Tsunami Krakatau

Joumalof volcanology
and geotheml resemh
Journal of Volcanology and Geothermal Research 64 ( 1995) 23-52
Inner structure of the Krakatau volcanic complex (Indonesia) from

gravity and bathymetry data
Christine Deplus”, Sylvain Bonvalotb, Darharta Dahrin”, Michel Diamenta, Hery Harjonoc,
Jacques Dubois”
“Laboratoire de Gravimetrie et Giodynamique, Institut de Physique du Globe de Paris, 4 Place Jussieu, 75252 Paris, France
bLaboratoire de Gkophysique, ORSTOM, 72 route d’Aulnay, 93140 Bandy, France
‘Geoteknologi LIPI, Jl Cisitu, Bandung, Indonesia
Received 30 April 1993; accepted 12 May 1994
Abstract
On 27 August 1883, the Krakatau volcanic complex (Indonesia) was the site of one of the most destructive
historical eruptions. Most of the volcano was destroyed and a new caldera also formed during this catastrophic
event. Since the date of the eruption, many geological studies of the superficial structures and eruption products
have been carried out. A debate on the scenario of the eruption and the way the volcano collapsed has developed
and still is unresolved.
In order to assess the inner structure of the volcanic complex, we carried out a detailed land and marine geo-
physical survey in the summer 1990. In this paper, bathymetry and gravity data collected during the survey as well
as literature data are compiled and analysed. Bathymetric data show that the caldera is characterized by a flat sea-
bottom at 240 m below sea level and by steep linear walls suggesting that the caldera collapse has been controlled
by pre-existing features. Moreover, the build-up of the young active volcano, Anak Krakatau, on the very edge of
the caldera could lead to mechanically unstable conditions which must be considered for hazard mitigation. The
Bouguer anomaly of the volcanic complex is characteristic of volcanoes with an explosive behaviour. 3-D gravity
modelling reveals the previously unknown geometry of the dense substratum of the proto-Krakatau and evidences
the presence of a collapsed structure beneath the caldera filled up with low-density material.
Finally, we point out a major weakness zone, oriented N150” on a line passing through the old and recent vents
of the volcano. This zone could have guided both the development of the volcanic activity and the emplacement
of the 1883 caldera. Furthermore, this weakness zone passing through the summit line of the pre- 1883 Krakatau
volcano has been introduced as a significant disruption surface of the volcanic edifice in the updated scenario of
the 1883 eruption that we propose.
1. Introduction of geological investigations, only few geophysi-

cal studies were carried out on the Krakatau vol-
The Krakatau volcano, located in the Sunda canic complex in order to constrain the inner
Strait (Indonesia), is famous for its 1883 erup- structure of the volcano. Furthermore, the 1883
tion which was one of the largest volcanic events eruption resulted in the formation of a large sub-
in historic time. Although this eruption has been marine caldera which was selected by Williams
discussed by many authors mainly on the basis ( 194 1) as the type example of a collapse caldera:
0377-0273/95/$09.50 0 1995 Elsevier Science B.V. All rights reserved

SSDI 0377-0273(94)00038-I
24 C. Deplus et al. /Journal of‘Volcano1og.v and Geothermal Research 64 (I 995) 23-52
“as the type of a caldera produced by collapsefol- ture of the volcanic complex, we carried out a
lowing the evisceration of a magma chamber by geophysical survey in the summer of 1990. This
explosions ofpumice, it seems best to select that study is part of a French-Indonesian pro-
of Krakatau, both because of the recency of its or- gramme devoted to the geodynamics of the Sunda
igin and the clarity of the evidence”. However, the Strait (see Diament et al., 1990 for earlier stud-
more recent bathymetric map of the 1883 cald- ies ). Our survey involves a part of an oceano-
era has been mainly established with data col- graphic cruise and land data acquisition on the
lected by Escher in 19 19. islands of the volcanic complex. New bathyme-
In order to detail the morphology of the 1883 tric data were collected over the Krakatau cald-
caldera and to better constrain the deep struc- era and around the islands. Side-scan sonar and
10%
105"E 110"E
Fig. I. Geodynamical context of the Krakatau volcanic complex. (a) Heavy solid line with black triangles=trench axis for the
subduction of the Indo-Australian plate beneath the South-East Asian plate; light solid lines=isobaths in meters; black trian-
gles =volcanoes distributed along the volcanic arc (Java and Sumatra) related to the subduction; SFZ= Sumatra Fault Zone.
(b) Detailed view of the Sunda Strait showing the location of the Krakatau complex on a N20” volcanic line extending from
Mount Rajabasa to Panditan island.
C. Deplus et al. /Journal of Volcanologyand Geothermal Research 64 (1995) 23-52 25
high-resolution seismic profiles were also re- tonic setting of the Sunda Strait is governed by
corded on some selected traverses. In addition, an extensional regime according to the results of
we realized 148 gravity measurements on the is- recent geological and geophysical investigations
lands of the complex and about 30 magnetotel- (Lassal et al., 1989; Diament et al., 1990; Har-
luric soundings in the audio-frequency range (8 jono et al., 1991).
Hz-23 kHz) and 4 in the low-frequency range
(less than 0.25 Hz).
This paper presents the new bathymetry and 3. Volcanic activity of Krakatau
gravity data collected during our survey. Bathy-
metry data analysis allows us to better constrain The geological map of the Krakatau volcanic
the present morphology of the 1883 caldera and complex (from Effendi et al., 1986) is shown on
to discuss the location of Anak Krakatau, the Fig. 2. The complex consists in four islands (Ser-
presently active center of the volcanic complex, tung, Panjang, Rakata and Anak Krakatau). This
on the edge of the caldera. Moreover, bathyme- configuration results from at least two historical
try combined with previous topographic map- destructive eruptions and from the recent
ping provide a numerical model of altitude for emergence of Anak Krakatau. History of the vol-
the reduction of gravity data. A model of the deep canic activity of Krakatau can be divided into
structure of the volcano is then proposed from three main periods:
gravity data modelling and is discussed in light
of the historical development of the complex.
6”OYS
2. Geodynamicalcontext
The Krakatau volcanic complex lies in the

Sunda Strait (Indonesia) between Java and Su- 6’06’5
matra islands and belongs to the volcanic arc re-

lated to the subduction of the Indo-Australian
plate beneath the South-East Asian plate (Fig. 1) .
Petrologic studies have shown that Krakatau dif-
6-09’S
fers from the common pattern of volcanoes in
Indonesia. Indeed, the eruptions cycles evolve
m4
from basaltic to dacitic composition (van Bem- m5
melen, 1941; Zen and Sudradjat, 1983; Camus et -6
al., 1987). The Krakatau volcanic complex is also 6”12’S

located on a N20” volcanic line which extends lOY20’E 1 OY23’E 1 OY26’E 105’29’E
across the Sunda Strait from PanaYtan island in
KRAKATAU VOLCANIC COMPLEX
the south to Mount Rajabasa in the north pass-
ing through Sebesi and Sebuku islands (Fig. 1). Fig. 2. Geological sketch map of the Krakatau volcanic com-
This line is outlined by shallow seismicity as re- plex (from Geological Survey of Indonesia). J = alluvial de-
posits; 2 = recent volcanic products of Anak Krakatau (ba-
vealed by world-wide seismological data (Har- salt and andesite); 3=pyroclastic deposits of the 1883
jono et al., 199 1). The volcanic activity along this eruption; 4 = Rakata basaltic dyke (latest volcanic activity of
line is supposed to be recent (Nishimura et al., Rakata); 5= Rakata basaltic rocks; 6=old volcanic substra-
1986) although there is no precise dating except tum of the proto-Krakatau (trydimite andesite).
on Krakatau. Presently, the Krakatau is the only E = Bootmans rock. Light solid line indicates the coastline of
the Krakatau island before the 1883 eruption: P=Perbuatan
active volcano of this line and displays signifi- crater, D=Danan crater. Note that the old vents (P and D)
cant seismicity which has been found to be of and the new one (Anak Krakatau) are distributed along a
tectonic origin (Harjono et al., 199 1) . The tec- line trending Nl50”E.
26 C. Depius et al. /Journal of Volcanology and Geothermal Research 64 (I 995) 23-52
Stage 1: Formation of a shield volcano (Pruto- koyama, 1981; Camus and Vincent, 1983a,b;
Krukutau) about 2000 m high and 10-l 2 km in Francis and Self, 1983; Sigurdsson et al., 1991a ).
diameter according to Verbeek ( 1885) and The chronology of the different events and the
Escher ( 19 19). Destruction of the primitive vol- origin of the tsunami (Latter, 198 1; Francis,
cano would have resulted in the formation of a 1985; Vincent et Camus, 1986; Yokoyama,
large sub-marine caldera which diameter is ap- 1987) have been recently largely debated in the
proximately 7 km. The islands of Sertung and scientific community.
Panjang and the basement of Rakata are rem- The eruption can be divided into two stages on
nants of the rim of this caldera. The destruction the basis of the nature and sequence of the de-
occurred about 60,000 years ago according to posits: a plinian phase followed by several large
Ninkovich ( 1979) on the basis of drilling data explosions and emission of a large volume of ig-
or only 1500 to 1700 years ago according to Judd nimbrite. It has been proposed that the caldera
( 1889) from historical documents. collapsed late in the eruption sequence subse-
Stage 2: Formation of a basaltic cone, 800 m quently to the emptiness of the magma chamber.
high, on the edge of the prehistoric caldera and The main differences between the two recent
covering the Rakata remnant. Then, growth of scenarios deal with the chronology and the way
two andesitic cones, Danan (450 m above sea- the volcano collapsed.
level) and Perbuatan ( 120 meters a.s.1.), within ( 1) In the first model (Self and Rampino,
the caldera. These three overlapping volcanoes 1981; Francis and Self, 1983; Sigurdsson et al.,
merged into one elongated volcanic island, 199 1a,b ), the culminating event was the genera-
Krakatau. The Krakatau island largely disap- tion of pyroclastic flows by gravitational collapse
peared on 27 August 1883, leaving only half the of the eruption column. Then, most of the Krak-
cone of Rakata and a small rock pinnacle, Boot- atau island disappeared into the sea when the roof
mans rock, exposed above sea level. The associ- of the magma chamber collapsed. Thus, the large
ated events were: emission of a huge volume of volume of ignimbrite was erupted before major
ignimbrite that covered the islands and the sur-
collapse of the volcano. The new caldera was
rounding sea-floor, formation of a submarine
formed within the prehistoric one. The authors
caldera and generation of large tsunamis that
also proposed that the tsunami were produced
killed more than 36, 000 persons on the neigh-
when the pyroclastic flows entered the sea.
bouring coastlines. Pyroclastic deposits have
(2) In the second model, Camus and Vincent
greatly increased the area of the islands of Ser-
( 1983a,b) proposed a Mount St. Helens type
tung and Panjang.
Stage 3: Emergence of Anak Krakatau (Child
collapse of the pre-existing volcano that pro-
ofKrakatau in Indonesian) on January, 1928. A duced a debris flow oriented towards the north
tuff ring (basic andesite) was constructed up to and northeast (note that the collapse of Rakata
152 m above sea level until 1959 (Sudrajat, would have followed the boundary of the prehis-
1982). Then, the volcanic activity migrated toric caldera). After this catastrophic event, ex-
southwestward and built a still active cone to an tensive sub-marine eruption of ignimbrite took
height of 200 m a.s.1.. The last eruption, that be- place ending with the collapse of a new caldera.
gan on 7 November 1992, consisted mostly of According to these authors, the 1883 caldera
basaltic andesite lava flows (Bulletin of the would be located northwest of Rakata and inter-
Global Volcanism Network, 1992 ) . sects the prehistoric one. In this model the giant
landslide would be responsible for the major
3. I. The 1883 eruption tsunami.
These models were proposed on the basis of
Several scenarios have been proposed for the field geological investigations and only few geo-
1883 eruption (Verbeek, 1885; Stehn, 1929; physical studies were performed on this volcanic
Williams, 1941; Self and Rampino, 198 1; Yo- complex concerning its inner structure.
4. Previous geophysical studies tween experimental results on formation of cra-

ters by conventional or nuclear devices and field
4.1. Gravity data observations concerning various caldera-form-
ing eruptions (eruptive sequences, volume of
The first geophysical study was based on land ejecta, size of caldera, gravity anomalies...), he
gravity data gathered in 1969 and 1982 (Yokoy- concluded that the energy required to form ex-
ama and Hadikusumo, 1969; Yokoyama, 198 1, plosion crater with radii exceeding 500 m is never
1987) on the shores of the Krakatau islands. Data available during volcanic eruptions and, thus,
reduction revealed a centrosymetric Bouguer that Krakatau caldera, as most calderas, could not
anomaly: its maximum amplitude is circular and have been generated by an explosion. As previ-
runs over the external islands of the complex ously mentioned, the idea that the Krakatau vol-
while the minimum is centered over the bathy-
105.Z’E 105.4-E 105.6-E
metric depression to the West of Anak Krakatau.
However, it can be noticed that the shape and r a
I I I I I
5.95
amplitude of the Bouguer anomaly is not well Sebesi
constrained due to the limited number of land
a
data and to the lack of marine gravity data. The
relative minimum is estimated by interpolation
to be about 9 mGal. It was interpreted by Yokoy-
ama ( 198 1) in terms of caldera deposits which 6.1’S
have a negative density contrast of -0.3 g/cm3
with the basement rocks and he proposed a fun-
nel-shape model of 4 km radius and 1 km depth.
He also suggested that this low-density infill
might be material which was ejected and fallen
back into an explosion crater and probably con- 6.23
sists of pumice, ash and rocks of the old volcano.
Yokoyama ( 198 1) has then proposed that the
most important factor in the disappearance of
two-thirds of the Krakatau island during the 1883
eruption was a large explosion. This was later w AK
kml /
discussed by Self and Rampino ( 1982) and
Scandone ( 1990) who argued that the eruption
was not of the type to eject a large amount of lithic
material explosively, sufficient to cause a 4-5-km
diameter caldera.
Instead of an explosive event, a more likely
10
1 7
GIL
mechanism was proposed by Verbeek as early as

1885. This mechanism leads to the disappear-
ance of the missing part of the volcano by engulf-
ment into a collapsed caldera. In a review paper,
McBirney ( 1990) mentioned that such an expla-
nation was already proposed by FouquC in 1879
to explain the formation of the Santorini caldera Fig. 3. Attenuation zones of S waves beneath the Krakatau
and was mainly supported by the fact that the volcanic complex (from Harjono et al., 1989). AK=Anak
Krakatau (a) Surface projection: superficial (9 km depth)
volume of older lithic rocks in the ejecta is much and deep (22 km depth) attenuation zones are indicated with
less than the missing structure. Scandone ( 1990) shaded areas and with a solid line respectively. (b) West-
confirmed this hypothesis. By comparison be- east vertical section: attenuation zones appear in shaded areas.
28 C. Deplus et al. /Journal of Volcanologyand Geothermal Research 64 (I 99.5) 23-52
cane would have been blown away by an explo- day). Due to the size of the vessel, data could be
sion is dismissed in the current scenarios. only collected in areas where the depth was large
enough (Fig. 4b); for example, we could not re-
4.2. Microseismicity data cord data just north of Anak due to the presence
of shoals. Our data set completes well the bathy-
A temporary seismological network was in- metric survey recently realized by Sigurdsson et
stalled in the Sunda Strait in 1984 in the frame- al. ( 199 1a,b) in shallow areas (less than 120 m ) .
work of the French Indonesian Program. More
than 600 local earthquakes were recorded (Har- 5.2. Numerical model of altitude
jono, 1988, Harjono et al., 1991). The analysis
of S waves attenuation allowed some of us (Har- The new bathymetric data were combined with
jono et al., 1989) to detect beneath the Krakatau those of two previous cruises (Corindon in 1983
volcanic complex two anomalous zones (Fig. 3 ) and Krakatau in 1985 ) also conducted in the
which are assumed to correspond to two distinct framework of the French-Indonesian program.
magma reservoirs. The deeper one is located just In the areas where we have no data, especially in
beneath the Moho and presents a large horizon- shallow areas, we used the bathymetric map pub-
tal extension. The upper one, located at a depth lished by the Volcanological Survey of Indonesia
of about 9 km, has a smaller extension but is large in 1982 on the basis of the data collected by
enough to allow the differentiation of the vol- Escher in 19 19. Compilation of these data allows
ume of dacite of the 1883 eruption (G. Camus, us to build a new bathymetric map extending up
pers. commun., 1988). It is divided into three to 20 km around the volcanic complex. The data
parts that were interpreted as magma pockets. collected by Sigurdsson et al. ( 199 1a) in shallow
The depth found for this upper zone agrees well areas indicate only small differences with our
with the depth proposed for the magma chamber compiled bathymetric map. Topography on the
of Anak Krakatau according to petrological and islands is given by both the altimetry data col-
geochemical results (Camus et al., 1987). Fi- lected during our land survey and contour lines
nally, as shown on Fig. 3, the present shape of digitized from a topographic map of the volcanic
this upper attenuation zone could be related to complex (Volcanological Survey of Indonesia,
the internal structure of the volcano, controlled 1982). A more detailed topographic map ( 198 1,
by the successive collapses of calderas, as pro- with corrections for the eruption of 1988) was
posed by Vincent et al. ( 1989 ). used for Anak Krakatau. We used a portable GPS
receiver to locate a few points on each island. The
horizontal accuracy of the GPS equipment was
5. Morphology of the Krakatau volcanic complex checked on a geodetic bench mark, built up on
(bathymetry and topography data analysis) Anak Krakatau during a GPS survey (M. Kas-
ser, pers. commun., 1990)) and is about 10 to 20
5.1. Marine data acquisition m. As a result of our accurate positioning in ab-
solute geographical coordinates, we had to cor-
The marine data were collected during the rect the previous published map by shifting and
Mentawai Cruise on board R/V Baruna Jaya III rotating some of the islands. This was possible
(Diament et al., 199 1) (Fig. 4a). We used a nar- because we located also some remarkable spots
row-beam echo-sounder for the bathymetry that on the islands. After these corrections, the com-
gives more accurate estimations of steep slopes parison between field altitude data (barometric
than classical equipment does. The navigation and topographic levelling) and topographic maps
was achieved using GPS and TRANSIT satellite shows only small discrepancies in the range of a
systems (the GPS system gave us a position every few meters. A numerical model of altitude
13 seconds when a good constellation of satel- (bathymetry and topography) has thus been re-
lites was available, that was about 22 hours a alized for a large area encompassing the Kraka-
-6”03
-6”09
-6”12
105”20 105” 23 105”26 105” 29
BATHYMETRY DATA DISTRIBUTION

Fig. 4. Bathymetry data distribution. Large dots are data gathered during the Mentawai oceanographic cruise on board R/V
Baruna Jaya III ( 1990). (a) Around the Krakatau area: small dots are data from previous cruises. (b) Detailed data distribution
over the Krakatau complex. Crosses are data from Corindon oceanographic cruise (1983).
tau complex in order to compute the terrain cor- quently toward the steep wall of the 1883 cald-
rections for the gravity data reduction. The cen- era. We discuss later the growth of Anak Kraka-
tral part of this model (6”12’S to 6”03’S and tau on the edge of the 1883 caldera and its
105”20’E to 105”29’E) is shown on Fig. 5 on a location on a probably unstable site.
3-D representation. This area will be also used
for the further gravity maps (anomaly and 5.3. Morphologyof the caldera
model). The 3-D view allows to identify the main
effects of the 1883 eruption: the steep concave Generalpatterns
cliff of Rakata probably collapsed along the A detailed bathymetric map of the caldera
boundaries of the prehistoric caldera, the ignim- contoured every 20 meters is shown on Fig. 7.
brite sub-marine deposits on the external sides The bathymetric depression is elongated in shape
of Sertung and Rakata islands (yellow shallow with two main directions: southwest-northeast
zones); the deep 1883 caldera (blue) resulting and west-east to the west and east of Anak Krak-
from the collapse of the roof of the magma atau, respectively. The deepest part (A on Fig.
chamber. 7) is located southwest of Anak Krakatau and
Note also the location of Anak Krakatau on the has a rectangular shape with a flat bottom (240
very edge of the 1883 caldera. A detailed view of m below sea level) and steep linear walls. Lead-
Anak Krakatau seen from the southeast is shown ing out this flat deep area towards southwest and
in Fig. 6. The old crater rim and the new active east are elongated grabens with floors deeper than
cone are clearly shown on this representation. 100 m. A careful examination of our bathyme-
This emphasizes the migration of the volcanic tric records to the west of the caldera shows that
activity in a southwestern direction and conse- the narrow graben cutting the western wall of the
30 C. Deplus et al. /Journal of Volcanologyand Geothermal Research 64 (1995) 23-52
KRAKATAU VOLCANIC COMPLEX
Rdtata
Ming (413 m)
Ansk Krakatau 1883csldera

/
ANAK KRAKATAU
“1” flaw I rrm

1883caldera (+lM m)
I -cn_-,
/
Fig. 5. Threedimensional Numerical Model of Altitude (NMA) of the Krakatau volcanic complex (view from the west). The
NMA is obtained from interpolation of bathymetry data gathered during oceanographic cruises and digitized data from existing
topographic and bathymetric maps within a 17 x 17 km area. A logarithmic color scale is used to represent the bathymetric and
topographic variations between - 250 m (bottom of 1883 caldera, in blue) and + 8 13 m (summit of Rakata island, in red); the
sea surface is approximately between the orange and yellow color levels.
Fig. 6. Three-dimensional Numeric Model of Altitude (NMA) of Anak Krakatau (see legend of Fig. 5 for NMA construction.
This view is from the South East. The logarithmic color scale represents the bathymetric and topographic variations between
-250 m (bottom of 1883 caldera, in blue) and + 200 m (summit of the active cone, in red).
C. Deplus et al. /Journal of Volcanologyand Geothermal Research 64 (199s) 23-52 31
-6”O’.
-Sol0
IO! 105’25
CALDERA BATHYMETRY
Fig. 7. Detailed bathymetric map of the 1883 caldera drawn from data gathered during the MENTAWAI cruise (1990) and
digitized data from existing bathymetric maps. Isolines are drawn every 20 m and solid lines represent curves - 100 and - 200
m.
caldera is prolonged by a linear depression 20 m above sea level, “the eruptions were not proceed-
deep (B on Fig. 7). We previously suggested ing from a single point but that rather there were
(Harjono, 1988; Diament et al., 1990) that six crater openings arranged in a line about 500
Krakatau lies just at the intersection of the N20” meters in the NNW-SSE” (i.e. N150” ). We pro-
volcanic line with a transverse feature, oriented pose that the N150” azimuth, along which cra-
N50”, that links Krakatau to the Sunda Strait ters are aligned, corresponds to a structural di-
graben. This N50” feature was interpreted as a rection that has partly controlled the caldera
fault on the base of its seismological and gravi- formation.
metric signatures. The bathymetric depression to Therefore, analysis of our new data confirms
the west of the caldera corresponds probably to that the shape of the caldera is associated to lin-
the termination of that fault. Since strong under- ear directions corresponding to tectonic fea-
water currents are reported in the vicinity of the tures. Thus, we agree with Williams ( 194 1) who
Krakatau complex, we believe that the forma- assumed that the collapse of the caIdera has been
tion of this narrow graben is a consequence of a controlled by preexisting features. In 1968, Oide
submarine erosion along the N50” fault. already noted that the Japanese Krakatau-type
The north and south walls of the deepest part calderas are mostly rectangular in outline and
are oriented N70” with slopes of about 20 O.The that their straight boundaries have received few
direction of these walls is roughly parallel to the attention.
one of the above-mentioned fault. The east and Although, our data do not extend very close to
west walls are oriented N150”, and the slope on the Rakata steep cliff (see Fig. 4b) we can pre-
the western one is smaller of about 7”. The cise the shape of the contour lines north of Rak-
N 150 o azimuth corresponds to the direction on ata. This shape may be related to a flow of ma-
which the former vents of the pre-1883 Krakatau terial coming from the cliff that could correspond
island, Danan and Perbuatan were lying (see Fig. to slumping or rolling down materials. This could
2). In addition, Stehn ( 1929) reported that in be in agreement with Francis and Self assump-
January 1928, when Anak Krakatau emerged tions ( 1983) which supposed that the cone of
Rakata was apparently left perched on the south- atau and he has estimated the rate of sedimen-
em rim of the caldera after the caldera collapsed, tation to be about 1 to 3.5 m per year between
then slumped into the sea. Moreover, two weeks 19 19 and 1940. According to the fact that the
after the paroxysm and 9 months later, rock-fall frequency of eruptions of Anak seems to be quite
dust clouds, that were initially mistaken for vol- constant, Sudrajat also proposed that this rate of
canic activity, have been observed. Such clouds sedimentation remained identical between 1940
were also seen in 1896 and 1897 (Simkin and and 1982. Our data show that the depth of the
Fiske, 1983). Rock-falls along Rakata cliff have bottom of the caldera did not really changed since
been also reported by Cool in 1908 (in Simkin 1940. That can be explained by both a slowing
and Fiske, 1983). Nevertheless, a detailed sur- down of the erosion of ignimbrite deposits and
vey of the submarine foot of the Rakata cliff is by the fact that the activity of Anak became ex-
necessary to precise this point. clusively aerial since 1960.
Changes in the caldera since 1883 5.4. Growth ofAnak Krakutau

The bathymetry of the inner part of the Krak-
atau complex after the 1883 eruption has been Anak Krakatau is born about midway between
known at different times. Soundings were started the former craters of Danan and Perbuatan,
shortly after the 1883 eruption (Verbeek, 1885 ) where the main vent for the 1883 eruption is
by the survey ship HydrograaJ: Then, numerous supposed to have laid (following the consensus
soundings collected in 1919 (bathymetric map opinion), immediately off the steep NE wall of
published by Escher in 19 19) and a few supple- the basin formed by the collapse of the 1883
mentary ones in 1922-1923 (Escher, 1928, in caldera. According to Stehn ( 1929), the position
Stehn, 1929 ) have shown that changes have oc- of the eruption point corresponded nearly to a
curred: the caldera had become deeper. Exami- former depth point of 188 meters (soundings of
nation of the submarine profiles has convinced 1919 ) . Rapid soundings at the end of January
Neumann van Padang (1933) that this deepen- 1928 have shown that the western slope of the
ing of the caldera represented a vertical sagging. volcano was considerably steeper than the east-
This strongly suggests a compaction of the ma- em. This was interpreted by Stehn ( 1929) as a
terials filling up the caldera. consequence of its position right off the steep wall
Comparison of our map with the one of Escher of the basin and also of the strong current that
(in Stehn, 1929) reveals some differences. The was generally running southwest to northeast.
bottom of the caldera is more flat: the two small Figure 8 displays the temporal evolution of a
25-m depressions flanking a N-S central ridge, SW-NE morphological profile from 1919 to the
that were shown on Escher’s map, are no more present day. We also show the location of our data
present. The walls of the caldera appear now to closest to Anak. Clearly, Anak Krakatau was built
be steeper and more rectilinear. Moreover, the on the flank of the caldera (slope estimated of
bottom of the caldera is now less deep: on aver- about 18’ in 19 19). The build up of new cones
age, 240 m b.s.1. instead of 250 m b.s.1. and the on the flank of collapse calderas was also ob-
maximum depth is 255 m b.s.1. instead of 279 m served in Campi Flegrei (Scandone et al., 199 1)
b.s.1.Both the erosion of ignimbrite deposits and and for Santorini (Heiken and McCoy, 1984).
the products from Anak might be advocated in In 1928, the southwestern slope of Anak was
order to explain this apparent inlill of the caldera. steeper as 29’. Nowadays, this slope seems to be
Sudrajat ( 1982), on the base of a few sound- still important and shows that the successive
ings obtained in 1940, already noticed that the eruptions did not resulted in a gentle filling of
deepest part of the caldera had become shallower the caldera. Indeed, our data show that at a dis-
from 279 m b.s.1. to approximately 250 m b.s.1.. tance of 900 m from the shore of Anak, the depth
He suggested that this was due to the sedimen- is 245 m b.s.1. and that there is no indication of
tation of the volcanic products from Anak Krak- a gentle rise. Therefore the actual slope of the
South-West North-East
d *
d- -d
I I
N _:- .., cd
d- . . pre 1883 -6
I I
s
Y 9 _..........................................’.. 9
-0 0
N
N _ N
dr - .- , ,,-__L -I- d
-2/-
d: I / I 4
0 I I I I I I I I
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0°
X (KM)
Fig. 8. Morphological evolution with time of Anak Krakatau along a SW-NE line. Dashed lines below the sea level are the
bathymetric profiles in 19 19 and 1928. Solid line is the actual morphology of Anak Krakatau along which the nearest bathymetry
data gathered during the MENTAWAI cruise 1990 (dots) have been reported. Dashed line above the sea level shows the mean
topography of the Krakatau island before the 1883 eruption.
western submarine flank of Anak Krakatau is at to the IGSN71 base of Waymuli, located on the
least 15”. Note that on Fig. 7, we used the bath- Southwest coast of Sumatra. We used the algo-
ymetric contour lines of the map published by rithm of Longman ( 1959) for the earth tide cor-
the Volcanological Survey of Indonesia on which rection. The accuracy of the gravity measure-
the southwestern flank of Anak is not really con- ments is about 0.05 mGa1.
strained by data. Nevertheless, our data ob- On each island, the gravity stations were tied
tained on board the Burma Juyu III show that with optical devices (theodolite and EDM) from
its southwestern flank is steep and since Anak is the network of approximately 30 stations located
growing towards the southwest, one cannot ex- with GPS. The absolute horizontal positioning
clude landslides along this flank. Several few me- of all gravity stations is within 10 to 20 m.
ters tsunamis occurred probably there in 1981 A precise levelling of the gravity stations has
(Camus et al., 1987; Sigurdsson et al., 1991b). been done with optical devices. The accuracy is
Obviously, a more detailed survey of the slope within a few centimeters. For some stations along
should be realized in the future. the shore of the islands and on Anak, the altitude
data were determined by barometric levelling.
The accuracy of the barometric levelling is better
6. Inner structure(gravitydata analysis) than 2 m, precision that is enough for the gravity
modelling of the deep structure of the volcano.
6. I. Lund data acquisition The irregular distribution of the gravity sta-
tions shown on Fig. 9b relates the field condi-
In 1990, we realized 148 new gravity measure- tions. Except Anak Krakatau, the vegetation is
ments on the Krakatau volcanic complex. The extremely dense on the islands due to the tropi-
gravity stations are located on a few profiles cal climate. Moreover, the topography is very
through Panjang, Sertung and Rakata islands and rough, cut with numerous incised valleys conse-
on a dense network on Anak Krakatau (Fig. 9). quence of the strong erosion over the pumice de-
Gravity readings were made with two Lacoste posits. Verbeek ( 1885) reported in October
and Romberg gravity meters (G and D model). 1883, two months after the eruption, that steep
A gravity base was installed on Sertung and tied sided gullies 40 m deep had been already cut into
34 C. Deplus et al. /Journal of’ Volcanologyand Geothermal Research 64 (I 995) 23-52
-6’03 - “/
,,j:#+._ ..... ;.;!;‘-

-r
105010 105020 105T30 lo!P4 0
105”20 105”23 105’26 105”29
GRAVITY DATA DISTRIBUTION

Fig. 9. Gravity data distribution. Data come from land surveys (this survey; Yokoyama, 1969, 1982) and oceanographic cruises
(Corindon’83 and Krakatau’lS). (a) Around the Krakatau area. (b) Detailed gravity data distribution over the Krakatau vol-
canic complex. Land data: small dots are data from this survey and large dots are data from Yokoyama and Hadikusumo, 1969;
Yokoyama, 1982. Marine data: symbols ‘ +’ and ‘x’ consist in data from Krakatau and Corindon cruises, respectively.
the deposits. As the islands are uninhabited, no good agreement between the two cruises since the
paths exist inside the islands and the optical discrepancies of free air anomaly at crossing
sightings in the forest was limited to 25 m, slow- points are less than 2.5 mGal, i.e. the accuracy of
ing down the surveyors progression to 500 m per marine gravity data. Comparison of our land data
day. On another hand, landing on the shores ap- with the nearby data of Yokoyama show differ-
peared to be quite difficult in some places, and ences of less than 1 mGa1 on free air values. A set
even impossible (along the western and south- of 180 land data and 337 sea surface data was
ern coast of Sertung for example), due to the used to compute the Bouguer anomaly map.
presence of crumbling sea cliffs.
Data reduction
6.2. Complete Bouguer anomaly The observed gravity value g is reduced to
Bouguer anomaly gABby:
Data compilation
In order to compute the Bouguer anomaly map gAB=g-y+ph-2nGph+T(p)
of the Krakatau complex, we used all data avail- (on land data)
able in this area (Fig. 9 ) . We added the land data (1)
previously published by Yokoyama and Hadi- gAB =g-y+2Wp--pw)d+ T(P)
kusumo ( 1969) and Yokoyama ( 1987) and the (sea surface data)
sea surface data recorded during Corindon and (2)
Krakatau cruises (Diament et al., 1990). All data where y is the normal gravity value in IGSN71
were tied to the IGSN7 1 system. The cross-over system, B is the vertical gradient of the normal
analysis of the sea surface gravity data shows a gravity (estimated at 0.3086 mGal/m), G the
C. Deplus et al. /Journal of Volcanology and Geothermal Research 64 (1995) 23-52 35
Newton gravitational constant, p the reference mGaI for land data and ? 2.5 mGa1 for sea sur-
density, p,,, the sea water density is 1.03 g/cm3 face data.
and T(p) is the terrain correction; h is the sta-
tion elevation for on-land data and d the depth Density determination
of the sea-floor below a sea-surface gravity mea- A critical point in the computation of Bouguer
surement. According to the definition of Bou- anomaly is the choice of the density for the slab
guer anomaly, the Bouguer anomaly refers to the and terrain corrections. As no rock density mea-
topographic surface for on-land data (Patella, surements are available for the whole Krakatau
1988) and to the sea-level for marine data. volcanic complex (islands and submarine de-
In order to take into account the large topo- posits), we applied the Nettleton ( 1976) and the
graphic and bathymetric variations of the stud- Parasnis methods ( 1962) on the gravity data.
ied area, we computed terrain corrections at These methods have only good application within
every gravity data location. For that purpose we areas characterized by homogeneous density and
used the numerical model of altitude (NMA) where there is no correlation between the topog-
previously described, extending up to 20 km raphy and the Bouguer anomaly (Williams and
around the volcanic complex. Using bathyme- Finn, 1985). They have to be used carefully in
tric and topographic contour lines, we repre- volcanic areas where the reliefs are made of het-
sented the NMA as a pile of prisms. Horizontal erogeneous structures. In the Nettleton method
geometry of each prism is directly given by the ( 1976), the density is adjusted to minimize the
shape of the contour lines and their thickness is correlation between the complete Bouguer
equal to the altitude difference between two con- anomaly and the topography. We have com-
tour lines. This thickness depends upon the ac- puted Bouguer anomaly maps for various densi-
curacy of the data used for the NMA and is of ties ranging from 1.9 to 2.7 g/cm3 with a 0.1-g/
12.5 m for Anak topography, 25 m for bathyme- cm3 step. The resulting maps show that no den-
try and 50 m for topography of the outer islands sity minimizes the correlation for the whole vol-
canic complex. In fact, a poor correlation is ob-
of the complex. The terrain correction is then
tained on the outer islands for high density values
computed at each data point by summing the
while the correlation is minimized over Anak for
gravity effects of horizontal faces of the prisms.
lower densities. This information will be later
The algorithm used to compute the gravity effect
used and discussed in the modelling of the Bou-
of a plane facet is given by Chapman ( 1979). We
guer anomaly.
will use the same computation method for the 3- Parasnis ( 1962) proposed an analytical ap-
D gravity modelling described later on. proach to estimate the average superficial den-
The total terrain correction at a data point is sity. From Eq. ( 1) one can write:
T(p) =@iand + (Pi -P) Tseambottom where TLandand
Tsea-bottomare terrain corrections, for unit den- Y=2rrGp,X+g,, (3)
sity, due to the emerged and immerged part of with Y=g-y+@+T(p,) andX=h.
the NMA respectively. Maximum values for the Assuming g*B to be a random error of mean
terrain correction are 9 mGa1 at the top of Anak value constant in the study area, the slope of the
and 8 mGal for points located close to the Rak- best-fit straight line for the plot of Y versus Xwill
ata cliff. The use of a steplike topographic model give pi for a given value of pr. The retained den-
instead of a regular slope model introduces an sity pi is such that it equals the assumed h_. Of
error smaller than 1 mGal. This error depends course this method is only valid in areas with
on the thickness of the prisms, for example on small variations of Bouguer anomaly compared
Anak the error is less than 0.4 mGa1 (Dahrin, to the variations of the slab corrections, that is in
1993). areas with important altitude changes and con-
For a given reference density, the final accu- stant geology. Therefore, the only zone where this
racy of the Bouguer anomaly is estimated at 2 1 method can be applied is the southwestern part
36 C’. Deplus ef al. /Journal of Volcanology and Geothermal Research 64 (I 995) 23-52
SOUTH-WEST ANAK
6”06’S
Y = -8.49895 + iHxY4/ *x
Density = 2.11 g/cm3
_y+--, , , , ,
I W2S’b lOY26’E
-2s 2.5 75 12s 175 225
HMO
Fig. 10. Density determination on Anak Krakatau by means of Parasnis method. The method is applied on data from southwest
of Anak Krakatau (black squares on right side). Over this geologically homogenous area, the topographic variations are large
while the Bouguer anomaly is nearly constant. On left side: free-air anomalies plus terrain correction plotted versus height of the
station. Varying the density for terrain correction by step of 0.1 g/cm3, the best fit is obtained for 2.1 g/cm3 (see text). This
value consisting in the mean density of recent volcanic products, is used as density reference for gravity reductions and gravity
modelling.
of Anak where many lava flows were emplaced Bouguer anomaly map
since 1970. The resulting density value is of 2.1 Figure 1la shows the Bouguer anomaly map
g/cm3 (Fig. 10). This low value characterizes re- for the density of 2.1 g/cm3. According to the
cent products from Anak and is comparable to data distribution, we preferred to draw the con-
the surface density found on other volcanoes tour map by hand rather than to use any inter-
(Williams and Finn, 1985 ). Moreover, it is in- polation algorithm that could introduce spu-
teresting to note that Verbeek ( 1885) gave a rious effect. This map reveals a ring of high
value of 2.2 g/cm’ for the density of the fine gravity over the external islands of the complex.
products of the 1883 eruption, these products The average amplitude of this anomaly is + 77
covering the sea-bottom and the outer islands of mGa1 with local maxima on Rakata, Sertung
the complex. Furthermore, a density close to 2.1 ( + 82 mGal), and Panjang ( +80 mGa1) is-
g/cm3 for the upper sedimentary layer around the lands. To the southwest, the amplitude of the
Krakatau complex has been deduced from veloc- positive anomaly decreases down to + 75 mGa1.
ities obtained by a refraction study west of Krak- A sub-circular low anomaly of + 54 mGal lies in
atau (Larue, 1983) and from drilling data in the the middle of the high ring. Note that Anak
eastern part of the Sunda Strait (Noujaim, Krakatau is located on the steep linear gradient
1976). Therefore, we decided to use a reference between the minimum and maximum anoma-
value of 2.1 g/cm3 for the reduction of gravity lies. This N 150” gradient underlines the already
data. mentioned direction passing through the ancient
m Above 82.0
I 78.0
80.0 - 80.0
82.0
I 74.0
76.0 - 76.0
78.0
72.0 - 74.0
70.0 - 72.0
68.0 - 70.0
I 64.0
66.0 - 66.0
68.0
I 60.0
62.0 - 64.0
62.0
I 56.0
58.0 - 58.0
60.0
m 54.0 - 56.0
m Below 54.0
105020 105O23 105O26 105O29
Fig 11. (a) Bouguer anomaly map of Krakatau volcanic complex (mGa1). The map is computed in IGSN7 1 reference system
for a 2.1 g/cm3 density reference. The coastline of the Krakatau island before the 1883 eruption appears in &shed line; R, D and
P denotes its old active centers Rakata, Danan and Perbuatan respectively. (b) Bouguer anomaly map of Rrakatau volcanic
complex superimposed over the 3D Numeric Model of Altitude. The color scale shows the variation of the Bouguer anomaly
from + 54 mGa1 (purple) to + 82 mGa1 (red). This graphic representation emphasizes the relations between the Bouguer anom-
aly and the morphology of the complex. No correlation remains between gravity anomaly and topography on Anak Rrakatau. In
the same way, there is no correlation along the ring between Bouguer anomaly and the bathymetry of the sea-floor. This indicates
that 2.1 g/cm3 is a good value for the average density of Anak and also for volcanic products covering the sea-bottom and the
outer islands. A remarkable feature is the correlation between gradients of the low gravity anomaly and the morphology of the
1883 caldera. This correlation indicates a probable superficial origin related with the caldera development.
38 C. Depius et al. /Journal of Volcanologyand Geothermal Research 64 (199.5) 23-52
- 6OO3
- 6OO6
- 6"09
Above -0.5
-1.0 - -0.5
-1.5 - -1.0
-2.0 - -1.5
-2.5 - -2.0
Below
6'12
-2.5
105O20 105O23 105O26 105O29

Fig. 12. Gravity effect of superficial magma chambers deduced from the seismological study of Harjono et al. ( 1989). The effect
is computed in mGal using a mean density of -0.2 g/cm’ and the 3-D geometry shown in Fig. 3. Horizontal extension of the
three superficial magma chambers at 9 km depth appears in dashed lines.
craters of Danan and Perbuatan and the present neath the outer islands. Finally, there is a clear
active one. Outside the positive ring, the Bou- spatial relationship between the minimum of the
guer anomaly is roughly of +65 mGal, consis- anomaly and the 1883 caldera morphology: the
tent with the regional trend deduced from the highest gravity gradients closely follow the steep
free-air gravity map of the Sunda Strait (Dia- flanks of the caldera. Such correlation is an in-
ment et al., 1990). dication that the source of this relative negative
No correlation remains between gravity anomaly is mainly superficial.
anomaly and topography on Anak Krakatau (Fig.
1lb). This conlirms that 2.1 g/cm3 is a good 6.4. Gravity interpretation (3-D gravity model)
value for the average density of Anak. In the same
way, there is no correlation along the ring be- Gravity anomaly over calderas
tween Bouguer anomaly and the bathymetry of The Bouguer anomaly map of the Krakatau
the sea-floor. This indicates that 2.1 g/cm3 is also presents similar characteristics to those obtained
a good value for volcanic products covering the on other volcanic complex such as Santorini vol-
sea-bottom and the outer islands. Therefore, the cano (Budetta et al., 1984) or Campi Flegrei
Bouguer anomaly primarily reflects density het- (Barberietal., 1991;Fedietal., 1991).Allthese
erogeneities beneath the sea-floor and cannot be maps exhibit a negative anomaly over a caldera
related to a deficient terrain correction. Further- surrounded by a ring of positive anomalies, Ac-
more, Anak Krakatau is underlined by a much cording to Rymer and Brown ( 1986), negative
smaller anomaly than the other islands of the anomalies are mostly observed on volcanic com-
complex indicating higher density values be- plexes with a long history of explosive eruptions
Fig. 13.3-D geometry of our best-fit density model. (a) High-density body (density contrast of + 0.5 g/cm3) interpreted as the
volcanic substratum of the proto-Krakatau. (b) Low-density body (density contrast of -0.2 g/cm’) interpreted as a mixing of
juvenile products of the 1883 eruption and lithic remnants of the destructed volcano filling up the 1883 caldera.
associated with caldera formation and emission between anomaly wavelength and caldera diam-
of highly silicic pyroclastic materials. Such vol- eter indicating a strong genetic relationship. Low-
canoes are found on continental crust in a sub- density caldera intill, often silicic pyroclastic
duction or rift environment (e.g., Long Valley, material, is probably responsible for the ob-
Yellowstone, Japanese and Indonesian volcan- served negative anomaly while denser volcanic
oes...). In each case, there is a close correlation rocks forming the volcano substratum are fre-
40 C. Deplus et al. /Journal of Volcanology and Geothermal Research 64 (1995) 23-52
Thickness (mews)
-6”12 I I IA
105"20 105"23 105"26 105"29
DENSE SUBSTRATUM (+0.5 g/cm3)
-6"03
b B\
-6"06
bL-
-6"09
. ./
Thickness (metres)
\
B’ 240 - 1500
I
240 1000
-6"12 i-
105"20 105"23 105"26 105"29
CALDERA INFILL (-0.2 g/cm3)
Fig. 14. Horizontal geometry of the 3-D density model shown in Fig. 13. The grey color scale indicates the thickness of the density
body below the sea surface. AA’, BB’ and CC’ gives the location of the cross-sections shown on Fig. 15. (a) High-density body:
dense substratum. (b) Low-density body: caldera inlill.
Fig. 15. Vertical sections of the 3-D density mode1 along lines AA’ (a), BB’ (b) and CC’ (c) (see location on Fig. 14). (I)
Observed and computed gravity anomalies in mGal. The observed values at gravity stations are shown with dots on the observed
anomaly (dashed line). Error bars are f 1mGal and ? 2.5 mGal for land and marine data, respectively. Light solid lines show
the gravity effects of low- and high-density bodies. The total computed anomaly (heavy solid line) tits the observed anomaly
within the precision of the data. A constant regional value of +62 mGal has been added to all computed gravity effects. (II)
Complete density model extending down to a depth of 8 km. The reference density model (on the right) is based on the results
of a refraction data analysis (Larue et al., 1983). All values ofdensity are in g/cm3. (III) Detailed superficial part of the density
model from 0 to 2 km depth.
Computed
+--- + Observed
N-l 0 1 2 3 4 5 6 7
(b) B B’
Computed
I
+~-._, Observed
9
0
9
hN
E
II
sz
No
d
-0.20
9
co m -0.10
9
0 9 m +0.50
0
In
”
-d Q
III
E2 9
-3 9
No
_; ”
0
cti 0
.!
5 6 7 6 9 10 11 12 13 14 15 16 17 18 19 20 21
X (km)
Fig. 15. (continued).
quently invoked as source of the positive ring. In density contrast of -0.2 g/cm3 and the 3-D ge-
the case of an active volcano, a low-density ometry deduced from the seismological study and
magma chamber within the uppermost few kilo- shown on Fig. 3. This effect is elongated in the
meters of the crust may also contribute to the east-west direction and is of small amplitude and
gravity low (Kane et al., 1976; Yokoyama, 198 1; large wavelength and thus contributes slightly to
Rymer and Brown, 1986; Scandone, 1990). the observed anomaly. Since the amplitude vari-
ation ( 1.5 mGa1) within the area is of the order
The gravity effect of magma chambers of the accuracy of our Bouguer anomaly, we will
In order to interpret the observed gravity not further consider this contribution in our mo-
anomaly over the Krakatau volcanic complex, we delling. Note that the deep attenuation zone re-
first determine the contribution to the gravity low vealed by the tomographic study and located at
of the magma bodies imaged by Harjono et al. the Moho is too large and too deep to produce
( 1989) from anomalous S waves. Although some any local gravity effect and thus will neither be
uncertainty exists for the density contrast of a taken into account. Introducing a denser crystal-
magma body, it is generally assumed to be nega- lised roof for a more realistic model of a magma
tive and less than 0.5 g/cm3 (Isherwood, 1976; chamber, as proposed by Scandone ( 1990 ) , will
Iyer, 1984; Blake and Ivey, 1986; Rymer and also decrease the amplitude of the gravity effect.
Brown, 1986; Brown et al., 1987; Scandone, Therefore, the observed gravity anomaly over the
1990). Figure 12 displays the gravity effect of the Krakatau area is mostly controlled by the sub-
superficial magma chambers, computed using a surface inner structure of the volcanic complex.
-10 1 2 3 4 5 6 7 i io 11 12 13 lk 15 1e
X (km)
Fig. 16. Gravity effect of two different 3-D models for the deep structure below the Krakatau volcanic complex shown on a
vertical section along line AA’ (see location on Fig. 14). Lower part: Model 1= model of caldera formation by ‘chaotic’ collapse;
Model 2 =model of caldera formation by ‘piston-type’ collapse. In the two models, the collapsed roof of the magma chamber is
supposed to lie at a depth greater than 8 km. The superficial structure is recalled with light lines. Upper part: Computed gravity
effects in mGa1 for models 1 (heavy solid line) and 2 (light solid line).
Subsurfacestructure 1984). So we choose 2.6 g/cm3 for the density of

The observed Bouguer anomaly over the the substratum of the proto-Krakatau. Accord-
Krakatau volcanic complex may be explained ing to the total density contrast (0.7 g/cm3) re-
with a subsurface model including a low-density quired for matching the steep gradients, we de-
caldera infill and the dense volcanic substratum duced a density of 1.9 g/cm3 for the caldera infill.
of the proto-Krakatau. In order to fit the steep Due to this contrast and as no direct field obser-
gravity gradients observed, a sub-vertical con- vation on the nature and the density of these
tact associated with a density contrast of about submarine materials are available, we suggest
0.7 g/cm3 has to be introduced. Examination of that they contain both juvenile pyroclastic prod-
the Bouguer anomaly maps computed with var- ucts and older lithic rocks of the destructed is-
ious density values reveals that a density of 2.6 land. The densities deduced from many studies
g/cm3 minimizes the variations of the Bouguer for these volcanic materials usually range be-
anomaly along the positive ring. Such a value is tween 2.0 and 2.2 (Nishimura et al., 1984; Ry-
similar to the one proposed for ancient volcanic mer et Brown, 1986; Alatorre-Zamura and Cam-
products on other volcanoes such as Mount St. pos-Enriques, 1991; Zubin et al., 1992). Our
Helens (Williams and Finn, 1985) or the Toba value indicates that the average density of the
caldera in the Sunda arc (Nishimura et al., filling-up materials tends towards low values for
Fig.17
m Ahove 82.0
I 78.0 -- 82.0
80.0 80.0
111 74.0
76.0 - 78.0
76.0
72.0 - 74.0
70.0 - 72.0
68.0 - 70.0
I 66.0
64.0 - 68.0
66.0
m 62.0 - 64.0
I 60.0
58.0 - 62.0
60.0
g 5”;;; I 5”;:;
I Below 54.0
106020 105O23 1tw26 lW26
Fig.18
- 6OO6
I bove 6.0
4.0 - 6.0
20 - 4.0
0.0 - 2.0
-20 - 0.0
-4.0 - -2.0
-6.0 - -4.0
slow -6.0
- 6O12
106020 106023 105O26 106026

volcanic products. It might be explained by an rocks form outcrops on some places of Rakata,
homogeneous light inlill or by an heterogeneous Sertung and Panjang (see Fig. 2) and consist of
inlill consisting of blocks or a combination of ‘trydimite andesite’. The comparison of the shape
both sources. of the islands with the contour lines of the posi-
Therefore, the density contrast with respect to tive anomaly confirms that the southeastern part
our reference value of 2.1 g/cm’ are about + 0.5 of Rakata and the north and northeastern part of
g/cm3 for the dense substratum and - 0.2 g/cm3 Panjang are formed with younger volcanic prod-
for the caldera infill. The geometry of our best- ucts emplaced during recent eruptions.
fit model is shown on Figs. 13 and 14. The sub- Another hypothesis might be advocated in or-
stratum is sub-cylindrical with a hole in the der to explain the origin of the positive anomaly.
centre. Its thickness is on average 1000 meters, Density measurements were performed on pyr-
its flat bottom lies at a depth of about 1225 oclastic rocks drilled in Campi Flegrei (Barberi
m.b.s.1. and its top at 225 m.b.s.1. except beneath et al., 199 1) and have shown an increase of den-
the outer islands where it forms their basement. sity with depth up to + 0.2 g/cm3 resulting from
As land data has been gathered mainly along the both compaction and hydrothermal circulation
shores, detailed density model of the internal part along circular faults. On the other hand, numer-
of the outer islands would be poorly constrained. ous well data obtained on a geothermal site in
Therefore, the upper limit of the dense substra- Costa-Rica (Miravalles volcano) did not show
tum lies arbitrarily at sea level. The middle any increase of density due to hydrothermal ef-
depression is filled with up to 1000-l 500 m of fect (Hallinan et al., 1992). Although one can-
light material. Figure 15 shows three vertical sec- not completely discard such an hypothesis for
tions across the density model along lines AA’, Krakatau, we favour the first hypothesis that is
BB’ and CC’ (see location on Fig. 14) and the supported by geological data.
corresponding observed and computed gravity
profiles. The fit between the observed anomaly Deep structure
and the computed one was considered as good if For the deepest part, we tested two different
smaller than the accuracy of the observed Bou- models, both compatible with the collapse of the
guer anomaly, that is 1 mGal for land data and top of a magma chamber. Scandone ( 1990) re-
2.5 mGa1 for marine data, when close to the data cently discussed various models of caldera for-
points. mation. He suggested that the shape of the grav-
In this model, we interpret the origin of the ity anomaly allows to discriminate between
positive anomalies as a dense substratum made piston-like or chaotic collapse. In the piston-like
of ancient magmatic rocks that form the base- model, the collapse of the roof of the magma
ment of the outer islands. This is in good agree- chamber is entirely transmitted up to the surface
ment with the hypothesis on the inner structure and results in the formation of a flat bottom
of the volcano made by Verbeek in 1885 on the caldera. Chaotic collapse is produced by the pro-
basis of geological investigations. He proposed gressive breaking away of the rocks from the top
that the outer islands were built by accumulation of the magma chamber. A chimney, with a den-
of deposits from successive eruptions on the sity lower than the original density, is then
remnants of the proto-volcano. These ancient formed. Depending upon the size of the magma
Computed gravity effect mGa1 of density 13- 15 ) over the whole

+62 mGa1 has added to computed gravity effect. is very the observed Bouguer
anomaly map (Fig. 11)
Fig. Residual map the computed effect of density from observed Bouguer
anomaly map (Fig. 11). is better than 2 mGa1 (equivalent to gravity data accuracy) volcanic complex
data available. Departures greater 2 mGa1 occurred outside the complex the gravity modelling is
constrained due the lack
46 C. Deplus et al. /Journalof Volcanology and Geothermal Research 64 (I 995) 23-52
chamber and the strength of the material, the the data points the misfits are smaller than 2
chimney can reach the surface and create a fun- mGa1.
nel shape caldera. The corresponding gravity
anomaly is negative for both models but the pis-
7. Discussion
ton-like model anomaly presents a U-shape
whereas the chaotic model anomaly is character-
7. I. Extension with depth of the bathymetric
ized by a V-shape. On this basis, Scandone depressions
( 1990) suggested that the observed gravity
anomalies over Krakatau, Aira (Japan ), Campi Modelling of the observed Bouguer anomaly
Flegrei (Italy) and Santorini (Greece) better lit reveals the deep structure of the volcanic sub-
the chaotic model. But such a discrimination stratum and allows us to image the extension with
based on observed gravity gradients seems in fact depth of the bathymetric depressions. A col-
difficult to evidence if the caldera is filled in with lapsed zone within the substratum lies below the
low-density material, which is the main source bathymetric depression characterized by a rect-
for the negative anomaly. angular shape and a flat bottom, southwest of
Figure 16 shows the two models tested. The Anak (A on Fig. 7 ) . On the contrary, the bathy-
first one is close to the chaotic model proposed metric depression located between Anak and
by Scandone ( 1990) and thus is characterized by Panjang (D) as well as the east-west trending
the presence of a chimney filled up with an het- graben north of Rakata (C) are not associated
erogeneous material. This chimney extends from with any deep collapse (see Figs. 13, 14 and 15 ).
the top of the magma chamber to the bottom of Note however that our gravity map is better con-
the caldera infill. The density contrast for the strained on the E-W graben that between Anak
material inside the chimney is taken as - 0.1 g/ and Panjang due to the data distribution (see Fig.
cm3 with respect to a reference crustal model. The 9). Nevertheless the observed gradients do not
second model is the piston-like subsidence of the advocate for any important gravity low between
crustal layers with an amplitude of 1500 meters, Anak and Panjang islands. As for the narrow gra-
that is roughly the entire collapse of the caldera ben south of Sertung (B), it is associated to a
in our model (bathymetric depression plus in- small depression in the substratum of about 100
fill). The reference density crustal model is de- m (see Figs. 13, 14 and 15 ). It can be attributed
duced from velocities obtained by a refraction both to a tectonic origin and/or to a horseshoe
study close to Krakatau (Larue, 1983; Diament shape of the ancient volcano such as the one ob-
et al., 1990) using Birch’s law (Nafe and Drake, served during the first stages of the growth of
1963). The gravity effect for the two models is Anak by Sudrajat ( 1982).
nearly identical with a maximum amplitude on Therefore, gravity data interpretation clearly
the order of 2 mGa1. According to our data dis- points out that only the rectangular part of the
tribution and precision, we cannot discriminate bathymetric depression located SW of Anak (A
between both mechanisms. Nevertheless, as the on Fig. 7) is related to a deep cave-in of the vol-
wavelength of this effect is comparable to the one canic substratum. In other words, the under-
of the observed Bouguer anomaly, we have re- ground structure is much smaller in diameter
tained one of these two models in order to take than the present bathymetric outline of the mor-
into account this small contribution in our inter- phological depression. Such a feature was also
pretative models shown on Fig. 15. recognized in most Japanese calderas ( Aramaki,
The gravity effect of the complete density 1984).
model (subsurface plus chimney) was also com-
puted on the whole area and is shown on Fig. 17. 7.2. Shape of the substratum depression
Wavelength and amplitude are comparable to the
ones of the observed Bouguer anomaly map. A The geometry of the substratum, as deduced
difference map (Fig. 18 ) confirms that close to from our model, is compatible with the emplace-
ment of one or several collapse caldera. To the ucts within the depression. For Krakatau, we
centre of the model, the depression in the sub- propose in this paper that this inlill probably
stratum has a funnel shape with two distinct contains both juvenile products and shattered
slopes. The upper one is moderate and prolongs rock fragments of the destructed volcano. In-
the inner parts of Rakata, Sertung and Panjang deed, the volume lost during the 1883 eruption
islands; it is covered with 2.1 g/cm3 density ma- (both above and below sea-level) has probably
terial (Fig. 15b). Such a geometry might be in- disappeared partly by engulfment and partly by
terpreted as the result of one or more pre-histor- gravitational failure, the debris lying either on the
ical collapses before the 1883 eruption. The sea-bottom or in the infill. So, one can write that
deeper slope is larger and is located just below the proximal volume of the juvenile products plus
the flanks of the rectangular part of the bathy- the volume of the debris of the destructed vol-
metric depression SW of Anak island; it limits cano must be roughly equal to the volume of the
the area filled up with a low density material (see deposits close to the volcanic complex plus the
Fig. 15b and 15~). Thus, the deep depression was volume of the inlill:
probably formed during the 1883 eruption.
VPProducts + VDetmis = I/Deposits + b/In~ (4)
7.3. Location of the zone of maximum collapse
with respect to the present magma bodies VDeposits has been recently estimated by Sig-
urdsson et al. ( 1991a) as 13.6 km3.
It is noteworthy to compare the zone of maxi- Vlnfillis deduced from our 3-D model, 11.4
mum collapse in our model (Fig. 14) with the km3. However, gravity modelling alone cannot
location of the magma chambers deduced from provide a definitive value for this volume. The
anomalous S waves (Fig. 3 ) . It is located above first uncertainty is related to the density con-
a zone where no seismic attenuation has been re- trast, a larger one yielding to a smaller volume.
ported, just between two magma bodies. This However, 1.9 g/cm3 seems to be a lower limit for
observation is consistent with a collapse of a part the mean density of the intill because even though
of the roof of a larger magma chamber during the it is composed of light products, its thickness re-
1883 eruption. Indeed, Vincent et al. ( 1989) al- sults in some amount of compaction. For exam-
ready proposed that a large reservoir would have ple, typical measured densities of the Cerro Galan
been separated in two magma pockets by the col- (Argentina) ignimbrite within the caldera ( 1200
lapse of its central part. Comparison of gravity m thick) range between 2.1 and 2.4 g/cm’
and seismological interpretations confirms this (Francis et al., 1983 ) . A second uncertainty deals
hypothesis. with its geometry itself: the horizontal extension
is well constrained by the location of the ob-
7.4. Volumes of the 1883 eruption served gravity gradients but the depth of its base
is not clearly resolved. Increasing this depth of a
Let us now compare the volume of the depres- few hundred metres slightly affects the calcu-
sion in the volcanic substratum, as given by our lated anomaly simply because the further the
gravity modelling, with respect to the one of the sources, the smaller is the gravity effect. So, 11.4
erupted products of 1883. 3-D gravity modelling km3 has to be considered as a minimum value.
requires a low-density funnel-shaped body below VDebris is more difficult to estimate because de-
the floor of the 1883 caldera. For Krakatau-type bris might come both from the lost structure and
calderas, silicic caldera infill has been currently the walls of the underground funnel-shaped
proposed and provide an adequate explanation structure. Moreover, according to the localisa-
(Rymer and Brown, 1986). Therefore, this intill tion of the funnel-shaped structure in the sub-
has to be taken into account for the estimation stratum with respect to the bathymetric depres-
of the volume of the erupted products. The dif- sion and to the pre-1883 volcano, we propose that
ficulty arises with the amount of juvenile prod- at most one half of the lost volume (i.e. 25 to
48 C’. Deplus et al. /Journal of‘ Lblcanology and Geothermal Research 64 (I 995) 23-52
50%) has foundered en masse and that the other The value of this density ratio is poorly con-
part has became fragments. So: strained but is commonly assumed to lie be-
tween 0.5 and 0.9. A mean ratio ranging between
VDebris =A* v,ost ‘kkmo/hebr,s with f <A < a
0.7 and 0.8 is in good agreement with the density
The lost volume is roughly 5 km3 above sea- values used in our gravity modelling: assuming
level (Yokoyama, 198 1) and 9 km3 below sea- an initial density of the magma of 2.6 g/cm3, the
level (Sigurdsson et al., 1991a). Assuming 2.5 g/ mean density of the products would be 1.9-2.1
cm3 and 1.9 g/cm3 for the densities of the vol- g/cm3.
cano and the debris IXSpCCtiVCly, vD&ris ranges Varying the density ratio between 0.7 and 0.8
between 9 and 13.5 km3. yields Ifmagma= 12-20 km”.
Therefore, he proximal volume for the erupted This value is slightly larger than previous esti-
products becomes: mations (9-10 km3 from Self and Rampino,
198 1) based on different assumptions. Never-
Vpproducts--V Deposits2 2. 5km3 = 1 1- 16km ’
theless one must keep in mind that, due to the
that is in the range of previous estimations (Ver- previously mentioned uncertainties, our estima-
beek, 1885). tions are only indicative.
Following the estimation of Self and Rampino The two important points to remind are:
( 198 1) for the distal products (6-9 km3), the ( 1) 3-D gravity interpretation requires a low-
volume of the total erupted products is then 17- density body beneath the floor of the 1883 cald-
25 km3. era and its volume has to be taken into account
It is common to compare the volume of within the estimations of the erupted products.
drawn magma with the lost volume. However, (2) The location of the underground collapse
such a comparison is only valid in terms of or- with respect to the pre- 1883 volcano implies that
ders of magnitude. In fact to equalize these two the lost volume should have disappeared both by
volumes requires first that all withdrawn magma engulfment and by gravitational failure, the de-
is erupted. It requires also a piston-type collapse bris lying either on the sea-bottom or in the inlill.
where the whole lost volume has settled down as
a block. In that case one should find the topog- 7.5. Consequences for the scenario of the
raphy of the volcano on the bottom of the cald- destruction of the pre-1883 volcano
era. This was observed after the formation of
Femandina caldera (Galapagos) in June 1868 The linear northeastern border of the deep
where a cone has come down by 300 metres depression in the substratum trends N150”E
(Scandone, 1991). In most cases of Krakatau ( Fig. 14 ). Its geometry is well constrained by the
type calderas, the bottom of the caldera is flat and location of our gravity data on Anak. It follows
modelling of gravity data yields to underground the direction along which old and recent vents of
funnel-shaped structures in which juvenile prod- Krakatau are distributed (dashed line on Fig. 2 ).
ucts and fragments of the destructed volcano ac- Thus, gravity interpretation favours the exis-
cumulate. The initial volume of the destructed tence of the postulated N 150’ tectonic weakness
volcano should have been increased by fragmen- zone that could have guided both the develop-
tation and the volume of collapse should be larger ment of the volcanic activity (vents of Danan,
than the lost volume. Therefore in most cases, Perbuatan and Anak have successively took place
the volume of withdrawn magma should be larger along this line) and the caldera emplacement
than the lost volume. during the 1883 eruption. Clearly, this major
Another difficulty arises when estimating the weakness zone passing through the summit line
volume of magma. If all withdrawn magma is of the pre- 1883 Krakatau volcano must be intro-
erupted, one can write: duced in the models of the 1883 eruption as a
Vmagma- VP/Products
significant disruption surface of the volcanic ed-
~PproductslPmagma.
ifice. Moreover, the southwards prolongation of
the N 150’ line intersects the Rakata coastline at ably weakened the volcanic edifice yielding to a
its northern cape (see Fig. 2). To the west of this slumping of its northeastern flank on the sea-
cape lies the spectacularly concave cliff extend- bottom. The bathymetric depression located be-
ing up to the summit of the island along which tween Anak and Panjang (D on Fig. 7), which
most of the northern part of the cone would have does not seem associated with any deep collapse,
slumped down; to the east, the shape of the may result from the slumping of this part of the
coastline and of the topographic and bathyme- volcano, the debris lying further north and east.
tric isolines (see Fig. 7 ) clearly image a smaller Indeed, hummocky sea-bottom topography has
slumping cliff. Therefore, the missing part of the been observed there (Camus et al., 1992). The
cone has probably slumped into the sea along two fact that the upper part of the hummocks con-
collapse zones located on each side of the N 150” sists of pyroclastic products (Sigurdsson,
line! 199 1a,b) is in good agreement with an avalanch-
Thus, in light of these new geophysical data, ing of the northeastern flank of the old volcano
we propose that the western part of the volcano before the eruption of ignimbrite.
had preferentially foundered by engulfment at the This event, creating a sudden lowering of the
time of the caldera collapse whereas the eastern lithostatic pressure on the vent, may be respon-
part had slumped on the sea-bottom towards sible for the onset of the generation of pyroclas-
northern and eastern areas. tic flows. Several units of pyroclastic flows have
What triggered the collapse of the volcano? been identified on the islands deposits (Sigurds-
Collapse of the roof of the magma chamber does son et al., 1991b). The caldera might have col-
not occur necessarily at the end of the eruption lapsed with successive phases related to these
sequence. As pointed out by Williams and Mc- different episodes of magma draining. We pro-
Bimey ( 1979) and McBimey ( 1990), “err+ pose that progressive collapse of the caldera and
tionsare not necessarilythe cause of collapse;in foundering of the western part of the volcano may
many instances it is the collapse that has trig- have occurred throughout the stage of emission
gered the eruption rather than vice-versa(Mc- of ignimbrite. The presence of the low-density
Call, 1963)“. In the same way, Druitt and Sparks infill below the caldera favours this hypothesis.
( 1984) proposed that eruptions producing large Finally, the northern face of Rakata slumped into
volume of ignimbrite involve two well-defined the sea as previously described.
stages. The first stage begins with an over-pres- Camus and Vincent (1983a,b), Vincent and
sured magma that produces pumice-fall depos- Camus ( 1986) and Camus et al. ( 1992) already
its. The authors show that only a small to mod- proposed a flank failure of the volcano to explain
erate volume can be erupted during this stage. the drastic change from plinian to ignimbritic
Then the pressure decreases to values less than activity. So we agree with them but our scenario
lithostatic pressure. Initiation of the second stage differs on several points:
is marked by the onset of the caldera collapse. - only a small part of the volcano would have
Substantial volumes of magma can then be dri- slumped into the sea and produced the debris av-
ven to the surface. In a generalised scheme of the alanche before the emission of ignimbrite (the
evolution of a caldera-forming eruption, Scan- northeastern part and not all the missing
done ( 199 1) proposed also that collapse begins structure).
during or at the end of the plinian phase. - the initial desequilibrium would be produced
This model can be applied to the Krakatau along the N 150” weakness zone and be related
1883 eruption where the two stages have been to the onset of the collapse of the roof of the
recognized: a plinian phase followed by the ex- magma chamber.
tensive emission of ignimbrites. We propose that - the northern face of Rakata would have disap-
the roof of the magma chamber began to collapse peared late in the eruption sequence as proposed
at the end of the plinian phase along the N150” by Francis and Self ( 1983 ) .
weakness zone. Movement along this zone prob- Therefore, analysis of the morphology of the
50 C. Deplus et al. /Journal qf Volcanologyand Geothermal Research 64 (I 995) 23-52
Krakatau complex and gravity data modelling do Blake, S. and Ivey, G.N., 1986. Magma-mixing and the dy-
not allow to discriminate between the two recent namics of withdrawal from stratified reservoirs. J. Vol-
canal. Geotherm. Res., 27: 153-178.
scenarios for the 1883 eruption (described in the Brown, G.C., Rymer, H. and Thorpe, R.S., 1987. The evolu-
Introduction part of this paper) but favour a tion of andesite volcano structures: new evidence from
‘mixed model’ in which the western part of the gravity studies in Costa Rica. Earth Planet. Sci. Lett., 82:
former Krakatau island had preferentially foun- 323-334.
dered by engulfment at the time of the caldera Budetta, G., Condarelli, D., Fytikas, M., Kolios, N., Pascale,
G., Rapolla, A. and Pinna, E., 1984. Geophysical pros-
collapse whereas the eastern part had slumped on
pecting on the Santorini Islands. Bull. Volcanol., 47( 3):
the sea-bottom. 447-466.
Bulletin of the Global Volcanism Network, 1992. Smithson-
ian Institution, vol. 17, no. 10, I5 pp.
Acknowledgements Camus, G. and Vincent, P.M., 1983a. Discussion of a new
hypothesis for the Krakatau volcanic eruption in 1883. J.
Volcanol. Geotherm. Res., 19: 167-173.
This study was financially supported by the Camus, G. and Vincent, P.M., 1983b. Un siecle pour com-
“Delegation aux Risques Majeurs” and by INSU prendre I’truption du Krakatoa. La Recherche, 149: 1452-
“Geosciences Marines”. The authors are in- 1457.
debted to Professor M.T. Zen (BPPT, Jakarta) Camus, G., Gourgaud, A. and Vincent, P.M., 1987. Petro-
for providing a constant support in this French- logic evolution of Krakatau (Indonesia): Implications for
a future activity. J. Volcanol. Geotherm. Res., 33: 299-
Indonesian Research Program. We thank the
316.
captain, officers and the crew of the Indonesian Camus, G., Diament, M., Gloaguen, M., Provost, A. and
R/V Baruna Jaya III and all the Indonesian and Vincent, P.M., 1992. Emplacement of a debris avalanche
French scientists who participated to the Men- during the 1883 eruption of Krakatau (Sunda Straits, In-
tawai oceanographic cruise. The field operations donesia). Geojournal, 28(2): 123-128.
Chapman, M.E., 1979. Techniques for interpretation of geoid
would not have been possible without the effi-
anomalies. J. Geophys. Res., 84(B8): 3793-3801.
ciency of the geophysical team of Edy Arsadi Dahrin, D., 1993. Etude bathymetrique et gravimetrique du
(LIPI, Bandung). We thank our colleagues Yves Detroit de la Sonde et du volcan Krakatau (Indontsie):
Albouy, Gilbert Juste, Germinal Gabalda (OR- Implications geodynamiques et volcanologiques. Thesis,
STOM, Bondy) and M.T. Zen Jr. (IPG, Paris) Univ. Paris VII, 335 pp.
for their assistance in survey preparation and Diament, M., Deplus, C., Harjono, H., Larue, M., Lassal, O.,
Dubois, J. and Renard, V., 1990. Extension in the Sunda
Dominique Rtmy (ORSTQM, Bondy ) for pro- Strait (Indonesia): a review of the Krakatau programme.
viding us the code for MNT representation. We Oceanol. Acta, 10: 3 l-42.
benefited of discussions with R. Scandone, I. Diament, M., Harjono, H., Arsadi, E., Bonvalot, S., Dahrin,
Yokoyama, P. Vincent and of comments on the D., Deplus, C., Dubois, J. and Zen Jr, M.T., 1991. A geo-
manuscript by M. Semet, R. Scandone and an physical study of the Krakatau volcanic complex. EUG
VI Strasbourg, Terra, 3( 1): 266.
anonymous reviewer. IPGP contribution 13 17. Druitt, T.H. and Sparks, R.S.J., 1984. On the formation of
calderas during ignimbrite eruptions. Nature, 3 10: 679-
681.
References Effendi, A.C., Bronto, S. and Sukhyar, R., 1986. Geological
map of Krakatau Volcanic Complex. Volcanological Sur-
Alatorre-Zamura, M.A. and Campos-Enriques, J.O., 1991. La vey of Indonesia.
Primavera Caldera (Mexico): Structure inferred from Escher, R.G., 1919. De Krakatau groep als Vulkaan. Handel.
gravity and hydrogeological considerations. Geophysics, Eerste Nederl. Ind. Natuurw. Congr., pp. 28-35.
56: 992-1002. Fedi, M., Nunziata, C. and Rapolla, A., 1991. The Cam-
Aramaki, S., 1984. Formation of the Aira Caldera, southern pania-Campi Flegrei area: a contribution to discern the
Kyushu, -22 000 years ago. J. Geophys. Res., 89(BlO): best structural model from gravity interpretation. J. Vol-
8485-8501. canal. Geotherm. Res., 48: 5 I-59.
Barberi, F., Cassano, E., Torre, P.L. and Sbrana, A., 1991. Francis, P.W., 1985. The origin of the 1883 Krakatau tsun-
Structural evolution of Campi Flegrei caldera in light of amis. J. Volcanol. Geotherm. Res., 25: 349-363.
volcanological and geophysical data. J. Volcanol. Geoth- Francis, P.W. and Self, S., 1983. The eruption of Krakatau.
erm. Res.. 48: 33-49. Sci. Am., 249: 146-159.
Francis, P.W., G’Callaghan, L., Kretzschmar, G.A., Thorpe, position of tephra-layers in deep-sea sediments off West-
R.S., Sparks, R.S.J., Page, R.N., de Barrio, R.E., Gillou, em Indonesia. J. Volcanol. Geotherm. Res., 5: 67-86.
G. and Gonzalez, O.E., 1983. The Cerro Galan ignim- Nishimura, S., Abe, E., Nishida, J., Yokoyama, T., Dharma,
brite. Nature, 301: 51-53. A., Hehanusa, P. and Hehuwat, F., 1984. A gravity and
Hallinan, S., Brown, G. and Rymer, H., 1992. Geophysical volcanostratigraphic interpretation of the Lake Toba Re-
assistance in geothermal field assessment and monitor- gion, North Sumatra, Indonesia. Tectonophysics, 109:
ing, Miravalles, Costa-Rica. Final Rep. (ODA Ref. FNI 253-272.
8788/541/093) Dep. Earth Sciences, The Open Univer- Nishimura, S., Nishida, J., Yokoyama, T. and Hehuwat, F.,
sity, Milton Keynes, England, 2 11 pp. 1986. Neo-Tectonics of the Strait of Sunda, Indonesia. J.
Harjono, H., 1988. Gt?odynamique du Detroit de la Sonde Southeast Asian Earth Sci., 1: 8 l-9 1.
(Indonbie): Apports des donnees de microsismicite et Noujaim, A.K., 1976. Drilling in high temperature and ov-
implications volcanologiques. Thesis, Univ. Paris-Sud, erpressured area Sunda Straits, Indonesia. Proc. Indone-
262 pp. sian Petroleum Association. Fifth Annu. Convention, pp.
Harjono, H., Diament, M., Nouaili, L. and Dubois, J., 1989. 211-214.
Detection of magma bodies beneath Krakatau volcano Oide, K., 1968. Geotectonic conditions for the formation of
(Indonesia) from anomalous shear waves. J. Volcanol. the Kmkatau-type calderas in Japan. Pacific Geol., 1: 119-
Geotherm. Res., 39: 335-348. 135.
Harjono, H., Diament, M., Dubois, J. and Larue, M., 199 1. Parasnis, 1962. Principles of Applied Geophysics. Methuen,
Seismicity of the Sunda Strait: evidence for crustal exten- London, 176 pp.
sion and volcanological implications. Tectonics, 10: 17- Patella, D., 1988. An unambiguous derivation of the Bou-
30. guer gravity anomaly. Boll. Geof. Teor. Appl., 30( 119-
Heiken, G. and MC Coy, F., Jr, 1984. Caldera development 120): 345-352.
during the Minoan eruption, Thira, Cyclades, Greece. J. Rymer, H. and Brown, G.C., 1986. Gravity fields and the
Geophys. Res., 89(BlO): 8441-8462.
interpretation of volcanic structures: Geological discrim-
Isherwood, W., 1976. Proc. 2nd U.N. Symposium on the De-
ination and temporal evolution. J. Volcanol. Geotherm.
velopment and Use of Geothermal Resources. U.S. Gov-
Res., 27: 229-254.
ernment Printing Office, Washington D.C., 2: 1065-1073.
Scandone, R., 1990. Chaotic collapse of calderas. J. Vol-
Iyer, H., 1984. Geophysical evidence for the locations, shapes
and sizes, and internal structures of magma chambers be-
Scandone, R., 199 1. A reply to the comments by Yokoyama
neath regions of Quaternary volcanism. Philos. Trans. R.
and De la Cruz-Reyna and some other ideas on the col-
Sot. London, Ser. A, 310: 473-510.
lapse of calderas and the generation of ignimbrites. J. Vol-
Judd, J.W., 1889. The earlier eruptions of Krakatau. Nature,
40: 365-366.
Scandone, R., Bellucci, F., Lirer, L. and Rolandi, G., 1991.
Kane, M.F., Mabey, D.R. and Brace, R.L., 1976. A gravity
and magnetic investigation of the Long Valley Caldera, The structure of the Campanian Plain and the activity of
Mono County, California. J. Geophys. Res., 81: 754-762. the Neapolitan volcanoes (Italy). J. Volcanol. Geotherm.
Larue, M., 1983. Corindon IX Cruise Report. Res., 48: 1-31.
Lassal, O., Huchon, P. and Harjono, H., 1989. Extension Self, S. and Rampino, M.R., 1981. The 1883 eruption of
crustale dans le Detroit de la Sonde (Indontsie ). Don- Krakatau. Nature, 294: 699-704.
&es de la sismique reflexion (Campagne Krakatau). CR. Self, S. and Rampino, M.R., 1982. Comments on “A Geo-
Acad. Sci. Paris, 309: 205-2 12. physical interpretation of the 1883 Krakatau eruption”
Latter, J.H., 1981. Tsunamis of volcanic origin: summary of by I. Yokoyama. J. Volcanol. Geotherm. Res., 13: 379-
causes, with particular reference to Krakatau. Bull. Vol- 386.
canal., 44: 467-490. Sigurdsson, H., Carey, S., Mandeville, C. and Bronto, S.,
Longman, I.M., 1959. Formulas for computing the tidal ac- 1991a. Pyroclastic flows of the 1883 Krakatau eruption.
celeration due to the moon and the sun. J. Geophys. Res., EOS Trans. Am. Geophys. Union, 72: 377.
64: 2351-2355. Sigurdsson, H., Carey, S. and Mandeville, C., 1991b. Sub-
McBimey, A., 1990. An historical note on the origin of cald- marine pyroclastic flows of the 1883 eruption of the
eras. J. Volcanol. Geotherm. Res., 42: 303-306. Krakatau Volcano. National Geographic Research and
Nafe, J.E. and Drake, C.L., 1963. Physical properties of ma- Exploration, 7: 310-327.
rine sediments. In: The Sea. McGraw-Hill, Wiley Inter- Simkin, T. and Fiske, R.S., 1983. Krakatau 1883: The Vol-
Sciences, London-New York, pp. 794-8 15. canic Eruption and its Effects. Smithsonian Inst. Press,
Nettleton, L.L., 1976. Gravity and Magnetics in Oil Pros- Washington, D.C., 464 pp.
pecting. MC Graw-Hill, New York, NY. Stehn, C.E., 1929. The geology and volcanism of the Kraka-
Neumann van Padang, M., 1933. De Krakatau voorheen en tau Group. Proc. Fourth Pacific Sci. Congress (Batavia),
thans. De Tropische Natuur, 22: 137-l 50. pp. l-55.
Ninkovich, D., 1979. Distribution, age and chemical com- Sudrajat, A., 1982. The morphological development of Anak
52 C. Deplus et al. /Journal of Volcanologyand Geothermal Research 64 (I 995) 23-52
Krakatau Volcano, Sunda Strait. Geol. Indonesia, 9: I- Regional Gravity and Magnetic Anomaly Maps. Society
11. of Exploration Geophysicists, Tulsa, OK, pp. 36 I-376.
van Bemmelen, R.W., 1941. Krakatau. Bull. East Indian Vol- Williams, H., 194 I. Calderas and their origin. Bull. Dept.
canal. Surv., 18: 53-60. Geol. Sci., Univ. Calif., 25: 239-346.
Verbeek, R.D.M., 1885. Krakatau. In: T. Simkin and R.S. Williams, H. and McBimey, A.R., 1979. Volcanology. Free-
Fiske (Editors), Krakatau 1883 : the volcanic eruption man, Cooper and Co, San Francisco, CA, 397 pp.
and its effects. Smithsonian Inst. Press, Washington, D.C., Yokoyama, I., 198 I. A Geophysical interpretation of the 1883
pp. 169-217. Krakatau eruption. J. Volcanol. Geotherm. Res., 9: 359-
Vincent, P.M. and Camus, G., 1986. The origin of the 1883 378.
Krakatau tsunamis, by P.W. Francis. Discussion, J. Vol- Yokoyama, I., 1987. A scenario of the 1883 Krakatau Tsun-
canal. Geotherm. Res., 30: 169- 177. ami. J. Volcanol. Geotherm. Res., 34: 123-132.
Vincent, P.M., Camus, G. and Gourgaud, A., 1989. Hy- Yokoyama, 1. and Hadikusumo, D., 1969. Volcanological
pothtse sur la structure du Krakatau. Bull. Sect. Volcanoi. survey of Indonesian volcanoes. Part 3: A gravity survey
Sot. Geol. France, 13. on the Krakatau Islands, Indonesia. Bull. Earthq. Res.
Volcanological Survey of Indonesia, 1982. Bathymetric map Inst., 47: 991-1001.
(Escher, 19 19) and topographic map ( 1979) of Kraka- Zen, M.T. and Sudradjat, A., 1983. History of the Krakatau
tau volcanic complex. volcanic complex in Strait Sunda and the mitigation of its
Williams, D.L. and Finn, C., 1985. Analysis of gravity data future hazards. Bul. Jurusan Geol., 10: l-28.
in volcanic terrain and gravity anomalies and subvolcanic Zubin, MI., Kozikev, AI. and Luchitskiy, AI., 1992. A grav-
intrusions in the Cascade Range, USA, and at other se- itationally derived model of Klyuchevskoi Volcano,
lected volcanoes. In: W.J. Hinze (Editor), The Utility of Kamchatka. Volcanol. Seismol., 12: 628-646.

Tsunami Krakatau

Uploaded by

Copyright:

Available Formats

Tsunami Krakatau

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Tsunami Krakatau

Uploaded by

Copyright:

Available Formats

Joumalof volcanology

and geotheml resemh

Journal of Volcanology and Geothermal Research 64 ( 1995) 23-52

Inner structure of the Krakatau volcanic complex (Indonesia) from

Received 30 April 1993; accepted 12 May 1994

1. Introduction of geological investigations, only few geophysi-

0377-0273/95/$09.50 0 1995 Elsevier Science B.V. All rights reserved

The Krakatau volcanic complex lies in the

matra islands and belongs to the volcanic arc re-

al., 1987). The Krakatau volcanic complex is also 6”12’S

4. Previous geophysical studies tween experimental results on formation of cra-

mechanism was proposed by Verbeek as early as

BATHYMETRY DATA DISTRIBUTION

Ansk Krakatau 1883csldera

“1” flaw I rrm

Changes in the caldera since 1883 5.4. Growth ofAnak Krakutau

,,j:#+._ ..... ;.;!;‘-

105”20 105”23 105’26 105”29

GRAVITY DATA DISTRIBUTION

Density = 2.11 g/cm3

105020 105O23 105O26 105O29

105O20 105O23 105O26 105O29

Fig. 15. (continued).

Subsurfacestructure 1984). So we choose 2.6 g/cm3 for the density of

106020 105O23 1tw26 lW26

106020 106023 105O26 106026

Computed gravity effect mGa1 of density 13- 15 ) over the whole

You might also like