Ground Improvement Proceedings of the Institution of Civil Engineers

Volume 166 Issue GI4 Ground Improvement 166 November 2013 Issue GI4
Pages 196–208 http://dx.doi.org/10.1680/grim.11.00016
Relative density correlations are not reliable Paper 1100016
criteria Received 06/04/2011 Accepted 07/02/2013
Published online 30/05/2013
Hamidi, Varaksin and Nikraz Keywords: foundations/geotechnical engineering/strength and testing of
materials
ICE Publishing: All rights reserved

Relative density correlations

are not reliable criteria
j
1 Babak Hamidi MEng(Hons), MIEAust j
3 Hamid Nikraz MEng, PhD, FIEAust, CEng, NPER
PhD candidate, Curtin University and member of ISSMGE Ground Professor & Head of Civil Engineering Department of Curtin
Improvement Technical Committee, Perth, Australia University, Bentley, Western Australia, Australia
j
2 Serge Varaksin MEng, CFMS
Chairman of ISSMGE Ground Improvement Technical Committee,
Paris, France

j
1 j
2 j
3

The concept of relative density was developed with the intention of appropriately defining looseness and denseness
of sand or sand–gravel soils in a meaningful way; however, there are sufficient amounts of research and case studies
to demonstrate the unreliability of this concept due to its large inherent errors. Nevertheless, this parameter found
its way early on as a ground improvement acceptance criterion based on the same philosophy that led to its
formation. As there was general agreement among engineers that, in any case, direct testing of relative density in
actual deep ground improvement projects was difficult, time consuming and costly, direct methods of relative density
measurement were abandoned in favour of correlation to other field tests. At first glance, this may have seemed to
work out quite well as the conceptual unreliability of relative density did not come into play, but a deeper look could
reveal that the proposed correlations are also as unreliable as the concept itself. This paper will discuss the reasons
why relative density correlations should not be used as ground improvement acceptance criteria, and alternative
reliable criteria will be proposed.

Notation ratio of about 78% from the theoretical free-fall

a Skempton parameter energy
b Skempton parameter OCR overconsolidation ratio
Co , C1 and C2 experimental coefficients qc cone penetration test (CPT) cone resistance
Coc overconsolidation coefficient (kPa)
Cu coefficient of uniformity qcNC CPT cone resistance for normally consolidated
Dd relative density (%) soil (kPa)
D50 mean particle size qcOC CPT cone resistance for overconsolidated soil
emax maximum index void ratio or the reference (kPa)
void ratio of a soil at the minimum index R roundness
density/unit weight ˜Nf correction term that is a function of fines
emin minimum index void ratio or the reference void content
ratio of a soil at the maximum index density/ ó9 effective vertical, horizontal or mean stress
unit weight depending on the overconsolidation ratio (kPa)
Fc fines content (%) ó v9 effective overburden pressure (kPa)
fshell shell correlation factor ö9 effective internal friction angle
KO ratio of effective horizontal to vertical stress for
an overconsolidated soil 1. Introduction
KONC ratio of effective horizontal to vertical stress for Hamidi et al. (2013) have made a comprehensive review of
a normally consolidated soil current ASTM test methods (ASTM, 2006a, 2006b) and studies
N standard penetration test (SPT) blow count carried out by numerous researchers to demonstrate the limit-
N78 SPT blow count corresponding to an energy rod ations of relative density, and to show that although the objective

196
Ground Improvement Relative density correlations are not
Volume 166 Issue GI4 reliable criteria
Hamidi, Varaksin and Nikraz

Ottawa sand
of developing this parameter was to bring the behavioural charac-
Del Monte white sand
teristics of soils together on a common basis in consistent and Monterey sand
practically useful relations and to provide a tool for communica- 1·4 Lapis Lustre sand
tion between engineers, due to its formulation, relative density is Crushed basalt Cu ⫽ 1·4
prone to errors with magnitudes of tens of per cent (Hamidi et 1·2
al., 2013). The amount of inaccuracy associated with relative emax
density is so great that it makes this concept unreliable as a
1·0

Void ratio
ground improvement acceptance criteria.

As the notion of relative density and the thought of describing 0·8

the characteristics of soils with one parameter is very attractive
indeed, this parameter found its way into ground improvement in
0·6 emin
the early days of its development. Yet, practical application of
relative density in ground improvement projects proved time
consuming and costly due to the extraction of high-quality 0·4 Very
samples of the treated soil. The unreliability of relative density angular Subangular Rounded
due to its formulation further complicated its usage; hence the Angular Subrounded
correlation of relative density to other commonly used field tests 0·2
0·1 0·2 0·3 0·4 0·5 0·6
appeared as an attractive means of applying the relative density Roundness, R
parameter without the actual usage of the problematic formula-
tion. Thus, it can be occasionally seen that in some projects the Figure 1. Density limits as a function of grain shape for laboratory
specification stipulates that the improved ground must exceed a fractions with Cu ¼ 1.4 (Youd, 1973)
minimum relative density value or curve based on standard
penetration test (SPT) or cone penetration test (CPT) correlations.
Correlations, themselves, are as questionable as the concept of others such as Tokimatsu and Yoshimi (1983), Skempton (1986),
relative density, and have been the subject of discussion by and Cubrinovski and Ishihara (1999) note the importance of the
experts over many years. In this paper the authors review a effect of particle diameter on relative density–field test relation-
number of well-established correlations that are commonly ships.
referred to, and will demonstrate the unreliability of relative
density correlations as acceptance criteria for ground improve- Holubec and D’Appolonia (1973) focused their study on the
ment. sphericity and the shape of four types of medium to fine sands
with the same gradation but varying particle shapes. What they
2. Relationship of relative density with soil have found is that granular soils at the same relative density can
characteristics have drastically different engineering properties and have con-
A fundamental question that crosses the mind is that, irrespective cluded that the use of relative density criteria in design, without
of the limitations and errors that are associated with relative considering particle shape, can result in poor or misleading
density concept and formulation (Hamidi et al., 2013), would it predictions of soil behaviour. As can be seen in Figure 3, Holubec
be true to say that if two soils had the same relative density, then and D’Appolonia show that while both emin and emax increase
they would possess the same physical characteristics and will with the increase of the coefficient of angularity, the increase in
behave the same? It would have been very satisfying if the reply emax is considerably more than that for emin and that the
to this question was positive; however, this is not the case, and difference between these two limits increases with the angularity
the use of relative density correlations based on average sand to of the soil particles.
predict soil behaviour without considering other parameters can
result in poor or misleading predictions. It can also been seen from the work of Holubec and D’Appolonia
that each of the four sands that they tested had separate and
Youd (1973) conducted a study on the maximum and minimum distinct relative density versus friction angle relationships
indexes of 13 specimens of different clean sands with less than whereas although the roundest particles had the least friction
5% fines passing the 0.075 mm sieve, and was able to identify angles and exhibited the minimum increase in friction angle with
clear relationships between maximum void ratio, emax , and mini- increasing relative density, the sands with angular particles had
mum void ratio, emin , with particle roundness, grain shape, range the highest friction angles with an intermediate increase with
of particle sizes defined by the coefficient of uniformity, and the increasing relative density. What can be assessed is that friction
type of gradational curve. Figures 1 and 2 show the relationship angle is a function of both relative density and particle shape. In
between limit void ratios with roundness and coefficient of fact, this research suggests that equally large differences in
uniformity. Youd did not identify a unique relationship between friction angle are possible with variations of particle shape as
mean particle diameter (D50 ) and limit void ratios; however, with changes in relative density.

197
Ground Improvement Relative density correlations are not
Volume 166 Issue GI4 reliable criteria
Hamidi, Varaksin and Nikraz

1·40

Olive sand
1·4

Southport sand
1·20
Maximum
void ratio
1·2
1·00
An

Ottawa sand
gu
lar R⫽

Glass beads
Void ratio, e
emax

1·0 0·1
7
0·80
0·20
0·8
Sub
ang
ular
0·25 0·60
Sub Minimum
0·6 roun 0·30 void ratio*
ded
0·35
Rou
nded 0·49 0·40 Vibration
0·8
0·70
Modified Proctor
0·20
0·6 Ang 1·0 1·2 1·4 1·6 1·8 2·0
ular R⫽0
·17 Round Angular
emin

Coefficient of angularity, E
Suba 0·20
ngula
0·4 r
0·30 0·25 Figure 3. Effect of particle shape on minimum and maximum void
Subro 0·35
unde ratios (Holubec and D’Appolonia, 1973). *Minimum void ratio
d Rounde 0·49
d 0·70 based on modified Proctor compaction test, except for glass
0·2
1 2 3 4 6 10 15 beads
Coefficient of uniformity, Cu

Figure 2. Generalised curve for estimating emax and emin from

Figure 4 shows the differences between effective peak and critical
gradational and particle shape characteristics. Curves are only
state friction angles at different effective vertical stresses for the
valid for clean sands with normal to moderately skewed grain size
four soils when relative densities were 20, 50 and 80%. It can be
distributions. (Youd, 1973)
seen that while the general trend is the reduction of the friction
angle differences at higher effective vertical stresses, the lines

Stress–strain relations are also influenced by particle shape.

15
Holubec and D’Appolonia show that the more angular the Dd ⫽ 0·8
particles are, the greater the failure strain is for a given relative Dd ⫽ 0·5
Difference between effective peak
and critical state friction angle: °

density. Holubec and D’Appolonia perform a miniature penetra- Dd ⫽ 0·2

tion test using a 12.5 mm diameter rod that was driven to a depth
10
of 20 cm in sands that were compacted to selected densities. The
data showed a marked increase in the penetration resistance with
the increase of angularity of the sand particles. For example they
observed that the required number of blows for a sand with
angular particles is nearly double the values for a subrounded 5
sand when relative density is 70%. Conversely, a blow count of
20 indicated a spread in relative densities from 66 to 86%.

More recently, Liu and Lehane (2012) have studied the behaviour 0
of four uniformly graded silica soils with similar mineralogical 10 100 1000
Effective vertical stress: kPa
compositions but with distinctly different particle shapes. The
soils were subjected to centrifuge CPT and direct shear tests at Figure 4. Difference between effective peak and critical state
different relative densities in dry and saturated states. In these friction angles in four different uniformly graded silica soils with
tests, gradation and material were controlled to be similar; hence similar mineralogical compositions at different relative densities;
any differences in test results could be attributed to the grain’s reconstructed from Liu and Lehane (2012)
sphericity, roundness and roughness.

198
Ground Improvement Relative density correlations are not
Volume 166 Issue GI4 reliable criteria
Hamidi, Varaksin and Nikraz

fitted to the plotted points of the four soils are distinctly different per a non-existent standard of its time. Today, it is known that
for any of the three relative density values. due to its formulation, relative density is prone to large errors
whereas Yoshimi and Tohno (1973) have shown, as an example,
It can be understood from the work of Youd (1973), Holubec and that if minimum index density increases by 1% then relative
D’Appolonia (1973), and Liu and Lehane (2012) that applying density will reduce by 14%. Thus, even small differences in limit
relative density correlations from one type of sand to the other, indexes originating from Gibbs and Holtz’s testing method could
even if the gradation and mineralogy are similar, could lead to have had significant impacts on the calculated relative densities.
very misleading results.
Unlike the rather well known Gibbs–Holtz relative density versus
3. Correlation of relative density with field overburden pressure diagram that has been referenced in numer-
tests ous publications, those who have actually seen the Gibbs and
By definition, correlation is a statistical relation between two or Holtz (1957) paper know that they published a relative density–
more variables such that systematic changes in the value of one penetration resistance chart (as reconstructed in Figure 5), not the
variable are accompanied by systematic changes in the other. better known chart of Figure 6 which first appeared in Earth
Correlations are not physical laws or theorems; they are simply Manual (Bureau of Reclamation, 1960) in accordance with Gibbs
statistical relations and only meaningful once their scatters, and Holtz. It is noted that the Bureau of Reclamation (Bureau of
deviations and variances are known. In general, empirical correla- Reclamation, 1998) does not use the term correlation, but rather
tions are derived from a set of specified data under special refers to the chart as the criterion for predicting the relative
conditions that are not necessarily applicable to other data, soil density of sand. In addition to the limitations that ASTM (2006a,
and conditions. In addition to the inherent drawbacks and limit- 2006b) has in place for the application of relative density, the
ations of relative density, when consideration of a correlation’s Bureau of Reclamation also limits the use of its chart to sands
applicable domain is not taken into account, inaccurate and containing less than 10% fines and no gravels.
unrepresentative outcomes should be expected.
Osterberg and Varaksin (1973) extracted frozen soil samples from
This paper will limit its review and discussion to some of the Lake Michigan and compared them with what was estimated by
better known relative density–field test relationships as discussing Gibbs and Holtz. Predictions were quite different from reality and
all studies is beyond the scope of a journal paper; however, the it was concluded that relative densities obtained from SPT using
trend of this discussion and its conclusions is equally applicable the Gibbs and Holtz chart had no relationship to the actual
to any other such correlation. Furthermore, the limitations and relative densities. This study was an indication that relative
drawbacks that are associated with any of the field testing density correlation of one sand cannot be extended to any other
methods are not considered herein as they are different issues and sand.
not directly associated with the unreliability of relative density.
Contrary to the general overconfidence by others, Holtz (1973)
3.1 Relative density–standard penetration test does not share the same unconditional and unlimited trust that
correlations others have for their chart. In the discussion on the results of
Although the first correlations between relative density and SPT Osterberg and Varaksin (1973), Holtz notes that everyone should
were qualitatively realised by Terzaghi and Peck (1948), probably recognise that the SPT is a relatively crude test and no one should
the best known and most referenced estimation method has been expect to determine the relative density of sands to the nearest
developed by Gibbs and Holtz (1957) based on data obtained 1% or anything like that, and further adds that he and Gibbs
from calibration chamber tests performed at the US Bureau of developed a set of correlations to take into account the effect of
Reclamation. Gibbs and Holtz performed tests on a fine-grained overburden pressures, and that they never indicated that the sets
and a coarse-grained sand by placing them at controlled densities of curves developed at that time were necessarily applicable to all
and moistures in a heavy steel tank, 0.9 m in diameter and 1.2 m cohesionless soils under all conditions. More importantly, he
in height. Overburden pressure was realised by load plates and clarifies that they had always laid stress on relative density trends
loading springs. The maximum density was determined by indicated by SPT values rather than the specific individual values.
vibrating the saturated material to constant density or by using This discussion suggests that the research of Gibbs and Holtz
extreme compaction hammer blows, whichever gave higher may have been blown out of proportion, and rather than acknowl-
values, in a container of known volume. Minimum density was edging the trend of relative density versus overburden pressure it
found by lightly pouring the dry material into a container of is used systematically for something that was never the intention.
known volume.
Meyerhof (1957) has formulated the coarse sand graph of Gibbs
Gibbs and Holtz carried out this research in 1957 (Gibbs and and Holtz (1957) and expressed it in the form of Equation 1 or
Holtz, 1957); that is, 12 years before ASTM published its first Equation 2 (after conversion to SI units); however, in practice this
standard on relative density (ASTM, 1969). Obviously the testing equation is often extended to most types of sands regardless of
procedure for measuring limit densities could not have been as the soil particle size and shape, gradation and mineralogy.

199
Ground Improvement Relative density correlations are not
Volume 166 Issue GI4 reliable criteria
Hamidi, Varaksin and Nikraz

100

kPa
275
î
Air-dry í
ì No pressure
Moist
Coarse sand

Air-dry 62 to 69 kPa
80 î
Air-dry í
152 to 186 kPa
Penetration resistance: blows/305 mm

Pa
Moist

8k
î
Air-dry

13
í
ì 248 to 324 kPa
60 Moist
0 kPa
î

Pa
í
Fine sand

Pressure as measured

5k
62 kPa

27
All tests on air-dry a
172 kPa ì kP
sand 69
40 331 kPa

a
8 kP
13 a
0 kP
20 Pa
69 k
a
0 kP

0
0 20 40 60 80 100
Relative density: %
0 15 35 65 85 100
Very Very
Loose Medium Dense
loose dense

Figure 5. Curves of relative density plotted against penetration

resistance for fine and coarse sands, reconstructed from Gibbs
and Holtz (1957)

0 0
Obviously, this will further cast doubt on the reliability of the
correlation as it is very clear in the original Gibbs and Holtz
0·5 24
Vertical pressure: ton force/ft2

research that the fine and coarse sand curves do not coincide.
Vertical pressure: %

1·0 48
1: N ¼ ð17 þ 0:25ó v9 ÞD2d
0
10

1·5 72
90
80
70 rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2·0 96 N
0
50 6 % Dd ¼
15 40 ity: 2: 17 þ 0:25ó v9
2·5 d ens 120
ative
Rel
3·0 144 where N is the SPT blow count, Dd is the relative density
0 10 20 30 40 50 60 70
expressed as a ratio (not percentage), ó v9 is the effective vertical
Penetration resistance, N: blows/305 mm
stress (kPa).
Figure 6. Criterion for predicting relative density of sand from the
penetration resistance test (Bureau of Reclamation, 1998) In lieu of Equation 1, Haldar and Tang (1979) have approximated
(1 ton/ft2 ¼ 107.3 kN/m2 ) Gibbs and Holtz’s relative density correlation using Equation 3
(after conversion to SI units).

200
Ground Improvement Relative density correlations are not
Volume 166 Issue GI4 reliable criteria
Hamidi, Varaksin and Nikraz

:
3: N ¼ 20D2d 5 þ 0:21ó v9 D2d exhibiting a small difference between the limit indexes, the
uncertainty in direct laboratory determination of in-situ relative
density (which is prone to large errors) is expected to be less than
Haldar and Tang who note the difficulty of obtaining data that estimations using Gibbs and Holtz’s relationship for normally
actually includes measurements of Dd , N values and ó v9 have consolidated deposits, and that a systematic bias appears to exist
plotted against calculated (using Gibbs and Holtz’s relationship) in Gibbs and Holtz’s prediction relationship.
relative density of what they could gather (Gibbs and Holtz, 1957;
Marcuson and Bieganousky, 1977; Moretto, 1954; Varaksin, 1970; Peck and Bazaraa (1969) proposed Equations 4 (after conversion
Wu, 1957) onto the graph of Figure 7. It can be seen that there is to SI units) for predicting relative density. Comparison of these
considerable spread about the 458 bisecting parity line. The relations will show that for equal N values, these equations will
majority of data is above the 458 line, suggesting that Gibbs and consistently yield higher relative density estimates than Gibbs
Holtz’s relationship may be non-conservative. Figure 7 also shows and Holtz’s relationship.
a mean line, an upper bound (line B) and a lower bound (line E).
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
For a given value of relative density calculated from Gibbs and N
Holtz’s relationship, the measured relative density may be Dd ¼ for ó v9 , 72 kPa
4a: 20 þ 0:84ó v9
assumed to follow a triangular distribution between the upper and
lower bounds and be symmetrical about the mean line. On this
basis, the mean measured relative density is only 75% of the
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
calculated value and the scatter around the mean has a coefficient N
of variation of 27% which is constant at any relative density Dd ¼ for ó v9 > 72 kPa
4b: 65 þ 0:21ó v9
calculated. Haldar and Tang conclude that except for sand

1·2

Mean line

1·0
Calculated relative density by Gibbs and Holtz's relationship

Line E

0·8

0·6 Lower bound Upper bound

line C line D

0·4

Gibbs and Holtz (1957)

Moretto (1954)
Varaksin (1970)
0·2 Wu (1957)
Upper and lower bounds for
Line B
Marcuson and Bieganousky (1977)
Line A

0
0 0·2 0·4 0·6 0·8 1·0 1·2
Measured relative density

Figure 7. Comparison of measured and calculated relative

density using Gibbs and Holtz’s relationship (Haldar and Tang,
1979)

201
Ground Improvement Relative density correlations are not
Volume 166 Issue GI4 reliable criteria
Hamidi, Varaksin and Nikraz

Lacroix and Horns (1973) note that in their experience Gibbs and 1 þ 2K O
C oc ¼
Holtz’s method of predicting relative density yields results that 7: 1 þ 2K ONC
are too high for heavily compacted fill. As noted by Peck and
Bazaraa (1969), they also agree that while Gibbs and Holtz
greatly underestimate N values corresponding to 100% relative Parameters KONC and KO are the ratio of effective horizontal to
density, Bazaraa’s (Bazaraa, 1967) relationship was in better vertical stresses, respectively, when the soil is normally consoli-
agreement with the blow count data. dated and overconsolidated. Skempton refers to Mayne and
Kulhawy (1982) for determining these coefficients as a first
Marcuson (1978), who refers to a study on four sands at various approximation according to Equations 8 and 9
relative densities under overburden pressures in the laboratory of
the US Army Waterways Experiment Station (WES), concluded 8: K ONC ¼ 1  sin ö0
that a simple family of curves relating N values, overburden
pressure and relative density for all sands under all conditions is
not valid. He also concluded that based on comparisons between
K O ¼ K ONC (OCR)sin ö
0
the relationships presented by Gibbs and Holtz, Bazaraa and 9:
WES, SPT is not sufficiently accurate to be recommended for
final evaluation of the density or relative density, unless site-
specific correlations are developed. where OCR is the overconsolidation ratio and ö9 is the effective
internal friction angle.
Skempton (1986) proposes the relationship between relative
density and N values in the general form of Equation 5, with a Tokimatsu and Yoshimi (1983), who were studying soil liquefac-
and b as two parameters. Skempton tested five different types tion, modified Equation 2 to Equation 10 by taking into account
of sand and proposed values for a and b ranging respectively the effect of fines content, Fc , and introducing ˜Nf as a correction
from 15 to 30 and from 17 to 24 in the imperial system (ton/ft2 term (refer to Table 1). However, they themselves did not
(1 ton/ft2 ¼ 107.3 kN/m2 ). It can be seen that the values for these demonstrate confidence in their proposed equation and note that
parameters can be respectively 100% and 40% more than the its application had yet to be proven.
least values. Inclusion of Bazaraa (1967) would even further
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
increase the figures respectively to 300% and 47%. This clearly u N ˜N f
Dd ¼ 0:21u
t þ :
suggests that the specific studies of one site cannot simply be ó v9 17
used on other sites. 10:
0:7 þ
98
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
N More recently, Hatanaka and Feng (2006) have carried out a
Dd ¼
5: a þ bó v9 comparative study on high-quality undisturbed samples recovered
by the in situ freezing method. The material used in this study
was less than 4.75 mm in size and D50 was less than 1 mm. They
Among his observations, Skempton (1986) notes that at a given
then compared the measured and calculated values of relative
relative density and overburden pressure, N values are higher for
density using Meyerhof (1957), Bazaraa (1967) and Tokimatsu
sands with larger grain sizes (D50 ). He also assesses that there is
and Yoshimi (1983). As shown in Figure 8, the estimated values
direct evidence that ageing of sand will increase the SPT blow
of relative density based on Meyerhof’s method were in the range
counts. This suggests that not only is relative density influenced
+15 to 45% of the measured values. Similar to Haldar and Tang
by numerous other parameters such as gradation, particle size,
(1979), this research also shows a large scatter of results about
overburden pressure, mineralogy and particle shape, and hence its
the prediction equation; however contrary to Haldar and Tang,
correlation in one sand may not be reliable for any other sand, but
here the scatter is mostly concentrated on the lower side of the
even the correlation for a soil at a specific time or state may be
bisecting parity line, suggesting that Meyerhof’s equation is
not valid and applicable at other times and conditions. Skempton
also identifies a relationship between the effects of overconsolida-
tion and relative density and introduces an overconsolidation Fines content: % ˜Nf
coefficient, Coc , into Equation 5 to derive Equation 6.
0–5 0
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 5–10 Interpolate
N 10- 0.1Fc + 4
Dd ¼
6: a þ C oc bó v9
Table 1. ˜Nf – fines content correlation (Tokimatsu and Yoshimi,
1983)
The coefficient Coc itself can be calculated from Equation 7.

202
Ground Improvement Relative density correlations are not
Volume 166 Issue GI4 reliable criteria
Hamidi, Varaksin and Nikraz

Relative density estimated from Meyerhof (1957): %

Measured Dd ⬎ 50% more appropriate to use the definition of relative density (ASTM,
2006a, 2006b) to calculate a value which is prone to error
Measured Dd ⬍ 50%
100 (Hamidi et al., 2013) rather than going to the trouble of
estimating a value which may be even more erroneous and
80 misleading.

Many other SPT-Dd correlations could be introduced and re-

60 viewed and this discussion continued endlessly; however, what is
⫺45% evident and established is that due to the numerous other
40 ⫹15% parameters that influence the outcome, SPT–relative density
correlations will have an unreliably large amount of scatter and
cannot be trusted as ground improvement criteria.
20

3.2 Relative density–cone penetration test estimations

0 Many researchers, for example Jamiolkowski et al. (2001), note
0 20 40 60 80 100
Measured relative density of undisturbed sample: % that the first CPT versus relative density correlation was pub-
lished by Schmertmann (1976). This is not entirely accurate as
Figure 8. Comparison of measured and calculated relative density the authors are aware of at least one previous publication by
using Meyerhof’s equation (Hatanaka and Feng, 2006) Schmertmann (1975) and a reference by Schmertmann (1978) to
Mitchell and Gardner (1975).

Although the authors agree that Schmertmann’s (Schmertmann,

1976) publication should be recognised as an early research on
generally underestimating relative density. Similar results were the subject, it has unfortunately been published in the form of a
also obtained when Tokimatsu and Yoshimi’s (Tokimatsu and contract report for WES and today this document is not even
Yoshimi, 1983) correlation was used. In this case, the range of available in WES’s own library (Report DACW 38-76-M 6646
estimated to measured relative density was from +25 to 20%. requested 16/03/2011). The authors have unsuccessfully at-
tempted to obtain a copy of the report through numerous sources,
Tatsuoka et al. (1978), who examined the accuracy of Meyerhof’s and to date have not met anyone who has actually read the
(Meyerhof, 1957) expression by studying the results of SPT on publication. It can only be speculated that many researchers who
normally consolidated sandy deposits and conventional undis- are more recently referring to this publication have also never
turbed samples, made a similar assessment and concluded that read the original publication and are simply citing information
Meyerhof’s formulation tends to underestimate the relative from other publications. This can be quite concerning due to
density of fine sands and silty sands. possible publication errors, misinterpretations or the fact that
each of the intermediate authors only extract that scope of work
For estimating relative density, Cubrinovski and Ishihara (1999) which is relevant to their own study. By reviewing Schmertmann
note the importance of grain size and propose Equation 11 (1978) and other research such as Villet and Mitchell (1981) the
authors understand that the updated correlation of Schmertmann
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
: rffiffiffiffiffiffi (1976) is the same as that which was later published in
N 78 ðemax  emin Þ1 7 98
Dd ¼ Schmertmann (1978).
11: 9 ó v9
Schmertmann (1978), who himself was aware that cone resistance
is significantly affected by grain size distribution, cementing,
where N78 is the SPT blow count corresponding to an energy rod lateral stresses, depth of overburden, compressibility, pore pres-
ratio of about 78% from the theoretical free-fall energy. sure and thin layer effects, modified and updated his earlier work
(Schmertmann, 1975) after performing 80 more tests in a
Cubrinovski and Ishihara report that by using Equation 11, in calibration chamber with a 1.2 m diameter on two artificial sands
84% of instances they were able to calculate relative density with opposite extreme crushabilities, two natural fine sands, and
within a deviation of 20% from the measured values, and more one natural and one artificial medium sand and proposed a
than half the calculations were within 10% deviation. Although correlation chart which only takes into account the effect of
this amount of accuracy is by no means sufficient to reliably vertical effective stress and is for normally consolidated, recent,
establish an acceptance criterion upon, in any case, here three uncemented fine SP sands. He also proposed a correction factor
parameters are involved, two (limit void ratios) of which are for converting overconsolidated sands’ cone resistances, qcOC , to
included in the original definition of relative density, and the third normally consolidated sands’ cone resistances, qcNC : Once simpli-
(the N value) is a function of the in situ density. It may be simply fied, the expression will take the form of Equation 12

203
Ground Improvement Relative density correlations are not
Volume 166 Issue GI4 reliable criteria
Hamidi, Varaksin and Nikraz

qcOC 3 : Table 2. After comparing their calibration chambers and chamber

¼ 1 þ ð4OCR0 42  1Þ
12: qcNC 4 boundary conditions, Villet and Mitchell conclude that these
large differences were due to soil type, thus suggesting that cone
resistance, vertical stress and relative density relationships are
As shown in Figure 9, there are significant differences between not unique and universal for all sands, but rather a function of
the relative density estimations of Schmertmann (1975) and the sand being penetrated. This conclusion is applicable to any
Schmertmann (1978). Although Schmertmann (1978) does not other research that defines a relationship between these para-
show the scatter of testing results, he does note that for clean meters.
quartz sands his chart is able to estimate relative density with a
standard deviation of about 10%. The standard deviation of In fact, had Villet and Mitchell used the same calibration
Schmertmann (1975) was 7%. This suggests that irrespective of chamber as Schmertmann, then the differences would have been
the scatter which one may experience in one study, there is a even larger. A number of researchers such as Parkin (1977) and
strong possibility that the correlation would be almost mean- Parkin and Lunne (1982) have studied the effects of calibration
ingless for another soil. chamber size and boundary conditions. Parkin’s results indicate
that the ratio of the calibration chamber diameter to the CPT
This has also been observed by Villet and Mitchell (1981) who cone diameter and boundary conditions affects the test results
performed a series of tests on four gradations of a commercially whereas either decreasing diameter ratio or maintaining constant
available windblown dune sand in a calibration chamber with a stress rather than constant volume boundary conditions results in
diameter of 0.76 m and height of 0.8 m. The sands were mainly lower cone resistance.
of quartz and feldspar grains and subrounded to subangular in
shape. In addition to the corrections that may be necessary due to the
geometry and boundary conditions of the calibration chamber,
The relationship of Villet and Mitchell is also shown in Figure 9. Bolton and Gui (1993) mention the simulation method of over-
The differences between Villet and Mitchell’s measurements and burden pressure by surcharging as another limitation of the
Schmertmann’s (1978) are significant and from as low as 20% in calibration chamber, as the effect of stress gradient due to the
dense sand with lesser vertical effective stress to 140% for loose self-weight of the soil cannot be simulated in this way. Bolton
sand subject to more vertical effective stress. This is shown in and Gui conclude that the results obtained from calibration
chambers always leave room for questioning, and propose the use
of centrifuge testing.
CPT cone resistance, qc: MPa
Irrespective of Bolton and Gui’s view, calibration chamber testing
0 10 20 30 40 50
0 has remained the dominant approach for developing relative
Villet and Mitchell (1981) density–CPT correlations. Baldi et al. (1986), who have proposed
Schmertmann (1978) one of the most popular correlations, performed calibration
50 Schmertmann (1975) chamber tests on Ticino sand and Hokksund sand. They proposed
an expression in the form of Equation 13
 
1 qc
Vertical effective stress: kPa

100 Dd ¼ ln
13: C2 C o ó 0C1
Dd

150 where Co , C1 and C2 are experimental coefficients.

100%

⫽
80%
60%

10
40%
20%
0%

0%
80%

200
Vertical effective qc ratios at relative densities of
60%
40%
0%
20%

80%

stress: kPa
60%

20% 40% 60% 80%

40%

250
20%

100 1.7 1.9 1.5 1.2

200 2.0 2.1 1.7 1.3
300 300 2.4 2.2 1.8 1.3

Figure 9. Comparison of relative density–qc relationships by Table 2. The ratio of qc measured in Villet and Mitchell (1981) to
Schmertmann (1975, 1976) and Villet and Mitchell (1981) qc predicted by Schmertmann (1978) (after Villet and Mitchell)

204
Ground Improvement Relative density correlations are not
Volume 166 Issue GI4 reliable criteria
Hamidi, Varaksin and Nikraz

Baldi et al. then refer to earlier work by themselves and define ó9 density in overconsolidated sands becomes less reliable because
as the effective vertical stress if the sand is normally consolidated of the inherent difficulties in proper assessment of effective
or as the effective horizontal stress or effective mean stress if the horizontal stresses for improved soils.
soil is overconsolidated. This was based on the fact that using
effective vertical stresses in overconsolidated sands led to an Relative density correlations are even more unreliable when it
overestimation of relative density, as also observed for SPT. comes to calcareous sands. Almeida et al. (1992) carried out
calibration chamber tests on the calcareous Quiou sand and
Although Baldi et al. propose experimental coefficients for ten concluded that for the same relative density, cone resistance in
cases of normally consolidated, overconsolidated and normally/ the calcareous sand was well up to half the value of qc measured
overconsolidated conditions for both Ticino sand and Hokksund in the silica Ticino sand. The observed trend in the differences
sand, for some unknown reason, the normally consolidated Ticino was greater for higher relative densities.
sand has become the better known correlation, as expressed in
Equation 14, and what will generally appear in one form or the More recently, Al-Homoud and Wehr (2006) report that, based on
other when relative density correlations are used as ground correlation charts of Robertson and Campanella (1985), a relative
improvement acceptance criteria. density of 60% was required for land reclamation projects (Palm
Jumeira) in Dubai. (The authors note that there are no such charts
 
1 qc in Robertson and Campanella’s cited publication and Al-Homoud
Dd ¼ : ln :
14: 2 41 157ó v90 55 and Wehr have probably made a mistake in their reference.) Due
to the difficulties in achieving the requested penetration resistance
in some zones of the compacted fill, it was felt necessary to
Had normally consolidated Hokksund sand gained fame, the verify whether silica sand-based correlation was equally applic-
expression would have had to be as shown in Equation 15. The able to calcareous sand.
difference between predicted relative density values using Equa-
tions 14 and 15 increases with effective vertical stress and the Al-Homoud and Wehr refer to the unpublished work of Gudehus
reciprocal of cone resistance and can be more than 20%. This is and Cudmani who had performed calibration chamber tests on
yet another example of the fact that relative density correlations Dubai’s calcareous sand and Karlruhe’s quartz sand. According to
are not unique and are dependent on soil type. Al-Homoud and Wehr the calibration chamber diameter and
height were respectively 0.95 m and 1.5 m. The diameter of the
 
1 qc CPT rod used in the test was 36 mm. Without entering into a
Dd ¼ : ln
15: 3 29 86ó v90:53 detailed discussion, Al-Homoud and Wehr state that a shell
correlation (correction) factor of 1.5 for depths greater than 8 m,
1.6 for depths of 4 to 8 m, and 1.7 for depths less than 4 m must
In line with the above, Jamiolkowski et al. (2001) have proposed be applied to Dubai sand. According to this study, these correla-
Equation 16 using Ticino sand, Hukksund sand and Toyoura sand. tion factors should be seen as the lower limit and conservative for
Here, the data scatter is more for each of the Ticino sand or calcareous sand if the material in the field is much coarser than
Hukksund sand equations. Equation 16 yields the lowest estimate the soil fractions used in the experiments since larger shells crush
of relative density as compared with Equations 14 and 15 in most more easily than very small ones. They also note that the
cases, except for loose sand at high effective vertical stresses. penetration resistance for Dubai sand was reported to be about
37% lower than that of Karlsruhe sand for a medium dense state.
" #
1 (qc =98:1) They also extend the best-fit exponential curves of Dubai sand to
Dd ¼ : ln : a number of other silica sands to derive a shell correlation factor
16: 3 10 17:68ðó v9 =98:1Þ0 50
fshell which is expressed in Equation 17

17: f shell ¼ 0:46Dd þ 1:3629

It should be noted that Jamiolkowski is well aware of the
unreliability of relative density as a criterion, and is merely
proposing expressions to better the estimate of relative density. Comparing Dubai sand with Karlsruhe sand demonstrates that, as
Jamiolkowski and Pasqualini (1992) note that quality control that previously mentioned, there is no unique relationship between
is based only on values of relative density can be insufficient to relative density and cone resistance. Similarly, the cone resistance
evaluate the ground modifications achieved by compaction. Even differences between Dubai sand and Karlsruhe sand does not
if effective horizontal stresses are used in the estimation expres- mean that the same differences would be observed between the
sion, they conclude that important factors such as compressibility, project criterion’s reference sand and Dubai sand. The same is
ageing and the presence of fines limit the use of correlations as a equally true for the expression that was derived in Equation 17.
guide for evaluating in situ density of clean, predominantly silica In fact, entering a relative density of 0.6 into Equation 17 will
sands of recent, uncemented deposits. Evaluation of relative yield a correlation factor of 1.6389 which, if used, would have

205
Ground Improvement Relative density correlations are not
Volume 166 Issue GI4 reliable criteria
Hamidi, Varaksin and Nikraz

resulted in about 10% overestimation of the relative density of 5. Conclusion

Dubai sand at depths greater than 8 m. In this paper, correlations of relative density to field tests have
been reviewed and it appears that these correlations do not have
Furthermore, the calibration chamber diameter to cone diameter the required reliability to be used as ground improvement
ratio in this testing programme was 26.4. This is almost half of acceptance criteria. It seems that users of relative density
the value that Parkin and Lunne (1982) propose for boundary correlations have more confidence in these relationships than the
effects to become negligible in normally consolidated dense researchers who have themselves developed them.
sands. With the small ratio that has been used by Al-Homoud and
Wehr, the difference between the calibration chamber test and the The relationships between relative density and field tests are not
project’s acceptance criteria may have simply been due to the unique and are strongly influenced by sand properties. Parameters
boundary effects. such as fines content, grain size, grading, grading curve shape,
grain shape, effective vertical or horizontal stresses, mineralogy,
compressibility, cementation, overconsolidation, age and crush-
4. Appropriate acceptance criteria ability all have significant impacts. This has been well observed
Hamidi et al. (2010a, 2010b) report a ground improvement in research and the wide scatter of data in the studies suggests
project whose initial acceptance criterion was based on rel- that proposed correlations may have a notable deviation from
ative density, but calculations were able to demonstrate that it even the data points that were used in the studies themselves.
was possible to meet the design criteria with a higher safety
factor without complying to the relative density criterion. In Comparison of studies performed by different researchers demon-
Nakilat Ship Repair Yard, a reclamation project that was part strates substantial differences between their estimation methods
of the expansion of Port of Ras Laffan in Qatar, the to the point that if all estimation equations are used, differences
specifications required relative density to be 60% based on could suggest that estimations were no better than wild guesses.
Baldi et al. (1986) with a correction factor of 1.94 applicable Fortunately, there is a much better way to develop a project’s
to the CPT cone resistance. Calculation of allowable bearing acceptance criteria, and that is to base acceptance directly on
capacity for the various treated soil profiles showed that even design criteria. This can be simply achieved by defining the
the worst-case profile would still result in a higher bearing design requirements (such as allowable bearing, settlement limits,
capacity than what was calculated for the 60% relative density liquefaction requirements, etc.) rather than stipulating relative
design profile. Likewise, calculations showed that for an density, or minimum test values.
assumed square footing subject to a load of 4000 kN equiva-
lent to a pressure of 200 kPa the relative density design curve REFERENCES
resulted in 40 to 80% more settlement than any of the treated Al-Homoud AS and Wehr W (2006) Experience of
soil profiles that had not satisfied the relative density specifi- vibrocompaction in calcareous sand of UAE. Journal of
cation. Geotechnical and Geological Engineering 24(3): 757–774.
Almeida MSS, Jamiolkowski M and Peterson RW (1992)
Hamidi et al. (2011) have studied the numerous ways that Preliminary result of CPT tests in calcareous Quiou sand.
acceptance criteria can be developed and have concluded that Calibration Chamber Testing: First International Symposium
the most appropriate method is to base acceptance criteria on on Calibration Chamber Testing (ISCCT1), Potsdam, NY, 28–
performance directly originating from design criterion without 29 June 1991, 41–53.
resorting to proxies and go-between parameters such as relative ASTM (1969) D 2049–6: Test method for relative density of
density. The best way to stipulate a design criterion is not by cohesionless soils. ASTM International, West Conshohocken,
developing a correlation that can be flawed or by developing a PA, USA.
performance specification that is based on a minimum test ASTM (2006a) D4253–00 (Reapproved 2006): Standard test
value or test curve, but rather to specify the very design methods for maximum index density and unit weight of soils
criterion or criteria. Minimum value specifications do not using a vibratory table 15. ASTM International, West
realistically take into account the mass behaviour of the ground, Conshohocken, PA, USA.
and while any results less than the minimum specified value ASTM (2006b) D4254–00 (Reapproved 2006): Standard test
can immediately lead to failure to reach the specifications, test methods for minimum index density and unit weight of soils
values above the requirements will literally go unaccounted for. and calculation of relative density 9. ASTM International,
If the requirement is bearing capacity, total or differential West Conshohocken, PA, USA.
settlement, creep, long-term settlement or liquefaction mitiga- Baldi G, Bellotti VN, Ghionna N, Jamiolkowski M and Pasqualini
tion, the best specification would logically be to specify the E (1986) Interpretation of CPT’s and CPTU’s – 2nd part:
design requirement. The minimum test value or test curve is Drained penetration of sands. Proceedings of the 4th
only one right solution and the most unlikely to occur among International Geotechnical Seminar Field Instrumentation
the thousands of probable acceptable soil profiles that can and in-situ Measurements, Nanyang Technological Institute,
develop after treatment. Singapore, 25–27 November 1986, 143–156.

206
Ground Improvement Relative density correlations are not
Volume 166 Issue GI4 reliable criteria
Hamidi, Varaksin and Nikraz

Bazaraa AR (1967) Use of Standard Penetration Tests for Geosynthetics, New Orleans, 25–28 February, 902–914,
Estimating Settlements of Shallow Foundations on Sand. PhD ASCE Geotechnical Special Publication No. 30. ASCE,
thesis, University of Illinois at Urbana, Urbana, Illinois, USA. Reston, VA, USA.
Bolton MD and Gui MW (1993) The Study of Relative Density Jamiolkowski MLO, Presti DCF and Manassero M (2001)
and Boundary Effects for Cone Penetration Tests in Evaluation of relative density and shear strength of sands
Centrifuge. Technical Report CUED/D-Soils TR 256. from CPT and DMT. Soil Behavior and Soft Ground
Cambridge University, Cambridge, UK. Construction. Geotechnical Special Publication No. 119.
Bureau of Reclamation (1960) Earth Manual. US Department of ASCE, Boston, USA, pp. 201–238.
the Interior, Denver, CO, USA. Lacroix Y and Horns HM (1973) Direct determination and indirect
Bureau of Reclamation (1998) Earth Manual, Part 1, 3rd edn. evaluation of relative density and its use on earthwork
US Department of the Interior, Denver, CO, USA. construction projects. Proceedings of a Conference on
Cubrinovski M and Ishihara K (1999) Empirical correlation Evaluation of Relative Density and its Role in Geotechnical
between SPT N-value and relative density for sandy soils. Projects Involving Cohesionless Soils: ASTM STP523-
Soils and Foundations, Japanese Society of Soil Mechanics EB.7744–1, Los Angeles, 25–30 June, 251–280.
and Foundation Engineering 39(5): 61–71. Liu QB and Lehane BM (2012) The influence of particle shape on
Gibbs KJ and Holtz WG (1957) Research on determining the the (centrifuge) cone penetration test (CPT). Géotechnique
density of sands by spoon penetration testing. Proceedings of 62(11): 973–984.
the 4th International Conference on Soil Mechanics and Marcuson WF III (1978) Determination of in situ density of sands.
Foundation Engineering, London, 35–39. Proceedings of a Conference on Dynamic Geotechnical
Haldar A and Tang WH (1979) Uncertainty analysis of relative Testing, ASTM STP 654, Denver, 28 June, 318–340.
density. Journal of Geotechnical Engineering, ASCE 107(7): Marcuson WF III and Bieganousky WA (1977) Laboratory
899–904. standard penetration tests on fine sands. Journal of
Hamidi B, Varaksin S and Nikraz H (2010a) Treatment of a Geotechnical Engineering, ASCE 103(6): 565–589.
hydraulically reclaimed port project by dynamic compaction. Mayne PW and Kulhawy FH (1982) K0 –OCR relationship in soil.
Proceedings of the 3rd International Conference on Journal of Geotechnical Engineering, ASCE 108(6): 851–
Problematic Soils (PS10), Adelaide, 7–9 April, 113–120. 872.
Hamidi B, Varaksin S and Nikraz H (2010b) Application of Meyerhof GG (1957) Discussion on research on determining the
dynamic compaction in Port of Ras Laffan expansion project. density of sands by spoon penetration testing. Proceedings of
Proceedings of the 6th Australasian Congress on Applied the 4th International Conference on Soil Mechanics and
Mechanics (ACAM6), Perth, Australia, 12–15 December, Foundation Engineering, London, 110.
Paper 1148. Mitchell JK and Gardner WS (1975) State of the art: in situ
Hamidi B, Nikraz H and Varaksin S (2011) Ground improvement measurement of volume change characteristics. Proceedings
acceptance criteria. Proceedings of the 14th Asian Regional of ASCE Conference on in situ Measurements of Soil
Conference on Soil Mechanics and Geotechnical Properties, Raleigh, North Carolina, pp. 279–345.
Engineering, Hong Kong, Paper No 404. Moretto O (1954) Subsoil investigation for a bridge over the
Hamidi B, Varaksin S and Nikraz H (2013) Relative density Parana river, Argentina. Géotechnique 4(4): 137–142.
concept is not a reliable criterion. Proceedings of the Osterberg JO and Varaksin S (1973) Determination of relative
Institution of Civil Engineers – Ground Improvement 166(2): density of sand below groundwater table. Proceedings of a
78–85. Conference on Evaluation of Relative Density and its Role in
Hatanaka M and Feng L (2006) Estimating relative density of Geotechnical Projects Involving Cohesionless Soils: ASTM
sandy soils. Soils and Foundations, Japanese Society of Soil STP523-EB.7744–1, Los Angeles, 25–30 June, 364–378.
Mechanics and Foundation Engineering 46(3): 299–313. Parkin AK (1977) The Friction Cone Penetrometer: Laboratory
Holtz RD (1973) Discussion on determination of relative density Calibration for the Prediction of Sand Properties. Norwegian
of sand below groundwater table. Proceedings of a Geotechnical Institute, Oslo, Norway, Internal Report No.
Conference on Evaluation of Relative Density and its Role in 52108–5.
Geotechnical Projects Involving Cohesionless Soils: ASTM Parkin AK and Lunne T (1982) Boundary effects in the
STP523-EB.7744–1, Los Angeles, 25–30 June, 376–377. laboratory calibration of a cone penetrometer for sand.
Holubec I and D’Appolonia E (1973) Effect of particle shape on Proceedings of the 2nd European Symposium on Penetration
the engineering properties of granular soils. Proceedings of a Testing, Amsterdam, 761–767.
Conference on Evaluation of Relative Density and its Role in Peck RB and Bazaraa AR (1969) Discussion of settlement of
Geotechnical Projects involving Cohesionless Soils: ASTM spread footings on sand. Journal of Soil Mechanics and
STP523-EB.7744–1, Los Angeles, 25–30 June, 304–318. Foundations Division, ASCE 95(3): 905–909.
Jamiolkowski M and Pasqualini E (1992) Compaction of granular Robertson PK and Campanella R (1985) Liquefaction potential of
soils – remarks on quality control. Proceedings of a sands using the CPT. Journal of Geotechnical Engineering,
Conference on Grouting, Soil Improvement and ASCE 111(3): 384–403.

207
Ground Improvement Relative density correlations are not
Volume 166 Issue GI4 reliable criteria
Hamidi, Varaksin and Nikraz

Schmertmann JH (1975) State of the art paper: measure of in situ liquefaction based on SPT-N value and fines content. Soils
strength. Proceedings of ASCE Conference on in situ and Foundations, Japanese Society of Soil Mechanics and
Measurements of Soil Properties, Raleigh, North Carolina, Foundation Engineering 23(4): 56–74.
pp. 57–138. Varaksin S (1970) Determination of Sand Density Below
Schmertmann JH (1976) An Updated Correlation between Groundwater Table. MSc thesis, Northwestern University at
Relative Density and Fugro-type Electric Cone Bearing qc . Evanston, IL, USA.
Waterways Experiment Station, Vicksburg, MS, USA, Vol. Villet WCB and Mitchell JK (1981) Cone resistance, relative
145, Contract report, DACW 38–76-M 6646. density and friction angle. Proceedings of a Symposium on
Schmertmann JH (1978) Guidelines for Cone Penetration Test, Cone Penetration Testing and Experience, ASCE National
Performance and Design. Federal Highway Administration, Convention, St Louis, October, pp. 178–208.
Washington, DC, USA, Vol. 145, Report FHWA-TS-78–209. Wu TH (1957) Relative density and shear strength of sands.
Skempton AW (1986) Standard penetration test procedure and Journal of the Soil Mechanics and Foundations Division,
the effects in sands of overburden pressure, relative density, ASCE 83(1): 1161–1123.
particle size, ageing and overconsolidation. Géotechnique Yoshimi Y and Tohno I (1973) Statistical significance of the
36(3): 425–447. relative density. Proceedings of a Conference on Evaluation
Tatsuoka F, Tokida K, Yasuda S et al. (1978) A method for of Relative Density and its Role in Geotechnical Projects
estimating undrained cyclic strength of sandy soils using involving Cohesionless Soils: ASTM STP523-EB.7744–1, Los
standard penetration resistance. Soils and Foundations, Angeles, 25–30 June, 74–84.
Japanese Society of Soil Mechanics and Foundation Youd TL (1973) Factors controlling maximum and minimum
Engineering 18(3): 43–58. densities of sands. Proceedings of a Conference on
Terzaghi K and Peck RB (1948) Soil Mechanics in Engineering Evaluation of Relative Density and its Role in Geotechnical
Practice. Wiley, New York, USA. Projects involving Cohesionless Soils: ASTM STP523-
Tokimatsu K and Yoshimi Y (1983) Empirical correlation of soil EB.7744–1, Los Angeles, 25–30 June, 98–112.

WHAT DO YOU THINK?

To discuss this paper, please email up to 500 words to the
editor at journals@ice.org.uk. Your contribution will be
forwarded to the author(s) for a reply and, if considered
appropriate by the editorial panel, will be published as a
discussion in a future issue of the journal.
Proceedings journals rely entirely on contributions sent in
by civil engineering professionals, academics and students.
Papers should be 2000–5000 words long (briefing papers
should be 1000–2000 words long), with adequate illustra-
tions and references. You can submit your paper online via
www.icevirtuallibrary.com/content/journals, where you
will also find detailed author guidelines.

208