coppice polar architecture

8/2/2019 Coppice Polar Architecture

1/18

Summary A multi-scale biometric methodology for de-scribing the architecture of fast-growing short-rotation woody

crops is used to describe 2-year-old poplar clones during thesecond rotation. To allow for expressions of genetic variabilityobserved within this species (i.e., growth potential, leaf mor-phology, coppice and canopy structure), the method has beenappliedtotwoclones:Ghoy(Gho)(Populus deltoides Bartr. exMarsh. Populus nigra L.) and Trichobel (Tri) (Populustrichocarpa Torr. & A. Gray Populus trichocarpa).

The method operates at the stool level and describes theplant as a collection of components (shoots and branches) de-scribed as a collection of metameric elements, themselves de-fined as a collection of elementary units (internode, petiole,leaf blade). Branching and connection between the plant units

(i.e., plant topology) and their spatial location, orientation,size and shape (i.e., plant geometry) describe the plant archi-tecture. The methodology has been used to describe the plantarchitecture of 15 selected stools per clone over a 5-month pe-riod. On individual stools, shoots have been selected fromthree classes (small, medium and large) spanning the diameterdistribution range. Using a multi-scale approach, empiricalallometric relationships were used to parameterizeelementaryunits of theplant,topological relationships andgeometry (e.g.,distribution of shoot diameters on stool, shoot attributes fromshoot diameter).

The empirical functions form the basis of the 3-D CoppicePoplar Canopy Architecture model (3-D CPCA), which recre-ates the architecture and canopy structure of fast-growingcoppice crops at the plot scale. Model outputs are assessedthrough visual and quantitative comparisons between actualphotographs of the coppice canopy and simulated images.Overall, results indicate a good predictive ability of the 3-DCPCA model.

Keywords: biometry, canopy structure, fish-eye lens, Populus

deltoides Populus nigra, Populus trichocarpa, seasonal dy-

namics, short-rotation woody crop.

Introduction

Populus spp. is one of the most promising fast-growing treespecies for producing fuel and fiber on short-rotation coppicesystems where crops are harvested every 3 years over a15-year lifespan (Grassi et al. 1990, Macpherson 1995,Ceulemanset al.1996).Large genotypicdifferences have beendemonstrated among natural and hybrid clones in productivity(Heilman and Stettler 1985, Pontailler et al. 1999, Laureysenset al. 2000, Tubby andArmstrong 2002), leaf area index (LAI)(DeBell et al. 1996, Heilman et al. 1996, Casella andCeulemans 2002) and other growth characteristics (e.g.,plantsoilnutrient interactions, site climate, pests and dis-eases) (Liu and Dickmann 1992a, 1992b, Stettler et al. 1996,Ceulemans and Deraedt 1999, Dickmann et al. 2001). Poplar

taxa differ considerably in canopy architecture and seasonaldynamics through shoot demography (Laureysens et al.2000),leaf phenology (Ceulemans et al. 1988, Deraedt andCeulemans1998), leaf andbranch morphology and spatial dis-tribution (Ceulemans et al. 1990, Dunlap et al. 1992, Casellaand Ceulemans 2002), and growth characteristics (Ridge et al.1986, Ceulemans et al. 1988, Pieters et al. 1999). In particular,productive hybrid poplar clones differ from less productivenatural ones in leaf size (maximum/average: 600/90 versus120/20 cm2 for Populus trichocarpa Torr. & A. Gray Populus deltoides Bartr. ex Marsh. hybrid clone Hoogvorstand P. nigra L. clone Wolterson, respectively), leaf inclinationangles (0 40 versus 0 80 on average for clones Hoogvorst

and Wolterson, respectively), and the distribution of thesefoliar characteristics within the canopy (Ceulemans andIsebrands 1996, Eckenwalder 2001, Casella and Ceulemans2002). Because plant productivity is directly related to theability to intercept photosynthetically active radiation (e i) (seeTable 1 for a list of symbols and their definitions) and convertit to biomass through photosynthesis (ec) (Cannell et al. 1988,Cannell 1989), leaf area density (LAD) andleaf attributes (i.e.,locations, orientations, sizes and shapes) (Takenaka 1994) arekey canopy parameters needed to describe the radiation re-gime within a canopy for simulating the mass and energy ex-

Tree Physiology 23, 11531170 2003 Heron PublishingVictoria, Canada

A method for describing the canopy architecture of coppice poplar

with allometric relationships

ERIC CASELLA1,2 and HERV SINOQUET3

1Forest Research, Mensuration, Alice Holt Lodge, Farnham GU10 4LH, U.K.

2Author to whom correspondence should be addressed ([email protected])

3UMR Physiologie Intgre de lArbre Fruitier et Forestier, INRA-Universit Blaise Pascal, Site de Croul, 234 avenue du Brzet, 63039

Clermont-Ferrand cedex 02, France

Received March 23, 2003; accepted May 24, 2003; published online November 3, 2003


2/18

change between a coppice crop and the atmosphere (Milne etal. 1992).

The models that have been developed to simulate mass andenergy exchange between plants and the atmosphere differmainly in their treatment of canopy architecture (Ross 1981,Chen 1992, Chen et al. 1994a, 1994b, Law et al. 2001,Sinoquet et al. 2001). Because of difficulties in estimating thespatialdistributionof leaf area withina canopy, therepresenta-tion of the vegetation has generally been simplified. In the tur-bid medium approach (Ross 1981), canopies have been

abstracted as a leaf gas and divided into horizontal layerswith uniform spatial distributions of the foliage (Law et al.2001, Casella and Ceulemans 2002). But these one-dimen-sional models are unsuitable for horizontal or vertically heter-ogeneous or discontinuous canopies with narrowly or widelyspaced plants and low leaf area in their earlier growth stages

(such as forest landscape or short-rotation woody crop(SRWC) poplar systems) (Law et al. 2001, Casella andCeulemans 2002). For these situations, canopy architecturecan be described (i) as a collection of individual crowns mod-eled as 3-D geometric shapes (Law et al. 2001), (ii) by model-ing the 3-D architecture of a population of plants usingstochastic (De Reffye et al. 1988, 1991, Whitehead et al.1990), fractal (Chen et al. 1994a) or L-system (Prusinkiewicz1986, Prusinkiewicz et al. 1994) methods, or (iii) by describ-ing accurately the geometry of each plant in situ, using the 3-Ddigitizing method (Sinoquet and Rivet 1997). The first ap-proach hasbeen successful in heterogeneous open-canopy for-est landscape systems (Law et al. 2001). Nevertheless, this

effort to consider the spatial heterogeneity of a canopy basedon crown envelopes remains fundamentally one-dimensionalat the crown or plot scales because the LAD within each fo-liage envelope is assumed to be uniformly distributed. The ap-proaches based on stochastic, fractal or L-system theoriesimproved the 3-D canopy architecture resolution by integrat-ing both topological (Hall et al. 1978) and geometric (Ross1981) notions of plant architecture. These methods can ap-proximate many species shapes for studying lightvegetationinteractions, but model outputs have never been preciselycompared with field measurements (e.g., canopy openness).The digitizing method provides a precise description of the3-D organization of every plant entity in space (i.e., from

internode to leaf). Unfortunately, this method cannot be ap-plied to poplar plants designed as SRWC systems because oftheir complex structure. To address this challenge, an alternateapproach was to conceptualize the underlying plant architec-ture (after Godin et al. 1999) and use it as a guide for modelconstruction and data acquisition.

The objectives of this study were to (i)propose a multi-scalebiometric methodology based on topological and geometricalinformation to deal with the 3-D architecture of fast-growingshort-rotation poplar crops, (ii) apply the methodology to twohybrid clones of different genetic origin: Ghoy (Gho, Populusdeltoides Populus nigra , characterized by a potential of pro-ductivity of about 4 Mg DM ha 1 year1) and Trichobel (Tri,

Populus trichocarpa Populus trichocarpa, characterizedby a high potential of productivity of about 11 Mg DM ha 1

year1), (iii) develop an empirical 3-D Coppice PoplarCanopyArchitecture model (3-D CPCA), and (iv) evaluate the modelquality from fish-eye photographs.

Materials and methods

Plant material and plantation layout

Poplar plants were measured during the second growing sea-son (2001) in a 0.5-ha experimental field in Farnham, U.K.

TREE PHYSIOLOGY VOLUME 23, 2003

1154 CASELLA AND SINOQUET

Table 1. Symbols and abbreviations.

Symbol Definition

A Axis. A1, an axis of order 1, is a shoot, andA2, an axis of order 2, is a branch

A Total leaf area per U2 (cm

2

)a Individual L area (cm2)ab Basal stool section area (cm

2)a10 Shoot section area at 10 cm above the point at

which the shoot joins the stool (cm2)a Inclination angle between A1 and A2 or A2

and P ()Cl The stool or the shoot diameter class numberdb Diameter at the bottom of a stool or shoot (cm)d10 Diameter at 10 cm above the point at which a

shoot joins a stool (cm)dt Diameter at the shoot tip (cm)DlI Internode elongation growth potentialDnL Leaf initiation rated Branch or petiole divergence angle ()

ec Canopy light conversion efficiencyei Canopy light interception efficiencyI Internode. An internode is morphologically

represented as a cylinder, bounded at itsends by nodes

l Length (cm)L Leaf bladeLAD Leaf area density (m2 m3)LAI Leaf area indexM Metameric unitN Node. A node consists of a possibly vegetative

bud, leaf or branchn Number ofw Random deviationP Petiole. A petiole is the slender stalk that

supports the leaf bladej Elevation angle of the normal to the vertical

for I, Se and L ()y Twist angle of the normal to the vertical for L

()S StoolSe 20-cm long segment on A1U1q Azimuth angle of the normal ()U Growth unit. U1 is the part of an axis that has

lengthened during the previous growingseason, U2 is the growing part of an axisduring the current year

w Width (cm)


3/18

(5111 N, 051 E; 115 m a.s.l.). The plantation is situated ona stagnogley soil surrounded by forest, and consists of 16 dif-ferent poplar (Populus spp.) clones belonging to different par-entage and hybrid groups (Armstrong 1997). The clones Ghoand Tri were used in this study (Table 2).

A randomized block design with three replications for eachclone was used, following the protocol prescribed by the U.K.Forestry Commission (Armstrong 1997). Hardwood cuttings

(25 cm) were planted in April 1996 in double rows (orientednorthsouth) with alternating inter-row distances of 0.75 mand 1.5 m (essential for mechanical harvests), and a spacingwithin rows of 0.9 m, to give a planting density of about10,000 stools ha 1. Plots were 9 11.5 m and contained10 rows of 10 stools each. In January 1997, stools were cutback to a height of 5 cm to create a multi-shoot coppice. Foreach clone, measurements were limited to 36 stools in the cen-ter of one plot.

Descriptions of plant topology and geometry

Two-year-old poplar plants were decomposed into compo-

nents (stool, S; axis, A; growth unit, U; and metamer, M) andelementary units (internode, I; petiole, P; and leaf blade, L).On S, axes of order 1 (A1) were shoots connected to S, andaxes of order 2 (A2) were branches connected to A1 (Fig-ure 1). Growth unit 1 (U1) was the part of A1 that had length-ened during the previous growing season (i.e., 2000), andgrowth unit 2 (U2) was the growing part (or shoot leader) ofA1 during the assessed year (i.e., 2001) (Figure 1). Becausefew lateral sylleptic A2 had lengthened on A1U1 during theprevious growing season (i.e., A2U1), only proleptic A2 weredescribed on A1U1 during the 2001 growing season (i.e.,A2U2) (Figure 1). Each U was described as a succession of M,themselves defined as I bounded at their ends by nodes (N).

Each N consisted of a vegetative bud, an A2U2 connected onA1U1,oranaxillarybud+P+LconnectedonU2(Figure1).With this plant topology definition, 3-D plant architecture

can be computer-generated if each elementary unit is given ashape, size, orientation and location in space. Basic geometricmodels (e.g., triangle, cylinder, frustum of a cone) are used torepresent the shape (i.e., external surface) and size of plantunits (e.g., the height of a triangle, l, and the width of its base,w). Detailed examples are available for tree architecture(Sinoquet and Rivet 1997, Godin et al. 1999), leaf geometryand canopy structure (Chen 1992, Sinoquet et al. 1998,Rakocevic et al. 2000, Dauzat et al. 2001, Sonohat et al. 2002).

In our study, S and P were represented as cylinders, U weredivided into a sequence of conic frustums (i.e., I), and L wereregarded as planar objects. A leaf blade prototype was createdforeach clone.Prototypes were represented as polygons with aset of four contiguous triangles for Clone Gho (Figure 1) andeight for Clone Tri, to fit the leaf blade shape and the allo-metric relationships between the leaf blade area (aL), the leafmidrib length (lL) and the leaf blade width (wL) as:

a C l wL L L= (1)

where Cis a regression coefficient.Finally, the representation of a basic geometric model in the

scene refers to its position (i.e., orientation and location) withrespect to a coordinate system (Figure 1). For L, its position inspace is defined by the direction of the axis of symmetry of thepolygon, or the leaf midrib, and the coordinates of the point ofattachment of L to P (see Figure 1 for more details).

Model description

The model operates at the stool level and describes the plantarchitecture as a collection of A1. Each A1 is in turn describedas a succession of components and elementary units definedby their size, shape, location, orientation and connection inspace, which can change over the course of the growing sea-son. From an input data file describing the locations (x and y)and basal diameters (db) of stools in a vegetative scene, themodeling process starts by generating individual plants (i.e.,stool + alive A1U1) as they were before bud burstin the springof 2001. Each plant is then used as an initial fixed geometricalarchitecture (in terms ofdbS; nA1U1, the number of A1U1connected to a stool; lA1U1, the A1U1 length; nM, the num-

ber of metamers of a A1U1; and qA1U1, the A1U1 azimuth)for the scene reconstruction over the growing season (by de-scribing the size, shape and position of each component car-ried by each M on each U). At the U and the leaf scales, themortality wasassumed to be negligible during theentire grow-ing season.

Initial plant architecture reconstruction The number ofA1U1 connected to S is calculated from dbS as:

n f dA U S)b1 1 = ( (2)

TREE PHYSIOLOGY ONLINE at http://heronpublishing.com

CANOPY ARCHITECTURE OF COPPICE POPLAR 1155

Table 2. The two Populus clones used in this study.

Name Code number Sex Parentage Parental Provenance Latitude Longitudecode number

Ghoy S.682-68 F P. deltoides S.9-2 Iowa, USA /Ontario, Canada 42 49 N P. nigra Ghoy Wallonnie, Belgium 51 N 4 E

Trichobel S.724-101 M P. trichocarpa V.235 Washington, USA 49 N 12230 WP. trichocarpa V.24 Oregon, USA 4530 N 12240 W


4/18

With the A1U1 numbered arbitrarily, shoot diameter values(d10A1U1) are calculated from shoot section area at 10-cmheight according to a relative shoot rank (i/nA1U1) as:

da

fi

ni10 1 1 2 1 1

A US

A Ub=

p(3)

where abS isthe basal stool section area and i =1,, nA1U1isthe A1U1 number. From d10A1U1, lA1U1 and nM are calcu-lated as:

l f dA U A U1 1 1 110= ( ) (4)

n f lM A U= ( )1 1 (5)

After numbering the M from the bottom to the top of A1U1,the elementary I units are described by their l, basal (db) andtop (dt) diameter, according to the relative M rank (i/nM) andthe relative M position ( l lj

iI A U

1

11 1

- or l lji I A U1 1 1 ) onA1U1, as:

l l fi

niI A U

M

=

1 1 (6)

d f dl

li

j

i

b I A UI

A U=

-10

1

1

1 11 1

, (7)

d f dl

li

j

i

t I A UI

A U=

10

11 11 1

, (8)

where i = 1, , nM is the M number.Additionally, the curved shape of A1U1 is described by

the evolution of the elevation angle (j) of every consecutive

20-cm-long segment (Se) along A1U1 (from plagiotropic Seat the bottom to orthotropic ones at the top of A1U1) as a func-tion oflA1U1:

jSe A Ui f l= ( )1 1 (9)

where i = 1, , nSe is the Se number and nSe is the total num-ber of Se along A1U1.

From Equation 9, each consecutive I on A1U1 takes a jvalue (jI) given by the j value of its corresponding Se (jSe),as:

j j

j j j

I Se

Se

rand Se3

4Se

1

4

i k

i

k k

i

i i

k

=

==

+

1

1 1

1,

; Se ... , SeSen ik n

=, ,2

(10)

where i = 1, , nM is the M number, k li ji= + I1 20 1 is

the corresponding Se number, x is the integer part ofx, andrand[a, b] is a random number between a and b.

Finally, A1U1 takes place in the scene after assigningqA1U1 a random number between 0 and 360.



Figure 1. Multi-scale codification of a poplar plant topology and de-scription of Euler angles (q, azimuth; d, divergence; j, elevation; a,inclination; and y, twist) for assessing and computer-generating thethree-dimensionalplant architecture of 2-year-old poplar within high-density coppice crops. Plant components and elementary units arestool (S), internode (I), petiole (P)and leaf blade (L). At a given scale,elementary units (e.g., I) of a plant component (e.g., axis of order 1growth unit 1, A1U1) are numbered from the bottom to the top of thecomponent(e.g., from thepoint of attachment of A1U1 on S to itstip).The diagram presented at the bottom of the figure details the geomet-ric3-D reconstructionof L ina scene,withrespect to a coordinate sys-tem. The compositepolygon(i.e., the set of fourcontiguous triangles)is a basic planar geometric model representing the shape and the size

of L, where lL is the leaf midrib length and wL the leaf blade width.The positionof L in space refers to its orientation and its position (co-ordinates of the point of attachment of L to P), with respect to thecoordinate system. The leaf midrib direction is given by three angles:q, a rotation aroundthez-axis; j, a rotation around they-axis; and y, arotation around the x-axis. Definitions of symbols are given in Table1.


5/18

May to September plant architecture reconstruction Fromthe initial geometrical plant architecture described above, andfor a selected date (e.g., May), lA1U2 is calculated for eachA1U1 forming the plant as:

l f lA U A U1 2 1 1= ( ) (11)

To take into account potential variations in the size and shapeof each A1U1 throughout the growing season (in terms ofdin-crements and jSe variations), d10A1U1 is recalculated as:

d f l l10 1 1 1 1 1 2A U A U A U= +( ) (12)

and then dbI, dtI and jI are recomputed from Equations 7, 8and 10, respectively.

At the A2U2 scale, the modeling process calculates the l ofevery A2U2 carried by each branched M on A1U1 as:

l fl

li

i

jA UI

A U2 2

1 1

1=

(13)

where i = 1, , nM 3 is the M number, and nM 3 is ex-plained by the presence of two consecutive vegetative budsandtheA1U2connectedtothelastthreeMatthetopofA1U1.At this scale, the branching rate along the entire branched por-tion of A1U1 is arbitrary fixed to 1.

Following Equations 5 and 6, A1U2 and A2U2 are, in theirturn, divided into consecutive I units, described by their l ac-cording to a relative M rank on A1U2 or A2U2. In contrastwith the geometrical description of A1U1, A1U2 and A2U2are represented in the vegetative scene as conic frustums of

height l (Equations 11 and 13, respectively), with db and dt val-ues given by:

d db tA U A U1 2 1 1= (14)

d f dt bA U A U2)1 2 1= ( (15)

d f lbA U A2U2)2 2 = ( (16)

d f dt bA U A U2)2 2 2= ( (17)

The final position and orientation of A1U2 and A2U2 areobtained after calculating q, j and a (insertion angle) valuesas:

q qA U A U1 2 1 1=

j jA U rand[ Se Se1 2 90= n , ]

qq d

A Urand[

A U A U M2 2

0 360 1

2 2 2 2 21i

i

i

i n=

=

+ =

-

, ],

, , ,K

aA U2 2 = c

where d is the A2U2 divergence angle and c is a constant.

At last, the A1U1 reconstruction is completed by describingthefoliage distributionalong each A1U2 andA2U2, at the leafscale. For each of them, total and individual leaf area values(AL and aL, respectively) are calculated according to lU2 andthe relative M rank on U2 as:

A f lL U= ( )2 (18)

a a fi

niL L M

=

(19)

where aL is the mean individual leaf area (= AL/nM) and i = 1,, nM is the M number. At this last scale, the leaf rate alongthe entire U2 was arbitrary fixed to 1, and temporal leafabscission on U2 was not taken into account.

For each individual leaf, lL, wL and petiole length (lP) val-ues are calculated as:

l f aL L)= ( (20)

w f lL L)= ( (21)

l f lP L)= ( (22)

The final position and orientation in the scene of L and P areobtained aftercalculatingq, a, j andtwist(y) angle valuesas:

qq d

Prand[

P P Mi i

i

i n=

=

+ =

-

0 360 1

21

, ],

, , ,K

q qL P=

aPMi

fi

n=

(23)

jLMi

fi

n=

(24)

yLMi

fi

n=

(25)

where i = 1, , nM is the M number and d is the P divergenceangle.

Finally, the plant structure is generated at metamer levelfrom dbS and Equations 125, parameterized from field mea-surements for a given clone and date. Model outputs are a col-lection of geometric objects.

Biometric measurements

Equations 225 were parameterized with multi-scale biomet-ric sampling. During the 2001 growing season, as the two pop-lar clones had only leafy A1U2 and A2U2 directly connectedto A1U1, plant topology and geometry were assessed (i) at theclone scale, (ii) at stool and A1U1 scales in March (before bud




6/18

burst), and (iii) at U2 (A1U2 and A2U2) and leaf (P and L)scales from May to September.

Initial plant architecture parameterization (Equations 210)

In mid-March, thedb of the 36 stoolswere measured above thegroundwithacaliper(dialMax-5921,SwissPrecision,Ashton,Switzerland)and the cumulatedbasal area wascalculated. Ad-

ditionally, nA1U1 was assessed (Equation 2). The cumulatedbasal area was divided into 10 (for baseline assessments) orfive (for detailed assessments) equal classes (Cl). For eachclass, themean basal stool diameter (dbS) was calculated and astool with a db close to the mean was selected and labeled fromthe 36 assessed stools. From these 15 selected and labeledstools, at the A1U1 scale, d10 (Equations 3, 4, 7 and 8) and l(Equations49and1113)weremeasuredoneachaliveA1U1with a tape and the caliper, respectively. Additionally, on S la-beledfor detailedassessments,q wasmeasuredwithacompass(KB-14/360 R, Suunto, Finland) for each A1U1. Using thesame selection method as above, 45 A1U1 were selected (30for baseline and 15 for detailed assessments) and labeled from

three size classes (1, small; 2, medium; and 3, large) spanningthe d10A1U1 distribution range of all labeled stools. Finally,nM was measured along each A1U1 connected on an S labeledfor detailed assessments (Equations 58), whereas lI (Equa-tions 68) and dt (Equations 7 and 8) were measured onlyalong labeled A1U1.

May to September plant architecture parameterization

(Equations 1125) Followingthesamemulti-scaleapproachas above,measurements were carriedout monthly fromMay toSeptember (during the first and last 12 days of each month forClones Tri and Gho, respectively). At the A1 scale, d10A1U1(Equation 12), dtA1U1 and dtA1U2 (Equations 14 and 15)were measured on all A1 connected on S labeled for detailedassessments.Additionally, the j value wasmeasured for everyconsecutive Se along A1U1 in May, July and September(Equations9and10).AttheU2scale, lA1U2(Equations11,12and 18) and lA2U2 (Equations 13, 16 and 18), nM (Equa-tion 58) and nL (the number of leaves) of every U2 weremeasured along each labeled A1U1. At the A2U2 scale q, a,db, dt (Equations 16 and 17, respectively), and lI (Equations68)weremeasuredononeselectedandlabeledA2U2inthreealongeach labeled detailed assessmentsA1U1.Finally, at theleaf scale, lL (Equations 2022), lP (Equation 22), qP, aP(Equation 23), jL (Equation 24) and yL (Equation 25) weremeasured at each leafy M on labeled U2 from May to Septem-ber.

Destructive sampling for leaf allometry

From the unselected experimental poplar plots, a set of leaveswas harvested separately for each clone during the entiregrowing season to parameterize Equations 1, 20 and 21.Leaves were sampled on A1U2 and A2U2 within the canopiesand lL, wL and aL were measured in the laboratory with a la-ser area meter (CI-203, CID, Camas, WA).

Hemispherical photographs

The hemispherical photography method was used routinely to

estimate seasonal dynamics of canopy openness. Photographswere taken during uniform overcast sky conditions with afish-eye lens (NikonAF Fisheye 16 mm f/2.8D, Tokyo, Japan)mounted on a manually operated camera (Nikon FM2, Tokyo,Japan). A total of 15 below-canopy pictures were made 15 cmabove thesoil surface(i.e.,top of thehemispherical lens) everyweek between May 1 and November 11. Five photographswere taken along the central row of the 36 assessed stools, and10 from its two adjacent inter-rows.

Empirical functions and random deviation

From the biometric measurements, empirical allometric rela-tionships were described with polynomial, asymmetric sig-moid or power functions. The regression model was chosenbased on the experimental data distribution, the best-fit valueof the correlation coefficient of the adjustment (r2) anda satis-factory random distribution of theresidues(i.e., pass theSIGNtest and/or the RUNS test). However, as an example, poplartrees are recognized to have a phyllotaxy angle (A2U2 angle

around the A1) with a divergence ratio of 0.4. But, ifdA2U2 isconstant (i.e., 144), a top-down view of the A1U1 will showonly the last fiveA2U2at its tip. To be more realistic, a randomdeviation w was applied to selected empirical relationships(Equations 2, 3, 11, 13 and 2225) and constants (dA2U2,aA2U2, dP and aP), and was calculated as:

w = -rand[ i i], (26)

where i is the standard deviation of a mean or the maximal orthe mean standard error of a constant pattern or of an empiricalfunction, respectively.

Hemispherical computer-generated images

Plant images were computer-generated from the model out-puts with the POV-Ray ray-tracing software program (Per-sistent of Vision Raytracer, Version 3.5). Each object orelement of the scene was scaled to the appropriate geometricdimensions (e.g., height l, base radius db/2 and top radius dt/2for a conic frustum). The object was then rotated and trans-lated according to its orientation and location in the scene(Figure 1). A specific texture was finally given to each object,describing its color. The resulting virtual coppice crop couldthen be looked at from any point of view, after having placed avirtual camera and light source in the scene. The light source(i.e., single and defined as direct sunbeams) was located a

large distance from the scene at the zenith. Simulated coppicecrop andhemispherical imageswere created with a virtual per-spective camera with a viewing angle of 80 and 180, respec-tively. Following the experimental hemispherical photographymethod, five virtual hemispherical images were simulatedalong the central row of the 36 stools and 10 from the two adja-cent inter-rows.

Hemispherical photographs and virtual hemisphericalimages were processed with HemiView (Delta-T Devices,Cambridge, U.K.), a computer program for analyzing hemi-spherical images. For both kinds of image, the sky-map wasconstructed by dividing the sky into an array of sky annuluses




7/18


8/18

defined by six zenith divisions (i.e., six annuli with 15 inter-vals).

Results

Clonal differences in model parameterization

In March 2001, dbS of Clone Gho was around 1.6 times

smaller than that of Clone Tri but with a similar mean numberof connected alive A1U1 (25 9 for Clone Gho and 22 6 forClone Tri). Compared with Clone Gho, a weak variation onnA1U1 versus dbS (Empirical function 2, Table 3) and largerdata ranges ofd10A1U1 (Empirical function 3, Table 3 andFigure 2) and lA1U1 (Empirical function 4, Table 3) werefound for Clone Tri. At the I level,however, the two clones hada similar data range ofnM as a function oflA1U1 (Empiricalfunction 5, Table 3). Additionally, taking into account theshoot diameter class, empirical relationships obtained be-tween the relative lI and its corresponding relative M rank onA1U1 (Empirical function 6, Table 3) showed similar data

rangesand data distribution shapes(i.e., trapezoidal sequentialpatterns) for the two clones (Figure 2). Internodes near theproximal and distal ends of every A1U1 were relatively short,but longer in their middle part (Figure 2), despite the presenceof a weak depletion related to unfavorable weather at the be-ginning of summer in 2000 (data not shown). Finally, bothclones had a large range in the curved shape of A1U1 accord-ing to its length (from plagiotropic to orthotropic shapes forthe shortest to longest A1U1) (Empirical function 9, Table 3

and Figure 2). But no correlation between the curved shape ofA1U1 and its position around the stool (in terms of locationand orientation, according to the double-row design of the ex-perimental plot) was found.

The growing season of Clone Tri began in early April, about26 days before that of Clone Gho. At the A1U1 scale, jSe val-ues of the shortest A1U1, which were between the rows, de-creased substantially during the first 2 months of the growingseason from May to September (Empirical function 9, Table 3and Figure 2). Additionally, for both clones, a common sub-model was used to describe diameter evolutions along every



Figure 2. Examples of empiricalallometric relationships obtained fromfield measurements: the shoot/stoolsection area ratio (a10A1U1i/abS) ver-

sus its relative shoot rank on S(i/nA1U1) (Equation 3, Table 3); therelative internode length (lIi/lAIU1)versus its relative metamer rank onA1U1 (i/nM) (Equation 6, Table 3);and the elevation angle value of everyconsecutive 20-cm long segment alongA1U1 (jSei) versus the shoot length(lA1U1) (Equation 9, Table 3) for pop-lar Clones Ghoy (Gho, left column)and Trichobel (Tri, right column). 59are month numbers (MaySeptember,respectively) and Cl = 13 the A1U1diameter class number (smalllarge, re-spectively). Definitions of symbols aregiven in Table 1.


9/18

A1U1 (d d a l l bi ji

I A U I A U= +10 11 1 1 1( ), where i = 1, ,nM is the M number, a = 0.871 0.008 and b = 1.029 0.004; n = 156, r2 = 0.998) (see Equations 7 and 8),despite theclones having different A1 size ranges (Empirical function

12, Table 3).At the U2 and leaf scales (Tables 4 and 5 and Figure 3),

strong contrasts were found between the two clones in theempirical allometric relationships observed. For both clones,lA1U2 and lA2U2 were always positively correlated to theshoot diameter class (Empirical functions 11 and 13, Table 4and Figure 3). But, in contrast with Clone Tri, Clone Ghoshowed a relatively constant lA1U2 for the largest A1U1 di-ameter class, highlighted by an asymmetric sigmoid functionfit between lA1U2 and lA1U1, compared with a power func-tion for Clone Tri (Empirical function 11, Table 4 and Fig-

ure 3). Similar regression models were used to describe theacropetal increase oflA2U2 on A1U1 for both clones (Empiri-cal function 13, Table 4 and Figure 3), although stronger datavariability was observed for Clone Gho (Figure 3). Addition-

ally, Clone Gho was characterized by the presence of un-branched M along the first half of its A1U1 (Empiricalfunction 13, Table 4 and Figure 3). However, both clones ex-hibited a branching rate of about 1 with common dA2U2values (141 13, n = 352) along the entire branched portionof every A1U1. Finally, aA2U2 had a range of 40 5 forClone Gho and 49 11 for Clone Tri.

Whereas a single relationship described nM versus lA1U2for the entire growing season, and for all A1U1 diameterclasses (Empirical function 5, Table 4), the description ofA1U2 and A2U2 entities (i.e., I) required relationships for



Table4. Empiricalfunctions at the current-year growth-unitscale (U2)for Clones Ghoy (Gho) and Trichobel (Tri) for the entire growingseason of2001. The relationshipsbetween Yand Xwere described with three regression models: (1) y = C0 + C1x + C2x

2; (2) y C x C C= +0 11 2/( ( / ) ); and (3)y C x C

C= +0 21 . Definitions of symbols are given in Table 1.

Scale Number Month Y [min, max] X [min, max] Cl Model C0 C1 C2 n r2

Clone Gho

A1U2 11 5 lI [0.2, 10] lA1U1 [54, 185] 13 2 8.892 92.53 2.547 43 0.7386 [0.2, 19] [54, 185] 13 2 16.89 79.11 3.966 41 0.8137 [0.2, 35] [54, 185] 13 2 31.69 86.49 4.360 40 0.81989 [0.2, 50] [54, 185] 13 2 41.87 87.57 3.977 83 0.767

A1U2 5 59 nM [6, 31] lA1U2 [0.2, 50] 13 3 5.173 0.442 189 0.962

A2U2 13 59 lA2U2i [0.2, 7]l

l

i

jI

A U1

1 1

[0.6, 1] 1 1 2.508 4.980 211 0.061

59 [0.2, 11] [0.5, 1] 2 1 14.62 42.13 22.410 875 0.1465 [0.2, 11] [0.4, 1] 3 1 19.58 64.70 39.34 214 0.3836 [0.2, 20] [0.4, 1] 3 1 18.56 57.15 26.96 215 0.4157 [0.2, 31] [0.4, 1] 3 1 18.61 54.05 18.87 191 0.36289 [0.2, 35] [0.4, 1] 3 1 15.04 39.43 4.530 376 0.374

A2U2 16 59 dbA2U2 [0.2, 0.5] lA2U2 [0.2, 45] 13 3 0.178 0.261 355 0.811

U2 15, 17 59 dtU2 [0.1, 0.2] dbU2 [0.2, 0.5] 13 1 0.370 372 0.338

Clone Tri

A1U2 11 5 lI [0.2, 9] lA1U1 [72, 301] 13 3 3.13E3 1.385 40 0.7386 [0.2, 34] [72, 301] 13 3 1.30E2 1.384 46 0.8137 [0.2, 72] [72, 301] 13 3 4.41E3 1.700 34 0.8198 [0.2, 101] [72, 301] 13 3 1.20E3 1.987 39 0.7679 [0.2, 145] [72, 301] 13 3 2.73E4 2.296 46 0.767

A1U2 5 59 nM [3, 32] lA1U2 [0.2, 145] 13 3 1.936 0.563 319 0.962

A2U2 13 59 lA2U2i [0.2, 2]l

l

i

jI

A U1

1 1

[0.2, 1] 1 3 1.357 2.281 610 0.061

5 [0.2, 7] [0.1, 1] 2 3 4.570 3.389 0.171 440 0.146[0.2, 13] [0.1, 1] 3 3 11.49 1.942 362 0.383

6 [0.2, 26] [0.1, 1] 2 3 25.83 7.360 0.306 457 0.415[0.2, 40] [0.1, 1] 3 3 38.56 2.784 0.155 353 0.36279 [0.2, 42] [0.1, 1] 2 3 38.53 9.258 0.469 1105 0.374

[0.2, 73] [0.1, 1] 3 3 67.78 3.911 1.294 404 0.36289 [0.2, 84] [0.1, 1] 3 3 75.86 4.232 1.559 767 0.362

A2U2 16 59 dbA2U2 [0.2, 0.9] lA2U2 [0.2, 73] 13 1 0.263 7.90E3 171 0.811U2 15, 17 59 dtU2 [0.1, 0.3] dbU2 [0.2, 0.9] 13 1 0.346 298 0.338


10/18

each A1U1 diameter class and observation date. Comparedwith the description of A1U1, lI and nM data ranges observedat theU2 scale were less comparable between theclones. Rela-tive lI ranges were smaller for Clone Gho than Clone Tri, sug-gesting a higher leaf initiation rate for Clone Gho.

Finally, at the leaf scale (Table 5), except for the distribution

ofjL along U2 (Empirical relationship 24, Table 5), A1U2and A2U2 characteristics (Table 5) were described with a sin-gle relationship for each clone. For both clones, AL of A1U2was larger than that of A2U2 and increased with increasinglA1U2 or lA2U2 (Empirical function 18, Table 5). The distri-bution ofaL along A1U2 (Figure 3) or A2U2 (Empirical func-tion 19, Table 5) showed similar data ranges and sequentialpatterns between the clones with the shape of the relationshipprogressing from parabolic in May to trapezoidal in Septem-ber. Additionally, the two clonesexhibited a common dP value(141 13, n = 352) along the entire leafy part of every U2.Compared with Clone Tri, leaves of Clone Gho were differentin many geometric characteristics: heart-shaped leaf blades

(wL = 0.809lL, r2

= 0.865; lL = 2043. aL, r2

= 0.912; andaL = 0.593lLwL, r2 = 0.988; n = 658) compared with lanceo-late ones for Clone Tri (wL = 0.582lL; l aL L= 2565. , r2 =0.976; and aL = 0.631lLwL, r2 = 0.996; n = 821), aL wasabout four times smaller, lP was 75% of lL versus 30% forClone Tri, and larger data ranges described jL (Empiricalfunction 24,Table 5) andyL,whichhadarangeof080(n =141) versus 0 20 (in May, n = 105) and 0 40 (July andSeptember, n = 386) for Clone Tri.

Evaluation of the virtual coppice crop reconstruction method

Throughout the entire growing season of 2001, model outputs

were compared with field measurements to test features rang-ing from the characteristics of individual components of aplant (Figure 4, Table 6) to the more complex structure of acanopy (Figures 57).

From May to September and for both clones, results re-ported in Figure 4 and Table 6 show that the modeling processrecreates the geometry of a plant (see also Figure 1). First, theseasonal time course of thesimulated total numberof potentialleafy M per A1 (i.e., M on U2 only) is strongly correlated withthe measured one (Figure 4). Second, as observed from the ex-perimental results, mean lA1U2 and lA2U2 derived frommodel outputs are positively correlated with the A1U1 diame-ter class (Table 6). Third, the clonal contrast in the acropetal

increase oflA2U2 on A1U1 is well recreated by the modelingprocess (Table 6). Finally, at the leaf scale, spatial and tempo-ral variations in AL are taken into account by the model (Fig-ure 1 and Table 6).

At the plot level, the quality of the spatial and temporal dis-tributions of every plant entity in a scene is shown by the syn-thesized images in Figure 5. Foreach clone, the reconstructionmethod seems to recreate the geometry of the canopy in accor-dance with the double-row design of the experiment. Clonaland temporal differences in coppice structure (e.g.,A1 geome-try) and canopymorphology (e.g.,L geometry), detailed in thepreceding paragraph, are clearly recreated (Figure 5) as the



Table5.Empiricalfunctionsattheleafscale(petiole,Pandleaflamina,Llevels)fo

rclonesGhoy(Gho)andTrichobel(Tri)fortheentiregrowingseasonof2001.TherelationshipsbetweenYandX

weredescribedwithtworegressionm

odels:(1)y=C0+C1x+C2x2+C3x3+C4x

4+C5x5+C6x6;and(2)y

C

xC

C

=

+

0

1

1

2

/(

(/

)

).DefinitionsofsymbolsaregiveninTable1

.

Number

Month

Y

[min,max]

X

[min,max]

Model

C0

C1

C2

C3

C4

C5

C6

n

r2

CloneGho

18

59

AL

[70,613]

lA1U2

[3,48]

1

71.48

4.901

0.128

84

0.968

18

59

AL

[2,265]

lA2U2

[0,35]

1

11.68

0.133

423

0.935

19

59

a

ai

L

L/

[0,2.68]

i/nA2U2

[0,1]

1

4.347

2.918

0.879

4126

0.500

22

59

lP

[0,6]

lL

[0,7]

2

6.310

4.093

1.920

386

0.774

23

59

aPi

[5,120]

i/nM

[0,1]

1

99.90

81.700

154

0.751

24

5,6,9

jLi

[85,80]

i/nM

[0,1]

1

55.87

79.30

141

0.302

CloneTri

18

59

AL

[200,5132]

lA1U2

[6,144]

1

21.86

21.14

0.107

92

0.990

18

59

AL

[5,1674]

lA2U2

[0,80]

1

52.19

4.314

0.168

2.93E3

2.43E5

7.76

E8

1311

0.997

19

59

a

ai

L

L/

[0,2.57]

i/nA2U2

[0,1]

1

0.767

2.026

3.027

0.986

5676

0.080

22

59

lP

[0,6]

lL

[0,22]

2

2.782

6.897

4.496

559

0.696

23

59

aPi

[10,80]

i/nM

[0,1]

1

66.14

443.400

1002.000

601.20

276

0.502

24

5

jLi

[85,0]

i/nM

[0,1]

1

5222

105

24

6,9

jLi

[70,90]

i/nM

[0,1]

1

23.72

29.12

386

0.090


11/18

images show the strong contrast in growth potential expressedby the two clones. To assess the spatial aL distribution, virtualhemispherical pictures were compared with real hemispheri-cal photographs (Figures 6 and 7, respectively). A visual in-spection (Figure 6) shows that A1, A2, leaf and canopy gapsizes, shapes and spatial distributions are correctly recreated

throughout the growing season for the two clones (Figure 6),implying that the modeling process recreates the spatial andtemporal aL distribution within the contrasting canopies.However, due to the restricted dimensions of the experimentalplots and the surrounding forest landscapes, a dense borderwas observed on the horizon of the real hemispherical photo-



Figure 3. Examples of empiricalallometric relationships obtained fromfield measurements: the shoot leaderlength (lA1U2) versus the shoot length(lA1U1); the branch length (lA2U2i,for the shoot diameter class 3 only) andthe individual/mean individual U2 leafblade area ratio (a aiL L/ ) versus itscorresponding relative metamer rank(i/nM) on A2U2 and A1U2, respec-tively, for poplar Clones Ghoy (Gho,left column) and Trichobel (Tri, rightcolumn). Numbers 59 indicate month(MaySeptember, respectively). Defi-

nitions of symbols are given in Table 1.

Figure 4. Measured and simulated totalnumber of leafy metamers, at the shootscale, over the 2001 growing season( = May, = June, = July, =August and = September). Cl = 13is the shoot diameter class number(smalllarge, respectively). Relation-

ships were fitted to the linear regres-sion y = ax + b. Dotted lines depict theconfidence interval at P < 0.05.


12/18

graphs (Figure 6). Taking this observation into account, theseasonal time course of the spatial aL distribution within eachcanopy was assessed by comparing canopy openness valuescalculated from the virtual and real hemispherical images foreachsky-map zenithdivision(Figure 7). A strong direct corre-lation was found between simulated and measured canopy

openness values for each clone for the first four zenith angles(between 0 and 60). But the border effect (exceptionallystrong between 75 and 95) was not recreated by the model be-cause virtual cropcanopies were limited to 100plants. By inte-grating these values over a restricted hemispherical viewingangle of 120, temporal coursesof themeasured andsimulated



Table 6. Basic structural plant and canopy parameters derived from the 3-D CPCA model outputs describing the poplar Clones Ghoy (Gho) andTrichobel (Tri). Results (MaySeptember) were generatedfrom a population of 36 virtual plants by the modelingoperation, linearlycorrectedbe-tween two consecutive months ( 6 days, taking into account the delay of 12 days between the assessment of the two clones), and displayed byshoot diameter class (Cl), relative branch position on A1U1 ( l lI A U 1 1) or within the canopies. Abbreviations: n = number of; l = length (cm);DlI = change in lI between May and September, the internode elongation growth potential; DnL = change in nL between May and September, theleaf initiation rate; and aL = leaf blade area (cm 2), characterizing an average shoot (A1U1), shoot leader (A1U2), branch (A2U2), metamer (M),internode (I) or leaf blade (L) of an individual or a population of plants. Definitions of symbols are given in Table 1. Bars over symbols denote

means.

Plant

A1U1 A1U2 A2U2

Cl n l nI lI l nI DlI aL l lI A U 1 1 n l nI DlI aL

Clone Gho

3 3 171 46 3.7 5.441.3 1126 2.4 9.319 0.751 11 5.718.9 817 1.5 7.9100.50.75 9 4.79.7 713 0.8 6.98

00.5 4 1.22.0 56 0.8 3.242 6 120 36 3.3 4.333.6 1024 2.1 9.516 0.751 8 3.85.4 710 0.5 5.66

0.50.75 8 1.72.5 57 0.4 3.9400.5 2 0.10.1 11 0.0 1.42

1 16 63 23 2.7 1.911.1 613 1.3 12.412 0.751 5 1.41.9 56 0.5 3.130.50.75 4 0.30.5 22 0.1 2.2300.5 0

13 25 89 29 3.1 2.919.9 817 1.9 11.014 0.41 13 1.73.2 69 0.5 5.16

Clone Tri

3 2 279 45 6.2 12.4126.7 729 5.2 39.3160 0.751 10 13.752.0 715 4.8 40.1660.50.75 9 6.314.1 68 3.9 30.038

00.5 24 1.02.4 24 0.7 19.0282 4 184 34 5.4 7.049.9 517 3.6 32.983 0.751 8 6.224.5 49 3.7 37.445

0.50.75 7 1.93.9 34 2.0 24.23600.5 18 0.31.0 13 0.4 11.418

1 15 86 21 4.1 2.510.6 26 2.0 28.940 0.751 4 1.83.8 34 2.0 25.7290.50.75 4 0.70.9 23 0.2 14.917

00.5 6 0.20.3 11 0.1 7.89

13 22 124 26 4.8 4.429.7 310 3.6 32.285 01 21 1.65.3 35 1.9 2632Canopy

A1U2 A2U2

Cl n nM DnL Layer depth (m) n nM DnL

Clone Gho

3 111 11932914 16 1.62.6 353 30398308 152 200 19314776 14 1.01.6 2079 1726531135 71 586 36157532 7 01.0 9231 4877466958 213 897 673815221 9 02.6 11663 69078106400 3

Clone Tri

3 78 5502243 22 2.54.4 300 25367327 16

2 148 7292468 12 1.62.5 1363 744015144 61 554 13813547 4 01.6 14603 3356551745 113 780 26598258 7 04.4 16267 4354174217 2


13/18

canopy openness values could be compared over the entiregrowing season for each clone (Figure 7). The modeling pro-cess recreates (i) weak row-related canopy closure dynamics

and the strong heterogeneity observed throughout the entiregrowing season for the canopy of Clone Gho (Figures 57),and (ii) the fast (less than 3 months) canopy closure of Clone



Figure 5. Examplesof virtual 2-year-old poplar crop images, computer-generated from the 3-D CPCA model outputs by the POV-Raysoftwareforpoplar ClonesGhoy (Gho) andTrichobel(Tri) in Mayand September2001. Crop canopies, consisting of 100 plantseach, are viewed from theSouth to the North. Axis lengths represent 1 m.

Figure 6. Examples of visual comparisons between fish-eye photographs (top line) and fish-eye images (bottom line) generated from the 3-DCPCA model outputs by the POV-Ray software for poplar clones Ghoy (Gho; ad, between-row location) and Trichobel (Tri; eh, inside-rowlocation). Crop canopies consist of 100 plants each.


14/18

Tri, which was constant (at about 90% compared to a maxi-mum of 60% for Clone Gho) after mid-June (Figures 57).

Discussion

Clonal characteristics

A considerable number of studies have shown that stands ofcertain highly productive poplars (often P. trichocarpa and itshybrids) quickly achieve high LAI values (up to 10 or more),whereas less productive ones (often Euramerican hybrids)may reach only one-third to one-half of those values (DeBellet al.1996, Heilman et al.1996, Casella and Ceulemans2002).But this genetic tendency is not always apparent because thedevelopment of plant LAI is often under strong environmentalcontrol (site climate, plantsoilnutrient interactions, pests

and diseases) (Liu and Dickmann 1992a, 1992b, Stettler et al.1996, Ceulemans and Deraedt 1999, Dickmann et al. 2001).Because leaf physiological processes (i.e., maximal photosyn-theticelectron transport capacity and carboxylation rate) seemto be relatively constant over the Populus spp. (Niinemets etal. 1998, Casella and Ceulemans 2002, Ripullone et al. 2003),canopy LAIdevelopment and leaf angular and spatial distribu-tions over a growing season seem to be the most pronouncedgenetictraits expressing the large range in yield of this species.Thus, under the same environmental conditions, these clonalcontrasts could be explained through the expression of con-trasting genotypic traits related to their potential to maximize

ei by an optimal coppice and canopy structure arrangementover the growing season (number of A1, A2 and L, and theirgeometry in space).The broad range of features detailed in this

paper and generated by the 3-D CPCA model (Table 6) arenow used to provide some explanations for the difference inperformance of our two clones. Although both the empiricalfunctions and the 3-D CPCA model do not consider the A1mortality and L abscission in the vegetative period, potentialLAI evaluated from model outputs could be used for a clonalcomparison between May and July because the leaf longevityof most poplarclones is between 90 and120 days (Zavitkovski1981, Ceulemans et al. 1992) and canopy closure of Clone Trioccurred in late July (Figure 7).

As expected, throughout the growing season of 2001, theclonal differences in coppice and canopy structures of our two

clones resulted in striking differences in plant phenology, andA1, A2, and L morphology and growth characteristics. Com-pared with Clone Tri, the growing season for Clone Gho wasshortest (about 5 months compared with 7 for Clone Tri), andbegan about 26 days later than for Clone Tri (Figure 7). InMarch, the two coppices showed a similar A1U1 density (Ta-ble 6) but, despite a similar number of I, the A1U1 of CloneGho were on average 1.4 times shorter than those of Clone Tri(Table 6). In contrast with several results for various natural(e.g., P. trichocarpa) and hybrid (e.g., Euramerican poplars)poplar clones (Ceulemans et al. 1990, Dunlap et al. 1992,Ceulemans and Isebrands 1996), a small number of lateral



Figure 7. Measured (mes) and simu-lated (sim) canopy openness over sixsky-map zenith divisions (, =015;, = 1530;, =3045;, = 4560; , =6075; and, = 7590) over the2001 growing season, and temporalcourses of canopy openness (lines =measured and symbols = simulated) forpoplar Clones Ghoy (Gho, left column)and Trichobel (Tri, right column).Open and closed symbols show be-tween- and inside-row locations, re-spectively. Vertical and horizontal barsindicate the standard deviation of themean (n = 5 or 10 for the inside- or be-tween-row locations, respectively).Letters indicate significant differences(t-test, P < 0.05) between row locations

and results obtained from field mea-surements and the modeling process.Relationships between simulated andmeasured canopy openness were fitted,for four zenith angles (060), to thelinear regression y = ax + b. Dottedlines depict the confidence interval atP < 0.05.


15/18

sylleptic A2 characterized the A1U1 of our two clones (


16/18

a quantitative assessment of reconstruction quality allows oneto evaluate easily the effect of combining several empiricalrules. For the first reason, such a criterion of reconstructionquality (e.g., canopyopenness in this study)must be defined atthe plant population scale, not the individual plant scale. Thecriterion can be used to refine the set of architectural rules and

identify the main architectural parameters that determine the3-D space occupation. In this study, we used a large set of re-construction rules in order to increase the reconstruction qual-ity (Figure 7). The next step would be to evaluate simplifiedreconstruction methods (i.e., by replacing some rules with as-sumptions, e.g., constant internode length), and then to definethe minimum data sets to be measured in the field for a givenreconstruction quality.

Conclusion

The 3-D CPCA model, based on a detailed set of empirical re-lationships obtained from field measurements, recreates the

spatial aL distribution within contrasting SRWC poplar cano-pies. This study shows that, as suggested by Casella andCeulemans (2002), foliage cannot be accurately described byassuming it to be randomly distributed within horizontal lay-ers ofPopulus canopy.

From a methodological point of view, the 3-D CPCA modelis likely to find several applications. (i) It could be used tomake sensitivity analyses of the reconstruction method toidentify the relevant parameters and define the minimum dataset to be measured in the field to feed the model. (ii) Becausethe 3-D CPCA model expresses the spatial geometry of stools,U2 and leaves of a population of plants, it could be used as aninput to 3-D modelsdealing with the distribution of light inter-

ception (e.g.,RATP, Sinoquet et al. 2001), rainfall interception(Gash 1979), wind speed (Daudet et al. 1999), photosynthesisand transpiration (e.g., RATP), and sap flow (Dauzat et al.2001) activitieswithin a canopy. With the accurate descriptionof the aL distribution within our canopies, a further applica-tion of the 3-DCPCA model could be the exploration of the ef-fects of foliage clumping on radiation interception. Error inradiation interception resulting from the assumption that fo-liage is randomly distributed could be estimated. (iii) The 3-DCPCA model could be used to test all these enumerated mod-els at different scales (from branch to canopy) to proposecomprehensive analysis of the relationships between leaf mor-phology and branchcanopy functions (e.g., light capture, ma-

nipulating shoot and branch inclination angles, petiole length,leaf shape, size and orientation) (Takenaka 1994) that will leadto increased productivity. Radiative transfer models could beusedto predictthe potentialgrowth of individual poplarclonesand ensure the efficient selection of those that will remain forthe final crop. Finally, (iv) data sets obtained with canopiesmeasured forseveral years could be used to test models of can-opy architecture dynamics (Sinoquet and Le Roux 2000).

Acknowledgments

This research was supported by a grant from Forest Research,an executive agencyof theForestry Commission, and by a Ma-

rie Curie Fellowship of the European Community programQuality of Life and Management of Living Resources undercontract number QLK5-CT-2001-50583. The authors grate-fullyacknowledge R. Ceulemans (University of Antwerp, Bel-gium) for useful discussions and review comments, as well asM. Baldwin (Forest Research) for improving the English of

this manuscript.

References

Armstrong, A. 1997. The United Kingdomnetwork of experimentsonsite/yield relationships for short rotation coppice. Note 294. For-estry Commission Research Information, Forestry Commission,Edinburgh, 6 p.

Barczi, J.F., P. de Reffye and Y. Caraglio. 1997. Essai sur lident-ification et la mise en uvre des paramtres ncessaires lasimulation dune architecture vgtale: le logiciel AMAPsim. InModlisationet Simulation de lArchitecture des Vgtaux. Eds. J.Bouchon, P. de Reffye and D. Barthlmy. INRA Editions, Paris,pp 205254.

Cannell, M.G.R. 1989. Physiological basis of wood production: a re-view. Scand. J. For. Res. 4:459490.

Cannell, M.G.R., L.J. Sheppard and R. Milne. 1988. Light use effi -ciency and woody biomass production of poplar and willow. For-estry 61:125136.

Casella, E. and R. Ceulemans. 2002. Spatial distribution of leaf mor-phological and physiological characteristics in relation to local ra-diation regime within the canopies of 3-year-old Populus clones incoppice culture. Tree Physiol. 22:12771288.

Chen, S.G. 1992. Modelingarchitectureand the light regime of poplarcanopies using fractal approach. Ph.D. Thesis, Dept. Biology,Univ. Antwerp, Belgium, 200 p.

Chen, S.G., R. Ceulemans and I. Impens. 1994a. A fractal-basedPopulus canopy structure model for the calculation of light inter-ception. For. Ecol. Manage. 69:97110.

Chen, S.G., B.Y. Shao, I. Impensand R. Ceulemans. 1994b. Effects ofplant canopy structure on light interception and photosynthesis. J.Quant. Spectrosc. Radiat. Transf. 52:115123.

Ceulemans, R. and W. Deraedt. 1999. Production physiology andgrowth potential of poplars under short-rotation forestry culture.For. Ecol. Manage. 121:923.

Ceulemans, R. and J.G. Isebrands. 1996. Carbon acquisitionand allo-cation.In Biology ofPopulus and its Implications for Managementand Conservation. Eds. R.F. Stettler, H.D. Bradshaw, P.E. Heilmanand T.M. Hinckley. NRC Research Press, Ottawa, Canada, pp355399.

Ceulemans, R., I. Impens and V. Steenackers. 1988. Genetic variationin aspects of leaf growth ofPopulus clones, using the leaf plasto-chron index. Can. J. For. Res. 18:10691077.

Ceulemans, R., R.F. Stettler, T.M. Hinckley, J.G. Isebrands and P.E.Heilman. 1990. Crown architecture of Populus clones as deter-mined by branch orientation and branch characteristics. TreePhysiol. 7:157167.

Ceulemans, R., G. Scarascia-Mugnozza, B.M. Wiard, J.H. Braatne,T.M. Hinckley, R.F. Stettler, J.G. Isebrands and P.E. Heilman.1992. Production physiology and morphology ofPopulus speciesand their hybridsgrown under shortrotation. I. Clonal comparisonsof 4-year growth and phenology. Can. J. For. Res. 22:19371948.

Ceulemans, R., A.J.S. McDonald and J.S. Pereira. 1996. A compari-son among eucalypt, poplar and willow characteristics with partic-ular reference to a coppice, growth-modeling approach. BiomassBioenergy 11:215231.




17/18

Daudet, F.A., X. Le Roux, H. Sinoquet and B. Adam. 1999. Windspeed and leaf boundary layer conductance variation within treecrown: consequences on leaf-to-atmosphere coupling and treefunctions. Agric. For. Meteorol. 97:171185.

Dauzat, J., B. Rapidel andA. Berger. 2001. Simulation of leaf transpi-ration andsap flow in virtualplants: model description and applica-tion to a coffee plantation in CostaRica. Agric. For. Meteorol. 109:

143160.DeBell, D.S., G.W. Clendenen, C.A. Harrington and J.C. Zasada.

1996. Tree growth and stand development in short-rotationPopulus plantings: 7-year results for two clones at three spacings.Biomass Bioenergy 4:253269.

Deraedt, W. and R. Ceulemans. 1998. Leaf phenology in the first andsecondyearof growthin a coppice biomass plantation. Biomass forEnergy and Industry, CEC, Carmen, Germany, pp 969971.

De Reffye, P., C. Edelin, J. Franon, M. Jaegger and C. Puech. 1988.Plant models faithful to botanical structure and development.Comput. Graph. 22:151158.

De Reffye, P., E. Elguero and E. Costes.1991. Growth units construc-tion in trees: a stochastic approach. Acta Biotheor. 39:325342.

Dickmann, D.I., J.G. Isebrands, J.E. Eckenwalder and J. Richardson.

2001. Poplar culture in North America. NRC Research Press,Ottawa, Canada, 397 p.

Dunlap, J.M., P.E. Heilman and R.F. Stettler. 1992. Genetic variationand productivity ofPopulus trichocarpa and its hybrids. V. The in-fluence of ramet position on 3-year growth variables. Can. J. For.Res. 22:849857.

Eckenwalder, J.E. 2001. Descriptions of clonal characteristics. InPoplar Culture in North America. Eds. D.I. Dickmann, J.G.Isebrands, J.E. Eckenwalder and J. Richardson. NRC ResearchPress, Ottawa, Canada, pp 331382.

Gash, J.H.C. 1979. An analytical model of rainfall interception byforests. Quart. J. R. Met. Soc. 105:4355.

Godin, C., E. Costes and H. Sinoquet. 1999. A method for describingplant architecture which integrates topology and geometry. Ann.

Bot. 84:343357.Grassi, G., G. Gosse and G. Dos Santos. 1990. Biomass for energy

and industry. Vol. 1. Policy, environment, production and harvest-ing. Elsevier Applied Science, London, 697 p.

Hall, F., R.A.A. Oldeman and P.B. Tomlinson. 1978. Tropical treesand forests. An architectural analysis. Springer-Verlag, Berlin,411 p.

Heilman, P.E. and R.F. Stettler. 1985. Genetic variation and produc-tivity ofPopulus trichocarpa and its hybrids. II. Biomass produc-tion in a 4-year plantation. Can. J. For. Res. 15:384388.

Heilman,P.E., T.M. Hinckley, D.A. Roberts and R. Ceulemans. 1996.Production physiology. In Biology ofPopulus and its implicationsfor management and conservation. Eds. R.F. Stettler, H.D. Brad-shaw, P.E. Heilman and T.M. Hinckley. NRC Research Press,

Ottawa, Canada, 539 p.Ivanov, N., P. Poissard, M. Chapron and B. Andrieu. 1995. Computer

stereo plotting for 3-D reconstruction of a maize canopy. Agric.For. Meteorol. 75:85102.

Laureysens, I., E. Casella and R. Ceulemans. 2000. Productivity,basal area and shoot density of different poplar clones in a coppiceculture. In Topics in Ecology: Structure and Function in Plants andEcosystems. Eds. R. Ceuleman, J. Bogaert, G. Deckmyn and I.Nijs. Univ. Antwerp, Wilrijk, Belgium, pp 2938.

Law, B.E., A. Cescatti and D.D. Baldocchi. 2001. Leaf area distribu -tion and radiative transfer in open-canopy forests: implications formass energy exchange. Tree Physiol. 21:777787.

Liu, Z. and D.I. Dickmann. 1992a. Responses of two hybrid Populusclones to flooding, drought and nitrogen availability. I. Morphol-ogy and growth. Can. J. Bot. 70:22652270.

Liu, Z. and D.I. Dickmann. 1992b. Responses of two hybrid Populusclones to flooding, drought and nitrogen availability. II. Gas ex-change and water relations. Can. J. Bot. 71:927938.

Lord, D., S. Morissette and J. Allaire. 1993. Influence of light inten-

sity, night air temperature and CO2 concentration on growth ofcontainerised black spruce seedlings grown in the greenhouse.Can. J. For. Res. 23:101110.

Macpherson, G. 1995. Home-grown energy from short-rotation cop-pice. Farming Press, Ipswich, U.K., 214 p.

Milne, R., M. Sattin, J.D. Deans, P.G. Jarvis and M.G.R. Cannell.1992. The biomass production of three poplar clones in relation tointercepted solar radiation. For. Ecol. Manage. 55:114.

Niinemets, ., O. Kull and J.D. Tenhunen. 1998. An analysis of lighteffects on foliar morphology, physiology, and light interception intemperatedeciduous woody species of contrasting shade tolerance.Tree Physiol. 18:681696.

Pieters, G.A., M.E. van den Noort and J.A. van Nijkerken. 1999.Growth adaptation of leaves and internodes of poplar to irradiance,

day length and temperature. Tree Physiol. 19:933942.Pontailler, J.Y., R. Ceulemans and J. Guittet. 1999. Biomass yield ofpoplar after five 2-year coppice rotations. Forestry 72:157163.

Prusinkiewicz, P.W. 1986. Graphical applications of L-systems. InProceedings of Graphics Interface 86. Ed. W. Davis. Canadian In-formation Processing Society, Toronto, Ontario, pp 247253.

Prusinkiewicz, P.W., W.R. Remphrey, G.C. Davidson and M.S.Hammel. 1994. Modeling the architecture of expanding Fraxinus

pennsylvatica shoots using L-systems. Can. J. Bot. 72:701714.Rakocevic, M., H. Sinoquet, A. Christophe, C. Varlet-Grancher.

2000. Assessing the geometric structure of a white Clover(Trifolium repens L.) canopy using 3-D digitizing. Ann. Bot. 86:519526.

Remphrey, W.R., A.G. Bartlett and C.G. Davidson. 2002. Shoot mor-phology and fate of buds in relation to crown location in youngFraxinus pennsylvanica var. subintegerrima. Can. J. Bot. 80:12741282.

Ridge, C.R., T.M. Hinckley, R.F. Stettler and E. van Volkenburgh.1986. Leaf growth characteristics of fast growing poplar hybridsPopulus trichocarpa Populus deltoides. Tree Physiol. 1:209216.

Ripullone, F., G. Grassi, M. Lauteri and M. Borghetti. 2003. Photo-synthesisnitrogen relationships: interpretation of different pat-terns between Pseudotsuga menziesii and Populus euramericanain a mini-stand experiment. Tree Physiol. 23:137144.

Ross, J. 1981. The radiation regime and architecture of plant stands.W. Junk Publishers, The Hague, 391 p.

Sinoquet, H. and X. Le Roux. 2000. Short term interactions betweentree foliage and the aerial environment: an overview of modeling

approaches available for tree structure-function models. Ann. For.Sci. 57:447496.Sinoquet, H. and P. Rivet. 1997. Measurement and visualizationof the

architecture of an adult tree based on dimensional digitizing de-vice. Trees 11:265270.

Sinoquet, H., S. Thanisawanyangkura, H. Mabrouk and P. Kasemsap.1998. Characterization of the light environment in canopies using3-D digitizing and image processing. Ann. Bot. 82:203212.

Sinoquet, H., X. Le Roux, B. Adam, T. Amglio and F.A. Daudet.2001. RATP, a model for simulating the spatial distribution of radi-ation absorption, transpirationand photosynthesis within canopies:application to an isolated tree crown. Plant Cell Environ. 24:395406.




18/18

Sonohat, G., H. Sinoquet, C. Varlet-Grancher, M. Rakocevic, A.Jacquet, J.C. Simon and B. Adam. 2002. Leaf dispersion and lightpartitioning in three-dimensionally digitized tall fescue-white clo-ver mixtures. Plant Cell Environ. 25:529538.

Stettler, R.F., H.D. Bradshaw, P.E. Heilman and T.M. Hinckley. 1996.Biology ofPopulus and its implications for management and con-servation. NRC Research Press, Ottawa, Canada, 539 p.

Takenaka, A. 1994. Effects of leaf blade narrowness and petiolelength on the light capture efficiency of a shoot. Ecol. Res. 9:109114.

Tubby, I. and A. Armstrong. 2002. Establishmentand managementofshort rotation coppice. Practice Note 7. Forestry Commission Re-search Information, Forestry Commission, Edinburgh, 12 p.

Whitehead, D., J.C. Grace and M.J.S. Godfrey. 1990. Architecturaldistribution of foliage in individual Pinus radiata D. Don crownsand the effect of clumping on radiation interception. Tree Physiol.7:135155.

Zavitkovski, J. 1981. Structureand seasonal distribution of litterfall inyoung plantations ofPopulus Tristis #1. Plant Soil 60:409422.


coppice polar architecture

Documents