Professional Documents
Culture Documents
is the assumption that p-*-oo when <j-*oo, while experi- The Model for the Hydraulic Properties
mental data indicates that soils can be compacted up to of the Compacted Soil
a finite maximal bulk density, depending on the specific The soil hydraulic properties consist of the WRC, which
soil and its initial water content (Amir et al., 1976; describes the relationships between the volumetric water con-
Faure, 1981). The second objective of our study is, tent, 0, and the matric potential V|/, and the hydraulic conductiv-
therefore, to present a more simple two-parameter model ity function which relates 0 to the hydraulic conductivity, K.
for p(o), based on a physical concept and satisfying a According to the Brooks and Corey (1964) model, the WRC
different boundary condition for a very high stress. is expressed by
As stated earlier, compaction affects soil hydraulic \-\.
J/a [9]
properties (Gill and Vanden Berg, 1967). Experimental
data relating the effect of the increase in the soil bulk > V|/a [10]
density on soil hydraulic properties is quite scarce. The with
trends observed are that soil compaction (i) decreases
the saturated water content and increases the air entry & = (6 - er)/(e, - er) [ii]
value (Croney and Coleman, 1954; Laliberte et al. , 1966; where 0S and 0r are the saturated and residual water contents,
Smith and Woolhiser, 1971; Libardi et al., 1982) and \l/a is the air entry value, and X is the pore-size distribution
(ii) decreases the saturated hydraulic conductivity (Dawi- index. The saturated water content, 0S, is generally assumed
dowsky and Koolen, 1987) and consequently, the soil to be equal or very close to the soil porosity. The residual
infiltrability (Akram and Kemper, 1979). water content, 0r, is defined as the water content at which the
water capacity C(n/) = d0/dv|/-K) and the soil water conductiv-
Approaches towards modeling the effect of the increase ity K(Qr) = 0. In some studies, it is defined as the water
in the soil bulk density on soil hydraulic properties content at the wilting point, identified in practice by \|/ =
are very limited (Baumhardt et al., 1990; Mualem and -1500 kPa (Rogowski, 1972; van Genuchten and Nielsen,
Assouline, 1989). They were developed to infer hydraulic 1985). Usually, it is regarded as an additional fitting parameter.
properties to the seal layer formed at the soil surface by The constants v|/a and X are also fitting parameters.
the action of raindrops. The advantage of the approach The unsaturated hydraulic conductivity function of the soil,
of Mualem and Assouline (1989) is that it is conceptual K(Se), can be defined, in terms of Mualem's (1976) model:
and addresses both the soil water retention curve and t>vc\ _ v cn+2+2/X
f*-\^e) — ^s'-^e
MOI
L^^J
the hydraulic conductivity function. The third objective
where Ks is the soil-saturated hydraulic conductivity, and n,
of this study is to apply this approach to modeling the a parameter accounting for the correlation between pores and
effect of compaction on the soil water retention curve the flow path tortuosity. Considering data from 45 soils, Mua-
and the hydraulic conductivity function. lem (1976) suggested that the best value for n might be 0.5.
When a compressive pressure P is applied to a homogeneous
soil of initial bulk density Pi at an initial water content 0i,
THEORY leading to a bulk density p(P,6i), the WRC of the compacted
The Dynamic Model of Compaction soil is
-Xc
Consider a homogeneous nonoverconsolidated soil volume (6sc - 0rc) (V|//V|/ac) [13]
of initial bulk density pi at a given volumetric water content where the subscript c denotes the new parameters which charac-
0j. One may expect that, as the soil is progressively compacted, terize the compacted state. According to Mualem and Assouline
it will be more difficult to compact it further. In other words, (1989), the different parameters for the compacted state can
the higher the soil bulk density, the smaller the increase of be denned in terms of p(P,Q\):
the bulk density, Ap, resulting from the increase of the com- (i) Considering that the volumetric water content at saturation
pressive pressure applied, AP. This can be expressed through of compacted soils is equal to the porosity, the saturated water
the linearly decreasing function content is
dp/dP = Ti(0,) - [4] 0SC = 1 - p(P,6i)/p, [14]
where r\ and £ are parameters dependent on the specific soil where ps is the solid particle density.
and water content. For the initial and boundary conditions (ii) Based on the data of Laliberte et al. (1966) and Smith
and Woolhiser (1971), the air entry value is given by the
P = 0; p = Pi(0i) [5] relationships
[6] Vac = V|/aH[p(/ ) ,0i)/Pi] P [15]
the solution of Eq. [4] becomes where |x and (3 are positive constants, found to be equal to
0.99 and 3.72, respectively.
(iii) The residual water content is considered to be mainly
[7] a function of the surface area of the soil particles and, thus,
where pmax(6i) is the highest bulk density reachable at the to be practically not affected by the soil compaction when
specific 0i. For each soil type, a specific water content permits expressed on a weight basis. This leads to the relation between
compacting the soil to a maximal bulk density. Denoting, in volumetric water content and bulk density:
that case, the specific values of pmax(0i) and £(6i) by piL, and 0rc = 6r [p(P,0i)/pJ [16]
£*, Eq. [7] becomes
(iv) Based on experimental data, the pore-size distribution
= pi+ (p*» - Pi)d - [8] index appears to decrease with increasing bulk density. How-
392 SOIL SCI. SOC. AM. J., VOL. 61, MARCH-APRIL 1997
ever, no quantitative relationship describing this relationship the samples at y, values of -1.0, -3.2, -10.0, -32.0,
is available. Therefore, it is assumed, as a first approximation, -100.0, -320.0, and -1000.0 kPa. Soil aggregates from
that this index decreases linearly with the change in bulk the samples at »|/j = —3.2 kPa were embedded in an epoxy
density: resin (Tessier, 1984). After hardening, thin sections were cut
with a diamond knife. Electron micrographs at low (lOOOOx)
magnification were made using a Philips (Model 420, Eindho-
where y is a positive constant dependent on the specific soil. ven, the Netherlands) transmission electron microscope, to
Another widely used model for the WRC is the van Genuch- characterize the soil constituents. A series of one-dimensional
ten (1980) relationship: compressive pressures was applied to the samples that had
been pre-equilibrated at the different \|/i. A piston was installed
Se(V|/) = 1/[1 + (a\\V\b}m [18] in the filtration apparatus, and pressures of 50, 100, 150, 200,
where a, b, and m are fitting parameters. In the particular 250, 300, 400, 500, 600, 800, and 1000 kPa were applied.
case where m is related to b by the equation m = 1 - I/b, Few minutes were generally necessary to reach equilibrium.
a = \|/a~' and b = K + 1. As a result, the relationships The volume change of the sample was evaluated by measuring
proposed above are also directly applicable when the van the piston displacement. Water was allowed to drain out of
Genuchten model is preferred. the filtration apparatus during the compression experiments.
Once the WRC is defined for the compacted state, the After equilibrium was reached, and when no drainage was
hydraulic conductivity at saturation, Kx, can be estimated by observed, the samples were extracted from the apparatus.
applyifig principles similar to those suggested by Mualem Water retention curves and saturated hydraulic conductivity
(1986). This leads to were measured on the samples pre-equilibrated at \\i, = -32.0
kPa, which had been compacted at 1000 kPa pressure. The
n+2 main drying curves of the saturated samples were obtained
6sc — 6rc using pressure cells for matric tensions up to —1.6 MPa.
[19J
e s -e r Dessicators with a range of relative humidity between 95 and
50% were used for matric tensions between —2 and —100
MPa. The main wetting curves were obtained using dry samples
Substitution of Eq. [13] into Eq. [19] yields, after integration at 50% relative humidity and applying the sample procedure
«_!_*> 1 T as for the drying curves but with increasing matric potentials.
_— Ac Volume change of the samples at the different matric potential
Asc
e s -e r steps of the WRC were measured according to the method
presented in Tessier and Berrier (1979), to account for possible
The unsaturated hydraulic conductivity of the compacted soil swelling and shrinking of the samples. As a result, the specific
can be defined as in Eq. [12]. The relationships presented bulk density of each sample at every step was monitored,
above define the hydraulic properties of the compacted soil and the volumetric water content corresponding to the matric
both in saturated and unsaturated conditions and express them potential at each step determined with accuracy.
in terms of properties of the initial uncompacted soil. The saturated hydraulic conductivity, K^, of the compacted
samples was measured using constant-head permeameters.
MATERIALS AND METHODS All the compression experiments as well as the WRC and
Ks measurements were taken in 10 replicates.
The two soil types were a Haplic Acrothox from the site The fitting procedure used was an iterative nonlinear regres-
of Cascavel and a Haplic Eutrothox from the site of Palotina. sion using the Marquardt-Levenberg algorithm to find the
Comparative studies in the region have shown that the B values of the parameters of the independent variable that gives,
horizon (120-150 cm deep) was not affected by external me- the best fit between the model and the data, i.e., that minimize
chanical stresses from tillage or traffic (Tavares-Filho, 1995). It the mean square errors between the observed and predicted
could represent the soil properties before intensive cultivation. values of the dependent variable (Glantz and Slinker, 1990).
Therefore, the 120 to 150-cm depth zone of each soil was The square root of this mean is defined as the norm index,
selected to study its compaction behavior. Some characteristics which is an indicator of the goodness of the fit reached.
of the two soils are given in Table 1. Undisturbed blocks, of
= 5000 cm3, were collected in the field. They were put into
plastic bags to preserve their humidity and kept in a refrigerator RESULTS AND DISCUSSION
( = 4°C) before treatment. Coarse fragments of the blocks were
gently fragmented and sieved. The < 5-mm fraction was kept, The Compaction Behavior of the Two Soils
and samples of 40 cm3 were placed in a filtration apparatus for The relationships between the final bulk density (p)
water retention measurements and for mechanical compaction
(Sala and Tessier, 1993). The samples were fully rehydrated and initial matric pressure (\|/i) are depicted in Fig. la
and brought to equilibrium at various matric potentials, \\i,. and Ib for four different compressive pressures. The
For this purpose, gas pressure was applied to the filtration initial bulk density of the two soils is approximately the
apparatus and selected filter pore sizes were used to prepare same. In spite of the high amount of clay, shrinking
Table 1. The physical and chemical properties of the soils of Cascavel and Palotina.
Soils Clay Silt Sand CECt PH OMt W P,t
_, meqlOOg-' m s~' Mgm-3
Cascavel 0.81 0.17 0.02 4.6 4.9 8.0 4 x 10-' 2.88
Palotina 0.83 0.11 0.06 6.6 6.5 4.0 3 x 10-' 2.94
t CEC = cation-exchange capacity; OM = organic matter; K, = saturated hydraulic property; p, = solid particle density.
ASSOULINE ET AL.: EXPERIMENTAL RESULTS AND MODELING OF COMPACTION EFFECT ON SOIL 393
are achieved when \|/j = -10.0 kPa for 100 kPa < P
< 300 kPa, and when \\i, = -32.0 kPa, for P > 400
kPa. It is interesting to note the similarity in the relative
compactibility of these two soils when \\i\ = —1.0 and
-1000.0 kPa and when \\it = -3.2 and -100.0 kPa.
Beyond the similarity in textural analysis, the two soils
present differences in physico-chemical properties, which
are known to affect soil stability.
The two soils differ in particle thickness and crystallin-
10.0 100.0 1000.0 ity of the respective clay fractions. Electron micrographs
-(INITIAL MATRIC POTENTIAL) (kPa) of thin sections of the two soils (Fig. 3) show that the
Fig. 1. Bulk densities obtained at different initial matric potentials, clay particles, including kaolinite and oxides, are coarser
vj/i, for four of the eleven compressive pressures applied: (a) Palotina in the Palotina soil (A) than in the Cascavel soil (B). It
soil, (b) Cascavel soil. The dashed lines represent the bulk density has been shown that the thinner the clay particles, the
of the uncompacted samples at each \|>i. larger the surface area in contact between soil constit-
uents, thus inferring a higher stability to the fabric (Tes-
resulting from drying is limited and only slight changes sier, 1991; Van Damme and Ben Ohoud, 1989).
in p are measured with the decrease in \|/j. The soils also differ in pH, CEC, and OM content
For the lowest pressure applied (50 kPa), the same (Table 1). As shown by Guerif and Faure (1979) and
compaction behavior is observed for both soils. The O'Sullivan (1992), the presence of OM decreases the
highest p is achieved for V|/i = — 1.0 kPa, in other words, soil compaction sensitivity to initial water content and
when the soil is saturated. For the two intermediate decreases the bulk density reached after compaction.
pressures presented (200 and 500 kPa), a different behav- The pH also affects the structure stability of strongly
ior is observed for each soil. For the Palotina soil, the weathered soils such as Oxisols where kaolinite and
highest p at 200 kPa applied pressure is achieved for oxides are dominant (El Swaifi, 1980; Me Bride, 1989;
v|/j = -10.0 kPa, and the highest p at 500 kPa is achieved Schwertmann and Taylor, 1989). Under the acidic condi-
for \j/j = —32.0 kPa, demonstrating clear effect of V|/i tions of the Cascavel soil, the surfaces of iron oxides
on p. On the contrary, for the Cascavel soil, the highest are mainly positively charged, while kaolinite surfaces
p is still achieved in both cases for \\i\ = —1.0 kPa, are negatively charged. The resulting attraction forces
with practically no effect of \\i, on p, especially in the between the oppositely charged soil constituents impart
P = 500 kPa case. For the highest pressure applied some physical stability to the clay aggregates. By con-
(1000 kPa), the maximal p of the two soils is obtained trast, in die Palotina soil, where pH is close to neutrality,
when \i/j = -32.0 kPa. However, the bulk density iron oxides have very low charges and therefore a lower
achieved in the Palotina soil (~ 1.55 Mg m~3) is higher stability is obtained.
than that of the Cascavel soil (~ 1.30 Mg m~3). In Fig. Also, because of the low pH of the Cascavel soil, the
2, the results obtained for all compressive pressures presence of free aluminum hydroxides was only observed
applied at each \\it are presented in terms of ratio between in this soil (Fig. 3). Free aluminum hydroxides act as
the bulk density of the Palotina soil and that of the a ligand and are more effective than iron oxides in
Cascavel soil. For the whole range of pressures, the maintaining the stability of soil aggregates (El Swaifi and
ratio is greater than one, indicating that for any given Emerson, 1975). The effectiveness of the free aluminum
W and P, the Palotina soil reaches a higher bulk density hydroxides and the iron oxides in stability is increased
than the Cascavel soil. The highest ratios, around 1.2, by the presence of OM (Edwards and Bremner, 1967).
394 SOIL SCI. SOC. AM. J., VOL. 61, MARCH-APRIL 1997
Fig. 3. Transmission electron micrographs of thin sections of the two uncompacted soils: (a) Palotina, (b) Cascavel (black surfaces represent
particles; arrows indicate aluminum hydroxides).
ASSOULINE ET AL.: EXPERIMENTAL RESULTS AND MODELING OF COMPACTION EFFECT ON SOIL 395
All these differences in soil composition, which deter- Table 2. The parameters, and the corresponding norm index,
resulting from fitting the three compaction models (Eq. [1],
mine the chemical bonding strength, are related to the [2], and [8]) to measured bulk density resulting from the applica-
sensibility of compressibility to the initial matric pres- tion of increasing compressive pressures to samples initially at
sure. They provide potential explanations for the higher y, = -32kPa.
compactibility observed in the Palotina soil. It is worth Soil type Model type Parameter 1 Parameter 2 Parameter 3 Norm
noting that, when differences in compaction behavior 3
Palotina Multiplicative a = -0.86 b = - 1.26 10-" <: = 7.010- 0.06
resulting from intensive cultivation are considered, the Logarithmic m = -0.35 d = 1.67 0.39
practical aspect of the sensitivity to initial matric pressure Proposed p,™,* = 1.57 tf i- = 4.1 10-' 0.04
3
can play a preponderant role. Cultivation practices are Cascavel Multiplicative a = -0.70 6 = -1.40 10-" <: = 6.110- 0.02
usually carried out only when the upper soil layer is dry Logarithmic m — —0.36 d = 1.85 0.37
Proposed Pm** = 1.33 !;"' = 3.5 10-3 0.02
enough to allow the use of heavy machines, that is, when
the matric potential in this layer reaches values between
—10 and —30 kPa. This is precisely the range of matric conditions can predict the maximal bulk density reached
potentials where the compaction of the Palotina soil is in the field (O'Sullivan, 1992; Hartge, 1986). In our
maximal within a wide range of compressive pressures case, the bulk densities obtained by the application of a
(Fig. 2). compressive pressure of 1000 kPa to samples at v|/i =
— 32 kPa are similar to the respective bulk densities
measured in the field in the upper layer of the cultivated
The Dynamic Change of the Soil Bulk Density soils. In such a case, p*ax can be evaluated independently
during Compaction so that only one parameter, £*, remains to be determined
The measured bulk densities obtained at increasing P, to define the dynamics of the compaction of the specific
for the samples initially at V|/, = —32 kPa, are depicted in soil studied.
Fig. 4. The fitted curves corresponding to the logarithmic The Effect of Compaction on the Water
model of Bailey and Vanden Berg (1967) and Larson et Retention Curve
al. (1980) (Eq. [1]), the multiplicative model of Bailey
et al. (1986) (Eq. [2]), and the proposed model (Eq. The main drying and wetting curves of the uncom-
[8]) are also presented. The values of the fitting parame- pacted and the compacted soils of Palotina and Cascavel
ters and of the corresponding norm index are given in are presented in Fig. 5a and 5b.
Table 2. The logarithmic model, which represents the
measured data very roughly, leads to the less satisfactory
results. The multiplicative model and the proposed model
have similar performances. However, on a conceptual
basis, the multiplicative model assumes that p-*oo with
an exponentially increasing rate when P-*oo, which does
not occur experimentally, while the proposed model
yields the experimentally correct value p = pmax(6i) as
P-*oo. Thus, one of the two parameters of the proposed
model [Pmax(6i)] has a physical meaning, effectively trans- - - 0 - INITIAL-drying
forming the model into an expression with only one —•—COMPAC - drying
empirical fitting parameter. The multiplicative model, • - Q- - INITIAL - wetting
on the other hand, requires three fitting parameters. It —•—COMPAC - wetting
has been shown that compression tests under laboratory
• - O - INITIAL - drying
—•—COMPAC - drying
• - n- - INITIAL - wetting
—•—COMPAC - wetting
Table 3. The parameters resulting from fitting the Brooks and Table 4. The parameters of the Brooks and Corey relationship
Corey relationship to the data of the main drying curve for for the main drying curve of the compacted soils (at P = 1000
the uncompacted soils.t kPa), computed using the relationships in Eq. [14] to [17].t
Soils Pi <t> es e, V, X Soils Pe ere Vac X,
3 3
Mg m~ 3 kPa Mg m- 3
-m m - kPa
Cascavel 0.96 0.67 0.58 0.33 -2.2 0.61 Cascavel 1.27 0.55 0.55 0.43 -6.2 0.46
Palotina 1.25 0.57 0.49 0.32 -2.5 0.45 Palotina 1.50 0.49 0.49 0.39 -4.9 0.38
t Pi = initial bulk density; <t> = soil porosity; O, = saturated water content; t PC = bulk density of compacted soil; <t>c = porosity of compacted soil;
Or = residual water content; i|/a = air entry value; X = pore-size distribution BSC = saturated water content of compacted soil; Grc = residual water
index. content of compacted soil; <)»„ = air entry value of compacted soil;
Xc = pore-size distribution index for compacted soil.