soilflex: a model for prediction of soil stresses and soil compaction due to agricultural field...
TRANSCRIPT
www.elsevier.com/locate/still
Soil & Tillage Research 93 (2007) 391–411
SoilFlex: A model for prediction of soil stresses and soil
compaction due to agricultural field traffic including
a synthesis of analytical approaches
Thomas Keller a,*, Pauline Defossez b, Peter Weisskopf c,Johan Arvidsson a, Guy Richard b,d
a Department of Soil Sciences, Swedish University of Agricultural Sciences (SLU),
P.O. Box 7014, 75007 Uppsala, Swedenb Institut National de la Recherche Agronomique (INRA), Unite d’Agronomie de Laon, Rue Fernand Christ,
02007 Laon Cedex, Francec Swiss Federal Research Station for Agroecology and Agriculture (FAL), Reckenholzstrasse 191, 8046 Zurich, Switzerland
d Institut National de la Recherche Agronomique (INRA), Soil Science Department of Orleans,
Avenue de la Pomme de Pin, BP 20619, Ardon, 45166 Olivet Cedex, France
Received 27 October 2005; received in revised form 19 May 2006; accepted 27 May 2006
Abstract
Soil compaction is one of the most important factors responsible for soil physical degradation. Soil compaction models are
important tools for controlling traffic-induced soil compaction in agriculture. A two-dimensional model for calculation of soil
stresses and soil compaction due to agricultural field traffic is presented. It is written as a spreadsheet that is easy to use and therefore
intended for use not only by experts in soil mechanics, but also by e.g. agricultural advisers. The model allows for a realistic
prediction of the contact area and the stress distribution in the contact area from readily available tyre parameters. It is possible to
simulate the passage of several machines, including e.g. tractors with dual wheels and trailers with tandem wheels. The model is
based on analytical equations for stress propagation in soil. The load is applied incrementally, thus keeping the strains small for each
increment. Several stress–strain relationships describing the compressive behaviour of agricultural soils are incorporated.
Mechanical properties of soil can be estimated by means of pedo-transfer functions. The model includes two options for
calculation of vertical displacement and rut depth, either from volumetric strains only or from both volumetric and shear strains. We
show in examples that the model provides satisfactory predictions of stress propagation and changes in bulk density. However,
computation results of soil deformation strongly depend on soil mechanical properties that are labour-intensive to measure and
difficult to estimate and thus not readily available. Therefore, prediction of deformation might not be easily handled in practice. The
model presented is called SoilFlex, because it is a soil compaction model that is flexible in terms of the model inputs, the constitutive
equations describing the stress–strain relationships and the model outputs.
# 2006 Elsevier B.V. All rights reserved.
Keywords: Model; Soil compaction; Soil stress; Soil displacement; Traffic; Contact area; Bulk density
* Corresponding author. Tel.: +46 18 67 12 10; fax: +46 18 67 27 95.
E-mail address: [email protected] (T. Keller).
0167-1987/$ – see front matter # 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.still.2006.05.012
1. Introduction
Soil degradation is a major environmental problem
worldwide. Soil compaction, i.e. decrease in pore space,
is one of the most important factors responsible for soil
T. Keller et al. / Soil & Tillage Research 93 (2007) 391–411392
Nomenclature
a half axis of the super ellipse (m)
A compactibility coefficient
Acontact contact area (m2)
Ai area carrying a point load (m2)
b half axis of the super ellipse (m)
B compactibility coefficient (kPa�1)
c cohesion (kPa)
C compactibility coefficient (kPa�1)
CA amplitude of the decay function (kN m�3)
Cc compression index (Mg m�3 kPa�1)
d rut depth (m)
D compactibility coefficient
E Young’s modulus of elasticity (kPa)
G shear modulus (kPa)
Hi horizontal point load (kN)
ji shear displacement (cm)
J1 first invariant of the strain tensor
k shear displacement coefficient (cm)
K modulus of the shear stress–strain curve
l(y) length of contact at certain position of y
(m)
m distance between the yield line and the
virgin compression line on the v� ln pplane
n parameter governing the shape of the
super ellipse
N specific volume at a mean normal stress
of 1 kPa
p mean normal stress (kPa)
Pi vertical point load (kN)
q deviator stress (kPa)
qa asymptotic value of the shear strength
(kPa)
qf failure deviator stress (kPa)
r distance between the point load and the
desired point (m)
Rf failure ratio
S1 desired degree of saturation (%)
Sk desired saturation at a reference stress (%)
v specific volume
vinit initial specific volume
vYL specific volume at the intersection of the
yield line and the recompression line
wðxÞ width of contact at certain position of x (m)
x horizontal distance along the driving
direction (m)
y lateral distance at right angles to the
driving direction (m)
z soil depth (m)
Greek letters
a order of the power-law function
g parameter governing the shape of the
decay function
d angle between the shear load vector and
the vertical plane that contains the posi-
tion vector from the shear load to the
desired point (8)DT slope of the bulk density versus degree of
water saturation curve (Mg m�3)
e strain
en natural volumetric strain
es shear strain
ev volumetric strain
ez vertical strain
u angle between the normal load vector and
the position vector from the point load to
the desired point (8)k recompression or swelling index (kPa�1)
k0 slope of the steeper recompression line
(kPa�1)
ln compression index (kPa�1)
m Poisson’s ratio
j concentration factor
r bulk density (Mg m�3)
rk reference bulk density (Mg m�3)
rs density of solids (Mg m�3)
r0 initial bulk density (Mg m�3)
s normal stress (kPa)
sk reference stress (kPa)
soct octahedral normal stress (kPa)
sr radial normal stress (kPa)
s1 major principal stress (kPa)
s3 minor principal stress (kPa)
sa applied stress (kPa)
t shear stress (kPa)
toct octahedral shear stress (kPa)
w angle of internal friction (8)
physical degradation (Pagliai et al., 2003). Therefore, it
is important to control the mechanical impacts of
agricultural machinery on soil structure in order to
reduce the risk of soil compaction.
Soil compaction models are important tools for
controlling soil compaction due to agricultural field
traffic. By means of such models, strategies and
recommendations for prevention of soil compaction
may be developed and specific advice may be given to
farmers and advisers.
Soil compaction models can be divided into two
categories, analytical and numerical (finite element)
T. Keller et al. / Soil & Tillage Research 93 (2007) 391–411 393
models (Defossez and Richard, 2002; Abu-Hamdeh and
Reeder, 2003). Analytical models are based on the work
of Boussinesq (1885), who developed equations for the
propagation of stress through elastic, homogeneous,
semi-infinite media. Frohlich (1934) introduced the so-
called concentration factor to account for the fact that
stresses in soils are more concentrated below the centre
of the applied load than in a fully elastic medium. From
the calculated stress, strain is then obtained by applying
a stress–strain relationship. Finite element models apply
continuum mechanics and may have the potential to
describe the mechanical behaviour of soil more
accurately. Stresses and strains are calculated simulta-
neously. However, compared with analytical models,
numerical models usually require more input para-
meters, which may be difficult to measure. A review of
existing soil compaction models can be found in
Defossez and Richard (2002).
Despite the limitations of the model assumptions,
analytical models have yielded satisfactory predictions of
stress propagation through soil and changes in soil
volume (e.g. Gupta et al., 1985; Johnson and Burt, 1990;
Koolen et al., 1992; O’Sullivan et al., 1999; Arvidsson
et al., 2001; Defossez et al., 2003; Keller and Arvidsson,
2004). Therefore, the analytical model approach is
justified. In addition, analytical models are usually easier
to use than finite element models, and can therefore be
used not only by experts in soil mechanics, but also by
e.g. agricultural advisors. This is a very important aspect
for the control of soil compaction in practice.
1.1. Characteristics of some existing analytical
models
In this section, we discuss the analytical soil
compaction models of Gupta and Larson (1982), van
den Akker (2004), Johnson and Burt (1990) and
O’Sullivan et al. (1999). The model of O’Sullivan
et al. (1999) is a further development of the model of
Table 1
Description of surface stress in some analytical soil compaction models
Reference Contact area Shape of stress d
Vertical stress
Gupta and Larson (1982) Elliptical Power-law functi
Johnson and Burt (1990) Rectangular Uniform or powe
(maximum stress
centre or under th
O’Sullivan et al. (1999) Circular Power-law functi
van den Akker (2004) Rectangular, elliptical,
or user-defined
Uniform, parabol
or user-defined
Smith (1985); therefore, we only discuss the former
here. The model of van den Akker (2004) is known as
‘SOCOMO’, while the model of O’Sullivan et al.
(1999) is known as ‘Compsoil’. The four models
discussed are widely used and referred to in the
literature (see Defossez and Richard, 2002).
The characteristics of the upper boundary condition
of the models discussed are presented in Table 1. In the
model of O’Sullivan et al. (1999), the contact area is
calculated from tyre and loading characteristics and
then converted to an equivalent circular area, while the
stress distribution is dependent upon soil water content
and bulk density as suggested by Sohne (1953).
The model of O’Sullivan et al. (1999) is one-
dimensional, i.e. stress is calculated under the tyre
centre only, whereas the models of Gupta and Larson
(1982), Johnson and Burt (1990) and van den Akker
(2004) are two-dimensional and stress is calculated in a
vertical plane parallel to the transverse or longitudinal
axis of the tyre.
O’Sullivan et al. (1999) use the analytical equations
proposed by Sohne (1953) to calculate the vertical stress
below a circular contact area. Gupta and Larson (1982),
Johnson and Burt (1990) and van den Akker (2004) use
the summation procedure as proposed by Sohne (1953).
Whereas Johnson and Burt (1990) and van den Akker
(2004) calculate the complete stress state including the
principal and octahedral stresses, Gupta and Larson
(1982) and O’Sullivan et al. (1999) only calculate the
major principal stress. O’Sullivan et al. (1999) use then
an empirical relationship to determine the intermediate
and minor principal stress in order to calculate the mean
normal stress.
Soil deformation is not calculated in the models of
Johnson and Burt (1990) and van den Akker (2004).
However, van den Akker (2004) estimates soil failure:
soil failure occurs when the precompression stress and/
or the shear strength are exceeded by the soil stresses.
Gupta and Larson (1982) describe volume change with
istribution No. of wheels
Horizontal stress
on (Sohne, 1953) Not considered 1
r-law function
either in the
e edge of the tyre)
Uniform or as a function
of soil shear strength
1
on (Sohne, 1953) Not considered 4 � 1
ic, trapezoidal, As for vertical stress 1
T. Keller et al. / Soil & Tillage Research 93 (2007) 391–411394
a relationship between the bulk density and the major
principal stress according to Larsson et al. (1980),
whereas O’Sullivan et al. (1999) use the relationship of
O’Sullivan and Robertson (1996) in terms of the
specific volume and the mean normal stress. O’Sullivan
et al. (1999) calculate vertical displacement from
volumetric strain, assuming the vertical strain to be
equal to the volumetric strain.
1.2. Objectives of the new model presented
in this paper
The analysis of the soil compaction models of
Gupta and Larson (1982), Johnson and Burt (1990),
O’Sullivan et al. (1999) and van den Akker (2004)
reveals that each of these models has its limitations.
Some models include stress–strain relationships, while
others do not. The same applies to the shear stress on
the soil surface. The contact area and the distribution
of stress over the contact area are described in many
different ways.
Therefore, the objectives of this work were to
include and further develop all useful aspects from these
existing models in a new, two-dimensional model for
prediction of stresses and compaction due to agricul-
tural field traffic. A further objective was to provide
model functions for estimation of the upper boundary
condition and soil mechanical parameters. The main
goal of the new model is to provide accurate predictions
of stress propagation and changes in bulk density. The
model, which we call ‘SoilFlex’, is presented in this
paper. We included different approaches for character-
isation of the stress–strain relationship that were
developed from compression tests on different agri-
cultural soils and comprise slightly different soil
mechanical parameters. The model allows for a direct
comparison between these stress–strain relationships.
We developed a model that allows for the simulation of
the passage of machine combinations (several wheels),
as well as different wheel configurations (dual wheels,
tandem wheels). Furthermore, we included functions to
estimate the contact area and the contact stresses from
tyre and loading characteristics, and pedo-transfer
functions to estimate soil mechanical properties that
are required for the stress–strain relationships. Our new
model is intended to be easy to use and can therefore be
used in practice, e.g. by agricultural advisers.
2. Structure of SoilFlex
SoilFlex is a two-dimensional model that calculates
the stress state, changes in bulk density and vertical
displacements in soil due to agricultural field traffic.
The model contains three main components. Firstly,
stress on the surface is described; both normal and shear
stresses are considered. Secondly, stress propagation
through soil is calculated analytically. Thirdly, soil
deformation is calculated as a function of stress. This
approach was also used by O’Sullivan et al. (1999). The
flow chart of SoilFlex is illustrated in Fig. 1.
2.1. Upper boundary condition: contact area and
contact stresses
The stresses are represented by an array of i point
loads, each having normal (Pi) and shear (Hi)
components, and acting at the centre point of their
respective areas, Ai. In SoilFlex, the vertical and
horizontal point loads are represented as matrices on
separate Excel sheets.
In addition to the possibility of entering any user-
defined contact area and stress distribution, the model
also offers several options to generate stress distribu-
tions (Fig. 1, Table 2). The shape of the contact area can
generally be described by a so-called super ellipse
(Hallonborg, 1996), which in an orthogonal co-ordinate
system is given as
xn
anþ yn
bn ¼ 1 (1)
where a and b are the half axes of the super ellipse and n
is a positive real number that governs the shape. For
a = b and n = 2, Eq. (1) defines a circle. In the general
case of a 6¼ b, the curve is a pure ellipse for n = 2,
whereas it grows towards a rectangle as n!1. In the
model, the contact width is assumed to be equal to the
tyre width (unless a circular contact area is chosen). The
contact area is calculated either (i) by dividing wheel
load by tyre inflation pressure, or from tyre and loading
characteristics according to (ii) O’Sullivan et al. (1999)
or (iii) Keller (2005). The parameter n of Eq. (1) has a
value of 2 in the cases (i) and (ii), and is a function of
tyre dimensions in case (iii). The distribution of vertical
stress can either be uniform, linear (the stress declines
linearly from a maximum at the contact area centre to
zero at the contact area edge), described by a power-law
function (the stress declines continuously from a max-
imum at the contact area centre to zero at the contact
area edge) according to Sohne (1953), or generated
from tyre properties according to Keller (2005). In all
cases, the wheel load is equal toP
Pi. If the stress
distribution is chosen according to Sohne (1953) in
SoilFlex, the order of the power-law function can be
calculated from soil conditions according to O’Sullivan
T. Keller et al. / Soil & Tillage Research 93 (2007) 391–411 395
Fig. 1. Flow chart of SoilFlex. For explanations see Table 2 (input parameters) and Table 3 (pedo-transfer functions (ptfs)).
T. Keller et al. / Soil & Tillage Research 93 (2007) 391–411396
Table 2
Options, characteristics and input parameters SoilFlex
Description Equations Input parameters Reference
Upper boundary condition
Contact area Distribution of vertical stress
Circular Uniform or linear Fwheel: wheel load (kg);
ptyre: tyre inflation pressure (kPa)
Power-law function Fwheel: wheel load (kg);
ptyre: tyre inflation pressure (kPa)
Sohne (1953)
Power-law function,
calculated from soil
conditions
Fwheel: wheel load (kg); ptyre: tyre inflation
pressure (kPa); ttyre: tyre type (X-ply or radial);
rsurf: bulk density surface layer (Mg m�3)
O’Sullivan et al.
(1999) and Sohne
(1953)
Elliptical Uniform Fwheel: wheel load (kg); ptyre: tyre inflation
pressure (kPa); wtyre: tyre width (m)
Power-law function Fwheel: wheel load (kg); ptyre: tyre inflation
pressure (kPa); wtyre: tyre width (m)
Sohne (1953)
Superelliptical Calculated from
tyre parameters
Eqs. (1)–(3) Fwheel: wheel load (kg); ptyre: tyre inflation
pressure (kPa); prec: recommended ptyre (kPa);
wtyre: tyre width (m); dtyre: tyre diameter (m)
Keller (2005)
User-defined User-defined
Contact area Distribution of
horizontal stress
Any shape Uniform or linear Ftraction: net traction (kN)
Calculated from
soil strength
Eq. (4) c: cohesion (kPa); w: angle of internal
friction (8); k: shear displacement coefficient
(cm); s: wheel slip (%)
Janosi (1962)
User-defined
Stress
propagation
Eqs. (5),
(A.1)–(A.16)
j: concentration factor Boussinesq
(1885), Cerruti
(1888), Frohlich
(1934) and
Sohne (1953)
Stress–strain
relationships
General input
parameters
rinit: initial dry bulk density (Mg m�3);
rs: density of solids (Mg m�3)
Larsson et al. (1980) Eq. (6) C: slope of VCL (Mg m�3 kPa�1);
S1: desired degree of saturation (%),
Sk: degree of saturation at sk and rk (%);
DT: slope of r vs. S curve at sk (kPa);
rk: reference bulk density (Mg m�3);
sk: reference stress (kPa)
Larsson et al.
(1980)
Bailey and
Johnson (1989)
Eqs. (7) and (8) A: compactibility coefficient; B: compactibility
coefficient (kPa�1); C: compactibility
coefficient (kPa�1); D: compactibility coefficient
Bailey and
Johnson (1989)
O’Sullivan and
Robertson (1996)
Eqs. (9)–(11) m: separation distance YL-VCLa (kPa);
N: specific volume at p = 1 kPa; k: slope
of RCLa (kPa�1); k0: slope of the ‘‘steeper’’
RCL (kPa�1); ln: slope of VCLa (kPa�1)
O’Sullivan and
Robertson (1996)
Shear failure Mohr-Coulomb Eq. (14) c: cohesion (kPa); w: angle of internal
friction (8)cited in Koolen
and Kuipers (1983)
Shear deformation Eqs. (20)–(24) c: cohesion (kPa); w: angle of internal friction
(degrees); G: shear modulus (MPa)
a YL: yield line (O’Sullivan and Robertson, 1996); VCL: virgin compression line; RCL: recompression line.
T. Keller et al. / Soil & Tillage Research 93 (2007) 391–411 397
Fig. 2. Geometrical relationship for the calculation of stress at a
desired point in the soil. Pi: vertical point load; Hi: horizontal point
load; sr: radial normal stress.
et al. (1999). Keller (2005) describes the transverse
distribution of vertical stress as
sðyÞ ¼ CA
�wðxÞ
2� y
�e�gððwðxÞ=2Þ�yÞ; 0 � y � wðxÞ
2
(2)
where CA and g are parameters and wðxÞ the width of
contact area at position x (for x = 0, wðxÞ is equal to the
tyre width). The longitudinal axis of the tyre along the
driving direction is assumed to be an axis of symmetry.
The longitudinal distribution of vertical stress is
described by a power-law function:
sðxÞ ¼ sx¼0;y
�1�
�x
lðyÞ2
�a�; 0 � x � lðyÞ
2(3)
where sx = 0,y is the stress under the tyre center, l(y) the
length of contact area at position y (for y = 0, l(y) is
equal to the tyre length), and a is a parameter. The
transverse axis of the tyre (at right angles to the driving
direction) is assumed to be an axis of symmetry. The
parameters of Eqs. (2) and (3) are calculated from
readily available tyre parameters (Keller, 2005).
Horizontal stress (shear stress) on the soil surface can
be either calculated from a given traction or from soil
strength, as defined by the cohesion, c, and the angle of
internal friction, w. Distribution of horizontal stress is
either uniform, linear from zero at the front edge of the
contact area to a maximum at the rear edge, or
calculated from the distribution of vertical surface stress
and shear strength. In the latter case, Hi, is calculated
according to Janosi (1962):
Hi ¼ Ai
�cþ Pi
Aitan ’
�ð1� e� ji=kÞ (4)
where ji is the shear displacement at Ai and k is the shear
displacement coefficient. The shear displacement is
calculated as ji = sxi, where s is the slip and xi is the
distance from the front edge of the contact area to the
position of Ai (Koolen and Kuipers, 1983). The slip, s, is
a model input parameter. For all distributions, net
traction is equal toP
Hi.
In contrast to many soil compaction models, the
input of e.g. dual and triple wheels, as well as double
and triple tandem wheels is possible and automatically
generated in SoilFlex. The additional parameters
needed for that are the gap between the dual/triple
wheels and the distance between the axles of double/
triple tandem wheels.
It is possible to input up to four different axles (one
axle can contain one single wheel or dual/triple wheels,
or tandem wheels) in SoilFlex to simulate multi-passage
of wheels.
2.2. Stress propagation
Calculation of stress is based on the work of
Boussinesq (1885), Cerruti (1888), Frohlich (1934) and
Sohne (1953). The contact area, Acontact, is divided into i
small elements with an area Ai each and a normal stress,
si, carrying the load Pi = siAi, which is treated as a point
load, following the approach of Sohne (1953). We
include shear stress on the soil surface in the same way,
i.e. each area Ai can also contain a shear point load, Hi.
The radial normal stress, sr,i, at depth z is then
sr;i ¼jPi
2pr2i
cosj�2 ui þjHi
2pr2i
sinj�2 ui cos di (5)
where j is the concentration factor (Frohlich, 1934), r
the distance from the point load to the desired point, u
the angle between the normal load vector and the
position vector from the point load to the desired point
and d is the angle between the shear load vector and the
vertical plane that contains the position vector from the
shear load to the desired point (Fig. 2). The propagation
of the stress components in the x-, y- and z-directions,
the principal stresses, the mean normal stress, p, the
deviator stress, q, and the octahedral shear stress, toct,
are calculated from Eq. (5), as given in Appendix A
(Eqs. (A.1)–(A.16)).
T. Keller et al. / Soil & Tillage Research 93 (2007) 391–411398
Fig. 3. Model of rebound and recompression according to O’Sullivan
and Robertson (1996) in terms of specific volume, v, and mean normal
stress, p (adapted from O’Sullivan et al. (1999)). VCL: virgin com-
pression line; RCL: recompression line; RCL0: steeper recompression
line; ln: compression index; k: recompression index; k0: slope of the
steeper recompression line.
2.3. Soil compaction: stress–strain relationships
We incorporate three different approaches (which can
be chosen optionally in SoilFlex, cf. Fig. 1 and Table 2)
for description of the sub-routine to calculate volumetric
strain and change in bulk density, namely the stress–
strain relationships by Larsson et al. (1980), Bailey and
Johnson (1989) and O’Sullivan and Robertson (1996),
which all describe the compressive behaviour of
agricultural soils. These three approaches may be
classified as ‘strain hardening models’, i.e. soil becomes
stronger due to straining. To our knowledge, the stress–
strain relationship of Bailey and Johnson (1989) has not
previously been included in any analytical soil compac-
tion model, but it has been incorporated in finite element
models (Bailey et al., 1995).
Larsson et al. (1980) describe volume change with a
relationship between the bulk density and the base 10
logarithm of the major principal stress:
r ¼ ½rk þDTðS1 � SkÞ� þ Cc log
�sa
sk
�(6)
where r is the compacted (final) bulk density corre-
sponding to an applied stress sa, rk the reference bulk
density corresponding to a reference stress sk on the
virgin compression line (VCL), DT the slope of the bulk
density versus degree of water saturation curve at sk, S1
the desired degree of saturation at sk, Sk the degree of
saturation corresponding to rk and sk and Cc is the
compression index, i.e. the slope of the VCL. Eq. (6)
describes the soil behaviour on the VCL only. We
assume the recompression index to equal zero in the
model of Larsson et al. (1980).
The model of Bailey and Johnson (1989) was
developed for cylindrical stress states and is given by
en ¼ ðAþ BsoctÞð1� e�CsoctÞ þ D
�toct
soct
�(7)
where en is the natural volumetric strain, soct the
octahedral (or mean) normal stress, toct the octahedral
shear stress, and A–D are the compactibility coeffi-
cients. For D = 0, the model of Bailey and Johnson
(1989) reduces to the model of Bailey et al. (1986),
which was developed for hydrostatic stress states. The
coefficients A–C have the same values in both models
(Bailey et al., 1995). In terms of bulk density, r, Eq. (7)
becomes (Bailey and Johnson, 1989):
ln r ¼ ln r0 ��ðAþ BsoctÞð1� e�CsoctÞ þ D
�toct
soct
��
(8)
where r0 is the initial bulk density.
The model of O’Sullivan and Robertson (1996) is
illustrated in Fig. 3. The VCL, the recompression line
(RCL) and the steeper recompression line (RCL0) are
given by
VCL : v ¼ N � ln ln p (9)
RCL : v ¼ vinit � k ln p (10)
RCL0 : v ¼ vYL � k0 ln p (11)
where v is the specific volume, p the mean normal
stress, N the specific volume at p = 1 kPa, ln the
compression index, vinit the initial specific volume,
k the recompression index, vYL the specific volume at
the intersection of the yield line and the recompression
line and k0 is the slope of the steeper recompression
line, which is given by (O’Sullivan and Robertson,
1996):
k0 ¼ffiffiffiffiffiffiffilnkp
(12)
The parameter m in Fig. 3 was found to be 1.3 (O’Sul-
livan and Robertson, 1996). Bulk density, r, is calcu-
lated from specific volume, v, according to
r ¼ rs
v(13)
where rs is the density of solids. Note that the VCL in
the model of O’Sullivan and Robertson (1996) is
defined in v� ln p space, whereas Larsson et al.
(1980) define the VCL in r � log s space, and hence
Cc in Eq. (6) is different from ln in Eq. (9).
T. Keller et al. / Soil & Tillage Research 93 (2007) 391–411 399
2.3.1. Pedo-transfer functions to estimate soil
mechanical parameters
In SoilFlex, the parameters of the stress–strain
relationships can be estimated by means of pedo-transfer
functions (ptfs). Table 3 gives an overview of ptfs for soil
mechanical parameters found in the literature. These
functions correlate soil mechanical properties with soil
physical properties, soil texture, etc. They may be used to
estimate soil mechanical properties or to analyse the
effect of e.g. soil texture on soil mechanical properties in
an average case. Not all ptfs of Table 3 were included in
SoilFlex, because some require many soil parameters
and/or complicated measurements.
The parameters of the model of Larsson et al. (1980;
Eq. (6)) can be estimated by the ptfs described in Gupta
and Larson (1982). Alternatively, the compression
index, Cc (Eq. (6)), can be estimated from the ptfs
developed by Lebert and Horn (1991) and Imhoff et al.
(2004). Parameters A–C of the model of Bailey and
Johnson (1989; Eq. (8)) can be estimated by the ptfs by
McBride (1989). We could not find any ptfs for
prediction of D (Eq. (8)). The parameters of the model
of O’Sullivan and Robertson (1996; Eqs. (9)–(11)) can
be estimated by the ptfs developed by O’Sullivan et al.
(1999) and Defossez et al. (2003).
2.4. Shear failure
In SoilFlex, shear failure can be calculated option-
ally (Table 2). Shear failure is calculated according to
the Mohr–Coulomb failure criterion. No failure of the
soil occurs if the following criterion is satisfied (Koolen
and Kuipers, 1983):
s3� s1 tan2
�45� � ’
2
�� 2c tan
�45� � ’
2
�(14)
where s1 and s3 are the major and minor principal
stress, respectively, w the angle of internal friction, and c
is the cohesion (kPa).
Note that the shear parameters cohesion and angle of
internal friction are no required input parameters for
model calculations of compaction, since shear para-
meters are not included in the calculations of changes in
bulk density (Section 2.3). We could not find any ptfs
for estimation of the shear parameters. Typical values
for w and c may be found in textbooks.
2.5. Vertical soil displacement and rut depth
Two options to calculate vertical soil displacement
are incorporated in SoilFlex. In the first option, uniaxial
strain state is assumed. According to Koolen and Kuipers
(1983), lateral strains are negligible in the subsoil, i.e.
the strain state in the subsoil is with good approximation
uniaxial. Assuming uniaxial strain state is a simple
approximation that has been used in analytical soil
compaction models (e.g. O’Sullivan et al., 1999).
However, different authors including O’Sullivan et al.
(1999), Defossez and Richard (2002) and Defossez et al.
(2003) have reported that this may result in under-
estimation of rut depth and that considering lateral
strains in the calculation would probably increase the
accuracy of the predictions. While lateral strains may
be small in the subsoil, they may be significant in the
topsoil. Therefore, a second option for calculation of
vertical displacement is included in the model. This
second option includes shear strains due to shear stresses
andallows for lateral deformation. To ourknowledge, this
aspect has not been considered in analytical soil com-
paction models, but is common in finite element models.
The first invariant of the strain tensor, J1, equals the
volumetric strain, ev, and yields (see e.g. Koolen and
Kuipers, 1983):
J1 ¼ ev ¼ e1 þ e2 þ e3 ¼ ex þ ey þ ez (15)
where e1, e2 and e3 are the normal strains in the principal
directions and ex, ey and ez are the normal strains in x-,
y- and z-direction. Note that ev is calculated as described
in Section 2.3.
In the first option, the strain state in soil is assumed to
be uniaxial, i.e. ex = ey = 0. Hence, Eq. (15) reduces to
ez ¼ ev (16)
where ez is the vertical strain.
In the second option, volumetric and shear strains are
calculated (following Atkinson and Bransby (1978), we
associate like invariants of stress and strain, i.e. shear
strain, es, with deviator stress, q, and volumetric strain,
ev, with mean normal stress, p). The strain state is
assumed to be biaxial (plane strain), i.e. the horizontal
strain in the plane of calculation is assumed to be zero.
This restriction has to be made due to numerical
reasons. Therefore, the model does not predict
horizontal displacements. However, this is an accep-
table simplification if the main goal is to predict vertical
displacements. Furthermore, the x-, y-, and z-directions
are assumed to be principal directions. Shear strain, es,
is then given as
es ¼ ez � ez (17)
where ez is the horizontal strain perpendicular to the
vertical plane of calculation. The vertical strain, ez, is
T.
Keller
eta
l./So
il&
Tilla
ge
Resea
rch9
3(2
00
7)
39
1–
41
14
00
Table 3
Pedo-transfer functions for estimation of soil mechanical properties
Reference Property No. of
samples
Variables Soil group
description
Regressions equations (pedo-transfer functions) r2
Gupta and
Larson (1982)
Slope of r vs. degree of
saturation curve at sk, DT
54 Silt content, %silt
(g 100 g�1); clay content,
%clay (g 100 g�1)
Fine-textured soils DT = 3.461 � 10�3 + 1.742 � 10�4 (%silt) � 2.980 � 10�6 (%silt)2 0.63
Coarse-textured soils DT = 3.217 � 10�3 + 3.251 � 10�4 (%clay) � 5.385 � 10�6 (%clay)2 0.70
Compression index, Cc
(uniaxial tests)
54 Clay content, %clay
(g 100 g�1)
Expanding clay Cc = 2.033 � 10�1 + 1.423 � 10�2 (%clay) � 1.447 � 10�4 (%clay)2 0.79
Non-expanding clay Cc = 1.845 � 10�1 + 1.205 � 10�2 (%clay) � 1.108 � 10�4 (%clay)2 0.89
McBride (1989) The three parameters of
the Bailey et al. (1986)
relationship: A–C
(uniaxial tests)
34 Sand content, SC (g kg�1);
clay content, CC (g kg�1);
organic matter, OM (g kg�1);
particle density, rP (Mg m�3);
Atterberg lower plastic limit,
LPL (g 100 g�1); plasticity
index, PI; initial bulk density,
rb (Mg m�3); initial void ratio,
e; gravimetric water content,
w (g g�1); degree of saturation, S
Non-plastic (low w) A = 0.184 � 0.0529 log CC � 0.203e 0.973
B ¼ �0:887þ 1:946 log rP þ 0:557 log rb þ 0:0895 log w 0.909
C ¼ 19:87� 10:11 log SCþ 1:355 log OMþ 14:55 log w 0.807
Plastic ðw>PLÞ A ¼ 12:88� 0:130wþ 0:003w2 0.927
B = �0.766 + 1.649 log rP 0.667
C = 16.35 � 1.34 log(OM � CC) � 7.574 log PI 0.987
Plastic
(w< PL, PI < 15)
A = 1.036 � 0.062 log(OM � CC) � 1.086 logL PL 0.968
B = �2.269 � 0.049 log(OM � CC) + 5.075 log rP + 0.190 log LPL 0.995
C = �175.8 + 348.7[log(CC)]�1 + 1.746 log(OM � CC)
+ 14.05 log PI + 11.11S
0.998
Plastic
(w< PL, PI � 15)A ¼ 6:14� 13:74½logðCCÞ��1 � 0:950 log LPL � 0:345 log w 0.967
B = 0.551 � 0.477 log LPL 0.614
C = 25.30 � 5.896 log SC � 8.622 log PI 0.974
Lebert and
Horn (1991)a
Precompression stress,
pc (uniaxial tests)
307 Internal friction, f (8); cohesion,
c (kPa); bulk density, rb (Mg m�3);
air capacity, Lk (%, v/v); available
water capacity, nFk (%, v/v);
non-available water capacity,
TW (%, v/v); saturated hydraulic
conductivity, kf (�103 cm s�1);
organic matter, OM (g 100 g�1)
Sand pc(pF = 1.8) = 438.10rb � 0.0008(f1.8)3 � 3.14TW
� 0.11(nFk1.8)2 � 465.60
0.778
pc(pF = 2.5) = 410.75rb � 0.0007(f1.8)3 � 3.41TW
� 0.35(nFk2.5)2 � 384.71
0.710
Sandy loam pc(pF = 1.8) = 169.30rb � 29.03(OM)0.5 + 6.45kf + 32.18 log(C1.8)
� 9.44f1.8 + 27.25sin(TW) + 119.74 log(nFk1.8) + 19.51
0.828
pc(pF = 2.5) = 89.50rb � 23.99(OM)0.5 � 2.89kf + 125.76 log(C2.5)
� 1.14f1.8 + 26.90 sin(TW) + 51.46 log(nFk2.5) + 77.25
0.874
Silt pc(pF = 1.8) =374.15rb � 4.10OM + 3.38Lk1.8 � 1.58(kf)�0.5
+ 1.79C1.8 + 1.09(TW) � 6.37(f1.8)0.67 + 0.088(nFk1.8)2 � 472.77
0.765
pc(pF = 2.5) = 460.71rb � 20.33OM + 9.08Lk2.5 � 2.38(kf)�0.5
+ 2.86C2.5 + 4.50(TW) � 20.96(f2.5)0.67 + 0.304(nFk2.5)2 � 610.62
0.847
Clay and clay
loam (<35%)
pc(pF = 1.8) = 0.843rb � 0.544(kf)0.33 + 0.022TW + 7.03 (C1.8)�1
+ 0.024f1.8 � 0.015nFk1.8 + 0.725
0.808
pc(pF = 2.5) = 0.844rb � 0.456(kf)0.33 � 0.026TW + 12.88 (C2.5)�1
+ 0.003f2.5 � 0.016nFk2.5 + 1.419
Clay and clay
loam (�35%)
pc(pF = 1.8) = 4.59rb � 1.02OM � 16.43(kf)0.33 + 0.31TW
� 1.57nFk1.8 + 3.55C1.8 + 1.18f1.8 � 18.03
0.774
pc(pF = 2.5) = 70.65rb � 0.55OM � 7.01(kf)0.33 + 1.32TW
� 1.08nFk2.5 + 1.72C2.5 + 1.05 f2.5 � 100.94
0.763
T.
Keller
eta
l./So
il&
Tilla
ge
Resea
rch9
3(2
00
7)
39
1–
41
14
01
Compression index, Cc
(uniaxial tests)
307 Clay and clay
loam (�35%)
Cc = 0.302 0.696
Clay and clay
loam (<35%)
Cc = 0.260 0.446
Sandy loam Cc = 0.235 0.523
Sand (�1%) Cc = 0.138 0.224
Sand (<1%) Cc = 0.145 0.146
Silt Cc = 0.168 0.510
Clay, loam and silt
with platy structure
Cc = 0.129 0.419
Kirby (1991)b Precompression stress, pc
(uniaxial tests)
170 Liquid index IL, void ratio at
pc, epc; saturation at epc, Sepc;
All soils pc = exp(5.856 � 4.352IL � 1.074epc) 0.682
Compression index, Cc
(uniaxial tests)
170 Void ratio at pc, epc;
saturation at epc, Sepc
All soils Cc = 0.229 + 0.1736epc � 0.400Sepc 0.514
Rebound index, k
(uniaxial tests)
170 Void ratio at pc, epc All soils k = �0.00905 + 0.0153epc 0.205
O’Sullivan
et al. (1999)
Specific volume
at p = 1 kPa, N
Gravimetric water content,
w (g 100 g�1)
Sandy loam N ¼ 2:430� 0:0055ðw� 11:2Þ2
Clay loam N ¼ 2:813� 0:0128ðw� 17:4Þ2
Slope of the VCL, l Specific volume at p = 1 kPa, N Sandy loam l = (N � 1.572)/17
Clay loam l = (N � 1.557)/26
Slope of the RCL, k Water content, w (g 100 g�1);
slope of the VCL, l
Sandy loam k ¼ l½0:119� ð0:082w=17ÞClay loam k ¼ l½0:119� ð0:082w=26Þ�
Defossez
et al. (2003)
Specific volume at
p = 1 kPa, N
Gravimetric water content,
w (g 100 g�1)
Loss N ¼ 1:9997þ 0:2629w� 0:02753w2 þ 0:00111w3
�1:6� 10�5w4
Calcareous N ¼ 2:4196þ 0:13767w� 0:02035w2 þ 0:00121w3
�2:4� 10�5w4
Slope of the VCL, l Gravimetric water content,
w (g 100 g�1)
Loss l ¼ 0:1402þ 0:00192wþ 0:00021w2 � 1:3� 10�5w3
Calcareous l ¼ 0:08051þ 0:03131w� 0:00502w2 þ 0:00031w3
�6� 10�6w4
Imhoff
et al. (2004)a
Precompression stress,
pc (uniaxial tests)
50 Clay content, CC (g 100 g�1);
initial bulk density, rb (Mg m�3);
water content, w (g g�1)
All soils pc ¼ �566:764þ 442:891rb þ 4:338CC� 773:057w 0.70
Compression index,
Cc (uniaxial tests)
Clay content, CC (g 100 g�1);
initial bulk density,rb (Mg m�3)
CC < 29.42 Cc = 0.248 + 0.006CC � 0.121rb 0.77
CC > 29.42 Cc = 0.416 � 0.121rb 0.77
r: bulk density; sk: reference stress on the virgin compression line (Larsson et al. (1980), Gupta and Larson (1982)); pF: logarithm of soil water potential in kPa; p: mean normal stress; VCL: virgin
compression line; RCL: recompression line.a The equations for the precompression stress were not included in SoilFlex.b Not included in SoilFlex.
T. Keller et al. / Soil & Tillage Research 93 (2007) 391–411402
Fig. 4. Deviator strain, es, as a function of deviator stress, q; measure-
ments (grey curve) of O’Sullivan and Robertson (1996) and approx-
imation (black curve) using Eq. (19). qa: asymptotic value of the shear
strength; qf: ultimate (failure) deviator stress.
obtained from Eqs. (15) and (17):
ez ¼ 12ðev þ esÞ (18)
Shear stress–strain relationships were measured by
O’Sullivan and Robertson (1996) in tri-axial pressure
cells. An example of a q � es curve is given in Fig. 4. We
propose to describe this relationship by the following
equation:
es ¼ �K ln
�1� q
qa
�(19)
where K is the modulus of the shear stress–strain curve
and qa is the asymptotic value of the shear strength. The
initial shear modulus, G, is the slope of the initial
section of the shear stress–strain curve. For tri-axial
strain conditions with ex = ey, the initial part yields
(Atkinson and Bransby, 1978):
G ¼ dq
3des
(20)
Hence, with respect to Fig. 4:
K ¼ qa
3G(21)
The shear modulus, G, may be calculated as (see e.g.
Atkinson and Bransby, 1978):
G ¼ E
2ð1þ mÞ (22)
where E is the Young’s modulus of elasticity and m is
the Poisson’s ratio. The asymptotic stress, qa, is defined
as
qa ¼qf
Rf
(23)
where Rf is the failure ratio, which is chosen to be
Rf = 0.9 (Brinkgreve, 2002) and qf is the ultimate (fail-
ure) deviator stress (kPa), which is obtained from
Eq. (14) as
qf ¼ s3½tan2ð45� þ 12’Þ � 1� þ 2c tanð45� þ 1
2’Þ (24)
where c is the cohesion and w is the angle of internal
friction. During elastic unloading a different path is
followed (Fig. 4), which can be formulated as
es � e0s ¼
K
qa
ðq� q0Þ (25)
where the superscript 0 indicates the state at the end of
unloading. The relationship for shear deformation pro-
posed here is similar to the hyperbolic relationship
proposed by Duncan and Chang (1970) for the hyper-
bolic model used in several finite element soil compac-
tion models (e.g. Chi et al., 1993).
From the vertical strain calculated either by
assuming uniaxial strain state (Eq. (16)) or by assuming
plane strain (Eq. (18)), vertical displacement and rut
depth are calculated by summation of vertical strains.
The bottom layer is fully fixed, i.e. it cannot move in any
direction. Therefore, it is important to choose the lower
boundary at a depth great enough where displacements
are zero. The rut depth, d, is given as
d ¼Xi¼n
i¼1
ez;i (26)
where ez,i is the vertical strain of layer i, i = 0 the bottom
layer and i = n is the surface layer.
2.6. Numerical aspects
The model is implemented in an Excel file. It
contains several macros that are written in Visual
Basic. As an option, the user may calculate the vertical
stress only, the stress state (without consideration of
soil deformation) or the complete stress state
including calculation of soil deformations and
displacements (Fig. 1). This has the advantage that
input and output can be chosen according to the
requirements needed. For example, for an a priori
comparative assessment of the impact of different
machines, the calculation of the vertical stress may be
sufficient. The computational time of such a calcula-
tion is short, and no soil mechanical parameters are
required. In a second step, site-specific calculations
(with the most suitable machine) may be made by
considering stress–strain relationships and calculation
of soil deformations.
T. Keller et al. / Soil & Tillage Research 93 (2007) 391–411 403
For the calculation of soil stress, the soil profile is
assumed to be homogeneous, i.e. the concentration factor
is the same at any depth. This might be a poor
approximation of a real soil. However, the introduction
of layers is in conflict with the analytical equations that
were developed by Boussinesq (1885) and Cerruti (1888)
for a semi-infinite, elastic, homogeneous, isotropic
medium. Frohlich (1934) introduced the concentration
factor to account for inelasticity of soil but the equations
are still only strictly valid for semi-infinite conditions.
Therefore, only one layer can be considered.
For the calculation of soil compaction and soil
deformation, the soil profile can be divided into a
number (maximum 10) of separate soil layers for
description of soil mechanical properties, i.e. input
parameters for the three different models to describe the
compressive behaviour of soil and the parameters
needed to describe the shear behaviour of soil. The
thickness of each layer can be chosen individually.
The load (surface stresses) is applied in a number of
increments as suggested by O’Sullivan et al. (1999). In
doing so, the strains are kept small for each increment,
thus overcoming conflicts with the basic assumptions of
the stress calculus. After each increment, displacement
is calculated and rut depth is updated before the next
increment. The contact area is divided into i small
squares, Ai, the size of which can be chosen (by default,
all Ai are squares of 0.05 m � 0.05 m).
The model is principally three-dimensional. By
default, stresses and strains are calculated in two
dimensions, namely in a plane perpendicular to the
driving direction (i.e. under the transverse axis of a tyre)
and/or in the driving direction (i.e. under the long-
itudinal axis of a tyre).
As mentioned elsewhere, the simulation of the
passage of several axles (each of which may contain one
single wheel or dual/triple wheels) is possible in
Fig. 5. Wheeling experiment on a moist Eutric Cambisol with a tyre of size 1
(a) measured vertical stress (symbols) and calculated stress (curve); (b) mea
uniaxial strain state (black curve), and assuming plane strain state (grey cu
SoilFlex. In this case, the soil undergoes elastic
unloading prior to the passage of the subsequent wheel;
the irrecoverable strains and displacements due to the
previous wheel form the initial soil conditions for the
subsequent wheel.
3. Model evaluation
3.1. Comparison of simulated and measured stress
and displacement
We compared measured with calculated vertical
stress (Fig. 5a) and vertical displacement (Fig. 5b)
due to a single passage of a wheel of a sugar beet
harvester (wheel load 86 kN, tyre inflation pressure
100 kPa) on a moist Eutric Cambisol. Soil parameters
are given in Table 4. A value for j of 5 was used. The
experiment is described in detail in Keller and
Arvidsson (2004). The stress distribution at the tyre–
soil interface was calculated according to Eqs. (2) and
(3). Bulk density was calculated applying the stress–
strain relationship of Robertson (1996; Eqs. (9)–(13)).
Vertical displacement was calculated using the two
options described in Section 2.5: either the vertical
strain was equal to the volumetric strain (Eq. (16)), or
Eq. (18) was used where strains due to shear stresses
are included in the calculation of vertical displace-
ment. The parameters used for the simulations are
given in Table 4.
The simulated stress agreed very well with the
measured stress (Fig. 5a). With both options, vertical
displacement in the subsoil was overestimated, whereas
it was underestimated in the topsoil, which resulted in
an underestimation of the rut depth (i.e. the vertical
displacement of the soil surface) (Fig. 5b). Similar
results were reported by Defossez et al. (2003) who
compared measured with simulated rut depth using
050/50 R32 with a load of 86 kN and an inflation pressure of 100 kPa;
sured vertical displacement (symbols) and calculated values assuming
rve). For soil parameters see Table 4. The error bars indicate S.E.M.
T. Keller et al. / Soil & Tillage Research 93 (2007) 391–411404
Table 4
Soil parameters used for simulations shown in Figs. 5 and 10
Parameter Symbol
(units)
Topsoil
(0–0.3 m depth)
Subsoil
(>0.3 m depth)
Texture (FAO) Loam Silty clay loam
Initial bulk density r0 (Mg m�3) 1.66 1.67
Model of O’Sullivan and Robertson (1996)
Specific volumea at p = 1 kPa N 1.811 1.777
Compression indexa ln (kPa�1) 0.068 0.060
Swelling indexa k (kPa�1) 0.0194 0.0137
Slope of the ‘steeper recompression line’a k0 (kPa�1) 0.0363 0.0286
Parametera m 0.8 0.9
Shear parameters
Cohesionb c (kPa) 60 100
Angle of internal frictionc w (8) 30 30
Elastic shear modulusd G (kPa) 500 800
a Calculated from oedometer tests by assuming s2 = s3 = 0.5s1 (Koolen and Kuipers, 1983).b Measured in situ by means of a shear vane apparatus.c Estimated.d Calculated from the Young’s modulus derived in oedometer tests by assuming a Poisson’s ratio of 0.3.
Fig. 6. Distribution of vertical stress on the soil surface (upper
boundary condition) below a tyre of size 710/70 R38 with a load
of 107 kN and an inflation pressure of 220 kPa in Gysi (2001) and with
SoilFlex.
‘Compsoil’ (O’Sullivan et al., 1999), which applies
Eq. (16). The underestimation of rut depth may be due
to the fact that ‘Compsoil’ neglects lateral displacement
(Defossez et al., 2003). However, rut depth was also
underestimated by considering lateral displacement
(Fig. 5b). This may be due to uncertain values of some
parameters, especially of the shear parameters. There-
fore, we performed a sensitivity analysis of the critical
mechanical parameters, which is described later in
Section 3.3.3.
3.2. Comparison of predicted with measured values
and with values predicted by other models
We compared simulations with SoilFlex with
measurements and simulations by Gysi (2001) who
measured and modelled stress, rut depth and change in
bulk density due to the passage of a single wheel of a
sugar beet harvester. Gysi (2001) used a critical state
soil mechanics model which was based on the finite
element method. Soil parameters for the simulations in
SoilFlex were taken from Gysi (2001). Avalue for j of 3
was used for the simulations in SoilFlex. The
distribution of the vertical stress on the soil surface
as used in Gysi (2001) and the distribution generated
from tyre parameters (Keller, 2005) used in SoilFlex are
shown in Fig. 6. The comparison of measured and
predicted mean normal stress and bulk density are
shown in Fig. 7. The values predicted by means of
SoilFlex agreed well with measurements by Gysi
(2001), as shown in Fig. 7a. It is interesting to note that
the mean normal stress calculated with SoilFlex using
the upper boundary condition (surface stresses) of Gysi
(2001) does not differ from the mean normal stress
calculated by Gysi (2001), who used a finite element
model. This shows that (i) the analytical approach for
calculation of stress distribution is justified and (ii) the
distribution of stresses on the soil surface strongly
affects the stress propagation. The latter is in agreement
with earlier studies (e.g. Keller and Arvidsson, 2004).
3.3. Sensitivity analyses
3.3.1. Concentration factor
The concentration factor, j (Eq. (5)), may often
be used as a fitting parameter, since it is a parameter
T. Keller et al. / Soil & Tillage Research 93 (2007) 391–411 405
Fig. 7. Wheeling experiment on a Eutric Cambisol with a tyre of size 710/70 R38 with a load of 107 kN and an inflation pressure of 220 kPa (Gysi, 2001);
(a) measured (white circles) and simulated (grey rhombi) mean normal stress, p, of Gysi (2001), simulated p with SoilFlex (black curve), and (black
broken curve) simulated p with SoilFlex using the upper boundary condition (surface stresses) of Gysi (2001); (b) measured initial bulk density (circles)
and final bulk density (triangles) of Gysi (2001), and simulated final bulk density with SoilFlex (black curve). The error bars indicate S.D.
that is not directly measurable. Horn (1990) showed
that j is greater the smaller the precompression stress
(i.e. the weaker the soil) and the greater the applied
load. This implies that j is not only dependent on
Fig. 8. Effect of concentration factor, j, on propagation of (a) the major pr
density with depth with the stress–strain relationship of (c) Larsson et al
characteristics as in Fig. 5.
soil properties, but also on the loading intensity.
Sohne (1953) and O’Sullivan et al. (1999) related j to
the bulk density and the water content of the soil in such
a way that j is greater the softer the soil. However,
incipal stress, s1, and (b) the mean normal stress, p; changes in bulk
. (1980) and (d) O’Sullivan and Robertson (1996); soil and loading
T. Keller et al. / Soil & Tillage Research 93 (2007) 391–411406
Trautner (2003) found an opposite behaviour, i.e. that j
is greater the harder the soil. Therefore, it is important
to analyse the sensitivity of the predicted stress
propagation on j.
It is well known that j changes the shape of ‘pressure
bulbs’ in such a way that the stress is more concentrated
around the load axis and therefore reaches greater
depths at greater values of j. It is worthwhile
mentioning that this holds true for the major principal
stress (Fig. 8a), but not necessarily for other stress
components. For example, the opposite is true for
horizontal stresses. Therefore, p may become smaller or
larger with increasing j, depending on the soil depth
(Fig. 8b) and as a consequence, the effect of j on soil
compaction is dependent on whether soil compaction is
described as a function of s1 (e.g. Larsson et al., 1980)
or p (e.g. O’Sullivan and Robertson, 1996) as shown in
Fig. 8c and d. This has to be considered when drawing
conclusions on the effect of the concentration factor on
soil compaction.
3.3.2. Compressive stress–strain relationship
We studied the effect of the stress–strain relation-
ship (as described in Section 2.3) on the change in
bulk density. Uniaxial strain state was assumed, i.e.
vertical displacement was calculated according to
Eq. (16). The soil mechanical parameters required for
the different models (i.e. the models of Larsson et al.
(1980), Bailey and Johnson (1989) and O’Sullivan
and Robertson (1996)) were measured for a chalky
soil with an initial bulk density of 1.1 Mg m�3 and an
Table 5
Measured soil mechanical parameters of a chalky soil used for simulations
Parameter
Model of Larsson et al. (1980)
Reference stress
Reference bulk density at sk
Degree of saturation corresponding to rk and sk
Desired degree of saturation
Slope of the bulk density vs. Sk curve at sk
Compression index
Model of O’Sullivan and Robertson (1996)
Specific volume at p = 1 kPa
Compression index
Swelling index
Slope of the ‘steeper recompression line’
Parameter
Model of Bailey et al. (1986)
Compactibility coefficient
Compactibility coefficient
Compactibility coefficient
initial water content of 0.20 g g�1 (corresponding to a
water potential of �10 kPa). The soil is described in
Defossez et al. (2003). The parameters of the model
of Larsson et al. (1980; Eq. (6)) were measured by
means of oedometer tests, while those of the models
of Bailey and Johnson (1989; Eq. (8)) and O’Sullivan
and Robertson (1996; Eqs. (9)–(13)) were measured
using tri-axial cells. Details of the soil mechanical
tests are given in Defossez et al. (2003). The
measured parameters are summarised in Table 5.
We used a value for j of 5. We did not measure the
compactibility coefficient D (Eq. (8)); therefore, we
applied the model of Bailey et al. (1986) rather than
the model of Bailey and Johnson (1989). For the
simulations we applied a total load of 10 kN with
a uniform surface stress of 100 kPa over a circular
area.
The results of the simulations are shown in Fig. 9.
The differences in simulated changes in bulk density
(Fig. 9a) are small between the simulations using the
models of Bailey et al. (1986) and O’Sullivan and
Robertson (1996) at depths shallower than 0.3 m. The
predicted change in bulk density was lowest when using
the model of Larsson et al. (1980). At depths greater
than 0.3 m, the model of Larsson et al. (1980) predicts
no volume change, while the model of O’Sullivan and
Robertson (1996) predicts volume changes down to
0.85 m depth. The model of Bailey et al. (1986) predicts
volume changes at any depth. This is obvious from
Eq. (8): as soon as soct > 0, r > r0 with the parameters
used (Table 5). An additional simulation using the
shown in Fig. 9
Symbol (units) Value
sk (kPa) 100
rk (Mg m�3) 1.25
Sk (%) 51.4
S1 (%) 44.2
DT (Mg�1 m3) 0.054
Cc (Mg m�3 kPa�1) 0.393
N 2.813
ln (kPa�1) 0.176
k (kPa�1) 0.0058
k0 (kPa�1) 0.0319
m 1.3
A �0.244
B (kPa�1) �1.62 � 10�4
C (kPa�1) 1.175 � 10�2
T. Keller et al. / Soil & Tillage Research 93 (2007) 391–411 407
Fig. 9. Changes in (a) bulk density and (b) vertical displacement of a chalky soil below a circular area that is uniformly loaded (load = 10 kN, contact
stress = 100 kPa) calculated with the stress–strain relationship of (light grey curve) Larsson et al. (1980); (dark grey curve) Bailey and Johnson
(1989); (black curve) O’Sullivan and Robertson (1996); (black broken curve) O’Sullivan and Robertson (1996) assuming k = k0 = 0.
model of O’Sullivan and Robertson (1996) was
performed with k = k0 = 0, as assumed for the other
models. This resulted in compaction to a smaller depth
compared with the ‘original’ simulation (k0 > k > 0,
Table 5), which demonstrates that in the model of
O’Sullivan and Robertson (1996) it is assumed that
compaction occurs even if the applied stress is
lower than the precompression stress (Fig. 3). The
differences in changes in bulk density are reflected
in the simulated vertical displacement and hence rut
depth (Fig. 9b).
3.3.3. Soil mechanical parameters
Based on the results of the simulation experiment
described in Section 3.1 and shown in Fig. 5, we
performed a sensitivity analysis to evaluate the two
Fig. 10. Measured (symbols) and calculated (lines) vertical displacement be
pressure of 100 kPa on a moist Eutric Cambisol (as in Fig. 5); (a) effect o
assuming uniaxial strain state: reference simulation with parameter values giv
(black curve) and ln = 0.07 (grey curve); (b) effect of the shear parameter
reference simulation with parameter values given in Table 4 (black broken c
(dark grey curve) and G = 100 kPa (black curve). The error bars indicate s
options for calculation of vertical displacement. For the
sensitivity analysis, we studied the following soil
mechanical parameters: the compression index, ln
(Eq. (9)); the cohesion, c (Eq. (24)); the angle of internal
friction, w (Eq. (24)); the elastic shear modulus, G
(Eq. (20)). Each of these parameters was analysed
individually, while the values of the other parameters
were as given in Table 4. The simulations were performed
with the stress–strain relationships of O’Sullivan and
Robertson (1996; Eqs. (9)–(13)). Loading characteristics
were as described in Section 3.1.
Fig. 10a shows the effect of the compression index,
ln, of the subsoil on the vertical displacement assuming
uniaxial strain state. We made simulations with
ln = 0.05 and 0.07, which had a strong effect on the
vertical displacement. Furthermore, the reference
low a tyre of size 1050/50 R32 with a load of 86 kN and an inflation
f the compression index, ln, of the subsoil on vertical displacement
en in Table 4 (black broken curve) and simulations made with ln = 0.05
s of the topsoil on vertical displacement assuming plane strain state:
urve), and simulations made with c = 10 kPa (light grey curve), w = 0
tandard error of the mean.
T. Keller et al. / Soil & Tillage Research 93 (2007) 391–411408
simulation assuming uniaxial strain state, with a value
for ln of 0.06 (Table 4) is shown in Fig. 10a.
Fig. 10b shows the effect of the cohesion, c, angle of
internal friction, w, and elastic shear modulus, G of the
topsoil on the vertical displacement assuming biaxial
strain state. Parameters c, w and G are the additional
input parameters needed when assuming biaxial strain
state compared with when assuming uniaxial strain
state. In addition to the reference simulation, we made
three different simulations where either c only, w only,
or G only was changed compared with the reference
simulation, and all other parameters had the values
given in Table 4. The values of the changed parameters
used in these three simulations were c = 10 kPa, w = 0
and G = 100 kPa, respectively. A decrease in soil
strength, i.e. a smaller c, w or G, significantly increases
shear deformation and thus the vertical displacement
compared with the simulated displacement obtained
from the measured soil parameters (reference simula-
tion assuming biaxial strain state) (Fig. 10b). Simula-
tions with higher c, w and G resulted in insignificantly
smaller displacements compared with the reference
simulation and are therefore not shown. This indicates
that the shear strength was probably overestimated.
Cohesion was obtained from measurements with a vane
shear apparatus, which may overestimate cohesion
(e.g. Schjønning and Rasmussen, 2000). Furthermore,
we did not measure w, and for the calculation of G we
had to assume a value for the Poisson’s ratio. Therefore,
there is a large uncertainty in the parameter values,
which may explain discrepancies between measured
and simulated vertical displacements.
Note that the reference simulations in Figs. 10a and
b slightly differ from each other, as also shown in
Fig. 5b. This is because the reference simulation shown
in Fig. 10a represents the reference simulation for
uniaxial strain state (no shear parameters needed),
while the one shown in Fig. 10b represents the
reference simulation for biaxial strain state (shear
parameters are required).
It may be concluded that accurate prediction of
displacements is strongly dependent on accurate values
of the soil mechanical parameters. However, some
parameters are difficult and/or work-intensive to
measure. Including shear deformations in the calcula-
tion of vertical displacement (Eqs. (17)–(25)) may have
the potential to provide more accurate predictions, but
requires additional soil parameters (c, w and G), which
in many cases are not readily available. Therefore, two
options are available in SoilFlex, which can be utilised
according to the available soil parameters and the
required accuracy.
4. Summary and conclusions
A two-dimensional soil compaction model, ‘Soil-
Flex’, is presented. We use the name SoilFlex, because
it is a soil compaction model that is flexible in terms of
the description of the stresses on the soil surface,
modelling of the stress–strain relationship, estimation
of mechanical soil properties using pedo-transfer
functions, and model output; and because the user
can e.g. easily add pedo-transfer functions to the model.
With the model, the mechanical flexibility of soil may
be studied. The flow chart of SoilFlex is illustrated in
Fig. 1. The model options and their respective input
parameters are summarised in Table 2. A copy of the
Excel spreadsheet is available by e-mail from the
authors. SoilFlex is not a commercial product. No
liability is accepted by the authors.
The model is easy to use, which makes it applicable
not only in education and research, but also for
agricultural advisers. This is of great importance for the
control of soil compaction in practice. With SoilFlex, it
is possible to simulate the passage of machine
combinations that are used in practice, including e.g.
tractors with dual wheels, which is a significant aspect
and an omission in previous models. The model allows
for a realistic estimation of the stresses at the tyre–soil
interface that can be estimated from tyre parameters.
This is not accounted for in either existing finite
element or analytical soil compaction models. How-
ever, it has been shown that this is an important aspect
for accurate predictions of stress propagation (e.g.
Keller and Arvidsson, 2004). Soil mechanical para-
meters required for characterisation of the stress–strain
relationship can be estimated by means of pedo-
transfer functions, which is a great help in many
situations where data are missing. However, there are
only a few pedo-transfer functions for soil mechanical
parameters available; the performances of which are
not well evaluated and therefore their reliabilities are
not well known.
With regard to the accuracy of predicted stresses and
volume changes, the analytical approach as used in
SoilFlex was shown not to be inferior to finite element
models. A weak point of the analytical solution may be
the concentration factor, as it is not a directly
measurable soil parameter. However, finite element
models include e.g. the Poisson’s ratio, which may be
very difficult to measure with standard testing equip-
ment. With the analytical approach, we can only
consider one homogeneous layer for stress propagation,
which may be a serious limitation. However, the error
may not be large in many cases. Nevertheless, for
T. Keller et al. / Soil & Tillage Research 93 (2007) 391–411 409
strongly heterogeneous soil conditions the model may
yield less accurate predictions, and it could be the
subject of future research to define the limits of
application of analytical models. In SoilFlex, we
consider strain due to shear stresses and lateral
displacement, thus increasing the accuracy of the
prediction of the rut depth, which is not done in previous
analytical models. However, computation results of soil
deformation strongly depend on soil mechanical
properties. The lack of easily accessible and represen-
tative soil mechanical properties is a major obstacle to
the prediction of soil deformation in practice. As stated
by van den Akker (2004), there is a need for
development of pedo-transfer functions for estimation
of soil mechanical properties. Some generally valid
aspects on the reliability and use of pedo-transfer
functions are discussed for soil hydraulic properties by
e.g. Wosten et al. (1999).
Future refinements of the model could include a
function for estimation of the traction (shear stresses on
the soil surface) from draught force requirement of tillage
implements, the incorporation of more (better) pedo-
transfer functions for estimation of the soil mechanical
parameters and the consideration of the factor of time
(deformation is a time-dependent process).
Agricultural field traffic is an important aspect of soil
management. Future applications may therefore include
coupling SoilFlex with SISOL (Roger-Estrade et al.,
2000), a model that simulates the structural changes of
soil over time as a function of management practices.
Acknowledgments
Mr. M.F. O’Sullivan of the Scottish Agricultural
College (SAC), Midlothian, U.K., is thanked for kindly
providing the code of the Compsoil model (O’Sullivan
et al., 1999). Dr. A.J. Koolen of the Wageningen
University, Wageningen, The Netherlands, and Mr.
Hubert Boizard of the Institut National de la Recherche
Agronomique (INRA), Estrees-Mons, France, are
thanked for helpful discussions. Financial support
was received from the French National Institute for
Agricultural Research (INRA), the Swedish Research
Council for Environment, Agricultural Sciences and
Spatial Planning (Formas) and the Swedish Farmers’
Foundation for Agricultural Research (SLF), which is
gratefully acknowledged.
Appendix A
The stress components in the x-, y- and z-directions at
depth z are calculated from Eq. (5), following the
approach of Sohne (1953):
sz ¼Xi¼n
i¼0
ðszÞi ¼Xi¼n
i¼0
sr;i cos2 ui (A.1)
sx ¼Xi¼n
i¼0
sh;i cos2 di ¼Xi¼n
i¼0
sr;i sin2 ui cos2 di (A.2)
sy ¼Xi¼n
i¼0
sh;i sin2 di ¼Xi¼n
i¼0
sr;i sin2 ui sin 2di (A.3)
txy ¼Xi¼n
i¼0
sh;i cos di sin di ¼Xi¼n
i¼0
sr;isin2 ui cos di sin di
(A.4)
txz ¼Xi¼n
i¼0
ti cos2 di ¼Xi¼n
i¼0
sr;i cos ui sin ui cos2 di (A.5)
tyz ¼Xi¼n
i¼0
ti sin2 di ¼Xi¼n
i¼0
sr;i cos ui sin ui sin2 di (A.6)
The principal stresses s1, s2 and s3 are then calculated
with help of the three stress invariants I1, I2 and I3 (see
e.g. Koolen and Kuipers, 1983):
I1 ¼ sx þ sy þ sz ¼ s1 þ s2 þ s3 (A.7)
I2 ¼ sxsy þ sxsz þ sysz � t2xy � t2
xz � t2yz
¼ s1s2 þ s1s3 þ s2s3 (A.8)
I3 ¼ sx þ sy þ sz þ 2txy þ 2txz þ 2tyz � sxt2yz
� syt2xz � szt
2xy
¼ s1s2s3
(A.9)
and (Bardet, 1997):
cos 3f ¼ 2I31 � 9I1I2 þ 27I3
2ðI21 � 3I2Þ3=2
(A.10)
to (Bardet, 1997):
s1 ¼I1
3þ 2
3
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiI21 � 3I2
qcos f (A.11)
s2 ¼I1
3þ 2
3
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiI21 � 3I2
qcos
�2p
3� f
�(A.12)
s3 ¼I1
3þ 2
3
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiI21 � 3I2
qcos
�2p
3þ f
�(A.13)
T. Keller et al. / Soil & Tillage Research 93 (2007) 391–411410
From the principal stresses, we calculate the mean
normal stress, p (also called octahedral stress, soct)
(see e.g. Atkinson and Bransby, 1978):
p ¼ soct ¼ 13ðs1 þ s2 þ s3Þ (A.14)
the deviator stress, q:
q ¼ 1ffiffi2p
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiðs1 � s3Þ2 þ ðs1 � s2Þ2 þ ðs2 � s3Þ2
q
(A.15)
and the octahedral shear stress, toct:
toct ¼ 13
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiðs1 � s3Þ2 þ ðs1 � s2Þ2 þ ðs2 � s3Þ2
q(A.16)
References
Abu-Hamdeh, N.H., Reeder, R.C., 2003. Measuring and predicting
stress distribution under tractive devices in undisturbed soil.
Biosyst. Eng. 85, 493–502.
Arvidsson, J., Trautner, A., van den Akker, J.J.H., Schjønning, P.,
2001. Subsoil compaction caused by heavy sugarbeet harvesters
in southern Sweden. II. Soil displacement during wheeling
and model computations of compaction. Soil Tillage Res. 60,
79–89.
Atkinson, J.H., Bransby, P.L., 1978. The Mechanics of Soil—An
Introduction to Critical State Soil Mechanics. McGraw-Hill,
London, 375 pp.
Bailey, A.C., Johnson, C.E., 1989. A soil compaction model for
cylindrical stress states. Trans. ASAE 32, 822–825.
Bailey, A.C., Johnson, C.E., Schafer, R.L., 1986. A model for agri-
cultural soil compaction. J. Agric. Eng. Res. 33, 257–262.
Bailey, A.C., Raper, R.L., Johnson, C.E., Burt, E.C., 1995. An
integrated approach to soil compaction prediction. J. Agric.
Eng. Res. 61, 73–80.
Bardet, J.P., 1997. Experimental Soil Mechanics. Prentice-Hall Inc.,
New Jersey.
Boussinesq, J., 1885. Application des potentiels a l’etude de l’equi-
libre et du mouvement des solides elastiques. Gautiher-Villars,
Paris, 30 pp.
Brinkgreve, R.B.J., 2002. PLAXIS Version 8, Material Models Man-
ual. Balkema, The Netherlands, ISBN: 90 5809 508 8.
Cerruti, V., 1888. Sulla deformazione di un corpo elastico isotropo per
alcune speciali condizioni ai limiti. Mathematica Acc. r. de’Lin-
cei, Rom.
Chi, L., Tessier, S., Lague, C., 1993. Finite element modelling of soil
compaction by liquid manure spreaders. Trans. ASAE 36, 637–
644.
Defossez, P., Richard, G., 2002. Models of soil compaction due to
traffic and their evaluation. Soil Tillage Res. 67, 41–64.
Defossez, P., Richard, G., Boizard, H., O’Sullivan, M.F., 2003.
Modeling change in soil compaction due to agricultural
traffic as function of soil water content. Geoderma 116, 89–
105.
Duncan, J.M., Chang, C.Y., 1970. Nonlinear analysis of stress and
strain in soils. J. Soil Mechanics and Foundations Division ASCE
95 (SM5) 1629–1653.
Frohlich, O.K., 1934. Druckverteilung im Baugrunde. Springer Ver-
lag, Wien, 178 pp.
Gupta, S.C., Hadas, A., Voorhees, W.B., Wolf, D., Larson, W.E.,
Schneider, E.C., 1985. Field testing of a soil compaction model.
In: Proceedings of the International Conference on Soil Dynamics,
Auburn, AL, USA, pp. 979–994.
Gupta, S.C., Larson, W.E., 1982. Predicting soil mechanical behaviour
during tillage. In: Predicting Tillage Effects on Soil Physical
Properties and Processes, American Society of Agronomy, pp.
151–178 (Special Publication 44, Chapter 10).
Gysi, M., 2001. Compaction of a Eutric Cambisol under heavy wheel
traffic in Switzerland: field data and a critical state soil mechanics
model approach. Soil Tillage Res. 61, 133–142.
Hallonborg, U., 1996. Super ellipse as tyre–ground contact area. J.
Terramech. 33, 125–132.
Horn, R., 1990. Structure effects on strength and stress distribution in
arable soils. In: Proceedings International Summer Meeting
ASAE, Columbus, OH, June 24–27, 1990, pp. 8–20.
Imhoff, S., Da Silva, A.P., Fallow, D., 2004. Susceptibility to compac-
tion, load support capacity, and soil compressibility of Hapludox.
Soil Sci. Soc. Am. J. 68, 17–24.
Janosi, Z., 1962. Theoretical analysis of the performance of tracks and
wheels operating on deformable soil. Trans. ASAE 5 133–134,
146.
Johnson, C.E., Burt, E.C., 1990. A method of predicting soil stress
state under tires. Trans. ASAE 33, 713–717.
Keller, T., 2005. A model to predict the contact area and the distribu-
tion of vertical stress below agricultural tyres from readily-avail-
able tyre parameters. Biosyst. Eng. 92, 85–96.
Keller, T., Arvidsson, J., 2004. Technical solutions to reduce the risk of
subsoil compaction: effects of dual wheels, tandem wheels and
tyre inflation pressure on stress propagation in soil. Soil Tillage
Res. 79, 191–205.
Kirby, J.M., 1991. Critical-state soil mechanics parameters and their
variation for Vertisols in eastern Australia. J. Soil Sci. 42, 487–
499.
Koolen, A.J., Kuipers, H., 1983. Agricultural Soil Mechanics.
Advanced Series in Agricultural Sciences, vol. 13. Springer-
Verlag, Berlin, 241 pp.
Koolen, A.J., Lerink, P., Kurstjens, D.A.G., van den Akker, J.J.H.,
Arts, W.B.M., 1992. Prediction of aspects of soil–wheel systems.
Soil Tillage Res. 24, 381–396.
Larsson, W.E., Gupta, S.C., Useche, R.A., 1980. Compression of
agricultural soils from eight soil orders. Soil Sci. Soc. Am. J. 44,
450–457.
Lebert, M., Horn, R., 1991. A method to predict the mechanical
strength of agricultural soils. Soil Tillage Res. 19, 275–286.
McBride, R.A., 1989. Estimation of density–moisture–stress func-
tions from uniaxial compression of unsaturated, structured soils.
Soil Tillage Res. 13, 383–397.
O’Sullivan, M.F., Henshall, J.K., Dickson, J.W., 1999. A simplified
method for estimating soil compaction. Soil Tillage Res. 49, 325–
335.
O’Sullivan, M.F., Robertson, E.A.G., 1996. Critical state parameters
from intact samples of two agricultural soils. Soil Tillage Res. 39,
161–173.
Pagliai, M., Marsili, A., Servadio, P., Vignozzi, N., Pellegrini, S.,
2003. Changes in some physical properties of a clay soil in central
Italy following the passage of rubber tracked and wheeled tractors
of medium power. Soil Tillage Res. 73, 119–129.
Roger-Estrade, J., Richard, G., Boizard, H., Boiffin, J., Caneill, J.,
2000. Modelling structural changes in tilled topsoil over time
T. Keller et al. / Soil & Tillage Research 93 (2007) 391–411 411
as a function of cropping systems. Eur. J. Soil Sci. 51, 455–
474.
Schjønning, P., Rasmussen, K.J., 2000. Soil strength and soil pore
characteristics for direct drilled and ploughed soils. Soil Tillage
Res. 57, 69–82.
Smith, D.L.O., 1985. Compaction by wheels: a numerical model for
agricultural soils. J. Soil Sci. 36, 621–632.
Sohne, W., 1953. Druckverteilung im Boden und Bodenverformung
unter Schlepperreifen. Grundlagen der Landtechnik 5, 49–63.
Trautner, A., 2003. On soil behaviour during field traffic. Agraria 372.
Ph.D. Thesis. Swedish University of Agricultural Sciences,
Uppsala, Sweden.
van den Akker, J.J.H., 2004. SOCOMO: a soil compaction model to
calculate soil stresses and the subsoil carrying capacity. Soil
Tillage Res. 79, 113–127.
Wosten, J.H.M., Lilly, A., Nemes, A., Le Bas, C., 1999. Development
and use of a database of hydraulic properties of European soils.
Geoderma 90, 169–185.