New Forests (2017) 48:67–82

DOI 10.1007/s11056-016-9556-2

Growth and wood properties of natural provenances,

local seed sources and clones of Eucalyptus camaldulensis
in southern India: implications for breeding
and deployment

M. Varghese1 • C. E. Harwood2 • D. J. Bush3 •

B. Baltunis4 • R. Kamalakannan1 • P. G. Suraj1 •

D. Hegde1 • R. Meder5

Received: 4 April 2016 / Accepted: 8 September 2016 / Published online: 22 September 2016
Ó Springer Science+Business Media Dordrecht 2016

Abstract Genetic trials at three dryland sites in southern India compared 183 families
from 4 superior natural provenances, 48 families from locally developed seed sources and
10 commercial clones of Eucalyptus camaldulensis. Three of the local seed sources were
seed production areas developed by phenotypic selection for growth from an initial broad
base of superior natural provenances, and two were clonal trials. The local seed sources
grew significantly faster to 3 years than the natural provenances and the clones. Mean
survival at 3 years of the natural provenances (72 %, across the three sites) was better than
that of local seed sources (67 %) and clones (50 %). The three types of planting materials
did not differ significantly in their wood basic density or NIR-predicted lignin content,
while clones had significantly higher NIR-predicted pulp yield. Site mean wood density
was highest (579 kg m-3), and pulp yield lowest (43.7 %) at the driest site where growth
was slowest, while at the wettest, most productive site, density was 517 kg m-3 and pulp
yield 46.5 %. Narrow-sense heritabilities and inter-site genetic correlations for growth and
wood traits were moderate to high. Genetic correlations between growth and wood traits
did not differ significantly from zero. Unpedigreed seed production areas developed from
an appropriate genetic base of best provenances, may provide a simple option to mass-
produce improved seed.

Electronic supplementary material The online version of this article (doi:10.1007/s11056-016-9556-2)

contains supplementary material, which is available to authorized users.

& M. Varghese
mvarghese1@rediffmail.com
1
ITC Life Sciences and Technology Centre (LSTC), Bangalore 560 058, India
2
CSIRO Land and Water, Private Bag 12, Hobart 7001, Australia
3
CSIRO Australian Tree Seed Centre, Black Mountain, ACT 2601, Australia
4
Weyerhaeuser, PO Box 9777, Federal Way, WA 98063-9777, USA
5
Meder Consulting, Bracken Ridge, QLD 4017, Australia

123
68 New Forests (2017) 48:67–82

Keywords Provenance  Seed production area  Clone  Genetic gain  Growth  Density 
Pulp yield

Introduction

India has about 4 million ha of eucalypt plantations (GIT-Forestry 2008) which provide a
major source of raw material for the nation’s pulp and paper industry. Most of southern
India experiences a long, hot dry season, and Eucalyptus tereticornis and E. camaldulensis
are the best-performing and preferred planting species (Ginwal 2014). Because of statutory
limits on the size of agricultural land holdings, Indian wood processing industries cannot
own or directly manage sufficient areas of plantations to meet their wood requirements.
The majority of Indian eucalypt plantations are managed by smallholder farmers, many of
whom plant the local land race (Mysore gum) that has developed primarily from early
introductions of E. tereticornis (Boland 1981). Seed of Mysore gum, collected from tree
crowns after plantations are felled, is available at very low cost. Planting stock of better
genetic quality could increase plantation productivity, improve farm incomes and help
meet the acute wood shortage faced by the industry.
Clones of E. camaldulpensis and E. tereticornis were developed by paper companies,
commencing in the 1980s. The best-performing clones that have been made available to
farmers have helped to increase wood yields during the last decade. However plantations of
these clones, mostly selections from provenance trials and the land race (Kulkarni 2013)
make up only about a quarter of the total area of eucalypt plantations. The number of
available, well-tested clones is small, and they do not provide a sufficiently broad or well-
documented genetic base to sustain breeding over multiple generations. There is a ten-
dency to plant a few selected clones across diverse ecosystems without proper evaluation
of clone-by-environment interaction, resulting in serious failures. Some eucalypt clones are
highly susceptible to pests and diseases such as the gall wasp Leptocybe invasa
(Senthilkumar et al. 2013). Infusion of new genetic material is therefore essential for
ongoing breeding to increase productivity and reduce risk.
A genetic improvement program for E. camaldulensis and E. tereticornis for
southern India commenced in 1996. Unpedigreed first-generation seed production areas
(SPAs) of E. camaldulensis using a mix of seed from 11 natural provenances from
Queensland, Northern Territory and Western Australia were established and pheno-
typically thinned to produce improved seed (Varghese et al. 2009). Provenance-pro-
geny trials testing a total of 188 open-pollinated families of E. camaldulensis from
northern Australian provenances were established at three locations. These were used
to identify superior provenances and families, and to select and test individual clones
for deployment and breeding (Varghese et al. 2008). The progeny trials provided
information on provenance performance (Varghese et al. 2008) and served as first-
generation breeding populations.
In a genetic gain trial bulk seedlots collected from three first generation SPAs grew
significantly faster to 3 years (Dbh 6.1 cm) than a seedlot of the local Mysore gum land
race (Dbh 5.4 cm). However, in this trial and in two other gain trials, the SPA seedlots had
no growth advantage over a bulk seedlot incorporating wild seed from three of the best
natural provenances of E. camaldulensis (Laura, Morehead and Kennedy River prove-
nances) or the best available commercial clones (Varghese et al. 2009). However, some
farmers in southern India purchased large quantities of bulked seed from the first

123
New Forests (2017) 48:67–82 69

generation SPAs and SSOs for planting, due to its clear superiority over the local Mysore
gum land race.
Eucalypt plantations in southern India are grown on short rotations (usually about
four years). Improving growth has been the major focus of improvement for E.
camaldulensis to date, with little attention to wood properties. Most wood is sold on
green weight, so there is no financial incentive for growers to produce wood with
improved wood properties. Neither is there an imperative to increase wood basic
density from current levels of about 550 kg m-3 in three-year-old trees, since this is
close to optimal for pulping (Varghese et al. 2008). However, the species has low pulp
yield compared to other commonly planted tropical eucalypts such as E. grandis and E.
urophylla (Dehon et al. 2013). It would be desirable to breed for increased pulp yield as
well as growth rate. Breeding and deployment strategies that benefit both the growers
and processing industries need to be devised.
In some circumstances (Griffin 2014) a seed-based deployment strategy might be
more effective than clonal forestry, which is the option currently favoured by most
Indian companies and researchers (Kulkarni 2013). Some E. camaldulensis planta-
tions established using seed from the first-generation SSOs and SPAs have been
phenotypically thinned to convert them into unpedigreed second generation SPAs as a
low-cost way of meeting ongoing seed requirements. If applied at large scale, this
strategy could quickly produce large quantities of seed at low cost, facilitating
replacement of the poorly-performing Mysore gum. However, information on the
performance of plantations based on second-generation SPAs is lacking. Also of
interest is the change in fecundity under domestication: it was shown (Varghese et al.
2009) that one generation of domestication of E. camaldulensis in India significantly
increased the proportion of fertile trees, compared to that in wild populations. This
finding has implications for seed yields and outcrossing rates in advanced-generation
SPAs.
To broaden the genetic base of E. camaldulensis for breeding, a total of 183 open-
pollinated seed families were imported from four natural provenances of E. camal-
dulensis among those ranked highest for growth in southern India (Varghese et al.
2008). These new introductions were tested in three contrasting dryland environments
in trials that also included selected families from local SPAs and currently-deployed
commercial clones. Here we report genetic parameter estimates for growth and wood
traits and the extent and nature of genotype-by-environment interaction, and compare
seed-based and clonal strategies for mass-producing genetically improved planting
stock.

Materials and methods

Genetic material tested

The field trials tested open-pollinated progenies of E. camaldulensis from natural prove-
nances in north Queensland, Australia, local south Indian seed sources of E. camaldulensis,
and commercial clones. A total of 183 open-pollinated families of six collections repre-
senting the Kennedy River, Laura River, Morehead River and Petford provenances
(Table 1) and 48 open-pollinated families collected from five local seed sources of E.
camaldulensis were tested. The natural provenances spanned two subspecies of E.

123
70 New Forests (2017) 48:67–82

Table 1 Natural provenances, local seed sources and clones of Eucalyptus camaldulensis tested in the trials
Source CSIRO Type No. of families Latitude Longitude Altitude
seedlot or clones (E) (m)
no.

Kennedy River 20654 Natural provenance 34 15°220 S 144°100 85

0
Laura River 20650 Natural provenance 80 15°34 S 144°28 100
Morehead River 19615 Natural provenance 28 15°150 S 143°34 60
Kennedy River Laura 20907 Natural provenance 20 15°180 S 144°10 60
Morehead River 19010 Natural provenance 7 15°020 S 143°40 60
Petford Area 16720 Natural provenance 14 17°240 S 145°02 490
SPA 1 2nd generation SPA 10 13°050 N 75°42 597
SPA 2 2nd generation SPA 10 15°26 N 76°41 473
SPA 3 2nd generation SPA 11 15°16 N 80°00 9
Clone trial 1 Clone trial 7 15°46 N 80°35 9
Clone trial 2 Clone trial 10 17°100 N 81°45 22
Commercial clones Clones 10

camaldulensis: subsp. acuta (Petford, Kennedy River and Morehead River provenances)
and subsp. simulata (Laura River) (McDonald et al. 2009).
The development of the local seed sources was as follows. Three were second-gener-
ation SPAs developed by low-intensity selective thinning of second-generation stands.
These had been established from an unpedigreed mix of seed collected from 25 pheno-
typically superior parent trees in the first-generation SPAs described by Varghese et al.
(2009), which incorporated provenance bulk seedlots from 11 natural provenances of E.
camaldulensis in northern Australia, collectively representing a total of over 500 open-
pollinated families. Two of these second-generation SPAs were sets of 49-tree plots of E.
camaldulensis within two genetic gain trials (Varghese et al. 2009) which were subse-
quently selectively thinned. The third was a selectively thinned 2 ha farmer-managed stand
of E. camaldulensis established from a 25-tree bulk from one of the first-generation SPAs.
This second-generation SPA has been used as a seed source for subsequent plantations.
The remaining seed sources were two clonal trials, each of which included 21 clones that
are planted commercially in southern India (Kulkarni 2013). Progenies from 17 of the
clones were collected. Ten of these 21 clones were also included as experimental treat-
ments in the field trials.

Field trials

Field trials were established at one coastal and two inland locations in southern India
(Table 2). All sites experience very hot, seasonally dry climates. The length of the annual
dry season, defined as consecutive months receiving less than 40 mm of rain, over the
study period (years 2010–2012) was 5 months at Ongole, 3–4 months at Kovai, and
7 months at MBNagar. Ongole represents a more productive site type, from which paper
companies procure large quantities of wood. Kovai and MBNagar are representative of
marginal sites where resource-poor farmers raise less-productive plantations using low
inputs.

123
New Forests (2017) 48:67–82 71

Table 2 Description of trial sites of Eucalyptus camaldulensis genetic field trials

Kovai Ongole MBNagara
(Mettupalayam)

Latitude (N) 11°190 15°160 16°440

0 0
Longitude (E) 77°56 80°00 77°590
Altitude (m) 426 9 498
Mean Annual Rainfall (mm) 760 1078 661
Mean annual temperature (°C) 26.9 28.5 26.7
Mean daily maximum temperature of hottest month (°C) 35.4 38.2 39.3
Mean daily minimum temperature of coldest month (°C) 19.7 19.6 15.6
Dry season (consecutive months with mean monthly rainfall 3 or 4b 5 7
\40 mm
Soil texture Loam Loam Sandy
loam

Temperature averages are for the period 1982–2012, taken from http://en.climate-data.org/, while rainfall
data are from rain gauges maintained near each trial site
a
Temperatures shown are for nearby Jadcherla, 16°46N, 78°09E
b
Bimodal rainfall distribution with two separate dry seasons separated by two wetter months

Randomized row-column incomplete block designs were used in each trial. There were
5 replications and 2-tree plots of each family or clone within each replication. Initial
spacing, along ploughed planting rows, was 3 m 9 1.5 m. The trials were planted during
the monsoon rains, in July–October 2009. To assist initial establishment, trials were flood
irrigated twice with about 100 mm of water during the first dry season. Fertilizer was
applied 3, 12 and 24 months after planting, at a rate of 100 g of NPK fertilizer (17:17:17)
per tree at each application. Weeds were controlled twice per year by ploughing between
tree rows, using tractors.
The trial at Kovai was located within a research station where management oper-
ations were implemented to a high standard, whereas the other two sites were on
farmers’ lands. Subsequent to layout and planting of the trials by researchers, man-
agement operations of vegetation control and fertilizer application were conducted by
the farmers themselves.

Measurement and assessment

Survival was assessed at 10 months and 3 years. Tree height and diameter at 1.3 m
(Dbh) were measured at 3 years, and the presence/absence of reproductive material
(buds or fruits) on each tree was noted. An increment core sampler (5.6 mm diameter)
was used to extract 2 core samples at breast-height from over 95 % of the surviving
trees, excepting the smallest trees. Wood density was estimated from one core sample
per tree, using the water displacement method. The other samples were ground to
produce a wood-meal sample for NIR reflectance spectroscopy. Pulp yield and lignin
content were estimated from the wood-meal samples, applying calibration models
developed for E. camaldulensis (Ramadevi et al. 2010). NIR predictions were made for
over 90 % of the surviving trees at Kovai and MBNagar, and over 60 % of the sur-
viving trees at Ongole.

123
72 New Forests (2017) 48:67–82

Wood volumes were estimated approximately for each site. Tree volume over bark was
calculated for each surviving tree from its height and stem basal area at breast height,
applying a form factor of 0.45, as determined for E. camaldulensis plantations in Laos
(Alder 2006). Stand volume at 3 years for each entire trial was then calculated by summing
individual tree volumes of all surviving trees, expressed on a per-hectare basis. Mean
Annual Increment (MAI) for the 3-year trial period was calculated by dividing stand
volume per hectare by 3.

Statistical analysis

A staged approach to data analysis was taken. First, performance of the three types of
planting material (natural provenances, local seed sources and clones) was compared.
Then, for additive genetic parameter estimation, individual-tree models were run on a
subset of the data that excluded the clones. For each of these approaches, individual sites
were first analyzed separately, after which an across-sites analysis was conducted. All
analyses were based on mixed models, implemented with restricted maximum likelihood,
of the following general form:
y ¼ Xb^ þ Zu^ þ e ð1Þ
where y is the vector of observations on the trait, b and u are vectors of fixed- and random-
effect estimates respectively, X and Z are incidence matrices for fixed and random model
terms and e is a vector of random residual effects.

Comparison of planting stock types

To test whether the three planting stock types (natural provenance, local seed source and
clones) differed significantly in their performance at each site, vector b contained sub-
vector estimates for fixed effects of replicate and type effects, and u contained sub-vectors
for the random effects of row, column and plot. For the assessment of planting types across
sites, additional terms for site and the interaction between site and type were included in
b. The across-sites model also incorporated separate estimates of residual variance for each
site in e. The significance of differences between types was tested using Wald-type F tests,
and the maximum standard error of the difference of the treatment (type) means was
determined; this applied to the comparison between local seed sources and clones, which
were less well-represented in the trials than the natural provenances. These analyses were
conducted using the REML directive in Genstat 17 (VSN International, Hemel Hempstead,
UK). Analysis of survival on the individual trees was conducted for model [1] using a
generalized linear mixed model for binomial distributions and the logit link function
implemented using the GLMM directive in Genstat 17.

Additive genetic parameter estimation

Additive genetic parameter estimates were based on individual-tree models taking data
from open-pollinated families only, i.e. the clones were excluded. For the single-site
analyses, vector b contained sub-vector fixed-effect estimates for replicate and individual
provenance or local seed source, and u contained sub-vectors for the random effects of
individual trees as well as design effects for planting row, column and plot. For the across-

123
New Forests (2017) 48:67–82 73

sites estimates, a subvector for site was included in b while a subvector for site-by-
individual-tree interaction was included in u.
As the breeding population contained families from wild provenances and first-gener-
ation clone trials and second-generation SPAs, which are likely to have different levels of
outcrossing, the additive relationship matrix included in b was adjusted to account for this
(Bush et al. 2011). The wild and SPA families were modelled with family-average selfing
rate s = 0.10. The choice of this value was influenced by the Butcher and Williams (2002)
estimate of s = 0.05 for wild E. camaldulensis progenies from north Queensland, and the
estimate of s = 0.12, with no evidence for biparental inbreeding, reported by Varghese
et al. (2009) for a first-generation SPA of E. camaldulensis in southern India. The selected
value is lower than that normally assumed for open-pollinated Eucalyptus progenies, for
which a coefficient of relationship of 1/2.5, corresponding to s = 0.3, as recommended by
Eldridge et al. (1993), is commonly used. Open-pollinated families from the clone trials
were harvested from trees that had been established in plots, each comprising 9 ramets of a
single clone. As the plots were not thinned at flowering time, it is possible that higher
levels of selfing than those in SPAs will have resulted. These families were therefore
modelled with family-average selfing rate s = 0.20.
Observed variance components were used to estimate the causal variance components,
individual-tree narrow-sense heritabilities and trait–trait correlations. Parameters were
estimated as follows:

r^2t ¼ r^2A = estimate of individual-tree (i.e. additive) variance,

r^2plot ¼ estimate of plot ðfamily - by - replicateÞvariance

r^2Phen ¼ r^2A þ r^2plot þ r^2e = estimate of the phenotypic variance where r^2e is the residual
variance of seedling progeny,
2 r^2
h^ ¼ 2 A = the estimate of individual-tree narrow-sense heritability
r^Phen

Individual-tree bivariate models were implemented to simultaneously estimate the

variance for each trait and the covariance between them. The additive genetic correlations
(^rA ) between traits within sites was then estimated as:
r^A1 A2
^rA ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi :
r^2A1 r^2A2

Conceptually, a single trait measured at different sites can be treated as if it were different
traits (Burdon 1977), and correlations between these are termed Type-B correlations (^ rB ).
Univariate, multi-site analyses were conducted using Eq. 1 as specified for the estimation of
genetic parameters with the addition of site as a fixed effect in b to simultaneously estimate
variance components for each site and covariance components between each pair of sites.
Separate error variances for each site were estimated in e. Type-B additive genetic correla-
tions were then estimated between pairs of trials for each trait (Burdon 1977):
r^Ai Aj
^rB ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffi ;
^2Ai r^2Aj
r

where r^Ai Aj is the covariance between additive effects at sites i and j, and r^2Ai and r^2Aj are
the additive genetic variance estimates ate the two sites, respectively. Values of ^ rB closer to

123
74 New Forests (2017) 48:67–82

unity indicate little GxE, while lower values indicate that GxE exists and that rank changes
may be occurring across sites.
All analyses in this section were implemented in ASReml 4 (VSN International, Hemel
Hempstead, UK).

Prediction of genetic gain

Genetic gain in the next generation, relative to the mean of the current generation was
predicted using the following formula (Falconer and Mackay 1996):
% gain ¼ i  CVA  h
where i = selection intensity in units of standard deviation; CVA = coefficient of additive
genetic variation; h = square root of within-provenance, individual-tree heritability.
Across-site estimates of heritabilities and additive coefficients of variation were used in
the gain calculations. It was assumed that superior trees across the three trials were selected
after adjustment for site, replicate and block effects at an intensity of 1 in 100 (2.665 units
of standard deviation) and grafted into a clonal seed orchard. Gains from single-trait
selection for Dbh, density and pulp yield were examined separately.

Results

Performance of natural provenances, local seed sources and clones

Early mortality was substantial at MBNagar where 23 % of trees had died by 10 months;
the corresponding mortalities at Ongole and Kovai were 17 and 9 % respectively
(Table 3). Survival at 3 years remained higher at Kovai (81 %) than at Ongole (63 %) and
MBNagar (66 %). Differences in survival among planting material types at 3 years were
significant at all three sites. The clones had the lowest survival at all locations. The across-
sites analysis showed significant (p \ 0.001) differences in survival among types at both
10 months and 3 years. Across the three sites, at 3 years natural provenances had similar
survival (72 %) to second-generation sources (67 %): both were substantially higher than
the mean survival of the ten clones under test (50 %).
Site mean annual increments in volume over bark (MAI) for the 3-year growth period
were estimated to be 12 m3 ha-1 at Ongole, 10 m3 ha-1 at Kovai and 5 m3 ha-1 at
MBNagar.
Diameter at breast height was highest at Ongole (8.1 cm), followed by Kovai (6.8 cm)
and MBNagar (5.8 cm). There were differences in Dbh among planting material types at
all three sites (Table 3). Local seed sources and clones grew better than natural prove-
nances at Kovai, while at Ongole and MBNagar provenances and seed sources outgrew
clones. In the across-sites analysis, local seed sources (Dbh 7.2 cm) outgrew natural
provenances (6.8 cm) and clones (6.5 cm). Differences among the three planting material
types were significant at each individual site and in the across site analysis (p \ 0.01 or
p \ 0.001). Similar rankings were observed for height, again with significant differences
among types. The interactions between site and type for Dbh and height were also
significant.
Wood properties differed between sites (Table 3). Wood basic density was highest (site
mean 579 kg m-3) and pulp yield lowest (43.7 %) at MBNagar, where tree growth was

123
New Forests (2017) 48:67–82 75

Table 3 Least-squares means for planting material types at three trial sites and for across-site analysis of
types, maximum standard errors of difference of type means, and probabilities of treatment differences for
type and sitetype interaction
Dbh Ht Surv. 10 m Surv. 36 m Density Pulp Lignin
(cm) (m) (%) (%) (kg m-3) (%) (%)

Kovai
Natural provs. 6.7 9.0 92.8 83.9 555 44.9 22.5
Local sources 7.3 9.4 91.3 76.4 549 45.0 22.7
Clones 7.3 9.8 66.0 60.0 572 45.5 22.8
Site mean 6.8 9.1 91.4 81.4 555 45.0 22.5
Max. s.e.d. 0.19 0.16 6 0.22 0.26
p (type) \0.001 \0.001 \0.001 \0.001 \0.001 0.112 0.333
Ongole
Natural provs. 8.1 10.3 83.0 64.0 518 46.5 23.8
Local sources 8.3 10.5 83.1 61.4 515 46.2 24.1
Clones 7.2 10.0 80.0 47.0 500 46.7 22.9
Site mean 8.1 10.4 82.9 62.7 517 46.5 23.8
Max. s.e.d. 0.27 0.18 10 0.32 0.43
p (type) 0.003 0.011 0.78 \0.01 0.048 0.105 0.014
MBNagar
Natural provs. 5.7 8.0 77.8 68.3 578 43.7 23.0
Local sources 6.0 8.4 80.2 64.3 581 43.7 22.8
Clones 5.5 8.1 53.0 44.0 578 44.2 22.9
Site mean 5.8 8.1 77.2 66.4 579 43.7 22.9
Max. s.e.d. 0.19 0.18 10 0.26 0.36
p (type) 0.005 0.007 \0.001 \0.001 0.951 0.029 0.596
Across-site
Natural provs. 6.8 9.1 84.5 72.1 550 45.0 23.1
Local sources 7.2 9.4 84.9 67.4 548 45.0 23.2
Clones 6.5 9.3 66.3 50.3 550 45.5 22.9
Overall mean 6.9 9.2 83.8 70.2 550 45.1 23.1
Max. s.e.d. 0.13 0.24 10 0.16 0.21
p (types) \0.001 \0.001 \0.001 \0.001 0.179 0.002 0.32
p (sitetype) \0.001 \0.001 0.06 0.359 \0.001 0.574 0.046

Surv survival, Pulp pulp yield

slowest, whereas density was lowest (517 kg m-3) and pulp yield highest (46.5 %) at
Ongole, where growth was fastest. Differences among the three types of planting materials
were relatively minor, although in the across-sites analysis the clones had significantly
(p = 0.002) higher mean pulp yield (45.5 %) compared to the provenances and local
sources (both 45.0 %). Significant differences among planting material types were
observed at two of the sites for wood density and one site for lignin, but type rankings were
not consistent across sites, so types did not differ significantly in the across-site analysis for
density or for lignin content.

123
76 New Forests (2017) 48:67–82

Differences in survival, growth, and wood properties among the six natural provenances
were minor, as were corresponding differences among the five local seed sources (Sup-
plementary Table 1), except that families sourced from clone trial 2 had a mean across-site
survival of 54 %, lower than the range of the other four local sources (65–75 %).

Genetic parameters and predicted gain from selection

Narrow-sense heritabilities for growth and wood traits differed among the sites (Table 4).
Heritabilities for height and Dbh were moderate at Kovai and Ongole (0.30–0.38), while at
MBNagar, heritability for height was only 0.14 while that for Dbh was not significantly
different from zero. Across sites, heritabilities were more precisely estimated, with lower
standard errors (0.20 ± 0.04 for height and 0.23 ± 0.03 for Dbh). Among the wood traits,
pulp yield had a broader range of estimated heritability (0.15–0.61) at individual sites than
did wood density (0.16–0.17) and lignin (0.09–0.48), the latter being non-significant at
Ongole and MBNagar. As with growth traits, the across-sites estimates of heritabilities for
wood traits were moderate and more precisely estimated (0.20 ± 0.04 for density,
0.37 ± 0.06 for pulp yield and 0.21 ± 0.05 for lignin).
Families were fairly stable in their performance across sites for all traits under study
(Table 5). Type-B (site–site) genetic correlations between the two inland sites, Kovai and
MBNagar, were close to 1 for height, density and pulp yield, 0.80 for Dbh and 0.74 for
lignin. Type-B correlations between the coastal site, Ongole and the two inland sites were
also close to 1 for density, over 0.8 for lignin and also high for Dbh (0.71–0.76), but lower
(0.61–0.63) for pulp yield.
Genetic correlations between height and Dbh were strongly positive (0.91 or greater) at
all three sites (Supplementary Table 2). Correlations between growth and wood traits were
generally weak and estimated with low precision. There were positive correlations between
Dbh and lignin at Kovai and Ongole. Density was not significantly correlated with growth
traits, pulp yield or lignin. Correlations between pulp yield and lignin were negative and
significant at all sites.
Predicted genetic gains from the progeny of a clonal seed orchard based on single-trait
selection and selecting the best 1 in 100 trees, as a percentage of the population mean were
11 % for Dbh, 4.6 % for basic density and 2.7 % for pulp yield (Table 6).

Reproductive status

At age 3 years, 39 % (323 of 900) surviving trees originating from the three local second-
generation SPAs carried flower buds and/or seed capsules, while only 12.6 % (444 out of

Table 4 Within-source (natural provenance or local seed source), individual-tree heritabilities for Dbh,
height, basic density, pulp yield and lignin for each trial, and across the three trial sites
Trait Kovai Ongole MBNagar Across sites

Ht 0.38 ± 0.08 0.30 ± 0.09 0.14 ± 0.07 0.20 ± 0.04

Dbh 0.36 ± 0.07 0.34 ± 0.09 0.16 ± 0.07NS 0.23 ± 0.04
Density 0.17 ± 0.06 0.16 ± 0.07 0.17 ± 0.07 0.20 ± 0.04
Pulp yield 0.61 ± 0.09 0.45 ± 0.12 0.15 ± 0.08 0.37 ± 0.06
Lignin 0.48 ± 0.09 0.18 ± 0.11NS 0.09 ± 0.07NS 0.21 ± 0.05

NS estimate not significantly different from zero as gauged by the Akaike information criterion

123
New Forests (2017) 48:67–82 77

Table 5 Type-B genetic correlations across sites

Trait Location

Kovai and Ongole Kovai and MBNagar Ongole and MBNagar

Ht 0.56 ± 0.17 0.97 ± 0.27 0.33 ± 0.32NS

Dbh 0.80 ± 0.14 0.74 ± 0.21 0.65 ± 0.25
Density 0.99a 0.99a 0.99a
Pulp yield 0.63 ± 0.16 0.99a 0.61 ± 0.26
Lignin 0.92 ± 0.33 0.84 ± 0.33 0.99a
a
Estimate restrained just below the upper theoretical limit of 1
NS
Estimate not significantly different from zero as gauged by the Akaike information criterion

Table 6 Predicted genetic gain in Dbh, basic density and pulp yield, based on single-trait selection of best
1 % of trees for a clonal seed orchard
Dbh (cm) Basic density (kg m-3) Pulp yield (%)

Mean (across 3 trial sites) 6.85 551 44.8

Across-site additive genetic variance 0.351 451 0.556
Coefficient of additive genetic variation (%) 8.6 3.9 1.7
Across-site heritability 0.23 0.20 0.37
Selection intensity (units of standard deviation) 2.67 2.67 2.67
Predicted gain (% of mean) 11 4.6 2.7
Predicted mean, next generation 7.6 576 46.1

3529) surviving trees from natural provenances were fertile (Petford provenance, which
was not represented at one site, was omitted from this comparison). This difference in
fertility is highly significant (v2 = 145.54, 1 d.f., p \ 0.001). Clones (22.5 % of trees
fertile), and families from the clonal trials (28.9 %), had levels of fertility that were
intermediate between those of SPA families and the natural provenance material. There
were major differences in fertility between the sites, with much higher percentages of
fertile trees at Kovai (21.2 %) and MBNagar (24.6 %) than at Ongole (4.5 %).

Discussion

Growth, survival and wood properties of different planting material types

The estimated MAIs of 10, 12 and 5 m3 ha-1 at Kovai, Ongole and MB Nagar were low,
reflecting the low rainfall and long dry season. Poor survival contributed to the low volume
increments. Trials at MBNagar and Ongole were managed by farmers and overall survival
at 3 years was only 63 and 66 % respectively, whereas at Kovai, located within a breeding
station with good maintenance, survival was 82 %. At all three sites, there was substantial
mortality by age 10 months, and subsequent ongoing mortality to 3 years.

123
78 New Forests (2017) 48:67–82

Growth rates were similar to those observed in three genetic gain trials at dryland sites in
southern India (Varghese et al. 2009) that compared best natural provenances and seed from
first-generation SPAs of E. camaldulensis. Mean annual rainfalls are lower and dry seasons
longer in much of southern India than in some other tropical regions where E. camaldulensis
is planted, for example in Vietnam (Kien et al. 2008) and Thailand (Pinyopusarerk et al.
1996). Annual rainfall at the six trial sites reported by these authors ranged from 1100 to
2300 mm, and growth rates at comparable ages were faster than those reported here. Higher
MAIs at 3 years, in the range 15–25 m3 ha-1, were reported for many individual E.
camaldulensis clones in an irrigated clonal trial on a site with deep alluvial soil at a subtropical
latitude of 31°N in Punjab state, northern India (Lal et al. 2006).
Pulp yield was lowest and wood basic density highest at MBNagar, the site with the
most severe dry season where growth was slowest, while at Ongole, where growth was
fastest, pulp yield was highest and density lowest. There was a 2.8 % difference in pulp
yield and 62 kg m-3 difference in density between these two sites (Table 4). This influ-
ence of site on wood properties is consistent with other studies. For example, air-dry wood
density of 10-year-old E. globulus was significantly higher, 648 kg m-3, at the driest of
three sites in Western Australia, than that at the wettest site, 590 kg m-3 (Downes et al.
2014). In a trial on a dry site in Tasmania, Australia, irrigation of E. globulus and E. nitens
reduced wood basic density by about 50 kg m-3, and increased pulp yield by about 2 %,
compared to that of non-irrigated trees (Downes et al. 2006).
The local seed sources displayed faster growth and better survival, compared to the mean
growth and survival of 10 commercially planted clones. Taking into consideration both growth
and survival, second-generation sources would probably produce slightly greater wood vol-
umes than best natural provenances. Across sites, the mean pulp yield of the clones, 45.5 %,
was slightly but significantly (p \ 0.001) higher than that of the natural provenances and local
seed sources (45.0 %). While significant variation among clones was evident (data not pre-
sented), their representation in these trials, with only 10 ramets per clone planted at each site
and overall mortality of 45 %, was not sufficient to reliably establish rankings of the clones for
growth, survival and wood properties. The mean Dbh and survival of the 17 progenies col-
lected from the two clonal trials, which each incorporated 21 clones including the clones that
were tested, were superior to the mean of the 10 tested clones, at all three trial sites (Sup-
plementary Table 1).

Genetic parameters and predicted genetic gain

The estimated heritabilities for height and Dbh were higher at Kovai and Ongole (range
0.30–0.38) than those typically reported for these traits in eucalypt species (Harwood 2014).
Values at MBNagar (0.14–0.16) were comparable to other estimates for E. camaldulensis.
Heritability estimates across sites for height and Dbh (Varghese et al. 2008) from three dryland
provenance-progeny trials in southern India, based on 188 families from tropical Australian
provenances of E. camaldulensis were much lower (0.07 ± 0.02) than those for the current
study (Table 5). Individual-site heritability estimates for height and Dbh at age 2 years in three
progeny trials in Thailand (Pinyopusarerk et al. 1996) were also much lower, after adjusting for
the different within-family coefficient of relationship assumed in the Thai study. The current
estimates were determined at 3 years whereas these two earlier studies reported heritabilities
determined at only 2 years and tested somewhat different sets of genetic material.
Additive genetic correlations between growth and wood traits were estimated with low
precision and in most cases were not significantly different from zero, despite the large
numbers of families under test and most of the trees in the trials being sampled for both

123
New Forests (2017) 48:67–82 79

growth and wood traits. The absence of strong genetic correlations implies that selection
and breeding to improve growth of E. camaldulensis is unlikely to lead to major shifts in
population means for wood traits. This conclusion is supported by the very similar mean
values for wood traits for the natural provenances and the local SPA seed sources and
clones (Table 4, Supplementary Table 1). Low, non-significant genotypic correlations
between Dbh and wood basic density were reported by Kien et al. (2008) in three clonal
trials of E. camaldulensis in Vietnam. Pulp yield and lignin were predicted from NIR;
although the calibrations used are good (Ramadevi et al. 2010), associated prediction errors
would tend to degrade the estimation of heritabilities for these two traits and genetic
correlations with and between them.
The lower predicted percentage gains in basic density and pulp yield from selection,
expressed as a percentage of the population mean, compared to the corresponding gain in
Dbh, were largely a consequence of the lower coefficients of additive genetic variation for the
wood traits, relative to that for Dbh (Table 6). Expressed as a change in pulp yield (percent of
dry wood weight basis), genetic gain from 44.8 to 46.1 % pulp yield was predicted for single-
trait selection at an intensity of 1 in 100 trees. While an increase of this magnitude would be
valuable to industry it is less than half the difference in site mean pulp yield between Ongole
(46.5 %) and MBNagar (43.7 %). This suggests that companies should evaluate the influence
of the growing environment on the quality of the wood that they source.
Realized genetic gains in growth from breeding would likely be much lower than predicted.
Progeny trials over-estimate genetic gains in stand-level productivity, because competition
effects accentuate differences in family rankings owing to small plot sizes. Realized gain is
typically lower in uniform competitive environments when the highest-ranked genotypes are
deployed in plantations, as demonstrated for clones by Stanger et al. (2011).

Strategies for improving the quality of planting stock and plantation growth
rates

In the dryland environments where the three trials were located, growth rates are limited by
available water. MBNagar, which received only 661 mm of annual rainfall on average and
had a harsh 7-month dry season in each year of the trial period, appears to be too dry for
profitable wood production without irrigation (MAI of only 5 m3 ha-1). Improving wood
mass produced per volume of water used in evapotranspiration, which includes not only
transpiration by the trees but water use by the understory and evaporation from the soil
surface (White et al. 2014) is a key factor. To increase wood production in these sites,
understanding the causes of mortality and growth-limiting factors is critical. Inter-row
ploughing for weed control, as practiced in these trials, may repeatedly disrupt the surface
root systems of the trees (Nambiar and Harwood 2014) and may adversely affect survival.
Alternative methods of weed control (e.g. use of herbicide) warrant investigation.
Regardless of the genetic material planted, sustainable site management practices are
required to maintain productivity over multiple short rotations. Retention of organic matter
and nutrients on-site are critical for improving and sustaining production, especially in
water and nutrient limited environments and short rotation cycles (Laclau et al. 2010;
Nambiar and Harwood 2014).
These trials, together with previously-reported results evaluating the progeny from first-
generation SPAs (Varghese et al. 2009), have confirmed that seed from unpedigreed SPAs
developed from appropriate, broad-based plantings of a mix of many families from the best
provenances of E. camaldulensis can grow faster than both best natural provenances and
the currently available commercial clones in these environments. Planting stock cost is

123
80 New Forests (2017) 48:67–82

another factor for growers to consider. Companies in India sell clonal stock at about Rs 4
per plant, and orchard-derived E. camaldulensis seedlings at Rs 1 per plant, so the addi-
tional cost of clonal over orchard seedling stock, at a stocking of 2200 plants per hectare
would be Rs 6600 (*USD 100 ha-1).
Based on available information (Varghese et al. 2008, 2009) both improved SPA seed and
selected clones should yield significantly greater wood volumes than the local Mysore gum
land race on dryland sites. The very low price of Mysore gum planting stock (Rs 0.25 per
open-rooted seedling) relative to clonal plants and the limited availability of improved seed
from pedigreed seed orchards are major contributing factors to its continued use. Large-scale
seed production from unpedigreed SPAs of broad genetic base derived from the best natural
provenances may help to replace Mysore gum and improve plantation productivity.
The three genetic trials reported here provide a sufficiently broad family-pedigreed base
to select individuals for inclusion in seed orchards, based on their breeding values for
growth, survival and wood traits, and also to select individuals for clonal testing with a
view to clonal deployment. By applying multi-trait index selection (e.g. for both Dbh and
pulp yield) it should be possible for a clonal seed orchard to deliver substantial genetic
gains in both productivity and pulp yield, simultaneously. However, it would be difficult
and expensive to dedicate sufficient land to clonal seed orchards to produce sufficient seed
for large-scale replacement of Mysore gum. An indirect strategy, whereby seed from a
clonal seed orchard is used to establish large areas of unpedigreed SPAs for bulk pro-
duction of seed to deliver improved growth and wood properties, merits consideration.
This study confirmed the observation by Varghese et al. (2009) that trees sourced from
local SPAs flower earlier and more heavily than those from natural provenances. Not all
trees flowered in the first- and second-generation SPAs (Varghese et al. 2009) and there is a
natural tendency for more-fecund trees to dominate pollen pools and contribute propor-
tionally more to seed collections. Consequently there is selection for fecundity as well as
for improved growth during successive breeding cycles. This suggests that second-gen-
eration plantings established from a well-documented, broad genetic base of the best
provenances are more suitable candidates for seed production than are predominantly first-
generation plantings (i.e. plantings from wild provenances). Higher proportions of flow-
ering trees and heavier flowering in second generation stands would yield more seeds and
reduce the likelihood of selfing. It would also enable seed collections to focus on the trees
of best phenotype thus increasing the effective selection intensity for growth that could be
attained in SPAs (Kamalakannan et al. 2007). Locating SPAs in environments conducive to
early, heavy flowering and seed production is also important, as witnessed by the major
differences in fertility between sites observed here.
Clonal forestry brings the advantage that non-additive variation as well as additive gains
from breeding can be exploited, so in principle a set of selected, well-tested clones could
deliver gains in both growth and wood quality greater than those predicted for the CSO
strategy. Some eucalypt clones developed in India have proved particularly susceptible to
pest and disease attack, while others have shown superior resistance (Kulkarni 2013;
Senthilkumar et al. 2013). Novel pests and disease outbreaks may be expected to emerge on
an ongoing basis (Garnas et al. 2012), and it has been argued (Dehon et al. 2013) that clonal
forestry offers the best pathway to risk reduction through breeding and deployment of
resistant interspecific hybrid genotypes. Broad-based seed sources do at least incorporate a
wide range of different genotypes, making complete failure of plantations as a result of pest or
disease outbreaks less likely than for plantations based on one or a very few clones.
The lead time required for clonal testing and the highly significant clone-by-site
interaction demonstrated in southern India make effective clonal deployment challenging

123
New Forests (2017) 48:67–82 81

(Varghese et al. 2008). Clonal eucalypt plantations should be focused on more productive
site types where currently available clones have been well-tested (Kulkarni 2013; Lal et al.
2006), while improved seed from well-documented CSOs and SPAs may be an appropriate
choice for dryland environments where plantation productivy is lower.

Conclusion

Genetic trials of E. camaldulensis in dry environments at three sites in southern India

allowed comparison of three different types of planting material: known best natural
provenances of the species, local seed sources developed from the best provenances, and
commercial clones. At age 3 years, the local seed sources grew faster than the best natural
provenances and grew faster and survived better than the commercial clones. The three
planting material types did not differ significantly in their wood density, although mean
NIR-predicted pulp yield of the ten clones was slightly higher. Unpedigreed seed pro-
duction areas, developed using a broad genetic base of the best provenances could deliver
genetically improved planting material at low cost.

Acknowledgments We thank John Doran, Washington Gapare, Sadanandan Nambiar and two anonymous
reviewers for their reviews of earlier drafts of the manuscript.

References
Alder D (2006) Analysis of the 2006 forest inventory for the Laos Industrial Tree Plantation Project:
Executive Summary, Revision 2.4. LTS International. https://www.researchgate.net/publication/
264873745. Analysis of the 2006 forest inventory for the Laos Industrial Tree Plantation Project
Executive Summary
Boland DJ (1981) Eucalypt seed for Indian plantations from better Australian natural seed sources. Ind For
107:125–134
Burdon RD (1977) Genetic correlation as a concept for studying genotype-environment interaction in forest
tree breeding. Silvae Genet 26:168–175
Bush D, Kain D, Matheson C, Kanowski P (2011) Marker-based adjustment of the additive relationship
matrix for estimation of genetic parameters-an example using Eucalyptus cladocalyx. Tree Genet
Genomes 7:23–35
Butcher PA, Williams ER (2002) Variation in outcrossing rates and growth in Eucalyptus camaldulensis
from the Petford Region, Queensland; evidence of outbreeding depression. Silvae Genet 51:6–12
Dehon G, Resende S, Resende M, Assis T (2013) A roadmap to eucalyptus breeding for clonal forestry. In:
Fenning TM (ed) Challenges and opportunities for the world’s forests in the 21st century. Springer,
Dordrecht, pp 394–424
Downes G, Worledge D, Schimleck L, Harwood C, French J, Beadle C (2006) The effect of growth rate and
irrigation on the basic density and kraft pulp yield of Eucalyptus globulus and E. nitens. N Z J For Sci
51:13–22
Downes G, Harwood C, Washusen R, Ebdon N, Evans R, White D, Dumbrell I (2014) Wood properties of
Eucalyptus globulus at three sites in Western Australia: effects of fertiliser and plantation stocking.
Aust For 77:179–188
Eldridge K, Davidson J, Harwood C, van Wyk G (1993) Eucalypt domestication and breeding. Clarendon
Press, Oxford
Falconer DS, Mackay TFC (1996) Introduction to quantitative genetics. Pearson Education Ltd, Harlow
Garnas JR, Hurley BP, Slippers B, Wingfield MJ (2012) Biological control of forest plantation pests in an
interconnected world requires greater international focus. Int J Pest Manag 58:211–223
Ginwal HS (2014) Eucalypt improvement: efforts and achievements in India. In: Bhojvaid PP, Kaushik S,
Singh YP, Kumar D, Thapliyal M, Barthwal S (eds) Eucalypts in India. Forest Research Institute,
Dehra Dun, pp 117–138
GIT-Forestry (2008) Cultivated Eucalyptus Global Map 2008. http://www.git-forestry.com

123
82 New Forests (2017) 48:67–82

Griffin AR (2014) Clones or improved seedlings of Eucalyptus? Not a simple choice. Int For Rev
16:216–224
Harwood C (2014) Classical Genetics and Traditional Breeding. In: Henry R, Kole C (eds) Genetics,
genomics and breeding of Eucalypts. CRC Press, Boca Raton, pp 12–33
Kamalakannan R, Varghese M, Lindgren D (2007) Fertility variation and its implications on relatedness in
seed crops in seedling seed orchards of Eucalyptus camaldulensis and E.tereticornis. Silvae Genet
56:253–259
Kien ND, Jansson G, Harwood C, Almqvist C (2008) Clonal variation and genotype by environment
interactions in growth and wood density in Eucalyptus camaldulensis at three contrasting sites in
Vietnam. Silvae Genet 59:17–28
Kulkarni HD (2013) Pulp and paper industry raw material scenario—ITC plantation a case study. Q J Indian
Pulp Pap Tech Assoc 25:79–89
Laclau J-P et al (2010) Organic residue mass at planting is an excellent predictor of tree growth in
Eucalyptus plantations established on a sandy tropical soil. For Ecol Manag 260:2148–2159
Lal P, Dogra AS, Sharma SC, Chahal GBS (2006) Evaluation of different clones of Eucalyptus in Punjab.
Ind For 132:1383–1390
McDonald MW, Brooker MIH, Butcher PA (2009) A taxonomic revision of Eucalyptus camaldulensis
(Myrtaceae). Aust Syst Bot 22:257–285
Nambiar EKS, Harwood CE (2014) Productivity of acacia and eucalypt plantations in South-East Asia 1.
Biophysical determinants of production: opportunities and challenges. Int For Rev 16:225–248
Pinyopusarerk K, Doran JC, Williams ER, Wasuwanich P (1996) Variation in growth of Eucalyptus
camaldulensis provenances in Thailand. For Ecol Manag 87:63–73
Ramadevi P, Meder R, Varghese M (2010) Rapid estimation of kraft pulp yield and lignin in Eucalyptus
camaldulensis and Leucaena leucocephala by diffuse reflectance near-infrared spectroscopy (NIRS).
South For 72:107–111
Senthilkumar N, Murugesan S, Thangapandian K (2013) Present status of eucalyptus gall insect, Leptocybe
invasa (Fisher and LaSalle) in Tamil Nadu. Curr Sci 104:1135–1136
Stanger TK, Galloway GM, Retief ECL (2011) Final results from a trial to test the effect of plot size on
Eucalyptus hybrid clonal ranking in coastal Zululand, South Africa. South For 73:131–135
Varghese M, Harwood CE, Hegde R, Ravi N (2008) Evaluation of provenances of Eucalyptus camaldulensis
and clones of E. camaldulensis and E. tereticornis at contrasting sites in southern India. Silvae Genet
57:170–179
Varghese M, Kamalakannan R, Harwood CE, Lindgren D, McDonald MW (2009) Changes in growth
performance and fecundity of Eucalyptus camaldulensis and E. tereticornis during domestication in
southern India. Tree Genet Gen 5:629–640
White DA et al (2014) Managing for water-use efficient wood production in Eucalyptus globulus planta-
tions. For Ecol Manag 331:272–280

123
New Forests is a copyright of Springer, 2017. All Rights Reserved.