Sie sind auf Seite 1von 19












ARTICLE IN PRESS DTD 5 1 2 3 4 5 6 7 8 9 Engineering Geology
ARTICLE IN PRESS DTD 5 1 2 3 4 5 6 7 8 9 Engineering Geology

Engineering Geology xx (2005) xxx – xxx

1 2 3 4 5 6 7 8 9 Engineering Geology xx (2005) xxx – xxx

Soil liquefaction during the Arequipa Mw 8.4, June 23, 2001

earthquake, southern coastal Peru´

Franck A. Audemard M. a, T , Juan Carlos Go´mez ,

Hernando J. Tavera b , Nuris Orihuela G.



a FUNVISIS, Venezuelan Foundation for Seismological Research, Apdo. Postal 76.880, Caracas 1070-A, Venezuela

b Instituto Geofı´sico del Peru´, Centro Nacional de Datos Geofı´sicos, Lima, Peru´

Received 26 June 2003; received in revised form 19 August 2004; accepted 24 December 2004

10 Abstract

11 The Arequipa June 23, 2001, earthquake with a moment magnitude of Mw 8.4 struck southern Peru´, northern Chile and

12 western Bolivia. This shallow (29 km deep) interplate event, occurring in the coupled zone of the Nazca subduction next to the

13 southeast of the subducting Nazca ridge, triggered very localized but widely outspread soil liquefaction. Although sand blows

14 and lateral spreading of river banks and road bridge abutments were observed 390 km away from the epicenter in the southeast

15 direction (nearing the town of Tacna, close to the Chile border), liquefaction features were only observed in major river valleys

16 and delta and coastal plains in the meizoseismal area. This was strongly controlled by the aridity along the coastal strip of

17 Southern Peru´. From the sand blow distribution along the coastal area, a first relationship of isolated sand blow diameter versus

18 epicentral distance for a single event is ever proposed. The most significant outcome from this liquefaction field reconnaissance

19 is that energy propagation during the main June 23, 2001, event is further supported by the distribution and size of the isolated

20 sand blows in the meizoseismal area. The sand blows are larger to the southeast of the epicenter than its northwestern

21 equivalents. This can be stated in other words as well. The area affected by liquefaction to the northwest is less spread out than

22 to the southeast. Implications of these results in future paleo-liquefaction investigations for earthquake magnitude and epicentral

23 determinations are extremely important. In cases of highly asymmetrical distribution of liquefaction features such as this one,

24 where rupture propagation tends to be mono-directional, it can be reliably determined an epicentral distance (between

25 earthquake and liquefaction evidence) and an earthquake magnitude only if the largest sand blow is found. Therefore,

26 magnitude estimation using this uneven liquefaction occurrence will surely lead to underrating if only the shortest side of the

27 meizoseismal area is unluckily studied, which can eventually be the only part exhibiting liquefaction evidence, depending on

28 the earthquake location and the distribution of liquefaction-prone environments.

29 D 2005 Published by Elsevier B.V.

30 Keywords: Liquefaction; Sand blows; Asymmetric energy distribution; Arequipa 2001 earthquake; Peru´

31 T Corresponding author. Fax: +58 2 257 99 77. E-mail addresses: (F.A. Audemard M.) 8 (J.C. Go´mez)8 (H.J. Tavera)8 (N. Orihuela G.).

0013-7952/$ - see front matter D 2005 Published by Elsevier B.V.




F.A. Audemard M. et al. / Engineering Geology xx (2005) xxx–xxx

32 1. Introduction

Peru´ vian coast at Atico, between Chala and Ocon˜a


( Fig. 1 ). After IGP, the epicenter coordinates are


33 Southern Peru´ was hit by a shallow Mw 8.4

16.20 8 latitude and 73.75 8 longitude ( Tavera, 78

34 (NEIC) earthquake on June 23, 2001. An earthquake

2002b ). Instead, NEIC reports the epicenter at 79

35 of such characteristics is a very likely candidate to

16.15 8 latitude and 73.40 8 longitude, while 80

36 have triggered induced effects. This paper discusses

Harvard CMT solution (energy centroid) places the


37 the results from a field survey in search of these

epicenter at 17.28 8 latitude and 72.71 8 longitude


38 effects that was initiated 12 days after the Arequipa

( Fig. 1 ). The main shock took place at the coupled


39 earthquake, starting in Lima and heading southeast

zone of the Nazca subduction under South America


40 along the coast between Pisco and Tacna (Peru´ vian

( Tavera, 2002c ), at a shallow depth of 29 km ( Tavera,


41 border-town with Chile). The survey target was

2002b ). This resulted in extensive damage along the


42 defined based on a preliminary intensity map raised

southern Peru´ (affecting the town of Ocon˜a, Camana´,


43 by Instituto Geofı´sico del Peru´ (IGP) that clearly

Mollendo, Arequipa, Moquegua and Tacna) and 88

44 placed the damage area southeast of Pisco. This

northern Chile coasts and as far inland as La Paz –


45 survey was only launched then, only when the road

Bolivia – ( Tavera, 2002a,b ). This inter-plate contact


46 network was partly functioning, although the Pan-

zone is the longest and one of the most active plate


47 American road running along the coast was inter-

boundaries worldwide. The main event was followed


48 rupted several times in occasion of almost all the

in the next 2 weeks by three other rather large


49 larger aftershocks during fieldwork. No air reconnais-

earthquakes on June 25 (Mw 6.8; NEIC), July 05 94

50 sance was carried out because choppers were attend-

(Mw 6.6; NEIC) and July 07 (Mw 7.5; NEIC); all


51 ing first priority tasks related to the emergency. The

located east to southeast of the main shock ( Tavera,


52 teamwork comprising personnel from IGP and FUN-

2002c ). Within the first 24 h, a total of 134 ML z 3.0


53 VISIS focused attention on all type of mass wasting

aftershocks were recorded, being all located southeast


54 and particularly on liquefaction, to whose description

of the June 23 main event ( Tavera, 2002b ). All these


55 this paper is devoted. Special attention was paid to

abovementioned seismological aspects clearly point 100

56 sand blows. No geotechnical investigations, such as

out that the energy directivity and rupture propagation


57 SPT or CPT, could be carried out due to logistic

history were SE-directed. In supplement to this, the


58 limitations and unavailable funding.

waveforms, their amplitudes and periods support this


59 In this paper, we have semi-quantitatively evaluated

and allow estimating that the energy was released 104

60 the liquefaction distribution of isolated sand blows

towards the southeast (77 8 120 8S; Tavera, 2002a ). 105

61 triggered by this earthquake, which has showed a very

Furthermore, the main shock itself comprises three


62 clear trend as we shall discuss herein, in comparison

sub-events separated by 5–6 and 36–40 s to the


63 with the seismological data of this Arequipa 2001

southeast of the first one ( Tavera, 2002a ). The third


64 event. Strong similarities in the rupture process are

and last of these three sub-events composing the main


65 revealed by, or deduced from, both – geologic and

shock lied right below the Camana´ region ( Tavera,


66 seismologic – approaches. Finally, the outcome of this

2002a ), which was the hub of the area affected by


67 evaluation is interpreted in terms of, and is extrapo-

tsunami waves at least as high as 6 m ( Carpio et al.,


68 lated to, its application to the past record of earth-

2002; Jaffe et al., 2003 ). It is very likely that the


69 quake-induced liquefaction and the assessment of past

whole rupture process (mono-directionality, rupture 114

70 earthquake magnitudes and epicenter locations from

nucleation, rupture progression) on the coupled zone


71 liquefaction features.

of the Nazca subduction is highly controlled by the


presence of the subducting Nazca ridge at the north-


western tip of the Arequipa 2001 earthquake rupture.


72 2. The earthquake and its aftershocks

The focal mechanism solution of the main earth-


quake images pure thrust slip along a NNW–SSE 120

73 The Arequipa earthquake struck the Southern Peru´

striking, 21 8 E-dipping fault plane ( Tavera, 2002c ; 121

74 coastal region with a moment magnitude (Mw) of 8.4

Fig. 1 ). This solution depicts well the rather shallow


75 at 20:33:13 on June 23, 2001. Its epicenter lied at the

dip (16–20 8 N) of the subduction slab on the south-


F.A. Audemard M. et al. / Engineering Geology xx (2005) xxx–xxx




Ocoña R.
Caraveli R.
San Juan
06 / 23 / 2001
Colca River
La Curva
Mw 8.4
Vitor River
SAND BLOW (diameter in m)
07 / 07 / 2001
Yauca R.
Acari River
La Yarada
Tambo R.
Grande River
Camana R.
Locumba River
Sihuas R.
Osmore R.
Sama River



Mw 7.5



Fig. 1. Map of the southern coast of Peru´, showing main rivers, surveyed road and main settlements. Relative location is shown in inset map. It also displays the focal mechanism

solutions from IGP for the main event and its main aftershock, as well as the different main earthquake epicenters reported by different agencies. Fin ally, it also shows the distribution of all reported liquefaction evidence split by type (sand blow, sand dike or lateral spread).




F.A. Audemard M. et al. / Engineering Geology xx (2005) xxx–xxx

124 eastern edge of the subducting Nazca ridge; coincid-

125 ing with where flat subduction in the northwest

126 changes to normal subduction in the southeast ( Jordan

127 et al., 1983; Mercier et al., 1992; Gutscher et al.,

128 2000 ). But the orientation of the fault planes in the

129 focal mechanism solution keeps an angle of some 30 8

130 with respect to the trench orientation ( Fig. 1 ),

131 although they are almost perfectly normal to the

132 GPS-derived convergence vector for the Nazca plate

133 in this region ( Kreemer et al., 2003 ). This excludes

134 any strain partitioning along this portion of the plate

135 boundary.

136 3. Investigation of liquefaction features in the

137 meizoseismal area

138 The Southern Peru´ coast is a very particular

139 landscape worldwide. It is much like a lunar land-

140 scape deprived of any vegetation cover due to

141 extreme aridity. It also lacks of soils, but does have

142 some wind-transported sands and volcanic ashes.

143 This arid strip can eventually extend as far inland as

144 a 100 km. Consequently, the search for liquefaction

145 features was narrowed down to very few and

146 localized Holocene sedimentary environments prone

147 to liquefy during strong ground shaking, such as

148 riverbeds and terraces, and alluvial, delta and coastal

149 plains. In fact, very few rivers and streams cross this

150 arid zone and pour into the Pacific. The most

151 important rivers in the study area from northwest

152 to southeast are rare: Acari, Yauca, Ocon˜a, Camana´,

153 Sihuas, Tambo, Osmore, Locumba and Sama ( Fig.

154 1 ). These rivers display running water, mostly fed

155 from far upstream into the Andes highlands. Most of

156 these river valleys exhibited both sand blows and

157 venting fractures, related or not to lateral spreading.

158 The two largest delta plains of the Tambo and

159 Camana´ Rivers had the most widespread liquefaction

160 distribution ( Fig. 1 ). There is no report of liquefac-

161 tion in the Ocon˜a valley by us because the crossing

162 was made at night. It is worth to mention that the

163 field survey was a natural transect along rupture

164 strike because main roads, coast and the subduction

165 trends are all parallel ( Fig. 1 ). In this sense, the

166 reported liquefaction distribution can bring additional

167 insights on the earthquake rupture process, as shown

168 in this case study.

Regardless of the causative process (hydraulic 169 fracturing or associated with lateral spreading), 170 utilized conduits (pre-existing or newly formed 171 cracks, root casts, burrows, rheologic boundaries or 172 any other pre-existing seal weakness) and liquefied 173 feature geometries (aligned or isolated blows, sand 174

dikes), all sand-venting features were investigated, but 175 special attention was paid to isolated sand blows 176

though. This venting feature, when perfectly isolated 177

on the ground, should be the most reliable measure of 178

the pressure exerted by hydraulic fracturing. This in 179

turn must keep a tight relationship with the earthquake 180

ground shaking through the cyclic loading of shear 181

waves when passing through well-saturated, near- 182

surface, cohesionless, granular sediments. In other 183

words, it should be a better gauge of the earthquake 184

energy than all other sand-venting features reported in 185

the literature in the occurrence of moderate-to-large 186

earthquakes. Instead, Obermeier (1996) and Oberme- 187

ier and Pond (1999) have proposed the use of sand 188

dike width as a tool for characterizing the triggering 189

paleo-earthquake in terms of magnitude and epicentral 190

location. We believe this approach can be of limited 191

applicability since sand dike width can be strongly 192

conditioned by the occurrence of lateral spreading, 193

which in turn is tightly linked to topographic 194

conditions, and not mainly to energy release. Charac- 195

terization of past topographies over large rough relief 196

areas can be a very difficult task. It may be that this 197

approach is very reliable in areas such as the 198

Mississippi valley – where the approach has been 199

developed – because most topographic irregularities 200

(read free faces or talus) in such a case are introduced 201

by the depth of river incision, which in very flat areas 202

must tend to be rather constant throughout. In that 203

respect, the use of isolated sand blows could have a 204

more general applicability in characterizing past 205

earthquakes. We feel that sand blows – either in the 206

geologic record or on ground surface in association 207

with contemporary earthquakes – can be as easily 208

characterized geometrically as sand dikes, by measur- 209

ing their base diameter and maximum thickness at the 210

blow mouth instead, although this analysis is only a 211

good semi-quantitative approximation, since both 212 parameters are strongly dependent on several intrinsic 213 factors of the liquefied material. We believe though 214 that the two most conditioning factors among many 215 others are (1) the percent per volume of sand in the 216


F.A. Audemard M. et al. / Engineering Geology xx (2005) xxx–xxx


217 mixture vented to the surface (or the amount of water

218 spouted out, which is proportional to the original

219 water saturation of the liquefied sand layer) and (2)

220 the vented sand grain size. These factors are inter-

221 connected. For instance, the larger the grains the more

222 water is contained in the pores, for any specific

223 sorting. These two factors combined control the cone

224 slopes of isolated sand blows when resting on flat

225 ground, as observed by Beltra´n and De Santis (1990) ,

226 and Audemard and De Santis (1991) . If the ground

227 surface has some irregularities (e.g., ditches, grooves,

228 scours, furrows) or slope irrespective of how gentle it

229 is inclined, the sand blow circular base is then

230 distorted. In such case, depending on whatever

231 controls the blow shape, either a measure of the

232 minimum diameter must be taken or the shortest and

233 longest base dimensions must be averaged. Other

234 factors intrinsic to the sand spelled out may also exert

235 some control on the blow shape such as grain sorting,

236 roundness and shape. However, very little grain-size

237 analyses have been carried out on spouted sands after

238 individual contemporary earthquakes. At most, some

239 pictures, and occasionally few measures, of sand

240 blows are typically reported in the literature. This is

241 the case of the present study, but this should change in

242 future field reconnaissance surveys if any reliable

243 quantitative analyses are intended. A very rare

244 exception to this is the survey made by Tuttle et al.

245 (2002) after the Bhuj, January 26, 2001, earthquake

246 (Gujarat, India) that gives figures of both diameter and

247 thickness of sand blows. For the Arequipa 2001

248 earthquake, we have also intuitively focused our

249 attention on the isolated sand blow diameters, as well

250 as on the visual evaluation of grain size of the

251 liquefied sands.

252 Additional factors, not only intrinsic to the

253 liquefiable sands, must actually be taken into account

254 to undertake any robust quantitative analysis – which

255 is not the aim of this rather rough first semi-

256 quantitative approximation – of the geometry of sand

257 blows as a reliable measure of released energy

258 through earthquakes. Depth of water table, as well

259 as depth of liquefaction-prone sand bodies, must be

260 well constrained, which requires specific geotechnical

261 studies and/or instrumentation (piezometer, SPT and

262 CPT, among others). Particular attention must be also

263 devoted to the sealing cap, as to thickness and pre-

264 existing likely weakness zones or features, such as

rotten roots, crab burrows, cracks, among several 265 others. In our study, two very strong limitations 266 precluded any quantitative analysis, thus leading us 267 to propose this semi-quantitative approach or hypoth- 268

esis instead. On one hand, in accessible areas such as 269 the coastal strip of Peru´, due to rather smooth relief 270 and fairly good road network, running water is only 271 limited to the larger river courses, as mentioned 272

earlier, and superficial water is definitely absent 273

elsewhere (water table can be typically be in excess 274

of 200 m, as pointed out by Wartman et al., 2003 ), 275

which substantially narrows down areas prone to 276

surface liquefaction, either triggered by earthquake, 277

artesian waters or any other process. This also reduces 278

the number of observational data (refer to Wartman et 279

al., 2003 , who reutilized most of the data presented 280

herein). On the other hand, far inland and away from 281

the coastal strip, where water is abundant, liquefac- 282

tion-prone areas are smaller, narrower and bounded to 283

narrow, sinuous river courses incised in a very steep, 284

inaccessible relief, inside the rough relief of the high 285

Andes (refer to Fig. 3-1 in Wartman et al., 2003 ). 286

Next, we shall describe all the documented 287

evidence of liquefaction observed from NW to SE, 288

adding further details to the original descriptions from 289

Audemard et al. (2001, 2002) and Gomez et al. 290

(2002) . This information was essentially gathered 291

along the South Pan-American road and other minor 292

coastal roads (for survey coverage refer to Fig. 1 ). 293

First, we shall deal with liquefaction features strictly 294

speaking, and later with occurrence of lateral spread- 295

ing, which is a combination of gravity-driven lateral 296

translation of almost flat-lying geologic units induced 297

by liquefaction at shallow depth. Reported lateral 298

spreading may have or may not have been accom- 299

panied by spouted sands to the surface. 300

4. Superficial soil liquefaction evidence


As foreseen prior to the field reconnaissance, 302

provided the main event intensity distribution and 303

the extreme aridity of the affected area, the occurrence 304

of liquefaction was only bounded to those few alluvial 305 valleys with permanent running water and active delta 306 plains. This includes the valleys of the Yauca, Ocon˜a, 307 Camana´, Tambo, Osmore, Locumba and Sama Rivers, 308 from northwest to southeast ( Fig. 1 ). Major relevant 309

F.A. Audemard M. et al. / Engineering Geology xx (2005) xxx–xxx6


Table 1


Surface evidence of liquefaction along the coastal strip of southern Peru´



Distribution and characterization of surface liquefaction features




Map label

Feature description

Depth to water table

Other relevant comments


( Fig. 1)

Sedimentary environment

Sand grain size

Size of largest


and liquefaction features

sand blow


Yauca River


–Sand boils on small channel

Very fine to fine

1.50 m in diameter,

Visible and running

First pile of Yauca bridge from the western abutment sunk due to liquefaction in a few tens of centimeters, sagging bridge deck and south banister


bars, in side riverbed

but typically 0.40 m

only 50 cm below

–Isolated sand blows with

sand blow base

almost circular base on flat




–Frequently darker and finer

material (mica rich) expelled,

probably when water pressure



–Sand-venting dikes


( b1 m long)

–Some sand blows lying on

open NE–SW trending ground

cracks, paralleling the channel


bar edges (lateral spread)


Ocon˜a River

Not visited

Camana´ delta


–Widespread liquefaction in

Very fine to

1.60–2.00 m in

Features were only

–Rock-armored flood-control earth embankment of its left riverbank was damaged at two places. –40-cm-across grayish sand blows were reported by a later party, resting on the tsunami deposits of the main event, attesting to a second liquefaction episode during a large aftershock; probably the Mw 7.5 event on July 07th, 2001


the riverbed and its active


diameter, being

some 50 to 60 cm

alluvial and delta plains

circular on flat

above visible running


–Plentiful light colored



liquefied sands resting on

pebbly-to-cobbly channel

bars, probably supplied

by well-oxidized sandy

channel bars

–Occasionally, sand-volcano

cones aligned along open

cracks, individually not longer

than 4 m and displaying an en


echelon array

–In the coastal plain of the

Camana´ delta, a 30-m-long

venting fracture, paralleling

both plowing furrows and 5 m

away of the river protection

embankment, spouted grayish sands in a cornfield, coming from recent anoxic organic-


rich delta deposits


F.A. Audemard M. et al. / Engineering Geology xx (2005) xxx–xxx



Between Mollendo


–Very scarce superficial

Fine to medium

Only 0.20–0.30 m

Shallow, since barrier


and Mejı´a

evidence of liquefaction,

along major axis


is b +3–4 m msl


except for small isolated sand

volcanoes along cracks on

top of the present-day sand

barrier, at b Urbanizacio´ n

Arizona Q


Golden Playa Discoteca


–Both vent fractures and

Very fine to fine.

Boils could be up

Spouted water left

Gravel road entrance fell

(between Mejı´a and

b apparently Q isolated sand

Significant gravel

to 0.70 m in

whitish stains; probably

into a few-meter-wide,

El Conto)


fraction (Contamination


brackish. Water table at

road-parallel ditch due to


–Blows aligned on top of

during upward

shallow depth ( b3–4 m)

3 small lateral spreads.

almost imperceptible fissures


No sand venting along

in gravel road carpet, but next

most of the largest cracks


to wide open cracks

–Grayish-colored sands

in macadam carpet

suggest a provenance from

anoxic bodies. Also dark-

colored spouting through


small cracks

–Capping of darker and finer

mica-rich fractions at the

isolated sand cone mouths

(last deposition due to high




Soccer field at La Curva


–Sand venting along ground

Fine to medium,

Not reliable because

Very shallow (b 2m deep)

Significant amount of spouted

north entrance

cracks, across a soccer field

light-colored sands

volcanoes were aligned

water table as seen in

water that almost crossed the


sitting on the Tambo river

atop an open ground

irrigation ditch nearby

entire soccer field length,

alluvial plain, due to lateral


because of a very gentle



ground-surface inclination

–Most of water and sand

vented by a single fracture,

whose cone was 2.5 m long

and about a 1 m wide, in

combination with a 8-m-long


volcanoes alignment


Tambo riverbed at the


–Numerous rather small

Medium, with a


m across but very

Largest sand blows were

The 6-m-high and

El Fraile bridge

( b 0.20 m across), mostly

significant fraction



than a few tens of

two-lane-wide Tambo-bridge


isolated, sand blows on coarse

of gravels at least


above running water

earth embankment exhibited

channel bars

in the largest blows


–Grayish-colored vented

longitudinal cracks. Also


axial cracking in the west


bridge abutment


Osmore delta (Ilo)


–Liquefaction in the Osmore

Very fine to fine.


to 1.00 m across

Base of blows only 0.50 m

Also, 1.35-m-long,


riverbed, near its delta mouth,

Light brown colored.

above running water



north of the town of Ilo.

venting fractures

–Liquefied sands rested on

pebbled-to-cobbled channel



(continued on next page)


F.A. Audemard M. et al. / Engineering Geology xx (2005) xxx–xxx8


t1.26 Distribution and characterization of surface liquefaction features

Table 1 ( continued )

t1.27 Locality

Map label

Feature description

Depth to water table

Other relevant comments


( Fig. 1)

Sedimentary environment

Sand grain size

Size of largest


and liquefaction features

sand blow


Ilo-La Yarada coast road


–Near the Locumba River,

Very shallow water table


longitudinal cracks in paddy–

field earth dams, suggesting

that liquefaction had also

happened at depth in this

delta plain

–No evidence of sand



–Artificially saturated




Sama River


–Isolated and aligned sand

Fine to medium

0.50 m in diameter

Shallow water table under

Two large craters at the retaining wall foot of the south bridge abutment, while the earth fill behind it cracked open and settled. Grain-size distribution clearly attests to a progressively dying-out water pressure

(Los Ban˜ os bridge)

blows, vent fractures and

Grain-size of topping

features ( b 0.50 m deep)


lateral spreads of modest

could vary from silt


dimensions in the riverbed

to very coarse sand

–Sand blows, atop bars in the

riverbed, were numerous but

not as frequent as in others


surveyed rivers

–Outward grain coarsening of

the volcano cones, at least as a

capping film, from silts and

black micas at the cone mouth

to coarse to very coarse sands


at the rim

–Open cracks paralleling

channel bar edges. Some of

these also trended obliquely

but their orientation was

controlled by sand extrusion

and differential settlement

produced by closely located





F.A. Audemard M. et al. / Engineering Geology xx (2005) xxx–xxx


310 observations as to the liquefaction evidence at the

311 different localities are summarized in Table 1 ,

312 following that same order.

313 In a general way in this study, we report surface

314 evidence of liquefaction from the Yauca riverbed (km

315 598 of the South Pan-American road; station 1 in Fig.

316 1 ) to as far southeast as the Sama River, along the

317 coastal strip of southern Peru´, near the Chilean

318 border. Instead, The Acari River, located only 15

319 km to the northwest of the Yauca River (station 0 in

320 Fig. 1 ), did not exhibit any visible liquefaction

321 feature, which sharply defines the northern extent

322 of the area exhibiting surface evidence of liquefac-

323 tion. Generally speaking, most of the reported

324 features were lying either inside the few riverbeds

325 crossed in the study area, atop channel bars ( Fig. 2 )

326 or in their alluvial or delta plains. As mentioned

327 earlier, there is no surface evidence of occurrence of

328 liquefaction out of these particular sedimentary

329 environments because the water table is usually too

330 deep (over a 100 m). Bars in the riverbeds were

331 typically few tens of centimeters above water level,

332 and most commonly in the order of 0.50 m, implying

333 that sand feeders could lie very shallow and were

334 well saturated. From the sedimentological viewpoint,

335 the dominant grain-size of liquefied material ranged

336 between very fine to medium sand (refer to Table 1 ).

337 Most sand blows exhibit good grain sorting.

However, the most densely spaced surface expres- 338 sion of liquefaction has been reported at the delta 339 plains of the both Camana´ and Tambo rivers. For 340 instance, close to the river mouth of the Camana´ 341

River, in its delta plain, a 30-m-long venting fracture 342 was seen ( Fig. 3 ). The spouted sand was grayish in 343 color, implying that the organic content was high, 344 significantly differing from the vented material 345

observed in the riverbed which was very light 346

colored. This also supports that the spouted sand 347

belonged to the recent anoxic organic-rich delta 348

deposits of this river. The inundation line of the 349

tsunami that affected the Camana´ region, with run- 350

ups of as much as 8.2 m high (Jaffe et al., 2003 ) and 351

surely above 6 m (Carpio et al., 2002 ), was very few 352

tens of meters away, which attests that this evidence 353

was just very luckily preserved. This line is over 500 354

m inland from the coast. The parallel orientation of 355

the venting fracture to that of the embankment 356

suggests that lateral spreading towards the riverbed 357

(free face effect) of the river protection embankment, 358

together with the underlying left riverbank as a unit, 359

may have played a major role in its formation.

The Tambo delta plain and the adjacent coastal 361


stretch display the most outspread evidence of 362

liquefaction of all. This is a straightforward conse- 363

quence of the size of the liquefaction-prone environ- 364

ment. The Tambo River delta is the largest of all in 365

ment. The Tambo River delta is the largest of all in 365 Fig. 2. Aligned sand

Fig. 2. Aligned sand volcanoes on very coarse channel bars in the Camana´ riverbed, close to its mouth (scale=1m). These bars lie only half a meter above running water.



F.A. Audemard M. et al. / Engineering Geology xx (2005) xxx–xxx

Audemard M. et al. / Engineering Geology xx (2005) xxx–xxx Fig. 3. A 30-m-long crack in

Fig. 3. A 30-m-long crack in a cornfield in the Camana´ floodplains vented out grayish sands, probably fed by an anoxic sand body of the

Camana´ overbank sequence. This crack paralleled 5 m away the flood-control embankment, implying that the riverbank and embankment as a

whole underwent lateral spreading.

366 this southern portion of the Peru´ coast. It extends

unequivocal evidence proving their concurrent occur- 391

367 between Mejı´a and Corio, for almost 25 km in length

rence ( Fig. 4 ). Sand blows or vented sand was sitting 392

368 along the seashore (see Fig. 1 for relative location).

on ground fissures or cracks, regardless of their 393

369 However, Holocene active coastal plains (including

opening width. Next, we shall describe the most 394

370 sand barriers, mud flats and salt flats) extend as far

relevant localities were this phenomenon was 395

371 northwest as south of Mollendo, for an additional strip



372 length of 10 km; totaling a 35-km coastal stretch of

373 Holocene and recent geologic environments of high

5.1. Camana´ River flood-control earth embankment


374 susceptibility to liquefaction. Besides, the Tambo

375 River alluvial plain stretches inland for some 25 km,

376 between the villages of Punta de Bombo´ n on the coast

377 (near La Curva) and El Fiscal (see Fig. 1 for relative

378 location); the latter being located along the South Pan-

379 American road.

380 5. Lateral spreading

381 This phenomenon was the most widespread and

382 common to the visited meizoseismal area of the June

383 23, 2001, earthquake. It both affected the natural as

384 well as the constructed environment, but road

385 embankments close to irrigation ditches happened to

386 be the most affected feature. It also obeyed the same

387 distribution pattern of superficial evidence of soil

388 liquefaction. However, not all surveyed lateral spreads

389 exhibited evidence of liquefaction on ground surface.

390 It mostly was otherwise. But, few places exhibited the

The left Camana´ River embankment in its delta 398

plain nearing its mouth suffered two strong sagging: 399

(1) where the river showed sandy channel bars in its 400

bed, both embankment shoulders were longitudinally 401

cracked for several tens of meter in length; (2) in a 402

straight section of the river course, the embankment, 403

protected by rock blocks, moved sideward into the 404

river as a rotational slide, inducing top sagging and a 405

semi-circular head scar. On the opposite side, a 1-m- 406

long open crack vented sand and water at the foot of 407

the embankment. Water was still ponded on July 06 408

and had run off for over 15 m on a very gentle slope. 409

These are supporting evidence that embankment 410

failure resulted from lateral spreading in association 411

with shallow liquefaction. Thickness (height) of the 412 embankment forced water escape to dart such a huge 413 compacted seal. The light color of the vented sand 414 suggests that it came from a well-ventilated sand bed. 415 Therefore, the sand was provided from the riverbed 416


F.A. Audemard M. et al. / Engineering Geology xx (2005) xxx–xxx


M. et al. / Engineering Geology xx (2005) xxx–xxx 11 Fig. 4. Aligned sand volcanoes between

Fig. 4. Aligned sand volcanoes between the edge of a soccer field on the left and the toe of the La Curva–El Arenal road embankment to the

right. Spouted water ran toward the viewer (station 7 on Fig. 1).

417 but from the floodplains, although the crack and the

on mudflats located behind the abovementioned sand


418 spouted sand and water were resting on the overbank

barrier, ground fissures paralleling shallow irrigation


419 deposits.

ditches were also reported. These are unequivocal 426

evidence of lateral spreading induced by very shallow


420 5.2. Coast road between Mollendo and La Curva

liquefaction in both conditions, although no venting


was reported. In particular, the sand barrier cracking


421 Between Mollendo and Mejı´a, shore-parallel tens-

(Fig. 5 ) should be attributed to sea-front relaxation,


422 of-meter-long fissures were observed atop of an about

while it was induced by lack of lateral confinement by


423 400-m-wide sand barrier near Mollendo. In crop fields

ditches in the cultivated land. In fact, the supporting


in the cultivated land. In fact, the supporting 432 Fig. 5. Axial crack along the sand

Fig. 5. Axial crack along the sand barrier, close to Mollendo (station 3 in Fig. 1), due to sea-front relaxation. No surface evidence of liquefaction was found here.



F.A. Audemard M. et al. / Engineering Geology xx (2005) xxx–xxx

433 evidence to this was found closer to Mejı´a, at

434 b Urbanizacio´ n Arizona, Q where small isolated sand

435 volcanoes were sitting on cracks opened in the sand

436 barrier deposits.

437 The road embankment sitting on the coastal-delta

438 plains of the Tambo River actually happened to be the

439 most damage feature of all by lateral spreading along

440 this road stretch. The embankment, built on the

441 mudflats, was about a 1.5 m high. Dozens of places

442 exhibited longitudinal cracks along the embankment

443 shoulders. Some of these cracks could be of the order

444 of 100 m in length. Few of these cracks could actually

445 cut the asphalt carpet, which had very little vertical

446 displacement. This would imply that the asphalt

447 (acting as a highly cohesion seal) could not be torn

448 apart. When road embankment lateral spreading was

449 more severe, one road lane (half way) could exhibit

450 asphalt carpet cracking. However, road embankment

451 suffered most extensively where road ran parallel to

452 irrigation ditches directly dug in the ground (sector El

453 Boquero´ n). Lateral spreading was here induced by the

454 void effect of the near ditch, in combination with the

455 liquefaction of shallow saturated cohesionless sands

456 of the Tambo plain. We could only report once that the

457 embankment collapsed along a 100-m-long narrow

458 graben-like feature at the El Boquero´ n, between El

459 Conto and La Curva (station 6 in Fig. 1 ). Here, the

460 road embankment was longitudinally trapped between

461 a concrete channel and a ground-dug ditch along both

462 sides ( Fig. 6 ). The embankment spread laterally due to

463 shallow liquefaction, thus compressing the ditch and

464 stretching the opposite shoulder of the embankment

( Fig. 6 ). Local farmers from Boquero´ n accounted to 465 vertical ejection of a mixture of water and sand in 466

their crop fields during the main earthquake.

Lateral spreading was very well imaged at the 468 water pump station located at the entrance of the 469 Golden Playa Discoteca dirt road, near the Wild Bird 470 Reservation of Mejı´a (station 5 in Fig. 1 ). Three small 471 coalescent lateral spreads coevally took place at this 472


b T Q road intersection ( Fig. 7 ). The dirt road bridged 473

over a ground-dug few-meter-wide ditch. During the 474

earthquake, two opposing smaller lateral spreads 475

moved sideward into the wide water-full ditch, which 476

in turn triggered a third one that moved away from the 477

main asphalt road. Fig. 7 illustrates this in detail. 478

5.3. Road along the Tambo River


Between the villages of La Curva and El Arenal, 480

three different road embankment sections underwent 481

sliding normal to the road, in association with lateral 482

spreading. All three failures occurred where the road 483

ran parallel to a shallow irrigation ditch. In all three 484

cases, over half of the asphalt carpet was affected. 485

Nevertheless the most prominent case occurred at the 486

northern entrance of La Curva (station 7 in Fig. 1 ). It 487

was also the largest of all single lateral spreads 488

reported during this survey. It was as big as a 489

traditional soccer field, measuring some 60 m wide 490

and 100 m long. It actually affected a soccer field 491

almost entirely and the 2-lane-road embankment 492

running along its southeast side. Both the embank- 493

ment and the soccer field sagged few tens of 494



Fig. 6. Schematic profile depicting the lateral spreading of the road embankment at Boquero´ n (between El Conto and La Curva; station 6 in Fig. 1). The embankment shoulder exhibited a longitudinal graben-like depression.


F.A. Audemard M. et al. / Engineering Geology xx (2005) xxx–xxx WATER- PUMP STATION MARSH
F.A. Audemard M. et al. / Engineering Geology xx (2005) xxx–xxx


Fig. 7. Schematic plan view of the three coalescent lateral spreads that affected the b TQ road intersection of the entrance to the Golden Playa

Discoteca (station 5 in Fig. 1).

495 centimeters, shared among several sub-parallel arcu-

whose cone was 2.5 m long and close to a 1 m wide,


496 ate cracks ( Figs. 4 and 8 ). Sagging was produced by

in combination with a set of aligned volcanoes


497 differential settlement after water and fine sands

stretching over some 8 m ( Fig. 4 ). These features 507

498 were spouted to the surface through a venting

were bordering the soccer field just at the foot of the


499 fracture and a set of aligned volcano mouths, along

road embankment, where the latter sank ( Fig. 4 ). 509

500 the edge of the embankment toe, between the soccer

This large mass motion on a liquefied layer was set


501 field and road embankment (Fig. 4 ). Amount of

by a small irrigation ditch running along the road 511

502 water was that much that ran off for over half the

embankment, on the opposite side to the soccer field.


503 length of the soccer field. The mixture of water and

However, the most severe damage to road embank-


504 sand was essentially vented by a single fracture,

ments occurred where a shallow irrigation ditch ran


ments occurred where a shallow irrigation ditch ran 514 Fig. 8. Arcuate cracks across the soccer

Fig. 8. Arcuate cracks across the soccer field of La Curva (station 7 in Fig. 1). It can be clearly distinguished the spouted light-colored sands from the darker ground. In left lower corner, sands from the aligned sand blows shown in Fig. 12 are also visible.



F.A. Audemard M. et al. / Engineering Geology xx (2005) xxx–xxx

515 next to the road, such as between La Curva and El

516 Arenal, between Cocachacra and El Fiscal. Embank-

517 ment locally cracked and spread laterally due to the

518 free face effect introduced by the shallow irrigation

519 ditch running parallel at the toe of the embankment

520 (similar to image in Fig. 6 ).

521 5.4. The Pan-American road along the Osmore valley

522 Far inland, the road embankment of the South Pan-

523 American road when crossing the Osmore River plain,

524 near Montalvo (south of Moquegua), was longitudi-

525 nally cracked by outward-directed lateral spreading

526 over tens of meters in length, implying that ground

527 liquefaction occurred beneath the embankment.

528 6. Discussion

529 From this field survey observations, the distribu-

530 tion and size of superficial liquefaction evidence

531 show their maximum values at Camana´, progres-

532 sively diminishing outward both in frequency and in

533 amount of spouted sand and water, both to the

534 northwest and the southeast ( Fig. 1 ). Although no

535 field reconnaissance could be performed upstream

536 along rivers due to logistic limitations, this survey is

537 a perfectly oriented profile along strike of the

538 causative fault (Nazca subducting slab under the

539 South America plate that runs parallel to both the

540 coast and surveyed Pan-American road) of the main

541 earthquake and the aftershock sequence trend. This

542 allows cross-plotting the size of the largest spotted,

543 isolated sand blows with epicentral distance. Argu-

544 ments on the selection of this liquefaction feature

with respect to others have been previously dis- 545 cussed in Section 3. However, some complications in 546 this correlation have arisen because of two different 547 types of uncertainties: one relates to which is the most 548 appropriate epicenter location for the calculation of 549

the epicentral distance; and the other to the reliability 550 that all reported liquefaction features actually were 551 triggered by the main event. The second concern is 552

due to the fact that the Mollendo–Tacna survey was 553

carried out in the same day after the largest (Mw 7.5) 554

b aftershock Q that occurred before dawn (4:38 am local 555

time) on July 07, 2001.

Regarding the first aspect, every agency, either 557


national or international, obtains different epicentral 558

solutions for a given earthquake. These differences 559

result from the used seismic datasets and the type of 560

epicenter determination. Since availability of these 561

solutions is very variable, we decided to test the 562

sensibility of this variable (epicentral determination) 563

in the calculations of epicentral distance (between 564

chosen epicenter and reported liquefaction evidence) 565

for three of the published epicenters (Table 2 ). We 566

selected the IGP, NEIC and Harvard solutions. The 567

USGS epicenter was discarded because it is located 568

between the two extreme solutions from IGP and 569

NEIC, and not substantially varying from those. 570

However, assuming that all reported isolated sand 571

blows were produced by the main event on June 23, 572

2001, relationships between sand blow diameter and 573

epicentral distance are not qualitatively affected, with 574

respect to either the IGP or NEIC epicenter, but are 575

from the quantitative viewpoint. Regardless of being 576

the epicentral solution provided either by a national 577

(IGP) or an international (NEIC or USGS) agency, the 578

liquefaction evidence spatial distribution for this 579


Table 2


Epicentral distance of largest reported sand blows at a certain locality, for different main earthquake epicenters



Epicentral distance (km)

Sand blow diameter (m)


IGP epicenter

NEIC epicenter

HARVARD CMT epicenter


Acari River






Yauca River Camana´ delta Punta Bombo´ n Osmore River (Ilo) Sama delta

























t2.11 The minus ( ) sign indicates that the evidence is located to the northwest of the respective epicenter.


F.A. Audemard M. et al. / Engineering Geology xx (2005) xxx–xxx


580 particular earthquake is definitely asymmetric in

581 qualitative terms, being the extent of the occurrence

582 of liquefaction always larger to the southeast. It can

583 also be stated that, for a given radius from the selected

584 epicenter, the farther the evidence is, the more

585 pronounced the asymmetry becomes. But when

586 comparing these epicentral solutions with the Harvard

587 epicenter, the latter one is well off the other solutions

588 ( Fig. 1 ) and this is one of the most accessible

589 earthquake catalogue worldwide. This is a significant

590 limitation if this kind of correlation is to be applied

591 and extrapolated to a worldwide dataset, as Castilla

592 and Audemard (under review) have intended. The

593 national seismic catalogues are not typically at hand,

594 except for residents perhaps. Worldwide earthquake

595 data are essentially provided by international agencies

596 such as: NEIC, USGS and Harvard.

597 Explanation for such a discrepancy between the

598 Harvard and other solutions is given by the rupture

599 history of the main event. Based on how the

600 Harvard epicentral solution is calculated, it is known

601 that it indicates where the largest energy has been

602 released during the earthquake. Conversely, the IGP

603 epicenter, determined from a national array, should

604 provide the earthquake rupture nucleation. As

605 mentioned in Section 3, the isolated sand blow size

606 was the chosen parameter because it should be a better

607 measure of the earthquake energy and duration,

608 regardless of all other parameters (intrinsic and/or

609 extrinsic to the liquefiable sand) that must be taking

610 into account, should the scope of the study be a

611 reliable quantitative analysis. In single-rupture events,

612 we believe that the Harvard and other agency

613 epicentral solutions should be much alike. But in the

614 particular case of the June 23, 2001, earthquake,

615 which comprises three different sub-events, the

616 Harvard solution pinpoints the location of the third

617 and largest sub-event in terms of energy release,

618 which was the largest of the three sub-events after

619 Tavera (2002a) . Tavera (2002a) indicates that this

620 third sub-event is lagging almost 40 s behind the first

621 one that should represent the rupture nucleation area.

622 The Harvard epicenter is 160 km southeast of the

623 nucleation (given by IGP epicenter). Then, this raises

624 other doubts: is the liquefaction evidence distribution

625 biased by the rupture process? What epicenter has to

626 be chosen to estimate the farthest liquefaction

627 evidence reported to the southeast: IGP (nucleation

and national agency) or Harvard (maximum energy 628 release and international agency) epicenter? Are the 629

farthest evidence to the southeast produced by the first 630 sub-event, the third sub-event or the addition of the 631 three sub-events? As to this, Atkinson et al. (1984) 632 state that seismic shaking duration plays a more 633 prevailing role than just the total energy release. In 634 that sense, we then picked up the IGP solution for this 635

investigation, although is not the most accessible one 636

when a worldwide evaluation is intended. However, 637

we can confidently say that the liquefaction distribu- 638

tion is necessarily influenced by the main-event 639

rupture progression towards the southeast (multi event 640

rupture history), since energy release and shaking 641

duration is progressively added in that direction. 642

With respect to the second issue, there is neither 643

clear evidence nor suspecting hints of occurrence of 644

several liquefaction episodes triggered by both the 645

main event and any of the largest aftershocks between 646

Mollendo and Tacna, although that region was visited 647

in the hours following the largest (Mw 7.5) aftershock 648

of July 07, 2001, whose epicenter was less than 50 km 649

southeast of Mollendo and close to the visited coast 650

strip (see Fig. 1 for relative location). However, 651

another scientific party during a later reconnaissance 652

photographed liquefaction features at Camana´, resting 653

on the tsunami deposits of the main event, which 654

could only be related to a large aftershock. If these are 655

actually liquefaction features triggered by the largest 656

aftershock, they are 160 km northwest of the after- 657

shock epicenter, but no younger features cutting older 658

ones was actually observed between Mollendo and La 659

Yarada (near Tacna), where should widespread lique- 660

faction have occurred due to closeness to the after- 661

shock epicenter (refer to Fig. 1 for relative location of 662

epicenter with respect to this coastal strip). 663

Consequently, for plotting the diameter of the 664

largest isolated sand blow of all spotted at each 665

locality versus epicentral distance ( Fig. 9 ), we have 666

assumed that all the reported features were solely 667

triggered by the main event on June 23, 2001 (no 668

evidence supporting otherwise), and that shaking 669

duration is the sum of the three individual sub-event 670

durations composing the main event that nucleated at 671 and propagated southeastward from the IGP epicenter 672 determined from the Peruvian seismologic network 673 data. The number of plotted datapoints of largest sand 674 blows is very few, only six, as a straightforward 675



F.A. Audemard M. et al. / Engineering Geology xx (2005) xxx–xxx



250 150 50 -100 -50 100 150 200 250 300 350 400 450 NW IGP
300 350
400 450



Fig. 9. Cross-plot of diameter (in centimeters) of largest reported isolated sand blows and epicentral distance (in kilometers) to the locality wher e

spotted. The graph shows the distance calculated to the liquefaction evidence from both the IGP and NEIC epicenters.

676 consequence of the very low likelihood of liquefac-

that all these sand blows have been fed from a similar


677 tion potential of the coastal region due to aridity and/

range of depth, since no geotechnical investigations


678 or deeply seated groundwater table. However, this

were carried out at the sites exhibiting superficial 706

679 plot shows that isolated sand blow distribution is

liquefaction features.


680 definitely highly asymmetrical. Liquefaction is

In addition, this distribution supports that energy


681 observed much farther to the southeast than to the

propagation and rupture progression were southeast


682 northwest ( Fig. 9 ). Fig. 9 does not report all other

directed, as determined from other different evidence


683 isolated sand blows measured during this investiga-

(shape of meizoseismal area based on intensity data,


684 tion, because they clearly lie under the drawn line.

rupture history derived from seismograms, distribu-


685 This gun-shot-type distribution below the line does

tion of tsunami waves, aftershock distribution, 713

686 not further support the cross-plot. So, we can

among others), although epicentral distance to the 714

687 confidently state that the along-subduction-strike

farthest liquefaction evidence to the southeast could


688 distribution seems not biased significantly by other

be biased by the rupture history, meaning that it


689 parameters proper to the liquefiable sand bodies or

might be actually only 250 km away from the


690 their environment of occurrence, from a semi-quanti-

causative earthquake (third sub-event with high 718

691 tative viewpoint. In fact, stream channels in their

energetic contribution) instead of 390 km from the


692 downstream section throughout the coastal region are

rupture nucleation. Nevertheless, we are convinced 720

693 very similar, as to the liquefaction-prone environ-

that the very long duration of this large (Mw 8.4)


694 ment. In addition, measured sand blows occur atop

earthquake resulting from the addition of the three


695 channel bars in all cases, which lie less than a meter

sub-events, combined with a strong SE directivity 723

696 above running water, implying that water table was

(linked to the rupture history) and a depth of about


697 actually very shallow when liquefaction took place

30 km, is responsible for such a long epicentral


698 (refer to Table 1 ). In the same way, all sand blow

distance to the farthest liquefaction evidence reported


699 cones used in this relationship are made of medium to

to the southeast. Eventually, this distance should be


700 fine grain, light-colored, rather well-sorted sands,

even over 390 km, since sand blows were still 50 cm


701 which would suggest that they were fed, environ-

across at the Sama River crossing. Instead, they were


702 mentally speaking, from similar sand sources (refer to

not found at a distance of 110 km to the northwest


703 Table 1 ). Nonetheless, we cannot confidently state

(station 0 in Fig. 1 ).



F.A. Audemard M. et al. / Engineering Geology xx (2005) xxx–xxx


732 The implications of this finding for paleoseismic

lead to a better application to the past liquefaction


733 assessments relying on paleo-liquefaction features



734 preserved in the recent geologic record, as to earth-

735 quake magnitude and epicentral determinations, are of

736 highest significance. For past earthquakes exhibiting

7. Conclusions


737 highly mono-directional rupture propagation, for

738 which there is no simple way of revealing this aspect

739 from geologic data, an epicentral distance and earth-

740 quake magnitude can be reliably estimated only if the

741 largest sand blow is found. This is no easy task

742 because recognition of the spatial distribution of these

743 paleo-liquefaction features is only possible in out-

744 crops, which are never as frequent and thorough as in

745 plan view. Liquefaction associated to the New Madrid

746 1811–12 sequence is a perfect example of this. Spatial

747 distribution of the larger sand blows induced by that

748 earthquake sequence is still recognizable from aerial

749 photos over a century later ( Obermeier et al., 1993;

750 McCalpin, 1996; Obermeier, 1996; Obermeier and

751 Pond, 1999 ), which eases their search in outcrops.

752 However, shape of areas affected by older liquefaction

753 events in the same region are not so well constrained

754 from paleoseismic investigations (e.g., Munson et al.,

755 1997; Obermeier, 1998; Tuttle et al., 1996; Guccione

756 et al., 2002; Cox et al., 2004 ). Therefore, those

757 earthquake magnitudes can be hardly estimated with

758 certain reliability, solely based on size of affected

759 area. Consequently, this very frequently implies that

760 pre-historical earthquake magnitudes derived from

761 this type of studies have to be conceived as lower

762 bounds, unless the largest fossil sand blow is hit;

763 unless the size and shape of another liquefaction

764 feature can come in help to constrain it better, such as

765 width and/or frequency of sand dikes as proposed by

766 Obermeier (1996) , Munson et al. (1997) and Ober-

767 meier and Pond (1999) .

768 Finally, if this type of magnitude assessment –

769 relying on fossil sand blow shape – is to be further

770 developed, a more quantitative analysis needs to be

771 performed on future liquefaction occurrence. In that

772 sense, near-future post-earthquake field surveys will

773 necessarily have to study liquefaction features in a

774 deeper insight. This shall imply that the practice of

775 geotechnical surveys will have to become a must, as

776 well as sampling for grain-size distribution of

777 liquefied sands. The measuring of sand blow shape

778 also needs to become a common practice. This new

779 understanding through data gathering will surely

Although the Arequipa Mw 8.4 June 23, 2001, 783 earthquake triggered widespread liquefaction along 784

the southern coastal strip of Peru´, extending even- 785

tually over 400 km from the epicenter, sand-venting 786

features and lateral spread were only reported in the 787

major river plains and their deltas, only where there 788

was running water. This fact was conditioned by the 789

extreme aridity of the region and localized extension 790

of the liquefaction-prone environments. Lateral 791

spread resulted to be the most frequent evidence of 792

liquefaction induced by the main event, irrespective 793

of any association with sand spouting or not. If any 794

other event than the main earthquake on June 23, 795

2001, such as the 7.5 b aftershock Q on July 07, 2001, 796

triggered liquefaction, this survey party did not find 797

unequivocal and undisputable evidence of more than 798

one liquefaction event, although part of the survey 799

was carried out in the same day after that largest 800

aftershock. On the contrary, from the sedimentary 801

viewpoint, it could be possible to detect that different 802

feeder beds liquefy during the main event. From the 803

organic content, sand beds from both the overbank 804

sequence and the channel bars spouted sand and 805

water to the surface. Most liquefied sands were very 806

fine to medium grain size but few volcanoes also 807

contained some gravel. In some cases, some outward 808

coarsening could be evidenced in the cone con- 809

struction, at least during the final stage of water– 810


The ideal orientation of the surveyed transect 812

paralleling the southern Peru´ coast allowed to 813

determine from the largest reported sand blow 814

diameter that this earthquake had a strong energy 815

directivity and an asymmetric rupture history directed 816

to the southeast, thus supporting the same finding 817

from several other very different approaches. The 818

implications of this finding are of outmost importance 819

for paleoseismic assessments relying on past lique- 820 faction events preserved in the geologic record of a 821 given region. Due to the limited exposure of past sand 822 volcanoes, the derived magnitude of pre-historic 823 earthquake must be actually considered as a lower 824

sand mixture ejection.



F.A. Audemard M. et al. / Engineering Geology xx (2005) xxx–xxx

825 bound, unless the largest of all sand blows is reliably

826 discovered.

827 Acknowledgements

828 This technical collaboration was only possible

829 through the cooperation between the governments of

830 Peru´ and Venezuela, to whom we are grateful. We also

831 want to thank our own institutions: the Instituto

832 Geofı´sico del Peru´ – IGP – and Fundacio´ n Venezolana

833 de Investigaciones Sismolo´ gicas—FUNVISIS. Our

834 thanks also go to all locals who accounted their lived

835 experience, eye-witnessed liquefaction or provided

836 useful information to the purpose of this contribution.

837 Drafting is Marina Pen˜a’s contribution. Finally, a

838 former version of this contribution was largely

839 improved by the comments and suggestions from

840 Dr. Ellis Krinitzsky – ENGEO Editor in chief – and

841 Dr. Christian Hibsch, who are much thanked.

842 References


844 Atkinson, G.M., Liam Finn, W.D., Charlwood, R.G., 1984. Simple

845 computation of liquefaction probability for seismic hazard

846 applications. Earthquake Spectra 1 (1), 107 – 123.

847 Audemard, F.A., De Santis, F., 1991. Survey of liquefaction

848 structures induced by recent moderate earthquakes. Bulletin of

849 the International Association of Engineering Geology 44, 5 – 16.

850 Audemard, F.A., Gomez, J.C., Vargas, J. 2001. Efectos geo-

851 lo´ gicos asociados al sismo de Arequipa del 23 de Junio del

852 2001, Departamento de Arequipa, Peru´ meridional: informe de

853 misio´ n. FUNVISIS’ unpublished report for Instituto Geofı´sico

854 del Peru´ and Ministerio de Ciencia y Tecnologı´a-Venezuela,

855 34 pp.

856 Audemard, F.A., Gomez, J.C., Vargas, J., 2002. Efectos geolo´ gicos

857 asociados al sismo de Arequipa del 23 de Junio del 2001,

858 Departamento de Arequipa, Peru´ meridional. In: Tavera, H.

859 (Ed.), El terremoto de la regio´ n sur del Peru´ del 23 de Junio del

860 2001. Instituto Geofı´sico del Peru´, pp. 175 – 204 (in CD).

861 Beltra´n, C., De Santis, F., 1990. Manifestaciones de licuacio´ n en

862 Falco´ n oriental, a consecuencia de los sismos de los meses de

863 abril y Mayo de 1989. Funvisis’ unpublished report. 34 pp.

864 Carpio, J., Zamudio, Y., Salas, H., 2002. Caracterı´sticas generales

865 del tsunami asociado al terremoto de Arequipa del 23 de Junio

866 del 2001 (Mw=8.2). In: Tavera, H. (Ed.), El terremoto de la

867 regio´ n sur del Peru´ del 23 de Junio del 2001. Instituto Geofı´sico

868 del Peru´, pp. 121 – 128 (in CD).

869 Castilla, R., Audemard, F.A., under review. Sand blows as tools for

870 magnitude estimation of pre-instrumental earthquakes. Journal

871 of Seismology.

Cox, R.T., Larsen, D., Forman, S., Woods, J., Morat, J., Galluzzi, J.,




Preliminary assessment of sand blows in the Southern


Mississippi embayment. Bulletin of the Seismological Society


of America 94 (6), 1 – 18.


Gomez, J.C., Audemard, F.A., Quijano, J., 2002. Efectos geolo´ gicos


asociados al sismo del 23 de Junio del 2001 en el sur del Peru´.


In: Tavera, H. (Ed.), El terremoto de la regio´ n sur del Peru´ del



de Junio del 2001. Instituto Geofı´sico del Peru´, pp. 159 – 174


(in CD).