Trees

, Volume 27, Issue 4, pp 865–877

Assessing internal epicormic dynamics in Quercus alba L. using CT scanning: the strong effects of shoot development and tree growth relative to progeny level genetic variation

Authors

  • Andrew Meier
    • Department of Forestry and Natural ResourcesPurdue University
    • Department of Forestry and Natural ResourcesPurdue University
Original Paper

DOI: 10.1007/s00468-013-0840-x

Cite this article as:
Meier, A. & Saunders, M.R. Trees (2013) 27: 865. doi:10.1007/s00468-013-0840-x
  • 219 Views

Abstract

Epicormic branches can be a serious silvicultural problem in many Quercus species because of the potential reduction in log value associated with their occurrence. The phenomenon is also problematic for tree improvement since the genetic component of epicormic branching has not been well quantified. The strong influence of ontogeny on epicormic development in Quercus is well established; however, the long-standing assumption that genetic variation also influences epicormics has not been rigorously tested. With trees from two, 25-year-old Quercus alba L. progeny tests in IN, USA, we used computed tomography scanning to characterize internal epicormic development. We sampled trees from upper and lower crown classes of families that had been classified as having low, medium and high numbers of epicormic sprouts. We also measured an array of variables related to growth and competition with the objective of assessing the relative impacts of genetics and vigor on epicormic development. Using generalized linear and linear mixed models, we found that ontogenetic and vigor variables were strongly associated with epicormic structure and development, and that the genetic effect was negligible. The total number of epicormics was most significantly influenced by the number of sequential branches that bore epicormics (p < 0.001) and the proportion of undeveloped epicormics was most significantly influenced by diameter increment (p < 0.001). We propose that a strong focus on individual tree vigor and form in tree improvement could minimize the impact of epicormic branching in Q. alba trees.

Keywords

White oakComputed tomography scanningEpicormic branchesGenetic effectsTree vigor

Introduction

The development of epicormic complexes on valuable oak trees can result in significant declines in log value (Meadows 1995; Fontaine et al. 2004; Miller and Stringer 2004). Many studies have quantified the epicormic branch response to silvicultural treatments in terms of various tree and site characteristics (Ward 1966; Spiecker 1991; Miller 1996; Fontaine et al. 2001; Kerr and Harmer 2001). However, informed decisions about tree quality, especially in terms of epicormics, cannot be made without knowledge of how the current tree condition will change over time. Both silvicultural selection of superior crop trees and phenotypic selection of improved genotypes in tree improvement programs often occur at relatively young tree ages (Gottschalk 1997; Wu et al. 2000); it is therefore imperative to understand the factors in young trees that are indicative of epicormic prevalence at maturity (Meier et al. 2012).

Epicormic branches develop from epicormic buds that, in temperate species, form on an annual shoot but remain dormant for a year or more (Fink 1983; Colin et al. 2012; Meier et al. 2012). This is in contrast to sequential branches, which develop with the expansion of an annual shoot from buds that have been dormant for less than 1 year (Nicolini et al. 2003; Meier et al. 2012). Many tree species, including the oaks, produce epicormic branches almost exclusively from preventitious epicormic buds (Fontaine et al. 1999). These buds can be either primary or secondary; primary buds form directly on an expanding annual shoot while secondary buds descend from a previously developed axillary bud (Fink 1980; Fontaine et al. 2001). Epicormic traces develop in the wood from the annual radial growth of associated epicormic structures (Fink 1980; Fontaine et al. 1999) and are accurate indicators of previous epicormic development (Spiecker 1991).

The number of epicormic branches on a tree fluctuates greatly from year to year, rendering epicormic branch counts at one point in time a poor estimate of future epicormic sprouting (Kerr and Harmer 2001; Morisset et al. 2012b). This is because epicormic branching is ultimately a function of four main developmental factors: epicormic bud establishment, bud persistence, bud sprouting and epicormic branch persistence (Kerr and Harmer 2001; Meier et al. 2012), which are themselves influenced by “ontogeny, environment including silviculture, and genetics” (Colin et al. 2010a). Consequently, any potential environmental or genetic control could act on any of these factors (Colin et al. 2010a; Meier et al. 2012) and accurate predictions of epicormic development are contingent upon a strong understanding of the relative influence of each of these components.

Environmental and silvicultural influences on epicormic branch development are well documented. Vigorous, dominant trees have repeatedly been shown to be less susceptible to epicormic branch formation than suppressed trees (Meadows 1995; Spiecker 1991; Miller and Stringer 2004; Dimov et al. 2006; Colin et al. 2008), and epicormic responses to thinning are largely a function of the epicormic composition (Colin et al. 2010b) prior to thinning (Ward 1966; Miller 1996). Stand density does not appear to have a strong direct influence on epicormic branch development in young oaks (Colin et al. 2008), though the intense competition characteristic of the stem exclusion stage of stand development has been associated with higher epicormic prevalence (Dimov et al. 2006). However, tree to tree variation within and among vigor classes has been noted (Ward 1966; Harmer 1992a; Meier 2012), indicating that current tree vigor is not the sole determinant of epicormic branch development.

One explanation for variation in the magnitude of epicormics among individual trees has been genetic variation within a species (Ward 1966; Miller 1996). The veracity of this assumption has not, however, been rigorously tested (Colin et al. 2010a; Meier et al. 2012). Bowersox and Ward (1968) found a significant difference in the number of epicormic buds that sprouted in stem cuttings from five different Quercus alba clones. However, this finding incorporated only the genetic influence on the epicormic response to severe wounding and abnormal development of epicormic composition. No other studies are available that provide quantitative descriptions of variation in any component of the epicormic composition at the clonal or progeny level in oaks, though a study with Pinus rigida found strong clonal variation (Kuser and Knezick 1985). Heritability estimates of qualitative differences in epicormic branching in Q. petraea and Q. robur progeny tests have indicated some level of genetic control over the epicormic trait (Jensen et al. 1997; Jensen 2000), but an in-depth description of provenance level variation in the epicormic composition in Q. petraea showed only minor genetic influence (Colin et al. 2010a).

Epicormic ontogeny, which is the development of epicormic structures over time, has been the subject of significant research in the last 15 years, revealing the overwhelming influence of past developmental processes on the current epicormic composition and potential in Q. petraea. There have been many important findings. First, epicormic complexes, i.e., clustered groups of individual epicormics structures (Meier et al. 2012), undergo significant development over time (Fontaine et al. 1998, 1999). Second, the epicormic sprouting capacity of a tree at a given time, also known as the epicormic potential, is dynamic and not static (Fontaine et al. 2001). Third, a significant proportion of epicormic buds are associated with sequential branch bases (Fontaine et al. 2001; Colin et al. 2010a). Fourth, the accurate prediction of epicormic development requires the quantification of all epicormic structures, not only epicormic branches (Colin et al. 2010b; Morisset et al. 2012a). Lastly, the current epicormic frequency is driven largely by the epicormic frequency at a very young age (Morisset et al. 2012b, c).

Until recently, descriptions of epicormic ontogeny required either long-term studies tracking individual buds, which have not been done to our knowledge, or laborious manual sectioning of logs to observe the structure of epicormic traces and thereby describe epicormic dynamics (Braham and Kellison 1987; Fontaine et al. 2004). However, the advent of non-destructive imaging technology, which is currently being developed and implemented for maximizing wood utilization in sawmills (Wei et al. 2011; Longuetaud et al. 2012), has facilitated significant advances in the visualization of internal wood structure. One of the most common imaging techniques, computed tomography (CT) scanning, has been used in a wide array of forestry research applications, including studies of wood anatomy (Wei et al. 2011), tree growth (Longuetaud et al. 2006) and dendrochronology (Okochi et al. 2007). The most recent of the studies of epicormic ontogeny in Q. petraea have used a CT imaging approach to comprehensively describe epicormic development within log sections from the pith to the bark (Colin et al. 2010c; Morisset et al. 2012a, b, c). These studies have provided significant advances in our understanding of epicormic dynamics and epicormic anatomy and ontogeny. However, most have had insufficient sample size (i.e., n ≤ 10) to conduct rigorous statistical analysis; Morisset et al. (2010c) is the notable exception with n = 35. Additionally, none of this research provided a quantitative assessment of genetic variation in epicormic dynamics at the progeny level.

Our study sampled a large enough number of trees from two Q.alba progeny tests to allow for comparisons of epicormic development among family groups, specifically variation in epicormic bud establishment and dormancy release. This is the first description of epicormic ontogeny in a North American species and the first study to provide a quantitative assessment of the progeny-level genetic variation in epicormic dynamics over time in any species. Our hypotheses were: (1) both the genetic effect, in terms of family epicormic class, and the vigor effect, in terms of crown class, would be significant factors influencing epicormic bud establishment and bud dormancy release; (2) the influence of growth related variables would introduce variability into the expression of the genetic effect; and (3) epicormic development is an extension of sequential branch development, and that sequential branch development would have a moderate influence on epicormic bud establishment.

Methods

Study trees

White oak logs used in this study were grown in progeny tests planted in the state of Indiana, USA (Meier 2012 for full description). Trees in these tests consist of a total of 70 families that originated from 15 stands in Indiana, Illinois and Missouri. Acorns were collected in 1982 from upper crown class trees of at least average stem quality (Rink and Coggeshall 1995; O’Connor and Coggeshall 2011) and grown in nursery beds for 1 year prior to out-planting at four locations in spring of 1984. Two of these four locations with similar stand histories were used for the current study: the Jasper-Pulaski Fish and Wildlife Area (JP) (41°09′N, 86°54′W) and the Harrison-Crawford State Forest (HC) (38°15′N, 86°15′W). At these locations, trees were planted in row plots of four adjacent trees per family. Progeny test sites were chemically treated prior to planting and in at least one subsequent year to reduce weedy competition. High initial mortality occurred in some areas of the progeny tests (Rink and Coggeshall 1995; O’Connor and Coggeshall 2011) because of excessive moisture; these areas were not replanted (M. Coggeshall, personal communication, 14 Sept 2012). The progeny tests have not been thinned since establishment, though trees at HC were pruned in the mid-1990s to a height of approximately 2.4 m (D. Sieg, personal communication, 30 Aug 2012). Site indices (height at 50 years; Carmean et al. 1989, p. 56) for JP and HC are 20 and 21 m, respectively. Mean dbh in early 2010 at JP was 16.9 cm [standard deviation (SD) = 3.8] and at HC was 14.3 cm (SD = 3.3); average stand basal area (m2 ha−1) was 21.8 (SD = 4.3) for JP and 22.2 (SD = 2.9) for HC.

Individual trees for the current study were removed as part of a crop tree release treatment implemented at both locations. Prior to treatment, families were assigned to epicormic classes (epiC; Meier 2012) based on previously recorded qualitative ratings of epicormic expression (S. Rogers, unpublished data). Families with ratings near or above the upper limit for the third quartile were classified as high epicormic families (epiC:H) while those with values within the first quartile were classified as low epicormic families (epiC:L). Medium epicormic families (epiC:M) were grouped around the mean. Three families were randomly selected from each of these groups for in-depth analysis of internal epicormic dynamics. Within selected families, individuals were further stratified between upper (dominant and codominant; cc:U) and lower crown classes (cc) (intermediate and overtopped; cc:L) in order to select trees of high and low vigor in each epicormic class. Our sample was limited to only those trees removed in the release treatment. It was therefore not possible to insure a uniform number of trees from each combination of epiC and cc (Table 1). Field measurements of growth and density variables (Table 1) were taken from January to March 2010 following tree selection but prior to felling. An approximately 1.2 m section was removed from each felled tree for CT scanning; the base of this section was located at breast height (1.37 m) of the standing tree.
Table 1

Descriptions and mean values for predictor variables and factors that were included in the full models

Variable

Description

Mean value (±SD) or distribution of factors

Factors

 fam

The numbered family for individual trees

see Fig. 6

 epiC

Epicormic class, derived from initial qualitative epicormic ratings from 2006 (L = low, M = medium, H = high)

L (15), M (17), H (15)

 cc

Crown class in 2010 (L = intermediate and suppressed, U = dominant and codominant) (Smith et al. 1997)

L (16), U (31)

 Ftg

The free-to-grow rating of the tree (Perkey and Wilkins 2001) in 2010, ranging from 0 (no lighted faces) to 4 (full sunlight)

0 (25), 1 (13), 2 (9)

 Site

One of two plantation locations, either JP or HC

HC (31), JP (16)

Covariates

 dbh27

Tree diameter (cm) at age 27, prior to the 2010 growing season

15.9 (±3.2)

 dbh14

dbh (cm) at age 14, calculated from scanned images

6.5 (±1.8)

 dbh12

dbh (cm) at age 12, calculated from scanned images

4.7 (±1.6)

 dbh10

dbh (cm) at age 10, calculated from scanned images

2.9 (±1.3)

 dinc2327

Mean annual diameter growth (cm yr−1) between ages 23 and 27

0.4 (±0.3)

 dinc10–14

Mean annual dbh growth rate (cm yr−1) between ages 10 and 14

0.9 (±0.2)

 ht27

Total height (m) at age 27, prior to the 2010 growing season

11.8 (±1.2)

 ht11

Total height (m) at age 11

3.4 (±0.7)

 ht3

Total height (m) at age 3

0.4 (±0.1)

 hinc23–27

Mean annual height growth (cm yr−1) between ages 23 and 27

0.3 (±0.2)

 hinc3–11

Mean annual height growth between ages 3 and 11 (m yr−1)

0.4 (±0.1)

 lc

A measure of live crown size, calculated as the product of mean crown radius (m) by length of live crown (m)

11.2 (±5.4)

 ba

Neighborhood basal area at the tree (m2 ha−1) as an average of the basal area at the location of four adjacent trees and the study tree

22.6 (±3.3)

 seqT

The total number of sequential branch knots

21.2 (±4.3)

 seqE

The total number of sequential branch knots with epicormics

13.8 (±5.4)

 seqDia

Mean maximum diameter(mm) of sequential branch knots

11.5 (±2.2)

 seqDistM

Mean distance to nearest sequential branch per tree (mm)

2.2 (±0.7)

 seqDisCV

Mean coefficient of variation of distance to nearest branch per tree (mm)

93.2 (±24.0)

Variables that were significant in any of the models are listed in bold; variables that were removed because of high collinearity are shown in italics

CT scanning and image analysis

Since high wood moisture content reduces contrast between epicormic traces and surrounding wood (Colin et al. 2010c), logs were allowed to air dry for a summer before log scanning commenced. Logs were scanned using a GE Lightspeed QX/i multi-slice helical CT scanner. Images were collected at 5 mm intervals; each scanned log consisted of approximately 240 images. This scanning interval was chosen because our objective was to sample a large number of logs rather than to develop a detailed reconstruction of internal epicormic structure for a few logs (Colin et al. 2010c; Morisset et al. 2012a, b, c). An initial log scan to compare the efficacy of 2.5 and 5 mm slices resulted in identical counts of epicormic structures; we therefore felt that our sample was representative of the epicormic population in individual logs. Scanner settings were held constant for all logs at 80 kV and 200 mA. Resulting images were analyzed manually using ImageJ software (Schneider et al. 2012) with the cell-counter plugin. Because there was some variability in the number of images captured per log, only the first 220 images were assessed, equivalent to 1.1 m of log length.

For all sequential branches within each scanned log, branch diameter (seqDia; mm) and mean distance between branches (seqDistM; mm) was recorded. All sequential branch knots were tallied as being either without (Fig. 1, Type 1) or with (Fig. 1, Type 2) associated epicormic traces. Epicormic structures were then tallied by their level of development using adjacent scanned images. Epicormic development was assessed based on morphological features in the wood indicative of bud dormancy release, especially sudden widening of the trace. Traces with no expansion were considered undeveloped epicormics (unDE; Fig. 1, Type 3), while those that had widened were considered developed epicormics. Of the developed epicormics, those less than 6.4 mm in diameter were considered to be small developed epicormics (smDE, Fig. 1, Type 4), while those larger than 6.4 mm were considered to be large developed epicormics (lgDE, Fig. 1, Type 5). Finally for each epicormic structure, origin was determined by following the trajectory of the epicormic trace to the most proximal point within the log. If the trace originated at the main pith, it was considered to be a primary trace, whereas if the trace was initiated in the knot of a sequential branch, it was considered to be a secondary trace and the preceding sequential branch was noted as a sequential branch with epicormics (seqE). In instances where the trace’s origin was not visible, the angle of the trace at the most proximal position was extrapolated visually to determine whether it was primary or secondary.
https://static-content.springer.com/image/art%3A10.1007%2Fs00468-013-0840-x/MediaObjects/468_2013_840_Fig1_HTML.jpg
Fig. 1

Examples of two CT scanned log image with tallied internal structures. Type 1: sequential branch knot without epicormics; Type 2: sequential branch knot with epicormics; Type 3: undeveloped epicormic trace; Type 4: small developed epicormic trace (<6.4 mm diameter); Type 5: large developed epicormic trace (>6.4 mm diameter)

These definitions of different epicormic types and epicormic development are based on descriptions of epicormic trace characteristics available in the literature (Fink 1980; Spiecker 1991; Fontaine et al. 1999, 2004; Colin et al. 2010c). The ubiquitous presence of parenchyma rays in oaks makes it nearly impossible to distinguish small diameter traces. We were, therefore, unable to count small epicormics below 1 mm in diameter (Colin et al. 2010c) unless an external structure that extended beyond the vascular cambium, such as a bud, was evident in the scanned image. The age of the pith at the lowest complete slice, which was located approximately at breast height (1.37 m), was estimated by counting growth rings from the cambium; this age was then used to estimate individual tree diameter (cm) in 1993 at age 10 (dbh10), 1995 at age 12 (dbh12) and 1997 at age 14 (dbh14).

Data analysis

We used generalized linear models (glm) to analyze our data since it was either count- or proportion-based; count data most often follows a Poisson related distribution, while proportion data is analyzed as a binomial distribution (Bolker et al. 2009). Because of the potential variation in the data that could be attributed to differences between families (n = 9), we analyzed the data initially in a generalized linear mixed model (glmm) framework with family as a random effect. Site was included as a fixed effect in all full models because, with only two sites in the study, meaningful random effect parameters could not be estimated. For all analyses, significance was assessed at α = 0.05.

Overdispersion can lead to a response variance that does not equal the response mean, the primary assumption of Poisson distribution (Zuur et al. 2009). To determine the appropriate glm data distribution, we tested dispersion as
$$ D = \frac{{\text {Var}_{\text{observed}} \times \left( {n - 1} \right)}}{{\text {Var}_{\text{theoretical}} }} $$
where n = the number of observations. If this test is significantly different from a χ2 distribution, overdispersion is present in the data (Wetherill and Brown 1991) and the data likely follows negative binomial or similar non-Poisson distributions (Zuur et al. 2009). To correct for overdispersion in our data, we used a constant overdispersion negative binomial regression model (NB-1) for count data (Hardin and Hilbe 2007). In our proportion glmm models, overdispersion was initially accounted by introducing an observation level random effect (Browne et al. 2005). When the family random effect did not contribute significantly to explaining model variance, we switched to a quasi-binomial glm model. In the NB-1 and quasi-binomial approaches, the expected value is scaled by a constant dispersion parameter (α for NB-1 and ϕ for quasi-binomial) to calculate model variance (Hardin and Hilbe 2007; Zuur et al. 2009).

Full models were developed based on both field collected data and data collected from log scan image analysis (Table 1). Some variables were not included in the full model because of collinearity with other variables (Table 1, italics), which we defined based on a Pearson correlation coefficient of |r| ≥ 0.7. The full models for the three response variables, the total number of epicormic traces (totTr), the relative proportion of undeveloped epicormics to total epicormic traces (propunDE) and the total number of sequential branches (seqE), were nearly identical with the exception of the response variable.

Stepwise model reduction was undertaken using the AIC criterion and analysis of deviance with a χ2 test for NB-1 models, and an analysis of deviance with an F-test for quasi-binomial models (Zuur et al. 2009). For NB-1, non-significant variables were removed from the models until AIC was minimized. In the interest of model parsimony, complex models were further reduced if the removal of marginally significant parameters did not cause a significant decrease in log-likelihood estimates. AIC values are not well defined for quasi-binomial models; for these, model reduction continued until variable removal caused a significant increase in the residual deviance, indicating that the value of the removed variable in the model was significantly different from zero (Zuur et al. 2009). Analysis of deviance was also used to compare both full and reduced models with and without random effects. In all cases, the family random effects were not significant and it was determined that the simpler model without their inclusion was preferable.

In all models, one tree with no observable epicormic traces was removed from the analysis because it consistently behaved as an outlier. In the totTr model, a second tree with extremely high influence was removed from the analysis because slow growth in the last decade made accurate estimation of pith age at breast height impossible, and therefore the dbh increment from 1993 to 1997 (dbhinc) could not be estimated reliably.

All analyses were performed using the R statistical computing program, version 2.15.0 (R Development Core Team 2012a). Overdispersion tests were carried out using the package qcc (Scrucca 2011), NB-1 glmms and glms were modeled using the glmmADMB package (Skaug et al. 2012) and binomial glmms were developed using the glmer function in the package lme4 (Bates et al. 2012). The function AICtab was used in the package bbmle (Bolker 2012) to compare AIC values for glmms. Binomial and negative binomial glms for creating plots were fitted using the glm function in the stats package (R Development Core Team 2012b) and the glm.nb function in the MASS package (Ripley et al. 2012), respectively.

Results

Total number of epicormic traces

Initial comparisons showed that the mean number of total epicormic traces was similar for the low (\( \overline{x} \) = 46.3, SD = 25.2) and medium (\( \overline{x} \) = 43.9, SD = 20.6) epiC while the high epicormic class was notably greater (\( \overline{x} \) = 61.8, SD = 31.6) (Fig. 2a). Differences between cc were not as evident, though the lower crown class did have a slightly higher mean (\( \overline{x} \) = 56.8, SD = 19.8) than the upper (\( \overline{x} \) = 47.1, SD = 29.3) (Fig. 2b). For both epiC and cc, intra-class variability was quite high. Generalized linear models identified only the interaction of epiC and cc:U as being a significant factor (Table 2); the number of sequential branches-bearing epicormics (seqE) was the most significant predictor (p < 0.001) with a strongly positive relationship between seqE and totTr (Fig. 3a). A slightly positive relationship between totTr and the height in 1986 at age 3 (ht3) was retained in the model (Table 2), though it was only marginally significant (p = 0.073; Fig. 3b); its removal caused a significant increase in the AIC when tested with an anova log-likelihood test. With seqE as a predictor, estimates of totTr (Fig. 3a) were lower for epiC:M (dashed lines) and epiC:L (dotted and dashed lines) in cc:U (black lines) than cc:L (gray lines), though the reverse was true for epiC:H (solid lines). Three of the potential interactions, epiC:M × cc:L, epiC:L × cc:U and epiC:H × cc:L, produced nearly identical predictions (Fig. 3a). With ht3 as a predictor, the pattern of estimates was similar (Fig. 3b), though all prediction lines showed stronger differentiation.
https://static-content.springer.com/image/art%3A10.1007%2Fs00468-013-0840-x/MediaObjects/468_2013_840_Fig2_HTML.gif
Fig. 2

Mean total number of epicormic traces (totTr), total sequential branch knots (seqT) and sequential branch knots with epicormics (seqE) per 1.1 m log section for a epiC and b cc and proportions of undeveloped epicormics (unDE), small epicormics less than 6.4 mm in diameter (smDE) and large epicormics greater than 6.4 mm in diameter (lgDE) for c epiC and d cc. Vertical bars in a and b represent standard errors from the mean. See text, Fig. 1, and Table 1 for further description of these variables

Table 2

Parameter estimates for generalized linear models of the total number of epicormic traces, the proportion of undeveloped traces, and the number of sequential branches with epicormics

Response variable

Dispersion test

Modeled distribution

Parameters

D

p value

Fixed effects

Estimate

SE

p value

totTr

14.06

<0.001

Negative binomial

seqE

0.079

0.007

<0.001

ht3

0.615

0.342

0.073

epiC:L

0.386

0.173

0.026

epiC:M

0.010

0.133

0.937

cc:U

0.196

0.117

0.093

epiC:L × cc:U

–0.606

0.203

0.003

epiC:M × cc:U

–0.348

0.180

0.053

propunDE

23.40

<0.001

Binomial

seqDia

0.094

0.045

0.043

dinc23–27

2.111

0.372

<0.001

epiC:L

–0.493

0.419

0.247

epiC:M

0.171

0.274

0.536

cc:U

–0.471

0.262

0.080

epiC:L × cc:U

1.344

0.482

0.008

epiC:M × cc:U

–0.021

0.394

0.957

seqE

2.08

<0.0001

Negative binomial

seqT

0.053

0.010

<0.001

ftg:1

–0.256

0.105

0.015

ftg:2

0.108

0.100

0.284

D is the test statistic to compare data dispersion to either a Poisson or a binomial distribution; the associated P value indicates whether the dispersion statistic is significantly different from the expected variance

https://static-content.springer.com/image/art%3A10.1007%2Fs00468-013-0840-x/MediaObjects/468_2013_840_Fig3_HTML.gif
Fig. 3

Predicted curves and observed values for the total number of epicormic traces (totTr) on a per meter basis plotted by a: the total number of sequential branches with epicormics per meter (seqE) and b: height in 1986 (hgt86). Symbol type and fill color indicate observed values for the various interactions of epiC and cc, line styles and colors correspond to different predicted curves. In a, the lines for epiC:L × cc:U, epiC:M × cc:L, and epiC:H × cc:L are nearly identical. Curves for the two predictors were drawn using glm submodels of the full reported model that included only the respective covariates. Note that the epiC:L × cc:L line is based on only two points

Number of sequential branches with epicormics

The mean number of sequential branches with epicormics (seqE) was highest for epiC:H (\( \overline{x} \) = 15.4, SD = 5.21). The means for epiC:L (\( \overline{x} \) = 12.7, SD = 5.55) and epiC:M (\( \overline{x} \) = 13.24, SD = 5.25) were similar to each other (Fig. 2a). These trends were not apparent for the total number of sequential branches (seqT), with mean seqT being similar across all cc and epiC (Fig. 2a, b). The reduced glmm model did not include either epiC, cc or their interaction as significant factors in the final model (Table 2), though the free-to-grow rating in 2010 (ftg) was retained. No trees were observed that had more than two free-to-grow quadrants; the difference in the number of seqE between trees with no free quadrants and two free quadrants was not significant, though the seqE value with one free-to-grow quadrant was significantly lower (p = 0.015). A strong positive relationship was evident between seqE was seqT (p ≤ 0.001); all other quantitative predictors were dropped from the model (Fig. 4).
https://static-content.springer.com/image/art%3A10.1007%2Fs00468-013-0840-x/MediaObjects/468_2013_840_Fig4_HTML.gif
Fig. 4

Observed values and predicted curves for the number of sequential branches with epicormics (epiC) regressed on the total number of sequential branches (seqT). Dots represent observed values. Line and symbol styles indicate the corresponding 2010 free-to-grow rating (ftg). Values are plotted on a per meter basis

Proportion of undeveloped epicormics

The proportion of undeveloped epicormics (propunDE) was higher for epiC:L (\( \overline{x} \) = 0.66, SD = 0.23) than for epiC:M (\( \overline{x} \) = 0.48, SD = 0.18) or epiC:H (\( \overline{x} \) = 0.44, SD = 0.13) (Fig. 2c). Mean propunDE was lower for cc:L (\( \overline{x} \) = 0.48, SD = 0.19) than cc:U (\( \overline{x} \) = 0.55, SD = 0.21) (Fig. 2d). When the interaction between epiC and cc was considered in addition to growth and developmental variables, the only interaction significant at α = 0.05 was epiC:L × cc:U (Table 2). In the final model, the magnitude of diameter growth between 2006 and 2010 (dinc 23–27) was the strongest predictor of propunDE (p < 0.001). As growth rate increased, there was a concurrent increase in value of propunDE. Higher mean sequential branch knot diameters (seqDia) also tended to be associated with an increased propunDE (Fig. 5). Generally, trees in the epiC:L had higher growth rates (Fig. 5, dotted bubbles) than epiC:M and epiC:H (Fig. 5, solid gray and solid black bubbles). The mean seqDia for epiC:H was lower than the mean for all trees combined (\( \overline{x} \) = 10.9, \( \overline{x} \)Alltrees = 11.5); means for the epiC:M (\( \overline{x} \) = 11.8) and epiC:L (\( \overline{x} \) = 11.8) were higher than the mean for all trees. The largest sequential branch diameters were found in epiC:M. Trees in the upper crown class (cc:U; Fig. 5, crossed bubbles) had larger sequential branches and higher growth rates than cc:L (Fig. 5, open bubbles).
https://static-content.springer.com/image/art%3A10.1007%2Fs00468-013-0840-x/MediaObjects/468_2013_840_Fig5_HTML.gif
Fig. 5

Bubble plot describing the association between the observed proportion of undeveloped epicormics (propunDE), 4-year dbh change (dbhCH) and the mean diameter of sequential branches (seqDia), as well as the distribution of epiC and cc. The central point within each bubble indicates its exact xy location. Bubble size is scaled by seqDia, with large bubbles being indicative of high seqDia values. For reference, bubble sizes associated with a subset of branch diameters are included on the bottom right corner of the graph. Bubble outline indicates epiC: low, broken black outline; medium, solid gray outline; high, solid black outline. Bubble type denotes cc: lower, open bubble; upper, crossed bubble. One value in which a decrease in dbh of 0.68 cm was recorded was plotted as dbhCH = 0

Family variation

In these models, models including and excluding family as a random effect were never significantly different (data not shown). However, when epicormic class (epiC) was removed from any model, large increases in the variance associated with the family random effect were noted. Variation within some families was quite high, but was moderate in others. Families with high totTr values tended to also be more variable than those with low values (Fig. 6a). Mean family values for propunDE were aligned most closely with epiC (Fig. 6b). Differences among families in seqE were not pronounced (Fig. 6c), with the exception of IL-01-01, which appeared to be somewhat lower. For these three traits, only IL-01-01 and IN-11-02 consistently exhibited values that could be associated with a lower tendency towards epicormic development. However, both of these families were represented solely by trees in the upper crown class (cc:U) (Fig. 6, red interior lines) and trees in epiC:L (Fig. 6, green beanplots) and epiC:M (Fig. 6, red beanplots) tended to be more strongly skewed towards the upper crown class than those in epiC:H (Fig. 6, blue beanplots).
https://static-content.springer.com/image/art%3A10.1007%2Fs00468-013-0840-x/MediaObjects/468_2013_840_Fig6_HTML.gif
Fig. 6

Beanplots of the distribution of a number of epicormic traces, b proportion of undeveloped traces, and c number of sequential branches with epicormics among individual families for the full 1.1 m log section. Outline color of individual plots denotes epicormic class (epiC) and the curvature of the outlines is the estimated kernel density. Solid red and dotted blue lines within the plots represent individual observations and line color indicates the crown class (cc) of each tree. Most colored lines indicate a single observation, but longer lines indicate two nearly identical observations. When line color is split, it indicates that the two observations were from different cc. Thick black lines indicate family means. In c, observations in the final two families corresponded exactly with the family mean. The dotted black line in the background represents the mean for all observations across all families. The letters in family labels indicate the state of origin of the mother tree: IL Illinois, IN Indiana, MO Missouri. Additional information on individual families, including family origin, can be found in Rink and Coggeshall (1995) and O’Connor and Coggeshall (2011)

Discussion

Bud establishment

The initial component of epicormic development is the formation of epicormic buds. In this study, we quantified this component in terms of totTr. We found that the direct influence of both the genetic and vigor components on totTr was much lower than the influence of sequential branches, particularly counts of sequential branches with associated epicormics (seqE). This suggests a statistically significant but minor direct influence of progeny level genetic variation compared to a strong influence of plant developmental dynamics on epicormic bud establishment.

Most studies of bud development in oaks have focused on sequential buds on annual shoots (Collet et al. 1997; Heuret et al. 2003), while there is little information available regarding the association between sequential branching and epicormic bud development. It has long been recognized that buds forming at the base of sequential branches are a component of the epicormic potential (Büsgen and Münch 1929); the presence of minute buds at the bases of branches has been noted in many hardwood species, including Acer saccharum (Church and Godman 1966), Liquidambar styracifula (Kormanik and Brown 1969), Q. rubra (Ward 1964), and Q. robur and Q. petraea (Spiecker 1991). Quantitative estimates show that more than 50% of epicormic buds are located at sequential branch bases in Q. petraea and Fagus sylvatica (Fontaine et al. 2001; Colin et al. 2010a, 2012), and epicormic buds continue to develop along the length of growing sequential branches (Remphrey and Davidson 1992; Colin et al. 2010c) that later migrate to the bole (Büsgen and Münch 1929; Church and Godman 1966; Colin et al. 2010c). As a result, an increase of one sequential branch has the potential to significantly increase the epicormic potential of a tree (Fontaine et al. 2001).

Though there was correlation between the total number of sequential branches (seqT) and seqE (r = 0.68), this did not exceed our collinearity threshold (|r| ≥ 0.70) so the full totTr model included both variables. In this study, the seqT did not have a significant impact on totTr when considered in conjunction with seqE (Table 2). It is intuitive that when more sequential branches develop, the overall value of seqE will increase, but it is apparent from our data that the relationship between seqE and seqT is not constant among all trees. This suggests that variability in epicormic bud establishment associated with sequential branches may influence variation in the overall epicormic potential of a given tree. Additionally, considering the fact that primary epicormic buds form directly on an expanding annual shoot and not on a sequential branch, the correlation between seqE and the number of primary epicormic traces was relatively high (r = 0.40). It seems possible that the same developmental controls regulating epicormic bud set on sequential branches also regulate epicormic bud set on the annual shoot. Our estimates of the total number of epicormic traces were slightly lower than those for Q. petraea (Colin et al. 2010c; Morisset et al. 2012c), though the variability in estimates was remarkably similar. Similarly, the strong influence of early developmental dynamics on subsequent epicormic composition found is this study was also evident in Q. petraea (Morisset et al. 2012a, b).

Sequential branches

There was no indication from our data that the number of sequential branches with epicormics (seqE) was at all influenced by differences in epiC, though our study was not designed to identify differences in sequential branch dynamics between genotypes. Instead, individual trees with high numbers of total sequential branches (seqT) also had the highest seqE. No other studies have quantified seqE but many have quantified changes in seqT in response to different treatments.

We did not expect that the influence of the free-to-grow rating at study establishment in 2010 (ftg; Table 1) would be an important factor in any of our analyses, since it is only a rough measure of current competition. It was particularly surprising that seqE that was significantly lower for trees that were free-to-grow on one side (ftg:1) than for those free-to-grow on zero (ftg:0) or two sides (ftg:2). It may be that the influence of ftg on seqE is related to the smaller sample sizes in ftg:1 (n = 13) and ftg:2 (n = 9) compared to ftg:0 (n = 25). A wider sample could potentially draw the means closer together.

An alternative explanation is related to the influence of competition on relative height growth. In general, these plantations are early in the stem exclusion stage of stand development but many individuals have been relatively free to grow since planting due to areas of high initial mortality (Rink and Coggeshall 1995; O’Connor and Coggeshall 2011). Higher ftg ratings are therefore largely indicative of lower competition throughout stand development in these plantations. Trees in the ftg:0 class would have likely experienced reduced height growth due to lower vigor and smaller crown sizes associated with intense competition. Trees in the ftg:2 class, on the other hand, may have allocated more growth to horizontal rather than vertical crown expansion, as is characteristic for species with weak epinastic control growing without adjacent competitors (Miller 2000). It is possible that seqE is highest in ftg:2 because of more rapid growth and lower associated control over axillary buds. In ftg:0, lower growth rates may have reduced average internode lengths, leading to a higher seqE per unit length.

Bud dormancy release

The ultimate silvicultural and tree improvement concern related to epicormics is the potential for reductions in wood quality. Therefore, from an economic perspective, bud dormancy release, quantified here in terms of the proportion of undeveloped epicormics (propunDE), may be a more important component of epicormic development than the total number of epicormic traces. This is because individual, undeveloped epicormic buds contribute little to log defect (Fontaine et al. 2004) while epicormic development in the form of branch formation or bud proliferation introduces substantial degrade (Meadows 1995; Fontaine et al. 2004). In this study, the trees with the most rapid recent growth (dbhCH) and largest mean sequential branch diameters (seqDia) were those with the lowest proportion of developed epicormic buds. Higher means for both of these variables were associated with trees in the upper crown class (cc:U) and are indicative of higher tree vigor. This suggests that, at least in unthinned stands of young Q. alba, the sprouting of epicormic buds is driven strongly by growth related parameters. The influence of recent growth rates should not be surprising considering that increased epicormic branch production has been associated with low diameter and volume increment (Nicolini et al. 2001; Colin et al. 2008). In Dicorynia guianensis, epicormic branching has even been proposed as an alternative measure of tree vigor since annual rings are lacking in this tropical species (Nicolini et al. 2003).

To our knowledge, the evidence that seqDia also influences the propensity for epicormic buds to sprout has not been previously described in the literature. Sequential branch diameter could be an indicator of greater physiological control over epicormic buds. Larger branches maintain greater leaf area (Zellers et al. 2012) and more vigorous meristems; as these organs develop they are the source for many of the auxins (Aloni et al. 2003) that are responsible for suppression of epicormic buds. It therefore seems possible that epicormic buds on or below these branches would be subject to higher auxin concentrations (Alden 1971; Meier et al. 2012) and less likely to sprout.

Genetics, vigor and ontogeny

There was indication from our data of a minor progeny-level effect on epicormic dynamics, at least when individual families were grouped by epicormic branch characteristics, though there was significant variation in epicormic characteristics within individual families and epiC. In general, epiC:L and epiC:M families did tend to have a lower number of epicormic traces and higher proportion of undeveloped epicormics. Inherent inconsistencies in qualitative methods of epicormic rating, on which our epiC groupings are based, may have led to the selection of families for study that were not the best representatives of contrasts in epicormic dynamics. However, families did appear to be grouped well by epiC for the variable propunDE (Fig. 6), suggesting that the epicormic ratings may have accurately assessed bud dormancy control. The prevalence of trees in the upper crown class in epiC:L and epiC:M also may have contributed to somewhat lower variability within those groups.

From this study, we can definitively say that any genetic effect on epicormic development is complex and highly confounded by the vigor effect. These findings at the progeny level are synonymous with those at the provenance level in Q. petraea (Colin et al. 2010a), but since our trees were open-pollinated progeny, we cannot eliminate the potential effect of unknown father trees, and therefore cannot conclusively counter the assumption that genetic variation strongly influences epicormic development. We also cannot draw range-wide conclusions from this study because we sampled from only nine families; these cannot be considered representative of the entire Q. alba population. Additionally, we could not capture the spectrum of environmental variation from only two sites and epicormic dynamics on other sites may vary.

However, if genetics do influence epicormic bud establishment, the expression may be mediated by genetic variation in other preceding developmental components, such as the tendency to produce multiple flushes (Harmer 1992b), which in turn, would influence sequential branch development and, potentially, epicormic bud establishment (Harmer 2000). Since seqE was a more significant predictor of totTr than seqT, and since there is some tree to tree variation in the proportion of seqE to seqT, genetic variation in the tendency of sequential branches to form epicormic buds may also be present. To our knowledge, this question has never been directly addressed in the literature and quantification of the genetic effect on seqE could provide significant insight into genetic and silvicultural selection of trees with a lower epicormic potential.

Conclusions

The data from this study with Q. alba validates previous studies of Q. petraea that demonstrated epicormic ontogeny and tree growth are much more important factors influencing epicormic development than genetics (Colin et al. 2010a). Our initial hypotheses were partially correct, though the magnitude of relative influences was the reverse of what we expected. Instead of an important effect of epicormic class and crown class with variation introduced by relative growth variables, we observed a strong effect of the growth and development variables with variation introduced by the genetic background of individual trees. Our hypothesis that sequential branch dynamics would influence epicormic dynamics was also verified, though, again, the magnitude of the influence was much stronger than we predicted.

Nevertheless, we feel that more evidence is needed to conclusively determine the impact of genetics on epicormic branch formation, particularly in light of the increasing effort to improve many fine hardwood tree species. For example, tree selection that focuses on growth may concurrently select against epicormic branching, not because of a superior genotype in regards to epicormics, but because better formed phenotypes tend to form fewer sequential branches and, in turn, fewer epicormic buds. Additionally, in faster growing phenotypes, a greater proportion of the buds that do form will fail to develop into epicormic complexes. However, improvement programs that rely only on the selection of superior genotypes while failing to implement treatments to maintain crown vigor are likely to continue to have problems with epicormic branching (Meadows 1995). In the longer term, these improvement programs must be able to account for the relative influence of genetics and tree vigor on sequential branch and flushing dynamics, and subsequent influence on epicormic potential, when making selections. Multi-decadal studies tracking individual epicormic structures in plantings with known family backgrounds are needed to assess the genetic effect on the evolution of the epicormic composition for these improvement efforts.

Acknowledgments

We would like to thank the Indiana Department of Natural Resources—Division of Forestry for access to the progeny tests from which plant material was obtained. In particular, we would like to thank Phil O’Connor for allowing us to use the white oak progeny tests in this study, and Dwayne Sieg at Harrison-Crawford State Forest and John Karstens at the Jasper-Pulaski State Tree Nursery for logistical support. Dr. Mark Coggeshall provided valuable insights into the history of the plantations. Jake Dyer, Daniel Moscosco and Sebastian Saenz helped with the installation of treatments and data collection. The log scanning portion of the project would have been impossible without the assistance of Dr. James Naughton and Donna Tudor at the Purdue University School of Veterinary Medicine. Finally, we would like to thank Drs. Charles Michler and Rado Gazo, and 2 anonymous individuals for reviewers of earlier versions of this manuscript. This project was funded by the Fred M van Eck Foundation for Purdue University, the Northern Research Station of the U.S. Forest Service, and the Hardwood Tree Improvement and Regeneration Center at Purdue University.

Copyright information

© Springer-Verlag Berlin Heidelberg 2013