Granular memory and its effect on the triggering and ...raman/papers2/FriedmannJGR.pdfGranular memory and its effect on the triggering and distribution of rock avalanche events S

Granular memory and its effect on the triggering and distribution of

rock avalanche events

S. J. FriedmannDepartment of Geology, University of Maryland, College Park, Maryland, USA

G. Kwon and W. LosertDepartment of Physics, International Solar-Terrestrial Physics and Institute for Research in Electronics and Applied Physics,University of Maryland, College Park, Maryland, USA

Received 27 August 2002; revised 13 February 2003; accepted 21 March 2003; published 15 August 2003.

[1] Rock avalanches are uncommon, large, catastrophic events in which granularmasses of rock flow at high speeds, commonly with unusually long runouts. Necessaryconditions for rock avalanche occurrence include high relief, steep slopes, and aprefractured (granular) groundmass. An additional parameter, antecedent shear of thegranular mass in the direction of failure, may be prerequisite. In many instancesthe basement mass of the eventual landslide underwent internal strain subparallel to thedirection of mass wasting, commonly focused along a basal surface (e.g., fault). Newexperimental data from laboratory experiments indicate that shear of a granular massproduces subtle anisotropy in grain arrangements that strongly affects the subsequentdistribution of deformation and rearrangement, the failure threshold, and the rate atwhich steady state conditions are reached. This phenomenon may help to explain thespatial clustering of rock avalanche deposits. If memory of prior shear is also relevanton larger scales, taking deformation history into account could greatly enhanceprediction of rock avalanche occurrence. INDEX TERMS: 1815 Hydrology: Erosion and

sedimentation; 1869 Hydrology: Stochastic processes; 7223 Seismology: Seismic hazard assessment and

prediction; 8010 Structural Geology: Fractures and faults; 9820 General or Miscellaneous: Techniques

applicable in three or more fields; KEYWORDS: Rock avalanche, granular flow, landslide hazard, seismic

triggering

Citation: Friedmann, S. J., G. Kwon, and W. Losert, Granular memory and its effect on the triggering and distribution of rock

avalanche events, J. Geophys. Res., 108(B8), 2380, doi:10.1029/2002JB002174, 2003.

1. Introduction

[2] Rock avalanches are spectacular and catastrophicexamples of natural granular flows of great magnitude.Workers have recognized deposits from these events for over100 years under a variety of circumstances and settings [e.g.,Heim, 1932; Shreve, 1968; Friedmann, 1997]. Tectonicallyactive regions constitute by far the most common setting forrock avalanching [Keefer, 1984a, 1993], and while someevents started due to undercutting [e.g., Daly et al., 1912;Heim, 1932], the vast majority are believed to have begun dueto ground accelerations associated with seismic events [e.g.,McSaveney, 1978; Hadley, 1978]. These events are commonin arid and semi-arid settings [e.g., Keefer and Wilson, 1989;Yarnold, 1993], although they occur in humid settings,undersea [e.g., Moore et al., 1989], and on other planets[e.g., Howard, 1973; Luchitta, 1979; McEwan, 1989].[3] They also represent a significant hazard, associated

historically with >20,000 deaths and major property de-struction in populated areas [e.g., Pflaker and Ericksen,1978]. Historic landslides in these areas have generated

eyewitness accounts that provide key insight into thephysics of rock avalanche transport. However, there areno eyewitness accounts of rock avalanche triggering. Be-cause these events are uncommon and difficult to predict,little is known about how to best predict their occurrence inhigh-risk regions, including seismically active areas.[4] This paper attempts to improve the understanding of

probability of event occurrence and distribution by examin-ing the role of basement deformation history on triggeringthrough a combination of field, literature, and experimentalapproaches. Field and literature studies focus on the directionof basal deformation relative to the direction of flow, as afactor that may influence avalanche probability and runoutlength. The experimental work specifically builds uponrecent advances in the physics of granular materials.

2. Rock Avalanche Occurrence

[5] Much previous work has focused on the physics oftransport for rock avalanches. Their unusual physics centerson a key feature: exceptionally long runout. On earth, rockand debris avalanches runout between 3 and 10 times theirfall heights; on Mars, over 25 times [Shaller, 1991]. Thisrequires extremely low internal angles of friction, often less

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 108, NO. B8, 2380, doi:10.1029/2002JB002174, 2003

Copyright 2003 by the American Geophysical Union.0148-0227/03/2002JB002174$09.00

ECV 8 - 1

than 10� (most natural landslides, for comparison, haveinternal fiction angles near 30�). This long runout, farbo-schung or drop height versus runout length (H/L), is whatmakes rock avalanches both extremely devastating anddifficult to predict. Despite hundreds of examples and obser-vations, there is no agreement among workers concerning thephysics that produce this effect. Many disparate transportmodels exist, including air lubrication [Shreve, 1968], inelas-tic collision [Bagnold, 1954;Campbell et al., 1995] frictionalmelting of the rocks themselves [Erissman, 1979], andfluidization due to internal vibrational modes of the rocks(known under the potentially misleading name of acousticfluidization) [Melosh, 1979, 1987]. Most of these models areconsistent with local observations but inconsistent withuniversal observations or physical constraints [Shaller,1991; Yarnold, 1993; Friedmann, 1997]. To date, few ofthese models have received direct tests through physical ornumerical experiments (exceptions include Campbell et al.[1995] and Hsu [1975]). No tests have attempted to deter-mine the importance of various processes within thesemodels, nor attempted to parameterize key inputs. Moreover,no model currently predicts under what circumstances longrunout will occur, nor what that runout will be.[6] Relatively little work has focused on the necessary

conditions for rock avalanche initiation. Keefer [1984b]documented three critical preconditions for flow initiationby examining the basement terrains of many rock avalanchesystems. He demonstrated that high relief (>150 m dropheight), steep slopes (>25�), and a prefractured (granular)groundmass were necessary conditions for flow initiation(Figure 1a). The last point, a prefractured groundmass, iscentral to this discussion because it implies that insightsfrom granular physics may help to better understand theinitiation process. In locations without a penetrativelyfractured basement, failure initiates in other modes (e.g.,rock slide, slump) that do not exhibit long runout. Keefer[1993] continued to assess the potential for rock avalanche

triggering by looking for factors that would increase thepotential for flow-failure mode (e.g., bedding surfacesdipping out of the rock face), producing a logic-tree forgauging the relative potential for flow initiation. Schmidtand Montgomery [1995] argue that some critical relief mustbe exceeded before deep-rooted failure can occur. In short,as relief increases, the force of the rock’s mass ultimatelyexceeds the local yield strength, determined by a variety ofgeological factors. This is consistent with Keefer’s [1984b]recognition of the minimum critical relief threshold for rockavalanche failure.[7] In recent and ancient settings, rock avalanche depos-

its exhibit a dense clustering of events wherein successiveevents occur directly on top of or adjacent to previousevents. Although many studies document this spatialclustering [e.g., Mudge, 1965; Shreve, 1968; Eppler etal., 1987], they have not examined the underlying causesthat lead to multiple events within a restricted spatialframework. Other aspects of the spatial distribution (e.g.,why an event initiates in one portion of the range versusanother) also have not received significant research. Thespatial distribution may hold further clues about theavalanche triggering mechanism and valuable informationfor hazard assessment.[8] To shed light on these key issues, we have chosen

three examples from the literature in which the spatialoccurrence of the rock avalanche event is noteworthy andpoints to additional root circumstances to failure. Thesestudies include two exceptionally well know cases, onerecent and one historic, and one ancient case with a largenumber of rock avalanche occurrences.

3. Blackhawk Landslide

3.1. Background

[9] Along with the Frank, Elm, and Sherman Glacierslides, the Blackhawk serves as an archetypical example of

Figure 1. Source characteristics for avalanching behavior. (a) Plot of relief (m) versus slope (degrees)for 47 rock avalanche deposits. All events require an initial slope >25� (after Keefer [1984b]). (b) Plot offailure angles for steady state avalanches. Height (h) is measured in grain diameters (d). For flows thickerthan 10 grain diameters, failure also requires a 25� slope (after Daerr and Douady [1999a]).

ECV 8 - 2 FRIEDMANN ET AL.: GRANULAR PHYSICS AND ROCK AVALANCHE TRIGGERING

rock avalanche deposits. One of the best studied examplesin the world [Shreve, 1968; Johnson, 1978; Shreve, 1989],the landslide sits in the northern San Bernardino Mountainsnear Apple Valley, CA, and comprises 3 � 108 m3 ofgranulated rock (Figure 2). It features many characteristicsclassically attributed to rock avalanche deposits, including apenetratively deformed granular framework, ‘‘jigsaw’’ brec-cias, and inherited stratigraphy and structures. Transversewaves of constant height and spacing transect the surface ofthe slide, developed either during or near the cessation offlow (Figure 2b). The flow is broadly lobate, and displaysprominent raised levees along the edges of flow. The slideran out very far, with an estimated internal angle of frictionnear 10�.[10] Despite it’s prominence in the landscape and litera-

ture, it is only the best exposed and expressed of severallandslides in that area. At least one and potentially two otherevents called the Silver Reef slide(s) occur adjacent to theBlackhawk [Shreve, 1968] (Figure 2e). Many of the samefeatures found within the Blackhawk are found in the SilverReef slides, including transverse flow waves, a penetrativelydeformed granular framework, long runout, ‘‘jigsaw’’ brec-cias, and inherited stratigraphy and structures. On the basis oftheir quality of preservation and shallow burial depth, theselandslides are probably late Pleistocene and close in age tothe Blackhawk, and most workers consider the Silver Reefevent(s) to be identical in origin and nature [Shreve, 1968;Shaller, 1991].

[11] Importantly, these landslides are the only Quaternaryrock avalanche deposits to be found nearby. The nearestdocumented rock avalanche deposit, the Martinez MountainLandslide, occurs over 150 km to the south [Jennings,1973]. However, the other preconditions for rock avalanchetriggering are maintained along the range front, includingsteep, high-relief terrain, fractured basement, and beddingsurfaces dipping out of the rock face. This region is wellknown for large seismic shaking associated with earthquakefaults and is part of an active fold/thrust belt severalhundred kilometers in length [e.g., Agnew et al., 1992;Hutton et al., 1991]. Despite the abundance of seismic-triggering events in the Pleistocene, there are no other rockavalanche events exposed along the San Bernardino Moun-tains, along strike, north or south side. This suggests thatlocal characteristics in Blackhawk canyon are also critical tothe initiation of rock avalanche behavior.

3.2. Basement Configuration

[12] The source regions for the Blackhawk and Silver Reeflandslides are well exposed and easily accessed reentrantswithin the San Bernardino mountain front (Figure 3). Severalworkers [Shreve, 1968; P. Sadler, unpublished maps] havemapped the basement source region for these landslidesand documented key stratigraphic and structural relation-ship. One relationship is noteworthy. Along much of therange front, older reverse faults crop out near the pied-mont. These are mostly north vergent, dipping to the south

Figure 2. Blackhawk Landslide, California. (a) Birds-eye view looking southward of the landslide andthe source region. Landslide is 1 mile wide. (b) Close-up image of transverse flow waves from landslidesurface. Ridges are commonly 4–6 m in relief and spaced �50 m apart. (c) Granular framework oflandslide with inherited bedding features; white bar is 2 m. (d) Jigsaw breccia within landslide (close-upof Figure 2c). (e) Trace of map view extent of Blackhawk and Silver Reef rock avalanche deposits, north= top. Note close spatial arrangement and potential for multiple Silver Reef lobes or events.

FRIEDMANN ET AL.: GRANULAR PHYSICS AND ROCK AVALANCHE TRIGGERING ECV 8 - 3

(Figure 4a). In Blackhawk Canyon reentrants, however,the Voorhies Thrust and its imbricates dip both to thesouth and north due to folding by lower structures. Thisgeometry is due to a local structural culmination.[13] Here, both the antecedent strata and the subjacent

bounding fault dip northward toward the adjacent basin(Figure 4b). This means the material above the fault under-went strain synthetic and sub-parallel to the vector of land-slide initiation. In other words, before exhumation, thebasement rocks underwent bulk shear in the same directionas the subsequent flow to a first order, moving out of therange front and downward into the basin. This is true onlywithin the Blackhawk source reentrant. Elsewhere along therange front, the basement rocks and their key structuralsurfaces dip in the opposite direction from previous shear.

4. Frank Landslide

4.1. Background

[14] The Frank Landslide in Alberta is Canada’s greatestlandslide disaster and one of its 10 greatest natural disasters(Figure 5). More than 12% of the population of Frank, at least70 people, died in the event, and 75% of the homes weredestroyed [Daly et al., 1912; Anderson, 1968]. The failurewas instigated by coal mining underneath the eventual failureplane, similar to the undermining of the Elm landslidebasement [Heim, 1932]. The rock avalanche flowed at anestimated 40 m/s, moving across the valley and up theopposite side 130 m and leaving a thickness of debris up to45 m. Volume estimates are on the order 3 � 107 m3.[15] Although the Frank event is the only major historic

rock avalanche in the vicinity of Turtle Mountain, there isevidence of previous major landslides there. Local NativeAmericans reportedly avoided Turtle Mountain, referring to

Figure 4. Simplified diagrams of basement structure inBlackhawk source region [after Shreve, 1968]. Each crosssection is roughly 1.5 km in length and runs N-S acrossnorthern San BernardinoMountains. (a) Cross section 1.3 kmeast of Blackhawk Canyon axis, along Miles Canyon. Notethe dip of Voorhies and Santa Fe thrusts out of the mountaindue to subjacent folding. (b) Cross section along east edge ofBlackhawk Canyon. Note the southward tip of the Voorhiesthrust and its imbricates shallow toward the landslide scarp,suggesting flattening and northward dip within the canyon,synthetic and subparallel to the landslide paleoflow vector.See color version of this figure at back of this issue.

Figure 3. Evacuated source area reentrants of Blackhawk and Silver Reef landslides, birds-eye view.Upper arrow points to Blackhawk Canyon. Note contact between light and dark units dipping at �15�into valley at right. This contact is the folded Voorhies thrust. The lower arrow points into GrapevineCanyon, the source area for the Silver Reef event. Arrow �0.5 miles in length.


it as ‘‘the mountain that walks.’’ There are many other typesof landslide deposits around the mountain as well, includingslumps and debris flows.


[16] Since the Frank event is the only clear rock avalancheevent that crops out near Turtle Mountain, there is no case forclustering of rock avalanches along strike or nearby. How-ever, the specific locus of failure is of interest (Figure 6).Early workers argued that the slide initiated along jointsurfaces within the basement [McConnell and Brock, 1904;Daly et al., 1912] or along bedding surfaces [e.g., Allen,1933; Norris, 1955]. The joint networks are critical to thestory in order to create a penetratively fractured basement,forming a granular mass. However, much of the length ofTurtle Mountain shares this characteristic with the slide area,making this condition nonunique. On this basis, workersargue that the likelihood of future events is relatively high.[17] Subsequent mapping of the source area for the slide

reveals important local distinctions [Cruden and Krahn,1978]. Within the slide area, the failed strata overlie a minorimbricate reverse fault that verges eastward out of the rangefront. Elsewhere along the range front, this fault dipswestward and the overlying rock mass moved to the east.Within the slide source region, however, the reverse faultdips eastward, so that the basal slip plane, the overlyinggranular mass, and the rock avalanche flow azimuth are allcoincident and verge eastward (Figure 6c). In addition,bedding-parallel flexural slip occurred along stratal bound-aries within the slide region, also with the same sense ofvergence (see their photo 12). Thus, within the sourceregion, there were multiple modes of shear and deformationdown and outward into the basin.[18] Cruden and Krahn [1978] also collected shear

strength measurements along various surfaces within theFrank event source region. They determined that the faultand flexural-slip surfaces were significantly weaker thanother anisotropic surfaces (joints and bedding planes), evenweaker than a diamond-saw cut. This is consistent withwork-softening along these surfaces in the direction ofvergence. This is reinforced by the observation that deep-seated creep apparently preceded the slide event.

5. Shadow Valley Basin

5.1. Background

[19] The Shadow Valley basin is a Miocene supradetach-ment basin of the eastern Mojave Desert [Friedmann et al.,1996; Friedmann, 1999]. It occurs above the KingstonPeak-Halloran Hills detachment fault, a west-vergent, low-angle normal fault [Davis et al., 1993]. This fault cuts older,east-vergent reverse faults in the Clark, Mesquite, andMescal Mountains, reversing their sense of shear (Figure 7).[20] The Shadow Valley basin contains many large vol-

ume mass-wasting deposits (30–50%), including megabrec-cias, glide-blocks, lahars, and debris flows [Friedmann etal., 1996; Friedmann, 1997; Bishop, 1997]. More than19 megabreccias crop out within the basin, representingnearly 10% of basin fill with a minimum volume of 3.5 �109 m3. All are interpreted to be rock avalanche deposits.This density and clustering of events is typical of othersupradetachment basins, many of which contain abundant,

Figure 5. The Frank Landslide, Alberta, Canada. (a) Theslide scarp and headwall on the east side of Turtle mountain.Photo courtesy of E-an Zen. (b) Birdseye view of thelandslide toward the east. Note railroad track crossinglandslide. Reproduced with the permission of the Ministerof Public Works and Government Services, Canada, 2003,and courtesy of Natural Resources Canada, GeologicalSurvey of Canada. (c) Granular framework for the slide.Reproduced with the permission of the Minister of PublicWorks and Government Services, Canada, 2003, andcourtesy of Natural Resources Canada, Geological Surveyof Canada.


large megabreccia deposits [e.g., Hodges et al., 1989;Topping, 1993; Yarnold, 1994; Friedmann and Burbank,1995; Ingersol et al., 1996; Miller and John, 1999].


[21] The source region for the Shadow Valley rockavalanches is well constrained. On the basis of directionalflow features within the deposits and their lithology, themegabreccias originated in a structural reentrant along theeastern basin edge [Friedmann et al., 1996; Friedmann,1997]. The primary fault surface that formed the reentrant iscurviplanar and corrugated, and rock avalanches preferen-tially formed within these corrugations (Figure 8).[22] The sense of slip along the basin-bounding fault was

westward, dipping gently out of the range front. Rockavalanches entering the basin from the range-bounding faultwould have flowed in the same direction, probably initiatingalong synthetic faults or fractures near the main fault. Thisis consistent with sense-of-shear indicators in both the rockavalanche deposits and basement structures within thereentrants.

6. Application of Granular Physics:Granular Memory

[23] Several mechanisms behind rock avalanche phenomenamay be understood from granular physics experiments.Studies of the physics of granular matter focus on particlesthat are orders of magnitude smaller (on the submillimeter tocentimeter scale), and often round. The aim is to reveal thephysical principles that are generic to a system of many,inelastic particles that interact through direct contact. Someeffects may be purely geometric, others due to frictionbetween particles. For a review of recent studies of granular

Figure 7. The Shadow Valley basin, eastern MojaveDesert. Map showing location of basin and KingstonPeak-Halloran Hills Detachment Fault (thick dashed line).A, Avawatz Mountains; CM, Clark Mountain; HH, HalloranHills; KPP, Kingston Peak Pluton; MM, Mesquite Moun-tains; MR, Mescal Range; N, Nopah Range; SM, ShadowMountains; SDVFZ, Southern Death Valley Fault Zone.

Figure 6. Several cross sections through Turtle Mountain,basement for the Frank event. In all cross sections west = Land east = R. (a) Through north peak of Turtle Mountain,after Daly et al. [1912]. (b) Through south peak of TurtleMountain, after Norris [1955]. Note prominent fold beneaththe Turtle Mountain fault. (c) Through southern slide area,after Cruden and Krahn [1978]. Note the small, imbricatethrust above the Turtle Mountain fault dipping out of thebedrock surface.


flow, see, for example, Jaeger et al. [1996], Clement [1999],and Rajchenbach [2000].[24] How can rock avalanche properties be scaled down

for comparison to simple granular shear flow models? Thescaling law depends on the specific contact forces used inthe model. For the simplest model of inelastic Hertzian

contact forces between spheres, if the lengths are scaled bya factor c, time and dissipation will have to scale as Sqrt(c),and the elastic constant requires a more complicated rescal-ing [Poeschel et al., 2001]. However, several granular shearflow properties appear robust, independent of shear rates,pressures, or particle properties. These flow properties, i.e.

Figure 8. (a) Map of eastern detachment, including truncated basement units. Both detachment andbasement strata dip toward the west into the basin parallel to the fault’s vergence and displacement.(b) Outcrop of a portion of the KHDF; image approximately 40 m high. Note thick band of shearedbreccia along fault. (c) Close-up of the fault; field of view approximately 2 m across. Note theasymmetric sense of shear (top to the left) and the granular nature of the gouge. (d) Strata from the lowerShadow Valley basin. The four dark ridges are stacked rock avalanche deposits. Image is approximately500 m wide and shows �150 m strata.


that shear is localized in a shear band, and that the shearstress is roughly independent of shear rate, may manifestthemselves also in rock avalanches. Other effects such asfrictional heating or comminution are not captured in scaledexperiments.[25] Here we describe recent work on the physics of

granular shear flow with a focus on the start of flow. Weapply these results to provide a possible explanation forclustering of avalanche locations and for the dependence ofslide regions on antecedent shear history.

6.1. Experimental Approach

[26] To investigate granular shear flows, the material(1 or 2 mm glass beads in most experiments) is placed into aCouette cell, which consists of two independently moveableconcentric cylinders, as shown in Figure 9a. The gap betweencylinders is 44 mm, and a glued-on layer of glass beadsroughens the cylinder walls. The motion of particles on thetop surface is imaged with a high-speed CCD camera that cancapture up to 3000 frames/s (see inset, Figure 9a). Particletracking software allows us to extract the motion of grains ata resolution of�1/20 of one pixel for up to�2000 grains perframe. This permits measurements of the time-dependentvelocity of particles averaged over different regions awayfrom the shear surface, as indicated in Figure 9b.[27] The height of the granular surface, a measure of the

average density, is determined by imaging a region similarto the one shown in the inset from a shallow angle.[28] In addition to steady state shear flow of the inner

cylinder, the start of shear flow is studied in great detail.Preliminary results are presented by Losert and Kwon[2001]; complete results and a detailed discussion of thephysics will be presented elsewhere (M. Toiya et al., Shearreversal in granular flows, submitted to Physical Review E,2003) (hereinafter referred to as Toiya et al., submittedmanuscript, 2003).

6.2. Results

[29] In steady state shear flow, velocity gradients areconfined to a thin shear band, roughly 5 particle diameterswide [Losert et al., 2000; Mueth et al., 2000]. Figure 9bshows particle velocity as a function of distance from theshear surface normalized by the shear rate for a large range ofspeeds and pressures, indicating that shear banding is notsensitive to pressure or shear rate. The location of the shearband is sensitive to the flow geometry and remains near thesmaller surface area of the inner cylinder at all times, evenwhen both cylinders, or only the outer cylinder are rotated[Losert and Kwon, 2001].[30] How does this shear band start? When the flow is

stopped and then restarted in the same direction (Figure 10a)the shear band forms immediately, and a steady state velocityprofile is reached before the inner cylinder is moved by oneparticle diameter. This implies that the configuration of grainsretains an imprint of the earlier shear history, leading to animmediate transition to steady state.[31] However, when the flow is restarted in the direction

opposite to the prior shear, particles far from the shearedsurface move significantly during the initial transient(Figure 10b) [Losert and Kwon, 2001]. The height ofthe granular column, which indicates the density of thematerial, is shown in Figure 11a as a function of time after

Figure 9. (a) Schematic and image of a granular shearcell. Granular material (1 mm glass beads) is confined in a44 mm gap between movable inner and outer cylinders. Thebottom of the shear cell can move with either cylinder. Flowof material is imaged from above with a high-speed CCDcamera (see inset). The average velocity of particles withintwo regions close to the inner cylinder will be indicated assolid and dotted lines respectively in subsequent figures. (b)Distance from the inner shear surface as a function ofvelocity (normalized by the shear rate). Shear is confined toa thin shear band, �5 particle diameters wide. The shear-band width is roughly exponential, independent of shearspeed (from 0.004 to 0.3 rotations/s) or pressure (up to morethan 100 N/m2).


a restart. When shear is restarted in the same direction, thedensity does not change. When shear is restarted in theopposite direction, however, the material compacts by about0.5 particle diameters over a characteristic length of severalparticle diameters. In other words, the granular frameworkcollapses significantly (Toiya et al., submitted manuscript,2003). The system then slowly dilates back to the originaldensity over longer timescales. Whether this gradual dila-tion is accompanied by a transient increase in shear strength

has not yet been determined and is the subject of currentexperimentation.[32] A potential mechanism for the behavior under shear

reversal is shown in Figure 11b. During shear, particlecontacts are preferentially in the direction of the main shearstress, at about 45�. When shear is restarted in the oppositedirection those contacts break, and the system can rearrangewithout much shear resistance and compact slightly.[33] This granular memory of the direction of prior shear

thus appears as the result of shear anisotropies. Two dimen-sional experiments on birefringent disks have shown that thenetwork of load bearing particle contacts becomes stronglyanisotropic (B. Behringer, personal communication, 2002) Inaddition, experiments have shown an additional kind ofgranular memory. Vertically shaken granular matter of ex-actly the same initial density will evolve differently, depend-ing on the magnitude of prior shaking [Josserand et al.,2000]. In free flows on a sand pile, the flow properties wereshown to depend on the pile preparation, in addition to theexpected dependence on pile density [Daerr and Douady,1999b]. For rocks, this implies that in addition to the densityof the arrangement, rock bodies may be stacked in a way thatreflects the direction and magnitude of prior shear.

7. Discussion

[34] The clustering effects of multiple rock avalancheevents and the specific occurrence of individual events canbe explained through the application of granular physics toflow initiation and transport. Specifically, the most recent

Figure 11. (a) A drop in layer height of �0.5 particlediameters accompanies shear reversal. Restart in the samedirection does not alter the layer height. The shear speed is0.25 particle diameters/s. (b) Possible particle rearrange-ment during shear reversal. During shear, particles prefer-entially are in contact along the shear direction; duringreversal, these contacts break, and particles can move untilnew contacts form along the new shear direction.

Figure 10. (a) Average velocity of particles in differentregions (two of the regions are highlighted in Figure 9a). Asshown below, shear is stopped and restarted in the samedirection at t = 0 s. The velocity in each region reachessteady state immediately. (b) Same as Figure 10a, exceptthat shear is started in the opposite direction. Particles farfrom the shear surface move significantly when the shear isstarted.


previous shear sense is a critical control on the subsequentshear for granular rock assemblages in terms of locus offailure, material yield strength, distribution of shear withinthe granular mass, and time needed to approach steady state.Importantly, the experimental results demonstrate that shearinitiation in the direction opposite the previous shear direc-tion requires the reorganization of a large granular volume toaccommodate the new direction of shear stress. It is reason-able to assume that all of these factors affect the initial failureconditions of rock avalanches. It is also reasonable to assumethat these factors affect the early history of granular interac-tion, specifically the density of the granular network and thenumber of collisions between grains during flow.[35] As argued by Bagnold [1954], Melosh [1987], and

Campbell et al. [1995], energy is lost through inelasticgranular collisions. This takes place during transport, andthe mode of macro-scale granular interactions distinguishesrock avalanches form other geostrophic flows.Melosh [1979]argues that rapid energy loss results in rapid jamming ofgrains, terminating runout. Following the experimentalresults, synthetic shear reduces the number of granularcollisions, thus reducing the initial lost energy during flow.This may enhance the likelihood of flow when other trigger-ing forces bring the granular mass to failure.[36] This is most important as the flow approaches steady

state conditions. Previous experimental results indicate thatat steady state, only a thin shear band of grains is requiredfor flow to occur, commonly no more than five graindiameters thickness [Losert et al., 2000] (Figure 9b). Sucha flow configuration would confine frictional or collisionaldissipation to a thin layer, thus potentially reducing overallenergy loss in the flow. This flow profile was predicted forrock avalanches by Friedmann [1997] and is consistent withcommon features of rock avalanche deposits such as jigsawbreccias, shear bands, inherited stratigraphy, and a basalmixing zone (Figure 12). It is also consistent with theSavage-Hutter granular flow model [Savage and Hutter,1989]. Granular flow initiated with synthetic shear defor-mation reaches steady state almost immediately. By com-parison, antithetic shear deformation produces a muchbroader zone of shear and compaction, with potentiallygreater frictional and collisional dissipation. It may alsotake much more time to evolve into a thin shear band, andmay not ever reach steady state in a gravity-driven flow.[37] If these observations are accurate and appropriate,

then the most-recent shear history becomes important torock avalanche triggering. Specifically, an event will initiatewith a much greater probability if and only if the sourceregion followed a deformation trajectory similar to the

landslide’s eventual course. In support of this hypothesis,many recent rock avalanche deposits crop out at the base ofnormal faults with a sense of shear sub-parallel to transport[e.g., Longwell, 1951; Burchfiel, 1966; Blair, 1999]. Thischaracterization can be used to better predict the futureoccurrence and distribution of rock avalanche hazards.Application of this precondition would significantly reducethe potential hazard areas, allowing policy makers and localworkers to focus prevention, mitigation, and evacuationstrategies in a small number of locations.

8. Conclusions

[38] The sense of shear in rock avalanche source regionscommonly matches the antecedent shear sense. This sug-gests a link and necessary precondition for rock avalanchetriggering.[39] Granular materials ‘‘remember’’ sense of shear due

to anisotropies in granular contact network. This ‘‘memory’’can help to explain the clustering and specific occurrence ofrock avalanches. This new precondition to rock avalancheinitiation may be used to create enhanced hazard assess-ments for rock avalanche potential.

[40] Acknowledgments. The authors would like to thank Jeff Harpfor field assistance, Peter Sadler for invaluable help and expertise in thefield, and E-an Zen. This manuscript was greatly improved through thereviews of D. K. Keefer, H. J. Melosh, and one anonymous reviewer.Special thanks go to the Univ. of Maryland Dept. of Physics and theCollege of Computer, Math, and Physical Sciences. This work was partiallyfunded by NASA grant NAG32736 (proposal 01-OBPR-02-075).

ReferencesAgnew, D., et al., Future seismic hazards in southern California; Phase I:Implications of the 1992 Landers earthquake sequence, 42 pp., Calif.Geol. Soc., Sacramento, 1992.

Allen, J. A., Report on the stability of Turtle Mountain, Alberta, Dep. ofGeol., Univ. of Alberta, Edmonton, Alberta, Canada, 1933.

Anderson, F. W., The Frank Slide Story, 60 pp., Frontiers Unlimited, Cal-gary, Alberta, Canada, 1968.

Bagnold, R. A., Experiments on gravity-free dispersions of large solidparticles in a Newtonian fluid under shear, Proc. R. Soc. London, Ser.A, 225, 49–63, 1954.

Bishop, K. M., Miocene rock-avalanche deposits, Halloran/Silurian Hillsarea, southeastern California, Environ. Eng. Geosci, 3, 501–512, 1997.

Blair, T. C., Form, facies, and depositional history of the North Long Johnrock avalanche, Owens Valley, California, Can. J. Earth Sci., 36, 855–870, 1999.

Burchfiel, B. C., Tin Mountain landslide, southeastern California, and theorigin of megabreccia, Geol. Soc. Am. Bull., 77, 95–100, 1966.

Campbell, C. S., P. Cleary, and M. Hopkins, Large-scale landslide simula-tions: Global deformation, velocities, and basal friction, J. Geophys. Res.,100, 8267–8283, 1995.

Clement, E., Rheology of granular media, Curr. Opinion Colloid InterfaceSci., 4, 294–299, 1999.

Cruden, D. M., and J. Krahn, Frank Rockslide, Alberta, Canada, in Rock-slides and Avalanches, part 1, Natural Phenomena, Dev. Geotech. Eng.,vol. 14a, edited by B. Voight, pp. 97–112, Elsevier Sci., New York, 1978.

Daerr, A., and S. Douady, Two types of avalanche behaviour in granularmedia, Nature, 399, 241–243, 1999a.

Daerr, A., and S. Douady, Sensitivity of granular surface flows on prepara-tion, Europhys. Lett., 47, 324–330, 1999b.

Daly, R. A., W. G. Miller, and G. S. Rice, Report of the commissionappointed to investigate Turtle Mountain, Frank, Alberta, Mem. Can.Geol. Surv., 27, 34 pp., 1912.

Davis, G. A., et al., The effect of pluton pinning on an active detachment:The Kingston Range/Halloran Hills detachment fault system, Geology,21, 627–630, 1993.

Eppler, D. B., J. Fink, and R. Fletcher, Rheological properties and kine-matics of emplacement of the Chaos Jumbles rock avalanche, LassenVolcanic National Park, California, J. Geophys. Res., 92, 3623–3633,1987.

Figure 12. Hypothetical flow profile derived from ob-servations of rock avalanche deposits (after Friedmann[1997]).


Erissman, T. H., The mechanics of large landslides, Rock Mech., 12, 5–46,1979.

Friedmann, S. J., Rock avalanches of the Miocene Shadow Valley basin,eastern Mojave Desert, California: Processes and Problems, J. Sediment.Res., Sect. A, 67, 792–804, 1997.

Friedmann, S. J., Sedimentology and stratigraphy of the Shadow Valleybasin, eastern Mojave, Desert, California, in Cenozoic Basins of theDeath Valley Region, edited by L. A. Wright and B. W. Troxel, Spec.Pap. Geol. Soc. Am., 333, 214–243, 1999.

Friedmann, S. J., and D. W. Burbank, Rift and supradetachment basins:Intracontinental extensional endmembers, Basin Res., 7, 109–127, 1995.

Friedmann, S. J., G. A. Davis, and T. K. Fowler, Geometry, paleodrainage,and geologic rates from the Shadow Valley supradetachment basin, east-ern Mojave, California, in Reconstructing the Structural History of Basinand Range Extension Using Sedimentology and Stratigraphy, edited byK. K. Beratan, Spec. Pap. Geol. Soc. Am., 303, 85–105, 1996.

Hadley, J. B., Madison County Rockslide, Montana U.S.A., in Rock-slides and Avalanches, part 1, Natural Phenomena, Dev. Geotech.Eng., vol. 14a, edited by B. Voight, pp. 87–128, Elsevier Sci., NewYork, 1978.

Heim, A., Bergsturz und Menschleben, 218 pp., Fretz and Wasmuth, Zurich,Switzerland, 1932.

Hodges, K. V., et al., Evolution of extensional basins and basin and rangetopography west of Death Valley, CA, Tectonics, 8, 453–467, 1989.

Howard, K., Avalanche mode of motion: Implications from lunar examples,Science, 180, 1052, 1973.

Hsu, K. J., Catastrophic debris streams (sturtzstroms) generated by rock-falls, Geol. Soc. Am. Bull., 86, 129–140, 1975.

Hutton, L. K., L. M. Jones, E. Hauksson, and D. D. Given, Seismotectonicsof southern California, in Neotectonics of North America, GSA DecadeMap, vol. 1, edited by D. B. Slemmons et al., pp. 133–152, Geol. Soc. ofAm., Boulder, Colo., 1991.

Ingersol, R. V., et al., The Mud Hills, Mojave, Desert: Structure, stratigra-phy, and sedimentology of a rapidly extended terrane, in Reconstructingthe Structural History of Basin and Range Extension Using Sedimentol-ogy and Stratigraphy, edited by K. K. Beratan, Spec. Pap. Geol. Soc.Am., 303, 61–84, 1996.

Jaeger, H. M., S. R. Naegel, and R. P. Behringer, Granular solids, liquids,and gases, Rev. Mod. Phys., 68, 1259–1273, 1996.

Jennings, C. W., (Compiler), Geological map of California, scale1:750,000, Calif. Geol. Surv., Sacramento, 1973.

Johnson, B., Blackhawk Landslide, California, in Rockslides and Ava-lanches, part 1, Natural Phenomena, Dev. Geotech. Eng., vol. 14a, editedby B. Voight, pp. 481–504, Elsevier Sci., New York, 1978.

Josserand, C., A. V. Tkachenko, D. M. Mueth, and H. M. Jaeger, Memoryeffects in granular materials, Phys. Rev. Lett., 85, 3632–3635, 2000.

Keefer, D. K., Landslides cause by earthquakes, Geol. Soc. Am. Bull., 95,406–421, 1984a.

Keefer, D. K., Rock avalanches caused by earthquakes: Source character-istics, Science, 223, 1288–1290, 1984b.

Keefer, D. K., The susceptibility of rock slopes to earthquake-induced fail-ure, Bull. Assoc. Eng. Geol., 30, 353–361, 1993.

Keefer, D. K., and R. C. Wilson, Predicting earthquake-induced landslides,with emphasis on arid and semi-arid environments, in Landslides in aSemi-arid Environment, edited by P. M. Sadler and D. M. Morton,pp. 118–149, Inland Geol. Soc., San Bernardino, Calif., 1989.

Longwell, C. R., Megabreccia developed down-slope from large faults, Am.J. Sci., 249, 343–355, 1951.

Losert, W., and G. Kwon, Transient and steady-state dynamics of granularshear flows, Adv. Complex Syst., 4, 1–9, 2001.

Losert, W., L. Bocquet, T. C. Lubensky, and J. P. Gollub, Particle dynamicsin sheared granular matter, Phys. Rev. Lett., 85, 1428–1431, 2000.

Luchitta, B. K., A large landslide on Mars, Geol. Soc. Am. Bull., 89, 1601–1609, 1979.

McConnell, R. G., and R. W. Brock, Report on the great landslide at Frank,Alberta, Canada, Annual Rep. 1902–1903, part 8, 17 pp., Can. Dep. Inter.,1904.

McEwan, A. S., Mobility of large rock avalanches: Evidence from theValles Marineris, Mars, Geology, 17, 1111–1114, 1989.

McSaveney, M. J., Sherman Glacier Rock Avalanche, Alaska, USA, inRockslides and Avalanches, part 1, Natural Phenomena, Dev. Geotech.Eng., vol. 14a, edited by B. Voight, pp. 197–258, Elsevier Sci., 1978.

Melosh, H. J., Acoustic fluidization: A new geological process?, J. Geo-phys. Res., 84, 7512–7513, 1979.

Melosh, H. J., The mechanics of large rock-avalanches, in Debris Flows/Avalanches: Processes, Recognition, and Mitigation, edited by J. E.Costa and G. F. Wieczorek, Rev. Eng. Geol., 7, 41–49, 1987.

Miller, J. K., and B. John, Sedimentation patterns support seismogenic low-angle normal faulting, southeastern California and western Arizona,Geol. Soc. Am. Bull., 111, 1350–1370, 1999.

Moore, J. G., et al., Prodigious submarine landslides on the HawaiianRidge, J. Geophys. Res., 94, 17,456–17,484, 1989.

Mudge, M. R., Rockfall-avalanche and rockslide-avalanche deposits atSawtooth Ridge, Montana, Geol. Soc. Am. Bull., 76, 1003–1014, 1965.

Mueth, D. M., et al., Signatures of granular microstructure in dense shearflows, Nature, 406, 385–389, 2000.

Norris, D. K., Blairmore, Alberta, Pap. Can. Geol. Surv., 55-18, 1955.Plafker, G., and G. E. Ericksen, Nevados Huascaran avalanches, Peru, inRockslides and Avalanches, part 1, Natural Phenomena, Dev. Geotech.Eng., vol. 14a, edited by B. Voight, pp. 197–258, Elsevier Sci., NewYork,1978.

Poeschel, T., C. Saluena, and T. Schwager, Scaling properties of granularmaterials, Phys. Rev. E, 64, 11,308–11,312, 2001.

Rajchenbach, J., Granular flows, Adv. Phys, 49, 229–256, 2000.Savage, S. B., and K. Hutter, The motion of a finite mass of granularmaterial down a rough incline, J. Fluid Mech., 199, 177–183, 1989.

Schmidt, K. M., and D. R. Montgomery, Limits to relief, Science, 270,617–620, 1995.

Shaller, P. J., Analysis and implications of large Martian and terrestrialslides, Ph.D. thesis, 451 pp., Cal. Inst. of Technol., Pasadena, 1991.

Shreve, R. L., The Blackhawk Landslide, Spec. Pap. Geol. Soc. Am., 108,51 pp., 1968.

Shreve, R. L., Blackhawk Landslide, southwest San Bernardino County,California, in Geological Society of America Centennial Field Guide-book—Cordilleran Section, pp. 109–114, Geol. Soc. of Am., Boulder,Colo., 1989.

Topping, D. J., Paleogeographic reconstruction of the Death Valley ex-tended region: Evidence from Miocene rock-avalanche deposits in theAmargosa Chaos basin, California, Geol. Soc. Am. Bull., 105, 1190–1213, 1993.

Yarnold, J. C., Rock-avalanche characteristics in dry climates and theeffect of flow into lakes: Insights from the mid-Tertiary sedimentarybreccias near Artillery Peak, Arizona, Geol. Soc. Am. Bull., 105, 345–360, 1993.

Yarnold, J. C., Tertiary sedimentary rocks associated with the Harcuvar corecomplex in Arizona (USA): Insights into paleogeographic evolution dur-ing displacement along a major detachment fault system, Sediment.Geol., 89, 43–63, 1994.

��S. J. Friedmann, Department of Geology, University of Maryland,

College Park, MD 20742, USA. ([email protected])G. Kwon and W. Losert, Department of Physics, IPST, and IREAP,

University of Maryland, College Park, MD 20742-3511, USA.


Figure 4. Simplified diagrams of basement structure in Blackhawk source region [after Shreve, 1968].Each cross section is roughly 1.5 km in length and runs N-S across northern San Bernardino Mountains.(a) Cross section 1.3 km east of Blackhawk Canyon axis, along Miles Canyon. Note the dip of Voorhiesand Santa Fe thrusts out of the mountain due to subjacent folding. (b) Cross section along east edge ofBlackhawk Canyon. Note the southward tip of the Voorhies thrust and its imbricates shallow toward thelandslide scarp, suggesting flattening and northward dip within the canyon, synthetic and subparallel tothe landslide paleoflow vector.

FRIEDMANN ET AL.: GRANULAR PHYSICS AND ROCK AVALANCHE TRIGGERING

ECV 8 - 4

Documents

Granular memory and its effect on the triggering and ...raman/papers2/FriedmannJGR.pdfGranular memory and its effect on the triggering and distribution of rock avalanche events S