Roads over lava tubes in Cheju Island, South Korea

Engineering Geology 66 (2002) 53 – 64
www.elsevier.com/locate/enggeo
Roads over lava tubes in Cheju Island, South Korea
A.C. Waltham a,*, H.D. Park b
a
Civil Engineering Department, Nottingham Trent University, Nottingham NG1 4BU, UK
School of Civil, Urban and Geosystem Engineering, Seoul National University, Seoul 151-742, South Korea
b
Accepted 18 January 2002
Abstract
There are 59 known lava tubes in the basalts of Cheju Island, South Korea. They have formed by either crusting or injection,
and the structural properties of the rock over them vary with their formative processes. Two tubes are crossed by roads that
stand on very thin rock arches. One road has been built on a bridging concrete slab, the other is unprotected, but both appear to
be stable. A ratio of 3:1 for tube width to roof thickness appears to ensure integrity under normal engineering loads. D 2002
Elsevier Science B.V. All rights reserved.
Keywords: Lava; Cave; Collapse; Ground stability; Highways
1. The shield volcano of Cheju Island
2. Lava tubes in basalt
Cheju Do (Cheju Island) lies 80 km south of
mainland Korea, almost midway between China and
Japan. It is 75 km long and 30 km wide (Fig. 1), and
consists entirely of a single shield volcano, which is
distinguished by having over 360 parasitic cones on
its flanks. The central peak of Hallasan rises to 1950
m. It last erupted in 1007 AD. On the higher slopes,
open grassland covers roughly the area formed by the
Hallasan volcanics, with woodland forming a zone at
lower altitude. Nearly a million people thrive on
farming and tourism in a well-developed coastal strip.
This is formed of old basaltic lavas, which generally
provide stable ground, except over the old lava tubes.
Though layers of scoria and aa (blocky) rubble can
create zones of loose and weak rock, most basalt
solidifies into a very strong rock. Good ground conditions are however compromised where lava tubes
have been left to form extensive voids just below the
ground surface. A lava tube may be formed where a
lava flow develops a cooled and solidified crust, while
molten lava continues to flow beneath the surface
(Kauahikaua et al., 1998). Alternatively, liquid basalt
may be injected beneath a crust of solidified basalt
from an earlier stage of the eruption; in this style, lava
flows are inflated by new material that arrives through
tubes and may solidify before ever reaching the surface (Walker, 1991; Hon et al., 1994). When an
eruption ceases suddenly, or when flow is switched
to another outlet, the molten lava may drain out, to
leave an empty tube down the axis of the lava flow.
Most extensive basaltic lava flows are tube-fed.
Heat loss to the atmosphere means that lava can rarely
*
Corresponding author. Tel.: +44-115-848-2133.
E-mail addresses: [email protected] (A.C. Waltham),
[email protected] (H.D. Park).
0013-7952/02/$ - see front matter D 2002 Elsevier Science B.V. All rights reserved.
PII: S 0 0 1 3 - 7 9 5 2 ( 0 2 ) 0 0 0 3 0 - 3
54
A.C. Waltham, H.D. Park / Engineering Geology 66 (2002) 53–64
Fig. 1. Main features of the topography and geology of Cheju Island.
flow on the surface for more than a few kilometres,
before it cools and solidifies. But a lava stream that
has a solidified roof crust suffers negligible heat loss,
and can flow underground for many kilometres. The
mechanisms of lava tube formation and extension are
a complex function of the lava’s rheology, its supply
rates and chemistry, and the ground topography
beneath the flow. Most active tubes observed on
Hawaii are less than a metre in diameter, and these
must be regarded as the normal state in shield volcano
basalts. However, lava tube caves, which are generally
2– 10 m in diameter, can form as the main feeder
within a long and mature flow. These may subsequently drain out to survive as open tubes—that are
known to be up to 40 km long. Though tubes are a
feature of most basalts, they only drain out to leave
open tubes where the lava’s viscosity and gradient
permit; others are sealed as the lava solidifies within
them or as hot lava deforms and closes in on them.
Many basalt flows have no open tubes, and some
tubes were filled by lava in later stages of the same
eruption. Many tubes at high level within older lavas
have been lost by subsequent surface lowering; there
are few long tubes recorded in lavas older than about
200,000 years.
Tubes formed by injection may lie beneath basalt
tens of metres thick. However, tubes formed by roofing of surface flows lie just below the surface, and
consequently provide a potential hazard for engineering construction. The crusts of rock over these tubes
are normally no more than a few metres thick. Nearly
all entrances to lava tubes are through sections of roof
collapse. Some of these collapses occurred at an early
stage, when lava was still flowing beneath; skylights
are commonly seen in active lava flows, where the
thin crust has fallen in, so that molten lava can be seen
flowing within the tube (Fig. 2). Other collapses occur
at later dates. The minor earthquakes that are common
on active volcanoes have caused tube roof collapses 5
or 10 years after the eruption ended and the lava
cooled, revealing tubes that had not previously been
seen or known. And some collapses occur much later
in response to weathering and weakening of the thin
basalt arches above the tubes.
A.C. Waltham, H.D. Park / Engineering Geology 66 (2002) 53–64
55
Fig. 2. A skylight into an active lava tube on Hawaii. The opening is 2 m across and yellow-hot lava can be seen 3 m below, flowing in a tube
about 5 m in diameter.
2.1. The lava tubes of Cheju
There are 59 known lava tubes on Cheju Island.
They have a combined length of over 42 km, and
eight tubes have lengths greater than a kilometre. All
but 700 m of the known tubes are in the Pyoseonri
lavas (Fig. 1). Most of the known tubes lie beneath the
developed lands of the coastal strip, and there are
undoubtedly more tubes that have no roof collapses,
as yet, and therefore remain inaccessible and undocumented. Three of the tubes visited by the authors are
worthy of note.
3. Manjang Gul
The longest single lava tube on Cheju is Manjang
Gul (located on Fig. 1). It has a mapped length of
nearly 8 km, though its accessible ends are merely
collapse zones, beyond which the open tube continues.
A notable feature of Manjang Gul is the large size of
its tube (in comparison to most others in the world);
much of it is 15 m wide and 10 m high (Fig. 3), and
some sections are 20 – 25 m wide. The entrance is
through a simple natural collapse, about 12 m across,
where the original tube rose slightly and the roof was
less than 4 m thick. The cave is now a popular tourist
site. Precautionary engineering works are modest, but
include a sprayed concrete lining under the entrance
arch where it is exposed to weathering above the show-
cave path. This entrance is in a high-level loop off the
main tube. Just up-flow of the tourist entrance, the two
tubes merge at the lower level, beneath solid rock over
10 m thick (Fig. 3).
Along the kilometre of tube that contains the tourist
trail, the floor is a pahoehoe surface with a uniform
gradient of 0.4j. Most of the roof appears to be a broad,
stable arch. Sections where the roof is only about 3 m
high have profiles that are almost circular arches.
Hanging from the smooth arches are small stalactites
of glassy basalt. These formed where a glaze dripped
from the roof after it was heated to melting point by hot
gasses above the molten lava stream within the tube.
Their presence indicates that the roof is in its original
state, with no subsequent collapse.
Some other sections of the roof are much higher, but
lack any collapse debris beneath them. The thick cover
of basalt over this tube suggests that it was formed by
injection, when the lava could flow over loops under its
own hydrostatic pressure. The crests of the loops may
have been modified by block plucking and roof erosion
by the lava when it was under pressure within its tube.
Roof blocks that collapsed into the hot lava were swept
away downstream; some blocks survive, welded into
the lava floor, and with no sign of roof collapse directly
above them. The uniform profile of the floor indicates
that any original rises were removed by thermal
erosion and entrenchment of the lava.
Other sections of tube contain high collapse domes
above extensive piles of blockfall, with individual
56
A.C. Waltham, H.D. Park / Engineering Geology 66 (2002) 53–64
Fig. 3. The large main tube in Manjang Gul. The roof structure appears to indicate some weathering along fracture lines, but there is no collapse
at this point.
blocks up to 4 m across. Some of these sites are
monitored, but there are no observed cases of new
rockfall, and it appears that the roofs have developed
into stable voussoir arches within the fractured rock.
Some high sections have roof joints, at a spacing of
about 1.5 m, along which there are zones of discoloration that widen at the joint intersections. It appears
that roof collapse has stopped upwards into basalt of
weathering zone II, with the first traces of spheroidal
weathering, where the ground surface is about 7 m
above the cavern roof.
4. Sung Gul
Almost at the western tip of Cheju (Fig. 1), Sung
Gul is a lava tube nearly a kilometre long beneath the
village of Shinchang; it is sometimes known as
Fig. 4. The collapse entrance to Sung Gul, between the houses of Shinchang.
A.C. Waltham, H.D. Park / Engineering Geology 66 (2002) 53–64
57
Fig. 5. Outline plan survey of the Sung Gul lava tube where it lies beneath the island’s perimeter road.
Shinchang Gul. It also extends beneath the island’s
main perimeter road.
The entrance to Sung Gul is an ancient collapse
hole, formed where the roof was thinnest over a gentle
upward loop in the tube profile. It lies between the
houses of Shinchang (Fig. 4), and there are steps
down to the cave floor, as villagers use it for cool
storage of farm produce. Down-flow, the tube is easily
followed for 400 m, as it is all 2– 4 m high and 5– 14
m wide (Fig. 5). The floor is a rough and partly
broken pahoehoe (smooth or ropy skinned) lava, and
the gradient is less than a degree.
Most of the roof of the Sung Gul tube is a broad,
flat arch, with small glaze structures that indicate it is
in its original state. At the upper end of the two oxbow loops, the tube is 14 m wide, with an almost flat
roof that shows no sign of collapse (Fig. 6). For much
of its length, the tube is lined with at least one shell of
lava, 200 –400 mm thick. This was formed after the
tube was partly drained, so that its roof cooled
slightly. A subsequent pulse of lava re-filled the tube,
and the shell was formed where it chilled against the
cooler roof. Some small sections of this shell have
fallen away (Fig. 7); most of these falls are old, and
Fig. 6. An almost flat roof 14 m wide just up from the down-flow ox-bow loop in the Sung Gul lava tube, with no sign of rockfall or collapse.
58
A.C. Waltham, H.D. Park / Engineering Geology 66 (2002) 53–64
Fig. 7. Breakdown of the chilled basalt shell that lines part of the Sung Gul lava tube.
there are no signs of collapse of the rock outside the
detaching shell.
Five short sections of the tube have more extensive
roof collapse (Fig. 5). Irregularly shaped blocks up to
2 m across have fallen away to create large piles of
breakdown under higher, domed sections of roof.
Some of these dome roofs appear to have developed
into stable arches within the jointed basalt (Fig. 8),
and there are no signs of ongoing collapse on any of
them.
The collapse section furthest down-flow extends
into an area of total roof failure (Fig. 9). This has
created a rocky depression at least 15 m long (now
partly backfilled) on the surface. Part of the tube
Fig. 8. The third collapse zone down-flow from the village entrance in the Sung Gul lava tube.
A.C. Waltham, H.D. Park / Engineering Geology 66 (2002) 53–64
59
Fig. 9. Sketch profile along the lower part of Sung Gul. The tube has an unbroken roof where it passes beneath the road, but has extensive
breakdown down-flow of the road as far as the zone of total collapse through which lies the access pipe.
survives round the north side of the collapsed blocks
(Fig. 5), and it appears that the roof failure occurred in
a very wide section of the tube. Access through the
blockfall is now via a 500-mm-diameter concrete pipe
that opens into the side of a concrete culvert built
across the collapse depression.
4.1. The road over Sung Gul
Cheju Island’s perimeter road crosses over the
Sung Gul tube between its two down-flow collapse
sections (Fig. 5). The road was a rough track until
1961, when it was paved. This was widened to two
lanes in 1974, and to four lanes in 1998. It was not
realised that the cave in Shinchang village extended
under the road until it was found in 1994 during the
site investigation for the last phase of road widening.
No diversion of the road was then considered, as the
line was already established.
Purely by good luck, the line of the road lay over
the most stable part of the cave, where a smooth arch
barely 6 –10 m wide rises 2 m above a clean floor of
ponded lava (Fig. 10). A shell of chilled lava lines part
of the walls; it is incomplete, but there are no fallen
blocks beneath, and any fallen shell sections may have
been carried away when the lava was still flowing.
Glaze drips on the roof show that it is in its original
state. Assessment of the cave in 1994 found that the
roof consisted of 3.5 m of rock and 2 m of soil. The
internal report (unpublished) concluded that ‘‘the rock
seems to be strong enough, but jointing and fractures
could be a potential problem for stability of the road’’.
Fig. 10. The stable arch in the Sung Gul lava tube, directly beneath the main road.
60
A.C. Waltham, H.D. Park / Engineering Geology 66 (2002) 53–64
Lack of data on the rock structure precluded any
quantitative analysis, but it was deemed necessary to
strengthen the ground.
The soil was removed, and a flat surface was
prepared by placing clean sand at least 100 mm deep
over the uneven rockhead. For a 20 m length of the
road over the cave, two layers of expanded polystyrene blocks (each 500 mm thick) were placed on the
sand for the full 11.2 m width of the road (Fig. 9).
Centred on this, a slab of reinforced concrete 400 mm
thick and 30 m long, also 11.2 m wide, completely
spans the cave (Fig. 5). The road pavement is 50 mm
thick over the concrete. The reinforced slab can act as
a single-span bridge element (Fig. 11), eliminating
any hazard from even a complete failure of the lava
tube roof.
4.2. The stability of Sung Gul beneath the road
Any analysis of rock stability over an underground
void relies on adequate information concerning the
density and pattern of the fractures within the rock
arch. These data are almost impossible to obtain for a
unique site, and can only be an approximation for a site
where exposures, comparisons and destructive test
data are available. As intact rock, basalt is very strong,
with UCS normally over 200 MPa, but it varies
enormously in its fracture patterns and consequently
in its rock mass strength.
The state and strength of the rock forming the roof
of a lava tube depends largely on how the lava stream
was roofed over, or whether it was formed by injec-
tion into an inflating flow. There are various possibilities, especially for the surface channel roofing:
1.
accretion onto the banks and levees on both
sides, until they coalesce over the flow, leaving
vertical fractures in the arch centre;
2. accretion and wedging of floating slabs,
leaving an irregular blocky structure;
3. crust formation by toes and small sheets of
pahoehoe lava, leaving a pattern of fractures
around sub-horizontal lava lenses (as in Fig.
2);
4. tube formation by injection beneath a thin or
thick roof of solid lava;
5. addition of shells inside the tube, formed when
lava pulses fill the tube and leave chilled
linings against the cooler roof and walls;
6. addition of subsequent lava layers to thicken
the roof.
Where tubes originated as surface channels, process nos. 1 –3 normally form the initial roof. This may
be thickened by process nos. 5 –6, or may be thinned
by subsequent melting away of part of the roof
(thermal erosion). In all cases, the ‘‘fractures’’ within
the basalt may be important potential weaknesses,
especially where they have opened up during cooling
contraction. Alternatively, they may weld while still
hot, to represent minimal structural weakness in what
appears to be very massive basalt. Fractures may
develop subsequently, under changes of tectonic stress
(which is common in an active volcano), or in res-
Fig. 11. Cross section through the Sung Gul lava tube and the concrete slab that carries the main road above it.
A.C. Waltham, H.D. Park / Engineering Geology 66 (2002) 53–64
ponse to weathering disturbance. Injection tubes,
formed by process no. 4, normally have more massive
rock roofs. Tensile fracturing of their roofs, caused by
inflation pressure, is recorded in the form of lava
tumuli and breakouts (Walker, 1991), but many of
these fractures are subsequently filled and welded by
lava inflow.
Most of the roof rock that is visible within the Sung
Gul tube does appear to be in good condition, with few
major structural weaknesses. There are fractures in the
basalt, but they are mostly tight and irregular. It therefore appears that the rock mass factor of the basalt is
high, and its mass strength under compression is close
to its intact strength. The internal shells that are falling
away from open fractures are not relevant to stable
arches that can form in the overlying rock; multiple
shells could exist (and remain invisible until they
partially fail), but concentric arched fractures would
have little influence on a compression arch formed
parallel to them. Fractures within the basalt have little
effect on tube roof integrity where they lie within the
zone of compression that creates a self-supporting arch
over a void. Rock that remains beneath the compression zone is in tension and is subject to deformation
under loading.
The rock arch over the Sung Gul tube has been
examined by Finite Difference Analysis within a
FLAC Ubiquitous Joint Model that was originally
applied to assessment of cave roofs in sandstone
(Waltham and Swift, in prep). A worst-case scenario
was created for the road over Sung Gul, with a cave
width of 12 m and a thickness of rock roof of 3 m.
The basalt UCS was taken as 160 MPa, and a rock
mass factor of 0.1 was applied as a reduced strength
on the ubiquitous joints. Failure of the tube roof
occurs under a vertical loading of 4000 kN applied
on a square pad of 0.5 m edge length. This is far in
excess of any possible highway loading or design
criteria. However, the figure relies on estimates of the
fracture weakness within the basalt, and is unrealistic.
The FLAC programme that models relatively homogeneous rocks is inappropriate for the investigation of
strong, fractured basalt. Further analysis was outside
the scope of this short field project.
The available data suggest that the concrete slab
may not have been entirely necessary to ensure
integrity of the road. Volcanologists recognise that it
is extremely difficult to interpret how a tube formed,
61
merely by inspecting its cold state; by implication, it
is equally difficult to appreciate the fracture state of an
in situ rock roof. The cost of the concrete slab was
small as a proportion of the project costs, and its
inclusion was a reasonable precaution. A layer of
heavy geogrid in the road sub-base offers a cheaper
precautionary measure against potential ground collapse; but this is inappropriate over such a large rock
void at shallow depth where any failure is likely to
cause instantaneous total collapse of an area perhaps 8
m in diameter. The alternative options of collapsing
the tube and building over the stabilised breakdown,
or filling the tube with concrete, both represented
higher costs than the simple bridging slab where
sound rock was available on each side.
The integrity of the smaller rock arch, just 500 mm
thick, over the Sung Gul entrance (Fig. 4) also relies
on a lack of critical fractures within the basalt. It could
not be adequately modelled on the FLAC programme.
The rock appears to be intact, and modelling intact
rock produces failure at far in excess of any construction loading. The arch thickens as the roof
descends immediately inside the cave, where it becomes even stronger. A crack within the wall of the
house on top of the cave indicates settlement at one
end, but this appears to be related to soil compaction
under the very shallow footings. The end wall of the
house appears to be completely safe, despite its
impressive location.
5. Cheamchon Gul
Close to the village of Hyopchae, Cheju’s old
perimeter road passes over the short lava tube of
Cheamchon Gul. A new section of road, further inland,
avoids the cave crossing (Fig. 1).
Only about 100 m of this tube is open and
accessible. Most of this is 8 m wide, with its roof
no more than 3 m below ground level. Just north of
where it crosses beneath the road, the tube is completely collapsed over a length of 40 m, and its
partially collapsed continuation is flooded. Adjacent
to the road crossing, the tube has two collapse
windows (Fig. 12). The larger is the entrance with
steps down into the cave, where a level floor has been
created for occasional use as a village night club. The
smaller skylight is in the soft shoulder of the road,
62
A.C. Waltham, H.D. Park / Engineering Geology 66 (2002) 53–64
Fig. 12. Sketch survey of Cheamchon Gul lava tube where it lies beneath the old main road.
Fig. 13. The two roof collapses that create the entrance and the skylight in Cheamchon Gul, seen from the tube’s continuation beneath the road.
A.C. Waltham, H.D. Park / Engineering Geology 66 (2002) 53–64
63
Fig. 14. Fracture opening recorded in the roof of Cheamchon Gul lava tube when vehicles passed over the road above (by courtesy of KICT).
immediately adjacent to the tarmac, from which it is
separated by a low steel fence. Parts of the tube roof
have an irregular profile left by past block collapse
(Fig. 13). The low wide profile of the tube suggests
that it may have originated by lava injection, beneath
2 –3 m of older basalt. Among the irregular fractures
in the roof rock, there appears to be a crude arched
layering that may represent inflated flows of pahoehoe
older than the tube.
The road stands on rock that is about 3 m thick,
except where it tapers to only 2 m thick at the
southern edge (Fig. 12). The fractured basalt appears
to have stabilised in a low compression arch. This
arch thins between the two skylights (Fig. 13), where
its more broken structure suggests any future failure is
most likely to occur. The rock under the road is in
better condition, less fractured, and forming a thicker
and wider arch. Exposed fractures in the roof of the
cave have been monitored under dynamic loading by
traffic on the road directly above. Strain gauges were
attached across two roof fractures, one at the tube
centre and one near the wall. When light trucks passed
over the cave, the fractures opened by about 7 Am
(Fig. 14); the effect of cars was not easily recognisable. After unloading, the fractures closed again, as
the entire rock mass was behaving as an elastic
material. This rock arch over the cave appears to be
stable, and does not warrant a reinforcing concrete
slab, even though part of the arch is thinner than that
over Sung Gul.
6. The lava tube hazard
Any construction project on young basalt lavas
may encounter a potential hazard from open tubes at
shallow depth. The statistical chance of a tube lying
beneath a particular engineered structure is extremely
small, but it is impossible to predict where they lie
beneath a mature landscape. Direct exploration and
mapping of tubes from accessible collapse entrances
is the cheapest method of establishing their positions,
but many tubes exist without known accessible
entrances. Borehole exploration is expensive. Geophysical searches can be productive, and various
investigators of lava tubes have preferred resistivity
(Kwon et al., 1998) or magnetic (Wood, 2001) surveys; these methods may be more economical than the
microgravity surveys that are most useful for void
detection in rocks less conductive than basalt.
Recorded collapses of lava tubes (in Hawaii) have
all been in the initial stages of road building, when
very large bulldozers provide point loadings far heavier than those of road traffic; their use appears to offer
a convenient means of load testing the ground before
a road is in use.
64
A.C. Waltham, H.D. Park / Engineering Geology 66 (2002) 53–64
The high strength of basalt means that most lava
tubes are stable even where their roofs are disconcertingly thin. A ratio of 3:1 for tube width to roof
thickness (in unweathered rock) would appear to be
adequate to establish a sound rock arch that will bear
all but the heaviest structures. This ratio should
provide an appropriate safety factor where the basalt
lava has only the low density of fractures that is
typically observed in tube roofs. Zones of more
fractured lava can account for weaker tube roofs. As
a by-product from a geophysical cavity search, a
measured field seismic velocity could indicate rock
mass quality by correlation with existing empirical
data.
Highways and light structures can stand in full
safety above the great majority of lava tubes, but
heavy structural loadings should only be applied to
basalt lava after appropriate investigation of the local
ground conditions.
Acknowledgements
The Royal Society generously provided a grant for
travel costs in support of the programme of academic
exchange between members of the Geohazards Group
in Nottingham Trent University and members of the
Department of Geological Engineering at Seoul
National University. The authors thank colleagues at
the Korean Institute of Construction Technology
(KICT) for contributing their data on the Cheamchon
Gul roof monitoring, Bill Halliday and Chris Wood
for helpful comments in discussion, and Jun Yong Lee
for assistance in the field. Part of this work was
supported by SNU Research Institute of Engineering
Science and BK project 21 in 2001.
References
Hon, K., Kauahikaua, J., Denlinger, R., Mackay, K., 1994. Emplacement and inflation of pahoehoe sheet flows: observations
and measurements of active lava flows on Kilauea Volcano,
Hawaii. Geological Society of America Bulletin 106, 351 – 370.
Kauahikaua, J., Cushman, K.V., Mattox, T.N., Heliker, C.C., Hon,
K.A., Mangan, M.T., Thornber, C.R., 1998. Observations on
basaltic lava streams in tubes from Kilauea Volcano, island of
Hawaii. Journal of Geophysical Research 103 (B11), 27303 –
27323.
Kwon, B.-D., Lee, H.-S., Oh, S.-H., Lee, C.-K., 1998. Application
of multiple geophysical methods in investigating the lava tunnel
of Manjanggul in Cheju Island. Economic and Environmental
Geology 31, 535 – 545.
Walker, G.P.L., 1991. Structure and origin by injection under surface crust of tumuli, lava rises, lava-rise pits and lava-inflation
clefts in Hawaii. Bulletin of Volcanology 53, 546 – 558.
Waltham, A.C., Swift, G., in preparation. Bearing capacity of rock
over underground cavities. Quarterly Journal of Engineering
Geology and Hydrogeology.
Wood, C.J. (Ed.), 2001. Laki Underground 2000. Bournemouth
University, 36 pp.