1124
Correlated responses of height increment and components of increment in 2-year-old
1
Douglas-fir
ZEKI
KAYA
2
Department of Forest Science, Oregon State University, Corvallis, OR
97331,
U.S.A. R. K. CAMPBELL
Pacific Northwest Forest and Range Experiment Station, United States Department of Agriculture, Forest Service, Corvallis, OR 97331, U.S.A. AND w.
T.
ADAMS
3
Department of Forest Science, Oregon State University, Corvallis, OR
97331,
U.S.A. Received March 29, 1989
Accepted May 5, 1989
KAYA, Z., CAMPBELL, R. K., and ADAMS, W.T. 1989. Correlated responses of height increment and components of
increment in 2-year-old Douglas-fir. Can. J. For. Res. 19: 1124-1130.
The consequences for growth and phenology of early selection for height or its growth components were evaluated
in 160 open-pollinated families of Douglas-fir
(Mirb.) Franco) from southwestern Oregon. Seed­
lings from two inland and two coastal populations (40 families each) were grown for two growing seasons in a common
garden. Predicted response to selection suggests that risk of low juvenile-mature correlation and maladaptation with
early selection would be less in the inland than in the coastal region. A phenological event that influences a common
growth pattern seems to account for the difference in response. Early bud set in the 1st year was genetically correlated
with larger overwintering buds in seedlings from both inland and coastal regions. These larger buds yielded a large
increment of predetermined growth in the 2nd year, followed by little or no free growth and early bud set. Seedlings
with late bud set in the 1st year had the converse pattern. Inland seedlings set buds much earlier on the average than
did coastal seedlings; hence seedlings from the two regions had different growth patterns. Risks that can attend early
selection for height generally would be decreased in both regions by selecting for predetermined growth, but several
qualifications are discussed.
(Pseudotsuga menziesii
KAYA, Z., CAMPBELL, R. K., et ADAMS, W.T. 1989. Correlated responses of height increment and components of
increment in 2-year-old Douglas-fir. Can. J. For. Res. 19: 1124-1130.
Les consequences d'une selection precoce, effectuee au niveau de Ia hauteur ou de ses composantes, sur Ia croissance
et Ia phenologie du Douglas taxifolie
(Mirb.) Franco) du sud-ouest de !'Oregon ont ete evaluees.
Les semis de deux populations de l'interieur et de deux populations cotieres (40 families chacune) furent cultives pen­
dant deux saisons de croissance dans le meme jardin. La reponse escomptee de Ia selection suggere que le risque inhe­
rent d'une faible correlation juvenile-adulte et d'une mesadaptation resultante de Ia selection precoce serait moindre
a l'interieur qu'en region cotiere. II semble qu'un evenement phenologique influencant un patron de croissance commun
re
soit responsable de Ia difference de reponse a Ia selection observee. L'aoiitement precoce de Ia 1 annee etait geneti­
quement correle avec Ia formation de gros bourgeons d'hivernage chez les semis de l'interieur de meme que chez ceux
de Ia region cotiere. Ces gros bourgeons resulterent en une forte augmentation de Ia croissance predeterminee de Ia
2e annee, suivie ou non d'une faible croissance non-predeterminee et d'un aoiitement hiitif. Les semis a aoiitement
re
tardif Ia 1 annee eurent un patron phenologique reciproque. En general, les semis de l'interieur ont forme leurs
bourgeons beaucoup plus tOt que ne le firent les semis de Ia region cotiere; les semis des deux regions avaient done
des patrons de croissance differents. Les risques inherents a une selection precoce en egard a Ia hauteur seraient moindres
dans chacune des regions si celle-ci etait pratiquee au niveau de Ia croissance predeterminee; plusieurs conditions sont
toutefois requises et sont discutees.
[Traduit par Ia revue]
(Pseudotsuga menziesii
Introduction
Early culling of individuals or families in tree-breeding
programs is often based on height increment alone. The rela­
tionship of early growth with later growth potential and
adaptability, however, has not been thoroughly explored.
Part of the difficulty of early testing is that patterns of
juvenile stem growth in conifers may differ from those in
mature trees (Cannell et a/. 1976).
Annual shoot extension can be separated into two com­
ponents: predetermined growth and free growth (Cannell
1Paper No. 2358, Forest Research Laboratory, Oregon State
University, Corvallis, OR.
2Present address: Department of Biology, Middle East
Technical University, 06531, Ankara, Turkey.
3 Author to whom all correspondence should be addressed.
Printed in Canada I lmprimC au Canada
1978). Predetermined growth (PG) results from expansion
of the primordial shoot contained in an overwintered bud,
whereas free growth (FG) results from expansion of primor­
dia initiated in the same growing season. Free growth may
follow predetermined growth without interruption (termed
continuous growth (CG) in this paper) or may occur as a
result of multiple flushing (lammas growth, LG).
Free growth may contribute substantially to height incre­
ment in young trees, but becomes increasingly less of a factor
in the growth of older trees; the extent of its role in juvenile
growth and the age at which it no longer is a significant fac­
tor varies with species and geographical source within species
(Pollard and Logan 1976; Cannell and Johnstone 1978). In
a study of shoot-growth patterns in trees from Rocky
Mountain provenances growing in southern Michigan,
KAYA ET AL.
Bongarten (1978) found that 8-year-old spruce (Picea
pungens Engelm.) no longer produced FG; in contrast, up
to 20o/o of the height increment in 15-year-old Douglas-fir
(Pseudotsuga menziesii (Mirb.) Franco) was derived from
FG, although this percentage varied among provenances.
If FG is the primary criterion of height superiority in
young trees, early selection may not identify individuals with
superior PG unless the genetic relationship between potential
for FG and PG is strong. Thus, one risk of early testing is
that genotypes with greatest growth potential beyond the
juvenile stage may not be identified. A second problem
occurs if height extension is genetically correlated with the
length of the vegetative growth period, as it appears to be
in Douglas-fir (Rehfeldt 1983; Campbell 1986). In this case,
early selection may choose families with longer growth
periods, increasing the risk of damage by early or late frost
(Rehfeldt 1979, 1983). One goal of selection might be to
obtain the largest increase in height increment with the least
change in vegetative growth period. Different degrees of
genetic correlation may exist among phenological traits and
height and its components. Thus, by selecting for com­
ponents of height, appreciable gains in height increment
might be achieved with little indirect change in phenology.
A first step in evaluating the implications of early selection
might be to examine the genetic relationships among height
growth, its components, and phenological traits in a single
annual increment. In this paper we report on these relation­
ships in the 2nd-year height increment (HI) of coastal
Douglas-fir (P. menziesii var. menziesil) families from
populations in southwest Oregon. We also examine the
effects of selection for HI on the components of increment
and growth period by estimating their correlated responses
to selection. Predicted changes in sizes of growth com­
ponents or growth period when selection is based on an
increment will reflect the relative degrees of genetic deter­
mination of components, as well as the genetic correlations
among components and total increment. Finally, because
separating PG and FG would be extremely labor intensive
on any kind of operational scale, we examined the relation­
ship between PG and apical bud size. In some species, the
number of needle primordia produced in the overwintering
bud and, therefore, the size of the bud, are correlated with
the amount of predetermined growth in the following year
(Clements 1970; Cannell 1974; Cannell and Willett 1975;
Cannell et al. 1976).
Variation among populations in the occurrence of
multiple flushing has been reported in coastal Douglas-fir
(Campbe11 1986; Loopstra and Adams 1989). In southwest
Oregon, a large difference was discovered between a region
(coastal) west of the Cascade Ranges and a region (inland)
between the coastal and Cascade Ranges (Loopstra and
Adams 1989). Thus, for this study, we sampled and com­
pared populations from the coastal and inland regions.
Materials and methods
The experimental procedure involved four steps: (i) sampling
1125
step in more detail because of its particular relevance to this paper.
Sampling procedures
Two populations were sampled from both the coastal and inland
regions in southwest Oregon. All four populations occurred within
42°N latitude, but the coastal populations were only 16 km from
the Pacific coast, whereas the inland populations were 66 and
166 km, respectively. One population within each region was from
a low elevation (coastal, 152-457 m; inland, 305-762 m) and one
from a high elevation (coastal, 457-762 m; inland, 1067-1372 m).
On average, the summer season precipitation in the coastal popula­
tions (34 em) was nearly twice that in the inland population (18 em).
Seeds were obtained from 40 parent trees in each population, with
Jots identified by parent.
Experimental methods
Open-pollinated offspring of parent trees (families) were planted
as newly germinated seeds in two nursery environments: a regularly
watered "moist" environment, and a "dry" environment, where
water was withheld for approximately 4 weeks in the first growing
season and 8 weeks in the second. In each environment, five­
seedling row plots of 160 families (4 populations X 40 families
per population) were randomly allocated to plot location in a
complete block design with three replicates (15 seedlings per family
in each environment). Ten traits expressing timing of the vegetative
cycle, total height increment, and components of height increment
were measured (Table 1). Family means for each trait were
estimated from the surviving trees in the family ( n
28.4).
To distinguish PG from FG in the 2nd year, we marked shoot
tips of seedlings with a black felt pen 3 weeks after bud burst
(Wiihlisch 1982). By 3 weeks, all needle primordia of the over­
wintering bud apparently had expanded; the mark remained on
needles for the growing season, and, in seedlings with LG, the top
of the mark always corresponded to the location on the shoot of
the temporary scales of the lammas bud. In other words, the ink
marked the last of the needles formed as primordia in the over­
wintering bud. It did not mark the first of the bud scales formed
as primordia in the lammas bud. Shoot growth from the bud scale
scar of the overwintering bud to the pen mark therefore was PG;
growth from the mark to the base of the final terminal bud was
FG. Free growth apparently was produced either by LG or by CG,
but not by both. Thus, seedlings with multiple flushing produced
FG only by LG. The growth occurring beyond the mark, but not
from an expanding lammas bud, was CG. Continuous growth as
defined in this study is equivalent to the "free growth" of
Bongarten (1978). We believe, as did Bongarten, that CG results
from expansion of primQrdia formed during spring growth from
the overwintering bud. A microscopic examination to test this
hypothesis was beyond the scope of the experiment.
The error in determining PG by the marking procedure probably
was not large. The certainty that we correctly measured PG was
greater in the inland populations than in the coastal populations.
Lammas growth made up a large proportion of FG in the inland
populations (see Results), and the presence of LG exactly defines
PG. In coastal populations, CG made up a large proportion of
FG, so PG had to be determined by the marking procedure. The
correlation between PG and size of the overwintering bud, however,
was equally strong in the two regions. Strong and consistent
correlations would be unlikely if PG had been poorly measured.
Statistical analysis
Knowledge of genetic correlation and heritabilities is required
the two regions; (ii) estimating genetic variation in the relevant
seedling traits by growing open-pollinated families in common
to evaluate the effect of selection for one trait on another trait.
Genetic correlations measure the extent to which two traits are
controlled by the same genes, whereas phenotypic correlations result
gardens; (iii) estimating genetic and phenotypic correlations among
seedling traits; and (iv) estimating correlated responses in com­
from the influence of both environmental and genetic causes of
association. Genetic correlations are estimated from components
ponents of HI and phenology when total increment was selected.
The first three steps are sketched briefly in a following section;
of genetic variance and covariance (Falconer 1981) substituted into
the standard equation for the product-moment correlation coef­
ficient. Heritabilities are estimated from components of variance.
Kaya (1987) provides a thorough treatment. We describe the final
CAN. J. FOR. RES. VOL.
1126
19, !989
TABLE I. Description of traits
BSET-1
BUDD
BUDH
BBUR
PO
LG
co
FG
HI
BSET-2
Trait
Units
Date of bud set in first growing season; recorded
weekly
Diameter of terminal buds in February; measured
with an engineering template
Height of terminal buds in February
Date of bud burst in second growing season;
recorded twice weekly
Predetermined shoot growth in second growing
season
Lammas growth in second growing season
Continuous growth in second growing season
Free growth (LG or CO) in second growing season
Height increment in second growing season
(PO + FG)
Date of bud set in second growing season,
recorded weekly
Days after
Jan. 1
Classes
(1-10)
mm
Days after
Jan. I
mm
mm
mm
mm
mm
Days after
Jan. I
"Number refers to first (!) or second (2) growing season.
Large standard errors, therefore, usually are associated with the
estimates of genetic correlation and heritability. The size of the
error depends on the number of genetic entries (open-pollinated
families in this case) and on the number of individuals within
families (Becker I984).
Preliminary analyses of our data indicated that standard errors
were too large for our purpose if based on the families within one
population; populations within regions could not be compared with
confidence because type II error was large. Sums of squares for
families within the two populations in a region therefore were
pooled. In addition, since family x environment interactions were
generally small or lacking (Kaya 1987), 4 data from the two envi­
ronments were treated as replicates by pooling the effects of
replicates within environments. Standard errors of genetic
parameter estimates were substantially reduced by pooling of envi­
ronments and families within populations.
The two regions were analyzed separately by the model
f( p); eifk
Zijk
= P- + r(th + Pi +
+
[1]
is the mean performance
where fL is the experimental mean,
in the kth replica­
of the ith family (j(p)) in thejth population
and
is the experimental error.
tion within environment
Estimates of components of variance and covariance were
obtained by equating mean squares (and cross products) to their
expectations and solving equations. Family heritabilities were
estimated according to Namkoong (1979) and their standard errors
as described by Becker (1984). Genetic correlations were calculated
according to Falconer (1981). Phenotypic correlations, on a family
mean basis, were estimated as in Foster
(1984). Standard
errors of both genetic and phenotypic correlations were computed
according to Becker (1984).
Z;Jk
(r(t)),
eijk
(p)
et a!.
Estimation of correlated response
Correlated response (CR) after selection quantifies the expected
effect of selection for a trait (y) in one generation on the perfor­
in the next generation (Falconer 1981).
mance of another trait
It is calculated as
(x)
[2]
CRx = i
i
..J7i'f: .J/iJ, Rgxy IJp x
h;
h;
where is the selection intensity,
and
are the heritabilities
of traits
and y, respectively, Rgxy is the genetic correlation
x
he family within populations X environment interaction was
significant
< 0.05) in only 2 of the 10 traits: BBUR and BUDH.
In both of these cases, the family within populations variance was
at least 1. 6 times greater than the estimated interaction variance.
(P
between the traits, and apx is the phenotypic standard deviation
of trait x. Direct response, on the other hand, is the expected effect
of selection for a trait
in one generation on the average
performance of the same trait
in the next generation. It is
calculated as
(x)
[3]
Rx =
(x)
ih; IJpx
The consequence (C) of selection for one trait on the performance
of a second trait can be calculated as a ratio of correlated to direct
response. The ratio is the expected genetic gain (CRx) in the trait
of interest
that is obtained by selecting for another trait (y),
divided by the expected gain (Rx) when selection is done directly
on trait
Consequence measures the change in performance after
indirect selection as a proportion of the change expected given direct
selection at the same intensity. In this paper, we will only be
is calculated as
concerned with selection among families, so
(x)
x.
C
where
traits.
h; and h; are the family heritabilities for their respective
Results
Genetic variation
All traits varied genetically (P < 0.05) among regions
(Kaya 1987} and, with the exception of CG in the inland
region, also varied among families within regions (Table 2).
Seedlings from inland populations set their final buds
approximately 27 days earlier than seedlings from coastal
populations in both growing seasons (Table 3). In the second
season, seedlings from inland populations burst bud only
1 day earlier, on average, than seedlings from coastal
populations. Estimated family heritabilities of phenology
traits (bud set in the 1st and 2nd year, BSET-1 and BSET-2,
and bud burst, BBUR) were always somewhat lower in the
inland than in the coastal region (Table 3).
In inland seedlings, bud height (BUDH) and bud diameter
(BUDD) were about twice those of coastal seedlings
(Table 3). As with the phenology traits, family heritabilities
for bud size were moderate and somewhat lower in the
inland region (Table 3). The HI of inland families was about
180Jo smaller than that of coastal families (Table 3). In inland
seedlings, FG made up 23% of HI and LG made up 66%
of FG. In coastal seedlings, FG made up 39% of HI, with
LG and CG contributing about equally to FG. Except for
KAYA ET AL.
1127
TABLE 2. Analyses of variance, degrees of freedom (df), and mean squares for
height growth and phenology traits in inland and coastal regions
Sources of variance
Environment
Replications
within
environments
Populations
Families
within
populations
Error
Inland region
BSET-1
BUDD
BUDH
BBUR
PG
LG
CG
FG
HI
BSET-2
df0
9 555.0
17.3
12.8
2 295.9
20 990.7
6 414.8
246.4
4 146.9
6 478.2
8 703.2
1
6.7
83.0
437.2
983.0
266.1
149.6
548.1
558.6
4
BSET-1
BUDD
BUDH
BBUR
PG
LG
CG
FG
HI
BSET-2
dfa
6 463.1
3 273.2
3.7
4.1
2 999.4
3.5
18.4
6 477.8
1.2
11
3
6
33
2
12 093.2**
1.1
18.9**
701.0**
8 803.8**
8 618.8**
467.1*
13 098.7**
43 379.9**
13 575.9
l
254.8*
0.7**
1.6**
18.9**
789.6**
191.5**
95.2
225.8**
1252.2**
378.22**
142.1
0.3
0.9
6.5
421.4
117.4
96.5
126.9
679.3
214.3
392
78
Coastal region
26 844.1
5 123.3
1 061.6
1 520.6
15 587.1
5 733.3
1
84.4
14 531.2
765.6
9 956.2
12 267.0
50 709.5
1 336.1
4
12 299.0**
16.7**
73.4**
1.2
17 710.4**
42.8
7 629.9**
8 816.0**
1 535.6
28 182.4**
1
392.8**
0.5**
2.1**
22.0**
572.3**
399.8*
755.0**
750.5**
1396.4*
583.6**
74
137.2
0.2
0.8
6.4
384.6
289.3
392.3
343.6
965.5
256.3
375
NoTE: All analyses were carried out on plot means. See Table I for description of traits.
"Missing plots caused by mortality resulted in reduced error df for both regions and a loss of
for families within populations in the coastal region.
•p < 0.05. ••p < 0.01. CO, family heritabilities for components of HI were similar
for seedlings from the two regions (Table 3). Inland families
did not differ significantly in CO (Table 2), perhaps because
very little CO was expressed in inland families in this experi­
ment (Table 3).
Correlations among traits
Genetic and phenotypic correlations were estimated
separately for inland and coastal regions (Tables 4-6).
Genetic correlations consistently were larger than phenotypic
correlations, but also had larger standard errors. Coeffi­
cients for genetic and phenotypic correlations were closely
correlated; only genetic correlations will be discussed
further.
Since BUDD and BUDH were strongly correlated in
inland and coastal regions (Table 4), discussion will be
limited to BUDD. Seedlings with large overwintering buds
in the 1st year generally produced more PO in the 2nd year.
Strong positive genetic correlations between BUDD and PO
occurred in seedlings from inland and coastal regions
(Table 4). In inland families, BUDD in the 1st year was not
correlated with the amount of FG in the next year, but
coastal families with large buds in the 1st year had less FG
{and less LG and CO) in the 2nd year.
Early BSET-1 was associated with larger BUDD in both
inland and coastal regions. The correlation was stronger in
4
df
coastal than in inland seedlings (Table 5). Families with early
BSET-1 also had more PO in the 2nd year; the correlation
was again stronger in coastal than in inland families.
Families with later BSET-1 produced more FG in the
2nd year.
Timing of BBUR g-enerally was poorly correlated with
components of HI in the second growing season (Table 5).
In inland seedlings, however, more LG occurred in seed­
lings with early BBUR. In seedlings of both regions, the
relative amounts of PO and FG strongly influenced BSET-2.
Families with more PO set final buds earlier. The correlation
was slightly stronger in coastal than in inland seedlings.
Delayed bud set provides more time for free growth. Thus,
genetic correlations between BSET-2 and FG were strongly
positive in inland and coastal regions (Table 5). In inland
seedlings, this strong association can be attributed mainly
to LG, because LG and BSET-2 were strongly correlated,
whereas correlation between CO and BSET-2 was weak or
lacking. In coastal seedlings, the strong relationship depends
on CO, because CO and BSET-2 were strongly correlated
but LG and BSET-2 were not.
Correlations between total HI and its components
depended on the population. In seedlings from inland
populations, HI was more strongly correlated with PO than
with FG (Table 6). In those from coastal populations, HI
was more strongly correlated with FG than with PO.
1128
CAN. J. FOR. RES. VOL. 19, 1989
2
TABLE 3. Estimated means (X) and family heritabilities (h ) of apical bud size,
vegetative growth period, and 2nd-year height growth components in seedlings
from inland and coastal regions
Inland
x
BSET-1 (days)
BUDD (classes)
BUDH (mm)
BBUR (days)
PO (mm)
LO (mm)
CO (mm)
FO (mm)
HI (mm)
BSET-2 (days)
271.59
1.96
4.11
105.47
80.31
15.93
8.38
24.31
104.61
177.67
Coastal
h2
0.44
0.50
0.44
0.66
0.47
0.39
x
(0.16)
(0.16)
(0.16)
(0.15)
(0.16)
(0.16)
Ll6
2.36
106.23
74.85
22.40
0.44 (0.16)
0.46 (0.16)
0.43 (0.16)
24.64
47.04
121.88
205.47
298.82
Both
h2
0.65 (0.16)
0.58 (0.17)
0.60
0.71
0.33
0.28
0.48
0.54
0.31
0.56
(0.16)
(0.16)
(0.17)
(0.17)
(0.16)
(0.16)
(0.17)
(0.16)
x
285.20
1.56
3.23
105.85
77.74
19.15
16.51
35.34
113.28
191.36
NoTE: See Table I for description of traits. BSET-1, BBUR, and BSET-2 are expressed as days after
January 1 and BUDD as classes measured with an engineering template. Standard error of heritability
is in parentheses. -, variance among families within populations was not significant.
TABLE 4. Estimates of genetic (Rg) and
phenotypic (Rp) correlations (SE in
parentheses) between apical bud size
(BUDD) and height growth components
in the 2nd year
BUDD
Inland
Coastal
BUDH
Rg
Rp
PO
Rg
Rp
LO
Rg
Rp
co
Rg
Rp
0.93 (0.07)
0.81 (0.02)
1.02 (0.03)
0.92 (0.01)
0.81 (0.12)
0.66 (0.04)
0.78 (0.18)
0.55 (0.05)
0.01 (0.26)
0.03 (0.06)
0.23 (0.30)
-0.06 (0.06)
-0.08 (0.07)
-0.34 (0.06)
-0.61 (0.17)
0.38 (0.05)
-0.52 (0.19)
FO
Rg
0.01 (0.25)
Rp
HI
0.08 (0.06)
Rg
Rp
0.66 (0.17)
0.49 (0.05)
-0.08 (0.28)
0.07 (0.06)
NOTE: See Table l for description of traits.
variance among families was not significant.
Growth consequences of selecting for height increment
Predicted effects of selection for HI on components of
increment differed in populations from the two regions
(Table 7). In inland seedlings, selection for HI indirectly
provided an expected genetic gain in PG that was 91 OJo of
the gain expected by direct selection (Table 7). Indirect gain
in FG would be about two-thirds as large as for PG. All
gain in FG is from an expected change in LG; inland families
did not vary significantly in CG.
Selection for HI in coastal families yielded a predicted
indirect gain in PG only 360Jo of that expected from direct
selection. Predicted indirect gain in FG was almost 60%
larger than in PG. Most of the gain in FG would come from
a response in CG, because the change in LG would be
expected to be small.
Phenological consequences of selecting for height
Selecting directly for HI in inland families produced more
predicted gain in HI than did indirect selection based on
either PG or FG (Table 7). Phenology of inland seedlings
would be altered only slightly. Matings among such selec­
tions would yield progeny with slightly earlier BSET-1
(Table 7) and earlier BBUR; BSET-2 would not be affected.
Selecting for PG instead of HI should produce about 93%
as much gain in HI as would be obtained by direct selec­
tion. BBUR would hardly be affected, but BSET-1 and
BSET-2 would be earlier. Selecting for FG, on the other
hand, would produce only about 63% as much gain in HI
as would direct selection. BBUR would be earlier and BSET
later, especially in the 2nd year.
Direct selection for HI among coastal families would pro­
duce progeny with earlier BBUR and later BSET. As well
as being inefficient for increasing HI, selecting for predeter­
mined growth would produce offspring with earlier BBUR,
BSET-1 and BSET-2. Selection for FG, an efficient way to
improve HI, shoul<:J. produce progeny with no change in
BBUR but with later BSET-1 and BSET-2.
Discussion
The main purpose of this study was to evaluate whether
early selection for height in Douglas-fir of western Oregon
and Washington might lead to later problems. Two kinds
of risk are involved: (i) early selection for height will not
select genotypes for taller trees at later ages (type 1); and
(ii) selection for height will adversely change the vegetative
growing period (type 2).
The extent of risk may depend partly on a height growth
pattern that is qualitatively common among all seedlings but
differs between regions in timing. In both regions, seedlings
with early BSET-1 produced larger overwintering buds.
These larger buds expanded in the second season to yield
a large increment of PG, followed by little or no FG and
early BSET-2. Seedlings with late BSET-1, in contrast, had
small buds, small amounts of PG, larger amounts of FG,
and late BSET-2. Earlier average BSET in inland seedlings
accounted for the difference between regions. This variation
among regions in growth pattern is not unique to the coastal
KAYA ET AL.
1129
TABLE 5. Estimates of genetic (Rg) and phenotypic (Rp) correlations (SE in parentheses) between height
growth components and phenological traits in inland and coastal regions
Inland
BSET-1
BUDD
Rg
Rp
PG
Rg
Rp
Coastal
BSET-2
BSET-1
BBUR
BSET-2
-0.53 (0.19)
-0.41 (0.05)
-0.97 (0.05)
0.83 (0.02)
0.21 (0.18)
-0.13 (0.06)
-0.84 (0.12)
-0.60 (0.04)
0.43 (0.24)
-0.22 (0.06)
-0.56 (0.20)
-0.42 (0.05)
-0.33 (0.23)
0.23 (0.06)
0.67 (0.22)
0.16 (0.29)
BBUR
-0.45 (0.19)
0.07 (0.19)
-0.46 (0.05)
O.o7 (0.06)
-0.29 (0.24)
0.18 (0.06)
-0.05 (0.20)
-0.18 (0.28)
-0.47 (0.20)
0.05 (0.06)
-0.32 (0.06)
0.80 (0.14)
0.65 (0.04)
0.12 (0.28)
0.02 (0.06)
0.14 (0.27)
0.02 (0.07)
0.86 (0.17)
0.13 (0.06)
0.11 (0.07)
0.71 (0.15)
0.47 (0.05)
-0.07 (0.20)
0.34 (0.06)
-0.05 (0.06)
0.48 (0.05)
0.19 (0.25)
0.17 (0.06)
0.26 (0.20)
-0.21 (0.06)
0.83 (0.13)
0.67 (0.04)
0.73 (0.14)
0.48 (0.05)
0.01 (0.19)
-0.03 (0.06)
0.89 (0.09)
0.73 (0.04)
-0.002 (0.26)
0.11 (0.06)
0.35 (0.27)
0.21 (0.24)
0.09 (0.06)
-0.17 (0.06)
0.08 (0.06)
0.43 (0.05)
LG
Rg
Rp
CG
Rg
Rp
FG
Rg
Rp
HI
Rg
-0.16 (0.26)
-0.15 (0.20)
Rp
-0.07 (0.06)
-0.16 (0.06)
0.34 (0.06)
0.42 (0.26)
0.26 (0.06)
NoTE: See Table 1 for description of traits. -, variance among families was not significant.
TABLE 6. Estimates of genetic (Rg) and phenotypic (Rp)
correlations (SE in parentheses) between height growth components
and 2nd-year height increment
HI
PG
Inland
PG
Rg
Rp
LG
Rg
Rp
CG
Rg
Rp
FG
Rg
Rp
HI
Rg
Rp
1.00
1.00
0.13 (0.27)
0.15 (0.06)
Coastal
1.00
1.00
-0.39 (0.39)
-0.13 (0.07)
Inland
Coastal
0.92 (0.04)
0.91 (0.01)
0.37 (0.31)
0.68 (0.04)
0.54 (0.21)
0.46 (0.06)
0.09 (0.41)
0.18 (0.08)
0.77 (0.17)
0.63 (0.04)
0.10 (0.31)
0.23 (0.06)
0.15 (0.06)
0.36 (0.06)
0.28 (0.23)
0.30 (0.31)
0.06 (0.06)
0.64 (0.04)
0.65 (0.04)
0.78 (0.12)
0.28 (0.06)
0.92 (0.04)
0.91 (0.01)
0.37 (0.31)
0.68 (0.04)
1.00
1.00
1.00
1.00
0.77 (0.13)
NoTE: See Table 1 for description of traits. -, variance among families was not
significant.
variety of Douglas-fir. When interior Douglas-fir was grown
in Michigan, both LG and CO varied among provenances
(Bongarten 1978).
If selecting for PO rather than FG incurs less type-1 risk,
as hypothesized by Logan and Pollard (1975) and Cannell
and Johnstone (1978), then risk is lower in some populations
and environments. In the environment of this experiment,
PO contributed most to variation in increment of inland
seedlings and FG to variation in increment of coastal seed­
lings. Type-1 risk, therefore, should be greater in coastal
populations, if the hypothesis is true. Early selection for
height is also likely to cause less type-2 risk in inland than
in coastal populations. In inland populations, predicted
BBUR and BSET were changed very little by selection. Selec­
tion in coastal populations added FG to HI, but also con­
siderably extended the vegetative period.
In this experiment, environment had little differential
effect on growth components. However, differential effects
are routinely achieved in forest nurseries (Lavender 1984).
Both test environment and population, therefore, may con­
dition the risk of early selection. A test environment
minimizing risk would induce PO at the expense of FG.
Second-year growth can be limited by restricted watering
(Duryea 1984), but whether stressing the seedling reduces
FG or PO is not clear. Patterns of growth observed in this
experiment indicate that any treatment inducing early
BSET-1 will reduce FG in the second growing season.
The implication of our finding for Douglas-fir breeding
in the Pacific Northwest depends on the validity of the
hypothesis that FG is limited to a transitory juvenile period.
Our study supplies no information on this point. General
observation suggests that " ... with increasing age, increas­
ingly larger parts of the shoot elongate in the form of
predetermined growth until eventually all shoot growth
develops as predetermined growth in the adult tree"
(Wi.ihlisch and Muhs 1986). In Douglas-fir, the role of FG
beyond the seedling stage is far from understood. Whereas
LG decreased with age in coastal Douglas-fir from British
Columbia and was much more frequent on trees less than
10 years old (Walters and Soos 1961), FG still made up a
significant proportion of the annual height increment of
several interior Douglas-fir provenances at age 15
(Bongarten 1978).
The implication of our results also depends on the
breeding population. In breeding populations derived from
inland sources, early selection for height should be quite
effective, unless progeny evaluation sites encourage FG.
Evaluation and selection in restrictive sites, however, may
produce strains unable to utilize the longer growing seasons
on favorable sites. In breeding populations from coastal
sources, early height may be a poor predictor of later per­
formance, particularly if FG is restricted to a few early years.
Selecting for early height superiority also may lengthen the
CAN. J. FOR. RES. VOL.
1130
y
TABLE 7. The consequence of selecting for trait
on the
performance of trait x, measured as the ratio of the correlated
response to direct response resulting from direct selection for x
19, 1989
A.B. Squillace. School of Forest Resources and Conservation,
University of Florida, Gainesville. pp. 313-318.
CANNELL, M.G.R., and JOHNSTONE, R.C.B. 1978. Free or
lammas growth and progeny performances in
Silvae Genet. 27: 248-254.
CANNELL, M.G.R., and WILLETT, S.C. 1975. Rates and times at
which needles are initiated in buds on differing provenances of
and
in Scotland. Can. J. For.
Res. 5: 367-380.
CANNELL, M.G.R., THOMSON, S., and LINES, R. 1976. An
analysis of inherent differences in shoot growth within some
north temperate conifers.
Tree physiology and yield improve­
ment. Edited by M.G.R. Cannell and F.T. Last. Academic Press,
New York. pp. 173-205.
CLEMENTS, J.R. 1970. Shoot responses of young red pine to
watering applied over two seasons. Can. J. Bot. 48: 75-80.
DURYEA, M.L. 1984. Nursery cultural practices: impacts on seed­
Forest nursery manual: production of bareroot
ling quality.
M.L. Duryea and T.D. Landis. Martinus
seedlings.
Nijhoff/Dr. W. Junk Publishers, Boston. pp. 143-164.
FALCONER, D.S. 1981. Introduction to quantitative genetics.
2nd ed. Longman, Inc., New York. Chap. 19.
FOSTER, G.S., CAMPBELL, R.K., and ADAMS, W.T. 1984.
Heritability, gain, and C effects in rooting of western hemlock
cuttings. Can. J. For. Res. 14: 628-638.
KAY A , Z. 1987. Genetic variation in shoot growth patterns of
Douglas-fir populations from southwest Oregon. Ph.D. disser­
tation, Oregon State University, Corvallis, OR.
LAVENDER, D.P. 1984. Plant physiology and nursery environ­
Forest nursery
ment: interactions affecting seedling growth.
M.L.
manual: production of bareroot seedlings.
Duryea and T.D. Landis. Martinus Nijhoff!Dr. W. Junk
Publishers, Boston. pp. 133-141.
LOGAN, K.T., and POLLARD, D.F.W. 1975. Mode of shoot
growth in 12-year-old black spruce provenances. Can. J. For.
Res. 5: 539-540.
LOOPSTRA, C.A., and ADAMS, W.T. 1989. Patterns of variation
in first-year seedling traits within and among Douglas-fir
breeding zones in southwest Oregon. Silvae Genet. In press.
NAMKOONG, G. 1979. Introduction to quantitative genetics in
forestry. USDA For. Serv. Tech. Bull. No. 1588. Chap. 3.
POLLARD, D.F.W., and LOGAN, K.T. 1976. Inherent variation in
"free" growth in relation to numbers of needles produced by
provenances of
Tree physiology and yield
improvement.
M.G.R. Cannell and F.T. Last.
Academic Press, New York. pp. 245-251.
REHFELDT, G.B. 1979. Ecological adaptations in Douglas-fir
var.
populations. I. North Idaho
and north-east Washington. Heredity, 43: 383-397.
1983. Genetic variability within Douglas-fir populations:
implications for tree improvement. Silvae Genet. 32: 9-14.
WALTERS, J., and Soos, J. 1961. Some observations on the
relationship of lammas shoots in the form and growth of
Douglas-fir seedlings. Faculty of Forestry, University of British
Columbia, Vancouver. Res. Pap. No. 40.
WDHLISCH, G. VON. 1982. By what method should predetermined
and free growth be measured? Proceedings of the IUFRO Joint
Meeting of Working Parties on Genetics about Breeding
Strategies Including Multiclona:l Varieties, 6-10 Sept. 1982,
Sensenstein. Lower Saxony Forest Research Institute,
Staufenberg-Escherode, Federal Republic of Germany. p. 172.
WOHLISCH, G. VON, and MUHS, H.-J. 1986. Influence of age on
sylleptic and proleptic free growth of Norway spruce seedlings.
Silvae Genet. 35: 42-47.
Picea sitchensis.
Selected trait
x
BSET-1
BBUR
PG
LG
CG
FG
HI
BSET-2
Inland
HI
FG
PG
Trait
y
Coastal
Inland
Coastal
Inland
Coastal
-0.40
0.22
1.00
0.42
-0.08
-0.23
0.38
0.51
0.19
-0.21
0.27
1.25
0
1.00
0.63
0.84
0.67
0.01
-0.38
0.78
1.16
1.00
1.02
0.87
0.15
-0.18
0.91
0.59
0
0.65
1.00
0
0.51
-0.32
0.36
0.09
0.62
0.59
1.00
0.56
NoTE: See Table 1 for description of traits.
vegetative period considerably in seedling stages, increas­
ing the probability of damage by frost or drought.
In coastal breeding populations, selecting for predeter­
mined growth would seem to be a desirable way to improve
height while decreasing risk. There is a danger, however.
Free growth is probably a juvenile character, even in coastal
Douglas-fir; even so, FG contributes some unknown quan­
tity, perhaps a significant proportion, to height at harvest.
If this contribution is large, selecting for increased predeter­
mined growth may select indirectly for less FG (Table 7) and
smaller height at harvest.
The ramifications of selecting for components of height
can be tested directly in long-term tests. Fortunately, both
PG and FG of Douglas-fir vary genetically, and it should
be possible to manipulate environments to emphasize either
type of growth. If early selection for PG is found to be
appropriate, our results suggest that bud size can be used
to select indirectly for PG. Bud size and PG were strongly
correlated and bud size is less expensive to measure,
especially in seedlings in which FG is continuous and
therefore easily confused with PG.
Picea sitchensis
In
In
Edited by
In
Edited by
Picea mariana. In
Edited by
Acknowledgments
Financial support for this research was provided by the
U.S. Department of Interior, Bureau of Land Management,
and the U.S. Department of Agriculture, Forest Service,
under the auspices of the Southwest Oregon Forestry Inten­
sified Research Program (grant No. PNW-80-85).
BECKER, W.A. 1984. Manual of quantitative genetics. 4th ed.
Academic Enterprises, Pullman, WA. pp. 57, 121-122.
BONGARTEN, B. 1978. Genetic and environmental variation in
shoot growth and other traits of blue spruce
Ph.D. thesis, Michigan State University, East Lansing.
CAMPBELL, R.K. 1986. Mapped genetic variation of Douglas-fir
(Picea pungens).
to guide seed transfer in southwest Oregon. Silvae Genet. 35:
85-96.
CANNELL, M.G.R. 1974. Production of branches and foliage by
and
provenance
young trees of
differences and their simulation. J. Appl. Ecol. 11: 1091-1115.
1978. Components of conifer shoot growth.
Proceedings
of the 5th North American Forest Biology Workshop,
C.A. Hollis and
13-15 March 1978, Gainesville, FL.
Pinus contorta
Pinus contorta
Picea sitchensis:
In
___
Edited by
(Pseudotsuga menziesii
___
glauca)