0

0 British Geological Survey

0
0 TECHNICAL REPORT
0 Engineering Geology Series

0
0
0 ENGINEERING GEOLOGY
0
OF TROPICAL RED CLAY SOILS
0 Cover illustration
0 (upper left)
Cracks in red clay Report WN/93/15
0 road embanlonent,
western Kenya.

0 (upper right) SUMMARY FINDINGS AND

Landslide at Puncek,
0 West Java,Indonesia. THEIR APPLICATION FOR
(Lower left) ENGINEERCNG PURPOSES
a Pit sampling, West
Java, Indonesia.
0 (Lower Right)
(ODA/BGS R&D Project 91 /18)
Red clay soil viewed
0 in thin section, K J Northmore, M G Culshaw, P R N Hobbs,
magnification x200. J R Hallam and D C Entwisle
0
0 Authors
Northmore, Culshaw,
0 Hobbs, Hallam,
Entwisle
0 BGS Keyworth A Report prepared for the Overseas Development Administration
under the ODA/BGS Programme o f Research and Development
0 in Developing Countries
Bibliographic
0 reference
NottbmDre, KJ,
e 3 - MG,
EDbbqPRN,
0 -JR&
-, D C. 1992.
a Engineering geology
of tropical red clay
soils: s u m m a r y
0 findings and their
application for
e engineering
purposes. British
0 Geological Survey
Technical Report
WN/93/15.
0
0
@ NERC copyrighf
0 1992 Keyworth,Nottingham British Geological Survey 1992
 Preface
This report is the fifth of a series of five main project reports describing the results of a
study into the engineering geological characteristics of tropical ‘red’ clay soils. T h e study
was funded by the Overseas Development Administration (ODA) under its programme
of ‘Research and Development’) part of the British Government’s provision of aid and
technical assistance to developing countries (Project 9 1/18) Engineering Geology-of
Tropical Red Clay Soils). T h e main objectives of the study were to:

i) investigate the geotechnical properties of a variety of red clay soils by means of

standard, specialised andlor new test procedures on both disturbed and
undisturbed samples.

ii) relate the geotechnical properties of the soils to their geological provenance)
position in the weathering profile and climatic and topographic conditions so
that a practical classification scheme could be developed for engineering use.

T h e project work was conducted) primarily, on samples obtained from selected study
sites in Kenya and Indonesia. In both countries) an extensive programme of field
sampling was carried out to obtain a wide variety of red clay soils from contrasting
climatic topographic and geological terrains) a t various depths within the weathering
)

profiles. Fundamental material and geotechnical properties of the collected soil samples
were determined from comprehensive laboratory testing programmes conducted a t the
British Geological Survey (BGS) laboratories (Keyworth, Nottingham) UK) and by
project counterparts in Kenya (the Ministry of Public Works, Materials Testing and
Research Department) Nairobi) and Indonesia (the Institute of Road Engineering)
Bandung).

T o complement the investigations undertaken on the Kenya and Indonesian soils,

additional samples of tropical red clays were also collected (through BGS geologists and
other contacts) from Fiji and Dominica) and returned to the UK for analysis and testing
. in the BGS laboratories. T h e field and laboratory test data determined for all the soil
samples is stored in a PC compatible database at BGS, Keyworth.

Overseas field and laboratory investigations could not have been effectively carried out
without the collaboration of project counterparts in Kenya and Indonesia.

In Kenya, the main project counterparts were the Ministry of Public Works, Materials
Testing and Research Department (MT&RD)j Foundations and Drilling Section,
Nairobi. The Department has a keen interest in red soil research and their collaboration,
support and assistance in all aspects of work undertaken in Kenya is gratefully
acknowledged.
? +

In Indonesia) the project counterparts were the Institute of Road Engineering (IRE),
Bandung, who’s valuable collaboration in site selection) sampling, and laboratory testing
is also gratefully acknowledged.

Acknowledgment is also made to the valuable support and assistance given by many
other organisations and individuals during the project work undertaken both overseas and
in the UK. T h e following deserve special mention:
Kenya
Transport Research Laboratory (formerly Transport and Road Research Laboratory),
Nairobi Office (Mr M J O’Connell) for invaluable advice, technical discussions, field
visits and generous assistance during the Kenyan fieldwork.

Coffee Research Foundation, Jacaranda Coffee Research Station, Ruiru (Mr P K

Michori) for granting access and permission to site boreholes within the Research Station _.
grounds and for every assistance during the drilling and sampling operations.

Sir Alexander Gibb & Partners (Africa), Nairobi (Mr R W M Carrington and Mr P
Karekezi), for generously supplying records of geotechnical test data and for extremely
useful on-site discussions and assistance during field investigations along the Eldama
Ravine-Nyaru road in west-central Kenya.

Mowlem Construction (East Africa) Ltd, drilling contractors, Dandora, Nairobi (Mr M
N Dobson) who’s experience, equipment, drill crew, back-up facilities and assistance
ensured that the drillingkampling operations in Kenya were a t least partially successful.

British High Commission, Aid Department, Nairobi whose assistance, advice, and
logistical support throughout the duration of the project work in Kenya is much
appreciated.

Indonesia
British Embassy, Jakarta for advice and logistical support during the fieldwork visits to
West Java.

Fiji
Department of Mineral Resources, Suva (Mr T Lawson/Mr J Davies) for supplying
samples of tropical clay soils from selected landslide areas in Fiji.

Dominica
Ministry of Communications and Works, Roseau (Chief Technical Officer), and J R
Hallam (BGS) for supplying samples of tropical clay soils from selected sites in
Dominica.

UK
Transport Research Laboratory (formerly Transport and Road Research Laboratory),
Overseas Division, Crowthorne (Mr D Newill) who’s keen interest in tropical soil
research, valuable advice, technical discussions and support throughout the study is
gratefully acknowledged.

British Geological Survey

Keyworth
Nottingham
NG12 5GG
UK
 CONTENTS

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Nature of tropical soils . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Project aims and objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Work carried out and reporting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2 4
Project study areas and sampling sit s . . . . . . . . . . . . . . . . . . . . . . . . .
Selection of study areas . . . . . . . . . . . .i . . . . . . . . . . . . . . . . . . . . . . . . . .
6
6
_.

Selection of sampling sites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

Types of soils investigated . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . .
' 9

3 Sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Investigation and development of sampling techniques . . . . . . . . . . . . . 10
Summary of test results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Recommendations for good sampling practice . . . . . . . . . . . . . . . . . . . . 15

4 Index properties and testing procedures . . . . . . . . . . . . . . . . . . . . . . . 16

Plasticity tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Moisture content . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
Shrinkage tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
Particle size determinations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
Particle density (specific gravity) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Recommendations for good testing practice . . . . . . . . . . . . . . . . . . . . . . 24

5 Mechanical properties and testing procedures . . . . . . . . . . . . . . . . . . 27

Specimen preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
Triaxial tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Oedometer tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
. Compaction tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
Ring shear tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
Summary of results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
Recommendations for good testing practice . . . . . . . . . . . . . . . . . . . . . . 41

6 Identification and classification of tropical red clay soils . . . . . . . . . . 42

Geological maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
Pedologicalmaps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
Engineering index properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
Indexing the mechanical behaviour of tropical red clay soils . . . . . . . . . . 48

References

Figures

Tables
FIGURES

1 Map showing simplified world distribution of tropical ‘red’ clay soils

2 Schematic representation of the tripod sampler
3 Sampler for tropical red clay soil - 0.5 m capacity
4 Schematic drawing of the sampler test instrumentation
5 Compression profiles for sampler tests at Mchana, Kenya
6 Composite graph of processed parameters for sampler test MCH 14
7 A-line plots for soils remoulded by hand-mixing and in the ‘greaseworker’
8 Triangular particle size classification charts for tropical red clays developed
over various types of parental bedrock
9 Progress of saturation of triaxial specimens
10 Triaxial compression stress-path plots
11 Triaxial consolidation ‘swell’ effect
12 Histograms of effective cohesion and angle of friction
13 Shear stresses showing a ‘bonding’ effect
14 Stress dependance of oedometer consolidation settlement
15 Voids ratio vs. (log) stress for oedometer settlement tests
16 Relationship of Compression Index, Cc, to relative voids ratio, eK
17 Application of Vaughan’s ‘bonded’ soil model
18 Compaction test plots
19 Results of ring-shear tests on remoulded Kenyan samples
20 Examples of tropical weathering profiles in igneous rocks
21 Data plotted on a combined ‘Plasticity/Activity’ chart for classifying tropical
clay soils

TABLES

1 Soil map units at project sample sites in relation to terrain and climate
2 Summary of drilling and sampling investigations at Ruiru, 1989
3 Summary of rpsults from mechanical properties test programme
4 Approximate correlation between the major classes of tropical residual soils
in the principal soil classification systems
1 Introduction

NATURE! OF TROPICAL SOILS

In the engineering literature, tropical soils (as a group of soils distinct from those formed
in areas subject to temperate climes) are sometimes described as falling into two broad
categories, tropical black soils and tropical red soils. T h e former category, which
includes those soils termed ‘black cotton soils’ or ‘black swelling clays’, comprises a
relatively distinct group of soils, rich in smectite clay minerals, whose engineering
behaviour is dominated by volume changes (that is, shrinking and swelling) when they
are subjected to changes in natural moisture content. However, there is some confusion
as to what materials comprise the so-called tropical ‘red’ soils and the terms used to
describe them.

Red clay soils are very common throughout the tropical and sub-tropical regions of the
world (Figure 1). However, it is a misconception to consider them as forming a distinct,
clearly defined soil type as they encompass a wide variety of soils whose material and
engineering properties vary considerably. This variation has been reflected in the ever
increasing literature on tropical soils since the 196O’s, which has seen red and reddish
brown soils described as ‘laterites’, ‘laterite soils’, ‘lateritic soils’, ‘non-lateritic tropically
weathered soils’, ‘latosols’ and ‘tropical red clays’.

Under certain conditions, hardened horizons, often referred to as ‘laterite’, may be

associated with the soil profiles. These form as a result of the accumulation of iron
which, in some areas, may develop into a continuous sheet of indurated ferricrete
(‘laterite’) forming a surface or near-surface duricrust. However, not all red clay soils
harden on exposure to form laterite (as defined by the original description of Buchanan,
1807) and it is unfortunate that the terms ‘lateritic clays’ and even ‘laterite’ are still used
by some engineers to describe any reddish tropical soil. Due to this confusion in
terminology, the soils described in this study are broadly referred to as tropical ‘red’
clays, despite the fact that they are not all red.

T h e tropical ‘red’ clay soils considered in this study are residual soils, that is they are
formed in situ by the intense weathering of parent material, both primary and
sedimentary, in tropical and sub-tropical climatic environments. This weathering process
primarily involves the progressive chemical alteration of primary minerals, the release of
iron and aluminium sesquioxides, increasing loss of silica and the increasing dominance
of new clay minerals (such as smectites, allophane, halloysite and, as weathering
progresses, kaolinite) formed from dissolved materials. Continued depletion of silica
under prolonged weathering in hot humid climatic zones may eventually cause alteration
of kaolinite to the aluminium oxide, gibbsite, as free alumina is formed in the soil profile.
At any particular site, mineral composition and microstructure will depend on numerous
factors such as the nature of the parent material, the age of the land surface (time for soil
formation), climate, topography and drainage conditions (Duchaufour, 1982). Only a
small proportion of sesquioxides will impart a red colouration, and because of the
progressive nature of the tropical weathering process and the numerous inter-related
factors controlling soil development at any particular locality, the division between red
and ‘non-red’ soils are not clear cut. For example, iron oxide occurs in the form of
2 SECTION ONE Introduction

haematite when the soil is seasonally desiccated, giving the characteristic red colour, and
as goethite in a constantly humid environment, giving the soil a brown or ochreous
colour. Also, particularly in mountainous areas, soils developed over the same parent
rock may grade imperceptibly from red to brown or yellowish brown soils as altitude and
topographic relief vary. For example, andosols are typically brown or yellowish brown
soils. They develop on recent volcanic ashes and, characteristically, contain the hydrated
amorphous clay mineral, allophane. Detailed mineralogical investigations in this study
have also recognised other amorphous aluminium silicate ‘gels’ present in ferrisols and -’

some ferrallitic soils. Their occurrence is restricted primarily to constantly humid (but
well-drained) regions and in many areas they are found only at higher elevations, where
high rainfalls and lower temperatures are apparently required for these soils to develop
and retain their hydrated condition (Morin, 1982). Where drainage is impeded,
smectite-rich black swelling clays tend to be formed rather than andosols. With a change
in climate (particularly the appearance of a marked dry season), which can occur a t
decreasing elevations, andosols sensu stn’cro give way to transitional (‘andic’) soils which ,
with increasing kaolinisation and crystallisation of hydroxides (gibbsite, goethite and
haematite), can eventually grade into truly ‘red’ ferrallitic soils, developed over similar
geological bedrock. This sequence of soil development has been described by Martini
(1 976) ,Duchaufour (1 982) and Morin (1 982) and others, and was clearly seen by the
present authors in the mountainous regions of Kenya and Indonesia.

Viewed in carefully prepared thin sections under the optical microscope, the fabric of
these soils was seen to comprise a porous, open structure of clay clusters (or ‘peds’),
roughly silt- to sand-sized, which are generally disseminated throughout with very finely-
divided iron oxides. Optical and electron microscope studies showed that the ‘clay
clusters’ are usually not “cemented” by iron oxide coatings (as has often been reported).
Rather, the clusters appear to be weakly “bonded”, possibly by a variety of mechanisms
within each type of soil. In some cases, such as many of the Kenyan nitisols (ferrisols),
the clay clusters are at least partly bonded by amorphous gel, which sometimes coats
entire peds or develops as isolated areas at the ped contacts. In other parts of the same
soil, individual peds are in direct contact, which, in the absence of amorphous gel or iron
oxide cement, are apparently joined by some form of ‘clay-like’ bond.

Prior to this study, no generally accepted methods had been established by which tropical
residual soils can be adequately classified in terms of their engineering behaviour by
means of simple ‘index’ tests. In addition, no methods exist whereby their undisturbed
(or mechanical) properties can be related to geological descriptions or to index tests, as
is the case for sedimentary soils of temperate regions (Vaughan et al. , 1988). In part, this
is probably due to the very diverse nature of residual soils, found even within the tropical
red clays of this study. However, much of the difficulty would also appear to stem from
the inapplicability of ‘conventional’ soil mechanics concepts, developed almost
exclusively from work on temperate soils, when applied directly to certain soils from
tropical zones. This inability to adequately classify tropical clay soils using conventional
soil engineering methods has led many workers to regard them as ‘problem’ soils (for
example, Morin and Todor, 1975) and particularly troublesome when encountered in
engineering works.
 3

Because the geotechnical properties of tropical residual soils appear to be largely

controlled by their composition (mineralogy) and microstructure, it would seem logical
that the first level of any classification system should take account of the factors giving
rise to their development. This entails the identification of the environmental factors
(climate, parent rock, drainage conditions, etc) and the geological/pedological processes
that lead to the formation of particular soil types. At the same time, the geotechnical
properties of the soils, based on tests appropriate to the soils themselves, and ‘the
engineering use to which the soils are put, must be determined. Such a ‘genetic’
classification, based on soil-forming processes, soil composition and microstructure in
addition to geotechnical properties, is likely to be complex, and may be unfamiliar to the
geotechnical engineer. However, only when all these inter-related elements are more
fully understood and documented will a workable classification system be derived.

The study of tropical soils has gone some way down this road, and progress continues as
engineers and engineering geologists take greater interest in these materials and more
information based on a better understanding of these soils becomes available (see, for
example, the Proceedings of the 1st) 2nd and 3rd Conferences on Geomechanics in
Tropical Soils, held in Brasilia, Singapore and Maseru in 1984, 1988 and 1992
respectively).

In late 1985, a Working Party convened by the Engineering Group of the Geological
Society was charged with the overall task of providing “a workable and practical
classification of residual soils which was scientifically based and suitable for engineering
use”. The report of this Working Party was published in 1990 (Anon., 1990). A revision
of this report is likely to be published in the near future.

T h e work undertaken in this study has concentrated on the identification, description

and engineering characteristics of tropical red clay soils. Significant effort was directed
to establishing the fundamental material composition (mineralogy and structure) and
‘typical’ engineering properties (on ‘undisturbed’ and disturbed bulk samples) of red clay
soils obtained from a variety of geological and topographical terrains and climatic zones.
This was deemed essential to confirm the sometimes confusing or contradictory
geotechnical descriptions and data found in the published literature, and to verify or
develop suggested new methods of test to more accurately determine meaningful
geotechnical parameters. T h e results of the study were applied towards advancing the
work hitherto directed (by the Working Party and others) towards deriving a practical
engineering classification.

PROJECT AIMS AND OBJECTIVES

T h e main aims of the study were:

1. T o investigate the natural development and distribution of various types of red

clay soils in relation to geology, climate, topography and elevation.

2. T o investigate the fundamental soil mineralogy and structure.

4 SECTION ONE Introduction

3. T o investigate the geotechnical properties of a variety of red clay soils by means

of standard, specialised andlor new test procedures on both disturbed and
undisturbed samples.

4. T o relate the geotechnical properties of the soils to their geological provenance,

position in the weathering profile and climatic and topographic conditions in
order that a practical classification scheme can be developed for engineering use. _.

WORK CARRIED OUT A N D REPORTING

T h e project investigations were undertaken primarily in Kenya and Indonesia in

collaboration with the Materials Testing and Research Department (MT&RD) of the
Ministry of Public Works, Nairobi, and the Institute of Road engineering (IRE),
Bandung, West Java. T h e work was carried out in two phases.

Phase 1
This was undertaken between 1986 and 1989 and, after initial desk studies, involved the
selection of study areas followed by an extensive programme of field sampling to obtain
a wide variety of red clay soils from contrasting climatic, topographic and geological
terrains. This was complemented by a comprehensive laboratory testing programme,
carried out a t the BGS laboratories in Keyworth and by the project collaborators in
Kenya and Indonesia, to establish fundamental material and geotechnical properties of
the collected soil samples.

Phase 2
Phase 2 of the study was undertaken between 1989 and 199 1 and involved the sampling
of selected deep red soil weathering profiles and further laboratory testing to investigate
the variation of soil characteristics and geotechnical properties with depth. In order to
investigate as wide a spectrum of red soil types as was practically possible, additional
samples of tropical red clays were also obtained, through BGS geologists and other
contacts, from sites in Fiji and Dominica and returned to the U K for analysis and testing
a t the BGS laboratories. All field and laboratory results were entered into an PC-
compatible database prior to undertaking data analysis and interpretation. T h e extensive
project data set is held a t BGS, Keyworth.

The results of the project investigations are presented in five main reports: this volume
and four preceding BGS Technical Reports:

WNl9311 1 Project background, study areas and sampling sites, by E:J Northmore,
M G Culshaw and P R N Hobbs.

WNl93112 Geotechnical characterisation: index properties and testing procedures,

by K J Northmore, D C Entwisle, P R N Hobbs, M G Culshaw and
L D Jones.
 5

WNJ93J13 Geotechnical characterikation: mechanical properties and testing

procedures, by P R N Hobbs, D C Entwisle, K J Northmore and
M G Culshaw.

WNJ93J14 Manual samplingmethods, by M G Culshaw, P R N Hobbs and

K J Northmore. _.

The results of the complimentary ODABGS R&D project 9 1/19 are presented in report:

WNJ93J27 Geotechnical borehole sampling o f tropical red clay soil, by J R Hallam

and K J Northmore.

These reports are augmented by a series of supplementary ‘factual’ reports detailing

results of field visits and geotechnical laboratory testing.

The results of the mineralogical and soil fabric analyses (including optical and scanning
electron microscope studies) and presented in further reports:

WGJ88J2 Mineralogy ofkaolinitic soils from Kenya and Indonesia, by S J Kemp,

A J Bloodworth and D J Morgan.

WGJ89J14 Mineralogy offirther kaolinitic soils from Kenya and Indonesia, by

S J Kemp.

WGf9 1J7 Mineralogical analysis o f tropical red clay soils, by S J Kemp.

WGf92f31 Petrographical examination of some tropical red soils from Kenya and
Indonesia, by A E Milodowski.
2 Project study areas and sampling sites
SELECTION OF STUDY AREAS
The main project investigations were centxed in Kenya and Indonesia (West Java). Initial
desk studies, including an extensive literature review, showed that these two regions, both
individually and in combination, were characterised by markedly contrasting topographic
and climatic zones resulting in a wide spectrum of tropical red clay soils developed over.-’
a variety of predominantly igneous (both volcanic and intrusive) rocks. As such, these
areas effectively enabled a considerable variety of red clay soils to be investigated within
the time, staffing and cost limits of the study. In addition, further tropical red clay
samples were obtained through contacts in Fiji (Viti Levu) and Dominica to supplement
the investigations of the soils from the main Kenyan and Indonesian study areas.
T h e choice of Kenya and Indonesia as principal study areas was also influenced by
logistical considerations. Resident BGS geologists, engaged primarily on regional
geological, resource and geochemical mapping projects in both countries, provided
valuable assistance in the project planning stages and facilitated contact with potential
collaborating organisations. Effective project collaboration with the Materials Testing
and Research Department (MT&RD) of the Ministry of Public Works & Housing in
Kenya, and the Institute of Road Engineering (IRE) in Indonesia was readily established
and maintained throughout and beyond the project study.

SELECTION OF SAMPLING SITES

T h e initial selection of regional sampling sites within the Kenya and West Java study
areas was largely based on the soil classifications and distributions presented on
pedological maps. Also taken into account were suggestions for sampling areas put
forward by the project collaborators and engineering consultants who, with their local
knowledge and experience, were able to highlight areas where current engineering works
provided exposures, cuttings or excavations to facilitate sample collection, and/or areas
where particular types of red clay soils had proved ‘problematical’ or ‘difficult’ during
recent construction projects. However, a wide-ranging regional approach to site selection
and sampling was necessarily moderated to allow time for an extensive laboratory testing
programme and careful on-site sampling of selected deep soil profiles.

Kenya
In Kenya, the selection of sample sites characterised by well-developed red soil profiles
was undertaken by reference to the ‘Exploratory Soil Map of Kenya’ (Kenya Soil Survey,
1982). This map adopts the FAO-UNESCO system for the purposes of classifying and
correlating soils. Despite the generalisations inherent in presenting the soil distribution
for the whole of Kenya at a scale of 1 : 1 million, the map and accompanying repodkey
was an extremely useful information source. T h e map attempts to present, in general
terms, the complex relationship between landforms, geology and soils using a
methodology developed by the Kenya Soil Survey from 1972 onwards. Thus, the map
legend provides information not only with respect to the soil mapping units, but also to
related landforms and bedrock geology. Accompanying the Exploratory Soil Map is an
Agro-Climatic Zone Map at the same 1: 1 million scale which, with the soil map, is
 7

intended primarily as a tool for assessing areas climatically suitable for crop cultivation.
This map provided additional information on the distribution of climatic zones (humid,
semi-humid, arid, etc.) and variations in rainfall and temperature with altitude.
On the basis of the information provided by these maps, two main areas, in central and
western Kenya (east and west of the Rift Valley, respectively), were identified as being
most suitable for the regional sampling programme. Both areas comprise large tracts of
fertile land which support relatively high population densities, and are favoured for
present and future infrastructural development (roads, transmission lines, etc).
Important existing and planned construction projects concerned with water supply
(dams, water pipelines, etc.) to feed the expanding needs of Nairobi were much in
evidence in the highlands of the central Kenya sampling area.
Sample collection was undertaken to obtain ‘disturbed’ and ‘undisturbed’ samples from
exposures, pits a n d o r boreholes to enable determination of geotechnical properties in the
laboratory, assessment of the variation of soil properties with depth and descriptions of
the weathering profiles. T h e first phase of sample collection involved obtaining
‘disturbed’ soil samples from 45 regional sites across the two sampling areas, the samples
being carefully sealed in polythene bags to retain their natural moisture contents. This
was followed by the collection of ‘undisturbed’ and further ‘disturbed’ samples from
seven trial pits (five in the central sampling area, and two in the western sampling area).
T h e trial pits were hand-excavated to a depth of c. 2 m, with an average of four
undisturbed samples taken from each pit. A further eight ‘undisturbed’ samples were also
made available to the project study from four deep trial pits excavated to depths ranging
from 2 to 6 m in the western sampling area to assess potential source material for the
Chemususu dam. The third phase of the Kenyan sample collection involved the recovery
of ‘undisturbed’ core samples from borehole sites in order that complete soil profiles from
surface to weathered bedrock could be investigated. Because the recovery of high quality
‘undisturbed’ samples, required for analysis of the in situ soil fabric and geotechnical
laboratory testing, was critical to the project investigation, the borehole sampling
programme was not as extensive as first envisaged, mainly due to the difficulties
encountered in recovering samples with minimal disturbance.
During the course of the pit sampling, it was found that despite their generally high clay
contents, almost all of the red soils tended to resemble relatively weak friable loams
(a result of their characteristic open-textured and ‘weakly-bonded’ ped-like structure).
Excess stress applied to this soil fabric when sampling resulted in severe compressional
disturbance. Therefore, great care was needed to obtain high quality samples with
minimal disturbance, particularly when sampling from boreholes. Because the problem
of sample disturbance was fundamentally important it was decided to select an easily
accessible site witb a ‘characteristic’ deep, mature red soil profile to investigate various
sampling techniques in order to achieve the optimum method of core recovery. This test
site was located in the grounds of the Jacaranda Coffee Research Station near Ruiru,
located in the volcanic footridges on the south-east facing slopes of the Aberdare
Highlands approximately 25 km north-east of Nairobi. A total of ten boreholes were
drilled a t four locations. Of these, six boreholes were used to test and assess the
effectiveness of various ‘push-sampling’ and rotary coring techniques. Continuously
sampled profiles of sufficient quality (that is,with little compressional disturbance) for
aeotechnical testing were obtained from three of the boreholes. ‘Undisturbed’ samnles
8 SECTION TWO Project study areas and sampling sites

were also taken from a trial pit excavated close to the boreholes, in order to compare the
effects of any disturbance between the pit and borehole samples (over a depth range from
ground surface to 2 m). Following these borehole sampling trials, it was found that most
standard borehole sampling techniques routinely employed in temperate clay soils, many
of which have been reported as being suitable for various tropical residual soils, are not
adequate for the recovery of high-quality ‘undisturbed’ cores in the tropical red clays
investigated in this study. Further work was clearly required to ascertain the optimum __
borehole sampling procedures for tropical red clay soil profiles. This resulted in a
separate ODA-funded R&D project (91/19) , during which field trials were undertaken
a t three locations in central Kenya in contrasting deep red soil profiles, developed on
Pleistocene and Tertiary volcanic rocks at elevations ranging from c. 1650 m to 2250 m.
Results of the investigations into sampling, from both pits and boreholes, are summarised
in Section 3 of this volume.

Indonesia
In Indonesia, the criteria used for the selection of regional sampling sites were similar to
those for Kenya, but with reference being made to the 1:250 000 scale Soil Map of West
Java (1966) for initial identification and distribution of residual red clay soils. T h e
pedological classification used for this map follows the Indonesian Classification System
of Dudal and Soepraptohardjo (1 957) which describes ‘red’ clay soils of interest to this
study in terms of two main groups: latosols and andosols. These soils are further divided
mainly on the basis of the colour of the B horizon, thus in the map legend, a distinction
is made between Red, Yellowish Red, Reddish Brown and Brown Latosols and Brown
and Yellowish Brown Andosols. These soil divisions could not be easily correlated with
the FAO-UNESCO soil units shown on the Kenya soils map but still served as a useful
guide in selecting sampling sites of well-developed ‘red’ soil profiles established on
various, dominantly volcanic, rocks in contrasting topographic and climatic zones.
Nineteen sites were sampled, encompassing a variety of ‘red’ clay soils in West Java. As
in Kenya, the regional sampling programme concentrated on the collection of bulk
(disturbed) bag samples taken from road cuttings or ‘borrow area’ excavations connected
with recent or current construction works. This was followed by the collection of
‘undisturbed’ and further ‘disturbed’ samples from nine trial pits (again, each hand-dug
by local labour to a depth of about 2 m) sited at selected locations to enable ‘typical’ soil
profiles to be described and sampled for geotechnical laboratory testing and analysis.
At the time of the Indonesian pitting programmes, it was not logistically feasible to carry
out borehole sampling operations. Therefore, to achieve the recovery of ‘undisturbed’
samples and description of soil profiles at greater depths, deeper pits had to be dug. A
fieldwork programme was undertaken in 1990 for this purpose, during which three trial
pits were hand-excavated to depths of c. 5 m. T h e Indonesian soil profiles revealed
during the excavation of these deep pits were found to have a much more layered
sequence than the almost uniform, structureless soil profiles encountered in the Kenyan
boreholes. Layering was observed in all the deep pit profiles and could be related to
differential weathering of the parent volcanic bedrock. In one particular pit, excavated
in an andosol soil developed on Younger Quaternary volcanic bedrock, the layered profile
was related to sequential weathering of successive layers of volcanic tuff.
 9

Fiji (Viti Levu)

From Fiji, bulk (‘disturbed’) samples of reddish brown tropical residual soil were
obtained from five sites in hilly terrain close to the Suva-Nadi highway, running along
the southern coast of the island of Viti Levu. The samples were collected by BGS and Fiji
Mineral Resources Department personnel, sealed in plastic bags to retain their field
moisture contents and freighted back to the UK for laboratory testing and analysis. All
the sites were involved in recent landsliding, thus some of the residual soils collected
from landslide debris were to some degree transported soils. Pedological identification
of these soils, was based on reference to the FAO-UNESCO Soil Map of the World
(Sheet X: Australasia, 1976).

Dominica (West Indies)

From Dominica, bulk (‘disturbed’) samples of reddish brown tropical residual soil were
obtained from five road cutting sites located in the northern, north-central and eastern
parts of the island. According to the pedological classification of Dominican soils by Lang
(1 967), these samples comprised dark brown ‘allophane latosolic’ and reddish brown
‘kandoid latosolic’ soils. Two additional sites were also sampled near Good Hope and
Castle Bruce on the eastern side of the island by staff from the Ministry of
Communications and Works, Roseau. Although precise locational details of these latter
sites are not known, they are apparently in landslide debris comprising reddish brown
‘kandoid latosolic’ and possibly ‘allophane latosolic’ soils, incorporating many highly
weathered, gravel-sized volcanic rock fragments. T h e Dominican soil samples were,
again, sealed in plastic bags and freighted back to the U K for laboratory testing and
analysis.

TYPES OF SOILS INVESTIGATED

From the regional and trial pit sampling surveys undertaken in Kenya and West Java, 9 3
‘undisturbed’ and 126 ‘disturbed’ (bag) samples were collected from 84 sites for
. laboratory description, testing and analysis. Thirteen ‘disturbed’ samples were obtained
from the five Fiji sites and the seven sites in Dominica. From the Kenya borehole
sampling programme, approximately 24 m of good-quality samples (three, virtually
complete, weathered soil profiles) were obtained from three sites at the Jacaranda Coffee
Research Station. Pedologically, the sampled soils included fersiallitic andosols,
ferruginous (ferrisol) and ferrallitic soils of the genetically-based French classification
system of Duchaufour (1982), adopted as a basis for engineering classification by the
Geological Society Engineering Group Working Party Report on Tropical Residual Soils
(Anon, 1990). Direct correlation between the various pedological soil classification
schemes is difficult, but approximate equivalent soil classes according to the Duchaufour
and widely-used FAO-UNESCO classification schemes for the soils obtained from
Kenya, Indonesia, Fiji and Dominica are shown in Table 1. Also given are summary
terrain and climatic details associated with the occurrence of each soil type and the
parental bedrock from which they were derived.
Detailed descriptions of all the project sampling sites, including trial pit and borehole
logs for all the Kenya and Indonesia subsurface investigations, are presented in project
report WN/93/11.
3 Sampling
Tropical red clay soils are geotechnically unusual in that, despite having clay contents
usually well in excess of 50°/0, they are friable and comprise an open, weakly-bonded
structure of silt-sized clusters of clay particles (termed ‘peds’). This structure gives a
high voids ratio and a low dry density. The soils are usually only partially saturated.
Because of this weak structure, these soils are liable to deform, break-up or crack if
subjected to relatively minor normal and/or lateral stresses during excavation or _.
sampling.
‘Undisturbed’ samples for geotechnical testing are obtained usually from tubes or split
tubes pushed or hammered into the ground, from rotary coring, or from hand-cut block
samples. Using the conventional techniques, all these methods can produce a significant
degree of disturbance in tropical red clay soils.

INVESTIGATION AND DEVELOPMENT OF SAMPLING TECHNIQUES

Techniques for obtaining samples of these soils With minimal disturbance, from both trial
pits and boreholes, have been developed progressively during four phases of project work:

i) manual sampling in plain plastic tubes.

ii) borehole sampling trials with a wide range of conventional methods.

iii) manual sampling with a new sampler incorporating a cutting shoe and axial
guidance.

iv) development of a new borehole sampler and field trials to establish the optimum
method of use.
The first three of these phases were carried out as part of the main research project
(91/18) and the fourth as a separate specific project (91119). For both manual and
borehole sampling, the common objective has been to obtain samples of approximately
100 mm diameter in a standard plastic tubing, with minimal disturbance. With manual
sampling, as carried out at surface exposures or in trial pits, the sampler length is
sufficient for 200 mm long specimens to be prepared for triaxial strength testing. The
borehole sampler can be configured to provide longer samples, usually of 0.5 m, or up
to 1.5 m when required.
Design of the samplers is based on the use of standard plastic pipe with an OD of
110 mm and a wall thickness of approximately 3 mm. For high quality results it is
important that the inner surface should be as smooth and regular as possible. This
material is preferred to metal tubing, which can corrode and be dented with reuse and
mishandling. The tropical clay soils are particularly subject to disturbance during
extrusion from a samhling tube. With plastic tubing the need for this procedure can be
obviated. On reaching the laboratory, the sample is removed by making a series of axial
cuts in the tube and then carefully removing the longitudinal tube segments.
T h e manual sampling method, for use in trial pits and other excavations, involves the
gradual insertion of the sampler into a prepared, in situ, truncated cone of soil.
It involves first digging a pit to a level about 0.2 m above the required sampling depth
 I1

and then deepening part of the pit to about 0.2 m below the sampling depth to leave a
‘step’ at one end. In an approximately 1 m wide pit, two tube samples can be taken side
by side. One half of the ‘step’ is gradually carved into an approximately 300 m m
diameter cuboid pillar; the top of the pillar is trimmed flat and the sampler is placed
vertically near the centre of the pillar. T h e pillar is trimmed to a truncated cone and the
sampler gradually pushed down into the soil cone while trimming away the excess soil.
When the sampler is slightly over-full, it is carefully removed by undercutting, the excess
soil is trimmed away flush with the ends of the sample tube and tight-fitting end caps,
preferably of plastic, are taped to each end to prevent moisture loss.
The method was used extensively in pit sites in Kenya and Indonesia. During the first
phase, the sampler comprised simply a thick-wall plastic tube with a chamfered cutting
edge machined a t one end. It was found that sample compression could be much
reduced by greasing the inner surface of the tube. For the second phase of manual
sampling, the method was modified to reduce operator error in the application of the
driving force (hand pressure) and to further reduce internal side friction along the tube
wall as the sample slides inside. T h e application of the driving force was improved by
using a modified photographer’s tripod which serves as a guide to a pushrod for driving
the sample (Figure 2). Internal side-wall friction is reduced by using a steel cutting shoe,
with a slightly smaller internal diameter, on the end of the plastic tube. As a result of
these modifications, only very small loads (applied by gentle hand pressure) are used to
ease the sampler into the soil cone, thus reducing the chance of disturbance to the soil
structure, the mpod ensuring that the load is applied parallel to the tube axis (see project
report WN/93/14).
For the first phase of borehole sampling, a heavy truck-mounted long-stroke hydraulic
rotary drill rig was equipped for drilling and sampling in soil and weathered bedrock to
a depth of about 15 metres. Sampling equipment comprised a standard British U100
sampler, a double tube core barrel with split inner tube and a retractor type triple tube
core barrel. T h e boreholes could be advanced by continuous flight rotary augers or full
hole drilling with air, foam, water or mud as the flushing medium.
During four weeks at a site near Ruiru, to the northeast of Nairobi, all the available
sampling techniques were attempted, but with only modest success. T h e results of the
drilling and sampling techniques used are summarised in Table 2 (see also Section 2).
It was found that the partially saturated soil, with clay peds bonded to form open porous
structure, readily deforms and compresses under stress to a more dense and de-
structured state.
T w o major problems were encountered with rotary flush coring. When in situ and
undisturbed, the soil has a remarkable permeability: to water, air, at least some muds and
possibly even foam. When remoulded during drilling, it transforms from a friable to a
highly plastic consistency and stubbornly adheres to metal, thus blocking the flush return.
Results with pushing drive samplers were erratic, but rather better. There was clearly a
benefit in lubricating the inner surface of a sampling tube.
T h e design and development of a sampler specifically for tropical red clay soils started
from the drive sampling approach that had achieved the best results during the drilling
12 Sampling

programme near Ruiru. This had utilised the inner barrel from the triple tube corebarrel,
fitted with a high quality plastic liner and a very sharp cutting shoe.
It was anticipated that the main problem in demonstrating the effectiveness of a sampler
would be in measuring the sample recovery (as the primary indicator of sample quality
or disturbance) with sufficient accuracy. Usual practice is to measure the overall length
of sample recovered after dismantling the sampler, and comparing this to length of
sample drive. However, in such weak and friable soils it is rarely possible to clean the --
borehole completely before sampling, or to define the physical ends of a sample with
much accuracy. At best, such measurements can provide only an indirect assessment of
any change in the length of the material sampled during the sampler drive.
The problem could be overcome if the level of the sample top could be monitored before
and after the sample drive. If this remains unchanged, the in situ sample recovery (as an
indicator of disturbance) must be 100°/~. Furthermore, by monitoring the level
throughout the sample drive, variations in the rate of sample recovery could be
investigated and possibly be related to other relevant parameters. In pursuing this
approach, the emphasis of the sampling project (91/19) was amended, from the mere
development of a sampler, to the study of the performance and effectiveness of the
sampler. For this study three parameters were monitored continuously during the
sampler drive: the vertical movement of the sampler, the drive force applied to the
sampler and level of the sample top.
T h e sampler designed a t BGS specifically for use in tropical red clay soils (Figure 3)
comprises five components: outer tube, liner tube, cutting shoe, sampler tube coupling
and sampler head. T h e four steel components were designed and dimensioned to suit a
standard PVC/CPE plastic liner tube, of 110 m m OD and 2.7 m m wall thickness, with
a very smooth bore. The sampler was manufactured for nominal sample lengths of either
1.5 or 0.75 metres, with a sample diameter of approximately 104 mm.
A schematic drawing of the test instrumentation is presented in Figure 4. Movement of
the soil sampler and of the sample top are measured, via outer and inner tubular
assemblies, by the ‘drive’ and ‘compression’ transducers respectively. A load cell a t the
upper end of the outer assembly monitors the applied load. T h e three analogue data
signals are converted to digital format and recorded on a laptop computer by simple icon-
driven acquisition software.
A test programme to study the effectiveness of the sampler and investigate the
significance of operating technique was carried out at three sites to the north of Nairobi.
These sites varied in elevation from 1650 m to 2250 m, and thus provided a range of soil
types as a result of the different climatic zones (see Section 2). T h e greater part of the
programme was conducted at the first site (Mchana, 1650 m), to study the significance
of different sampler configurations and operating techniques, and thus derive an optimal
sampling procedure. The effectiveness of this procedure (sample compression of 1 o/‘ or
less) was then confirmed by shorter sampling programmes at the second and third sites.
The raw digital field data recorded during each test comprised 1000 values each for drive
distance, sample compression and drive force, together with the times a t which each
value had been measured. These data had first to be adjusted for calibration, zero and
other corrections.
T h e following output parameters were then computed, with drive distance rather than
test time as the primary parameter:
Drive (D). Penetration or vertical movement of the sampler (mm).

Compression (C) Downward movement of the inner rod assembly,

representing sample compression (mm). .-
Drive force (L) Load applied to the sampler (kN).

Drive Rate (DR) Incremental rate of sampler movement (mm/s)

Specific Recovery (SR) Incremental sample recovery: (6D-S C)/QD (%)

T h e last parameter, Specific Recovery (Hvorslev, 1949), is a particularly sensitive
measure of the sample recovery/compression at any incremental point during the sample
drive. This parameter, and to a less extent others such as Drive Rate, is very noisy when
computed directly from the corrected digital data. Sophisticated smoothing routines were
applied to the derived data, so as to reduce noise as far as practicable, whilst retaining as
much of the finer detail as possible. These parameters have been presented graphically
for the 36 tests where digital data were recorded (see report WN/93/27 which details the
investigations undertaken into borehole sampling during both research projects).

SUMMARY OF TEST RESULTS

T h e testing programme, carried out a t three distinct sites in tropical red clay soil,
demonstrated the dependence of compressional sampling disturbance on the following
factors:

Sampler design
Cutting shoe profile. T h e cutting shoes for the BGS sampler were designed and
manufactured with a very shallow taper profile and sharp edges, such that care is
needed in their handling, to avoid injury to the personnel or damage to the shoes.
They proved to be capable of taking samples with virtually no detectable compression.
Some shoes had additionally been machined with fluted serrations, as a possible
means of further reducing penetration resistance or friction. This refinement gave no
discernable benefit. In contrast, the blunter and less tapered shoe of the standard
U 100 sampler caused severely sheared annular disturbance with any operational
technique.
Internal clearance. This clearance is achieved by machining the internal diameter
of the cutting shoe to a slightly smaller size than the bore of the sample tube. Ideally
this clearance, which is expressed as a percentage of the sample diameter, should
relieve much of the friction between the sample and the tube, but without the sample
becoming loose within the tube. (Relief of such friction also depends on having a
smooth bore in the liner tube). The BGS sampler could be assembled to give internal
clearances of approximately 0.25, 0.50 or 0.75%. T h e largest of these was found to
be too great, the sample clearly being loose in the liner. There is little evidence from
which to choose between the two smaller clearances and an intermediate value of
0.3% to 0.4% is suggested for future work.
14 Sampling

Operational technique
Lubrication. The benefit of greasing the inner surface of the liner, anticipated from
the earlier sampling programmes, was evident from the results of several tests run
without lubrication, where the onset of severe compression (Specific Recoveries of
only 50% or less) occurred quite abruptly. In one test it occurred after little more than
300 m m of sample drive (test 18, Figure 5).
_.
Drive rate. T h e drive, or penetration, rate of the sampler may vary from virtually
zero to the maximum feed rate of the rig in use. Before this testing programme the
drive rate had not been considered a major factor. Extensive research had apparently
established (at least for saturated soils) that the drive rate should invariably be as fast
as possible (Hvorslev, 1949). Whilst not always clear and consistent, the test data
indicate that sample compression is generally minimised with a drive rate of less than
15 m d s . Figure 6 shows an inverse relationship between Specific Recovery and drive
rate in test M C H 14. With high drive rates (Figure 6, tests 6, 7 and 12)the Specific
Recovery can drop to 80% or less. T o allow some margin, rates of 10 mm/s or less
would be advisable (c. 1 minute for a 0.5 m drive length).

Drive Length. Although a sample length of c. 0.5 m is generally adequate for

geotechnical laboratory testing, lengths of c. 1.5 m are much more economic if there
is a requirement for core sampling of the complete sequence, as may be the case in
engineering geological studies. The longer 1.5 m sampler was generally used, in order
to investigate the dependence of sample quality on drive length, at least to an
indicative level for engineering geological applications, but with much greater
certainty for geotechnical sampling. T h e majority of the tests demonstrate that drive
length is a factor of major significance, as most of the other factors are cumulatively
dependent on this length. Most have only a minimal effect within the initial 500 m m
of a sample drive, but may need to be closely controlled to achieve a satisfactory
sample of 1500 mm length. In most instances the length of the sampler to be used will
be determined by external considerations, such as the purpose and economics of the
sampling project.

Drive force. Like drive rate, this factor is not directly controlled during sampling,
but results from the interaction between the resistance to sampler penetration and the
characteristics of the drilling rigs hydraulic system. In this test programme, the
monitoring of drive force was a relatively minor consideration. It was carried out to
establish the general force levels required, with the possibility that it might reveal some
unsuspected features of interest. The majority of the tests show that the drive force for
the new sampler, after the initial 200-300 m m of penetration, was in the range
15-25 kN (1.5-2.5 tonnes). In those tests where the force rose significantly above this
range, it generally indicated the onset of significant sample compression.
T h e influence of soil type was most evident as a factor. In the more allophanic soils
found at the higher elevations of Kibubuti (1960 m), and Mabroukie (2250 m), sample
compression never exceeded 10 mm for a 1.5 m sample drive. Although the majority of
tests were conducted with the optimal technique, the specific use of an un-greased liner
or a high drive rate had little deleterious effect. These types of soil are evidently more
tolerant of sampling technique.
RECOMMENDATIONS FOR GOOD SAMPLING PRACTICE
T o sample tropical red clay soils with minimal disturbance, either in a pit or a borehole,
the following factors should be observed:
0 the essential components of a sampler are a plastic sample tube and a steel cutting
-*
shoe.
0 the sample tube should have a smooth bore and a minimal variation in wall
thickness. A wall thickness of 2.5-3.0 m m is recommended.
0 the internal surface of the sample tube should be lubricated, preferably with a thin
coating of silicone grease.
0 the cutting shoe should have a sharp edge and the shallowest practicable taper.
0 the internal diameter of the shoe (and thus the diameter of the sample) should be
approximately 0.3-0.4% less than that of the sample tube. It should have a very
close or interference fit on the outer diameter of the sample tube, to ensure that
the bores of the two components are as near coaxial as possible.
the sampler should be pushed into the soil with an even force, such that the
penetration rate is less than 10 m m per second.

For manual sampling, in shallow pits, the sampler should be driven by a push-rod,
laterally supported in a simple framework, such as a lightweight tripod, to ensure that the
axis of the sampler remains constant. It is recommended that the potential sample is
initially prepared as a tapered pedestal within the pit. This pedestal is then progressively
trimmed away as the sampler is pushed downwards into it.

For borehole sampling, with a drilling rig, further factors should be observed:
0 the borehole should be drilled, reamed and cleaned by mechanical techniques,
such as rotary augering, that minimise any disturbance, by percussion or flushing
media, of the material to be sampled.
0 the sampler should be push-driven in a single pass, adjusting the drive force so as
to achieve a reasonably steady rate of penetration. The drive should be terminated
if there is a rapid increase in this force.
for longer samples, of up to 1.5 m length, it is essential that the detailed design
criteria and operating technique are strictly observed.
For more critical applications, where it is desirable to quantify any sample disturbance,
a simple lightweight inner rod assembly should be incorporated to monitor any
movement in level of the sample top. This technique has been used successfully during
this project for quantifying disturbance during borehole sampling, and should be equally
applicable to the manual sampling technique.
4 Index properties and testing
procedures
Following conventional engineering practice developed for temperate sedimentary
soils, it is generally sufficient for the purposes of engineering classification to consider __
their simple index properties, which can be assessed easily, such as their particle size .
distribution, consistency (plasticity) limits or density. Test procedures for the
determination of these index tests are defined in British and ASTM standards.

Conventional index tests are carried out on samples of disturbed, remoulded soil and are
thus independent of the state of the soil in situ. However, for sedimentary soils, they not
only give a good guide to how the disturbed soil will behave when used as a construction
material under various conditions of moisture content but also give an indication of the
in situ mechanical properties of the soil. In effect, standard engineering practice, by
regarding the particle size and consistency limits as the basis of soil classification, is
asserting that the influence of mineralogy, chemistry and origin of a soil on its
mechanical behaviour is adequately measured by these simple index tests (Schofield and
Wroth, 1968). This is not the case for tropical residual red clay soils. T h e
mineralogical composition and microstructure has a pronounced influence on the
engineering behaviour and measurement of the index properties of tropical red clays,
which is generally much more complex and important than with sedimentary soils.

T h e usefulness of conventional index tests for characterising tropical residual soils in

general, and tropical red clays in particular, has been questioned many times in the
published literature. There have been two major concerns. Firstly, the problem of
obtaining meaningful and repeatable index property measurements by following
procedures laid down by current testing Standards (eg. BS1377 and ASTM). Secondly,
the concern that, unlike the great majority of temperate soils, the in situ character of
tropical red clay soils is so destroyed by the test procedures that the index parameters
cannot give an indication of the ‘mechanical’ properties of the undisturbed soil. Problems
in achieving repeatability of meaningful index test measurements relate largely to
‘standard’ sample preparation requirements prior to testing and in measuring the
moisture state of the soil as part of the test requirements. For example, because of the
mineralogical composition of tropical red clays, even partial air-drying a t ambient
laboratory temperatures may change the structure and physical properties of these soils.
Some of these changes are not reversed when the soil is re-mixed with water. These
structural changes are reflected in sometimes drastic changes in the index properties
derived from plasticity, shrinkage and particle size tests, or, sometimes, in particle
density (specific gravity).

It has become increasingly clear that conventional index testing procedures, such as those
defined in the British and ASTM standards, are not necessarily applicable to tropical
residual soils without some modification or change in emphasis. In recent years, various
modifications to these tests procedures have been attempted, usually based on local soil
engineering experience, and new indexing procedures have been put forward (for
example, Vargas 1988; Vaughan, 1988; Vaughan et al., 1988). However, many
engineering manuals in current use in developing countries continue to be based on
current British or ASTM standard testing procedures with little or no modifications
tailored to account for the particular characteristics of residual soils in the tropics.
Particle size and consistency limit test data continue to be reported widely in the
published literature, much of which has been utilised by many workers as a basis to
formulate classification schemes intended for local and more universal usage. It is to be
regretted that much of this reported index data is not accompanied by detailed
descriptions of sample preparation and test procedures employed in their determination.
This has led to confusion when attempting to correlate apparently similar test data
between soil types and to erroneous comparison of various soil classification schemes.

Investigation of the index properties of a wide spectrum of tropical red clay soils were
conducted to:

assess the applicability of conventional, modified and new sample preparation and
index testing procedures and to evaluate these procedures in achieving more
accurate and repeatable results, and

assess the implications of these results for classifying tropical red clays in
engineering terms and their applicability in indexing the mechanical behaviour of
these soils.

T h e study findings with respect to sample preparation and test procedures to obtain
repeatability of test results are summarised below. T h e implications of index test
parameters for the engineering classification of tropical red clays are summarised in
Section 6.

PLASTICITY TESTS (liquid limit; plastic limit)

The study results have shown that consistent and repeatable plasticity-related index test
measurements can be obtained on a spectrum of red clay soils provided care is taken in
their preparation, or pre-treatment, prior to testing. Drying of the soil in any way prior
to testing will effect the test results by ‘hardening’ the clustered clay structure, making
it generally more difficult to break down and leading to non-homogeneity of remoulded
soil mixes and inconsistency in measured results. Pre-drying the soil may also drive off
water held within the clay minerals and amorphous gels characteristically found in these
soils (particularly hydrated halloysite and allophane). This can change the structures of
these minerals and, in effect, cause the test sample to be totally unrepresentative of the
natural soil. Sensitivity to pre-drying is variable for different soil types. Allophanic
andosols (‘young‘ soils) are by far the most sensitive and the most ‘mature’, dominantly
kaolinitic, soils the least sensitive. Between the two ‘extremes’, soils will have a variety
of clay mineral compositions and correspondingly variable drying sensitivities. These
soils can comprise mainly ferrisols and ferrallitic soils, but also some andosols with weak
allophanic properties. Therefore, it is recommended that no pre-drying of any tropical
red soil is undertaken, and plasticity tests are carried out on soils prepared from their
natural (or as received) water contents.

All of the soils investigated in this study were, to a greater or lesser extent, sensitive to
the degree of manipulation of the soil-water mix. Repeatable results can only be attained
if the red clay soils are completely remoulded prior to testing. This can be achieved by
I8 SECTION FOUR Index properties and testing procedures

thorough hand-mixing or by a simple mechanical mixing device such as the

‘greaseworker’. A comparison of the plasticity characteristics (shown by means of
conventional ‘plasticity’, or ‘A-line’ charts) for a variety of red clay soils prepared by
standard hand-mixing and ‘non-standard’ greaseworker-mixing is shown in Figure 7
(a & b). T h e results are for soils tested from their natural state (with no pre-drying) and
with hand-mixing being undertaken for a t least 60 minutes until it was considered that
a smooth, homogeneous paste was achieved (as required by BS1377 testing standards). _.
T h e results clearly show that mixing in the greaseworker enables more thorough
remoulding (‘ultimate’ mixing) than can normally be achieved by standard hand mixing.
This is reflected in the data points for nearly all the red soil types being shifted closer to
and higher up the A-line (that is, becoming more clay-likehess silt-like and more plastic).
Data points for the andosol soils are the exception to this general trend and remain
clustered well below the A-line in manner characteristic of high plasticity silts, whether
mixed by hand or by the greaseworker. The general shift of the data points for these soils
farther away from the A-line following greaseworker mixing reflects the decrease in
measured liquid limit with increased remoulding, and was found to be unique for the
andosols soils.

Comparative tests on subsamples are advisable to check the thoroughness of the

remoulding. However, when hand-mixing is employed it is recommended that the soils
are initially mixed for at least 60 minutes and that comparative tests are done by more
than one operator. In this study, completely remoulded soil mixes were obtained after
500 ‘pumps’ of the greaseworker mixing handle. This device allows more controlled and
rapid mixing of the soils than conventional hand-mixing of the soil-water paste using
spatulas and a glass plate. T h e use of this, or a similar, mixing device is recommended
wherever possible. However, caution must be used when dealing with allophanic
andosols. Provided these soils are not dried prior to preparation, thorough remoulding
can be achieved quite effectively by hand mixing. Additional working in the greaseworker
was often found to result in a decrease in the liquid limit of these andosols, although
values still fell within the bounds of ‘extremely plastic silts’ on the A-line chart
(Figure 7). For these soils it is suggested that thorough hand-mixing alone is sufficient
for meaningful results. A comparison of the liquid limits determined from increasing
degrees of hand-mixing or between hand and greaseworker (or similar) mixes may be a
useful guide to identify the presence of allophanic soils, as no other soils investigated in
this study recorded a decrease in liquid limit with increasing manipulation.

Both hand and greaseworker mixing was made easier when the soil was mixed up in its
wettest state (above the liquid limit) and carefully dried back during testing to obtain the
required points on the ‘moisture content versus log cone penetration plot’. This ‘drying-
back‘ technique gave identical results to the conventional ‘wetting-up’ technique.

Prolonged remoulding to achieve destructuring of the soil in the laboratory will generally
not be achieved in the field. Because of this it has been suggested by some workers that
only a minimum amount of soil remoulding should be undertaken prior to testing, in an
effort to avoid complete breakdown of individual aggregations of soil particles. However,
the study results have shown that partial breakdown of the soil structure by incomplete
remoulding will give inconsistent plasticity data. Partial breakdown (or destructuring)
of these soils in a ‘controlled’ and repeatable manner is extremely difficult, if not
impossible, to achieve in the laboratory and cannot hope to duplicate field conditions
with any accuracy.

With regard to testing procedures, some difficulty was encountered in determining

accurate plastic limits using the conventional thread-rolling technique, as most of the
soils tended to start crumbling before the required 3 m m thread diameter. When this
occurred, and provided the thread remains intact, consistency of measurement was best
achieved by continued rolling to the required diameter, and the plastic limit moisture
content determined accordingly. It is recommended that this technique is adopted when
undertaking conventional plastic limit tests on these soils, and that the procedure
followed is described along with the test result. Slightly increased consistency was
obtained by adopting a method described by Harison (1988) for using the cone
penetrometer to determine the plastic limit, which tended to give slightly lower values
than the conventional method. However, practical difficulties with this method remain,
as does the uncertainty as to how plastic limits determined by the cone penetrometer
relate to those determined by the present standard test. More detailed studies and
probably equipment redesign or modification (for example, a heavier cone) are required
before this test can be recommended as a standard alternative to the current testing
procedure.

T h e plasticity chart shown in Figure 7 b presents the most consistent and repeatable
plasticity data obtained in the study and is derived from thoroughly remoulded
greaseworked soil mixes which were not subject to drying prior to sample preparation and
testing. T h e most striking aspect of this plot is the distinct clustering of the data points
in terms of their bedrock pedogenesis. T h e most intensively weathered (dominantly
kaolinitic) soils plot at the lower end of the A-line whilst soils which have undergone
progressively less intensive weathering plot parallel to and higher up the A-line. T h e least
intensively weathered soils are andosols formed over Younger Quaternary volcanics
which form a distinct cluster well below the A-line. The figure clearly shows the influence
of soil genesis, and how far the soils have progressed along the tropical weathering path,
on the engineering properties of these soils.

MOISTURE CONTENT

Despite the presence of hydrated halloysite and amorphous allophane in many of the red
soils investigated, no significant differences in natural moisture content determinations
were found between samples oven dried at 5 0 ° C or 105°C. T h e lower temperature is
that recommended by Anon. (1990) to prevent water of crystallisation being driven from
the clay mineral structure. When compared to four temperate soils (with no water of
crystallisation to drive off), the red clay soils showed the same percentage difference in
moisture content measurements for samples dried at 40, 50 and 60"C, with respect to
moisture contents for the same soils dried at 105°C. Unless there is irrefutable evidence
that significant 'structural' water is driven off from comparative tests on samples dried
at 50 and 105"C, drying at the latter conventional temperature is advised. This applies
to all determinations of moisture content, including those related to other tests such
plasticity, shrinkage and density tests.
20 SECTION FOUR Index properties and testing procedures

SHRINKAGE TESTS (shrinkage limit; linear shrinkage)

The shrinkage limit (volumetric shrinkage) is the water content at which a soil ceases
to shrink when gradually dried. Tests were conducted according to the ASTM standard
method (BS 1377 subsidiary method) on eleven remoulded Indonesian ‘red and brown
latosols’ and one ‘brown andosol’ by project collaborators in that country. The soils were
not predried prior to remoulding and placing in the shrinkage limit apparatus. Shrinkage _.
limits for the latosol soils were very consistent with values ranging from 2 3 to 32%.
Equivalent plasticity indices for these soils ranged from 33 to 48%. N o appreciable
change in shrinkage limit was noted between test samples remoulded following a
minimum hand-mixing of 10 minutes or after thorough hand-mixing of 60 minutes. All
of these soils were developed over ‘Older Quaternary volcanics’ and are probably
equivalent to ferrisols and/or ferrisol-ferrallitic soil associations in the pedological soil
classification of Duchaufour (1982). These data follow closely those determined by
Dumbleton and Newill (1962) who quote shrinkage limits ranging from 23 to 30%, with
equivalent plasticity indices of between 34 to 46%, for. The shrinkage limits of the
Indonesian latosols agreed very well with the range quoted by Dumbleton and Newill
(1 962) for halloysitic ‘laterite clays’ from Kenya (ws = 27 to 30%).

The Indonesian andosol soil contained an appreciable amount of amorphous allophane.

The shrinkage limit measured for this soil was 82.7%, much higher than those obtained
for the latosol soils. Equivalent liquid limit and plasticity index values of the soil were
201% and 149%, respectively. As with the latosol soils, the degree of remoulding
(10 minutes versus 60 minutes of hand-mixing) had no appreciable change on the
shrinkage limit. The shrinkage limit value agrees well with the 50 to 100% range quoted
by Maeda et al. (1977) for ‘typical’ allophane soils. However, lower values may be
expected for some andosol soils with ‘low’ allophanic properties (that is, in soils where,
due to increased weathering, the proportion of amorphous allophane decreases with
respect to the presence of crystalline halloysitic and/or kaolinitic clay minerals). Maeda
et al. (1977) also found that remoulding does not change the value of the shrinkage limit.

Linear shrinkage (or one-dimensional shrinkage) tests give an indication of the amount
of shrinkage by determining the change in length of a semi-cylindrical bar sample of soil
when it dries out, following remoulding a t a water content near the liquid limit. Because
the majority of the red clays obtained during the current study were found to contain
‘moisture sensitive’ halloysite a n d o r allophane clay minerals, linear shrinkage tests were
conducted by drying the test specimens at both 40°C and the standard drying
temperature of 105-1 10°C. However, for all red soil types, no significant difference in
linear shrinkage was observed between samples tested at the two drying temperatures.

T h e effect of remoulding on linear shrinkage values was examined by testing soils

remoulded to a smootfi fhomogeneous’ paste by thorough hand-mixing and by more
controlled extended mixing using the greaseworker apparatus. As was found with the
liquid and plastic limit tests, increased remoulding resulted in a corresponding increase
in linear shrinkage for the great majority of tropical red clays. Linear shrinkage values for
three UK soils (Gault Clay, London Clay and Mercia Mudstone) showed no difference
with increased remoulding.
 21

T h e greaseworker allowed more controlled and thorough remoulding of the soil-water

mixes and, hence, more consistent linear shrinkage data. For the ‘younger’ soils
developed over the Quaternary volcanics measured linear shrinkage values fell within the
higher range (23 to 30%), whereas the ‘older’ soils developed over Precambrian
Basement rocks fell within a lower range (5 to 15%). Values for soils developed over
Tertiary volcanics tended to dismbute relatively evenly between, and overlap with, these
two end ranges.

Despite the large amount of shrinkage on drying indicated by the linear shrinkage values
for those soils developed over the Quaternary volcanic soils and many soils developed
over the Tertiary volcanics, one would expect little visible cracking for these soils in their
‘undisturbed’ natural state in the field. Because of the high voids ratios associated with
these soils due to their aggregated soil fabric comprising ‘clay clusters’ (or ‘peds’) with
variable amounts of finely divided iron oxides, it is likely that the shrinkage would be
taken u p in the small inter-cluster spaces. Maeda et al. (1977) stated that this is
particularly true for allophane soils (andosols) which generally have a low natural
cohesion which decreases on drying. However, it should be pointed out that shrinkage
cracking may occur in reworked and compacted red clay soils.

In addition to indicating the amount of shrinkage, the linear shrinkage test can provide
an approximate estimate of the plasticity index (Ip ) for soils where the liquid and plastic
limits are difficult to determine (Head, 1992). Therefore, this may be useful for those
tropical red clay soils from which it is sometimes difficult to obtain reproducible results.
Despite some scatter, the following approximately linear relationship between plasticity
index (Ip) and Linear shrinkage (Ls) was obtained for a wide spectrum of soils:

Ip = 2.5 Ls - 15.2
This relationship is very similar to other (but not all) correlations between linear
shrinkage and plasticity index determined for other ‘typical’ red tropical clays in Ghana
(Gidigasu, 1976; (Newill, 196 1a). In most cases the scatter of points is probably too large
to justify the use of a single ‘universal’ formula which converts the linear shrinkage to the
plasticity index. However, when doubt exists over the reliability of the plasticity index
obtained by normal testing methods, an appropriate local correlation of I,, with L,
developed for the particular soils under investigation, may usefully be used to check
results. Linear shrinkage and plasticity are greatly influenced by the degree to which the
soils are remoulded during testing and on whether the soil is pre-dried or tested from its
natural state. Therefore, care is needed to ensure that meaningful and reproducible
plasticity index data are acquired before developing any correlations with linear shrinkage
for a particular red soil type.

Because accurate measurement of the plastic limit (needed for calculation of the
plasticity index) can be difficult to determine for some red clay soils, correlation of linear
shrinkage values (L5 ) with liquid limits (w,.) were also examined. T h e results showed
a reasonably good linear relationship between all soil types, but with the andosol soils
clearly forming a clear ‘grouping’ distinct from the more mature ferrisol and ferrallitic
soils. For all the red clays except the andosols the overall relationship between linear
shrinkage and liquid limit was:
w,. = 3.9 L, + 6.3
22 SECTION FOUR Index properties and testing procedures

When the soils were considered in terms of their development over different parental
bedrock the following relationships were established:

Soils develoDed OK

Younger Quaternary volcanics wL = 6.6 Ls

(andosols) _-
Timescale and intensity
Older Quaternary volcanics WL = 2.9 Ls + 39.8
of weathering
(ferrisols & ferrallitic soils)
increasing
Tertiary volcanics wL = 2.7 LS + 29.8
I (ferrisols & ferrallitic soils)

i Pre-Cambrian Basement wL = 2.2 Ls + 24.5

! complex [mainly granites]
i (ferrallitic soils)
Similar correlations have been obtained between linear shrinkage and such properties as
clay content, liquid and plastic limits, optimum moisture content, field moisture
equivalent and volumetric shrinkage for soils developed mainly over phyllites and granites
from Ghana (Gidigasu, 1976). These, and the derived relationships found in this study,
clearly indicate the influence of soil genesis on their engineering properties.

Experiments undertaken on nine soil samples to examine which soils shrink irreversibly
on drying showed that only andosols with significant allophane content showed virtually
no additional shrinkage when re-wetted and re-dried. Dominantly kaolinitic ferrallitic
soils showed almost completely reversible shrinkage characteristics. Between these two
extremes, soils with a mixed allophanic, halloysitic and kaolinitic clay showed variable
reversibility on re-wetting and re-drying, with linear shrinkage values ranging from 30 to
60% of the original test value.

PARTICLE SIZE DETERMINATIONS

With the exception of the allophanic andosol soils, no difficulty was encountered in
undertaking meaningful particle size analyses, even for pre-dried samples. Dispersion of
the soils was effectively achieved using ‘standard’ sodium hexametaphosphate dispersant,
provided the dispersant-soil suspension was shaken for at least 12 hours, or overnight.
Comparative tests using the same dispersant but mixing in a high-speed mechanical
stirrer for 15 minutes (as specified in the ASTM and 1967 British Standard) showed that
the latter did not enable full dispersion of the soil particles and is not recommended.
Although, pre-drying did not significantly affect the results for these soils, it is
recommended that pre-test preparation and testing are performed on soils at their natural
state (not pre-dried). Pre-drying significantly effected the particle size distributions of the
andosol soils, and in practice a testing programme may be undertaken without realising
allophanic andosols may be involved. Testing the soil without pre-drying also maintains
consistency with the recommended preparation procedures for other index tests.

With respect to the allophanic andosol soils, the problem was not in dispersing the soils,
but in preventing them re-flocculating (a condition also reported by Wesley, 1973).
Maeda et al. (1977) stated that different allophane soils react differently to different
 23

dispersion treatments. If this is the case, it is suggested that grain-size analysis has a
limited usefulness in characterising these soils for engineering purposes, as continuous
experimentation will be necessary to obtain maximum dispersion. However, not all
andosols have pronounced allophanic properties and, in the first instance, it is
recommended that particle size distributions are carried out following the procedures
adopted for other red clay soils using ‘standard’ dispersant. Wherever possible, the grain
size analysis should be done on field-moist samples.

Typical particle size distributions for a variety of tropical red clay soils from Kenya and
Indonesia, developed over various types of parent volcanic bedrock, are shown in
Figure 8.

PARTICLE DENSITY (specific gravity)

Particle densities varied widely for all the red soil types investigated (2.66-2.95). Results
for samples which were tested from their natural moisture content (no pre-drying)
generally gave higher measured particle densities than those subjected to oven pre-drying
(at the standard 105-1 10°C). The reason for this possibly lies in the loss of water of
crystallisation (or intra-particle or inter-layer water) from the clay minerals on oven-
drying, and hence a reduction in the density of the solid matter. No clear distinction was
found between pedological groups which either do, or do not, show a reduction in
measured particle density on drying. However, the Kenyan ferrisols and ferrallitic soils
developed on Tertiary volcanics and Precambrian Basement Complex rocks (mainly
granites) are probably the least susceptible as a whole. T h e Indonesian ferrisoUferrallitic
soils (‘latosols’) developed over Older Quaternary volcanics appear to be most susceptible
to reduction in particle density with drying, and andosols possibly less so. This does not
agree with the findings of Wesley (1973) for similar tests on Indonesian ‘latosols’ and
andosols. However, experimental error may be considerable, resulting in scatter
unrelated to pedology.

No clear correlation could be established, in this study, between the reduction in particle
density on pre-drying and the percentage of hydrated halloysite present in the soil, as was
described by Newill (1 96 1b) for a halloysitic red clay from Kenya. In this respect, the
results agree with Wesley’s findings for Indonesian red clays which showed little or no
reduction in particle density with pre-drying. Mineralogical analysis on similar
Indonesian red clays investigated in this study showed that many of these soils contained
appreciable amounts of hydrated halloysite.

Errors in the determination of particle density, or incorrect assumptions of particle

density, result in incorrect determinations of clay content, voids ratio, porosity, and
hence consolidation parameters. It is recommended that particle densities are carried out
a t natural, or ‘as received’ moisture contents, rather than on pre-dried samples. T h e
moisture content a t which the test is carried out should be measured and quoted.
Discrepancies reported for pre-dried results may originate from differences in the degree
of ped breakdown of the dried sample prior to the test, rather than to the loss of
crystallisation, or intra-particle, water per se.
24 SECTION FOUR Index properties and testing procedures

RECOMMENDATIONS FOR GOOD TESTING PRACTICE

General
0 F o r all index tests on tropical red clay soils, it is vitally important to report the
sample preparation and test procedures undertaken to obtain the measured result.
Without this information, comparison with test results on similar soils from other -*

sites are suspect and can lead to erroneous assumptions regarding classification
of the soil for engineering purposes.

Preparation of disturbed samples

0 It should be assumed that all tropical red clay soils will be affected in some way
by drying. Care should be taken to preserve the natural (as sampled) moisture
contents by ensuring adequate sealing of field samples to prevent moisture loss.
O n return to the laboratory the samples should be kept sealed, and stored in a
cool environment, until required for testing. Classification tests should then be
carried out on the soil with no additional drying prior to testing.
0 T h e majority of residual red clay soils have a friable texture which generally
allows samples to be disaggregated, or broken down, sufficiently by light finger
pressure to allow sub-division by accepted riffling or quartering procedures to
obtain representative samples for testing.

Recommendations for preparation procedures for individual test requirements are

given below. T h e method of preparation should always be reported.

Moisture content
0 It is recommended that moisture content determinations are carried out at the
‘conventional’ oven-drying temperature of 105-1 10°C.

Plasticity tests
T h e standard fall cone apparatus (cone penetrometer) should be used in
preference to the Casagrande apparatus for the liquid limit test.

The soil should be mixed by adding water to the soil in its natural (or as received
state) and worked as necessary for the liquid and plastic limit tests. Drying of the
soil prior to mixing should be avoided.

Easier and more efficient mixing is achieved by mixing the soil at a moisture
content above the liquid limit (that is, in its wettest state), followed by gradual
drying back to achieve the required data points on the moisture content versus
cone penetration plot. At least five points should be determined.

Repeatable results can only be attained if the red clay soils are completely
remoulded prior to testing. The amount of manipulation to which the soil is
subjected determines the extent to which the soil is broken down and has a
significant effect on the measured test results. T h e sensitivity of the soil to
remoulding should be verified using a range of mixing times prior to testing. It is
 recommended that conventional hand-mixing should be undertaken initially for
a t least 60 minutes and, if necessary, additional mixing carried out to ensure
thorough remoulding of the soil-water paste. A simple mechanical mixing device
such as the ‘greaseworker’ provides the most thorough and efficient means of
mixing the soil, and is independent of variations in strength/energy of the
operator. It is highly recommended that such a device is used. _.
T h e one-point liquid limit test is not advised for use with these soils.

0 Difficulty may be encountered in determining accurate plastic limits using the

conventional thread-rolling technique, as some red soils tend to start crumbling
before the required 3 m m thread diameter. When this occurs, and provided the
thread remains intact, consistency of measurement can best be achieved by
continued rolling to the required diameter, and the plastic limit moisture content
determined accordingly (the procedure should be reported along with the test
result). Where an intact soil thread of c. 3 mm cannot be achieved by continued
rolling of the soil, the soil should be recorded as ‘non-plastic’.

Shrinkage tests
0 Linear shrinkage test on remoulded red soils are affected by the degree of
remoulding. Procedures for the thorough mixing of the sample prior to testing
should follow those described for the liquid limit test.

0 T h e linear shrinkage should be determined at the conventional drying

temperature of 105-1 10°C.

0 It is important to distinguish between those soils which shrink irreversibly on

drying and those which do not. This may be achieved by re-wetting and
remoulding the samples used in the linear shrinkage test. Repeat testing of the
same samples enables calculation of the percentage of additional shrinkage, if any.

Particle size distribution

Pre-drying of the soil should be avoided. The initial sample should be weighed
and a subsample taken for moisture determination, so that the initial dry mass
can be calculated.

Disaggregation of the soil sample prior to sieving and sedimentation analysis

should be done using ‘standard’ sodium hexametaphosphate dispersant. T h e soil-
dispersant solution should be placed in a sealed flask and gently agitated for at
least 8 hours (or overnight) in a flask shaker, to ensure proper disaggregation.

Following disaggregation, the soil (plus dispersant) should be washed through a

63 p sieve and collected. Gentle agitation by light finger pressure may be used
to facilitate washing through of the fine particles. T h e material remaining on the
63 pn sieve should be wet sieved through the normal range of sieves. Additional
material passing the 63 pn sieve should be added to that collected from the initial
wet sieving.
26 SECTION FOUR Index properties and testing procedures

Sedimentation analysis of the fine fraction should be undertaken according to

standard procedures.
0 Consistent and meaningful particle size distributions are difficult to obtain for
allophanic andosol soils. Continuous experimentation is necessary to obtain
maximum and permanent dispersion. However, not all andosols have pronounced
allophanic properties and, in the first instance, it is recommended that particle __
size distributions are carried out following the procedures adopted for other red .
clay soils using ‘standard’ dispersant.

Particle density (specific gravity)

0 Particle densities should be carried out on soils at natural, or ‘as received’
moisture contents, rather than on pre-dried samples. T h e moisture content at
which the test is carried out should be measured and quoted.
0 After testing the soil may be oven-dried at the ‘standard’ temperature of
105-1 1 0 ° C to determine its dry mass for use in the equation for calculating
particle density.
5 Mechanical properties and testing
procedures
Mechanical properties ones dependent on applied stresses or conditions. They are
distinguished from index properties, which may be considered as characteristic of the soil
solids rather than of the solid/void structure, and are usually derived from tests carried
out on disturbed or deliberately reworked samples. In addition, compaction properties
and residual strength tests are briefly considered. These tests were carried out on
disturbed samples.

Mechanical properties, unlike the index properties, are capable of taking into account the
natural soil fabric or structure (including voids) , and stress conditions, and tend to be
more complex and time consuming, with difficulties often experienced in specimen
preparation. The main part of the mechanical properties test programme dealt with
triaxial and oedometer tests, which, despite the use of computer control and datalogging,
was designed to follow standard procedures recognised worldwide. Determinations of
strength, deformation, saturation,permeability and consolidation properties were made
as part of the triaxial test procedure. All triaxial tests were of isotropically consolidated,
undrained compression (CIU or ICv> type, and were carried out on specimens saturated
(by raising of back pressure) as part of the preliminary test procedure. With the exception
of four tests on Kenyan specimens, triaxial tests were multi-stage (multi-CIU) ,that is,
all test stages were carried out on the same specimen. In addition, two sets of samples,
one from Kenya and one from Indonesia, were prepared in both destructured (that is,
slunied and re-sedimented) and compacted (standard Proctor) forms for triaxial testing
in order to allow comparison with the undisturbed condition. The results of a limited
number of compaction and ring shear tests are also considered. The test results are
summarised in Table 3.

Tests to determine mechanical properties have been selected in order to reflect standard
practice worldwide, or to extend or modify that practice in areas where the need for
improvement is perceived. Whilst modern apparatus has been used, particularly in the
case of the triaxial tests with computerised datalogging and control, the fundamental
nature of the tests is reproducible in most soils testing laboratories. No attempt has been
made to carry out unsaturated tests for, without special equipment for measuring air and
water phases separately, the test is unreliable (Fredlund and Rahardjo, 1993). However,
problems have been experienced in the saturation process.

For the determination of index properties, the inadequacy of applying standard practice
to tropical red clay soils has been discussed in Section 4. This also applies, but to a lesser
extent, to the measurement of mechanical properties. Multi-stage isotropically
consolidated, un'drained (multi-CIU) triaxial strength and multi-specimen one-
dimensional oedometer consolidation tests have been successfully carried out on these
soils. Procedures described in Head (1986) were followed with minor exceptions. A total
of 26 triaxial and 100 oedometer tests were carried out. In the main, 100 mm diameter
X 200 mm long maxial specimens and 50 or 75 mm diameter X 20 mm thick oedometer
specimens have been used. A small number of conventional mplicated 50 mm x 100 mm
(CIU) triaxial specimens were tested , but these proved less successful.
28 SECTION FIVE Mechanical properties and testing procedures

In most cases, ‘mechanical’ samples have matching ‘index’ samples so that correlations
can be made between them. Such relationships have been described in the literature and
are here investigated and examples given. Unusual consolidation results are described
from some triaxial tests. These appear to relate to the saturation regime in the triaxial
test, affecting the normal test procedure. The phenomenon of collapse, that is sudden
densification on flooding under a constant applied stress, has been investigated in the
oedometer. The effects on mechanical properties of total destructuring of the soil --
followed by resedimentation, and of standard compaction have also been investigated by
special triaxial preparation techniques and SEM study. These have led to conclusions
about the nature of bonding within the soil structure.

SPECIMEN PREPARATION

Difficulties have been experienced in the preparation of test specimens from

‘undisturbed’ samples of tropical red clay soils of both allophanic and halloysitic types.
These stem from their high clay content and stickiness at natural moisture content and
yet, at the same time, their friable and relatively weakly bonded ped structure and low
density. Any form of trimming of specimens at natural moisture content, be it by cutting,
turning, scraping, either by hand or machine, tends to be extremely difficult. T h e use of
normal trimming tools tends to ‘pluck out’ or ‘peel’ large sand-sized clusters or
aggregations of clay. This is particularly the case for the halloysite-rich types. T h e use of
pushdriven tube samples, taken from either pits or boreholes at the same diameter as the
maxial specimens (see Section 3), greatly facilitated preparation, as trimming was then
confined to the specimen ends. It was found that for the wetter and stickier soils a small
scalpel was best suited for trimming, whereas a hacksaw blade could be used for drier or
less sticky soils.

Oedometer specimens were easier to prepare, it being usually possible to obtain four or
five specimens per tube sample. Nevertheless, the problem of plucking-out or peeling-off,
even when using a sharp blade, was ever present. A process of careful paring, rather than
a single clean cut, had to be employed. A cheese wire, typically used for cutting clay soils,
was not usually successful due to the low density of the soils. In some cases, where a level
end surface could not be achieved on a specimen, fine uniform quartz sand was used as
a filler. Handling of the 100 mm triaxial specimens was particularly difficult. The
problem of open or loosely infilled termite or root holes in triaxial test specimens was
dealt with by filling with a soil paste where possible. If such holes intersecting the
specimen surface were left unfilled, membrane entry into them caused volume errors and
membrane failure (Bohac and Feda, 1992). A single fresh standard latex membrane was
used for each test. No filter drains were used on the specimen sides. Drainage from one
end of the triaxial specimen during isotropic consolidation was used, though drainage
from both ends would have been beneficial. Traditional methods of estimating the
duration of undrained compression stages, based on c, values obtained from isotropic
consolidation stages, were not used. This was due to the difficulty in calculating c, using
the actuators Also, it is likely that such calculations are inappropriate for these soils;
probably producing too rapid a test rate.

Various types and sizes of tube were used during the project for sampling in hand-dug
trial pits. Initially, sharpened aluminium or thick-wall plastic tubes (100 to 104 mm
 29

bore) were used. Large diameter (150 mm) cardboard tubing was also tried in order to
obtain adjacent multiple 50 mm maxial specimens, but with limited success. T h e design
and operation of a new sampler for manual use in trial pits has been described in
Section 3. This design added a sharp, reusable, cutting shoe to a plain ended plastic tube
of 250 m m length and ensured straight alignment of the tube. T h e length was intended
to provide a 2: 1 triaxial specimen after end trimming. In the laboratory, the tubes weFe
cut from the sample by hacksaw or grinding wheel. This operation required several stages
to prevent sticking of the sample to the tube segments, despite the use of release agents
(subsequently, a vibrating saw jig was developed, which removes plastic tubes more
accurately and rapidly, causing less damage to the sample). Extrusion of samples was
avoided throughout. Plastic tube samples from trial pits were not waxed, but weighed on-
site and sealed with plastic end caps and vinyl tape and packed in padded boxes. Storage
was in a humidity and temperature controlled room (9"C, 80% RH). Under these
conditions storage periods of up to five years have resulted in moisture losses of about 5%
accompanied by shrinkage away from the liner wall of about 0.5 mm.

Destructured specimens were prepared by slurrying, achieved by mixing with water in

a 5 litre mixer and a grease-worker (see Section 4), and then sedimenting in plastic tubes
with side-drainage filters. A period of 3 to 6 months dewatering and air drying was
allowed under a small axial stress. Laboratory compacted samples were prepared by
Standard Proctor method but using a plastic sampling tube, as used for undisturbed
sampling, rather than a Proctor mould. This was to allow removal of the specimen
without extrusion.

In the case of four Kenyan trial pits, traditional triplicate (50 X 100 mm) triaxial
specimens were hand trimmed from single 150 m m diameter cardboard tube samples
(see project report WN/93/13). This method had limited success due to the difficulty in
using the cardboard tube, loss of moisture through the tube, and frequent disaggregation
and slumping of the test specimen during the extension stage of the triaxial test. This was
believed to be due to the small size of the specimen and its inability to withstand large
. changes in effective stress.

TRIAXIAL TESTS

Method
Single multiple-stage, 100 x 200 mm, isotropically-consolidated , undrained triaxial tests
(multi-CIU) were used throughout, with the exception of four Kenyan trial pit samples
for which traditional triplicate 50 x 100 m m specimens were prepared (CIU). T h e
apparatus used for the test program was a 100 m m Bishop & Wesley type hydraulic
triaxial cell (confining stresses to 1700 Wa, deviator stresses to 5200 Wa), forming part
of an integrated computer-controlled stress-path testing system (for soils).

As a first stage, saturation in steps of 50 minutes by elevated back-pressure was used

(Black and Lee, 1973). Saturation was monitored by checking Skempton's pore
parameter B, a saturation of between 96 and 99.9% being assumed when B = 95% (Baldi
e t al., 1988). Stages of isotropic consolidation (usually 24 hours) were followed by
undrained compression and extension. In the case of the multi-stage tests, more
consolidation stages (between three and ten) than compression stages (between three and
30 SECTION FIVE Mechanical properties and testing procedures

five), were usually carried out. Compression (axial loading) was stress-controlled
whereas, for software-related reasons, extension (axial unloading) was strain-controlled.
Permeability tests were carried out using a direct constant flow rate method (Olsen et al.,
1985) at a variety of junctures in the triaxial test, but usually post saturation, at effective
stresses of between 20 and 30 kPa. No filter drains were used, and drainage was to one
end of the specimen only via a porous disc (drainage could have been allowed from both __
ends in order to speed the test). Cell and pore water volume changes were measured
throughout. Triaxial parameters were calculated from standard linear Mohr-Coulomb
plots.

Consolidation in the triaxial test can be problematic with these high voids ratio,
heterogeneous soils, particularly where termite or root holes are present. Deformations
during consolidation may be large, even at low effective stresses, and not necessarily
uniform. This may result in considerable alteration of specimen shape and misalignment
or tilting of the top cap and ball seating prior to the compression stage of the test (Baldi
et al., 1988).

A perceived problem in the multi-stage C I U triaxial testing of sensitive soils, or those

undisturbed soils where structure dominates geotechnical behaviour, is that the structure
of the test specimen changes during the test (Anon, 1990). This is certainly the case with
the tropical clays described here and the results reflect the changes taking place in the
specimen during the test. It is less easy to infer the causes of these changes at each stage.
Similar changes take place during the oedometer test.

Results
Results of triaxial tests may be displayed statistically as derived test parameters, for
example cohesion, angle of internal friction, (Table 3), or graphically as saturation versus
stress, Mohr-circle or stress-path plots, and isotropic consolidation. Examples are given
in Figures 9, 10 and 11, respectively. Selected parameters derived from triaxial and
oedometer tests have been plotted against each other or against ‘index’ parameters.

T h e progress of saturation of triaxial specimens is shown in Figure 9 as plots of

Skempton’s B value versus back pressure for Kenyan and Indonesian samples. It will be
noted that starting B values vary widely, the Kenyan sample plots are less scattered than
the Indonesian ones, and some specimens required in excess of 500 kPa to achieve a B
of 0.9. Some Kenyan specimens show an initial reduction in B at low stresses.

The results of several extended time period isotropic consolidation stages at low effective
confining stresses suggest that a two-phase process is taking place in some specimens
whereby a ‘swelling‘ process either is contemporaneous with, or closely follows, a normal
secondary consolidaGon process. Similar behaviour has been observed with a partially
saturated Gault clay specimen from the U K using the same apparatus and method.
However, the tropical clay results show that, whilst the volume of the specimen continues
to reduce as part of the normal consolidation process, the specimen also begins to take
in pore water at a steady rate for many days or weeks. Examples are given in Figure 1 1.
The latter process starts, or is detectable, at between six and ten hours after the start of
consolidation, and is particularly marked in the case of Kenyan specimens. On current
evidence it seems likely that this ‘swelling’ process is the result of incomplete saturation
 of the microstuctural peds at low effective confining stresses (< 100 Wa), so that whilst
the macrostructure is consolidating the microstructure is swelling; the volume changes
involved in the latter become more significant as normal consolidation diminishes with
time.

Study of specimens in the scanning electron microscope (SEM) taken before and after
the consolidation stage described above show a denser, more fissured soil fabric after
consolidation, compared with before. Soil peds consisting of partially decomposed,
porous, rock fabric were observed surrounded by a thin, but continuous layer of gelified
clay. T h i s was bounded by fissures. It is conceivable that such peds could be ‘prised
open’ by the stresses associated with consolidation, thus freeing new void space for
saturation from the adjacent macro voids. In this way water intake (or ‘swelling’) could
take place while the consolidation process, involving the closing of macro voids, was still
underway.

Experiments were conducted which appeared to rule out procedural factors, such as
membrane leakage, in this ‘swelling’ process. Estimates of long-term membrane leakage
(Lerouil et al., 1988) suggest volume changes an order of magnitude smaller than those
measured, for the same time period. Swelling was not observable, that is as a separate
process contemporaneous with consolidation , in any conventional oedometer tests due
to the fact that strains due to swelling and consolidation cannot be distinguished from
one another.

A problem experienced with the type of stepper-motor actuator used on the back pressure
line during the early consolidation stages was that the actuator could not keep pace with
the very rapid rate of consolidation of the average red clay soil. Thus, the initial part
(typically the first minute) of the deformation versus time plot was incorrect. This
problem is magnified by the use of large test specimens. A solution to this problem would
be to use a larger actuator or a different form of volume measurement for the
consolidation stages of the triaxial test, for example, the more traditional constant
, pressure, oil-water or compressed air systems combined with an in-line, passive volume
measuring device. Alternatively, the measurement of volume could be separated from the
control of pressure. Typically, a 100 x 200 m m specimen of allophane-rich clay expelled
a total of 300 ml of pore water, i.e. approximately 20% of its original bulk volume, during
the consolidation stages of a triaxial test; most of this amount being expelled within the
first minute of each stage. This resulted, in extreme cases, in a considerable alteration
of the specimen shape, from a right cylinder to an ‘apple-core’ shape. In contrast, field
and laboratory compacted specimens gave small pore water volume changes of only
2 to 4%.

Permeability results show an exponential reduction in permeability with time, up to

8 hours, beyond which a state of equilibrium is established. Results for saturated,
undisturbed specimens show high permeabilities of 5.6 x 10-8 to 1.1 x 10-5 m/s for
Kenyan and 2.2 x 10-8 to 5.6 x 10-6 m/s for Indonesian samples. These values are
comparable with those of silt or very fine sand. Constant flow rate permeability test
stages, carried out during triaxial tests, revealed a 100 to 200 fold decrease in post-
saturation (water) permeability with destructuring, but only a two to three fold decrease
with laboratory compaction, compared with the undisturbed state. T h e test is particularly
sensitive to fluctuations in environmental temperature. Permeabilities may be
32 SECTION FIVE Mechanical properties and testing procedures

significantly increased by bioturbation (Blight, 199 1). Comparison between field and
laboratory permeability data depend on their relative states of saturation.

Values of effective cohesion, c’, are very variable ranging from 0 to 97 kPa, but are
generally high. Values of effective angle of internal friction, #, range from 11 to 41 O

(Figure 12). Again, these values are high when compared with temperate soils of similar
plasticity or clay fraction, and are more typical of silts and fine sands. This generally
agrees with the findings of Wesley (1977) who also tested soils from Java. T h e effective --
friction angle, +I, is typically about double the total friction angle, .@
,, However, the
effective cohesion, c’, is either similar to the total cohesion, c,, or slightly lower.
Interestingly, the Indonesian total cohesion results appear to consist of two populations,
one from 10 to 20 kPa and the other from 30 to 50 kPa. Wesley (1977) noted that
allophane-rich soils gave higher strengths than halloysite-rich soils. This is not borne out
by the data presented here. Values of effective friction angle within the Chemususu group
of samples are very similar. Axial strain at the yield point ranged from 1.5 to 3.0%, with
the exceptions of the allophane-rich Indonesian samples (Pits 7, 9, and 1 1) and K89/3F
Kenyan samples, which had strains of around 5 to 6%.

The two sets of undisturbed / destructured / compacted samples (Indonesia Pit 1 1 , 4 m

and Kenya K89/3F/lB, 0.8 m) revealed surprisingly similar stress-path and strength
behaviour despite large structural differences. They showed that the estimated effective
shear strengths of undisturbed and compacted specimens for the Indonesian set (an
allophane-rich soil) are similar, with the destructured lower. In the Kenyan case
(a halloysite-rich soil) the compacted shear strength is much higher than the
desuuctured, which is itself higher than the undisturbed. However, if effective friction
angles alone are compared, the Indonesian destructured specimen has the highest of the
set while the Kenyan compacted specimen has the highest; the undisturbed specimen
having the lowest in both cases. These trends are matched by the dry densities. If these
results reflect the overall situation it suggests that reworked material, for example fill,
should be capable of greater slope stability than undisturbed material. However, this does
not take into account the large permeability and self-weight differences between the states
which mitigate against stability. A typical engineered fill is probably in a state part way
between the laboratory compacted and destructured states. In addition, some earthwork
fills placed using a sheepsfoot roller have revealed a partially sheared or remoulded fabric.

A familiar problem in conducting multi-stage triaxial tests is in determining the correct

axial stress at which each stage of compression should be terminated (Head, 1986). This
is particularly difficult for the first stage where no pattern of behaviour has been
established. With structural changes taking place during the consolidation and
compression stages of the test a pattern may not be established. On-line test plots of
stress-ratio and pore pressure are useful here. T h e choice of whether to use multi-stage
or multi-specimen tests is influenced by structural alteration, lithological and structural
variability, and specimen size.

Transitions from below to above bonding yield, with each stage of increasing effective
confining stress, can be seen in the shape of some triaxial compression stress-paths, for
example Kenya Chemususu Pit 3,5.7 m and Indonesia Pit 11, 4.0 m (Figure 10). These
are analogous to ‘over-consolidated’ and ‘normally-consolidated’ behaviour in
 33

sedimentary soils, that is, a transition from ‘dry’ to ‘wet’ of critical with increasing
effective stress. T h e transition occurs at stresses of between 50 and 150 kPa. This may
also represent the gradual transition from an open, bonded structural state to one more
typical of a temperate clay soil, The effect of bonding as described by Vargas (1973), can
also be seen in a plot of triaxial maximum shear stress versus effective average stress
(Figure 13). Here a ‘bonding‘ or ‘apparent overconsolidation’ effect is seen as a flattening
of the plot line below a yield stress or bond strength. This is most clearly seen in. &e
Kenyan undisturbed (halloysite-rich) specimen below t’ = 150 kPa. It is not seen in the
Indonesian undisturbed (allophane-rich) specimen which has a straight plot line.

Failure modes during the compression stages were typically a combination of barrelling
and shearing, developing contemporaneously. Shearing often consisted of either one pair
of conjugate shears or one major and one minor pair. In some cases shearing was not
clearly defined but rather a kind of ‘stratified’ densification and spreading occurred. In
other cases complex shear zones and some tensile ‘columnar’ failure elements were seen.
It was frequently observed that shearing was accompanied by tilting of the upper part of
the specimen. This is not uncommon for tests using a ball-jointed top cap. Few
specimens developed a classic single shear plane by the end of final compression
(Indonesia Pit 1,2.0 m and Pit 4,2.0 m are ones that did). More often, multiple shallow-
angled shears, or an absence of discrete shears, were noted. T h e allophane-rich
specimens, in particular Indonesia Pit 11, 4.0 m, showed drastic volume reduction
resulting in an apple-core shape, followed by barrelling and multiple shears. T h e two
laboratory compacted specimens did not exhibit shears or significant barrelling, but
appeared to compact further during the compression stage of the test.

OEDOMETER TESTS
Method
Multiple oedometer consolidation tests were carried out on push-driven tube samples.
T h e samples from trial pits matched those taken for triaxial tests in most cases.
. Conventional front-loading, dead-weight, lever-arm type apparatus were used, capable
of applying axial compressive stresses of up to 2000 kPa to the specimens mounted in
open, fmed-ring cells, with drainage at both ends, and with computer-controlled logging
of deformation. T w o high capacity tests were carried out in a converted lever-arm type
tensile testing machine capable of applying up to 32 MPa axial compressive stress on a
75 m m diameter specimen.

T h e use of multiple tests on specimens taken from the same ‘undisturbed’ sample
enabled an examination of collapse behaviour, and the effects of pre-drying the specimen.
Each set of tests included at least one where flooding of the test cell took place at the start
of the test (that is, the British Standard method). For ‘collapse’ tests the specimen was
flooded at stresses of between 100 and 200 kPa as a separate test stage inserted into the
normal 24 hour loading cycle. Oedometer tests were also carried out at the MT&RD,
Nairobi, Kenya. These used shorter time increments than 24 hours, and mostly dealt
with shallow samples (- 1.O m depth) not matching those used at the BGS.
34 SECTION FIVE Mechanical properties and testing procedures

Results
An example of the progress of settlement with time in the consolidation test is shown in
Figure 14. This shows a plot of settlement versus (log) time for all the loading stages for
Kenya Pit 1, 2.0 m (C). It will be noted that a t low stresses primary consolidation is
almost instantaneous. At moderate stresses it is very quick, and a t high stresses it
resembles that of a conventional temperate soil plot, that is, the plot has a recognisable _.
parabolic primary phase followed by a secondary phase, the two separated by a point of
inflection. T h e transition appears to have taken place between 800 and 1000 kPa.

T h e voids ratio versus (log) normal stress plots (‘e - log P’) are shown in Figure 15.
Many samples showed the tendency to exhibit true collapse to a greater or lesser extent,
in particular the shallow samples (1 .O m depth) tested at MT&RD and a shallow road
embankment sample (field compacted). Perhaps surprisingly, the samples with highest
voids ratio did not show collapse. Many other samples showed a marked deviation from
the smooth voids ratio versus log stress curve, at an axial stress (typically between 100
and 200 kPa), apparently unrelated to the stress at which flooding took place.

Initial undisturbed voids ratios were highly variable, from 1.14 to 5.36 for Indonesia and
from 0.92 to 3.47 for Kenya, allophane-rich soils tending to have the highest values in
both cases. These high voids ratios (and low densities) are accompanied by high natural
moisture contents (greater than 150% in one case). Variation in initial undisturbed voids
ratio between specimens taken from the same tube sample, but for different purposes,
was between 1 and 18%, but typically around 5%. For both Kenyan and Indonesian
samples it is the allophane-rich soils which showed the most variability, probably the
result of lithological variability and sample disturbance. In three Indonesian trial pits
(9,10, and 1 1) the voids ratio was found to increase with depth, rather than decrease.
T h e reverse was the case for those Kenyan samples taken from the same borehole.

T h e coefficient of consolidation, cv, was impossible to calculate in most cases, a t low

effective stresses, because of the very rapid primary consolidation. In fact, up to half of
the total consolidation settlement took place typically within the first few seconds of the
test. An alternative indicator of consolidation rate has been used, that is percentage
consolidation within the first minute. Allophane-rich samples tend to have a marked
reduction in consolidation rate (c,) with increasing stress, whereas halloysite-rich samples
tend to maintain a high rate, at least up to 1000 Wa. However, in general a t high
effective stresses the settlement versus time plots begin to resemble those of a temperate
clay, that is, having a measurable parabolic primary portion to the curve and a smooth
transition to the secondary stage. Where a sequence of samples with varying depth was
available, it was seen that primary consolidation rate tended to decrease with depth for
a given applied stress.

Compressibility, mV, is found to range from ‘high’ to ‘very low’. There tends to be an
overall decrease in compressibility with increasing applied stress. It is notable that above
applied stresses of about 500 kPa the mv versus log stress curves tend to coalesce in the
case of specimens taken from the same sample, irrespective of voids ratio, whereas they
do not, with one exception, for specimens taken from different depths at the same
location. T h e Compression Index, Cc, is generally ‘moderate’ to ‘high’ (between 0.20
and 2.08), and highest for Indonesian samples and allophane-rich samples. Swelling
 3s

index, CE, which measures the extent of recovery on unloading, ranges from 0.02 to
0.14, and may be considered ‘low‘ to ‘very low‘ compared with temperate clay soils. This
is probably due to non-recoverable strain associated with major densification of the
macro-pores in a ped-type soil structure. It is notable that the high-pressure oedometer
tests, (Indonesian Pits 1, 2.0 m and 5, 2.0 m) which take the soil to much lower voids
ratios than normal tests, give higher CE values than the matching low pressure tests,
indicating a degree of elastic recovery.

Vaughan (1 988) defined ‘relative voids ratio’, em as follows:

where: e = natural voids ratio

eopt = voids ratio a t optimum compaction
et = voids ratio a t liquid limit
A reasonable positive correlation has been obtained between Compression Index, CC,
and relative voids ratio, eR ; the more plastic soils have both higher compressibility and
higher relative voids ratio (Figure 16). T h e problem of finding the ‘correct’ liquid limit
with which to derive eR is discussed in project report WN/93/12.

The consolidation process may be usefully divided into ‘primary’ and ‘secondary’ stages,
the primary representing the expulsion of water from macro-pores and rapid, large-scale
densification due to compression of those voids, and the secondary representing visco-
elastic creep of the solid phase. T h e coefficient of secondary consolidation, C,, ranges
(at a confining pressure of 200 kPa) from 0.004 to 0.052 for Indonesian, and 0.003 to
0.134 for Kenyan samples. Values of C, tend to rise and become more diverse between
about 100 and 200 kPa. Allophane-rich samples’ C, values rise particularly sharply
(Indonesia Pit 7, 2.0 m and Pit 9, 2.0 m). This rise appears to be related to the yield
. point seen on the voids ratio versus log stress plots. There does not appear to be a clear
correlation between C, and depth, or between C, and pre-drying.

A small number of ‘pre-drying’ experiments were carried out as part of the multiple
oedometer tests. Here, one of the five test specimens from an individual sample was pre-
dried in air prior to testing, resulting in reductions in saturation of between 1 1 and 53%.
Some of these specimens collapsed and others swelled, with no clear pattern emerging.

The geotechnical implications of bonding in soils are described by Vaughan (1 988) and
Lerouil and Vaughan (1990). The relationship of an undisturbed, bonded oedometer soil
specimen to its ‘destructured’ state line is shown applied to a test plot example
(Indonesia Pit 10, 5 m) in Figure 17. Such behaviour is seen in over consolidated clays,
cemented soils, weak rocks, and dense sands, and was noted in the oedometer test
results, particularly in the following cases: pre-dried Kenyan specimens, high-pressure
test (Indonesia Pit 5J2.0m) in which a clear decrease in compressibility is seen at about
800 kPa, and normal tests (Indonesian Pits 10, 5.0 m and 1 1, 4.0 m). In some cases the
bonded part of the e - log P curve may be obscured by collapse. Two yield points have
been observed in some triaxial (first stage) test stress-strain curves.
36 SECTION FIVE Mechanical properties and testing procedures

Scanning electron microscopy (SEM) has revealed something of the nature of bonds, and
their ability to sustain a sizeable void network up to effective stresses in the range 50 to
200 kPa. Both the fundamental mineralogy and the structure of these soils are complex.
T h e minerals halloysite and allophane exist together, and the former in more than one
form (see report WG/92/31). These may be tubular, platy, spheroidal, or block. Peds may
consist of agglomerations of clay particles or of partially decomposed parent rock
particles with a coating of gel-like clay. Peds may be connected by ‘bridges’ of clay. _.
Abundant pores in the narrow size range of 0.005 to 0.015 pm have been found in-
helicities with small particles (< c. 0.08 pn in width) but not in helicities with larger
particles (0.1 pm in width). Individually, tubular particles have the combined properties
of high voids, high surface area, and high strength. In abundance these may impart the
same properties to the soil mass. T h e accessibility of tubular particles to water is also of
interest. Is such water ‘bound water’ or ‘free water’?

Results of destructuring experiments suggest that the destructuring process is not simply
one of densification and loss of bonding. It is possible that fresh bonds, perhaps of a
different nature, may be established within a destructured clay soil. T h e precise nature
of the undisturbed bonding, and the effect on it of applied stress, may have an important
influence on mass permeability and saturation. T h e effect of compaction also has
profound influence on the soil structure, possibly in a detrimental way in some
applications. For example, ‘reconstituted’ oedometer test specimens (from Kenya Pit 5,
Indonesia Pits 10 & 12) have higher initial voids ratios than the equivalent undisturbed
specimens. A normalising parameter, the voids index, Ivy was suggested by Burland
(1990) to relate the different fundamental states of a soil to its structure. This is a similar
concept to that of Vaughan (1988) but applies to consolidation rather than compaction
and liquid limit. T h e voids index is defined as follows:

Iv = (e - e,,,) / Cc

where: Cc is the compression index (= e,,, - e,,,,)

e is the initial voids ratio,
e, is the voids ratio at a consolidation stress of n kPa
T h e voids index ranges from 0.03 to 0.60 and tends to decrease with depth. T h e
investigation into the phenomenon of collapse has been notable for the significantly larger
collapses (as measured by the collapse index, CI, in the oedometer consolidation test)
obtained by MT&RD compared with BGS. This is probably due to MT&RD testing
shallower samples (- 1.O m depth) having lower initial degrees of saturation. BGS sample
K89/3F/1 (0.6 m), for example, collapsed in a similar manner to the MT&RD samples.
Kenya Pit 4, 2.0 m and some pre-dried specimens showed negative C I values; that is,
they swelled. It is likely that the net biological disturbance (e.g. shrub rootlets, fungi,
termites) is more significant at shallow depths. Collapse measured in the oedometer test
appeared generally to be independent of flooding but more dependent on stress. A small
number of field collapse tests were carried out in Indonesian pits 10, 11, & 12.
Significant collapse was noted (on flooding at 100 kPa) at a depth of 3 m in Pit 11.

The coefficient of secondary consolidation, C,, characterises the long term consolidation
behaviour o f a soil. A notable rise in the value of C, is typically found beyond stresses of
between 100 to 200 kPa, particularly for allophane-rich soils (for example, Kenya Pit 3,
 37

Indonesia Pits 7 & 9). These stresses match those measured for yield point, which again
lie typically between 100 and 200 kPa.

Throughout the project every attempt has been made to minimise sample disturbance,
recognised as a particular problem with weakly bonded soils (Vaughan, 1988) and soils
with high voids ratios, However, whilst it is impossible to eliminate this problem, neither
has it been possible to quantify sample disturbance in the manner used for temperate
soils (Schmertmann, 1953). T h e absence of bonding inferred from an oedometer test
may indicate either that the bonding has been destroyed by sample disturbance or that
there was no bonding. Every indication is that bonding does exist in these soils, and that
sample disturbance is minimal. De-stressing due to exhumation cannot readily be
mitigated. However, both the trial pitting and drilling procedures have been carefully
considered, executed, and monitored (see Section 3).

COMPACTION TESTS
Compaction tests were carried out on Kenyan and Indonesian samples using both the
Proctor and the Dietert apparatus (Institute of Road Engineering, 1987; Wachira, 1988).
Compared with typical temperate clay soils, values of Maximum Dry Density (MDD)
were low, and values of Optimum Moisture Content (OMC) high. O n a plot of moisture
content versus dry density (Figure 18) the Kenyan Pits 5 and 7 data are closely grouped
with OMC’s of 44 to 48% and MDD’s around 1.2 Mg/m3, whereas samples ES4, 2.0 m,
Pit 3, 1.0 m y and Pit 3, 2.0 m were widely scattered, albeit on a common curve. T h e
sample with the highest MDDAowest O M C was ES4, 2.0 m (Tertiary volcanic origin).
Indonesian data were similar to the Kenyan but without the very high and very low
values. All Indonesian compaction data were from samples of ‘Older Quaternary
Volcanic’ origin. Unfortunately, there were no representatives of ‘Younger Quaternary
Volcanics’ soils. For Indonesian samples values of O M C were found to be between 0 and
10% lower than natural moisture content (NMC) measured in the field, whereas for
Kenyan samples they were between 5% higher and 5% lower. Voids ratio values after
. compaction were used to calculate ‘relative voids ratio’ (Vaughan, 1988).

Problems are frequently experienced in reaching target MDD’s on road construction

projects in Kenya. Considerable shrinkage problems are also experienced if soils are
compacted near their natural moisture content. Longitudinal cracks were frequently
observed parallel with embankment flanks. These were generally absent on adjacent
sections of road in the same soil type in cut. In some cases (for example, the Eldama
Ravine-Nyaru road) drying to below optimum had to be carried out to preclude
shrinkage. A variation of the linear shrinkage test, using a compacted specimen, could
usefully be run in parallel with the compaction test. Typically, OMC’s are between 80
and 105% of the.T$MC’s. In Indonesia, experience of compaction with red clay soil fills
has generally been good, compared with the other soils available. T h e red soils tend to
be compacted slightly wet of O M C despite relatively low MDD’s, and some longitudinal
cracking of embankments has been observed, as in Kenya. It would appear that whilst
reworking of soils in excavation will lead to a reduction in shear strength (Belloni et al.,
1988), this is not as severe as the reduction produced by embankment construction
where reductions in undrained strength of up to ten times are reported (Anon, 1992).
38 SECTION FIVE Mechanical properties and testing procedures

Successful compaction using light compactive effort is reported by Knight et al. (1982)
and Rouse et al. (1986).

Cutting slopes and embankment slopes are often constructed in tropical clay soils at
angles of 30 to 40°, that is, much steeper than an equivalent temperate clay slope. This
means that commonly used stability charts do not apply (Wesley, 1977). In the case of
the cutting slopes the strength is achieved by virtue of the natural bonded structure which
imparts a significant effective cohesion (Vaughan, 1988). This strength may be --
unaffected by transient saturation, a t least within the normal climatic regime, bearing in
mind that the undisturbed soil is highly permeable and will allow effective drainage of the
most severe tropical storm flows. Wesley (1 977) suggested that allophane-rich soils are
fundamentally stronger than halloysite-rich soils, despite their higher moisture content
and plasticity. He attributed this to their fundamentally less-developed weathering state.
The embankment (fill) case is different in that the soil is no longer undisturbed, and has
a new, denser structure possibly containing discontinuities induced by construction plant.
This new structure may be as strong as the undisturbed structure (this is suggested by
the special triaxial test data) but less as the result of bonding. This will be a function of
the extent to which excavation and placement have destroyed the original structure. T h e
compactive effort and method are key factors. T h e destruction may be only on a macro
scale. However, the new structure will certainly be denser and less permeable. It will
probably also be more prone to swelling and shrinkage, and dilatant behaviour when
subject to shear.

RING SHEAR TESTS

A small number of Bromhead ring-shear tests were carried out on Kenyan soils in order
to measure the residual strength of a remoulded sample (Njaibu, 1988). This is a useful
parameter where large-scale strains along a particular slip plane or zone of movement,
take place, either within an initially undisturbed soil profile or within an embankment or
other reworked soil fill material. T h e tests were carried out on soil pastes prepared from
disturbed samples, both at natural moisture content, and pre-(air)dried. Values of
effective residual friction angle, and effective residual cohesion, c i , ranged from
20.1 to 3 1.2" and 0 to 1 1.O kPa, respectively. T h e effect of pre-drying was to increase
slightly and decrease c i , with the sole exception of Pit 7, 2.0 m. Compared with
temperate soils of similar plasticity these values are particularly high. T h e versus PI
relationship does appear to correspond with data given by Boyce (1985). Lupini et al.
(1981) described the influence on residual strength of clay mineral morphology.
Smectites and kaolinite are dominantly plate minerals and provide low resistance to shear
a t high values of strain. O n the other hand, halloysite and allophane are not
characteristically plate and provide a greater resistance to shear at large strains (Anon,
1990). This is suggested by the results of the tests (Figure 19) when compared with
envelopes for typical temperate clays.
 39

DISCUSSION

T h e voids and structure of the tropical red clay soils are key to their undisturbed
mechanical behaviour. This structure, in contrast to that of sedimentary soils, is
dependent on the parent rock and the weathering processes to which it has been
subjected. Voids ratios are generally high and may decrease or increase with depth. Voids
in these residual soils are produced as a result of mineral dissolution, alteration, and
biological disturbance. T h e effects of high voids ratios are seen as high consolidation
rates and amounts, high permeabilities, and some of the properties of a silt. This permits
free drainage and high effective strength above the water table. T h e coefficient of
consolidation is difficult to determine at low and moderate stresses. T h e effects of
geological provenance are also seen in some aspects of mechanical behaviour.

Scanning and transmitting electron microscope investigations (SEM & T E M ) have

shown both tubular and plate halloysite structures, bonding of peds largely by clay
bridges, and a ‘flowing’ geliferous clay structure (allophane ?).Little influence due to
iron compounds has been found. Mineralogical quantification of halloysite and allophane
content has proved elusive. Collapse has been observed in some oedometer test
specimens, particularly in shallow specimens. Collapse appears to be depth (and possibly
method) related, and may in turn, be related to bio-disturbance. Pre-drying oedometer
experiments have proved inconclusive; most specimens having swelled on flooding.
Primary consolidation in the triaxial and oedometer tests is extremely rapid. T h e effects
of bonding have been observed in some triaxial and oedometer tests. Unusual
consolidation behaviour, for apparently saturated specimens, has been noted at low
effective confining stresses in the triaxial test. This appears to involve rapid primary
consolidation, followed by slow swelling accompanied by continuing secondary
consolidation. This appears to be due to incomplete saturation, and reflects the bi-modal
macro- and micro- structure, respectively.

The effect of destructuring and compaction have been investigated and compared with
the undisturbed state for two triaxial specimens. T h e results are surprising inasmuch as
. the destructuring process, whilst reducing voids ratio dramatically has not changed the
strength significantly. Further laboratory testing of this kind is required. There does not
appear to be a clear distinction between allophane-rich and halloysite-rich soils in terms
of strength. Residual strength was high compared with temperate soils of similar
plasticity.
40 SECTION FIVE Mechanical properties and testing procedures

SUMMARY OF RESULTS

Tropical red clay soils have a high voids ratio / low dry density / high mass
permeability; the highest voids ratios are for allophane-rich soils.

0 Tropical red clay soils exhibit a two-phase structure giving clay-like behaviour on
the micro-scale and silt-like behaviour at the macro-scale. -.
0 Destructuring and resedimentation results in a large decrease in voids ratio but
no significant strength increase. T h e effect of bond breakage counters the
densification effect.

Strength
Tropical red clay soils are bonded and have yield points typically a t stresses
between 100 and 200 kPa.

Stress paths may be interpreted in terms of yield.

Large volume changes should be anticipated in consolidated, undrained tests.

The two-phase structure of tropical red clay soils results in incomplete saturation
of micro-pores using conventional Skempton B saturation criteria.

Apparent contemporaneous swelling and consolidation is observed during the low

effective stress consolidation stages of the triaxial test.

Effective strength behaviour is relatively ‘brittle’ with peak strength achieved at

around 2% strain.

Residual shear strength is higher than for temperate clays of similar plasticity.
This may be due to the mainly non-platy clay mineral morphologies of halloysite.

Consolidation
Consolidation rate is extremely high at low stresses. Most primary oedometer
consolidation is completed within 30 seconds. Primary and secondary
consolidation are more easily distinguished a t high stresses.
0 Many tropical red clay soils do not exhibit true collapse, but do exhibit large
volume changes, particularly post-yield with allophane-rich soil. Yield appears
to be independent of collapse.
0 A reasonable positive correlation is found between relative voids ratio and
compression index in the oedometer test. More data are required to more fully
determine the felationship.

Compaction
0 Compaction data show very low maximum dry densities and high optimum
moisture contents, particularly for allophane-rich soils.
RECOMMENDATIONS FOR GOOD TESTING PRACTICE
0 High quality push-driven sampling, combined with suitable preparation
techniques, should be used for triaxial and oedometer test specimens. T h e use of
extrusion should be avoided. The need for trimming should be minimised.
_.
Strength testing
Large specimens should be used for all triaxial tests.

Multi-stage triaxial tests can be used successfully.

Consolidation testing
Multiple oedometer tests may be useful to evaluate the effects of pre-drying and
collapse.
0 A high-speed data logger is required at low stresses.

Compaction testing
0 Separate sub-samples must be used for each stage of the compaction test. This
means that large samples (c. 25 kg) must be collected. Excessive pre-drying
should be avoided.

T o reduce the need for large samples, non-standard methods (for example, the Dietert
test) may be useful.
6 Identification and classification of
tropical red clay soils
The main objective of a good classification scheme is to enable identification, description
and grouping of soils with similar characteristics to be easily undertaken in a systematic
manner, and without recourse to complicated and costly test procedures. Traditional
geotechnical classification schemes have been developed for temperate soils, which are
often little altered sedimentary or transported deposits. For tropical residual soils, the
nature and properties of the soils are directly related to the weathering of rock masses in
situ and are intimately associated with the mineralogy of the parent materials, the nature
of the tropical climate and on drainage conditions. Unlike most temperate soils, the
products of tropical weathering can, under certain conditions, contain minerals with
unusual properties and possess a highly porous structure comprising clusters or ‘peds’
of clay minerals disseminated throughout with finely divided iron oxides. T h e clusters
tend to be weakly ‘bonded’ and have a significant role in controlling engineering
behaviour and in the preparation of samples for geotechnical index testing.

The influence of soil pedogenesis on the engineering properties of tropical soils is critical.
As such, it is logical that it should form the basis of any systematic engineering
classification scheme for these soils. As stated by the Geological Society Working Party
Report on Tropical Residual Soils (Anon., 1990), the importance of recognising the
interaction of the soil-forming factors for engineering classification is that the groups of
secondary minerals which form the basis of the pedogenetic classifications are all
characteristic of distinct geotechnical behaviour. Anon. (1990) proposed a systematic
engineering soil classification scheme based on the pedogenetic system of Duchaufour
(1982). T h e classification includes both ‘black’ and ‘red’ tropical soils and also
concretionary weathering products or duricrusts (for example, ferricrete, alucrete,
gypcrete, silcrete). The ‘black’ soils are those formed from a variety of parent materials
under conditions of impeded drainage and termed ‘vertisols’ in the major pedological soil
classifications and known locally by such names as ‘black cotton soils. ‘Black’ soils and
‘cretes’ were not investigated in this study and are not referred to further. However, it
should be recognised that such soils can be intimately associated with tropical red soils
under appropriate local topographic and drainage conditions. Based on Duchaufour’s
scheme, the classification of Anon. (1 990) consider tropical red clays to comprise those
contained within ferruginous, ferrisol and ferrallitic profiles. In the present study,
andosols were considered as forming a distinct type of so called ‘red’ tropical soil. This
was because soils designated as andosols in Kenya (Kenya Soil Survey, 1982), were
distinctly reddish in colour over large areas of that country, grading transitionally into
the more characteristic brown andosol soils at higher elevations.
T h e classification of Anon. (1 990) requires recognition of geological, climatic,
topographic and soil forming factors, and the role and limitations of geotechnical index
tests for the geotechnical evaluation of tropical residual soils. T h e comments given below
are directed to the use of geological and pedological maps as sources of ‘environmental’
information and the role of geotechnical index properties for the engineering
classification of tropical red clay soils. T h e comments are made with the aim of
 43

amplifying the classification scheme of Anon. (1990) with respect to tropical red clays
developed from the weathering of igneous rocks under well-drained conditions.

GEOLOGICAL MAPS
Residual red clay soils in the tropics and sub-tropics can occur over all types of bedto-&
However, the unique properties of the soil at any particular location will have developed
in response not only to the underlying geological parent material, but also to the
combined effects of other environmental factors such as climate (especially rainfall and
temperature regimes), water movement (drainage conditions), topography, vegetation
and age of the land surface (how long the soil has been forming). When interpreted with
respect to these other factors, geological maps can provide information very useful to the
preliminary engineering assessment of the residual soils developed over the mapped
parent materials.

For example, geological information can give an initial indication of the general type and
form of the weathered soil profile which may develop over different types of bedrock
under tropical weathering conditions. Townsend (1 985) and Wesley (1 988) both
commented on the different weathering profiles which can form over acid igneous and
metamorphic rocks (granite, gneiss and phyllites) and over basic igneous rocks (basalts,
pyroclastics). In the former case, the rocks are composed dominantly of quartz and
feldspars. T h e quartz is very resistant to weathering and has an important role in
influencing the texture of the secondary products by remaining as quartz particles. T h e
zone of alteration in these quartz-rich rocks is usually quite thick, forming a sequence of
successively altered weathering zones ranging from completely weathered residual soil
through to successively less altered rock with depth in the profile (Figure 20, a & b). In
basic igneous rocks, the rock minerals (olivines, pyroxenes and calcic plagioclases)
weather rapidly into soils which often provide an abrupt contact with the parent rock,
with only a very thin zone of transition material (Figure 20c). In layered or interbedded
sequences of both igneous and sedimentary rocks (such as multiple basic lava, tuff,
pyroclastic flows or quartz-rich sandstones interbedded with clays) the soil profiles are
likely to be more complicated, reflecting both the weathering sequence and the
differences in the parent rock (Figure 20d). T h e presence of faults, also identified from
geological maps, indicate areas where tectonic movements have displaced adjacent blocks
of strata. These are zones of weakened, disturbed or broken rock which enable deep
penetration of downward percolating waters and enhance weathering and alteration of
the rock mass. Therefore, fault zones are likely to be characterised by thicker weathering
profiles than adjacent strata.

With regard to the type of residual red soil formed over various types of parent rock
under well-drained conditions, the sesquioxides of iron and aluminium can occur due to
the weathering of practically all igneous rocks and kaolinite clays are the eventual end
product, with clay minerals of the smectite group being a precursor in the weathering of
basic rocks. Pyroclastic and other extrusive volcanic rocks can also produce allophanes,
helicities and metahalloysites in addition to kaolinite, particularly under cooler climatic
conditions encountered at higher altitudes. T h e type and proportions of these various
minerals at any location (which impart to the soil its particular engineering
characteristics) will depend on a number of environmental factors. For this reason the
44 SECTION SIX Identification and classification o f tropical red clay soils

information from geological maps should be supplemented by an assessment of the

environmental influences on the weathering process, such as site elevation, slope angles,
morphological history and drainage conditions (for a given site with similar bedrock and
climatic conditions, impeded drainage will tend to result in the formation of smectitic
clays, such as ‘black cotton soils’, whereas free draining conditions result in ‘red clay
soils’).
_.
PEDOLOGICAL MAPS

Pedological, or soil, maps are a very useful means by which tropical soils of different
types can be differentiated. However, because different classification schemes are used
in different countries, care is needed in comparing like soils identified under different
schemes. In addition, the quality and detail shown on pedological maps varies greatly.
However, once understood, they can provide the geotechnical engineer with a sound
basis by which different types of residual soils can be distinguished in broad terms and
provide an initial indication of soil types likely to show similar engineering properties.

Three of the most widely used pedological classifications systems are the French
(Duchaufour, 1982), the American Soil Taxonomy (Soil Survey Staff, 1975) and the
FAO-UNESCO (1969, 1988) schemes. The FAO-UNESCO system is used as the basis
for the Soil Map of the World, and, as such, is the most universal of the major
classifications. However, it is a system developed to overcome gaps in national
classification systems in order to accommodate world soils, and although internationally
accepted, it is to some extent a compromise. Many of the diagnostic concepts of the
FAO-UNESCO system have been derived from the U S ‘Soil Taxonomy’ system, but
each rely on subtle changes in the soil profile to distinguish soil groups and, to the lay
person, employ complex terminology. The French system of Duchaufour, based on
phases of the weathering process and the resulting mineral composition of the soil, is
perhaps most readily understood by the engineering geologist or geotechnical engineer.
It is for this reason that the Geological Society Working Party (Anon., 1990) in preparing
a guide to the classification and behaviour of tropical residual soils, based its approach
on the Duchaufour scheme. T h e use of this scheme also helps to clarify the complex
terminology applied to tropical residual soils, which has arisen by the unsystematic use
of soil science names applied to particular soil groups by engineers and other non-soil
scientists. In particular, there is considerable confusion in the engineering literature over
the use of terms for ‘red tropical soils’. Only a small proportion of sesquioxides will
impart a red colouration, but the findings of this study have shown that variable amounts
of these iron and aluminium oxides and different associations of clay minerals, will result
in a wide variation of geotechnical properties. Therefore it is important to distinguish
between different types of red soils. In the Duchaufour scheme ‘red tropical soils’ are
identified as those contained within the ferruginous, frrrisol and ferralliric profiles.
However, ‘andosol’ soils, although generally brown, can also have a red colouration and
are often closely associated with other red clays in field situations (for example, in
Kenya). For this reason, andosols, were considered as part of the spectrum of ‘tropical
red clays’ investigated in the present study. In the Duchaufour system, these comprise
those soils termed ‘fersiallitic andosols’. The ‘red soil’ terminology adopted by Anon
(1990) is recommended but, as in this study, may be sensibly expanded in countries such
 45

as Kenya, on the basis of soil colour, to include fersialliric andosols . Summary

characteristics of the fersiallitic andosol soils, ferruginous, ferrisol and ferrallitic red soil
types, which represent increasingly intensive phases of the tropical weathering cycle, are
as follows:
Fersiallitic soils. Fersiallitic soils (sensu smcto) are formed mainly in sub-tropical or
Mediterranean-type climates with mean temperatures of 13 to 20" C, rainfall 500-'to
1000 mm, and a hot dry season; tropical sub-types are also known. 2:l clays are
dominant and the main new clay mineral formed is smectite, especially where drainage
is impeded, but kaolinite may form on older well-drained land surfaces and silica-poor
parent materials, such as basalt. O n recent volcanic rocks, particularly ashes, tuffs and
pyroclastic material fersiallitic andosols are formed under perennially wet and cooler
climatic conditions found at higher altitudes. These immature soils are generally brown
(but may be reddish brown or brownish red) in colour, highly porous and of low bulk
density, and contain amorphous, or imperfectly crystallized, clay minerals (allophanes)
produced by the rapid weathering of volcanic glass. As silica is lost during the
development of andosols, allophanes are replaced by disordered fibrous clay minerals
(imogolites) and eventually by the 1: 1 clay mineral, halloysite and possibly kaolinite
(Mohr e t al., 1972). Andosols may have a high content of organic matter due to the
formation of allophane-humus complexes. They have enormous water-holding capacity
which can be reduced by prolonged desiccation, often irreversibly. They have high cation
exchange capacity, are very clay and iron rich and indurate on drying. Quantin (1985)
described andosols from Vanuatu which have developed in less than 5000 years, whilst
Yamada (1977) reported that these soils can be developed in between 500 and 1500
years, depending on site conditions.

Ferruginous soils. These soils are formed in climatic zones which are either more
humid (humid sub-tropical, without a dry season) or slightly hotter (tropical with dry
season) than the Mediterranean-type areas. They are more strongly weathered than
fersiallitic soils, but orthoclase and muscovite typically remain incompletely altered.
Kaolinite is the dominant 1: 1 clay mineral formed and gibbsite is usually absent; 2: 1 clay
minerals (for example, smectites) are subordinate and tend to occur in clay-rich horizons
due to preferential lessivage. On older land surfaces and more permeable base-rich parent
rocks, soils transitional between ferruginous and ferrallitic soils, known as ferrisols, may
be formed. These are more deeply weathered and tend to have thicker profiles (often
greater than 3 m) than typical ferruginous soils but the weathering of primary minerals
is not complete. In very humid zones with ferrallitisation dominant, they tend to occur
at higher altitudes (mountain slopes) where lower temperatures slow down
ferrallitisation, or where erosion on steep slopes rejuvenates the soil profiles and prevents
their complete (ferrallitic) development.

Ferrallitic soils. These soils are formed in the hot humid tropics where annual rainfalls
exceed 1500 m m and mean temperatures are above 25" C, with little or no dry season.
They represent the final phase of soil development in a hot, humid climate and the soil
profiles may be several metres thick. All primary minerals except quartz are altered by
hydrolysis in near-neutral or slightly acid conditions, and much of the silica and bases are
removed in solution. Remaining silica combines with alumina to form kaolinite, but with
insufficient silica, excess alumina results in the formation of gibbsite. Most ferrallitic soils
46 SECTION SIX Identification and classification of tropical red clay soils

probably take 10 000 or more years to form, with development being more rapid on
silica-poor rocks such as basalt than on silica-rich rocks like granite or gneiss. Increased
silica content of the parent rock is usually reflected by kaolinite formation in the subsoil.
This zone of kaolinite formation is often poorly drained and mottled with white, red and
ochreous patches (plinthite) and can harden irreversibly on exposure; it may be overlain
by a ‘lateritic’ horizon enriched with iron mobilised from acidic near-surface (and
upslope) horizons or by fluctuations in the water table. T h e formation of kaolinite is --
encouraged by the poor drainage, whereas free drainage removes dissolved silica and
favours the development of gibbsite. T h e iron-enriched horizon may be moderately or
stxongly indurated (hardened), pisolitic from the welding of concretions or vesicular from
precipitation of iron in a polyhedral network of fissures and subsequent removal of softer
interstitial material. O n silica-poor basic rocks on well-drained slopes, these soils are
generally less deep with little or ‘slowed down’ neoformation of kaolinite. However, this
clay is stable and tends not to be degraded, as it is bound within aggregates ‘cemented
or ‘bonded’ by iron oxides. This results in the formation of a friable, aggregate soil
structure of high porosity, which does not harden irreversibly on exposure. Concretions
are usually absent but when present are usually soft.

From the foregoing, it is clear that soil maps, when available, can provide a guide to the
occurrence and distribution of tropical soils of use to the engineer, provided the general
process of weathering leading to soil composition and fabric in the tropics is appreciated,
and the basis of the soil classifications employed on the various maps is readily
understood. However, the limitations of such maps must also be recognised. For
example, Buurman (1980) stated that, especially on volcanic rocks, there is considerable
doubt about soil classification because of textural differences arising from stratification
of the parent material. The divisions between different soil types as shown on pedological
maps are rarely, if ever, sharply defined in field situations. Local variations of weathering
grade and soil type, both areally and with depth, are not possible to present at regional
mapping scales, and may make preliminary interpretations for engineering use somewhat
tentative. Soil maps do, however, provide a valuable indication of the general occurrence
of broad soil types. For example, in this study, the pedological soil maps available in
Kenya and Indonesia enabled rapid identification of those areas where fersiallitic andosol
and ferrallitic soil development dominated. In the scheme of Duchaufour (1 982), these
soil types represent the ‘end members’ of sesquioxide-rich (red) soil development in hot
tropical and sub-tropical climates, the ferrallitic soils developing from long periods of
in tense weathering, and the andosol soils under less intensive weathering conditions
which have inhibited completion of the ferrallitisation process. T h e marked differences
in the mineralogy and structure of these soil types are reflected in their engineering index
properties. Ferruginous soils represent an intermediate stage of soil development between
that of fersiallitic andosols and ferrallitic soils, and as such may be expected to show a
transition in mineralogy and index properties between these soil types, that is:

fersiallitic andosols --* ferruginous soils ( --. ferrisols) --. ferrallitic soils

Because of the important effect of climatic control (rainfall and temperature), the general
relationship between the broad development of these different soil types with decreasing
elevation in Kenya and Indonesia was fairly clear. However, the boundaries between the
main soil types were not clear cut in the field, and the gradual transition of one into the
 47

other was not clearly represented on the soil maps. Local topographic, climatic, and
erosional effects also give rise to different stages of soil development within each of the
broad soil zones. These local and broad transitional variations in soil types make it
difficult to accurately predict engineering properties at the site specific level. However,
they can provide a useful initial means of assessing the broad distributions of soils likely
_-
to show distinct or similar engineering properties (for example the presence of allophanic
andosol soils).

Not all pedological classification schemes follow that of Duchaufour (1982) and
correlation between them and the French system can be difficult. Approximate
correlations between the Duchaufour scheme and the FAO-UNESCO and American Soil
Taxonomy systems are shown in Table 4. Approximate correlations between the FAO-
UNESCO and French (Duchaufour) schemes for soils obtained and investigated in this
study from sample sites in Kenya, Indonesia, Fiji and Dominica are given in Table 1,
along with related terrain and climatic details.

ENGINEERING INDEX PROPERTIES

T h e usefulness of conventional index tests such as plasticity limits and particle size for
the identification and engineering classification of tropical residual soils has been
questioned on a number of occasions. This is primarily because of the dependence of
such tests on sample preparation (which has led to inconsistent test results) and the
destruction of the natural soil structure which occurs during preparation (with the
implication that index tests give no indication of the properties of the undisturbed soil
in situ). T h e results from this study have shown that, with simple modifications to
preparation and test procedures, meaningful and repeatable index property
measurements can be achieved. As also found by Wesley (1988)) the study has shown
that index properties are very useful for identifying and ‘grouping’ tropical soils likely to
possess distinctive or similar engineering properties. For example, the position which a
soil occupies on the conventional plasticity (or A-line) chart was found to be particularly
useful in distinguishing different types of red clay soils (Figure 7b). However, problems
can arise when attempts are made to directly relate other engineering properties to index
test measurements for these soils in different structural states.

A primary purpose of an engineering classification system based on simple index tests is

to rapidly identify and group soils into ‘classes’ with similar index properties. It can then
be reasonably inferred that, under similar conditions, the engineering behaviour of the
soils in each class will be similar. In effect, once meaningful and repeatable liquid and
plastic limit data are obtained, the position of the soils on the plasticity chart identify the
‘footprint’ of those soils likely to show similar behaviour in similar undisturbed or
‘destructured’ states under similar conditions of moisture content.

The grouping of soils on the plasticity chart reflect the soil mineralogies which, in turn,
are dependent on the degree of soil pedogenesis. Therefore, the plasticity chart can be
used to classify the pedological soil divisions of Duchaufour in more quantitative
engineering terms. This is illustrated in Figure 21, which shows the position of the
fersiallitic andosol, ferrisol and ferrallitic soils investigated in this study when plotted on
the plasticity chart. T h e soils are also distinguished in terms of their development over
48 SECTION SIX Identificationand classification of tropical red clay soils

different types of bedrock. Vargas (1 985) showed that classification of tropical clays by
means of the A-line plasticity chart can sometimes be enhanced by incorporating a plot
of ‘activity’. Activity (defined as the ratio of plasticity index to percentage clay fraction,
% < 2 p ) can further indicate the nature of the dominant type of clay minerals present
in the soil. For example, smectites are ‘high activity’ clays (activity >1.25) whilst
kaolinite-type clay minerals are ‘low activity’ clays (activity < 0.75). This combined plot
(also shown in Figure 21) would appear to be quite useful for distinguishing red clays _-
from other tropical soils (for example, smectite-rich vertisols or ‘black cotton soils’, and
a variety of saprolitic soils), but is perhaps of more limited use in distinguishing between
different types of red clay soils, particularly those developed over igneous rocks (virtually
all of which, in this study, possessed activities < 0.75). Of the very few data points
plotting in the range of ‘high activity’ clays (>1.25), all are andosol soils. However,
because of the difficulty in obtaining meaningful clay contents of these soils from particle
size distribution tests, the activity values for these soils are unreliable.

The significant influence of parent igneous rock on the red soil groupings is clearly shown
in Figures 7 b and 21. Similar relationships between soil engineering properties and
parent rock type have been reported elsewhere (by Gidigasu (1976) for red soils
developed over phyllites and granites in Ghana, and by Tuncer and Lohnes (1 977) for
red soils from Hawaii and Puerto Rico). This has led to a number of descriptive or
classification methods being developed for local use in particular formations. However,
as noted earlier, particular types of red soils develop not only in response to parental
bedrock, but also to other environmental factors. Because of this, locally developed
classifications may not be applicable to soils developed in other terrains and under
different intensities of weathering.

Soils formed from similar parental rock under similar conditions of weathering can be
expected to have similar index properties, and this study has shown that the plasticity
chart is an effective means by which these similar groups of soils can be distinguished.
Comparison of soils on the basis of index data obtained for the complex spectrum of
tropical clay soils worldwide, developed over a variety of rock types under different
climatic conditions, can only be achieved if consistent standardised procedures of sample
preparation, such as those recommended elsewhere in this report. In practice, the
plasticity chart should provide an effective means for identifying the ‘footprint’ of soils
likely to show similar behaviour for local use during engineering projects. It could also
be effective in distinguishing tropical red clay soils developed in a variety of terrains at
a country-specific level.

INDEXING THE MECHANICAL BEHAVIOUR OF TROPICAL RED CLAY

SOILS
Having established the ‘footprints’ of similar red soil groups from the plasticity chart, the
question remains as to how this data can be used to assess their in situ mechanical
behaviour. For the tropical red clay soils considered in this study, in situ behaviour will
be controlled by both their mineralogy and structure (that is, the aggregated ped-like soil
fabric) and the ‘bond’ strength between aggregated peds. The mineralogy can be
indicated by plasticity data. Therefore, when grouped or classed together on the plasticity
chart, it is reasonable to expect that soils with the same in situ structural state will behave
 49

as other soils within each particular group. Soil structure can be indicated by in situ
densities, or voids ratios, which will vary according to how much the soil is destructured.

Vaughan et al. (1988) proposed that the in situ mechanical behaviour of residual soils can
be usefully assessed empirically by indexing the in situ voids ratio (e) of the soil in
relation to the voids ratio of the soil at two ‘standard’ de-structured states, in a manner
similar to the indexing of sands by relative density or of sedimentary clays by liquidity
index. It is suggested that the voids ratio at the liquid limit (eL) is used as the higher of
the two standard voids ratios and the standard lower voids ratio is measured at the
optimum dry density (e,,,), as determined from the proctor compaction test. T h e plastic
limit, as currently measured, is not suitable for establishing the second lower standard
voids ratio. Vaughan et al. (1988) defined the relationship between these voids ratios at
the in situ and ‘standard’ destructured soil states as the relative voids ratio (eR ),
equivalent to the liquidity index for sedimentary soils, as follows:

The use of this index requires mal in the field to determine fully its practical applications.
T o o few compaction tests (to determine eopt) were undertaken in this study to allow
detailed conclusions of the effectiveness of the above relationship to ‘index’ the
mechanical properties of undisturbed red soils investigated here. However, for the limited
results obtained (albeit using the small-scale Dietert apparatus and not the standard
Proctor compaction method specified by Vaughan, et al.), there appeared to be a good
relationship between eK and ‘compression index’ determined from oedometer
consolidation tests. The ‘relative voids ratio’ appears to be a promising method for
engineering correlation of tropical residual soils in their undisturbed state. Moreover, its
basis is not related to stress history which is irrelevant to residual soil behaviour. It also
provides a link between easily determined index properties and undisturbed soil
behaviour.
REFERENCES

American Society for Testing and Materials (ASTM). 1990. Annual Book ofASTM
Standards, Vol. 04.08, Soil and Rock; Building Stones. (Philadelphia: ASTM).
Anon. 1990. Geological Society Engmeering Group Working Party Report: Tropical
Residual Soils. QuarterlyJournal offigineenng Geology, Vol. 23, No. 1.
Baldi, G, Hight, D W, and Thomas, G E. 1988. A re-evaluation of conventional maxial
testing methods. 219-262 in Advanced Tnaxal Testing ofsoil and Rock. STP 977.
Donaghe, R T, Chaney, R C, and Silver, M L (editors). (Philadelphia: American _.
Society for Testing Materials).
Belloni, L, Morris, D , Bellingeri, G A and Purwoko, I. 1988. Proceedings ofthe2nd
International Conkrence on Geomechanics in Tmpical Soils, Singapore, December I 988.
Vol. 1, 343-349. (Rotterdam: A A Balkema)
Black, K D and Lee, K L. 1973. Saturating laboratory samples by back pressure. Journal of
the Soil Mechanics & Foundation Division. Amenkan Society of Civil Engineers. Vol. 99,
Pt SM1, 75-93.
Blight, G E. 1991. Tropical processes causing rapid geological change. 485-490 in
Quaternary Engincenng Geology. Engineering Gcology Special Publication No. 7.
Forster, A, Culshaw, M G, Cripps, J C, Little, J A, and Moon, C F (editors).
(London: Geological Society).
Bohac, J and Feda, J. 1992. Membrane penetration in maxial tests. Technical Note.
Geotechnical TestingJournal, GTJODJ, Vol. 15, No. 3, 288-294.
Boyce, J R. 1985. Some observations on the residual strength of tropical soils. Proceedings
of the 1st International Conference on Geomechanics of Tmpical Soils, Brasilia, Fcbruary,
1985. Vol. 1, 229-237. (Brazilian Society for Soil Mechanics).
British Standard 1377. 1990. British standard methods of test for soils for civil engincenng
puqoses. (London: British Standards Institution).
Buchanan, F. 1807. A journeyh-om Madras through the countries ofMysorc, Canara and
Malabar[in 1800-18011. Vol. 3. (London: The East India Company).
Burland, J B . 1990. O n compressibility and shear strength of natural clays. Gcotcchnique,
Vol. 40, NO. 3, 329-378.
Buurman, P. 1980. Red soils in Indonesia, a state of knowledge. 1-12 in Red Soils In
Indoncsia. Buurman, P (editor). Agricultural Research Reports 889, Bulletin No. 5.
(Bogor: Soil Research Institute).
Duchaufour P. 1982. Pedologv, Pedogenesis and Classification. (English Edition
Translation: Paton T R). (London: George Allen & Unwin).
Dudal, R and Soepraptohardja, M. 1957. Soil classificationin Indonesia. General
Contributions of the Agricultural Research Station, Bogor, No. 148, 3-1 6.
Dumbleton, M J and Newill, D. 1962. A study of the properties of 19 tropical clay soils and
the relation of these properties with the mineralogical constitution of these soils. Bn-tish
Road Kcsearch Laboratory, Note 44, 14pp. (unpublished).
Food and Agricultural Organisation of the United Nations. 1969. Dcfinitions ofsoil
units for the soil map of the world. I : 5 000 000. Soil map of the world. FAO-UNESCO
Project. (Rome: F.A.O.)
Map Shcct II: North America (2 sheets): 1972
Map Shcct III; Mexico and Central America: 1972
MapShect IV. * . - South America (2 sheets): 1971
Map Shcct .W: Africa (3sheets): 1973
Map Sheet WI: South Asia (2 sheets): 1974
Map Sheet WII: North and Central Asia (3 sheets): 1977
Map Sheet IX: Southeast Asia: 1976
Map Shcct X: Australasia (2 sheets): 1976
- . 1988. Soil map of the world, revi.wd legend, 1:5 000 000. (Rome: F.A.O.)
Fredlund, D G and Rahardjo, H. 1993. Soil Mechanics forunsamrated soil.9. (John Wiley
and Sons).
Gidigasu, M D. 1976. Laterite soil engineering, Pedogenesis and engineering principles.
Developments in geotechnical engineenng, No. 9. (Amsterdam: Elsevier).
Harison, J A. 1988. Using the cone penetrometer for the determination of the plastic limit of
soils. Geotechnique, Vol. 38, No. 3, 433-438.
Head, K H. 1986. Manual ofsoil laboratory testing. Volume 3: Effectivestress tests.
(London: Pentech press).
_.
- . 1992. Manual of soil laboratory testing. Volume I : Soil classificationand compaction
tests. 2nd edition. (London: Pentech press).
Hvorslev, M J. 1949. Subsurface exploration and sampling of soils for civil engineering
purposes. (Vicksburg, Mississippi: Waterways Experiment Station).
Institute of Road Engineering. 1987. Laboratory test results: study on red clay soils in W.
Java.
Kenya Soil Survey. 1982. Exploratory soil map and agro-climatic zone map of Kenya 1980,
Scale 1:1 000 000.(by Soebroek, W G, Braun, H M H, and van der Pouw, B J A).
Exploratozy Soil Survey ReportNo. E l . (Nairobi: Kenya Soil Survey, Ministry of
Agriculture-National Agricultural Laboratories) .
Knight, D J. 1986. Geotechnical properties and behaviour of the Monasavu halloysite clay,
Fiji. CIayMnerals, Vol. 21, 31 1-332.
Lang, D K. 1967. Soil and land use surveys, No. 21, Dominica. Regional Research Centre,
University of the West Indies, Trinidad.
Lerouil, S, Tavenas, F, Larochelle, P, and Tremblay, M. 1988. Influence of filter paper
and leakage on maxial testing. 189-201 in Advanced Triiaxial Testing ofsoil and Rock.
STP 977. Donaghe, R T, Chaney, R C, and Silver, M L (editors). (Philadelphia:
American Society for Testing Materials).
Lerouil, S, and Vaughan, P R. 1990. T h e general and congruent effects of structure in
natural soils and weak rocks. Geotechnique,Vol. 40, 467-488.
Lupini, J F, Skinner, A E, and Vaughan, P R. 1981. T h e drained residual strength of
cohesive soils. Geotechnique,Vol. 31, No. 2, 181-213.
Maeda, T, Takenaka, H, and Warkentin, B P. 1977. Physical properties of allophane
soils. Advances in Agronomy, 29, 229-264, in Andosols, Kim H T a n (editor), Benchmark
Papers in Soil Science Senes, Vol. 4, 1984 (New York: Van Nostrand Reinhold).
Martini, J A. 1976. Journal of the Soil Science Society o f h e n c a . Vol. 40, Pt 6, 895-900.
Mohr, E C J, Van Baren, and Van Schuylenborgh, J. 1972. Andosols. Tropical Soils:
a comprehensive study of their genesis, 3rd ed., Mouton-Ichtiar Baru-Van Houve, T h e
Hague, T h e Netherlands, 1972. 397-418 in Andosols, K H T a n (editor), Benchmark
papets in soil science, Vol. 4, 1984 (New York: Van Nostrand Reinhold).
Morin, W J. 1982. Characteristics of tropical red residual soils. 172-198 in Proceeding,s of
the ASCE GeotechnicalEngineering Division Speciaf r y Confirence on Engineering and
Construction in 7hpical Soils, Honolulu, Hawaii.
- and Todor, P C. 1975. Laterite and lateritic soils and other problem soils of the tropics.
Engineenng Evaluation and High way Design Smdx USAID 3682. Lyon Associates,
Baltimore, Maryland, USA.
Newill, D. 1961a. An investigation of the linear shrinkage test as applied to tropical soils in
relation to plasticity. British Road Research Laboratory, Note 4106, 6 pp (unpublished).
- .No.4,
1961b. A laboratory investigation of two red clays from Kenya.
302-318.
Geotechnique, Vol. 11,

Njaibu, M W. 1988. Residual strength characteristics and geotechnical index properties of

some tropical red clay soils from Kenya. Unpublished MSc thesis, University of
Birmingham, UK.
Olsen, H W, Nichols, R W, and Rice, T L. 1985. Low gradient permeability measurements
in a triaxial system. Gcotrchniqoe, Vol. 3 5 , Pt 2, 145-157.
Quantin, P. 1985. Characteristics of the Vanuatu andosols. 99-105 in VolcanicSoils:
Weathenng and Landscape Relationships of Soils on Tephra and Basalt. E Fernandez
Caldas and Dan H Yaalon (editors), Catena Supplement 7, Selecredpapers ofthe
Congress0 Intemacional de Suelos Volcanicos, La Laguna, Tenen2, Canary Islands,
Spain, July 1984.
Rouse, W C, Reading, A J, and Walsh, R P D . 1986. Volcanic soil propemes in Dominica,
West Indies. Engineenng Geology, Vol. 23, 1-28.
Schmertman, J H. 1953. Estimating the true consolidation behaviour of clay from
laboratory test results. Proceedings of the American Sociey of Civil Engineers, Vol. 79,
Separate 3 11, 26pp.
_.
Schofield, A and Wroth, P. 1968. Cn’u’calstate soilmechanics. 310pp. (London:
McGraw-Hill) .
Soil Survey of Indonesia. 1966. Soil map oftheprovince of WestJava, 1:250 000. Soil
Survey Staff, Bogor.
Soil Survey Staff.1975. Soil taxonomy: a basic system ofsoil classificationfor making and
interpreting soil surveys. U.S.Department o f Agriculture Soil Conservation Service.
(London: Castle House Publications).
Townsend, F C. 1985. Geotechnical characteristics of residual soils. Journal of
Geotechnical Engineenng, Vol. 1 1 1, No. 1, 77-94.
Tuncer, E R, and Lohnes, R A. 1977. An engineering classification for certain basalt-
derived lateritic soils. Engineenng Geology, Vol. 11, N o . 4, 319-339.
Uehara, G. 1982. Soil Science for the tropics. 13-26 in ASCE GeotechnicalEhgineenng
Specialty Confernce on Engineering and Construction in Tmpical and Residual Soils,
Honolulu, Hawaii, 1982.
Vargas, M. 1973. Structurally unstable soils in southern Brazil. Proceeding ofthe 8/21
International Conference?? Soil Mcchanics & Foundation Enginenng, MOSCOW, Vol. 2.2,
239-246.
- . 1985. T h e concept of tropical soils. Proceedings of the 1 s t International Confetence on
Geomcchanics in Tropical Laten’tic and Saprolitic Soils, Brasilia, Vol. 3, 10 1-1 34.
- . 1988. Characterisation, identification and classification of tropical soils. Ihceedings of
the 2nd International Confetenceon Geomechanics in Tropical Soils, Singapore, Vol. 2,
71-75.
Vaughan, P R. 1988. Characterising the mechanical properties of in-situ residual soil.
I’roceedings of the 2nd International Confirence on Geomechanics in Tropical Soils,
Singapore, December 1988. Vol. 2, 469-487. (Rotterdam: A A Balkema).
- , Maccarini, M, and Mokhtar, S M. 1988. Indexing the engineering properties of
residual soil. QuarterlyJournal of Engineenng Geology, Vol. 2 1, No. 1, 69-84.
Wachira, J N. 1988. A laboratory investigation of tropical red clay soil from Kenya, with
reference to cracking. Unpublished MSc thesis, University of Birmingham, UK.
Wesley, L D. 1973. Some basic engineering properties of halloysite and allophane clays in
Java, Indonesia. Geotechnique, Vol. 23, Pt. 4, 471-494.
- . 1977. Shear strength properties of halloysite and allophane clays in Java, Indonesia.
Geotechnique, Vol. 27, Pt. 2, 125-136.
- 1988. Engineering classification of residual soils. Proceedings of the 2nd International
Confirence on Geomechanics in Tropical Soils, Singapore, December 1988.Vol. 1,
77-84. (Rotterdam: A A Balkema)
Yamada, S . 1977. Distribution and morphology of soils derived from volcanic ash in Japan.
1-9 in Soils Dcn’ved From VofcanicAsh in Japan. Ishizuka, Y and Black, C A (editors).
Centro Unternacional de Mejoramiento de Maiz y Trigo, Mexico.
a
a
a
a
a
a
a
a
a
0
a
a
0
0
a
Figures
a
a
a
0
a
a
a
a
0
a
0
a
a
a
a
a
0
0
0
0
0
0
e
MandreI
0
0
0
Pushrod
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
Figure 2. Schematic representation of the tripod sampler
0
0
0
Sampler Head
0
0
0
0 Coupiing
0
0
0
0
0
0
0
0
0 Outer Tube
0
0
0 Liner Tube
0
0
0
0
0
0
e
0
0
Cutting Shoe
0
0
0
0 Figure 3. Sampler for tropical red clay soil - 0.5 m capacity (shown 1/3 scale)
0
0
0
drive head
p.’
/ pulley below
mast top
0 of rig\
0 \ /
e --
0 elastic shock cord
--I-
e
, - non-stretch
0 /’
cord
0
0 load cell - -
~ I
I 1
compression
0 transducer
a
slotted tube
e - ~-

0
inner rod ..___-
0 assembly

/
e drive
transducer
0 cord anchored
to drill rig
e rack bar
casing
0 ,,’tube

a I
l i
a t
0
0
0
0
0
0
0
e
0
0 Figure 4. Schematic drawing of the sampler test instrumentation
0
0
0 0

0
0 0
h

0
5:
0
a 0
(3

0 0
T-
U
C
0 0, (0
cn
0
0
0
e -
c
i7j
e 8
a 0
IC

a h
m
W

0
0 0
cc)

a zU
a C
8
a
0
0
a n
U 0

a
a 0
r-
a
0 5:
e 0
(3

m 0
r
a 0,
U
C
lu
cn
0
0
0
Skempton's 6 vs. Back pressure (triaxial saturation) KENYA
1 .00

0 0.90

0.80

0.70

0 0.60

m 0.50

0.40

0.30

0.20

0.10

0.00-I
0 200 400 600 800 1000
BPr. (kPa)

Skempton's B vs. Back pressure (triaxial saturation) INDONESIA

1 .00

0.90

0.80

0.70

0.60

m 0.50

0.40

0.30

0 0.20

0.10

0.00
0 100 200 300 400 500 600 700 800 900 loo0
BPr. (kPa)

0 Figure 9. Progress of saturation of triaxial specimens

0
0
0
0
0 INDONESIA PIT 11,4.0m (Consolidation 2OkPa)
0 0.01 0.10 1 .00 10.00 100.00 1000.00 1oooo.00

0
0
0
0
- - - -Destr.
0 - i - Compac.
0
0
0
0 Time (rnins.)

0
KENYA K89/3F, 0.8m (Consolidation 10 20 kPa) -
0 3000-

0 1000 - -

0 -,ami
p- 0.1 e 1 10
d
ia
\
0
+Undist.
0 (--t-.Deotr.l
1 - - Compac. I
0 a
v)
-7000
4

0
0
-9000 t \
-1 loo0
0
0
0
0
0
0 Figure 1 1. Triaxial consolidation kwell' effect
0
0
0
0
0
Effective cohesion KENYA Effective friction angle, $', KENYA
0
0
0
0
0
0
0
0
0
0 10 20 30 40 50
0
C (kPa) (I' (degrees)
0
0
0
0
e Effective cohesion INDONESIA Effective friction angle, $', INDONESIA
0 e;zzd BGS (1.7 - 3.8m depth) -
E Z Z l BGS (1.7 3.8m depth)

e 0 IRE (1 .Om depth) 0 IRE (1.0m depth)

0 8 Counts

0 1
7 1
0
0
e 4-

3-
0
2-
a
1-
0
0 0 10 20 30 40 50 60 70 80 90 100 0 10 20 3 0 4 0 50

+' (degrees)
0 C (kPa)

a
0
Fimre 12. Histoprams nf pffertivp rnhpqinn nnrl nnule nf frirtinn
0
0
0
0
0
0
0
0
Maximum shear stress vs. Effective average stress (triaxiai) for Kenya
0 K89/3F/1 B,1 .Om & Indonesia Pit 11, 4.0m
0 300
0
0
0 /
/

0
0
--t- KEN(undi!
0 - - Q - - KEN(dest1
- 9- - KEN(com
0 --t-IND (undi!
- - .c3 - - IND (desti
- 4-- IND (comi
0
a
0
0
0 O !
0 100 200 300 400 500 600
0 t' (kPa)

0
0
0
0
0
0
0
0
Figure 13. Shear stresses showing a 'bonding'effect
0
0
0 Oedometer consolidation, Kenya Pit 1, 2.0m (C)
0 1.8
10 kPa
0
0
1.7
0
100
0
0 1.6

0
200
0
1.E
0
0
0 1. d

a
0 400
1.:
0
e
0 1.: 600

0
0 800
1.
0
1000
0
1200
0 1.
1400
0
1600 kPa
0
0. 1 1 1 1 1 1 1 1 1
0 10’ 102 7 03 104 105 Seconds
0 1 1
1 hour 24 hours

0
0
0
Figure 14. Stress dependance of oedometer consolidation settlement.
0
0
0
0
e
0

0
a Compression index, C, (oedom.) vs. Relative voids ratio, eR
Key: 2 = soils developed on Older Quaternary volcanics
0 3 = soils developed on Tertiary volcanics
0
0
e
1
0
0
a 0.8
0

e
0
0 0.6
0

0
a 0.4 0 0
e 0 INDON., 2
0 KENYA, 3
0
e 0.2

a -0.2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6

0 eR

0
0
0
0
0 Figure 16. Relationship of Compression Index, C,, to relative voids ratio, e,
 Voids ratio vs. Effective stress (oedometer) Indonesia Pit 10, 5.0m
3

'\\\ / c r l t i l state line

2.5

1.5

1
1 10 100 1000 10000
stress (kPa)

Figure 17. Application of Vaughan's 'bonded' soil model

0
0
0
COMPACTION TEST PLOTS
0
KENYA +Pit 3, 1.Om (Dn)
0 16.0 -
+Pit 3, 2.0m (Dn)
*Pit 5, 1.6m (Dn)
0
0
0
0
s
E
2
14.0
15.0:
13.0

12.0 --
7 - - t - .Pit 5, 1.6m (Dr)
-e-- Pit 5, 1.6m (Pn)
+Pit 7, 2.5m (Dn)
- - A- - - Pit 7, 2.5m (Dr)
-o-ES4, 2.0m (Dn)
Y,
0 s 11.0
.-
v)
--

0 g 10.0 --
e --
9.0
0
0
0
0
7.0
8.0
6.0 -/
1
20
I

30
I

40
I

50
I

60
I

70
I
I

80
I
90
Moisture content (%)
0
0
COMPACTION TEST PLOTS - INDONESIA
0
0 16.0
--t-Pit 1, 2.0m (Pn)
0 - - c) - .Pit 2, 2.0m (Pn)
0 14.0 --f- Pit 3, 2.0m (Pn)
+Pit 4, 2.0m (Pn)
0 s 13.0 ~

--t-Pit 5, 2.0m (Pn)

0
3E
W
12.0 - - t - .Pit 6, 2.0m (Pn)
.-s 11.0 - ---e-- Pit 8, 2.0m
0 v)
8
g 10.0 ~

0
I 2
Q 9.0 -

0
I
8.0 -

0 7.0
0 6.0 C-- +. . +--
20 30 40 50 60 70 80 90
0
Moisture content (%)
0
0
0
Figure 18. Compaction test plots
0
0
0
0
0
0 Residual friction angle vs. Plasticity Residual friction angle vs. Clay
index (Kenya) size fraction (Kenya)
0
0
0
.\I35
30 e

0
0
0
I v-
ISkempton (1964)
5
0 5

0 0 10 20 30 40 50 60 70 80 0 I 0 2 0 3 0 4 0 5 0 60 70 80
% CLAY
0 PI (70)
ONMC (BS1377) D r - d r y (BS1377) e N M C ( GWmr
d y( G W
r 1-
0
0 Ring shear test plots (Kenya) Normal stress vs.
Shear stress
0
0
09 --
0 0 8 --
0 07 --

0 cI
m
06.- - - 0 - - (Air dry)
p1t5

-%
+Ptt7 (NMC)
05.- - - cf - -Pit7(krdry)
0 04.- +ES4 (NMC)

0 0 3 --
0 2 --
0
01 --

0 0 1
0
0; (kPa)
0
0
0
0
0
Figure 19. Results of ring-shear tests on remoulded Kenyan samples
 A. Idealised weathering profiles - B. Example of a complex
without corestones (left) and profile with corestones
with corestones (right)

..............
- Wealherddisintegrated rock
- Rock discoloured by weathering

Layer 3
(eg. ashltuff~lava)

VI
Resrdual soil Layer 2
(eg. ashltuffflava)

Weathered rock Layer 1

( e g . ash/tuff/laval
I Fresh rock

C. Example of a profile, without D. Layered sequence of parent

corestones, with a sharp volcanic rocks reflected in
transition from rock to soil weathering profile

Figure 20. Examples of tropical weathering profiles in igneous rocks.

(A & B after BS 5930:1981 and Anon, 1990)
0
0
0
.
c \'

0
0
0
0
0
0
0

0
0
0
0
0

0
0
0
0
0
0
0

0
0
0

0
0
0
0
rn
0
0
0
0
0

0
0
0
0 Tables
a
a
0
0
0
0
0
0
0
0
0
0

0
0
0
0
0
 a
D
s
.-
r
n N
m
d
n :3
? c4
d m r
N N N

.............. ..............
-m -m >- m
0
L - -
C
0 0 2 : :
m
a cdi
m 9 x
%:2
v,

........
U)

.............
v)

.............. ............. 3C
.I

+ M
8 0
80
c,
1"

fi
N z
l cn

0
c
"f
2 0
B
0
0 He
c
3 e

c,
0
............. .............. ............. d)
.-
E 2
E
0 d
c
I
63
'5
a :jj
6
c g c
c 0

gz
m E
E 2
ki5
2
32 Br E
.............
U
m
a
*
w-
3
e
.-0
s .
= a
0 0
I?
c

PQ
.m E
t.5
0 Q m
I-m
............. ..............
0 0
m -
N t
0
m-
a
0
m 0
cug - p+ Lo
c 0
c

0 oc
g9
' U
0 0
8 %

0 0
c
Z h 0)
c I-
c c

0 ........ ..............

0
.-
5
0 -0
a 5
-
a 2 -a0 0

0 2 ;
.- 0
-.-.-
c

U t 7 E
* o c 7
.............. u
........a a
.............. U
9)
..................

. .
d ) Q
5
-
a '5 - li
-a0
s 1?(
0 0
U .$ 4 U
0

a 2:
.............. ........ a
0
0
0
0
0
0
0
0
8
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
I)
I)
0
0
0
0
0
C
I)
I)
I)

0
0
0
0 2
a,
0 -
.-C
0, 0)
m c
.- c
.-
0 8 2 0
L
0
0
L

0-CI) 0

0 s3 L,
m
c
2
0 a,
3
c

0 -d,Q
.-
c
L

0
0
0
0
0
0
0
0 m

0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
-0 U a,
0 aoo

0
0
0
0
0
ex 2
0
0
0
0
I)
0
0
Table 4. Approximate correlation between the major classes of tropical residual
0 soils in the FAO-UNESCO, French (Duchaufour), and American soil classification
systems [based on Duchaufour, 1982; Uehara, 1982; Morin & Todor, 1975; Anon,
0
19901.
0
FAO-UNESCO CLASSIFICATION FRENCH CLASSIFICATION U.S.A. SOIL TAXONOMY
0 (FAO 1988) (Duchaufour 1982) (Soil Survey Staff 198 1)

0 Cambisols (some), Luvisols (some), Fersiallitic soils Alfisols, lnceptisols

Calcisols, Alisols
0 ........................................................................................................................................................................................................
Andosols Fersiallitic Andosols lnceptisols landepts)
........................................................................................................................................................................................................
0 Luvisols (some), Alisols, Lixisols. Ferruginous soils Alfisols, Ultisols
Plinthosols
0 ........................................................................................................................................................................................................
Nitosols, Acrisols, Cambisols (some), Ferrisols Ultisols, Oxisols
Lixisols, Luvisols (some), Plinthosols

0
0
Ferralsols. Plinthosols

Vertisols

Podzols
I Ferrallitic soils

Vertisols

Podzols
I Oxisols
.......................................................................................................................................................................................................
Vertisols

Spodosols

0
0
0
0
0
0
0
0
a
0
a
0
0

0
I)
0
0