The Physics of Debris Flows PDF

THE PHYSICS OF DEBRIS FLOWS
Richard M. Iverson
CascadesVolcano Observatory
U.S. Geological Survey
Vancouver, Washington
Abstract. Recent advancesin theory and experimen- sorted,water-saturatedsedimentinvariablymove as an

tation motivatea thoroughreassessment of the physics unsteadysurgeor seriesof surges.Measurementsat the
of debrisflows.Analysesof flowsof dry, granularsolids base of experimentalflows show that coarse-grained
and solid-fluidmixturesprovidea foundationfor a com- surge fronts have little or no pore fluid pressure.In
prehensivedebrisflow theory,and experiments provide contrast,finer-grained,thoroughlysaturateddebrisbe-
data that revealthe strengthsand limitationsof theoret- hind surgefronts is nearly liquefiedby high pore pres-
ical models.Both debrisflow materialsand dry granular sure,which persistsowing to the great compressibility
materialscan sustainshearstresses while remainingstat- andmoderatepermeabilityof the debris.Realisticmod-
ic; both can deform in a slow,tranquil mode character- els of debrisflowsthereforerequire equationsthat sim-
ized by enduring,frictionalgraincontacts;andboth can ulate inertial motion of surgesin which high-resistance
flow in a more rapid, agitatedmode characterizedby fronts dominatedby solid forcesimpede the motion of
brief, inelasticgrain collisions.In debrisflows,however, low-resistancetails more strongly influenced by fluid
pore fluid that is highlyviscousand nearlyincompress- forces. Furthermore, because debris flows characteristi-
ible,composed of waterwith suspended siltandclay,can cally originate as nearly rigid sedimentmasses,trans-
stronglymediate intergranularfriction and collisions. form at least partly to liquefiedflows,and then trans-
Grain friction, grain collisions,and viscousfluid flow form again to nearly rigid deposits,acceptablemodels
may transfer significantmomentum simultaneously. must simulate an evolution of material behavior without
Both the vibrationalkinetic energyof solidgrains(mea- invokingpreternaturalchangesin materialproperties.A
suredby a quantitytermed the granulartemperature) simplemodel that satisfiesmost of these criteria uses
and the pressureof the interveningpore fluid facilitate depth-averagedequationsof motion patterned after
motion of grainspast one another,therebyenhancing thoseof the Savage-Huttertheoryfor gravity-drivenflow
debrisflow mobility.Granular temperaturearisesfrom of dry granularmassesbut generalizedto includethe
conversionof flow translationalenergy to grain vibra- effectsof viscouspore fluid with varyingpressure.These
tional energy,a processthat dependson shear rates, equationscan describea spectrumof debrisflowbehav-
grainproperties,boundaryconditions,and the ambient iors intermediate between those of wet rock avalanches
fluid viscosityand pressure.Pore fluid pressuresthat and sediment-ladenwater floods.With appropriatepore
exceedstaticequilibriumpressuresresult from local or pressuredistributions the equationsyield numericalso-
global debris contraction.Like larger, natural debris lutions that successfully predict unsteady,nonuniform
flows,experimental debrisflowsof --•10m3 of poorly motion of experimentaldebrisflows.
1. INTRODUCTION 1987],but the necessityof interactingsolid and fluid

forces makes a broader, more mechanistic distinction.
Debris flows occur when massesof poorly sorted By thisrationale,manyeventsidentifiedasdebrisslides,
sediment,agitatedand saturatedwith water, surgedown debris torrents, debris floods, mudflows, mudslides,
slopesin response to gravitationalattraction.Both solid mudspates, hyperconcentrated flows,and laharsmaybe
andfluid forcesvitallyinfluencethe motion,distinguish- regardedasdebrisflows[cf.Johnson,1984].The diverse
ing debrisflowsfrom related phenomenasuchas rock nomenclaturereflectsthe diverseorigins,compositions,
avalanches and sediment-laden water floods. Whereas and appearancesof debris flows, from quiescently
solid grain forcesdominatethe physicsof avalanches, streaming,sand-richslurriesto tumultuoussurgesof
andfluid forcesdominatethe physicsof floods,solidand boulders and mud.
fluid forcesmustact in concertto producea debrisflow. Interaction of solid and fluid forces not only distin-
Other criteriafor definingdebrisflowsemphasizesedi- guishesdebris flows physicallybut also gives them
ment concentrations,grain size distributions,flow front unique destructivepower. Like avalanchesof solids,
speeds,shearstrengths,and shearrates [e.g.,Beverage debris flows can occur with little warning as a conse-
and Culbertson,1964; Varnes, 1978; Pierson and Costa, quenceof slopefailure in continentaland seaflooren-
This paper is not subjectto U.S. copyright. Reviewsof Geophysics,35, 3 / August 1997
pages 245-296
Publishedin 1997 by the AmericanGeophysicalUnion. Paper number 97RG00426
ß 245 ß
246 ß Iverson' PHYSICS OF DEBRIS FLOWS 35 3 / REVIEWS OF GEOPHYSICS
Figure1. Digitallyenhanced
photographs
of thepathof the 2300m3 Oddstaddebrisflow,whichoccurred
January4, 1982, in Pacifica,California. The flow destroyedtwo homesand killed three people. The source
area slopes26ø.The flow path slopes21ø on averageand extends170 m downslope.Depositsat the baseof
the flowpath havebeenremoved[Schlemon et al., 1987;Wieczorek
et al., 1988;Howardet al., 1988].(Modified
from USGS [1995],courtesyof S. Ellen and R. Mark.)
vironments,and they can exert great impulsiveloadson transform to a flowing, liquid-like state, but eventually
objectsthey encounter.Like water floods, debrisflows theytransformbackto nearlyrigid deposits.New models
are fluid enoughto travel long distancesin channelswith and measurementsthat clarify the physicalbasisof de-
modestslopesand to inundatevast areas.Large debris bris flow behavior from mobilizationto depositionare
flows can exceed 109 m3 in volume and release more the focusof this paper.
than1016J of potential energy, butevencommonplace Includingthis introduction,the paper has 10 sections.
flowsof--'103 m3 candenude vegetation, clogdrainage- Section 2 describesthe net energeticsof debris flow
ways,damagestructures, andendangerhumans(Figure1). motion, the variability of debris flow mass, and the
The capricioustiming and magnitudeof debrisflows challengesthese phenomenapose for researchers.In
hamper collection of detailed data. Scientific under- section3 a compilation of key observations,data, and
standinghas thus been gleanedmostlyfrom qualitative conceptssummarizesqualitativelythe factorsthat con-
field observationsand highly idealized, first-generation trol debrisflows'mass,momentum,and energycontent.
experimentsand models.However, a new generationof In section 4, scaling analysesassistidentification and
experimentsand models has begun to yield improved classification of debris flow behavior on the basis of
insight by simulating debris flows' key common at- dimensionlessparameters that distinguish dominant
tributes. For example, all debris flows involve gravity- modes of momentum transport in solid-fluid mixtures.
driven motion of a finite but possiblychangingmassof In section 5 a retrospectiveof traditional, one-phase
poorly sorted, water-saturatedsediment that deforms modelsfor momentumtransportin debrisflowsexplains
irreversiblyand maintains a free surface. Flow is un- why suchmodels are incompatiblewith current under-
steady and nonuniform, and is seldom sustainedfor standing.In section 6, mass, momentum, and energy
morethan104S.Peakflowspeeds cansurpass 10m/sand conservationequationsfor two-phasedebris-flowmix-
are characteristically
sogreatthat bulk inertial forcesare tures establisha theoretical framework that highlights
important. Total sediment concentrationsdiffer little the variable compositionof debrisflows and the impor-
from those of static, unconsolidated sediment masses tance of solid-fluidinteractions.In section7 a relatively
and typicallyexceed50% by volume. Indeed, most de- complete analysisof an idealized debris-flowmixture
bris flows mobilize from static, nearly rigid massesof moving steadily along a rough bed helps clarify the
sediment,laden with water and poisedon slopes.When complicatedinterplaybetweenlocal solid and fluid mo-
mass movement occurs, the sediment-water mixtures tion, boundaryforces,and mechanismsof energydissi-
35, 3 /REVIEWS OF GEOPHYSICS Iverson' PHYSICS OF DEBRIS FLOWS ß 247
pation and momentum transport. In section 8 a less What is the outcomeof the experimentwhen the inter-
completeanalysisof unsteadydebrisflow motionfocuses sticesbetweenthe sedimentgrainsare filled with viscous
on persistenceof nonequilibriumfluid pressuresthat fluid?A logicalpossibility,suggested
by the behaviorof
differ with proximity to debris flow surge fronts. In elastic spheres,is that the viscousfluid will increase
section 9, numerical calculationsusing a simplified, dissipation and reduce runout. However, experience
depth-averaged routingmodelthat emphasizes the ef- with water-saturateddebris flows showsthat the pres-
fects of Coulombgrain friction mediatedby persistent ence of viscousfluid increasesrunout even though the
nonequilibriumfluid pressuresindicatethat the model fluid dissipatesenergy.Interactionsof viscousfluid with
can predict the velocitiesand depths of experimental dissipativesolid grains of widely varying sizesproduce
debris flows. Section 10 summarizesthe strengthsand this behavior and merit emphasisin efforts to under-
limitationsof current understandingand suggests prior- stand debris flow motion.
ities for future research.AppendicesA-C providesome As the precedingthought experimentimplies, debris
key mathematicaldetails omitted in previoussections, flow motioninvolvesa cascadeof energythat beginswith
and a completesummaryof mathematicalnotationfol- incipientslopemovementand endswith deposition.As
lows in a separatenotation section. a debris flow movesdownslope,its energy degradesto
Becausethis paper emphasizesphysical aspectsof higher entropy statesand undergoesthe followingcon-
debrisflow motion, it includesonly incidentalcoverage versions:
of importanttopicssuchasdebrisflowhabitats,frequen-
cies,magnitudes,triggeringmechanisms,hazard assess- bulk gravitationalpotential energy
ments,engineeringCountermeasures,
morphologyand -• bulk translational kinetic energy
sedimentologyof debrisflow deposits,and the relation-
ship betweendebrisflows and other massmovements. • grain vibrational kinetic energy
Severalpreviousreviewsand compilations,suchasthose
by Takahashi[1981, 1991, 1994], Innes [1983], Costa + fluid pressureenergy -• heat
[1984], Johnson[1984], Costa and Wieczorek[1987],
Here right pointingarrowsdenote conversionsthat are
Hooke [1987],Pierson[1995], and Iversonet al. [1997]
irreversible,exceptin specialcircumstances,
whereasthe
treat thesesubjectsmore completely.In addition,video-
two-way arrow denotes a conversionthat apparently
tape recordings[Costaand Williams,1984;SaboPublicity
involvessignificantpositivefeedback.The detailsof this
Center,1988]revealmanyqualitativeattributesof debris
energy cascadeencompassvirtually all the important
flows, and summariesby Iversonand Denlinger[1987],
issuesof debrisflow physics.Before pursuingthesede-
MiyamotoandEgashira[1993],Savage[1993],andHutter tails, however, it is worthwhile to consider debris flow
et al. [1996]introducesomeof the quantitativeconcepts
elaborated here.
energeticsfrom a broader perspective.
The net efficiencyof debris flows, and of kindred
phenomenasuchasrock and snowavalanches,describes
2. BULK ENERGETICS AND RUNOUT EFFICIENCY
conversionof gravitationalpotential energyto the work
doneduringdebrisflow translation.The more efficiently
this conversionoccurs,the less vigorouslyenergy de-
The energeticsof debris flows differ dramatically
grades to irrecoverable forms such as heat, and the
from thoseof a homogeneoussolid or fluid. The inter-
farther the flow runsout before stopping.Net efficiency
actions,and not merely the additive effectsof the solid
can be evaluated by integrating an equation that de-
and fluid constituents, are important.A simplethought scribes motion of the debris flow center of mass as a
experimenthelpsillustratethis phenomenon:
function of time. Alternatively, as was recognizedorigi-
Considerfirst a very unrealisticbut simplemodel of a
nallybyHeim [1932]for rock avalanches, the outcomeof
debris flow. A mass of identical, dense, frictionless elas-
the integration can be obtained without specifyingan
tic spheresflowsdowna bumpy,rigid inclineand onto a
equationof motion by equatingthe total potential en-
horizontal runout surface, all within a vacuum. The
ergy lost during motion, M#H, to the total energy de-
spheresjostle and collideas they acceleratedownslope,
gradedto irrecoverableformsby resistingforces,M#R,
but no energydissipationoccurs,and the flow runs out
that work throughthe distanceL to makethe debrisflow
forever. Then fill the spacesbetweenthe sphereswith a
stop:
viscousfluid less densethan the spheres(e.g., liquid
water), and repeat the experiment.Owing to viscous M#H - M#RL (1)
shearing,the mixture losesenergy as it moves down-
slope,and runout remainsfinite. The fluid retardsthe Here M is the debrisflow mass,# is the magnitudeof
motion. Next, replace the elastic sphereswith rough, gravitationalacceleration,and R is a dimensionless net
inelastic sedimentgrains, and repeat the two experi- resistancecoefficient,which incorporatesthe effectsof
ments.In the vacuumthe collectionof grainsrunsout a internal forcesbut which dependsalsoon externalforces
finite distance and stops owing to energy dissipation that act at the bed to convertgravitationalpotential to
causedby grain contactfriction and inelasticcollisions. horizontal translation. The coordinates H and L de-
248 ß Iverson:PHYSICSOF DEBRISFLOWS 35 3 / REVIEWSOF GEOPHYSICS
summarizestypicalL/H valuesinferred from the distal

limits of debris flow source areas and deposits.The
tabulatedL/H valuescan be comparedin only the
.
': ....Distal
limits
. broadest sense because the data were collected on debris
Y
flowswith diverseoriginsand flow path geometriesby
.œ
investigatorswith diverse objectives.Nonetheless,the
data of Table 1 indicatethat L/H increasesroughlyin
proportionto the logarithmof volume for debrisflows
withvolumes
greaterthanabout10s m3 but thatL/H
Figure2. Schematic
cross
section H andL fordebris remainsfixed at ---2-4 for smaller flows. Data for dry
defining
flow paths.Strictly,H and L are definedby linesthat connect rock avalanches exhibit similar trends but indicate that
dry avalanchestypically have only about half the effi-
the sourcearea centerof massand the depositcenterof mass.
In practice,H and L are commonlyestimatedfrom the distal
ciency(L/H) of debrisflowsof comparablevolume [cf.
limitsof the sourcearea and deposit. Scheidegger,1973; Hsu, 1975; Davies, 1982; Li, 1983;
Siebert, 1984; Hayashi and Self, 1992; Pierson, 1995].
Theseempiricaltrendsare noteworthy,but case-by-case
scribedisplacement of the debrisflow centerof mass
variationsin debris-flowbehaviormake runout predic-
during motion:H is the vertical elevationof the debris
flowsource
abovethedeposit,
andL is thehorizontal
tiononthebasis
of onlyL/H ratherquestionable.
Rigorousevaluationof L/H from center-of-massdis-
distancefrom sourceto deposit(Figure 2).
Eventhoughall debrisflow energyultimatelyde- placements
undercontrolled
initialandboundary
con-
ditions has been possibleat the U.S. GeologicalSurvey
gradesto heat, thermodynamicdata provide few con-
straints
for evaluatingR in (1). Theequationshows that (USGS) debrisflow flume (Figure3) [Iversonand
a debrisflow'stotal energydissipationper unit massis LaHusen,
1993].Experiments
in whichabout10m3of a
givenby #H, which impliesabout 10 J/kg of heat pro- water-saturated,poorly sorted, sand-graveldebris flow
ductionper meter of flow descent.Even without heat mixture is suddenlyreleasedfrom a gate at the head of
loss,this 10 J/kg sufficesto raise the temperatureof a the flume yield L/H --- 2 for unconfined runout but
typicaldebrisflow mixture only about 0.005øC.Conse- L/H > 2 for channelizedrunout (Figure 4). These
quently,debrisflow temperaturemeasurements in open, valuessurpassthe L/H for runout of similarsand-gravel
outdoor environments,with unrestrictedheat exchange massesnot saturatedwith water [Major, 1996].When the
and ambienttemperaturesthat vary Widely,yield little sand-gravelmix is replacedby well-sortedgravel, how-
resolutionof energydissipation due to flow resistance. ever, the influenceof water on the outcomeof experi-
Instead, debris flow physicsconventionallyemphasizes ments changes: drygravelproduces L/H > 2, butwater-
thepurelymechanical behaviorof anisothermal system, saturated gravel produces L/H < 2. Thus water
and this paper followsthat convention. enhances the mobility of poorly sorteddebrisflow sed-
The mechanicalphenomenathat governR must be iments in a manner not manifested by mixtures of well-
quantified in detail to understandand predict debris sorted gravel andwater, and experiments with water-gravel
flow motion, but evaluation of net efficiencyfrom the mixturesprovidea poor surrogatefor experimentswith
aftermathof a debrisflow is far simpler.Dividing each realistic debris-flow materials.
sideof (1) by M#HR yields Effectsof water-sedimentinteractionsposechalleng-
ing problemsthat consumemuch of the remainder of
1/R = L/H (2) this paper, but effectsof debrisflow massare even more
Which showsthat the net efficiency,defined as l/R, enigmatic. According to equations (1) and (2), debris
inci•easesas the runout distanceL increasesfor a fixed flow massshould not affect runout efficiency,but the
descentheight, H. Thus net efficiencymay be deter- data of Table 1 contradict this inference. The Cause of
mined from surveysof debris flow source ai•easand this contradiction is difficult to resolve because debris
depositsthat yield the value of L/H. flowsand avalanchescan changetheir massand compo-
Rigorousevaluationsof L/H from debrisflows'cen- sition while in motion and can spreadlongitudinallyto
ter-of-mass displacementshavebeenrare,butfieldmap- changetheir massdistribution[cf. Davies, 1982]. Some
ping of debrisflow pathsand detailedmeasurementson debris flowsgrowseveralfoid in mass Owing to bedand
experimentaldebrisflowsdemonstratethree important bank erosion [Piersonet al., 1990] and others decline
p6ints [cf. Vallanceand Scott,1997]: (1) L/H of water- substantiallyin solidsconcentrationas a rest•ltof mixing
saturated debris flows exceeds that of drier sediment with streamwater[Pierson
andScott,1985].Changes in
flowswith comparablemasses,(2) Large debris flows debrisflow massor compositionhavebeenidentified
appear to have greater efficiencythan small flows,and somewhat interchangeably
by the terms"bulking"(in-
(3) L/H dependson runoutpath geometryandboundary creaseof massor solidsconcentration)and "debulking"
conditionsthat determine, for example, the extent of (decreaseof massor solidsconcentration),but more
erosion, sedimentation, and flow channelization. Table 1 preciseterminologyis desirablebecausechangesin de-
35, 3 /REVIEWS OF GEOPHYSICS Iverson: PHYSICS OF DEBRIS FLOWS ß 249
TABLE 1. EstimatedValuesof Total Flow Volume,RunoutDistanceL, DescentHeight H, and EfficiencyL/H

of Various Debris Flows
Flow
Volume,
FlowLocation Date Reference m3 L, m H, m L/H Origin
Mount Rainier, Osceola circa 5700 B.P. Vallance and Scott ---109 120,000 4,800 25 landslide and down-
mudflow [1997] stream erosion
Nevados Huascaran, Peru May 31, 1970 PlafkerandEricksen ---108 120,000 6,000 20 landslide
[1978]
Nevado del Ruiz, Colombia, Nov. 13, 1985 Piersonet al. ---107 103,000 5,190 20 pyroclastsmelting
Rio Guali [1990] snow
Mount St. Helens, South May 18, 1980 FairchiMand Wigmosta '--107 44,000 2,350 19 wet pyroclastic
Fork Toutle [1983] surge
Mount St. Helens, Muddy May 18, 1980 Pierson[1985] '-'107 31,000 2,150 14 wet pyroclastic
River surge
Wrightwood,Calif., Heath Canyon May 7, 1941 SharpandNobles ---106 24,140 1,524 16 landslide
[1953]
Three Sisters,Oreg., Separation 1933 J. E. O'Connor et al. '--10 6 6,000 700 9 glacierbreakout
Creek (manuscriptin flood
preparation,1997)
Mount Thomas,NZ, BullockCreek April 1978 Pierson[1980] •10 s 3,500 600 6 landslide
Wrightwood,
Calif.,HeathCanyon May 1969 Mortonand Campbell ---l0 s 2,700 680 4 landslide
[1974]
Santa Cruz, Calif., Whitehouse Jan. 4, 1982 Wieczorek
et al. [1988] ---10s 600 200 3 landslide
Creek
Pacifica, Calif., Oddstad site Jan. 4, 1982 Howard et al. [1988] ---103 190 88 2 landslide
USGS debris flow flume Sept. 25, 1992 Iversonand LaHusen ---10• 78 41 2 artificial release
[1993] from flume gate
Mostdataarefor flowsthatwereobserved duringmotionor withinhoursof deposition. With the exception
of theOsceolamudflow,all flows
apparentlymaintaineda relativelyconstantmass(withina factorof 2) from initiationto deposition.The Osceolais includedin the tabulation
becauseit is the largestwell-documented
debrisflow in the terrestrialgeologicrecord.
bris-flowmass,independentof changesin composition, changewith the bed andbanks,whichmaydiffergreatly

might influenceefficiency. in differentlocalities.Despite the lack of clear resolu-
Attemptsto use elementaryenergybalancesto pre- tion, recognitionof the fundamentaleffectsof external
dict effectsof masschangeon debris flow efficiency forceson debrisflow efficiencyis essential,for otherwise
encounterdifficulties,which can be tracedto assump- it may be tempting to attribute differencesin runout
tions implicit in equation(1) and in similarlysimple solelyto differencesin flow compositionand rheology.
momentum balances [cf. Cannon and Savage,1988; Section7 delvesmore deeplyinto the mechanicaleffects
Hungr, 1990;Erlichson,1991].It might seemfrom (1), of external forces.
for example,that loss of massduring motion should
increaseefficiencybecausethe potentialenergyinitially
availableto power the motion,M#H, staysfixed,while 3. MASS, MOMENTUM, AND ENERGY CONTENT:
the work doneby resistingforcesapparentlydeclinesas DESCRIPTION AND DATA
the flow massdeclines.The problemwith this logiclies
in the assumptionthat R remainsconstantor decreases An empiricalpicture of debris flow physicscan be
as the flow massdeclines.This assumptionwould be drawn from a combination of real-time field observa-
correct if R depended only on internal forces, but R tionsand measurements [e.g.,Okudaet al., 1980;Li and
dependsalsoon the externalforcesthat causethe flow Yuan, 1983;Johnson,1984;Pierson,1980,1986],detailed
massto decline.Lossof debrisflow massrequiresthat observations and measurements duringcontrolledfield
work be done on the flow by the banks and bed to and laboratoryexperiments[e.g.,Takahashi,1991;Khe-
decelerateand depositthe lost mass.This work addsto gai et al., 1992;Iversonand LaHusen,1989, 1993],and
the work that wouldbe doneoverthe samepath length analysesof debrisflow pathsand deposits[e.g.,Fink et
in the absenceof deposition.The critical questionis al., 1981;Pierson,1985, 1995;Whippleand Dunne, 1992;
whether the additional work is less than the energy Major, 1996]. Furthermore,videotapecompilationsof
savingsaccruedby leaving massbehind. Universal an- debrisflow recordings providemanyqualitativeinsights
swersto this questionare perhapsunattainable.The [Costaand Williams,1984;SaboPublicityCenter,1988].
sameis true for the questionof whether massgain will Relativelylittle detailedinformationis availablefor sub-
increasebulk mobility and runout. In each case, mass aqueousdebrisflows,but mostaspectsof their behavior
changedependson work done during momentumex- (other than their tendencyto hydroplane,entrain sur-
250 ß Iverson- PHYSICS OF DEBRIS FLOWS 35 3 / REVIEWS OF GEOPHYSICS
ß
: fGate
ß
-..•.m'.'• L Endoffixed
':•:.'•.:
..
x,;4 B channel
walls
Figure 3. U.S. Geological Survey

(USGS) debrisflow flume [Iversonet
al., 1992]. (a) Photographof an ex-
periment in progress,May 6, 1993.
(b) Schematicverticalcross-sectional
profile.
rounding water, and transform to dilute density cur- data by themselvesadd little to the understandingof
rents)appearsimilarto thoseof their subaerialcounter- debrisflow physics.Suchunderstandingrequiresdata on
parts [Priorand Coleman,1984;Weirich,1989;Mohriget debris properties that are rigorouslymeasurableonly
al., 1995;Hamptonet al., 1996].This summaryfocuseson during motion.
the subaerialcaseand particularlyon inferencesdrawn Few acceptabletechniquesexist to measureproper-
from detailed experimentaldata. ties of flowing debris, even simple properties such as
bulk density.Grosslyinvasivetechniquessuchas plung-
3.1. MaterialProperties ing bucketsor sensorsinto debris flows conspicuously
Somepropertiesof debrisflow materialscanbe mea- change the state of the debris, and the inconsistent,
sured readily and accuratelyin a static state, whereas noisy, dirty character of debris flows has discouraged
other propertiesdependon the characterof debrismo- attempts to use noninvasivetechniquessuch as ultra-
tion. The most readily measuredstatic property is the sonic,X ray, lasersheet,or magneticresonanceimaging
grain sizedistribution.Abundantgrain sizedata demon- that are useful for probing simpler solid-fluidmixtures
stratethat individualdebrisflowscan containgrainsthat [Lee et al., 1974;Malekzadeh,1993;KytomaaandAtkin-
range from clay size to boulder size. However, many son,1993;Abbottet al., 1993].The mostconcertedefforts
publishedgrain size distributionsare biasedbecause to determine properties of flowing debris have relied
they ignore the presenceof cobblesand bouldersthat either on real-time measurements at the boundaries of
are difficultto sample[Major and Voight,1986]. None- debrisflowsin artificial channelsor on postdepositional
theless,it is clear that sand, gravel, and larger grains measurements on desiccated debris flow sediment sam-
composemost of the massof debrisflows and that silt ples reconstitutedby adding water [Takahashi,1991].
and clay-sizedgrainscommonlyconstitutelessthan 10% Precisereal-time measurements havebeen possibleonly
of the mass[e.g.,Daido, 1971; Costa,1984; Takahashi, with experimental flows that contain sediments no
1991;Pierson,1995;Major, 1997].Grain sizedata reveal coarserthan gravel [e.g.,Iversonet al., 1992]; measure-
the oversimplification of debrisflow modelsthat assume mentson reconstitutedsampleshavegenerallyexcluded
a singlegrain size or a preponderanceof fine-grained sediment coarser than gravel and have also involved
sediment[e.g.,Coussotand Proust,1996], and they rein- uncertaintiesabout appropriatewater contentsand de-
force the notion that a diversityof grain sizesmay be formation styles[e.g., Phillipsand Davies, 1991; Major
critical to debrisflow behavior.Beyond this, grain size and Pierson,1992].
35 3 /REVIEWS OF GEOPHYSICS Iverson: PHYSICS OF DEBRIS FLOWS ß 251
8-31-94
0 10 METERS
ISOPACHS IN CENTIMETERS
Figure 4. Isopachmaps of depositsthat formed at the base of the USGS debrisflow flume during three
experiments
in whichnearlyidentical
volumes
(---10m3) of water-saturated
sandandgravel
werereleased
from
the gate at the flume head. In eachmap the shadedarea denotesthe positionof a nearlyhorizontalconcrete
pad adjacentto the flume base. Differencesin positioningof deposits,which indicate differencesin flow
runout,are attributableto differentdistancesof flow confinementby rigid channelwalls [after Major, 1996].
Graphs of flow depth and total basal normal stress flowsinferred from depositsseldomrange outside 1800
recorded simultaneouslyat fixed cross sectionshave to 2300kg/m3 [cf.Costa,1984;Pierson,
1985;Majorand
been used to estimate the average bulk density p of 17oight,1986]. The data of Table 2 imply that deposit
experimentaldebrisflowsin the USGS debrisflow flume densitiesprovide crudely accurate estimatesof debris
(Figure 5). Measuredbasal fluid pressuresvary some- flow densitiesbut that relatively low density (dilute)
what asynchronously with the basal total normal stress debrisflowsmay producedepositsthat yield deceptively
(astheydo in largernaturaldebrisflows[e.g.,Takahashi, high estimatesof flow density.The data also indicate
1991]), and bulk densityestimatesbased on the fluid that the volume fraction of solid grains in debris flows
pressurealone may be inaccurate.Further complicating typically ranges from about 0.5 to 0.8, although more
the picture,debrisflowsinvariablymoveasone or more dilute flowsare possible.The wide variety of grain sizes
pulsesor surges,and steady,uniform flow seldom, if and shapesin debrisflowsallowsthem to attain densities
ever, occurs.The relationshipbetweenflow depth,basal that.substantially
surpass
thoseof randompackings
of
fluid pressure,and basalnormal stresschangesmarkedlyidentical spheres [Rodine and Johnson, 1976], which
have solid volume fractions no greater than 0.635
as surgespass(Figure 5) [cf. Takahashi,1991].Only for
brief intervalswhen flow is nearly steadyand uniform[Onada and Liniger, 1990]. The ability of debris flow
(implyingnegligiblevelocitynormalto the bed) canthe solidsto exhibit dense,interlockedpackingsas well as
averagebulk densitybe estimatedwith confidencefrom loose, high-porositypackings has significant ramifica-
the measuredbasal normal stress{r and a simple static
tionsfor mixturebehavior[Rogerset al., 1994].
force balance,{r = p#h cos 0, where 0 is the bed slope Rheometricinvestigationsof debrisflow mixturesre-
and h is the flow depth measurednormal to the bed. constitutedby adding water to samplesof debris flow
Employingthis force balanceand data from Figure 5 for
depositshavedemonstratedthat mixturebehaviorvaries
an intervalwhen nearly steadyflow occurred(between markedlywith subtlevariationsin solidvolume fraction
18.1 and 18.3 seconds)yieldsthe densityestimatep = (concentration),shear rate (an approximatesurrogate
2100kg/m3.Similarly computedestimates
for additional for kinetic energycontent), and grain size distribution
flumedebrisflowsrangefrom 1400to 2400kg/m3, (particularlythe silt and clay content, which strongly
whereasmean bulk densitiesof samplesexcavatedfrom influencessolid-fluidinteractions)[O'Brienand Julien,
freshdepositsof the sameflowsrangeonly from 2100 to 1988;Phillipsand Davies,1991;Major and Pierson,1992;
2400kg/m 3 (Table2). Bulkdensitiesof naturaldebris Coussotand Piau, 1995]. Such behavior evokesstrong
252 ß Iverson: PHYSICS OF DEBRIS FLOWS 35, 3 /REVIEWS OF GEOPHYSICS
Ai al
0.20
' I ' I ' I '
03 0.4 Debris Debrisflowdepth
0.15
I- 0.3
• 0.2
0.10
:• 0.1
0.05
-0.1 , I , I , I , I 0
.000
03
.• 8000 Total basal normal stress
0
03 6000 .000 _
'I'I
z 4000 000
r.• 2000
000
rr 0
03 -2000 0
i i i
3000 Basalpore 2OOO _ Basal

polre
pressure
I I
- _
:• • 2000 1500
LU• 1000
1000
500
11 13 15 17 19 21 18.10 18.15 18.20 18.25 18.30
TIME, IN SECONDS TIME, IN SECONDS
Figure 5. Representativemeasurements of flow depth,basaltotal normalstress,and basalfluid pressure

made at a crosssection67 m downslopefrom the head gate at the USGS debrisflow flume. Data are for a
debrisflowof 9 m3of water-saturated
loamysandandgravelreleased
August31,1994.Theflumegateopened
at t = 6.577 s. All data were sampledat 2000 Hz. Depth measurements
were madewith a lasertriangulation
system.
Totalstresses
weremeasured
witha loadcellattached
to a 500-cm
2steelplatemounted
flushwiththe
flume bed and roughenedto matchthe textureof the surroundingconcrete.Pore pressures were measured
with a transducermountedflushwith the flumebed. The transducerdiaphragmwasisolatedfrom debrisflow
sedimentby a number230 meshwire screen,andthe transducerport wasprefilledwith water to retardentry
of fine sedimentparticles.(a) Data for the entire eventduration.(b) Details of the samedata for a 0.2-s
interval when flow was nearly steady.
analogiesbetween debrisflow mixturesand better un- the diameter of a sphereof identicalvolume), friction
derstoodmixturessuchas ideal, densegases[cf. Camp- coefficient
tan q)a(whereq)ais the angleof sliding
bell, 1990].In densegasesthe concentrations
of distinct friction,whichdependson grain shapeand roughness),
molecularspecies,their kineticenergies(temperature), and restitutioncoefficient'e (which varies from 1 for
and their interaction forces determine bulk mixture perfectlyelasticgrainsto 0 for perfectlyinelastic)[Spie-
propertiessuchas density,flow resistance, and the pro- gel, 1967,p. 195]. The granularsolidsas a whole occupy
pensityfor changesof state.Similarly,the bulk proper- a fraction vs of the total mixture volume and have a
ties of debrisflow mixturesdependfundamentallyon the distributionof • that characteristicallyspansmany or-
concentrations, kinetic energies,and interactionsof dis- ders of magnitude.The fluid componentof the mixture
tinct solid and fluid constituents[cf. Johnson,1984, pp. ischaracterizedbyitsmassdensity pT(assumedlessthan
289-290]. Therefore the followingdescriptioneschews Ps), effectiveviscosity
ix, and volumefractionvT. At
the traditional practice of assumingthat debris flow meannormalstresses typicalin debrisflows(<100 kPa),
solidsand fluids are inextricablyjoined to form a single- the solid and fluid constituentsare effectivelyincom-
phasematerial;insteadit emphasizes the distinctprop-
pressible,
andvariations in Vs/V
f greatlyexceedthosein
erties and interactions of debris flows' solid and fluid
Ps/PT.
Twoadditional properties linkthebehavior of the
constituents. solidandfluid:the volumefractionsobeyVs+ vf = 1
The salientmechanicalpropertiesof a solidgrain are (thusthemixturedensity obeysp = PsVs + pfl)f),anda
diametera (definedas parametersuchas the hydraulicpermeabilityk charac-
its massdensityPs,characteristic
TABLE 2. Comparison of Bulk Densities of Experimental Debris Flows and Their Deposits
Laser Mean Bed Mean Bulk Dried Bulk Calculated Mean

Mean Flow Measurement Stresson DensityFrom Densitiesof SaturatedDensity
Depthh, TimeInterval, 500-½m
2 Plate, BedStress, DepositSamples, of Deposit,
ExperimentDate Material m s Pa kg/m3 kg/m• kg/m•
April 19, 1994 sand-gravelmix 0.05 17.0-17.2 1000 2400 1870 2200
1930
1850
1830
1930
May 25, 1994 loam-gravelmix 0.05 16.0-16.5 600 1400 1630 2100
24.0-24.5 1770
Aug. 31, 1994 loam-gravelmix 0.08 18.1-18.3 1400 2100 2050 2200
1910
1680
1770
2260
2050
2460
Bulk densitieswere determinedon the basisof (1) simultaneous

measurementsof flow depthandbed normalstressduringintervalsof nearly
steadyflow and (2) averagevaluesof depositdensitiessampledby the excavationmethod,as describedby Blake [1965].
terizes the resistance to relative motion of solids and time and length scales.Rodine and Johnson[1976], for
fluid [Iversonand LaHusen, 1989]. Table 3 summarizes example, used a length scale approach and suggested
the definitionsand typicalvaluesof theseproperties,and that all grainswith • < •' effectivelyact like fluid asthey
Figure 6 showsthat key properties(e.g., fluid volume exert forceson a grainwith diameter•'. This appliesfor
fraction and permeability)can be stronglyrelated. any arbitrary •' and resultsin distinctionsbetween solid
Definition of distinct solid and fluid properties and fluid constituentspurely relative to the choiceof •'.
prompts two difficult questions:(1) What effectively However, an absolute distinction between solid and fluid
constitutesthe fluid fraction, when a debris flow may constituentsis necessaryfor applicationof formal mix-
contain solidsof any size, includingcolloidal and clay ture theories [Atkin and Craine, 1976] and can be de-
particlescarried in solutionand suspension? (2) If the duced if time as well as length scalesare considered.
fluid fraction includes fine solid particles, can it be If the duration to of a debrisflow is long in compar-
characterized by the simpleproperties [9fandix? isonwith the timescalefor settlingof a grain of diameter
Criteria for distinguishing the effectivefluid and solid • in static,pure water with viscosityixw,the grain must
fractionsin debrisflowscanbe developedon the basisof be consideredpart of the solid fraction. Such a grain
requireseither sustainedinteractionswith other grains
or fluid turbulence to keep it suspendedin the debris
TABLE 3. Typical Values of Basic Physical Properties of
Debris Flow Mixtures flow mixture(Figure7). On the otherhand,if a graincan
remain suspendedfor timesthat exceedto as a result of
Propertyand Unit Symbol TypicalValues only the viscousresistanceof water, the grain may act as
part of the fluid. Timescalesfor debris flow durations
Solid Grain Properties range from about 10 s for small but significantevents
Massdensity, kg/m3 Ps 2500-3000 (e.g.,Figure1) to 104sforthelargest. Thetimescale for
Mean diameter,m • 10-s-10
Frictionangle,deg d•s 25-45 grain settling can be estimated by dividing the charac-
Restitution coefficient e 0.1-0.5 teristic settling distance or half thickness,h/2, of a
debris flow by the grain settlingvelocity Vsctestimated
Pore Fluid Properties
Massdensity, kg/m 3 pT 1000-1200 from Stokes'law or a more generalequationthat accounts
Viscosity,Pa s • 0.001-0.1 for grain inertia [Vanoni,1975].Thus if h/(2tz)Vsct)< 1,
the debris flow duration is large compared with the
Mixture Properties
Solid volume fraction Us 0.4-0.8 timescalefor settling.The half thicknessof debrisflows
Fluidvolumefraction Vf 0.2-0.6 ranges from about 0.01 m for small flows to 10 m for
Hydraulicpermeability, m2 k 10-•3-10-9 large ones. Thus h/2tD • 0.001 m/s, which implies
Hydraulicconductivity, m/s K 10-7-10-2 vsct< 0.001 m/s for grains to act as part of the fluid.
Compressive stiffness, Pa E 103-10 s Settling velocitiesof 0.001 m/s or lessin water require
Friction angle, deg d• 25-45
grainswith diameterslessthan about 0.05 mm [Vanoni,
254 ß Iverson- PHYSICS OF DEBRIS FLOWS 35, 3 / REVIEWS OF GEOPHYSICS
10-9 ___ I I I
_
10-10•-
EXPLANATION
_
10-11• _
ß Sandandgravelmix
_
_
-h' Loamandgravel
mix
10-12 ß MountSt. Helens,1980
ß
_
_ & Osceola Mudflow

_
ß
[] Mount St. Helens < 10mm
10-13 •-_
[]
_
_
/• Osceola<10 mm
_
10-14
_
10-15 I i I
0.20 0.30 0.40 0.50 0.60
POROSITY
Figure 6. Hydraulic permeabilitiesof representativedebris-flowmaterialsas a functionof porosity(fluid

volumefraction)uf.Testswithsediments
sieved
to removegrains>32 mm(solidsymbols)
wereconducted
in
a compactionpermeameter, and the volume fractions depicted for each material representthe full range
achievablein the deviceunder very low (--•2 kPa) effectivestress.Tests at lower volume fractions(open
symbols)wereconductedundercompression in a triaxialcellusingonlythe sedimentfraction<10 mm [Major,
1996]. Each material exhibitsan approximatelyexponentialdependenceof permeabilityon volume fraction,
i.e.,k = koexp(auf)wherekoanda areconstants.
Grainsizedistributions
of allmaterials
aregivenbyMajor
[1996].
1975, p. 25]. This critical grain size correspondsquite by Pdry-- PsUfines(

1 -- Us)+ PsUs,
whichcanbe manip-
well with the silt-sandboundary of 0.0625 mm, and it ulated to yield a simple expressionfor U•nes:
also falls in the range where settlingis characterizedby
(Pdry/Ps)
-- Us OtUs Ot
grainReynolds
numbers
(NRey----•pfVset/bCw)
muchless
than 1, so that viscousforcesdominate grain motion. By Urines
= i -- Us i -- Us (Ps/Pdry)(1
+ a) - 1
this rationale a useful but inexact guideline statesthat (4)
grains larger than silt size compose the debris flow Here a = PsUfines(1
- %)/p•% is the massof fine grains
solids,whereasgrainsin the silt-clayfinesfraction act as divided by the massof coarserclastsin disaggregated,
part of the fluid. Analysesof fluids that drained from dried sedimentsamples;equivalently,(100a)/(1 + a) is
depositsof four debris flows at the USGS debris flow the masspercentageof fines in suchsamples.Estimates
flume provide empirical support for this guideline:the of Urine
sfrom(4) andpffrom(3) areinexactbecause
(4)
sedimentmassin each debrisflow includedonly 1-6% assumesthat the sampledportion of the depositlosesno
grainsfiner than sand,but more than 94% of the sedi- fines during drainage. By judiciously samplingwhere
ment massin eachsampleof the effluentfluid consisted drainagehas been minimal, the estimationerror can be
of grainsfiner than sand(Table 4) [Major, 1996].
Incorporation of fine grainsinfluencesthe massden-
sityof debrisflowfluid,pf, definedas
pf- psUfines
q'-pw(1-Urines) (3)
where Urine s is the volume fraction of fluid occupiedby
fine (i.e., silt and clay) grains,Pwis the massdensityof
pure water, and Psis the massdensityof fine grains(for
simplicity assumedequal to that of the coarser sedi-
ment).Direct measurements
of pf of effluentfluidsin
flume experimentsyield valuesthat range from 1030 to
1110kg/m
3 (Table4). Wheredirectmeasurementsare Figure 7. Schematicdiagram illustrating the distinctionbe-
impossible,
estimates
of pfcanbemadefrom(3) andthe tween a small grain that remains suspendedexclusivelyby
dry bulk densitiesand grain size distributionsof debris viscousforcesand thus can act as part of the fluid (grain A)
flow deposits.These estimatesexploit the fact that the and a large grain that requiresinteractionwith other grainsto
dry bulk densityof undisturbeddepositsamplesis given remainsuspended(grain B).
35, 3 /REVIEWS OF GEOPHYSICS Iverson:PHYSICSOF DEBRISFLOWS ß 255
TABLE4. Densities
pf andVolumetric
Sediment
Concentrations
Vsediment
of theFluidFractionin FourExperimental
Debris Flows at the USGS Debris Flow Flume
CalculatedFrom Deposit
MeasuredFrom EffluentFluid Samples Samples*
Experiment
Date Material
PT'
kg/m
3 Vsed
.... t
Sediments
Consisting
ofFines,
wt% 1)fines
pT,
kg/m
3
April 19, 1994 sand-gravel

mix 1030 0.02 100 0.02-0.05 1030-1080
April 21, 1994 sand-gravel
mix 1040 0.025 99.7 0.02-0.05 1030-1080
June21, 19947 sand-gravel
mix 1160 0.095 94.2 0.02-0.05 1030-1080
July20, 1994 sand-gravel-loam
mix 1110 0.064 94.6 0.07-0.12 1120-1200
Measureddensitiesare thoseof fluid that drainedfrom debrisflow depositswithin the first few minutesfollowingdeposition.Calculated
densities
are obtainedfromequations(3) and(4) andthe grainsizedistribution anddriedbulkdensityof depositsedimentsamples obtained
several
hoursafterdeposition.
All calculations
assume
Os= Pfines
-- 2650kg/m
3 andPw= 1000kg/m3;
alsoassumed
isthevaluePary= 1900
kg/m3,whichis the meanof depositdriedbulkdensities
inferredfromtensof measurements.
*Calculationsfor depositsemploytherangeof otvaluesinferredfromnumerous grainsizeanalyses of deposits:
0.01-0.02for the sand-gravel
mix and 0.03-0.06 for the sand-gravel-loam
mix.
?Someof the June21, 1994,fluid leakedfrom the samplejar duringtransit,andthe resultingfluid lossmaybe responsiblefor the relatively
large1)sediment
andpœvalues
measured
thei•eafter.
minimized.Comparison of pf loidal particlesare significant[Coussot,1995].Nonethe-

of directmeasurements
with estimatescalculatedfrom depositpropertiesand less,an expressionsuchas (5), whichpredictsincreased
equations(3) and (4) showsthat estimationerrors of effective Newtonian viscosityas a consequenceof in-
about 10% are common(Table 4). creasedfinesconcentrationin the fluid fraction,provides
The presenceof fine grainsin the pore fluid also a useful guideline. Viscometrictests of suspensions of
influencesthe effectivefluid viscosity.The influenceis only the finesfractionfrom debrisflow sedimentspro-
complexandhasbeenthe objectof systematic research vide empirical support for such a guideline but also
dating at least to Einstein [1906], who deducedthe reveal complicationsthat remain incompletelyresolved
well-knownequationix/ixw= 1 + 2.5x)fines , in which ix [O'BrienandJulien,1988;Major andPierson,1992;Cous-
is the effectiveviscosityof the fine-grainsuspension and sot and Piau, 1994].
ixwis the viscosityof the fluid alone.Einstein'sequation
appliesto dilute suspensions of chemicallyinert spheres 3.2. Debris Flow Mobilization
thatsatisfy
1)fines
( • 0.1andNRey<< 1, conditions
that Successful models of debris flows must describe the
are roughlymet by the fluidsin the experimentaldebris
flows characterized in Table 4. Some natural debris flows
mechanicsof mobilizationaswell asthoseof subsequent
flow and deposition.Althoughdebrisflowscanoriginate
havehigherconcentrations of fines,however[Majorand
by variousmeans,aswhenpyroclasticflowsentrainand
Pierson,1992], so treatmentsmore general than Ein-
melt snowand ice [Piersonet al., 1990] or when abrupt
stein'sare necessary.Although numerousinvestigators
floodsof water undermineand incorporateample sedi-
[e.g.,FrankelandAcrivos,1967]havededucedequations
ment (J. E. O'Connor et al., manuscriptin preparation,
to predictthe effectiveviscosityof concentrated
suspen-
1997) origination from slope failures predominates.
sionsof fine spheres,otherinvestigationshaveexplained
Hence mobilization is defined here as the processby
why no suchequationcanbe expectedto work well for
which a debris flow developsfrom an initially static,
the full range of Vnnes and all conceivableflow fields
apparentlyrigid massof water-ladensoil, sediment,or
[Batchelor and Green,1972;Acrivos,1993].For the spe-
rock. Mobilization requiresfailure of the mass,sufficient
cial case of gravity-drivensettling,in which buoyancy water to saturate the mass, and sufficient conversion of
and drag dominatesolid-fluidinteractionforces,an em-
gravitationalpotentialenergyto internal kinetic energy
piricalformuladevelopedby Thomas[1965]predictsthe
to changethe styleof motionfrom slidingon a localized
viscosityof suspensions with diverseconcentrationsrel-
failure surfaceto more widespreaddeformationthat can
atively well [Polettoand Joseph,1995]. This formula
be recognizedasflow. These three requirementsmay be
reducesto Einstein'sequationin the low-concentration
limit and has the form
satisfiedalmost simultaneously,and the mechanicsof
mobilizationare understoodmoderatelywell [Ellen and
ix/ixw: 1 + 2.5Urines
+ 10.05V2fines Fleming,1987;Andersonand Sitar, 1995].Iversonet al.
[1997] discussthe mechanicsof mobilizationin detail,
+ 0.00273 exp (16.6X)•nes) (5) whereasthe followingdiscussionsummarizesonly some
Among the shortcomings of this and similarformulasis rudiments.
the neglectof shearrate effectsthat are especiallypro- Debris flowscan result from individualslopefailures
) 0.4, if graingeometriesdeviategreatly or from numerous small failures that coalesce down-
nouncedif 1)fines
from spheres,or if physicochemical influencesof Van stream. In exceptionalcases,failure can occur almost
der Waals or electrostaticforcesbetween clay and col- grain by grain, as it might during sappingerosionor
256 ß Iverson' PHYSICS OF DEBRIS FLOWS 35, 3 / REVIEWS OF GEOPHYSICS
analyses
of failedslopesindicatethat the well-kriown
Coulomb criterion adequately describesthe state of
stresson surfaceswhere frictional failure occurs[e.g.,
Lambe and Whitman, 1979]. In its simplestform, the
Coulombcriterion may be expressedas
= -p) tan (b+ c (6)
Here ß is the average shear or driving stressOn the

failure surface,and the resistingstrengthdependson the
averageeffectivenormal stress((r - p), bulk friction
angle•, and cohesionc on the samesurface.The bulk
frictionangledependson the frictionangleof individual
grains,(•g, and alsoon the packinggeometryof the
assemblage of grainsal6ngthe failuresurface.
During
failure, cohesivebondsare graduallybroken,sothat c •-
0 obtainsin failed zonesof even clay-richsoils[Skemp-
ton, 1964, 1985]. Thus as failure proceeds,(• and the
effectivestress,here definedsimplisticallyas the differ-
ence of the total ½ompressive normal stress(r and pore
fluid pressurep [cf.Passmanand McTigue,1986],deter-
minetheresistance
to motion.Thevalueof (• might
changesomewhatas grainsrearrangeduringfailure [cf.
Hungr and Morgenstern, 1984;Hanesand Inman, 1985],
but changesin effectivestressdue to stressfield rotation
and pore pressurechangeare generallymore significant
[Sassa,1985;Andersonand Sitar, 1995].
In some debris flows the water necessaryto saturate
the masscomesfrom postfailuremixingwith streamsor
other surface water, but in most debris flows, all water
necessary for mobilizationexistsin the masswhen fail-
ure occurs.Indeed, many debris flows are triggered by
changesin pore pressuredistributionsthat result from
infiltration of rain or snowmeltwater that precipitates
slopefailure [e.g.,Sharpand Nobles,1953;Sitar et al.,
1992]. To aid mobilizationin these circumstances, the
debrismay contractasfailure proceeds[Ellenand Flem-
ing, 1987]. Contraction producestransient excesspore
pressures that help weakenthe massand enhancethe
transformationfrom localizedfailure to generalizedflow
[Bishop,
1973;Iverson
andMajor,1986;Eckersley,
1990;
Iversonet al., 1997]. Contractionduring failure has tra-
Figure 8. Photographsof advancingfronts of debris flow ditionallybeenregardedas atypicalof naturaldebris
surges.(a) Nojiri River, Kagoshima,Japan, September10, becauseonlyverylooselypackedsoilsexhibitcontractive
1987.Flowis about20 m wideand2-3 m deep.(photocourtesy behavior during standardlaboratory compressiontests
JapanMinistryof Construction.)(b) JiangJia Ravine,Yunnan, [cf. Casagrande,
1976;Sassa,1984;Andersonand Sitar,
China,June24, 1990.Flowisabout12m wideand2•3 m deep 1995]. However, recent experimentationhas shownthat
(photocourtesyK. M. Scott.)(c) USGS debrisflowflume,July even dense soils may undergovolumetric contraction
20, 1994. Flow is about 4 m wide and 0.2 m deep.
duringfailure that occursin an extensionalmode [Vaid
and Thomas,1995]. Extensional(active Rankine state)
failure doesindeed occurduring mobilizationof exper-
sedimentimpactby a waterjet [Johnson, 1984].Failure imental debrisflows [Iversonet al., 1997], and contrac-
on all scales,from singlegrainsto great landslides,is tion of water-saturateddebrisduring extensionalslope
resistedprimarilyby strengthdue to grain contactfric- failuremightthusexplainthe apparentenigmaof debris
tion [Mitchell,1978]. Cohesivestrengthdue to soil ce- flowsthat mobilizefrom hillslopedebristhat is relatively
mentationor electrostaticattractionof clay particles dense[cf.Ellen and Fleming,1987].
maybeimportant
in somecircumstances,
however.
Re- Transformationfrom localizedfailure to generalized
sults from experimentalsoil and rock mechanicsand flow might occurwithout debriscontractionif sufficient
35, 3 / REVIEWS OF GEOPHYSICS Iverson' PHYSICS OF DEBRIS FLOWS ß 257
energyis availableto agitate the failing mass.This type Rock spray

of transformationcan occurin dry granularmaterialsas
arrives
-• f Flow.
front
a rrlves
well as debrisflows [Jaegerand Nagel, 1992;Zhang and

0.20
I--Debris-flow
depth
Campbell,1992].For example,a landslidethat becomes
agitatedand disaggregated as it tumblesdown a steep 0.15[-
o.o ]
• I
slopecan transforminto a debrisflow if it containsor
acquiressufficientwater for saturation.Some of the
largest and most devastatingdebris flows originate in
0.05
o
thismanner[e.g.,Plafkerand Ericksen,1978;Scottet al.,
1995]. -0.05 , I , I
8ooo
/I'- Total
' basalnormalstress
3.3. Debris Flow Motion
Followingmobilization,debrisflowsappear to move
like churningmassesof wet concrete.The largestflows
can transportboulders10 m or more in diameter.How-
4000
2000
f
ever, large-scaleexperimentaldebrisflows (Figure 3) o
that contain clasts no larger than 5 cm in diameter -2000
exhibit the same qualitative features as larger natural
3000
flows.Theseexperimentalflowsyield the most detailed, - Bas•l
pore
pressure I -
quantitativedata (e.g., Figure 5) and provide the best
Do• 2000
_
evidence for much of the behavior summarized here.

• • 1000 • i -
Virtually all debrisflows move downslopeas one or
more unsteadyand nonuniform surges.Commonly, an n_Z 0
abrupt bore forms the head of the flow, followedby a -1000 ,t I , I I rll ,
12.75 13.25 13.75 14.25 14.75
graduallytaperingbodyand thin, more waterytail [e.g.,
Pierson,1986; Takahashi,1991] (Figure 8). When mul- TIME, IN SECONDS
tiple surgesoccurin individualdebrisflows,eachexhib-

its a conspicuous head and tail [Jahns,1949;Sharpand Figure 10. Data from Figure 5a plottedon an expandedtime
Nobles,1953;Pierson,1980;Davies,1988, 1990]. Graphs base to show details during arrival of the debris flow front.
Note that a sprayof tumblingrocksprecedesthe arrivalof the
of flow depth or dischargeversustime illustrate the
flow front, whichoccursat t • 14 s. Pore pressuresdo not rise
generallyirregularcharacterof thesesurges(Figure 9)
appreciablyuntil the deepestpart of the flow front passesthe
[Takahashi,1991; Khegai et al., 1992; Ohsumi Works measurement cross section.
Office, 1995]. Observationsduring experimentsat the
USGS debris flow flume show that surges can arise
spontaneously, without extraneousperturbationsof the
flow. The resultinglow-amplitudesurfacewavesresem-
1000
ble roll wavesthat form in open channelflows of water
on steepslopes[e.g.,Henderson,1966]. In experimental
800 debris flows, larger waves tend to overtake and canni-
balize smallerwaves,as may be anticipatedfrom kine-
600 matic wave theory [Lighthilland Whitham,1955]. Con-
400
sequent coalescenceof wave fronts can produce a
200 sequenceof large-amplitudesurges,which may them-
o0 60 120 180 240 300 360 selvesbecomeunstable.Although other processes,such
as transientdammingor episodicslopefailures, might
200
also generate surges[Jahns,1949], intrinsicflow insta-
150
bility and wave coalescencesuffice.
100 Heads of debris flow surgeshave several distinctive
attributes[Takahashi,1991]. Pore fluid pressuresmea-
50 sured at the base of surge heads are close to zero,
o whereas fluid pressuresin the flow bodybehindthe head
600 1200 1800 2400 3000 3600
commonlyapproachor even exceedthosenecessaryto
TIME, IN SECONDS
balancethe total normal stressand liquefy the sediment
Figure 9. Measurementsof debris flow discharge,which il- mass(Figure 10). Surgeheadsgenerallycarrythe great-
lustrate multiple surgeswithin a flow event. (a) Data from est concentrationof large sedimentclastsand incidental
Name River, Japan.(Redrafted from Takahashi[1991];copy- items, such as downed trees, mangled bridges, or dis-
right A. A. Balkema).(b) Data from ChemolganRiver, Kaza- tressedautomobiles.The heads appear to remain rela-
khstan[Khegaiet al., 1992]. tively dry and to restrain downslopeflow of the more
258 ß Iverson: PHYSICS OF DEBRIS FLOWS 35, 3 / REVIEWS OF GEOPHYSICS
fluid, water-saturated debris that follows. Pore fluid does ular media (and debrisflows)cannotmimicthe abilityof
not escapevisibly by draining through surge heads in a gas to maintain constantagitation and flow resistance
moving debris flows, even though the permeability of in the absenceof energyexchangewith the environment.
headsmay be great owing to the concentrationof large Instead, granular temperature requires bulk deforma-
clasts. tion and dependson flow interaction with boundaries
Large clastsaccumulateat surgeheadsby two means: that impart externalforces.Granular temperaturesand
they can be incorporatedand retained there if the flow boundaryforcescannotbe specifiedindependentlybut
acquiresthe clastsin transit, or they can migrate to the mustbe determinedhand in hand as part of rigorous
head by preferential transport.Migration and retention mathematicalmodels[Hui and Haft, 1986].(2) As gran-
of large clastsappear to result chieflyfrom kinetic siev- ular temperature increases,stressesand flow resistance
ing similar to that describedby Middleton [1970]. In becomeincreasinglyrate dependent.At highergranular
kinetic sieving, selective entrainment or transport of temperaturesthe massactsmore like a fluid and lesslike
large clastsoccursbecausegravityand boundarydragdo a frictional solid. (3) Formal applicationof kinetic the-
not sufficeto force the claststhrough small voids that ory to granular media resultsin severelymathematical
open and closeas the agitateddebrisdeforms.As small formulations[e.g.,Lun et al., 1984],whichhavenot been
grainstranslocatethroughvoids,largegrainsaccumulate adapted to inertial flows of solid-fluidmixturessuchas
as a residue near the flow surface and snout. Both debrisflows,althoughGarcia-Aragon [1995]hasinitiated
physical and numerical experiments demonstrate the work along these lines.
efficacy of kinetic sieving in dry granular materials Granular temperature not only plays a key role in
[Bridgwateret al., 1978;Rosatoet al., 1987;Savage,1987; kinetic theories but also indicates whether the instanta-
Vallance,1994],whereaslittle experimentalor theoreti- neous,collisionalgrain interactionspostulatedin such
cal evidence supportsan alternative, dispersivestress theoriesare an appropriateidealization.Many dry gran-
mechanismproposedby Bagnold [1954] [Iversonand ular flows involve enduring,frictional grain contactsas
Denlinger,1987;Vallance,1994].Nonetheless,grain size well asbrief graincollisions[Drake,1990;Walton,1993],
segregationmechanismsin debrisflowsmay be compli- and even the most advancedtheoreticaldescriptionsof
cated and may involve more than one process[Suwa, thesetypesof flowsare relativelyrudimentary[Anderson
19881. and Jackson,1992]. Enduring frictional contactsneces-
Agitationof flowingdebrisinfluencesnot onlykinetic sarilyexistduringat leastpart of a debrisflow'sduration,
sievingbut alsothe bulk densityof the debris(Table 2) for contacts must be sustained as a flow mobilizes from
and the ability of grains to move past one another. a static massor forms a static deposit [cf. Zhang and
Improvedunderstandingof the influenceof agitationon Campbell, 1992]. Moreover, at any instant, part of a
the mobility of flowing granular materials has consti- debris flow may move in a collision-dominatedmode,
tuted a major advanceof the last2 decades[e.g.,Savage, whereas other parts may be friction dominated. The
1984;Campbell,1990;Jaegerand Nagel, 1992].The role relative importance of collisional,frictional, and fluid-
of agitation can be characterizedby defining instanta- mediated grain interactionsis a central problem of de-
neous grain velocitiesvs as the sum of mean •s and bris flow physicsand is analyzedin sections4, 7, and 8.
fluctuatingv} components.The intensityof fluctuations Stressmeasurementsat the bases of experimental
and degreeof agitationis then measuredby a mechan- debris flows at the USGS flume provide compelling
ical quantity that has come to be known, following evidenceof nonzerogranulartemperatures.Both Fig-
Ogawa[1978],astheg?anular
temperature
T. The gran- ures 5 and 10 show fluctuations in total normal stress
ular temperaturemay be interpreted as twice the fluc- associated
with grain agitation,althoughthe fluctuations
tuationkineticenergyper unit massof granularsolids are difficultto interpretbecausea large sensingelement
and defined as (500cm2)measured
theaveraged
effects
ofmany(- 10s)
simultaneousgrain interactions.However,contempora-
T = (v;2) = (Vs- rs)2) (7) neousmeasurements
witha 1-cm
2 sensing
elementre-
veal stressfluctuationsat a length scalecloseto that of
where angle brackets denote an appropriate average the largest grains (gravel) in the experimentaldebris
suchasthe ensembleaverage.The granulartemperature flow (Figure 11). Stressfluctuationsdetected by the
playsa role analogousto that of the moleculartemper- 1-cm 2 sensor, but not thosedetected by the 500-cm 2
aturein the kinetictheoryof gases[Chapmanand Cowl- sensor,had amplitudesas large as or larger than the
ing, 1970]. Like the moleculartemperatureof a gas, a mean stress.The presenceof theselarge-amplitudefluc-
higher granular temperature reducesbulk densityand tuations,which apparentlyresult from individualgrains
thereby enhancesthe ability of a granular massto flow. sliding,rolling, and bouncingirregularly along the bed
However, a higher granular temperaturealso requires and contactingthe sensor,indicatesthat the effectsof
higherratesof energydissipation,becausegrainvelocity boundaryslip on stressescan be substantial.If debris
fluctuationscauseinelasticgrain collisionsor inter- flowstranslatedsmoothlydownslopeas steady,laminar
granularfluid flow that dissipatesenergy.This energy flowswithout boundaryslip,no stressfluctuationswould
dissipationhasthree importantramifications:(1) Gran- occur.If stressfluctuationsresultedsolelyfrom fluctua-
35, 3 / REVIEWSOF GEOPHYSICS Iverson' PHYSICS OF DEBRIS FLOWS ß 259
500
tions in mean flow quantitiessuch as flow depth, the L ' I ' I ' I [ I ] [
magnitudeof the fluctuationsrelativeto the mean stress • 400
z
would not changewith the sensorsize. Granular stress o
fluctuationsnecessarilyare accompaniedby grain-scale • 300
pore fluid pressurefluctuations[Iversonand LaHusen, z 200
z
1989]. Furthermore, solid and fluid stressfluctuations
with time and length scalesmuch larger than those of r,.) 100
individual grain interactionsalso may occur owing to O
LL 0
developmentof interlockedgrain clustersthat move as
more or lesscoherent blocks.Theoretical results[Sheri -100
andAckerman,1982], computationalexperiments[Hop- 1.0
kins et al., 1993] and physicalexperiments[Iversonand • 0.8

B 1 cm2 plate
LaHusen,1989;Drake, 1990]all point to the existenceof z
o
such clusters.
• 0.6
Pore fluid pressure and granular temperature play z
synergistic
roles,asis indicatedby the debrisflow energy z 0.4
cascade described in section 2. The combined influence
0 0.2
of granular temperature and pore pressureon flow re- O
LL 0
sistanceappears to control debris mobility. In turn,
debrisflow motion generatesboth granulartemperature -0.2
I I
15 16 17 18 19 20
and nonequilibrium(nonhydrostatic) fluid pressures.A
critical distinction existsbetween the means by which TIME, IN SECONDS
granular temperaturesand nonequilibriumfluid pres-

Figure 11. Contemporaneousmeasurementsof bed normal
sures arise, however. Steady debris flow motion can
forceon (a) a 500-cm
2 plateand(b) a 1-cm
2 platemounted
produce and sustaingranular temperaturesby conver- flush with the flume bed. The scalefor force in Figure 11b is
sion of flow translational energy to grain fluctuation 500 times the scale in Figure 11a, so that the scaled force
energy,whereasanalogousconversionof flow energyto amplitude
isthesamein eachplot.The1-cm 2platewaslocated
fluid pressureenergyis problematic.The mostgenerous 0.5 m downslope
fromthe 500-cm 2 plate,producinga small
estimateof suchconversionassumesthat all thermody- time lag betweenthe measurements
in Figures11a and 11b.All
namic heat generatedby debris flow motion produces data were sampledat 2000 Hz during the debris flow flume
fluid pressure.Then a typicaldebrisflow heatingrate of experimentof August31, 1994 (see Figures5 and 10).
0.005øCper meter of flow descent(seesection2) canbe
multipliedby the thermal pressurizationfactor for con-
finedwaterat 20øC(6 x 10sPa/øC)to estimate
3000Pa pressuresin debris flows, then debrisflows are funda-
or about 0.3 rn of excesspressurehead generatedper mentallyunsteadyphenomena,and limited light can be
meter of flow descent.However, such a generousesti- shedon the phenomenaby steadystate rheometricex-
mate neglectsthe fact that water at 20øCcan accommo- perimentsand theoreticalmodels.Anecdotal evidence
date a temperature increaseof 0.005øCby expanding suggeststhat this may indeed be the case;steadydebris
only 0.0000001%,with no attendant pressureincrease. flow motion is virtually never observedin nature, and
In the agitated,unconfinedenvironmentof a debrisflow, steadymotion of debris flow slurriesis notoriouslydif-
constraintson expansionare minimal, and substantial ficult if not impossibleto achievein experimentalappa-
thermal pressurizationof fluid thus appearsunlikely. ratus [e.g.,Phillipsand Davies, 1991;Major and Pierson,
Yet fluid pressures---o#h (roughlydouble the hydro- 1992]. Section7 providesa detailedmechanicalevalua-
staticporefluidpressure ---o/jh),highenough to liquefy tion of hypotheticalsteadymotion, and section8 shows
the sedimentmass,are common at the base of experi- why unsteadydebris flow motion appearsmore viable
mental debrisflows(Figures5 and 10). Sustainedhigh mechanically.
fluid pressuresreduce intergranularfriction and influ-
encegrain collisionsassociated with high granulartem- 3.4. Debris Flow Deposition
perature.Understandingthe origin and effectsof high Depositionconstitutesa specialcaseof unsteadyde-
fluid pressuresappearsvital to understandingdebris bris flow motion. Deposition occurswhen all kinetic
flow behavior. energy degradesto irrecoverableforms. Complete en-
An obviouspossibilityis that sedimentconsolidation ergydegradationoccursfirstwhen granulartemperature
produces high pore pressuresin debris flows [cf. falls to zero in the coarse-graineddebristhat collectsat
Hutchinson,1986].However,consolidationrequiresthat debrisflow snoutsand lateral margins,where leveesmay
the debriscontractmonotonically,a conditionthat can- form. This coarse debris consequentlycomposesthe
not be sustainedin steadydebrisflow motion. Thus the perimeterof mostdebrisflow deposits(Figure 12). De-
debrisconsolidationhypothesisis, at once,both routine posited coarse debris lacks high pore pressuresand
and radical. If consolidationgives rise to high pore typicallyformsa dam that impedesand eventuallyhalts
(Figure 13) [Major, 1996] [cf. Hampton, 1979;Pierson,

1981].Subsequentdecayof interior pore pressures,with
attendant consolidation(i.e., gravitationalsettling) of
the solidsand drainageof fluid, marksthe final stagesin
a debris flow's transition from fluid-like to solid-like
behavior.
The timescalefor pore pressuredecayis defined by
the quotient of a pore pressurediffusioncoefficient,
D - kE/ix, and the squareof the characteristicdrainage
path length.Here E is the compositestiffness(reciprocal
of the compressibility)of the debrismixture. Measure-
mentsand modelingby Major [1996] showthat drainage
is dominantly vertical in typical debris flow deposits,
which have lengthsand widths that greatly exceedtheir
thicknessh. Thus the characteristicdrainagepath length
is h, which yieldsthe pore pressurediffusiontimescale
tdif = h21x/kE (8)

Becausehigh pore pressureshelp sustaindebris mobil-
ity, it is temptingto equate the diffusiontimescaletdiff
with the debris flow duration or timescale for which
mobilityis sustained,tt>[Hutchinson,1986].Three prob-
lemscomplicatethis interpretation,however.First, pore
fluid pressuresare but one phenomenonthat influences
mobility;debrisflow mixturescan flow in the absenceof
high pore pressureif they have sufficientgranulartem-
perature.Second,nonequilibriumpore pressuresmaybe
smallor absentat the front of debrisflow surges(Figures
Figure 12. Photographsof snoutsof debris flow deposits,
whichshowconcentrationsof coarseclastsand bluntlytapered 5 and 10), so that pore pressurediffusion is locally
marginmorphology.
(a) Lobeof a small(---1000m•) debris irrelevant.Finally, the definition of tdiffincludesa com-
flow that partly crossedthe scenichighwaynear BensonState positestiffnesscoefficientE, whichhasthe propertiesof
Park, Oregon, February 7, 1996. (b) Vertical crosssection an elasticmodulusin small-strainproblems[Biot, 1941]
througha marginallobe of an experimentaldebrisflow at the but which has more complicatedpropertieswhen defor-
USGS debris flow flume, October 8, 1992. mations are large and irreversible [e.g., Helm, 1982].
Section8 addressesthis issuequantitativelyand shows
how the evolvingcompressibility of debrisflow materials
the motion of ensuingfiner-graineddebris that retains can influencepore pressurediffusionand play a key role
higher pore pressures.Alternatively, wetter, more mo- in debrisflow physics.
bile debrismay have enoughmomentumto overrideor
breach the dam of previouslydepositeddebris, so that
depositscan developby a combinationof'forwardpush-
ing, mass "freezing," vertical accretion, and lateral 4. MOMENTUM TRANSPORT:
shuntingof previouslydepositedsediment[Major, 1997]. SCALING AND DIMENSIONAL ANALYSIS
Experimentalobservationsof idealized debrismixtures
indicatethat freezing generallyoccursfrom the bottom To build a quantitativebackgroundfor analyzingde-
up, rather than the top down, as debris comesto rest bris flow physics,it is usefulto ignoretemporarilysome
[Vallance,1994].Thus neither the thicknessof deposited of the complexitiesdescribedin the precedingsection
lobesnor that of leveesprovidesa goodindicatorof the and considermomentumtransportduringsteady,simple
dynamicbehaviorof the movingdebris as a whole [cf. shearingof an unbounded,uniform mixtureof identical,
Johnson,1970]. Instead, the complexinterplay between dense spherical grains and water. Initially restricting
the resistanceof the first-depositeddebris and the mo- attention to an unbounded domain and a single grain
mentumof subsequently arrivingdebrisproducesdepos- diameter b vastlysimplifiesthe analysis,becauseit un-
itsthatardinitiallydryandstrongat theirperimeter,wet ambiguouslyestablishesthe dominantlength scaleas b.
and weak in their interior, and conspicuously heteroge- Scalingconsiderations then sufficeto draw rudimentary
neousin their resistanceto motion. Indeed, pore fluid conclusionsabout momentum transport and the atten-
pressuresin the centerof a depositcan remain elevated dant state of stress in the mixture. Associated dimen-
well abovehydrostaticlevelsand maintain the sediment sional analysisdefines dimensionlessparameters that
in a nearly liquefied state long after depositionoccurs can be usedto classifydebrisflowsand identify limiting
35, 3 / REVIEWSOF GEOPHYSICS Iverson:PHYSICSOF DEBRISFLOWS ß 261
SAND-GRAVEL MIX 04-19-94 LOAM-GRAVEL MIX 08-31-94

5000 ........• ........• ........• ........• ......., ............ • ........• .......• ........• ........• ........
- Deposit
thickness - Deposit
thickness
-
4000 0.12m
i 0.115
m :
3000 --[ -_
2000 --
1000 i•
0 .I•..M•,•,,.•• -
0.1 1 10 102 103 104 105 0.1 1 10 102 103 104 105
TIME, IN SECONDS TIME, IN SECONDS
..............
Total normal stress
Fluid
pressure
Figure13. Measurements
of totalbasalnormalstress
(on a 500-cm
2 plate)andbasalporepressure
during
depositionof debrisflow sedimentswith differentgrain sizedistributions
at the USGS debrisflow flume.
Measurements were madethroughportsin the runoutpad at the flume base(Figure 4), and depositswere
centeredover the measurement ports.Depositinteriorswere liquefiedby highpore pressureat the time of
emplacement, and pore pressuressubsequently decayed.High pore pressures persistedmuchlongerin the
depositthat containedloamwith about6% (by weight)silt and clay-sized particlesthan in the depositthat
lackedloam and containedabout2% (by weight)silt and clay-sizedparticles[afterMajor, 1996].
stylesof behavior.The multiplicityof relevant dimen- various stresses(solid grain shear and normal stress,
sionlessparametersalso revealswhy nearly intractable fluid shear and normal stress,and solid-fluid interaction
problemsarise in attemptsto "scaledown" debrisflow stress)that accompanymomentumtransportin the mix-
mixturesto the sizeof laboratoryapparatus.Suchscaling ture are representedcollectivelyby 5;. Adapting the
problemsmay partly explainwhy very divergentviews approachusedby Savage[1984]for drygrainflows,these
about debris flow physicshave arisen from different stressesare postulatedto depend functionally on the
approachesto experimentationand modeling(see sec- mixture shear rate 4/ and on 12 additional variables
tion 5). discussed in section 3 and listed in the notation section:
Figure 14 depictsschematicallya representativere-
gion within a uniform grain-watermixture undergoing Z = •(•/, a, ps,phg, !•, k, T, E, vs,vf, d),e) (9)
steady,uniform shearingmotion in a gravityfield; the Variables not included in (9) might influence stresses
alsobut are assumedto havelessimportancethan those
included.
As a preliminary step, dimensionalanalysisreorga-
nizes(9) into a morefundamentalandcompactrelation-
ship that involvesonly dimensionless parameters.The
first 10 variables in (9) have units comprisingthree
physicaldimensions:mass,length, and time. The last
four variablesin (9) are intrinsicallydimensionlessand
are superfluousin dimensionalanalysis.According to
the BuckinghamII theorem [Buckingham,1915], any
physicallymeaningful relation between 10 variables
comprisingthree dimensionsmust reduceto a relation
between7 (= 10 - 3) independentdimensionless pa-
rameters. Definition of these parameters depends on
choicesfor the characteristiclength,mass,and time. For
Figure 14. Schematic diagram of a steady, uniform, un- the simplesystemdepictedin Figure 14, the choicesare
boundedshearflow of identicalsolid spheresimmersedin a obvious:the characteristiclength is 8, the characteristic
Newtonian fluid. This flow is too simple to representdebris mass isps•3, andthecharacteristic
timeis1/4/.These,in
flows,but it providesa basisfor assessing scalingparameters turn, determine a characteristic
velocity v -- •/8, which
that influence stresses. describesthe speedat which grainsmove past one an-
262 ß Iverson: PHYSICS OF DEBRIS FLOWS 35, 3 /REVIEWS OF GEOPHYSICS
other and at which fluid movesto accommodategrain adjacentfluid increasesindependently.At depthN8 the
motion. With these choices,standard methods of dimen- quasi-staticsolid stressis estimatedby
sionalanalysis[e.g.,Bridgman,1922]appliedto (9) yield
rs(q)'• Nus(Ps
- p•)#• tan 4) (13)
__:
q/282ps•('•t28
g , q/82ps
• , ps T 82'
k q/2•-2ps
Pf, q/282' E) whereN, the numberof grainsaboveand includingthe
layer of interest, accountsfor the effectsof the overbur-
denload,andUs(Ps
- PT)#isthebuoyant
unitweightof
The right-hand side of this relation lists six dimension- this overburden.Additional(nonhydrostatic)
fluid pres-
less parameters that determine the dimensionless suremayalsomediateTs(q)but is characterized sepa-
stresses,
•/q/282ps.
The significance
of the firstright- rately(below)bythesolid-fluid
interaction
stress,
Ts_f.
hand-side parameter was first enunciated by Savage The quasi-staticfluid stressderives from Newton's
[1984], and accordinglyit has been dubbedthe Savage law of viscosity:
number [Iversonand LaHusen, 1993]. The secondpa-
rameter is a variationof a parameterfirst investigatedby rf(q): •f• • (14)
Baghold[1954], commonlycalled the Bagnoldnumber
[Hill, 1966]. The third parameter is the ratio of solid In thisequation,
•f appears
because
onlythisfractionof
the m•ture undergoesviscousshear.
densityto fluid density,which rangesonly from about 2
to 3 in debrisflows.The fourth parameteris the granular The solid-fluid
interaction
stressTs_f resultsfrom
relative motion of the solid and fluid constituents. Al-
temperature scaledby the square of the characteristic
shearvelocity4/8 [cf. Savage,1984].The fifth parameter thoughTs_f mayinvolvebothinertialand quasi-static
(viscousdrag)components, a detailedanalysisbyIverson
is the permeability divided by the grain diameter
[1993] showsthat viscouscoupling surpassesinertial
squared;it reflectsthe role that grain size and packing
couplingin materialssimilarto thosein debrisflows,and
play in solid-fluidinteractions.The sixth parameter is
that neglect of inertial couplingis generallyjustified.
the compositemixture stiffness(resistanceto dilation
Viscouscouplingresultsin drag that generatesa force
and contraction) divided by the characteristicstress
per unit volumeof m•ture • v(•/k), whichproducesa
q/282ps.
stress•vS(•/k). Thus, expressedin termsof the shear
The significanceof the parametersin (10) can be
rate • = v/8, the interaction stresscan be estimated as
clarified by analyzingtheir relationshipto estimatesof
solid, fluid, and solid-fluid interaction stresses in the
mixture. These stresses have both shear and normal
components;in turn, each of these componentsmay
have both quasi-staticand inertial components.For This interactionstressresultsfrom grain-scalefluid flow
brevity,this analysiswill focusexclusivelyon shearcom- driven by grain rearrangementsduring steadyshearing
ponentsof stress,which are generallyof greatestinter- motion at the rate 4/[cf. Iversonand LaHusen, 1989].If
est. A similar analysisis easilyconductedfor the normal motion were unsteadyand net volume changewere to
stress components. occur, an additional viscous interaction stress would
The solidinertialstressTs(i)andfluidinertialstress arise in concertwith net pore pressurediffusion(see
TT(i)bothscalelike the productof the mass(solidor discussion followingequation(8)).
fluid) per unit volume and the squareof the character- The chief significanceof (11)-(15) lies in the ratios
isticvelocity,
v2 ---q/282. Thustheymaybeestimated
by that they form. For example,divisionof the character-
isticstressTs(i)by Ts(q)showsthat a Savagenumber
rs(i) .• •)spsq128
2 (11) Nsav (here modified to accountfor the solid friction
angle,overburdenload, and hydrostaticbuoyancy)may
rf(i).---.
13fpf'•t282 (12) be definedby the ratio of inertial shearstressassociated
with grain collisionsto quasi-staticshear stressassoci-
The first of theserelationshipsshowsthat the character- ated with the weight and friction of the granularmass
isticstressusedto scale2; in (10) is essentially the solid
grain inertia stress.This is the stresstransmittedby grain
collisions[cf.Iversonand Denlinger,1987]and explicated Nsav
=N(ps
- p•)
#tan4) (16)
by Baghold[1954]. The secondrelationshipshowsthat
fluid can also sustain inertial stresses, in a mannerSimilarly,
division
of Ts(i)byTf(q)shows thata Bagnold
roughlyanalogousto that of Reynoldsstressesin turbu- number NBag may be defined by the ratio of inertial
grain stressto viscousshear stress:
lent flow of pure fluid. The fluid-inertia stresswas ig-
nored by Baghold[1954].
The quasi-static
solidstressTs(q)is associated
with NBag •)s
= 1-- sps82q/
19 • (17)
Coulombslidingand enduringgraincontacts(seeequa-
tion (6)). This stressincreasesasdepthbelowa horizon- wherein the factor rs/(1 - rs) resultsfrom the substi-
tal datum increasesbut decreasesif staticpressurein the tution
vf - 1 - vs anddiffers fromthefactorX•/2 =
- us )] originallyusedby Bagnold[1954] solid-fluid interactions. At least this many processes

(ii, is the maximumvalue of Ils achievablein a dense- must affect stressesin debris flows with more compli-
packed configuration). Division of Ts(i)byTf(i)producescated constituentsand kinematics [Iversonand Den-
a "massnumber"Nmassthat describesthe ratio of solid linger, 1987]. It thus appearsunlikely that any simple
inertia to fluid inertia in the mixture: rheologicalmodel can accuratelyrepresentall stressesin
debris flows. Nonetheless,simple but valid approxima-
Ils Ps
tions may be attainable if in some circumstancesonly a
Nmass
= I - IlsPf (18) subset of these stresses dominate.
Rough but useful assessments of the relative impor-
Divisionof rs_f by Ts(i) producesa quantityhere tance of different stressgeneration mechanismsin de-
termed the "Darcy number,"
bris flowscan be accomplishedby calculatingrepresen-
tative valuesof the dimensionlessparametersdefined in
NDa
r= IlsPs4lk (19) (16)-(21) and comparingthese values with those for
simpler systemsin which stressgenerationis better un-
which describesthe tendency for pore fluid pressure derstood. This processis analogousto assessingopen
developedbetweenmovinggrainsto buffer grain inter- channelflow of water on the basisof Froude and Reyn-
actions.
Furthermore,division of NDar by E/(•/2•2ps ) oldsnumbers.Table 5 listsvaluesof (16)-(21) computed
(whichappears
in (10)),yields(4ll,82)/(ilskE),a dimen- for a representativespectrumof debris flows, ranging
sionlessparameter that describesthe ratio of the time- froma 10-m 3flowin theUSGSflumeto theprehistoric
scalefor diffusivepore pressuredissipationacrossthe Osceola mudflow of '"109m3 (seeTable1). Valuesof
distance8 to the timescalefor pore pressuregeneration some variables need to be estimated to make these
by grain interactions, 1/4/ [cf. Iverson and LaHusen, computations.For example,Table 5 lists a fixed • of 1
1989].The diffusiontimescale !,82/kE resembles the mm for all flows. Although all flows listed in Table 5
consolidation timescaledefinedin (8). The chief differ- consistedpredominantly of grains of about this size
enceisthat!,82/kEcharacterizes grain-scale
diffusion in (sand),larger grainsmight be more significantin some
a flow that is macroscopically steady,whereas(8) char- instances(as in a boulder- or cobble-richdebris flow
acterizesdiffusionduringunsteadyconsolidation through- snout). Nonetheless,the values of (16)-(21) listed in
out the entire debris flow thickness. Table 5 provide some idea of the range of values for a
Additional dimensionlessparametersof interest can variety of debris flows.
be obtainedby formingratiosof the parametersdefined Althoughinterpretationof the valuesof (16)-(21) in
in (16)-(19). For example,a versionof the well-known Table 5 is limited by a dearth of relevant data, some
grain Reynoldsnumber can be expressedas guidelinesexist.For example,Savageand Hutter [1989]
reviewed a variety of experimental evidence and con-
NBag Pf•t•2 cluded that grain collision stressesdominate grain fric-
NRey
--Nmass
- [z (20)
tion stressesin dry granularflows if Nsav is greater than
about 0.1. Similarly,Bagnold's[1954] experimentsdem-
and the ratio of the Bagnoldnumber to Savagenumber
onstratedthat in neutrallybuoyantmixturesof spherical
forms a version of the friction number identified by
grainsand liquid (whereNsav --->oo),collisionalstresses
Iversonand LaHusen [1993]:
dominateviscous
stresses
if NBagexceeds
roughly200.
NBag Ils N(ps- p•)#8 tan 4) (This differsfromBagnold's[1954]valueof 450,because
Nfric
= Nsa
v 1- Ils 4/ix (21) Bagnold
included
thefactorX•/2ratherthanIls/(1- Ils)
in his evaluation.) Apparently no experimental data
This numberexpressesthe ratio of shearstressborne by bearing on transition values of Nmassare available, al-
sustainedgrain contactsto viscousshearstress.It resem- thoughthe qualitativeinfluenceof Nmass is obviousfrom
bles the well-knownBinghamnumber, which describes its definition: grain inertia becomesunimportant as the
the ratio of stressborne by shearstrengthto stressborne densityor concentrationof grains approacheszero. In
by viscousflow in viscoplasticmaterials.The key differ- contrast,many data pertain to grain Reynoldsnumbers,
encebetweenthe friction number and Binghamnumber NRey.Typically,
fluidflowwithrespectto grainsbegins
lies in the fact that the friction number characterizes to show inertial effects and deviate significantlyfrom
stresses borneby distinctsolidand fluid phases,whereas idealviscous
(Stokesian)
behavior
forNRey> 1 [Vanoni,
the Bingham number characterizesstressesin a one- 1975]. Fewer data are available for Nfric and NDar, al-
phasematerial that exhibitsboth viscosityand strength. though Iverson and LaHusen [1989] reported experi-
The dimensionless groupsdefinedby (16)-(21) dis- ments with 1000 < NDar < 6000, in which large fluid
tinguish five processesof momentum transport (i.e., pressurefluctuationsevidencedstrongsolid-fluidinter-
stressgeneration)in a steadyshear flow of a uniform actions.Values of NDar at least this large probablyapply
mixture of identicalgrainsand water: (1) inertial grain for most debrisflows(Table 5).
collisions,(2) graincontactfriction,(3) viscousshear,(4) With these guidelinesfor interpretation, the tabu-
inertial (turbulent) fluid velocityfluctuations,and (5) lated values of dimensionlessparameters in Table 5
264 ß Iverson- PHYSICS OF DEBRIS FLOWS 35, 3 / REVIEWSOF GEOPHYSICS
TABLE 5. Estimation of DimensionlessParametersThat CharacterizeStressesin a Range of Prototypical

Debris Flow Mixtures
DebrisFlow Prototype
USGSFlumeExperiment Oddstad DebrisFlow SouthToutleRiver, OsceolaMudflow,
Parameter (Sand-Gravel) (Figure1), Jan.4, 1982 May 18, 1980 circa5700B.P.
Dimensional Parameters
8, m 0.001' 0.001' 0.001'* 0.001'
h = NS, m 0.1 1 5 20
v, m/s 10 10 20 20
4/, 1/s 100 10 4 1
Ps, kg/m3 2700 2700 2700 2700
pf,kg/m3 1100 1100 1100 1200
ix, Pa s 0.001 0.01' 0.01' 0.1'
g, m/s2 9.8 9.8 9.8 9.8
k, m2 10-• 10-•* 10-•2 10-•2
E, Pa 104* 104* 104* 104*
vs 0.6 0.6 0.6 0.6
vf 0.4 0.4 0.4 0.4
q>,deg 40 30 30 30
Dimensionless Parameters
Nsav 0.2 2 X 10-4 6 X 10-6 i X 10-7
NBag 400 4 0.2 0.4
N .... 4 4 4 4
NDar 600 60,000 2 X 106 6 X 107
NRey 100 i 0.04 0.0l
Nfric 2 X 103 2 X 104 3 X 104 4 x l0 s
Data sourcesfor approximaterangeof valuesof dimensional
parametersare as follows:h, v, and 4/,kinematicreconstructions
and direct
observations
(s.eereferences
citedin Table1); pf,Ps,ix,k, Us,andvf, Table2, Table4, andFigure6 of thispaper;8, E, andqb,Figure20 of
thispaperanddatafromMajor[1996].In all casesthetypicalshearrate4/isestimatedfromthequotientof thetypicalflowspeedv anddepth,
h.
*Valuesof dimensional
parameters
for whichthe tabulatedvaluemayvaryor err by morethanan orderof magnitude.
paint a reasonablyconsistent
pictureof the factorsapt to
transport(suchas thosedescribedin sections5 and 9)
influence stressesin debris flows. For thin, fast flows on
canbe placedin an appropriatecontext.
steepslopes(e.g.,flowsat the USGS debrisflowflume), In principle,the valuesof key dimensionless param-
high shear rates cause both the Savagenumber and eters also facilitate discrimination of debris flows from
Bagnoldnumber to be moderatelylarge; however,the related phenomena.For example,by selectingthe pa-
tabulated
values
Nsav = 0.2andNBag= 400approximate
the respectivetransitionvalues0.1 and 200. Thus grain
collisionsmight be expectedto transmit most stressin
suchflows,but frictionandviscosityalsomay contribute
current;
significantly.For larger flows with greater depths and Muddy
waterflood
smaller shear rates, the situation is more clear-cut. Small
valuesof Nsav andNBag indicatethat collisions
likely Dryrock
avalanche
transmit negligiblestressin suchflows and that friction
andviscositydominate.Largevaluesof the frictionnum-
ber suggest that frictionalshearstressesprobablyexceed
viscousshearstresses, but smallgrainReynoldsnumbers
and largevaluesof NDar indicatethat viscousdragasso- Npar
ciatedwith solid-fluidinteractionsis likely to be impor-
tant. The picturechangesin partsof debrisflows(such Earthflow
as heads of surges)where grains coarser than sand

predominate.If shearrates are constantand 8 increases, Figure 15. Classificationschemethat distinguishesdebris
flowson the basisof valuesof the dimensionless
parameters
friction increasinglydominatesviscosity,but collisions
Nsav,NBag
, andNDar. Debrisflowsoccupy
a broadandimpre-
increasinglydominate friction. Thus individual debris ciselydefineddomain(shaded)in the centerof thisparameter
flows may include regionswhere different momentum space.
If oneor moreof theparameters
Nsav,NBag
, andNDar
transportprocesses dominateor whereseveralprocesses hasa valueeitherverylargeor small,debrisflow maygrade
contributealmostequally.With this knowledge,models into other typesof sedimenttransportprocesses,
as indicated
that include only one or two processesof momentum by labeled regionsin the diagram.
TABLE 6. Comparison of Models' Ability to Explain Physical Phenomena That Typify Debris Flows
Typeof Model
Coulomb Grain
BagnoM Flow With
Bingham Grain Variable Pore
Phenomenon l,qscoplastic Flow Pressure
Flow mobilizesfrom rigid slopefailure without changesin constitutiveproperties X

Fluid pressuresin flow can differ from the mean pressureand affect apparent X
strengthor flow resistance
Flow can exhibit a "rigid" plug of undeformingmaterial X X
Flow can lack a "rigid" plug of undeformingmaterial X X
Flow is unsteadyand nonuniform,with a blunt snoutand tapered tail X X X
Flow can transportlarge claststhat do not settle out X X X
Flow producesgrain size segregation
Flow agitation can affect apparent strengthor flow resistance X
Boundaryslip occursat the bed X
Flow strengthensand halts rapidlywhen pore fluid is drained from beneathit X
Deposit interior can remain weak and unable to supportloadswhile deposit X
perimeter becomesrigid
"X" denotephenomenathat can be explainedat least qualitatively.Assessmentof the Binghamand Bagnoldmodelsassumesthat th,e
standardforms as describedby Johnson[1984] and Takahashi[1981] apply.The model of Coulomb grain flow with variable pore pressureis
describedin section9 of this paper.
rametersNsav,NBag
, andNDar asthosemostlikelyto 5. TRADITIONAL (RHEOLOGICAL) MODELS
vary significantlyfrom flow to flow, a classification OF MOMENTUM TRANSPORT
schemesuchas that shownin Figure 15 can be devised.
In this schemevarious phenomena that can resemble Models of two distincttypes,viscoplasticand inertial
and transform to or from debris flows, such as rock grain flow, traditionally have provided the theoretical
avalanchesor turbidity currents,representlimiting cases framework for most debris flow research.Each type of
in whichoneor moreof theparameters
Nsav,NBag
, and model postulatesa unique rheologicalrelation between
NDar has a value that is either very large or very small. the shear stressand shear strain rate in flowing debris
The parameter spaceintermediatebetweentheselimit- mixtures.Suchpostulatesconflictwith data showingthat
ing casesincludesthe varietyof behaviorsthat constitute solid and fluid stressesin debris flows vary asynchro-
the processof debrisflow. At present,sucha classifica- nous!y(Figures5, 10, and 13) and with inferencesthat
tion has utility chieflyas a conceptualtool; it illustrates varyingpore pressuresand granulartemperaturesinflu-
the hybrid characterof debris flows and indicatesthat ence debrisbehavior.Consequently,this sectionavoids
debris flow behavior likely cannot be discriminatedon detailed review of traditional rheologicalmodels (pro-
the basisof a few simple measures,such as shear rate vided previously by Johnson [1984] and Takahashi
and solidsconcentration.A more rigorousinterpretation [1991]), and instead summarizestheir strengthsand
remainselusivebecausethe parameterspaceboundaries shortcomings. Table 6 comparesqualitativeattributesof
betweenvariousprocesses identifiedin Figure 15 remain debrisflowsthat can be explainedwith traditional mod-
vaguelydefined. elsanda model
thatemphasizes
solid-fluid
interactions.
Although the appeal of simpledimensionalmethods Later sectionsof this paper provide a more quantitative
and classifications is clear, it is important to recognize perspective.
their limitations. Because the foregoing dimensional The first systematicefforts to developa physicalun-
analysisassumes veryidealizedkinematics(uniformsim- derstandingof debrisflowswere thoseof Johnson[1965]
ple shear flow), it neglectsvariationsin granular tem- and Yano and Daido [1965], who recognizedindepen-
perature and volume fraction, and it neglects energy dentlythat debrisflowsexhibitpropertiesof both viscous
conversionand dissipationthat necessarilyoccurat flow fluids and plastic solids.This marked a significantstep
boundaries.Perhapsmost importantly, it neglectsthat forward,becauseearlier, descriptivework did not clearly
debris flows virtually always include grains of widely distinguishthe mechanicsof debrisflowsfrom those of
rangingsizes,developpore pressuresthat exceedhydro- muddywater floods.Johnson[1965, 1970, 1984] adopted
staticvalues,and occuras unsteady,nonuniformsurges. the simplestmechanicalmodelthat combinesplasticand
Analysesmore sophisticatedthan simple scaling and viscousattributes: that of a Bingham, or viscoplastic,
dimensionalmethodsare therefore needed to develop continuum[cf. Bird et al., 1982]. This model describesa
better insight and appropriate models. The following single-phasematerial that remainsrigid or elasticunless
sections describe traditional and more recent ap- deviatoricstressesexceeda thresholdvalue, the plastic
proachesto this problem. yield strength.Where stressesexceedthe yield strength,
the material flows like a viscous fluid. At a stress-free grainsmay slide,collide and roll along flow boundaries.
surfaceof an open channelflow suchas a debrisflow, a Grains visible on the surfaceof debris flows may either
Bingham material translateslike a rigid solid. jostle energeticallyor lock togetherto form an appar-
As was recognizedby Johnson[1965, 1970, 1984], ently rigid plug, dependingon the granulartemperature,
Bingham modelsfor debris flows can be generalizedto which in turn dependson flow speed,composition,and
allowyield strengthto dependon Coulombfriction (and boundary conditions. Realistic models of debris flow
hence on the mean stress)and viscosityto depend on physicsneed to accountfor these phenomena.
deformationrate [Iverson,1985;Coussotand Piau, 1994, To accountfor grain interactions,Takahashi[1978,
1995], but applicationsof Bingham models to debris 1980, 1981] exploited the seminal findingsof Baghold
flows have almost invariably assumedfixed viscosities [1954] to developan inertial grain flow model of debris
and yield strengths[e.g.,Fink et al., 1981]. Most appli- flows.
Bagnold's
[1954]experiments
employed
an en-
cations have also assumedsteady, uniform flow. For closedannularshearcell to evaluatethe effectsof grain
example, steadystate balancesof driving and resisting interactionsin rapidly shearing,concentratedsuspen-
force have been used to infer fixed yield strengthsfrom sionsof uniform, solidspheresimmersedin a Newtonian
the thicknessof depositeddebris flow lobes [Johnson, fluid of identical density.From his experimentsand a
1984; Whipple and Dunne, 1992]. Bingham strengths simple analysisof binary grain collisions,Bagnold in-
treated in this manner are equivalentboth conceptually ferred that shearand normal stressesin the suspensions
andmathematicallyto Coulombstrengths(equation(6)) variedeither quadraticallyor linearlywith the shearrate,
in which 4) = 0 and cohesionalone controlsyielding. If depending
onthevalueofNBag.
Baghold
[1954]usedthe
this equivalencyhas a sound physicalbasis, Bingham terms "grain inertia" and "macroviscous,"respectively,
strengths should increase as the fines content of the to describe the regimes where quadratic and linear
debrisincreases,and small-scaleexperimentswith debris stress-strainrate behavior obtained. Subsequentshear
mixtures composedof only fine sediment and water cell experimentsby others largely confirmedBagnold's
indeed produce this behavior [Johnson,1970; O'Brien results [e.g., Savageand McKeown, 1983] and also
and Julien, 1988; Major and Pierson, 1992]. However, showedthat the dynamic friction angle relating shear
large-scaleflume experimentswith mixturesof predom- and normal stressesin rapidly shearinggranularmateri-
inantly sand, gravel, and water, with a fines content of als differed little from the staticCoulomb friction angle
only a few percent (comparableto most natural debris describedby (6) with c = 0 [e.g.,HungrandMorgenstern,
flow mixtures) show that increasedfines content de- 1984;Savageand Sayed,1984; Sassa,1985]. Takahashi's
creaseslobe thicknessand apparent strength,because [1978, 1980, 1981] influential contributionto debrisflow
the fines help sustainhigh pore pressuresthat reduce physicsinvolved applicationof Bagnold'sstress-strain
frictional resistanceand enhancelobe spreading[Major, rate relationsfor the grain inertia regime. Other inves-
1996]. This revealsa fundamentalshortcomingof fixed- tigators [e.g., Davies, 1986] advocatedBagnoM's[1954]
yield-strengthBingham models: such models simulate formulas for the macroviscousregime as a model for
the rheologyof the water-plus-finesfraction of debris- debrisflows.Unfortunately,use of Bagnold's[1954]for-
flow mixtures, whereas observationsand data show that mulas for either regime is problematic,for Bagnold's
interactionsof coarsesedimentgrainswith one another resultsreflect the specialconditionsof his experiments.
and with adjacentfluid stronglyaffectdebrisflow behav- Adoption of Bagnold's[1954] formulas as constitutive
ior [cf. Costa and Williams, 1984; Major and Pierson, equationsfor general flow fields leads to contradictory
1992]. results.
Even if posedand used in a very generalform liver- Flow of a solid-fluid debris mixture in a channel
son, 1985, 1986a, b], Bingham models have significant enclosedby parallel,verticalplatesillustratesthe type of
limitations [cf. Johnson,1984]. They assumethat mo- contradictionsthat can arisein applyingBagnold's[1954]
mentum transportand energydissipationin debrisflows equations.Flow is driven by a longitudinalbody force,
occursexclusivelyby viscousshearing.They neglectthe suchas that due to gravity.Regardlessof flow rheology,
fact that rate-independentenergydissipationcan occur symmetrydictatesthat the mixture'sshearrate vanishes
when sediment grains contact one another or flow at the flow centerline(Figure 16). Bagnold'sequations
boundaries [e.g., Adams and Briscoe,1994], and they for both the graininertia and macroviscous regimesthen
neglectfluid flow relative to the granularassemblage.In require that the normal and shear stressalso vanish at
this respect,Binghammodelsrepresenta limiting type the centerline.However, a vanishingnormal stresscon-
ofbehavior inwhichNBag ---> oc(seeFigure tradictsthe presenceof the body force that drivesthe
0 andNDar --->
15), whichmay providean adequatedescriptionof phe- flow. Bagnold's experimentslacked this contradiction
nomena such as slow, creeping earthflowsbut not of becausegranular pressuregradientsdue to gravity or
debris flows. Bingham models also generally employ other forcesindependentof shearingwere absentin his
fluid-mechanicalno-slip boundary conditions.No-slip apparatus. In Bagnold's experiments with neutrally
boundariesrequirea Binghammaterialto leavea con- buoyantspheres,he intentionallycamouflagedthe effect
tinuouslayer of depositedsedimentalong its path, but of gravity on the solid grains, imposed the shear rate,
debrisflow pathscommonlylack suchdeposits.Instead, and measured the stress. In debris flows, in contrast,
gravityimparts a stressthat increaseswith depth below

the surface,and the shear rate responds.Shearing, in
turn, can modify the granular pressuregradient by in-
fluencingthe local granulartemperature,grain concen-
U
tration, and possiblythe pore-fluidpressure[cf.Hui and
Haft, 1986;Johnsonet al., 1991]. Thus Bagnold'sresults,
obtainedwith fixed concentrationsand shearrates,pro-
videvaluableinsightbut not a valid constitutiveequation
for debrisflows.Section7 caststhis interpretationquan-
titatively.
Takahashi's[1978, 1980, 1981] applicationsof Bag-
nold's [1954] equationsassumethat Nsav --->0%that no
boundaryslipoccurs,that grainsare uniformlydispersed
in the flow, and that interstitial fluid sustains no excess
pressure.As a consequence, grain collisionstressesmust
increaselinearlywith depth to balancethe gravitational
stress.In turn, this requirement mandates a specific
distributionof shearrate that excludesthe possibilityof
"locked" or unyieldingdebris that does not shear [cf.,
Iversonand Denlinger,1987]. Not only doesthis predic-
tion contradictobservationsand the viscoplasticmodel,
it also contradictsthe fact that frictionallylocked mate-
rial must be present during the early and late stagesof
debris flows,when material is mobilized or deposited
[Iversonet al., 1997;Major, 1996]. Figure 16. Schematic diagram depicting steady flow of a
Shortcomingsof the viscoplasticand inertial grain solid-fluidmixture betweenparallel vertical plates.A longitu-
flow modelshave motivatedalternativeapproaches,but dinal body force (e.g., gravity)drivesthe flow. Dependingon
the distributionof granulartemperature,a variety of velocity
none has produced a widely embracedadvance.One
profilesare possible,includingvelocityprofileswith an appar-
approachmelds the equationsof the viscoplasticand ently rigid plug, as shown,but all velocityprofilesmusthave a
inertial grain flow models[Chen, 1987, 1988a,b]. This zero velocitygradient at the channelcenterline [cf. Hui and
yields a formulation with numerous adjustable coeffi- Haft, 1986;Johnsonet al., 1991].Consequentzero shearrate at
cients and with unresolvedphysicalissues,described the centerlineinvalidatesuse of Bagnold's[1954] formulas as
above,that lurk behind the mathematics.A similar ap- constitutiveequationsbecausethe equationsrequire zero nor-
proachcombinesthe viscoplasticand grain inertia mod- mal stressin the presenceof zero shearrate, whichcontradicts
the presenceof the body force that drivesthe flow.
els in a linear sumwithout reconcilingthe models'phys-
ical contradictions [O'Brien et al., 1993]. Other
approachesabandonthe effort to includedetailed rheo-
6. MASS, MOMENTUM, AND ENERGY
logicaldescriptions,and adopthydraulicapproximations CONSERVATION IN DEBRIS FLOW MIXTURES
similar to that used in water flood routing. Traditional
hydraulicapproachesdo not considerthe dynamicsof Mass and linear momentum balances for debris flows
debris mobilization, deformation, and deposition,and can be borrowed with only minor modificationfrom the
instead use empirical coefficientsto parameterize the relativelymature field of continuummixture theory [At-
momentum distribution and energy dissipation in kin and Craine, 1976]. Under this rubric, separatebut
reacheswhere debrisflow is fully developed.Hydraulic stronglycoupled equationsdescribemassand momen-
formulationsthat employ depth-averaged"shallowwa- tum conservation for the debris flow's solid and fluid
ter" momentum balances [Macedonio and Pareschi, constituents,and the solid and fluid equations are as-
1992; Carusoand Pareschi,1993;Hunt, 1994] as well as sumedto apply at all locationssimultaneously.Angular
kinematicwaveapproximations[Weir,1982;Vignauxand momentum equationscan also be formulated but are
Weir, 1990;Arratano and Savage,1994] have been pre- rendered unnecessaryby assumingall stresstensorsto
sented. Calibrated hydraulic models hold promise for be symmetric.Similarly,balancesof thermodynamicen-
practical forecastsof debris flow speedsand shoreline ergy are rendered redundantby assumingthe mixture is
inundation,but they necessarilyneglect key facets of isothermal.However, an equation for grain fluctuation
debris-flowbehavior.Developmentof improvedhydrau- energy(granulartemperature)may be necessaryto de-
lic models(section9) requiresexplicitconsiderationof scribesolid phase motion, and fluid fluctuation energy
the physicalprocessesthat control mass, momentum, (turbulence)may be embeddedin the fluid momentum
and energyfluxesin debrisflows. equation by includingReynoldsstresses.
The mixture theory massconservationequationsfor solid.The last terms in (26a) and (26b) arise from the
the solid and fluid constituentsare, respectively, nonzerotermson the right-handsidesof (22a) and(22b)
and accountfor momentumchangedue to masschange.
O(psVs)/Ot
+ V. (ps%Vs)
= ms (22a) However, they do not account for forces that enable
O(pfvf)/Ot
+ V ß(pfv•v•)
= m• (22b) masschange,and they assumethat massentersor leaves
with zero momentum.Mixture theory equationssimilar
in whichvx and vf are the solidand fluidvelocities,to (22)and(26),butwithmx = mf = 0, appearto have
respectively,
andmxandmf are the respective ratesof been first applied to phenomenalike debris flows by
solid and fluid mass addition, per unit volume. These Shibataand Mei [1986a,b].
equations are coupled becausethe volume fractions Addition of the momentum conservationequations
mustobeyv + vf = 1. Additionof (22a)and(22b) (26) for the solid and liquid constituentsfor the case
yields an equivalentmassconservationequationfor the mx = mf -- 0 yieldsa momentum equationfor thebulk
mixture, mixture
Op/Ot+ V. (pv) = ms+ mf (23) p[Ov/Ot

+ v. Vv] = V. (Ts+ TX+ T') + pg (27a)
in which the mixture mass density p and velocity v are in which
defined by
T' = -PsVs(Vs-v)(Vs- v) - pfvf(vf- v)(vf- v)
p = psVs
+ p•vf (24a) (27b)
¾= (PsVs¾s
q- pfVfYf)/p (24b) is a contribution to the mixture stress that results from
motion of the solid and fluid constituents relative to the
These definitionsshowthat the relevantmixturevelocity
is that of the center of mass, not volume, of a mixture mixture as a whole. Mathematically, T' arisesbecause
volume element. the convective acceleration terms on the left-hand sides
An importantspecialcaseof massconservationexists of (26a) and (26b) do not sum to yield the mixture
if no masschange
occurs
(mx = mf = 0) andthe solid convectiveaccelerationgivenby v. Vv in (27a). Physi-
and fluid constituentsare individually incompressible. cally,T' indicatesthat stresses
in a two-phasedebrisflow
Then additionof (22a) and (22b) resultsin the alterna- mixture representedas a one-phasematerial are more
tive forms complicatedthan those obtained by summingthe solid
andfluidstresses,
Tx + Tf. Exceptfor the complicated
V.v•(v•- Vs)+ V'Vs = 0 (25a) gtressterm, the summedmomentumconservationequa-
tion (27a) has the standardform for a single-phase
V.v:0 (25b) continuum.
Equation (25a) is noteworthybecauseif the standard The basicmixturetheoryequations(22) and (26) hold
expression
for the fluidspecific
discharge
q = vf (vf - three significant advantagesover comparable single-
vx) is substitutedin the first term, the equationmatches phase equations:(1) They explicitlyaccountfor solid
the standardcontinuity equation for deformingporous and fluid volume fractionsand masschangesand thus
media undergoingeither quasi-static[Bear,1972,p. 205] can explicitlyrepresentdiverseor evolvingdebris flow
or inertial [Iverson,1993] motion. Thus an analogybe- compositions. (2) They includeseparatesolidand fluid
tween debris flow mixtures and porous media can be stress tensors, which have relatively straightforward
exploited.Equation (25b) matchesthe standardconti- physicalinterpretations.In contrast,single-phasemod-
nuity equationfor an incompressible,single-phasecon- elsrely upon a stresstensorthat amalgamatesthe effects
tinuum. of solidsand fluidsand their interactions.This amalgam-
The mixture theory momentum conservationequa- ated stressformulation may necessitateuse of numerous
tions are poorly constrainedparametersto describethe mixture
rheology.(3) The mixturemomentumequationscontain
PsVs[OVs/Ot
+ Vs' Vvs] = V. Ts + PsVsg
+ f- msVs an explicit solid-fluidinteractionforce. Such a force is
(26a) lackingin single-phasemodels,which embedits effect in
the amalgamatedstresstensor.Becausesolid-fluidinter-
pfvf[Ovf/Ot
+ vTßVv/] = V.T/+ pfvfg- f- mfvf actionsdiffer from point to point within debrisflowsand
play a key physicalrole (e.g., Figures5, 10, and 13),
(26b) explicitrepresentationof their effectsis desirable.
in whichg isgravitational
acceleration,
TxandTf arethe
solid phase and fluid phase stresstensors,respectively, 6.1. Quasi-StaticMotion
and f is the interactionforce per unit volumethat results Somepropertiesof the interactionforce aswell as of
from momentum exchangebetweenthe solid and fluid the solidand fluid stresses
canbe clarifiedby considering
constituents.Sign conventionsdefine normal stressesas the specialcaseof quasi-staticmotion with incompress-
positivein tension and f as positivewhen it acts on the ible constituents. Quasi-static motion occurs when the
inertial (left-hand side) terms in (26) are negligible, tution of (29), (30), and (31) into (28a), which yields,
whichimplieszero granulartemperature.This wouldbe after some algebraicmanipulation
the case,for example,during inception or cessationof
debrisflow motion. In these situationsthe masschange (Ds- P)l)sg-- •7pdev--
0 (33)
terms in (22) and (26) are likely negligible as well.
whereTe = Ts + TT + pl is the effectivestressand
Equations(22) then reduceto (25), and the momentum
Ts + TT is the total stressas classically
definedby
balances(26) reduceto Terzaghi[1949].This resultimpliesthat the total stressis
V. Ts+ ps(1-v•)g + f: 0 (28a) relatedto the solidstressandpore fluid pressureby Ts +
TT = Ts - vfIp. Moreover,(33) demonstrates
that
V. %+ p•vg- f= 0 (28b) mixture theory subsumesthe standardtheory of quasi-
Under quasi-staticconditionsin a granularmedium, an static porous media as a special case. Thus standard
appropriateconstitutiveequationfor the pore fluid as- theoriesfor slopefailure (which instigatesdebrisflow)
sumesthat only isotropicfluid pressurep contributesto and depositconsolidation(whichconcludesdebrisflow)
derive naturally from mixture theory. Single-phasethe-
the fluid stress[Bear, 1972]. Thus
ories of debris flow lack this generality and power of
T• = -v•Ip (29) explanation.
Mathematical details of slope failure and deposit
whereI istheidentitytensorandvf isincluded because consolidationtheories are too lengthyto present here,
p existsonlywithinthe fluid,whereas TTis assumed to but some key conceptswill be outlined to clarify how
act throughoutthe mixture.
mixture theory providesa unifying framework. Prior to
Even with specification of (29), to evaluate(28b) it is
failure, granular slopedebrismay be regardedas static,
necessaryto specifythe interactionforce f. In the most
and V ßVs= 0 is satisfied.Then (32) reducesto a readily
general case of rapid motion, f might include a wide
solvedLaplace equation for P dev,provided that k/ix is
variety of phenomena such as buoyancy,drag, added
constant.This is the procedureusedin most slopesta-
mass,lift, and Basset,Faxen, and grain diffusionforces
bility analyses[Bromhead,1986]. Following determina-
[Johnsonet al., 1990]. However, for analysisof quasi-
tion of P dev,effective stressesat failure must be calcu-
staticmotion of debrisflows,f dependschieflyon buoy-
lated using(33) and an appropriateconstitutivemodel,
ancy and fluid drag that resultsfrom relative, creeping
suchas a Coulombplasticitymodel (equation(6)) for
motion of the solid and fluid phases[cf. Iverson,1993]:
effective stresseson prospectiveslip surfaces[Savage
and Smith, 1986;Iversonand Major, 1986].Alternatively,
f- -pVv•+ -•- (v•- Vs) (30) elasticitymodels can be used to determine a static ef-
fectivestressfield that canbe usedto infer the potential
Herethebuoyancy force-pfVsgisincluded implicitly
in for Coulombfailuresin slopes[Iversonand Reid, 1992;
the sumof -pVvf andthe gravityforcepfl)fg,andthe Reid and Iverson,1992]. In either case, (32) and (33)
drag force is a function of fluid viscosityix, granular provide the basicbalanceequations.
phasehydraulic permeability k, fluidvolumefractionvf, Mixture theory also subsumesthe theory of consoli-
andrelativevelocityvf - Vs[cf.Johnson et al., 1990]. dation of debrisflow deposits.For small displacements
The ramificationsof (30) canbe clarifiedby combin- the relation V'vs = O½/Otapplies,where ½is the volu-
ing (28b), (29), and (30) and rearrangingtermsto yield metric strain (dilatation) of the solidphase.Employing
this relation, a standardporoelasticconstitutiveequa-
q = vl(vf- Vs): -- •7Pdev (31) tion that relatessoliddilatationand pore pressure[Biot,
1941;Rice and Cleary,1976] can then be substitutedin
in whichPdev-- P - p/Jzisthefluidpressure deviation (32) to yield a diffusion equation for nonequilibrium
fromtheequilibrium orhydrostatic pressure p/Jz,where pore pressure.If the solid and fluid constituentsare
z is the vertical depth below a horizontal datum. Note individuallyincompressibleand k/ix is constant,the re-
that (31) is simply a statementof Darcy's law [Bear, sultingdiffusionequationhasthe simpleform [Chandler
1972].Substitutionof (31) in (25) yieldsan equationthat and Johnson,1981]
governsthe nonequilibriumpore pressure
OPdev kE
k ot ix
V2pdev
: 0 (34)
V' •7Pdev:V' Vs (32)
in which E = K• + 4G/3 is a composite stiffness
Solutionsto (31) can be obtainedif V'Vs is known or modulusthat dependson the conventionalelasticbulk
specifiedin terms of Pdev.Quasi-staticstagesof debris (Kb) and shear(G) moduli of the granularcomposite.
flow initiation and depositioneach involvephenomena The groupkE/ix servesasa pore pressurediffusivityand
that allow this specification. appearsin the pore pressurediffusiontimescaleidenti-
A useful form of the granular phase momentum fied in (8). Modelingandmeasurements byMajor [1996]
equationfor quasi-staticconditionsresultsfrom substi- confirmthat this linear diffusionmodel representspost-
270 ß Iverson: PHYSICS OF DEBRIS FLOWS 35, 3 / REVIEWSOF GEOPHYSICS
depositionalconsolidationof debris flow depositsrea- granular temperatureon debris flow dynamicsare not
sonablywell (Figure 13). However, changesin perme- yet rigorouslyquantified,theyare boththeoreticallyand
abilityand stiffnessmayproducenonlinearbehaviorthat empiricallyidentifiable.Section 7 providesfurther in-
is especiallyimportant in the early stagesof consolida- sight.
tion, when the solid grainsare looselypackedand the A better basisexistsfor evaluatingthe dynamicinter-
mixture undergoeslarge strains. This nonlinearity is actionforce f duringinertial debrisflow motion.Drag
analyzedin section8. arguably constitutesthe most significantinteraction
force in most solid-fluidmixture flows [Johnsonet al.,
6.2. Inertial Motion 1990].Thus the Darcian drag describedby (30) might
The mixture theory approachprovidesa complete representthe most important interactionsin inertial as
frameworkfor predictingquasi-static phenomenaduring well asquasi-static
stagesof debrisflows.Comparisonof
initiation and depositionof debris flows, but can the the experimentalresultsof IversonandLaHusen[1.989]
theory represent inertial debris flow motion with non- with calculationsthat use the model of Iverson[1993]
zero granular temperature?In kinetic theories of dry showthat Darcy couplingalonecanyield excellentpre-
granularflow, the conceptof granulartemperatureleads dictionsof fluid pressureseven when grain Reynolds
to a balanceequationfor the fluctuationenergyof the numbersfall well abovethe Stokesflow limit. Moreover,
solidgrains,whichmustbe satisfiedalongwith momen- an analysisby DiFelice [1994] of diverseexperimental
tum and mass balances.The physicalmotivation and data on fluid dragforcesin both dilute and concentrated
interpretationof granulartemperatureequationswere suspensions of spheresshowsthat the total dragdepends
describedby Haft [1983, 1986], and mathematicalcon- stronglyon the solidvolumefractionus but surprisingly
nectionsof such equationsto classicalkinetic theory weakly on the grain Reynoldsnumber over the range
were established byJenkinsand Savage[1983]andLun et 10-2 < NRey
< 104(where
NRey
depends
ontheabso-
al. [1984].A typicalform of suchan equationis givenby lutevalueof the relativesolid-fluid
velocity,
vf - Vs).
Campbell[1990]: Thus, as a first approximation,a simple Darcy-drag
model may be valid for debris flows.
1
•PsVs[OT/Ot
+ vs' VT] - -V-j - Ts:Vvs-F (35) Finally, appropriateboundaryconditionsas well as
constitutive
equations thatrelatethestresses
Tf andTs
whereinT = (v•2) is thegranular
temperature,
j is the to thevelocities
vf andVsmustin generalbespecified
to
conductiveflux of granular temperature from highly solvethe momentumequations(26). Appropriatecon-
agitatedto less-agitatedregionswithinthe flow,Ts:17Vsis stitutiveequationsfor a Newtonianfluid phaseare well
the rate of generationof granulartemperaturevia work known,but appropriateequationsfor the granularphase
performed by the stresses,Ts, and F is the rate of are lacking.Althoughbeginnings havebeenmadealong
degradation of granular temperature into thermody- theselines [e.g.,Shenand Ackermann,1982],rigorous
namic heat as a result of dissipativegrain interactions. formulationsanalogous to thosefor collisionaldry grain
An importantimplicationof (35) is that grainfluctuation flows[e.g.,Lun etal., 1984]havenot beendeveloped[cf.
energycannot be specifiedas a simple function of the Garcia-Aragon,1995].To build insight,the next section
localshearrate and solidvolumefraction;instead,gran- considerssolid,fluid, and boundaryeffectson momen-
ular temperatureis a field variablethat maydependin a tum transportfrom the perspectiveof elementaryme-
complicatedway on boundaryconditionsand transport chanics.
phenomena.
Rigorousapplicationof (35) requiresknowledgeof a
diffusioncoefficientfor j as well as appropriatebound- 7. GRAIN, FLUID, AND BOUNDARY
ary conditionsand constitutiveparametersto determine INTERACTIONS: ANALYTICAL SOLUTIONS
Ts and F. For purely collisional flows with identical FOR IDEALIZED, STEADY FLOW
sphericalgrains characterizedonly by their size (8),
density(Ps), and coefficientof restitution(e), the nec- The balanceequationsof the precedingsectionpro-
essaryinformationcan be deducedfrom kinetic theory vide a quantitativebut rather generalpictureof debris
(seethe reviewby Campbell[1990]).For flowsin which flow mixture dynamics.To gain more detailed under-
enduring,frictional grain contactsmay play an impor- standing,solutionsand not merely balanceequations
tant role, the theory is less complete [Andersonand mustbe investigated. Althoughsolutionsfor quasi-static
Jackson,1992]. If, in addition, a viscousintergranular slope failure and deposit consolidationproblemsare
fluid is present,satisfactorytheory is lackingentirely. abundantin the literature,solutionsfor boundaryvalue
However, hueristicanalysesindicate that conductionof problemsthat containall the dynamicvariablesin (22),
fluid pressurefluctuationsthat occurif there is nonzero (26), and (35) are unavailable.This sectiontherefore
granular temperature can play an important, perhaps considersprimitiveformsof the balanceequationsthat
dominant,role in mixture momentumtransportwhen admit explicitanalyticalsolutions.These solutionsclar-
viscousfluid is present[Jenkinsand McTigue,1990;Mc- ify physicaleffectsof solid-fluidinteractionsand bound-
Tigueand Jenkins,1992]. Thus althoughthe effectsof ary conditions,and they reveal the significance of solid
and fluid velocities normal to the bed, which are ne-

glectedin most practicalmodels(see section9). As a
by-product,
thesolutions
quant'ify
theshortcomings
of
Bagnold's[1954] formulasas constitutiveequationsfor
debrisflows.They also indicate that presenceof a vis-
cous fluid phase reducesthe net efficiencyof steady,
uniform motion of a granular material. This result con-
tradicts the observation that viscous fluid enhances de-
brisflow efficiency(sections2 and 3), and thusrevealsa
limitation of steadystate theories;it supportsthe idea
that debris flows may be fundamentallyunsteadyphe-
nomena.
Guidancefor simplifyingdebrisflow mixture theory

comesfrom work on closelyrelated granularflows.For
example,Andersonand Jackson[1992] have shownthat
in steady,uniform,gravity-drivenflowsof dry grainsthat
interact throughboth collisionsand friction, significant
variationsin granular temperature,solid volume frac-
tion, and mean grain velocitycommonlyoccuronly near
the bed; grainsfar from the bed "lock" and translateas
a rigid body if there is much dissipationand little con- Figure 17. Schematicvertical crosssectionof a representa-
duction of granular temperatureinto the flow interior. tive segmentof an idealizeddebrisflow (and translatingcoor-
Jenkinsand Askari [1994] have exploited this fact to dinate system(x', y)) that move downstreamwith constant
velocity Vx. Within the segmenta single grain of massm2
analyze the dynamicsof unsteady,dry, granular flows
exchanges momentumwith the bed and the overlyinggrains,
with nonzero granular temperaturesconcentratedin a
which have collectivemassm•. All momentum transfer and
thin basal shear layer. Similar lockingbehaviorneces- unsteadymotion occur in the y direction. The fluctuating
sarilyoccursin debrisflowsduringinitiation and depo- motion of m2 is restrictedto a vertical domaindefinedby the
sition and appearsto occur when "rigid" plugs form characteristicgraindiameter8 andmeanfree paths. All grains
duringsustaineddebrisflow motion [Johnson,1984]. are surroundedby fluid lessdensethan the grains.Gravityacts
With this background,considera hypotheticalsteady, in the -y direction.
uniform debrisflow movingdowna rough,impermeable,
fixed bed of infinite extent and uniform inclination.By
definition,no variation of any quantityoccursin direc- layer of grains,of collectivemassm I (>>m2) , moves
tionsparallelor transverse to the slope(Figure 17). The onlybetweeny = 0 and the upper flow surface(Figure
flow translateswith velocity Vx and for simplicity has 17). Interactionsof the shear layer and locked layer
velocityfluctuationsv' with onlyy components.A mov- occurexclusively at y = 0. No fluid or solidmassmoves
ing coordinatex' = x - Vxt that translateswith the between the shearedand locked domains,and grainsin
steady flow in the x direction provides a convenient both domainsare rigid. Thesesimplifications reducethe
frame of reference. Shearingoccursonly between the continuum-mixtureproblem to a discrete two-body
bed and a singlelayer of grainsandfluid. A thick, locked problemthat entailsexplicitanalysisof momentumex-
layer of grains that moves downslopeas a rigid body changeat the domain boundary.The analysisparallels
overliesthe shearinglayer. The goal is to understand that of BagnoM[1954]but differsby includingthe effects
how momentumcommunicatedby the shearlayer to the of gravity,a free surface,and dissimilarmassesm • and
lockedlayer and bed is influencedby the shearrate and m2, all of which existin debrisflows.
material properties such as solid and fluid densities, The idealizationdescribedabove simplifiesthe gov-
grain friction and restitutioncoefficients,and fluid vis- erningequationssubstantially. Grain fluctuationenergy
cosity.This understandingmay then be extrapolatedto is conductedfrom the basal shear layer to the locked
more general caseswhere many layers of grains shear layer by momentumexchangebetweenthe layers,but
past one another. the time-averagedconductiondoesnot changethe gran-
Substantialsimplification can be achievedwith little ular temperatureof either layer. Thus the fluctuation
sacrificeof relevance if the fluctuating motions of the energy equation (35) becomesunnecessary;terms on
shearingand locked layersoccupyspecifieddomainsin both the left and right sidesof (35) sumto zero for both
y. It is convenientto definethe boundarybetweenthese the shearlayer and the locked layer. Furthermore,the
domainsasy = 0. Thusin the coordinatesystem(x', y), massbalanceequations(22) applicablein eachdomain
assumethat a shearlayer grainwith characteristicdiam- are satisfiedtrivially, and the x componentsof the mo-
eter 8 and massm 2 movesonly in the domain between mentum equations(26) reduce to simple steadystate
y = 0 andy = -8 - s, where s can be viewed as the balances.The momentum equationsfor motion of the
meanfree path of grainoscillation.The overlyinglocked solidsand fluid in the y direction reduce to
272 ß Iverson' PHYSICS OF DEBRIS FLOWS 35, 3 /REVIEWS OF GEOPHYSICS
dv txvf buoyant (the case of Baghold [1954]). If the fluid is

psl)s
•-: 1)s(Ps-
pf)•]y •- v (36a) inviscid or absent, .4 = 0 as well.
Equation (37a) hasthe solution
dp txuf
v (36b) C1 B
u- A exp(-At) +• t + C2 (38)
in which v designates
the solidfluctuationvelocityin the
of g. (Hereprimes in which C• and C2 are constantsdeterminedby initial
y directionand#yisthey component
andsubscripts are omittedon v to simplifythe notation.) conditions.If B = 0, this solution lacksthe secondterm,
No term involvingthe solid stressTs appearsin (36a) but if .4 = 0 (indicatingan inviscidfluid), an entirely
because the solid masses in each domain are treated as different solutionapplies:
discretebodies.Instead,the time-integratedgrain iner- 1Bt 2 +Cit+C2
u =•_ (39)
tia force• ps%(dv/dt)dt communicated by impulsesat
domainboundariespredictsthe time-averagedsolidnor- It is instructiveto examinefirst the predictionsof this
mal stress and obviates the need for a continuum stress
solution for .4 = 0 and then to compare them with
term [cf. Iverson,1993]. In addition, (36b) lacksfluid predictions of the morestronglynonlinearsolution(38).
accelerationterms. This omissionis justified if the di-
mensionless groUP IXV•8/kpav
x (where Paistheadded-7.1. Inviscid Case (A = O)
massdensityof solidgrainsaccelerating throughadja-
centfluid) hasa valuesubstantially
greaterthan 1, which Initialconditions
determine
thevalues
ofC• andC2
indicates that solid-fluid interaction needed to completethe inviscidsolution(39) for the
forces are domi-
nated by viscousrather than inertial effects,a condition motion of grain massesm• and m2. For the upper,
probablysatisfiedin many debrisflows [Iverson,1993]. lockedmass(m •), appropriateinitial conditionsare
Furthermore,(36a) and (36b) are only partly coupled; u(0) = 0 (40a)
(36a) can be solvedexplicitlyfor the solid fluctuation
velocity,whichcanthenbe inputto (36b) to solvefor the v(O) = du/dt(O) - Vo (40b)
pore fluid pressuredistribution.The pore fluid pressure
gradientdp/dy is negativeunder staticconditions(v = which give C• = v0, C2 = 0, and the solutionsfor
0) because y is reckoned positiveupwardand #y is position and velocity
negative; dp/dy increaseswhen the solid fluctuation
velocityv is positiveand decreaseswhen v is negative.
I t2 + rot
u - •B (41a)
However,(36b) showsthat there is no tendencyfor net
excessfluid pressuresto developif a net upwardor v = Bt + Vo (4lb)
downwardsolid velocityis absent.Fluid pressuresthat
These are simpleballistictrajectoryequationsfor the
fluctuate so that the time-averagedexcesspressureis
oscillatingmotionof m •, whichis sustainedby impulses
zero havebeen measuredin laboratoryexperimentswith
from m2. From (41a) it is easyto seethat m • returnsto
idealized debris flow mixtures [Iversonand LaHusen,
1989] and predictedwith a model similarto (36) that itsinitialposition, u - 0, aftera timetcycle:
couplesinertial grain motion to pore pressurediffusion tcycle
-- -2(vo/B) (42)
[Iverson,1993].
Solutionsof the y directionmomentumequationfor At t = tcycle, m• collides withm2, andthenrepeatsits
the solids(equation(36a)) hold the key to understand- trajectory.
ing this idealizeddebrisflow and can be describedbestif The oscillationsof m2 are more complicated,because
the equationis first recastas they must supplyenoughmomentumto sustainthe os-
cillationsof m 1 and alsosatisfy(39), (42), and a condi-
d2u du
tion for interaction with the bed, which includes both
dt2+ A •- = B (37a)
collisionaldissipationand frictional dissipationdue to
slip at the bed. Thusfor m2, pairedevaluations of (39)
in whichu = f v dt is the soliddisplacement
andA and are required,one for upwardmotion and one for down-
B are definedby ward motion. Paired initial conditions are also needed,
which can be written in terms of the grain positionu;
A = (37b) downwardboundvelocityVdown; upwardboundvelocity
vspsk
Vup;
arrivaltimeat thetopof thedomain
boundary,
tup;
and arrivaltime at the bottomof the domainboundary,
tdown.The initial conditionsare
Downward bound
NotethatB isgenerally
negative because #yisnegative
andpf < ps,butB - 0 if the solidgrainsare neutrally U(tup
)=0 (43a)
Vdown(tup
) = V0dow
n= --eVup(tup
) (43b)
Upward bound
U(tdown
) = --S (43c)
Vup(tdown)
= V0up -{-t• (43d)
= --e•down(tdown)
in whichV0u
pandV0dow
naretheinitialvaluesof Vupand
Vdown
, respectively,and tp is a very importantquantity
with dimensionsof velocity.
Values of tp measure the net conversionof transla- Figure 18. Schematicdiagramdepictingconversionof slope
tional momentumm 2vx into fluctuationmomentumm 2v parallel translationvelocityinto slope normal fluctuationve-
that resultsfrom interaction of m 2 with the rough bed locity as a grain interactswith a rough bed. The slopenormal
(Figure 18). Grain fluctuationenergygeneratedby velocity generated minus velocity lost due to frictional and
working of the bed shearstressminus grain energylost collisionalenergydissipationdeterminestp.
to inelastic andbedfrictionequals•1 m2t•2.
collisions
Becausegraininteractionswith the bed andone atiother Thus (44), (45), and (46) form a closed,slightlynonlin-
generallydissipateenergy,a positivevalue of tp is nec- ear setof equationsthat danbe solvedexplicitly.
essaryto conductfluctuationenergyawayfrom the bed Solutionsof the set (44)-(46) yield key quantities
andpreventthemassfromlocking frictionally. such as s and tp, and also facilitate evaluation of the
The restitutioncoefficiente in (43b) appliesto the granulartemperature, granularstressand Bagnold's
collision
of theshearlayergrainm2 andtheoverlying[1954]"dispersive
pressure."The exactsolutionsare
massm1-An exactanalysisof thecollision
shows thate algebraicallycumbersomeand are statedin Appendix B
should
bereplacedbyaii effective
coefficient
of restitu- asequations(B1)-(B6).Validapproximationsof (B1)-
tion whichdiffersslightlyfrom the true coefficient(Ap- (B6) can be obtainedby further exploitingthe assump-
pendixA). However,for the conditionm• >> m2 as- tion m2/m• << 1. As m2/m• -• O, the exactexpression
sumed here, the difference between the effective and
for tp(equation(B6)) reducesto
true valuesof e is negligible.
Evaluationof (39) for the initialconditions
(43) yields q•= vo(m•/m2)(1- e2) (47a)
upward bound and downwardbound solutionsfor the
Substituting
(47a)into(B1)-(B5)thenyieldstheapprox-
position
andVelocity
of m2,whichat timestdown
andtup imations
reduce to
1 2e
tup= (47b)
2
down
-{-V0downtdown:
--S (44a) B v01 + e
tdown
= tup/e (47c)
•1Bt2ui
• + VOuptup
=s (44b)
V0dow
n: -- (m•/m2)ev0 (4 7d)
Btdown
-'l-V0dow
n'-' e v0up
-{---
e (44c) t/Oup= v0 l+e +• (47e)
Btup
+ V0u
p: e V0døwn (44d) S •
B(l+e) l+e + (47f)
Thesefourequationscontainsixunknowns,tup, tdown
, The approximationerror is zero for all valuesof m •/m 2
V0up,
V0down,
Sandtp;thereforetwoadditional
equationsif e - 1 but growsas e --, 0.
arerequired
for clostire.
Constraints
on collisionsbe- The simpleequations (47a)-(47f)describe the essen-
tweenm2 andm • providethe necessary
equations.Col- tial physicsof the inviscidcase;Table 7 listssomenu-
lisionsoccurat timetcycle,
as definedby (42), so it is mericalvaluesthat satisfythe equations.The thbulated
necessarythat values
demonstrate
thefollowing
effects:
(1) If v0,B; e,
andm 2 are held fixed,increasedgrain fluctuationspeeds
tup-{-tdown-tcycl
e= -2(v0/B) (45)
and increasedmean free paths are required to sustain
Moreover, collisionsmust conserve momentum, which motionif the overburdenmassm • is increased.Collision
for m a >> m2 requires(AppendixA) frequencies
andtcyde
remainconstant,
butmorefluctu-
ation energyis requiredto supportthe increasedover-
--v0ml(l + e) + (m•- em•)[--VOdown(l/e)] burdenandpreventtheentiremassfromlockitig.(2• If
t/0down =
e = 1, thentup= tdown,
andmeanfluctuation
speeds
are
(46) identicalin the upward and downwarddirections;no
TABLE7. Examples
ofParameter
Values
ThatSatisfy
Equations
(47)
SpecifiedValues ComputedValues
Case e m•/m2 Vo,m/s B, m/s2 ql,m/s V0d
.... m/s V0up,
m/s tup, S td.... S S, m
1 1 100 0.01 -5 0 - 1 1.01 0.002 0.002 0.00201

2 1 1000 0.01 -5 0 - 10 10.01 0.002 0.002 0.02001
3 0.5 1000 0.01 -5 7.5 -5.00 10.0067 0.00133 0.00267 0.013
4 0.9 1000 0.01 -5 1.9 -9.00 10.0095 0.00189 0.00211 0.019
5 0.9 1000 0.01 -1 1.9 -9.00 10.0095 0.00947 0.0105 0.095
Cases1-5 were selectedto illustrate the effectsof variationsin e, m•/m2, and B.
inputof fluctuationenergyfrom the bedis required,and ThesumVup(tup

) + Vdown(tdown
) measures
theasymme-
q•= 0. Valuesof e smallerthan 1 produceasymmetrical try of the collisionspeedof a shearzone grainwith the
fluctuations,in which mean upward speedsof grains overlyingmassand underlyingbed. If grains are neu-
exceeddownwardspeeds.Then more input of fluctua- trally buoyant(B = 0) and the dissipationrate equals
tion energyfrom the bed is neededto sustainmotion,as the rate of fluctuation energy generation by shearing
is reflectedby increasedvaluesof q•.(3) Reductionsin (q•= 0), this asymmetry is zero, and the collisionspeed
the magnitudeof B, which reflect increasedbuoyancy dependsonly on shearkinematics.This is preciselythe
forces,decreasethe frequencyof grain contactsbut do caseinvestigatedby Bagnold[1954]. However,in a de-
not affectcontactvelocities.They increasethe meanfree brisflow with nonzeroB and q•,collisionsare asymmetric
path s, which reducesthe solidvolumefractionif all exceptin theveryspecial
casethatq•= -Btcycl
e = 2v0.
other factors are constant. This specialcaseis similarto that in Takahashi's[1978,
As characterizedby this simplemodel, granularnor- 1980, 1981] applicationof Bagnold'sequations;the ap-
mal force, which can be equated with the "dispersive plication is appropriateonly if the effectsof gravity
force" of Baghold[1954], resultsfrom the time average settling, asrepresented by thevelocity Btcycle,
arepre-
of impulsesdue to graininteractions.This time average cisely offset by generationof fluctuationvelocity, as
can be calculatedusingthe impulse-momentum princi- representedby q•. In the more general case of q• 4=
ple [e.g.,Spiegel,1967],which for the impulsecommu- -Btcyde , asymmetry of graininteraction forcescauses
nicatedby m i duringmomentumexchange with m2 can dispersive forceto dependon all the kinematicphenom-
be written as ena and material propertiesthat affect flow dynamics
and granular temperature.Bagnold's[1954] equations
do not provide a valid descriptionof flows in such
[-m •vo- mlv0] = Bin1
Favg
-- tcycle
tcycle Fdt=]cycle circumstances.
Finally, the inviscidmixture formulationpermitsex-
(48) plicit evaluationof the granulartemperatureT, which
dependson the temporalaverageof the grainfluctuation
whereFavgis thetime-averaged impulse forceandF is
the instantaneous impulseforce.Accordingto (48) the velocity v. This averageis simplythe quotient of the
time-averagedimpulseis simplythe buoyantweight of mean-free-path distances andhalf the time requiredfor
massm •. Consequently, the normalstressor dispersive onecycle of grainmotion,
tcycle.Thuscombining (7) and
force communicatedby grain interactionsdependson (45) yields an equationfor granular temperature,
the buoyantweight of the overlyingmaterial, and it
dependson grain dynamicsonly insofar as dynamics
determinethe mean free path of grain motion,s (see xtcyc,•/
= v• (50)
(47f)), or equivalently,
%. In a "gravity-free"
casesuchas
that of Bagnold[1954],the solidmassconcentration Usis Alternatively,by substituting(B5) and (B6) into (50),
specifiedratherthan determinedby the physicsof grain the temperaturecan be expressedas a functionof only
interactions,and the dispersiveforce need not balance the fundamentalquantitiesq•, e, and ml/m2,
the body force due to gravity.
A single quantity that summarizeshow dispersive
force dependson gravitycan be calculatedby adding T = (1- e)2(1 + e)6 • + 1+ e (Sl)
(44c) and (44d) and then using(43b) and (43d) to
eliminateV0dow n and V0upin favorof Vup(tup ) and which showsthat the granulartemperatureincreasesas
Vdown(tdown), yielding e, q•, and m2/mi increase.In the limiting caseof e -• 1,
T -• oounlessq• -• 0 and friction due to bed slip
1
dissipatesfluctuationenergyat the samerate it is gen-
Vup(tup ) -- i + e (Btcyc,
) d-Vdown(tdown e+ ½) (49) erated.
7.2. Viscous Case (A > 0) Consequently,increasedviscosityreduces the average

The viscousterm A in the equationof motion (37a) fluctuationvelocityof m i but increasesthe frequencyof
addsmathematicalcomplexitythat reflectsthe physical collisionsbetweenm i and m2, implyingthat the mean
complexityof debrisflows.With A > 0, completeana- free path s of m i is reducedby the presenceof viscous
lytical results analogousto (47) cannot easily be at- fluid. In turn, this result demonstrates that the viscous
tained, but quantitative inferences about the role of fluid decreasesthe granulartemperature,asis expressed
viscousintergranularfluid can nonethelessbe drawn. by (42) and (43). Reduction of granular temperature
The appropriateinitial conditionsfor the viscouscase occursbecausesome of the grain fluctuation energy is
are the same as for the inviscid case. For the rigidly convertedto fluid pressureenergy,which involvesdissi-
lockedupper massm•, substitutionof (40) and (43) in pative viscousforces.
(38) leadsto the positionand velocitysolutions If viscousforces dissipateenergy,how does viscous
fluid enhancedebrisflow efficiency?This questionlies at
B/A - Vo B the heart of debris flow physics.Viscositycausesgrains
u = A [exp(-At) - 1]]+ • t (52a) to interact lessenergetically(with commensuratelyless
dissipation)but more frequently,so the effect of viscos-
v = (Vo- B/A) [exp (-At)] + B/A (52b) ity on net dissipationdue to grain interactionsappears
Like the solutionsfor the inviscidcase(41), theseequa- ambiguous.The issuecan be clarified by assessing the
tionsdescribeballisticpaths,but they differ by including rate of conversion of downslope translational momen-
the effectsof viscousdrag.An important specialcaseof tum to grain fluctuation momentum that is required to
(52b) occurswhenAt is large enoughthat exp (-At) • sustainsteadymotion. To do so, an explicit expression
0; then v = B/A is a good approximation,which indi- for ql/tcycl e mustbeobtained. Thisrequires repetitionof
cates that viscousand gravity forces balance and that the stepsusedto obtain (44a)-(44d) for the shearzone
solid grains descendat their terminal velocity. This is grain m 2 in the inviscidcase,which producesanalogous
comparableto the settlingvelocitydescribedin section3 expressionsfor the viscouscase:
as a basisfor distinguishinggrains that act as discrete
B/A - V0dow n B
solidsfrom those suspendedas part of the fluid.
Further effects of viscousdrag can be evaluatedby
A [exp(--Atdown)-- l] + • tdown -- --S
(56a)
notingthatat timet = tcycle,
themassmi returnsto its
original position,u = O. Using these values in (52a) B
yields B/A
- V0u
Ap[exp
(-Atup
)--1]q-
•-tup
--S (56b)
Bicycle
1 q,
exp(--A/cycle)
= v0_ B/A+ i (53) (V0dow
n--B/A) exp(--Aidown)
+ B/A = e
V0u
p-[-
and substitutingthis resultinto (52b) yieldsthe velocity (56c)
of m1 at timetcycle,
V(/cycle)
= Bicycl
e-t-v0 (54) ( V0u
p-- B/A) exp( -A tup
) q-B/A = V0dow
n (56d)
This equation matchesthe analogousequation for the In conjunctionwith (45) and (46), thesefour equations
inviscidcase,(4lb). form a set of six equations in six unknowns, which
The fact that (54) applies in both the viscousand determinethe motion of m 2. This stronglynonlinearset
inviscidcaseshassignificantimplications,whichare clar- does not admit simple, explicit solutions,but it does
ified by comparingviscousand inviscidmixturesmoving yieldusefulinformation.
Expressions
for exp(-Atup)
at the samerate and undergoingsimilarinternal motion. and exp (--Aidown)can be readily obtainedfrom (56c)
A reasonablecriterion for identifying similar internal and (56d), andthesecansubstitutedinto (56a) and (56b)
motion focuseson the momentum exchangedduring to yield
collisionof m i and m 2. This momentum exchange,de-
scribedby (46), is the only facet of internal motion that (l/e)Vodown
+ V0up
-t-Btup= sA (57a)
can be characterizedindependentlyof viscosity;thus
similar flows can be regarded as those for which Vx is (l/e)Voup
+ V0down
-Jr-
Bidown-qde= -sA (57b)
identicaland the valuesof m 1, m2, e, V0down, V0 in (46)
areidentical.
If v0 isidentical andinviscid Addition of thesetwo equationsproducesa simpleand
in •heviscous
cases,
thenthemagnitude mustbe smallerin importantresult: an equationidenticalto (41) derived
of V(tcycle)
the viscouscase becausem • dissipatesenergywhile it for the inviscidcase.Combinationof (41) with (43b) or
describesits ballistic trajectory in the viscouscase,but (43d) and rearrangementof termsyieldsalternativeex-
not in the inviscidcase.Thus (54) demonstrates that pressionsfor t•/tcycle,
onebasedonupwardboundgrain
velocities
fromtakeoff(V0up) to impact(Vup(tup))
and
tcycle(ViSCOUS)
< tcycle(inviscid) (55) one basedon analogousdownwardbound velocities:
x x
Inviscid fluid Inviscid fluid

e=l e<l X Collision
withmz
KEY:
VupOup
I &•0down
Vdown
(tdown)
/////2////
V0up
ß
x //////5'///
Bed
Viscous
fluid
e=l
I •• Viscous
fluid
e <1
V0dow
n > B/A
V0down: B/A
U0down
< B/A
///
ii
: : :
•0down
>B/A
U0down= B/A
tdOdown<B/A
Figure19. Pictorialsummaryof relativefluctuationvelocitiesof grainm 2 duringonecycleof verticalmotion

in mixturesof elastic(e = 1) and inelastic(e < 1) grainswith viscousand inviscidfluids.Lengthsof arrows
depicttherelativespeeds
of m2:Vou
p istheinitialupwardvelocity
of m2 asit departs
thebed,andVup(tup
)
is its upwardvelocityjust beforeit contactsthe overlyingmassof grains,m •; V0down is the initial downward
velocityof m 2 just after it contactsm •, and Vdown(tdown)
is its downwardvelocityjustbeforeit contactsthe bed.
Dottedlinesat the endof a motioncycledenotethe deficitin fluctuationvelocity(due to energydissipation),
which mustbe resuppliedby q•to sustainthe fluctuatingmotion.B/A is the terminal descentvelocityof m 2
in a viscous fluid.
movestoward the bed, the term in bracketsin (58b) is

= -e
tcycle tcycle tgup(/up
) -- •-t90u
p -- eB (58a) negative,
t•/tcycl
edecreases,
andsomeenergysavings
are
accrued.The effect is mostpronouncedfor largegrains
that have large terminal velocities,and it disappearsas
½_1+e[tgdown(tdown)
_•-t90down
tcycle tcycle ]__
B (58b) grain size diminishes.Moreover, the effect can occur
onlyif e < 1, for equating(58a) and (58b) for the special
For both the viscousand inviscidcases,equations(58) case e = 1 leads to
quantifythe rate of momentumconversionper unit mass
needed to sustainsteadymotion. Terms on the right- --[Vup(tup) -- V0up]
= [Vdown(tdown)
-- t90down] (59)
hand sidesof (58a) and (58b) distinguishthe effectsof
fluid density(representedby B, which has a negative whichshowsthat the term in bracketsin (58b) mustbe
value) from thoseof viscosity.Increasesin fluid density positiveif e = 1. Thusonlyif grainsare sufficientlylarge
cause B -• 0, and this decreasesthe rate of momentum (i.e., have sufficientlylarge terminalvelocities)canvis-
conversion t•/tcycl
e requiredto sustainsteadymotion. cosityenhanceefficiency,and this effect occursonly as
Effects of viscosityenter (58a) and (58b) in a more grainsmove toward the bed. Figure 19 summarizespic-
complicated manner,throughbothtcycl e andthe grain toriallythe variouspossibilities
for energysavingsandloss
velocityterms in brackets.Increasedviscositydecreases duringgrainvelocityfluctuationsfor viscousand inviscid
tcycl
e andhencetendsto increase e. In (58a)the cases with e = 1 and e < 1. In all cases it is evident that
t•/tcycl
term in bracketsmustbe negativebecauseV0up must the net effect of viscosityis increaseddissipation.
exceed Vup(tup
) unless
thereisnoviscosity
andnoeffec- The precedinganalysisshowsthat viscousmediation
tivegravity(A = 0 andB = 0). Thisshows thatt•/tcycl
e of dissipativegrain interactionscannotbe expectedto
is positiveunlessA = 0 andB - 0. In (58b), however, enhancethe net efficiencyof steadydebris flow. This
the term in bracketscan be either positiveor negative, seems to contradict the most basic observation of debris
depending on whether the downward bound takeoff flow behavior,that the interstitialliquid phaseenhances
velocity V0down exceedsthe terminal velocity inferred net mobility(see section2). However,the analysisalso
from (56c), B/A. If V0down exceedsthe terminal fall showsthat increasesin buoyancy(expressedby dimin-
velocity,then the graindeceleratesasit descends toward ishedmagnitudesof B) can enhanceefficiency.Indeed,
the bed,the term in bracketsin (58b) ispositive,andthe asB -• 0, grain contactforcesdue to gravitydiminish,
valueof t•/tcycl
eincreases. nislessthan and grain collisionforces are increasinglybuffered by
However,if V0dow
the terminal velocity, then the grain acceleratesas it viscosity.Pore fluid pressuresthat exceed hydrostatic
35, 3 / REVIEWSOF GEOPHYSICS Iverson: PHYSICS OF DEBRIS FLOWS ß 277
pressuresmimic the conditionB -• 0, and pore pres- The effects of granular temperature and nonlinear
sureshigh enoughto produceliquefactionand mimicthe diffusionon pore pressuresin debrisflowsappear inex-
conditionB = 0 have been measuredin experimental tricablyrelated. In a hypotheticalsteadydebrisflow such
debrisflows(Figures5, 10, and 13). However,(36b) indi- as that consideredin section7, brief or enduring inter-
catesthat suchpressurescan existonly transiently.Thus actionsof fluctuatinggrainsproducetime-averagedcon-
understandingunsteadybehaviorof debrisflow mixtures tact forcesthat exactlybalancethe buoyantweightof the
appearsvital for understandingdebrisflow motion. grains themselves(see equation (48)). Effects of the
fluctuations on the distribution of normal stress in the
mixture are straightforward:stronger fluctuations in-
8. UNSTEADY MOTION AND HIGH PORE crease the fluid volume fraction and reduce the mixture
PRESSURES bulk density[cf.Jenkins,1994],and althoughfluctuations
dilate the debris to a greater degree than is possible
Previoussectionspoint to two key phenomena that statically,the effect of fluctuationson the time-averaged
characterizeunsteady,nonuniform debris flow motion: stress distribution in the debris is identical to that of a
(1) Fluid pressuresgreater than hydrostaticpressures static reduction in bulk density.Gravity-drivenconsoli-
existin debrisflowsand can enhanceflow efficiency,but dationin the presenceof grainfluctuationscan therefore
cannotexistduring steady,uniform motion. (2) Debris proceed much as it does in quasi-staticsediment.Con-
flows move as a surge or series of surges,in which solidation simply requires an attendant reduction of
coarse-grainedheads that lack high fluid pressurere- granulartemperature.However, declininggranulartem-
strict the downslopemotion of finer-graineddebristhat perature with accompanyingchangesin fluid volume
maybe nearlyliquefiedby highfluid pressure(Figures8, fraction can produce large changesin permeability and
9, and 10). A coherenttheorythat predictsthe coupled compressibilitythat render the ongoing consolidation
evolutionof these phenomenais currentlyunavailable. stronglynonlinear. Thus it is reasonable,as a first ap-
This section examines some rudiments of the individual proximation,to embed the effectsof granular tempera-
phenomenawithout consideringcoupling. ture implicitly in a nonlinear consolidationmodel. This
approachis followed here.
8.1. Developmentand Diffusionof High Fluid Consider a debris mass moving sufficientlysteadily
Pressures that equations(28) are a good approximationto (26).
Momentumbalancessuchas (366) and (36b) imply This implies that bulk accelerationsare negligible.As-
that pore pressures greaterthan hydrostatic (-p•h) sumealso that (29) and (30) providean adequatede-
can persist only if the sediment mass contractsvolu- scription of fluid stressesand solid-fluid interaction
metricallyor (in the one-dimensional case)if there is a forces.This implies that the fluid carriesno shearstress
net flux of sedimenttoward the bed. Bulk densityand and impartsforce to the solidsvia buoyancyand Darcian
flow depth data from the USGS debrisflow flume indi- drag only. Equation (32) then describesthe relative
cate that both of thesephenomenaare common.Debris solid-fluidmotionthat producesconsolidationand pore
flow elongationthat causesa flux of sedimenttowardthe pressurediffusion.It is convenientto work with a form
bed involvesa complicatedcombination of shear and of thisequation in whichthefluidvolumefractionvf is
normal strains that is difficult to assess. Contraction the quantity that diffuses.This "porosity diffusion" is
involves volumetric strains that lend themselves to analogous
to the diffusionof voidratio (vf/Vs)usedin
straightforwardanalysisand are thus the focushere. classicalanalysesof soil consolidation,and it is coupled
Contractionof a moving debris flow massproduces to pore pressurediffusionin a straightforwardmanner
pore pressurediffusionanalogousto that which occurs [Gibsonet al., 1967].
duringconsolidationof deposits.Consequently,a diffu- The first step in the analysisinvolvesreplacementof
siontimescalelike tdiff(equation(8)) describesthe per- the solidvelocitydivergenceV'vs in (32) with a more
sistenceof high pore pressures.If this timescaleequals usefulquantity.If the solidand fluid densitiesare assumed
or exceeds the debris flow duration, then volumetric constant,the mass-conservation equationfor solids(226)
contraction(consolidation)and attendantpore pressure can be manipulatedto yield V-vs = (-1/Vs)(Dvs/Dt),
diffusioncan explain the existenceof locally high pore in which D/Dt designatesthe material time derivative
pressures.However, two factors complicateestimation followingthe motion of the solids[Birdet al., 1960;.4tkin
of an appropriatediffusiontimescale:(1) Consolidation andCraine,1976].ThenVscanbe replaced by 1 - vf,
in movingdebrisflowsis resistednot only by pore fluid and the resultingexpression canbe substitutedinto (32)
pressurebut alsoby fluctuatinggrain motions(granular to yield
temperatures)that help keep the sedimentdilated. (2)
The dilated, highly compressiblestate of the moving k D1)•
debrisvirtuallyensuresthat consolidationwill be accom- (1--1)f)V ' • •7pdev = Dt (60)
paniedby large strainsand attendantchangesin perme-
ability and compressibility,which can causepore pres- For sediment-watermixtures under gravity loads, it is
suresto diffuse nonlinearly. reasonable to assume that vf is a functionof onlythe
1.0 mixturesreasonablywell [cf. Lambe and Whitman,1979,

I I I I
p. 320; Beenand Sills, 1981].

0.8
•_•••.iluteclay
.-'..•...•
and
silt
slurries
(2) --
--
For problems of gravitational consolidation under
constantload, the compressibility
OUf/OPdevin (61)
• ...,,,... /-%."'•,
• --
matches
thecompressibility
C - -Ovf/O(tr- p) defined
!- 0.6
-- Osceola
•
(3)• • •Z) . • •
• .. . .....
-- in terms of effective stress.This is apparent from the
o definition of effective stress,which can be written as
0 0.4
(y - p - (y - p•Jz- Pdev (64)
_ Mount
St.Helens
•'"•-•• ',•• •
0.2
North
Fork
Toutle
River/ <Q• • -- where the secondterm on the right-hand side is the
-- andUSGSFlume L USGSFlume
hydrostaticcomponentof the pore pressure,which re-
_ Sandandgraveland Sandandgravelmix(3) -- mains constant. For estimation of consolidation behav-
Ioa•mix
(3) I I I ior, the total stress(• can also be regarded as constant
0.1 1 10 100 1000 10,000
becausethe total stresschangeslittle in comparisonwith
EFFECTIVE STRESS, kPa
the pore pressureduringconsolidation. Thus0(tr -p) •
0(--Pdev), and the compressibilityestimate defined by
Figure 20. Data trends illustratingthe approximatepropor-
tionality between fluid volume fraction (porosity) and the
(62) can be substitutedinto (61), yielding
logarithm of effective stressin various soils and suspensions.
- ,9
Numbersidentify data source:1, Lambe and Whitman[1979]; V. Vv• • Dt (65)
2, Beenand Sills [1981]; 3, Major [1996]. p•C
This maybe interpretedas an advection-diffusion

equa-
tionforvf, in whichtheadvection
velocityisthevelocity
effectivestress,and thus ofp dev-Then applicationof the Vsof the referenceframe for the material time derivative
chainrule, •7pdev
-- (OPdev/OVf)•7Vf,
allows(60) to be Dvf/Dt [Iverson,
1993][cf.AtkinandCraine,1976].Thus
recast
as
it is possibleto exploit the fact that in a coordinate
k(i -- 'U
f) OqPdev D'uf systemthat moveswith velocityVs,(65) transformsinto
V. V.•- (61) a standarddiffusionequationthat is readilysolvedliver-
• ovf Dt
son, 1993] [cf. Ogata, 1970].
in whichthefactorOPd•v/OV•
playstheroleof a stiffness Theexpression
k(1 - vf)/p•C,whichplaystheroleof
modulus(reciprocalcompressibility)of the mixture. the hydraulicdiffusivityin (65), can be used to under-
The conventionalexpressionfor the compressibility
C stand the character of nonlinear consolidation and esti-
of sediment-watermixtures describesthe change in mate the timescaleof nonlinearpore pressurediffusion.
porevolume• dueto changes in effectivestress,
C = The expressioncanbe written in a more explicitform by
-Ou•/O(tr- p). (Here (• is the total normalstress, noting that the permeabilityof most sedimentmixtures,
defined as positivein compression,which is related to includingdebrisflow mixtures(Figure6), is an exponen-
thestressesin section 6 by(• - -5• (trT•+ trT•), wheretial functionof porosityor fluid volumefraction,k - k 0
tr denotesthe trace of the tensor.) For large strains,a exp(avf) [cf.Lambeand Whitman,1979,p. 286;Been
reasonablepostulate is that this compressibilityvaries and Sills,1981].Use of this expressionand (62) leadsto
inverselywith the effectivestress, an estimateof the hydraulicdiffusivityDh,
C = ,/(tr - p) (62) k(1 -•/) k0(1-v•)(tr-p)
where , is a positiveconstanttypicallysmaller than 1. p,C KI& Dh= - exp(av/) (66)
Relationship(62) implies that the mixture growspro- which showsthat the diffusivityis also an exponential
gressivelyless compressibleas consolidationproceeds
function of fluid volume fraction. The implicationsof
and effectivestressincreases.Combiningthe definition
(66) can be clarifiedby notingthat (63) can be usedto
C = -0•f/0(tr - p) with (62) and integratingthe rewrite the exponentialfunction as
resultingequationshowsthat (62) alsoimpliesthat the
fluid volumefraction(porosity)declineslogarithmically exp(av/)- exp[a(•/0- KIn I((• - p)/((• - P)o)]
as effective stress increases: (67)
= exp(av/0)/[(tr- p)/((• - p)0]a"
•/= •/0- "In I((• -p)/((• -P)01 (63)
Substituting(67) into (66) then yields
inwhichUfoand((• - P)0 arecharacteristic
valuesof the
fluid volume fraction and effective stress that derive
(1 - •/)k0 exp(a•/0)
fromthestipulation
thatvf = Ufowhen(• - p = ((• - P)0. Dh -- ((y_ p)•K((y__p)l--a. (68)
Data plotted in Figure 20 indicate that a logarithmic
relationshipsuch as (63) describesthe volume change An obviousimplicationof the lastfactorin (68) is that
behavior of a variety of soils,slurries,and debris flow the character of pore pressure diffusion depends on
35, 3 /REVIEWS OF GEOPHYSICS Iverson- PHYSICS OF DEBRIS FLOWS ß 279
TABLE 8. Hydraulic Diffusivities D h and Diffusion Timescales tdifrfor Various Debris Flow Materials as Computed From
Available Data and Equation (69)
(• - P)0,
Material ko,m2 a Ufo I• Pas •: Pa h,m Dh,m2/s tdiff
,S
USGS flume sand-gravelmix 10-13 20 0.3 0.001 0.02 200 0.1 4 X 10 -4 20

USGS flume loam-gravelmix 10-13 10 0.4 0.001 0.03 200 0.1 4 X 10 -s 300
Toutle River, May 18, 1980 10--14 15 0.45 0.01 0.02 2000 1.0 9 X 10--s 10,000(3 hours)
Osceola, circa 5700 B.P. 10-•4 10 0.55 0.01 0.03 2000 1.0 2 X 10-s 50,000(14hours)
*For comparison,
Major [1996]usedquasi-static
consolidation
experimentsto determinethe followingbestfit valuesof hydraulicdiffusivities:
USGSflumeloam-gravel
mix,1 x 10-6 m2/s;ToutleRiverdebrisflow,2 x 10-6 m2/s;andOsceolamudflow,4 x 10-7 m2/s.
whether or not aK > 1. If aK > 1, the diffusivity 2000 Pa, whichgiveC --- 2 x 10-s Pa-•. Thiscom-
decreases as the effective stress increases because re- pressibilityis roughly 1000 times greater than that of
ducedpermeabilitymore than compensatesfor reduced typical granular soil and many orders of magnitude
compressibility.This is likely to be true for highly di- greaterthan that of rock [cf.Lambeand Whitman,1979].
lated, highlycompressible sediment-watermixturessuch Table 8 lists representativevalues of D h and tdiff
as debrisflow mixturesthat are fully liquefied. In con- calculatedusing(69), data from Figures6 and 20, and a
trast,for a • < 1, which is typicalof less-dilatedmixtures representativerange of debrisflow materials and thick-
such as most soils, the diffusivityincreasesas effective nesses.The pore pressure diffusion timescalesrange
stressincreases,becausereducedcompressibilitydomi- from tens of secondsto >10 hours, and they generally
natesreducedpermeability.Values of a maybe obtained exceed the duration of motion of the corresponding
from plots of permeabilityas a function of porosityor debrisflows.This result is noteworthyfor two reasons.
fluid volume fraction. For debris flow mixtures, such a First, it demonstratesthat consolidationprovidesa rea-
plot (Figure 6) yieldsvaluesof a that rangefrom about sonableexplanationfor sustainedhigh pore pressuresin
10 to 20 (Table 8). Values of • may be obtainedfrom debris flows. Second, it demonstratesthat the large
(67) and plots of porosity(fluid volume fraction) as a compressibilityof debrisflow mixturesunder low effec-
functionof effectivestress(Figure 20), which showthat tive stressapparently contributesvitally to debris flow
values 0.02 < • < 0.04 are typical (Table 8). These mobilityby enablingeffectivehydraulicdiffusivityvalues
valuesindicatethat a• < 1 probablycharacterizesmost to be surprisinglylow, much lower, for example, than
debrisflow mixturesbut that valuesof a K approachand thoseof most granular soilsand fractured rocks [cf. Li,
might even exceed 1 when mixturesare highly dilated 1985;Roelofts,1996]. The large compressibility and low
[cf. Major, 1996]. Values closeto 1 imply that the diffu- diffusivityresultfrom the wide diversityof grain sizesin
sivitydependsweakly on the effectivestressand that a debris flows and from dilation of debris flow sediments
fixed diffusivity may provide reasonableestimatesof that attendsproduction of nonzero granular tempera-
consolidationbehavior.This conclusionis supportedby tures.Thus effectsof widelyranginggrain sizes,granular
measurementsand modeling of quasi-staticconsolida- temperature,and high pore pressuremay play synergis-
tion of experimentaldebrisflow deposits[Major, 1996]. tic and perhaps inseparableroles in sustainingdebris
Estimation of the timescale for consolidation in mov- flow mobility.
ing debris flows requires a characteristicvalue of the
variablediffusivitygivenby (68), whichcan be obtained 8.2. DebrisFlow SurgesWith NonuniformFluid
by assumingthat the effectivestressequalsthe charac- Pressures
teristic effective stress((r - P)0 and that the fluid Concentration of coarse clasts at the heads of debris
volume fraction equals the characteristicvolume frac- flow surgesgives them hydraulic diffusivitiesthat may
tionUfo.Substitution
of thesevaluesin (68)leadsto the greatly exceedthose of most debrisflow material. This
characteristicdiffusivity may explain, in part, why surge heads appear unsatur-
(1 - v•0)k0
exp(av•0) ated and exhibitlittle or no pore fluid pressure(Figure
D h= (or--P)0 (69) 10). Interactionof surgeheadswith the nearlyliquefied
material behind them playsa key role in determiningthe
The timescale for consolidation of debris flow mixtures unsteady,nonuniform character of debris flow motion
withthisdiffusivity
is givenbytdiff'-' h2/Dh,whereh is and the extent of debris flow runout. Parts of debris
the debrisflow thickness.Note that (69) impliesthat the flowsthat remain nearlyliquefiedprovidelittle frictional
effective compressibilityof the debris flow mixture is resistanceto motion, whereas surge heads can provide
givenby C = K/(rr - P)0 (see (62)). By this definition, much frictional resistance.Other forms of flow resistance,
the effectivecompressibilityof a debrisflow bodythat is associatedwith viscousflow of pore fluid and inelastic
1 m thick and 90% liquefied, for example,can be esti- grain collisions,may alsovary spatially,as is reflectedby
matedfrom the typicalvaluesK --- 0.04 and ((r - P)0 '" variationsin Savagenumbersand Bagnoldnumbers.
Furthermore,Ph • Pbis usuallya reasonableapproxi-

mationfor debrisflows,sothat (71) canbe expressed
as
h/l=tan d• - tanO (72)
Although quantitativepredictionsof (71) and (72)

must be interpreted with great caution owing to the
manyfactorsneglectedin the analysis,qualitativetrends
predicted by (71) and (72) provide some insight for
interpretingboth debrisflow behaviorand predictionsof
Figure 21. Schematicvertical crosssectionof the rigid body
more elaboratemodels.For example,(72) showsthat on
model of a debris-flowsurge,with geometricparametersde-
steep slopes(where 0 -• d•), h/l -• 0 is required to
fined [cf. Coleman,1993; Whipple,1994].
sustain steady motion; this implies that debris flow
surgeswill accelerateon steepslopesunlessthe lengthof
Rigorous assessment of the interaction of relatively the surge head is very long in relation to its height.
dry surgeheadswith nearlyliquefieddebrisbehindthem Moreover, for surgeswith identical values of l, surges
requiresnumericalanalysisof unsteady,nonuniformde- with the largest h will acceleratefastest and overtake
bris flow motion, as described in section 9. However, smaller surges,which may help explain surge coales-
simple steadystate analysesprovide some insightinto cence.On low-angleslopes,where 0 -• 0, h/l --• tan d•
the problemand a frameworkfor interpretingnumerical is required to sustain steady motion; this implies that
results.Such analyses[Coleman, 1993; Whipple,1994; surgeswill decelerateand stop on low-angleslopesun-
Major, 1996]assumethat the surgehead actsas a trans- lessh/l > tan d•canbe maintained,whichrequiressurge
latingrigid body,with Coulombresistanceat its baseand headsto be short and steep.Typicallyd• --- 30ø,so that l
a completelyliquefied mass pushing it from behind. greater than about h/0.6 sufficesto stop motion. Data
These analysesignore other forms of resistance,includ- suchas thoseof Figures5 and 10 indicatethat the length
ing all resistanceassociated with internaldeformation; of surgeheadstypicallyexceedsh/0.6, so that the fric-
they ignore all inertial and time-dependenteffects,in- tional resistanceof surgeheadsappearscapableof halt-
cludingthe evolutionof the debrisflow shape;and they ing debris flow motion as slope anglesdecline toward
ignore multidimensionaleffectsthat cannot be repre- zero. Data suchas thoseof Figures5 and 10 alsoreveal
sented with a one-dimensional force balance. the oversimplificationof the rigid surge head model,
With the caveatsdescribedabove,considerthe simple however.More realisticassessment of the role of surge
model of a debrisflow surgedepictedin Figure 21. The headsrequiresa model suchasthat describedin the next
surgemovessteadilyon a uniform slopeinclinedat the section.
angle 0. The surgehead has a triangularcross-sectional
shapewith height h equal to the debris flow thickness
measurednormal to the slope.The lengthI of the surge
9. HYDRAULIC MODELING AND PREDICTION
head is measuredparallel to the slope.The massof the
OF DEBRIS FLOW MOTION
surgeheadisthen• phlW,
wherePhisthebulkdensity
of
the head and w is its breadth normal to the plane of the
page.The basalshearforce -rand normalforce (r due to Models that employ hydraulic theory simplifications
provide the most sophisticatedtool for practical fore-
theactionofgravity
1
ontheheadaresimply-r= • ph#hlw casts of debris flow runout and inundation limits. Such
sin0 and{r - • ph#hlwcos0. SlopeparallelCoulomb
resistanceto basalslidingof the head is describedby models also have scientificimportance,for at present
tan d•,and the slopeparallelforce of the liquefieddebris they constitutethe state-of-the-art method for predict-
flow bodypushingagainstthe upslopeface of the head is ing unsteady,nonuniform motion, one of debris flows'
mostobviousand readilymeasuredattributes.Hydraulic
described by• pb#h2w cos0,where Pbisthedensity of models are distinguishedprimarily by the use of depth-
the liquefied body. This expressionassumesthat the
streamlinesof flow parallel the slope.Steadymotion of averaged equations of motion, which omit some key
the head then requires that the slope parallel forces physicalphenomena. In particular,becausesuchmodels
acting on the head sum to zero: ignorevelocity components normal to the bed, they can
include solid-solid and solid-fluid interaction effects in
1 1
only a rudimentaryway. This precludesrigoroustreat-
• ph#hlw
sin0 - • ph#hlw
cos0 tand•
ment of the evolution of granular temperatures and
cosO- O (70) nonhydrostaticpore pressures(see sections7 and 8).
1
+ •pbgh2w
Efforts to build more sophisticationinto hydraulicmod-
Combinationof termsreduces(70) to [cf. Whipple,1994] els continue[e.g.,Jenkinsand Askari, 1994].
To date, severaltypesof hydraulicmodelshave been
h Ph presented. They differ primarily in the type of slope
(tanO- tan0) (71)
l parallel momentumbalanceemployed.The simplestap-
proachusesthe kinematicwaveapproximation,in which andPTaswell asthe totaldebrisflowmassanddensity

a steadystatemomentumbalancereplacesthe dynamic are constant,
momentumbalance for unsteadyflow [e.g., Weir, 1982;
Arratano and Savage,1994]. More elaborate dynamic
wave models retain the full momentum balance [e.g., I0
pfuf
• (q/uf q
+ Vs)+ --' V(q/u•
+ Vs)
Yamashitaand Miyamoto, 1991;Macedonioand Pareschi,
1992;Hunt, 1994; Shiehet al., 1996] but differ in their
representationof stresses that resistmotion. Most begin +Vs.
V(q/u•
+rs)
1- V.T•
+p•u•g
- f (73)
with the one-phase Bingham or Baghold model de- Darcy'slaw providesan estimateof the largestplausible
scribedin section5 but ultimately lump the Bingham q in (73) becausedata (Figures5, 10, and 13) showthat
and Baghold stressesinto a bulk flow resistancecoeffi- hydraulichead gradientsin debrisflowscommonlyap-
cient (e.g., Manning'sn) similarto thoseusedin water proach but seldomexceedliquefaction-inducing gradi-
flood routing. Bulk resistancecoefficientshave been ents,which roughlyequal 1. Thus the hydraulicconduc-
thoroughlycalibratedfor water floodsbut not for debris tivityK = plak/Ix providesa goodestimateof the
flows. Calibration is problematic for debris flows be- maximum magnitudeof q, and the conductivityrarely
causethe mechanismsof momentumtransportand en- exceeds---0.01m/sfor debrisflow materials(seeTable 3
ergy dissipationin debrisflows (solid friction, inelastic and Figure 6). In contrast,v• typicallyexceeds1 m/s.By
collisions,pore fluid flow) may differ significantlybe- thisrationale,
v• generally
exceeds
q/vfbymorethanan
tween events, whereas the chief mechanism in water order of magnitude,and discardingterms in (73) that
floods (hydrodynamicturbulence)is universal.More- containq/vf in favorof thosecontaining v• yieldsthe
over, Manning-type flow resistancecoefficientsamal- approximation
gamate the effects of internal and boundaryresistance
and cannot represent static resistancethat is present pfvf[OVs/Ot
+ rs' •7Vs)]= •7.Tf + pfvfg- f (74)
duringdebrisflow initiation and deposition.Indeed, the
fact that debris flows exhibit both solid and fluid behav-
This equationimpliesthat inertial forcesaffectingfluid
motion are practicallyindistinguishablefrom those af-
ior means that debris flow models require initial and
fectingsolidmotion, exceptinsofaras the fluid and solid
boundary conditions that differ fundamentally from
those for water floods.
massesper unit volume of mixture differ. Small differ-
ences in solid and fluid velocities can nonetheless have
To accountfor debrisflows'variablecomposition,the
very significant(albeit noninertial)effects.
possibilityof boundaryslip, and the mechanicsof initi-
A simplifiedmomentum equation for the solid-fluid
ation and depositionaswell asflow, the hydraulicmodel
mixtureresultsfrom adding(74) and (26a). This yields
describedhere uses internal and basal friction angles
and pore fluid viscosityto characterizeflow resistance. p[OVs/Ot
+ Vs'VVs]= V. (Ts+ Tf) + pg (75)
This facilitatesrigorousmodel testsbecausevalues of
friction anglesand fluid viscositycan be measuredinde- in which p is the mixturedensitydefinedby (24a). The
pendentlyrather than calibrated.Fluid effectsalsoenter solid-fluidinteractionforce f doesnot appearexplicitly
the model by mediating internal and basal friction. in this equationbut residesimplicitlyin the combined
Stressdue to grain collisionsis neglected,so the model solid-fluid
stress
tensor,T• + Ti.
does not represent the full spectrumof debris flow The assumption q/uf<< vsalsoproduces
a simplified
behaviordepictedin Figure 15. The mathematicalfor- mass-balance
equationfrom (25a),
mulation and solutiontechniqueare basedon a modifi- V.Vs=0 (76)
cation of the hydraulic theory for dry granular flows
developedby Savageand Hutter [1989, 1991]. Hungr Thus (75) and (76) constituteapproximategoverning
[1995] has describedan approachthat is in some re- equationsfor debrisflowsthat maintainconstantmass
spectssimilar. and densityas they move. These equationsdiffer from
analogousequationsgoverningmotion of a one-phase
9.1. Relationship to Mixture Theory granularsolidin onlytwo respects:they involvethe total
To clarify the assumptions of the hydraulicformula- mixturedensity
p andtheinfluence
of fluidstress
Ti.
tion, it is useful to establishits relationshipto the mix- The relative simplicityof (75) and (76) simplifies
ture theorydescribedin section6. Simplificationof the subsequentmanipulations,which consistof specializing
mixturemomentumequations(26) can be achievedby the equationsto two spatial coordinates,followed by
focusingon the motion of the solidsand analyzingthe scalingand depth averaging.The stepsare essentially
motionof the fluid relativeto that of the solids,just asin identicalto thosedescribedby Savageand Hutter [1989,
quasi-staticporous media problems[e.g., Bear, 1972]. 1991], who additionally generalized the equations to
Then the pertinentfluid velocityis the specificdischarge accommodatecurvilinear coordinates.The steps can
dividedby the fluidvolumefraction,q/•f = vf - vs. also be generalizedto three spatial coordinates[Lang
Substitutingthis expressionin the fluid momentum and Leo, 1994],but for brevityonly the two-dimensional
equation(26b) yields,for the specialcasein which ps approachis summarizedhere. For two-dimensionalflow
y scaleis that of free fall in the x direction,(1--/g)1/2,

becausethe potential for free fall drives debris flow
motion. These time and length scalesin turn lead to
differing
velocity
scales
forthex direction,
(g•)1/2,and
y direction,
E(g7)
1/2,whichimplythatVx>> Vy.The
scales for stresses are the stresses that would exist at the
baseof a steady,uniform flow of depth h, p#h sin 0 for
shearstress
andp#• cos0 for normalstressandpore
Figure 22. Schematicvertical crosssectionof an unsteady, pressure.
deformingdebrisflow surgemovingdown an inclinedplane. Scaled(i.e., normalized)equationsresultfrom multi-
The flow depth h and depth-averagedvelocity Vx vary as
functions of x and t.
plyingeachtermof (77a)by(l-/#)1/2,
dividing
eachterm
of (77b) and (77c) by p#, and takingthe limit as e -• 0.
This yieldsgoverningequationsthat differ from equa-
acrossan infinitelywide planar surfacethat slopesat the tion (2.10) of SavageandHutter [1989]onlyby including
fluid stresses,
angle0 (Figure 22), (75) and (76) reduceto
aVx aVy
0Vy
aVx+
ox =0 (77a) O•+ • = 0 (78a)
(Ox _ Ors(x)or(xx) øVx+vx
I = Ors(xx) + OVx
05- OVx
= sinO
(1 oy /
Plat+ vxb--x-x
+ vyay/ - ax by - Ox
--
aT•(yx)
Oy
+ p# sin 0 (77b) + e cos0 i Ox Ox
P•,at+ vy•-•+Vxax/ - ay a•- Oy

0=cos
0 Ors()
oy or/(.)
t (78c)
1
In these equationsand all subsequentequations,all
aT•(xy) quantities are normalized by the appropriate scaling
Ox
p# cos0 (77c) variables,as describedabove,but for brevity the original
(dimensional)notation is retained [cf. Iverson,1997].
To streamlinethe presentation,these and subsequent Equations(78) havetwo keyproperties'(1) The y direc-
equationsincorporateseveralchangesin notation.They tion momentumbalance(78c) has a simpleform iden-
omit the subscripts that denotessolidphasevelocities, tical to that for steady,uniformflow;integrationof (78c)
becauseall v refer to the solidphase;velocitysubscripts showsthat the total normal stressat any depth is simply
x andy denoteslopeparalleland slopenormalCartesian the static stressp#(h - y) cos 0. (2) The x direction
components, respectively(Figure 22). Signconventions momentumbalance(78b) includeslongitudinalnormal
for stresscomponentshave been reversedso that com- stressgradienttermsprecededby the smallparametere,
pressionand left-lateral shear are positive,following which apparentlyindicatesthat such terms can be ne-
Savageand Hutter [1989]. Subscriptsin parenthesesde- glected.However,aswasexplainedby SavageandHutter
note the Cartesiancomponentsof the solid and fluid [1989],neglectof longitudinalnormal stressgradientsis
stresses;the first subscriptindicatesthe normal to the untenablebecauseit producesa stressfield identicalto
plane upon which the stresscomponentacts,and the that for steady,uniform flow,whichnegatesany hope of
secondsubscriptindicatesthe directionof action. The modelingsurge-likemotion. The physicalrationale for
shear stresssubscripts(xy) and (yx) are interchange- retaining this term becomesmore apparent when the
able, however, becausestresstensorsare assumedto be
equationsare integratedover the flow depth.
symmetric. The final stepin simplifyingthe governingequations
involvesdepth integration,which incorporatesconstitu-
9.2 Normalization,Depth Averaging,and tive assumptionsabout stressesand producesequations
ConstitutiveAssumptions withoutexplicity dependence.The processis reasonably
A key stepin simplifying(77a)-(77c) involvesscaling straightforward but ratherprotracted,andthe resultsare
that is similar but not identical to the well-known shal-
simplysummarizedhere. The processinvolvesrepeated
low water or Saint-Venant scaling [cf. Vreugdenhil, applicationof Leibniz'srule for integratingderivatives
1994]. As describedby Savageand Hutter [1989] and and incorporateskinematicboundaryconditions,which
Iverson[1997],two lengthscalesexist,the characteristic state that mass neither enters nor leaves at the free
flowlengthl in thex directionandthe characteristic flow surface(wherey - h) and bed (wherey - 0),
depth h in the y direction.The parametere = h/l
oh oh
describesthe ratio of theselength scalesand is deemed
generallymuchsmallerthan 1. The characteristic time- 0•-+ v••xx - vy- 0 at y - h(x,t) (79a)
35, 3 /REVIEWS OF GEOPHYSICS Iverson-PHYSICS OF DEBRIS FLOWS ß 283
vy=0 at y=0 (79b) relatingTs(xx)

to rs(yy),obtained
fromclassical
Rankine
[1857] earth pressuretheory. Earth pressuretheory ap-
It also involves the assumptionthat the debris flow plies to effectivestresses
borne by the solidconstituents
surface is free of all stresses,
[Lambeand Whitman, 1979] and presumesa relation
Tsjj: = =0 (80)
Ts(xx)
= kact/pass
rs(yy) ( 83)
at y=h(x,t)
and it employs depth-averagedvelocities and normal in whichkact/pass
isanearthpressure
coefficient
thathas
stressesdefinedby different valuesdependingon whether the flow is "ac-
tively" extending(OVx/OX > 0) or "passively"compress-
ing (OVx/OX< 0). For deformationthat includesboth
Vx-•
lI0•
vxdy (81a) bed slip and internal slip, earth pressuretheory leadsto
the expression
for kact/pass
presented
withoutderivation
by Savageand Hutter [1989],
Ts(xx)
-• Ts(xx)
dy (8lb)
1 7 [1 - cos2 (bint(ldr-tan2 (bbed)]
1/2
kact/pass
=2 COS 2 4)int - 1
(84)
Tx(yy)
-• Tx(m
dy (81c)
and derivedin Appendix C. The "-" in "7" appliesto
the activecoefficient,and the "+" appliesto the passive.
(81d) Hungr [1995] suggestedthat the earth pressurerelation
T•(xx
): • T•(xx
)dy (83) might be generalizedby includingthe effects of
elastic compliance,but noted that the elastic modulus
has negligibleinfluenceon model predictions.
T•(yy)
:• T•(yy)
dy (81e) Equations (78)-(84) provide all information neces-
sary to complete the formulation of the hydraulic equa-
tions. Integration of (78a) from y = 0 to y = h, with
applicationof the kinematicboundaryconditions(79),
dy - •Vx-2
vx2 (81f)
producesa depth-averagedmassconservationequation,
As wasnotedby Savageand Hutter [1989],valuesof • in Oh O(h•)

(81f) that deviatefrom unity provideinformationabout Ot
+ Ox
:0 (85)
the deviationof the verticalvelocityprofile from unifor-
mity. If a debrisflow movesexclusively by basalslip,• = Integrationof (78c) from y = 0 to y = h yieldsa steady
1 applies.At the other extreme, • = 6/5 appliesfor a momentum balance in the y direction, which statesthat
debrisflow with no basalslidingand a parabolicvelocity the sum of the nondimensional solid and fluid stress
profile indicativeof laminarviscousflow. As a result,the balancesthe y componentof the nondimensionaltotal
assumption• • 1 appearsgenerallyjustified.
mixture weight
More important than the details of the internal ve-
locity field is the constitutivedescriptionof stressesin
Ts(yy)
+ T•(yy)
- h(x, t) - y (86)
(78b). For the granularsolidsthe stressmodelusedhere
is the simpleCoulombrule givenby (6). The appropriate
nondimensional form of the Coulomb rule is This in turn leadsto nondimensionalexpressions for the
total (solid plus fluid) normal stressat the bed and for
Ts(vx)--sgn (v0Ts(mcot 0 tan (bbed/int(82) the y direction depth-averagedtotal normal stress,
in which4)bed/intindicatesthe appropriatefrictionangle
for bed slipor internaldeformationand sgn(Vx) denotes Ts(m+Tf(m-h at y-0 (87)
the sign(+ or -) of Vx.In (82), cot 0 appearsbecauseof
the different scalingsfor shear and normal stresscom- 1
ponents.Although (82) doesnot containpore pressure (h -y) dy = •h

(88)
effectsexplicitly,
it doessoimplicitly
because rs(yy
) and
Tf(yy)arerelatedby(78c).Thusasfluidpressures Tf(yy)
growin magnitude, themagnitudes ) andTs(yx) Integration of the normalized x direction momentum
of rs(yy
diminish. equation(78b) from y - 0 to y = h yieldsthe result [cf.
The Coulomb rule alsoleadsdirectlyto an expression Savageand Hutter, 1989, equation(2.24)]
0 0
at(hVxx)
+ • (•hVxx
2)= h sin0 -e cos0• (hT•(xx))
= -e cos0• (hTf(m) (93)
+ (rs(yx)y:0)
sin0 + (rf(yx)ly:o)
sin0 The utilityof (92) and (93) is enhancedby evaluatingthe
integral(81e)to calculate
rœ(yy)
for a condition
inwhich
0
the fluid pressureincreaseslinearly from zero at the
- ecos0•-•(hTs(xx)) (89)
- ecos0• (hTf(xx)) debrisflow surfaceto a maximumof Pbedat the bed (a
condition consistentwith the hydraulictheory assump-
Terms on the right-handside of (89) can be inter- tions).The integrationshowsthat O(hrf(yy))/Ox
=
pretedasfollows.The first term representsthe gravita- h(OPbed/OX),
and thissubstitution
is usedin (94) below.
tional driving stress.The secondterm representsfric-
tionalresistance to slipat the baseof the flow andcanbe 9.3. GoverningEquations
and AuxiliaryConditions
evaluatedby applyingthe Coulombequation(82) and The final form of the x directionmomentumequation
the normal stressequation(87) at the flow base, resultsfrom incorporating(85) and (90)-(93) in (89),
assuming • = 1, collectingandcancellinglike terms,and
(Ts(yx)
y:0)sin0 = -sgn (•)(h -Pbed)COS
0 tan4)bed dividingby h, which yields
(90)
where h - Pbedis the nondimensional

basal effective + Vx• = sin0 - sgn(•) 1 - cos0tan(bbed
stress
andpbed
-- rf(yy)y=0
isthenondimensional
basal
pore pressure.The third term on the right-handsideof
(89) representsflowresistancedueto shearof the fluid l
at the flow base.It can be evaluatedusingNewton'slaw 0
of viscosity(14), whichyields COS
0•-•[kact/pass(h
-Pbed)
-t-Pbed]
(94)
[(Tf(vx))
•=0]sin0 = -,f(ax/ay)lv:o (91) Together,(94) and (85) form a set of two equationsin
where • is the appropriate,nondimensional depth-aver-
two unknowns,Vx(X,t) andh(x, t), whichcanbe solved
agedviscosity, givenby• - •/[p#•2/(#•)1/2].Applica- providedthat the basalporefluidpressure Pbed(X,t) and
the necessaryinitial and boundaryconditionsare speci-
tion of (91) requiresknowledgeof the fluid velocity
fied. The need to specifyrather than predictbasalpore
gradient at the bed,(OVx/Oy) y=0,whichis generallypressures is inherentto the hydraulicmodel;fluid pres-
unknown but can be obtained from estimates of the
sure deviationsfrom hydrostaticvalues result from ve-
verticalvelocityprofile.The estimates are constrainedby
locitycomponents normalto the bed (seesection8), and
assumingno slip of fluid at the bed and a mean fluid
neglectof suchvelocitiesin (78c) precludesthe possibil-
velocityof •. For example,if the velocityprofile is
ity of predictingnonhydrostatic pressures.Thus inclu-
linear,then(OVx/Oy) y=0- vx/h.If thevelocity profile is sionof nonhydrostatic pressuresPbed(X,t) may seemto
parabolic,thena simpleanalysis of laminarflowdownan
contradictthe hydraulicmodel assumptions. The inclu-
inclineshows that(OVx/Oy) y=0= 3•/h [cf.Birdetal., sion is justified,however,on the groundsthat the con-
1960, pp. 37-40]. If the velocityprofile is blunt, with
solidationprocessresponsiblefor generatingnonhydro-
shearstronglyconcentratednear the bed, a good de-
static fluid pressures(section8) typicallyoperateson
scriptor is (OVx/Oy)
y=0= (n + 2)(vx/h),wheren = 1 timescalessubstantiallylongerthan the debrisflow du-
indicatesa parabolicprofileandn > 1 indicatesblunter
ration(Table 8). Thusasa firstapproximation, highpore
profiles;thisform is usedbelow.The fourthterm on the
pressures, onceestablished, maybe assumed to persistin
right-handsideof (89) represents the longitudinalstress
debrisflows,and pore pressuresmay be treated as pa-
gradientdue to interactionof solid grains.It can be
rametersin hydraulicmodel calculations.
evaluatedusing(83) and (88), yielding
Inspectionof the individualtermsin (94) revealshow
0 the hydraulicmodel encapsulates debrisflow physics.
-e cos0• (hTs(xx)) The inertialterms on the left-handsideof (94) showthat
both rigid body accelerationsand convectiveaccelera-
tionsmaybe important.On the right-handsideof (94),
: -ekact/pass
(;OS(}•-• hTf(yy)(92) if the first two terms are viewed in isolation,they depict
a static balance of forces identical to that used in infinite
As is indicated
by the presence in (92), the slopestabilityanalysesfor casesin whichthere is zero
of Tœ(yy)
longitudinalsolid stressgradientis mediatedby fluid cohesionand an arbitrary distributionof pore pressure
pressure.The final term in (89) representsthe longitu- [cf. Iverson,1990, 1992]. If the last term on the right-
dinal stressgradient due to the fluid pressurealone. hand side is included, this static force balance assumesa
Becausefluid pressureis isotropic,it can be rewritten form comparableto that of two-dimensionalslopesta-
withTœ(yy)in
placeof Tœ(xx
), bilityanalyses that usemethodsof slices,andin thiscase
35, 3 / REVIEWS OF GEOPHYSICS Iverson: PHYSICS OF DEBRIS FLOWS ß 285
the intersliceforcesare representedby depth-averaged Boundary conditionsstipulate that the height of the
Rankine stresses. Thus the model subsumes classical deformingmassis zero at the front margin(x = xr) and
modelsof the staticsof landslideswith spatiallyvaried rear margin (x = xR),
pore pressuresas a limiting case,whichappliesto incip-
h(xr, t) = 0 (96a)
ient debris flow motion. The third term on the right-
hand side of (94) representsthe effectsof shearresis- h(xR,t) = 0 (96b)
tance due to fluid viscosity.The motion of a frictionless
Thesezero-depthboundaryconditionsare connectedto
but viscousmassis representedby the specialcasewhere
the velocitiesat the front and rear flow marginsby the
•)bed/int= 0 or, alternatively,
Pbed= h (in whichcasethe relations
massis completelyliquefiedby pore pressure).
The final term is perhapsthe most interestingand Vx•= dxr/dt (97 a)
importantterm in (94), for it describesthe longitudinal
stressvariation that accompaniesvariations in flow VxR = dxR/dt (97b)
depth and surge-likemotion. The term showsthat a Finally,the pore pressuredistributionpbed(X , t) mustbe
great change in debris flow behavior occurs as Pbed specified.
rangesfrom 0 to h. IfPbed = h and the sedimentmass
behaveslike a liquid,normalstresses are isotropic,equal 9.4. Solutionsand Comparisons With Data
to the staticpressure,and independentof the local style In general,the nonlinearityof the equationset (85)
of deformation. If Pbed = 0 and the debris behaves and (94) necessitates numericalsolutions.The numeri-
like a Coulomb solid, normal stressesare anisotropic, cal solutionssummarizedhere employedthe Lagrangian
and the longitudinal normal stressdepends strongly finite difference schemedevelopedfor dry Coulomb
on whether the sediment mass is locally extending flowsby Savageand Hutter [1989, 1991],who determined
(OVx/OX > 0) or compressing (OVx/OX < 0) asit deforms that this schemewassuperiorto variousEulerian meth-
and moves downslope.For example, in a typical case ods. The Lagrangianschemewas modified to account
with •)int = 40øand •)bed= 30ø,(84) indicatesthat the for the effectsof pore fluid pressure,as representedin
value of the active (extending)and passive(compress- (94), but the viscousshearterm involving• in (94) was
ing) earth pressurecoefficientsare 0.82 and 4.0, respec- omitted for three reasons.(1) Scalinganalyses[Iverson,
tively. In this case, longitudinal stressesin regions of 1997] indicate that this term is commonlyorders of
extendingflowwill be 18% lessthan in a liquid of density magnitudesmallerthan other termsin (94). (2) Lack of
p, but longitudinalstressesin regionsof compressing knowledgeof the appropriaten value makesevaluation
flow will be 4 times greater than in a liquid. Conse- of the viscousshear term uncertain, and it is undesirable
quently,the modelpredictsthat stronglocalgradientsin to introducea poorly constrained"fitting" parameter in
the longitudinalnormal stresscan occurfor two reasons: the model. (3) Omissionof the viscousshearterm re-
either the style of deformation changeslocally from ducesthe model to a straightforwardforce balance in
extendingto compressing,or the pore pressurevaries which Coulomb friction providesall resistanceto mo-
locallyfrom high to low. Thus, dependingon the defor- tion, and fluid stressesmerely mediate the Coulomb
mation style and pore pressuredistribution,the model friction. This facilitates comparisonof model results
expressed by (85) and (94) can representunsteadyflow with those for the dry Coulomb flows of Savageand
Hutter [1989, 1991]. Moreover, becausethe Coulomb
behaviorthat rangesfrom that of a granularavalanche,
bed friction and internal friction anglescanbe measured
asmodeledby Savageand Hutter [1989, 1991],to that of
independentlyand the pore pressuredistributioncan be
a liquid surge,as modeledbyHunt [1994].Furthermore,
measured during debris flow flume experiments,the
the front of a fully developeddebrisflow may act like a mnctel nrnx•icteq tr•e nrectic'ticmq c•œ ex, nerlmentallxr nh-
F'•' '" ' 1: 1: 'J '-'
compressinggranular solid and support high lateral
servedflow velocitiesand depthsand not merely cali-
stresses,while the trailing flow acts more like a fluid.
brated fits of data. Comparisonof predictionsand data
This phenomenonexplainshow debrisflow surgeswith then indicate whether the omitted viscous shear term
steepsnoutsand graduallytaperedtails can move down- might be essential.
streamwith only modestattenuation.
To make predictionsfor debris flow flume experi-
The initial and boundaryconditionsusedin conjunc- mentswith sedimentmixturescontainingabout 2% silt
tion with (85) and (94) are identicalto thosedescribed and clay,43% sand,and 55% gravelby weight,the values
by Savageand Hutter [1989].The initial conditionsspec- •)int: 42øand •)bed-- 28øwere inferred from quasi-static
ify the zero velocityand staticgeometryof the massthat measurementsof the critical anglesfor motion of dry
mobilizes into a debris flow, sedimenton a tilting table. The bed friction angle was
established as the mean of numerous measurements of
Vx(X,0) = 0 (95a) the tilt requiredfor basalslipof a tabularsedimentmass
placedon a concreteslabwith a surfacetextureidentical
h(x, O) = ho(x) (95b) to that of the flume bed. The internal friction anglewas
tative of many others but was distinguishedby particu-

larly high resolution measurementsof vx and h. The
model initial conditionssimulated the July 24, 1995,
conditions
priorto flowrelease:
a 9.4-m
3 heapof sedi-
ment with a verticaldownslopeface and a gentlysloping
LONGITUDINAL PFIOFILE OF $UFIGE :
surface
hada drybulkdensityof 1800kg/m3 andwas
! :
: .. saturatedwith water at hydrostaticpressure.
Figure 24 illustratesthe predicted evolution of the
• i • .• - - Liquefaction-
-:
speedand shapeof the debrisflow surgethat developed
when the 9.4-m3 mass of water-saturated sediment was
releasedfrom the gate. Predictionsindicatethat a blunt
snoutquicklydevelopsand attenuatesonly modestlyas
the debrisflow advancesdownslope.In contrast,simu-
LONGITUDINAL PROFILE OF PORE PRESSURE
lationswith spatiallyuniform pore pressuresor with no
pore pressuresproduce a finely tapered leading edge
Figure 23. Spatialdistributionof basalpore fluid pressuresin rather than a blunt snout [cf. Savageand Hutter, 1989].
a movingsurge(expressedas a fraction of the pressurere- The snout acceleratesas long as it travels on the uni-
quired for completeliquefaction)used in model calculations formly inclined31ø slope.However,the nearlyliquefied
for nodeswhere Vx> 0. The pore pressuredistributionelon- tail behindthe snoutelongatesmarkedlyand accelerates
gateswith the surge,but the pressuremagnitude(relative to
lessrapidlythan the snout.This causesthe surgeprofile
the liquefactionpressure)remainsconstant.
to stretchin lengthand attenuatein heightwith increas-
ing time and travel distance,as commonlyobservedin
experimentsandin nature (seeFigure3). The simulated
establishedin a similarway from measurementsin which surgesubsequentlycompressesas the debrisflow decel-
the slip was constrainedto occur within the sediment erates and comes to rest on the flatter runout surface.
massby barriers that prevented basal slip [Lill, 1993]. One test of model performanceinvolvescomparisonof
FollowingSavageand Hutter, [1989],the measuredstatic the predictedand observedtime and distanceof runout.
friction angleswere assumedto be goodapproximations The distallimit of the experimentalflow was 101 m from
of the applicabledynamicfriction angles.Measurements the flume gate, or 18.5 m acrossthe runout surface
of pore fluid pressuresshowedthem to be hydrostaticat before stopping,whereasthe model predictedthat the
the time of flow release but to rise within a few seconds flow would stop after travelinga total distanceof 91 m,
to near-liquefactionlevels [Iversonet al., 1997] and to or 8.5 m acrossthe runout surface.The experimental
then evolve to a state of near-zero pressureat the flow flow required about 11 s to reach its distallimit, whereas
snout and near-liquefactionpressurein the flow body the model predicted that 9.5 s would be required to
(Figures5, 10, and 11). Figure23 depictsthe distribution reach the distal limit. A causeof the model'soverpre-
of pore fluid pressures,relative to total basal normal dictionof averageflow front speedand underprediction
stress, used in model calculations for all nodes where of the flow front runout is visible in videotape and
Vx > 0. This distribution mimics, in a simplifiedway, photographicrecordingsof the experiment:as it pro-
measured pore pressure distributions.Pore pressures
were assumedto be hydrostaticwhere Vx = O.
2.5
A strongtendencyexistedfor numericalsolutionsto
rr _ _
exhibit oscillatorybehavior, even when the Lagrangian LM

• 2.0 __
Computeddebris-flow
motion
__
schemewas optimized to suppressnumerical instability LM

- t=O
by includinga small numericalviscosityas describedby z 1.5 __ ii;
Savageand Hutter [1989].The "best,"or leastoscillatory, - •
solutionswere obtained by using a discretizationthat Q_ 1.0
includedabout 80 spacenodesand a dimensionlesstime - •
step size of about 0.001. For a typicalflume experiment, • o.5 -- , t=2s t=4s t=6s --
o
in which the debrisflow length is of the order of 10 m,
o
this translatesto a dimensionaltime stepof about 0.001 _
0 0 10 20 30 40 5O 6O 7O
s and spacediscretizationof about 0.1 m. About 30,000
METERS DOWNSLOPE FROM GATE
time steps were needed to simulate a typical flume
experiment.
Figure 24. Predictedvertical cross-sectional profiles of the
To facilitate comparisonswith experimental data, experimental debris flow of July 24, 1995. Profiles illustrate
model predictionsare presentedin termsof dimensional continuousaccelerationand elongation of the debris flow
rather than normalized variables. Comparisonswith surge during 6 s of motion down a uniform 31ø slope. The
data from the USGS flume experimentof July 24, 1995, profile of the staticsedimentheap denotedby t - 0 represents
are emphasizedbecausethis experimentwas represen- the initial condition.Vertical exaggerationis 20x.
gresseddownslope,the flow segregatedinto three major 0.7
surgesplus severalminor surges(Figure 25), whereas

.....i•........
i.....
i....2meters
below
gate
......
0.6
the model simulatesone coherentsurgeonly. Segrega-
0.5
tion into multiple surgesappearsto retard the average
speedof the flow front but increaseits ultimate runout. 0.4
............. -- Model-::........
Depleted in size and momentumby lossof massto the 0.3
trailing surges,the flow front deceleratesquicklyon the
0.2
runout surface.Succeedingsurgesovertakethe deceler-
ating front and impart momentumthat pushesthe front o3
n'
0.1
forward and delaysits stoppage. l- 0 I ' I ' I '-'-I....' I I'' I I I I

More quantitativetestsof the model can be made by •
(.)
0.5 .... •
,
- - -' ' ' 33 meters

below
gate....
comparingpredictedand measuredgraphsof flow depth z
as a function of time at fixed measurement cross sec- :3• 0.4 ,
tions. Figure 25 comparespredictionswith records of n 0.3

flow depth measuredat three downstreamcrosssections • .
• 0.2 o
duringthe flume experimentof July 24, 1995. Measure- 0

-• 0.1
mentsat a crosssection2 m downslopefrom the release
gate were made with an ultrasonictransceiverthat pro-
• , ,
duced data with a time resolution of about 0.1 s and 0.4
"-•'': ..... !...... i....... :..... 67meters
, , ,
below
gate
depthresolutionof about0.05 m. Measurementsat cross 0.3
sections33 m and 67 m downslopewere made with ..... i.......!.... i.... :,.......i......'.....
0.2
high-precisionlaser triangulationsystemsthat provided
a time resolutionof 0.001s (the samplingfrequency)and
depthresolutionof 0.001m. Comparisonof the data and
model predictions show that the model predicts the
0.1
8 10
....
12
:i'....
'....
'....
..•--i. .... :..... !....
,•i , i , i ,
14
"I
.....
16 18 20 22 24
speedof the advancingdebrisflow front very well and TIME, IN SECONDS

the overall shape of the debris flow surge reasonably
well. However,it fails to predict someimportant details. Figure 25. Comparison of measured and predicted flow
depthsasa functionof time for the experimentaldebrisflow of
The model underestimatesthe attenuationof the surge
July 24, 1995. Predictionsand data are shownfor three cross
front, it overestimates the attenuation of the waveform
sections,2 m, 33 m, and 67 m downslopefrom the gate at the
tail, and it doesnot simulatebreakupof the surgeinto flume head. Flow commenced at the recorded time t = 9.324
multiple surges.The first two difficultiesmight be par- s, and flow through all crosssectionswas on a uniform 31ø
tially remediedby includingthe viscousshearterm in the slope.
model, and this is a logicalnext step.Accuratesimula-
tion of the instabilitythat causesa single surgeto de-
volve into multiple surgesis more difficult becauseit adequate model for predicting bulk motion of debris
requires unambiguousdistinction between numerical flows.The model may be especiallyappropriatefor less
and physicalinstabilities.This will likely require a new energeticdebrisflowsin which shearrates are lower and
type of analysis. graincollisions
figurelessprominentlythanin the high-
Overall,the success of the simplemodel that employs speedflows at the USGS flume.
(85) and(94) asgoverningequationsappearsbetterthan
expected.The relatively rapid, thin debris flows in the
USGS flume are characterizedby moderatelylarge Sav- 10. CONCLUSION
age numbersand Bagnold numbers(Table 5), so a
model that emphasizesCoulombfriction and excludes Debris flows are gravity-drivensurges of roughly
explicittreatment of grain collisionsmight be expected equal volumes of water and poorly sorted sediment,
to perform poorly. However, collision-dominatedflows thoroughlymixedand agitated.Phenomenasuchasfine-
obey a relation betweenshearand normal stressesthat sedimentgravitycurrentsor wet rock avalanches
can
mimicsthe quasi-staticCoulomb equation [Savageand superficiallyresembledebris flows but lack the strong
Hutter, 1991]. The ratio of shear to normal stresses, solid-fluidinteractionsthat producedebrisflows'unique
rather than their absolutemagnitudes,determinesthe attributes. Interaction of solids and fluid gives debris
valueofkact/pass in (94),andthetermcontaining flowsbulk mobilities(L/H) that commonlyexceedthose
kact/pass
establishesthe surge-likecharacterof the flow. There- of comparablysizedrock avalanches by 100%. Juxtapo-
fore althoughCoulomb stresseslack the shear rate de- sition of solid and fluid forces allows debris flow mate-
pendence of collisional stresses,they can mimic the rials to slide or lock frictionallyas well as to flow. Most
effects of collisionalstressesin some respects.A mo- debris flows commenceas landslidestriggered by in-
mentum equation as simple as (94), with appropriate creased pore water pressures,and most terminate as
accountingfor viscousshear effects,might provide an slowlyconsolidatingsedimentdeposits.Between these
288 ß Iverson: PHYSICS OF DEBRIS FLOWS 35, 3 / REVIEWS
OF GEOPHYSICS
quasi-staticstagesof motion, debris flow behavior typi- might be required to understanddebrisflow behaviorin
cally is influencedby inertial forces and by a combina- suchcircumstances. Furthermore, the apparentsensitiv-
tion of grain friction, grain collisions,and viscousfluid ity of debrisflow motionto nuancesin the positioningof
flow. Investigationsof theseinfluencesindicate that tra- individualbouldersor flow path obstacles
hintsthat
ditional Bingham and Bagnoldmodelsof debrisbehav- debrisflow motion may be mechanistically chaotic,at
ior should be supplantedby models that account for least in some instances.Deterministic prediction of
interactions of solid and fluid constituents. some aspectsof debris flow behavior may therefore
Data from large-scale flume experimentsprovide prove impossible.
ciues to the character of momentum transfer in debris Future experimentationand modeling can probably
flows. Measured stresses at the base of debris flows illuminateat least sixkey aspectsof debris-flowphysics
change rapidly, and the relative magnitude of stress withoutresolvingthe possibilityof chaosor adoptingthe
fluctuations increases as the area of the measurement methodologyof discrete-particlemodeling:
surface decreases.This indicatesthat individual grains 1. Flow regimesin which grain friction, grain colli-
or groupsof grainsand adjacentfluid interactdynami- sions,and fluid viscositydominate can in principle be
cally with the flow boundary, and probably with one discriminatedon the basisof transitionvaluesof dimen-
another, and corroboratesthe visual impressionthat sionless parameters suchasNBag , Nsav,andNDar. Rig-
debris flows canviolate no-slipboundary conditions and orousexperiments,in which one parameter is systemat-
developsubstantial grainfluctuationenergy,or granular icallyvaried while othersare held constant,shouldhelp
temperature. Measured basal fluid pressures,which define the transition values. Once defined, these values
change asynchronouslyrelative to the basal total can guide application of simplified models of debris
stresses,indicate that headsof debrisflow surgesgener- flowsdominatedby only one or two typesof momentum
ally lack much fluid pressure,whereasthe finer-grained transfer.
tailsof surgesare nearlyliquefiedby highfluid pressure. 2. Development of physicallybasedunderstanding
Interior fluid pressuresremain elevated at near-lique- and modeling of mass gain and loss by moving debris
faction levels even during deposition,indicating that flowsis essentialfor realisticpredictionsin manycircum-
depositionresultsmainly from resistanceat flow heads stances.Incorporatingmasschangeterms in mathemat-
and margins. ical modelsis straightforward,but how to predict the
Mixture theoryprovidesan appropriatemathematical magnitudesof masschangeon the basisof boundaryand
framework for investigatingdebris flows. It indicates flow properties remains unclear. Better understanding
that fluid pressuresgreaterthan hydrostaticcannotper- of erosionand sedimentationby debrisflowswill prob-
sistduring steady,uniform debrismotion. Instead, high ablyrequire systematicexperimentation.
fluid pressuresresult from debriscontraction(consoli- 3. Debris flowscan move as a singlesurge,but they
dation), which must be accompaniedby local reduction commonlybreak into a seriesof surgesof roughlysimilar
of granular temperatureand by globallyunsteadymo- magnitude.Surgefronts carry the largestpercentageof
tion. This suggeststhat debris flow motion may be a large clastsand commonlyform the deepestpart of the
fundamentallyunsteadyphenomenon.As yet, however, flow, andformationand segregationinto multiplesurges
no comprehensive model existsto calculatethe coupled, therefore have great implicationsfor hazards due to
simultaneousevolution of pore pressuresand granular impactand inundation.Althoughdevelopmentof surges
temperaturesin unsteadydebrisflows.Nonetheless,es- from infinitesimalflow perturbationshasbeen observed
timates of characteristictimescalesfor dissipationof under various field and experimentalconditions,the
excesspore fluid pressuresin debrisflowsshowthat they physicsremain poorly understood.More experiments
typicallyexceedflow durations.Depth-averagedmodels and analysesare needed.
of debrisflow motion can therefore exploitthe assump- 4. Pore-fluid pressuresexert a stronginfluenceon
tion that pore fluid pressuresremain elevated for the debris flow mechanics. The influence can be modeled
durationof an event.A model of this type, derivedby simplisticallyby includingrealisticpore pressuredistri-
generalizingthe Savage-Huttermodel of dry flows of butionsin appropriatelyformulated hydraulicmodels
Coulomb material, predictsthe behaviorof experimen- (e.g., section9). However, a rigorousunderstandingof
tal debris flows reasonablywell. pore pressureeffectsrequires a fully coupled model in
Experimentaldebrisflows,like manyin nature,typi- whichpore pressuresand granulartemperaturesevolve
cally have includedno clastswith dimensionscompara- contemporaneously from initiation throughinertial mo-
ble to or greater than the flow depth and have been tion and subsequentdeposition.
confined to channelswith simple geometries.Models 5. Grain sizesortingthat selectivelymovesthe larg-
basedon classicalcontinuummechanicsappearto work est claststo the surfaceand front of debris flow surges
well for describingsuchflows.However, natural debris may play an essentialrole in controllingthe pore pres-
flows may encounter channels so tortuous or entrain sure distribution.Sophisticated continuummodelsmay
boulders solargethatcontinuum mechanical assump-be able to accountfor sortingphenomena.
tions fail. Discrete particle models,comparableto the 6. Models and experimentalteststhat provide un-
dry avalanchemodel of Cleary and Campbell [1993], derstandingof debris flows' responseto three-dimen-
35, 3 / REVIEWS OF GEOPHYSICS Iverson' PHYSICS OF DEBRIS FLOWS ß 289
sionaltopographyare necessaryfor hazard forecastsin TABLE A1. Values of e' Calculated From (A5) for Various
many areas. Quasi-three-dimensionalmodels that cap- Values of e and ml/m 2
ture some elements of debris flow behavior have been
developedrecentlyby O'Brienet al. [1993],Lang and Leo Value
of Value
ofe'
m l/m 2 e = 0.1 e = 0.5 e = 0.9• e = 1.0
[1994], and Hungr [1995], but further work to expand
upon thesemodelsand includethe effectsof solid-fluid 10 0.017 0.443 0.886 1.0
interactions is desirable. 100 0.091 0.494 0.899 1.0
1000 0.099 0.499 0.900 1.0
Comparisonof the tabulatede' valueswith the associatede values

APPENDIX A: DERIVATION OF EQUATION (46) showsthe error of the approximatione - e'.
Conservationof linear momentumby m 2 during col-

lision with m• can be expressedwith good approxima-
Equation (46) is obtainedfrom (A2) by using (A7)
tion by (46) if m• >> m2. The approximationcan be
derivedfrom exact equationsfor momentumconserva- and replacingVup(tup
) with the equivalent
expression
tion during collisionof two inelasticbodiestravelingin
--VOdown(1/e). The validityof the approximationcan be
investigatedempiricallyby comparingvalues of e with
colinearpaths,whichmay be statedas [Spiegel,1967,p.
thoseof e' calculatedusing(A5) (Table A1). The tab-
200-201]:
ulated valuesindicatethat the approximationerror in-
(ml- em2)(-Vo)+ m2(1+ e)Vup(tup) creasesas values of m•/m 2 and e decline. The error is
U0:
ml +m2 lessthan 2%, however,if m•/m2 > 10 and e > 0.9, or
if m•/m2 > 1000 and e > 0.1. The approximation
ml(1 + e)(--Vo)+ (m2- eml)Vup(tup) appearsusefulaslong astheseor comparableconditions
(A2)
U0døwn
-- ml q-m2 are satisfied.
Here m • has velocity v0 prior to the collisionand -v 0

afterthecollision;
m2 hasvelocityVup(tup
) priorto the APPENDIX B' EXACT SOLUTIONS FOR BEHAVIOR
collisionand U0dow
n after the collision. OF GRAIN m2
An effective coefficient of restitution e' for behavior
of m 2 during the collisionmay be defined exactlyas
Exact, simultaneoussolution of (44a)-(44d), (45),
U0down and (46) yields
e': - (A3)
Vup(tup)
An exact expressionfor e' may be obtained by first tup= B v0 1 + e + (e- l) +l+e
solving
(A1) for Vup(tup),
whichyields
taown
= B! Vo1 +2e (e-1) 1+ e (B2)
(2m•
Vup(tup)-
Vom2(•
2[_ 1-;)
e)+ 1+ (A4) rOdown
= -(m•/m2)evo (B3)
Substituting(A4) into (A2) thenyieldsan expressionfor
Voaow
n that is linear in %. Then substitutingboth this
expressionfor Voaow n and (A4) into (A3)yields, after p= ev0 1 + e + m
U0u + 1+ e (B4)
somemanipulation,
e'= 2(m1/m2)
-2 + 2e[2 + (ml/m2)]
+ 3-e + (1- e)(m2/ml)
(A5)
s= 2B[(
•,2-vø--
1+ eq• m-2112-e2vo2(
+ Vo
m2/ (BS)
This equationis exact.Simplifiedversionsof (AS) exist
for the specialcasewhereinm • = m2, whichleadsto the
qJ= v0(1
- e) 2---
roll+ m2
l+e
e2+ 2 + ml
• (1+ e)) m2
result
ß 2 m21+e2e+ i )-1 (B6)
mll+
3e- 1
e': 3-e (A6)
The equationfor qJ,(B6), is criticalbecauseqJappearsin
most of the other equations and has great physical
andfor the specialcasem • >> m2, whichgivesthe result significance.If e - 1, then qJ- 0, which implies that
e' • e (A7) grain fluctuationenergylost to bed friction exactlybal-
ancesthe productionof fluctuationenergyby workingof
This approximation neglectsall termsin (A5) that are of the bed shear stress. If e < 1, as is true for inelastic
the order of 1 and smaller and retains terms of the order sedimentgrains,then qJ> 0, whichimpliesthat produc-
of m i/m 2. tion of fluctuationenergymust more than compensate
290 ß Iverson' PHYSICSOF DEBRISFLOWS 35, 3 / REVIEWSOF GEOPHYSICS
Shear
stress
Ts(yz)
/I
/ I
Normal
stress
/
/
/
/
FigureC1. Mohr stresscircleandCoulombfailureenvelopes

for a granularmaterialthat is simultaneously
slippingalonga bed and failing internally.The radiusof the stresscircledefinesthe maximuminternal shear
stress, 'rmax.
1
for bed frictionin order to sustainthe oscillatingmotion normalstress, rs(mean
) = • (rs(xx)q- rs(yy)).This
of grain m 2. quantity,like the quantitiesb and *max,is definedgraph-
icallyin FigureC6. Physically,'rma
x is the maximumshear
stressattainablein the failing mass,and b is the differ-
APPENDIX C: DERIVATION OF EQUATION (84) encebetween
they direction
normal
stress
rs(yy
) andthe
mean normal stress.
The earth pressurecoefficientequation(84) can be The initial stepsin the derivationconsistof substitut-
derived with reference to the Mohr stress circle and ing (C4) into (C2), substituting(C1) into the resulting
Coulomb failure envelopesdepictedin Figure C1 [cf. equation,and then substituting (C3) into this equation.
Savageand Hutter, 1989]. The diagram illustratesthe This producesthe result
stateof stressin a Coulombmassthat is simultaneously
slidingalongthe bed, where the friction angleis 4)bed, [rs(m_ rs(mean)] 2__r•(mean
2 ) sin24)int 2 tan24)bed
-- r•(yy)
and failing internally,where the friction angle is 4)int. (C6)
The basicequationsnecessaryto obtain the expression Regroupingterms in this equation,usingthe identity
for kact/pass
are 1 - sin2 4)int= COS
2 4)int,anddividing
all termsby
Ts(yx)
= Ts(yy)
tan 4)bed (C]) T$2(vv)(
1 + tan24)b•d)
thenyields
Tma
x = rs(mean
) sin 4)int (C2)
CO82
4)int
(Ys(mean)12
- 2 (Ys(mean)l
1 + tan24)bed
\ Ts(vv)
/ 1 + tan24)bed
\ Ts(vv)
/
rs(mean
) q-b = Ts(yy) (C3)
+ :0 (c7)
2 = 'rmax
b2+ T•(yx) 2 (C4)
Thisisa quadratic
equation
inrs(mean)/rs(yy),
whichmaybe
rs(xx) rs(mean) solvedby the standardquadraticformula,yielding
kact/pass
= rs(yy
)- - l q-2 Ts(yy) (C5) rs(mean)] • [1 -- COS
24)int(]q-tan24)bed)]
1/2
Equations(C1) and (C2) statethe Coulombfailurerule rs(m COS2
4)in
t (C8)
for bed slip and internal slip, respectively.Equations
(C3) and (C4) statesimplegeometricrelationsevident
which is obtained after some algebraicsimplification.
in Figure C1. Like (C1) and (C2), (C3) and (C4) apply
Substitution of (C8) into (C5) thenyields(84). Notethat
for both the active(extensional)and passive(compres-
this derivationand Figure C1 assume4)bed( 4)int;oth-
sional) failure states.Equation (C5) providesa useful
erwise,4)bedis irrelevantbecauseall deformationoccurs
alternative
definitionof kact/pass
in termsof the mean internally.In the eventthat 4)bed) 4)int,the term involv-
ing 4)bedin (C8) and (84) can be ignored, and (84) viscouspower law coefficient.
reducesto the standardform of the Rankine equation number of grainsabove slip surface.
for deforminggranularmedia without basalsliding[e.g., NBag Bagnold number.
Lamb and Whitman,1979, p. 164]. NDar Darcy number.
Nfric friction number.
mass mass number.
NOTATION NRey grain Reynoldsnumber.
Nsav Savagenumber.
factor characterizingdependenceof P porefluidpressure,
M/L T2.
permeability on porosity. Pbed normalizedpore fluid pressureat bed.
A
viscous
relaxation
rate,equalto vftX/Vspxk,
1/T. P dev deviation
ofp fromhydrostatic
value,M/LT2.
b differencebetweeny direction and mean q subscriptdenotingquasi-staticcomponentof
normalstress,
M/L T2. stress.
B buoyancy-adjusted
gravity,equal to q Darcianspecificdischargeof pore fluid,L/T.
(l - PT/Ps)
•ly,L/T2. R bulk flow resistance coefficient.
½ cohesivestrength,
M/L T2. mean free path of grain motion, L.
c mixturecompressibility,
L T2/M. t time, T.
c• constantof integration,L/T. tD time duration of debris flow from initiation
c2 constantof integration,L. to deposition,T.
d total derivativeoperator. tdiff timescalefor pore pressurediffusion, T.
D material derivativeoperator (following tup time of upward motion of grain m 2 between
motion of solids). successivecontacts, T.
Dh hydraulic
diffusivity,
L 2IT. tdown time of downwardmotion of grain m 2
e coefficientof restitutionof solid grains. between successivecontacts, T.
compositemixture stiffness,equal to l/C, tcycl
e tupq- tdown
, T.
M/L T 2. Ts solidphasestress
tensor,M/LT 2.
f solid-fluidinteractionforce per unit volume TT fluidphasestress
tensor,
M/LT2.
of mixture,M/L 2T2. Ts_
f solid-fluid
interaction
stress
tensor,
M/LT2.
F magnitudeof instantaneousgrain impulse Te effective
stress
tensor,M/L T2.
force,ML/T 2. T' extra stress tensor in solid-fluid mixture
Favgtime-averaged
valueofF, ML/T2. modeledassinglephase,
M/L T2.
g gravitational
acceleration,
L/T 2. •r depth-averaged
stress
component, M/L T2.
•7 magnitudeof g, L/T 2. u displacementof solid grain from initial
#y y component
ofg,L/T2. position,L.
h debris flow thicknessnormal to bed, L. v velocitymagnitude,L/T.
h characteristic value of h, L. v mixture velocity,L/T.
H vertical distance of debris flow descent from Vs solid phasevelocity,L/T.
source area, m. vf fluidphasevelocity,
L/T.
i subscriptdenotinginertial componentof Vs time-averagedmean value of Vs,L/T.
stress. v} fluctuation of Vsabout its mean value, L/T.
I identity tensor. Vy component
of Vsnormalto bed,L/T.
j conductiveflux of granulartemperatureper Vx componentof Vsparallel to bed, L/T.
unit volume,M/T 3. Vx depth-averagedvalue of Vx,L/T.
k hydraulic
permeability,
L2. /)setgrain settlingvelocity,L/T.
kact/pass
Rankineearthpressure
coefficient. v0 initial velocity of grain m • following contact
K hydraulicconductivity,L/T. with m2, L/T.
l length of head of debrisflow surge,L. Uup upward-boundvelocity of grain m2, L/T.
l characteristiclength(parallelto bed) of /)down downward-boundvelocityof grain m2, L/T.
debris flow surge,L. UOup initialvalueof Vupfollowing
graincontact
L horizontal distance of debris flow runout with bed, L/T.
from source area, L. U0down initial value of/)downfollowing grain contact
ms massinflux rate of solidsper unit debris with overlyingmass,L/T.
flowvolume,M/TL 3. w debris flow width, L.
my massinfluxrate of fluidper unit debrisflow x coordinatedirected parallel to bed, L.
volume,M/TL 3. x' x coordinate that translatesdownslopewith
m • massof grainsoverlyingbasalshearzone,M. velocity Vx,L.
m 2 massof grain within basalshearlayer, M. y coordinate directed normal to bed, L.
M total mass of debris flow, M. z vertical coordinate, L.
ot masspercentageof finesin sedimentmixture. Walder, Mark Reid, Jim Vallance, Bill Savage,Ron Shreve,
•/ shear strain rate, 1/T. StephenGrand, Neal Iverson,and Sue Knightprovideduseful
F rate of degradationof granulartemperature critiquesof the manuscript.
JamesA. Smith was the editor responsiblefor this paper.
to heatperunitvolume,
M/L T3.
He thanksDave McTigue, William Savage,and Ronald Shreve
g characteristicgrain diameter,L. for their technicalreviewsand Steven Grand for servingas a
ß debrisflow aspectratio, equal to h/l. cross-disciplinaryreferee.
0 slope angle.
K sediment-watermixture compression
coefficient. REFERENCES
X Bagnold's[1954]linear concentrationof solids.
• dynamicviscosityof pore fluid with Abbott, J., L. A. Mondy, A. L. Graham, and H. Brenner,
suspendedfine sediment,M/L T. Techniquesfor analyzingthe behaviorof concentratedsus-
•w dynamicviscosityof pure water, M/L T. pensions,in ParticulateTwo-PhaseFlow, edited by M. C.
Roco, pp. 3-32, Butterworth-Heinemann,Newton, Mass.,
• momentum distribution coefficient for 1993.
depth-averagedflow. Acrivos,A., The rheologyof concentratedsuspensions
of non-
p massdensity
of debrisflowmixture,
M/L3. colloidalparticles,in ParticulateTwo-Phase
Flow, editedby
Pa added-massdensityof solidsaccelerating M. C. Roco, pp. 169-189, Butterworth-Heinemann,New-
ton, Mass., 1993.
throughfluid,M/L 3. Adams, M. J., and B. J. Briscoe, Deterministic micromechani-
Pb massdensityof body of debrisflow surge, cal modelingof failure or flow in discreteplanesof densely
M/L 3. packed particle assemblies:Introductory principles, in
pf massdensity
of debrisflowfluid GranularMatter, edited by A. Mehta, pp. 259-291, Spring-
constituents,
M/L 3. er-Verlag, New York, 1994.
Anderson, K. G., and R. Jackson,A comparisonof some
Ph mass
density
ofheadofdebris
flowsurge,
M/L3. proposedequationsof motion of granular materials for
Ps massdensityof debrisflow solid fully developedflow down inclinedplanes,J. Fluid Mech.,
constituents,
M/L 3. 241, 145-168, 1992.
Pw massdensity
of purewater,M/L3. Anderson, S. A., and N. Sitar, Analysis of rainfall-induced
debrisflows,J. Geotech.Eng., 121,544-552, 1995.
Pdrydry(dehydrated)
bulkdensity
of debrisflow Arattano, M., and W. Z. Savage,Modelling debris flows as
mixture,M/L 3. kinematicwaves,Bull. Int. Assoc.Eng. Geol.,49, 3-13, 1994.
cr totalnormalcompressivestress,
M/L T2. Atkin, R. J., and R. E. Craine, Continuum theories of mixtures:
5; stress(generic),
M/L T2. Basictheory and historicaldevelopment,Q. J. Mech.Appl.
'r shearstress,M/L T2. Math., Part 2, 29, 209-244, 1976.
Bagnold, R. A., Experimentson a gravity-freedispersionof
q'maxmaximum shearstress,M/L T2. large solidspheresin a Newtonianfluid under shear,Proc.
T granulartemperature,L2/T2. R. Soc. London, Ser. A, 225, 49-63, 1954.
vf volumefractionof porefluid,(equalto Batchelor, G. K., and J. T. Green, The determination of the
porosityin saturatedmixture). bulkstress
in a suspension
of spherical
particles
to orderc2,
J. Fluid Mech., 56, 401-427, 1972.
vs volume fraction of granularsolids
Bear, J., Dynamicsof Fluids in PorousMedia, 764 pp., Dover,
(vs + vf = 1 in saturated
mixture). Mineola, N.Y., 1972.
1)finesvolumefractionof fine grains(silt plusclay). Been, K., and G. C. Sills,Self-weightconsolidationof softsoils:
q> bulk friction angle. An experimentaland theoreticalstudy,Geotechnique, 31,
q>a grainfrictionangle. 519-535, 1981.
•)bed friction anglefor slidingalongbed. Beverage,J.P., and J. K. Culbertson,Hyperconcentrations of
suspendedsediment,J. Hydraul. Div. Am. Soc. Civ. Eng.,
•)int friction angle for internal deformation. 90(HY6), 117-128, 1964.
• fluctuationvelocitygeneratedby grain Biot, M. A., General theory of three-dimensionalconsolida-
interaction with bed, L/T. tion, J. Appl. Phys.,12, 155-164, 1941.
X7 gradient operator, 1/L. Bird, R. B., W. E. Stewart, and E. N. Lightfoot, Transport
X7. divergenceoperator, 1/L. Phenomena,780 pp., John Wiley, New York, 1960.
Bird, R. B., G. C. Dai, and B. J. Yaruso, The rheologyand flow
0 partial derivativeoperator. of viscoplastic materials,Rev. Chem.Eng., 1, 1-70, 1982.
• functionaloperator(generic). Bishop,A. W., The stabilityof tips and spoilheaps,Q. J. Eng.
0 subscriptdenotingreferencestate. Geol., 6, 335-376, 1973.
ß scalarproduct of tensorsand dyads[cf. Bird Blake, G. R., Bulk density,in Methodsof SoilAnalysis,edited
et al., 1960]. by C. A. Black et al., pp. 374-390, Am. Soc. of Agron.,
Madison, Wis., 1965.
Bridgman,P. W., DimensionalAnalysis,Yale Univ. Press,New
Haven, Conn., 1922.
ACKNOwLFDGMENIS. Rick LaHusen was instrumen-
Bridgwater,J., M. H. Cooke, and A.M. Scott, Inter-particle
tal in devisingand implementing electronic data collection percolation:Equipment developmentand mean percola-
systemsat the USGS debrisflow flume. Jon Major collected tion velocities,Trans.Inst. Chem. Eng., 56, 157-167, 1978.
and consolidatedthe data on debris flow depositsformed Bromhead,E. N., TheStabilityof Slopes,374 pp., Chapmanand
during flume experiments.Dave McTigue, Jon Major, Joe Hall, New York, 1986.
35, 3 / REVIEWS OF GEOPHYSICS Iverson: PHYSICS OF DEBRIS FLOWS ß 293
Buckingham,E., Model experimentsand the form of empirical Eckersley, D., Instrumented laboratory flowslides,Geotech-
equations,Trans.ASME, 37, 263, 1915. nique, 40, 489-502, 1990.
Campbell,C. S., Rapid granularflows,Annu. Rev.Fluid Mech., Einstein,A., A new determinationof moleculardimensions(in
22, 57-92, 1990. German),Ann. Phys.,19, 289-306, 1906. (Englishtransla-
Cannon,S. H., and W. Z. Savage,A mass-change
model for the tion by A.D. Cowper in Investigationson the Theoryof
estimationof debris-flowrunout, J. Geol., 96, 221-227, 1988. BrownianMovement,editedby R. Furth, pp. 38-62, Dover,
Caruso, P., and M. T. Pareschi, Estimation of lahar and lahar- Mineola, N.Y., 1956.)
runout flow hydrographon natural beds,Environ.Geol., 22, Ellen, S. D., and R. W. Fleming, Mobilization of debris flows
141-152, 1993. from soil slips, San FranciscoBay region, California, in
Casagrande, A., Liquefaction and cyclic deformation of DebrisFlows/Avalanches: Process,Recognition,and Mitiga-
sands--A criticalreview,HarvardSoilMech.Ser.,88, 51 pp., tion, Rev. Eng. Geol., vol. 7, edited by J. E. Costa and G. F.
1976. Wieczorek, pp. 31-40, Geol. Soc. of Am., Boulder, Colo.,
Chandler, R. N., and D. L. Johnson, The equivalence of 1987.
quasi-staticflow in fluid-saturatedporousmedia and Biot's Erlichson, H., A mass-changemodel for the estimation of
slowwave in the limit of zero frequency,J. Appl. Phys.,52, debris-flow runout, A second discussion:Conditions for the
3391-3395, 1981. application of the rocket equation, J. Geol., 99, 633-634,
Chapman, S., and T. G. Cowling, Mathematical Theory of 1991.
NonuniformGases,3rd ed., 423 pp., CambridgeUniv. Press, Fairchild, L. H., and M. Wigmosta, Dynamic and volumetric
New York, 1970. characteristicsof the 18 May 1980 lahars on the Toutle
Chen, C., Comprehensivereviewof debrisflow modelingcon- River, Washington, in Proceedingsof the Symposiumon
ceptsin Japan,in DebtisFlows/Avalanches: Process,Recog- Erosion Control in VolcanicAreas Tech. Mem. 1908, pp.
nition,and Mitigation,Rev.Eng. Geol.,vol. 7, editedby J. E. 131-153, Jpn. Public Works Res. Inst., Min. of Constr.,
Costa and G. F. Wieczorek,pp. 13-29, Geol. Soc.of Am., Tokyo, 1983.
Boulder, Colo., 1987. Fink, J. H., M. C. Malin, R. E. D'Alli, and R. Greeley,
Chen, C., Generalizedviscoplasticmodelingof debrisflow, J. Rheological properties of mudflows associatedwith the
Hydraul.Eng., 114, 237-258, 1988a. spring1980 eruptionsof Mount St. Helens volcano,Wash-
Chen, C., General solutionsfor viscoplasticdebris flow, J. ington, Geophys.Res.Lett., 8, 43-46, 1981.
Hydraul.Eng., 114, 259-282, 1988b. Frankel, N. A., and A. Acrivos, On the viscosityof a concen-
Cleary, P. W., and C. S. Campbell, Self-lubricationby long trated suspensionof solid spheres,Chem. Eng. Sci., 22,
runout landslides:Examinationby computersimulation,J. 847-853, 1967.
Geophys.Res.,98(B12), 21,911-21,924,1993. Garcia-Aragon,J. A., Granular-fluidchuteflow: Experimental
Coleman, P. F., A new explanation for debris flow surge and numericalobservations, J. Hydraul.Eng., 121,355-364,
phenomena(abstract),Eos Trans.AGU, 74(16), Spring 1995.
Meet. Suppl., 154, 1993. Gibson,R. E., G. L. England, and M. J. L. Hussey,The theory
Costa,J. E., Physicalgeomorphology of debrisflows,in Devel- of one-dimensionalconsolidationof saturatedclays,1, Fi-
opmentsandApplicationsof Geomorphology, editedby J. E. nite nonlinear consolidationof thin homogeneouslayers,
Costa and P. J. Fleisher, pp. 268-317, Springer-Verlag, Geotechnique,17, 261-273, 1967.
New York, 1984. Haft, P. K., Grain flow as a fluid-mechanicalphenomenon,J.
Costa,J. E., and G. F. Wieczorek (Eds.), DebtisFlows/Ava- Fluid Mech., 134, 401-430, 1983.
lanches:Process,Recognition,and Mitigation, Rev. Eng. Haft, P. K., A physical picture of kinetic granular fluids,
Geol., vol. 7, 239 pp., Geol. Soc. of Am., Boulder, Colo., J. Rheol., 30, 931-948, 1986.
1987. Hampton, M. A., Buoyancyin debrisflows,J. Sediment.Petrol.,
Costa,J. E., and G. P. Williams,Debris-flowdynamics(video- 49, 753-758, 1979.
tape), U.S.Geol.Surv.OpenFileRep.,84-606,22 min., 1984. Hampton, M. A., H. J. Lee, and J. Locat, Submarine land-
Coussot,P., Structural similarity and transition from Newto- slides,Rev. Geophys.,34, 33-59, 1996.
nian to non-Newtonianbehaviorfor clay-watersuspensions, Hanes, D. M., and D. L. Inman, Experimentalevaluationof a
Phys.Rev. Lett., 74, 3971-3974, 1995. dynamicyield criterionfor granularfluid flows,J. Geophys.
Coussot, P., and J.-M. Piau, On the behavior of fine mud Res.,90(B5), 3670-3674, 1985.
suspensions, RheoLActa, 33, 175-184, 1994. Hayashi, J. N., and S. Self, A comparisonof pyroclasticflow
Coussot, P., and J.-M. Piau, The effects of an addition of and debris avalanchemobility,J. Geophys.Res., 97(B6),
force-free particles on the rheologicalproperties of fine 9063-9071, 1992.
suspensions, Can. Geotech.J., 32, 263-270, 1995. Heim, A., Bergsturzund Menschenleben, 218 pp., Fretz und
Coussot,P., and S. Proust, Slow unconfinedspreadingof a Wasmuth, Ztirich, Switzerland, 1932.
mudflow,J. Geophys. Res.,101(Bll), 25,217-25,229,1996. Helm, D.C., Conceptualaspectsof subsidencedue to fluid
Daido, A., On the occurrenceof mud-debris flow, Bull. Disaster withdrawal,in RecentTrendsin Hydrogeology, edited by T.
Prey.Res.Inst. Kyoto Univ.,Part 2, 21(187), 109-135, 1971. N. Narasimhan,Spec.Pap. Geol. Soc.Am., 189, 103-139,
Davies, T. R. H., Spreadingof rock avalanchedebrisby me- 1982.
chanical fluidization, Rock Mech., 15, 9-24, 1982. Henderson, F. M., Open ChannelFlow, 522 pp., Macmillan,
Davies,T. R. H., Large debrisflows:A macroviscous phenom- New York, 1966.
enon, Acta Mech., 63, 161-178, 1986. Hill, H. M., Bed forms due to a fluid stream,J. Hydraul. Div.
Davies,T. R. H., Debris-flowsurges--A laboratoryinvestiga- Am. Soc. Civ. Eng., 92, 127-143, 1966.
tion, Mitt. 96, 122 pp., Versuchsanst.ftir Wasserbau,Hy- Hooke, R. L., Mass movement in semi-arid environments and
drologieund Glaziologie,Ztirich, Switzerland,1988. the morphologyof alluvialfans,in SlopeStability--Geotech-
Davies, T. R. H., Debris-flowsurges--Experimentalsimula- nical Engineeringand Geomorphology,edited by M. G.
tion, N. Z. J. Hydrol., 29, 18-46, 1990. Anderson and K. S. Richards, pp. 505-529, John Wiley,
DiFelice, R., The voidage function for fluid-particleinterac- New York, 1987.
tion systems,Int. J. MultiphaseFlow, 20, 153-159, 1994. Hopkins, M. A., J. T. Jenkins, and M. Y. Louge, On the
Drake, T. G., Structuralfeaturesin granularflows,J. Geophys. structure of three-dimensional shear flows, Mech. Mater.,
Res.,95(B6), 8681-8696, 1990. 16, 179-187, 1993.
Howard, T. R., J. E. Baldwin II, and H. F. Donley, Landslides mobilization from landslides,Ann. Rev. Earth Planet. Sci.,
in Pacifica,California, causedby the storm, in Landslides, 25, 85-138, 1997.
Floodsand Marine Effectsof the Stormof January3-5, 1982, Jaeger,H. M., and S. R. Nagel, Physicsof the granular state,
in the San FranciscoBay Region,California,edited by S. D. Science, 255, 1523-1531, 1992.
Ellen and G. F. Wieczorek, U.S. Geol. Surv. Prof Pap., Jahns,R. H., Desert floods,Contrib.499, pp. 10-15, Calif. Inst.
1434, 163-184, 1988. of Technol., Pasadena, 1949.
Hsu, K. J., Catastrophicdebrisstreams(sturzstroms)gener- Jenkins,J. T., Rapid granularflow down inclines,Appl. Mech.
ated by rockfalls,Geol. Soc.Am. Bull., 86, 129-140, 1975. Rev.,Part 2, 47(6), S240-S244,1994.
Hui, K., and P. K. Haft, Kinetic grain flow in a verticalchannel, Jenkins,J. T., and E. Askari, Hydraulic theoryfor a debrisflow
Int. J. MultiphaseFlow., 12, 289-298, 1986. supportedby a collisionalshear layer, Proceedings of the
Hungr, O., A mass-changemodel for the estimationof debris- International Associationfor Hydraulic ResearchInterna-
flow runout: A discussion,J. Geol., 98, 791, 1990. tional Workshopon Floodsand InundationsRelatedto Large
Hungr, O., A modelfor the runout analysisof rapid flow slides, Earth Movements,Trent,Italy, 1994, A6.1-A6.10, Int. Assoc.
debris flows, and avalanches,Can. Geotech.J., 32, 610-623, for Hydraul. Res., Delft, Netherlands,1994.
1995. Jenkins,J. T., and D. F. McTigue, Transport processesin
Hungr, O., and N. R. Morgenstern,High velocity ring shear concentratedsuspensions: The role of particlefluctuations,
testson sand, Geotechnique,34, 415-421, 1984. in TwoPhaseFlowsand Waves,edited by D. D. Josephand
Hunt, B., Newtonian fluid mechanics treatment of debris flows D. G. Schaeffer, pp. 70-79, Springer-Verlag, New York,
and avalanches,J. Hydraul. Eng., 120, 1350-1363, 1994. 1990.
Hutchinson, J. N., A sliding-consolidationmodel for flow Jenkins,J. T., and S. B. Savage,A theoryfor the rapid flow of
slides, Can. Geotech. J., 23, 115-126, 1986. identical, smooth,nearly elastic particles,J. Fluid Mech.,
Hutter, K., B. Svendsen, and D. Rickenmann, Debris-flow 130, 187-202, 1983.
modeling:A review,ContinuumMech. Thermodyn.,8, 1-35, Johnson, A.M., A model for debris flow, Ph.D. dissertation,
1996. Pa. State Univ., State College, 1965.
Innes, J. L., Debris flows,Prog.Phys.Geogr.,7, 469-501, 1983. Johnson,A.M., PhysicalProcesses in Geology,557 pp., W. H.
Iverson, R. M., A constitutiveequation for mass-movement Freeman, New York, 1970.
behavior, J. Geol., 93, 143-160, 1985. Johnson,A.M., Debris flow, in Slope Instability,edited by
Iverson, R. M., Unsteady, nonuniform landslide motion, 1, D. Brunsdenand D. B. Prior, pp. 257-361, JohnWiley, New
Theoretical dynamicsand the steadydatum state,J. Geol., York, 1984.
94, 1-15, 1986a. Johnson,G., M. Massoudi,and K. R. Rajagopal,A review of
Iverson, R. M., Unsteady, nonuniform landslide motion, 2, interaction mechanismsin fluid-solid flows, Tech. Rep.
Linearizedtheory and the kinematicsof transientresponse, DOE/PETC/TR-90/9, U.S. Dep. of Energy, 54 pp., Pitts-
J. Geol., 94, 349-364, 1986b. burgh Energy Technol. Cent., Pittsburgh,Pa., 1990.
Iverson, R. M., Groundwater flow fields in infinite slopes, Johnson,G., M. Massoudi, and K. R. Rajagopal, Flow of a
Geotechnique,40, 139-143, 1990. solid-fluidmixturebetweenflat plates,Chem.Eng. Sci.,46,
Iverson, R. M., Sensitivityof stabilityanalysesto groundwater 1713-1723, 1991.
data, in Landslides(Proceedings of the SixthInternational Khegai, A. Y., N. V. Popov, P. A. Plekhanov, and V. A.
Symposiumon Landslides,1), edited by D. H. Bell, pp. Keremkulov, Experiments at the Chemolgan debris-flow
451-457, A. A. Balkema, Rotterdam, Netherlands, 1992. testingground,Kazakhstan,LandslideNews,6, 27-28, Jpn.
Iverson, R. M., Differential equationsgoverningslip-induced LandslideSoc., Kyoto, 1992.
pore-pressurefluctuations in a water-saturated granular Kytomaa, H. K., and C. M. Atkinson, Sound propagationin
medium, Math. Geol., 23, 1027-1048, 1993. suspensions and acousticimagingof their microstructure,
Iverson, R. M., Hydraulic modeling of unsteadydebris-flow Mech. Mater., 16, 189-197, 1993.
surgeswith solid-fluid interactions,in Proceedingsof the Lambe, T. W., and R. V. Whitman, Soil Mechanics,SI Version,
First International Conferenceon Debris-FlowHazardsMiti- 553 pp., John Wiley, New York, 1979.
gation,Am. Soc. of Civ. Eng., New York, in press,1997. Lang, R. M., and B. R. Leo, Model for avalanchesin three
Iverson, R. M., and R. P. Denlinger, The physicsof debris spatial dimensions,CRREL Rep. 94-5, 23 pp., U.S. Army
flows--A conceptualassessment, in Erosionand Sedimen- Corps of Eng. Cold Reg. Res. and Eng. Lab., Hanover,
tationin thePacificRim, edited by R. L. Beschtaet al., IAHS N.H., 1994.
Publ., 165, 155-165, 1987. Lee, J., S.C. Cowin, and J. S. Templeton III, An experimental
Iverson, R. M., and R. G. LaHusen, Dynamic pore-pressure studyof the kinematicsof flow throughhoppers,Trans.Soc.
fluctuationsin rapidly shearinggranularmaterials,Science, Rheol., 18, 247-269, 1974.
246, 769-799, 1989. Li, J., and J. Yuan, The main features of the mudflows in
Iverson, R. M., and R. G. LaHusen, Friction in debris flows: Jiang-JiaRavine, Z. Geomorphol.,27, 326-341, 1983.
Inferences from large-scaleflume experiments,Hydraulic Li, T., A model for predictingthe extentof a major rockfall,Z.
Engineering'93 (Proceedingsof the 1993 Conference of the Geomorphol.,27, 473-482, 1983.
HydraulicsDivision of the American Societyof Civil Engi- Li, V. C., Estimation of in-situ hydraulic diffusivity of rock
neers),vol. 2, 1604-1609,Am. Soc.of Civ. Eng., New York, masses,Pure Appl. Geophys.,122, 545-559, 1985.
1993. Lighthill, M. J., and G. B. Whitham, On kinematic waves,1,
Iverson, R. M., and J. J. Major, Groundwaterseepagevectors Flood movement in long rivers,Proc. R. Soc. London, Set.
and the potential for hillslopefailure and debrisflow mo- ,4, 229, 281-316, 1955.
bilization, Water Resour. Res., 22, 1543-1548, 1986. Lill, T., A critical evaluationof a new methodfor determining
Iverson, R. M., and M. E. Reid, Gravity-driven groundwater the angle of internal friction for cohesionlesssediments,
flow and slope failure potential, 1, Elastic effective-stress B.S. thesis,37 pp., Beloit Coll., Beloit, Wis., 1993.
model, Water Resour. Res., 28, 925-938, 1992. Lun, C. K., S. B. Savage,D. J. Jeffrey, and N. Chepurny,
Iverson, R. M., J. E. Costa, and R. G. LaHusen, Debris-flow Kinetic theories for granular flow: Inelastic particles in
flume at H. J. AndrewsExperimentalForest, Oregon, U.S. Couetteflow and slightlyinelasticparticlesin a generalflow
Geol. Surv. Open File Rep., 92-483, 2 pp., 1992. field, J. Fluid Mech., 140, 223-256, 1984.
Iverson, R. M., M. E. Reid, and R. G. LaHusen, Debris-flow Macedonio, G., and M. T. Pareschi, Numerical simulation of
some lahars from Mount St. Helens, J. VolcanoL Geotherm. Creek and Muddy river lahars, Mount St. Helens, Wash-
Res., 54, 65- 80, 1992. ington, Geol. Soc.Am. Bull., 96, 1056-1069, 1985.
Major, J. J., Experimentalstudiesof depositionby debrisflows: Pierson, T. C., Flow behavior of channelized debris flows,
Process,characteristics of depositsand effectsof pore-fluid Mount St. Helens, Washington,in HillslopeProcesses, ed-
pressure,341 pp., Ph.D. dissertation,Univ. of Wash., Seat- ited by A.D. Abrahams, pp. 269-296, Allen and Unwin,
tle, 1996. Winchester, Mass., 1986.
Major, J. J., Depositionalprocessesin large-scaledebris-flow Pierson,T. C., Flow characteristicsof large eruption-triggered
experiments,J. Geol., 105, 345-366, 1997. debris flows at snow-clad volcanoes: Constraints for debris-
Major, J. J., and T. C. Pierson,Debris flow rheology:Experi- flow models, J. Volcanol. Geotherm. Res., 66, 283-294, 1995.
mental analysisof fine-grainedslurries,WaterResour.Res., Pierson,T. C., and J. E. Costa, A rheologicclassificationof
28, 841-857, 1992. subaerial sediment-water flows, in Debris Flows/Avalanches:
Major, J. J., and B. Voight, Sedimentologyand clast orienta- Process,Recognition,and Mitigation,Rev. Eng. Geol., vol. 7,
tionsof the 18 May 1980 southwest-flanklahars,Mount St. edited by J. E. Costaand G. F. Wieczorek,pp. 1-12, Geol.
Helens, Washington,J. Sediment.Petrol.,56, 691-705, 1986. Soc. of Am., Boulder, Colo., 1987.
Malekzadeh,M. J., Flow of liquid-solidmixturesdowninclined Pierson, T. C., and K. M. Scott, Downstream dilution of a
chutes,Ph.D. thesis,212 pp., McGill Univ., Montreal, Que., lahar: Transition from debris flow to hyperconcentrated
Canada, 1993. streamflow, Water Resour. Res., 21, 1511-1524, 1985.
McTigue, D. F., and J. T. Jenkins,Channel flow of a concen- Pierson, T. C., R. J. Janda, J. C. Thouret, and C. A. Borrero,
trated suspension, in Advancesin Micromechanics of Gran- Perturbationand meltingof snowand ice by the 13 Novem-
ular Materials,editedby H. H. Shen,pp. 381-390, Elsevier, ber 1985 eruption of Nevado del Ruiz, Colombia, and
New York, 1992. consequentmobilization,flow, and depositionof lahars,J.
Middleton,G., Experimentalstudiesrelatedto the problemof Volcanol. Geotherm. Res., 41, 17-66, 1990.
flyschsedimentation,Geol.Assoc.Can. Spec.Pap., 7, 253- Plafker, G., and G. E. Ericksen, Nevado Huascaran ava-
272, 1970. lanches, Peru, in Rockslidesand Avalanches, vol. 1, Natural
Mitchell, J. K., Fundamentalsof Soil Behavior,422 pp., John Phenomena,edited by B. Voight, pp. 277-314, Elsevier,
Wiley, New York, 1976. New York, 1978.
Miyamoto, K., and S. Egashira, Mud and debris flows, J. Poletto, M., and D. D. Joseph,Effective densityand viscosity
Hydrosci.Hydraul.Eng.Jpn.Soc.Civ.Eng., SI-2, 1-19, 1993. of a suspension, J. Rheol., 39, 323-343, 1995.
Mohrig, D., G. Parker, and K. X. Whipple, Hydroplaningof Prior, D. B., and J. M. Coleman,SubmarineSlopeInstability,
subaqueousdebris flows (abstract), Eos Trans. AGU, in SlopeInstability,edited by D. Brunsdenand D. B. Prior,
76(46), Fall Meet. Suppl.,F277, 1995. pp. 419-455, John Wiley, New York, 1984.
Morton, D. M., and R. H. Campbell, Spring mudflows at Rankine, W. J. M., On the stability of loose earth, Philos.
Wrightwood, southernCalifornia, Q. J. Eng. Geol., 7, 377- Trans. R. Soc. London, 1857.
384, 1974. Reid, M. E., and R. M. Iverson, Gravity-drivengroundwater
O'Brien, J. S., and P. Y. Julien,Laboratoryanalysisof mudflow flow and slope failure potential, 2, Effects of slope mor-
properties,J. Hydraul.Eng., 114, 877-887, 1988. phology,material properties,and hydraulicheterogeneity,
O'Brien, J. S., P. Y. Julien, and W. T. Fullerton, Two-dimen- Water Resour. Res., 22, 939-950, 1992.
sionalwater flood and mudflowsimulation,J. Hydraul.Eng., Rice, J. R., and M.P. Cleary, Some basic stressdiffusion
119, 244-261, 1993. solutionsfor fluid-saturatedelasticporousmediawith com-
pressibleconstituents,Rev. Geophys.,14, 227-241, 1976.
Ogata, A., Theory of dispersionin a granular medium, U.S.
Rodine,J. D., andA.M. Johnson,The abilityof debrisheavily
Geol. Surv.Prof. Pap., 411-1,34 pp., 1970.
freighted with coarse clastic materials to flow on gentle
Ogawa, S., Multitemperature theory of granular materials, in
slopes,Sedimentology, 23, 213-224, 1976.
Proceedings of the U.S.-JapanSeminaron Continuum-Me-
Roeloffs, E., Poroelastic techniquesin the study of earth-
chanicsand StatisticalApproachesto theMechanicsof Gran-
quake-inducedhydrologicphenomena,Adv. Geophys., 37,
ular Materials,pp. 208-217, GukujutsuBunken Fukyukai, 133-195, 1996.
Tokyo, 1978. Rogers, C. D. F., T. A. Dijkstra, and I. J. Smalley, Particle
Ohsumi Works Office, DebrisFlow at Sakurajima,2, 81 pp., packing from an Earth scienceviewpoint, Earth Science
Jpn. Min. of Constr.,Kagoshima,1995. Reviews,36, 59-92, 1994.
Okuda, S., H. Suwa, K. Okunishi, K. Yokoyama, and Rosato, A., K. J. Strandburg,F. Prinz, and R. H. Swendsen,
M. Nakano, Observations on the motion of a debris flow Why the Brazil nuts are on top: Size segregationof partic-
anditsgeomorphological effects,Z. Geomorphol.,suppl.35, ulate matterby shaking,Phys.Rev.Lett.,25, 1038-1040, 1987.
142-163, 1980. Sabo Publicity Center, Debris flow at Sakurajima Volcano,
Onada, G. Y., and E.G. Liniger, Random loosepackingsand videotapewith audioin English,OhsumiWork Off., Kyushu
the dilatancyonset,Phys.Rev.Lett., 64, 2727-2730, 1990. Reg. Constr.Bur., Jpn. Min. of Constr.,Kagoshima,1988.
Passman,S. L., and D. F. McTigue, A new approachto the Sassa,K., The mechanismstarting liquefied landslidesand
effectivestressprinciple,in Compressibility Phenomenain debrisflows,in Proceedings of theFourthInternationalSym-
Subsidence, editedby S. K. Saxena,pp. 79-91, Eng. Found., posiumon Landslides,vol. 2, pp. 349-354, Int. Soc.for Soil
New York, 1986. Mech. and Found. Eng., Toronto, Ont., Canada, 1984.
Phillips, C. J., and T. R. H. Davies, Determining rheological Sassa,K., The mechanismof debrisflow, in Proceedings of the
propertiesof debrisflow material, Geomorphology, 4, 101- Eleventh International Conferenceon Soil Mechanicsand
110, 1991. Foundation Engineering,pp. 1173-1176, A. A. Balkema,
Pierson,T. C., Erosion and depositionby debrisflowsat Mt. Rotterdam, Netherlands, 1985.
Thomas,North Canterbury,New Zealand, Earth Surf. Pro- Savage,S. B., The mechanicsof rapid granular flows,Adv.
cesses,5, 227-247, 1980. Appl. Mech., 24, 289-366, 1984.
Pierson,T. C., Dominant particle supportmechanismsin de- Savage, S. B., Interparticle percolation and segregationin
bris flowsat Mt. Thomas,New Zealand, and implications granular materials:A review, in Developmentsin Engineer-
for flow mobility,Sedimentology, 28, 49-60, 1981. ing Mechanics,edited by A. P.S. Selvadurai,pp. 347-363,
Pierson, T. C., Initiation and flow behavior of the 1980 Pine Elsevier, New York, 1987.
Savage,S. B., Mechanicsof debrisflows,in HydraulicEngineer- Takahashi,T. (Ed.), Japan-Chinajoint researchon the pre-
ing '95 (Proceedings of the;[993 Conference of theHydraulics vention from debris flow hazards,Res. Rep. 03044085, 195
Divisionof the•lmedcan Societyof Civil Engineers),vol. 2, pp., Grant-in-Aid for Scientific Research, Jpn. Min. of
1402-1407, Am. Soc. of Civ. Eng., New York, 1993. Educ. Sci. and Cult. Int. Sci. Res. Program,Tokyo, 1994.
Savage,S. B., and K. Hutter, The motionof a finite massof Terzaghi, K., Stressconditionsfor the failure of concreteand
granularmaterial down a rough incline,•. Fluid Mech., !99, rock, Proc. Am. Soc. Test.Mater., 45, 777-801, 1945.
177-215, 1989. Thomas,D. G., Transportcharacteristics of suspension,
VII, A
Savage,S. B., and K. Hutter, The dynamicsof avalanchesof note on the viscosityof Newtoniansuspensions of uniform
granular materials from initiation to runout, I, Analysis, sphericalparticles,J. ColloidSci.,20, 267-277, 1965.
Acta Mech., 86, 201-223, 1991. U.S. GeologicalSurvey(USGS), Debris flow hazardsin the
Savage, S. B., and S. McKeown, Shear stressesdeveloped San Francisco Bay region, U.S. Geol. Surv. Fact Sheet,
duringrapid shearof denseconcentrations of large spher- 112-95, 4 pp., 1995.
ical particlesbetweenconcentriccylinders,or.Fluid Mech., Vaid, Y. P., and J. Thomas,Liquefactionand postliquefaction
;[27, 453-472, 1983. behaviorof sand,J. Geotech.Eng., 121, 163-173, 1995.
Savage,S. B., and M. Sayed,Stressesdevelopedin dry cohe- Vallance, J. W., Experimentaland field studiesrelated to the
sionlessgranularmaterialsshearedin an annularshearcell, behaviorof granular massflows and the characteristics of
Z Fluid Mech., ;[¾2,391-430, 1984. their deposits,Ph.D. dissertation,197 pp., Mich. Technol.
Savage,W. Z., and W. K. Smith, A model for the plasticflow Univ., Houghton, 1994.
of landslides,U.S. GeologicalSurveyProf.Pap., ;[385,32 pp., Vallance, J. W., and K. M. Scott, The Osceola mudflow from
1986. Mount Rainier: Sedimentologyand hazard implicationsof
Scheidegger, A. E., On the predictionof the reachand velocity a huge clay-richdebrisflow, Geol. Soc.Am. Bull., 109(2),
of catastrophiclandslides,Rock Mech., 5, 231-236, 1973. 143-163, 1997.
Scott,K. M., J. W. Vallance, and P. T. Pringle, Sedimentology, Vanoni, V. A. (Ed.), Sedimentation Engineering,
745 pp., Am.
behavior, and hazards of debris flows at Mount Rainier, Soc. of Civ. Eng., New York, 1975.
Washington,U.S. Geol.Surv.Prof.Pap.,;[5¾7,56 pp., 1995. Varnes, D. J., Slope movementtypesand processes,in Land-
Sharp, R. P., and L. H. Nobles, Mudflow of 1941 at Wright- slides--Analysisand Control, edited by R. L. Schusterand
wood, southern California, Geol. Soc. Am. Bull., 6¾, 547- R. J. Krizek, Spec. Rep. Natl. Res. Counc. Transp. Res.
560, 1953. Board, 176, pp. 11-33, Natl. Acad. of Sci., Washington,
Shen, H., and N. L. Ackerman, Constitutiverelationshipsfor D.C., 1978.
solid-fluid mixtures,or.Eng. Mech. Div. Am. $oc. Civ. Eng., Vignaux, M., and G. J. Weir, A general model for Mt. Rua-
;[08, 748-763, 1982. pehu lahars,Bull. Volcanol.,52, 381-390, 1990.
Shibata, M., and C. C. Mei, Slow parallel flows of a water- Vreugdenhil, C. B., Numerical Methods for Shallow-Water
granule mixture under gravity, I, Continuum modeling, Flow, 261 pp., Kluwer Acad., Norwell, Mass., 1994.
Acta Mech., 63, 179-193, 1986a. Walton, O., Numerical simulationof inelastic,frictionalparti-
Shibata,M., and C. C. Mei, Slow parallel flows of a water- cle-particleinteractions,in ParticulateTwo-PhaseFlow, ed-
granulemixture under gravity,II, Examplesof free surface ited by M. C. Roco, pp. 884-911, Butterworth-Heinemann,
and channel flows, Acta Mech., 63, 195-216, 1986b. Newton, Mass., 1993.
Shieh, C.-L., Jan, C.-D., and Y.-F. Tsai, A numerical simula- Weir, G. J., Kinematicwave theoryfor Ruapehulahars,N. Z.
tion of debris flow and its application,Nat. Hazards, ;[3, J. Sci., 25, 197-203, 1982.
39-54, 1996. Whipple, K. X., Debris-flow fans: Processand form, Ph.D.
Shlemon,R. J., R. H. Wright, and D. R. Montgomery,Anat- dissertation,205 pp., Univ. of Washington,Seattle, 1994.
omy of a debrisflow, Pacifica,California, in DebrisFlows/ Whipple, K. X., and T. Dunne, The influenceof debris-flow
Avalanches:Process,Recognition,and Mitigation, Rev. Eng. rheology on fan morphology,Owens Valley, California,
Geol.,vol. 7, editedby J. E. Costaand G. F. Wieczorek,pp. Geol. Soc. Am. Bull., 104, 887-900, 1992.
181-200, Geol. Soc. of Am., Boulder, Colo., 1987. Wieczorek,G. F., E. L. Harp, R. K. Mark, and A. K. Bhatta-
Siebert,L., Largevolcanicdebrisavalanches;
Characteristics
of charyya,Debris flows and other landslidesin San Mateo,
sourceareas, deposits,and associatederuptions,Z Volca- Santa Cruz, Contra Costa, Alameda, Napa, Solano,
nol. Geotherm. Res., 22, 163-197, 1984. Sonoma,Lake, and Yolo counties,and factorsinfluencing
Sitar, N., S. A. Anderson, and K. A. Johnson, Conditions for debris-flow distribution, in Landslides, Floods and Marine
initiation of rainfall-induced debris flows, in Stabilityand Effectsof the Storm of January$-5, 1982, in the San Fran-
Performanceof Slopesand EmbankmentsII Proceedings, pp. ciscoBayRegion,California,editedby S. D. Ellen and G. F.
834-849 Geotech. Eng. Div., Am. Soc. of Civ. Eng., New Wieczorek, U.S. Geol. Surv.Prof. Pap., 1434, 133-162, 1988.
York, 1992. Wierich, F. H., The generationof turbidity currentsby sub-
Skempton,A. W., Long-termstabilityof clayslopes,Geotech- aerial debris flows, Geol. Soc.Am. Bull., 101,278-291, 1989.
nique, ;[¾,75-101, 1964. Yamashita, S., and K. Miyamoto, Numerical simulation
Skempton,A. W., Residual strengthof clays in landslides, method of debris movementswith a volcanic eruption,
foldedstrataand the laboratory,Geotechnique, $5, 3-18, 1985. paper presentedat Japan-U.S. Workshop on Snow Ava-
Spiegel,M. R., TheoreticalMechanics,Schaum'sOutlineSeries, lanche, Landslide and Debris Flow Prediction and Control,
368 pp., McGraw-Hill, New York, 1967. Jpn. Sci. and Technol.Agency,Tsukuba,1991.
Suwa,H., Focusingmechanismof large bouldersto a debris- Yano, K., and A. Daido, Fundamentalstudyon mud-flow,Bull.
flow front, Trans.orpn.Geomorphol.Union,9, 151-178, 1988. DisasterPrey. Res. Inst. Kyoto Univ., 14, 69-83, 1965.
Takahashi, T., Mechanical characteristicsof debris flow, or. Zhang, Y., and C. S. Campbell, The interface betweenfluid-
Hydraul. Div. Am. Soc. Civ. Eng., ;[0¾,1153-1169, 1978. like and solid-like behaviour in two-dimensionalgranular
Takahashi,T., Debris flow on prismaticopen channel,or.Hy- flows,J. Fluid Mech., 237, 541-568, 1992.
draul. Div. Am. Soc. Civ. Eng., ;[06, 381-396, 1980.
Takahashi,T., Debris flow,Annu. Rev. Fluid Mech., ;[3, 57-77,
1981. R. M. Iverson,CascadesVolcano Observatory,U.S. Geo-
Takahashi,T., DebrisFlow, 165 pp., A. A. Balkema, Brook- logical Survey, 5400 MacArthur Boulevard,Vancouver,WA
field, Vt., 1991. 98661. (e-mail: riversøn@usgs'gøv)

The Physics of Debris Flows PDF

Uploaded by

Copyright:

Available Formats

The Physics of Debris Flows PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The Physics of Debris Flows PDF

Uploaded by

Copyright:

Available Formats

THE PHYSICS OF DEBRIS FLOWS

Abstract. Recent advancesin theory and experimen- sorted,water-saturatedsedimentinvariablymove as an

1. INTRODUCTION 1987],but the necessityof interactingsolid and fluid

summarizestypicalL/H valuesinferred from the distal

TABLE 1. EstimatedValuesof Total Flow Volume,RunoutDistanceL, DescentHeight H, and EfficiencyL/H

bris-flowmass,independentof changesin composition, changewith the bed andbanks,whichmaydiffergreatly

Figure 3. U.S. Geological Survey

3000 Basalpore 2OOO _ Basal

TIME, IN SECONDS TIME, IN SECONDS

Figure 5. Representativemeasurements of flow depth,basaltotal normalstress,and basalfluid pressure

Laser Mean Bed Mean Bulk Dried Bulk Calculated Mean

Bulk densitieswere determinedon the basisof (1) simultaneous

_ & Osceola Mudflow

Figure 6. Hydraulic permeabilitiesof representativedebris-flowmaterialsas a functionof porosity(fluid

1975, p. 25]. This critical grain size correspondsquite by Pdry-- PsUfines(

April 19, 1994 sand-gravel

minimized.Comparison of pf loidal particlesare significant[Coussot,1995].Nonethe-

= -p) tan (b+ c (6)

Here ß is the average shear or driving stressOn the

energyis availableto agitate the failing mass.This type Rock spray

well as debrisflows [Jaegerand Nagel, 1992;Zhang and

evidence for much of the behavior summarized here.

tiple surgesoccurin individualdebrisflows,eachexhib-

andAckerman,1982], computationalexperiments[Hop- 1.0

kins et al., 1993] and physicalexperiments[Iversonand • 0.8

granular temperaturesand nonequilibriumfluid pres-

(Figure 13) [Major, 1996] [cf. Hampton, 1979;Pierson,

tdif = h21x/kE (8)

SAND-GRAVEL MIX 04-19-94 LOAM-GRAVEL MIX 08-31-94

- us )] originallyusedby Bagnold[1954] solid-fluid interactions. At least this many processes

TABLE 5. Estimation of DimensionlessParametersThat CharacterizeStressesin a Range of Prototypical

as heads of surges)where grains coarser than sand

Flow mobilizesfrom rigid slopefailure without changesin constitutiveproperties X

gravityimparts a stressthat increaseswith depth below

Op/Ot+ V. (pv) = ms+ mf (23) p[Ov/Ot

and fluid velocities normal to the bed, which are ne-

Guidancefor simplifyingdebrisflow mixture theory

dv txvf buoyant (the case of Baghold [1954]). If the fluid is

1 1 100 0.01 -5 0 - 1 1.01 0.002 0.002 0.00201

Cases1-5 were selectedto illustrate the effectsof variationsin e, m•/m2, and B.

inputof fluctuationenergyfrom the bedis required,and ThesumVup(tup

7.2. Viscous Case (A > 0) Consequently,increasedviscosityreduces the average

Inviscid fluid Inviscid fluid

Figure19. Pictorialsummaryof relativefluctuationvelocitiesof grainm 2 duringonecycleof verticalmotion

movestoward the bed, the term in bracketsin (58b) is

1.0 mixturesreasonablywell [cf. Lambe and Whitman,1979,

p. 320; Beenand Sills, 1981].

This maybe interpretedas an advection-diffusion

USGS flume sand-gravelmix 10-13 20 0.3 0.001 0.02 200 0.1 4 X 10 -4 20

Furthermore,Ph • Pbis usuallya reasonableapproxi-

Although quantitativepredictionsof (71) and (72)

proachusesthe kinematicwaveapproximation,in which andPTaswell asthe totaldebrisflowmassanddensity

y scaleis that of free fall in the x direction,(1--/g)1/2,

P•,at+ vy•-•+Vxax/ - ay a•- Oy

vy=0 at y=0 (79b) relatingTs(xx)

As wasnotedby Savageand Hutter [1989],valuesof • in Oh O(h•)

ponents.Although (82) doesnot containpore pressure (h -y) dy = •h

where h - Pbedis the nondimensional

tative of many others but was distinguishedby particu-