American Journal of Botany 94(9): 1506–1514. 2007.

A GENERIC 3D FINITE ELEMENT MODEL OF TREE ANCHORAGE

INTEGRATING SOIL MECHANICS AND REAL ROOT SYSTEM
ARCHITECTURE1

LIONEL X. DUPUY,2,3,6 THIERRY FOURCAUD,4 PATRICK LAC,2 AND ALEXIA STOKES5

2
Université Bordeaux 1, Mixed Research Unit: Unité Sciences du Bois et des Biopolymères (UMR US2B), Talence, F-33405
France; 3Institut pour le Développement Forestier (IDF), Paris, F-75007 France; 4Centre de Coopération International en
Recherche Agronomique pour le Développement (CIRAD), Mixed Research Unit: Botanique et Bioinformatique de
l’Architecture des Plantes (UMR AMAP), TA A-51/PS2, Boulevard de la Lironde, 34398 Montpellier Cedex 5, France; and
5
Institute National de la Recherche Agronomique (INRA), Mixed Research Unit: Botanique et Bioinformatique de l’Architecture
des Plantes (UMR AMAP), TA A-51/PS2, Boulevard de la Lironde, 34398 Montpellier Cedex 5, France

Understanding the mechanism of tree anchorage in a forest is a priority because of the increase in wind storms in recent years
and their projected recurrence as a consequence of global warming. To characterize anchorage mechanisms during tree uprooting,
we developed a generic finite element model where real three-dimensional (3D) root system architectures were represented in a 3D
soil. The model was used to simulate tree overturning during wind loading, and results compared with real data from two poplar
species (Populus trichocarpa and P. deltoides). These trees were winched sideways until failure, and uprooting force and root
architecture measured. The uprooting force was higher for P. deltoides than P. trichocarpa, probably due to its higher root volume
and thicker lateral roots. Results from the model showed that soil type influences failure modes. In frictional soils, e.g., sandy soils,
plastic failure of the soil occurred mainly on the windward side of the tree. In cohesive soils, e.g., clay soils, a more symmetrical
slip surface was formed. Root systems were more resistant to uprooting in cohesive soil than in frictional soil. Applications of this
generic model include virtual uprooting experiments, where each component of anchorage can be tested individually.

Key words: biomechanics; poplar; Populus; root anchorage; tree stability; windthrow.

In recent years, attention has focused on the consequences of
climate change on the composition and sustainability of forest
ecosystems around the world. One likely consequence is an
increase in the number and intensity of wind storms (Bloom-
field, 2000), which would cause major economic damage and
disturbance to temperate forests. Although much experimental
research has been carried out on tree anchorage and
biomechanics, particularly since the storms Lothar and Martin
hit Europe in 1999 (Cucchi et al., 2004; Achim et al., 2005;
Nicoll et al., 2005; Stokes et al., 2005; Brüchert and Gardiner,
2006; Peltola, 2006), mechanistic numerical models of such
phenomena are lacking (Blackwell et al., 1990; Dupuy et al.,
2005a, b; Fourcaud et al., 2007). The use of such models is
three-fold: they can be used to investigate the contribution of
different plant or soil parameters to tree anchorage; virtual
experiments can be carried out, thus reducing destructive and
expensive fieldwork; and finally, the models can be incorpo-
rated into decision support systems for forest managers
(Gardiner et al., 2000; Cucchi et al., 2005). Although the
initial development of such models is time consuming, once
calibrated, they could be adapted fairly easily to different
species in a variety of situations. To our knowledge,
1
Manuscript received 5 January 2007; revision accepted 10 July 2007.
The authors thank M. Maitrot and C. Drenou (IDF) for help in the
measurements of root architecture and F. Charnet (IDF) for data on soil
properties at Chinon. Funding was provided by a French governmental
CIFRE bursary, the poplar ECOFOR project (no 2002.06), and the E.U.
project Eco-Slopes (QLK5-2001-00289). Authors are also grateful to the
two anonymous reviewers for very constructive advice that helped
improve this manuscript.
6
Author for correspondence (e-mail: lxd20@cam.ac.uk), present
address: Department of Plant Sciences, Downing Street, Cambridge
CB2 3EA, UK, phone: þ44 01223-7-66545
1506
mechanistic tree anchorage models based on a description of
root architecture have not yet been linked to field data, but
would be a major step toward developing a powerful tool to
predict tree uprooting during storms.
Vulnerability to uprooting has been traditionally investigated
at the population level, by correlating forest damage to certain
stand characteristics, i.e., tree species, silviculture, soil type, and
wind speed (Putz et al., 1983; Ruel, 2000; Cucchi and Bert,
2003; Mayer et al., 2005). However, this empirical analysis of
windthrow is incomplete because root system characteristics
and other variables influencing tree stability (Khuder et al.,
2007) are not always taken into account. Experimental studies
whereby trees are winched until failure and the force necessary
to uproot or cause stem breakage is measured are useful in that
they provide detailed data at the tree level (Fraser, 1962; Stokes,
1999; Cucchi et al., 2004; Nicoll et al., 2005; Stokes et al., 2005;
Peltola, 2006). Nevertheless, even if root system architecture is
measured, data are often incomplete because roots are usually
damaged during the tests (Stokes et al., 2007). The role of root
system architecture in tree anchorage has been alluded to in
studies of root system morphology, but without the appropriate
experimental data, it is difficult to quantify the influence of each
parameter during the uprooting process (Nicoll and Ray, 1996;
Mickovski and Ennos, 2003; Danjon et al., 2005; Di Iorio et al.,
2005; Soethe et al., 2006; Khuder et al., 2007).
While root architecture strongly affects tree anchorage, soil
physical and mechanical properties are major determinants of
uprooting. Because the root–soil plate of a tree is a composite
structure, soil type and its interaction with the root system will
contribute to determining the overall anchorage capacity of a
tree (Moore, 2000). When soil is waterlogged, both shear
strength and the root–soil bond are reduced, and therefore, the
entire system fails more easily during wind loading (Coutts,
1986). In dry soils or soils with a high shear strength, trees tend
September 2007] D UPUY ET AL .—3D TREE ANCHORAGE MODEL
to be better anchored but the stem breaks during loading,
whereas trees in waterlogged soils or soils with a lower shear
strength are usually uprooted (Coutts, 1986; Moore, 2000). In
simulations of uprooting experiments of simplified model trees,
soil internal friction angle modified the position of the root
system rotation axis during overturning (Dupuy et al., 2005b).
Such an adjustment to the mechanics of the composite root–soil
system in turn strongly influences the force required to cause
failure of this system. However, although the models by Dupuy
et al. (2005a, b) and others (Ennos, 1990; Niklas et al., 2002;
Fourcaud et al., 2007) explicitly consider soil and root system
properties, the models do not use real root system architectures
in uprooting simulations. Hence, testing models that can
accurately predict overturning behavior of trees in different
types of soils is a priority.
The development of automatic and generic numerical
methods that use accurate descriptions of plant morphology
for biomechanical analyses can be extremely useful, and such
models have already been applied to both real and simulated
architectures subjected to mechanical loading in different
environments (Fourcaud et al., 2003a, b; Sellier et al., 2006).
We focused on the use of finite element modelling (FEM) to
simulate the uprooting mechanism of realistic root architecture
in different soil types.
The validation of FEM models with real data is, however,
particularly difficult as mechanical testing often greatly
damages the root system. Therefore, trees selected for complete
uprooting were different from trees used for root system
measurements. We also chose two poplar cultivars (Populus
trichocarpa Torr. Gray and P. deltoides Batr. Marsh) to
illustrate interspecific variations in root morphological proper-
ties and anchorage (e.g., branching density, rooting depth)
(Harrington and DeBell, 1996).
Winching tests were carried out in the field, and the force
required to cause failure was recorded. Root system architec-
ture was then measured, and these data were fed into the
model, which calculated the force necessary to uproot the trees
in two types of soil. Results were compared with experimental
data to better understand the major parameters influencing tree
uprooting during a wind storm.
MATERIALS AND METHODS
Tree material and site—To provide data from real trees, which could be
used to test the 3D model developed later, we performed static bending tests
and measured root architecture of forest trees. We chose poplars (Populus sp.)
because poplar plantations suffered severe losses during the 1999 storms in
Europe (Bergonzini and Laroussinie, 2000). The poplar trees selected for this
study were planted as live pole cuttings, each pole being several meters long
and pushed into the ground at a depth of 1 m. Adventitious lateral roots then
grow from this pole, which can be considered analogous to a central tap root.
Therefore, root systems of plantation-grown poplars are unusual but simpler to
describe than seed-grown, tap-rooted conifers such as maritime pine (Pinus
pinaster Ait., Danjon et al., 2005). Because these data were to be fed into the
3D root and soil model in a first attempt to calibrate the model with real data,
we preferred to use simple systems, thus reducing computation time.
Six 7-yr-old trees were used, three each of two North American poplar
cultivars: ‘Beaupré’ (Populus trichocarpa Torr. and Gray) and ‘Raspalje’
(Populus deltoides Batr. ex Marsh). Stem diameter at breast height (DBH) was
0.24 6 0.01 m (mean 6 SE) for Beaupré and 0.28 6 0.01 m for Raspalje.
Trees were planted at a density of 200 stemsha1 in two neighboring stands
located in northwest France (Moulin de Bariteau, Chinon, 0814 0 E, 47810 0 N).
The study area was flat with an elevation of 37 m. Mean annual precipitation is
820 mmyr1, and the prevailing wind direction is northwest. Mean wind speed
ranges from 4.5–5.5 ms1 (Météo-France). The soil consisted of a clayey sand
1507
Fig. 1. Diagrams of pulling tests for trees in the field. (a) The trees
were winched sideways using a motorized winch. The winch was attached
to the base of an anchoring tree at the longest possible distance to the
winched tree to obtain a small deviation from pure horizontal forces. The
height of the cable attachment was only 6.0 m so as to induce root failure
instead of stem breakage. (b) The finite element case study was formed of
a block of soil 2.0 m deep and 4.4 m wide. The underside face was built in
and lateral faces were blocked horizontally. A displacement was imposed
on the rigid trunk at a height of 6.0 m.
(69% sand, 21% clay, and 10% loam) and lacked a hard pan in the top 1.0 m (F.
Charnet, Institut pour le Développement Forestier, personal communication).
During the investigation, the ground water table was 1.2 m deep. Soil was
sampled at a depth of 0.30 m from the root–soil plate of trees winched
sideways. The wet soil mass was measured immediately, and samples were
then oven-dried for 5 d at 1058C to calculate soil moisture content (g water/100
g soil). In both plantations, soil moisture was 21.7 6 1.4% (Raspalje) and 21.7
6 0.9% (Beaupré) and thus was not thought to influence anchorage between
trees.
Static bending tests—The method used for uprooting trees was similar to
that used by Coutts (1983), Moore (2000), Mickovski and Ennos (2003), Cucchi
et al. (2004), and Nicoll et al. (2005). Two trees of each cultivar were winched
sideways until anchorage failure, which occurred when the tree was deflected
;108 from the vertical, forming an 808 angle with the soil. One other tree for each
species was winched sideways almost to the point of failure, i.e., the tree was
deflected 58 at the stem base, then allowed to return to the vertical. Trees that
were bent 58 were not damaged and were used for measuring root system
architecture, which required intact root systems. A motorised winch (type Hit-
Trac 16B, Habegger, Switzerland), with a 16 kN maximal strength capacity, was
used to winch trees sideways. The winch was attached to the base of an anchoring
tree at the longest possible distance to the winched tree to obtain a small deviation
from pure horizontal forces (Fig. 1). The cable was attached to the winched tree at
only 6.0 m so as to induce root failure instead of stem breakage. The tension
applied to the winched tree was measured by a load cell (type K25H-20kN
Scaime S.A., France), attached between the winch and the anchoring tree, and
was recorded every second using a data logger (type Almemo 2290–8, Ahlborn,
Germany). An inclinometer was attached to the tree stem at a height of 6.0 m to
measure the stem deflection angle during the winching test. Deflection angles
were recorded every second with a second data logger, which was synchronized
and simultaneously activated with the load cell data logger. Trees were winched
along the direction of the westerly prevailing wind.
Root systems—Root systems of trees that were winched but not uprooted
(stem deflection of 58) were then excavated. Those trees that had been uprooted
could not be used to measure roots because the root systems were damaged. A
mechanical shovel was used to cut a circular trench with a radius of 2.0 m
around the trunk, and the root system was lifted out of the soil. An air spade
(Rizzo and Gross, 2000) was then used to remove the soil and clean the root
system. Root system architecture was measured manually in the field, and
during these measurements, each root system was discretized into small
sections. Following a recursive pathway along the central tap root and each
1508 A MERICAN J OURNAL OF B OTANY [Vol. 94

TABLE 1. Mechanical properties of soils and roots used in the analysis.

The soils were modelled using Mohr–Coulomb yield criteria, and the
roots were modelled using the von Mises yield criterion. Refer to list
of symbols at end of Discussion for definitions.
Property Frictional soil Cohesive soil Root

E (MPa) 20 20 2000
m 0.3 0.49 0.3
c (kPa) 1.5 20 —
/ (8) 25 0 —
MOR (MPa) — — 15
c (kN/m3) 20 20 0

lateral root, we measured the distance (link) between each root base and
daughter root, along with the diameter at the root section base and halfway
along the link. The depth, branching angle, and azimuth of each link were also
noted, as well as any abrupt changes in azimuthal direction. Data were recorded
and encoded simultaneously in the software ARCHIROOT (Dupuy, 2003,
www.archiroot.org.uk). ARCHIROOT stores data in the format of a directed
Tree Graph (TG) using the AMAP MTG format as described by Danjon et al.
(1999). The data were used to compute the total root biomass and the distribution
of diameter size and to produce 3D images of the root system (Fig. 1).

Finite element modelling—Soil and roots material definition—Modelling

was carried out using the FEM software Abaqus (ABAQUS, Inc., http://www.
abaqus.com/). Soil was modelled as an elastic perfectly plastic material using a
linear Mohr Coulomb model available in the Abaqus software (see Abaqus
user’s manual for details). For frictional material, the model was used with
nonassociated flow, with a dilation angle w ¼ 0 (see Whitlow [1995] for an
introduction to soil mechanics). Root material was considered to be elastic
linear with a plastic threshold modelled by a von Mises yield criterion
(Kachanov [2004]).
Results from the 3D mechanical model were then to be compared with the
experimental data from the static bending tests. However, information
concerning the mechanical properties of soil at the field site was not available.
Therefore, we used theoretical values (Whitlow, 1995), which were considered
representative of the site conditions. Because the soil to be modelled comprised
70% sand particles, we chose the following values: Young’s modulus E ¼ 20
MPa, Poisson’s coefficient m ¼ 0.3, density c ¼ 20k N/m3, and friction angle /
¼ 258 (Table 1). A low cohesion value c ¼ 1.5 kPa was also used to account for
the presence of about 20% clay in the soil (Table 1).
Anchorage simulations were then also performed using a model of
undrained pure clay-like soil so that the root–soil mechanical response could
be analyzed in different conditions, where the cohesion c ¼ 20 kPa, m ¼ 0.49,
and / ¼ 08. The remaining mechanical parameters were kept the same as in the
preliminary soil model (Table 2).
The wood material of roots was characterized by a Young’s Modulus of 2.0
GPa, and the yield stress (rr) associated with the von Mises criterion equalled
15 MPa.
Geometry and meshing—The root architecture data, which were recorded in
the TG files, provided the topology and geometry of the whole root structure.
These data are the list of consecutive nodes, the type of connection (link) these
nodes had with the antecedent node of the same branching order, and the
associated geometrical attributes.
In this FEM model, root systems were discretized using 2-node linear beam
elements B21 available in the structural element library of ABAQUS. The FEM
TABLE 2. Lowest and highest values of c, /, w, and MOR used to analyze
the sensitivity of root and sand-like soil parameters on resistance to
uprooting. Refer to list of symbols at end of Discussion for
definitions.
Parameter Low High
c (kPa) 1.5 2 Therefore, a node to node interaction between soil and roots was
/ (8) 20 25
w (8) 0 5
MOR (MPa) 12.7 15.0
Fig. 2. Graphs of root architecture data from Beaupré (Populus
trichocarpa) and Rasplje (Populus deltoides) cultivars’ root systems used
to simulate uprooting. (a) The Beaupré cultivar’s root system had a high
branching density but a low root volume. (b) The Raspalje cultivar’s root
system had fewer but larger roots. XX and ZZ correspond to width and
depth, respectively.
mesh of the root system was automatically derived from the architectural data
using a computer program specifically dedicated to this task (see flowchart in
Fig. 2). This program gave a generic character to the meshing process, so that
the program is applicable to any branching structure encoded with an AMAP
MTG format.
The initial discretization was based on the nodes defined in the TG files, i.e.,
resulting from the field measurements. The automatic mesh generation
consisted of sequentially reading the pathway traced in the TG file and
creating a separate list of nodes (NODE) as well as a list of beam elements
(BEAM). The format of these elements must be consistent with the B21
element type chosen in this study. It should be noted that this section of the
program can be changed easily if another type of element or FEM software is
used. Each line of NODE contains a node number and its spatial coordinates,
whereas each line of BEAM contains an element number and the number of the
two associated nodes delimiting the extremities of the element.
The soil domain consisted of a parallelepipedic block discretized with 20
nodded brick elements, i.e., the 3D continuum element C3D20 in the ABAQUS
elements library.
Root–soil interaction—The following stage consisted of creating the
physical links between the soil and roots. Different techniques exist to model
explicit surface to surface interaction through the use of constraints
implemented with Lagrange multipliers or interface elements. Such methods
are computationally extensive and may sometimes lead to convergence
difficulties. However, the hypothesis of a rigid root–soil interaction was
supported by field observations that most of the tree roots remained embedded
in the mass of soil after uprooting, suggesting that most of the soil lifted during
uprooting was trapped in the network of roots and that little relative
displacement between soil and roots occurred.
implemented using kinematic coupling (see the ABAQUS theory manual).
Let Xr denote the position of a root node (master node) and Xs be the coupled
soil position (slave) in reference configuration. The initial reference
configuration is expressed as the translation vector N:
September 2007] D UPUY ET AL .—3D TREE ANCHORAGE MODEL 1509

TABLE 3. Summary of experimental and numerical pulling tests for

Beaupré (Populus trichocarpa) and Raspalje (Populus deltoides)
cultivars.
Beaupré Raspalje

Angle Displacement Force Angle Displacement Force

Test (8) (m) (kN) (8) (m) (kN)

Quick test 2 0.2 6.3 4.1 0.4 9.5

Complete test 1 5.9 0.6 8.3 10.5 1.1 12.3
Complete test 2 11 1.1 9.2 8.7 0.9 15.1
Simulation sand 9.6 1 10.5 9.6 1 15.9
Simulation clay 9.6 1 19.4 9.6 1 31.3

were created, with an increasing number of soil elements, i.e., 256 (8 3 8 3 4)

elements, 720 (12 3 12 3 5) elements, 1372 (14 3 14 3 7) elements, and 2976
(18 3 18 3 9) elements. These different meshes were used to analyze two root
systems in the frictional soil. The evolution of the reaction force was related to
the number of elements. Calculation time (CPU) was monitored to find an
optimal number of elements for a reasonable CPU.

Sensitivity analysis of material properties—Because we did not possess any

Fig. 3. Flowchart of the mesh generator program. BEAM and NODE
are lists defining nodes and beam elements, respectively, in the FEM
format of the input file. The program reads the MTG file incrementally and
finds at each line i the branching order o(i) and the spatial coordinates
needed to write a new beam element (,i. and ,i, j. in the BEAM and
NODE list). N(o) represents the number of the anterior node of the oth
branching order. The operator  indicates the insertion of a new line ,i.
or ,i, j. in the NODE or BEAM list.
N¼ Xs  Xr ð1Þ
In the current configuration, the root node position xr and the soil node position
xs are linked with the following relationship:
xs ¼ xr þ Cr  N; ð2Þ
r
where C denotes the rotation matrix associated to the root node rotation. The
constrained degrees of freedom are expressed as a function of the master node
degrees of freedom. This reduces the total number of degrees of freedom in the
tangent stiffness matrix and is therefore more computationally efficient than
explicit surface to surface interaction methods.
Soil–root node pairs were connected using the ABAQUS command *TIE.
Boundary conditions and bending test simulation—The reconstructed root
system model was placed at the center of a parallelepipedic domain of soil, 4.4
3 4.4 m wide and 2.0 m deep. Boundary conditions were defined to reproduce
the action of the surrounding infinite earth: translation on lateral faces was fixed
horizontally, and the underside face was built in (Fig. 3). A rigid beam 6 m long
was fixed to the top of the root system, representing the tree trunk beneath the
crown. To simulate uprooting, lateral displacements of 1.0 m were applied
incrementally on the rigid trunk up to a height of 6.0 m. The reaction force was
recorded at this point, and the force–displacement curve then determined. The
maximum force at the end of the lateral displacement was used to compare the
soil–root anchorage efficiency. The field of equivalent plastic strain was
monitored in the vertical plane in the direction of pulling. The plastic area
around the roots was also visualized in the complete root system, and the
motion of the stump was monitored for each increment. These variables were
used to understand the different mechanisms occurring during uprooting.
Mesh size sensitivity analysis—A preliminary analysis of the influence of
mesh size on the calculation was also carried out. Four different soil meshes
quantitative data on soil and root mechanical properties at the field site for the
static bending tests on poplar trees, a sensitivity analysis was conducted to
study the influence of small variations in frictional soil properties (c, /, w) and
root mechanical parameters. Each parameter of the initial configuration was
varied successively by 15% (Table 2). The sensitivity S of a parameter a was
then quantified as the percentage increase in anchorage resistance for an
increase in 1% of the specific parameter:
DF=F
S¼ ð3Þ
Da=a0
RESULTS
Static bending tests—Of the four trees winched sideways
until failure, the stem of two of each cultivar broke at heights of
2 m; the remaining two trees, which were deflected to 5 rather
than 108, failed through uprooting. The mean force that caused
failure was 8.75 kN in Beaupré poplars and 13.5 kN in
Raspalje (Table 3). Because we were simulating uprooting
behavior in these trees, we could not compare results from the
model with data from trees that had failed in the stem;
therefore, only two trees could be used to compare observed
and predicted uprooting results. For the two trees, which were
winched sideways but not to failure, only the initial part of the
force displacement curve could be compared with results from
the model.
Description of root systems—The root system of the
Beaupré cultivar consisted of a main vertical root 1.11 m deep
with a basal diameter of 0.31 m. This central root bore
numerous secondary laterals with an average density of 35
laterals/m and a mean diameter of 32.0 6 0.02 mm. The root
system had a total volume of 0.12 m3 and had 115 structural
roots (branching order ,5, 90% were smaller than 50 mm in
diameter), with a total root length of 94 m (Fig. 1a).
The root system of the Raspalje cultivar had fewer but larger
roots than the Beaupré. One massive central vertical root with a
length of 0.95 m and a basal diameter of 0.34 m bore secondary
laterals with an average density of 24 laterals/m and a mean
diameter of 0.55 6 0.04 m. The root system had a total volume
of 0.21 m3 and had 82 structural roots (branching order ,5)
with a total root length of 82.0 m (Fig. 1b).
1510 A MERICAN J OURNAL
Fig. 4. Influence of the soil mesh on the uprooting simulations. The
circles (*) and crosses (x) indicate the maximum reaction force for
Beaupré and Raspalje, respectively, for a given mesh size. The solid line
(—) indicates the calculation time (CPU) required for a given mesh size.
Mesh size sensitivity analysis—The study of mesh size
influence on the numerical results demonstrated that the
calculated reaction force opposed to a given stem displacement
decreased with the size of the soil elements (Fig. 4). The
calculated reaction force vs. mesh size did not stabilize to a
constant value with the range of element size used in this
sensitivity analysis (Fig. 4); however, the absolute difference
between the mechanical responses of the two root systems
remained constant. On the other hand, tthe calculation time
(CPU) increased exponentially with the number of soil
elements; therefore, a mesh with 1372 elements was finally
used for the comparative analysis. This choice provided a good
compromise between the accuracy of the numerical results and
the calculation time required for the simulations.
Static bending test simulations—The displacement applied
at the top of the rigid stem induced a bending moment at the
center of the bole. The roots on the ‘‘leeward’’ side (the side
held in compression during bending) tended to plunge
downward, which locally compressed the soil. On the
‘‘windward’’ side (the side held in tension during bending),
roots were pulled upward from the soil, decreasing soil
compression on this side of the tree. The failure areas were
highlighted by the field of equivalent plastic strain (Fig. 5) and
took the form of a slip surface. In cohesive soil, the slip
surfaces were symmetric and circular (Fig. 6a), whereas in
frictional soil, permanent deformations expanded further on the
windward side than on the leeward side (Fig. 6b). The plot of
the displacement field of a strip of soil in the plane of pulling
(centered on the stem) showed that the kinematics of uprooting
in these trees followed a rather regular rotation around an axis
at half the rooting depth (Fig. 6).
The differences in anchorage resistance between the two trees
were similar between both the experimental tests and computer
simulations (Fig. 7a, b). In the field, the Raspalje cultivar was
30% more resistant than the Beaupré. Simulations in the
frictional soil produced a difference of 40% between the two
cultivars, and a difference of 50% uprooting resistance resulted
when the simulations were carried out in cohesive soil. When
comparing the simulations to experimental data, we found that
20% less force was required to uproot the poplars in the
OF B OTANY [Vol. 94
Fig. 5. Simulated fields of plastic strain generated in the frictional soil
(a) and in the root system (b) during the uprooting of the Raspalje cultivar.
In (a), permanent deformations were greater on the windward side of the
tree. In (b), failure of root elements appeared in the region of soil plastic
failure.
experiment than in the simulation with the frictional soil (Table
3). Simulations with the cohesive soil resulted in an uprooting
resistance 70% higher than that measured in the field.
The influence of small variations in soil and root properties,
i.e., c, /, w, and MOR, showed that the cohesion c was the
most influential parameter with regard to uprooting, with a
sensitivity of 0.5 (Fig. 8). The properties / and MOR had a
similar effect on uprooting resistance, with a sensitivity of less
than 0.3 (Fig. 8), while w had the least influence on uprooting
resistance (Fig. 8).
DISCUSSION
Understanding how wind forces are transferred to the soil via
the root system and what determines the anchorage resistance
of trees is crucial for predicting windthrow in forests. To date,
our knowledge and ability to predict tree uprooting has been
limited by our ability to understand the nature of the root–soil
interaction in the context of real root system architectures. Root
morphologies are usually very complex and their variability
poorly understood. Also, by current experimental approaches
we are unable to measure the essential mechanical variables
(e.g., strains, stresses, and displacement fields) in both roots
and soils during in situ tree pulling tests. Finite element and
other numerical methods used in physics and engineering are
potentially very powerful tools for obtaining quantitative in-
depth information on such mechanisms, and the model
presented here was a first attempt in using such a method to
understand the mechanical properties of root–soil composites.
In this paper we primarily described the development and
testing of the finite element method in tree anchorage studies.
The developed model allowed us to simulate the uprooting of
two different poplar cultivars while accounting for an accurate
geometrical and topological description of their two distinct
root system architectures. To compare the model’s predictions
with the anchorage of real trees, destructive winching tests
September 2007] D UPUY ET AL .—3D TREE ANCHORAGE MODEL
Fig. 6. Displacement fields at soil nodes in a vertical plane along the
direction of pulling for the Beaupré cultivar in (a) clay-like soil and in (b)
sand-like soil.
were conducted on four additional trees. The use of cultivars
from the poplar plantation at Moulin de Bariteau aimed at
reducing the morphological variability among species (Har-
rington and DeBell, 1996), while the Beaupré and Raspalje
cultivars were chosen for their morphological differences.
Fig. 7. Simulated curves of force vs. displacement (with a mesh size of
1372) for (a) the Beaupré and (b) the Raspalje cultivars. Experimental
pulling tests are represented with a dashed line, the simulation in frictional
soil is represented with circles, and the simulation in cohesive soil is
represented with squares.
1511
Fig. 8. Results of the sensitivity analysis describing how properties
affected the force required to uproot trees. (a) Soil cohesion, c, had the
most influence on uprooting. (b) Friction angle, /, and (c) root strength
MOR, rr, had a similar influence, with sensitivities of 0.27 and 0.26,
respectively. (d) Dilation, w, had the least influence on resistance to
uprooting.
However, significant intraspecific variability occurred during
the pulling tests, and the model predictions can only provide
indications about the differences in anchorage between
cultivars. More trees need to be tested to validate our model,
and more elaborated poro-viscoplastic models could be
developed to incorporate the effect of pore water in the
mechanisms of uprooting.
The properties of the field soil (clayey sand) used are
between those of sand and clay, and yet the model simulations
predicted much higher resistance than was measured in the
field. Results from the simulations seem, therefore, to have
overestimated the actual uprooting resistance. Apart from the
fact that the trees that were winched to failure in the field were
not those for measured for root architecture, the difference in
uprooting resistance between simulated and real anchorage
tests could be explained in several ways. The rigid node-to-
node interaction used in the model does not enable consider-
ation of sliding or opening of the root–soil surface interaction
during failure. Studies on the influence of interface properties
have also shown that roughness of the interface on foundation
piles can account for 15–50% of the variation in resistance,
depending on the type of loading and pile configuration (Chen
and Martin, 2002; Jeong et al., 2004). Therefore, the rigid root–
soil adhesion used in our model can be expected to artificially
increase the uprooting resistance. Finally, the material model
used for roots and soils is probably not suitable for large
deformations. Highly anisotropic and fragile models, which
can involve failure of individual root elements, will more
realistically represent root breakage.
Although the generic 3D uprooting model developed in this
study does not yet include all the elements necessary to
accurately predict uprooting resistance, it can nevertheless
provide useful information on how root system structures
interact within a soil. In the simulations of uprooting in clay
soil, failure in the soil was symmetrical on both the windward
1512 A MERICAN J OURNAL
and leeward sides of the root system. In analogous studies on
laterally loaded piles on cohesive clay soils (Yang and Jeremic,
2002), deformations were also symmetrical. However, in the
same study on a frictional soil, Yang and Jeremic (2002) found
that the plastic deformation not only occurred close to the soil
surface but also expanded laterally further away from the pile.
In our model of uprooting in a frictional soil, plastic
deformation was greater on the windward than on the leeward
side of the root system. Although this phenomenon occurs in
real situations of tree overturning (Coutts, 1983, 1986; Crook
and Ennos, 1996; Cucchi and Bert, 2003), it has not been
explained comprehensively. We show that soil mobilized on
the windward side of the tree was subjected to low or negative
isostatic pressure when lifted up by roots. Therefore, failure
occurred and uprooting was induced.
The geometrical structure of the root systems also affected
soil behavior during overturning in our simulations. The two
root systems, which were from two different cultivars, were
distinctly different from each other with regard to the number
and diameter of first order lateral roots borne on the central tap
root. The Raspalje cultivar possessed few but large lateral
roots, whereas the Beaupré had twice as many smaller lateral
roots and also more total roots (115 structural roots in Beaupré,
85 structural root in the Raspalje root system). Because the
stiffness of a root is a function of its diameter to the fourth
power, numerous small roots will decrease anchorage rigidity
but increase resistance in tension (Stokes et al., 1995). The
‘‘group effect’’ of large lateral roots near the tree trunk will
cause plastic failure in the soil to occur further from the tree
axis, thus augmenting resistance to failure (Coutts, 1983;
Dupuy et al., 2005a). The simulated failure zone was larger
with the Raspalje than with the Beaupré cultivar. This
difference in the size of plastic failure in the soil is probably
due to the effect of the large lateral roots in the Raspalje tree.
Clonal, morphological differences affect the anchorage
efficiency of poplar trees in a plantation (Harrington and
DeBell, 1996). Uprooting resistance was higher in the Raspalje
cultivar than the Beaupré in both the numerical and real
uprooting tests. This increase in anchorage efficiency can be
explained by the high root volume and number of large lateral
roots in the Raspalje tree. The resistance of lateral roots held in
tension during uprooting is a major component of anchorage
(Coutts, 1983, 1986; Stokes et al., 1995; Crook and Ennos,
1996) because of the high tensile resistance of wood (Genet et
al., 2005). Numerous thin lateral roots, even if they are not
large, will augment uprooting resistance. However, our model
does not yet include the anisotropic material properties of root
wood, and including such properties is a priority for future
modelling studies.
Although root architecture has long been considered a major
component of root anchorage (Coutts, 1983, 1986; Danjon et
al., 2005; Dupuy et al., 2005a, b; Khuder et al., 2007), we
showed that soil cohesion c was also an important factor
affecting uprooting resistance. The weight of the system alone
was not sufficient to produce enough isostatic pressure to
prevent soil failure according to the Mohr–Coulomb theory.
Additionally, the soil that is lifted on the windward side of the
tree during uprooting is subject to low or negative isostatic
pressure, thus increasing sensitivity to variations in soil shear
strength, which was triggered by the soil cohesion c. Soil
friction angle also influenced uprooting resistance, but dilation
angle less so. Chen and Martin (2002) showed that these two
parameters were more important than soil cohesion when
OF B OTANY [Vol. 94
considering the efficiency of landslide stabilizing piles, but pile
failure is concerned with soil cohesion in compression in the
direction of pulling. In our model, soil failure was predominant
on the windward side of the tree, where roots and soil were
held in tension during overturning.
Changing root strength had little effect on overturning
resistance in the model. To our knowledge, no quantitative
studies of the influence of root strength on anchorage
efficiency of trees has been done, although indicators of its
importance, particularly in shallowly rooted forest trees, are
available (Stokes and Mattheck, 1996).
With regard to the technical aspects of the generic 3D model
described here, CPU time was high, and the problem of
convergence does not yet enable models to be compared
readily to real data. Nevertheless, such a numerical tool is very
promising and can be applied to any root architecture from
which topology and geometry have been accurately described.
Recording the structure with an AMAP MTG format also
allows the topological data of the root system to be read
automatically and incorporated in the FEM anchorage model.
The numerical approach developed in our study can also be
used to explore in detail the role of each of the structural
elements involved in tree uprooting. In particular, how any
given root interacts within the root–soil matrix can be
determined by isolating and removing that root from the
model, or by adding new roots (Dupuy et al., 2005a; Fourcaud
et al., 2007). Such virtual experiments are easier to perform
than complex field experiments, whereby root systems are
physically manipulated (Khuder et al., 2007).
LIST OF SYMBOLS
BEAM: Structure recording beam definition in FEM input file
NODE: Structure recording node definition in FEM input file
i: line number in Multi Tree Graph (MTG) file
o(i): Branching order of the root at the ith line in the MTG
file
N(o): Number of the anterior node of branching order o
, .: A line in the NODE or BEAM list
: Addition of a new line in the NODE or BEAM list
Xs: Soil nodal coordinates of the sth soil node
Xr: Root nodal coordinates of the rth root node
þ: Symbol indicating the initiation of a new branch in
the MTG file
,: Symbol indicating a new portion of root in the MTG
file
MTG: Multiscale Tree Graph. Architectural data file for the
coding of plant structure
c: Soil cohesion
/: Soil friction angle
w: Soil dilation angle
m: Poisson coefficient
c: Volumetric weight
LITERATURE CITED
ACHIM, A., J.-C. RUEL, AND B. GARDINER. 2005. Effect of precommercial
thinning on the resistance of balsam fir to windthrow. Canadian
Journal of Forest Research 35: 1844–1853.
September 2007] D UPUY ET AL .—3D TREE ANCHORAGE MODEL
BERGONZINI, J. C., AND O. LAROUSSINIE. 2000. Les écosystèmes forestiers
dans les tempêtes. ECOFOR, Ministère de ’Agriculture et de la Pêche,
Paris, France.
BLACKWELL, P. G., K. RENNOLLS, AND M. P. COUTTS. 1990. A root
anchorage model for shallowly rooted Sitka spruce. Forestry 63: 73–
92.
BLOOMFIELD, J. 2000. The potential impacts of global warming on
America’s forests. Critical findings for forest lands from the First
National Assessment of the potential consequences of climate
variability and change. Website www.climatehotmap.org/impacts/
forests.html [accessed 1 January 2007].
BRÜCHERT, F., AND B. GARDINER. 2006. The effect of wind exposure on the
tree aerial architecture and biomechanics of Sitka spruce (Picea
sitchensis, Pinaceae). American Journal of Botany 93: 1512–1521.
CHEN, C.-Y., AND G. R. MARTIN. 2002. Soil-structure interaction for
landslide stabilizing piles. Computer and Geotechnics 29: 363–386.
COUTTS, M. P. 1983. Root architecture and tree stability. Plant and Soil 71:
171–188.
COUTTS, M. P. 1986. Components of tree stability in Sitka spruce on peaty
gley soil. Forestry 59: 173–197.
CROOK, M. J., AND A. R. ENNOS. 1996. The anchorage mechanics of deep
rooted larch, Larix europea 3 L. japonica. Journal of Experimental
Botany 47: 1509–1517.
CUCCHI, V., AND D. BERT. 2003. Wind-firmness in Pinus pinaster Aı̈t.
stands in southwest France: influence of stand density, fertilisation
and breeding in two experimental stands damaged during the 1999
storm. Annals of Forest Science 60: 209–226.
CUCCHI, V., C. MEREDIEU, S. A. F. DE COLIGNY, J. SUAREZ, AND B. GARDINER.
2005. Modelling the windthrow risk for simulated stands of maritime
pine. Forest Ecology and Management 213: 184–196.
CUCCHI, V., C. MEREDIEU, A. STOKES, S. BERTHIER, D. BERT, M. NAJAR, A.
DENIS, AND R. LASTENNET. 2004. Root anchorage of inner and edge
trees in stands of maritime pine (Pinus pinaster Ait.) growing in
different podzolic soil conditions. Trees—Structure and Function 18:
460–466.
DANJON, F., D. BERT, C. GODIN, AND P. TRICHET. 1999. Structural root
architecture of 5-year-old Pinus pinaster measured by 3D digitising
and analysed with AMAPmod. Plant and Soil 217: 49–63.
DANJON, F., T. FOURCAUD, AND D. BERT. 2005. Root architecture and
windfirmness of mature Pinus pinaster Aı̈t. New Phytologist 168:
261–269.
DI IORIO, A., B. LASSERRE, G. S. SCIPPA, AND D. CHIATANTE. 2005. Root
system architecture of Quercus pubescens trees growing on different
sloping conditions. Annals of Botany 95: 351–361.
DUPUY, L. 2003. Modélisation de l’ancrage racinaire des arbres forestiers,
Ph.D. thesis, Bordeaux I, Bordeaux, France.
DUPUY, L., T. FOURCAUD, AND A. STOKES. 2005a. A numerical investigation
into factors affecting the anchorage of roots in tension. European
Journal of Soil Science 56: 319–327.
DUPUY, L., T. FOURCAUD, AND A. STOKES. 2005b. A numerical investigation
into the influence of soil type and root architecture on tree anchorage.
Plant and Soil 278: 119–134.
ENNOS, A. R. 1990. The anchorage of leek seedlings: the effect of root
length and soil strength. Annals of Botany 65: 409–416.
FOURCAUD, T., L. DUPUY, D. SELLIER, P. ANCELIN, AND P. LAC. 2003b.
Application of plant architectural models to biomechanics. In B. Hu
and M. Jaeger [eds.], Proceedings of international symposium on
plant growth modeling, simulation, visualization and their applica-
tions, Beijing, China 2003, 462–479 Tsinghua University press,
Beijing, China.
FOURCAUD, T., J.-N. JI, Z.-Q. ZHANG, AND A. STOKES. 2007. Understanding
the impact of root morphology on overturning mechanisms: a
modelling approach. Annals of Botany: in press.
FOURCAUD, T., P. LAC, P. CASTÉRA, AND P. DE REFFYE. 2003a. Numerical
modelling of shape regulation and growth stresses in trees. II.
Implementation in the AMAPpara software. Trees—Structure and
Function 17: 31–39.
FRASER, A. I. 1962. The soil and roots as factors in tree stability. Forestry
23: 90–95.
1513
GARDINER, B. A., H. PELTOLA, AND S. KELLOMAKI. 2000. Comparison of two
models for predicting the critical wind speeds required to damage
coniferous trees. Ecological Modelling 129: 1–23.
GENET, M., A. STOKES, F. SALIN, S. B. MICKOVSKI, T. FOURCAUD, J.-F.
DUMAIL, AND R. VAN BEEK. 2005. The influence of cellulose content on
tensile strength in tree roots. Plant and Soil 278: 1–9.
HARRINGTON, C. A., AND D. S. DEBELL. 1996. Above- and below-ground
characteristics associated with wind toppling in a young Populus
plantation. Trees—Structure and Function 11: 109–118.
JEONG, S., J. LEE, AND C. J. LEE. 2004. Slip effect at the pile–soil interface
on dragload. Computer and Geotechnics 31: 115–126.
KACHANOV, L, M. 2004. Fundamentals of the theory of plasticity. Dover
Publications Inc., Mineola, New York, USA.
KHUDER, H., A. STOKES, F. DANJON, AND K. GOUSKOU. 2007. Is it possible to
manipulate root anchorage in young trees? Plant and Soil 294: 87–
102.
MAYER, P., P. BRANG, M. DOBBERTIN, D. HALLENBARTER, J.-P. RENAUD, L.
WALTHERT, AND S. ZIMMERMANN. 2005. Forest storm damage is more
frequent on acidic soils. Annals of Forest Science 62: 303–311.
MICKOVSKI, S. B., AND A. R. ENNOS. 2003. Anchorage and asymmetry in
the root system of Pinus peuce. Silva Fennica 37: 161–173.
MOORE, J. R. 2000. Differences in maximum resistive bending moments of
Pinus radiata trees grown on a range of soil types. Forest Ecology
and Managment 135: 63–71.
NICOLL, B. C., A. ACHIM, S. MOCHAN, AND B. A. GARDINER. 2005. Does
steep terrain influence tree stability? A field investigation. Canadian
Journal of Forest Research 35: 2360–2367.
NICOLL, B. C., AND D. RAY. 1996. Adaptive growth of tree root systems in
response to wind action and site conditions. Tree Physiology 16: 891–
898.
NIKLAS, K. J., F. MOLINA-FREANER, C. TINOCO-OJANGUREN, AND D. J.
PAOLILLO JR. 2002. The biomechanics of Pachycerus pringlei root
systems. American Journal of Botany 89: 12–21.
PELTOLA, H. M. 2006. Mechanical stability of trees under static loads.
American Journal of Botany 93: 1501–1511.
PUTZ, F. E., P. COLEY, D. K. LU, A. MONTALVO, AND A. AIELLO. 1983.
Uprooting and snapping of trees: structural determinants and
ecological consequences. Canadian Journal of Forest Research 13:
1011–1020.
RIZZO, D. M., AND R. GROSS. 2000. Distribution of Armillaria on pear root
systems and a comparison of root excavation techniques. In A. Stokes
[ed.], The supporting roots of trees and woody plants: form, function
and physiology, vol. 87, 305–311. Kluwer, Dordrecht, Netherlands.
RUEL, J.-C. 2000. Factors influencing windthrow in balsam fir forests:
from landscape studies to individual tree studies. Forest Ecology and
Management 135: 169–178.
SELLIER, D., T. FOURCAUD, AND P. LAC. 2006. A finite element model for
investigating effects of aerial architecture on tree oscillations. Tree
Physiology 26: 799–806.
SOETHE, N., J. LEHMANN, AND C. ENGELS. 2006. Root morphology and
anchorage of six native tree species from a tropical montane forest
and an elfin forest in Ecuador. Plant and Soil 279: 173–185.
STOKES, A. 1999. Strain distribution during anchorage failure of Pinus
pinaster Ait. at different ages and tree growth response to wind-
induced root movement. Plant and Soil 217: 17–27.
STOKES, A., M. ABDGHANI, F. SALIN, F. DANJON, H. JEANNIN, S. BERTHIER,
A. KOKUTSE, AND H. FROCHOT. 2007. Root morphology and strain
distribution during tree failure on mountain slopes. In A. Stokes, I.
Spanos, J. E. Norris, and L. H. Cammeraat [eds.], Eco- and ground
bio-engineering: the use of vegetation to improve slope stability.
Developments in plant and soil sciences, vol. 103, 165–173. Springer,
Dordrecht, Netherlands.
STOKES, A., A. H. FITTER, AND M. P. COUTTS. 1995. Responses of young
trees to wind and shading: effects on root architecture. Journal of
Experimental Botany 46: 1139–1146.
STOKES, A., AND C. MATTHECK. 1996. Variation of wood strength in tree
roots. Journal of Experimental Botany 47: 693–699.
STOKES, A., F. SALIN, A. D. KOKUTSE, S. BERTHIER, H. JEANNIN, S. MOCHAN,
L. DORREN, N. K. KOKUTSE, AND T. FOURCAUD. 2005. Mechanical
1514 A MERICAN J OURNAL OF B OTANY [Vol. 94

resistance of different tree species to rockfall in the French Alps. YANG, Z., AND B. JEREMIC. 2002. Numerical analysis of pile behaviour
Plant and Soil 278: 107–117. under lateral loads in layered elastic–plastic soils. International
WHITLOW, R. 1995. Basic soil mechanics. Longman Scientific & Journal for Numerical and Analytical Methods in Geomechanics 26:
Technical, Harlow, Essex, UK. 1385–1406.