Você está na página 1de 16

Available online at www.sciencedirect.

com

Soil & Tillage Research 97 (2007) 91–106


www.elsevier.com/locate/still

Mechanical behaviour of an undisturbed soil subjected


to loadings: Effects of load and contact area
Mathieu Lamandé a,*, Per Schjønning a, Frede A. Tøgersen b
a
University of Aarhus, Faculty of Agricultural Sciences, Department of Agroecology and Environment,
Research Centre Foulum, P.O. Box 50, DK-8830 Tjele, Denmark
b
University of Aarhus, Faculty of Agricultural Sciences, Department of Genetics and Biotechnology,
Research Centre Foulum, P.O. Box 50, DK-8830 Tjele, Denmark
Received 30 May 2007; received in revised form 30 August 2007; accepted 1 September 2007

Abstract
This paper deals with the stresses and deformations induced by loading an undisturbed soil. Measurements of vertical stress and
displacement were performed in two dimensions in large soil bins. The experiment included four loading treatments with two loads
(F = 43 kN and 2F = 85 kN) and two contact areas (A = 0.45 m2 and 2A = 0.90 m2). The loads were applied by stiff plates to a very
homogeneous and undisturbed loamy soil established in the soil bins 12 years prior to the tests. Twenty-one pressure transducers and
nine displacement transducers were inserted horizontally from an inspection tunnel with minimal disturbance of soil. Vertical stress
measurements included four depths (30, 50, 70 and 90 cm) and two to seven positions from the centre of load. Measured vertical
stresses were compared to vertical stresses calculated with the Söhne model. Measured vertical displacements were used to evaluate
the structural failure criterion. Doubling the contact area reduced stresses and strains in the topsoil, but load determined stresses and
strains in the subsoil, which supports the elasticity theory. The Söhne model predicted the stresses in the soil profile with a maximum
bias of 7 and 27 kPa for the low and high load, respectively. The best prediction was obtained with a concentration factor n  5–6
and n  8–9 for the low and high load, respectively. This indicates that the concentration factor depends on both soil mechanical
characteristics and loading conditions. The model underestimated stresses when the soil deformation was large. No plastic strain
was recorded if the calculated major principal stress was smaller than the precompression stress. For soil directly under the load
axis, strain increased with increases in both the calculated major principal stress, if higher than precompression stress, and in the
calculated minor principal stress. In contrast, for soil at the periphery of the loaded area plastic strain related better to the calculated
minor principal stress and did not relate to the calculated major principal stress. This indicates that shear failure may be important,
especially near the limits of the loaded area. More studies are urgently needed to evaluate the performance of the Söhne model in the
field and to better understand the mechanical behaviour of pedologically mature soils.
# 2007 Elsevier B.V. All rights reserved.

Keywords: Soil compaction; Elasticity theory; Söhne model; Stress concentration; Soil deformation; Precompression stress; Shear failure

1. Introduction cause soil deformation detrimental to key soil pro-


cesses. Due to significant problems in stress measure-
For heavy machinery, stresses in the soil profile may ments for undisturbed soil profiles, few comprehensive
exceed the internal strength of the soil and thereby evaluations of stress transmission in real soils loaded
with agricultural machinery have been performed and
compared to predicted stresses according to the
* Corresponding author. Tel.: +45 8999 1831; fax: +45 8999 1719. elasticity theory (Gupta et al., 1985; Keller and
E-mail address: Mathieu.Lamande@agrsci.dk (M. Lamandé). Arvidsson, 2004). Boussinesq (1885) described the

0167-1987/$ – see front matter # 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.still.2007.09.002
92 M. Lamandé et al. / Soil & Tillage Research 97 (2007) 91–106

propagation of principal stresses for the case of a comparison of stress and strength because principles of
vertical point-load resting on a semi-infinite elastic stress transmission are academic in a soil conservation
medium. Fröhlich (1934) introduced the concentration context if the stress at specific horizons does not induce
factor n in the Boussinesq equation to take into account persistent deformation. Prevention of compaction has to
that soil is not a truly elastic material, and that stresses be based on knowledge of the relation between stress
are more or less concentrated around the load axis, and strain at any point in the soil and more specifically
depending on soil strength. One crucial consequence of the stress limits at which permanent deformation of soil
assuming elasticity is that stresses acting on a given occurs. Horn (1981) proposed the precompression stress
point in the soil profile are additive. The vertical stress for this purpose. The precompression stress is the
component at that point may thus be calculated as the maximum major principal stress a soil can resist without
result of several surface point loads (Söhne, 1953). major plastic deformation and compaction (e.g. van den
The summation procedure by Söhne (1953) led to the Akker and Schjønning, 2004). It is obtained at the point
following statement: if the normal stress remains with maximum curvature of the stress–strain curve (e.g.
unchanged and the contact area is multiplied by a Gregory et al., 2006). A mechanically applied stress
factor b, stresses will reach b times deeper into the soil above the precompression stress results in a non-
(Koolen and Kuipers, 1983). Where the same force per reversible compression (the structural failure criterion;
unit area is applied, the stress in the subsoil will increase Dawidowski and Koolen, 2004). Several soil compac-
with the size of the loaded area (at a given point in the tion models are based on the combination of the
subsoil: the larger the contact area, the higher the prediction of stress according to the elasticity theory
stresses). Speaking in terms of wheels, a wide tyre with and the prediction of compaction according to the
double the wheel load and contact area compared with a structural failure criterion (e.g. van den Akker, 2004;
narrow tyre will impose higher stresses in the subsoil Keller et al., 2007).
than the narrow tyre (Olsen, 1994). This is a key issue The first objective of this study was to evaluate the
when considering technical solutions to the compaction Boussinesq–Fröhlich theory of stress transmission in a
problem in agriculture. Modern, wide tyres are able to structured, undisturbed soil. More specifically, we
carry very high loads without inducing extreme stresses wanted to (i) quantify the effects of load and contact
in the very topsoil. We need to know, which stresses are area on vertical stresses in the soil profile, and (ii)
transmitted to the subsoil. evaluate the concentration of stresses along the load
As soils become softer, the concentration factor n is axis. The second objective was to better understand the
anticipated to increase (e.g. Koolen and Kuipers, 1983), mechanical behaviour of an undisturbed soil subjected
yielding more concentrated and deeper reaching to stresses. We quantified the impact of stresses on
stresses under the load axis. Söhne (1958) suggested physical properties and compared it with the predicted
n values for different soil firmnesses – from 4 to 6 for impact using data from compression and shear tests on
hard to soft soil, respectively – resulting from empirical undisturbed soil samples.
combinations of both the bulk density and the water
status of soil. Ram (1984) determined concentration 2. Material and methods
factors experimentally for a remoulded loamy soil with
three different bulk densities at a constant water content. 2.1. Soil
The concentration factor decreased from 5.4 to 1.5 with
an increase of bulk density from 1.24 to 1.63 Mg m3, The experimental part of this study took place in the
supporting the trend proposed by Söhne (1958) but with semi-field facilities at Research Centre Foulum, Den-
much lower estimated n values for a hard soil. Horn mark. These include four lanes of soil bins each 40 m
(1993) observed that the concentration factor was not long, 2.7 m wide and 1.5 m deep. Each lane is divided
only related to the soil mechanical characteristics but into 25 units by concrete walls to 30 cm below the soil
also to the conditions of loading (load and contact area). surface, which enables realistic tillage operations with
To our knowledge, the relation between the concentra- all traffic on the concrete walls. The bins were filled
tion factor and the soil mechanical characteristic has not with soil in 1993. Large quantities of soil were
been evaluated in two dimensions for a naturally excavated from three Danish locations with different
structured soil. soil texture, respecting its vertical pedological differ-
Compaction of soil will only occur if the stresses entiation. The soil was air-dried and crumbled to
exceed the strength of the soil. Any discussion on aggregates <20 mm. The soil bin units were filled with
soil compaction per se should of course include a soil in 10 cm increments and packed to the dry bulk
M. Lamandé et al. / Soil & Tillage Research 97 (2007) 91–106 93

Table 1
Textural composition of the investigated sandy loam soil
Horizon (m) Organic matter (g 100 g1) Texture (g 100 g1)
<2 mm 2–20 mm 20–63 mm 63–200 mm 200–2000 mm
A (0–0.2) 2.4 17.7 13.1 17.8 29.5 19.5
B (0.2–0.6) 0.5 20.8 15.0 14.5 27.0 22.3
C (0.6–1.3) 0.3 21.9 16.9 11.6 26.7 22.6

density found in the field (A horizon, 1.5 Mg m3; B cause it to be displaced relative to the other section
and C horizons, 1.6 Mg m3) by careful application of resulting in a wider diameter by wedging the cylinder in
the exact weight of soil. the hole (Fig. 1). Vertical stress was transmitted to the
The soil was exposed to years of freeze-thaw and load cell by a plastic piston (3.14 cm2), the top of which
dry-wet cycles as well as frequent tillage, allowing soil followed exactly the curvature of the upper half of the
structure to develop. The annual excess precipitation cylinder. A screw at the end of the transducer housing
amounted to 200–400 mm percolating the soil primarily could activate the wedge system. Vertical displacement
during the winter. Tillage was performed to 0.2 m depth of the soil was measured using the method described by
with a one-furrow moldboard plough, pulled by a tractor Arvidsson and Andersson (1997). A pressure transdu-
driving on the top of the concrete walls. Secondary cer, situated outside the soil, was connected to an oil-
tillage was carried out with a rotary harrow and drilling filled container situated in the soil by a plastic tube.
of seeds with a traditional drill. The soil was cropped Changes in oil pressure, recorded by the pressure
primarily with small grain cereals. However, at the transducer, were proportional to the vertical displace-
beginning of the period, alfalfa (Medicago sativa L.) ment of the soil.
was frequently grown due to its high potential for deep Horizontal holes were made for the insertion of
rooting in virgin soil. We selected these test conditions transducer housings from an inspection tunnel next to the
because we wanted a structured soil with no potential soil bins. We used a hydraulic jack to push a stainless steel
textural gradients or high heterogeneity of field soils. soil corer (54 mm outer 1, 50 mm inner 1) two metres
Our tests were applied to a sandy loam soil derived from horizontally into the soil. The jack was attached to a level
Weichselian glacial till (Table 1). trestle located in the inspection tunnel to ensure
horizontality of holes. The edge of the corer tube was
2.2. Construction and insertion of transducers sharpened, and the first few millimetres given a smaller
inner diameter. This was to minimize compaction around
We constructed cylindrical steel transducer hous- the corer and friction between the soil and the inner part
ings (52 mm 1, 80 mm length) to accommodate a load of the tube. The corer was extracted for every 0.5 m bored
cell (DS Europe Series BC 302) and a small oil-holding and emptied. The amount of material was compared to
container. The cylinder was constructed from two the expected to check that only a minimum compaction
close-fitting parts, where a push on one section will occurred during the coring phase. Each transducer

Fig. 1. Transducer housing (52 mm 1, 80 mm length) accommodating a load cell (LC) and an oil-holding container (C). When the two parts are
pushed relative to each other, the diameter of the housing is expanded, enabling a good contact between soil and the load cell piston. See text for a
detailed description of the insertion procedure.
94 M. Lamandé et al. / Soil & Tillage Research 97 (2007) 91–106

housing was then inserted one by one into the hole with a
specially constructed rest. The wedge system was
activated by a long screwdriver and ensured a good
contact between soil and the load cell. The wedge was
activated until a predetermined low force, as read on the
computer screen, had confirmed that contact between the
transducer and the soil had been achieved. A number of
transducer housings could be installed in the same hole,
allowing several vertical stress and vertical displacement
measurements across the loaded plates (see Section 2.3).
Cables and tubes from the inner transducer housings were
taken through their outside neighbours along pre-bored
holes. All transducer housings were able to move Fig. 2. The load was delivered by a wagon to a stiff plate through a
vertically independently of each other. See Lamandé hydraulic jack (8637 kg on axle and 0.45 m2 plate in this situation).
et al. (2006) for more details about the measuring system.
Kirby (1999a,b) showed that the compaction of the placed on the soil surface in the centre of the soil bin. The
surrounding soil during the insertion of sensors, the load was applied twice: the first time with a full load for
difference in stiffness between the sensors and the soil, 0.5 s and the second time with a full load for 3 s a few
and the size and shape of the sensor could lead to an minutes later. Twenty-one stress transducers, nine of
overestimation of stresses. With the methodology them including an oil-filled container connected to
developed here, the soil disturbances relating to the pressure transducers, were inserted at four depths below
insertion of the sensors were minimized as described, the loaded area (seven at 0.3 and 0.5 m depth, five at
and sensor size was optimised (by minimizing the 0.7 m and two at 0.9 m depth). The four lines of
height to length ratio). In addition, the cylindrical shape transducers were installed with different distances from a
of the sensor avoided sharp corners. However, the vertical line in order to avoid disturbance of the stress
transducer housings have a higher stiffness compared to field to be monitored. The stress- and displacement
the soil, which influences the accuracy of the stress transducers were located at four distances from the centre
measurements. of the loaded area (0, 0.08 (0.9 m depth: 0.04), 0.24,
0.40 m; Fig. 3). We installed ‘dummy’ transducer
2.3. Treatments and test procedure housings between those carrying transducers. Measure-
ments were acquired in 2 ms bursts at 2 kHz (500 ms)
Our study was performed in 2005 (12 years after during the loading event. Typical measurements of
establishing the soil bins) and included four loading vertical stress and displacement at 0.3 m depth under the
treatments. A reference treatment (AF) with a well- centre of the plate are shown on Fig. 4 for a test where
defined load (F = 43 kN) and contact area (A = 0.45 m2) 4377 kg was delivered to the plate of 0.45 m2.
was compared to three other treatments: one doubling the
load (A2F), one doubling the contact area (2AF) and one 2.4. Measurements of soil mechanical and physical
doubling the load as well as the contact area (2A2F). Six properties
soil bins were available for the investigations. This
allowed two replicate soil bins for two treatments (AF We sampled 100-cm3 soil cores (61 mm inner 1,
and 2AF) and one bin for each of the two other treatments 34 mm height) from the 8–12, 23–27, 38–42, 58–62, 78–
(A2F and 2A2F). The loads were applied to rigid 0.45 m2 82 and 98–102 cm depths below the loading surfaces
(0.45 m  1.0 m) and 0.90 m2 (0.90 m  1.0 m) plates. (sample group I). The cores were collected in a line across
A mesh of load cells could detect no deflection of the the loaded area with 11 cm between core centres.
plates when loaded onto a concrete floor (data not Additional cores (sample group II) were sampled at the
shown). We are thus able to anticipate the vertical same depths but next to the loaded area (close to the
pressure applied to the plate to be uniformly distributed. concrete walls of the bin). The cores from sample group I
The load was delivered by a wagon (4377 or 8637 kg) that were used for determination of the dry bulk density.
was lifted by a hydraulic jack (Fig. 2). The top 10 cm of Six cores per depth from group II were saturated with
the soil was carefully scraped off to obtain a completely water on tension tables and used for determination of
level and horizontal surface. This ensured a perfect the soil water characteristics with a standard technique.
contact between the plates and the soil. The plate was The parametric model of van Genuchten (1980) was
M. Lamandé et al. / Soil & Tillage Research 97 (2007) 91–106 95

Fig. 3. A cross section of a soil bin. Twenty-one stress transducers, nine of them including an oil-filled container, were inserted at four depths below
the loaded area (0.3, 0.5, 0.7 and 0.9 m). The transducer housings were located at four distances from the centre of the loaded area (0 or 0.04, 0.08,
0.24, 0.40 m). Empty housings were inserted to fill up the bored hole throughout its length.

fitted to the retention curves using function nls in the function was fitted to the compression data using nls
statistical programming environment R (R Development procedure in R. Precompression stress was estimated as
Core Team, 2005) to calculate the water potential from the point of maximal curvature derived from the
the volumetric water content at the time of compaction curvature function. We finally used twelve cores per
tests. Porosity for each soil horizon was calculated from depth from group II for measurement of shear strength
our measurement of bulk density and particle density with the torsional shearing device described by
measured for this soil by Jacobsen (1989). Schjønning (1986). For each depth, two replicate soil
Six cores per depth from group II were taken to a cores were subjected to six levels of normal load (range
confined compression test at field water content. We 36–186 kPa). The torque was measured during the
used strain-controlled stress application as suggested by rotation of a shear annulus with a mean rate of
Koolen (1974) and applied the technique described by 46 mm min1. The maximum shear strength, tmax, was
Schjønning (1991). Precompression stress Pc was estimated from the relation between strain and shear
estimated using the method described by Gregory strength. Cohesion c and angle of internal friction w
et al. (2006). The asymmetrical Gompertz sigmoidal were calculated from the Coulomb model using normal
stress as effective stress (Schjønning, 1986).

2.5. Calculations

2.5.1. Vertical stress


Vertical stresses was calculated by summation
(Söhne, 1953) for the 21 positions of vertical stress
measurement in the soil profile (Section 2.3):
Xn Xn
nPi
ðs z Þi ¼ 2
cosn ui
i¼1 i¼1
2pr i

where sz is the vertical stress, P is the vertical point


Fig. 4. Example of measured vertical stress and displacement during
load, r and u are the polar coordinates and n is the
the loading procedure (first loading for treatment AF, 0.3 m depth, concentration factor introduced by Fröhlich (1934). We
under the centre of the plate). used the statistical programming environment R with
96 M. Lamandé et al. / Soil & Tillage Research 97 (2007) 91–106

n = 6. In addition, the concentration factor was esti- 2A2F) were applied only to one test bin unit. A pre-
mated for each treatment with a non-linear least squares condition for a correct test of treatment effects by the
fitting (function nls in R). The root mean square error described model is thus identical initial physical con-
(RMSE) and bias of prediction was used for evaluating ditions within all soil bin units. Due to the meticulous
the fit to measured data: procedure followed when establishing the test units, we
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi find this assumption fulfilled. Similar approaches have
1X n been described by Crowder and Hand (1990). We used
RMSE ¼ d 2 and the SAS 8.0 package (SAS Institute Inc., Cary, NC,
n i¼1 i
USA) for these calculations.
1X n
bias ¼ di where d i ¼ s predicted  s measured
n i¼1 2.5.2. Major principal stress
We used SoilFlex (Keller et al., 2007) to calculate the
Treatment effects were analysed by a mixed model major principal stress using the analytical equations
taking into account the fixed effects of treatment and developed by Söhne (1958). Calculations were per-
depth and the random effects of test bin unit nested formed for four conditions corresponding to the four
within treatment and the interaction between depth and treatments (combination of two contact areas and two
test bin unit. We note that two treatments (A2F and loads) assuming a uniform contact pressure. The stress

Table 2
Selected soil properties for the sandy loam soil studied

Figures in brackets denote standard error (n = 6).


a
Calculated from the field water content and the water retention characteristics by use of the van Genuchten (1980) equation. bDegree of saturation.
c
Bulk density giving rise to water saturation (rt(s = 1) = (1  Vw)  rr, where Vw is field water content and rr is particle density).
d
From Jacobsen (1989); Campbell b gives the slope of a log–log plot of water potential and water content.
M. Lamandé et al. / Soil & Tillage Research 97 (2007) 91–106 97

Table 3
Soil mechanical properties
Depth (m)
0.1 0.25 0.4 0.6 0.8 1
Precompression stress, Pc (kPa) 47 (14) 51 (8) 43 (8) 40 (7) 38 (5) 33 (7)
Cohesion, c (kPa) 13 (6) 27 (6) 34 (4) 29 (5) 19 (4) 29 (9)
Internal angle of friction, w (8) 40 37 40 40 43 36
Tan w 0.85 (0.05) 0.76 (0.05) 0.83 (0.04) 0.85 (0.04) 0.93 (0.03) 0.73 (0.07)
Figures in brackets denote standard error (n = 6).

distribution on the contact area between a soil and a The Gompertz function fitted well to the data from
rigid body might not be uniform but definitions in the the uniaxial confined compression tests. We estimated
literature are contradictory. Craig (2005) mention a precompression stress between 33 and 51 kPa for the six
stress distribution concave parabolic on clay and convex depths (Table 3). These values were low compared to
parabolic on sand, while Balakrishna et al. (1992) soils with similar texture (e.g. Arvidsson and Keller,
reports studies where the stress distribution was 2004). Even if the dry bulk density in the soil bins had
concave parabolic on sand. Calculated major principal been identical to the original soil, the structure of the
stress was compared to the precompression stress soil in the bins was young (12 years) and the soil had
(Section 2.4), in order to test its ability to represent the never been trafficked. The highest precompression
structural strength of the soil. values were obtained for the two upper soil layers (0.1
and 0.25 m at 47 and 51 kPa, respectively, probably
because of better structural regeneration from effects of
3. Results climate, tillage, root growth and other biotic processes.
Soil cohesion, c, was estimated to range between 13
3.1. Soil characteristics and 34 kPa for the six depths (Table 3). As for Pc, these
estimates are lower than often seen for agricultural soils
The soil bulk densities were close to the targets close to field capacity (e.g. Schjønning, 1991, 1999),
aimed at in 1993: 1.6 Mg m3 for original B and C which may be explained by a poorly developed
horizons and 1.5 Mg m3 for the A horizon (Table 2; structure. We measured a lower cohesion in the plough
also see Fig. 3). The lower value of the very topsoil is layer (0–0.2 m) than in the subsoil. A higher cohesion
due to annual tillage. These data reflect the tendency of was expected in the plough layer than in the other
graded, morainic soils to gain a high density by natural horizons because it increases with decreasing degree of
processes (Ehlers and Claupein, 1994). The values of saturation S and increasing matric head h (Koolen and
the b-parameter suggested by Campbell (1974) indicate Kuipers, 1983; Schjønning, 1999). The angle of internal
that the soil in the bins is less structured than soil in the friction, w, was rather similar for all depths (Table 3)
field (Table 2). The soil thus represents an intermediate and generally tended to be high compared to values for
step between a remoulded soil column in the laboratory the top 40 cm of seven Danish soils including the one
and the pedologically mature soil in the field. There tested in this study (Schjønning, 1991, 1999).
were only small differences in the water-holding
capacity of the different soil layers (Table 2; only 3.2. Stress
selected potentials are shown). The water content at the
time of compaction treatments corresponded to an Measured stresses are presented in Fig. 5 and
increasing (less negative) water potential with depth Table 4. The highest stresses were measured at 0.3 m
(Table 2). The data indicate that the soil was drained to a depth near the load centre (black and dark-grey circles)
lower water content than field capacity, which is and the lowest stresses were measured at 0.9 m depth
normally approximately 100 hPa matric potential for and/or far from the centre (all grey tones rhombi; light
Danish soil types (Schjønning and Rasmussen, 2000). grey and white circles and squares). For the treatments
The degree of saturation increased with depth and including the large loading plate (2AF and 2A2F) the
ranged from 0.37 to 0.65. For all depths the soil should decrease in stress with depth was lower than predicted,
be compacted to bulk densities above 2 g cm3 to reach especially at 0.9 m depth in treatment 2A2F. The range
saturation (Table 2). in predicted values was low for these treatments
98 M. Lamandé et al. / Soil & Tillage Research 97 (2007) 91–106

Fig. 5. Calculated vertical stress according to Söhne (1953; n = 6) related to measured vertical stress for the four treatments (AF, A2F, 2AF, 2A2F) at
four depths: 0.3 m (circles), 0.5 m (squares), 0.7 m (triangles), 0.9 m (rhombi); and at four distances from the load centre: below the centre (black),
0.8 m (dark grey), 0.24 m (light grey), and 0.4 m (white).

because, according to the summation principle in the and 2A2F, 48 kPa for 2AF, and 188 kPa for A2F. At
Söhne model, for a given depth, the sensors are situated 0.3 m depth, the measured stress averaged for the
closer to the same isobar. sensors located close (0, 0.04 and 0.08 m) to the load
The four combinations of contact area and load axis were grouped in three, with significant differences
resulted in three theoretical contact stresses applied to reflecting the contact area stresses of the treatments
the soil surface (black squares in Fig. 6): 95 kPa for AF (2AF < AF = 2A2F < A2F; Fig. 6). At 0.5 and 0.7 m

Table 4
Mean measured vertical stresses (kPa) for the four treatments, the four soil depths, and the five positions relative to the load axis
Treatments Depth (m) Position across the plate (m)
0a 0.04b 0.08b 0.24b 0.4 b
AF 0.3 88 (1) 88 (10) 56 (6) 13 (3)
0.5 58 (0) 52 (8) 51 (7) 14 (3)
0.7 49 (8) 43 (2) 36 (5)
0.9 29 (4)
2AF 0.3 34 (9) 34 (10) 49 (9) 36 (5)
0.5 28 (8) 61 (8) 35 (6) 27 (2)
0.7 54 (16) 30 (6) 25 (4)
0.9 17 (3)
c
2A2F 0.3 NA 75 (10) 106 (5) 94 (10)
0.5 66 79 (16) 63 (18) 67 (8)
0.7 42 55 (15) 68 (0)
0.9 101 (5)
A2F 0.3 208 179 (6) 118 (32) 0 (4)
0.5 141 134 (12) 93 (11) 43 (9)
0.7 119 96 (0) 80 (1)
0.9 83 (1)
Figures in brackets denote standard error.
a
For AF and 2AF, n = 2; for 2A2F and A2F, n = 1.
b
For AF and 2AF, n = 4; for 2A2F and A2F, n = 2.
c
Missing value due to poor contact between soil and sensor.
M. Lamandé et al. / Soil & Tillage Research 97 (2007) 91–106 99

Fig. 6. Vertical stress in the soil profile for sensors located maximum 0.08 m from the load axis. Left: average measured stress with standard error;
for each depth, different letters indicate significant differences between treatments (P = 0.05). Right: the same data as on the left plus the Söhne-
predicted vertical stress (lines, n = 6) for the same coordinates.

depth, the A2F treatment gave a significantly higher concentration factor from 4 to 12. The bias indicates
stress than the other treatments, while these could not be model underestimation of stresses for the A2F
distinguished statistically. At 0.9 m depth, the measured treatment, which is a quantification of the observations
stress was significantly lower for AF and 2AF than for in Fig. 6 (right) for n = 6. Variation in the concentration
A2F and 2A2F. Inclusion of the Söhne-predicted factor could not account for the problem (Fig. 7, right).
stresses in the plot revealed that generally the No or only a minor bias was found for the three other
measurements followed the predictions, including a treatments (AF, 2AF and 2A2F). While the treatments
fit to the stress in the contact area calculated from the with the low load had a minimum RMSE for n  6 (5.5
load and loading area (Fig. 6, right). An exception is the and 6.4 for 2AF and AF, respectively), a much higher
A2F, where we measured higher stresses than predicted. concentration factor was needed to optimise the fit for
The root mean square error (RMSE) and bias of the high load (9.2 and 8.5 for A2F and 2A2F,
predictions for all the positions for the four treatments respectively) (Fig. 7, left).
are presented in Fig. 7. In terms of the root mean square In order to have a more detailed picture of the fit by
error (RMSE), the model fitted better for the low Söhne’s model in two dimensions, the RMSE, the
(43 kN, AF and 2AF) than for the high load (85 kN, coefficient of variance (CV, the relative RMSE), and bias
A2F and 2A2F), irrespective of the contact area (Fig. 7, were calculated for stress transducers grouped according
left). For the two treatments with the large contact area to three locations in the soil profile (Table 5). In these
(2AF and 2A2F), and in a smaller extent for AF, the calculations, we used a concentration factor of n = 6
RMSE was only slightly affected by changing the because this value would normally be chosen for

Fig. 7. Root mean square error (RMSE) and bias of prediction according to Söhne (1953) calculated for different values of the concentration factor n
for all positions in the soil profile.
100 M. Lamandé et al. / Soil & Tillage Research 97 (2007) 91–106

Table 5
Root mean square error (RMSE), coefficient of variance (CV, relative root mean square error) and bias of prediction according to Söhne (1953)
calculated for n = 6 for stresses measured in three locations in the soil profile: close to the load axis (0 and 0.08 m) at 0.3 and 0.5 m depth (load axis,
top), far from the load axis (0.24 and 0.4 m) at 0.3 and 0.5 m depth (periphery, top) and close to the load axis (0, 0.04 and 0.08 m) at 0.7 and 0.9 m
depth (load axis, bottom)
Locations in the soil profile AF 2AF 2A2F A2F
Load axis, top RMSE (kPa) 17 21 21 30
CV (%) 23 55 28 18
Bias (kPa) 3 5 9 27
Periphery, top RMSE (kPa) 15 13 27 33
CV (%) 43 36 33 52
Bias (kPa) 7 3 16 11
Load axis, bottom RMSE (kPa) 9 17 35 20
CV (%) 24 57 49 21
Bias (kPa) 1 2 8 18

modelling a ‘soft’ soil such as the one tested here. In other prediction accuracy is due to an overestimation of stress
words, the data in Table 5 quantify the absolute and in the load axis of upper soil layers but an under-
relative accuracies of prediction for specific parts of the estimation in the deeper layers and in the periphery of
loaded profile. The first group comprises stresses the top layers (Table 5). It ought to be added, however,
measured close to the load axis (0 and 0.08 m) at the that for 2A2F, the average 8 kPa bias covers a
0.3 and 0.5 m depths (load axis, top), the second group considerable overestimation (bias = +17 kPa) at 0.7 m
the stresses measured far from the load axis (0.24 and depth and a dramatic underestimation (bias = 45 kPa)
0.4 m) at the 0.3 and 0.5 m depths (periphery, top), and at 0.9 m depth (Fig. 6). Finally, the data in Table 5
the third group the stresses measured close to the load indicate that the Söhne predicted stresses for transducers
axis (0, 0.04 and 0.08 m) at the 0.7 and 0.9 m depths (load located vertically below or outside the edge of the
axis, bottom). The prediction accuracy in absolute values small plate (periphery, top) are lower than measured (7
(RMSE) calculated for all stress sensors at a concentra- and 16 kPa bias for the AF and A2F treatments,
tion factor of n = 6 was greater for the high (27 kPa) respectively).
than the low load (15 kPa), and was nearly independent
of the loaded area (Fig. 7). However, from Table 5 it 3.3. Stress impact
appears that close to the load axis the accuracy is
influenced by soil depth. For the AF and A2F the RMSE The vertical normal strain, e1, calculated as the
is smaller for the deep horizons than for the topsoil. That vertical displacement relative to the thickness of
is, the Söhne model gives better predictions for the deeper specific soil layers was significantly larger for the
layers. In contrast, for the 2A2F treatment the prediction A2F treatment for all soil layers except the 0.7–1.3 m
accuracy is better at the top than at depth. The prediction (Fig. 8). For the deepest layer (0.7–1.3 m), the AF and
accuracy relative to the measured stress level (CV) is 2AF treatments had similar vertical strains, which again
similar at the top and at the bottom of the soil profile near were lower than strains for the A2F and 2A2F
the load axis for AF and 2AF. For A2F and especially for treatments. However, the trend was not significant.
2A2F, the relative accuracy is better at the top than at the Thus, the differences in strain between treatments seem
bottom of the profile near the load axis. For the two to be related to the specific pressures for the upper
treatments including the low load (AF and 2AF), the layers and to the load for the lowest layer.
bias is generally small (7 to +5 kPa, Table 5) and Throughout the soil profile dry bulk densities were
independent of the location in the soil profile. This means higher for the A2F than for other treatments (Fig. 9).
that the predictions are central estimates of the stress and The treatment effect was significant when averaged for
that the prediction accuracy (RMSE) relates to stochastic the profile (P = 0.02), but not at specific soil depths.
variability. In contrast, there is generally a considerable
bias for the estimates below the high loads (Table 5). It 3.4. Criterion for structural failure
appears that the underestimation of stresses already
noticed from Fig. 7 (right) for the A2F is true for all parts We used the SoilFlex model (Keller et al., 2007) to
of the soil profile. For the 2A2F treatment, the poor calculate the major and minor principal stresses s1 and
M. Lamandé et al. / Soil & Tillage Research 97 (2007) 91–106 101

Fig. 8. Vertical normal strain in the soil profile for the four treatments
(AF, 2AF, A2F, 2A2F) for four layers (0–0.3 m, 0.3–0.5 m, 0.5–0.7 m,
0.7–1.3 m). Bars indicate standard error. Different letters indicate
significant differences between treatments (P = 0.05). In A2F, strains Fig. 10. Vertical normal strain related to the difference between major
were underestimated because the maximum range of some transducers principal stress (s1) and precompression stress (Pc), and to the ratio of
was reached. minor principal stress (s3) by major principal stress (s1) at the load
axis (black symbols) and at 0.24 m from the load axis (grey symbols)
and for three layers (circles, 0.3–0.5 m; squares, 0.5–0.7 m; triangles,
s3 at all soil depths and at two positions under the 0.7–1.3 m). Figures in brackets denote standard errors (n = 2 at load
loaded plates (the load centre and 0.24 m from the load axis; n = 4 at 0.24 m from load axis).
centre). We point out that we never measured any
significant strain for s1 values lower than Pc (data not precompression stress (s1  Pc). At the load axis for
shown). There were missing values for A2F because of treatments AF and 2AF, the stress–strain relationship
vertical displacements over the sensors’ range. Due to was close to what could be expected from the
the low values of precompression stress of our soil, only compression test, i.e. stress exceeding the precompres-
few predicted major principal stresses were lower than sion stress gave rise to an increase in strain. A regression
the precompression stress (treatment 2AF). analysis for AF and 2AF data with major principal
For the two treatments with the high load (A2F and stresses higher than the precompression stress explained
2A2F), the increase in measured strain was not 47% of the variation and indicated no significant
proportional to the increase in stress exceeding the interception on the strain axis (i.e., vertical strain = 0 at
s1 = Pc).
In Fig. 10 we have related both the difference
between s1 and the precompression stress (s1  Pc) and
the ratio s3/s1 to the vertical normal strain for the
corresponding three soil layers. The data relate to the
first load test for treatment AF. The strains measured
were on average about the same for sensors positioned
at the load axis (black symbols) and 0.24 m from this
axis (grey symbols), but they were associated with
different stress levels above the precompression stress.
For example, for the upper layer, the average strain at
the load axis was 0.032 and it was related to an excess
stress s1  Pc = 40 kPa. In contrast, the average strain
at 0.24 m from the axis was exactly the same (0.032) but
was related to s1  Pc = 0 kPa. For both positions the
ratio s3/s1 increased with decreasing depth.
Fig. 9. Dry bulk density in the soil profile after the two loadings for In Fig. 11 the estimated concentration factor is
the four treatments (AF, 2AF, A2F, 2A2F). Averages based on four
cores sampled under the loaded plates. Bars indicate standard error. At related to the average vertical displacements for the
a given depth, the differences between treatments were not significant whole profile measured during the two loadings in the
(P = 0.31). four treatments. For both loadings the concentration
102 M. Lamandé et al. / Soil & Tillage Research 97 (2007) 91–106

2A2F treatment (corresponding to a high load carried


by a wide wheel), the model overestimated the stress in
the topsoil at the load axis but underestimated the
stresses in the deepest layer along the load axis
(bias = 45 kPa at 0.9 m depth) as well as in the upper
soil layers far from the load axis (Table 5).

4.2. Concentration of stress along the load axis

According to the Fröhlich theory, a concentration


factor matching the soil mechanical state should
improve prediction accuracy. The n value for a medium
that follows the Hooke’s law is 3. For all treatments, the
best fit with the Söhne model was found for
Fig. 11. Estimated concentration factor against average vertical dis-
concentration factors higher than 3. The vertical stress
placement for the whole profile for the four treatments (black circles,
AF; white circles, 2AF; black triangles, A2F; white triangles, 2A2F) transmission in our soil was thus more vertical and more
and for the two loadings (‘‘1’’, first loading; ‘‘2’’, second loading). concentrated around the load axis than in an isotropic
and elastic medium. This may be interpreted as a plastic
factor increased with the vertical displacement. For behaviour of the soil (Smith et al., 2000). Estimated
each treatment, the estimated concentration factor was concentration factors for treatments with the low load
lower for the second loading than for the first loading. (2AF and AF) were close to the value of n = 5 given by
Koolen and Kuipers (1983) for a soft soil, or n = 6 given
4. Discussion by Söhne (1958) for a wet soil. This is in accordance
with: (i) the observed structure of the soil, which was 12
4.1. Stress transmission in the soil profile years old and thus had limited development; (ii) the
history of the soil: the soil had never been trafficked
Our results confirm the implication from the after installation in the bins. In contrast, the estimates of
summation principle in the model (Olsen, 1994): in concentration factor n were much higher than expected
the upper soil layers (<0.3 m depth) the measured for the two treatments with the high load (A2F and
stresses were close to those in the contact area, while the 2A2F). Söhne’s description of firmness relates to
load determined the level of stress in the deeper soil combinations of both bulk density and water content
layers (Fig. 6). The result is in accordance with Dexter of soil (Söhne, 1953). In our study, the estimated values
et al. (1988) and Smith and Dickson (1990). It has of n were different even if the initial dry bulk density
practical implications in terms of the development of and water content were not different between treat-
practices avoiding soil compaction. It means that the ments. It means that the concentration factor is not only
only way to reduce stresses in the subsoil would be to influenced by soil mechanical characteristics but also by
reduce the weight of the machinery. the characteristics of load application. We varied only
Even if the general principle of the model is valid, the two aspects of load application: the load and the contact
accuracy of predicted stress in a given part of the soil area. The concentration factor increased significantly
profile may be poor. The data in Table 5 indicate that the with load (P = 0.04). This has previously been observed
deviation from the 1:1 line for 2AF in Fig. 5 was not due also by Horn (1993). The decrease in concentration
to any structure in data: there was no trend in bias and factor with the increase of contact area mentioned by
the variability (RMSE) was about identical for the three the same author was not significant in our study
parts of the loaded profile. Thus, the predictions for the (P = 0.16). Neither did we find a significant interaction
first loadings with the small load seem to be rather between load and contact area (P = 0.98). Smith et al.
unbiased (7 to +5 kPa) with an accuracy (RMSE) of (2000) also recognized that the concentration factor was
8–20 kPa (Table 5). In contrast, the predictions for the dependent on the loading characteristics. The higher
first loadings with the high load were rather uncertain values of n for the high loads may be due to the non-
(RMSE 20–35 kPa; Table 5). Very importantly, our negligible deformation of the subsoil during the
results also show that the Söhne model underestimated application of load for the A2F and 2A2F treatments
stresses for all parts of the loaded profile when loading a (Fig. 8), which is contradictory to model assumptions.
small area with a high load (A2F; Table 5). For the When considering all the loadings, the model-optimis-
M. Lamandé et al. / Soil & Tillage Research 97 (2007) 91–106 103

ing concentration factor seemed to increase non- soil behaviour. Guérif (1984) showed that the compres-
linearly with the vertical displacement (Fig. 11). When sibility of the whole soil profile was a combination of
the vertical displacement was small or null, the the relative compressibility of each soil layer. Con-
concentration factor was close to 5.5 for the first sidering soil layers A and B where A is the upper layer,
loading and 4.5 for the second loading (2AF, Fig. 11). he measured that compaction occurred in layer B when
Our calculations showed that adjustment of the the rigidity of the upper layer A overcame the
concentration factor led to small improvements of the compressibility of layer B. We measured an increase
predictions (Fig. 7). in bulk density under 0.4 m depth for the two treatments
Some authors have reported stress distributions and with the high load only (A2F and 2A2F). Only for these
soil deformation due to stresses that could not be two treatments did the final dry bulk density in the
accounted for by the Söhne model (e.g. Danfors, 1974; plough layer (0–0.2 m) reach the dry bulk density at
Trautner and Arvidsson, 2003). Obviously such results 0.25 m depth. The rigidity of a material can be due to
were due to breakdown of one or more preconditions for its structure, to the aggregation of particles and/or to
the validity of the Söhne model. Håkansson and Reeder the degree of saturation because the water is not
(1994) hypothesized that ‘‘the occurrence of deep compressible (Hadas, 1994). Taking into account the
vertical cracks in dry swelling/shrinking soils lead to the low degree of saturation in the plough layer (S = 0.37,
transmission of stresses into deep subsoil layers without Table 2), we assume that the rigidity of this layer
attenuation’’. Soil is a heterogeneous assemblage of overcame the compressibility at 0.25 cm depth because
structural units held together with physico-chemical of structural changes due to compaction. Below 0.4 m
and biological bonding agents (e.g. Dexter, 1988). depth, the degree of saturation was higher (S = 0.57,
Studies in fragmentation of soils have shown that the Table 2), and compaction may have occurred due partly
strength of structural units is very scale-dependent (e.g. to structural changes and partly to the incompressibility
Skidmore and Powers, 1982). Based on these facts, it of water even if the final bulk density was lower than a
might be expected that soil would not obey the value giving rise to water saturation (Table 2). For the
requirements for the elastic Söhne model also at high A2F and 2A2F treatments, we measured average
water contents. Thus, Trautner (2003) suggested that a vertical displacements for the soil profile of 34.2 mm
distinct element model designed for describing stresses and 7.4 mm, respectively. These significant deforma-
in pavements and constructed roads might better predict tions took place within a few seconds. The soil
stress transmission in soil (Ullidtz, 1998). Such a model hydraulic conductivity has most probably not allowed
would imply a much more concentrated (vertical) stress the distribution of water among soil pores, and parts of
transmission than the Söhne model based on elasticity. the pore system may thus have been saturated. More
In our study we recorded stresses also for the sensors studies are needed to elucidate soil behaviour when
located outside the small loading plate (AF and A2F; heavily loaded. However, irrespective of the mechanical
0.24 and 0.4 m; Fig. 5). We also note that these stresses and physical processes involved, the Söhne model
were even higher than predicted by the Söhne model seems to be a poor tool for prediction of stresses in soils
(Table 5). In addition, our results indicate that a high loaded above some upper limit that may be different
concentration factor (yielding a high concentration of from soil to soil.
stress along the load axis) was not always able to
account for bias in the predictions. The Söhne model in 4.4. Precompression stress for prediction of soil
its present appearance did not perform perfectly when structural failure
the loading induced soil deformation. The model might
perhaps be improved by allowing for a deformation- The concept of precompression stress implies that
dependent choice of the concentration factor. loading a soil above this limit will induce irreversible
deformation, while stresses below the limit will cause
4.3. Soil mechanical behaviour at high loads only elastic strain (Lebert and Horn, 1991; Horn, 1993).
The stress state in soil during loading events may be
Our results have indicated that the Söhne model for different to that in an uniaxial confined compression
the low load could provide central estimates of stresses test (e.g. Keller, 2004), and the suitability of such a
in the soil profile with a bias of about 5–7 kPa. However, test in soil compaction contexts has been questioned
the model had problems for both treatments with the (e.g. Arvidsson and Keller, 2004; Keller et al., 2004).
high load. It appeared that even a regulation of the However, in this study we never recorded any
concentration factor could not account for the observed significant soil deformation at major principal stresses
104 M. Lamandé et al. / Soil & Tillage Research 97 (2007) 91–106

below the precompression stress. In addition, at the load on the load. The accuracy of the quantitative prediction
axis there was a positive correlation between soil was much poorer for the high than for the low load,
vertical normal strain and the stress exceeding the which we hypothesize was due to large vertical
precompression stress (black symbols in Fig. 10). This deformations measured in the heavily loaded soil. It
indicates that the mechanical behaviour of soil under the means that the concentration factor depends on both soil
centre of a loaded area may be comparable to what properties and on loading conditions. The model for
happens during a compression test. It is important to stress propagation has to be improved to accurately
note, however, that the above was only observed for a predict soil compaction especially when loading
moderate load. At high loads, we could not identify any impacts on soil physical properties are significant. This
structure in data. Moreover, the ratio s3/s1 increased may perhaps be accomplished by a feedback function
with increasing strain and was smaller than 0.5, which is correcting the stress propagation from the expected soil
the ratio of an ideal uniaxial compression (Koolen and deformation. However, in this study the prediction
Kuipers, 1983) (Fig. 10). This indicates that strains accuracy at high loads was unsatisfactory even with
measured at the load axis may result from both regulation of the concentration factor. Our results
compression and shear deformation. indicate that the precompression stress may be a
The mechanical behaviour of soil seemed to be relevant stress limit for prediction of soil failure at
rather different to expectations when the distance from isotropic stress conditions (at the load axis). However,
the load axis increased (grey symbols on Fig. 10). This the results also indicate that shear failure may contribute
can be partly explained by the slight underestimation of significantly to deformation of soil at a distance from
the major principal stress for these positions (Table 5). the load axis.
However, the high strains we have observed at levels of
vertical stress equivalent to the precompression stress
Acknowledgements
(s1  Pc  0 kPa) are probably due to shear failure.
This hypothesis is supported by the increase of strain
This work was financed partly by the Danish
with the ratio s3/s1 (Fig. 10). As we only measured
Research Council for Technology and Production
vertical stress and vertical deformation, this hypothesis
Sciences and partly by the Faculty of Agricultural
cannot be supported further. Future studies are
Sciences, Aarhus University. The development of the
encouraged to include measurements of horizontal
described methodology was inspired by ideas of Dr.
stress and strain.
Andreas Trautner. The technical assistance of Finn H.
Our data indicated that the stresses measured at the
Christensen, Michael Koppelgaard and Stig T. Ras-
second loading were very close to those measured at the
mussen is gratefully acknowledged. The trailer used for
first loading (data not shown). Nevertheless, the
applying the loads was kindly made available by Dr. J.
additional, plastic strains measured after the second
Arvidsson, the Swedish University of Agricultural
loading were as large as those measured after the first
Sciences. The soil compaction models SOCOMO and
loading (Fig. 11). This means that the ‘‘precompres-
SoilFlex were kindly provided by Dr. J.J.H. van den
sion’’ of the soil at the first loading did not prevent
Akker and Dr. T. Keller, respectively. The authors are
further compaction with the same stress. Taking into
indebted to Dr. K. Kristensen for advice on the
account the discussion above, we interpret this as an
statistical analyses.
influence of loading time. The first loading event
induced a full load for about 0.5 s, while the second was
allowed to impact the soil for approximately 3 s. This References
further complicates the evaluation of the precompres-
sion stress as a suitable parameter for predicting Arvidsson, J., Andersson, S., 1997. Determination of soil displace-
structural failure during mechanical loadings. ment by measuring the pressure of a column of liquid. In:
Proceedings of the 14th ISTRO Conference, Pulawy, Poland,
pp. 47–50.
5. Conclusions Arvidsson, J., Keller, T., 2004. Soil precompression stress I. A survey
of Swedish arable soils. Soil Till. Res. 77, 85–95.
Our results showed that stress transmission in a Balakrishna, C.K., Srinavasa Murthy, B.R., Nagaraj, T.S., 1992. Stress
structured, undisturbed soil could be qualitatively distribution beneath rigid circular foundations on sands. Int. J.
Numer. Anal. Meth. Geomech. 16, 65–72.
described by the Boussinesq–Fröhlich equation: the Boussinesq, J., 1885. Application des potentiels à l’étude de l’équi-
contact stress determined the stresses in the topsoil, libre et des mouvements des solides élastiques. Gauthier-Villars,
while stresses in the subsoil were primarily dependent Paris, 30 p.
M. Lamandé et al. / Soil & Tillage Research 97 (2007) 91–106 105

Campbell, G.S., 1974. A simple method for determining unsaturated compaction due to agricultural field traffic including a synthesis of
conductivity from moisture retention data. Soil Sci. 117, analytical approaches. Soil Till. Res. 93, 391–411.
311–411. Kirby, J.M., 1999a. Soil stress measurement: Part I. Transducer in a
Craig, R.F., 2005. Craig’s Soil Mechanics, seventh ed. Spon Press, uniform stress field. J. Agric. Eng. Res. 72, 151–160.
London and New York, 447 p. Kirby, J.M., 1999b. Soil stress measurement: Part II. Transducer
Crowder, M.J., Hand, D.J., 1990. Analysis of repeated measures. beneath a circular loaded area. J. Agric. Eng. Res. 73, 141–149.
Monographs on Statistics and Applied Probability 41. Chapman Koolen, A.J., 1974. A method for soil compactibility determination. J.
and Hall, London, 257 p. Agric. Eng. Res. 19, 271–278.
Danfors, B., 1974. Packing I Alven (Compaction in the Subsoil). Koolen, A.J., Kuipers, H., 1983. Agricultural Soil Mechanics.
Report S24, Swedish Inst. Agric. Eng., Uppsala, Sweden, 91 p. Springer, New York (USA).
Dawidowski, J.B., Koolen, J., 2004. Computerized determination of Lamandé, M., Schjønning, P., Drøscher, P., Steffensen, F.S.F., Ras-
the preconsolidation stress in compaction testing of field core mussen, S.T., Koppelgaard, M.K., 2006. Measuring vertical stress
samples. Soil Till. Res. 31, 277–282. and displacement in two dimensions in soil during a wheeling
Dexter, A.R., Horn, R., Holloway, R., Jakobsen, B.F., 1988. Pressure event. In: Proceedings of the 17th Conference of the International
transmission beneath wheels in soils on the Eyre peninsula of Soil Tillage Research Organization, Kiel, Germany, September
South Australia. J. Terramech. 25, 135–147. 2006, pp. 293–297.
Dexter, A.R., 1988. Advances in characterization of soil structure. Soil Lebert, M., Horn, R., 1991. A method to predict mechanical strength
Till. Res. 11, 199–238. of agricultural soils. Soil Till. Res. 19, 275–286.
Ehlers, W., Claupein, W., 1994. Approaches towards conservation Olsen, H.J., 1994. Calculation of subsoil stresses. Soil Till. Res. 29,
tillage in Germany. In: Carter, M.R. (Ed.), Conservation Tillage in 111–123.
Temperate Agroecosystems. Lewis Publishers, Boca Raton, FL. R Development Core Team, 2005. R: A language and environment for
Fröhlich, O.K., 1934. Druckverteilung im Baugrunde (Pressure dis- statistical computing. R Foundation for Statistical Computing,
tribution in foundation soils). Springer Verlag, Wien. Vienna, Austria, ISBN 3-900051-07-0, URL: http://www.R-pro-
Gregory, A.S., Whalley, W.R., Watts, C.W., Bird, N.R.A., Hallett, ject.org.
P.D., Whitmore, A.P., 2006. Calculation of the compression index Ram, R.B., 1984. Pressure measurement in the soil under the load. Soil
and precompression stress from soil compression test data. Soil Till. Res. 4, 137–145.
Till. Res. 89, 45–57. Schjønning, P., 1986. Shear strength determination in undisturbed soil
Guérif, J., 1984. The influence of water-content gradient and structure at controlled water potential. Soil Till. Res. 8, 171–179.
anisotropy on soil compressibility. J. Agric. Eng. Res. 29, 367– Schjønning, P., 1991. Soil mechanical properties of seven Danish soils
374. (in Danish). Report No. S2176, The Danish Institute of Plant and
Gupta, S.C., Hadas, A., Voorhees, W.B., Wolf, D., Larson, W.E., Soil Science, Copenhagen, 33 pp.
Schneider, E.C., 1985. Field testing of a soil compaction model. Schjønning, P., 1999. Mechanical properties of Danish soils—a review
In: Proceedings of the International Conference on Soil Dynamics, of existing knowledge with special emphasis on soil spatial
Auburn, AL, USA, pp. 979–994. variability. In: Van den Akker, J.J.H., Arvidson, J., Horn, R.
Hadas, A., 1994. Soil compaction caused by high axle loads— (Eds.), Experiences with the Impact and Prevention of Subsoil
review of concepts and experimental data. Soil Till. Res. 29, Compaction in the European Community. Proceedings of the First
253–276. Workshop of the Concerted Action ‘Experiences with the Impact
Horn, R., 1981. Eine Methode zur Ermittlung der Druckbelastung von of Subsoil Compaction on Soil, Crop Growth and Environment
Böden anhand von Drucksetzungsversuchen. Zeitschrift für Kul- and Ways to Prevent Subsoil Compaction’, 28–30 May, 1998,
turtechnik und Flurbereinigung 22. Verlag Paul Parey, Berlin und Wageningen, The Netherlands, pp. 290–303.
Hamburg, pp. 20–26. Schjønning, P., Rasmussen, K.J., 2000. Soil strength and soil pore
Horn, R., 1993. Mechanical properties of structured unsatured soils. characteristics for direct drilled and ploughed soils. Soil Till. Res.
Soil Tech. 6, 47–75. 57, 69–82.
Håkansson, I., Reeder, R., 1994. Subsoil compaction by vehicles with Skidmore, E.L., Powers, D.H., 1982. Dry soil aggregate stability:
high axle load – extent, persistence and crop response. Soil Till. energy based index. Soil Sci. Soc. Am. J. 46, 1274–1279.
Res. 29, 277–304. Smith, D.L.O., Dickson, J.W., 1990. Contributions of vehicle weight
Jacobsen, O.H., 1989. Unsaturated hydraulic conductivity for some and ground pressure to soil compaction. J. Agric. Eng. Res. 46,
Danish soils: methods and characterization of soils. Tidsskrift for 13–29.
Planteavls Specialserie nr. S2030. Smith, R., Ellies, A., Horn, R., 2000. Modified Boussinesq’s equation
Keller, T., 2004. Soil compaction and soil tillage-studies in agricul- for nonuniform tire loading. J. Terramech. 37, 207–222.
tural soil mechanics. PhD thesis, Swedish University of Agricul- Söhne, W., 1953. Druckverteilung im Boden und Bodenformung unter
tural Sciences, Uppsala. Schleppereiffen (Pressure distribution in the soil and soil defor-
Keller, T., Arvidsson, J., 2004. Technical solutions to reduce the risk of mation under tractor tyres). Grundlagen der Landtechnik 5, 49–63.
subsoil compaction: effects of dual wheels, tandem wheels and Söhne, W., 1958. Fundamentals of pressure distribution and soil
tyre inflation pressure on stress propagation in soil. Soil Till. Res. compaction under tractor tyres. Agric. Eng. 39, 276–282, 290.
79, 191–205. Trautner, A., 2003. On soil behaviour during field traffic. PhD thesis,
Keller, T., Arvidsson, J., Dawidowski, J.B., Koolen, A.J., 2004. Swedish University of Agricultural Sciences, Uppsala.
Precompression stress II. A comparaison of different compaction Trautner, A., Arvidsson, J., 2003. Subsoil compaction caused by
tests and stress-displacement behaviour of the soil during wheel- machinery traffic on a Swedish Eutric Cambisol at different soil
ing. Soil Till. Res. 77, 97–108. water contents. Soil Till. Res. 73, 107–118.
Keller, T., Défossez, P., Weisskopf, P., Arvidsson, J., Richard, G., Ullidtz, 1998. Modelling Flexible Pavement Response and Perfor-
2007. SoilFlex: A model for prediction of soil stresses and soil mance. Polyteknisk Forlag, 205 p.
106 M. Lamandé et al. / Soil & Tillage Research 97 (2007) 91–106

van den Akker, J.J.H., 2004. SOCOMO: a soil compaction model to B.T. (Eds.), Managing Soil Quality: Challenges in Modern Agri-
calculate soil stresses and the subsoil carrying capacity. Soil Till. culture. CABI Publishing, Wallingford, UK, pp. 163–184.
Res. 79, 113–127. van Genuchten, M.Th., 1980. A closed-form equation for predicting
van den Akker, J.J.H., Schjønning, P., 2004. Subsoil compaction and the hydraulic conductivity of unsaturated soils. Soil Sci. Soc. Am.
ways to prevent it. In: Schjønning, P., Elmholt, S., Christensen, J. 44, 892–898.

Você também pode gostar