.!18 0.153 loam aggregates repacked 0.98 0.63 750-1500 -4.4 Bimodal Bimodal Multimodal van Genuchten Trimodal Bimodalll Bimodalô Bimodal** Os,cm3cm-3 0.556 0.473 Or' cm3cm-3 0.044 0.132 wl 0.54 0.14 -• 0.0252 0.6634 al, cm nl 2.482 2.060 rnl 0.243 0.515 w2 0.46 0.86 - • O.00006 O.O011Ot2, cm n2 2.272 1.286 rn2 0.825 0.223 W3 -1 Ot3, ½m rt3 m2 0.403 0.633 0.633 0.633 0.000 0.160 0.160 0.154 0.21 0.68 0.65 0.62 0.715 0.4422 0.3320 0.2799 1.349 4.675 3.123 2.849 0.210 0.304 0.680 1.0130 0.16 0.32 0.35 0.38 0.0093 0.0119 0.0112 0.0084 1.550 3.471 2.527 1.741 0.355 0.422 0.604 1.000 0.62 0.00017 1.818 0.450 *Source:D. R. Hampton(personalcommunication,1989). ?Source:DeJonget al. [1992]. $Source:SchjOnning[1985]. õSource:Smettemand Kirby [1990]. IlSoconstraint. ôHere,rni = 1 -- 1/ni. **Here mi = 1. Jonget al., 1992].Thepore-sizedistributionis againbi- modal,but the secondarypore systemis now lesspro- nouncedandshiftedtowardlargerporesizes,ascompared tothepreviousqxample.Forthissoilwecandeducethe existence,but notthe actualshapeof the secondarypore system,becausethe measurementsare insufficientin the importantrangenearsaturation.Forthepurposeofanalyz- ingbimodalconductivityestimationsof suchporesystems weassumeherethat the fittedbimodalfunctionis a valid modelofthesoil'strueretentionpropertiesevenintherange whereno data are available. As a third example, the reten- tionpropertiesofanundisturbedsampleofR0nhavesandy 10am[SchjCnning,1985]aredepictedinFigure8(left).Ina strictsense,thisporesystemisunimodal.Nevertheless,it cannotbeadequatelydescribedbytheclassicalsigmoidal retentioncurves.Thecontinuous,almostlineardecreaseof thewatercontentfrompFOtopF3indicatesaconsiderable poredensityovera widerangeofporesizes,whereasthe mainporesystemislocatedin therangeof finepores. Acomparisonoftheconductivityestimatesfromtheuni- andthemultimodalretentionfunctions(Figures6(right)to8 (right))showsthefollowing. 1. Conductivityestimatesfromuni-andmultimodalde- scriptionsoftheretentioncurvecandifferbyordersof magnitude.ThesedifferencesarevisibleinboththeK(½) and the K(0) plane. 2. Thelargestdifferencesoccurin thewet moisture range,whichis veryimportantfor solutetransportpro- cesses. 3. ThedifferencesintheestimatedK(½)curvesarisein mostcasesfromdifferentslopesnearsaturation.Asthesoil dries,thecurvesremainapproximatelyparallel(Figure7)or approacheachotheragain(Figures6and8),providedthe sameresidualwatercontentisapproachedbytheretention functions. Simplifiedsigmoidalretentionfunctionscannotreflectthe smallbutimportantchangesinwatercontentthattakeplace inporerangesdistantfromthemainporesystem.The limitingcaseforthisinabilityisretentionfunctionsofthe typeof(2),wherebothwatercontentandconductivityare
  • 8. 218 DURNER: HYDRAULIC CONDUCTIVITY ESTIMATION FOR SOILS constant between saturation and the air entry point. Func- tions of thistype representdiscontinuouspore-sizedistribu- tions and are by definition not adequatefor conductivity estimations in loamy and clayey soils. Whether the conductivity estimation by a nonoptimal retention function will tend to lead to an overestimation or an underestimationcan be simply evaluatedby comparing the slope of the fitted retention curve with the slopeof the measured data. If the fitted retention curve underestimates the true water release near saturation (as in Figures 2 (top left), 2 (top right), 2 (bottomright), 7 (left) and8 (left)), the predictedchangeofK(½) will aswell be smallerthanthetrue change.Consequently,if matchedto Ks, thepredictedcurve results in an overestimation. In studies where conductivity predictionshave been comparedwith measurements,this type of predictionerrorhas,indeed,frequentlybeenfound [e.g., Parkesand Waters, 1980].If, on the otherhand,the fitted retention curve indicatesa steeperslope than the measuredwater content data (as in the wet rangeof Figures 2 (bottomleft) and6 (left)),thepredictedchangeofK willbe larger than the change that were predictedby a more accurateretentionfunction. Accordingly,thetwo estimated K(•) curvesof the morainicsoil(Figure8) approacheach otheragainatdrierconditions,sincetheunimodal0(½)curve hasa steeperslopethanthemultimodaloneat intermediate pore sizes. Usingconductivityestimationsfrommultimodalretention curvesas an investigationtool, someof the classicalprob- lems of the predictionmethodswill be discussedin the followingsections,namely(1) theroleof theparameterOr, (2)thequestionwhethertheparticularmathematicalformof the retention function is of relevance, and (3) the sensitivity of conductivitypredictionson the slopeof the retention curve near saturation, which will be found to be most important. Role of 0r and Os Throughoutthehistoryof conductivityestimations,the searchfor a "correct" residualwater content0r hasbeena continuoussourceof vexation.Statementsonthe roleof Or for conductivitypredictionsrangefrom"veryimportant" [Sidiropoulosand Yannopoulos,1988]to "unimportant" [VogelandCislerova,1988].Partly,thismaybeduetothe factthatthereis noagreementonthedefinitionand,conse- quently,thecorrectdeterminationofOr[Sidiroupoulosand Yannopoulos,1988;Lucknetet al., 1989;Nimmo,1991; NielsenandLuckner,1992].Mualem[1976],whoconsidered the correctvalueof the residualwatercontentto be impor- tantfortheprediction,devoteda considerablepartofhis classicalpaperto thedeterminationof Or-VanGenuchten [1980]proposedtosetOrequaltoameasuredwatercontent at a highsuction.Recentreviews[vanGenuchtenand Nielsen,1985;NielsenandLuckher,1992]recommendtreat- ing Orasa purefittingparameter. From(6)we canseethatin principletheconductivity predictioninthewetrangeisnotaffectedbythevalueofOr, sincethehydraulicconductivityofa soilisdeterminedat anysaturationbyonlyafewpercentofthelargestwater- filledpores.Durner[1992]confirmedthisindependency usingwaterretentiondataofa loamfromRoskilde,Den- mark(similartoFigure8).Contrarytoaunimodalmodel,a changeof Orfrom0.0to 0.19hadnoinfluenceonthe conductivitypredictionintherangeoflargepores,sinceit didnotaltertheshapeofa fittedbimodalretentioncurvein thatrange.Anyinfluenceof Orinunimodalhydraulicmodels is thereforecausedby interferencesof the curveshape parameters.For the vanGenuchtenretentionmodel,asan example,a decreaseof thevalueof Orcausesa shiftofthe parametern toward a smallervalue. A similareffectis causedby an increaseof the parameterOs.Thereforeifthe interest lies purely in the wet moisture range, we can disregarddatapointsin thedry rangeif thisleadstoabetter descriptionin the rangeof interest. Effectsof parametercorrelationscanbe shownbydataof Kablan et al. [1989], who comparedconductivitymeasure. mentsfor a fine sandwith conductivity estimatesbythevan Genuchten-Mualem model. In the fitting procedureforthe retentioncurve,the parameterOswas setequalto thelargest singlemeasuredwater content.Sincefield datawitha considerablescatter were used, this value was far higher thanan optimallyfitted value.As a consequence,the0pti. mizedvaluesof Orand n bothbecametoo small,andthe predictedconductivityfunctioncorrespondinglyunderesti. mated the decrease in conductivity at low saturation,thus leadingto the conclusionthat the predictivemodelworked well at highsaturations,but failed at low saturations.The exampleexhibitsa dilemmain fittingproceduresforthe retention curve. To get a best overall fit, all parameters shouldbe simultaneouslyoptimized [compareNielsenand Luckner,1992].An independentdetermination,asisdesir- ablefor parametersof physicalsignificance,e.g., intrans- portmodels,ishereinappropriate.Ontheotherhand,ifan optimizedvalueof Osislowerthatactuallyobservedvalues, informationrelevantfor K(0) at largeporesmaybemasked bytheoverallfittingprocedure[Mualem,1992]. Mathematical Formulation of the Retention Function Asisexpressedby(6),theconductivitypredictionmodels estimatetherelativeconductivityfunctionpurelyfromthe shapeoftheretentionfunction.Thuswecanpostulatethat thespecificmathematicalformofa retentionmodelhasno influenceon the conductivitypredictionas longasit pro- videsaproperdescriptionofthemeasureddata.Weverify thisbydescribingastronglyaggregatedloamwithabimodal poresystem(databySmettemandKirkby[1990])usingthree differentretentionmodels,basedonthebimodalformof(4): (1)theunconstrainedmodelwithvaryingmiandhi,(2)the constrainedmodelwithmi" 1- l/hi, and(3)asimplified formwithmi = 1.Forcomparison,wealso•interpolatethe retentiondatalinearly,withextrapolationsin thedryrange to watercontentzeroat pF7. To definethepore-size distributiontowardthelargepores,themeasuredvalueof0 atpFOistakenasOs.Figure9showstheretentioncharac- teristic,theunderlyingpore-sizedistribution,andthepre- dictedK(½)curve,asrepresentedbythislinearinterpola- tion.Thepore-sizedistributionhasaveryruggedshapeand isnoncontinuous.Suchadescriptionofthehydraulicprop- ertieswouldcauseproblemsin numericalsimulations [Hampton, 1990]. Acomparisonofallfourretentionmodelsandtherelated estimatedconductivityfunctionsisshowninFigure10.The agreementamongthefourmodelsaswellastheagreement betweentheestimatedandmeasuredconductivitiesisal- mostperfect(Figure10(right)).We concludethatany smoothfunctionwhich describesthe retentiondataaccu- ratelyappearstobesuitableinthiscase,regardlessofits
  • 9. DURNER:HYDRAULICCONDUCTIVITYESTIMATIONFORSOILS 219 0 o.4 0.2 pF ,, i i i i 15oo 150 15 3 poreradius !.5 0 I 2 pF 1 1oø •_ 10'2'• > I0"•• 0,2 Figure9. Hydraulic properties of AggregatedLoam [SmettemandKirkby, 1990].(Left) Experimentalretention data(solid circles), linear interpolation (solid line), and pore-sizedistribution(shadedarea),asrepresentedby the linearinterpolated retention curve. (Right) Estimated con- ductivityfunction. 0 ! 2 8 4 5 0 5 pF I 2 3 4 pF 10ø 10•-• 10.48 10.• '• Figure11. Dependenceof conductivityestimationsonthe asymptotic slope of the retention curve near saturation. (Left) Retentioncurves-dottedline, Brooksand Corey function(2) (a-t = 10 cm, A = 0.15);solidline, van Genuchtenfunction(3),withconstraintrn= 1- 1/n(a- t = 10.5cm, n = 1.15, nm = 0.15); dashedline, vanGenuchten functionwithoutconstraint(a-i = 10.5cm,n = 1.0, nrn= 0.15). (Right)Predictedconductivitycurves. mathematicaltype. As will be shownnext, thisis, however, onlytrueaslongasthe functionsdo really correspond,and donotdifferin their asymptoticbehaviortoward saturation. Slopeof the RetentionCurve Near Saturation Theasymptotic slope of the retention function near satu- rationis the most sensitive determinant for conductivity estimationsif smoothretention functions (i.e., functions that representcontinuouspore-size distributions) are used. This originatesfrom the use of the Hagen-Poiseuillelaw in the predictionmodels,which statesthatthe flux densityin pores ofacertainsizeis inverselyproportionalto thesquareof the poreradius.Consequently,all statisticalconductivitypre- dictionmethodsintegrateover the inverseof • [Mualern, 1992].The smallerthe suctioniswhere changesin the water contenttake place, the greater is its contribution to the conductivityintegral(e.g., (7)). If retentionfunctionswith a distinctair entrypointare usedfor the conductivityestima- tion[Brooksand Corey, 1964;Campbell, 1974],the inte- pF pF o 3 o 1 2ß 0.6' 1 2 i i' i i , 1500 150 15 porerad:us(!ran) ß5 0 0.2 0.4 0,6 water content 10ø 10'2 10'4 Figure10. Conductivityestimationfor AggregatedLoam [SmettemandKirkby,1990]fromfourmathematicallydif- ferentretentionfunctions.(Left)Fittedretentionmodels: linearinterpolated(solidline),bimodalretentionfunction (equation(4))withrni allowedtovary(dottedline),mi = 1 -' 1/ni(dashedline),andmi = 1(dashed-dottedline).See Table1 forcoefficientsof themodels.(Right)Estimated conductivityfunctionsandmeasuredconductivitydata.The fourretentionmodelsyieldalmostidenticalestimates.The agreementbetween estimates and measurements is excel- lent. grandin (7), 1/½dS•, is bounded,sincedS[, = 0 between saturationandtheair entrypressure.Thisisdifferentfor any retentionfunctionwhich approachessaturationasymptoti- cally. Now the valueof the integranddependson how fast dS•, approacheszero as 1/0 approachesinfinity, i.e., on the slopeof theretentionfunction.The absolutesensitivityof K to Se, definedby Sabs= dK/dSe, becomesinfinitenear saturation.For the closed-formconductivityfunction of the van Genuchten-Mualem model, g = - - s/m)m]2 (9) this absolutesensitivityis givenby 2Se ' Sabs-'rSe ~q- l/m)1-m' (10)(1 --Se whereg(Se) = [1 - (1 - S•I/")m].Whentheeffective saturationapproachesunity, the first right-handterm andthe numeratorof the secondterm in (10) approachfinite values, whereas the denominator of the second term becomes zero, becausethe value of the parameter m is boundedby 0 < m < 1. This causes[$abs[to approachinfinity.Due to this sensitivity characteristic,even small changesin water con- tent near saturation(indicatingthe existence of large pores) are very importantfor the shapeof the estimatedconductiv- ity function. Measurementsclose to Osmust therefore be taken at small pressureintervals and as accurately as possi- ble. Reliable conductivity estimations become impossible if measurements near saturation are missing, regardless whether the pore systemis heterogeneousor not. To illus- trate this, we have fitted three different unimodal functions to hypotheticalretentiondata(Figure 11 (left)). The reten- tion models are the Brooks and Corey function, the van Genuchten function with m = 1 - l/n, and the van Genuchten function with n = 1.0. Since there are no measureddata between saturationand Se = 0.8 at pF2, eachfunctiondescribesthe dataequally well. However, the correspondingpore-sizedistributions,whichareidenticalin the fine-porerange, are very differentin their asymptotic behaviorin the large-porerange.As was discussedabove, the Brooks and Corey function(dashedline) restrictsthe
  • 10. 220 DURNER:HYDRAULICCONDUCTIVITYESTIMATIONFORSOILS sizeof thelargestporesconsideredby theconductivity predictionmodelto theequivalentradiusoftheairentry pressure(r --- 0.15 mm in the givenexample).The two smoothfunctionsrepresentacertainamountofporesinthe large-porerangeandcanevenconsiderporesizesthatare physicallyabsurd.Theconductivityestimatesbasedonthe threefunctionsareshiftedagainsteachotherby ordersof magnitude, and without additional measurementsbetween saturationandpF2 thereis noway to decidewhichmodel represents the soil best. The accuracyof watercontentmeasurementsisgenerally limited by the error of the measurementmethod. Unfortu- nately,it becomesexperimentallymoreandmoredifficultto measuretheretentioncharacteristicof a soilaccuratelyand reproduciblyat very low suctions.It isfurtherquestionable whetherretentioncurves,indeed,representthe pore-size distributionof a porousmediumat high saturation,where the nonwettingphaseis notcontinuous[Corey,1992].This meansthat the conductivityestimationbecomesby defini- tion inaccuratewhen approachingsaturation,even if we assumethat the predictive model itself is free of error and its applicability does not depend on the pore sizes. As a consequence, conductivity estimationsfrom retention func- tionsthat representa certainporespaceintherangeof very smallsuctions(say, belowpFO) are unreliable,regardlessof whether unimodal or multimodal functions are used. This is, for example, generally the case if the parametern in the constrainedvan Genuchten equation(m = 1 - l/n) is close to unity. ComparisonsBetweenMeasured and Estimated Conductivities To assess whether conductivity estimations from more accurate retention functions are of advantage in practice, comparisonsof unimodal and multimodal predictions with experimental data are necessary.To be of statisticalsignif- icance, suchcomparisonsmustrest on a broaddatabaseof well-documented measurements of high quality. Existing data have frequently been measuredon homogenizedand packedsoilsamples,anda varietyof measurementmethods have been used. To date, most of the publishedempirical validation studies on conductivity estimations used rela- tively small databasesof the authors' own measurements. This mightbe onereasonwhy manyof thesestudiesresulted in a new proposed"optimal" value of the tortuosityor pore-interactionfactorr [e.g.,Kunzeetal., 1968;Greenand Corey, 1971;Jackson, 1972; Ghosh, 1977;Ragab et al., 1981;Talsma, 1985;Alexanderand Skaggs, 1986]. In caseswhere measureddata appear basically reliable, often the retention measurements near saturation are insuf- ficient. Then, the hydrologicallyimportantsecondarypore systemis not strictlydetermined,aswasdiscussedabove. Therefore estimationsby a flexible retention function are uncertain, and a better or worse agreementwith measure- mentsmightbe achievedat random.Durner[1992]showed for Beit Netofa clay (whichservedvan Genuchten[1980]as an examplefor possiblefailuresof the van Genuchten- Mualem model) that a conductivityestimationbased on a bimodal retentionfunction led to a very goodagreementwith measureddata. However, since for this soil no retention meaœurementsare available in the pressurerangebetween saturationandpF2, the saturatedwatercontentwasarbi- trarilyset2%largerthanthelargestmeasuredwatercontent. A differentvaluefor 0• mighthavechangedtheestimation considerably,leadingtoa badagreement.Note,however, thatit wasnotpossibleto cometo a satisfactoryagreement betweenmeasuredandestimatedconductivitiesusingthe unimodal van Genuchten-Mualem model. Sinceparticularlyforheterogeneoussoilswithsecondary poresystemsthe problemof measurementsnearsaturation is furthercomplicatedby macropores,air encapsulations, and hysteresiseffects,we do not attempt to compareinthis paper unimodaland multimodalconductivitypredictions with measurements.It appearsdoubtfulwhethera validation attemptin this classicalway is viable with presentlyavail- able data. Conclusions The existenceof a secondarypore systemgreatly affects the shapeof the estimatedhydraulic conductivitycurvein the wet range. Estimation differencesdue to differentreten- tion modelscan be larger than differencesdue to the useof alternative prediction models [Childs and Coilis-George, 1950; Burdine, 1953; Mualem, 1976]. Consequently,the ignoranceof a secondarypore system may lead to large errors. We have shown for soils with heterogeneouspore systemsthat the commonlyappliedunimodalretentionfunc- tions are inadequate for conductivity estimations, because they lack the necessaryflexibility in the rangeof largepores. Their use may lead to wrong results in flow and transport simulationsor in parameterestimationprocedureswherethe inverseproblemis solved[Durner, 1991].Validation studies of hydraulic conductivity prediction methods shouldbe reevaluatedif the appliedretention modeldid not accurately describe the measured data. Particularly for loamy and clayey soilsthe reasonsfor the frequent failure of conduc- tivity prediction methods should be reconsidered.In some casesit may not be the failure of the predictive model,but the use of inadequate retention functions which leadstoa discrepancybetweenestimatesandmeasurements.Forsoils with very wide pore-sizedistributions,unimodalretention modelsare sometimesforcedto representa nonnegligible part of the pore spacein the range of unrealisticallylarge poresizes.In suchcases(indicated,e.g., by thevanGenu- chtenparametern beingcloseto unity) the conductivity predictionnear saturationis by definitionunreliable. The almostperfectagreementbetweenmeasuredand estimatedconductivitiesfor theloamysoilof Smettemand Kirkby[1990,Figure10b]indicatesthatconductivitypredic- tionmethodsbasingonstatisticalpore-domainmodelshave a high successpotential not only for soils with narrow pore-sizedistributions,but as well for soilswith heteroge- neousporesystems.As comparedto unimodalretention functions,theuseofthemultimodalretentionfunction(orof anyotherflexibleretentionfunction)leadsto morereliable conductivityestimatesifit improvesthefitofthemeasured retentiondataandif theshapeofthefittedretentionfunction in the rangeoflargeporesis strictlydeterminedby measure- ments.Particularlyfor validationsof flow and transport modelsonlaboratoryscale,thehydrauliccharacteristicof thesoilsampleshouldbedescribedasaccuratelyaspossi- ble. Theslopeofthewaterretentioncurvenearsaturationis themostimportantpropertyindeterminingtheshapeofthe estimatedconductivityfunction.Evensmalluncertaintiesin
  • 11. DURNER: HYDRAULIC CONDUCTIVITY ESTIMATION FORSOILS 221 measuredwatercontentsnear saturationlead to largeesti- mationuncertainties.Whena retentioncharacteristichasno distinctairentrypoint,the measuredsaturatedconductivity isgenerallynot an appropriatematchingvaluefor the predictedconductivityfunction.Thisconclusion,whichwas alreadystatedfor unimodalporesystems[e.g., van Genu- chtenandNielsen, 1985;Lucknet et al., 1989;Nielsenand Lucknet,1992],is even more importantfor heterogeneous poresystems.Theuseofmultimodalfunctionscanhelpto explainthisproblem;however,it cannotsolveit. Wemayexpectthat the basicconceptsthat underliethe statisticalconductivitypredictionmodelsgenerallyholdless forsecondarypore systems as compared to textural ones. Theshapeandorganizationof voidsin aggregatedsoilscan beverydifferent,dependingon soilgenesisand aggregation state[WangandNarasimhan, 1990].As an example,a net of planarvoidswill behavehydrologicallyquite differently fromthe pore spacebetween sphericalaggregates,or the porespaceduetomicrorootchannels.Nevertheless,eachof theseporesystemsmay have a similarly shapedretention characteristic.Therefore we shouldregard •-in the correc- tionfunction SJ as a purely empirical fitting parameter [Roulieret al., 1972; Russo, 1988]. For use in parameter estimationprocedures, Nielsen and Lucknet [1992] even suggestto decouple the retention and the conductivity function,and to use the closed-form conductivity equation of the van Genuchten-Mualem model (9) as an empirical expressionfor K(Se). For soils with heterogeneouspore systems,we expectthe bestresultsby stillusingthe shapeof theestimatedconductivity function while treating the two parametersKs and ß as unknowns. By this strategy the informationon the pore system containedin the retention functionwill be usedto fullest advantage.Close to saturation theconductivitiesof soilscontaininglarge pores shouldbe directlymeasured.This means, however, that the attractive- nessofthe predictionmethodsis considerablyreduced. Propertiesof heterogeneouspore systemscannotbe iden- tifiedbyparameterestimationtechniqueswheretheinverse problemis solved,if a priori unimodalhydraulicproperties areassumed.Due to the ill-posednessof the mathematical problem,it appearsontheotherhandimpossibleto usemore flexiblehydraulicfunctionsand to identify the increased numberof coefficientsby the inversemethod.Thereforethe retentioncurvesof soilswith heterogeneouspore systems shouldbedirectlyfittedto measureddata.Preferably,these measurementsare obtainedfrom a transient flow experiment bysimultaneouslymeasuring½and0inhighresolution[e.g., Sobczuketal., 1992].Theflowcharacteristicsof theexper- imentcanthenbeusedto optimizethetwoconductivity- relatedunknowns,Ksandr, bysolvingtheinverseproblem. Finally,wewanttorecallthatthereliabilityofconductiv- itypredictionmethodsisgenerallyrestrictedbysomefun- damentalproblems,e.g.,theassumptionof isotropicorga- nizationofporesindependentoftheirsize,thevalidityofthe Hagen-Poiseuillelawforflowinlargepores,theproblemof airentrapment,andtheproblemofhysteresis.Thespecific findingsinthispaperforheterogeneousporesystemsare neverthelessapplicable.Theresultsarefurtherindependent ofthechoiceoftheparticularstatisticalconductivitypredic- tionmodel.Theextensionto thefieldscalerequiresthe considerationofspatialandtemporalvariabilityaswellas theconsiderationofscaleproblemsanderrorsinducedby themeasurementmethod.Thefurtherdevelopmentofsuit- ablemeasurementmethodsto identifyheterogeneouspore systemsappearsthereforeto be of primary importance. Acknowledgments.The authorthanksDuaneHampton,Kalama- zoo,Michigan,OleJacobsen,Tinglev,Denmark,andRichardDe JongandG. ClarkeTopp,Ottawa,Ontario,Canada,whoprovided thesoilphysicaldataofsomeoftheexamplesoilsinthispaper.The authorgratefullyacknowledgesthegrantof a long-termfellowship oftheEuropeanEnvironmentalResearchOrganization(EERO). References Addiscott,T. M., andR. J.Wagenet,Conceptsof soluteleachingin soils:A reviewofmodellingapproaches,J. SoilSci.,36,411-424, 1985. Alexander,L. R., andR. W. Skaggs,Predictingunsaturatedhydrau- lic conductivityfrom the soil water characteristic, D'ans. Am. Soc. Agric. Eng., 29, 176-184, 1986. Anderson,S. H., andD. K. Cassel,Statisticaland autoregressive analysisof soil physicalpropertiesof Portsmouthsandyloam, Soil Sci. Soc. Am. J., 50, 1096-1104, 1986. Barraclough,D., A usablemechanisticmodelof nitrate leaching,I, The model, J. Soil Sci., 40, 543-554, 1989. Bear,J., andY. Bachmat,Introductionto TransportPhenomenain PorousMedia, Kluwer Academic,Hingham, Mass., 1991. Bloemen, G. W., Calculation of hydraulic conductivitiesof soils from texture and organic matter content, Z. Pfianzenerniihr. Bodenkd., !43,581--605, 1980. Boels, D., J. B. H. M. Van Gils, G. J. Veerrnan, and K. E. Wit, Theory and system of automatic determination of soil moisture characteristics and unsaturated hydraulic conductivities, Soil Sci., 126, 191-199, 1978. Brooks, R. H., and A. T. Corey, Hydraulic propertiesof porous media,Hydrol. Pap., 3, pp. 1-27, Colo. State Univ., Fort Collins, 1964. Bruce, R. R., and R. J. Luxmoore, Water retention: Field methods, in Methods of Soil Analysis, part 1, Physical and Mineralogical Methods, 2nd ed., edited by A. Klute, pp. 799-824, ASA and SSSA, Madison, Wisc., t986. Brutsaert, W., Probabilitylawsfor pore size distributions,Soil Sci., IOI, 85-92, 1966. Brutsaert, W., The permeabilityof a porous mediumdetermined from certain probability laws for pore-size distribution, Water Resour. Res., 4, 425-434, 1968. Burdine, N. T., Relative permeability calculationsfrom pore size distribution data, Trans. Am. Inst. Min. Metall. Pet. Eng., 198, 71-78, I953. Campbell, G. S., A simple methodfor determiningunsaturated conductivityfrom moistureretentiondata, Soil Sci., I 17, 311-314, 1974. Carsel,R. F., L. A. Mulkey, M. N. Lorber, and L. B. Baskin,The pesticideroot zonemodel:A procedurefor evaluatingpesticide leachingthreatsto groundwater,Ecol. Modell., 30, 49-69, 1985. Chen,C., andR. J. Wagenet,Simulationof water andchemicalsin macroporesoils,I, Representationof the equivalentmacropore influence and its effect on soil-water flow, J. Hydrol., 130, 105-126, 1991. Chen, C., D. M. Thomas, R. E. Green, and R. J. Wagenet, Two-domain estimation of hydraulic properties in macropore soils, Soil Sci. Soc. Am. J., 57, 680-686, 1993. Childs, E. C., and N. Coilis-George,The permeabilityof porous materials,Proc. R. Soc. London, Ser. A, 201,392-405, 1950. Corey,A. T., Pore-sizedistribution,in Proceedingsof theInterna- tional Workshop,IndirectMethodsfor EstimatingtheHydraulic Propertiesof UnsaturatedSoils,editedbyM. T. vanGenuchten, F. J. Leij, andL. J. Lund, pp. 37-44, Universityof California, Riverside, 1992. Currie,J. A., andD. A. Rose,Gasdiffusionin structuredmaterials: The effect of tri-modal pore-size distribution, J. Soil Sci., 36, 487-493, 1985. DeJong,R., G. C. Topp,andW. D. Reynolds,Theuseofmeasured andestimatedhydraulicpropertiesin the simulationof soilwater contentprofiles:A casestudy,inProceedingsof theInternational Workshop,IndirectMethodsfor EstimatingtheHydraulicProp- ertiesof UnsaturatedSoils, pp. 569-584, editedby M. T. van
  • 12. 222 DURNER:HYDRAULICCONDUCTIVITYESTIMATIONFORSOILS Genuchten,F.J.Leij,andL.J.Lund,UniversityofCalifornia,Riverside, 1992. D'Hollander,E.H., Estimationoftheporesizedistributionfrom themoisturecharacteristic,WaterResour.Res.15 107-112,I979. ' ' Dixon,R. M., Controllinginfiltrationinbimodalporoussolids: Air-earthinterfaceconcepts,FundamentalsofTransportPhe- nomenainPorousMedia,vol.2, pp. 107-117,Universityof Guelph,Ontario,1972. Durner,W.,VorhersagederhydraulischenLeiffiihigkeitstrukturi- erterB6den,BayreutherBodenkundlicheBerichte,20, 1-180,1991. Durner,W.,Predictingtheunsaturatedhydraulicconductivityusing multi-porositywaterretentioncurves,inProceedingsof the InternationalWorkshop,IndirectMethodsforEstimatingthe HydraulicPropertiesof UnsaturatedSoils,editedbyM. T. van Genuchten,F.J.Leij,andL.J.Lund,pp.185-202,Universityof California,Riverside,1992. Farrel,D. A., andW. E. Larson,Modelingtheporestructureof porousmedia,WaterResour.Res.,8, 699-706,1972. Gardner,W.R.,Thepermeabilityproblem,Soil$ci.,117,243-249, 1974. Gerke,H.H.,andM.T. vanGenuchten,Adual-porositymodelfor simulatingthe preferentialmovementof waterandsolutesin structuredporousmedia,WaterResour.Res.,29,305-319,1993. Ghosh,R.K., Determinationofhydraulicconductivityfrommois- tureretentiondata,SoilSci., 124,122-124,1977. Gillham,R. W., Hydraulicpropertiesofa porousmedium:Mea- surementandempiricalrepresentation,SoilSci.Soc.Am.J.,40, 203-207, 1976. Green,R.E., andJ. C. Corey,Calculationofhydraulicconductiv- ity: A furtherevaluationof somepredictivemethods,SoilSoc. Am. Proc., 35, 3-8, 1971. Hampton, D. R., Coupled heat and fluid flow in saturated- unsaturatedcompressibleporousmedia,Ph.D. thesis,Colo.State Univ., Fort Collins, 1989. Hampton,D. R., A methodtofit soilhydrauliccurvesin modelsof flow in unsaturatedsoils,in ComputationalMethodsin Subsur- face Hydrology,editedby G. Gambolatiet al., Computational MechanicsPublishers,Boston,Mass., 1990. Hantschel,R., W. Durner,and R. Horn, Die Bedeutungyon Porenheterogenit•itenffr dieErstellungderpF-WG-undK-0- Kurven, Desch.BodenkdI.Mitt., 53, 403-409,1987. Haverkamp,R., M. Vauclin,J. Touma,P. J. Wierenga,andG. Vachaud, A comparison of numerical simulationmodels for one-dimensionalinfiltration,SoilSci. Soc.Am. J., 41,285-294, 1977. Hornung,U., Identificationof nonlinearsoilphysicalparameters from an input-outputexperiment,in Workshopon Numerical Treatmentsof Inverse Problemsin Differentialand Integral Equations,editedby P. Deyuflhardand E. Haier, pp. 227-237, Birkhauser, Boston, Mass., 1983. Jackson,R. D., On the calculationof hydraulicconductivity,Soil $ci. Soc. Am. Proc., 36, 380-382, 1972. Jacobsen,O. H., Unsaturatedhydraulic conductivityfor some Danishsoils,Methodsandcharacterizationof soils(in Danish), Tidsskr.Planteavl, 93(S2030), 1-60, 1989. Jarvis,N.J., P.-E. Jansson,P. E. Dik, andI. Messing,Modelling water andsolutetransportinmacroporoussoil,I, Modeldescrip- tion and sensitivity analysis,J. Soil Sci., 42, 59-70, 1991. Jeppson,R. W., W. J. RawIs, W. R. Hamon, andD. L. Schreiber, Use of axisymmetricinfiltration modeland fielddatato determine hydraulic propertiesof soils, Water Resour. Res., I1, 127-138, 1975. Kablan, R. A. T., R. S. Mansell, S. A. Bloom,andL. C. Hammond, Determinationsof unsaturatedhydraulicconductivityfor Candler Sand, Soil Sci., 148, 155-164, 1989. King, L. G., Descriptionof soil characteristicsfor partially satu- rated flow, Soil Sci. $oc. Proc., 359-362, 1965. Kool, J. B., and J. C. Parker, Development and evaluation of closed-formexpressionsfor hysteretic soil hydraulicproperties, Water Resour. Res., 23, 105-114, 1987a. Kool, J. B., and J. C. Parker, Estimating soil hydraulicproperties from transient flow experiments: SFIT User's Guide, Electric Power Research Institute, Palo Alto, Calif., 1987b. Kool, 5•. B., and J. C. Parker, Analysis of the inverse problem for transientunsaturatedflow,WaterResour.Res.,24,817--830,1988. Kool,J.B.,J.C.Parker,andM. T. vanGenuchten,Parameter estimationforunsaturatedflowandtransportmodels:Areview,J. Hyclrol., 91,255-294, 1987. Kunze,R.J.,G.Uehara,andK.Graham,Factorsimportantinthe calculationofhydraulicconductivity,SoilSci.Soc.Am.Proc.,32,760-765, 1968. Lafolie,F.,R.Guennelon,andM. T. vanGenuchten,Analysisof waterflowundertrickleirrigation,II, Experimentalevaluation, Soil Sci. Soc. Am. J., 53, 1318-1323, 1989. Laliberte,G.E.,Amathematicalfunctionfordescribingcapill• pressure--desaturationdata,Bull.Int. Ass.Sci.HydroI.,I4, 131-149, 1969. Loague,K., andR.E. Green,Statisticalandgraphicalmethodsfor evaluatingsolutetransportmodels,Overviewandapplications,J. Contam. Hydrol., 7, 51-74, 1991. Luckner,L., M.T. vanGenuchten,andD.R.Nielsen,Aconsistent setof parametricmodelsfor thetwo-phaseflowof immiscible fluidsinthesubsurface,WaterResour.Res.,25,2187-2193,1989. Marshall,T. J., A relationbetweenpermeabilityandsizedistribu. tion of pores,J. Soil Sci., 9, I-8, 1958. Michiels,P.,R. Hartmann,andE. DeStrooper,Comparisonsoflhe unsaturatedhydraulicconductivityof a coarse-texturedsoilas determinedin thefield,in thelaboratory,andwithmathematical models,Soil Sci., 147, 299-304, 1989. Millington,R. J., andJ.P. Quirk,Permeabilityof poroussolids, Trans. Faraday Soc., 57, 1200-1207, 1961. Mualem,Y., Anewmodelforpredictingthehydraulicconductivity of unsaturatedporousmedia,WaterResour.Res.,12,5!3-522, 1976. Mualem,Y., Hydraulicconductivityof unsaturatedporousmedia: Generalizedmacroscopicapproach,Water Resour.Res.,14, 325-334, 1978. Mualem,Y., Hydraulicconductivityof unsaturatedsoils:Prediction andformulas,in Methodsof SoilAnalysis,part 1, Physicaland MineralogicalMethods,2nded.,editedbyA. Klute,pp.799-824, ASA and SSSA, Madison, Wisc., 1986. Mualem,Y., Modelingthe hydraulicconductivityof unsaturated porousmedia,in Proceedingsof the InternationalWorkshop, IndirectMethodsfor Estimatingthe HydraulicPropertiesof UnsaturatedSoils,editedbyM. T. vanGenuchten,F. J.Leij,and L. J. Lund,pp. 15-36, Universityof California,Riverside,1992. Mualem,Y., andG. Dagan,Hydraulicconductivityof soils:unified approachto the statistical models, Soil Sci. Soc. Am. J., 42, 392-395, 1978. Nielsen,D. R., and L. Luckner,Theoreticalaspectsto estimate reasonableinitial parametersand rangelimits in identification proceduresfor soil hydraulicproperties,in Proceedingsofthe International Workshop,Indirect Methodsfor Estimatingthe HydraulicPropertiesof UnsaturatedSoils,editedby M. T. van Genuchten,F. J. Leij, andL. J. Lund,pp. 147-160,Universityof California,Riverside, 1992. Nielsen, D. R., D. Kirkham, and E. R. Perrier, Soil capillary conductivity:Comparisonofmeasuredandcalculatedvalues,Soil Sci. Soc. Am. Proc., 24, 157-160, 1960. Nimmo, J. R., Comment on the treatment of residual water content in "A consistentsetof parametricmodelsfor thetwo-phasett0• of immisciblefluidsin the subsurface"by L. Luckheretal., WaterResour.Res., 27, 661-662, 1991. Othmer,H., B. Diekk•ger, andM. Kutilek,Bimodalporosityand unsaturatedhydraulicconductivity,SoilSci., 152,139-149,1991. Parker,J. C., andM. T. vanGenuchten,Determiningsoilhydraulic propertiesfromone-stepoutflowexperimentsbyparameteresti- mation,II, Experimentalstudies,Soil Sci. Soc. Am. J., 49, 1348-1359, 1985. Parkes,M. E., andP. A. Waters,Comparisonof measuredand estimatedunsaturatedhydraulicconductivity,Water Resour, Res., 16, 749-754, 1980. Peters,R. R., and E. A. Klavetter, A continuummodelforwater movementin an unsaturatedfractured rock mass, WaterResour. Res., 24, 416-430, 1988. Pruess,K., J.S.Y. Wang,andY. W.Tsang,Onthermohydrologi½ conditionsnearhighlevelnuclearwastesemplacedinpartially saturatedfracturedtuff,2, Effectivecontinuumapproximation, WaterResour.Res., 26, 1249-!261, 1990.
  • 13. DURNER:HYDRAULICCONDUCTIVITYESTIMATIONFORSOILS 223 Puckett,W. E., J.H. Dane,B. F. Hajek,Physicalandmineralogical datatodeterminesoilhydraulicproperties,SoilSci.Soc.Am. J., 49,831-836,1985. Purcell,W. R., Capillarypressures--theirmeasurementusingmer- curyandthecalculationof permeabilitytherefrom,Trans.Am. Inst.Min. Metall. Pet. Eng., 186, 39-48, 1949. Ragab,R.,J.Feven,andD. Hillel,Comparativestudyofnumerical andlaboratorymethodsfor determiningthe hydraulicconductiv- ityfunctionof a sand,SoilSci.,131,375-388,1981. Rogowski,A. S., Estimationofthesoilmoisturecharacteristicand hydraulicconductivity:Comparisonof models,Soil Sci., 114, 423-429, 1972. Ross,P. J., and K. J. R. Smettem, Describingsoil hydraulic propertieswithsumsof simplefunctions,SoilSci.Soc.Am. J., 57,26-29, 1993. Roulier,M. H., L. H. Stolzy,J. Letey, andL. V. Weeks,Approx- imationof fieldhydraulicconductivityby laboratoryprocedures onintactcores,Soil Sci. Soc. Am. Proc., 36, 387-393, 1972. Russo,D., Determiningsoil hydraulicpropertiesby parameter estimation:On the selectionof a modelfor the hydraulicproper- ties,WaterResour. Res., 24, 453-459, 1988. Sakellariou-Makrantonaki,M., C. TzimopouIos,andD. Gouliaras, Analysisof a closed-formanalyticalmodelto predictthehydrau- lic conductivityfunction, J. Hydrol., 92, 289-300, 1987. Saxton,K. E., W. J. RawIs,J. S. Romberger,andR. I. Papendick, Estimatinggeneralizedsoil-water characteristicsfrom texture, SoilSci. Soc. Am. J., 50, 1031-1036, 1986. SchjCnning,P., Por½sitetsforholdi landbrugsjord,I, Modeller og jordtypeforskelle,(in Danish), Tidsskrift.Plantearl., 89, 411-423, 1985. Schofield,R. K., ThepF of thewaterin soil, Trans.3rd Int. Congr. SoilSci., 2, 38-48, 1935. Schuh,W, M., and R. L. Cline, Effect of soil properties on unsaturatedhydraulic conductivity pore-interaction factors, Soil Sci.Soc. Am. J., 54, 1509-1518, 1990. Sharma,M. L., and G. Uehara, Influence of soil structure on water relationsin low humic Iatosols, I, Water retention, Soil Sci. Soc. Am. Proc., 32,765-771, 1968. Sidiropoulos,E., and S. Yannopoulos, Sensitivity analysis of closed-formanalyticalhydraulicconductivitymodels,J. Hydrol., IOI, 159-172, 1988. Simmons,C. S., D. R. Nielsen, and J. W. Biggat, Scalingof field-measuredsoil-waterproperties,I, Methodology,ttilgardia, 50, 1-25, 1979. Smettem,K. R. J., andC. Kirkby, Measuringthe hydraulicprop- ertiesof a stableaggregatedsoil,J. Hydrol., 117, 1-13, 1990. Sobczuk,H. A., R. Plagge,R. T. Walczak,andC. Roth,Laboratory equipmentandcalculationprocedureto rapidlydeterminehyster- esisof somehydrophysicalpropertiesunder nonsteadyflow conditions,Z. Pflanzenern•ihr.Bodenkd., 155, 157-163, 1992. Stephens,D. B., A comparisonof calculatedandmeasuredunsat- uratedhydraulicconductivityof two uniform soilsin New Mex- ico, in Proceedingsof the International Workshop,Indirect Methodxfor EstimatingtheHydraulicPropertiesof Unsaturated Soils,editedbyM. T. vanGenuchten,F. J. Leij, andL. J. Lund, pp.249-262,Universityof California,Riverside,1992. Stephens,D. B., andK. R. Rehfeldt,Evaluationof closed-form analyticalmodelstocalculateconductivityinafinesand,SoilSci. Soc.Am.•., 49, 12-19,1985. Su,C.,andR.H. Brooks,Soilhydraulicpropertiesfrominfiltration tests,paperpresentedat WatershedManagementProceedings, Irrig.andDrain.Div.,Am.Soc.Civ.Eng.,Logan,Utah,Aug. 11-13, 1975. Talsma,T., Predictionof hydraulicconductivityfromsoilwater retentiondata,SoilSci., 140, 184-188,1985. Tsuji,G. Y., R. T. Watanabe,andW. S. Sakai,Influenceof soil microstructure on water characteristics of selected Hawaiian soils,SoilSci. Soc.Am. Proc.,39, 28-33, 1975. vanGenuchten,M. T., A closed-formequationfor predictingthe hydraulicconductivityofunsaturatedsoils,SoilSci.Soc.Am. J., 44, 892-898, 1980. vanGenuchten,M. T., andF. J. Leij, Onestimatingthehydraulic propertiesof unsaturatedsoils,in Proceedingsof theInterna- tionalWorkshop,IndirectMethodsfor EstimatingtheHydraulic Propertiesof UnsaturatedSoils,editedbyM. T. vanGenuchten, F. J. Leij, andL. J. Lund, pp. 1-14, Universityof California, Riverside, 1992. van Genuchten,M. T., and D. R. Nielsen, On describingand predictingthe hydraulicpropertiesof unsaturatedsoils, Ann. Geophys.,3, 615-628, 1985. Vereecken,H., J.Maes,J. Feyen,andP. Darius,Estimatingthesoil moistureretentioncurvefromtexture,bulk density,andcarbon content, Soil Sci., 148, 389-403, 1989. Visser,W. C., An empiricalexpressionfor thedesorptioncurve,in Waterin the UnsaturatedZone, Proceedingsof the Wageningen Symposium,vol. 1, editedby P. E. RijtemaandH. Wassink,pp. 329-335, IASH/AIHS, Unesco, Paris, 1968. Vogel, T., and M. Cislerova, On the reliability of unsaturated hydraulic conductivity calculated from the moisture retention curve, Transp. PorousMedia, 3, 1-15, 1988. Wang, J. S. Y., and T. N. Narasimhan,Hydrologic mechanisms governingfluid flow in a partially saturated fractured porous medium, Water Resour. Res., 21, 1861-1874, 1985. Wang, J. S. Y., and T. N. Narasimhan, Fluid flow in partially saturated, welded-nonmeldedtuff units, Geoderma, 46, 155-168, 1990. W6sten, J. H. M., and M. T. van Genuchten, Using texture and other soil properties to predict the unsaturated soil hydraulic functions, Soil Sci. Soc. Am. J., 52, I762-1770, 1988. Wyllie, M. R. J., and G. H. F. Gardner, The generalizedKozeny- Carman equation, World Oil, 146, 210-228, 1958. Yuster, S. T., Theoretical considerationsof multiphaseflow in idealizedcapillarysystems,WorldPetr. Congr.Proc., 3rd, 2, 437-445, 1951. Zachmann, D. W., P. C. Duchateau, and A. Klute, The calibration of the Richardsflow equationfor a draining column by parameter identification, Soil Sci. Soc. Am. J., 45, 1012-1015, 1981. Zachmann, D. W., P. C. Duchateau, and A. Klute, Simultaneous approximationof water capacityand soil hydraulicconductivity by parameteridentification,Soil Sci., 134, 157-163, 1982. W. Durner, Institute of Terrestrial Ecology, Soil Physics,Federal Institute of Technology,Grabenstr. 3, Zfirich, CH-8952 Schlieren, Switzerland. (ReceivedDecember8, 1992;revisedSeptember3, 1993; acceptedSeptember16, 1993.)