Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/245294542

Determination of Atterberg Limits: Uncertainty and Implications

Article  in  Journal of Geotechnical and Geoenvironmental Engineering · March 2006


DOI: 10.1061/(ASCE)1090-0241(2006)132:3(420)

CITATIONS READS
6 1,511

1 author:

Alvaro Enrique Gutiérrez García


Universidad de la República de Uruguay
5 PUBLICATIONS   6 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Stabilité des murs de soutènement compte tenu des déplacements. 07/1988 - 09/1992 View project

All content following this page was uploaded by Alvaro Enrique Gutiérrez García on 30 June 2016.

The user has requested enhancement of the downloaded file.


TECHNICAL NOTES

Determination of Atterberg Limits:


Uncertainty and Implications
Alvaro Gutiérrez1

Abstract: Arthur Casagrande made one of the most important contributions to Geotechnical Engineering; ordering and presenting clearly
the existing differences between objectives for civil engineering soil classification and soil classification schemes intended for other
purposes. However, more than 50 years after the Unified System of Soil Classification 共USSC兲 was proposed, one of the main ideas
expressed by Casagrande: “the plasticity chart representation of the plasticity parameters in different soil moistures, belonging to a
common geological origin, is a straight line, parallel to the A line,” has received little scrutiny. The main purpose of this technical note is
to begin a revision of Casagrande’s proposal under a probabilistic approach, suggesting some modifications to the application of the
plasticity chart. Regression analysis is proposed as a valid technique to express the linear behavior of the Atterberg limits for a given soil.
The problem considering plasticity index as a probabilistic variable is exposed demonstrating that the correct representation of the
plasticity chart is in terms of liquid and plastic limits. Data from the Libertad–Dolores clays formations from Uruguay are presented,
demonstrating the application of the proposed changes.
DOI: 10.1061/共ASCE兲1090-0241共2006兲132:3共420兲
CE Database subject headings: Soil classification; Statistics; Standards; Quality control; Uncertainty principles.

Introduction relations, and basing them on established geotechnical engineer-


ing principles. In this paper we start from the beginning: the
Casagrande developed a basis for soil characterization by study- Atterberg limits and the Casagrande plasticity chart.
ing an important number of soil types from around the world,
evaluating their Atterberg limits and summarizing the plasticity
properties of the soils in the plasticity chart. Application of the Mathematical Error of Plasticity Chart
probabilistic approach to geotechnical engineering started in the
1960s. However, some gaps still exist with respect to updating the Casagrande developed a basis for soil characterization by study-
classical deterministic approach and utilizing probabilistic meth- ing an important number of soil types from around the world,
ods. Harr 共1997兲 and Magnan 共1982兲 demonstrate application of evaluating their Atterberg limits and summarizing the plasticity
probabilistic techniques; however, to date no soil-specific data properties of the soils in the plasticity chart based on the variables
have been produced rendering detailed correlations between soil 共WL,WP兲 resulting from independent tests, but using a represen-
parameters and engineering properties. The Bothkenar clay re- tation WL,IP, where IP= WL− WP.
search program, presented in Geotechnique (1992), refers to a Consider two random variables x and y, related by regression
complete set of geotechnical laboratory and field tests, but corre- analysis 共Canavos 1988兲
lations of soil parameters were not studied. The great number of y = a 0 + a 1x 共1兲
parameters involved makes its very difficult to define a compre-
hensive model relating observed soil parameters to soil behavior. Consider the variable changes: x⬘ = x , y ⬘ = x − y. In statistics, y ⬘ is
The spatial variability of soil parameters increases the complexity called a “spurious” variable: an artificial correlation by creating a
of the problem. new variable using a linear combination of basic ones.
A rigorous probabilistic approach is necessary to drastically In the case of the Atterberg limits, the rigorous representation
reduce the number of variables, by studying the accuracy of cor- would be: WL,WP 共x , y兲 instead of x⬘ = WL, y ⬘ = IP= WL− WP as
it was presented by Casagrande. The regression results with x⬘
1 and y ⬘ are
Professor, Dept. of Civil Engineering, Facultad de Ingenierı́a, Univ.
de la República, C. P. 11.300 Montevideo, Uruguay. E-mail: agutierr@ a 1⬘ = 1 − a1 共2兲
fing.edu.uy
Note. Discussion open until August 1, 2006. Separate discussions
must be submitted for individual papers. To extend the closing date by
a0⬘ = − a0 共3兲
one month, a written request must be filed with the ASCE Managing
Editor. The manuscript for this technical note was submitted for review 共1 − R2兲
and possible publication on January 9, 2004; approved on September 9,
1 − R ⬘2 = 2 共4兲
共␴x/␴y兲 − 2R␴x/␴ y + 1
2005. This technical note is part of the Journal of Geotechnical and
Geoenvironmental Engineering, Vol. 132, No. 3, March 1, 2006. From Eq. 共4兲, R⬘2 depends on the R and ␴WP / ␴WL values.
©ASCE, ISSN 1090-0241/2006/3-420–424/$25.00. Casagrande’s coordinates system 共WL,IP兲, displays a false reality:

420 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MARCH 2006


Fig. 1. Location of Libertad–Dolores formations in Uruguay and principal sites explored

R⬘2 values are different from the R2, calculated from the correct granulated, or lense forms, in a percentage from 2 to 4%. The
coordinates system 共WL,WP兲. Casagrande observed that all the formation colors are generally brown or beige. These sediments
soils he had studied could be represented by a straight line were deposed in a continental environment with semiarid climate
practically parallel to the A line 共a⬘1 = 0.73兲. However, Eqs. 共2兲 and and variable humidity. Its contacts are discordant at the base with
共4兲 show that the mathematical transformation to 共a⬘0 , a⬘1兲 is other formations and with the crystalline basement. Maximum
independent of the data involved, depending only on the variable thickness is estimated as 30 m.
changes of the coordinates system. Thus, the representation of the
A line in the new coordinates system is the straight line
Dolores Formation
WP= 14.6+ 0.27 WL.
The Dolores Formation is essentially clayey silt with sand and
gravel, and includes calcium carbonate in dispersion or granular
Libertad–Dolores Clays in Uruguay „Bossi et al. form. The formation was originated in a glacial quaternary event
1998… that occurred between periods of marine deposition. These depos-
its are also continental with conditions and mechanisms similar to
Silty clays from Libertad and Dolores formations cover an impor- the ones in Libertad.
tant geographic area in the south of Uruguay 共see Fig. 1兲 where
more than half of the population lives.
Methodology
Libertad Formation
The study was mainly developed with data collected by the
The Libertad formation is essentially composed of quartz and Geotechnical Dept. of the Engineering School, and other local
feldspar and is inundated by massive mudstones, with coarse laboratories, corresponding basically to three different sources:
sand, fine gravel, and gravel homogeneously distributed in a per- 1. “Area Piloto,” defined as a territorial area within Montevideo
centage lower than 1%. The dominant lithologies are sandy Dept. 共10 km2兲.
clayey silt, and clayey silt, while the presence of loess is less 2. A territorial strip of more than 100 km in length, correspond-
frequent. Carbonaceous matter can be present in powdered, ing to the foundation soil of the high tension electric line

Table 1. Regression Analysis Results


Mean ␴ COV
Site identification N Regression data 共%兲 共%兲 共%兲
W P =8.71+ 0.27 WL WL 37 10.4 28
Area Piloto 共1兲 68
R2 =0.74 WP 19 3.2 17
Mvdeo–San Carlos WP = 11.06+ 0.24 WL WL 44.5 12.1 27
107
high-tension line共2兲 R2 =0.61 WP 21.6 3.6 17
WP = 10.65+ 0.26 WL WL 48.6 12.2 25
COVICO 共3兲 52
R2 =0.81 WP 23.4 3.6 15
WL 29.9 2.7 9
CONAPROLE 共4兲 28 —
WP 19.1 1.6 9
WL 39.8 6.8 17
“26 de Octubre” 共5兲 28 —
WP 21.1 2.9 14
WP = 13.79+ 0.24 WL WL 52.2 12.5 24
Montevideo A 共6兲 32
R2 =0.58 WP 26.4 3.8 14
W P =9.76+ 0.28 WL WL 43.6 12.4 28
General model 417
R2 =0.69 WP 22.0 4.2 19

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MARCH 2006 / 421


Table 2. Comparison with Other Soils from World Quality Control on Laboratory Tests
Identification N Mean ␴ COV 共%兲
Different local laboratories provided the data presented in this
WL 55.1 12.2 22.1
Schultze 共1972兲 75 paper. It is interesting to compare the statistic parameters obtained
WP 21.7 3.4 15.7 关mean value, standard deviation, coefficient of variation 共COV兲,
WL 104 27 26.0 trend line兴 with those obtained in other soils. Subsequently,
Baghery 共1980兲 163
WP 47.3 10.9 23.0 the dispersion between results from one operator and from the
Leman Lake clay WL 43.0 12.8 29.8 different local laboratories is evaluated.
36
共Dysli 1976兲 WP 17.5 7.8 44.6
WL 43.6 12.4 28.4
Libertad–Dolores clays 417 Comparison with Results from Other Soils
WP 22.0 4.2 19.1
Casagrande did not present the equations of the straight lines
representing the different soils studied by him, preferring in some
cases to represent them with a zone of the plasticity chart instead
between Montevideo and San Carlos cities; and of the straight line representation. Subsequently, this theme was
3. Specific sites of approximately 10,000 m2, located in not often considered by other researchers. As a consequence, only
Montevideo Dept.: a limited number of correlations between WL and WP are avail-
• COVICO Cooperative buildings; able in the literature 共see Table 2兲. It can be seen that Schultze’s
• Experimental site “Estación Montevideo A-UTE. Cno. 共1972兲 and Dysli’s 共1976兲 parameter results are close to those
Paso del Sauce”; obtained in Uruguayan Laboratories, while Bagherys 共1980兲
• Industrial complex “Conaprole, Las Higueritas” 共DNV results are higher for plastic and liquid limits.
Montevideo laboratory兲; and Casagrande found, for American soils, the following relation-
• Cooperative town “26 de Octubre 共Millán y Garzón兲” ship Costet and Sanglerat 共1975兲:
共Lemac Laboratory兲.
PI = a WL − b with 0.70 ⬍ a ⬍ 0.80 and 13 ⬍ b ⬍ 17 共6兲
Regression Analysis Results Also, according to Costet and Sanglerat 共1975兲 for French and
Each source was studied separately with a specific regression Spanish soils
analysis 共Table 1兲. All tests were executed before 2000, using
essentially the “three-point” method and the dry procedure. It is PI = 0.70 WL − 9 共7兲
possible then to obtain a general straight model to relate WL with 2
R =not provided or
WP,
WP = 0.30 WL + 9
WP = 9.76 + 0.28 WL 共5兲
The formula for Libertad–Dolores clays becomes
with R2 = 0.69. Applying “F” to the regression model and “t” sta-
tistic tests to the “a” and “b” coefficients indicates that the model PI = 0.71 WL − 9.76 R⬘2 = 0.93 共8兲
is a reasonable representation of the actual soil. 关Statistical tests
or
used to verify the accordance between the model and data, F
Snédecor test qualifies the model, t test verify the nonnullity of
WP = 0.28 WL + 9.76 R2 = 0.69
the terms “a” and “b” 共null hypothesis: a or b = 0, results for a
significance degree of 0.05.兴 and for Lac Leman clays

Table 3. Uruguayan Laboratories WL, WP Values versus Depth Results


Prof. DNV DNV Geot. Geot. cov
共m兲 Mont. Salto Lemac I II Geoproyectos Mean ␴ 共%兲
1.0 42 37 32 40 37 53 40.2 7.1 17.8
1.5 39 38 32 41 35 51 39.3 6.5 16.6
2.0 37 31 34 36 34 48 36.7 5.9 16.2
WL 2.5 42 38 35 41 37 47 40.0 4.3 10.7
3.0 44 43 40 43 38 53 43.5 5.2 11.9
3.5 51 42 45 52 48 66 50.7 8.4 16.5
4.0 54 43 46 52 47 55 49.5 4.8 9.8
1.0 23 28 17 22 20 21 23.2 4.3 18.4
1.5 23 26 18 20 19 21 22.3 3.3 14.6
2.0 22 23 19 18 19 21 21.0 2.1 10.0
WP 2.5 19 24 19 20 17 22 21.3 2.3 11.0
3.0 20 26 19 22 19 23 22.7 2.9 13.0
3.5 26 28 20 23 21 21 24.3 3.5 14.4
4.0 24 26 20 23 22 20 23.2 2.7 11.7

422 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MARCH 2006


Fig. 2. Chart of plasticity for Libertad–Dolores clays, WL - WP representation 共left兲, WL - IP classical representation from Casagrande 共right兲

PI = 0.4 WL + 8.3 R⬘2 = 0.92 共9兲 Quality Control Tests and Rejection Criteria Proposal
Prediction intervals and a spindle can be defined, beyond which
tests results may be rejected, because the sample tested does not
or belong to the formations mentioned, or the Atterberg test has not
been carried out correctly. For a given value of WP0, and a con-
fidence level of 99%, the boundary WL0 curves of the spindle are
WP = 0.60 WL − 8.3 R2 = 0.96 shown in Fig. 2.
The straight line model and its spindle provides the opportu-
nity for quality control of test data. If a value is accepted as valid,
Except for Lac Leman clays, all cases considered confirm that a⬘1 it is added to the data bank, redefining the straight line and the
is near to the 0.73 A line coefficient. Due to the little and incom- spindle. This feedback mechanism increases the model perfor-
plete statistical information provided by the literature, it is not mance. As can be noticed from Fig. 2 共left兲, eight points were
possible to achieve a more in-depth analysis At this time it may rejected 共indicated by a circle兲.
only be observed that the results for the Libertad–Delores clay are A straight line parallel to the A line can represent this soil on
similar to those reported by others. the IP-WL representation in agreement with Casagrande 共Casa-
grande 1947兲. A higher correlation coefficient is obtained from
Casagrande coordinates system 关Fig. 1 共right兲兴 compared with the
Comparison of Results among Local Laboratories correct coordinates system 关Fig. 1 共left兲兴.

To estimate the internal dispersion of the tests results, an operator


from the Geotechnical Dept. made nine tests using a 5 kg sample, Conclusions
taken from the “Area piloto.” The mean value, standard deviation,
and COV are 共55.2, 1.7, 3.1兲 and 共24.0, 0.8, 3.4兲 for WL and WP, At this stage of the study, it is clear that a straight line can rep-
resent the Atterberg limits of soils from the geological Libertad
respectively.
and Dolores formations. The model is not static, and new values
Soil samples from different depths at the UTE site were ex-
will be incorporated defining a new equation and spindle, as in
tracted in order to compare results obtained by several leading the following proposed procedure:
laboratories in Uruguay 共Table 3兲. Similar results of Atterberg 1. Make a large number of Atterberg limit tests on a soil from a
limits with depth were observed in all laboratories. However, geological formation;
DNV Salto has a tendency to obtain higher values of WP. 2. Plot the results in a WL-WP coordinates system and define
The D4318-00 ASTM standard 共ASTM 1984兲 accepts for a the characteristic trend line and its R2;
single operator a COV between 1.3 and 2.4 for plastic limit and 3. Choose the rejected points and define the spindle without
between 0.9 and 1.8 for liquid limit. For a multilaboratory deter- considering them;
mination, the COV must be between 0.9 and 2.0 for plastic limit 4. Recalculate the trend line and its R2 with the acceptable
and between 1.0 and 1.3 for liquid limit. This implies that points to define the general model. To consider more data
Uruguayan results are in order but not close to the ASTM stan- from a new site as from the same geotechnical “formation”;
dards. However, no specification is provided by the standard 5. Calculate the specific characteristic trend line of the new site
about number of tests and size of the sample considered. and its R2 and compare it with the general model. If the
specific and the general trend are close, it can be assumed
All the Atterberg test results presented for the Libertad–
that they come from the same geotechnical “formation”.
Dolores clays were performed with the dry procedure before the
Then, recalculate the characteristic trend line and its R2 with
ASTM 2000 standards modification, where only the wet method the acceptable points to redefine the general model.
is valid. This is another possible explanation for the high disper- This procedure could be extended to determine correlations
sion of the data compared to the standards. A survey of local among other geotechnical variables and parameters.
laboratories to verify the procedures employed would be done and Casagrande’s Chart with a WL-WP representation is defini-
more specific information concerning the number of tests, size of tively a basis for a geotechnical identification and the point of
sample, and number of operators should be included in future departure if the goal is to achieve a correct statistical analysis of
editions of the standards. data.

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MARCH 2006 / 423


Notation soils.” D4318, Philadelphia, 573–583.
Baghery, S. 共1980兲. “Probabilités et statistiques en mécanique des sols.
The following symbols are used in this technical note: Analyse probabiliste de la stabilité des tassements de remblais sur sols
compressibles.” PhD thesis, Ecole Nationale des Ponts et Chaussées,
Cov 共x , y兲 ⫽ covariance between two variables x , y; Paris.
n ⫽ number of tests; Bossi, J. et al. 共1998兲. Memoria explicativa de la carta geológica del
PI ⫽ plasticity index; Uruguay a escala 1:500.000.
R ⫽ correlation coefficient; Canavos G. C. 共1988兲. Applied probability and statistical methods,
Var共x兲 , Var共y兲 McGraw–Hill, New York.
⫽ variance of variables x , y, respectively; Casagrande A. 共1947兲. “Classification and identification of soils.” Proc.
WL,WP ⫽ Atterberg limits of liquidity and plasticity, Am. Soc. Civ. Eng., 901–991.
respectively; Costet, J., and Sanglerat, G. 共1975兲. Curso práctico de mecánica de sue-
WL0 , WP0 los, Omega, Barcelona, Spain.
⫽ Atterberg limits of liquidity and plasticity Dysli, M. 共1976兲. “Gazoduc suisse romand. Aspects géologiques et géo-
points defining boundaries of acceptability techniques du tracé souslacustre.” Communication du Laboratoire de
region 共spindle兲; and Géotechnique No. 37, EPFL, Lausanne, Switzerland.
␴x , ␴y , ␴WL , ␴WP Géotechnique. 共1992兲. “Eighth Geotechnique Symp. in print. Bothkenar
⫽ standard deviation from variables x , y, soft clay test site: Characterization and lessons learned.” Proc. Geo-
technique 8th, Symp., XLII共2兲 .
WL,WP, respectively.
Harr, M. E. 共1997兲. Reliability based design on civil engineering, Dover,
New York.
Magnan, J. P. 共1982兲. Les méthodes statistiques et probabilistes en méca-
References nique des sols, Presses de l’ École Nationale des Ponts et Chausées,
France.
American Society for Testing and Materials 共ASTM兲. 共1984兲. “Standard Schultze, E. 共1972兲. “Frequency distributions and correlation of soil prop-
test methods for liquid limit, plastic limit, and plasticity index of erties.” Proc. 1st ICASP, Hong Kong, 371–388.

424 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MARCH 2006


View publication stats

You might also like