Structural Analysis and Analogue Modeling of the Kinematics and Dynamics of Rockslide Avalanches

Home , Aucanquilcha, Breccia, Chimborazo, Lastarria, Llullaillaco, Parinacota (volcano)

Structural analysis and analogue modeling of the kinematics and dynamics of rockslide avalanches

Thomas Shea* Benjamin van Wyk de Vries* Laboratoire Magmas et Volcans, Université Blaise Pascal, Clermont-Ferrand, 63000, France

ABSTRACT ics: hummocky, nonhummocky, dominantly dynamics, because the link between deposit and extensional, and dominantly compressional emplacement is unclear. We present a structural analysis of subaer- rockslide avalanches. The models require ial natural and analogue rockslide avalanches. a brittle core and surface that spreads and Faults in Rockslide Avalanches Such deposits often have well-developed contracts by adjustment on large numbers of faults, folds, and hummocks. These structures faults that bottom into a low-friction décol- A striking feature of well-preserved rockslide- can be used to determine the kinematics and lement layer. Spreading is accommodated by avalanche deposit surfaces is their morphology, dynamics of emplacement. Large-scale terres- normal and strike-slip faults, while on decel- which, through ridges and escarpments, shows trial rockslide avalanches show large runout eration, thrust faulting generates thickening. the presence of a complex succession of faults distances compared to their fall height. Most To be realistic, any physical predictive model and folds. One superb example of this is the attempts to explain this phenomenon invoke must take into account these fundamental Socompa rockslide avalanche (Central Andes, fl uidizing mechanisms or lubricating agents kinematic and structural aspects. Chile and Argentina; Fig. 1), where the entire to reduce forces opposed to momentum, surface displays a dense network of normal, especially at the base. However, the proper- Keywords: rockslide-avalanche deposits, struc- strike-slip, and thrust faults (Wadge et al., 1995; ties and mechanics of low friction are still tures, faults, analogue modeling, avalanche van Wyk de Vries et al., 2001, 2002; Kelfoun and poorly understood. Any model for motion transport, kinematics, hummocks. Druitt, 2005; Kelfoun et al., 2008). The interior and emplacement must integrate geometric, of this avalanche deposit, as seen in road cuts, morphologic, and structural features, all cru- INTRODUCTION shows that the fault structures incise deeply into cial in constraining kinematics and dynam- the deposit interior (e.g., Fig. 3 in van Wyk de ics. Here we fi rst examine the morphological Rockslide Avalanches Vries et al., 2001; Figs. 1B, 1C herein). Other and structural features displayed by 13 natu- examples include the avalanches of Mombacho ral rockslide avalanche deposits; we then use Large-scale rockslide avalanches are part of Volcano (Nicaragua; Shea et al., 2007), Pari- simple and well-constrained analogue mod- the Earth mass-movement processes. Subaerial nacota (Clavero et al., 2002), and Flims (Pol- els involving the slide of stratifi ed granular examples on Earth involve volumes of as much let and Schneider, 2004). In each case, road material down smooth, curved ramps. These as 100 km3, and their deposits cover areas as cuts show either original material or granular differ from previous analogue models in that large as thousands of square kilometers, reach- rockslide-avalanche breccia layers being offset we concentrate on observing the structures ing distances >100 km (Stoopes and Sheri- by numerous faults. Displacement is localized produced by brittle deformation and use a dan, 1992). They cause considerable human on narrow shear zones in the breccia, where the low-friction sliding surface. Models show that and material loss, directly or through second- simple shear of the fault zone is accommodated. variations in the sliding surface curvature, ary events such as tsunamis, river obstruction, Such a structural pattern is common in faults in lateral profi le, roughness, and modifi cations lahars, or magmatic eruptions (Siebert, 1984; granular materials, seen in sandbox analogue in material cohesion can successfully repro- Siebert et al., 1987). Although the onset of models of volcano deformation (e.g., Merle duce the majority of rockslide-avalanche fl ank collapse was photographed at Mount St. and Borgia, 1996) or in natural breccia or pyro- deposit features. After discussing the geo- Helens in the 1980 eruption (Voight, 1981), clastic successions deformed by slow tecton- metrical and dynamic similarity between rockslide-avalanche formation and motion have ics (e.g., Borgia and van Wyk de Vries, 2003). experiments and natural examples, we pro- never been clearly observed; as a result, study is These observations allow us to characterize the pose a model for structure formation and a mostly done through their deposits, or by theo- structures observed here as faults, as this is the fourfold classifi cation based on model and retical modeling. Important gaps thus arise in closest and most suitable broad defi nition avail- natural deposit morphology and dynam- understanding the emplacement kinematics and able. The data indicate that large-scale brittle

*Shea, corresponding author: present address: Geology and Geophysics Department, SOEST, University of Hawaii at Manoa, Honolulu, USA; [email protected]. van Wyk de Vries: [email protected].

Geosphere; August 2008; v. 4; no. 4; p. 657–686; doi: 10.1130/GES00131.1; 15 ﬁ gures; 2 tables; 2 supplemental tables; 1 supplemental ﬁ le; 2 animations.

Downloaded from http://pubs.geoscienceworld.org/gsa/geosphere/article-pdf/4/4/657/3336451/i1553-040X-4-4-657.pdf by guest on 30 September 2021 Shea and van Wyk de Vries Bolivia large normal fault with back-facing normal A major strike-slip component fault at terminus 20° 3250 m isolated back thrust area, 3000 m formed like median scar before distal lobe active distal lobe has many bookshelf Socompa faults and a strike-slip boundary 30° 3100 m strike-slip and graben unit

structures join distal

n i structures here t

g r 3455 m eastward spreading

e indicated by grabens

C 3100 m northeast unit lineations continue 50° through median scarp 80° abrupt change in direction of NE unit large -throw normal faults 3600 m well-preserved early thrusts, cut by conjugate 3100 m strike-slips 3500 m spreading debris units longitudinal thrusts area of backflow sliding off mountain range 200 m -high debris units continue through fault scarp median scarp but are deflected buried scar Median scarp starts as trace en-echelon folds, thrusts 3580 m and strike-slips

longitudinal grooves 3450 m large-throw (<300 m) Toreva faults levée and marginal strike-slips scar infilled with later W / N distal lavas and pyroclastics margin lobe strike-slip/ areas of dense graben normal faulting 6061 m 3600 m east Key Normal faults Strike-slip faults Major thrust faults al im ox Folds Grooves/striae SB RIF pr N Toreva Levees scar Lithological repetitions Later cover 5 km

B F C SBa F F

SBb

repetition of vertical stratigraphy F F F F Transport direction RIF

Figure 1 (continued on next page). (A) Volcán Socompa (Central Andes, Chile and Argentina) rockslide-avalanche deposit structural map with main features (Guilbaud, 2000; van Wyk de Vries et al., 2002). The map is made from detailed aerial photographs, a digital elevation model (DEM), and ﬁ eld observations. Only the main structures are shown. (B) Outcrop photograph of normal faults developed in breccia at Socompa (from van Wyk de Vries et al., 2001). These are proximal normal faults that are deformed by latter strike-slip faults as shown in Figure 2A. (C) Vertical lithological repetitions (man for scale) at a marginal site shown in vertical cut, normal to transport direction of the dense banding seen in F. The repetition is similar to that seen in the analogue models. The white (basal) layers are Reconstituted ignimbrite facies (RIF; dominantly ﬁ ne materials from the remobilized ignim- britic substratum), and the gray (upper) is the Socompa Breccia (SB; blocks + matrix) (van Wyk de Vries et al., 2001).

658 Geosphere, August 2008

Downloaded from http://pubs.geoscienceworld.org/gsa/geosphere/article-pdf/4/4/657/3336451/i1553-040X-4-4-657.pdf by guest on 30 September 2021 Structural analysis and analogue modeling of rockslide avalanches D RIF 1 km N

scar s

s T

raft

ms 1 km

RIF N

1 km

Key normal fault thrust fault strike-slip fault DAD unit boundary (in E) thrust bands (in F)

Figure 1 (continued). Three detailed structural maps of the Socompa deposit, made from aerial photographs after fi eld inspection (van Wyk de Vries et al., 2001, 2002). Structures are shown at a greater resolution. Note that there are numerous crosscutting relationships between faults that indicate multiple generations of fault structures. Some crosscutting relationships have displacements too small to show on the map, others have displacements of several kilometers. (D) Proximal zone including the Toreva block margin (T) and edge of collapse scar (scar). Here Toreva blocks have 10–100-m-spaced large normal faults that become denser in the direction of transport. (The Torevas transform over a short zone of ~500 m to a fully brecciated rockslide avalanche.) Normal faults (illustrated in B) are displaced and deformed by later strike-slip faults and broad zones of shear that at small scale, are made up of many smaller discrete fault planes. S indicates small salar deposits on the deposit surface. (E) Central zone of deposit cut by the major Median Scarp (ms) thrust belt (Guil- baud, 2000; Kelfoun et al., 2008). Note that the proximal structures to the southeast are cut and the deposit banding is defl ected across the median scarp. Note also that islands or rafts of normally faulted blocks are preserved in the predominantly strike-slip deformed zone to the northwest of the median scarp (raft). (F) Distal area with fi ne-scale banding, cut by thrusts and strike-slip faults. Note that the outer margin of the deposit is composed of the RIF basal facies. The banding is interpreted to originate from the multiple thrusting of lower RIF facies and upper SB facies (van Wyk de Vries et al., 2001, 2002; Kelfoun et al., 2008). Note the large displacement strike-slip faults to the northeast, and the dense late-formed strike-slip faults: some of these clearly displace the earlier thrust faults, while on some lineaments no offset can be seen. The thrusts may have originated initially as normal faults that were subsequently rotated and inverted into thrusts during the deceleration phase. In this area the slide was riding up a topographic slope and there is also major late-stage sliding in the reverse direction (Kelfoun et al., 2008), with a second set of structures produced. DAD—Debris Avalanche Deposits.

Geosphere, August 2008 659

Downloaded from http://pubs.geoscienceworld.org/gsa/geosphere/article-pdf/4/4/657/3336451/i1553-040X-4-4-657.pdf by guest on 30 September 2021 Shea and van Wyk de Vries

deformation occurs in the main mass during ridges, and troughs cover their surfaces, and even though they are crucial in defi ning basic rockslide- avalanche emplacement. their respective outlines differ signifi cantly. constraints. For example, if the original pre- Any kinematic or dynamic model for rockslide- collapse gross stratigraphy is preserved within Rockslide-Avalanche Kinematics avalanche emplacement should account for all a deposit, turbulent transport models involving these morphological and structural features, as signifi cant large-scale mixing are not feasible. Socompa and other deposits like Blackhawk indicated in Pudasaini and Hutter (2007), or Similarly, if surface faults form throughout (Fig. 2A), Chaos Jumbles (Fig. 2B), or Mom- Cruden and Varnes (1996). However, few stud- emplacement, a brittle-type behavior must be bacho (Fig. 2C) are morphologically and geo- ies (e.g., Belousov et al., 1999; Clavero et al., accounted for in physical models. In turn, the metrically very different from each other. Vari- 2002; van Wyk de Vries et al., 2002; Kelfoun depth reached by these faults may provide ous combinations of undulations, hummocks, and Druitt, 2005) have focused on these aspects, information on the transition between a brittle

125° 115° 105° A Confining topography Slump Idaho block 1327 m California 2035 m

35° Blackhawk 939 m Blackhawk canyon Collapse source Mexico 115° 105° 1428 m

1010 m 1943 m

Strike-slip faults Thrust faults/anticlines N Normal faults Lateral levees 1 km Key

B 125° 115° 105°

y Idaho graph po Calif Chaos g to inin Jumbles Lassen peak f o Highway n 1807 m co rnia 35°

e l c 1887 m a t s b Mexico o 115° 105° c i h p a 1953 m r g o p o T 1841 m 1 2074 m Collapse source ing topogr confin aphy 2450 m

Strike-slip faults Thrust faults/anticlines N Normal faults Striations 2042 m 500 m 1 Crags Lake Key

Figure 2 (continued on next page). Structural maps. (A) Non-volcanic rockslide avalanche, Blackhawk (California, USA). (B) Volcanic rockslide avalanches: Chaos Jumbles (California, USA).

660 Geosphere, August 2008

20°N El Diamante

duras Hon 15°

Nicaragua 82 m Puerto C. Rica Asese Mombacho

34 m

Aguas Strike-slip faults calientes Thrust faults Normal faults Hummocks N 2km Key 451 m Lake Nicaragua Northern avalanche 344 m scar

1161 m 100 m

Southern Volcan Mombacho avalanche summit at 1345 m scar 812 m El Crater

543 m

Santa 98 m Agua Rosa agria

Las Lagunas scarp (anterior regional fault)

88 m

34 m

64 m

Figure 2 (continued). (C) Mombacho North and South (Nicaragua). Note the differences in surface structure organization where Blackhawk and Chaos Jumbles have ridged and fault-rich surfaces, contrary to those from Mombacho North and South, both ridge poor and hummock rich.

Geosphere, August 2008 661

Downloaded from http://pubs.geoscienceworld.org/gsa/geosphere/article-pdf/4/4/657/3336451/i1553-040X-4-4-657.pdf by guest on 30 September 2021 Shea and van Wyk de Vries

crust and a basal liquefi ed and/or fl uidized low- ROCKSLIDE AVALANCHES: the block facies forms the upper section of the friction slide layer. PROBLEMS deposit and covers the mixed facies (Crandell, 1989; Shaller, 1991; Belousov et al., 1999; Remote Sensing Analysis and Modeling On Earth, large mass movements occur in Clavero et al., 2002, 2005; Reubi and Hernán- four major environments: on submarine vol- dez, 2000; Shea et al., 2007); consequently, a Through detailed remote sensing analysis canoes, on continental slopes, on continental gross inverse grading may exist, which opposes and analogue modeling of rockslide-avalanche volcanoes, and on mountain chains. This study the use of the term “unsorted” in the defi nition deposit geometry, internal structures, and sur- focuses on subaerial collapses, although sub- of Crandell (1989). In contrast to debris fl ows face morphology, our approach constrains marine ones may behave in a similar fashion, as or mudfl ows, rockslide avalanches comprise a rockslide-avalanche kinematics and allows us the deposit morphology is similar (e.g., Oehler signifi cantly lower proportion of fl uids, and are to make predictions about their emplacement et al., 2004). The basic phenomenon involves therefore said to be unsaturated. dynamics. Remote sensing is generally used failure and collapse of a rock mass that acceler- because deposits are large and often inacces- ates until it reaches a slope suffi ciently low or Rockslide-Avalanche Mobility sible. However, the surface morphology records long to be brought to a stop. The term “debris clearly the structures that are developed within avalanche” has been used rather loosely, Studies on rockslide-avalanche dynamics the deposit, allowing structural maps to be hence the need to defi ne it more precisely. We have generally focused on their long runout. In created. Field verifi cation has been made on chose the term rockslide avalanche, as it best general, runout is related to a friction coeffi cient, Socompa and Mombacho deposits (van Wyk de describes the brittle and fl uidized nature of the defi ned as the tangent to the slope connecting Vries et al., 2001; Shea et al., 2007), and pro- phenomena. To a fi rst approximation we use the pre-collapse and post-collapse centers of vides the ground truth for the structures. Other the defi nition of Crandell (1989, p. 1), who gravity of the rock mass (Fig. 3A). Preexisting deposits also have fi eld descriptions of faulting, described rockslide avalanches as “rapidly topography being diffi cult to evaluate, Heim such as Mount St. Helens (Glicken, 1998), Pari- moving unsorted mixtures of blocks and matrix, (1932) proposed to approximate this coeffi cient nacota and Ollagüe (Clavero et al., 2002, 2005), mobilized by gravity consequently to [volcano] to the tangent of the slope joining the scar high and those described by Shaller (1991). fl ank destabilizations.” This defi nition is gen- point to the furthest point reached by the rock- This paper also presents a series of ana- eral enough to include non-volcanic rockslide slide avalanche: he named this “fahrböschung,” logue experiments that reproduces the slide of avalanches. When the rock mass collapses, it generally referred to as “apparent friction coef- a stratifi ed mass of sand with variable cohesion transforms into a moving rockslide avalanche fi cient,” and calculated dividing fall height H down a simple curved ramp. Resulting experi- due to fragmentation processes or an initially by the horizontal runout L. Scheidegger (1973) ments were photographed, fi lmed, studied in fragmented state (Voight et al., 1983; Glicken noticed that a direct relationship exists between cross section, and characterized both quantita- 1991), and travels over the landscape at speeds non-volcanic rockslide-avalanche volumes V tively and qualitatively. The models are scaled up to 300 km/h, while thinning and spreading. and their apparent friction coeffi cients: H/L and we compare results to natural examples. When the mass loses suffi cient kinetic energy, decreases systematically with volume increase. The models deform in the frictional granu- it comes to rest, depositing material tens of Ui (1983) reached similar conclusions regard- lar regime, the deformation is ruled by brittle meters in thickness over distances often greater ing volcanic mass movements. With the avail- mechanics, and faulting is generated as in the than 10 km. Rockslide-avalanche deposits fre- able data, he further suggested that volcanic natural examples. We show that this approach quently expose a bimodal lithology, described H/L ratios were smaller than non-volcanic can be used effectively in the interpretation of by Ui (1983) and Glicken (1991) as block facies ratios at similar volumes. McEwen (1989) com- structures in the fi eld. and mixed facies (matrix and blocks). Often, pared terrestrial rockslide-avalanche volumes

A B Extraterrestrial (35) Non-volcanic (115) Volcanic (47) L Lg 0.7 T-NV 0,6

ET 0,4 T-V

farböschung H H/L Hg 0 real 0.001 10000 0,2

0 Post-collapse 0,001 0,01 0,1 1 10 100 1000 10000 gravity center Volume (km³) n=197

Figure 3. (A) Comparison of apparent friction coefﬁ cient (farböschung) with real friction coefﬁ cient. H and L measure height and

length using the uppermost point of the collapse scar and the distant tip of the deposit, whereas H g and L g relate to the original and ﬁ nal gravity centers of the involved rock mass. (B) Plot of H/L versus volume for the gathered rockslide-avalanche data from different environments. Volume is in log scale and n is the number of data points. Subplot displays ﬁ tted trends for all three environments, where ET = extraterrestrial, T-NV = terrestrial non-volcanic, and T-V = terrestrial volcanic.

662 Geosphere, August 2008

and H/L ratios with their Martian equivalents, than this, no signifi cant difference in terrestrial emplacement was associated with intense erup- and discovered that on average, Martian H/L volcanic and non-volcanic behavior can be dis- tive activity, which can potentially mask impor- ratios are lower than terrestrial H/L ratios for tinguished, which slightly diverges from the tant structures. In the well-documented case of similar volumes. He suggested that such varia- conclusions of Ui (1983), but is in agreement Mount St. Helens, the very complex topographic tions arise from gravity differences, as Martian with those of Shaller (1991). The conclusion is setting, the reworking of materials by erosion, events need a much thicker fl ow to reach identi- important because it means that volcanic and and the pyroclastic and lahar cover prevent clear cal runout. non-volcanic events can essentially be treated as surface structure observation. In order to verify the validity of such obser- one phenomenon. vations, we gathered an up-to-date database Plan-View Shape (sources: Howard, 1973; Lucchitta, 1979; GEOMETRICAL, STRUCTURAL, Francis et al., 1985; Siebert et al., 1987; McE- AND MORPHOLOGICAL Deposits created by large-scale collapses wen, 1989; Shaller, 1991; Schenk and Bulmer, CHARACTERISTICS OF ROCKSLIDE- may partly or fully adopt four distinct shapes: 1998; Clavero et al., 2002; Shea et al., 2007) AVALANCHE DEPOSITS (1) fan shaped (e.g., Ollagüe, Fig. 4E; Tetivi- and compiled measurements for 35 extrater- cha, Fig. 5A; Llullaillaco “northern subunit,” restrial (Moon and Mars), 47 terrestrial volca- We characterized a total of 13 terrestrial and Fig. 5B) with linear proximal and/or distal mar- nic, and 115 terrestrial non-volcanic rockslide 1 extraterrestrial rockslide avalanches in terms gins and variably lobate distal margins; (2) single avalanches. Figure 3B plots H/L versus V data of structural, geometrical, and morphological lobes (e.g., Carlson, Fig. 4A; Martinez Moun- for each separate environment and gathers them features (Table 1; Figs. 1, 2, 4, and 5; see also tain, Fig. 4B; Blackhawk, Fig. 2A; Mombacho into a single graph for comparison. Regression Supplemental File1) prior to the experimental North and South, Fig. 2C; Socompa, Fig. 1; lines (H/L ≈ 0.25V-0.13 for extraterrestrial and work. Their structural maps were built using a Olympus Mons, Fig. 5D) with well-curved lat- for H/L ≈ 0.14V-0.16 terrestrial) have too much combination of available imagery, digital eleva- eral and distal margins; (3) tongue shaped (e.g., scatter for further quantitative use (0.36 < R2 < tion models, available literature (Shreve, 1968; Aucanquilcha, Fig. 4C; Llullaillaco, Fig. 5B; 0.63), but nonetheless show general trends. Note Eppler et al., 1987; Francis and Wells, 1988; Socompa “East subunit,” Fig. 1) with parallel that in such a large database, some of this scat- Shaller, 1991; Wadge et al., 1995; Guilbaud, lateral margins and a curved distal margin; and ter may be attributed to differences in volume 2000; van Wyk de Vries et al., 2001, 2002; (4) irregular: i.e., heterogeneous depositional and runout calculation, as well as measurement Clavero et al., 2002, 2005; Shea et al., 2007), topography such as valleys or topographic highs techniques. As suggested by Legros (2002), we and fi eld observations. The general guidelines and/or barriers often prevent rockslide-avalanche also plotted L versus V (i.e., disregarding H) we used to choose these deposits were a func- deposits from adopting fan-shaped, single lobe, and, while R2 values obtained are somewhat bet- tion of the following: or tongue-shaped outlines (e.g., Chaos Jumbles, ter (less scatter), the relative positions of trends 1. The presence of structures on their surface Fig. 2B; Parinacota, Fig. 5C). were similar to when H was included. (no signifi cant erosion or reworking); Overall, terrestrial and extraterrestrial trends 2. The quality and/or resolution of photographic Profi les show similar general behavior. However, extra- and/or remote sensing data available; terrestrial data are systematically shifted toward 3. Their runout (generally H/L ≤ 0.3); In cross section the profi les may appear as smaller H/L ratios at similar volumes. This con- 4. The environment (at least one for each terres- dominantly: (1) uniform (e.g., Aucanquilcha, fi rms previous observations made by McEwen trial volcanic, terrestrial non-volcanic, and Chaos Jumbles, Lastarria, Mombacho South, (1989). In all cases, the data also confi rm the extraterrestrial); Llullaillaco, Socompa East, and perhaps Olym- volume effect described by Scheidegger (1973) 5. The range of volumes they cover (0.026– pus Mons) with constant thickness through- and Ui (1983), although above certain volumes 40 km3 for terrestrial deposits and one out the deposit; (2) distally raised (e.g., Carl- (~1 km3 for terrestrial and ~10 km3 for extrater- extraterrestrial >500,000 km3); son, Blackhawk, Mombacho North, Ollagüe, restrial), H/L decreases more smoothly. 6. The absence of associated volcanic eruption; Socompa West), where the frontal margin is Contrary to hypotheses made by Ui (1983), 7. The state of preservation of the deposit (e.g., thicker than the proximal; (3) proximally raised volcanic H/L ratios are not signifi cantly lower no signifi cant erosion, reworking of mate- (e.g., Parinacota) with thickness decreasing with than non-volcanic H/L at a given volume. How- rial, or eruptive material cover); distance; (4) irregular (e.g., Tetivicha, Martínez ever, comparing the two environments is not 8. Their diverse but relatively well constrained Mountain), where thickness varies signifi cantly, entirely conclusive because 73% of the non- and fairly simple topography-deposit rela- unrelated to distance from the source. volcanic data we currently have is restricted to tionship; volumes <0.1 km3, whereas 95% of the volca- 9. Their profi les (at least one of each uniform, Constituent Textures nic data available are >0.1 km3. Possibly, the distally raised, proximally raised); lack of data for volcanic mass movements of V 10. Both hummock-rich and hummock-free The brecciated state of the materials enclosed < 0.1 km3 relates to their frequent transforma- cases need to be represented; and in the deposit is probably the most common tion into lahars. In conclusion, the comparison 11. The quantity of data available from previous internal characteristic of a deposit. In addition, between terrestrial volcanic and non-volcanic studies. the deposits always show a cohesive travel data is only able to demonstrate that volcanic For example, Mount St. Helens, Shiveluch, mode; for example, roads or houses located a rockslide avalanches reach higher volumes Ontake-San, and Bezymianny rockslide ava- few meters from the margins (the latter may more frequently than non-volcanic events. Other lanches were discarded, given that their measure several hundreds of meters in height)

1If you are viewing the PDF of this paper or reading it offl ine, please visit http://dx.doi.org/10.1130/GES00131.SF1 (Supplemental File 1) or the full-text article on www.gsajournals.org to view Supplemental File 1, a Google Earth (.kml) placemark fi le containing the locations of the investigated rockslide avalanches. To view the fi le, you will need Google Earth, which can be downloaded at http://earth.google.com.

Geosphere, August 2008 663

Downloaded from http://pubs.geoscienceworld.org/gsa/geosphere/article-pdf/4/4/657/3336451/i1553-040X-4-4-657.pdf by guest on 30 September 2021

Shea and van Wyk de Vries Lateral levees levees Lateral

arched ridges ridges arched

Traverse Traverse

ridges ridges

Transverse Transverse

ridges ridges

Longitudinal Longitudinal

Torevas Torevas

Hummocks Hummocks

faults faults

Strike-slip Strike-slip

Normal faults faults Normal Thrusts Thrusts CA—curved regions; frontal toward le; CF—curved

raised; PR—proximally raised; U—uniform; Unconf.— raised; U—uniform; raised; PR—proximally

repetitions repetitions

Lithological Lithological

striae striae

amphitheater amphitheater

Collapse Collapse

orientation orientation

Clast Clast

textures textures

Hackly Hackly Jigsaw breccia breccia Jigsaw

ATION AND STRUCTURAL CHARACTERISTICS incorporation incorporation

Substratum Substratum

grading grading

Gross inverse inverse Gross

stratigraphy stratigraphy Preserved Preserved O CF O O – x O O – – – – O – O O O O O x x O – – O O O x x x x x x O O x O O x O O – –

D O O O O x x x O O O O x O O O CF O CF O D O x O x O O O x x O O C

Confinement Confinement South, partly K partly at North at North Profile Profile West West East, DR at

TABLE 1. ROCKSLIDE-AVALANCHE INFORM ROCKSLIDE-AVALANCHE TABLE 1.

Shape Shape CF O O – x O O – – – – – – – Unconf.? U? L L U, DR Uncof. at Uncof. DR L U,

) ) (km 3

00 Volume Volume 1.88

(South) (South) (North); (North);

Figure Figure at U L,T 1 17

5B 1-2 F,T U K,2 – O O – – – – x O O O O O CF 2B 0.026 2B 0.026 I U K,D M – x – – O x O O x O O x O O O M 4A 0.1 L DR C M O O – – O x M O O O x O O O CF O CF O O O x 4A 0.1 O L DR O C – O O 4B 0.24 O x O L M M I O O O x D x x O O M x – – O x O x M 4E 0.6 (?) (?) F DR Unconf. O 4E 0.6 O O O – x – x O x x O O O O x CF 2A 0.3 L DR K M – O – – x x M O O O x O O x O x O O x O 2A 0.3 O DR L O – K – x x O M – M 5C x O x 6 O O I PR O O K,D O x O – x O x – O – 4D 0.091 ± F U K – – – – – – O – O O O x x O O CA O 4C 1(?) T CA U Partially O O x U F x O K O – – ± – – O O 0.091 – 4D 2C 1.2 5D >500,0 5 A – F I K – – O – – – O – O – O O – O O x x x O O – O O – A – O F I – O – – K O – 5 –

(Country) (Country) Name Name

(USA) (USA) (Chile and (Chile and Mountain Mountain (USA) (Chile and (Chile and (Chile and (Chile and Bolivia) Argentina) Argentina) Argentina) Argentina) (Bolivia) (Bolivia) (USA) (Nicaragua) (USA) (Chile and Mons (Mars) Mons (Mars) (Chile and (Chile and Bolivia) Jumbles Jumbles Bolivia) Bolivia)

T—terrestrial; TV—terrestrial volcanic; E—extraterrestrial; L—lobate; F—fan shaped; T—tongue shaped; I—irregular; DR—distally DR—distally I—irregular; shaped; T—tongue shaped; F—fan L—lobate; E—extraterrestrial; volcanic; TV—terrestrial T—terrestrial; Environment Environment Note: T Carlson T Martinez T Blackhawk T Blackhawk toward amphitheater. amphitheater. toward TV Llullaillaco TV Llullaillaco E Olympus E Olympus TV Socompa TV Socompa unavailab dash (–)—data observed; x—not O—yes; M—monolithologic; 2—divided; D—deflected; C—channelized; K—confined; Unconfined; TV Parinacota TV Parinacota TV Chaos TV Lastarria TV Aucanquilcha TV Aucanquilcha

TV Tetivicha TV Tetivicha TV Mombacho TV Mombacho TV Ollagüe TV Ollagüe

664 Geosphere, August 2008

California 2491 m

2513 m 35° Canalizing valley wall 2220 m

Mexico Slump blocks 115° 105° 1 2052 m 2304 m Strike-slip faults 2612 m Canalizing valley wall

Thrust faults Slump blocks Normal faults 2495 m 2186 m Lateral levees

2842 m 1 Carlson Lake N

500 m Nf-118 Road Key

1st deflecting wall B Collapse source topograp 408 m ing hy fin on C

528 m 287 m 55 m

125° 115° 105°

Idaho 103 m California 447 m phy gra po 251 m to 2nd d ng 35° Martinez Mt. eflecting wall Confini

Strike-slip faults Normal faults N 500 m Mexico 115° 105° Thrust faults Lateral levees Key

Bolivia 4081 m C

20° Confining topography Aucan- 4496 m Chile -quilcha 4043 m Aucanquilcha 30° summit and collapse source

4826 m Argentina 3983 m

4478 m 50° Confining topography 80° 4029 m

Strike-slip faults Normal faults N Secondary or Lithological Thrust faults posterior debris repetitions avalanche 1 km Lateral levees Key

Figure 4 (continued on next page). Structural maps. (A) Non-volcanic rockslide avalanche, Carlson (Idaho, USA). (B) Non-volcanic rockslide avalanche, Martinez Mountain (California, USA). (C) Volcanic rockslide avalanche, Aucan- quilcha (Central Andes, Chile).

Geosphere, August 2008 665

Downloaded from http://pubs.geoscienceworld.org/gsa/geosphere/article-pdf/4/4/657/3336451/i1553-040X-4-4-657.pdf by guest on 30 September 2021 Shea and van Wyk de Vries Bolivia D 4892 m

To 20° pographic o 4488 m bstacle

Chile 4837 m 5661 m Lastarria 30°

Collapse source

Argentina Lastarria scar

Strike-slip faults 4458 m 50° Thrust faults N Normal faults Ancient 80° 4746 m scoria cone 4539 m Lateral levees 1 km Key Failure plane striae

Ollagüe (town) E Strike-slip faults 3701 m Thrust faults

Normal faults N Hummocks

Post-avalanche lava flows Key

ll 1 km a

Buenaventura

c i

h 3739 m 3806 m p

o 4108 m p

T ad

Collapse source

Bolivia Volcan Ollagüe

üe main ro main üe Ollag 20° 3843 m Ollagüe

Chile

30° La Poruñita 3700 m cone

4368 m

Argentina

50° Salar de Carcote 3877 m 80° Confining topography

Figure 4 (continued). (D) Volcanic rockslide avalanche, Lastarria (Central Andes, Chile and Argentina). (E) Volcanic rockslide avalanche, Ollagüe (Central Andes, Bolivia and Chile).

666 Geosphere, August 2008

are frequently unaffected by their emplacement Structures are common in unconsolidated and low resis- (Shaller, 1991). Heim (1932) stated that this tance rocks, for example tephra, and are also cohesive behavior arises from the presence of Faults are often seen at outcrop in rockslide formed in analogue models of volcano deforma- fi ne interstitial powders between larger particles avalanches, and may cut massive blocks, fraction using sand or other granular material (e.g., in the mass. tured blocks (block facies), and matrix breccias. Merle and Borgia, 1996; Donnadieu and Merle, Frequently, the deposits preserve original Two examples from Socompa at outcrop are 1998; Merle et al., 2001). large-scale stratifi cation. Heim (1932) was the shown in Figures 1B and 1C. Many other exam- On the surface of deposits, long ridges, fi rst to make this observation at Elm, Switzer- ples have been given, including Glicken, (1998), escarpments, and troughs are the morphologi- land. This represents strong evidence against Siebe et al. (1992), Clavero et al. (2002), Siebert cal expression of such faults (e.g., van Wyk de intense large-scale mixing or turbulent transport. et al. (2006), Shea et al. (2007), and Bernard et Vries et al., 2001). Normal faults are seen on Many deposits also show a gross inverse grad- al. (2008). The faults record brittle deformation the surface as sharp, generally straight escarp- ing that is explained either by stronger shearing of incorporated parts of the intact edifi ce, as ments, often with interior layers exposed, or and fragmentation occurring at the base, or clast well as brittle deformation of the newly formed exhumed, at the scarp base. Thrust faults form segregation in a vibrating moving mass (Savage breccia. In nearly intact, resistant material, the clear thrust anticlines, and are rounded, elongate and Lun, 1988; Gray and Thornton, 2005; Shea faults can form discrete planes, with fault gauge ridges, often sinuous along strike. Strike-slip et al., 2007). A certain fraction of substratum and slickensides. In less resistant and brecci- faults are lineations that have little topographic and/or bedrock can be torn, dragged, and incor- ated material, they generally form broader shear relief, but are often associated with pull-aparts, porated by the mass during travel. zones (Figs. 1B, 1D). Such broader shear zones pop-ups, and splays. Normal and thrust faults

Bolivia A 3673 m Cerro Guachacollo 20° Salar Coipasa 4950 m

Topographic obstacle Chile Tetivicha

30° 3766 m

Argentina 3750 m

50° Cerro 80° Tetivicha Collapse scar 5020 m

Strike-slip faults Thrust faults Normal faults N 3716 m Hummocks 3655 m 5 km Key Salar Uyuni

Figure 5 (continued on next page). Structural maps of volcanic rockslide avalanches. (A) Tetivicha (Central Andes, Bolivia).

Geosphere, August 2008 667

Downloaded from http://pubs.geoscienceworld.org/gsa/geosphere/article-pdf/4/4/657/3336451/i1553-040X-4-4-657.pdf by guest on 30 September 2021 Shea and van Wyk de Vries

B Strike-slip faults 4384 m C on fin Thrust faults in g top ography Lateral levees Volcan Llullaillaco collapse source Lithological repetitions

Hummocks ct bje o g n 3827 m ti 1 km c N le f Key e Cerro Rosado D Bolivia 5280 m 3755 m 20°

Chile Llullai- Salar 30° -llaco Llullaillaco 4512 m

4310 m Bulldozed Argentina C sediments on fin in g 3755 m

South-eastern 50° anterior scarp

80°

N 500m

Bulldozed sediments

Steep margins SALAR LLULLAILLACO

Small displacement strike-slip Thrust Normal Early large strike-slip Lithological banding LATE STRUCTURES EARLY STRUCTURES Chronology of formation Key

Figure 5 (continued on next page). (B) Llullaillaco (Central Andes, Chile and Argentina).

668 Geosphere, August 2008

4583 m

4718 m

Small

underlying Cotacotani Cerro

obstacle? lakes Choquelimpie

4571 m

4617 m

hic ob rap sta Topog c le

Cerro Guaneguane Parinacota

5060 m

Bolivia

20°

C o

Chile Parina-

n f

cota i n

30°

h o

Argentina

y n

o 4402 m

50° o

r a

80° p h Strike-slip faults y

Thrust faults/anticlines 4384 m Normal faults Hummocks Toreva blocks

Post-avalanche cover (lava flows and lahars)

1 km N Key 4420 m

Figure 5 (continued on next page). (C) Parinacota (Central Andes, Bolivia and Chile).

Geosphere, August 2008 669

Downloaded from http://pubs.geoscienceworld.org/gsa/geosphere/article-pdf/4/4/657/3336451/i1553-040X-4-4-657.pdf by guest on 30 September 2021 Shea and van Wyk de Vries

3 to 7-km- high scarps 150 km

U C Later cover

Y Olympus Mons

L hides scarp? -2000 m summit caldera

~25000 m

7000 m

-2800 m

region Extension-dominated

~ 0 m 500 m

c om p re ss Strike-slip faults ion -do North mina Thrust faults Other debris ted region avalanches Normal faults Key

Figure 5 (continued). (D) Extraterrestrial Martian rockslide avalanche, Olympus Mons West (Mars).

often relay to strike-slip faults, showing coeval structural features, such as the Median Scarp Blackhawk, Parinacota, Socompa) are massive deformation. Good examples of strike-slip fault (Fig. 1) and the major strike-slip faults remained portions of intact material, which slid and under- relays are seen on Socompa (Fig. 1) and some active throughout the emplacement (van Wyk de went slight backward rotational movements were analyzed in detail by Kelfoun et al. (2008). Vries et al., 2002; Kelfoun et al., 2008). Nor- without being signifi cantly brecciated (Reiche, Probably the most spectacular example of fault- mal faults are generally located toward the cen- 1937; Wadge et al., 1995; van Wyk de Vries et ing on a deposit surface is found at Socompa tral or proximal regions of the deposits, while al., 2001; Clavero et al., 2002). Longitudinal or (Fig. 1), where the entire surface is cut by thrust faults mostly appear in distal regions (this oblique ridges may be observed oriented paral- numerous generations of faults. These faults are is true for all our observed rockslide-avalanche lel or oblique to transport, respectively, while seen in outcrop (Figs. 1B, 1C; van Wyk de Vries deposits, except Mombacho North and South, transversal ridges show perpendicular directions. et al., 2001, 2002; Kelfoun and Druitt, 2005; which do not display thrust faults). Thrusts may Arched ridges are curved concavely or convexly Kelfoun et al., 2008), but are best mapped using generate vertical recurrence of the stratigraphic transversal to slope direction, while marginal aerial images and satellite photos due to the fi eld succession (Fig. 1C; Socompa, Aucanquilcha, accumulation ridges (or “bulldozer facies”; observation scale. At Socompa, the upper part of Llullaillaco, and Ollagüe). Belousov et al., 1999) form when dragged mate- the deposit shows normal, thrust, and strike-skip Hummocks are also an expression of strong rial piles up at the borders of the rockslide ava- faults, while the lower part (which is infrequently stretching around blocks by faults (Glicken, lanche during travel. The rockslide avalanche’s exposed) shows horizontal banding, occasional 1998; Glicken et al., 1981; Voight et al., 1981; underlying substratum is often folded, eroded, diapirs, and fl uid-like mixing textures compat- Shea et al., 2007). Numerous rockslide- or faulted by the moving rock mass. Similar to ible with a basal liquidized zone (van Wyk de avalanche surfaces exhibit rounded or conical the rockslide avalanche, the substratum may Vries et al., 2001). Thus the surface behaved as hummocks with circular and/or elliptic bases be affected by slip surfaces or imbricate ridges a brittle crust (affected by faults) and the base as (e.g., Mombacho, Llullaillaco Northern subunit, (Belousov et al., 1999; Clavero et al., 2005). a liquefi ed and/or fl uidized zone, along which Parinacota, Ollagüe, and Tetivicha) and sizes fre- the mass slid. quently reaching one hundred meters. According ANALOGUE MODELING OF The faulting at Socompa shows distinct gen- to Voight et al. (1981) and Glicken et al. (1981), ROCKSLIDE AVALANCHES erations (Fig. 1), indicating that as the rockslide hummocks result from the creation of horsts avalanche spread, fault sets were partly replaced and grabens during transport and spreading. Most models have focused on the transport by new surface structures. However, some large Slump blocks or Toreva blocks (e.g., Carlson, physics and emplacement dynamics of rockslide

670 Geosphere, August 2008

avalanches, and some have tried to model them This type of “structural to dynamics” differs signifi cantly from dry granular materi- numerically directly (e.g., Kelfoun and Druitt, approach has been used in tectonic and volcano- als. Several characteristics modify their prop- 2005) or indirectly (e.g., Campbell, 1989, 1995; tectonic analogue models for a long time (e.g., erties, such as electrostatic interactions and the Pouliquen and Renaut, 1996; Staron et al., 2001). Merle and Borgia, 1996, and references therein) strong effect of humidity. The chosen analogue While analogue modeling is used extensively in and it can provide the necessary kinematic and material must therefore be mostly composed numerous domains of geology to study structures, structural information to investigate motion. of particles >100 μm. Our sand is 99.7% it has not been applied in this way to rockslide Once such kinematic information is established, silica, with 99.3% above 100 μm, its elastic- avalanches. The current literature on analogue a second and future step would be to investigate ity is equivalent to most silicate materials, its models of rockslide avalanches was compre- the dynamics of the basal layer. Here the basal internal friction angle is 33°, and its density hensively summarized by Pudasaini and Hutter layer is simulated by a low-friction surface, is ~1500 kg m–3. Numerical and/or analogue (2007). Ramps of differing shape and dimension which provides an analogue for the natural slide experiments frequently choose to use spherical have been used with various granular materi- layer. Because the exact process or combination beads in their initial conditions. This is, how- als. However, no model has yet been designed of processes that cause low friction in a sliding ever, not the case in natural deposits, where to reproduce the surface morphology and struc- mass are still not completely understood, this fragments are more angular than rounded; ture or explore the internal distribution of strati- approach provides a simple analogue to what- hence our choice of nonspherical sand par- fi ed material. In none of the described models ever low-friction conditions actually exist. The ticles rather than glass beads. Sand layers are are faults seen, and many of them have deposit approach is similar to the numerical simula- dyed with paints of different colors and can be thickness distributions very different from natural tions of Kelfoun and Druitt (2005) that vary the used accordingly as markers without modi- examples. For example, the experimental depos- coeffi cient of friction in and at the base of their fying the granular medium properties (Don- its in Denlinger and Iverson (2001) and Iverson simulated rockslide avalanche. nadieu, 2000). The cohesion can be increased and Denlinger (2001) are piled up in a mound at A rockslide avalanche involves a gravity- by adding an interstitial fl uid, such as water the base of the slide slope, a distribution found in driven downward slide of rocky fragments that (capillary cohesion; Hornbaker et al., 1997), small rockslides but not in larger events like those spreads until the initial kinetic energy acquired or by adding cohesive powder materials, such studied here. While such models can be useful for from the fall is entirely dissipated by intergranu- as plaster. Note that the addition of very small verifying physical models of granular fl ow, they lar interactions and basal friction. As fault struc- amounts of water does not simulate saturation are not so valuable for exploring the kinematics tures may only appear in the frictional granular and processes such as lahar generation. of rockslide analogues. In particular, they do not regime, the modeling of rockslide avalanches reproduce the observed structural morphology of requires a system that can realistically repro- Choice of Model the surface of natural events. duce analogues to such a state. In this study, we Modeling large-scale landslides poses numer- assume a plug-fl ow type of confi guration, where We chose a geometrically simple model rep- ous scaling problems: scale reduction inevitably the main body suffers a different deformational resenting the failure, transport, and emplace- creates the loss of a fundamental property: the regime than the base (e.g., Kelfoun and Druitt, ment plane. The common characteristic in volume. Consequently, the only ways to obtain 2005). It is therefore crucial to stress that our every sliding surface is slope variation. Failure equivalent runout at smaller scales is to inject intention is to model a granular slide (domi- usually occurs on a surface slope that is above a certain additional quantity of energy to the nantly frictional regime), in which brittle-type or equal to the internal friction angle of the col- system or to lubricate the basal sliding plane. structures can form, rather than a granular fl ow lapsing materials. These materials accelerate Therefore, laboratory experiments without this (dominantly collisional regime). until they reach a gently sloped depositional external energy contribution or basal friction Sand has been often used in analogue model- surface. Following these simple observations, reduction will not be realistic in modeling high ing of volcano-tectonic phenomena (Merle and our laboratory model is a curved ramp built horizontal travel for a given fall height. Even Borgia, 1996; van Wyk de Vries and Borgia, with a fl exible aluminum sheet (4 mm thick, so, with a reduced basal friction, low runout 1996; van Wyk de Vries and Merle, 1996; van 150 cm wide, and ~300 cm long), in which experiments can be used to approach the prob- Wyk de Vries and Francis, 1997; Donnadieu the initiation and deposition slopes could be lem by focusing on deposit structure formation and Merle, 1998; Davies et al., 1999; Merle and modifi ed (Fig. 6A). Other materials previously and deposit morphology, and so to investigate Donnadieu, 2000; Donnadieu, 2000; Lagmay et tested (e.g., varnished wood, Plexiglas, or the transport type and deformation chronology al., 2000; Merle et al., 2001; Cecchi et al., 2004; PVC) are less convenient because they induce (kinematics) from slide initiation to material Oehler et al., 2005; see also Pudasaini and Hut- undesired electrostatic interactions with the runout. This is our approach here. We stress ter, 2007, and references therein). Sand has an sliding material. The basal friction angle of that the models developed here do not attempt internal friction angle similar to most natural the aluminum sheet is relatively low, 24°, to explore the physical processes of the low- rocky materials and a low dry-state cohesion that measured by progressively tilting the surface friction basal décollement: this is simply repro- make it an appropriate tool for laboratory-scale with a pile of grains of known thickness on it. duced as a smooth sliding plane. Thus, the study reduction. However, most deposits enclose rock Note that in choosing a low-friction surface concentrates on the kinematics and structural fractions varying from the micron to the multi- we are replicating the equivalent of a highly development of the upper brittle layer. As the meter scale, and by defi nition, granular media deformed simple shear basal layer in a plug- physics of the low-friction basal layer are still must have diameters >~80–100 μm (Duran, fl ow–type confi guration. Three different ramps unknown, but the mechanisms of fault forma- 1997). Because the sand used for laboratory (later referred to as models 1, 2, and 3) were tion are better constrained, we believe this is a experiments ranges from 60 to 600 µm, scale tested using initiation slopes between 45° and suitable approach. In addition, the surfaces of reduction inevitably entails a sacrifi ce of end- 30° and deposition slopes varying between 0° rockslide avalanches provide abundant struc- member fractions that are present in nature. and 18° (Fig. 6B). Three source boxes of dif- tural data that can be compared with the models, Any particle with a diameter <100 μm is ferent volumes are used (1250 cm3, 2600 cm3, while basal layers are harder to study. classifi ed as powder, the behavior of which and 5250 cm3) preserving a constant thickness

Geosphere, August 2008 671

Downloaded from http://pubs.geoscienceworld.org/gsa/geosphere/article-pdf/4/4/657/3336451/i1553-040X-4-4-657.pdf by guest on 30 September 2021 Shea and van Wyk de Vries

A B 30° Model 1 Adjustable 110 cm initiation slope 18°

40° n

o i Model 2 t 85 cm a i t i 12° n I m c Adjustable 0 9 depositional 2 slope 45° 90 cm n Model 3 o iti os ~0° ep D 150 cm C

γ α A, W, e h H τ ρ V, o, θ

substratum β

Figure 6. (A) Illustration of the ramp used for running scaled laboratory experiments. (B) Different ramp conﬁ gurations, models 1, 2, and 3. (C) Main variables used in dimensioning our analogue model; see Table 2 for terms and values.

of 5 cm. A box-type geometry was preferred certain geometric and dynamic parameters must 2. Variables associated with intrinsic material ρ τ over a wedge or a bowl-shaped initial collapse be scaled. This fundamental step allows deter- properties: density , cohesion o, the inter- structure, because its simple symmetry allows mining of the conditions necessary to ensure nal friction angle θ. better tracking of deformation and changes proportional correspondence between forces 3. Dynamic variables: the basal friction angle γ, in shape geometry. The boxes can be closed acting in nature and those acting in a laboratory fl ow velocity u, and gravity g. from above and possess a frontal trapdoor. The environment (Ramberg, 1981). Accordingly, we The variables in 2 and 3 are very poorly restriction of particle motion when manipulat- defi ne eight dimensionless numbers that serve understood, or even unknown in natural rock- ing the boxes is crucial; when the latter are tilted to compare dynamic and geometric features in slide avalanches, which limits the scaling following the initiation plane, the sand must not both nature and the laboratory. possible for dynamic comparisons. However, move until the frontal trap is opened. This is Nearly all physical quantities can be expressed general statements can be made with caution ensured by a compressive lid tightly closed and with the primary dimensions of length [L], mass about the response of the models to changing tied to the lower part of the source boxes. The [M], and time [T], except for those that are said parameters that may be applicable to natural main advantage of the system is that it allows dimensionless and do not require unit specifi ca- cases. The geometrical and structural param- the emplacement of slope-parallel material tions. The following three groups of variables eters can be handled with the most confi dence, strata that stay in place when tilting the boxes. affect the system (Fig. 6C; Table 2). as these are best known in natural examples, This setting gives the possibility to follow the 1. Variables related to the deposit geometry and best constrained in the models. Both model deformation that affects different layers, and and the sliding environment (source and and natural prototypes deform in the frictional characterize the structures formed (e.g., thrust, depositional surface): horizontal runout regime, which is strain rate insensitive. Thus, normal, or strike-slip faults). L, deposit thickness e, area A, width W, as long as the models remain in this regime, total fall height H, fall height from source velocity and strain rate scale differences will Scaling to beginning of depositional surface h, the not be signifi cant (see discussion in Middle- failure surface average slope α, the depo- ton and Wilcock, 1994). These parameters are In order to guarantee maximum similarity sitional surface average slope β, and the discussed in the section on deformation and between reality and laboratory analogue models, initial volume V. velocity evolution.

672 Geosphere, August 2008

TABLE 2. LIST OF SCALING VARIABLES INVOLVED IN ROCKSLIDE AVALANCHE MOTION, EMPLACEMENT, THEIR DIMENSIONS AND THEIR VARIATION RANGE BOTH IN NATURE AND IN THE ANALOGUE EXPERIMENTS

Variables Dimensions Nature Model Scaling intervals L—avalanche runout [L] 0,1–100 km 0,001-0,003 km L*=Ω*.T*=1/300-1/15000 H—total fall height [L] 500–5000m ~1m 1/100-1/4000 h—fall height before arrival on [L] 500–3000m 0,5–0,9m 1/500–1/6000 depositional slope W—deposit width [L] 1–20000m 0,3–0,7m 1/3–1/30000 A—covered area [L2] 0,1–700 km2 ~1m2 1/106–1/108 V—volume [L3] 0,001–100 km3 1,2.10-12–5,25.10-12 km3 ~1/1012 e—deposit thickness [L] 5m–300m 0,005–0,03 m 1/5000–1/10000 u—avalanche velocity [LT-1] 20–100m.s-1 1 m.s-1 Ω*=1/20–1/100 Ø— clast diameters [L] 60μm–50m 60–600μm 1/1–1/70000 ρ—material density [M L-3] 1300–2600 kg.m-3 1300–1600 kg/m-3 1/2–1/1 g—gravity [L T-2] 9,81 m.s-2 on earth 9,81 m.s-2 1/1 θ—internal friction angle – 30° 33° 1/1 t—emplacement time [T] 30–500 sec 2–3 sec T*=1/15–1/150 γ—basal friction angle – 0–30° 24° ~1/1 α—failure plane average slope – 20–50° 30–45° ~1/1 β—depositional surface average – 0–20° 0–18° ~1/1 slope -1 -2 τo—cohesion [M L T ] 2000–100000 Pa 0–250 Pa 1/400–1/2000

L The number of variables acting on such a Π = equivalent to dimensionless area. The inverse of 4 1 Π complex phenomenon is large; hence to allow 6 could be considered as a factor of dimension- a simple and well-defi ned model, certain impor- V 3 , (1) less volume: tant variables must be selected. For our experi- − Π = Π 1 τ 6 ' 6 ments we consider that essential variables are or dimensionless runout. Scaling of o is as fol- . (5) α β θ τ ρ tan( ), tan( ), tan( ), L, H, A, V, o, , and g. lows: Other variables do not change as signifi cantly in Calculating variation intervals for these val- τ τ ρ experiments and thus are taken as constants. This f( o) = o,V, , g; ues in the model and in nature (Table 2) allows gives ten important variables, three of which are testing for similarity. Figure 7 shows that our τ Π selected as repetitive variables, each showing we choose to give o the exponent 1. The dimen- model numbers overlap with those from independent dimensions (i.e., no variable can be sions are: natural deposits, with a slight overestimation Π expressed by the product of the two others) and of dimensionless resistance 5 in the models. whose behavior is less interesting for the study [ML–1T–2]1[L3]a[ML–3]b[LT–2]c This is expected, as our basal friction is not sig- than the others (Middleton and Wilcock, 1994). nifi cantly lower than in nature, while the driving The three repetitive variables are V, ρ, and g. –1 + 3a + 3b + c = 0 for L force is greatly reduced. τ α β θ Combining L, H, A, V, o, tan( ), tan( ), tan( ) In the models the quantifi ed elements are the with them produces our dimensionless pi (Π) 1 + b = 0 for M, so b = –1 geometric parameters length, width, thickness, numbers. Tan(α), tan(β), tan(θ) are expressed as as well as structure orientation and structure dip. [L]·[L–1] and are dimensionless. Thus the three -2 - 2c = 0 for T, so c = –1 Other morphological parameters are described ∏ α fi rst pi numbers are: 1 = tan( ), or dimension- qualitatively, such as deposit geometry and the ∏ β =−1 less acceleration; 2 = tan( ), or dimensionless hence a . presence of structural features like hummocks ∏ θ 3 deceleration; and 3 = tan( ), which stays con- or lateral levees. We tested the infl uence of ini- stant and will be ignored. tiation and deposition slopes, material cohesion, For the length scaling, we start with the func- Thus: volume, confi nement, the presence of an uncon- tion: f(L) = L,V, ρ, g. L is given an exponent of τ solidated substratum, and particle size. Filmed Π = 0 1, the dimensions are:[L]1 [L3]a[ML-3]b[LT-2]c . 5 1 experiments serve to analyze the evolution of ρ 3 For the products to be dimensionless, the sum of gV (2) deformation (kinematics) and fl ow velocity. The the exponents from [L], [M], [T] must be 0: objective of our model is not to reproduce the can be considered the dimensionless resistance. extreme runout of natural events, but the forma- 1 + 3a – 3b + c = 0 for L To those fi ve pi numbers we can add: tion and evolution of structures, in order to gain insights into their kinematics. Even so, the ramp H b = 0 for M Π = used has suffi ciently low basal friction to permit 6 1 H/L values as low as 0.33. V 3 –2c = 0 for T, c = 0 , (3) Model Reproducibility 1 hence a =− . equivalent to dimensionless initial energy; and 3 A Before results can be interpreted, experi- Π = ments must have an acceptable reproducibility 7 H 2 In this way we defi ne: , (4) and show the same geometric, morphologic,

Geosphere, August 2008 673

Downloaded from http://pubs.geoscienceworld.org/gsa/geosphere/article-pdf/4/4/657/3336451/i1553-040X-4-4-657.pdf by guest on 30 September 2021 Shea and van Wyk de Vries

ratio was 0.37, and the mass covered an area of structures were frequent and well marked ∏ 0.36 1.19 1 ~4500 cm2. Analysis of the deposits shows that at 5250 cm3. The surface ridge wavelength ∏ 0 0.36 2 numerous ridges appeared on the surface; some was nevertheless similar, unaffected by vol- 0.21 21.5 were parallel to oblique to transport direction, ume changes. ∏4 70 depending on their location along the central ∏ 0 97.7 5 axis or near the lateral levees, respectively; oth- Effect of Cohesion ∏ 0.1 30 6 ers were transverse and slightly convex toward 0.001 3.7 ∏6’ the deposit toe. Cohesion can be increased in the initial material in two ways, resulting in both cases in ∏ 0.4 77 7 Characterization of Recurrent Structures a lobate deposit with shorter lateral levees than when using normal sand. Increases in cohe- Figure 7. Compared Π number variation Initial material stratifi cation allowed dis- sion were achieved by adding a certain propor- ranges between the model and nature. The tinction of structures. After transport and tion of plaster to the sand mixture. The effect lower values for each environment are emplacement, the sand mass preserved general of plaster content on L, and therefore H/L, is always on the left and upper values are on stratigraphic order, but recorded signifi cant reported in Figure 11C. Plaster varied between the right side. Note that the Π number value deformation. Transverse and longitudinal cross 1% and 0% in our experiments, corresponding intervals from the model are generally well sections (Figs. 9 and 10) show that thrust faults to cohesions of ~2–10 Pa (Donnadieu, 2000). included into the Π number value intervals (or folds) mostly affect distal zones of the The effect of adding plaster was apparent from from nature, except for Π , which reaches 5 deposit; their dips vary between 15° and 60° the smallest amount: it decreased L, therefore higher values in the model than in nature. and decrease toward the front (cross-sections, increasing considerably H/L. Adding more Figs. 10A, 10B and 11A). On the deposit sur- plaster, however, did not signifi cantly modify face, these were expressed as closely packed the rockslide-avalanche behavior. A rheologi- and dynamic characteristics for identical start- low-amplitude ridges. Lithological repetitions cal threshold was thus reached. Moreover, if ing parameters. Thus standard experiments were also appeared as alternating colored stripes pure plaster is sandwiched between cohesion- conducted before each set of runs to ensure good at the surface with orientations parallel to less layers, the deposit shows geometrical char- reproducibility. In addition, for each experiment thrust structures. Close to the lateral borders, acteristics (e.g., A, L, H/L, e, W) similar to the two or three reiterations were made to average structures had a strike-slip component. On a standard experiment. On the deposit surface, measurements. To avoid cohesion changes due slide-surface parallel cut (Fig. 9A), the lateral however, lumps comparable to hummocky to variations in ambient humidity, each group strike-slip zones were easily distinguished. topography emerge (Fig. 8C), large in central of parameters was tested on the same day. After They showed subvertical to 45° dips toward zones and smaller toward lateral edges. each experiment, the sliding surfaces were the center of the deposit, and their orienta- Alternatively, cohesion may be increased cleaned with an antistatic product to ensure con- tions varied from parallel to oblique (±35°) to by introducing very small quantities of water stant basal friction and avoid parasitic superfi - transport direction. Slight frontal digitations into the initial material (capillary cohesion); cial electrostatic charges. Each time the boxes (Figs. 8A, 8B, 8H, 8J) corresponded to zones this results in even stronger bonds between were fi lled, careful manual shaking guaranteed displaced by strike-slip structures. Similar to grains than previously (cohesion >200 Pa). the most compact particle arrangement. This the standard experiment, the profi le was dis- The humid layer also produced hummocky minimized intergranular spaces and ensured tally raised. Proximal zones were affected by topography on the deposit surface (Fig. 8D). minimum grain remobilization during tilting. normal faults (Figs. 9C, and 10C, 10D) that This time, however, hummock size remained A total of 50 experiments were characterized in were usually more diffi cult to distinguish than almost constant throughout the deposit. In detail while others served as standards. thrust faults because the deposit was much cross section (Fig. 10C), the brittle humid lay- thinner. At the back of certain deposits, small ers showed boudinage structures between nor- RESULTS isolated transport-parallel ridges (or striae) mal sandy layers and had migrated upward. could be observed (Figs. 8D, 8F, 8G). Whether humid sand or plaster was used, Experimental results (geometric measure- each hummock was structurally isolated from ments, structural characteristics, comments) are Effect of Volume the others and formed its own topographic reported in Supplemental Table 12. domain. In both cases the cohesive layer can Figure 11B plots deposit length L as a be placed on top or between noncohesive lay- Standard Experiment function of initial volume (1250, 2600, or ers, but never at the base, or else the mass is 5250 cm3). This relation demonstrates that prematurely stopped. The no. 1 ramp had an initiation slope α = an increase in initial volume also increased 45°, a horizontal emplacement slope β = 0° and L for a constant H, thus slightly decreasing the Effect of Confi nement a material volume of 2600 cm3. The obtained H/L ratio. This effect was more pronounced deposit (Fig. 8A) had an oval lobate shape, with from 1250 cm3 to 2600 cm3 than from 2600 cm3 A rockslide avalanche can be confi ned and/ lateral levees, a maximum length of 241.5 cm, a to 5250 cm3. Also, volume increase ampli- or channelized in two ways: during its down- maximum width of 52.5 cm, and a thickness of fi ed structure density and visibility (Fig. 8J): ward movement along the failure surface by its 1.85 cm. The profi le was distally raised, the H/L while sporadic and indistinct at 1250 cm3, own amphitheater edges, or subsequently by

2If you are viewing the PDF of this paper or reading it ofﬂ ine, please visit http://dx.doi.org/10.1130/GES00131.S1 (Table S1) or the full-text article on www. gsajournals.org to view Supplemental Table S1.

674 Geosphere, August 2008

A B Thrust faults/ anticlines Normal faults

Strike-slip faults Sliding plane striations

Lateral levees/ strike-slip-rich zones

C D Hummocks

10 cm scale Key

Figure 8. Deposits from laboratory scaled rockslide-avalanche experiments. (A) Experi- ment 3, pure white sand, standard experiment for model 1 ramp; the surface shows various structures including lateral levees, but the majority cannot be characterized due to the lack of a E F reference, such as material layering. (B) Experi- Axis of ment 5, three colored sand layers, model 1 ramp; bend thrust, normal, and strike-slip fault systems can now be observed and characterized (see Figs. 9

channel walls and 10 for cross sections). (C) Experiment 14, pure plaster layer between colored sand layers, model 1 ramp; small and numerous hummocks appear on the deposit surface, roughly aligned with thrust faults. (D) Experiment 15, humid sand layer between colored normal sand layers, model 1 ramp; larger and better-deﬁ ned hum- G H mocks are observed. (E) Rockslide-avalanche channelizing experiment with valley-like side- walls on each side. Note the well-marked lateral levees. (F) Mass deﬂ ecting experiment. The origi-

Substratum nal trajectory is shown by the acceleration plane striations. A subunit forms near the deﬂ ecting wall with high concentrations of strike-slip structures. The white dotted line shows the bend, and corresponding arrows show the bend direction. (G) Experiment 16, model ramp 1, black nor- bulldozed mal sand substratum under colored sand lay- I J ers. Note the mixed black and red sand frontal zone, which corresponds to important thrust and accumulation regions. Fault density is also generally higher. (H) Experiment 25, white sand, standard experiment for model 2 ramp. Trans- verse ridges are arcuate toward the front and the deposit shape is elongated with a lobate front. (I) Experiment 43, white sand, standard experiment for model 3 ramp; extensive structures (normal faults) now make up almost half of the tongue- shaped deposit, and lateral levees are extremely affected by strike-slip faults. (J) Experiment 31, three colored sand layers, model 1 ramp, highest initial volume; the density and/or concentration of structures increases with volume. For all experiments, 10 cm scale is on right.

Geosphere, August 2008 675

Downloaded from http://pubs.geoscienceworld.org/gsa/geosphere/article-pdf/4/4/657/3336451/i1553-040X-4-4-657.pdf by guest on 30 September 2021 red sand blue sand Shea and vanred Wyk sand de Vries white sand A initial B blue sand white sand stratigraphy Longitudinal cross-section

Transverse cross-section Transverse cross section C

“ S

h a e v c e a co f extension mpress d r ion ” u s s t u c r a fa t c in e

Longitudinal cross section Strike-slip Plane-parallel Thrust cross section

Figure 9. Cross sections from some scaled experiments. (A) Types of cross section with base-parallel cross section on left. Note that the lithological repetitions follow thrust-fault orientation. (B) Transverse and (C) longitudinal cross sections with preserved layering affected by various structures.

red sand red sand A initial B initial blue sand blue sand white sand white sand stratigraphy stratigraphy compression

Transverse cross section Longitudinal cross section

red sand red sand initial C D initial wet sand blue sand white sand white sand black sand stratigraphy stratigraphy

Longitudinal cross section Longitudinal cross section bulldozer facies

Figure 10. Various longitudinal and transverse cross sections from the experiments. (A) Longitudinal cross section of experiment 7 run on model 1 ramp, with layered colored sand; dominantly compressive structures are observed through the proﬁ le. (B) Transverse cross section of experiment 8 carried out on model 1 ramp; exclusively thrust faults appear. Note that the thrust-fault dip variations form center of mass to lateral margins. (C) Longitudinal cross section of experiment 32 run on model 2 ramp, where a humid sand layer placed between normal dry layers underwent boudinage processes and formed hummocks on the deposit surface. (D) Longitudinal cross section of experiment 16, model 1 ramp; the substratum layer (black sand) covering the sliding plane is entrained, accumulated, and piled up in frontal zones of the deposit.

676 Geosphere, August 2008

A Thrust dip variation vs. Cross section distance% B Length vs. Volume

Uncertainty: dip= ± 3° 60 260 Model 1

Extensive 250 Model 2 40 structures (no meas- 240 dip (°) -urements)

20 Med. volume Length (cm) 230 Small volume Uncertainty: L= ± 0.2% V= ± 5% Large volume 220 0 0 1000 2000 3000 4000 5000 6000 0 20 40 60 80 100 % section horizontal distance Volume (cm³)

C Length vs. Cohesion D Deposit thickness vs. Area covered 250 2.1 Model 1 230 1.7 Model 2 210 Model 3 1.3 190

Length (cm) 0.9 170 Thickness (cm) Uncertainty: Thickness e= <5% A= ± 15% 150 0.5 0246810 0 2000 4000 6000 8000 10000 12000 % Plaster Area (cm²)

E Length vs. Channelization 260

240

220 Length (cm) Length

200 020406080100 Channelized ramp fraction %

Figure 11. (A) Thrust faults dip variations in three deposits from model 1 ramp (corresponding to the three volumes) measured throughout longitudinal cross sections. X-axis represents the percentage of the distance from the beginning of the ramp curvature to the tip of the deposit. Measurements are only made for thrust faults since normal faults affect the thinner portions of the deposit, and are more difﬁ cult to characterize precisely. (B) Length versus volume graph for two of the three ramps (models 1 and 2) and for the three different volumes used in the experiments. (C) Length versus cohesion graph with cohesion expressed as plaster content ranging from 0% to 10%. (D) Effect of channelization on deposit length; x-axis refers to channelized percentage of the total sliding plane. The acceleration and deceleration planes are separated by a bar. (E) Deposit thickness versus area covered for the three ramp models. In a way, the relationship between area and thickness represents the spreading capacity of a mass. The arrow shows that the thinner the deposit and the greater area it covers, the greater the spreading capacity. For all plots, n is the total number of experiments, and the number of experiments used to average a given measurement are in parentheses.

surrounding valleys. Both scenarios were tested initiation plane), the resulting deposit thickness e (initiation plane and depositional plane), lat- on model no. 1 by placing lateral walls sepa- was slightly lower (1.6 cm instead of 1.7 cm in eral levees were more abrupt and longer than rated by 32 cm. Measurements are subsequently standard unconfi ned experiments) and its width in the standard experiment (Fig. 8E), the profi le expressed as a function of percentage of ramp slightly decreased (52–53 cm instead of 55 cm). was still distally raised, and the front showed channelized; accordingly, 100% represents a The distances reached did not show signifi cant small digitations. Thickness stayed unmodifi ed total confi nement whereas 0% denotes an uncon- changes (247–249 cm instead of 242 cm); con- (1.65 cm) and the distance L was slightly lower fi ned ramp (Fig. 11E). When the limiting walls sequently, the H/L ratio stayed approximately (236 cm), though the deposit extended further were placed only on the acceleration plane (e.g., equal. When the ramp was fully channelized back up the ramp. Structures were the same as in

Geosphere, August 2008 677

Downloaded from http://pubs.geoscienceworld.org/gsa/geosphere/article-pdf/4/4/657/3336451/i1553-040X-4-4-657.pdf by guest on 30 September 2021 Shea and van Wyk de Vries

previous experiments, but there were more strike- curvature (difference α – β = 45°) than model produced by no. 3 had uniform profi les. In turn, slip faults, especially toward the channel walls. 3 (α – β = 12°). Figure 11D shows area A varia- the no. 2 ramp formed intermediate profi les, Normal faults were still located at the back of the tions as a function of thickness e. In particular, slightly raised distally. deposit and thrust faults in the frontal zones. when β rose close to the basal friction value of the aluminum pane, the sand was emplaced Deformation and Velocity Evolution Effect of Topographic Obstacles over a greater surface area and the deposit was thinner. Figures 8A, 8H, and 8I show standard Surface deformation and velocity data were To characterize the effect of defl ection by experiments for ramps 1, 2, and 3; while model acquired by video cameras. The major portion nearby topographic heterogeneities, one dis- no. 1 generated lobed ovals, those of no. 2 are of the 2600 cm3 box was fi lled with white sand, tal corner of the ramp was bent so as to form more elongated, and those of no. 3 are tongue while the uppermost 0.5 cm had four colored an oblique-to-fl ow valley wall. The resulting shaped. In all cases the same types of structures equal-area rectangles that served as markers to deposits (Fig. 8F) were asymmetric and the dis- appeared; only their concentration and distribu- follow deformation and variations in rockslide- tances reached slightly lower than without a top- tion varied. Accordingly, the ramp with higher avalanche shape. Figure 12 and Animations 1 ographic obstacle (227.5–239 instead of 242 cm curvature (i.e., no. 1) had deposits affected by and 2 illustrate the evolution of these 4 surfaces in the standard experiment), although deposit higher densities of thrust faults toward the front during the 2.40 s of transport, subdivided into width and thickness were similar. The obstructed than the no. 2 and no. 3 model ramps. Models 2 0.08 s intervals. Velocity was measured at the side of the deposit displayed a higher density of and 3 showed higher concentrations of normal front, center, and back of the moving mass. strike-slip structures and a higher lateral levee faults toward the back and center of the deposit. The front accelerated earlier than the back of than the obstacle-free side; furthermore, a Increases in curvature also were positively cor- the mass, which reached similar velocities later. clear tongue-shaped subunit emerged from the related with lateral strike-slip fault intensity. In The average calculated velocity is ~1 m s–1 for obstructed side, traveling further than the rest cross section, all deposits generated by the no. 1 model 1, 0.85 m s–1 for model 2, and 0.5 m of the mass. Striae or isolated ridges present on ramp showed distally raised profi les while those s–1 for model 3. In Figure 12, the four zones the initiation plane testifi ed to the fl ow direction before defl ection by the topography. time (s) 0 0.08 0.16 0.24 0.32 0.40 0.48 0.56 0.64 0.72 0.80 0.88 0.96 1.04 1.12 Presence of Mobile Substratum Material on 0 Deceleration RB G N 0.3 distance (m) To simulate dry substrata, a 3 mm layer of 0.6 granular substratum (colored sand) was placed 0.9 over the depositional plane. The obtained deposit 1.2 (Fig. 8G) was emplaced only out to 205 cm (compared to 242 cm without substratum), was 1.20 1.28 1.36 1.44 1.52 1.60 1.68 1.76 1.84 0 considerably thicker (2.3 cm compared to 1.7), and retained a similar width (53 cm compared 0.3 to 55 cm). It was lobe shaped, well rounded, 0.6 symmetric, and its lateral levees were promi- 0.9 nent and wide. Two new families of conjugate 1.2

ridges appeared on the deposit surface, with 1.5 orientations varying between ±25°–45° relative 1.8 to ﬂ ow direction. The front had higher concentrations of thrust structures, which deformed not only the collapsed material, but also the under- 1.92 2.00 2.08 2.16 2.24 2.32 2.40 0 lying and entrained substratum. In cross section (Fig. 10D), the dip of thrust faults decreased 0.3 toward distal zones; in particular, at the bull- 0.6 dozed substratum zones, the low dips were 0.9 expressed on the surface by curved ridges point- 1.2 ing toward the front. 1.5

1.8 Ramp Curvature 2.1 Three different ramps were tested in order 2.4 to determine the infl uence of the relationship initiation plane/depositional plane on the rockslide-avalanche features. Models no. 1–no. Figure 12. Deformation stages in the moving granular scaled rockslide avalanche mea- 3 had respective initial slopes of α = 45°, 35°, sured through video images. The X-axis shows total elapsed time (0.08 s intervals) and 30° and depositional slopes of β = 0°, 12°, and the Y-axis shows the distance reached. The initial material was covered by a fi ne and 18°, covering large curvature intervals. (3–4 mm) layer of various colors of sand separated in four equal-area rectangles. This Accordingly, model no. 1 showed a stronger allowed following surface deformation throughout the fl ow.

678 Geosphere, August 2008

compressional stage and tends to spread later- direction. They seem to form when the granular ally while undergoing longitudinal shortening. mass stretches in the fi rst phase of the collapse, Velocities at the center of the mass are always and they are progressively converted or replaced higher than at the edges. by thrust faults as the compression generated at The measured velocities allow us to calculate the front during deceleration is able to propagate bulk strain rates of ~ε ≈3 s–1 in our experiments. suffi ciently far back. This is calculated using a strain of ~9 (20 cm is stretched to 200 cm) in an emplacement of Volume ~3 s. Typically maximum strain rates in nature are ~ε ≈6 × 10−2 s–1, taking a maximum strain As in nature, the laboratory experiments show of ~19 (1 km stretched into 20 km) and a rapid that an increase in initial material volume results emplacement of 300 s. Thus, strain rates are in an H/L decrease; however, the range covered Animation 1. Video sequence of one of the higher in models than nature. The strain rate does not allow defi nitive interpretation. Superfi - analogue experiments involving a thin col- over individual structures, rather than the deposit cial structures remain identical in location and ored sand surface with four equal areas of as a whole in both models and nature, would be orientation when increasing the initial collapse distinct colors above a uniform white sand much less than the bulk strain rate. As long as volume; only their concentration increases. This layer. These types of experiments were used the material stays in the frictional regime, both observation supports our scaling: scale reduc- to characterize deformation during motion analogue and natural examples will form Mohr- tion only has an effect on structure density and and emplacement. To view the animation, Coulomb faults that are rate insensitive. not on their nature and position. you will need Windows Media Player or a multimedia player such as Real Player. If INTERPRETATION Cohesion you are viewing the PDF of this paper or reading it offl ine, please visit http://dx.doi Structures Adding plaster in the granular mixture .org/10.1130/GES00131.SA1 (Animation 1) increases overall cohesion, and tends to pre- or the full-text article on www.gsajournals We review the recurrent structures that appear maturely stop the moving mass while favoring .org to view Animation 1. in our experiments. Thrust faults are the most the formation of compressive structures over ubiquitous, and testify to the compression extensive structures. A very small percentage affecting the frontal and some central regions of is suffi cient to greatly modify the behavior of the mass. In certain experiments they are per- the experiment. By analogy, we can assume two pendicular to fl ow direction, in others they are phenomena are opposed in nature: the formation curved, pointing toward the front and with a ten- of fi ne materials by fragmentation processes dency to parallel the general transport direction should theoretically increase cohesion and thus at the lateral margins of the deposit. They are reduce mobility. In most examples this does not seen to form early, as soon as the mass leaves seem to be the case. We must therefore imagine the acceleration plane for the deceleration plane. that this fi ne rock powder is maintained, at least This early formation is illustrated on video by at the base, in a fl uidized or liquidized state dur- the appearance of lithological repetitions in ing a certain period. In turn, in the bulk part of frontal zones, aligned with thrust structures. The the rock mass, fi ne material may well increase front of the mass slows down fi rst and produces cohesion and favor brittle behavior, akin to the a ramp effect, where material behind accumu- experimental rockslide avalanches. Animation 2. Close-up of Animation 1. The lates and shoves the front forward. If one, or several, cohesive layers (plaster or late-forming strike-slip faults can be eas- The thrust faults are frequently crosscut by humid sand) are sandwiched between normal ily seen just before movement completely strike-slip fault systems, generally subparallel sandy sequences, brittle domains separate from ceases. To view the animation, you will to transport (±15°–20°). Consequently, the latter each other and form hummocky topographies need Windows Media Player or a multime- appear to form after the majority of thrust faults; on the deposit. This observation is crucial in that dia player such as Real Player. If you are however, some are seen to develop alongside the it shows a single deposit could readily contain viewing the PDF of this paper or reading it thrusts, as lateral ramp structures or relay faults. evidence of extension (hummocks) associated offl ine, please visit http://dx.doi.org/10.1130/ The velocity variations causing strike-slip fault- with compressive structures (thrust faults). GES00131.SA2 (Animation 2) or the full- ing can be created by slide surface heterogene- text article on www.gsajournals.org to view ities or by intrinsic differences (e.g., layer thick- Confi nement Animation 2. ness, grain size). They initially form parallel to fl ow, but are subsequently moved toward lateral When the granular slide is partially canalized edges as the bulk of the mass moves forward. on the acceleration plane, the mass spreads nor- During rotation they also acquire a partial thrust mally as soon as it leaves the valley walls. This preserve good lateral symmetry through time; component. In addition, early-formed thrust confi rms that a moving rock mass will initiate the front is strongly stretched during the fi rst faults are rotated at the edge and develop strike- its spreading phase in the reduced-slope depo- phase then gradually joined by the back area, slip motion. sitional surface rather than on the high-slope which in turn undergoes strong stretching. A Normal faults generally materialize at the acceleration plane. In addition, the deposit certain fraction decelerates on the sides, form- back of the model. They are not perfectly linear, reaches similar distances to its unconfi ned ing the lateral levees. Then the front enters a but their orientation stays perpendicular to fl ow version. Hence, partial channel confi nement

Geosphere, August 2008 679

Downloaded from http://pubs.geoscienceworld.org/gsa/geosphere/article-pdf/4/4/657/3336451/i1553-040X-4-4-657.pdf by guest on 30 September 2021 Shea and van Wyk de Vries

parallel to slope direction on the acceleration depositional plane relationship (ramp curva- 8. During the stopping phase, numerous small plane does not signifi cantly modify the slide ture). Furthermore, proximal normal fault cover strike-slip faults form and crosscut the H/L ratio. When the slide is completely cana- and lateral strike-slip density both increase at majority of thrust structures with small lized by slope-parallel valley walls, its runout lower curvature (e.g., model 3). In cross sec- displacements. also remains comparable. Shaller (1991) tions, model 1 deposits show distally raised pro- reached a similar conclusion concerning cana- fi les, whereas those produced by model 3 have MODEL VALIDITY lized runouts. On the other hand, confi nement uniform, homogeneous profi les. Model 2 depos- by valley walls has a strong effect on lateral its form intermediate profi les. In this section we investigate the analogy of structure formation. Strike-slip structures are the internal structures, kinematics, and dynamics accentuated toward both sides, and lateral Kinematics of our models to natural rockslide avalanches. levees are longer, more abrupt, and better marked in the topography. Velocity variation studies demonstrate that Dynamic and Geometric Similarity upon release, the front of the granular fl ow Sliding Plane Heterogeneities accelerates earlier than the back (i.e., the lat- Figure 13 illustrates various relations between ter reaches similar velocities thereafter). Aver- the previously defi ned Π numbers. When signifi cant sliding plane topographical age calculated velocity is 1 m s–1 for model 1, –1 –1 Π variations are present on one side of the fl ow 0.85 m s for model 2, and 0.5 m s for model Dimensionless initial energy 6 versus Π path, these tend to defl ect the mass and provoke 3. Possibly the average slide velocity is posi- dimensionless runout 4 the formation of tongue-shaped subunits in fron- tively correlated with an increase in initiation Data follow a diffuse trend, and for each envi- tal regions. These subunits are bordered by long plane slope. A detailed time analysis of the ronment (extraterrestrial, volcanic, non- volcanic, strike-slip structures that accommodate velocity evolution of deformation affecting the moving and experiments) we can outline individual ten- differences created by the defl ection. The oppo- mass (Fig. 12) allows us to propose the follow- dencies. The experiments plot on the general site side of the mass is able to spread freely, even ing general kinematic model. trend confi rming the validity of measured runout if it was defl ected. 1. Movement initiates from the front, which lengths for such fall heights, considering the undergoes extension with intensity depend- small volumes used. (See Fig. 13A; n = 234 with Dry Substratum on Deceleration Plane ing on initial slope (i.e., higher initial slope 187 natural examples and 47 experiments.) results in a longer and more intense exten- Π When the depositional surface is covered sion period). The back starts moving later. Dimensionless volume 6’ versus Π by a dry, unconsolidated, granular substratum, 2. The whole mass is in movement; fl ow veloci- dimensionless area 7 the slide is slowed. The substratum is rapidly ties between the front and the back are now Generally the data defi ne a relatively well pushed and accumulated at frontal regions of the comparable. constrained trend with positive slope. The moving mass, absorbing and dissipating kinetic 3. Spreading fi rst occurs at the front of the mass, power law regression line shows a good R2 = energy, thus acting as a brake. New families of as this is the zone that fi rst reaches the 0.76. Experimental data are included in this contraction-generated surface ridges formed depositional plane. Velocities are always trend, toward the lower values. Therefore there because the mass was unable to spread freely. higher at the center of the mass than at lat- is a good geometrical similarity between the Note that the substratum used here was only on eral margins. analogue models and nature in terms of area the depositional plane, thus these results can 4. During motion, the material at the front is covered by the deposits for a given collapse vol- simulate only structures in the frictional regime. pushed sideways and forms primary lat- ume. We calculate an equation in the form Rockslide avalanches are known to incorporate eral levees. These levees still are in motion material, possibly increasing the runout, and this and are progressively modifi ed as the mass 2 Π =15Π 3 incorporation occurs in the fl uidized basal shear continues its trajectory. 7 6' , (6) zone (e.g., Hungr and Evans, 2004; Bernard et 5. Frontal velocity starts to decrease; the fi rst al., 2008). This process has not been modeled compressive structures form while velocity which strongly resembles the area and volume here. The structures generated here are more at the back is still maximal and extension relationship described by Dade and Huppert akin to the bulldozed facies of the rockslide- still affects these regions. The central part (1998) and Kilburn and Sorensen (1998); their avalanche front (Belousov et al., 1999). of the mass still pushes the material side- equations apply to Bingham type materials, ways and provokes the formation of sets of and just like these authors we fi nd it diffi cult to Slide Plane Curvature strike-slip faults near lateral levees. explain why such a relation applies to granular 6. While the front still moves forward, frontal materials in our frictional model. (See Fig. 13B; When the depositional slope is high (e.g., compression propagates toward the back n = 157 with 47 experiments and 110 natural when it approaches the basal friction angle), of the mass (ramp effect), and the central deposits.) the area covered by the deposit is greater and part slows down. During this phase, long its thickness is reduced compared to the stan- strike-slip faults appear in various zones to Structural and Morphological Similarity dard experiment. Thus, spreading capacity is accommodate relative velocity differences positively correlated with depositional slope already present throughout the fl ow. Shape increase. Model 1 experiments form oval-shaped 7. When kinetic energy is insuffi cient at the front Laboratory deposits show three distinct shapes deposits, while model 2 deposits are more elon- to prevail over frictional forces, the mass (lobe shaped, irregular, and tongue shaped) out gated, and model 3 deposits are tongue shaped. stops. The last compressive and strike-slip of the four possible, which furthermore depend As a consequence, there is a simple correlation structures appear where still mobile por- on the initiation versus depositional plane ratio, between slide deposit shapes and the initiation- tions are blocked by immobile regions. surface obstacles, and confi nement. Fan-shaped

680 Geosphere, August 2008

deposits such as Ollagüe (Fig. 4E), Tetivi- eva blocks such as Parinacota (Fig. 5C). These show a higher variability and density of surface cha (Fig. 5A), and the northeastern margin of deposits may be formed by retrogressive failure structures than unconfi ned equivalents. The Carl- Llullaillaco (Fig. 5B) are not obtained in our not modeled here. Uniform profi les are observed son (Fig. 4A) deposit and both Llullaillaco north- models. Tongue-shaped deposits comparable to at Mombacho South (Fig. 2C), Chaos Jumbles ern and southern defl ected subunits (Fig. 5B) Aucanquilcha (Fig. 4C), to the northeastern sub- (Fig. 2B), Aucanquilcha (Fig. 4C), Llullaillaco display a higher concentration in longitudinal unit of Socompa (Fig. 1), and the southeastern (Fig. 5B), the eastern side of Socompa (Fig. 1), ridges at lateral margins, just as in our granu- extremity of Llullaillaco (Fig. 5B) can be repro- Lastarria (Fig. 4D), and inferred for Olympus lar canalized experiments (Fig. 8E). Defl ected duced in our experiments with integral chan- Mons landslides on Mars (Fig. 5D). In labora- rockslide avalanches often form accumulation nelization of the rockslide avalanches, or with a tory experiments, uniform profi les are achieved levees and/or strike-slip fault–rich zones (e.g., model 3 type ramp with relatively low initiation for deposits that have fewer compressive struc- northeastern zone of Chaos Jumbles, Fig. 2B; planes (~30°) and relatively high depositional tures, on ramps with relatively steep depositional eastern zone of Martinez mountain, Fig. 4B; surfaces (18° in model 3). Tongue-shaped sub- planes (e.g., model 3). Distally raised profi les northern side of Parinacota, Fig. 5C; frontal units also develop in the laboratory when the occur at Mombacho North (Fig. 2C), Ollagüe levees at Socompa, Fig. 1), very similar to those mass is defl ected by topographical obstacles (Fig. 4E), at the western side of Socompa illustrated by analogue models when a portion of (Fig. 8F). The lobe-shaped deposits are prob- (Fig. 1), and at Blackhawk (Fig. 2A). These are the sliding surface is obstructed (Fig. 8F). ably the most common both in the models and reproduced in the majority of our experiments in natural examples. associated with high concentrations of compres- Deposit characteristics sive structures, when a model 1–type ramp with Natural and analogue model deposits share Profi le a high initiation slope and a low depositional the following morphological and structural Experiments successfully reproduced three slope is used. In these, compression provokes aspects. out of the four types of profi les observed in piling-up of the material, thus adding thickness 1. Preserved gross stratigraphy (all color-layered natural deposits: i.e., distally raised, uniform, to distal zones. experiments, Mombacho North and South, and irregular profi les. Proximally raised profi les Ollagüe, Socompa). are never obtained in analogue deposits, and in Confi nement 2. Incorporation of substratum by bulldozing nature these seem to characterize deposits with Deposits resulting from defl ected, lightly or (experiment in Fig. 8G and cross-section a signifi cant proportion of slump and/or tor- heavily confi ned rockslide avalanches tend to Fig. 10D, Carlson, Blackhawk, Mombacho

A ∏4 vs ∏6 Experiments Non-volcanic 15 15 20 Extraterrestrial 10 10 6 Terrestrial non-volcanic 6

16 ∏ 5 ∏ 5 Terrestrial volcanic Experiments 0 0 H 12 010203040 010203040 ∏ = Initial energy 6 1 ∏4 ∏4 V 3 8 Runout Extraterrestrial Volcanic 15 15 4 10 10 6 6 ∏ 0 ∏ 5 5 0 5 10 15 20 25 30 35 40 0 0 L 0 10203040 010203040 ∏4 = 1 ∏4 ∏4 V 3

B 1000 Extraterrestrial Volcanic Non-volcanic 100 Experiments 2 A H

= 10 7 ∏ Figure 13. Π number similarity diagrams. (A) Initial energy 1 Π versus runout Π for the laboratory experiments and the

Area 6 4 three natural environments (extraterrestrial, non-volcanic, and Volume volcanic). Data for each environment are also shown and con- 0,1 0,001 0,01 0,1 1 10 100 strained to individual trends on independent graph for clarity. Π Π V (B) Dimensionless area 7 versus dimensionless volume 6' for = ∏6' 3 all four environments. The general trend deﬁ ned by the data H and the linear regression equation are shown.

Geosphere, August 2008 681

Downloaded from http://pubs.geoscienceworld.org/gsa/geosphere/article-pdf/4/4/657/3336451/i1553-040X-4-4-657.pdf by guest on 30 September 2021 Shea and van Wyk de Vries

North, Ollagüe, Tetivicha, Llullaillaco, logue model demonstrated that collapses com- deposits may have distally raised profi les due Parinacota, and Socompa). posed of initially heterogeneous materials (e.g., to ramp and/or bulldozer structures, which in 3. Presence of initiation and/or failure plane the alternation between scoria layers and lava turn could derive from low depositional slopes striations (experiments in Figs. 8D, 8F, fl ow units in volcanic environments, the differ- (i.e., 0° in model 1) associated with relatively 8G; Chaos Jumbles, Lastarria, Mombacho ent lithologies in mountain chains, and the plas- high acceleration slopes (i.e., 45° in model 1). North, and Tetivicha). ter and/or sand alternation in the experiments) This is effectively the case at Mombacho North, 4. Vertical recurrence of lithologies (color- tend to generate hummocky surfaces like those Ollagüe, and Blackhawk. Their profi les are all, layered experiments, models 1 and 2; at Mombacho, Parinacota, Ollagüe, and Tetivi- to some extent, distally raised, and their initial Aucanquilcha, Socompa, Llullaillaco) cha. The latter are likely to obscure the possibly acceleration planes have high slopes compared to associated with thrust structures (central to existing ridged structures. their deceleration plane. In contrast, experiments distal transverse ridges seen in regions of show that relatively high-slope depositional all experiments; Chaos Jumbles, Carlson, Bulldozer Structures planes (i.e., 18° in model 3) associated with rela- Martinez Mountain, Blackhawk, Last- tively low acceleration planes (i.e., 30° in model arria, Aucanquilcha, Ollagüe, Tetivicha, Experiments also showed that emplacement 3) tend to generate uniform profi les. This obser- Llullaillaco, Parinacota, Socompa, and of a rockslide avalanche on a dry, unconsolidated vation is verifi ed at Lastarria, Aucanquilcha, Olympus Mons). substratum in the depositional area, creates bull- Chaos Jumbles, and Llullaillaco, where depo- 5. Strike-slip structures and longitudinal ridges dozer structures in its frontal regions similar to sitional planes have higher slopes than at most (all experiments and all natural deposits those described at Shiveluch by Belousov et al. other sites. The situation is somewhat different at except at Mombacho where hummocky (1999). Other natural examples fi t this model, Socompa, where a uniform profi le dominates the topographies hide most existing longitudi- such as Blackhawk, Llullaillaco (fi ne pyro- eastern regions (5°–10° depositional slope) and nal ridges). clastic deposits acting as substratum under the a raised profi le dominates the northwestern areas 6. Hummocky surfaces (Figs. 8C and 8D; southern subunit were pushed and accumulated (slope opposed to fl ow direction). Mombacho North and South, Parinacota, at the front; J.E. Clavero, 2006, personal com- Ollagüe, Llullaillaco, Tetivicha). mun.), Ollagüe (salar deposits; Clavero et al., Strong Relationship Between Deposit 7. Oblique ridges (toward lateral margins in 2005), and Tetivicha (salar deposits). At Sher- Shape and Surrounding Topography all experiments; Chaos Jumbles, Marti- man (Shreve, 1966), large amounts of blocks of nez Mountain, Aucanquilcha, Lastarria, ice and snow from the underlying glacier were When our scaled models are neither confi ned Llullaillaco, Socompa). bulldozed by the traveling rock mass. Particular or canalized, and are carried out on a low decel- 8. Lateral levees (most experiments; Carlson, care must nevertheless be taken to avoid confus- eration slope coupled with a high acceleration Martinez Mountain, Blackhawk, Lastarria, ing substratum entrained and bulldozed during slope (i.e., model 1), the resulting deposits are Llullaillaco, Socompa). the fl ow, and substratum involved in the initial well spread laterally and have lobate spoon The analogue model partially succeeds in failure (e.g., Socompa, van Wyk de Vries et shapes comparable to those of Mombacho reproducing normal faults and/or proximal al., 2001; Mombacho, Shea et al., 2007). Note North and South, Tetivicha, Ollagüe, and Olym- transverse ridges; always in proximal regions that we did not model entrainment of a water- pus Mons deposits. The inverse situation, rela- in the experiments, located in various regions in saturated substrate (because of capillary cohe- tively low acceleration slope and/or relatively nature. It also crudely reproduces gross inverse sion issues at the laboratory scale), which is high deceleration slope (i.e., model 3), produces grading when hummocky topography is gener- thought to undergo liquefaction effects and tongue-shaped deposits similar to Aucanquil- ated and fi ne particles tend to gather between enhance mobility (Hungr and Evans, 2004). We cha, Lastarria, Chaos Jumbles, and Martinez and under large analogue block portions (e.g., did not model substrata erosion and incorpo- Mountain. The only notable exception to this is cross section, Fig. 10C). ration on the acceleration plane that may also Blackhawk, which has an elongated shape, even induce bulking (e.g., Bernard et al., 2008). though its initial slope is high and its deposi- IMPLICATIONS OF MODEL AND tional slope low. The difference between lobe- NATURAL STRUCTURES Ramp structures and deposit profi les shaped and/or spoon-shaped and fan-shaped deposits could arise from general dynamics of Ridged or Hummocky Surface? In our scaled model, when the depositional the rock mass during its emplacement: a mass slope decreases, compressive structures increase decelerating strongly at its front could produce Our experiments show that dominantly fi ne- and propagate toward the center of the deposit, a lobe aspect, whereas a mass completely free to grained and homogeneous granular material resulting in a frontal ramp effect. This piling up spread in most directions and not slowed down slides tend to generate deposits with well-marked of materials differs signifi cantly from the bull- by its front would produce a fan-shaped aspect, superfi cial structures (dominantly ridged) com- dozer structures described here and is more simi- as suggested by Shreve (1966) for the Sherman parable to those of Lastarria, southern Llullail- lar to the structures described by Shreve (1968) landslide. Unfortunately our experiments are not laco, and Aucanquilcha. Even though its initial at Blackhawk. The bulldozer effect derives from able to confi rm this hypothesis, perhaps because material was heterogeneous in size and type, the ripping off, transport, and marginal accu- of insuffi cient kinetic energy when our granular Socompa could also be included in this cat- mulation of unconsolidated or fragmented sub- fl ow approaches the depositional plane or sim- egory, knowing that the fi ne substratum portion stratum during fi nal deceleration, whereas the ply because basal friction is still too high com- of the deposit actually makes up 90% of the total ramp effect corresponds to internal imbricate pared to natural large-scale landslides. Irregular volume (van Wyk de Vries et al., 2001); thus its layers in frontal portions of the mass generated shapes such as those at Parinacota or Carlson behavior was probably dominated by the latter by thrust faults during the deceleration phase. deposits are provoked by strong confi nement fi ner material and formed ridge structures on the In both cases, fault dip tends to decrease toward or channelization. The Carlson rockslide was deposit topography. On the other hand, the ana- the deposit front. Accordingly, natural avalanche fully canalized in a 300-m-wide by ~1-km-long

682 Geosphere, August 2008

valley and was only able to spread after leav- amplitude–long-wavelength structures. This and lateral push. They are modifi ed as the mass ing the valley walls. At Parinacota, the northern observation is less valid when focusing on trans- progresses forward, and the oblique ridges near area is obstructed by high topographic barri- port-parallel structures such as strike-slip faults, them observed on the video form to accommo- ers whereas the southern regions allowed the which only form thin, low-amplitude discon- date the lateral push. mass to advance with only slight confi nement. tinuities on top of and within the deposit. The At Ollagüe, the southern zone that entered the question still arises, do superfi cial structures in Two Morphological and Two Dynamic Salar de Carcote reached lower distances than natural rockslide-avalanche deposits represent Types of Rockslide Avalanches the northern deposit that spread outside the salar internal structures formed during fl ow, or sur- (salt fl at). The salar is quite unconsolidated, face waves? From our experiments, we deduce From the above interpretations, and as a and probably absorbed kinetic energy from the that a majority of early forming strike-slip general summary, we can separate rockslide arriving mass, in a similar fashion to our experi- faults are preserved after runout; nevertheless, avalanches into two textural categories, hum- ments when granular substratum was on top of transverse structures (normal and thrust faults) mocky and ridged, and two dynamic types, the sliding surface. Water is probably another mostly represent the last deformation phases in dominantly extensional and dominantly com- medium that opposes free spreading by absorb- the regions where they appear. pressive rockslide avalanches. Figure 15A illus- ing signifi cant kinetic energy. The Mombacho trates the formation of these two pairings (type North deposit (Las Isletas) entered Lake Nicara- Active Levees 1—dynamic pole 1 and morphological pole 1; gua and was possibly slowed down in this way. type 2—dynamic pole 2 and morphological pole At Socompa, eastern hills provoked the defl ec- Analogue model lateral levees appear either 1; type 3—dynamic pole 2 and morphological tion of the moving mass to the northeast, which when material is deposited and pushed aside by pole 2; type 4—dynamic pole 2 and morphologi- probably produced the median scarp described moving material at the front (unconfi ned rock- cal pole 2). Generally, a collapse occurring on a by Francis et al. (1985), Wadge et al. (1995), and slide avalanches), or when the mass is cana- highly curved sliding surface will form a laterally van Wyk de Vries et al. (2001). lized, in which case the levees are more abrupt well spread, distally raised deposit with mostly and bounded by numerous strike-slip faults. In compressive structures (type 1), whereas one Structural Genesis nature, levees at Blackhawk, Carlson, Martinez occurring on a moderately curved sliding surface Mountain, Lastarria, Llullaillaco, and Socompa will display a laterally restrained uniform deposit Our kinematic characterization shows that the are generally associated with confi ning environ- affected dominantly by extension (type 2). The front of the slide stops prior to its center, but after ments. Channelization or confi nement, whether major difference between these dynamic poles the back, which agrees with observations made in nature or in the models, tend to increase lon- is in the propagation of compression during the by Shreve (1968), but only partly agrees with the gitudinal ridge and/or strike-slip structure den- slide stopping phase. Thus at Mombacho South, numerical model of Kelfoun and Druitt (2005), sity, particularly near the margins. On the other Ollagüe, Tetivicha, Socompa, Llullaillaco, and where the mass freezes fi rst in its proximal then hand, natural unconfi ned deposits do not show Aucanquilcha, compression is concentrated in at its distal zones. The difference may be related levees akin to our experiments. Consequently, frontal zones and did not propagate backward. In to their dominantly compressive or dominantly rockslide-avalanche levees could differ sig- contrast, at Blackhawk, Chaos Jumbles, Carlson, extensional affi nities. In our experiments, exten- nifi cantly from dense pyroclastic fl ows or lava and Martinez Mountain, compression affects a sion initially affects the mass, and is later replaced fl ows in the dynamics of their formation. Where great fraction of the deposit from front often up by compression, which indicates that normal the later levees are usually thought to represent to central zones. faults could be converted into thrust faults. This a recording of the highest thickness reached by Furthermore, collapses composed of het- strong connection between the two explains the the fl ow in zones that become stationary, they erogeneously sized and/or heterogeneously possibility of observing compressive features are associated in our case to in-motion processes competent material will generate a deposit associated with boudinage structures in the same deposit. Lateral velocity variations throughout the mass are probably responsible for the formation of strike-slip faults, two different families Acceleration Depositional of which can be distinguished: large, infrequent, plane plane minor strike-slip faults early strike-slip structures, from late, smaller but lateral levees abundant strike-slip faults. In all cases, the video images show that many structures form early dur- thrust faults/anticlines ing fl ow and not just during the stopping phase. lithological repetitions Figure 14 summarizes the progression of structure formation inferred from our models. major strike-slip faults STRUCTURE normal faults/troughs Is Superfi cial Morphology Representative primary hummocks break into smaller hummocks? of Internal Structures? Initiation Stop Transverse surface ridges formed in our ELAPSED TIME experiments do not share the same spacing when compared to internal thrusts from the inte- Figure 14. Summarized diagram of structure genesis in laboratory scaled rockslide rior of the deposit. Most likely, surface granular avalanches during fl ow. The X-axis represents the transport time line, from initia- rearrangements take place progressively as the tion to stoppage, and the Y-axis simply separates distinct structure types. The bar mass slides, preventing the formation of high represents the limit between the acceleration plane and the deceleration plane.

Geosphere, August 2008 683

Downloaded from http://pubs.geoscienceworld.org/gsa/geosphere/article-pdf/4/4/657/3336451/i1553-040X-4-4-657.pdf by guest on 30 September 2021 Shea and van Wyk de Vries

Dynamic pole 1 Dynamic pole 2 Dynamic pole 1 Dynamic pole 2 Morphological pole 1 Morphological pole 1 Morphological pole 2 Morphological pole 2

β β 1< 2 ac α >α a 1 2 c α c c 2 el e er α l atio 1 er n d atio ecele n deceleration ration β β 1 2

Ridged Ridged Hummocky Hummocky

Normal f. dominant Thrusts f. dominant

Examples: Examples: Examples: Examples: Blackhawk Aucanquilcha Mombacho North Mombacho South Carlson Socompa Ollagüe Parinacota Martinez Mountain Lastarria Tetivicha Chaos Jumbles Llullaillaco Olympus Mons?

B B: Blackhawk Ca: Carlson 40 Compressive So 1000 Compressive Mt: Martinez Mountain Mn 30 Extensional Extensional Ga So A: Aucanquilcha ) Ga 2 100 A So: Socompa La: Lastarria I 20 A Ms F Mn: Mombacho North La B

L (km) L Ms Sh Ms: Mombacho South F (km A 10 O 10 La Mt O Mt O: Ollagüe F: Flims Sh B I Ca I: Iriga Ga: Galunggung Ch Ca Mn Ch 0 1 Sh: Sherman F: Flims 0.01 0.1 1 10 100 0.01 0.1 1 10 100 V (km3) V (km3)

Figure 15. (A) Formation of the two morphological types of rockslide avalanches (hummocky and ridged) and the two dynamic types (extensional and compressive) with typical associated structures and their natural occurrences. (B) (L) Length versus volume (V) and area (A) versus volume plots (in both cases V axis is in logarithmic scale). Some of the studied rockslide avalanches (Parinacota, Tetivi- cha, Olympus Mons, and Llullaillaco) were not plotted due to insufﬁ cient data. Instead, some other known natural examples not detailed in this study (Galunggung, Sherman, Flims, and Iriga) were taken as additional data points because their extensional or compressive natures are more conspicuous. Just like the other rockslide avalanches, they were the objects of structural and geometrical analysis before being classiﬁ ed into their categories. Note that except for Mombacho South in both graphs, all compressive rockslide avalanches are below their extensional homologues, meaning that they reach shorter distances and cover smaller areas.

684 Geosphere, August 2008

ACKNOWLEDGMENTS showing a hummocky, ridge-poor and structure- tural features seen in natural examples, and the poor topography (types 3 and 4), whereas col- runout in the experiments is similar to that of We thank Tim Druitt, Valentin Troll, Tim Davies, lapses composed of homogeneously sized and/ natural small-volume events. However, because and Frances Garland for discussion and comments. or homogeneously competent rock fractions of the different balance between driving force We also thank two anonymous reviewers for their tend to display a hummock-poor, ridged, and and friction in the model (which is too high), the careful examination, as well as Thomas Walter and structure-rich surface on their deposit (types rockslides stop earlier in the models than many Clause Siebe for their useful corrections. 1 and 2). Our hypothesis of hummock forma- volcanic examples. Even so, model parameters REFERENCES CITED tion consequently differs from that proposed by such as dimensionless volume and dimension- Clavero et al. (2002) and Clavero et al. (2005) less surface area show relationships similar to Belousov, A., Belousova, M., and Voight, B., 1999, Multiple for Parinacota and Ollagüe, respectively, which those found in Dade and Huppert (1998) and edifi ce failures, debris avalanches and associated erup- suggests that they derive from zones originally Kilburn and Sørensen (1998). tions in the Holocene history of Shiveluch volcano, Kamchatka, Russia: Bulletin of Volcanology, v. 61, separated by faults and not from differences in The structures form from the initiation of the p. 324–342, doi: 10.1007/s004450050300. material sizes and cohesion and/or lithology. slide, as the material is released. Normal faulting Bernard, B., van Wyk de Vries, B., Barba, D., Robin, C., Hummocks do not appear on monolithologic is developed fi rst as the slide extends. Thrusts are Léyrit, H., Alcaraz, S., and Samaniego, P., 2008, The Chimborazo sector collapse and debris avalanche: deposits; hence a minimum cohesion differ- produced when the front area starts to deceler- Deposit characteristics as evidence of emplacement ence must exist within initial material. Certain ate more than the back of the slide. Differential mechanisms: Journal of Volcanology and Geothermal factors may infl uence these major groups, such movements then create strike-slip faults that Research (in press). Borgia, A., and van Wyk de Vries, B., 2003, The volcano- as the presence of an unconsolidated or compe- remain active until the slide halts. Final deposit tectonic evolution of Concepcion, Nicaragua: Bulletin tent substratum or even surrounding topography surfaces can be grouped into two textural types, of Volcanology, v. 65, p. 248–266. Campbell, C.S., 1989, Self-lubrication for long runout land- (mountains, valleys, and plains). hummocky or ridged. The hummocky textures slides: Journal of Geology, v. 97, p. 653–665. The validity of our approach can be summa- are produced by differences in the cohesion of Campbell, C.S., Cleary, P.W., and Hopkins, M.J., 1995, rized onto length versus volume and area versus the initial layering. Ridged textures are produced Large-scale landslide simulations: Global deformation, velocities and basal friction: Journal of volume plots (Fig. 15B), where we separate a cer- when the initial material is homogeneous. Geophysical Research, v. 100, p. 8267–8283, doi: tain number of the studied rockslide avalanches The structures in the fi nal deposits can also 10.1029/94JB00937. according to their structural type: dominantly be grouped into two kinematic types, compres- Cecchi, E., van Wyk de Vries, B., and Lavest, J.M., 2004, Flank spreading and collapse of weak-cored volcanoes: extensional or dominantly compressive (see sional or extensional. The limit between the Bulletin of Volcanology, v. 67, p. 72–91, doi: 10.1007/ Supplemental Table 23). Unfortunately, in some two is controlled by the topographic profi le of s00445-004-0369-3. Clavero, J.E., Sparks, R.S.J., and Huppert, H.E., 2002, Geo- cases (Parinacota, Tetivicha, Olympus Mons, and the slide. If it goes from steep to fl at, then the logical constraints on the emplacement mechanism Llullaillaco) data are insuffi cient for plotting, so regime is compressional; if the profi le is more of the Parinacota avalanche, northern Chile: Bul- we added some deposits not described in detail regular, then extension dominates. Primary letin of Volcanology, v. 64, p. 40–54, doi: 10.1007/ s00445-001-0183-0. in this paper (Flims, Sherman, Galunggung, and results deduced from our classifi cation showed Clavero, J.E., Polanco, E., Godoy, E., Aguilar, G., Sparks, S., Iriga) for which the data are available. Similar that natural compressional rockslide avalanches van Wyk de Vries, B., Pérez de Arce, C., and Matthews, to the described avalanche deposits, a structural have lower runout and cover a smaller area than S., 2005, Substrata infl uence in the transport and emplacement mechanisms of the Ollagüe debris avalanche, north- analysis and classifi cation were made for these their extensional equivalent at similar volumes. ern Chile: Acta Vulcanologica v. 16, p. 31–48. avalanches following the same method. Length Our analysis shows that a brittle plug fl ow Crandell, D.R., 1989, Gigantic debris avalanche of Pleis- tocene age from ancestral Mount Shasta Volcano, alone was chosen instead of the H/L coeffi cient model or basal slide-dominated model fi ts California, and debris-avalanche hazard zonation: U.S. for plotting due to the identical meaning it pro- well with both natural and model rockslide Geological Survey Bulletin 1861, 32 p. vides, but with less scatter (Legros, 2002). The avalanches. Mixing fl ow, turbulence, complete Cruden, D.M., and Varnes, D.J., 1996, Landslide types and processes, in Turner A.K., and Shuster, R.L., eds., resulting graphs show that in both cases almost fl uidization, or simple shear throughout the ver- Landslides: Investigation and mitigation: Transporta- all compressive data are located below exten- tical profi le are not appropriate models. Very tion Research Board and National Research Council sional data (except for Mombacho South). This low friction at the base or in a thin basal layer Special Report 247, p. 36–75. Dade, W.B., and Huppert, H.E., 1998, Long-runout rock- confi rms that for the same volume, a compres- is necessary to allow the very long runouts. A falls: Geology, v. 26, p. 803–806, doi: 10.1130/0091- sive rockslide avalanche will tend to reach lesser component of motion resistance comes from the 7613(1998)026<0803:LRR>2.3.CO;2. Davies, T.R.H., McSaveney, M.J., and Hodgson, K.A., 1999, distances and cover smaller areas than exten- brittle plug of the slide, which spreads by bulk A fragmentation-spreading model for long runout rock sional rockslide avalanches. pure shear and faulting only so long as the driv- avalanches: Canadian Geotechnical Journal, v. 36, ing forces are greater than the resistance. p. 1096–1110, doi: 10.1139/cgj-36-6-1096. Denlinger, R.P., and Iverson, R.M., 2001, Flow of variably CONCLUSIONS The experiments and natural examples show fl uidized granular masses across three-dimensional that our structural approach can provide much terrain 2, Numerical predictions and experimental Natural rockslide avalanches are full of needed information on rockslide-avalanche kine- tests: Journal of Geophysical Research, v. 106, no. B1, p. 553–566, doi: 10.1029/2000JB900330. fault structures that can be successfully repro- matics. To model the emplacement of a rockslide Donnadieu, F., 2000, Destabilization of volcanic edifi ces duced using a sand-based granular slide model. avalanche, such structural work is a crucial fi rst by cryptodomes: Analog modeling and numerical approach [Ph.D. thesis]: Université Blaise Pascal, Cler- Any model of rockslide-avalanche kinemat- step in determining the conditions of emplace- mont-Ferrand II, 256 p. ics or dynamics must take such structures into ment. For predicting runouts and paths, the Donnadieu, F., and Merle, O., 1998, Experiments on the account, and the analogue models provide a topography and potential source material must indentation process during cryptodome intrusions: New insights into Mount St. Helens deformation: Geology, useful way of exploring their development and be characterized, and such data can be used for v. 26, p. 79–82, doi: 10.1130/0091-7613(1998)026<00 signifi cance. The models reproduce all struc- input into any physical model of emplacement. 79:EOTIPD>2.3.CO;2.

3If you are viewing the PDF of this paper or reading it ofﬂ ine, please visit http://dx.doi.org/10.1130/GES00131.S2 (Table S2) or the full-text article on www. gsajournals.org to view Supplemental Table S2.

Geosphere, August 2008 685

Downloaded from http://pubs.geoscienceworld.org/gsa/geosphere/article-pdf/4/4/657/3336451/i1553-040X-4-4-657.pdf by guest on 30 September 2021 Shea and van Wyk de Vries

Duran, J., 1997, Sables, Poudres et Grains: Paris, Eyrolles, p. 1111–1114, doi: 10.1130/0091-7613(1989)017<11 Siebert, L., 1984, Large volcanic debris avalanches: Char- 251 p. 11:MOLRAE>2.3.CO;2. acteristics of source areas, deposits, and associated Eppler, D.B., Fink, J., and Fletcher, R., 1987, Rheologic Merle, O., and Borgia, A., 1996, Scaled experiments of eruptions: Journal of Volcanology and Geothermal properties and kinematics of emplacement of the Chaos volcanic spreading: Journal of Geophysical Research, Research, v. 22, p. 163–197, doi: Jumbles rockfall avalanche, Lassen Volcanic National v. 101, p. 13,805–13,817, doi: 10.1029/95JB03736. 10.1016/0377-0273(84)90002-7. Park, California: Journal of Geophysical Research, Merle, O., and Donnadieu, F., 2000, Indentation of volcanic Siebert, L., Glicken, H., and Ui, T., 1987, Volcanic hazards v. 92, p. 3623–3633, doi: 10.1029/JB092iB05p03623. edifi ces by the ascending magma, in Vendeville, B., et from Bezymianny- and Bandaï-type eruptions: Bul- Francis, P.W., and Wells, G.L., 1988, Landsat Thematic al., eds., Salt, shale and igneous diapirs in and around letin of Volcanology, v. 49, p. 435–459, doi: 10.1007/ Mapper observations of debris avalanche deposits Europe: Geological Society of London Special Publi- BF01046635. in the Central Andes: Bulletin of Volcanology, v. 50, cation 174, p. 43–53. Siebert, L., Alvarado, G., Vallance, J.W., and van Wyk de p. 258–278, doi: 10.1007/BF01047488. Merle, O., Vidal, N., and van Wyk de Vries, B., 2001, Exper- Vries, B., 2006, Large-volume volcanic edifi ce failures Francis, P.W., Gardeweg, M., O’Callaghan, L.J., Ramirez, iments on vertical basement fault reactivation below in Central America and associated hazards, in Rose, C.F., and Rothery, D.A., 1985, Catastrophic debris volcanoes: Journal of Geophysical Research, v. 106, W.I., et al., eds., Volcanic hazards in Central America: avalanche deposit of Socompa volcano, north Chile: no. B2, p. 2153–2162, doi: 10.1029/2000JB900352. Geological Society of America Special Paper 412, p. Geology, v. 13, p. 600–603, doi: 10.1130/0091- Middleton, G.V., and Wilcock, P.R., 1994, Mechanics in the 1–26, doi: 10.1130/2006.2412(01). 7613(1985)13<600:CDADOS>2.0.CO;2. Earth and environmental sciences: Cambridge, Cam- Staron, L., Vilotte, J.P., and Radjai, F., 2001, Friction and Glicken, H., 1991, Sedimentary architecture of large volca- bridge University Press, 455 p. mobilization of contacts in granular numerical ava- nic-debris avalanches, in Fisher R.V., and Smith G.A., Oehler, J.F., Labazuy, P., and Lenat, J.F., 2004, Recurrence lanches, in Kishino, Y., ed., Powders and grains: Lisse, eds., Sedimentation in volcanic settings: SEPM (Soci- of major fl ank landslides during the last 2-Ma-history Netherlands, Swets and Zeitlinger, p. 451–454. ety for Sedimentary Geology) Special Publication 45, of Reunion Island: Bulletin of Volcanology, v. 66, Stoopes, G.R., and Sheridan, M.F., 1992, Giant debris ava- p. 99–106. p. 585–598, doi: 10.1007/s00445-004-0341-2. lanches from the Colima volcanic complex, Mexico: Glicken, H., 1998, Rockslide-debris avalanche of May Oehler, J.F., van Wyk de Vries, B., and Labazuy, P., 2005, Implication for long-runout landslides (100 km) and haz- 18, 1980, Mount St. Helens Volcano, Washington: Landslides and spreading of oceanic hot-spot and arc ard assessment: Geology, v. 20, p. 299–302, doi: 10.1130/ Geological Society of Japan Bulletin, v. 49, no. 2/3, shield volcanoes on Low Strength Layers (LSLs): An 0091-7613(1992)020<0299:GDAFTC>2.3.CO;2. p. 55–106. analogue modelling approach: Journal of Volcanology Ui, T., 1983, Volcanic dry avalanche deposits: Identifi cation Glicken, H., Voight, B., and Janda, R.J., 1981, Rockslide- and Geothermal Research, v. 144, p. 169–189, doi: and comparison with non-volcanic debris stream depos- debris avalanche of May 18, 1980, Mount St. Helens 10.1016/j.jvolgeores.2004.11.023. its: Journal of Volcanology and Geothermal Research, volcano: IAVCEI Symposium on Arc Volcanism, Pollet, N., and Schneider, J.-L., 2004, Dynamic disintegra- v. 18, p. 135–150, doi: 10.1016/0377-0273(83)90006-9. Tokyo and Hakone: Tokyo, Terra Scientifi c Publishing tion processes accompanying transport of the Holocene van Wyk de Vries, B., and Borgia, A., 1996, The role of Co., p. 109–110. Flims sturzström (Swiss Alps): Earth and Planetary basement in volcano deformation, in McGuire, W.J., Gray, J.M.N.T., and Thorton, A.R., 2005, A theory for par- Science Letters, v. 221, p. 433–448, doi: 10.1016/ et al., eds., Volcano instability on the Earth and other ticle size segregation in shallow granular free-surface S0012-821X(04)00071-8. planets: Geological Society of London Special Publi- fl ows: Royal Society of London Proceedings, ser. A, Pouliquen, O., and Renaut, N., 1996, Onset of granular cation 110, p. 95–110. v. 461, no. 2056, p. 1447–1473. fl ows on an inclined rough surface: Dilatancy effects: van Wyk de Vries, B., and Francis, P.W., 1997, Catastrophic Guilbaud, M.N., 2000, Structural analysis of the emplace- Journal de Physique II, v. 6, p. 923–935, doi: 10.1051/ collapse at stratovolcanoes induced by gradual volcano ment of the Socompa debris avalanche [M.S. thesis]: jp2:1996220. spreading: Nature, v. 387, p. 387–390. Université Blaise Pascal, Clermont-Ferrand, France. Pudasaini, S.P., and Hutter, K., 2007, Avalanche dynamics: van Wyk de Vries B., and Merle, O., 1996, The effect of vol- Heim, A., 1932, Bergsturz und Menchenleben: Zurich, Vier- Dynamics of rapid fl ows of dense granular avalanches: canic constructs on rift fault patterns: Geology, v. 24, teljahrsschrift, v. 77, no. 20, 218 p. Netherlands, Springer, 602 p. p. 643–646, doi: 10.1130/0091-7613(1996)024<0643: Hornbaker, D.J., Albert, R., Albert, I., Barabasi, A.L., and Ramberg, H., 1981, Gravity, deformation and the Earth’s TEOVCO>2.3.CO;2. Schiffer, P., 1997, What keeps sandcastles standing: crust in theory, experiments and geological applications van Wyk de Vries, B., Self, S., Francis, P.W., and Keszthe- Nature, v. 387, p. 765, doi: 10.1038/42831. (second edition): London, Academic Press, 452 p. lyi, L., 2001, A gravitational spreading origin for the Howard, K., 1973, Avalanche mode of motion: Implications Reiche, P., 1937, The Toreva-block, a distinctive landslide Socompa debris avalanche: Journal of Volcanology from lunar examples: Science, v. 180, p. 1052–1055, type: Journal of Geology, v. 45, p. 538–548. and Geothermal Research, v. 105, p. 225–247, doi: doi: 10.1126/science.180.4090.1052. Reubi, O., and Hernández, J., 2000, Volcanic debris ava- 10.1016/S0377-0273(00)00252-3. Hungr, O., and Evans, S.G., 2004, Entrainment of debris in lanche of the upper Maronne valley (Cantal Volcano, van Wyk de Vries, B., Gilbaud, M-N., Self, S., and Druitt, rock avalanches: An analysis of a long runout mecha- France): Evidence for contrasted formation and trans- T., 2002, A brittle crust during the emplacement of the nism: Geological Society of America Bulletin, v. 116, port mechanisms: Journal of Volcanology and Geother- Socompa debris avalanche?: Nice, European Geosciences p. 1240–1252, doi: 10.1130/B25362.1. mal Research, v. 102, p. 271–286, doi: 10.1016/ Union general assembly, 1st, abs. EGU04-A-02057. Iverson, R.M., and Denlinger, R.P., 2001, Flow of variably S0377-0273(00)00191-8. Voight, B., 1981, Time scale for the fi rst moments of the fl uidized granular masses across three-dimensional Savage, S.B., and Lun, C.K.K., 1988, Particle size segrega- May 18 eruption, in Lipman, P.W., and Mullineaux, terrain 1, Coulomb mixture theory: Journal of Geo- tion in inclined chute fl ow of dry cohesionless granular D.R., eds., The 1980 eruptions of Mount St. Helens, physical Research, v. 106, no. B1, p. 537–552, doi: solids: Journal of Fluid Mechanics, v. 189, p. 311–335, Washington: U.S. Geological Survey Professional 10.1029/2000JB900329. doi: 10.1017/S002211208800103X. Paper 1250, p. 69–86. Kelfoun, K., and Druitt, T.H., 2005, Numerical modelling of Scheidegger, A.E., 1973, On the prediction of the reach and Voight, B., Glicken, H., Janda, R.J., and Douglass, P.M., the emplacement of Socompa rock avalanche, Chile: velocity of catastrophic landslides: Rock Mechanics, 1981, Catastrophic rockslide avalanche of May 18, in Journal of Geophysical Research, v. 110, no. B12, v. 5, p. 231–236, doi: 10.1007/BF01301796. Lipman, P.W., and Mullineaux, D.R., eds., The 1980 B12202, doi: 10.1029/2005JB003758. Schenk, P.M., and Bulmer, M.H., 1998, Origin of mountains eruptions of Mount St. Helens, Washington: U.S. Geo- Kelfoun, K., Druitt, T.H., van Wyk de Vries, B., and Guilbaud, on Io by thrust faulting and large-scale mass move- logical Survey Professional Paper 1250, p. 347–377. M-N., 2008, The topographic refl ection of the Socompa ments: Science, v. 279, p. 1514–1517, doi: 10.1126/ Voight, B., Janda, R.J., Glicken, H., and Douglass, P.M., debris avalanche: Bulletin of Volcanology (in press). science.279.5356.1514. 1983, Nature and mechanics of the Mount St. Helens Kilburn, C., and Sørensen, S.-A., 1998, Runout lengths Shaller, P.J., 1991, Analysis and implications of large Mar- rockslide avalanche of 18 May 1980: Geotechnique, of sturzstroms: The control of initial conditions tian and terrestrial landslides [Ph.D. thesis]: Pasadena, v. 33, p. 243–273. and of fragment dynamics: Journal of Geophysi- California Institute of Technology, 586 p. Wadge, G., Francis, P.W., and Ramirez, C.F., 1995, The cal Research, v. 103, no. B8, p. 17,877–17,884, doi: Shea, T., van Wyk de Vries, B., and Pilato, M., 2007, Socompa collapse and avalanche event: Journal of Vol- 10.1029/98JB01074. Emplacement mechanisms of contrasting debris ava- canology and Geothermal Research, v. 66, p. 309–336, Lagmay, A.M.F., van Wyk de Vries, B., Kerle, N., and Pyle, lanches at Volcán Mombacho (Nicaragua), provided by doi: 10.1016/0377-0273(94)00083-S. D.M., 2000, Volcano instability induced by strike-slip structural and facies analysis: Bulletin of Volcanology, faulting: Bulletin of Volcanology, v. 62, p. 331–346, doi: 10.1007/s00445–007–0177–7. doi: 10.1007/s004450000103. Shreve, R.L., 1966, Sherman landslide, Alaska: Science, v. 154, Legros, F., 2002, The mobility of long-runout landslides: p. 1639–1643, doi: 10.1126/science.154.3757.1639. Engineering Geology, v. 63, p. 301–331, doi: 10.1016/ Shreve, R.L., 1968, The Blackhawk landslide: Geological S0013-7952(01)00090-4. Society of America Special Paper 108, p. 1–47. Lucchitta, B.K., 1979, Landslides in Valles Marineris, Mars: Siebe, C., Komorowski, J.-C., and Sheridan, M.F., 1992, Journal of Geophysical Research, v. 84, p. 8097–8113, Morphology and emplacement of an unusual debris- doi: 10.1029/JB084iB14p08097. avalanche deposit at Jocotitlán volcano, central Mex- MANUSCRIPT RECEIVED 24 MAY 2007 McEwen, A., 1989, Mobility of large rock avalanches: Evi- ico: Bulletin of Volcanology, v. 54, p. 573–589, doi: REVISED MANUSCRIPT RECEIVED 06 APRIL 2008 dence from Valles Marineris, Mars: Geology, v. 17, 10.1007/BF00569941. MANUSCRIPT ACCEPTED 29 APRIL 2008

686 Geosphere, August 2008

Downloaded from http://pubs.geoscienceworld.org/gsa/geosphere/article-pdf/4/4/657/3336451/i1553-040X-4-4-657.pdf by guest on 30 September 2021