Biomass and Bioenergy 21 (2001) 237–247

Multivariate approach for integrated evaluation of

clonal biomass production potential

P.J. Tharakan

a;∗

, D.J. Robison

b

, L.P. Abrahamson

a

, C.A. Nowak

a

a

Faculty of Forestry, College of Environmental Science and Forestry, State University of New York, 340 Illick Hall,

1 Forestry Drive, Syracuse, NY, 13210, USA

b

Department of Forestry, North Carolina State University, Jordan Hall, Raleigh, NC, 27695, USA

Received 21 June 2000; accepted 9 April 2001

Abstract

Evaluating the performance of clones to be used in short rotation intensive culture (SRIC) plantations for biomass production

is critical for identifying superior clones and matching them with sites on which they will perform best. This will lead to

increased production and a strengthening of the commercial prospects of these plantations. The primary objective of this study

was to use a multivariate approach to evaluate the relative clonal performance of 38 willow and hybrid poplar clones, deployed

in a genetic selection trial based on a coppice rotation system established in central New York State (NY) in 1997. Cluster

analysis was conducted using survival, several individual plant growth attributes, and insect defoliation, all measured during

or at the end of 1998. Two linear functions developed using discriminant analysis, comprising primarily of attributes related

to tree vigor and site adaptability; tree volume index and length of growing period, explained most of the variation (98.5%)

among the clusters. Eight of the 38 clones evaluated are expected to be high biomass producers, and are recommended for

more extensive clone-site trials and commercial scale plantations across central NY and the northeastern United States (US).

The results of this study indicate a possible approach to more e:ective juvenile selection in tree improvement programs, and

insights for a re;nement of the current SRIC tree ideotype. c

2001 Elsevier Science Ltd. All rights reserved.

Keywords: Willow; Hybrid poplar; SRIC; Coppice systems; Energy plantations; Genetic selection; Ideotype; Cluster

analysis; NY

1. Introduction

Short rotation intensive culture (SRIC) hardwood

tree plantations are being promoted in the United

States (US) and in other countries, as an economically

∗

Corresponding author. Tel.: [phone]; fax: 1-315-

470-6934.

E-mail address: [e-mail] (P.J. Tharakan).

and ecologically viable source of biomass for bioen-

ergy generation and biobased products [1,2]. These

plantations can also provide pulpwood, render envi-

ronmental bene;ts such as bio-;ltration and phytore-

mediation, help mitigate the e:ects of anthropogenic

carbon emissions, and create rural employment op-

portunities [3]. Among the di:erent hardwood tree

species, primarily willow (Salix spp.), and hybrid

poplar (Populus spp.) have been identi;ed as the most

0961-9534/01/$ - see front matter c

2001 Elsevier Science Ltd. All rights reserved.

PII: S0961-9534(01)00038-1

238

P.J. Tharakan et al. / Biomass and Bioenergy 21 (2001) 237–247

promising candidates for plantation establishment in

the northeastern US [4]. Development of superior ge-

netic material that is resistant to insects and diseases,

and is capable of high biomass production rates, is

critical for strengthening the commercial prospects of

SRIC plantations. Evaluating clonal relative perfor-

mance helps to identify superior clones, match them

with sites on which they will perform best, and pro-

vide feedback that can aid further tree improvement.

The traditional approach to clonal testing involves a

multi-stage selection process that culminates in com-

patibility trials that are aimed at identifying sets of

clones which can be advantageously grown together in

blocks or mixtures so as to help circumvent the prob-

lems associated with monoclonal plantations [5,6].

The last stage, however, is rarely implemented with

a few notable exceptions [7]. According to this

approach, during the initial stages, selection trials

containing many clones in small sized tree plots are

established and the clones are evaluated based on

univariate characteristics such as tree height, or insect

and disease resistance. In terms of a speci;c selection

approach, a more integrated alternative to the univari-

ate approach to clonal selection has been suggested,

one that involves the use of multivariate statistical

techniques to combine a wide range of desirable traits

into indices that provide a more comprehensive eval-

uation of the future growth potential and adaptability

of the clones [8]. Most studies in SRIC that have

used a multivariate approach have been conducted on

non-coppiced hybrid poplar plantation systems, and

reported varying degrees of success [9–11]. Similar

studies have not been attempted with coppiced willow

systems.

Owing to small tree plots, clonal selection trials

were considered to be unsuitable for obtaining in-

formation on the rotation age, area-wide (plantation

scale) biomass potential of the clones. Weak corre-

lations between juvenile and mature traits in most

tree crops meant that the best approach was to make

initial selections, plant the clones in larger clone-site

trials and make ;nal selections based on assessments

of area wide production. At every step, selection was

considered to be most e:ective when done at no less

than half the rotation age [12]. This approach is time

consuming and poses impediments to the rapid de-

velopment and testing of clones, both of which are

critical for the successful commercialization of bioen-

ergy plantations [13]. Recent studies in SRIC settings

indicate that, unlike traditional plantation forestry

where attempts at juvenile selection have often proved

ine:ective, SRIC might be more amenable to an early

and comprehensive estimation of clonal biomass pro-

duction potential [13,14]. Time and environmental

factors are largely responsible for thwarting juvenile

selection attempts. The short rotational nature and

intensive culture regime in SRIC reduces the impact

of both these agents [15].

Changes in growth patterns and in relative alloca-

tion of photosynthate to reproductive versus vegeta-

tive growth, from juvenile to mature phases, may have

lesser of an inJuence on the performance ranking of

genotypes, making phenotypic correlations more reli-

able [13].

SRIC is considered to be amenable to ideotype

based breeding [13]. Desirable tree attributes that may

contribute to high biomass productivity can be synthe-

sized together in the form of a model tree or an “Ideo-

type” [16] and provides a clear focus towards which

the tree improvement program can strive. Breeding

based on the ideotype concept has met with a fair de-

gree of success in ;eld crops [17]. A fairly detailed

SRIC hybrid poplar ideotype [16] and a preliminary

one for SRIC willow [18] have been proposed. Promi-

nent among the desirable morphological characteris-

tics speci;ed in these models are tall straight stems

with large diameters, coppicing ability, ;ve or more

stems per tree and long foliage retention period (grow-

ing period). In genetic selection trials, several clones

with diverse attributes are deployed in a homogeneous

environment, thus o:ering an opportunity for assess-

ing the advantage conferred by the variation in di:er-

ent phenotypic traits among the superior clones, and

for testing the existing ideotype.

This paper reports on the use of cluster analysis

technique to evaluate the relative performance of 38

willow and hybrid poplar bioenergy clones on a mul-

tivariate basis and separate them into growth poten-

tial classes at the end of the ;rst growing season of

a three-year coppice rotation. It is proposed that this

approach may aid in e:ective juvenile selection for

SRIC plantations. Rotation length, large plot ;eld tri-

als of these same clones also established in central NY,

will be used in future years to examine the accuracy

of the growth potential predictions made in this study.

Individual attributes of the clones identi;ed as having

P.J. Tharakan et al. / Biomass and Bioenergy 21 (2001) 237–247

239

superior growth potential were examined to see if they

;t the existing de;nition of an SRIC Ideotype.

The study was established in central New York

State (NY) in 1997, by the State University of New

York College of Environmental Science and Forestry

(SUNY-ESF). Based on the coppice rotation system,

the objective of the trial was to assess relative clonal

performance based on a suite of tree growth attributes.

A mix of tested and previously untested clones were

deployed in the trial with the objective of testing the

adaptability and productivity of untested clones, and

evaluating their performance relative to superior per-

formers that have been tested in this region [19–21].

The hybrid poplar clones had previously been tested,

in an uncoppiced system, in central New York region

and=or elsewhere in North America [22,23]. Some of

the willow clones, including several Salix eriocephala

clones, S. dasyclados (SV1) and S. discolor (S365),

had been tested at SUNY-ESF in earlier trials [19,20].

Two clone series had not been previously tested in this

region: S. purpurea (94001–94015); native to Europe

and naturalized in the NY region and, the SX series

(native to Japan). Willow clone SV1 and poplar clone

NM6, based on their consistent good performance in

earlier trials, were used as reference clones for this

study [19–21].

Clonal performance was monitored during 1997,

and post coppice in 1998. The results of the establish-

ment year 1997 are reported elsewhere [24,25]. Cop-

picing resulted in signi;cant changes in growth rates.

Clonal mean tree biomass for all clones increased from

34:4 g ± 9:5 in 1997, to 262:0 g ± 99:6 in 1998. The

rank correlation between clonal rankings based on tree

biomass, in 1997 and 1998, was moderate (R

s

= 0:64):

2. Materials and methods

2.1. Study site

This study was established in 1997 at SUNY-ESF’s

Genetics Field Station in Tully, New York (42 47

30

N, 76 07

30W). The soil was a well-drained Palmyra

gravelly silt loam (Glossoboric Hapludalf) that is

highly productive (capable of corn production of

10 o:d:t ha

−1

yr

−1

[26]. Total precipitation (rainfall)

and number of frost-free days during the 1998 growing

season period was 69 cm and 250 days, respectively.

2.2. Site preparation, plot establishment and

maintenance

In summer 1996, the existing woody overstory was

removed and the coarse roots were extracted. Herba-

ceous vegetation on the site was killed in fall 1996

with glyphosate (Roundup

J

, Monsanto Inc., St Louis,

MO) and 2,4-dichlorophenoxyacetic acid (2,4-D) at

1:0 and 0:5 kg ai ha

−1

, respectively. The site was

then ploughed, cross-disked and raked to a level sur-

face. Pre-emergent weed control was accomplished

in spring 1997 with oxyJurofen (Goal

J

1.6E, Rohm

and Haas Inc., Philadelphia, PA) at 1 kg ai ha

−1

.

The trial was planted in late April 1997, on approx-

imately 0:4 ha, as a randomized complete block de-

sign with four replications of 38 willow (Salix spp.)

and hybrid poplar (Populus spp.) clones. Plots were

planted with 48 cuttings of a single clone in an 8 × 6

array at 0:9 m × 0:6 m spacing. Cuttings were hand

planted with 25 cm dormant cuttings, Jush with the

soil surface. In December 1997, at the end of the ;rst

growing season, the trees were cutback (coppiced) at

2–4 cm above the ground with a power brush cutter to

promote coppice regrowth, as is typically done in this

production system [19]. In spring 1998, the trees grew

as ;rst-year coppice on one-year-old root systems. In

June 1998, the trial was fertilized with 120 kg ha

−1

of nitrogen as 3-month slow release fertilizer (sulfur

coated Urea). An 80 percent weed-free environment

(visual estimate of cover) throughout the trial (1997–

1998) was achieved with hand weeding, application

of Juazifop-p (Fusillade

J

, Zeneca Inc. Wilmington,

DE) at 0:84 kg ai ha

−1

to control grasses, and wick

application of glyphosate (1.2% solution).

2.3. Clones deployed in the trial

The 38 clones used in the selection trial were of

diverse parentage and geographical origin and repre-

sented a wide range of habits, yield capabilities and

insect resistance (Table 1).

2.4. Measurements

2.4.1. Survival and growth

Percent survival was determined in each plot at the

end of the 1998 growing season (;rst year of cop-

pice growth). We de;ne a tree as the aggregate of all

240

P.J. Tharakan et al. / Biomass and Bioenergy 21 (2001) 237–247

Table 1

List of willow and poplar clones deployed in the genetic selection trial at Tully, NY

Clone name

Parentage

Origin

a

SA2

Salix alba (S. alba)

Yugoslavia, Novi Sad

SV1

S. dasyclados

Canada, Ontario

S365

S. discolor

Canada, Ontario

S287

S. eriocephala (erio)

Canada, Ontario

94001

S. purpurea

USA, NY (New York)

94003

S. purpurea

USA, NY

94004

S. purpurea

USA, NY

94005

S. purpurea

USA, NY

94006

S. purpurea

USA, NY

94009

S. purpurea

USA, NY

94012

S. purpurea

USA, NY

94013

S. purpurea

USA, NY

94014

S. purpurea

USA, NY

94015

S. purpurea

USA, NY

PUR12

S. purpurea

Canada, Ontario

PUR34

S. purpurea

Canada, Ontario

SH3

S. purpurea

Germany

SX61

S. sachalinensis

Japan

SX64

S. miyabeana

Japan

SX67

S. miyabeana

Japan

S185

S. erio 16 × S. erio 24

Canada, Ontario

S19

S. erio 16 × S. erio 307

Canada, Ontario

S25

S. erio 16 × S. erio 276

Canada, Ontario

S301

S. interior 62 × S. erio 276

Canada, Ontario

S546

S. erio 16 × S. erio 24

Canada, Ontario

S557

S. erio 16 × S. erio 24

Canada, Ontario

S566

S. erio 28 × S. erio 24

Canada, Ontario

S599

S. erio 39 × S. pet 47

Canada, Ontario

S625

S. erio 39 × S. int 42

Canada, Ontario

S646

S. erio 28 × S. erio 24

Canada, Ontario

S652

S. erio 19 × S. erio 23

Canada, Ontario

Carolina

Populus deltoides × Populus nigra cv Carolina

USA, Michigan

DN5

P. deltoides × P. nigra

Canada, Ontario

DN70

P. deltoides × P. nigra

Canada, Ontario

DN74

P. deltoides × P. nigra

Canada, Ontario

DN34

P. deltoides × P. nigra cv. Eugenei

Canada, Ontario

NM5

P. nigra × P. maximowizii

Canada, Ontario

NM6

P. nigra × P. maximowizii

Canada, Ontario

a

Denotes the place where the collections or crosses were made rather than the geographical or botanical origin of the clone. For example,

S. purpurea clones were imported from Europe in colonial times and have since been naturalized to Ontario, Canada and the Northeastern

USA.

the stems that sprout from a single stool. We mea-

sured stem diameter (±0:1 mm) at 2:5 cm from ground

or point of emergence from the cutting, with a digi-

tal caliper. Stem height (length) (±2:5 cm) was mea-

sured from the base of the stem to the terminal bud. A

strati;ed random sample was used to select individ-

ual stems from each of the four center trees in each

plot for stem measurements. Following Ballard’s pro-

cedure [27], stems comprising a tree were divided into

two diameter strata (small and large) by visual es-

timation, and 50 percent of the total number of the

stems in each stratum were randomly sampled. The

total number of large and small stems in the tree was

counted and the mean stem diameter and height of

the tree was calculated using the weighted average of

the two strata. Dry tree biomass was estimated using

P.J. Tharakan et al. / Biomass and Bioenergy 21 (2001) 237–247

241

allometric relationships between stem diameter and

dry weight [25].

2.4.2. Bud break, leaf senescence and insect

defoliation

We recorded the timing of bud break and leaf senes-

cence on all the trees in each plot using a discrete

ranking scale at intervals of 10 ± 3 days. Bud break

surveys were started in April 1998, and leaf senes-

cence surveys were begun in October 1998. Buds on

each stool were scored as “bud break stages”: 0, buds

tightly closed; 1, swollen buds; 2, buds opening, leaf

tips clearly visible; 3, individual leaves expanding;

and 4, leaves expanded, stem elongation initiated. Leaf

senescence was scored as “senescence stages”: 1, no

leaf discoloration; 2, leaf discoloration, no leaf fall; 3,

leaf fall beginning; 4, moderate leaf fall; 5, heavy leaf

fall and 6, complete leaf loss. Growing period was de-

;ned as the total number of days between bud break

(stage 3) and leaf senescence (stage 6) [28]. Once a

month and immediately after an incident of insect in-

festation, plot-wide surveys were made to assess in-

sect defoliation. The extent of defoliation was scored

on a discrete ranking scale: 0, 1–10%; 1, 11–25%;

2, 26–50%; 3, 51–75% and; 4, 76–100% defoliation.

2.5. Statistical methods

Hierarchical cluster analysis was used to group

clones on a multivariate basis into growth potential

classes [29]. The analysis was performed using sur-

vival, defoliation, growing period, and tree volume

index—de;ned as D

2

H × N (diameter squared ×

height × number of stems), which is a modi;cation

of D

2

H, a version of the growth index that has found

wide acceptance in plantation forestry [30]. The anal-

ysis was also conducted after excluding the clones

that had sustained insect defoliation greater than 26

percent (“2” on ranking scale). This was because

heavy defoliation was clone speci;c and not uniform,

thereby probably confounding comparison of growth

among clones. In all the other aspects of the trial, all

clones were treated equivalently.

Ward’s Minimum Variance method was used to

construct the clusters (classes) and the actual number

of clusters was discerned using the pseudo t

2

statistic

and cubic clustering criterion [31]. Analysis of vari-

ance using the distance (Mahalonobis’ D

2

) among

growth potential class means (centroids) was used to

test the null hypothesis that growth potential classes

were equal. Following cluster analysis, discriminant

analysis was used to assess the contribution of individ-

ual variables to centroid separation. The e:ectiveness

of the classi;cation procedure was assessed through

misclassi;cation probabilities calculated by cross val-

idation [32]. We used the general linear model pro-

cedure of analysis of variance to evaluate univariate

di:erences among individual clones and the growth

potential classes. Tukey’s mean studentized range test

was used to determine signi;cant mean separations

among growth potential classes for each of the vari-

ables used in the cluster analysis. All statistical anal-

yses were performed using SAS Version 6 [33], at a

critical level of ˙= 0:05.

3. Results

We found signi;cant di:erences among clones

(P ¡ 0:05) in mean survival, stem dimensions and

number of stems per tree, length of the growing period

and defoliation index values at the end of the 1998

growing season (Table 2). All clones had greater than

85 percent survival and the mean overall survival

for the trial was over 96 percent. Mean clonal stem

diameter ranged from 0.8 to 1:9 cm, and mean clonal

height ranged from 0.9 to 2:2 m. Coppicing in the

winter of 1997 resulted in a new Jush of sprouts in

spring 1998 and by the end of the season, the number

of stems per tree ranged from 3 to 17.

The willow clones broke bud early, initiating shoot

growth by the second or third week of April, while the

hybrid poplar clones broke bud in mid May. Clonal

di:erences were also evident in leaf senescence. While

most of the hybrid poplar clones lost all their leaves

by late October, some willow clones (SV1, PUR12

and SH3) did not undergo leaf senescence until the

middle of December. Length of the growing period

ranged from 200 to 260 days among clones. Defo-

liation was caused by several insects: Gypsy moth

(Lymantaria dispar), Japanese beetle (Popillia japon-

ica), the willow leaf beetle (Calligrapha bigsbyana)

and imported willow leaf beetle (Plagiodera versicol-

ora). There were two major incidents of insect defo-

liation by the willow sawJy (Nematus ventralis), in

mid-June and mid-July 1998.

242

P.J. Tharakan et al. / Biomass and Bioenergy 21 (2001) 237–247

Table 2

Clonal means

a

(SD) for stem size and growth characteristics in 1998. The clones have been ordered by tree biomass

Clone

Oven dry

b

Stem

Stem

Stems per

Percent

Growing

Defoliation

tree

diameter

height

tree

survival

period

index

d

biomass (g)

(cm)

(m)

(days)

c

PUR12

477.6 (86.3)

e

1.0 (0.1)

1.7 (0.1)

16.2 (3.0)

99.0 (1.2)

260 (4)

1.0 (0.8)

94005

447.8 (56.6)

1.0 (0.1)

1.8 (0.1)

16.9 (3.0)

98.5 (3.1)

252 (5)

1.3 (1.3)

SX67

427.8 (151.2)

1.3 (0.2)

2.2 (0.3)

7.4 (0.9)

97.4 (3.1)

242 (7)

1.8 (1.0)

SV1

404.0 (72.8)

1.2 (0.1)

1.6 (0.2)

9.6 (1.5)

96.2 (1.7)

260 (0)

0.2 (0.2)

S365

392.0 (136.7)

1.2 (0.1)

1.5 (0.2)

12.3 (3.5)

92.7 (5.5)

256 (3)

0.5 (1.0)

94009

387.0 (77.1)

0.9 (0.1)

1.5 (0.1)

16.2 (3.3)

98.4 (2.0)

238 (25)

1.3 (1.0)

NM6

367.4 (158.1)

1.6 (0.4)

2.0 (0.3)

7.1 (2.0)

93.8 (2.9)

212 (4)

0.0 (0.0)

S546

356.6 (89.0)

1.2 (0.2)

1.5 (0.3)

10.0 (1.3)

89.6 (6.1)

244 (10)

1.3 (1.0)

94001

355.4 (69.7)

1.1 (0.0)

1.8 (0.0)

12.9 (2.5)

99.5 (1.1)

241 (7)

1.0 (0.8)

PUR34

353.0 (144.1)

0.9 (0.2)

1.7 (0.3)

16.1 (1.1)

96.9 (6.3)

208 (7)

0.8 (0.5)

SX61

333.1 (243.2)

1.3 (0.2)

2.0 (0.7)

7.6 (3.1)

98.4 (2.0)

252 (3)

2.3 (1.5)

94003

306.1 (104.3)

1.0 (0.1)

1.6 (0.3)

9.2 (2.6)

97.9 (1.7)

258 (6)

1.3 (0.5)

94004

295.5 (87.0)

1.0 (0.3)

1.6 (0.4)

11.2 (5.7)

96.9 (2.7)

246 (8)

1.3 (1.3)

S301

287.6 (30.7)

1.0 (0.1)

1.6 (0.1)

10.9 (2.0)

97.9 (1.7)

245 (12)

1.5 (0.6)

NM5

287.4 (89.6)

1.6 (0.4)

2.0 (0.5)

5.5 (1.5)

99.0 (1.2)

213 (7)

0.3 (0.1)

94013

280.4 (30.9)

0.9 (0.0)

1.7 (0.1)

13.0 (0.9)

99.5 (1.1)

239 (9)

0.3 (0.5)

CARO

269.3 (49.7)

1.4 (0.1)

1.7 (0.1)

6.5 (1.4)

98.4 (1.1)

211 (10)

0.0 (0.0)

S287

268.0 (73.9)

0.9 (0.1)

1.4 (0.2)

13.1 (1.9)

98.9 (2.1)

234 (7)

1.3 (1.0)

94014

263.9 (157.4)

0.9 (0.1)

1.5 (0.3)

12.1 (3.8)

99.0 (2.1)

238 (19)

1.0 (0.8)

SA2

242.3 (156.2)

0.9 (0.2)

1.4 (0.4)

14.0 (2.7)

98.4 (1.1)

214 (7)

0.5 (0.2)

SH3

231.8 (187.6)

0.8 (0.3)

0.9 (0.7)

10.0 (4.9)

98.4 (1.1)

263 (0)

0.3 (0.3)

S557

226.6 (65.5)

1.0 (0.0)

1.2 (0.2)

9.1 (3.3)

86.5 (8.4)

256 (7)

2.0 (1.4)

SX64

224.5 (75.9)

1.1 (0.2)

1.5 (0.4)

8.9 (2.7)

100.0 (0.0)

254 (10)

2.3 (1.5)

94015

216.0 (83.4)

0.8 (0.1)

1.5 (0.2)

12.7 (1.7)

96.4 (3.1)

231 (19)

1.8 (1.0)

94006

208.7 (53.4)

0.9 (0.1)

1.6 (0.2)

12.8 (5.5)

99.5 (1.1)

246 (8)

3.0 (1.4)

DN74

200.6 (81.6)

1.3 (0.3)

1.6 (0.4)

6.6 (1.5)

99.5 (1.1)

213 (7)

0.0 (0.0)

DN5

175.9 (61.9)

1.6 (0.2)

1.7 (0.2)

3.9 (1.2)

92.2 (4.6)

200 (0)

0.0 (0.0)

DN34

174.9 (76.3)

1.4 (0.2)

1.4 (0.3)

3.6 (0.7)

88.6 (3.6)

214 (5)

0.0 (0.0)

S599

173.9 (45.6)

0.9 (0.1)

1.2 (0.1)

9.2 (4.7)

95.9 (3.8)

245 (10)

2.3 (1.3)

S185

169.2 (149.0)

1.0 (0.4)

1.1 (0.5)

5.8 (2.8)

95.3 (4.3)

244 (20)

1.8 (0.5)

S652

164.0 (48.6)

0.9 (0.1)

1.0 (0.1)

7.8 (2.4)

94.8 (4.0)

254 (11)

2.5 (0.6)

DN70

157.9 (68.5)

1.9 (0.3)

1.8 (0.2)

3.0 (0.9)

94.3 (3.5)

210 (0)

0.0 (0.0)

S625

157.7 (134.2)

0.9 (0.3)

1.0 (0.4)

5.5 (2.9)

91.2 (7.8)

252 (10)

1.5 (0.6)

S25

150.6 (146.4)

1.0 (0.4)

1.2 (0.4)

5.1 (1.7)

92.2 (4.6)

255 (8)

2.5 (1.0)

94012

150.0 (33.0)

0.7 (0.1)

1.4 (0.0)

10.3 (3.7)

100.0 (0.0)

226 (20)

2.7 (0.6)

S566

147.6 (41.3)

1.0 (0.1)

1.1 (0.2)

5.4 (1.0)

98.42 (2.0)

247 (16)

2.3 (1.0)

S19

122.1 (57.6)

0.8 (0.1)

1.0 (0.2)

7.1 (4.2)

96.9 (5.0)

255 (5)

1.8 (1.3)

S646

100.9 (38.2)

1.0 (0.2)

1.0 (0.2)

4.7 (1.4)

89.6 (10.2)

231 (0)

2.5 (0.6)

a

Mean of all the stems in strati;ed sample that comprise a tree.

b

(MSE, F-statistic, P values): Tree biomass (37911:91; 3:40; 0:0001), Stem diameter (2:71; 7:26; 0:0001), Stem height (42:10; 4:98; 0:0001),

Stems per tree (56:57; 7:54; 0:0001), Survival percent (49:87; 3:54; 0:0001), Growing period (1817:90; 14:02; 0:0001) and defoliation index

(3:18; 5:89; 0:0001). 37 df were used in the entire analysis.

c

Growing period is the number of days between bud break and leaf senescence.

d

Defoliation index ranking scale, where: 0, 1–10%; 1, 11–25%; 2, 26–50%; 3, 51–75%; 4, 76–100% defoliation.

e

Values in parentheses are SDs.

Cluster analysis revealed four distinct growth po-

tential classes (Fig. 1). These classes were signif-

icantly di:erent based on pairwise comparison of

the distances among the four growth potential class

centroids (Mahalonobis distance, F-test, P = 0:0001).

A two-dimensional scatter plot of the linear discrim-

inant functions (LDF) that di:erentiate the clusters

revealed that LDF 1 is primarily responsible for

P.J. Tharakan et al. / Biomass and Bioenergy 21 (2001) 237–247

243

Fig. 1. Cluster analysis dendrogram for ;rst year coppice growth tree attributes among clones (1998). Clusters were constructed using

Ward’s minimum variance method.

discriminating among clusters, and accounted for 97

percent of the variation. LDF 2 accounted for 1.5

percent of the variation (Fig. 2, Table 3). LDF 1 is

weighed most heavily by tree volume index and to

some extent by defoliation index. LDF 2 is constituted

primarily by length of the growing period (Table 3).

Repeating the analysis after excluding clones that had

been defoliated more than 26 percent did not result in

signi;cant changes. While there was reassignment of

a few clones between clusters 2 and 3, there was no

change in the composition of clusters 1 and 4. Tree

volume index continued to be the primary contributor

to LDF 1, which now accounted for over 98 percent

of the variation. LDF 2 continued to be constituted

primarily by length of the growing period.

The clustering procedure did not di:erentiate

among genera or species. Except for cluster 2, which

contained only willows all other clusters contained,

both willow and hybrid poplar clones (Fig. 1). There

were signi;cant di:erences (P ¡ 0:05) among the

clusters with respect to all the tree attributes consid-

ered (Table 4). Tree volume index production varied

the most. Average tree volume index of the clones

in cluster 1 was the highest and was over four times

that of the clones in cluster 4. Cluster 1 also exhibited

excellent survival, moderate defoliation and a long

growing period. Both clusters 2 and 3 had excellent

survival and moderate tree volume index production,

with cluster 3 producing signi;cantly more. Although

cluster 3 had a shorter growing season and lesser

amount of defoliation relative to cluster 2, these dif-

ferences were not statistically signi;cant. Cluster 4

performed the poorest with relatively low survival,

extensive defoliation and lowest tree volume index

production of all the clusters, despite a long growing

period. Clones NM6 and SV1, which were selected

as reference clones for this study, were classi;ed into

clusters 1 and 3, respectively. Cross validation anal-

ysis of the cluster membership indicated that clone

94004 had been misclassi;ed from cluster 2 to 3

244

P.J. Tharakan et al. / Biomass and Bioenergy 21 (2001) 237–247

Fig. 2. Projection of clonal mean vectors by cluster membership for the ;rst two linear discrimination functions (LDF). LDF 1 is primary

discriminator and explains 97.0% of variation (principally dependent on stem volume index). LDF 2 explains 1.5% of the variation

(principally dependent on length of the growing period).

Table 3

Standardized coeTcients of the clonal variables of ;rst year cop-

pice growth (1998) that comprise the ;rst and second linear dis-

criminant functions (LDF).

LDF1

LDF2

CoeTcients

CoeTcients

Variables

Tree volume index

−1.035

0.105

Defoliation index

0.847

−0:495

Percent survival

−0:357

−0:146

Length of growing period

−0.158

0.702

Statistics

Eigen values

19.90

0.35

Cumulative variation explained

97.0

98.5

(correct classi;cation rate of 98.0%). Analysis with-

out the defoliated clones resulted in a lower correct

classi;cation rate (94.0%). In addition to clone 94004,

clone 94015 was misclassi;ed from cluster 4 to 3.

4. Discussion

These results are from the end of ;rst year coppice

of a three-year coppice rotation and are speci;c to the

coppiced SRIC system used. The clonal recommenda-

tions, however, may be valid for a larger geographical

region than central NY, as the site conditions are rep-

resentative of a signi;cant area across NY [34], and

some sites across the northeastern US.

The signi;cantly larger mean tree volume index

production of cluster 1 clones can be related to ex-

cellent survival, coppice vigor (number of stems per

stool) and long growing period (Tables 2 and 4,

Fig. 1). Cluster 1, in addition to the reference clone

NM6, contained seven other clones, ;ve of which

(94001, 94005, PUR12, SX61 and SX67) were being

tested in this region for the ;rst time. These new

clones appear to be well adapted to the central NY

region. Based on these ;ndings it can be provision-

ally concluded that clones in cluster 1 have the best

P.J. Tharakan et al. / Biomass and Bioenergy 21 (2001) 237–247

245

Table 4

Clusters and the average values of clonal attributes in each cluster measured on the ;rst year coppice growth (1998). Values followed by

di:erent letters within each column di:er signi;cantly at ˙= 0:05 according to Tukey’s studentized range test.

Clusters

Number

Tree volume

Growing period

Survival

Defoliation

of clones

index (cm

3

)

(days)

(percent)

index

1

8

2808.6a

246ab

97.3ab

a

1.0ab

2

7

1430.8b

241ab

98.0a

1.6ab

3

11

1962.0c

224a

96.3ab

0.6a

4

12

691.0d

247b

93.7b

1.8b

a

(MSE, F-statistic, P value): Survival (36:50; 3:79; 0:0190), Defoliation (3:45; 5:59; 0:0031), Growing period (1192:64; 3:32; 0:0313),

Stem volume (7722484:2; 6:10; 0:0001). 3 df were used in the entire analysis.

potential to be excellent biomass producers in this

region and are recommended for larger scale trials.

Despite not performing as well as the cluster 1 clones,

the relatively good growth of clones in clusters 2 and

3 makes them good contenders for biomass produc-

tion. Since changes in clonal performance rankings

between the ;rst year and end of rotation have been

known to be most signi;cant in intermediate perform-

ing clones [14], it may be premature to reject cluster

2 and cluster 3 clones. The ;nal decision to retain the

clones or discard them should be made at the end of

the present rotation, prior to larger scale testing. The

poor performance of cluster 4 clones in terms of tree

volume index production, despite a long growing pe-

riod, may be attributed to relatively high defoliation

damage and low coppice vigor. These clones may not

be suited for biomass production purposes.

Results of the discriminant analysis indicate that, of

the di:erent variables used to estimate clonal growth

potential, the variables that were most useful in dis-

criminating among growth potential classes were tree

volume index, length of the growing period and defo-

liation index (Table 4). Tree volume index and length

of growing period relate to tree vigor and adaptation

to local climatic conditions, respectively. The calcu-

lation of stem volume index included the number of

stems per tree, in addition to D

2

H as in coppice SRIC

systems, the number of stems put out by a clone af-

ter coppicing has been linked to its ability to occupy

the site, deploy large leaf area and intercept light [35].

Signi;cant clonal di:erences in extent of defoliation

were evident, with some of the S. eriocephala clones

and SX clones su:ering most damage. It is diTcult to

assess the impact of defoliation on clonal performance

or on the results of the cluster analysis, as these defo-

liation surveys are merely indicative of initial clonal

resistance to a few insect species (especially N. ven-

tralis) present in the landscape during the period of

the study. No inference can be made of overall dis-

ease or insect resistance, or clonal tolerance to defo-

liation. Percent survival was not very important for

the purposes of discrimination, given the overall high

survival percentage in all the clusters.

Monitoring one or two single phenotypic character-

istics such as stem height precludes a more integrated

assessment and prediction of growth potential, as sin-

gle characteristics may often be poorly correlated with

rotation age biomass production [10]. Multivariate

techniques like cluster analysis allow us to circumvent

these limitations by allowing for a more comprehen-

sive evaluation of the clones based on a suite of vari-

ables that may be more indicative of future production

potential. The variables considered in this study relate

to growth and vigor, disease resistance and site adapt-

ability. Together they may be considered as indicative

of the area-wide (plantation scale) production poten-

tial of a clone. Clones NM6 and SV1, which were

identi;ed as having good growth potential in this

study, have proven to be good area-wide producers in

larger trials conducted by SUNY-ESF [9,19]. By ex-

tension, we suggest that the evaluation technique used

in this study could be used to rapidly assess area-wide

production potential of clones in selection trials. Thus,

adopting this approach may allow us to circumvent

the clone-site trials and move directly to large-scale

demonstration trials, at least in cases where plantation

sites are fairly homogeneous, thereby addressing the

need for quick and e:ective clonal selection proce-

dures in SRIC. Ongoing studies with the same clones

planted in larger scale trials will provide the empirical

evidence necessary to test the validity of this propo-

sal. This makes a case for the systematic monitoring

246

P.J. Tharakan et al. / Biomass and Bioenergy 21 (2001) 237–247

of demonstration plantations and pre-commercial

plantations, so as to aid in data collection, hypothesis

testing and adaptive management [36].

Cluster 1 clones were evaluated to assess the accu-

racy of the existing SRIC ideotype. All the clones in

this group had over 90 percent survival (Table 2) and

similar stem diameters and heights, except for clones

SX67 and NM6 that had substantially larger stem di-

ameter and height. There was wide variation in num-

ber of stems per tree, defoliation percent and length

of growing period. Clone NM6 had very few stems

per tree (7.1), negligible defoliation and short grow-

ing period (212 days), on the other hand clone SX67

had few stems per tree (7.4), the highest defoliation in

the group (1.8) and a moderately long growing period

(242 days). Clones PUR12, 94005 and 94001 were

characterized by large to moderate number of stems

per tree (16.2, 16.9, 12.9), moderate defoliation (1.0,

1.3, 1.0) and long growing period (260, 252, and 241

days), respectively.

All the clones in cluster 1 satis;ed some of the

existing SRIC ideotype speci;cations, they were char-

acterized by at least ;ve tall stems with large diameters

per tree and moderate to long growing period (foliage

retention). The wide variation in growing period, ex-

tent of defoliation and number of stems per tree, how-

ever, meant that there was no clear combination of

di:erent levels of these variables that was indicative

of superior performance potential. Moreover, related

work with these clones has demonstrated the existence

of large di:erences in ecophysiological characteris-

tics such as leaf area, speci;c leaf weight and above

ground biomass partitioning patterns [25], all of which

have been shown to inJuence clonal biomass pro-

duction potential [37,38]. Further work is needed to

characterize the e:ects of these variables, and several

other morphological, physiological and ecophysiolog-

ical characteristics, on area-wide biomass production.

Gaining an insight into the relationship between these

di:erent variables and production will facilitate the

inclusion of speci;cations on these attributes, thus

re;ning the existing SRIC ideotype.

Acknowledgements

The authors wish to thank the Biomass Feedstock

Development Program of the US Department of

Energy under contract DE-AC05-00OR22725 with the

University of Tennessee-Battelle LLC. (Subcontract

number 19X-SW561C) for funding this study. We

acknowledge the critical manuscript reviews pro-

vided by R. D. Yanai, G. Shriver, D. Bickelhaupt and

T. Volk. Special appreciation is extended to D. Bick-

elhaupt, R. Filhart, J. Zuckerbraun and M. Appleby

for assistance in the ;eld and the laboratory.

References

[1] Hall DO, Mynick HE, Williams RH. Carbon sequestration

versus fossil fuel substitution: alternative roles for biomass

in coping with greenhouse warming. PU=CEES Report No.

225, Princeton University, Princeton, NJ, 1990. 27pp.

[2] Noon CE, Daly MJ. GIS-based biomass resource assessment

with BRAVO. Biomass and Bioenergy 1996;10(2–3):101–9.

[3] Abrahamson LP, Robison DJ, Volk TA, White EH,

Neuhauser EF, Benjamin WH, Peterson JM. Sustainability

and environmental issues associated with willow bioenergy

development in New York (USA). Biomass and Bioenergy

1998;15(1):17–22.

[4] White EH, Abrahamson LP, Gambles P, Zsu:a L.

Experiences with willow as a wood biomass species. In:

Klass DL, editor. Energy from Biomass and Wastes XII,

Institute of Gas Technology, New Orleans, LA. 1988.

p. 227–44.

[5] Zobel Z, Talbert J. Applied forest tree improvement. USA:

Wiley, 1984, 505pp.

[6] Libby WJ. Testing for clonal forestry. Annales Forestales

Vol. 13=1–2. Ontario Ministry of Natural Resources (1983)

New Forests in Eastern Ontario. Hybrid Poplar Science and

Technology Series 1987, Vol. 1(336), p. 69–75.

[7] Parker SR, Arnold GM, Royle DJ. Plantations designs for

short rotation coppice that will stand the test of time. In:

Modeling in applied biology: spatial aspects, Bristol, UK, 25

–27 June 1996. p. 236–46.

[8] Zuuring H, Wray P, Promnitz L, Gordon JC. Use

of controlled-environment and statistical techniques to

maximize discrimination among clones. North Central Forest

Experiment Station, General Technical Report NC, No. 81,

1983. p. 17–44.

[9] Abrahamson LP, White EH, Nowak CA, Briggs RD,

Robison DJ. Evaluating hybrid poplar clonal growth potential

in a three-year-old genetic selection ;eld trial. Biomass

1990;21:101–14.

[10] Walsh TA, Burk TE, Isebrands JG. Development and

evaluation of quantitative functions for early selection of

Populus clones. Biomass and Bioenergy 1996;11(2=3):

151–9.

[11] Ares A, Gutierrez L. Selection of poplar clones for the

lower valley of the Colorado River, Argentina. Forestry

1996;69(1):75–82.

[12] Demeritt Jr ME. Growth of hybrid poplars in Pennsylvania

and Maryland clonal tests. USDA Forest Service,

P.J. Tharakan et al. / Biomass and Bioenergy 21 (2001) 237–247

247

Northeastern Forest Experiment Station, Research Note

NE-302, 1981.

[13] Stettler RF, Bradshaw Jr HD, Zsu:a L. The role of

genetic improvement in short rotation forestry. In: Mitchell

CP, Ford-Robertson JB, Hinckley T, Sennerby-Forsse L,

editors. Ecophysiology of short rotation forest crops. England,

Elsevier Applied Science, 1992. p. 285–304.

[14] Lo M, Abrahamson LP, White EH, Manion PD. Early

measures of basal area and canker disease predict growth

potential of some hybrid poplar clones. Canadian Journal of

Forest Research 1995;25:1113–8.

[15] Zsu:a L. Experiences in vegetative propagation of Populus

and Salix and problems related to clonal strategies. In: Baker

FWG, editor. Rapid propagation of fast growing woody

species. London: Redwood Press Ltd, 1992. p. 86–97.

[16] Dickmann DI. The ideotype concept applied to forest trees.

In: Cannell MGR, Jackson JE, editors. Attributes of trees

as crop plants. Huntingdon, UK: ITE, Monks Wood, Abbots

Ripton, 1985. p. 89–101.

[17] Adams MW. Plant architecture and yield breeding. Iowa State

Journal of Research 1982;56:225–54.

[18] Koski V, Dickmann DI. Tree ideotype. Biomass and

Bioenergy 1992;2(1=6):71–5.

[19] Kopp KF, White EH, Abrahamson LP, Nowak CA, Zsu:a L,

Burns KF. Willow biomass trials in central New York State.

Biomass and Bioenergy 1993;5(2):179–87.

[20] Kopp KF, Abrahamson LP, White EH, Burns KF, Nowak

CA. Cutting cycle and spacing e:ects on biomass production

by a willow clone in New York. Biomass and Bioenergy

1997;12(5):313–9.

[21] White EH, Abrahamson LP, Maynard CA, Briggs RD,

Robison DJ, Nowak CA. Biotechnology and genetic selection

of fast-growing hardwoods. Energy Authority Report 89-11.

Report submitted to New York State Energy Research

and Development Authority. State University of New York

College of Environmental Science and Forestry, 1989. 30pp.

[22] Lo M, Abrahamson LP. Principal Component Analysis to

evaluate the relative performance of nine-year-old hybrid

poplar clones. Biomass and Bioenergy 1996;10(1):1–6.

[23] Robison DJ, Ra:a RF. Productivity, drought tolerance

and pest status of hybrid Populus: Tree improvement

and silvicultural implications. Biomass and Bioenergy

1998;14(1):1–20.

[24] Tharakan PJ, Abrahamson LP, Isebrands JG, Robison

DJ. First year growth and development of willow and

poplar bioenergy crops as related to foliar characteristics.

Proceedings of BIOENERGY 98, Madison, Wisconsin,

October 4–8, 1998. Vol. 1, p. 1170–81.

[25] Tharakan PJ. Clonal performance evaluation and production

physiology of willow and poplar bioenergy crops. MS thesis,

State University of New York College of Environmental

Science and Forestry, Syracuse, NY, 1999. 196pp.

[26] Hutton FZ, Rice CE. Soil survey of Onondaga county. New

York. Soil Cons. Serv., 1977. 235pp.

[27] Ballard BD, Stehman SV, Briggs RD, Volk TA, Abrahamson

LP, White EH. Aboveground biomass equation for ;ve

Salix and one Populus clones. Interim Status Report No. 1,

Submitted to Niagara Mohawk Power Corporation, Syracuse,

NY. State University of New York College of Environmental

Sciences and Forestry, Syracuse, NY, 1998. 28pp.

[28] Ceulemans R, Scarascia-Mugnozza GE, Wiard BM, Braatne

JH, Hinckley TM, Stettler RF, Isebrands JG, Heilman PE.

Production physiology and morphology of Populus species

and their hybrids grown under short rotation. I. Clonal

comparisons of 4-year growth and phenology. Canadian

Journal of Forest Research 1992;22:1937–48.

[29] Morrison DF. Multivariate statistical methods. New York:

McGraw-Hill Book Company, 1976. 415pp.

[30] Hatchell GE, Berry CR, Muse HD. Nondestructive indices

related to aboveground biomass of young loblolly and sand

pines on ectomycorrhizal and fertilizer plots. Forest Science

1985;31(2):419–27.

[31] Duda RO, Heart PE. Pattern classi;cation and scene analysis.

New York: Wiley 1973. 482pp.

[32] Johnson RA, Wichern DA. Applied multivariate statistical

analysis. New Jersey: Prentice-Hall, Inc., 1992. 642pp.

[33] SAS Institute Inc. User’s guide-volume 1, ACECLUS-

FREQ. Version 6. 4th ed. 4. Cary, NC: SAS Institute Inc.,

1990.

[34] Park G. E:ects of clone, fertilization, cutting cycle and

spacing on carbon content of Willows. Ph.D. thesis, State

University of New York, College of Environmental Science

and Forestry, 1996. 251pp.

[35] Ceulemans R, McDonald AJS, Pereira JS. A Comparison

Among eucalypt, poplar and willow characteristics with

particular reference to a coppice, growth-modeling approach.

Biomass and Bioenergy 1996;11(2=3):215–31.

[36] Nowak CA, Volk TA, Ballard B, Abrahamson LP, Filhart RC,

Kopp RF, Bickelhaupt DH, White EH. The role and process

of monitoring willow biomass plantations. Proceedings of

the Fourth Biomass Conference of the Americas, Biomass:

A Growth Opportunity in Green Energy and Value-Added

Products, Oakland, CA, August 26–29, 1999. p. 1711–6.

[37] Cannell MGR. Physiological basis of wood production:

a review. Scandinavian Journal of Forest Research 1989;

4:459–90.

[38] Scarascia-Mugnozza GE, Ceulemans R, Heilman PE,

Isebrands JG, Stettler RF, Hinckley TM. Production

physiology and morphology of Populus species and their

hybrids grown under short rotation. II. Biomass components

and harvest index of hybrid and parental species clones.

Canadian Journal of Forest Research 1997;27:285–94.