Hort review 2006 (1)

See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/245374431
Analysing Fruit Tree Architecture-Consequences for
tree management and fruit production
Article · June 2006
CITATIONS
126
READS
5,255
3 authors:
Some of the authors of this publication are also working on these related projects:
Integrated Pest Management View project
FSPM Apple View project
Evelyne Costes
French National Institute for Agriculture, Food, and Env…
328 PUBLICATIONS   4,813 CITATIONS
SEE PROFILE
Pierre-Eric Lauri
French National Institute for Agriculture, Food, and Env…
SEE PROFILE
Jean Luc Regnard
L'institut Agro | Montpellier SupAgro
SEE PROFILE
All content following this page was uploaded by Jean Luc Regnard on 11 April 2015.
The user has requested enhancement of the downloaded file.

1
Analyzing Fruit Tree Architecture:
Implications for Tree Management
and Fruit Production
E. Costes, P. É. Lauri, and J. L. Regnard
UMR 1098—Biologie du développement des Espèces Pérennes
Cultivées
Equipe INRA-Agro.M “Architecture et Fonctionnement
des Espèces Fruitières”
INRA, 2 Place Viala, 34000 Montpellier Cedex 1, France
I. INTRODUCTION
II. ARCHITECTURAL ANALYSIS
A. General Concepts
B. Defining Architectural Models of Fruit Tree Species
1. Identifying Shoot Types
2. Analyzing Branching
3. Examples of Architectural Analysis in Fruit Trees
4. Describing the Intra-Species Variability of Tree Architecture
C. Quantitative Studies of Fruit Tree Topology
1. Primary Growth
2. Branching Patterns
3. Location of Flowering
4. Meristem and Shoot Mortality
D. Describing Fruit Tree Form
1. Measuring a 3D Form
2. Models for Representing Whole Tree or Row Form
3. Modeling Axis Form Changes
4. Models for Representing the Organ Distributions within Canopy
III. CONSEQUENCES OF TREE ARCHITECTURE FOR TREE TRAINING, ORCHARD
MANAGEMENT, AND FRUIT PRODUCTION
A. Initial Choices of the Grower and Young Tree Training
1. Rootstock Effects
2. Tree Development and Initial Fruit Production
1
Horticultural Reviews, Volume 32, Edited by Jules Janick
ISBN 0-471-73216-8 © 2006 John Wiley & Sons
C
O
P
Y
R
I
G
H
T
E
D
M
A
T
E
R
I
A
L

B. Adult Tree Training
1. Fruit Load Effects on Tree Growth, Architecture, and Ecophysiology
2. Consequences for Fruit Thinning
3. Implementation of Adult Tree Training Procedures
IV. CONCLUSIONS
V. GLOSSARY
LITERATURE CITED
I. INTRODUCTION
High yield performance of fruit crops results from the integration of var-
ious components, which, pieced together, constitute the “orchard sys-
tem puzzle” (Barritt 1992). They are implemented at two different time
scales: a set of initial choices determines the basic components of the
orchard during its life-span (support system, tree arrangement and qual-
ity, density, rootstock and cultivar), and a set of annual procedures that
is closely related to the training system but evolves from one year to the
next (pruning, training, and thinning practices).
These components are strongly related to each other and they need to
be assembled properly to ensure good economic results (Hoying and
Robinson 2000). Provided that techniques are compatible, a large range
of combinations is possible. In all cases, however, the training system
should integrate the following objectives: (1) light capture needs to be
optimized at the orchard scale, in order to obtain a high biomass pro-
duction (e.g., Jackson 1980); (2) canopy porosity to light (Lakso 1994)
should be as high as possible to improve light distribution between
fruiting structures (Lakso and Corelli-Grappadelli 1992; Wünsche and
Lakso 2000) and to lower the variability in fruit quality; (3) biomass must
be partitioned to fruiting shoots, as demonstrated in apple (Lespinasse
and Delort 1993) or avocado (Thorp and Stowell 2001); and (4) compe-
tition with vegetative sinks by inappropriate heading cuts, which stim-
ulates tree growth and vigor as shown in kiwi (Miller et al. 2001), should
be avoided.
It is therefore of major importance to develop training concepts that
optimally combine training and management systems at the orchard
scale and training methods at the tree scale (pruning, bending). At both
levels, an accurate knowledge of growth, branching, and flowering
processes within the tree canopy, i.e., tree architecture, is thus required
to optimize tree manipulation adapted to the plant material. In the first
section, we will present the main concepts that are used in architectural
analysis and illustrate how they have been introduced and applied to
fruit species, from both a qualitative and quantitative point of view. The
2 E. COSTES, P. É. LAURI, AND J. L. REGNARD

second section will present the consequences of fruit tree architectural
analysis for tree and orchard management, especially regarding the
manipulation of both vegetative and floral organs at the tree level.
II. ARCHITECTURAL ANALYSIS
In the last 20 years, architectural analysis of plants has led to the devel-
opment of new approaches to horticulture from the acquisition of knowl-
edge about tree development to the study of intra-species variations of
related characters and, in more applied aspects, to the improvement of
fruit tree management at the orchard level. Architectural analysis was
introduced in a botanical and forestry context by Hallé and co-workers
(Hallé and Oldeman 1970; Hallé et al. 1978) by observing the whole tree
with a particular focus on the dynamics of development. From these
studies, a general comprehensive framework of the invariant features
and rules that are responsible for a tree’s architecture has been extracted.
This procedure has been shown to apply to all plant species (Hallé et al.
1978), while the rules are defined at the species scale.
Applications developed in horticulture have focused mainly on two
within-tree scales: (1) organ arrangement, including both vegetative and
floral organs, and their relative equilibrium, and (2) fruiting branches
and whole tree behavior. These two scales constitute a basic framework
that is then used to interpret the effect of agronomical practices at the
tree and orchard scales.
A. General Concepts
For many years, the diversity of plant morphology has been fascinating
scientists and has been extensively studied from both a scientific and
philosophic point of view, e.g., Goethe (1790) and Arber (1950). Plant
form diversity results from differences in organ morphology and from dif-
ferences in constructional organization (Bell 1991). The constructional
organization of trees, also called architecture, results from the activity of
the meristems. All plant organs are made of cells and tissues, which ini-
tially organize within a meristematic zone. Thus, a tree, whatever its final
size, is initially constructed by the activity of at least two primary meri-
stems (one for the aerial part, one for the root system) or possibly more,
and to a lesser extent, by the activity of secondary meristems, which are
responsible for the diameter increment of woody axes (Hallé et al. 1978).
After observation of the aerial meristem activity in many species,
Hallé and co-workers proposed a classification of the aerial organization
1. ANALYZING FRUIT TREE ARCHITECTURE 3

of trees into four main categories. The first two categories separate trees
constructed by a single aerial meristem (monoaxial trees) or several
meristems (polyaxial trees) (Fig. 1.1). Polyaxial trees are thus split into
three subcategories based on the differentiation state of axes produced
by meristems: (1) all meristems have a similar activity and produce
equivalent non-differentiated axes, (2) different meristems have differ-
ent potentialities and produce different axis types (i.e., differentiated
axes), and (3) a given meristem changes its activity with time and pro-
duces mixed axes, i.e., axes whose basal and top parts have different
organization.
Within these four main categories, finer classes are considered that are
named “architectural models” and are dedicated to famous botanists.
The model definitions are based on the concept of “axis differentia-
tion,” which combines five main morphological criteria all related to the
meristem activity: (1) the growth direction associated with phyllotaxy
makes it possible to distinguish plagiotropic from orthotropic axes. Pla-
giotropic axes are characterized by a horizontal to oblique growth direc-
tion with alternate or distic phyllotaxy and a plane symmetry, while
orthotropic axes combine vertical growth with a spiral phyllotaxy and
axial symmetry; (2) the growth rhythm can be either continuous or
rhythmic. In case of rhythmic growth, the portion of axis developed dur-
ing the same growing period is called a growth unit (GU); (3) the branch-
ing mode (monopodial versus sympodial), position (acrotonic versus
basitonic, i.e., long shoots located respectively in the top or basal part
of the bearer shoot), and dynamics (immediate or sylleptic vs. delayed
or proleptic); (4) the sexual differentiation of meristems; and (5) the
polymorphism of axes that allows distinguishing between short (brachy-
blasts), medium (mesoblasts), and long (auxiblasts) shoots.
The proposed classification, composed of 23 models, provided a
framework for analyzing plant architecture and led to observations of the
Fig. 1.1. Example of monoaxial (a) and polyaxial trees (b) belonging, respectively, to Hol-
tum and Troll models (Source: Hallé, Oldman, and Tomlinson. Tropical Trees and Forests,
p. 84, 97. With permission of Springer Science and Business Media.)
(a) (b)

developmental dynamics of trees, at the whole tree scale, and the iden-
tification of phenomena that are invariant with respect to the environ-
ment (Hallé et al. 1978). This approach showed that it was possible to
account for the variability of all higher plants by defining a limited
number of developmental patterns, defined at the whole tree and axis
scales (Edelin 1981; Remphrey and Powell 1987; Caraglio and Edelin
1990; Thorp and Sedgley 1993).
In parallel, other concepts emerged from the analysis of plants at both
more detailed and more global scales than axes, and from the observa-
tion of the repetitive nature of tree construction that results from repe-
titions of similar organs or sets of organs. At a more detailed scale, a
particular focus has been put on metamer or phytomer repetition (White
1979; Barlow 1989) since this entity is composed of a node and its
leaf(ves) and axillary bud(s) plus the subtending internode, thereby con-
stituting the basic element of plant construction. However, the repeated
entities are not exactly similar and their development was shown to
change with their position within the tree structure and during plant
development. Different concepts have been proposed to account for the
existence of different metamer states and bud fates: “morphogenetic
program” and internal correlation (Nozeran 1984), “age state” (Gatsuk
et al. 1980), or “physiological age” of the meristems (Barthélémy et al.
1997). Even though the changes in bud fate and entity states are specific
to each species and lead to the differentiation of axes (e.g., orthotropic
versus plagiotropic axes; short versus long axes, etc.), general rules
established for a Rauh architectural tree model are as follows (Bar-
thélémy et al. 1997; Fig. 1.2): (1) an increase in shoot length and in axil-
lary shoot development during an initial period (observed in seedlings
and called “establishment growth”); (2) a period of stability during
which specific branching gradients can be observed (such as acrotony);
and (3) a progressive decrease in shoot length growth and in axillary
shoot development towards the final development stage or aging.
At the whole-tree scale, two concepts have been introduced in order
to define the branching system. First, the concept of “organization plan”
has been proposed to account for the hierarchic level between the con-
stitutive axes of a tree (Edelin 1991). The terms hierarchic versus pol-
yarchic were introduced to indicate a hierarchy between main shoots
and their laterals, respectively, or conversely the absence of hierarchy.
In forest trees, one easily can observe trees that develop in a hierarchic
way for a few years before developing a fork and becoming polyarchic.
Second, the concept of excurrent versus decurrent trees has been intro-
duced, in relation to apical dominance, in forest trees (Brown et al. 1967).
These terms, which refer to a definitive main stem producing lateral

branches (excurrent) or a main stem that spreads and becomes indis-
tinguishable from the uppermost branches (decurrent), have probably
been among the most commonly used in classifications of forest and fruit
trees.
We will present how these concepts, whether applied at the tree,
branching system, axis, or metamer scales, have been used in fruit tree
species. First, qualitative concepts will be considered to define the archi-
tectural models of several fruit species. Second, quantitative studies
will be described to highlight architectural edification rules as well as
morphological gradients in fruit species. The following section will thus
develop how the architecture of fruit trees has been or can be analyzed.
The last section will present how these results were integrated to
improve tree training and led to the proposal of new training systems.
Fig. 1.2. Schematic representation of architectural gradients in a tree belonging to the
Rauh model (Source: Barthélémy et al. 1997).

B. Defining Architectural Models of Fruit Tree Species
1. Identifying Shoot Types. The identification of architectural models in
a given species first requires identifying the categories of shoots that are
observed within the tree structure, on the basis of morphological crite-
ria. As with many other species, fruit trees exhibit a polymorphic devel-
opment of axes (Fig. 1.3). Usually, two main shoot categories, i.e., short
and long shoots, are distinguished by a simple visual observation (Cham-
pagnat 1965; Zimmerman and Brown 1971). Champagnat (1965) used
three criteria to define short shoots (or brachyblasts): a limited length,
(a) Spurs
(b) Medium shoots
(c) Long shoots
Fig. 1.3. Heteroblastic development of shoots in fruit tree: Example of shoot types in an
apple branch. (a) Spurs, (b) medium shoots, and (c) long shoots (Original drawing from J.
M. Lespinasse reprinted by his courtesy).

a limited number of organs, and possibly a limited life span. An equiv-
alent definition, based on morphological characters at the metamer level
and on shoot growth dynamics, has been proposed (Rivals 1965, 1966,
1967): short shoots are made of organs contained in the resting bud
(also called preformed organs), which do not elongate after bud burst.
In the horticultural context, different terms have been used for short
shoots, according to the species and to their floral or vegetative fate
(Champagnat 1954b; Forshey et al. 1992). In pome fruit species, short
shoots have been named “dards” when they are strictly vegetative or
“spurs” when they are usually floral. In stone fruits they have been
named “bouquets de mai” or “clusters” with respect to their early ces-
sation of growth after bud burst.
By contrast, long shoots are possibly made of (1) preformed organs
solely, whose internodes elongate, or (2) preformed organs followed by
neoformed organs resulting from apical growth (Rivals 1965, 1966,
1967). In the first case, the final length is limited and the corresponding
shoots often have been considered to be an intermediate category, that
of mesoblasts. In a horticultural context, other terms have been used for
intermediate shoots according to the fruit species, such as “brindles” in
apple or pear, and “chiffonnes” or “mixed shoot” in stone fruit species
(Boyes 1922; Champagnat 1954a). In the second case, shoots result from
both internodal growth of the preformed shoot and apical growth, which
produces a neoformed part, and generally form long shoots. These shoots
have been named auxiblasts but occasionally are called “water shoots”
in horticulture when growth is extended and/or rate is high. The term
“extension shoots” is also widely used for long shoots in both stone and
pome fruits.
However, most of the horticultural terms have the disadvantage of not
strictly fitting with a clearly defined biological phenomenon such as
internodal growth and apical growth, or with shoots composed of pre-
formed or neoformed organs. In addition, horticultural terms are often
ambiguous with respect to the perennial development of shoots. Indeed,
terms can refer either to the result of annual growth or the total growth
occurring over several years. For instance, a “spur” can simply be a one-
year-old short shoot or consist of a perennial set of branched shoots,
which have all remained short. As a consequence, there is no adequate
term in horticulture that can be used to speak of a short annual shoot
occurring in the second, third, or fourth year of growth on a long axis.
Considering more detailed levels of organization such as growth units
(GU) and metamers, even though there are no equivalent horticultural
terms, is thus necessary to describe precisely the development of fruit
trees over successive years.

2. Analyzing Branching. In addition to the identification of shoot types,
architectural analysis of a tree requires studying whole tree develop-
ment, analyzing the relative position of the shoots one to another, i.e.,
tree topology, all over the tree ontogeny. Architectural analysis is made
difficult for fruit trees because training systems generally alter tree archi-
tecture, often by pruning. Pruning cuts promote local re-growth, which
interacts with the natural growth and branching patterns. Thus it
appears more convenient, at least as a first step of investigation, to ana-
lyze the architectural development of trees grown with minimal train-
ing, in particular without severe pruning over several years. The
interactions between tree architecture and the agronomic practices then
can constitute a second step in the investigations, based on the knowl-
edge of the tree developmental rules. Section III will detail how this sec-
ond step can be carried out.
In temperate climates, investigations of the branching process of fruit
trees have focused on one-year-old shoots and on the winter period,
because key events, such as apical dominance and bud dormancy, occur
at this level of organization and during this period of time. These two
main events lead to correlative relationships between buds and dictate
the organization of the one-year-old shoot. According to Cline (1997,
2000), apical dominance is defined as the control exerted by the shoot
apex on the outgrowth of axillary buds. Its morphological consequence
is the inhibition of axillary buds during the growing season when they
are formed, often described under the term “bud dormancy.” Research
on tree morphogenesis (Crabbé 1980; Champagnat 1989) has initiated the
reappraisal of this generic term and led to the differentiation of three dif-
ferent physiological states of the bud—ecodormancy, paradormancy,
and endodormancy (Lang et al. 1987)—depending on whether bud inhi-
bition is conditioned on environmental causes or on a correlation with
other tree parts, or lies within the bud itself, respectively. Apical dom-
inance involves different correlative mechanisms mediated by auxin and
cytokinins (Cline 2000; Sussex and Kerk 2001) and the nutritional sta-
tus of the axillary buds (Champagnat 1989).
However, there are exceptions to the general pattern of apical domi-
nance since some lateral buds can develop during the growing season
in which they are formed, producing sylleptic shoots (Champagnat
1954b) (Fig. 1.4). For a review of terms used for branching, see Caraglio
and Barthélémy (1997). The capability of axillary buds to develop into
sylleptic shoots depends on the parent shoot relative growth rate
(Génard et al. 1994; Lauri and Costes 1995) and on the stage of tree matu-
rity. In fruit trees, sylleptic shoots mainly develop during the early
developmental years (Crabbé 1987), such as “feathers” in the nursery,

and these branches are considered to be advantageous for young tree
establishment (Maggs 1960; van Oosten 1984). Naturally occurring
sylleptic shoots have been shown to be naturally located in a median
position along the bearing shoot in apple (Costes and Lauri 1995; Costes
and Guédon 1997) and peach trees (Lauri 1991; Fournier 1994). This dis-
tribution makes it possible to select among these shoots, retaining some
of them and pruning those located in the lower part of the trunk or pos-
sibly including other criteria for further tree training.
Except in the case of sylleptic development, axillary buds develop
along the one-year-old shoot, after having passed the three stages of
dormancy previously described. In this case, they are called proleptic
or delayed shoots (Crabbé 1987), and the axillary shoot development
strongly depends on the bud position along the bearer shoot (see Fig.
1.4). In temperate fruit trees, the distribution of laterals most commonly
corresponds to an acrotonic distribution, i.e., the longest laterals are
located near (just below) the apical (distal) end of the one-year-old shoot
(Champagnat 1965; Champagnat et al. 1971). Such a distribution has
been described in many species, like apple (Crabbé 1987; Cook et al.
1998b), apricot (Costes et al. 1992), plum (Cook et al. 1998a), and wal-
nut (Solar and Stampar 2003). Quantitative investigations and modeling
of lateral distribution are described in section II.C.2 (Branching Patterns).
(a)
-1-
(b) (a) (b)
-2-
-3-
(a) (b) (c)
Fig. 1.4. Axillary shoot positions and associated terminology regarding branching:
1 (a) monopodial, (b) sympodial; 2 (a) sylleptic, (b) proleptic; 3 (a) acrotonic, (b) mesotonic,
(c) basitonic (Source: Caraglio and Barthélémy 1997).

3. Examples of Architectural Analysis in Fruit Trees. In the following
paragraphs suppress applications of architectural analysis will be illus-
trated for two main fruit species with contrasting architecture, apple and
cherry tree. Complementary elements will also be provided regarding
two other Prunus species, apricot and peach tree.
For the apple tree, numerous studies in the literature must be com-
piled to obtain a complete description of its architectural development.
Primary growth is usually rhythmic (Zanette 1981; Abbott 1984) and the
successive leaves spread along an axis with a spiral phyllotaxy whose
angle varies from 3/8 to 2/5 (Abbott 1984; Pratt 1990). Branching remains
monopodial before the occurrence of flowering and lateral branches are
displayed according to acrotony (Crabbé 1987). Thus, during the juve-
nile phase, which can be defined in a young seedling tree as a state char-
acterized by the inability to flower (Miller 1988), or during the vegetative
state of a non-flowering scion, the apple tree develops according to a
Rauh model (Lauri and Térouanne 1995).
The different shoot categories that can be identified within an apple
tree are often classified into two or three types, based on their length and
on their constitutive growth unit types (see Fig. 1.3). Short axes (or spurs)
are composed only of short GUs whose constitutive metamers elongate
slightly or not at all. As a rule of thumb, the length of each constitutive
GU is less than 5 cm. Two types of short GU can be distinguished accord-
ing to whether the apical meristem is differentiated into an inflorescence
(flowering GU) or not (vegetative GU). In the case of floral GU, a leafy
basal part is followed by a floral distal part (Fulford 1966a, b; Abbott
1984). This GU, whose diameter is often increased by the presence of an
inflorescence and fruit development, is usually named a “bourse.”
“Brindle” or medium shoots are constituted of GUs whose lengths reach
6 to 20 cm (Lespinasse and Delort 1993). Long GUs, also called exten-
sion shoots, correspond to GUs whose apical meristem has a prolonged
activity, leading to the development of neoformed metamers.
Floral differentiation, which occurs in the terminal position of axes,
ends the monopodial phase. Branching on the floral GUs is immediate
and sympodial ( Crabbé and Escobedo 1991). Because of the change in
branching mode, from monopodial to sympodial, the architectural
model of apple tree evolves from the Rauh to the Scaronne model (Lauri
and Térouanne 1995). In addition, adult trees usually exhibit a poly-
archic organization resulting from both their sympodial branching and
gravimorphic reactions. Indeed, long shoots begin to bend usually after
fruiting has begun with long re-growth developing in the upper part of
curved axes (Crabbé and Lakhoua 1978). However, this tendency varies
greatly among genotypes (Lauri et al. 1995).

It is worth noting that at least two other fruit species have a roughly
similar architectural development to that of apple tree: pear tree, which
also belongs to the Rosaceae family, and walnut tree from the Juglan-
daceae family. Indeed, walnut trees have a rhythmic growth, and a
monopodial and acrotonic branching until flowering occurs (Sabatier et
al. 1998; Sabatier and Barthélémy 1999; Solar and Stampar 2003).
Terminal flowering is a discriminant feature between apple as opposed
to Prunus species, since in these latter species flowers differentiate in
lateral positions along one-year-old shoots, while the apical bud remains
vegetative. Two cases can be distinguished that correspond to cherry or
to apricot and peach trees, respectively (Fig. 1.5). In cherry trees, flow-
ers differentiate in lateral buds located on the preformed zone of the one-
year-old shoots. Thus, after bud burst, flowers are located on the basal
part of short and long shoots and are clearly separate from vegetative
buds. In both apricot and peach trees, floral bud differentiation can
occur in meristems located along the one-year-old shoots either directly
in an axillary position or on prophylls (i.e., the first two foliar organs of
a shoot) of axillary buds (see Fig. 1.6a).
Flower or inflorescence
Vegetative bud
1-year-old shoot
2-year-old shoot
Short shoots
(a) (b)
Fig. 1.5. Location of floral differentiation with respect to vegetative buds and shoot
organization in two Prunus species: (a) cherry tree—complete separation between floral
and vegetative zones; (b) apricot or peach tree—flowers and vegetative buds associated at
same nodes.

Flowering position, combined with the terminal meristem behavior,
defines different tree architectures. In cherry and peach trees, the ter-
minal meristem remains alive, leading to a monopodial branching sys-
tem, while in apricot (or plum) trees the terminal meristem usually dies
after each growth period, leading to sympodial branching. Thus, cherry
and peach trees both correspond to the Rauh model. In particular, the
cherry tree corresponds very strictly to the Rauh model definition, since
it is constituted of clearly defined short and long shoots and exhibits a
pronounced hierarchic structure. Moreover, the acrotonic gradient is
particularly abrupt, leading to long shoots located solely on the upper-
most nodes of the annual shoots and thus to a rhythmic distribution of
long branches along the main axis. By contrast, in the peach tree, Lauri
(1991) observed a more polyarchic organization resulting from the devel-
opment of basitonic reiterated complexes.
The apricot tree also exhibits a polyarchic architecture based on
entirely sympodial branching and including a continuum of shoot types
between short and long shoots (Costes 1993). The apricot tree provides
an example of the Champagnat model, with a definite and rhythmic pri-
mary growth, sympodial branching, and long shoots that naturally bend.
Shoot bending leads to the development of re-growth from short shoots
or latent buds located on the upper side of the curved axes. This pattern
has been shown to repeat with tree aging, leading to the formation of suc-
cessive reiterated complexes whose size decreases from the center to the
periphery of the trees (Fig. 1.6). Thus, the polyarchic organization can
be exhibited very early in the apricot tree ontogeny. However, this whole
tree organization depends on the cultivar, for some apricot cultivars,
such as ‘Stark Early Orange’, exhibit a dominant central trunk through-
out the life of the tree (Costes et al. 2001a) (Fig. 1.7).
4. Describing the Intra-Species Variability of Tree Architecture. While
the aim of architectural studies is to extract invariant features that may
adequately define the architecture of a given species, the variability that
exists among cultivars in growth, branching, and flowering location
must also be explored to propose optimized training methods, adapted
to the different behaviors observed within a given species.
Different criteria have been used to qualitatively classify cultivars
within the different fruit tree species. A pioneer study in this domain was
proposed by Bernhard (1961), who first attempted to type apple trees
according to both the overall tree growth pattern—i.e., direction of growth
of scaffold branches, from upright to weeping—and their fruiting types
(types I to IV) (Fig. 1.8). Type I apple cultivars mostly bear fruits on
spurs that are branched on “old wood,” whereas type IV cultivars mostly

carrier shoot (annual shoot n)
flowering shoots (annual shoots n+1
and axillary short twigs)
axillary bud alone
axillary bud with two
flowers on its prophylls
Short twigs (spurs)
Second growth unit: GU2
First growth unit: GU1
annual
shoot
n+1
annual
shoot
n
Fig. 1.6. Apricot tree architecture. (a) Organization of annual shoots showing the sym-
podial branching and the location of floral buds; (b) schematic representation of apricot
tree at adult stage showing the progressive decrease in size of the successive branching
systems with tree aging (Source: Costes 1993).
(a)
(b)

Erect
Weeping
Stark
Early
Orange
Goldrich
Orange
Red
Palsteyn
Slender Thick
Shoot slenderness
Shoot
bending
Bergeron
Sortilège
Fantasme
Lambertin
Harcot
Comédie
Fig. 1.7. Qualitative classification of apricot varieties observed in France, based on two
main criteria represented respectively in X and Y axes: (i) the shoot slenderness, (ii) shoot
bending.
ideotype 1 ideotype 2 ideotype 3 ideotype 4
Fig. 1.8. Apple ideotypes from spur (type I) to weeping trees (type IV), as defined by
Lespinasse (1992) based on previous studies on “fruiting types” from Bernhard (1961) and
Lespinasse (1977).

bear fruit at the terminal positions on brindle-type shoots. Lespinasse
and colleagues (Lespinasse 1977; Lespinasse and Delort 1986) later
included a third parameter, the position of the scaffold branches along
the trunk from basitonic to acrotonic. Lespinasse (1992) subsequently
proposed including all spur-type cultivars in type II, restricting type I
to cultivars that exhibit a typical columnar habit (mainly produced by
English breeding selection programs; Tobutt 1985, 1994). Thus all spur-
type apple cultivars will hereafter be considered to belong to type I/II.
These spur-type cultivars are characterized by a temporal and spatial dis-
junction between vegetative growth and fruiting since they usually have
strong, erect shoots with no or little terminal fruiting. By contrast, type
IV (tip bearing type) cultivars develop fruit in terminal positions on all
types of shoots, including water-shoots. These architectural features are
related to fruiting pattern, with alternate vs. regular fruiting patterns,
respectively (Lauri et al. 1995, 1997a, b). Between these two extremes,
types II and III (standard type) have an intermediate growth and fruit-
ing pattern.
Classifications based upon similar criteria, i.e., branch orientation,
position of flowering and lateral branches, have been proposed in other
fruit species, such as walnut by Germain (1990, 1992). Qualitative clas-
sifications also have been proposed on the basis of shoot types or mix-
ing shoot type with branching density, as the phenotypic classes of
peach cultivars proposed by Scorza (1984) and further studied by Bassi
et al. (1994) or in pear trees by Sansavini and Musacchi (1994).
C. Quantitative Studies of Fruit Tree Topology
The existence of generic rules and the repetitive nature of plant con-
struction led different scientists to introduce mathematics into plant
architectural studies. Different formalisms were proposed to simulate
plant growth processes (Borchert and Honda 1984; Fisher and Weeks
1985; Prusinkiewicz and Lindenmayer 1990; Fisher 1992; Prusinkiewicz
et al. 1997; Barczi et al. 1997). On the other hand, tree structure was
explored in order to quantify the general rules of tree architecture devel-
opment that were first highlighted from a qualitative and conceptual
point of view. Pioneer research in this area was performed on coffee trees
by introducing stochastic modeling of meristem activity (de Reffye
1981a, b, c). Four processes were considered: (1) primary growth, i.e.,
the dynamics of metamer emergence; (2) branching, i.e., the probability
of a given axillary meristem to elongate into a shoot; (3) flowering, i.e.,
the probability of a given terminal or axillary meristem to develop into
a flower; and (4) the probability of meristem mortality.

According to Godin et al. (1999), the measurements made to charac-
terize plant structure and the different organs plants are composed of can
be organized into two categories: (1) those dealing with the form or the
spatial location of organs and therefore defining the organ geometry and
(2) those that enumerate the organs and determine their relative con-
nections, therefore defining tree topology. Topological descriptions have
been designed to simultaneously integrate several organization levels
using Multi-scale Tree Graph (MTG) as an underlying model (Godin and
Caraglio 1998). Specific softwares, such as 3A (Adam et al. 1999) and
AMAPmod (Godin et al. 1997), are currently freely available to collect
data for plant architectural databases, and explore them with appropri-
ate statistical tools, respectively.
1. Primary Growth. The number of metamers per axis has been modeled
by binomial distributions on both orthotropic and plagiotropic axes of cof-
fee trees (de Reffye 1981c). In this approach, the two parameters of a bino-
mial distribution, p and n, represented respectively the probability that a
new metamer emerged from the terminal meristem and the total number
of leaves that potentially could be developed by a given axis type. This sto-
chastic approach was adapted to fruit trees with rhythmic growth, such as
litchi (Costes 1988) and apricot (Costes et al. 1992). In the case of rhythmic
growth, the number of metamers developed per growth unit was modeled
either by a Poisson distribution, in the case of entirely preformed organs,
or by a mixture of two distributions, in the case of growth units composed
of both preformed and neoformed organs (de Reffye et al. 1991). For apri-
cot, the mixture included a binomial distribution for preformed organs and
a negative binomial distribution for the neoformed part of the shoot.
This formalism made it possible to demonstrate that the progressive
decrease in annual shoot length over time resulted from a decrease in
the number of neoformed organs, while the number of preformed organs
remained almost invariant (Costes et al. 1992, Fig. 1.9). This result was
consistent with the observations of Rivals (1965), who assumed a con-
stant number of preformed primordia in resting vegetative buds, this
number being at the first order specific to the species. However, within
a given species, the number of preformed organs was shown to depend
on bud location within branching systems (Costes 2003). Similarly, the
acrotonic distribution of axillary shoots was shown to correspond to a
decrease in the neoformation development of these shoots according to
their position from the distal end of the initial shoot. These concepts
were useful for explaining the different shoot types in Actinidia (Selez-
nyova et al. 2002) or the progressive decrease of successive GUs along
axes in apple trees (Costes et al. 2003b).

2. Branching Patterns. In temperate fruit trees, axillary buds develop
during two main periods (during the current growing season producing
a sylleptic shoot, or during the following growing period producing a
proleptic—or delayed—shoot) and at different locations along the orig-
inal shoot. The median distribution of sylleptic shoots organizes the
branching pattern along the main shoot in three successive zones that
are observed from the base to the top: (1) not branched, (2) branched, and
(3) not branched zones. In addition, axillary shoots can be divided into
three types according to their length (short, medium, and long). Thus, the
class of Markovian models was selected, since it emerged as a reference
Number of GUs
1st GU 1987
1st GU 1988
1st GU 1989
1st GU 1990
1st GU 1991
2nd GU 1987
2nd GU 1988
2nd GU 1989
2nd GU 1990
2nd GU 1991
Number of GUs
Number of metamers Number of metamers
Fig. 1.9. Distributions of the number of metamers per growth unit (GU) during the suc-
cessive years of growth (from 1987 to 1991) and for two successive GUs. The 1st GUs are
represented by a set of histograms on the left and the 2nd GUs by a set of histograms on
the right of the figure. For each histogram, the total number of metamers per GU was mod-
eled as a mixture of binomial distribution representing the number of preformed organs
of the shoot and a negative binomial distribution for the number of neoformed organs. The
number of neoformed metamers decreased progressively with years in both 1st and 2nd
GUs (Source: Costes et al. 1992).

for analyzing successions of homogeneous zones in discrete sequences,
in both computational molecular biology and plant architecture domains
(Guédon et al. 2001). More precisely, hidden semi-Markov chains
(Durbin et al. 1998) were used to represent both the succession of zones
and the proportion of shoot types within the branched zone. At the first
level, a semi-Markov chain represents the succession of zones and the
length of each zone, the successive zones being connected by transition
probabilities. The second level consists in associating each zone with a
discrete distribution representing the different probabilities of different
types of axillary shoots. The resulting organization of axillary shoots
according to specific zones along parent shoots has been demonstrated
and modeled in several fruit species, including apple (Costes and Gué-
don 1997; Costes et al. 1999; Costes and Guédon 2002), peach (Fournier
et al. 1998), and Actinidia (Seleznyova et al. 2002).
This modeling approach also has been applied to the distribution of
both sylleptic and proleptic axillary shoots in one-year-old apple trees,
comparing a set of cultivars belonging to contrasting architectural types
according to the classification of Lespinasse (1992). All were shown to
present a similar organization in successive branching zones that dif-
fered one from the other by their composition of axillary shoot types
(Costes and Guédon 2002). Roughly, all the cultivars exhibited six suc-
cessive zones that could be described from the distal end of the shoot
as follows: in the most distal zone, long proleptic shoots were observed
mixed with latent buds and short shoots; the second zone was occupied
mainly by lateral bourses mixed with latent buds; the third zone corre-
sponded exclusively to sylleptic shoots. These first three zones spread
over the upper half of the bearer shoot. The basal half of the shoot com-
prised the three remaining zones: two unbranched zones flanking a
large branched zone where long proleptic shoots and spurs were mixed
with latent buds (Fig. 1.10).
The different cultivars were shown to differ by the length of each zone
and the relative proportion of the axillary shoot types within each zone.
The long lateral shoots, which can appear in the nursery, were observed
in three zones: the most distal, the median zone (corresponding to the
sylleptic zone), and the basal branched zone. Thus, the total number of
long shoots as well as their relative position along the trunk differed
according to cultivar. On average, all cultivars developed more than ten
long laterals, except ‘Wijcik’. Most of the long shoots were located in the
proximal zone in ‘Reinette Blanche du Canada’ and in ‘Fuji’. In ‘Belrène’,
the long proleptic shoots were located equally in the distal and proxi-
mal zones. A high number of long shoots in the distal zone were
observed in ‘Granny Smith’, reflecting an acrotonic behavior. ‘Imperial

20
0.80
0.69
0.23
0.08
0.83
0.1
0.92
0.08
1
0.92
0.08
1
(73.93)
100
50
100
50
100
50
100
50
100
50
0
1
2
3
4
(75.06)
0.93
0.07
1
0.87
0.13
0.93
0.07
1
0.12
0.37
0.63
1
100
50
100
50
100
50
100
50
100
50
100
50
0
1
2
3
4
(
73.00)
0.57
0.43
1
1
1
1
1
0.06
0
1
2
3
4
100
50
100
50
100
50
100
50
100
50
Latent
bud
(symbol
0)
Spur
(symbol
1)
Long
delayed
shoot
(symbol
2)
Most
probable
axillary
shoot
in
the
state
Axillary
floral
shoot
(symbol
3)
Sylleptic
shoots
(symbol
4)
Diffuse
mixture
of
spurs
and
long
delayed
shoots
Transition
probability
0.1
1
1
1
100
50
100
50
100
50
100
50
0
1
2
3
symbols
100
50
100
50
Fuji
(III)
Granny
(IV)
Wijcik
(57.26)
0.15
0.25
0.75
0.32
0.68
0.68
0.14
0.86
0.37
0.63
0.94
0.88
0.20
Reinette
(II)
symbols
symbols
symbols
Fig.
1.10.
Simplified
representation
of
the
branching
patterns
along
one-year-old
trunks
in
four
apple
cultivars:
‘Fuji’
(type
III),
Granny
Smith
(type
IV),
and
Reinette
Blanche
du
Canada
(type
II),
and
‘Wijcik’
(type
I,
compact).
Branching
patterns
are
modeled
by
hidden
semi-Markov
chains
represented
as
follows:
each
branching
zone
is
represented
as
a
state
and
its
length
is
represented
by
its
mean
number
of
nodes;
transitions
between
states
are
represented
by
arrows,
with
transition
probability
noted
nearby;
the
proportion
of
latent
buds
(symbol
0),
short
(1),
long
(2),
floral
(3),
and
sylleptic
(4)
axillary
shoots
within
each
zone
is
represented
by
a
histogram
attached
at
the
right
of
the
considered
zone.
The
total
mean
num-
ber
of
nodes
per
shoot
is
noted
between
brackets
at
the
bottom
of
the
diagram
(Source:
Costes
and
Guédon.
2002.
Annals
of
Botany,
Modeling
Branching
Patterns
on
1-Year-Old
Trunks
of
Six
Apple
Cultivars,
Vol.
89,
p.
520,
Figure
6.
With
permission
of
Oxford
University
Press.)

Gala’, ‘Granny Smith’, and ‘Elstar’ exhibited the highest mean values of
long shoots in the median zone. In ‘Granny Smith’, however, the zones
were densely branched and separated by latent buds, while in ‘Imper-
ial Gala’, the long shoots were more equally distributed and mixed with
latent buds (Costes and Guédon 1997, 2002). The branching pattern of
‘Imperial Gala’ along two-year-old trunks seemed more adapted to the
further branching organization of the adult tree.
3. Location of Flowering. Plant architectural descriptions previously
showed that flowering is often linked to the branching process. In most
temperate fruit species, flowers or inflorescences differentiate on leafy
shoots a few weeks after metamer expansion (Foster et al. 2003). Thus,
flower organogenesis occurs when and where sylleptic shoots can grow
(Crabbé 1987). However, they bloom the following year and finally fruits
that are currently borne on one-year-old shoots develop at the same
time as proleptic shoots.
Therefore, flowering distribution along shoots has been modeled
according to the same philosophy as branching, exploring the number
of flowers associated at each node rank, either with sylleptic shoots (on
current year shoots) or with proleptic shoots (on one-year-old shoots).
In both peach and apricot trees, because of the axillary position of flow-
ers and the possible vegetative or floral fate of the main axillary bud, two
variables must be simultaneously considered in order to represent,
respectively, the main axillary bud fate and the number of associated
flowers.
In the peach tree, long, medium, and short one-year-old shoots were
compared by analyzing the number of lateral flowers relative to the cen-
tral bud fate, as either a sylleptic shoot or as a flower (Fournier 1994,
Fournier et al. 1998, Fig. 1.11). Whatever their type, peach tree shoots
were highly structured from the base to the top, and this organization was
described as a succession of zones. The proximal and distal zones, which
contained latent buds and no flowers, were present in all shoot types.
Similarly, the zone that contained central buds, which had differentiated
into flowers, was always located in the upper half of the shoot. Two
median zones contained one associated flower with short sylleptic shoots
or vegetative buds. An additional zone, which contained two or more lat-
eral flowers, was observed in the median part of the longest shoots only.
This floral zone also contained long sylleptic shoots. Thus, the number
of zones that contained associated flowers, as well as the number of
flowers per node and the number of sylleptic shoots, increased with
shoot vigor. However, the number of flowers also can be affected by root-
stocks since an intermediate growth rhythm has been shown to promote

22
(a)
(b)
(c)
X1:
type
of
axillary
development
3
sylleptic
shoot
2
vegetative
bud
1
central
flower
0
blind
node
X2:
number
of
lateral
flowers
0
0
0
0
0
0
3
3
3
3
2
2
2
1
1
1
1
2
2
2
1
2
1
0
0
0
0
0
0
0
0
0
0
0
0
2
2
1
10
20
30
X2
0
0
0
0
0
0
3
3
3
3
2
2
2
1
1
1
1
X1
2
2
2
1
2
1
0
0
0
0
0
0
0
0
0
0
0
0
2
2
1
10
20
30
10
20
30
Zone
with
latent
buds
and
no
flower
Zone
with
vegetative
buds
and
0
to
1
lateral
flower
per
node
Zone
with
sylleptic
shoots
and
0
to
2
lateral
flowers
per
node
Zone
with
vegetative
buds
and
1
central
flower
per
node
Number
of
nodes
Fig.
1.11.
Branching
patterns
of
(a)
long,
(b)
medium,
and
(c)
short
mixed
shoots
of
peach
tree,
modeled
by
hidden
semi-Markov
chains
that
asso-
ciate
the
number
of
lateral
flowers
(X2
variable)
to
the
axillary
bud
fate
(X1
variable).
The
length
of
each
zone
is
represented
by
the
mean
num-
ber
of
nodes
(Source:
Fournier,
D.,
Costes,
E.
and
Guédon,
Y.
1998.
A
Comparison
of
Different
Fruiting
Shoots
of
Peach
Tree.
Acta
Hort.
(ISHS)
465:557–566
http://www.actahort.org/books/465/465_69.htm).

more floral differentiation in peach trees grafted on St Julien rootstock
than on those grafted on more vigorous rootstocks (Edin 1982).
In apricot, the association between flowers and axillary proleptic
shoots was only studied on long annual shoots (Costes and Guédon
1996; Costes et al. 1999). As previously, two variables were considered:
the number of flowers associated with each node and the type of pro-
leptic axillary shoot that developed at this node. Along these long shoots,
only the basal part did not bear any flowers. Two-thirds of the upper part
of the shoots were potentially floral. However, the transitions between
the number of flowers per node were gradual: they increased progres-
sively from one to three or more flowers, from the base to the top of the
shoot, and then decreased symmetrically from three flowers to two and
then one flower per node. As previously described in the peach tree,
sylleptic shoots were observed more frequently in the zone that con-
tained three flowers or more. Such an increasing gradient was also
observed in annual shoots composed of several GUs in apricot trees, with
the second and third GUs bearing more flowers than the first GU within
annual shoots (Clanet and Salles 1974).
4. Meristem and Shoot Mortality. In all trees, different shoot categories
usually exhibit different life spans. Shoot death is a general phenome-
non that may occur in large branches in forest trees or, more usually in
short shoots, from the year of bud burst to several or many years later
(Bell 1991).
Meristem mortality has been modeled by an exponential distribution,
considering death probability as a constant (de Reffye 1981b). In tem-
perate fruit trees, among the different shoot categories, short shoots usu-
ally have the shortest life span. Thus they have been the most studied
organs with respect to mortality. These studies were mainly focused on
apple by Lauri et al. (1995, 1997a, b), who demonstrated that spur death,
also called “extinction,” is a precocious phenomenon in lateral devel-
opment and depends upon the cultivar. Moreover, spur death was
shown to be correlated positively to the capability of each cultivar to
bear fruits regularly in the remaining branchlets through the “bourse-
over-bourse” phenomenon, which is defined as the proportion of fruit-
ful laterals that give rise to a fruitful lateral the following year. This
phenomenon was first described in apple (Lauri et al. 1997a; Fig. 1.12)
and, more recently, demonstrated in pear (Lauri et al. 2002). Spur extinc-
tion was also shown to occur at a constant rate over years and to be
higher for spurs on medium shoots than on long shoots in both ‘Fuji’ and
‘Braeburn’ (Costes et al. 2003b). Thus, it appears that spur extinction is
an interesting horticultural trait, specific of cultivar.

D. Describing Fruit Tree Form
Tree form is another criterion that defines tree architecture, even though
little information about plant geometry has been initially included in archi-
tectural model definitions. The variability of tree forms within a given
species has been described as a continuous phenomenon, from upright to
weeping (Bernhard 1961; Lespinasse 1977; Scorza 1984; Sansavini and
Musacchi 1994). However, tree and shoot form remains a quite vague con-
cept, and is quite difficult to measure and formally describe, especially
when the considered trees do not exhibit an extreme behavior.
At the whole tree scale, tree form can be defined according to canopy
volume and to the branching organization. It can be evaluated through the
overall tree hierarchic organization using the concepts of hierarchy vs.
polyarchy introduced by Edelin (1991) and used, for instance, to describe
two-year-old apple trees (De Wit et al. 2002). It also can be evaluated
through the concepts of excurrent vs. decurrent trees (Brown et al. 1967).
At more detailed scales, tree form relies on that of its constitutive
organs: for instance, a weeping tree can be viewed as a set of weeping
long axes, while an erected tree is a set of erected axes. Stem form and
orientation are important components of intra-specific fruit tree archi-
tectural diversity and have a qualitative and quantitative impact on fruit
production: bending or tilting stems increase flowering, reduce vigor,
and modify the branching pattern of the stems (Wareing and Nasr 1958).
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
bourse-over-bourse
extinction
Granny Smith
Red Winter
Jonagold
Royal Gala
Melrose
R. des Reinettes
O. Spur
Delicious
Fuji
Braeburn
Golden Delicious
Fig. 1.12. Relationship between bourse-over-bourse and extinction for the various cul-
tivars. Each point represents the mean value for the couples of years 1-2 and 2-3. Bourse-
over-bourse is defined as the proportion of fruitful laterals that give rise to a fruitful
lateral the following year. Extinction is defined as the proportion of laterals that abort
(Source: Lauri et al. 1997a. Reproduced with permission of the Journal of Horticultural Sci-
ence of Biotechnology.)

Moreover, taking into account stem form and orientation allows one to
tackle problems related to the within tree heterogeneity: (1) when leaves
or fruits are considered, spatial position and aggregation are input vari-
ables with which to estimate light interception within the canopy;
(2) when shoots are considered, it becomes possible to study the geom-
etry of axes and its changes with time.
In the following paragraphs, we introduce briefly the techniques that
are presently available to collect 3D coordinates of plant organs and we
describe the different models of forms that have been proposed to rep-
resent either the whole tree, the axes, or the tree organs at more detailed
scales. According to the scale considered, applications dealing with
light interception or shoot bending prediction are mentioned.
1. Measuring a 3D Form. Different digitizing techniques have been
developed involving articulated arms and sonic and magnetic methods
to measure the 3D coordinates of plant constituents (Sinoquet et al.
1997). Depending on the study goals, a method based on digital 3D mea-
surements can be applied either to leaves or to axis segments (Fig. 1.13).
A further step, which consists of coupling plant topological description
to that of constituent geometry, was achieved by coupling a one-scale
description of plant topology to sonic digitizing (Hanan and Room 1997)
(a) (b)
Fig. 1.13. Example of three-dimensional representation of digitized apple trees acquired
using a BD magnetic Polhemus digitizer (Adam et al 2001). (a) Branch digitized at leaf scale
and visualized with VegeStar software (Source: Massonnet et al. 2004); (b) Whole tree dig-
itized at woody segment scale, and reconstructed by AMAPmod software (Unpublished
data from Costes and Sinoquet).

or by coupling a multi-scale tree graph representation of plant topology
to magnetic digitizing (Godin et al. 1999). In both methods, the follow-
ing softwares are available for data acquisition: 3A (Adam et al. 1999)
and Floradig (Hanan and Wang 2004).
Despite these software developments, the acquisition of organ 3D
coordinates remains time consuming, and simplification procedures are
currently under investigation. Casella and Sinoquet (2003) proposed
the use of allometric relationships between shoot length, number of
leaves, and leaf area to reconstruct 3D architecture from sampling of 3D
coordinates. Another solution could be to reconstruct 3D coordinates
from stereoscopic photographs coupled to automatic processing that
allows an automatic extraction of morphological parameters (Kaminuma
et al. 2004). A good correlation between direct 3D measurements of tree
canopies and 3D reconstruction has also been recently proposed, which
relied upon the calculation of gap fractions from series or peripheral pic-
tures (Phattaralerphong and Sinoquet 2005).
2. Models for Representing Whole Tree or Row Form. The question of
whole canopy representation has been developed mainly in the context
of physical exchanges between canopy and the environment, especially
light interception. In these approaches, canopy structure has been con-
sidered at the whole tree, row, or orchard scales, and simple geometri-
cal models are sometimes considered as sufficient (e.g., Li et al. 2002).
Jackson and Palmer (1972), who pioneered the calculation of light inter-
ception in orchards, first considered solid, non-transmitting, and non-
reflecting hedgerows of different forms, latitudes, and times within the
year. The hedgerows were considered either as triangular, truncated tri-
angular, or rectangular in cross-section. Palmer and Jackson further
refined this modeling approach by considering the tree canopies to be
transmitting turbid media according to the law of Beer-Lambert (Palmer
and Jackson 1977; Jackson and Palmer 1979; Jackson 1980).
Both simpler and more complex models have also been proposed to
represent the whole tree canopies. For instance, a two-dimensional (2D)
model for representing orchard rows in light interception estimations for
different fruit tree species has been proposed by Annandale et al. (2004)
(Fig. 1.14). Refinements also have been introduced by considering each
tree individually, with tree shape being approximated as conic, para-
bolic, cylindrical, or as intermediate between a cone and a cylinder
(Wagenmakers 1991). Excellent estimations of light interception have
been obtained for symmetrical and elliptical canopies, but discrepancies
occur with asymmetric canopies or when the assumption of a uniform
leaf area distribution was not valid. This drawback can be overcome by
a three-dimensional (3D) model that makes it possible to account for

canopy asymmetry, such as the model proposed by Cescatti (1997) in a
forest context. However, this approach has not yet been applied in a hor-
ticultural context.
3. Modeling Axis Form Changes. Except for mutants, in which bending
was described during the first year of growth (Monet et al. 1988), main
changes in axis form occur in one-year-old shoots when the thin and
long to medium shoots are fruiting for the first time. Thus, a weeping
habit can be assumed to result from the individual shoot propensity to
bend under its own weight and the fruit load.
The elaboration of stem form was first studied in a forest context, and
is the purpose of tree biomechanics (Castera and Morlier 1991; Fournier
et al. 1991a, b; 1994). Stem form depends on several factors related to
its growth habit. The first factor is the primary direction of elongation
of the apex, which can be modified by subsequent re-orientations of the
stem. The weight of wood, axillary shoots, leaves, and fruits causes
bending of the stem. The intensity of this bending depends on the
amount and location of loads, dimensions of the stem, and mechanical
properties of wood. Secondary growth creates an increase in stem rigid-
ity, and the relative dynamics of loading and diameter growth plays an
important role in the final shoot form (Fournier et al. 1994). Another
effect of diameter growth is active re-orientation of the stem, due to the
maturation of new wood layers and more specifically to the action of ten-
sion wood (Archer 1986).
Fig. 1.14. Simplified ellipsoidal representation of whole tree form and hedgerow used
in a 2D solar radiation interception model (Source: Annandale, J. G., N. Z. Jovanovic,
G. S. Campbell, N. D. Santoy, and P. Lobit, 2004. Two-dimensional Solar Radiation Inter-
ception Model of Hedgerow Fruit Trees, 207-225. With permission of Elsevier.)

A modeling approach carried out in three contrasting varieties of apri-
cot trees showed that the main factors involved in final shoot form were:
(1) its initial geometry (in particular its slenderness and inclination),
and (2) the distribution of loads along the shoot (Alméras 2001; Alméras
et al. 2002) (Fig. 1.15). The dynamics of cambial growth also impacts re-
orientation, which corresponds to an up-righting movement after harvest,
since lignification stiffens the shoot during the period of maximal cur-
vature due to fruit development. By contrast, the mechanical properties
of the wood (i.e., its modulus of elasticity and the presence of tension
wood) have a small impact on the final shoot form (Alméras et al. 2004).
These results suggest that the variables related to shoot morphology
are the first targets to evaluate the propensity of a shoot to bend among
different genotypes. Diameter and shoot length constitute elementary
variables for characterizing the shoot form, and are putative descriptors
of the genetic variability (Kervella et al. 1994), despite the fact that they
may be plastic under various environmental conditions (Fournier et al.
2003). Other variables, such as the variation in shoot curvature over two
years, shoot slenderness, or branching angles, could also be relevant but
need to be confirmed by further studies.
-
0.5
0
0.5
1
-
0.5
0
0.5
1
-
–0.5
0
0.5
1
0 0.5 1
Simulated
T1
T1
T1
T1
T2
T0 T0
T0 T0
T2
T2
T2
0.5
0
0.5
1
–0.5
0
0.5
1
0 0.5 1
(a) (b)
Measured
Measured Simulated
Fig. 1.15. Observed and simulated dynamics of long shoot form during the second
growth season, from blooming time (T0) to physiological fruit drop (T1) and a few days
before harvest (T2). Example of a shoot belonging to (a) ‘Lambertin’ and (b) ‘Modesto’ cul-
tivars (Source: Alméras, 2001).

4. Models for Representing the Organ Distributions within Canopy. In
horticulture, there is a major interest in quantifying the heterogeneity of
organ environment within the canopy, especially considering the het-
erogeneity of leaves and fruits during their development. Indeed, the
organs of the same plant may be subject to contrasting environmental
conditions, especially for light distribution, and this may result in dif-
ferential responses, e.g., in terms of carbon assimilation potential of the
leaves and fruit coloring.
Two main types of plant representations that account for the organ dis-
tribution within the canopy have been proposed. Firstly, individual trees
can be split into voxels resulting from a spatial discretization of the space
occupied by the tree. Leaf area density (LAD) in a voxel is then assumed
to be uniformly and randomly distributed and spatial variation of leaf
area density is therefore accounted for by the inter-voxel differences in
LAD. This approach has been used for modeling light capture with a tur-
bid medium analogy and has been used to compute radiation balance at
canopy, plant, and shoot scale (Sinoquet et al. 1991). Secondly, organs
can be explicitly described, with their shape, size, orientation, and spa-
tial co-ordinates being taken into account in 3D plant mock-ups. These
mock-ups may be provided by either digitizing methods or simulation
softwares. Thus, methods based on polygon projection or on Monte Carlo
ray-tracing can be used for modeling light capture at the organ scale.
These approaches have been recently reviewed by Godin et al. (2005).
Finally, the numerous studies that have been carried out on fruit tree
architecture provide nowadays a large set of concepts, methods, and
techniques to quantify both the tree topology and geometry, with the
possibility to choose between different scales according to specific goals.
This overall knowledge constitutes a framework that also benefits
orchard and tree training systems, since growth, branching, and flow-
ering processes can be explicitly taken into account in their conception.
III. CONSEQUENCES OF TREE ARCHITECTURE FOR
TREE TRAINING, ORCHARD MANAGEMENT, AND
FRUIT PRODUCTION
Training systems have drawn considerable attention over the past 40
years since they must combine different purposes. Those have been
revealed to be more or less conflicting, since economic conditions have
varied over time. The main purposes, especially in intensive orchards,
are the following: (1) a rapid achievement of a developed canopy struc-
ture to reach orchard maturity and maximum fruit production within a

few years; (2) an optimal capture of light to optimize carbon gain and
fruit yield per hectare; (3) a fair distribution of intercepted light within
the aerial system of the tree to minimize the spatial heterogeneity of
local vegetative growth and fruit quality; and (4) management of tree
shape and fruit load with minimal pruning, to take advantage of the nat-
ural trends of the cultivar and reduce the economic cost of this manual
operation.
This last point is of major importance since training systems initially
conceived to improve light interception by the tree overall may stimu-
late growth of vigorous water-shoots, i.e., reiterated complexes, on the
upper side of scaffold branches. If not removed, these shoots acting
mainly as assimilate sinks may also thwart the benefits of high illumi-
nation within the tree by decreasing light interception by fruiting shoots.
On the other hand, an unpruned tree, in which vigor is well-distributed
to fruiting shoots, quickly begins production but, in most cases, results
in an overcrowded canopy after some years and eventually fruit size and
quality are reduced.
Training methods have then been particularly developed at the tree
scale to manipulate both the vegetative and the fruiting components.
Pruning vegetative shoots at different positions in the tree or/and at dif-
ferent phenological stages is used both for the building of the tree struc-
ture, according to a specific tree shape, and to optimize light distribution
within the canopy as in cherry (Flore et al. 1996) or in apple (Barritt
1992). Bending or tying down branches is often used with two objectives.
One is to maintain the tree in the allotted space in relation with the tree
management system. It is preferred to heading cuts for the control of tree
growth and shape and is currently used in particular training proce-
dures, as described for Solaxe (see section III B 3). A second objective
of bending is to reduce vegetative growth of the branch and promote
flowering. However, the effects of bending on flowering and fruiting
remain controversial and, depending on the experiment, orienting entire
trees or individual branches horizontally or downward either increases
(Tromp 1970; Wareing 1970) or does not have a consistent effect (Long-
man et al. 1965; Mullins 1965) on flower bud formation and fruiting. It
has been shown in apple that both the time and genotype influence the
branch response to bending (Fig. 1.16) (Lauri and Lespinasse 2001).
A. Initial Choices of the Grower and Young Tree Training
Reducing the amount of vegetative growth discarded by pruning should
be a main objective of training procedures, as shown in both apple and
pear (Forshey et al. 1992). Especially during the early stages of tree

development, training a tree to obtain a shape different from its natural
growth habit may delay initial fruit production and requires consider-
able care. Attention may also be needed to maintain the framework of
the new allotted shape (Preston 1974). Training the tree with less prun-
ing and taking into account the natural growth and fruiting habit of the
tree (Lespinasse 1977, 1980; Forshey et al. 1992) or vine (Possingham
1994) is of major importance and may lead to higher yield performance.
The consequences of initial management choices on the young tree
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0.1
1.0
10.0
0.1
1.0
10.0
0.1
1.0
10.0
0.1
1.0
10.0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
A X.3318 1-year-old wood B X.3318 2-year-old wood
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
D 'Chantecler' 2-year-old wood
C 'Chantecler' 1-year-old wood
Logarithmic
mean
of
length
(cm)
Relative position
23 June
11 July
4 Aug.
22 Jan.
Control
Fig. 1.16. Effects of time of bending and age of wood on logarithmic mean of the length
of laterals (vegetative and bourse-shoots) relative to the position on the shoot (0 = basal to
1 = subapical) for genotypes (A, B) X.3318 and (C, D) ‘Chantecler’. Only positions with at
least 15 laterals were considered. Vertical bars represent ±1 SE when larger than symbol
size (Source: Lauri and Lespinasse 2001).

construction is discussed through two examples dealing with the choice
of the rootstock and training system.
1. Rootstock Effects. Among the numerous initial decisions faced by
growers when establishing an orchard, that of rootstock is crucial. The
grafting of scion cultivars on various selected rootstocks allows the
grower to increase orchard density and tree productivity (for historical
points of view see Fallahi et al. 2002). Indeed, dwarfing rootstocks
reduce the whole tree volume and promote earlier flowering (Lockard
and Schneider 1981; Larsen et al. 1992; Barritt et al. 1995).
A wide range of rootstocks that promote various tree volumes is avail-
able in many fruit species, though not in all (Webster 1997). Important
efforts have been devoted to comparing rootstock/scion performance in
different climatic areas. From the 1980s, several national programs involv-
ing evaluation of different apple rootstocks, cultivars, training systems,
and planting sites have been established in the USA, such as NC-140
(www.nc140.org; Fernandez et al. 1991; Perry and Fernandez 1993; Bar-
ritt et al. 1995; Marini et al. 2001; Robinson 2003), as well as in northern
Europe (Callesen 1997), and in New Zealand (White and Tustin 2000).
Detailed interpretations of dwarfing rootstock effects on the develop-
ment of the aerial part of the trees have been made. They addressed two
main questions: (1) is the reduction of aerial growth due to a delay of leaf
emergence rate or to a shorter period of growth? and (2) what variables
are involved in the reduction of shoot length and in the structural changes
of branching systems? The first question has been addressed in apple
(Costes and Lauri 1995) and peach cultivars (Weibel et al. 2003) grafted
on different rootstocks. In both studies, the length of the growing period
was shown to be reduced by dwarfing rootstocks. It also has been demon-
strated that rootstocks reduce the internode length (Seleznyova et al.
2003; Weibel et al. 2003). However, different results have been obtained
regarding the effect of dwarfing rootstocks on the mean number of nodes
per shoot. In peach, Weibel et al. (2003) indicated that differences in
shoot length were related primarily to internode length rather than to the
number of nodes, whereas Seleznyova et al. (2003) attributed the differ-
ence in apple branch size to a reduction in both the length of internodes
and the number of nodes that are neoformed within long growth units.
Average internode length per extension unit depends on unit node num-
ber, with internodes being shorter for units with fewer nodes.
Rootstock not only affects shoot length and number of nodes but also
branching density and location, and branch characteristics. On cherry,
Schaumberg and Gruppe (1985) showed that rootstocks from the Giessen
series altered the number of flowers per bud but not the number of buds

per spur of ‘Hedelfinger’ sweet cherry cultivar. More recently Maguylo
et al. (2004) showed that the effects of the rootstock on growth and flow-
ering of ‘Hedelfinger’ may be split into two components, vigor and geno-
type, with an increase in the number of spurs and the number of flowers
per spur as vigor increases in the Giessen series, and a decrease in both
variables as vigor increases in the other rootstocks. The effect of root-
stock on branching pattern was also studied in apple relative to differ-
ent axis types, with annual shoots sampled on four- to nine-year-old
trees (Hirst and Ferree 1995a), three-year-old fruiting branches (Selez-
nyova et al. 2003), and along six-year-old trunks (Costes et al. 2001b).
In all these situations, the percentage of budbreak of axillary buds on
extension growth units was unaffected, regardless of the rootstock. Thus,
differences in the number of axillary annual shoots per branch were
shown to result mainly from that of the number of nodes developed dur-
ing the previous year (Costes et al. 2001b). This led to the interpretation
that the effect of rootstock on aerial growth is cumulative and superim-
posed year after year. The changes induced in branching patterns,
including both the floral and vegetative development of the axillary
shoots, are currently being analyzed, such as applying Hidden-Semi
Markov chain models to assess the structural differences induced by a
range of rootstock/interstock combinations (Seleznyova et al. 2004).
Despite the interest in using dwarfing rootstocks to control tree vol-
ume and height, counterproductive effects have also been noticed in the
case of excessive dwarfing effects, for instance in cherry (Webster and
Lucas 1997; Moreno et al. 2001), peach (Layne et al. 1976; Bussi et al.
1995; Bussi et al. 2002), and apple (Marini et al. 2002). Due to cultivar-
rootstock interactions, the highest yields at the tree scale are usually not
obtained in the most dwarfed trees but in larger trees. A positive rela-
tionship between rootstock vigor and cumulative yield has been
observed (Warrington et al. 1990). Evidence of a positive correlation
between shoot growth and flower bud formation has also been provided
in apple, at least under specific conditions of water and nutrient sup-
ply (Decker and Hansen 1990). Thus, the choice of the growth level,
through the use of rootstocks of various vigor, has to be considered with
other variables, such as the intrinsic flowering pattern of the scion, to
obtain long-term tree efficiency.
The possibility of reducing initial orchard investment through less
expensive plant material has been explored through the use of micro-
propagated, own-rooted trees in Malus (Webster et al. 1985; Quamme
and Brownlee 1993) and Prunus (Quamme and Brownlee 1993). In
Pyrus, it has been argued that this procedure would be interesting to
avoid “pear decline,” often related to grafting, or more generally to the

graft incompatibility phenomenon (Thibault and Hermann 1982), even
though this last argument is no longer applicable to OH × F rootstock
selections. As a general trend, self-rooted trees are similar in size to trees
grafted on a semi-vigorous or vigorous rootstock, with similar delays for
beginning fruit production (Hirst and Ferree 1995b). However, great dif-
ferences exist among scion genotypes. Some own-rooted apple cultivars
such as ‘Greensleeves’ have precocious cropping efficiency higher than
that of ‘Greensleeves’ on MM.106, although not as high as that of
‘Greensleeves’ on M9 (Webster et al. 1985). For other cultivars (e.g.,
‘Cox’s Orange Pippin’), growing own-rooted trees was much less satis-
factory (Webster et al. 1985). It is likely that the branching and flower-
ing patterns have to be taken into consideration when evaluating the
potential for growing trees on their own roots. A recent study (Maguylo
and Lauri 2004) showed that own-rooted apple genotypes belonging to
fruiting type IV with downward-oriented branches and a high frequency
of bourse-over-bourse may reach similar or higher cumulated yields
than grafted trees in the fourth year of growth. For these genotypes,
strong vegetative growth before the first crop significantly reduces
branch breakage due to overloading and wind, as compared to trees
grafted on M9. It thus makes it possible to grow self-supporting trees
needing only a minimal support, with a large root system that ensures
good anchorage and possibly enhances water and mineral uptake.
2. Tree Development and Initial Fruit Production. The choice of the
training system usually is made at the early stage of orchard planting and,
in some cases, has early implications in the nursery, in particular through
the selection or pruning of long shoots along the trunk. The manipula-
tions to be carried out usually are described step by step via training
schemes (Fig. 1.17). Major changes in training systems were introduced
in the 1970s when physiological and architectural rules of tree develop-
ment were integrated progressively to training concepts. Indeed, the
training systems proposed at that time strongly differed from older ones
that were mainly influenced by esthetic considerations (Loreti and Pisani
1990). Due to the high number of systems proposed, many studies and
discussions have been devoted to the comparison of their relative bene-
fits. According to Robinson et al. (1991b), training systems can be roughly
divided into two categories according to whether they (1) apply the nat-
ural shape of the trees, such as multiple and central leader (Barritt and
Dilley 1989), vertical axes (Lespinasse and Delort 1986), and slender
spindle forms (Wertheim 1985), or (2) restrict the canopy in a geometric
form such as the A, V, or T forms (van den Ende and Kenez 1985; Lakso
et al. 1989). Some of these tree training systems have been conceived as

technical strategies to ease harvest or pruning mechanization, e.g., Tatura
or MIA. The early development of the canopy, the initial fruit production
and the resulting yield efficiency have been compared for many training
systems and fruit crops, including apple (Robinson et al. 1991a; Barden
and Marini 1998), cherry (Kappel and Quamme 1993), plum (Wustenberg
and Keulemans 1997), and peach (Myers 1994). However, it is difficult
to make unbiased comparisons because climatic and horticultural con-
ditions usually differ from one experimental site to another (Tustin et al.
1997). It is also difficult to take into account the economic efficiency of
these systems (DeJong et al. 1999) without simulating the weight of dif-
ferent inputs (cost of trees, labor) or outputs (fruit prices). Moreover, the
genotypic variability of tree habit makes comparisons between training
systems difficult, since genotypes may react differently (Bassi et al. 1994;
Lauri and Lespinasse 2000).
An important goal of young tree manipulations is to reduce the length
of the unproductive period for obvious economic reasons. The counter-
productive effect of heading the central leader on the early production
of the trees has been emphasized by different authors (Lespinasse 1977,
1980; Barden and Marini 1998; Lauri and Lespinasse 2000). By contrast,
taking advantage of the natural branching habit and promoting feather-
ing has been demonstrated to be a possible strategy to reduce the dura-
tion of the unproductive period. This arose from studies that were
Fig. 1.17. The Solaxe apple training system during the first four years (Source: Lauri, P. E.
and Lespinasse, J. M. 1998. The Vertical Axis and Solaxe Systems in France, Acta Hort.
(ISHS) 513:287–296.) http://www.actahort.org/books/513/513_34.htm.

carried out either during the juvenile period of seedlings, in the context
of breeding programs, especially in apple and pear trees (Visser 1965;
Visser and De Vries 1970; Zimmerman 1972; Zimmerman 1977), or in
young grafted trees (Maggs 1960). As mentioned previously, in the con-
text of commercial fruit production, the development of sylleptic shoots,
or feathering, can be promoted by the application of different chemicals
(Preston 1968; Quinlan 1978; Wertheim 1978; Miller 1988). The evi-
dence of a relationship between feathering and earlier initial cropping
led to the development of “preformed” nursery trees. In pear (Costes et
al. 2004), the number of long sylleptic shoots has been correlated with
the number of inflorescences developed per tree in the third year of
growth, based on a set of cultivars with different habits and durations
of the unproductive period. The most accurate predictive variable of ini-
tial flower formation was the difference in the number of sylleptic lat-
erals during the first two years of growth. This suggests that management
of young trees should take advantage of the genotypic differences in
sylleptic laterals during the first years of growth in order to reduce the
length of the unproductive period. The combination of a supported cen-
tral leader having a selection of long sylleptic shoots along the trunk rep-
resents a key step towards training systems with early efficiency and low
labor cost.
B. Adult Tree Training
Once the tree structure is established, the main focus of all training sys-
tems is to annually balance the fruit number and weight with vegetative
growth. In many perennial crops, an excess of fruit at the expense of veg-
etative growth may lead to irregular cropping, alternating between large
and small crops in consecutive years. Thus, training methods at the tree
scale (pruning, bending) aim at directing vegetative growth towards
fruiting sinks through precocious growth cessation that optimizes the
carbon budget of the tree with regard to fruiting (Sansavini and Corelli-
Grappadelli 1992; Lauri and Kelner 2001) and reduces heterogeneity
between shoots (Lespinasse 1996). The balance between the vegetative
growth and the flowering/fruiting components also refers to the concept
of crop load, which has been recently reviewed by Bound (2001) and
Wünsche and Ferguson (2005). Since there is not a unique definition of
crop load, various expressions have been proposed: ratio of fruit buds
to vegetative buds or number of buds per meter of frame-wood in pear
(Helsen and Deckers 1984), number of leaves per fruit in peach (Ben
Mimoun et al. 1998), or number of fruit per canopy volume in citrus
(Bound 2001). An easy-to-use method for expressing crop load is to
consider it the ratio of the number of fruits (or alternatively weight of

fruits) per trunk cross-sectional area (TCSA) (Abbott and Adam 1978) or
even branch cross-sectional area (BCSA) (Abbott and Adam 1978; Lauri
et al. 2004b). Wünsche and Ferguson (2005) here prefer the term of
“yield efficiency,” and discuss its validity in the context of tree aging.
Nevertheless, large variations may be observed in the relationships
between the intensity of alternate bearing and values of TCSA, as shown
in apple (Goldschmidtreischel 1996), suggesting that TCSA alone is not
sufficient and other parameters should be considered, such as canopy
spur leaf area (Sansavini and Corelli-Grappadelli 1992; Wünsche et al.
1996; Lakso et al. 1999).
1. Fruit Load Effects on Tree Growth, Architecture, and Ecophysiology.
In what follows, fruiting will be considered through its strong effects on
individual tree growth and architecture: (1) fruit load modifies the par-
titioning of available carbohydrates and water economy in a short term,
i.e., during annual cycle, and (2) heavy yields affect tree vigor in a longer
term by reducing cumulated growth over years (Regnard et al. 2002) and
the fruiting potential possibly inducing an alternate bearing. Reaching
an equilibrium between both growth and fruiting is thus one of the main
objectives of the fruit grower, as noted by Forshey and Elfving (1989).
When trees are young, newly formed biomass is allocated preferen-
tially to growth and directed towards the scaffold establishment and the
development of the root system. Biomass investment in fruit progres-
sively increases as the tree ages (Cannell 1985). When the orchard
reaches its adult phase, and provided that environmental and cultural
practices are at the optimum, the fruit yield compared to total annual
biomass increment—namely the harvest index—can reach seventy per
cent or more in peach (Cannell 1985) or apple (Lakso 1994). In any case,
biomass acquisition and utilization should be considered at the whole
plant level in terms of functional balance, which requires modeling
approaches (e.g., Cannell and Dewar 1994; DeJong and Grossman 1994).
Considering the large numbers of flowers or inflorescences that a
fruit tree usually bears, regulation of fruit load is needed. This implies:
(1) knowledge of the normal rates of fruit set that are compatible with
vegetative growth equilibrium, (2) assessment of the fruit set ratio after
bloom, (3) a comparison of fruit/leaf ratios, and (4) use of efficient and
low-cost practices to thin excess fruitlets. Branch pruning combined
with thinning are, in fact, the one key control strategy for regulating fruit
load.
When trees are overloaded, fruits act as major sinks for carbohydrates
(Ho 1988) and divert a major proportion of photoassimilates. This can
be detrimental to primary and secondary vegetative growth of the
aerial system, and can starve the root system (Lenz 2001). Numerous

studies have demonstrated that altering growth can result in modifica-
tions of tree architecture. In peach, a reduction of primary shoot growth
is observed in stems with subtending fruit compared to shoots with no
fruit (Berman and DeJong 1997). Furthermore, shoot length and weight
decrease exponentially as the crop load increases, while the relative gain
in trunk girth decreases linearly (Blanco et al. 1995). In apricot, high fruit
loads were shown to affect primary growth rhythmicity and growth
resumption during the growing season, and to enhance the acrotonic pat-
tern of branching (Costes et al. 2000). However the long-term effects of
the spatial distribution of fruits within the canopy on tree architecture
have not specifically been investigated. In three- and four-year-old apple
trees, heavy fruit loads led to higher allocation of biomass to fruit, and
lower allocation to new shoot growth, which resulted in reduced thick-
ening of branches and less root growth (Palmer 1992). Experimental
defruiting of young apple trees can result in higher leaf weight on area
basis, longer shoots, and greater increases in TCSA compared to normal
fruiting (Erf and Proctor 1987), while the presence of fruits conversely
reduces leaf and root dry weight up to 45% (Buwalda and Lenz 1992).
In sweet cherry, current-season growth appears to be a greater sink for
photosynthates than fruits, but fruiting is recognized as a factor that
reduces shoot growth (Kappel 1991). In highbush blueberry, Maust et al.
(1999b) demonstrated that high flower bud density decreased vegetative
budbreak, new shoot dry weight, leaf area, and leaf area to fruit ratios.
Excessive crops can lead to biennial bearing. A classic study carried
out in seeded vs. seedless apples by Chan and Cain (1967) demonstrated
the specific role played by seeds in inhibiting floral initiation. The effect
of seeds, which peaks from 6 to 10 weeks after full bloom, is generally
attributed to gibberellin synthesis, which is assumed to counteract the
floral process within apical buds of brachyblasts (Crabbé 1987; Crabbé
and Escobedo-Alvarez 1991). Excessive fruit load during “on” years
also lowers the amount of vegetative growth and consequently potential
photosynthesis and C-assimilate supply by sources to sinks, ultimately
resulting in decreased floral initiation during the growing season. More-
over, it has been frequently noticed that during the following year, when
the return bloom is low (“off year”), the quality of flower buds and the
effective pollination period (Williams 1965) are often reduced. Combi-
nation of both phenomenons strongly reduces fruit yield. The biennial
cycle is then auto-reproducible, high yield and poor growth alternating
with low yield and vigorous growth. The propensity for biennial bear-
ing is particularly important in apple (type I and II), pear, plum, olive,
and citrus, although there are important differences among cultivars. A
thorough analysis of the strong alternate bearing of types I and II apple
cultivars was developed by Lauri et al. (1995, 1997a).

Source-sink relationships and particularly carbon budgets have
received considerable attention in fruit trees (Blanke and Notton 1992;
Wibbe et al. 1993; Wibbe and Blanke 1995), leading to the development
of carbon balance models, e.g., in apple (Johnson and Lakso 1986a;
1986b) and peach (Grossman and DeJong 1994). Distribution of C-
assimilates is subject to complex mechanisms involving many aspects,
including the distance between sources and sinks as described in both
kiwifruit (Lai et al. 1989) and apple (Palmer et al. 1991), the capacity of
the translocation system as noted in peach (DeJong and Grossman 1995)
and more generally sink strength variations (Ho 1988). Moreover, the
stimulation of the assimilation process by the fruit itself has been
demonstrated, e.g., in peach (DeJong 1986; Bruchou and Génard 1999)
and apple (Gucci et al. 1995; Giuliani et al. 1997; Palmer et al. 1997;
Wünsche et al. 2000; Untiedt and Blanke 2001). Relations between crop
load and water economy of the tree also must be considered, for it has
frequently been shown that high fruit load results in modifying the
trade-off between carbon assimilation and transpiration, as the tree
meets the increasing sink demand by increasing the assimilation rates
and concomitantly transpiring water at a higher rate. Conversely, defruit-
ing or harvesting trees produce a decrease in sink demand that suddenly
results in lower assimilation rates and secondarily higher water use effi-
ciency (Chen and Lenz 1997; Pretorius and Wand 2003).
Although fruit nitrogen demand is not generally large, except in nut
crops, high fruit loads can deplete nitrogen availability during early
shoot growth. This in turn can limit the vegetative growth extension rate
(Forshey 1982; Sadowski et al. 1995).
High crop loads also result in competition between fruits, which
affects their development and final size (Denne 1960; Goffinet et al.
1995; Maust et al. 1999a) and generally lowers their quality (Kelner et
al. 2000; Link 2000; Wünsche et al. 2000; Bound 2001; Wünsche and Fer-
guson 2005). Comparing severely thinned vs. unthinned peach trees,
Grossman and DeJong (1995) suggested that heavy flower suppression
could give a fair estimate of the potential relative growth rate of the
remaining fruits, while fruit growth was source-limited on unthinned
trees. A similar approach led Lakso et al. (1995) to develop an expolin-
ear model for apple fruit potential growth under non resource-limited
conditions.
2. Consequences for Fruit Thinning. In response to the necessity for fruit
load regulation, biennial bearing avoidance, and fruit quality improve-
ment, thinning methods have received considerable attention in recent
years, as attested by successive reviews (Williams 1979; Miller 1988;
Dennis 2000; Bangerth et al. 2000). In apple, where chemical thinning

has been commonly applied since the 1950s, the suppression of excess
fruitlets is performed up to 30 days after full bloom. Thinning effects are
optimal during this period because fruit-to-fruit competition for pho-
toassimilates is limited and excess fruits abscise before the detrimental
effect of fruit on floral initiation for the next year can be observed.
Chemical thinning often operates by anticipating the June drop. Con-
sidering the differential genotypic sensitivity of different apple cultivars
to thinning agents and also the numerous factors that can affect tree
responses, the ultimate choice of a thinning program includes several
parameters that must be adjusted by the fruit grower: active ingredient(s),
concentration, wetting agents, spray volume per ha, and time of appli-
cation. Some specific objectives have to be kept in mind and achieved.
For example, lateral fruits borne on one-year-old shoots are generally
undesirable (and are thinned by chemicals) because their potential
growth is much lower than that of fruits borne in terminal positions
either on mesoblasts or auxiblasts (Jackson 1970; Lespinasse 1970; Mar-
guery and Sangwan 1993). As chemical thinning procedures to date
have not produced totally predictable results, the grower also must
decide whether additional manual fruit thinning must be performed. In
apple, the strategy of fruiting shoot removal, also called artificial extinc-
tion, has proven to be effective in reducing biennial bearing and fre-
quently in improving fruit color and size, and will probably receive
increasing attention in the future (see section III B.3).
3. Implementation of Adult Tree Training Procedures. There is interest
in developing a better knowledge of tree architecture, i.e., growth and
fruiting patterns, to develop training procedures adapted to commercial
tree fruit species and cultivars. Over the last four decades, the evolution
of cultivation of apple is a good example of how the knowledge of tree
architecture can be used to improve tree training. Indeed, pioneering
work of Bernhard (1961) and Lespinasse and co-workers (Lespinasse
1977, 1980; Lespinasse and Delort 1986) ranked apple cultivars more or
less linearly according to their fruiting type, from type I to IV. In relation
to this classification, three fruiting zones have been defined, each corre-
sponding to the successive stages of branch development over time
(Lespinasse 1977; Lauri and Lespinasse 2000). Moreover, in the apple
tree, terminal flowering greatly varied depending on the cultivar
(Lespinasse 1977) and this trait has great consequences on branch bend-
ing and consequently on the distribution of vegetative growth on parent
branches. The morphological expression of the apple tree architecture is
then highly dependent on the cultivar. Similarly, although acrotonic
branching is a common trait of all cultivars, the density of branches, espe-

cially short laterals in median and possibly proximal positions, is related
to regularity of bearing (Lauri et al. 1995). The aim of training is there-
fore to use the variability existing among cultivars in growth, branching,
and flowering location to optimize training methods. These studies have
led to the introduction of two main improvements in apple tree training
in France in the last decades. They involve the regulation of both branch
growth and crop load by controlling fruiting lateral density.
From Renewal Pruning to the Free Growing Fruiting Branch. Renewal
pruning was a cornerstone of the Vertical Axis training system pro-
posed by Lespinasse (1980). However, observations in commercial
orchards showed that regular heading back of the branch to develop new
shoots may lead to an imbalance between vegetative growth and fruit-
ing, especially under vigorous conditions, and also to an increase in the
proportion of fruit on one-year-old wood (Lauri and Lespinasse 2000).
An opposing concept has been proposed, involving the removal of com-
peting shoots on the upper-proximal part of the branch to invigorate dis-
tal fruiting organs. This new concept was integrated into the Solaxe
training system, without any heading of the trunk, and minimal prun-
ing and shoot bending of lateral branches (if necessary) to better control
branch growth and tree shape (Lespinasse 1996; Lespinasse and Lauri
1996; Lauri and Lespinasse 2000).
Crop Regulation via Artificial Extinction and Centrifugal Training. The
positive relationship between natural extinction of some laterals and the
increase of bourse-over-bourse trend of other laterals observed in regu-
lar bearing cultivars suggested that lateral density in a branch is in some
way physiologically related to the development of the other laterals
(growth and flowering frequency). From this result, it has been hypoth-
esized that “artificial extinction” practices, i.e., thinning out of young
fruiting laterals (Lauri et al. 1997b; Lauri and Lespinasse 2000; Lauri
2002; Lauri et al. 2004b), implemented in alternate bearing, usually
spur-bound, cultivars would improve sustainability of the remaining
fruiting laterals over the years (Lauri et al. 2004a).
Artificial extinction is carried out more specifically on the underside
as well as the proximal part of the fruiting branch, and around the ver-
tical trunk where shaded laterals have poor vegetative development and
low fruit set and size, color, and soluble solids (Tustin et al. 1988; Rom
1992). This procedure, called centrifugal training (Fig. 1.18), favors fruit-
ing in the peripheral zone of the canopy and significantly improves
light interception by fruiting shoots as well as canopy porosity (Willaume
et al. 2004). From a biological point of view, centrifugal training does not

act only as a fruit load-regulating technique, since some photosynthate
sources (leaves) and sinks (fruits) are removed at the same time. It is
therefore not fully comparable to chemical thinning, or hand thinning,
of flowers and fruitlets during which only generative organs are
removed, while all the leaves are kept.
Results on cultivars ‘Gala’ and ‘Braeburn’ showed that centrifugal
training improved and homogenized fruit size and return bloom as com-
pared to Vertical Axis or Solaxe systems (Larrive et al. 2000, 2001; Crété
Light well brought about
by centrifugal training:
extinction is carried
out along the trunk
and on the proximal
part of branches to
improve light
penetration within
the canopy
Extinction on the underside
of branches to increase
light penetration through
the canopy
Fruiting zone in the
upper three-quarters
of tree canopy
No branching below
1-1.2 m to permit
development of the
fruiting branches
Fig. 1.18. Centrifugal training concept to improve light distribution (gray arrows) in the
tree (Source: Lauri 2002).

et al. 2002; Ferré et al. 2002; Lauri et al. 2004a). A possible interpreta-
tion of these effects would be the moderate length increase of bourse-
shoots that are brought about by the decrease of branching density (Lauri
et al. 2004b). From a physiological point of view, this manipulation
would improve the autonomy of the fruiting shoot with regard to car-
bohydrate acquisition and allocation, leading to a higher return-bloom
potential (Lespinasse and Delort 1993).
Implementation of centrifugal training in commercial orchards is now
under development and evaluation in various places in France and in
other countries (e.g., Italy, Diemoz et al. 2002, Neri and Sansavini 2004;
Argentina, Rodriguez 2003) through the impetus of the Mafcot network
(Lauri et al. 1999). Artificial extinction and its development as a train-
ing procedure through the setting up of centrifugal training is one among
several manipulations (artificial bending, pruning methods, etc.) that the
grower may use in the orchard. It has been shown that the relevance of
centrifugal training depends on the cultivar. Although the “light well”
(see Fig. 1.18) is not necessary for ‘Granny Smith’ training, it is recom-
mended for colored cultivars such as ‘Pink Lady®’ (Hucbourg et al.
2004a, 2004b). These changes in training concepts already have practi-
cal consequences on the desired nursery tree structure: it is now rec-
ommended to plant trees that are unbranched up to 100–120 cm from
the ground (Fig. 1.18) in order to keep the branches growing down-
wards as perennial structures (Lauri 2002). Specific studies are now car-
ried out to end up in an overall “LITE planting system” (Lauri et al.
2004b) that optimally combines rootstock/cultivar material, planting
distances, and tree height.
This concept of crop regulation through detailed pruning of fruiting
shoots, minimizing pruning of structural wood, is now under develop-
ment in other species, in particular cherry (Claverie and Lauri 2005;
Lang et al. 2004a; Lauri and Claverie 2005).
IV. CONCLUSIONS
In order to constitute its own “orchard puzzle system,” each grower must
consider numerous possible choices. Other criteria than those detailed
above will of course be considered, particularly the socio-economic
context that is specific to each species and production area. Even though
these criteria were not addressed in this review, they may be of major
importance and complementary to those related to tree architecture.
Similarly, we did not detail all the agronomic practices that can be
performed in an orchard or at the tree scale. Our primary aim was to

Hort review 2006 (1)

Recommended

Recommended

More Related Content

What's hot

What's hot (19)

Similar to Hort review 2006 (1)

Similar to Hort review 2006 (1) (20)

More from Boonyong Chira

More from Boonyong Chira (16)

Recently uploaded

Recently uploaded (20)

Hort review 2006 (1)