Clay Minerals (1986) 21, 311-332

GEOTECHNICAL
OF

THE

PROPERTIES

MONASAVU

AND

HALLOYSITE

D.

BEHAVIOUR
CLAY,

FIJI

J. K N I G H T

Associate and Head of Geotechnical Engineering, Sir Alexander Gibb & Partners, Earley House,
London Road, Earley, Reading RG6 1BL
(Received 11 December 1985; revised 27 February 1986)

A B S T R A C T : Geotechnical properties and behaviour are described of the very wet,

halloysitic, residual clay found in the dense rain forest at Monasavu, Viti Levu, Fiji, where the
annual rainfall can exceed 5 m. The tropical climatic conditions have caused deep weathering of
sandstones and produced a highly plastic clay with low density and a natural moisture content
greatly in excess of the standard compaction optimum. This clay was found to contain halloysite
which was 'amorphous' rather than crystalline. The material was used in this natural state in an
85-m high rockfill dam at Monasavu Falls as an unusually soft core, the construction of which
involved unconventionally light compaction by low-ground-pressure-tracked dozers. Its resulting
behaviour in terms of three-dimensional total and effective stresses, stress paths and
deformations throughout the construction, impounding and full reservoir stages was closely
monitored. This behaviour is examined in the light of the clay's classification, mineralogical,
compaction and engineering properties determined before and during construction. Despite its
unusual properties, it is concluded that the clay is a good engineering material, behaving like
others containing halloysite in the more common tubular form, and, moreover, that the high
natural moisture content is of positive benefit.

The purpose o f this paper is to describe the geotechnical properties and behaviour of the
residual clay material used for the core of M o n a s a v u Dam, Fiji, and to examine this use
and behaviour in the light of those properties determined both prior to and during
construction.
M o n a s a v u D a m is an 85-m high rockfill/clay core e m b a n k m e n t d a m situated at
M o n a s a v u Falls in Viti Levu, the main island o f Fiji (Fig. 1). It was constructed for the Fiji
Electricity Authority between 1979 and 1982, with impounding completed in 1983
(Knight, 1984).
The dam is situated on the wet side of the N a d r a u Plateau in the Central Highlands at an
elevation of 6 7 0 - 7 5 0 m, immediately u p s t r e a m of the substantial cliff where N a n u k u Creek
once formed the spectacular 126-m high M o n a s a v u Falls (Fig. 2). This plateau is part of a
mountainous barrier forming a transition zone between the wet east and dry west zones o f
the island. Although the latter experiences a m a r k e d dry season in the winter months the
annual variations of rainfall in the wet zone are less pronounced. Average monthly rainfalls
at M o n a s a v u from September 1980 to D e c e m b e r 1984 are shown in Table 1, from which
it m a y be deduced that the 19-month core construction period experienced a total rainfall
of 7536 mm.
1986 The Mineralogical Society

D. J. K n i g h t

312

PA C I F I C

SOUTH
N

(a)
0
6"

c~

q
c,

Bo

Monosavu

Lautoka
(b)

.o

- ~ - - ' ~ ~

Suva

Ground levels(ft}
~;:.~ A bove El.2000

lmAbo~

EI.3ooo

FIG. 1. Location map. (a) Fiji Islands. (b) Viti Levu.

TABLE 1. Monasavu rainfalldata.

Month

1980

1981

1982

1983

1984

Period average

Jan
Feb
Mar
Apr
May
June
July
Aug
Sept
Oct
Nov
Dec

--------444~
435
430
147

716
398
343
444
166
107
45
200
140
776
329
469

1072"
447
4285
320
175
305
338
515
145
176
841
429

537
1302"
867
293
208
102
158
477
168
368
432
416

254
228
724
457
390
510
78
196
59
155
429
354

645
594
591
379
235
256
155
347
191
382
492
363

TotM

--

5191

5328

3834

4622

* Cyclones.
~" Start of core construction.
$ Finish of core construction.

4133

Geotechnical properties of Monasavu clay

313

\
ersion
bnel

f
ee

i9.3

Spillway.

ITI

FIG. 2. Plan of dam and location of clay borrow area.

GEOLOGY
The dam site is underlain by a flat-lying sequence o f sedimentary rocks, belonging to the Ba
Group, into which is intruded a thick monzonite sill, which forms the Monasavu Falls
300 m downstream of the dam axis and constitutes the dam foundations in the river. At the
dam site a conglomerate occurs immediately above the sill and this is overlain by a
sequence of almost horizontally bedded sandstones of various grain sizes which form the
dam abutments (Fig. 3). The sand grains consist of reworked tuffaceous material and rock
fragments. Weathering of these lithic sandstones in the tropical climatic conditions has
produced a weathered profile varying in thickness from 3-25 m, most of which comprises
clay.
Table 2 summarizes the five zones o f differing rock qualities, o f which the RI and R I I
materials represent sound rock, and R I I I to RV the overlying weathered profile. The

TABLE2. Rock quality grades in Monasavu sandstone.

Rock quality grade
RV
RIV
RIII
RII
RI

Description
Soft, red-brown silty clay. (Residual clay: no trace of intact rock structure.)
Firm to stiff red-orange to brown silty clay with gravel-sized rock fragments in banded and
small block form. Completely weathered rock, with relic structures visible in excavations.
(Dam core material.)
Banded, weak clayey sandstone of variable hardness.
Discoloured massive sandstone, recovered as long sticks of core from boreholes.
Fresh grey-massive sandstone.

D. J. Knight

314
L.

d SL.pillwa~y

Embankment

-[ I-

Instrumented sections

Core borrow area

-800

800-

I
~"

!
i

~ ....

750-

_ _ ~

~ . ~

(-

99 700

Core
excavation
S

*.

LtJ 650LEFT
600

.
. . . . r'#ck" ~ ,
o
~
. . . . . .

.~
~

\ 7' ~ \

+300

~5
I

+200

'

. . . .

"

-700

* Monzonite

-650

~5

RIGHT
I

+100

-I00

Chainoge (m)
FIG.

750

~,. . . . . . . . . . . . . . . .
o~~ Conglomerate

'

--

.-'-""

.
:
- / ~ -

. . . . . . . . . . . . .

Diversion tunnel ~o
I

+L00

I I_ -ooGL-LL
,eve - " T r
I,

:
,

600
-200

eMain instrument
locations

3. Geological profile, longitudinal section along dam core trench and location of
instrumented sections.

interface between RIII and RII is sharp. The terms RI to RV, whilst specific to the site, are
based iargely on standard grades.
The bulk of the weathered profile comprised RIV and RV, with RIII constituting only a
small part of the total thickness above sound rock. Within the RIV zone shiny slickensided
surfaces of several square metres were found in excavations. Small conchoidal features as
well as larger joints exhibited the typical black staining of manganese oxide.
It is of interest to compare the RIV and RV materials with those described by Wesley
(1973) in Java, Indonesia. There the pedology from the parent rock of volcanic origin to
the fully residual soil shows less weathered yellow Andosols occurring above the 1000 m
elevation and more weathered red-brown Latosols occurring below that otherwise arbitrary
elevation. The Monasavu clays are the same as the latter in both colour and elevation.
INVESTIGATIONS
Investigations for a suitable core construction material concentrated on the RIV material
as it was soon realized that it represented the only potential source of abundantly available
material. The investigations comprised both field trials, involving core material selection
and proving, and laboratory testing.

Core material selection and proving

Initial investigations by large trial excavations in potential sources of core material
revealed the weathered succession (Table 2) to be of adequate thickness for practical
working, and the RIV material, drier than the RV, was selected for proving trials. On
excavation the RIV material appeared as lumps of clayey weathered rock; as expected,
drying altered its properties and working affected its behaviour, so that the common
classification tests were inadequate as a unique description of the material.
The results of early classification tests performed up to 1979, prior to the start of dam
construction, are given below. They represent samples of potential core material found

Geotechnicalproperties of Monasavu clay

315

within the RIV weathering grade. These results may be compared with the measured
properties of the clay in the constructed dam core (Table 5).
Natural moisture content: Air dried, average = 72%
Oven dried, average -- 83%
Natural dry density:

Range = 0-79-0.98 t/m 3

Average = 0.86 t/m 3

Specific gravity:

Range = 2.28-2.65
Average = 2.46

Liquid and plastic limits, and particle-size analyses, were also done in considerable
number but, as the testing methods were varied during this stage of the investigations, it is
inappropriate to quote values.
Uncertainty existed about the feasibility of constructing a clay core with this highly
plastic material having such a high natural moisture content. Trials involving various types
of plant and methods were therefore made to see whether it could be satisfactorily placed
and compacted. The wet climate precluded prior drying of the material to near its optimum
moisture content, and it soon became apparent that it would have to be used in its natural
condition. A trial embankment constructed in one of the large trial excavations in drizzly
weather and wet ground conditions demonstrated that a low-ground-pressure-tracked
D6-bulldozer could successfully borrow fresh material, spread and then compact it in thin
layers without sinking excessively below the surface. This technique broke down the lumps
and partially remoulded the RIV material into a homogeneous layer, so that the product
resembled a composite of small discrete chunks (up to 100 ram) within a soft matrix and it
proved physically possible to construct an RIV clay bank in these conditions. Undisturbed
block samples were taken to measure permeability and determine shear strength and
consolidation parameters, the values of which were subsequently verified from the main
core construction described later.

Laboratory testing
The laboratory testing programmes were designed to include the following tests on
disturbed and 'undisturbed' samples of the natural and recompacted soil and the clay from
the as-constructed core.
Classification tests:

Natural moisture content and dry density; liquid and plastic

limits; specific gravity; particle-size analysis; dispersivity.

Mineralogical analyses:

X-ray diffraction and scanning electron microscopy.

Compaction tests:

Maximum dry density; optimum moisture content.

Engineering design tests:

Consolidated-drained triaxial compression tests; consolidatedundrained triaxial compression tests with pore-pressure
measurement; oedometer consolidation tests; permeability
tests.

Construction control tests: Unconfined compressive strength; classification tests as above.

The majority of the testing was carried out in Fiji, with some of the early work being
done mostly in Australia. All construction control testing was done in accordance with

316

D. J. Knight

either Australian or British standards. Because of the suspected, and eventually confirmed,
presence of halloysite, investigations were carried out on the effect of different drying
methods on moisture content determination, involving (i) oven drying at 110~ (ii) oven
drying at 80~
(iii) flame drying and (iv) air drying. For consistency and comparison
purposes, all moisture contents during the construction control stage were measured by
oven drying at 110~ which resulted in the greatest apparent moisture content value.
Specific gravity was also found to increase with the efficiency of the drying method.
To examine the effect on the optimum moisture content (Frost, 1967; Wesley, 1973),
two different methods of sample preparation for the standard compaction tests were used
involving (i) natural drying to a moisture content below the expected optimum and the
subsequent addition of water to achieve the required moisture content range, and (ii)
natural drying ('drying back') through the testing range until the optimum moisture content
was reached. In controlled comparative tests it was found that method (ii) gave average
values of optimum moisture content 9% higher and maximum dry densities 2% lower than
those determined from method (i).
The early engineering design testing was done on potential core material recompacted to
maximum dry density at optimum moisture content, as was originally intended for the dam
core. Following the radical change in specification whereby the clay was only lightly
compacted at its high natural moisture content, a series of triaxial, permeability and
consolidation tests was carried out on 100 mm diameter undisturbed specimens procured
directly from the as-constructed core, from the bottom of 0.5 m deep pits at 16 different
locations when the dam was at a height of 23 m. The results of these were used to check the
parameters adopted for design. A further set of twenty 100 mm diameter samples was also
tested in unconfined compression at various times up to 28 days after sampling to
investigate the possibility of strength increase with time. A marginal increase was noted,
but it would be inappropriate to attribute much significance to this.
The routine construction control testing was performed on a series of six 100 mm
diameter core cutter samples for classification testing and an adjacent series of four 38 mm
diameter unconfined compressive strength test samples.
MINERALOGY
Mineralogical analyses of the Monasavu clay were performed at different stages of the
project by a variety of organizations, all principally concerned to confirm the suspected
presence of halloysite, in either its hydrated or 'meta' form. The well known effects (Frost,
1967; Wesley, 1973) of this mineral on testing procedures for the routine laboratory tests,
on the practicalities of construction and on the engineering behaviour of soils containing it
made it highly desirable to determine its presence.
Techniques used were X-ray diffraction (XRD) and scanning electron microscopy
(SEM), mostly on samples from the dam core borrow area or the dam itself, together with
some others on less highly weathered rock material. Table 3 summarizes details of the six
series of mineralogical analyses performed and Table 4 summarizes results of the analyses
in terms of the minerals identified.
Series 1. The presence of hydrous aluminium oxides and amorphous materials was
noted. It seems possible that these could be gibbsite and allophane respectively.
Series 2. Despite repeated efforts, Slansky (1978) could only obtain relatively poor
XRD patterns and this he ascribed to the nature of the material rather than to experimental

Geotechnieal properties of Monasavu clay

317

factors. This would appear to indicate its amorphous nature. The measured basal spacings
for the soil were 9.0 A (very broad) for air-dried and 11.5 A for glycolated specimens.
After heating to 400~ two spacings were shown, a stronger one of contracted dehydrated
halloysite at 7.4 A and a weaker one of halloysite still hydrated at I0.2 A.
Series 3. The report (AMDEL, 1979) noted a low abundance of crystalline minerals,
suggesting the presence of allophane or another amorphous component.
Series 4. XRD indicated that the principal minerals present in the whole sample were a
7 A kaolinite or metahalloysite phase, maghemite and hematite. The clay fraction was
examined to characterize the 7 A phase. Behaviour during various pre-treatments
suggested that this phase was a poorly-ordered kaolinite rather than a normal
metahalloysite, although the wet-state basal spacing of 10.5 A was very similar to that of a
halloysite. Subsequent SEM showed no trace of the tubular morphology typical of
halloysite, but rather a honeycomb texture (Fig. 4) often developed by smectitic clay
minerals. However, doubts existed about the precise location of the sample, and thus
whether it was the clayey sandstone parent material of the RIV clay, or rather decomposed
monzonite. It was therefore decided to resolve the question of halloysite presence by further
SEM work in Australia on fresh material taken from the dam clay core borrow area.
Series 5. Both XRD and SEM were carried out, the former indicating halloysite but the
latter giving inconclusive data (Figs 5a~:l) with respect to 'tubular' halloysite.

FIG. 4. Scanningelectronmicrograph: series4 ( 11 000).

318

D. J. Knight

I
<.

tl-

II

II

II

.1~

.a
g

,..J

~g
=~Zr, J

e~

~ ~
~.=

>

zg~

.a
0

9o . >

a. N ~

.a

_
N r

=:~

"d

Nr
0

.N
"0

r~

g,

g.

g
>-

>,

"-d

a:.o

,,~

d
0

N
9" o

""

me~

o=g,~
r

"d

Geotechnical properties of Monasavu clay

319

Jill.

~~~
~ .~ . ,~ .~

~'~

~ N "N

p~
o

C~

~.~ 0

P~
r

C;

C)

-~ .~

~ 2 "d
v.~

. ~
c~
P~

C~

_o

c~

"d

8
~Q

~ "~ .~
.g

N~
Z

D . J . Knight

320

TABLE 4. Summary of minerals

Minerals identified
Halloysite
Series

Sample No.

Date of
Report

30/5/77

2/6/78

WIII(A)I
WIII(A)2

Am

22/6/79
--

WIV 1
WIV 2

Gi

Ep

CD - CD - -

15/7/81

D D
D D
*?*?
1

C
D
E

Not stated

D D
D D
Tr Tr
A Tr

Tr/A
A

Tr/A Tr
Tr/A Tr

WV 1
WV 2
7/3/80

Hm

Tr

H~

T r

Not stated

Between
Nov. 1981
and April 1982

Details not available

Notes:
2. Key to other symbols used

1. Mineral key
Am =
Ep =
F
=
Gi =
H
=
Hh =
Hm =
Hem =
K
~

Monoclinic amphibole,
Epidote.
Feldspar.
Gibbsite.
Halloysite.
Hydrated halloysite.
MetahaUoysite
Hematite.
Kaolin.

MH - Maghemite.
P
= Palagonite, 'used to
describe a poorly
crystalline clay, probably
a smectite precursor'.
Sm = Smectite (montmorillonite).
Sp = Spine[, 'probably chromite
or magnesio-chromite'.
V
= Vermiculite.

9 Unnumbered sample.
* Identification of mineral
in report or reference.
- - Absence of mention in
report or reference.

Geotechnical properties of Monasavu clay

321

identified in Monasavu clay samples, Fiji.

(See note 1)

Hem

MH

Sm

Sp

Remarks

Meta-halloysite 60-70% by weight of fine fraction.

Hydrous aluminium oxides and amorphous materials
also present, pH 6.45 of bagged bulk sample as
received.
m

See Slansky (1978). Report refers to hydrated

halloysite as being the only crystalline clay mineral
present in the soil.

Tr?
Tr?
Tr
Tr

CD D See note 3 for key to semi-quantitative abbreviations

CD D used. Data summarized from AMDEL (1979). WlV
(later changed to RIV) represents the dam core
material. In each column, the left-hand symbol
relates to the bulk sample and the right-hand
symbol to the fine fraction.

Tr
Tr
T

Tr
*

9.

4
3

m
m

Last remark for series 3 applied also to series 4.

--

--

2
1

---

4
3

Details not available

3. Key to semi-quantitative abbreviations

D = Dominant. Used for the
component apparently most
abundant, regardless of its
probable percentage level.
CD = Co-dominant. Used for two
(or more) predominating
components, both or all of
which are judged to be
present in roughly
equal amounts.

--

--

Infrared spectrometry also performed on bulk

sample. Numbers indicate probable order of
occurrence of mineral identified. Data summarized
from Eggleton (1981).

Halloysite present in spherical blocky 'cabbage'

or cluster form rather than in tubular stick form.

~ Accessory. Components
judged to be present
between the levels of
roughly 5-20%.
Tr = Trace. Components judged
to be <5%.

D.J. Knight

322

(a)

(b)

(c)

(d)

FIG. 5. Scanning electron micrographs: series 5, sample E. (a) x2000; (b) x3500; (c) x250;
(d) x 350.

Geotechnieal properties of Monasavu clay

323

Series 6. This continued inconclusiveness regarding the presence of halloysite led to the
sixth and final series of tests on a sample procured by the author from the dam core
during construction. This confirmed that halloysite was present, but in spherical blocky
'cabbage' or cluster form rather than in the more common tubular form.

Conclusion on halloysite presence

The lengthy testing series established that the Monasavu clay, as was always suspected
by its laboratory moisture content testing behaviour, is halloysitic but of a different form to
that usually found, tending to be more amorphous than crystalline.
It is of interest to note the weathering stages by which clays containing the amorphous
mineral allophane change to kaolin. The first stage shows a mostly allophane presence, with
some halloysite, whilst in the next stage the allophane more or less disappears and
halloysite dominates. The hydrated halloysite then becomes dehydrated to metahalloysite
before the final kaolin stage is reached. There are hints of some of these stages from the
above-described mineralogical analyses carried out on the Monasavu clay. Frost (1967)
alluded to a number of these factors in relation to the irreversible property changes
involved in the pre-testing preparation of some tropical soils.

CONSTRUCTION

OF DAM CORE

The design of the dam and its core specification have been described elsewhere (Knight
et al., 1982; Knight et al., 1985). A radical change was made at the beginning of construction from working to a conventional end-product specification to a wholly methodrelated specification based on what was practically achievable, with a record kept of the
achieved product in terms of classification data and unconfined compressive strength.
The essential points relating to the developed construction procedures for this clay core
are summarized below. It is stressed, however, that the successful execution of such
procedures depends much on the practical experience of the construction supervisory staff
with wet core techniques.

Borrow area
Core material was taken from a large area on the right abutment constituting required
excavations for the right flank of the embankment and for the spillway (Figs 2 and 3). The
area drained well after rain and plant access was afforded by the sound rock benches
underlying the weathered material.
Exploitation
Bulldozers worked on fairly flat slopes across the full depth of the RIV material
(Fig. 6a), towards loaders and trucks waiting at the base of the slope. This ensured, as well
as drainage after rain, good mixing of the material in terms both of its inherent natural
variation through the depth range, and of the rock nodule/plastic matrix proportions. The
material was taken directly to the dam for placement.

324

D.J. Knight
S :

(a)

(b)

FIG. 6. Clay core construction. (a) Borrow area exploitation. (b) Compaction of soft clay core
with tow-ground-pressure-tracked D6 bulldozer.

84. . . . .

Geotechnical properties of Monasavu clay

325

Placing, spreading and compaction

The fill was brought to the edge of the core by 20 tonne rear dump trucks, with
spreading and compaction carried out by low-ground-pressure ('swamp') tracked
D6-bulldozers (Fig. 6b). These placed the clay in 100 mm thick layers and naturally
carried out much of the compaction. The final texture at depth below the grooved surface
comprised discrete uncrushed RIV nodules up to 100 mm size surrounded by a remoulded
clay matrix.

Surface sealing and drainage

The frequency and suddenness of rain necessitated pre-sealing of the constructed clay
surface to allow early resumption of filling. Early methods involved a 10% crossfall to the
watertight grooved surface, with some hand drainage; this gave way later to the use of a
small loader with wide street tracks, combined with a smaller crossfall.

Limitations
As experience was gained it was found that core placement could take place under an
increasing range of conditions, including mist or low rainfall. In one month, with only two
rainless days and over one metre of rainfall, about 15 000 m 3 of clay were placed.
Nevertheless, some limits were set. Accordingly, no filling was allowed after heavy rain
until the surface had been adequately drained. Filling was stopped when the bulldozer
tracks penetrated more than about 150 mm into the fill or 'floated' on it with excessive
rebound.

GEOTECHNICAL

PROPERTIES

OF CLAY IN DAM

CORE

The routine construction procedures resulted in the measurement of the classification and
undrained strength properties of the as-constructed dam core (Table 5).
The mean wet density of 1.52 t/m 3 is only about 75% of the value normally attained for
a clay core. The mean unconfined compressive strength of 34 kN/m 2 implies a mean
undrained shear strength of only 17 kN/m 2, assuming virtually saturated conditions. This
demonstrates vividly the very low strength of the core at its initially constructed stage.
In addition to the routine classification testing, sixteen sets of undisturbed samples from
100 mm diameter core cutters were taken from the 23 m height of the as-built core
between October 1980 and February 1981 for engineering property measurement,
correlation of results with the routine testing, and checking of design assumptions for
parameters and behaviour. In consolidated-drained triaxial compression a slightly curved
Mohr-Coulomb envelope develops, with 4' = 30~ and c' = 15 kN/m 2, and a yield strain
above 10% and sometimes 20%. In the quick undrained conditions such large strains were
usual. The coefficient of compressibility m v from oedometer tests was generally within the
range 0.2-0.1 m2/MN for pressures between 100 and 1600 kN/m 2. The indicated
'preconsolidation' left by the compaction process has been between 150 and 200 kN/m 2.
Permeability values of 1 x 10-8 cm/s were obtained.

D. J.

326

Knight

TABLE 5. S u m m a r y of core properties as constructed.

Property

Unit

No. of results

Mean

Standard deviation

%
t/m 3
t/m 3
%
%
%
%
%

1750
1750
1750
1642
340
332
332
332
-238

76
0.86
1.52
2-4
96
59
107
48
0- 35
2-66

7
0.05
-1.6"
4
-9
9
-0.11

kN/m 2

1008

Classification
Moisture content
Dry density
Wet density
Air voids
Passing 0.075 m m
Plastic limit
Liquid limit
Plasticity index
Liquidity index
Specific gravity

Strength
Unconfined compression

34

0.99
58

0.05
6

Standard compactiont (drying down)

Maximum dry density (MDD)
Optimum moisture content (OMC)

t/m 3
%

5
5

Relative compactiont
Mean dry density/mean M D D
Mean moisture c o n t e n t - - m e a n OMC

%
%

---

87
18

---

* Abnormal distribution.
~ For comparison only.

MEASURED

BEHAVIOUR

OF CLAY

CORE

Monitoring system
The dam core was instrumented to measure deformations, pore pressures and
three-dimensional stresses during construction, impounding and operation, by means of
hydraulic settlement cells, hydraulic piezometers and total earth pressure cells concentrated
in three cross-sections and up to four horizons (Figs 3, 7).
Internal vertical movement of the core was monitored by a settlement cell system able to
accommodate 2.3 m of differential settlement, and two deformation tubes were installed
either side of the core. Electrical extensometers were installed longitudinally along the top
of the core but, despite one metre of cover, the core placement process overstrained them.
Surface movement was measured by means of survey and level stations on the crest and
downstream slope.
Pore pressures were monitored by high-air-entry hydraulic peizometers installed across
the core width, and stresses measured by sets of three total earth-pressure cells. One cell
measured vertical stress, a second horizontal stress in an upstream-downstream direction,
and a third horizontal stress in a cross-vaUey direction. Adjacent hydraulic piezometers
permitted effective stresses to be deduced. All instrumentation was read regularly.
Reservoir impounding began in April 1982, after which the reservoir rose steadily
through 70 m height before the first spilling in March 1983 (Fig. 9). The final 8 m was
considerably hastened by cyclone Oscar (Table 1). The following sections summarize and
review the measured behaviour throughout construction, impounding and the first year of

Geotechnical properties of Monasavu clay

~-~

750

g
[
~
2
"-- 700 Cloy blonket 1 ~

84

Vories

Rockfill

.700

/f

am

UJ
/

75 2 (includes settlement olEowance]

1-3

7~s Fu,t sopp,y Lo~,

680

Monzonite

Monzonite

650 9

Key to instruments
(exaggerated for clority)
9 HydrQulic piezometer
I Hydrou~ic settlement ce~
9 Group of eorth pressure cells
(meesuring in 3 directions)
- - - - - D e f o r m c t i o n tube

327

650

FIG. 7. Dam cross-section and instrumentation at Ch. 175.

reservoir operation. They thus amplify and extend the information up to initial impounding
given by Knight (1982), and summarize the data from Knight et al. (1985).

Deformations (internal and external)

Most of the settlement was accomplished by the end of impounding, and comparative
settlements for the three cross-sections showed good consistency. The largest internal
settlement occurred as expected at the centre of El. 705 at Ch. 175, where the measured
settlement exceeded 1.8 m and continued slowly thereafter. This cross-section's composite
behaviour at the centre of the core is shown on Fig. 8a, and indicates a consistent pattern
of settlement with elevation. Behaviour of its mid-height zone is shown in more detail on
Fig. 8b to both natural and exaggerated scales, illustrating the proportionate vertical
movement of the core in relation to its width as well as the differential vertical
displacements across it. Differential settlement is indicated at the downstream edge
reflecting the differing depths of clay to the interface, and possible hang-up on the much
stiffer filter. Overall the soft clay core appears to have squeezed down satisfactorily over
about at least a 12 m horizontal width.
The maximum crest settlement was 0.33 m and, in the upstream-downstream direction,
the maximum external horizontal displacement 0.1 m, occurring half way down the
downstream slope, both at Ch. 175.

Construction pore pressures, dissipation and impounding

The general pore-pressure pattern for all core piezometers is typified on Fig. 9, showing
an initial increase of piezometric level with dam fill, followed by slow dissipation of
maximum construction values. Pore pressures increased again with impounding, rising
suddenly to reflect the remaining 8 m of rapid filling, since when they either slowly dropped
or remained steady. Piezometric level contour patterns within the core cross-sections were
satisfactory.
The maximum construction stage pore pressure ratio (ru) was 1.14, where ru is the ratio

D. J. Knight

328
750-

P~QI

1983
Mar1984

730

(a)

680 y

_/_

~ i n

1
Settlement (m)

Hydraulic settlement ceils

cross section (see F i g . 7 } ~

Core

"Filters and - - Downstream rockfi II

drain

0~ May1981
Mar 1984

Natural scale
I

v 0"

May
1981

Sept

1981

U3

Exaggerated

Mar 1982
Mar 1983
Mar 1984

20

vertical scale
upstream

~.

downstream 20

Distance from Dam CE (m)

FIG. 8. Settlement behaviour at Ch. 175. (a) Settlement vs. elevation. (b) Settlement profile at
El. 705.
of porewater pressure at a particular point to its total overburden pressure. With
dissipation r u fell, to average values just before impounding within the range 0.43-0-55.
Just before being affected by the rising reservoir these values had dropped to between
0.08 and O. 30, suggesting a relatively rapid consolidation of the clay core.

Total and effective stresses

The unusually soft core was always regarded as potentially advantageous, in being
unable to resist applied shear stresses sufficiently to permit arching and hang-up between
the rockfill shells on each side, and unable to sustain cracks during deformation which

Geotechnical properties of Monasavu clay

329

,,~ 750i

Fill l e v e ~ f ~ ~

u 715-

680 .1980//

198

"~

19%2

~9%3

19~54

FIG. 9. Piezometriclevelvs. time at Ch. 175.

would induce hydraulic fracture. Sherard (1981) has for long advocated this philosophy
and has given a summary of its use (mainly in Europe).
Hydraulic fracture is avoided when seepage pressures are everywhere exceeded by the
minimum total stress (approximately the lower of the two horizontal total stresses). This
requires positive effective stress throughout, preferably increasing with time. The safest
condition is that the minimum horizontal total stress should exceed the full reservoir head
at the particular elevation. The purpose of the earth pressure cells was to verify these
predictions and to quantify the behaviour.
Total vertical stresses (Fig. 10) rose steadily throughout construction, and continued to
rise through impounding for El. 680 and El. 705. The corresponding total horizontal
stresses showed a similar pattern but at lower absolute values, and exceeded the reservoir
head except at El. 680 where, however, the local steady seepage pore pressure was
exceeded. This was particularly encouraging considering the unusually low bulk density of
the core material. A high and reasonably constant ratio, 0.8, of total horizontal to vertical
stress was obtained, denoting proximity to uniform all-round stress conditions.

1000
Ch. 125
Ch.175~800 Ch.210

Impoundir&
Et.680
End of
I'-"
~'-"
construction

600
8

/
I~0

EI.735

1981

1983

FIG. 10. Totalvertical stress vs. time.

1984

330

D. J. Knight

100-

(Mobilised angle of shearing

~ /
~ , resistance)

Effective I
,~/~,~ J
st resses / . Z',7
.fl.).

I Toter /
I st resses /

Z
J

50-

~
/

"

End of construction
Pore pressure

I
v

(O-v

200

400
and (0-v h )/2 (kN/m 2)

600
CIv=Verticalstress
Oh= Horizontal stress

NOTE.Pressure readings below

lOOkN/m 2 considered to be of
low accuracy

FIG. 11. Stress relationships for Ch. 175 at El. 705.

The measured effective vertical and horizontal stresses throughout showed a generally
steady increase, whilst the ratio of effective horizontal to vertical stress with time generally
increased consistently to between 0.6 and 0-7.
Stress p a t h s

Total and effective stress paths were monitored throughout and after the entire impounding period (Fig. 11, which in effect treats the entire core as an 80 m high triaxial test
specimen). These show stress paths in a 89 v - ah) against 89 v + ah) space. A reasonable
approximation is that o v represents the major principal stress and a h the minor principal
stress. Steadily increasing values of equivalent deviator stress are shown for each case, as
well as the variation of the mobilized angle of shearing resistance r The latter is initially
just below the available value of 30 ~ (or more) indicated by triaxial tests, but subsequently
drops below 20 ~ The consistency of the measured stress behaviour is again demonstrated,
as well as the occurrence of expected stress redistribution. The ability of the low-strength
clay core to yield appears to have reduced hang-up in the upper regions, allowing the
self-weight of the core to be supported within itself rather than by shear and arching to the
shells.

DISCUSSION

AND

CONCLUSIONS

Properties

High values of liquid limit, plastic limit and natural moisture content imply in many soils
a low drained strength and difficult working (Terzaghi, 1958). For the Monasavu clay, and
other similar tropically weathered soils, these characteristics have no such dire portents,
and to that extent conventional classification systems can be misleading.
The clay's natural moisture content was unquestionably high in relation to the material's
optimum moisture content, and thus was early recognized as being difficult or impossible to

Geotechnieal properties of Monasavu clay

331

compact by conventional means. The impracticability of drying led to its use at its high
natural moisture content, necessitating placement and compaction with very light plant.
Even so, excessive trafficking could overwork the material by releasing water from its
micro-pores into the soft matrix surrounding the relic nodules (cf. Vaughan, 1982).
The very low dry density of 0.86 t/m 3 led to the low bulk density of 1.52 t/m 3, and to
the possibility that it would produce unacceptably low total horizontal stresses. However,
the achieved stresses were adequate, and thus the low density did not in the end become a
problem.

Behaviour
The essential behaviour of the core during construction, impounding and early operation
is summarized as follows.
A maximum vertical deformation exceeding 1.8 m occurred in the middle of the highest
section, as anticipated. Other settlements were self-consistent, showing that the main
central zone of the soft core continued to settle well in relation to the stiff rockfill shells.
Crest settlement was modest, being only about 0.4% of the dam height, indicating fairly
rapid consolidation.
Despite the fairly rapid consolidation, common with such residual soils and causing
some pore pressures at the start of impounding to be less than the steady seepage values,
the pore pressure values and patterns were satisfactory.
Measured stresses were satisfactory, showing a steady rise of effective stresses
throughout construction and impounding. Total stresses, accompanied by settlement,
continued to rise after the end of construction. At all except one location the lowest total
horizontal stress exceeded the full reservoir pressure at the particular elevation and in all
cases exceeded the seepage pore pressure. There is thus good evidence that arching is small
and thus of the absence of conditions conducive to cracking and hydraulic fracturing.
The overall stress behaviour is consistent not only within itself but also with
measurements from other instruments.
In order that the core's long-term behaviour can be measured, a separate trial bank of
similar material to the core was constructed in the reservoir. It is hoped that this will, when
reservoir conditions permit, indicate whether the clay is subject to any time-hardening
process not attributable solely to consolidation, such as cementation.

Conclusions
The following conclusions are drawn:
1. The tropically weathered clayey sandstone used as the core for Monasavu Dam is a
good engineering material, despite its very high natural moisture content, high plasticity
and low density. This conclusion is very similar to the one drawn by Wesley (1973) for
some halloysitic and allophanic clays in Java, Indonesia.
2. The clay is halloysitic but apparently tending to be more amorphous than crystalline,
with the halloysite present in a spherical blocky 'cabbage' or cluster form rather than as the
more usual tubular sticks. This difference does not, however, appear to affect the
geotechnical engineering behaviour of the clay in comparison with that experienced with
the tubular morphology.

332

D. J. K n i g h t

3. The clay possesses a high intrinsic strength and displays v e r y satisfactory b e h a v i o u r

b o t h during and after c o n s t r u c t i o n .
4. T h e material is c a p a b l e o f successful use within an e m b a n k m e n t d a m c o r e in its
natural wet state. Such a state is, m o r e o v e r , not o n l y c a p a b l e o f practical use, but also has
considerable benefits in c o m p a r i s o n with a lower m o i s t u r e c o n t e n t state, in terms o f its
s t r e s s - s t r a i n behaviour.

ACKNOWLEDGEMENTS
The author thanks the Fiji Hectricity Authority, its General Manager, Mr D. S. Pickering, CBE, and its
Consulting Engineers Gibb Australia for their kind permission to publish this paper. He is also most grateful to
his colleagues Mr N. M. Worner of Gibb Australia, for access to detailed construction notes, and to Mr G. L.
Smith, FSCET, of Sir Alexander Gibb & Partners for assistance with some of the data research involved. The
dam was constructed by the Special Projects Division of the Fiji Hectricity Authority. The first three power
projects for which it was constructed were funded by, inter alia, IBRD, EIB, ADB, Australian Development
Assistance Bureau as well as from Government grants, internally generated funds and local and overseas
commercial loans.
Soils testing during the investigation stage was carried out by Longworth and McKenzie Pty Ltd of Sydney,
Australia; a limited amount of construction stage soils testing was carried out by Golder Associates of Suva,
Fiji. Mineralogical testing was shared between: Department of Mines, Geological Survey of New South Wales;
The Australian Mineral Development Laboratories, Frewville, Adelaide, South Australia; The Australian
National University, Canberra, ACT. Some mineralogical testing was also arranged through Dr P. R. Vaughan
and Dr H. Shaw of Imperial College of Science and Technology, London.

REFERENCES
THE AUSTRALIANMINERAL DEVELOPMENT LABORATORIES(1979) Clay mineralogy of six soil samples.
Unpublished Report GS Q4797/79 for Sir Alexander Gibb & Partners Australia, 22nd June 1979.
EGGLETONA. (1981) Fiji clay investigation. Unpublished Report for Sir Alexander Gibb & Partners Australia,
15th July 1981, The Australian National University, Canberra, ACT.
FROST R.J. (1967) Importance of correct pretesting preparation of some tropical soils. Proc. 1st Southeast
Asian Regional Conference on Soil Engineering, Bangkok, 43-53.
KNIGHT D.J., WORNERN.M. & McCLUNG J.E. (1982) Materials and construction methods for a very wet clay
core rockfill dam at Monasavu Falls, Fiji. Proc. 14th Int. Cong. on Large Dams, Rio de Janeiro, 4,
293-303.
KNIGHT D.J. (1982) Discussion, question 55: Materials and construction methods for embankment dams and
cofferdams--testing methods and quality control. Proc. 14th Int. Congr. on Large Dams, Rio de Janeiro,
5,598-604.
KNIGHT n.J. (1984) Monasavu Dam in Fiji. Introductory note, ICE/BNCOLD, London, March 1984, 1-6.
Also discussion paper, 'Monasavu Dam, Fiji', BNCOLD News and Views, 27, 12-13.
KNIGHT D.J., NAYLORD.J. & DAVIS P.D. (1985) Stress-strain behaviour of the Monasavu soft core rockfill
dam: prediction, performance and analysis. Proc. 15th Int. Congr. on Large Dams, Lausanne, 1,
1299-1326.
SHERARDJ.L. (1981) Building embankment dams in areas of high rainfall. Syrup. on hydro-electric development
in the Amazon Region, Sdo Paulo, Brazil.
SLANSKYE. (1978) Clay mineral analysis of two specimens from Fiji (for Longworth and McKenzie Pty Ltd).
Geological Survey Report No. GS 1978/153 (unpublished). Geological Survey of New South Wales,
Department of Mines.
TEaZAGm K. (1958) Design and performance of the Sasumua Dam. Proc. Inst. Civil Engineers 9, 369-394.
VAUOHAN P.R. (1982) Design and construction with wet fills. Special Lecture to Associacdo Brasileira de
Mecdnica dos Solos, Sdo Paulo, April 1982.
WESLEY L.D. (1973) Some basic engineering properties of halloysite and allophane clays in Java, Indonesia.
Geotechnique 23, 471-494.