Marine and Petroleum Geology 67 (2015) 419e438

Contents lists available at ScienceDirect

Marine and Petroleum Geology

journal homepage: www.elsevier.com/locate/marpetgeo

Research paper

Submarine landslide deposits of the historical lateral collapse of Ritter

Island, Papua New Guinea
Simon Day a, Pilar Llanes b, c, *, Eli Silver b, Gary Hoffmann d, Steve Ward b, Neal Driscoll e
a
Dept. of Earth Sciences, University College London, London, UK
b
Earth and Planetary Sciences Dept., University of California, Santa Cruz, CA 95064, United States
c
Dept. Geodina
mica, Geolo

gicas, Universidad Complutense de Madrid, Madrid 28040, Spain
d
CFD Lab, Mech. Engineering, University of California, Santa Barbara, CA 93106, United States
e
Scripps Institution of Oceanography, La Jolla, CA 92093, United States

a r t i c l e i n f o a b s t r a c t

Article history: The March 13th 1888 collapse of Ritter Island in Papua New Guinea is the largest known sector collapse
Received 29 January 2015 of an island volcano in historical times. One single event removed most of the island and its western
Received in revised form submarine flank, and produced a landslide deposit that extends at least 70 km from the headwall of the
11 May 2015
collapse scar. We have mapped and described the deposits of the debris avalanche left by the collapse
Accepted 14 May 2015
using full-coverage multibeam bathymetry, side-scan sonar backscatter intensity mapping, chirp
Available online 22 May 2015
seismic-reflection profiles, TowCam photographs of the seafloor and samples from a single dredge.
Applying concepts originally developed on the 1980 Mount St. Helens collapse landslide deposits, we
Keywords:
Submarine landslide
find that the Ritter landslide deposits show three distinct morphological facies: large block debris
Debris avalanche avalanche, matrix-rich debris avalanche and distal debris flow facies. Restoring the island's land and
Ritter submarine topography we obtained a volume of 4.2 km3 for the initial collapse, about 75% of which is
Volcano collapse now forming the large block facies at distances less than 12 km from the collapse scar. The matrix-rich
Tsunami facies volume is unknown, but large scale erosion of the marine sediment substrate yielded a minimum
total volume of 6.4 km3 in the distal debris flow and/or turbidite deposits, highlighting the efficiency of
substrate erosion during the later history of the landslide movement. Although studying submarine
landslide deposits we can never have the same confidence that subaerial observations provide, our
analysis shows that well-exposed submarine landslide deposits can be interpreted in a similar way to
subaerial volcano collapse deposits, and that they can in turn be used to interpret older, incompletely
exposed submarine landslide deposits. Studying the deposits from a facies perspective provides the basis
for reconstructing the kinematics of a collapse event landslide; understanding the mechanisms involved
in its movement and deposition; and so providing key inputs to tsunami models.
© 2015 Elsevier Ltd. All rights reserved.

1. Introduction references therein) suggest that island or coastal volcano lateral

Lateral collapses is a nearly ubiquitous feature of volcanoes in
almost all volcanic island arcs. Numerous collapse generated
landslide deposits with volumes of cubic kilometers to tens of cubic
kilometers occur both in volcanic island arcs such as the Lesser
Antilles (Deplus et al., 2001; Le Friant et al., 2003) and Kermadec
arcs (Wright et al., 2006), and in back arc settings such as the Sea of
Japan (Satake and Kato, 2001). The available data (Table 1 and
* Corresponding author. Dept. Geodina
mica, Geolo
gicas, Universidad Complu-
tense de Madrid, Madrid 28040, Spain.
E-mail address: pllanes@geo.ucm.es (P. Llanes).
collapses with volumes of 1 or more cubic kilometers and
maximum thicknesses of 500 m or more occur at a global frequency
of the order of ~1 per 100 years. Smaller landslides removing
summit dome complexes or thin layers from the volcano flanks
(e.g., 2002 Stromboli; Bonaccorso et al., 2003) occur more
frequently. Significantly, the events in Table 1 indicate a broad
correlation between the size of the collapse event and the size of
the tsunami produced, as most of the deep-seated collapses of
Table 1 produced large destructive tsunamis.
The March 13th 1888 Ritter Island collapse is the largest ocean-
entering volcano lateral collapse, in volcanic arc settings, for which
we have written accounts of both the collapses and the resulting

http://dx.doi.org/10.1016/j.marpetgeo.2015.05.017
0264-8172/© 2015 Elsevier Ltd. All rights reserved.
 420
Table 1
Historical lateral collapses and large landslides at island and coastal volcanoes in volcanic arcs for which written descriptions of both the source events and the resulting tsunamis are available. A. Large, deep-seated lateral
collapses. B. Small-volume, shallow-seated volcano flank landslides that produced damaging tsunamis. Note the recent dates of these events, the historical record may be incomplete for the period prior to 1900, compared to the
record of lateral collapses.

Date Volcano Landslide volume, if known Maximum local tsunami runup Distant tsunami runups Landslide type References
(examples)

S. Day et al. / Marine and Petroleum Geology 67 (2015) 419e438

1640 Komagatake, Hokkaido, Japan 0.3 km3 >8 m Ocean entry by landslide from Katsui et al. (1975)
subaerial volcano lateral
collapse
1741 OshimaeOshima, Sea of Japan. 2.4 km3 (large block facies of 34 m ?; certainly >15 m (on 3 m e 4 m on Korean coast, Partly submarine volcano Kato (1997) and Satake and
deposit); 2.5 km3 (collapse scar) coasts 60e80 km distant) 1200 km distant lateral collapse Kato (2001)
1792 Unzen, Japan 0.34 km3 (collapse scar volume) >10 m none: collapse into enclosed Ocean entry by landslide from Geographical Survey Institute
bay. subaerial volcano lateral (1982)
collapse
1883 Augustine, Alaska ~0.5 km3 >19 m? 6 me8 m, ~100 km distant Ocean entry by landslide from Beget and Kowalik (2006) and
subaerial volcano lateral Kienle et al. (1987)
collapse
1888 Ritter Island, Papua New ~4 to 5 km3 (Johnson (1987)); >15 m (on coasts up to 50 km 8 m at Hatzfeldhafen, 370 km Partly submarine volcano Cooke (1981) and Johnson
Guinea 4.2 km3 (our estimate) distant) distant; 4.5 m at Rabaul, lateral collapse (1987)
540 km distant
1933 Harimkotan, Kuriles ~1 km3 (collapse scar volume) 20 m Significant damage on adjacent Ocean entry by landslide from Gorshkov (1970)
islands subaerial volcano lateral
collapse
B
1928 Paluweh, Indonesia Volume poorly known due to 3 waves, from 5 to 10 m Ocean entry by subaerial Neumann van Padang (1929)
eruption landslide at start of large
explosive eruption
1966 Tinakula, Solomon Islands <0.01 km3 Small local waves only Ocean entry by landslide from Latter (1981) and Johnson, R.W.
small failure near summit and Workshop Organizing
Committee (1999)
1979 Ili Werung, Indonesia 0.05 km3 9m Ocean entry by subaerial Pararas-Caryannis (1979)
landslide
3
2002 Stromboli, Italy 0.02 km 10 m 2 m (140 km distant) Two thin slope e parallel Bonaccorso et al. (2003) and
landslides, one subaerial and Tinti et al. (2006)
one submarine
 S. Day et al. / Marine and Petroleum Geology 67 (2015) 419e438 421

Figure 1. Location of Umboi, Sakar and Ritter Islands over digital terrain model (DTM) constructed from multibeam data gridded at 24 m. Shaded-relief image of the topography is
from SRTM data. Inset shows location in the New Britain volcanic arc, northeast of Papua New Guinea. 60 of latitude ¼ 11.1 km.

tsunamis (Table 1). It produced a large, regionally destructive
tsunami that was witnessed by literate observers who timed the
waves with watches and provided detailed accounts (summarized
in Cooke, 1981). That data has been key to modeling the generation
of the tsunami by the landslide and its later propagation (Ward and
Day, 2003). Understanding how submarine landslides from volcano
collapses generate these tsunamis is a key motivator for studying
their kinematics and ultimately their mechanics. There is a need for
detailed studies of young, well exposed landslide deposits on the
ocean floor that are associated with known tsunamis in order to
provide a basis for the interpretation and modeling of older,
incompletely exposed landslide deposits from other island arc
volcanoes.
In this paper we describe the submarine landslide deposit from
the 1888 collapse of Ritter Island (Figs 1 and 2). We show that while
the collapse was a single event, the deposit is subdivided into
distinct facies as a result of rheological transformations during
movement; and that the geometry of the deposit and associated
erosional features can be used to reconstruct the kinematics of
landslide movement. Historical descriptions, sketches and mea-
surements of pre-collapse Ritter Island define its geometry above
sea level in more detail than for any other volcano lateral collapse,
and these descriptions also constrain the sequence of events
(Cooke, 1981; Johnson, 1987; Ward and Day, 2003). These data have
provided us with a unique opportunity to analyze the deposits of a
submarine volcano collapse landslide in the context of independent
knowledge of the event that produced it and the associated
tsunami. Computer simulations illustrating the type of Ritter Island
landslide processes evidenced in this paper can be found here:
http://es.ucsc.edu/~ward/Ritter-new.mov.
http://es.ucsc.edu/~ward/Ritter-new-pn.mov.
http://es.ucsc.edu/~ward/ritter-tube.mov.
2. Ritter Island before 1888 and today
2.1. Field observations made on Ritter Island
Descriptions by Johnson et al. (1972) and our onshore field ob-
servations in 2006 reveal that the subaerial exposed part of Ritter is
a volcanic sequence dominated by rubbly or spatter-fed lavas with
some massive flow cores, and intercalations of finer-grained scoria.
Steeply east dipping sequences form the whole of the volcanic
stratigraphy, from the rim of the collapse scar to the oldest sub-
aerial lavas, exposed adjacent to hyaloclastite tuffs with thin scoria
422 S. Day et al. / Marine and Petroleum Geology 67 (2015) 419e438
Figure 2. (Top) Photographs of Ritter taken in 2006, courtesy of John Holder. Left: view from the southwest. Right: view from the southeast. (Bottom left) Sketch of Ritter Island in
1835 looking from the south, modified from Jacobs (1844). Arrows show how the modern island was used to estimate the scale of the sketch. (Bottom right) Sketch of Ritter Island in
1835 looking from the south, modified from Woodworth (unpublished, courtesy of James Fairhead, University of Sussex, UK).
beds that are the structurally lowest and oldest unit of the
remaining subaerial part of the edifice. These rocks are cut by
numerous irregular, broadly NeS striking dikes and outward dip-
ping sheet intrusions up to several meters thick, and some NW
striking dikes. Fumarolic alteration is present in the lower part of
the exposed sequence. The island is incised by gullies up to several
tens of meters wide, with steep to vertical walls up to a few tens of
meters high. Similar gullies filled by lava flows and discrete boulder
brecciate units form minor unconformities within the pre-collapse
volcanic sequence. However, there are no marked unconformities
extending along the face of the collapse scar that would record past
collapse events, nor are there flat lying, ponded lava flows that
would indicate a collapse scar buried within the intact remnant of
the island, as seen elsewhere (Day et al., 1999; Carracedo et al.,
1999, 2002). There is no evidence in the remaining part of the is-
land for recent but pre-1888 large lateral collapses of Ritter, com-
parable to that which occurred in 1888. Historical indications of
much smaller pre-1888 landslides from the island are discussed
below.
2.2. Historical observations from Ritter Island
Cooke (1981) summarized the observations of activity at Ritter
Island from William Dampier's sighting of an eruption in 1700
through to the German colonization of northern New Guinea that
started in the 1880s. These accounts indicate that Ritter experi-
enced at most intermittent eruptive activity in the years just before
the collapse, but numerous accounts of fumaroles indicate hydro-
thermal activity. Later collection of reports from communities on
Umboi Island indicated that the collapse of the volcano around 6
am on March 13th 1888 was preceded by explosions accompanied
by earthquakes, but there was no mention of an eruption (Fisher,
1939, 1957). Thus, the collapse was most likely associated with
phreatic activity, and would have been of the Bandai type in the
classification of Siebert et al. (1987).
The collapse produced tsunami waves that were recorded at
observation points between 100 and 600 km from the volcano
between about 6:30 am and 8:30 am local time on March 13th, and
continued for 40e150 min with dominant wave periods around
3 min. No observer described more than one series of tsunami
waves. Ward and Day (2003) showed that the times of arrival and
durations of the tsunami at these different sites are consistent with
the propagation, and progressive frequency dispersion with dis-
tance, of ~3 min period tsunami waves produced by a single land-
slide from Ritter Island. No further tsunamis were recorded at the
various observation points in the days to months immediately after
the event. The historic evidence is consistent with only one major
collapse or landslide event occurring in 1888.
Earthquakes and submarine explosive activity immediately
west of Ritter Island in 1972 and 1974 marked the growth of a new
submarine volcanic vent within the collapse scar (Cooke et al.,
1976; Cooke, 1981). Two small tsunamis occurred in these years,
the only ones to be recorded from Ritter between 1888 and the time
of our survey in 2004. Any young, large volume landslide deposits
in the area NW of Ritter Island are considered to be associated with
the single 1888 collapse.
Despite the lack of large volume tephra falls associated with the
events of 1888, it was supposed for many years that an explosive
eruption and vertical caldera collapse had occurred at Ritter Island
in 1888 (Fisher, 1957; Cooke, 1981). Following the 1980 May 18th
Mount St. Helens lateral collapse and recognition of the importance
of lateral collapses at volcanoes, the Ritter event was investigated
by Johnson (1987). This single-beam echosounder survey of the
immediate area of the island revealed a broadly west facing open
horseshoe shaped collapse scar and large bathymetric highs,
interpreted by Johnson as landslide blocks, allowing him to esti-
mate a collapse scar volume of 4e5 km3. However, the survey did
not extend far beyond the collapse scar and so the full run-out
distance, extent and structure of the deposits associated with the
collapse has been an important gap in knowledge of the 1888 Ritter
Island event.
 S. Day et al. / Marine and Petroleum Geology 67 (2015) 419e438 423

2.3. Restoration of pre-collapse island topography
The description of Ritter in 1700 (Dampier, 1729) shows the is-
land to have been a steep, gullied cone with a height about half of
the adjacent extinct volcano of Sakar Island that rises to 998 m.
Accounts in the 19th Century emphasized the extremely steep-
sided nature of the subaerial part of Ritter volcano with measure-
ments of about 780 m height and 1.83 km EeW width (Dumont
d’Urville, 1832), implying slopes significantly in excess of 45 . To
estimate the island's dimensions we compared the sketch and
description of the island, made in early 1835 by Jacobs (1844), with
the modern island and matched the present island's features to
gullies in Jacobs' sketch (Fig. 2). Scaling the sketch to the modern
island with a maximum height 140 m yields a pre-collapse height of
around 750 m and an EeW width of around 1.5 km. These pre-
collapse dimensions are consistent with the ~45 slope of the sur-
viving eastern flank of the island. Because the island is still about
1.9 km long it must have been elongated NeS. Another sketch of the
island made on the same voyage in 1835 by Woodworth (Fig. 2)
confirms the overall steep slopes of the island. Both sketches sug-
gest that the western side of the island (to the left in both cases)
was steeper and perhaps had already experienced small landslides
of less than 0.1 km3 comparable to those in Table 1B. Comparison
with data for similar events such as the 2002 Stromboli landslide
(Bonaccorso et al., 2003) suggests that deposits from such land-
slides would not have gone more than a few kilometers from Ritter
and would therefore be either buried under the 1888 deposits or
would have been eroded and entrained into the 1888 debris
avalanche. Our bathymetric data, discussed below, indicates that
the submarine slopes of the volcano down to 700 m below sea level
are mainly symmetrical and an upward extrapolation of these
slopes to sea level implies that the pre-collapse island had an NeS
length of a little over 2 km. We use this restoration of the pre-
landslide topography together with an extrapolation of the sub-
marine collapse scar in a later section to approximate the volume of
the sector of the island collapsed during the 1888 event.

Figure 3. Side-scan imagery of study area from MR-1 sonar data gridded at 16 m. White dotted line outlines landslide deposits. BS marks location of the black streak. Yellow lines
locate chirp seismic reflection profiles acquired and red lines locate segments of those shown in the paper. Transects using the underwater digital towed camera system “TowCam”
are marked with dashed blue line. Blue star show location of a dredge site. Low backscatter light and high backscatter dark. 60 of latitude ¼ 11.1 km. (For interpretation of the
references to colour in this figure legend, the reader is referred to the web version of this article.)
424 S. Day et al. / Marine and Petroleum Geology 67 (2015) 419e438

3. Methods

During NovembereDecember of 2004 we conducted a

geophysical research cruise along the Bismarck volcanic arc aboard
the R/V Kilo Moana. We acquired swath bathymetry data with a
Simrad EM-120 hull-mounted echo sounder, operating at 12 kHz
(±0.75 kHz). The beams are transmitted across a total swath
opening of up to 150 , depending on depth. Acoustic backscatter
intensity maps were obtained with the HAWAII MR1 Sonar, a
portable shallow-towed system with swath width up to 7.5 times
the water depth. The MR1 sonar transducers are housed in a 4.8 m
long, 1600 kg vehicle that is towed beneath the surface mixed layer
(80e120 m) and requires a minimum water depth of 500 m,
limiting our work at shallower depths. The Hawaiian Mapping
Research Group processed both the bathymetry data and the
acoustic images on board using Simrad's software. We used a me-
dian filter and gridded the bathymetry and the acoustic data to 24
and 16 m, respectively. High-resolution Compressed High Intensity
“Radar” Pulse (CHIRP) sub-bottom sonar data were collected with
an X-Star towed 1e5.5 kHz instrument (Scripps Institution of
Oceanography). CHIRP data were acquired by transmitting
approximately every 5e10 m along selected transects with pene-
tration up to several tens of meters. CHIRP data are unmigrated.
Digital photographs of the seafloor were taken using the digital
“TowCam” system (Woods Hole Oceanographic Institution) along
four transects over parts of the Ritter deposits, each of the tows
lasting approximately 4 h from deployment to recovery (see Fig. 3
for locations). The camera is towed 100e300 m behind the ship at
speeds of ¼ to ½ knot, at an elevation of about 5 m above the Figure 4. Bathymetry map of study area gridded at 24 m. Contour interval is 100 m.
seafloor. Elevation is controlled by the winch operator in response Bathymetric ridges from Umboi and Sakar constricting the valley are marked with R1
to measurements by the camera and so is prone to variations over and R2 respectively. R3 marks the smaller ridge that the fast-moving part of the
avalanche passed over. 60 of latitude ¼ 11.1 km.
uneven seabed. The TowCam records depth, altitude, water tem-
perature, water clarity or turbidity and it was set to record over-
lapping seabed photos every 10 s. It allowed us to document the
location and distribution of a variety of different lithologies and
structures exposed on the seabed, and distribution of megafauna Lipman, 2004). We see no evidence to indicate that the collapse
(animals visible in camera were typically >1 cm). Correlation of the scar was produced by multiple failure events.
camera images with the bathymetric, acoustic backscatter and The collapse scar has been partly filled by a post collapse vol-
CHIRP data provided ground truth and enabled better interpreta- cano with a width of 1600 m and a height of 250 m (from 550 to
tion of the various types of acoustic data.

4. The submarine collapse scar and the post-collapse volcanic

edifice

The collapse scar has a 2 km long headwall in an NeS direction

and extends to the western foot of the volcano at 900 m depth,
where it is up to 4 km wide with the sidewalls defining a sector arc
of about 60 (Fig. 4). The north and south walls of the collapse scar
are steep, rising 100e200 m above the level of the post-collapse fill,
and with very high backscatter intensity. This could be suggestive
of exposure of bedrock on the cliff faces (Fig. 5), or either less than a
meter of drape, depending on the frequency and incidence angle as
well as sediment attenuation properties, following Mitchell (1993).
The spoon or scallop shaped geometry of this surface may suggest
that the landslide began as a single rotational failure.
The single continuous tsunami reported by the eyewitnesses
indicates that the largest part of the collapse scar was generated in
a single event; but the failure may have consisted of a number of
blocks moving in rapid succession upon a single failure surface and Figure 5. Side-scan image detail of the large block debris avalanche facies from MR-1
thereafter moved as a single unit. No lateral boundaries between sonar data gridded at 16 m. Largest blocks are marked as B, post-collapse volcanics as V
distinct lobes of similar facies deposits are seen in the Ritter Island and pre-collapse volcanics as V’. Areas of sediment scouring are labeled (see also
Fig. 12, “scours to bedrock by landslide” in inset legend). R1, R2 and R3 mark bathy-
deposits. If multiple collapse events had occurred, deposits from metric ridges constricting the valley, the latter being the smaller ridge that the fast-
the earlier collapse events would be expected to show a partial moving part of the avalanche passed over. Low backscatter light and high back-
cover from deposits associated with the later events (Yokose and scatter dark. Topography from SRTM data. 60 of latitude ¼ 11.1 km.
 S. Day et al. / Marine and Petroleum Geology 67 (2015) 419e438 425

300 m of depth). This and other volcanic vents on the surface of the backscatter intensity extending from the base of the post collapse
landslide deposit are distinguishable from collapse blocks because edifice to the NW and SW may also be features of post collapse
they have conical shapes and low backscatter intensity (V in Fig. 5). eruptions, such as deposits from syneruptive volcaniclastic density
Because the volcanic vent within Ritter crater is at a shallow water currents (Fiske et al., 1998).
depth, we were able to compare it with the cone parameters given
by Mitchell et al. (2002). Representing the cone width versus its
5. Submarine landslide facies
height we obtained a very good fit for a submarine cone (see Fig. 4
in Mitchell et al., 2002), another argument to support its volcanic
Glicken (1991, 1996) subdivided the morphological and sedi-
origin. The summit vent of the large edifice within the collapse scar
mentological characteristics of the deposits from volcano collapses
is located at the sites of eruption plumes that broke the surface of
into distinct facies. He described the landslide facies of the Mount
the sea during the 1972 and 1974 explosive volcanic eruptions
St. Helens deposit in great detail, including block size distribution,
(Cooke et al., 1976). Eruptive activity inside the scar left by a lateral
abundances of discrete blocks in contrast with matrix material,
collapse is common because ascending magmas take advantage of
deposit thickness variations and depositional slopes. We compared
the weakened areas created by collapse structures, where there is
Glicken's landslide facies concepts to our submarine observations
also a loss of confining stress, erupting within it (e.g. the Lesser
and divided the deposit into three facies: large block avalanche,
Antilles, Boudon et al., 2007). The flanks of the main post-collapse
matrix-rich debris avalanche, and distal debris flow facies. In the
volcano are smooth, lacking discrete flank vents, and have uniform
following sections we describe and interpret the features of each of
moderate backscatter intensity. The smooth lobes of low
these distinct facies, before synthesizing the results into an overall
interpretation of the movement history of the landslide. Such an
approach will yield new insights into how submarine landslides
disintegrate and evolve downslope away from their source.

5.1. Large block debris avalanche facies

5.1.1. Depositional features

The large block facies extends 12 km from the headwall of the
collapse scar, filling the full width of the valley between Sakar and
Umboi Islands as far as a constriction in the valley formed by two
overlapping ridges trending NEeSW across the valley sides (R1 and
R2, Fig. 4). The margins and top surface of the large block facies
deposit are at a depth of about 900 m on its eastern and northern
sides, but descend towards the gap between the ridges, where the
top of the deposit is at 1240 m depth at its furthest downslope limit,
to the northwest. The average slope along the length of this facies is
about 2.1. The facies is characterized by large discrete irregular
blocks or hummocks up to more than 1 km across and over 100 m
high, with generally high backscatter intensity (B in Fig. 5). The
distribution was determined by contrasting acoustic responses of
the lava blocks and the fine products surrounding them, as well as
the angular shapes of the blocks, which provide surfaces facing the
beams, hence high backscatter. The largest of the blocks (B in Fig. 6)

Figure 6. 3D image of Ritter insular margin. Letter B marks location of the largest Figure 7. Part of CHIRP seismic line 7/0 across the large block avalanche facies (from
debris avalanche block. Note how the large block facies has been confined near the shot point 900 to the NE). The collapse blocks produced strong seabed diffractions and
island by a constriction. Letter V marks location of the main post-collapse volcanic lack internal structures. Location of the lines is shown in Figure 3. TWT denotes two
cone. Vertical exaggeration is 2. way time.
426 S. Day et al. / Marine and Petroleum Geology 67 (2015) 419e438

Figure 8. Side-scan image detail of matrix-rich debris avalanche facies from MR-1 sonar data gridded at 16 m. Low backscatter light and high backscatter dark. M for sediment
mesas with marked troughs around their margins in some cases. We interpret the mesas to be erosional features related to a landslide or flow moving from SE to W. M* for the best
example of a sediment mesa (see text for details). Line 24 in Figure 11 shows a chirp profile crossing two of the mesas, from shot points 100e250 and 750e1000, the latter being
mesa M*. M0 for sediment mesas along the SW margin of the facies deposit and without troughs developed around them. We attribute the latter to be the result of uneven upslope
retreat of scarps. BS marks location of the black streak. 60 of latitude ¼ 11.1 km.

is located in the mouth of the collapse scar and its asymmetry is
consistent with it being a displaced and back-rotated coherent
segment of the flank of the volcano. The majority of the other large
blocks are concentrated toward the southern and western margins
of the deposit. The high slopes and irregular surfaces of the blocks
produced strong seabed diffractions and the characteristic lack of
seismic penetration over blocks deposits. The absence of a sedi-
mentary cover over these large blocks in the seismic profiles (Fig. 7)
is consistent with the recent age of the landslide deposit.
The overall geometry of the large block facies deposit suggests
that it was ponded in the valley to the east of the constriction. This
ponding is similar to that seen in the large block facies of the Mount
St. Helens landslide deposit (Glicken, 1991, 1996) and others like
that of Mount Shasta (Crandall et al., 1984; Crandall, 1988).
5.1.2. Erosional features
On the lower slope of Umboi Island there is a laterally extensive
region of uniform high backscatter intensity with a sharp upslope
contact (named “Scours” in Fig. 5) sub-parallel to the 900 and 800
bathymetric contours. This contact undulates, curving upslope and
over a small ridge (R3 in Fig. 4) and reaching its maximum elevation
at a depth of 700 m, which is 200 m above the level of the large
block facies deposit (R1 in Fig. 4). The contact is continuous across
several fans of high backscatter material trending down the slope
from Umboi Island, but it is discontinuous where larger fans are
present. The high backscatter zone also extends over the crest of
the larger ridge (R1 in Fig. 4) that bounds the large block facies
deposit at the constriction and downslope into the matrix rich
facies deposit area.
 S. Day et al. / Marine and Petroleum Geology 67 (2015) 419e438 427

Figure 9. (Top) Chirp line 2/7 showing matrix-rich facies of Ritter debris avalanche and its transition towards the east to a layered sedimentation at the lower insular slope of Sakar
Island. (Middle) Continuation of Chirp Line 2/7 showing large blocks and a buried landslide deposit from Sakar Island. (Bottom) Line 3/1 across the insular slopes of Umboi and Sakar
Island showing the matrix-rich facies of Ritter and the channel eroded by Ritter debris avalanche. Diffractions related to hummocks. Location of the lines is shown in Figure 3. TWT
denotes two way time.

5.2. Matrix rich debris avalanche facies
The transition from block-rich to matrix-rich facies occurs
where the gap between Umboi and Sakar is constricted between
two ridges (R1 and R2 on Fig. 4). Similarly, the transition from large
block facies to matrix-rich facies in the 1980 Mount St. Helens
debris avalanche deposit occurs at a constriction in the North Fork
Toutle River (Glicken, 1991, 1996). In this section we describe the
nature of depositional and erosional features in this part of the
landslide, the latter consisting of steep-sided mesas and grooved
terrains to either side of the matrix-rich facies deposit itself.
5.2.1. Depositional features
The matrix rich facies is characterized by a distinctive speckled
or mottled pattern in the backscatter intensity map (Fig. 8), and as a
coarse blocky and non-stratified mass in high-resolution seismic
data (Shots 2800e3500 in Line 2/7, shots 700e1300 in Line 3/1
Fig. 9). The patches of high backscatter intensity are 100e300 m
across. Both features are typical of the irregular hummocky ter-
rains, found in other submarine debris avalanche deposits
(Holcomb and Searly, 1991; Moore et al., 1989, 1994; Watts and
Masson, 1995). We interpret the high backscatter intensity
patches to be areas where volcanic material is exposed on the
seabed, since high backscattering arises because the volcanic rock is
rough at the scale of the acoustic wavelength, and also volcanic
blocks have high impedance contrast relative to the sediment de-
posits surrounding them.
The acoustic character, a lack of CHIRP penetration, and discrete
areas of diffractions allowed us to map the lateral extent of the
matrix rich facies, differentiating it from the sediments of the lower
slopes of Umboi and Sakar Islands. Along the lower slope of the
islands we observed well-layered reflections, interpreted as bedded
sediments. Diffractions are present in areas of high slope. The
thickness of acoustically laminated sediments varies depending on
slope and location relative to the islands. Line 3/1 in Figure 9 shows
penetration up to 0.016 s two way travel time along the slope of
Umboi and 0.008 s two way travel time along the slope of Sakar.
This acoustic penetration is equivalent to 12 and 6 m using a
nominal water velocity of 1500 m/s to convert for thickness.
A series of lobes with the characteristic speckled appearance in
the backscatter intensity map occur at the downslope end of the
matrix rich facies deposit. These lobes are located in flat-floored
channels between steep-sided, flat-topped features referred to
here as mesas. TowCam images of these distal lobes show that they
are partly covered in fine-grained, bioturbated (hence soft and
unconsolidated) marine sediment. Hummock tops encountered
along the TowCam lines 5 and 6 (Fig. 3) that run across the lobes are
a few tens of meters to several tens of meters across, and several
meters high, composed of poorly sorted, angular blocks mainly less
than a meter across, but ranging in size from centimeters to over
five meters long (Fig. 10aec). No banding or layering is observed in
the blocks, nor are there any clear variations in block characteristics
at any one hummock or mound. Polygonal-section columnar
blocks, interpreted as being from a thick columnar-jointed sub-
strate lava flow or sheet intrusion, are exposed in one of the
mounds of TowCam 5 (Fig. 10b), whereas other mounds in the same
TowCam line showed pitted clasts that appear to be coarsely ve-
sicular lava or scoria (Fig. 10c). The clasts form a habitat for abun-
dant sessile filter-feeding organisms like white sponges, soft corals,
anemones and crinoids that require hard substrates for their at-
tachments and are not observed in sediment-draped areas. In
summary, we interpret the mounds as debris avalanche hummocks
428 S. Day et al. / Marine and Petroleum Geology 67 (2015) 419e438
Figure 10. Photographs taken with the “TowCam” system. Each photo shows an area
around 7 m across. Blue-white and pink objects are seabed organisms (brittle stars,
soft corals and anemones). A) Poorly sorted angular rock clasts from a hummock
within the matrix-rich debris avalanche facies (TowCam 5). B) Polygonal-jointed
columnar blocks in one of the hummocks within the matrix-rich debris avalanche
facies. The polygonal jointing suggests either dike rocks or part of a thick subaerial lava
flow (TowCam 5). C) Steep side of a hummock composed of pitted clasts apparently
coarsely vesicular (TowCam 5). D) Angular clasts, dark and pitted from the Black Streak.
The clasts are uniform in size, around 10e20 cm across, in contrast to the very poorly
sorted clasts of the debris avalanche hummocks, suggesting sorting in a submarine
explosive eruption (TowCam 6). E) Top of the mesa M* scarp (Fig. 8 for location)
(TowCam 5). F) Near vertical scarp of mesa M* with a steep slope partly covered by
cohesive stratified sediment blocks (TowCam 5). G) Scalloped erosional grooves in
cohesive sediments to the east of the debris avalanche termination (TowCam 5). H)
Clast alignment seen in the distal facies following the landslide movement direction
and indicative of pervasive shearing during deposition (TowCam 3). I) Cohesive sedi-
ment clasts seen in the distal facies (TowCam 4). J) Same clasts as seen in H taken from
the dredge sample. K) Scalloped features at the crest of a scarp at the western margin
of the distal deposit, similar to those seen in the grooved terrain (TowCam 4). (For
interpretation of the references to colour in this figure legend, the reader is referred to
the web version of this article.)
or fragmented blocks composed of different types of volcanic rock
(and perhaps some intrusive rocks) from the Ritter Island edifice.
5.2.2. Erosional features
5.2.2.1. Submarine canyon. Along most of the length of the matrix
rich facies deposits there is a 1 km wide and 20e40 m-deep sub-
marine canyon. It starts at the valley constriction that limits the
block and matrix facies deposits and it is easily recognized out to
the 1700 m isobath. The five CHIRP profiles crossing the channel
image a strong reflection with no penetration, indicating the lack of
a sedimentary sequence in the canyon (e.g., line 3/1 in Fig. 9) and
demonstrating that this canyon was formed by erosion. This canyon
abruptly truncates the gullies in the northeast slope of Umboi, and
the knickpoints present on them may be the result of erosion on the
steeper lower gradient, comparably to the ones described by
Mitchell (2006). This could support the canyon being young.
5.2.2.2. The “Black Streak” feature. Between the distal lobes of the
debris avalanche deposit there is an elongated feature with very
high backscatter intensity that we informally named the “Black
Streak” (BS in Fig. 8). It is 1 km wide and rises 20 m above the
surrounding terrain. TowCam images over its crest show that it is
composed of angular clasts that are dark and pitted, appearing
vesicular or scoriaceous (Fig. 10d). An abundant and diverse fauna
of sponges, soft corals and crinoids is present, similar to that seen
on the debris avalanche blocks. Unlike the poorly sorted clasts in
the debris avalanche deposit, the clasts in the “black streak” are
uniform in size, 10e20 cm across. The CHIRP profile crossing it
shows its relief in cross-section, with strong echoes underlying
transparent facies and debris avalanche deposits covering its
eastern side (Line 24, Fig. 11).
5.2.2.3. Mesas. To either side of the distal lobes of the matrix-rich
debris avalanche deposit there are several flat-topped surface fea-
tures rising above the surrounding seabed and with a mesa-like
appearance (M in Fig. 8). These mesas are unique to this area, and
similar features were not found anywhere else in the length of the
Bismarck arc covered by the research cruise (Silver et al., 2009;
Llanes et al., 2009). The best-defined example (M* in Fig. 8) has a
near vertical scarp, 50 m-high, with a steep slope partly covered by
cohesive stratified sediment blocks (Fig. 10e,f). Along its margins,
except that to the NW, there are marked troughs, several m-deep
and almost 100 m-wide. Images taken in these troughs revealed a
scattering of lithic clasts similar to those seen in the matrix-rich
debris avalanche deposit hummocks. Two of those mesas were
also crossed by CHIRP profile 24 (Fig. 11), revealing the presence of
sedimentary layering within them, in contrast to the acoustically
transparent adjacent debris avalanche deposits.
Along the SW margin of the matrix rich debris avalanche facies
deposit, several mesas are developed on a marked slope above the
debris avalanche deposit, without troughs developed around them
(M0 in Fig. 8; Line 3/1 in Fig. 9). The contrast between these and the
line of mesas running across the path of the landslide is considered
further in Section 7.
5.2.2.4. Grooved terrain. To the north and east of the matrix rich
facies deposits and on the tops of the sediment mesas, we observed
a lineated or grooved terrain in the backscatter and TowCam 5
images (mapped with dotted areas in Fig. 12). A correlation of sets
of TowCam images and the camera altimeter allowed us to estimate
that the largest grooves imaged were several tens of meters across
and several meters deep and seem to be small-scale counterparts of
larger linear structures (several thousand meters long) that pro-
duce the fabric seen in the backscatter data. The walls of the
grooves are steep, occasionally overhanging, and characteristically
 S. Day et al. / Marine and Petroleum Geology 67 (2015) 419e438 429

Figure 11. Selection of CHIRP profiler images from different parts of the deposit. Line 24 is a profile across the matrix-rich debris avalanche showing the pre-avalanche sediments,
the “black streak” and sediment mesas. Lines 3/25, 2/4 and 2/5 are segments along the distal facies, while Line 22 shows a segment across it. Red bars mark thickness of the
avalanche, labeled in meters using a conservative velocity of 1500 m/s to calculate the thickness (if a velocity of 1600 m/s will be used, the values will increased by a 6.25%). Location
of the lines is shown in Figure 3. TWT denotes two way time. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this
article.)

scalloped suggesting that the grooves are cut into cohesive sedi-
ment with significant strength (Fig. 10g). In rare cases, planar
sedimentary bedding is visible in groove walls. A sessile fauna
dominated by near-transparent barrel-shaped sponges occurs on
some groove walls but the diverse fauna seen on the debris
avalanche deposit blocks is not present. There are sparse, scattered
lithic angular clasts a few tens of centimeters across, in the floors of
some of the grooves, also very similar to the clasts forming the
blocks of the matrix rich debris avalanche facies. Apart from these
clasts, the groove floors are largely covered by soft sediment. The
best exposures of grooves are on top of the sediment mesas,
although they are also exposed in the intervening areas. Isolated
high backscatter intensity features, several tens of meters across,
are also present in the grooved terrain and may be comparable with
the high backscatter intensity “speckles” produced by debris
avalanche hummocks in the matrix-rich debris avalanche facies.
5.3. Distal debris flow facies
The transition between the matrix rich debris avalanche facies
and the distal deposits to the north appears sharp in the backscatter
intensity map but is gradual without a sharp boundary that can be
differentiated in the seismic data. In the latter, the hummocks that
characterized the matrix rich facies with presence of debris
avalanche blocks decrease to leave a relatively flat seabed of
acoustically transparent material in the distal deposits. This type of
transition is common in debris avalanches that transform to debris
flows, while in debris avalanche lobes where this transformation
does not occur the lobe has an abrupt front (Capra et al., 2002). We
mapped an approximate limit for the transformation, based on the
difference in backscatter intensity. The distinction between the two
facies is also supported by comparison of TowCam images from
tows 5 and 6 (matrix facies) with tows 3 and 4 (distal facies). Here
we discuss the depositional and erosional features of the distal
facies.
5.3.1. Depositional features
The distal landslide deposit consists of a laterally extensive
sheet extending over a broad, almost flat submarine plain with an
average slope of 0.11, less than 10% of the slopes on which the
debris avalanche facies deposits were emplaced. Within the distal
deposit there are areas with differential high or low backscatter
intensity that correspond to variations on the slope but that do not
show acoustic variations in the CHIRP profiles. The margins of the
deposit are ill-defined and in at least some places are covered by
more recent sediment (see below). It is likely that the distal deposit
extends out of the northern limit of backscatter mapping beneath
the “undivided marine sediments” unit of Figure 12. Figure 11
shows representative portions of seismic lines along and across
the distal facies, where the deposit is recognized as a transparent
unit beneath a strong echo and overlying a sedimentary sequence.
In contrast to the matrix-rich debris avalanche facies, the distal
facies lacks the characteristic hummock facies typical of debris
avalanches. To estimate the thickness of the deposits it is necessary
to use a velocity for the sediments, being a common value for debris
avalanche sediments between 1600 and 1800 m/s (eg. Urgeles et al.,
1997; Collot et al., 2001). We used a velocity of 1500 m/s based on
Hamilton's formulas (1985) for other volcanic forearcs in the Pacific
Ocean, being a good choice for the top 40 m. The minimum thick-
ness of the deposit is about 2 m in areas with somewhat greater
slope. The deposit is 5e10 m thick in the areas where we recog-
nized a reflection beneath the transparent acoustic facies, but in
many areas it may be thicker, because the CHIRP system did not
image a reflection beneath the deposit.
The TowCam images across the distal deposits show that the
seabed exposes areas of soft sediment, as indicated by the presence
of bioturbation and the lack of sessile epifauna, and areas with
430 S. Day et al. / Marine and Petroleum Geology 67 (2015) 419e438
Figure 12. Morphological interpretation of Ritter insular margin and adjacent areas. See inserted legend for interpretation of the different morphological features.
abundant dark, rounded clasts 10e20 cm across partially covered
by lighter sediment (Fig. 10h,i). In the transitions from areas with
low to high clast abundance, we observed clast alignments and
concentrations in ridges. The clast alignments are oriented SW to
NE on the seabed (Fig. 10h) parallel to a fabric seen in the back-
scatter intensity data for the same area. The dredge indicates that
the clasts are cohesive silt-grade sediment (Fig. 10j, see Fig. 3 for
location of the dredge). The size of the clasts in the dredge is
comparable to the size of the dark clasts in the camera images.
Comparison with the sizes of the scallops seen in the grooved
terrain and an inferred similarity between the silt-grade sediments
in the clasts found in the dredge and the cohesive sediments seen in
the grooved terrain and the walls of the sediment mesas suggests
that the distal deposits contain a high proportion of sediments
eroded by the landslide. In contrast, only four small lithic clasts
were present in dredge sample: a reworked dacitic pumice pebble,
two clasts of andesitic and dacitic composition and a basaltic clast
with Mg-rich olivine and chrome spinel phenocrysts. This last clast
is more primitive in composition than any of the basalts known
from the adjacent volcanic islands (Johnson et al., 1972).
Near the northwestern margin of the distal deposit we mapped
a series of features in the backscatter that were recognized as
sediment waves (Figs. 3 and 12; Hoffman et al., 2008). Varied
backscatter strength defines the eastern edge of the deposit. The
orientation of some features, including a small field of sediment
waves, suggest currents associated with Dampier Strait to the SE
are responsible for transporting sediment and reworking the sea-
floor in this region (Fig. 3).
5.3.2. Erosional features
No erosional features were observed over the main area of the
distal deposit. However, its southern margin extends along the foot
of the line of sediment mesas discussed in the previous section,
while its western margin is located at the foot of a scarp some tens
of meters high eroded into marine sediments that was partly
imaged in TowCam 4 (Fig. 3) crossing that margin. Scalloped fea-
tures at the crest of the scarp, similar to those seen in the grooved
terrain, were imaged in TowCam 4 (Fig. 10k). It is therefore possible
that the southern part of the distal deposit covering an area
15e20 km (EeW) by about 10 km (NeS) lies within an area where
mass erosion or failure of sediment in the substrate excavated the
seafloor to a depth of up to 20e40 m (the height of the highest
sediment mesas above their surroundings).
6. Volume estimates
6.1. Volume of the scar
The initial volume of a volcanic island flank collapse can be
estimated, where a well-defined escarpment or amphitheater ex-
ists, by subtracting a grid of the present topography from a grid of
the calculated pre-collapse surface. The latter is usually obtained by
extrapolating the topographic and bathymetric contours across the
embayment (e.g., Urgeles et al., 1997). For the Ritter collapse, where
most of the subaerial part of the island and the inferred convex
submarine flank has disappeared in a single event, it is essential to
restore the pre-collapse topography.
 S. Day et al. / Marine and Petroleum Geology 67 (2015) 419e438 431

Figure 13. A) Topographic and bathymetric reconstruction of Ritter Island and its insular margin before its collapse on March 13th, 1888. B) Topography of Ritter Island and insular
margin after its collapse but without the post-collapse volcanic vents. C) Thickness of displaced slide mass, obtained by subtracting the latter to the former reconstructions.

Restoration of the pre-collapse topography of Ritter together
with an extrapolation of the submarine collapse scar as a single
curved surface approximates the initial pre-collapse shape of the
island and its submarine failed flank (Fig. 13a). To create the pre-
collapse grid it was necessary to remove the landslide blocks, the
post-collapse volcanic eruptions and to smooth the erosion of the
submarine valley assumed to have been excavated during the
event. Subtracting the reconstructed slide surface (Fig. 13b) from
the reconstructed pre-collapse topography we obtained a total
volume of 4.2 km3 for the slide mass involved in the collapse of
Ritter. The contoured difference between these two surfaces used
to calculate the volume is shown in Figure 13c. This also provides
the initial wedge-and-cone shape of the landslide mass at the start
of its movement.
6.2. Volume of the deposits
In comparing collapse scar and deposit volumes, it is important
to recognize the occurrence of dilation during failure and landslide
movement. The deposits of the well constrained 1980 Mount St.
Helens debris avalanche were estimated to have experienced
around 20% dilation over their pre-collapse volume (Glicken, 1996).
Gridding the inferred thickness of the large block facies in the
avalanche within the limits of the Ritter deposit as mapped pro-
vides a volume of 4 km3. By correcting the latter value for 20%
dilation during the collapse and using Glicken's method, we
conclude that a volume of 3.2 km3 of the initial collapse material
would have been necessary to form the large block debris
avalanche facies. Since we have obtained a volume of 4.2 km3 for
432 S. Day et al. / Marine and Petroleum Geology 67 (2015) 419e438

Table 2
Main characteristics of selected landslides of different sizes for comparison of average thickness with Ritter landslide. Silver et al. (2009) calculated a thickness of 46 m for the
large block part of the avalanche. In this work we obtained an average of 7e10 m of thickness for the distal deposit. The available data does not allow us to estimate the
thickness of the matrix rich part of the landslide.

Landslide Location Setting Area Volume Average thicnkess Age References

2 3
Ritter Ritter Island, Volcanic Arc 100 km 4.6 km 46 m year 1888 Silver et al. (2009)
Papua New Guinea
2 3
El Golfo Debris Avalanche El Hierro Island, Oceanic island 1.500 km 150e180 km 100 m 13e17 ka Masson et al. (2002)
Canary Islands
2 3
Canary Debris Flow El Hierro Island, Oceanic island 40.000 km 400 km 10 m 13e17 ka Masson et al. (2002)
Canary Islands
Saharan Debris Flow El Hierro Island, Oceanic island 48.000 km2 1100 km3 23 m 60 ka Masson et al. (2002)
Canary Islands
Guimar Debris Avalanche Tenerife Island, Oceanic island 1.600 km2 120 km3 75 m 780e840 ka Masson et al. (2002)
Canary Islands
2 3
Anaga Debris Avalanche Tenerife Island, Oceanic island 502 km 36 km 72 m >600 ka Llanes et al. (2003)
Canary Islands
Cumbre Nueva Debris Aval. La Palma Island, Oceanic island 780 km2 95 km3 122 m 125e536 ka Masson et al. (2002)
Canary Islands
OshimaeOshima Debris Aval. Japan Sea Volcanic arc 69 km2 2.5 km3 36 m year 1741 Satake and Kato (2001)
Socompa Andean Volcanic Chain Stratovolcano 490 km2 >15 km3 31 m 10.000e500 yr Francis et al. (1985)
St. Lucia Debris Avalanche Lesser Antilles Arc Volcanic arc 2000 km2 hundreds km3 180e 225 m <100e200 ka Deplus et al. (2001)
Dominica Debris Avalanche Lesser Antilles Arc Volcanic arc 3500 km2 hundreds km3 180e 225 m <100e200 ka Deplus et al. (2001)
Martinique Debris Aalanche Lesser Antilles Arc Volcanic arc 800 km2 hundreds km3 180e 225 m <100e200 ka Deplus et al. (2001)
Kezymyannyi Debris Aval. Kamchatka Stratovolcano 36 km2 0.5 km3 30e14e9 m year 1956 Belousov and Belousova
(1998)
Nuuanu Debris Avalanche Oahu Island, Oceanic island 23.000 km2 5.000 km3 >2000 m 2.1e2.2 My Moore et al., 1998 and
Hawaii Islands McMurtry et al. (2004)
(age)
Storegga Debris flow Norwegian Continental 4.600 km2 5.000 km3 1.000 m 8.100 y BP Haflidason et al. (2005)
continental margin

the Ritter Island subaerial and submarine scar, we infer that a frequency CHIRP seismic data taken over the distal facies (Fig. 11)
fraction of about 76% of the initial landslide material never got show thicknesses of 7e10 m for much of the region for which we
beyond the constriction in the valley between Sakar and Umboi have data. If we assume a conservative average thickness of 8 m, the
Islands, and is now emplaced as the large block facies deposit. Only minimum volume of the distal deposit should be 6.4 km3. This
about a quarter of the initial landslide material from Ritter itself conservative value would imply that at most 15% of the distal de-
reached farther downslope to contribute to the matrix-rich debris posit is composed of lithic clasts originating from the initial
avalanche facies and to the distal deposit. collapse of the Ritter Island edifice, and likely less because of the
We also examined other reconstructions of pre-collapse finite although unknown volume of the matrix-rich debris
topography. Our most unrealistic assumptions did not result in avalanche facies deposit. It follows that the small part of the initial
more than a factor of 1.5 change to our resulting estimate of slide landslide that reached as far downslope as the mouth of the valley
volume, but did result in a factor of 2 change in our estimate of large between Sakar and Umboi, and the basin plain beyond, was highly
block facies deposit volume. The latter is in itself unrealistic, since efficient in eroding and mobilizing the sediment substrate over
there is an upper limit on the physically possible porosity of the which it moved.
large-block facies deposit, at about 35%, that also limits its dilation.
Therefore, we consider it very likely that the true slide volume falls
7. Discussion
between 0.8 and 1.25 times our preferred estimate, and the true
deposit volume falls between 0.8 and 1.5 times our estimate.
7.1. Instability and triggering of edifice collapse
However, quantifying the actual uncertainty in our estimate is
beyond the scope of available data.
Our data do not provide direct evidence on the processes that
Estimates of the volume of the matrix rich debris avalanche
triggered the March 13th 1888 lateral collapse of Ritter Island, but
facies deposit and of the volume of material eroded by the passage
they provide potential controls on the geometry of the collapse and
of the landslide before its emplacement are uncertain because none
hence factors that may have made the edifice susceptible to failure.
of the chirp lines provided deposit thickness in this area. We infer
The collapse differs from most subaerial volcano collapses in that
that most of the distal deposit is largely not from the initial collapse
the scar cuts down to the base of the edifice and perhaps into the
but rather from entrained sediment, which is consistent with the
substrate. Subaerial lateral collapses typically cut part way down to
low abundance (<1%) of lithic clasts in the dredge sample. Esti-
the base of the pre-collapse edifice, like the 1980 Mount St. Helens
mation of the volume of the distal deposit is complicated by the
collapse (Siebert et al., 1987). In a few subaerial volcanoes and other
sediment cover on the eastern margin and by its uncertain north-
submarine or island volcanoes, the failure occurs at the base of the
ern limit. It is possible that the distal deposit extends out of the
edifice or in the substrate. At the Socompa volcano, it has been
northern limit of backscatter mapping (Fig. 3). Within the mapped
argued that the reason for failure at the base of the volcano lies in
region the deposit area is about 800 km2.
pre-collapse deformation of weak, water-saturated sediments at
The thickness of the Ritter debris avalanche deposit varies
the base of the edifice (Francis et al., 1985; Van Wyk de Vries et al.,
within the mapped region. Silver et al. (2009) estimated an average
2001). This deformation is manifested in a series of anticlines
thickness of 46 m for the block-rich part of the debris avalanche
extending beyond the foot of the volcano, the crests of which form
deposit. For comparison, examples of thickness estimated by other
ridges that now project through the landslide deposit (Van Wyk de
authors in selected landslides can be found in Table 2. High
Vries et al., 2001). The Ritter volcano was most likely built upon
 S. Day et al. / Marine and Petroleum Geology 67 (2015) 419e438
marine sediments that could have provided a weak basal layer that
localized the failure at the base of the volcano, although we did not
find any direct evidence of pre-collapse deformation in the sur-
viving submarine flanks of the volcano. Other examples of vol-
canoes cutting down to the base of the edifice are the failure of
Unzen in 1792 (Siebert et al., 1987) and the 1888 collapse scar at
Bandai, that also cuts unusually deep into the edifice (Siebert et al.,
1987). Both of these lateral collapses were associated with phreatic
hydrothermal explosions, similar to those at Ritter before its
collapse, and hot groundwater discharges, rather than magmatic
eruptions.
There is no evidence that tectonic triggers such as large-scale
fault movements or gravitational settling-related faulting initi-
ated the Ritter landslide. Environmental factors such as high-
intensity rainstorms, which can cause a reduction in strength of
the edifice due to saturation (McGuire, 2003), were also absent. An
example of this is the recorded aseismic movement of the Hilina
Slump coinciding with a major rainstorm (Cervelli, 2002). The
collapse lacked an associated large-volume magmatic eruption, but
was likely associated with phreatic or phreatomagmatic explosions,
because observations of vapor emission at Ritter were reported
only a week before the collapse (Cooke, 1981), indicating that a
shallow hydrothermal system was present within the edifice
(Siebert et al., 1987). The Bandai and Unzen types of lateral collapse
may be triggered by pore fluid pressurization deep within the
edifice due to fluid input and/or heating from magmatic intrusions
emplaced well below the base of the edifice, rather than within the
edifice as in the case of the 1980 Mount St. Helens cryptodome
(Day, 1996; Reid, 2004). In these cases, the locus of pressurization,
deformation and failure may be in low permeability layers near or
at the base of the volcano: this could be another mechanism by
which a substrate of compacted deep marine sediments could ac-
count for the geometry of the 1888 Ritter Island collapse structure.
There is no geological evidence in our data for multiple or
retrogressive failures of the Ritter Island volcano within the 1888
Figure 14. Photograph and accompanying interpretation of the scouring walls of Coldwater creek eroded by the fast moving lobe of the 1980 Mount St. Helens debris avalanche
(taken in 2004). We interpret that the scouring mark found in the slope of Umboi Island was formed by the fast moving lobe of the Ritter avalanche, analogous to the scouring of
Coldwater creek.
433
collapse event, but we cannot exclude the possibility that lateral
collapse at Ritter involved successive blocks beginning to move tens
of seconds apart. Retrogressive failures with blocks beginning to
move a few tens of seconds apart and successive failures a few
minutes apart have been directly observed by eyewitnesses at
Mount St. Helens in 1980 (Voight, 1981) and Stromboli in 2002
(Bonaccorso et al., 2003) respectively, but the effects in the deposits
are subtle and ambiguous (Glicken, 1996; Ward and Day, 2006).
Evidence that the collapse at Ritter Island began as a single failure
or as a rapid sequence of retrogressive failures no more than a few
tens of seconds apart, is principally provided by the single tsunami
that it produced (Cooke, 1981; Ward and Day, 2003).
7.2. Erosion by the landslide
7.2.1. Scouring
The undulating high backscatter feature found on the slope of
Umboi Island (Fig. 5) lacks a distinct bathymetric expression and
the upslope boundary of the very high backscatter zone has no
expression in the chirp profile that crosses it, unlike the boundaries
of the other deposits that we associate with the 1888 Ritter Island
collapse in this paper. Similarly, it appears unlikely that it is a de-
posit originating from farther upslope, since all other such deposits
evident in the backscatter data from Umboi and also Sakar trend
down the slope. Furthermore, it is not an artifact due to the angle
relative to the sonar beam, because in that case the upslope of the
contact would then have the higher backscatter not the lower one.
Instead, in the light of the undulating nature of the upslope contact
and the way it passes upslope over of the small ridge, an alternative
interpretation is that it is an erosional scour feature comparable to
the scouring of the walls of Coldwater creek by a fast-moving lobe
of the 1980 Mount St. Helens debris avalanche (Voight et al., 1981;
Fisher et al., 1987; Glicken, 1996). The two are shown in Figures 5
and 14 respectively.
434 S. Day et al. / Marine and Petroleum Geology 67 (2015) 419e438
Such an interpretation also explains why a similar undulating
backscatter feature is not seen at the same elevations on the slopes
of Sakar, as would be the case if it was a “high tide” mark reflecting
the maximum thickness of a slow-moving landslide. The 1888
collapse scar is orientated oblique to the valley, pointing more to-
wards Umboi and away from Sakar. Therefore, as it moved west-
ward at high velocity, the leading part of the landslide would have
obliquely impacted the side of Umboi, running up the slope and
reaching its greatest height as it passed over small ridges on the
slope. Similar deflections of the scour line over small ridges are
seen in Coldwater Creek (Fig. 14), providing a further point of
similarity between the two features.
7.2.2. Black Streak
The very high intensity backscatter character of this unit,
together with its differential relief in the bathymetry and the uni-
formity of the size clasts imaged with the photographs (in strong
contrast to the wide variation in clast sizes seen in debris avalanche
deposits, but typical of scoriaceous volcanic deposits), lead us to
interpret the Black Streak as a volcanic unit. We infer that this
feature is the core of an in-situ submarine volcanic vent whose top
was eroded off by the leading part of the landslide.
7.2.3. Mesas
The development of scarps and troughs in the mesas located to
either side of the distal lobes of the matrix-rich debris avalanche
deposit (M in Fig. 9) lead us to interpret them as erosional features
related to a flow moving from SE to NW, along the groove trend
seen to the SE. The formation of the mesas must predate
emplacement of the matrix-rich debris avalanche facies lobes in the
troughs between the mesas, but the association between them
indicates that they are both related to the Ritter landslide event.
However, the mesas and troughs located along the SW margin of
the matrix rich facies deposit (M0 in Fig. 9, Line 3/1 in Fig. 10) may
not have originated by a flow moving down-slope, but instead they
may be the result of uneven upslope retreat of scarps developed at
the lateral margin of the moving debris avalanche, as a result of
undermining the foot of the slope by erosion along the valley axis,
followed by development of progressive slumping upslope. We
consider the formation of the other mesas further below in the
context of the formation of the distal debris flow (section 7.4).
7.2.4. Grooved terrain
The TowCam images discussed in section 5.2.2 indicate that the
grooved terrain was incised into weak but cohesive marine sedi-
ments. Similar scalloped, but overall linear, grooves eroded in
cohesive water-saturated sediment have been produced experi-
mentally by high velocity water flows simulating tidal flows over
littoral muds (Winterwerp and Van Kesteren, 2004). The grooves
imaged in our work are very similar to those produced in these
experiments in the intermediate-stress bulk erosion regime. We
infer that the grooved terrain was generated by bulk erosion of
older marine sediments at the base of a high velocity component of
the Ritter debris avalanche that extended over a wider front than
that part which deposited the matrix-rich debris avalanche deposit.
This inference is consistent with the occurrence in the photo-
graphed groove floors of sparse, scattered lithic angular clasts a few
tens of centimeters across, very similar to the clasts in the matrix
rich debris avalanche facies (Fig. 10); and of the isolated mound
features, several tens of meters across and with high backscatter
intensity, that are also present in the grooved terrain and which we
interpret as isolated debris avalanche blocks.
Similar grooved features have been described in the continental
slope off New Jersey (Farre and Ryan, 1985) and in Lake Superior
(Viekman et al., 1992). However, those others were much more
regular and linear than the ones that we observe in the Ritter, and
are at least partly produced by spatial variations in sedimentation.
The latter is indicated by consistent development of coarser sedi-
ments in the furrows, contrasting with finer grained sediment in
the inter-furrow areas; and by development of rippled sediment on
the furrow margins in some cases. In contrast, the Ritter grooves
have steep and scalloped walls exposing bedded cohesive sedi-
ment. The only distinctive sediment in the groove floors of Ritter,
draped with later sediment, is formed by a scattering of lithic clasts
comparable to blocks in the imaged hummocks in the matrix rich
facies deposit of the landslide.
7.3. Inferred landslide kinematics
The 1888 collapse of Ritter left a deposit composed of three
different facies, a division that implies three distinct stages in the
movement and rheological evolution of the landslide. As was done
by Glicken (1991, 1996) for the 1980 Mount St Helens collapse de-
posits, we use the distribution of these facies and the different
types of erosional structure, together with the overall geometry of
the deposit to reconstruct the movement and evolution of the
Ritter landslide.
The grooved terrain and the sediment mesas indicate the
occurrence of extensive erosion, of at least the top 20 m of the
existing sediment sequence in these areas. The association of the
grooved terrain with isolated debris avalanche blocks, the occur-
rence of scattered lithic clasts in the smaller grooves imaged with
the TowCam, and the channelization of the debris avalanche lobes
between the sediment mesas, indicate that the erosion took place
during the landslide, and that deposition of the matrix rich facies
debris avalanche deposit took place after the substrate erosion. We
conclude that the erosional features were produced by the fast
moving leading part of the landslide that, after being deflected off
the submarine slope of Umboi in a “bobsled” fashion similar to that
seen in some subaerial landslides in mountain valleys (Evans et al.,
1989), reached beyond the constriction and spread out over a front
up to 10 km wide (corresponding to the width of the grooved
terrain and the line of sediment mesas). The distribution of the
grooved terrain and the sediment mesas indicate that it was this
part of the landslide that generated the substrate erosion and the
debris flow. In contrast the matrix rich debris avalanche was
deposited from the slower, trailing part that was confined to the
axis of the valley between Umboi and Sakar.
The landslide created a tsunami that devastated the coast of
Sakar and Umboi Islands to more than 15 m of elevation (Cooke,
1981) implying high landslide velocities (Ward and Day, 2003).
Run-up of landslides onto valley walls can be also used to estimate
landslide velocities, by equating the potential energy gained in the
run-up to the minimum kinetic energy of the moving material at
the base of the slope. This argument was applied to landslides by
Heim (1882) and had notable applications to the debris avalanche
of Mount St. Helens (Voight et al., 1983) and to the Osceola mudflow
at Rainier (Vallance and Scott, 1997). In Ritter, some landslide ma-
terial ran up onto the submarine slope of Umboi, implying that it
had sufficient kinetic energy to climb uphill. Assuming that the
general level of the landslide was at 900 m and, applying this cri-
terion, it suggest that the material that climbed up to 200 m over
the small ridge (Fig. 4) would have had a speed of 63 m/s. This is
actually a high estimate because of the fact that it neglects the
particleeparticle interaction within the debris avalanche that
would tend to push particles up the slope. Ward and Day (2003)
obtained a more conservative average velocity of at least 40 m/s
for Ritter landslide from their tsunami modeling. However, the
63 m/s estimated velocity is not necessarily that of the landslide as
a whole, since it is likely to have fragmented much earlier in its
 S. Day et al. / Marine and Petroleum Geology 67 (2015) 419e438
motion and the front may have been traveling significantly faster
than the tail.
7.4. Transformation from debris avalanche to debris flow
The SWeNE orientation of clast alignments seen on the seabed
of the distal facies (TowCam line 3, Fig. 3) parallel to a fabric seen in
the backscatter intensity data, probably follows the landslide
movement direction in this area and indicates pervasive shearing
during deposition (Fig. 10g). These clast alignments may be anal-
ogous to ridges developed near the margins of experimental water-
saturated debris flows (Major, 1997) and to longitudinal shear
structures identified in backscatter intensity data (but not imaged
with photographs) from the Saharan debris flow by Masson et al.
(1993). These morphological similarities, together with the thin
margins of the deposit, overall low thickness/length and thickness/
width ratios of the deposit and the low slope angle of the plain
upon which the distal facies deposits were emplaced, indicate that
the distal facies was emplaced as a high mobility debris flow or a
series of flow lobes.
Debris flow deposits extending downslope from debris
avalanche deposits or other types of landslide have been identified
in many studies (for example, Piper et al., 1999; Gee et al., 1999;
Scott et al., 2001), implying rheological transformations. These
transformations often occur by erosion and entrainment of water
rich sediments, glacier ice and other debris from the substrates over
which the landslides move. Where there is abundant substrate that
is easily entrained, as in the floors of glaciated valleys, the total
volume of both volcanic and non volcanic landslides can increase
by more than 1000% over the initial failure volume (Plafker and
Ericksen, 1978; Scott et al., 2001; Erismann and Abele, 2001; Kaab
et al., 2003), a process known as bulking (Scott et al., 2001). Bulk-
ing increases the total landslide runout distance and the landslide
mobility, given by the decrease in drop height/run-out distance
ratio and commonly interpreted as effective friction coefficient.
Understanding the development of landslide facies and the role of
bulking is therefore key to the understanding of landslide
mechanics.
The transformation of the leading part of the Ritter debris
avalanche into the debris flow was also related to substrate erosion
and entrainment, as suggested by the dominance of sediment
intraclasts in the dredge sample and in the TowCam images, in
contrast to the lithic clast dominated mounds seen in the matrix
rich debris avalanche facies. The exposed areas of substrate erosion
in the grooved terrain and the sediment mesas upslope of the
present transition from debris avalanche deposit to debris flow
deposit are likely sources of some of the entrained sediment. In
addition, the scarps along the western margin of the deposit indi-
cate that another source is the 20 km by 10 km area bounded by the
scarp to the west and the sediment mesas to the southeast and now
covered by the upslope, proximal parts of the debris flow deposit
itself (Fig. 12). The lack of distinct raised sediment mesas within the
latter area implies that such erosion may have involved complete
removal of the substrate by mass failure, the most complete and
highest shear stress regime mechanism of substrate erosion iden-
tified by Winterwerp and Van Kesteren (2004).
We identify two possible causes for the transition from mass
erosion to mass failure. First, the coincidence of this change from
mass erosion in the grooved terrain, to mass failure around and to
the N of the sediment mesas, with the margin of the near-flat plain
on which the debris flow was deposited, may be due to a transition
from stronger or coarser sediment on the submarine slopes of
Umboi and Sakar to weaker, finer grained sediment on the plain
that was more susceptible to failure under the impact of shear
stresses from the fast-moving leading part of the landslide.
435
Alternatively, Talling et al. (2007) showed how debris flows
could deposit very little sediment and be locally very erosive in
large areas far from the sources, and then sediment deposition
could be triggered by a remarkably small but abrupt decrease in
sea-floor gradient from 0.05 to 0.01. In this case, our data show
that there is an order of magnitude reduction in slope between the
area of debris avalanche (1.6e2.1 ) and the area of debris flow
deposition (0.1 ).
The coincidence of the facies transition and the break in slope
could indicate that the debris-flow deposition started due to rapid
deceleration of the flow, because deceleration reduces the turbu-
lent mixing and sand settles more rapidly, as interpreted by Talling
et al. (2007); however, this process does not explain the onset of
mass failure at the break in slope. The two mechanisms may not be
mutually exclusive, because conservation of momentum requires
that a sudden bulking of the landslide due to mass failure of the
substrate would be accompanied by a reduction in its velocity.
Finally we note that the distal end of the landslide may have
extended farther north and out of the area that we mapped, further
enhancing the contrast in mobility between the debris avalanche
facies and the debris flow facies. Debris flow and turbidite deposits
associated with 1888 Ritter collapse may be found farther to the
north and west in the Bismarck Sea Basin than the area mapped,
similarly to what has been identified associated with the 1929
Grand Banks continental slope failure (Piper et al., 1985) and with
the Canary (Talling et al., 2007; Hunt et al., 2011) and Hawaii
(Kanamatsu et al., 2002) landslides.
7.5. Comparison of the submarine and subaerial collapse deposits
Comparison of submarine and subaerial debris avalanche de-
posits is not straightforward, but instead fairly difficult and un-
common because: 1) The submarine deposits have to be really fresh
and largely unsedimented, therefore from recent events. In older
deposits the sediment drape obscures small-scale features and only
large blocks are seen projecting above the seabed. For example, if
we compare the Ritter deposits with the older debris avalanche
deposits that Silver et al. (2009) found along the Bismarck arc, none
of them have matrix-rich facies deposits or distal debris flow units
exposed on the seabed, probably because they are buried. Also, in
older deposits often the collapse scar itself has been buried by
subsequent volcano growth, or grossly modified by erosion, not
allowing seeing the proximal area deposits. 2) Studies of subaerial
deposits commonly focus on small-scale (<~10 m across, or <100 m
across) features exposed both on deposit tops in recent deposits
and in cross-section in outcrops incised into older deposits. There is
therefore a scale disconnecting these outcrop scale studies and
both the large-scale (typically remote sensing based) studies of
morphologically fresh deposits such as Socompa and especially
Mount St. Helens 1980, with our data set. 3. Comparisons should be
made among subaerial and submarine deposits of either valley-
confined or unconfined debris avalanches. In subaerial studies,
there is a clear recognition that valley-confined debris avalanche
deposits show a range of morphological differences to laterally
unconfined debris avalanche deposits (e.g. Glicken, 1991 for a
review).
The Mount St. Helens 1980 and the Ritter 1888 are a great
subaerial/submarine pair of deposits to compare because they are
the best exposed recent cubic-kilometer volume class deposits in
their respective categories, they are both valley-confined and of a
similar scale. Even the height differences between upslope and
downslope deposit limits are very similar: Large block facies:
1150 me700 m elevation at Mount St. Helens, 800 m water depth to
1200 m water depth at Ritter; Matrix rich facies: 700 me350 m
elevation at Mount St. Helens, 1200 m water depth to 1700 m water
436 S. Day et al. / Marine and Petroleum Geology 67 (2015) 419e438
depth at Ritter. The transition may be so sharp in both cases
because of the involvement of a constriction in the valley walls. The
difference between the two is the distal debris flow at Ritter, that
we believe is due to the contrast between substrates: almost
entirely bedrock at Mount St. Helens, whereas the Ritter debris
avalanche transforms into the debris flow as soon as it encounters a
thick sequence of easily eroded basin-floor sediments on a flat
basin floor. This could be evidence of a hydraulic jump or it might
be due to a change in the substrate or it might be due to a poorly
understood change in the dynamics of the debris avalanche as the
slope changes, and we do not have data to support either of those
possibilities.
Our detailed analysis of the Ritter landslide deposits revealed
that if looking at its broad aspects, they are not that different as the
deposits of the Mt. St. Helens event. Maybe some retarding effect on
the initial movement could be expected in Ritter, since the tsunami
generation represents a transfer of energy to the water column,
which doesn't occur subaerially, air being much less dense. How-
ever, Ward & Day (2003) showed that only a small fraction, about
8%, of the potential energy released by the landslide, is required to
generate the observed Ritter Island 1888 tsunami. At the high but
subcritical (Froude number <1 but not <<1) velocities inferred for
the landslide in that paper, this wave-making energy will be the
dominant component of the energy transferred from the landslide
to the water column. All of the rest of its energy is dissipated in
internal work (fragmentation of the landslide), substrate erosion
and entrainment, and basal friction. These latter processes are
common to both subaerial and submarine landslides, hence they
would produce broadly similar deposits if they involved similar
materials and initial geometries. Work done in substrate erosion
and entrainment is dependent upon the nature and erodibility of
the substrate and the main difference is that submarine landslides
are much more likely to encounter easily eroded substrates; hence,
in the Ritter example, the presence of the distal debris flow deposit
with abundant sediment intraclasts, in contrast to the minimal
development of entrainment and debris flow deposits at Mount St.
Helens.
An interesting comparison between the deposits of subaerial
and submarine deposits comes from looking at if the block fields
are any different or indistinguishable in both cases. For instance,
the blocks of the Canary island landslide deposits were found to be
significantly larger than the blocks seen in subaerial collapses
(Mitchell et al., 2002). The latter authors explained as a function of
ambient water pressure, because opening a block rapidly would
involve work against hydrostatic pressure, while slow opening, in
contrast, would allow water to permeate and pressurize fluid in
cracks. The largest blocks in the large block facies of the Ritter Is-
land deposit are an order of magnitude larger than those in the
Mount St Helens 1980 deposit, but are comparable to the largest
blocks seen in other subaerial stratovolcano lateral collapse de-
posits, such as Socompa (van Wyk de Vries et al., 2001). It would be
expected that a broad correlation should exist between block size
and the dimensions of lateral collapses, not only because the latter
limit the maximum possible sizes of blocks but perhaps also
because rates of change in slope within and at the lip of the collapse
scar, which affect the bending forces that blocks experience as they
go over these changes in slope, may be less in the larger collapse
structures. However, it seems that there is a very wide scatter in the
ratio between maximum block size and collapse scar dimensions
both in subaerial and submarine lateral collapses of different sizes.
This may reflect both the lithologies and strength characteristics of
the materials making up the slide mass, but also the pore fluid
pressure distribution within it (Day, 1996), which may differ sys-
tematically between stratovolcanoes and ocean island volcanoes
(with, typically, rift zone lateral collapses). It will be hard to assess if
submarine landslides tend to create larger block sizes than sub-
aerial ones until there are many more detailed descriptions of block
size distributions in submarine volcano lateral collapse deposits
from island arc stratovolcanoes.
8. Conclusions
1) The collapse of Ritter Island in 1888 removed most of the island
and was probably associated with phreatic or phreatomagmatic
explosions. Based on our mapping, the collapse scar cut down
into the base of the edifice. We conclude that the collapse
resulted from a single event because we found no field evidence
either onshore or offshore that more than one collapse occurred
then or in the past. Historical accounts and tsunami modeling
also are consistent with a single event.
2) The deposits of the landslide generated by the collapse compose
three distinct facies, distinguished by their morphological
characteristics in our various datasets: large block debris
avalanche, matrix-rich debris avalanche and debris flow distal
facies. The transitions from large block to matrix-rich facies and
from the latter to distal facies are related to major changes in the
slopes and to the constriction in the valley at the western end of
the large block facies deposit.
3) TowCam photographs of the sea-bottom provided important
constraints on the acoustic imaging and are very useful for
studying landslide facies in subdued areas. Characteristics seen
in the photos provided valuable information for mapping the
facies of the avalanche including: morphology of the deposit,
clasts aspect, distribution and abundance, presence/absence and
type of benthic organisms indicative of substrate strength, and
signs of erosion such as scalloped walls.
4) Volumetric calculations based on the restoration of a modeled
original topography led to a total volume estimate of 4.2 km3 for
the initial Ritter Island collapse. After correcting for dilation, we
infer that about 75% of the material from the initial collapse now
forms the large block facies at distances of less than 12 km from
the collapse scar, filling an area between the islands of Umboi
and Sakar and topographically constricted by volcanic ridges
from those islands. The volume of the matrix-rich facies is un-
known, but the distal debris flow facies deposits have a mini-
mum volume of 6.4 km3, as a result of bulking of the landslide
through sediment entrainment. This bulking highlights the ef-
ficiency of substrate erosion during the later history of the
landslide movement, when much of its original volume had
already been deposited in the large block facies deposits
upslope.
5) Recent and well-exposed submarine landslide deposits can be
interpreted in terms of morphological facies and transitions in
similar ways to subaerial volcano collapse deposits. In turn,
these deposits can provide a basis for the interpretation of older
and incompletely exposed landslide deposits. Understanding
the mechanisms and kinematics of collapse events can improve
tsunami modeling.
Acknowledgments
This work was supported by National Science Foundation grant
OCE-0327004 to Eli Silver and Steve Ward, and grant OCE-0328278
to Neal Driscoll. Participation of Pilar Llanes was possible thanks to
grants “Becas del Amo 2008” from Universidad Complutense de
Madrid and “Movilidad para Profesores 2011” from Caja Madrid
Foundation. We thank Dan Fornari for the use of TowCam. We are
also grateful for the advice and support of many people in carrying
out this research, including Bruce Appelgate, Jennifer Engels, Steve
Saunders, Hugh Davies, James Robins, and Amilia Lyons. We are
 S. Day et al. / Marine and Petroleum Geology 67 (2015) 419e438
indebted to the officers, crew, and scientific party of the R/V Kilo
Moana for their cooperation at sea. Figures 1, 3, 4, 5, 8 and 13 were
drawn using GMT software (Wessel and Smith, 1995). Constructive
reviews of Dr. N. Mitchell and an anonimous reviewer have
improved this paper.
References
Beget, J.E., Kowalik, Z., 2006. Confirmation and calibration of computer modeling of
tsunami produced by Augustine Volcano, Alaska. Science of Tsunami Hazards
24, 257e266.
Belousov, A.B., Belousova, M.G., 1998. Bezymyannyi eruption on March 30, 1956
(Kamchatka): sequence of events and debris avalanche deposits. Volc. Seis. 20,
29e47.
Bonaccorso, A., Calvari, S., Garfi, G., Lodato, L., Patane, D., 2003. Dynamics of the
December 2002 flank failure and tsunami at Stromboli volcano inferred by
volcanological and geophysical observations. Geophys. Res. Lett. 30 (18), 1941.
http://dx.doi.org/10.1029/2003GL017702.
Boudon, G., Le Friant, A., Komorowski, J.-C., Deplus, C., Semet, M.P., 2007. Volcano
flank instability in the Lesser Antilles Arc: diversity of scale, processes, and
temporal recurrence. J. Geophys. Res. 112, B08205. http://dx.doi.org/10.1029/
2006JB004674.
Capra, L., Macías, J.L., Scott, K.M., Abrams, M., Gardun ~ o-Monroy, V.H., 2002. Debris
avalanches and debris flows transformed from collapses in the Trans-Mexican
Volcanic Belt, Mexico d behavior, and implications for hazard assessment.
J. Volcanol. Geotherm. Res. 113 (1e2), 81e110.
Carracedo, J.C., Day, S.J., Guillou, H., 1999. Quaternary collapse structures and the
evolution of the western Canaries (La Palma and El Hierro). J. Volcanol. Geo-
therm. Res. 94, 169e190.
Carracedo, J.C., Badiola, E.R., Guillou, H., de la Nuez, J., Perez Torrado, F.J., 2002.
Geology and volcanology of La Palma and El Hierro, Western Canaries. Estud.
Geol. 57, 175e273.
Cervelli, P., Segall, P., Johnson, K., Lisowski, M., Miklius, A., 2002. Sudden aseismic
fault slip on the south flank of Kilauea volcano. Nature 415, 1014e1018.
Collot, J.Y., Lewis, K., Lamarche, G., Lallemand, S., 2001. The giant Ruatoria debris
avalanche on the Northern Hikurangi Margin, New Zealand: result of oblique
seamount subduction. J. Geophys. Res. 106 (B9), 19271e19297.
Cooke, R.J.S., McKee, C.O., Dent, V.F., Wallace, D.A., 1976. Striking sequence of vol-
canic eruptions in the Bismarck volcanic arc, Papua New Guinea, 1972e1975. In:
Johnson, R.W. (Ed.), Volcanism in Australasia. Publ. Elsevier, Amsterdam,
pp. 149e172.
Cooke, R.J.S., 1981. Eruptive history of the volcano at Ritter Island. In: Johnson, R.W.
(Ed.), Cooke-Ravian Volume of Volcanological Papers, Geological Survey of
Papua New Guinea Memoir, vol. 10, pp. 115e123.
Crandall, D.R., Miller, C.D., Glicken, H.X., Christiansen, R.L., Newhall, C.G., 1984.
Catastrophic debris avalanche from ancestral Mount Shasta volcano, California.
Geology 12, 143e146.
Crandall, D.R., 1988. Gigantic Debris Avalanche of Pleistocene Age from Ancestral
Mount Shasta Volcano, California, and Debris Avalanche Hazard Zonation. In:
United States Geological Survey Bulletin 1861, 32 pp.
Dampier, W., 1729. A Voyage to New Holland. Publ. Nonsuch Press, Gloucester.
Republished with an Introduction by J. Spencer 2006.
Day, S.J., 1996. Hydrothermal pore fluid pressure and the stability of porous,
permeable volcanoes. In: McGuire, W., Neuberg, J., Jones, A. (Eds.), Volcano
Instability on the Earth and Terrestrial Planets, Geological Society of London
Special Publication, vol. 110, pp. 77e93.
Day, S.J., Heleno da Silva, S.I.N., Fonseca, J.F.B.D., 1999. A past giant lateral collapse
and present day flank instability of Fogo, Cape Verde Islands. J. Volcanol.
Geotherm. Res. 94, 191e218.
Deplus, C., Le Friant, A., Boudon, G., Komorowski, J.C., Villemant, B., Harford, C.,
Segoufin, J., Cheminee, J.L., 2001. Submarine evidence for large e scale debris
avalanches in the Lesser Antilles Arc. Earth Planet. Sci. Lett. 192, 145e157.
Dumont d’Urville, J.S.C., 1832. Voyage de la Corvette L'Astrolabe, execute par ordre
du Roi, pendant les annees 1826e1827e1828e1829, sous le commandement de
M.J. Dumont D'Urville, Capitaine de Vaisseau, Tome Quatrieme. Histoire du
Voyage. Publ, Tastu, Paris.
Erismann, T.H., Abele, G., 2001. Dynamics of Rockslides and Rockfalls. Publ.
Springer-Verlag, Berlin.
Evans, S.G., Clague, J.J., Woodsworth, G.J., Hungr, O., 1989. The Pandemonium Creek
Rock Avalanche, British Columbia. Can. Geotech. J. 26, 427e446.
Farre, J.A., Ryan, W.B.F., 1985. 3-D view of erosional scars on U.S. mid-Atlantic
continental margin. Am. Assoc. Petrol. Geol. Bull. 69, 923e932.
Fisher, N.H., 1939. Report on the Volcanoes of the Territory of New Guinea. In:
Geological Bulletin # 2. Administration of the Territory of New Guinea.
Fisher, N.H., 1957. Catalogue of the Active Volcanoes of the World. Part V: Melanesia.
Publ. International Volcanological Association, Naples.
Fisher, R.V., Glicken, H.X., Hoblitt, R.P., 1987. May 18, 1980, Mount St. Helens de-
posits in South Coldwater Creek, Washington. J. Geophys. Res. 92,
10.267e10,283.
Fiske, R.S., Cashman, K.V., Shibata, A., Watanabe, K., 1998. Tephra dispersal from
Myojinsho, Japan, during its shallow submarine eruption of 1952e1953. Bull.
Volcanol. 59, 262e275.
437
Francis, P.W., Gardeweg, M., Ramirez, C.F., Rothery, D.A., 1985. Catastrophic debris
avalanche deposit of Socompa volcano, northern Chile. Geology 13, 600e603.
Gee, M.J.R., Masson, D.G., Watts, A.B., Allen, P.A., 1999. The Saharan debris flow: an
insight into the mechanics of long runout debris flows. Sedimentology 46,
317e335. http://dx.doi.org/10.1046/j.1365-3091.1999.00215.x.
Geographical Survey Institute, 1982. Coastal area basic survey report, Shimabara
area. Geographical Survey Institute of Japan Technical Report D-3 No. 38, text
and 1:25 000 map.
Glicken, H.X., 1991. Sedimentary architecture of large volcanic debris avalanches. In:
Fisher, R.V., Smith, G.A. (Eds.), Sedimentation in Volcanic Settings, vol. 45. So-
ciety of Economic Mineralogists and Petrologists Special Publication,
pp. 99e106.
Glicken, H.X., 1996. Rockslide e Debris Avalanche of May 18, 1980, Mount St. Helens
Volcano, Washington. United States Geological Survey Open File Report 96-677.
Gorshkov, G.S., 1970. Volcanism and the upper Mantle: investigations in the Kurile
island arc. Publ. Plenum Press, N.Y., 385 pp.
Haflidason, H., Lien, R., Sefrup, H.P., Forsberg, C.F., Bryn, P., 2005. The dating and
morphometry of the Storegga Slide. Mar. Pet. Geol. 22, 123e136.
Hamilton, E.L., 1985. Sound velocity as a function of depth in marine sediments.
J. Acoust. Soc. Am. 78, 1349e1355.
Heim, A., 1882. Der Burgsturz von Elm. Z. Dtsch. Geol. Ges. 34, 74e115.
Holcomb, R.T., Searle, R.C., 1991. Large landslides from oceanic volcanoes. Mar.
Geotech. 10, 19e32.
Hunt, J.E., Wynn, R.B., Masson, D.G., Talling, P.J., Teagle, D.A.H., 2011. Sedimento-
logical and geochemical evidence for multistage failure of volcanic island
landslides: a case study from Icod landslide on north Tenerife, Canary Islands.
Geochem. Geophys. Geosyst. 12, Q12007. http://dx.doi.org/10.1029/
2011GC003740.
Jacobs, T.J., 1844. Scenes, Incidents and Adventures in the Pacific Ocean, or the
Islands of the Australasian Seas, During the Cruise of the Clipper Margaret
Oakley Under Capt. Benjamin Morell. Publ. Harper & Bros., New York.
Johnson, R.W., Taylor, G.A.M., Davies, R.A., 1972. Geology and Petrology of Quater-
nary Volcanic Islands off the North Coast of New Guinea. Bureau of Mineral
Resources, Geology and Geophysics Record 1972/21.
Johnson, R.W., 1987. Large scale volcanic cone collapse: the 1888 slope failure of
Ritter Volcano. Bull. Volcanol. 49, 669e679.
Johnson, R.W., Workshop Organizing Committee, 1999. Volcanic Cone Collapses and
Tsunamis: issues for Emergency Management in the South West Pacific Region.
Report, IAVCEI workshop on Ulawun Decade Volcano, Papua New Guinea. 40
pp.
Kaab, A., Wessels, R., Haeberli, W., Huggel, C., Kargel, J.S., Khalsa, S.J.S., 2003. Rapid
ASTER imaging facilitates timely assessment of glacier hazards and disasters.
Eos 84 (13), 117e121.
Kanamatsu, T., Herrero e Bervera, E., McMurtry, G.M., 2002. Magnetostratigraphy of
deep-sea sediments from Piston cores adjacent to the Hawaiian Islands:
implication for ages of turbidites derived from submarine landslides. In:
Takahashi, E., Lipman, P.W., Garcia, M.O., Naka, J., Aramaki, S. (Eds.), Hawaiian
Volcanoes: Deep Underwater Perspectives, American Geophysical Union
Geophysical Monograph, vol. 128, pp. 51e63.
Kato, Y., 1997. Topography and geology of the sector collapse deposit of the Oshima
e Oshima Island. JAMSTEC Journal of Deep Sea Research 13, 659e667.
Katsui, Y., Yokoyama, I., Fujita, T., Ehara, S., 1975. Komagatake. Report of the vol-
canoes in Hokkaido, part 4. Committee for the Prevention of Natural Disasters,
Sapporo, 194 pp.
Kienle, J., Kowalik, Z., Murty, T., 1987. Tsunami generated by eruption from Mount St.
Augustine volcano, Alaska. Science 236, 1442e1447.
Latter, J.H., 1981. Tsunamis of volcanic origin: summary of causes, with particular
reference to Krakatoa, 1883. Bulletin Volcanologique 44, 467e490.
Le Friant, A., Heinrich, P., Deplus, C., Boudon, G., 2003. Numerical simulation of the
last flank e collapse event of Montagne Pelee, Martinique, Lesser Antilles.
Geophys. Res. Lett. 30 (2), 1034. http://dx.doi.org/10.1029/2002GL015903.
Llanes, P., Mun ~ oz, A., Mun
~ oz-Martín, A., Acosta, J., Herranz, P., Carbo

, A., Palomo, C.,
2003. Morphological and structural analysis in the Anaga offshore massif, Ca-
nary Islands: fractures and debris avalanches relationships. Mar. Geophys. Res.
24, 91e112.
Llanes, P., Silver, E., Day, S., Hoffman, G., 2009. Interactions between a transform
fault and arc volcanism in the Bismarck Sea, Papua New Guinea. Geochemistry
Geophysics Geosystems 10 (6), Q06013. http://dx.doi.org/10.1029/
2009GC002430.
Major, J., 1997. Depositional processes in large-scale debris-flow experiments.
J. Geol. 105 (3), 345e366.
Masson, D.G., Huggett, Q.J., Brunsden, D., 1993. The surface texture of the Saharan
debris flow deposit and some speculations on debris flow processes. Sedi-
mentology 40, 583e598.
Masson, D.G., Watts, A.B., Gee, M.J.R., Urgeles, R., Mitchell, N.C., Le Bas, T.P.,
Canals, M., 2002. Slope failures on the flanks of the western Canary Islands.
Earth Sci. Rev. 57, 1e35.
McGuire, W.J., 2003. Volcano instability and lateral collapse. Revista 1, 33e45.
McMurtry, G.M., Watts, P., Fryer, G.J., Smith, J.R., Imamura, F., 2004. Giant landslides,
mega-tsunamis, and paleo-sea level in the Hawaiian Islands. Mar. Geol. 203,
219e233.
Mitchell, N.C., 1993. A model for attenuation of backscatter due to sediment accu-
mulations and is application to determine sediment thickness with GLORIA
sidescan sonar. J. Geophys. Res. 98, 22477e22493.
438 S. Day et al. / Marine and Petroleum Geology 67 (2015) 419e438
Mitchell, N.C., 2006. Morphologies of knickpoints in submarine canyons. Geol. Soc.
Am. Bull. 118, 589e605.
Mitchell, N.C., Masson, D.G., Watts, A.B., Gee, M.J.R., Urgeles, R., 2002. The
morphology of the flanks of volcanic ocean islands: a comparative study of the
Canary and Hawaiian hotspot islands. J. Volcanol. Geotherm. Res. 115, 83e107.
Moore, J.G., Clague, D.A., Holcomb, R.T., Lipman, P.W., Normark, W.R., Torresan, M.E.,
1989. Prodigious submarine landslides on the Hawaiian ridge. J. Geophys. Res.
94, 17465e17484.
Moore, J.G., Normark, W.R., Holcomb, R.T., 1994. Giant Hawaiian landslides. Annu.
Rev. Earth Planet. Sci. 22 (1), 119e144. http://dx.doi.org/10.1146/
annurev.ea.22.050194.001003.
Moore, J.G., Normark, W.R., Holcomb, R.T., 1994. Giant Hawaiian landslides. Annu.
Rev. Earth Planet. Sci. 22, 119e144.
Neumann van Padang, M., 1929. De Noordelijke doorbraak in den Papandajan
kraterwand. De Mijningenieur 10, 55e57.
Pararas-Carayannis, G., 1979. Tsunamis in Indonesia. Tsunami Newsletter 12, 32.
Plafker, G., Ericksen, G.E., 1978. Nevados Huascaran avalanches, Peru. In: Voight, B.
(Ed.), Rockslides and Avalanches, Natural Phenomena, vol. 1. Elsevier, Amster-
dam, pp. 277e314.
Piper, D.J.W., Sparkes, R., Mosher, D.C., Shor, A.N., Farre, J.A., 1985. Seabed Instability
near the Epicenter of the 1929 Grand Banks Earthquake. Geolog. Survey of
Canada, Open File 1131, 29 pp.
Piper, D.J.W., Cochonat, P., Morrison, M.L., 1999. The sequence of events around the
epicentre of the 1929 Grand Banks earthquake: initiation of debris flows and
turbidity currents inferred from sidescan sonar. Sedimentology 46, 79e97.
http://dx.doi.org/10.1046/j.1365-3091.1999.00204.x.
Reid, M.E., 2004. Massive collapse of volcano edifices triggered by hydrothermal
pressurization. Geology 32, 373e376.
Satake, K., Kato, Y., 2001. The 1741 Oshima-Oshima eruption: extent and volume of
submarine debris avalanche. Geophys. Res. Lett. 28, 427e430.
Scott, K., Macias, J.L., Naranjo, J.A., Rodriguez, S., McGeehin, J.P., 2001. Catastrophic
Debris Flows Transformed from Landslides in Volcanic Terrains: Mobility,
Hazard Assessment and Mitigation Strategies. U. S. Geol. Survey Prof. Paper
1630.
Siebert, L., Glicken, H., Ui, T., 1987. Volcanic hazards from Bezymianny and Banda
type eruptions. Bull. Volcanol. 49, 435e459.
Silver, E., Day, S.J., Ward, S.N., Hoffmann, G., Llanes-Estrada, P., Driscoll, N.,
Appelgate, B., Saunders, S., 2009. Volcano collapse and tsunami generation in
the Bismarck volcanic arc, Papua New Guinea. J. Volcanol. Geotherm. Res.
http://dx.doi.org/10.1016/j.jvolgeores.2009.06.013.
Talling, P.J., Amy, L.A., Wynn, R.B., 2007. New insight into the evolution of large-
volume turbidity current: comparison of turbidite shape and previous
modeling results. Sedimentology 54, 737e769.
Tinti, S., Maramai, A., Armigliato, A., Graziani, L., Manucci, A., Pagnoni, G.,
Zaniboni, F., 2006. Observations of physical effects from tsunamis of December
30, 2002 at Stromboli volcano, southern Italy. Bulletin of Volcanology 68,
450e461.
Urgeles, R., Canals, M., Baraza, J., Alonso, B., Masson, D.G., 1997. The most recent
megaslides on the Canary Islands: the El Golfo debris avalanche and the Canary
debris flow, west El Hierro Island. J. Geophys. Res. 102, 20305e20323.
van Wyk de Vries, B., Self, S., Francis, P.W., Keszthelyi, L., 2001. A gravitational
spreading origin for the Socompa debris avalanche. J. Volcanol. Geotherm. Res.
105, 225e247.
Vallance, J.W., Scott, K.M., 1997. The Osceola Mudflow from Mount Rainier: sedi-
mentology and hazard implications of a huge clay-rich debris flow. Geol. Soc.
Am. Bull. 109, 143e163.
Viekman, B.E., Flood, R.D., Wimbush, M., Faghri, M., Asaka, Y., Van Leer, J.C., 1992.
Sedimentary furrows and organized flow structure: a study of Lake Superior.
Limnol. Oceanogr. 37, 797e812.
Voight, B., 1981. Time scale for the first moments of the May 18 eruption. In:
Lipman, P.W., Mullineaux, D.R. (Eds.), The 1980 Eruptions of Mount St. Helens,
Washington, United States Geological Survey Professional Paper 1250,
pp. 69e86.
Voight, B., Glicken, H.X., Janda, R., Douglass, P.M., 1981. Catastrophic rockslide
avalanche of May 18. In: Lipman, P.W., Mullineaux, D.R. (Eds.), The 1980
Eruptions of Mount St. Helens, Washington, United States Geological Survey
Professional Paper 1250, pp. 347e377.
Voight, B., Janda, R.J., Glicken, H.X., Douglass, P.M., 1983. Nature and mechanics of
the Mount St. Helens rockslide-avalanche of 18 May 1980. Geotechnique 33,
243e273.
Ward, S.N., Day, S.J., 2003. Ritter Island Volcano e lateral collapse and tsunami of
1888. Geophys. J. Int. 154, 891e902.
Ward, S.N., Day, S.J., 2006. Particulate kinematic simulations of debris avalanches:
interpretation of deposits and landslide seismic signals of Mount Saint Helens,
1980 May 18. Geophys. J. Int. 167 (2), 001e1004.
Watts, A.B., Masson, D.G., 1995. A giant landslide on the north flank of Tenerife,
Canary Islands. J. Geophys. Res. 100, 24487e24498.
Wessel, P., Smith, W.H.F., 1995. New Version of the Generic Mapping Tools (GMT).
American Geophysical Union. http://www.agu.org//eos_els.
Winterwerp, J.C., Van Kesteren, W.G.M., 2004. Introduction to the physics of
cohesive sediment in the Marine Environment. Dev. Sedimentol. 56, 466.
Wright, I.C., Worthington, T.J., Gamble, J.A., 2006. New multibeam mapping and
geochemistry of the 30 e35 S sector, and overview, of southern Kermadec arc
volcanism. J. Volcanol. Geotherm. Res. 149, 263e296.
Yokose, Lipman, 2004. Emplacement mechanisms of the South Kona slide complex,
Hawaii Island: sampling and observations by remotely-operated vehicle Kaiko.
Bull. Volcanol. 66, 569e584.