Estimation of genetic parameters, provenance performances, and genotype by environment interactions for growth and stiffness in lodgepole pine (Pinus contorta)

ABSTRACT

Genetic parameters, performance of provenances, and genotype by environment interaction (G × E) for diameter at breast height (DBH), survival, and modulus of elasticity of time-of-flight (MOE_tof) (an indirect measure of stiffness), were investigated in six lodgepole pine progeny trials, aged 33–36 years, within three breeding zones in northern Sweden. Provenances of Yukon origin had the highest growth but lowest stiffness at higher latitude, while those of British Columbia (BC) origin grew faster at lower latitudes and had highest stiffness within zone 5. Combined-site heritability estimates ranged from 0.09 to 0.19 for DBH, from 0.19 to 0.27 for MOE_tof, and from 0.13 to 0.26 for survival. Type-B genetic correlations ($r_b$) were generally high for all studied traits, except for DBH and survival in zone 4 ($r_b$ = 0.74 and 0.40, respectively) and for MOE_tof in zone 2 ($r_b$ = 0.46). On the basis of the results obtained in this study, G × E for stiffness in northern Sweden and unfavourable growth-stiffness genetic correlation should be considered in selective breeding programmes of lodgepole pine. To achieve the highest stiffness for lodgepole pine, provenances of Yukon origin should be planted at lower latitudes and those of BC origin should be planted at lower elevations within the tested breeding zones.

KEYWORDS:

• Genotype by environment interaction
• lodgepole pine
• Pinus contorta
• genetic parameters
• acoustic stiffness
• provenance performance

Introduction

Lodgepole pine (Pinus contorta Dougl.) is native to North America, where it covers more than 26 million hectares of forest land in western Canada and the United States (Critchfield 1980). Its natural distribution area falls between latitude 31° N in Baja California and 64° N in the Yukon in Canada. It grows from sea level to 3900 m of elevation (Koch 1996). Lodgepole pine is the major commercial species in its natural habitat, and its wood and fibre properties are suited for the production of high-quality lumber and pulp-wood products. Due to its rapid growth and wide range of environmental tolerance, it has been introduced as an exotic species for multiple purposes in both northern and southern hemisphere (Elfving et al. 2001).

Large-scale establishment of the inland variety of lodgepole pine (P. contorta var. latifolia) plantations in Sweden was initiated during the 1960s and 1970s by transferring seed and breeding materials from provenances of lodgepole pine in Canada mirroring similar climate regions in northern Sweden (Hagner 1971, 1993). Early observations revealed that provenances that were obtained from interior British Columbia (BC) and Yukon performed best (Lindgren et al. 1993). It was also discovered that in contrast to the native Scots pine (Pinus sylvestris L.), lodgepole pine performs better with a 2–5° northward transfer compared to its latitudinal range in Canada (Hagner and Fahlroth 1974; Elfving et al. 2001). The significant breeding programme for lodgepole pine in Sweden was started in the late1970s. The intention was to make Sweden self-supporting with genetically good lodgepole pine base material and seed sources. The programme included plus-tree selection in western Canada, genetic testing of open-pollinated families in field tests in Sweden, and establishment of six seedling seed orchards, one for each orchard zone (Ericsson 1994).

According to results of Swedish experimental plantations, lodgepole pine is superior over Scots pine in terms of growth, productivity, damage resistance, and survival (Elfving and Norgren 1993; Elfving et al. 2001; Fries et al. 2017). Despite such advantages, stem breakage and natural mortality due to wind and snow is a major problem in lodgepole pine plantations (Hansson and Karlman 1997; Elfving et al. 2001). This may be attributed to a greater biomass of foliage (Norgren 1996), poor silvicultural practices which has caused root deformation and later instability (Rosvall 1994), or lower stiffness and lower bending strength of the stem (Andersson 1987). However, such drawbacks can be avoided through proper silvicultural practices and incorporation of wood quality traits, particularly wood stiffness, as an important trait in advanced breeding programmes.

Wood stiffness, measured as its modulus of elasticity (MOE), is the most important trait for construction timber. However, direct measurement of the bending MOE, known as static MOE, is destructive, time consuming, and expensive (Gapare et al. 2010). Recent works have shown that measuring stiffness of standing trees based on acoustic waves is highly efficient, and there is a growing interest to use acoustic tools in tree improvement programmes for wood quality traits (Wang et al. 2007; Lenz et al. 2013; Chen et al. 2015). Additionally, recent study by Hayatgheibi et al. (2017) reported very high genetic correlation between MOE measured acoustically and those obtained based on SilviScan technology using two trials of lodgepole pine (Hayatgheibi et al. 2017).

The Swedish breeding strategy for lodgepole pine is generally similar to that of Norway spruce (Picea abies L. Karst.) and Scots pine, except that lodgepole pine appears to be less sensitive to transfers as well as the genetic conservation of lodgepole pine within the breeding programme is of lower priority than Norway spruce and Scots pine. In addition to production, quality, and general resistance to biotic and abiotic stresses, another important breeding objective for lodgepole pine, is adaptation and preparedness for any future climatic change (Danell et al. 1993).

The importance of using seed sources adapted to their plantation environments for reforestation has long been recognized (Wu and Ying 2004, Wang et al. 2006). Climate is considered as a force of natural selection leading to local adaptation and differentiation of tree populations along climatic gradients. In the face of rapid climate change, it is becoming more important to construct and allocate deployment populations to environments where they will be well adapted (Rehfeldt et al. 2006; Wang et al. 2010). Further, a detailed knowledge of genetic control of breeding objective and selection criteria traits (Wu et al. 2008), levels of genotype by environmental interactions (hereafter referred to as G × E), and genetic correlations between traits is required to implement an effective breeding strategy (Wu et al. 2007; Silva et al. 2014).

In forestry, when G × E occurs, the phenotypic response of genotypes vary when grown under different climatic and edaphic conditions (Bridgwater and Stonecypher 1978). The lack of consistency in the response of genotypes across environments may be due to either change in the ranking of genotypes and/or alterations in the scale (Lynch and Walsh 1998).

The rank-change interactions complicates the design of breeding and deployment strategies as tree breeders must decide whether to select for stability of performance, whereby the rate of population improvement is slow, or to select genotypes for specific environments to maximize genetic gain in that site, but with greater programme costs (Namkoong et al. 2012).

In trees, G × E has been extensively studied, followed by the recent review (Li et al. 2017), and significant G × E has been reported, primarily for growth traits, in nearly all commercially important species such as radiata pine (Pinus radiata D. Don) (Wu and Matheson 2005; Baltunis et al. 2010), loblolly pine (Pinus taeda L.) (Li and McKeand 1989), and Norway spruce (Chen et al. 2017).

To date, several studies have investigated genetic control of growth, survival, form traits (Fries 1986; Ericsson et al. 1994), and wood quality traits (Hayatgheibi et al. 2017) in Swedish lodgepole pine plantations. However, G × E for growth and wood quality has not been yet quantified in lodgepole pine.

This study was the first to make use of all available growth and wood stiffness data from six lodgepole pine progeny trials in Sweden to explore G × E for these traits. The specific objectives were to (1) estimate heritability for tree diameter, survival, and wood stiffness; (2) estimate genetic and phenotypic correlations between tree diameter and wood stiffness; (3) evaluate possible genetic gain and correlated genetic response for diameter and stiffness; (4) investigate G × E for tree diameter, survival, and wood stiffness using six progeny trials within three breeding zones in northern Sweden; (5) compare performance of provenances for growth and stiffness within and among breeding zones.

Materials and methods

Study materials and measurements

The Swedish base breeding population of lodgepole pine originated from natural stands distributed in the interior of Yukon, BC, and Alberta in western Canada, locations well-matched with the harsh climatic conditions which occur in much of northern Sweden. Trials within breeding zone two (Mader and Jovan) originated from four provenances in Yukon (Carmacks, Frances Lake, Watson Lake, and Whitehorse), trials within breeding zone four (Övra and Hemmingen) originated from three provenances in BC (Fort St. John, Prince George, Fort Nelson) and one provenance in Yukon (Watson Lake), and trials within breeding zone five (Lagfors and Hemmesmark) originated from two provenances in BC (Fort St. John and Prince George), were selected for this study (Figure 1(A,B)). Trials within each breeding zone included ∼200 common half-sib families, while there were no common families among breeding zones. Detailed characteristics of the trials were described in Table 1.

Figure 1. Locations of six lodgepole pine progeny trials (within three breeding zones) in northern Sweden (A) and their seed origins (provenances) distributed in Yukon and British Columbia (BC) in Western Canada (B).

Table 1. Site descriptions for six lodgepole pine progeny trials assessed for DBH, MOE_tof and survival.

	Mader	Jovan	Övra	Hemmingen	Lagfors	Hemmesmark
Site_ID	F816367	F816366	F806373	F806371	F796374	F796375
Breeding zone	2	2	4	4	5	5
Year of establishment	1981	1981	1980	1980	1979	1979
Latitude	65° 47′ N	64° 58′ N	63° 57′ N	64° 45′ N	62° 46′ N	64° 09′ N
Longitude	19° 00′ E	17° 18′ E	16°46′ E	19° 38′ E	17° 08′ E	20° 19′ E
Elevation (m)	490	465	350	285	220	160
Soil	Sandy-silty moraine	Sandy-silty moraine	Sandy-silty moraine	Sandy-silty moraine	Sandy morain	Sandy morain
Year of DBH-MOE measurement	2014–^dSep. 2013	2014–Sep. 2013	2014–July 2014	2015–June 2014	2015–June 2014	2015–Sep. 2014
Total no. of provenances	4^a	4^a	4^b	4^b	2^c	2^c
Total no. of families	211	212	178	179	214	214
Total no. of sampled trees and families for DBH	1943-211	5547-212	4329-178	4662-179	6289-214	3129-214
Total no. of sampled trees and families for MOEtof	1174-108	1241-102	1250-100	1242-101	1332-115	1325-111
Total no. of sampled trees and families for survival	5022-211	8640-212	5490-178	5489-179	6904-214	4594-214

Note: MOE_tof: modulus of elasticity of time of flight; DBH: diameter at breast height.

^aCarmacks, Frances Lake, Watson Lake, Whitehorse.

^bFort Nelson, Watson Lake, Fort St. John, Prince George.

^cFort St. John, Prince George.

^dSeptember.

The experimental design for each trial was a randomized complete block (RCB) layout with five replicates. Each family was planted in 10-tree and 12-row plots in each block. Tree spacing was 2 m between rows and 1.5 m within rows.

The response variables measured for each tree included three traits assessed at the ages 33-36 years: diameter at breast height (DBH), survival (1 = alive trees, 0 = dead trees), and acoustic velocity (AV) on standing trees. The Hitman ST300 tool (Fiber-gen, Christchurch; New Zealand) was used to measure the acoustic velocity of standing trees by inserting two sensor probes (transmitter and receiver) into the outer wood of the tree. The distance was measured with a laser beam and an acoustic wave was passed through the stem by striking the transmitter probe with a steel hammer. More detailed description of AV measurements has been given in (Hayatgheibi et al. 2017).

The dynamic modulus of elasticity (MOE_tof) of standing trees can be obtained using the AV according to the Young’s equation (Wang et al. 2001):(1) $\mathrm{MOE}=\rho v^2$(1) where $\rho$ is the green density (kg/m³) (Bucur 2006), $v$ is the velocity of the wave (m/s) and the constant green density of 1000 kg/m³ is usually used for calculating dynamic MOE in standing trees (Lenz et al. 2013). In this study, the constant green density of 1000 kg/m³ was used.

Climate data

Climate information for the six trials is shown in Table 2. Monthly mean temperature and precipitation for each trial from the year of planting to the year of DBH measurement were extracted from a high-resolution (4 × 4 kilometre grid size) climate data base in Sweden (http://luftweb.smhi.se/). Mean annual temperature (MAT), mean annual precipitation (MAP), mean annual precipitation as snow (MASF) (mean precipitation of temperature <0°), mean temperature and precipitation during growing season (April to October) (MTGS, MPGS, respectively), mean temperature and precipitation during summer (May to August) (MTSUM, MPSUM, respectively), and mean temperature in the coldest months (January to March) (MTCM) were calculated.

Table 2. Climate information (from 1979 to 2015) for six lodgepole pine progeny trials within three breeding zones located in northern Sweden.

Zone	Trial	MAT (°C)	MAP (mm)	MASF (mm)	MTGS (°C)	MPGS (mm)	MTSUM (°C)	MPSUM (mm)	MTCM (°C)
2	Mader	–0.2	681	262	6.7	463	10.3	309	–10.4
2	Jovan	1.2	684	229	7.7	468	11.2	315	–8.6
4	Övra	1.2	728	257	7.7	490	11.1	316	–8.6
4	Hemmingen	1.3	666	237	7.8	444	11.4	287	–8.5
5	Lagfors	2.8	716	237	8.82	472	12	288	–6.4
5	Hemmesmark	2.7	763	283	8.68	475	11.9	288	–6.6

Notes: Mean annual temperature (MAT), mean annual precipitation (MAP), mean annual precipitation as snow (MASF) (mean precipitation at temperature <0°), mean temperature during growing season (April to October) (MTGS), mean precipitation during growing season (April to October) (MPGS), mean temperature during summer (May to August) (MTSUM), mean precipitation during summer (May to August) (MPSUM), mean temperature in the coldest months (January to March) (MTCM).

Statistical analyses

Genetic analysis was done in two stages: individual- and combined-site analysis using ASReml statistical software package (Gilmour et al. 2009). The mean DBH and mean MOE_tof among provenances at six progeny trials were compared using one-way ANOVA and Tukey HSD test (Sokal and Rohlf 1995).

Single-site analysis

Firstly, a single-site univariate analysis was performed to estimate the genetic variance components and individual-tree narrow-sense heritability estimates for DBH, MOE_tof, and survival. The family linear mixed-effects model (model 1) fitted to these data was as follows:(2) $y_{jklm}=\mu +P_k+B_j+F_{l\left(k\right)}+B{F}_{jl\left(k\right)}+e_{jklm}$(2) where y is the vector of observation on the m-th tree from the l-th family within the k-th provenance in the j-th block, µ is the overall mean, $P_k$ and $B_j$are the fixed effects of the k-th provenance and the j-th block, respectively. The variable $F_{l\left(k\right)}$ is the random effect of the l-th family within the k-th provenance and $B{F}_{jl\left(k\right)}$is the random interactive effect of the j-th block and the l-th family within the k-th provenance and$e_{jklm}$ is the random residual effect. DBH and MOE_tof were analysed as continuously distributed variable traits, while survival is a threshold trait with two categories (alive or dead). Therefore, the probability of survival was modelled with a generalized linear mixed model using a logit link function (Nakagawa and Schielzeth 2010; Xiong et al. 2010).

The individual-tree narrow-sense heritability ($h_i^2$) for each trait at each trial was calculated using the variance components from the univariate single-site analysis, assuming these open-pollinated families are half-sib families (Falconer and Mackay 1989).(3) $h_i^2=\frac{{\hat{\sigma }}_A^2}{{\hat{\sigma }}_P^2}=\frac{4\times {\hat{\sigma }}_f^2}{{\hat{\sigma }}_f^2+{\hat{\sigma }}_{bf}^2+{\hat{\sigma }}_e^2}$(3) where ${\hat{\sigma }}_A^2$ is the additive genetic variance,${\hat{\sigma }}_P^2$ is the phenotypic variance, ${\hat{\sigma }}_f^2$ is among-family variance, ${\hat{\sigma }}_{bf}^2$ is the family by block variance, and ${\hat{\sigma }}_e^2$ is the residual variance.

Bivariate, single-site analysis was then carried out to estimate the additive genetic and phenotypic correlations (type-A) between traits $a_1$and $a_2$($r_{\left(a\right)}$) within a site as follows:(4) $r_{\left(a\right)}=\frac{\widehat{Co{v}_{\left(a1,a2\right)}}}{\sqrt{{\hat{\sigma }}_{\left(a1\right)}^2\times {\hat{\sigma }}_{\left(a2\right)}^2}}$(4) where ${\widehat{Cov}}_{\left(a1,a2\right)}$ is the estimated phenotypic or additive genetic covariance between traits $a_1$ and $a_1$, ${\hat{\sigma }}_{\left(a1\right)}^2$ is the estimated additive genetic variance for trait$a_1$, and ${\hat{\sigma }}_{\left(a2\right)}^2$ is the estimated additive genetic variance for trait $a_2$.

Variances are not independent of the scale and mean of the respective traits (Sokal and Rohlf 1995). Therefore, to relatively compare genetic variances of the different traits across sites, the genetic coefficient of variation was calculated as:(5) ${\widehat{CV}}_A=\frac{\widehat{{\sigma }_A}}{\overline{X}}\times 100$(5) where, $C{V}_A$ is the coefficient of additive genetic variation, ${\sigma }_A$ is the standard deviation of the additive genetic variation, and $\overline{X}$ is the mean of the trait.

Combined-site analysis

In order to determine the extent of G × E for DBH and MOE_tof, combined-site (two sites within each zone), univariate analysis were conducted and type-B genetic correlations (Burdon 1977) for these traits were estimated using a linear mixed-model (model 2) as follows:(6) $y_{ijklm}=\mu +S_i+B_j+P_k+S{F}_{il\left(k\right)}+SB{F}_{ijl\left(k\right)}+F_{l\left(k\right)}+e_{ijklm}$(6) The definitions for the model effects are as for model 1 in Equation (2), except $S_i$ is the fixed effect of the $i$-th site; $S{F}_{il\left(k\right)}$ is the random interaction effect of the$i$-th site with the$l$-th family; $SB{F}_{ijl\left(k\right)}$ is the random interaction effect of the $i$-th site with the j-th block and the l-th family. The overall mean ($\mu )$, $S_i$, $B_j$, and $P_k$ were considered as fixed effects and other terms as random.

For the analysis and estimation of type-B genetic correlation, both heterogeneous additive genetic variance $\sim \mathrm{MVN}\left(0,G\otimes \mathit{I}\right)$ and error variances $\sim \mathrm{MVN}\left(0,\mathit{R}\otimes \mathit{I}\right)$ were included in the model$G=\left[\begin{array}{cc}{\sigma }_{a1}^2& {\sigma }_{a1}\\ {\sigma }_{a2}& {\sigma }_{a2}^2\end{array}\right]R=\left[\begin{array}{cc}{\sigma }_{e1}^2& 0\\ 0& {\sigma }_{e2}^2\end{array}\right]$where,${\sigma }_{ai}^2$ and ${\sigma }_{ai}$ are the additive genetic variances and covariances, respectively; ${\sigma }_{ei}^2$ is the error variance for each site; $\mathit{I}$ is the identity matrix equal to the number of observations at each site and 0 indicates no site–site error co-variance.

Type B genetic correlation of additive effects across sites ($r_B)$ was calculated as:(7) $r_B=\frac{{\sigma }_{a1a2}}{\sqrt{{\sigma }_{\left(a1\right)}^2\times {\sigma }_{\left(a2\right)}^2}}$(7) where ${\sigma }_{a1a2}$ is the covariance between the additive effects of the same trait at different sites; ${\sigma }_{a1}^2$ and ${\sigma }_{a2}^2$ are estimated additive variances for the same traits at different sites. Standard errors for heritability, type-A and type-B genetic correlations were estimated using ASReml (Gilmour et al. 2009) and the significance of these parameters was inferred from the estimated standard error.

Heritability estimates from single-site analysis are biased upward because the estimates of additive genetic variance contain variance components of G × E if G × E component is significant (White et al. 1993). Unbiased estimates of heritability and type-A genetic correlations were estimated using univariate and bivariate combined-site analysis, respectively.

The genetic gain expressed as percentage in direct selection of trait i was estimated as:(8) $\mathrm{\Delta }{G}_i=i\times h_i^2\times C{V}_i$(8) where $i$ is the selection intensity of 1% ($i$ = 2.67), $h_i^2$is the narrow-sense heritability of the trait and $CV$ is the coefficient of variation of the phenotypic effect (calculated as the phenotypic standard deviation divided by the mean of the specific trait). The correlated response of the target trait $t$ based on indirect selection of correlated trait i, was calculated as:(9) $C{R}_t=i\times h_i\times h_t\times r_A\times C{V}_t$(9) where $i$ is the selection intensity of 1% ($i$ = 2.667), $h_i\mathrm{and}{h}_t$ are the square root of narrow-sense heritability for the selected trait and the target trait, respectively. $r_A$ is the additive genetic correltion between the traits and $C{V}_t$ is the phenotypic coefficient of variation for the target trait $t$.

Results

Summary statistics for the traits

Summary statistics for DBH, MOE_tof and the survival rate of trees at each trial are presented in Table 3. The greatest mean DBH was at Mader (132 mm) and the lowest was at Hemmesmark (105.2 mm). Similarly, mean DBH values at Övra and Lagfors were greater than those at Hemmingen and Jovan (Table 3). Average MOE_tof was lowest at Mader (11.2 GPa) and were greatest at Lagfors and Hemmesmark (15.9 GPa and 15.1 GPa, respectively). Highest survival was recorded at Hemmingen and Jovan (83% and 82%, respectively), while Mader and Hemmesmark had the lowest survival (51% and 66%, respectively). Genetic coefficient of variation ($C{V}_A)$ for DBH ranged from 7.4% to 13.7% and 7.2% to 11.8% for MOE_tof with the lowest coefficients at Lagfors and Övra and the greatest at Hemmesmark and Mader for both traits (Table 3).

Table 3. Descriptive statistics for DBH, MOE_tof and the survival rate at six progeny trials of lodgepole pine.

	Mader	Jovan	Övra	Hemmingen	Lagfors	Hemmesmark
DBH (mm)
Mean	132	116.8	130.5	114.1	128.1	105.2
Min–Max	44–218	29–209	30–232.5	21.5–211	40–250	15.5–219
${\sigma }_A^2$	242.4	112	98.7	151.5	89.2	207.4
${\sigma }_p^2$	755	648	1003	884	744	912
$C{V}_A$ (%)	11.8	9	7.6	10.8	7.4	13.7
MOEtof (Gpa)
Mean	11.2	13.7	13.7	13.6	15.9	15.1
Min–Max	5.4–17.9	7.9–18.8	5.5–21	7.7–22	9.2–25.6	8.4–20.8
${\sigma }_A^2$	1.75	1.4	0.98	1.64	1.3	2.05
${\sigma }_p^2$	4.0	2.9	5.4	3.6	4.6	3.5
$C{V}_A$ (%)	11.8	8.6	7.2	9.4	7.2	9.4
Survival (%)
Survival rate (%)	51	82	74	83	75	66

Note: DBH: diameter at breast height; MOEtof: modulus of elasticity of time of flight; ${\sigma }_A^2$: Additive genetic variance; ${\sigma }_p^2$: phenotypic variance; $C{V}_A$: Coefficient of additive genetic variation.

Provenance performances

The mean DBH and mean MOE_tof of lodgepole pine provenances assessed at six progeny trials within three zones are presented in Figures 2 and 3, respectively. Additionally, the boxplots of DBH and MOE_tof for each provenance assessed at the six lodgepole pine progeny trials are presented in supplementary materials, Figure 1 and Figure 2, respectively. Provenances Frances Lake and Whitehorse had greatest mean DBH at both Mader and Jovan, while Carmacks and Watson Lake had the lowest mean DBH. The mean DBH of Whitehorse was significantly greater than that of Watson Lake at Mader (p-value < .05) whereas such difference was not significant at Jovan (p-value = .08).

Figure 2. Diameter at breast height (DBH) (mm) means for provenances of lodgepole at the six progeny trials (within three breeding zones) in northern Sweden. The vertical lines in the middle of the boxes are error bars.

Figure 3. Modulus of elasticity (MOE) (GPa) means for provenances of lodgepole at the six progeny trials (within three breeding zones) in northern Sweden. The vertical lines in the middle of the boxes are error bars.

At Övra, mean DBH of the provenances Fort St. John and Prince George were significantly greater than those of Fort Nelson and Watson Lake (p-value < .05), while there was no significant difference among mean DBH of provenances at Hemmingen, except for the significant difference between Prince George and Fort St. John (p-value = .01).

At Lagfors the provenance Prince George had slightly and significantly greater mean DBH than Fort St. John, whereas such difference was non-significant at Hemmesmark.

At Mader, the mean MOE_tof of Whitehorse was significantly lower than those of Carmacks and Watson Lake (p-value < .05) whereas mean MOE_tof was not significantly different among other provenances. At Jovan, mean MOE_tof of Carmacks and Frances Lake were significantly greater than those of Watson Lake and Whitehorse (p-value < .05) while other differences were not significant.

Mean MOE_tof was similar among all provenances at Övra and Hemmingen, except that at Hemmingen, mean MOE_tof of Fort St. John was significantly greater than MOE_tof of other provenances (p-value < .001).

As was observed for mean DBH, mean MOE_tof of Prince George was slightly greater than that of Fort St. John at Lagfors (P-value = 0.02), whereas such difference was non-significant at Hemmesmark.

Individual-tree narrow-sense heritability estimates (${\hat{h}}^2)$

Heritability estimates obtained for DBH, MOE_tof and survival at each site are presented in Table 4. Heritability for DBH ranged from 0.10 to 0.32 with the lowest estimates obtained at Övra and Lagfors (0.10 and 0.12, respectively), while greatest estimates obtained at Mader and Hemmesmark (0.32 and 0.23, respectively). Heritabilities obtained for DBH at Jovan and Hemmingen were similar (0.17). For MOE_tof, lowest ${\hat{h}}^2$ were observed at Övra and Lagfors (0.18 and 0.29, respectively), while they were moderate to high at other trials (0.58 $\ge {\hat{h}}^2\ge$ 0.44). Survival at Mader, Lagfors and Hemmingen had the greatest ${\hat{h}}^2$ (0.74, 0.58, 0.43, respectively), whereas it was less heritable or non-significant at Jovan, Övra, and Hemmesmark (0.24, 0.13, 0.06, respectively).

Table 4. Phenotypic ($r_p$) and additive genetic ($r_a$) correlations between DBH and MOE_tof, narrow-sense heritability estimates ($h^2$) for DBH, MOE_tof, and survival at the six progeny trials of lodgepole pine.

Site	DBH-MOEtof correlation		$h^2$
	$r_a$	$r_p$	DBH	MOEtof	Survival
Mader	–0.59 (0.14)	–0.34 (0.03)	0.32 (0.09)	0.44 (0.11)	0.74 (0.15)
Jovan	–0.12 (0.17)	–0.06 (0.03)	0.17 (0.03)	0.47 (0.11)	0.13 (0.09)
Övra	–0.35 (0.20)	–0.20 (0.03)	0.10 (0.04)	0.18 (0.09)	0.24 (0.12)
Hemmingen	–0.45 (0.13)	–0.29 (0.03)	0.17 (0.04)	0.46 (0.11)	0.43 (0.15)
Lagfors	–0.20 (0.14)	–0.05 (0.03)	0.12 (0.04)	0.29 (0.10)	0.58 (0.12)
Hemmesmark	0.21 (0.15)	–0.05 (0.03)	0.23 (0.08)	0.58 (0.11)	0.06 (0.12)

Note: MOE_tof: modulus of elasticity of time of flight; DBH: diameter at breast height.

Combined-site analysis showed heritability estimates that ranged from 0.09 to 0.19 for DBH, 0.19 to 0.27 for MOE_tof, and 0.13 to 0.26 for survival (Table 5). When comparing among traits, lowest ${\hat{h}}^2$ were obtained for DBH using both single-site and combined-site analysis. In general, heritabilities observed for combined-site analysis were lower than heritabilities obtained using individual-site analysis, particularly for ${\hat{h}}^2$ obtained for MOE_tof at combined trials Mader and Jovan.

Table 5. Phenotypic ($r_p$) and additive genetic ($r_a$) correlations between DBH and MOE_tof, type-B genetic correlations ($r_b)$, genetic gain $(\mathrm{\Delta }{G}_i$) and correlated genetic response ($C{R}_t$) narrow-sense heritability estimates ($h^2$) for DBH, MOE_tof, and survival based on combined-site analysis.

				DBH-MOEtof
Site	DBH	MOEtof	Survival	$r_a$	$r_p$
Mader-Jovan				–0.26 (0.17)	–0.22 (0.02)
$h^2$	0.19 (0.03)	0.19 (0.08)	0.24 (0.06)
${\sigma }_A^2$	129	0.68	0.37
${\sigma }_p^2$	675	3.56	1.5
$r_b$	0.94 (0.14)	0.46 (0.17)	0.99 (0.00)
$\mathrm{\Delta }{G}_i$ (%)	10.9	7.6	–
$C{R}_t$ (%)	–2.8	–2.0	–
Övra-Hemmingen				–0.34 (0.14)	–0.25 (0.02)
$h^2$	0.09 (0.03)	0.27 (0.07)	0.13 (0.09)
${\sigma }_A^2$	88	1.24	0.17
${\sigma }_p^2$	965	4.6	1.36
$r_b$	0.74 (0.24)	0.99 (0.00)	0.40 (0.30)
$\mathrm{\Delta }{G}_i$ (%)	6.1	11.3	–
$C{R}_t$ (%)	–3.6	–2.2	–
Lagfors-Hemmesmark				0.27 (0.15)	–0.03 (0.02)
$h^2$	0.12 (0.03)	0.25 (0.07)	0.26 (0.07)
${\sigma }_A^2$	96	1.04	0.36
${\sigma }_p^2$	834	4.11	1.39
$r_b$	0.90 (0.28)	0.94 (0.34)	0.99 (0.00)
$\mathrm{\Delta }{G}_i$ (%)	7.7	8.7	–
$C{R}_t$ (%)	3.0	1.6	–

Note: MOE_tof: modulus of elasticity of time of flight; DBH: diameter at breast height.

Correlations between DBH and MOE_tof

Additive genetic ($r_a$) and phenotypic$(r_p$) correlations obtained between DBH and MOE_tof for individual-site analysis are presented in Table 4. As expected, correlation estimates obtained between these two traits were mostly negative (–0.12$\ge r_a\ge$ –0.59) except for the positive and non-significant genetic correlation obtained at Hemmesmark (0.21 $\pm$0.15). In general, absolute values of the genetic correlation estimates were larger than the corresponding phenotypic estimates (Table 4).

Greatest negative$r_a$ and $r_p$ were observed for Mader (–0.59 and –0.34, respectively) and for Hemmingen (–0.45 and –0.29, correspondingly), while $r_a$was less negative and associated with high standard errors at other three sites (Övra = –0.35 $\pm$0.20, Jovan = –0.12 $\pm$0.17, and Lagfors = –0.20 $\pm$0.14).

Genetic and phenotypic correlations between DBH and MOE_tof were also calculated using combined-site analysis (Table 5). As similarly observed for single-site analysis, DBH was negatively correlated with MOE_tof at Övra-Hemmingen ($r_a=-0.34\pm 0.14)$and Jovan-Mader (r _a = −0.26 ± 0.17), while such correlation was positive and non-significant at Lagfors-Hemmesmark ($r_a=0.27\pm 0.15)$. Similarly, phenotypic correlations were moderate and negative at Övra-Hemmingen ($r_p=-0.25\pm 0.02)$and Mader-Jovan ($r_p=-0.22\pm 0.02)$, while it was nearly zero at Lagfors-Hemmesmark $\left(r_p=-0.03\pm 0.02\right)$.

Genetic gains and response to selection

The genetic gains observed for direct traits were generally moderate (Table 5), with the greatest gain for DBH obtained at breeding zone 2 (e.g. $\mathrm{\Delta }{G}_{\mathrm{DBH}}$ = 10.9%) and the greatest gain for MOE_tof obtained at breeding zone 4 (e.g. $\mathrm{\Delta }{G}_{\mathrm{MO}{\mathrm{E}}_{tof}}$ = 11.3%). As expected, due to the negative genetic correlation between growth and stiffness, selection based on DBH generated slight negative genetic gains for MOE_tof in breeding zone 2 and 4 ($C{R}_t$ = –2%, –2.2%, respectively).

G × E and type-B genetic correlations (${\hat{r}}_B$)

The extent of G × E was investigated for all three traits (Table 5). A value of ${\hat{r}}_{B_{\mathrm{ADDITIVE}}}$ close to one indicates little genotype by environment interaction, while a low ${\hat{r}}_{B_{\mathrm{ADDITIVE}}}$ indicates extensive genotype by environment interaction, implying parental rankings may change across sites. In general, type-B additive genetic correlations for all three traits were very high across paired-sites (Table 5). For example, ${\hat{r}}_B$ for DBH and MOE_tof were mostly greater than 0.9 except for ${\hat{r}}_B$ of DBH at Övra-Hemmingen (${\hat{r}}_B$ = 0.74) and MOE_tof at Mader-Jovan (${\hat{r}}_B$ = 0.46). For survival, ${\hat{r}}_B$ was mostly close to one except at Övra-Hemmingen (${\hat{r}}_B$ = 0.40).

Discussion

Trial means and provenance performances

This study used measures of DBH, MOE_tof, and survival from three sets of lodgepole pine open-pollinated trials within three breeding zones in northern Sweden originating from provenances distributed in Yukon and BC in western Canada.

Several studies have found that natural population variation of lodgepole pine is well related to latitude and longitude (Wu and Ying 2004) or temperature (Wang et al. 2006, 2010) and frost-free period (Rehfeldt 1988). Similarly, a steep elevational genetic differences has been demonstrated for the inland variety in its natural range, especially below latitude 56° (Ying et al. 1985; Ying 1991).

For lodgepole pine in Sweden, it has been evidenced that survival is largely dependent on the latitude of origin, whereby seed sources of northern origin had higher survival along with the higher rate of seed production (Lindgren et al. 1994). Additionally, it has been reported that height growth is associated with altitude of origin, whereby fastest growth was achieved for provenances about 800 m, and there was a decline in growth above this threshold (Lindgren 1983; Fries and Lindgren 1986).

In general, interactions of species and provenances with environments are usually not pronounced within a region, unless there are large differences in climates or soils (White et al. 2007). In our study, soil and climatic conditions were mostly similar within a breeding zone. However, within breeding zone 2, Mader had lower temperatures and higher snow fall. Similarly, within breeding zone 4, Övra had higher precipitation, particularly during growing season, and within breeding zone 5, Hemmesmark had higher snow fall (Table 2). In addition, within zones 4 and 5, trials at lower latitudes had higher elevations.

Our results generally indicate that the mean performance of provenances, particularly for mean DBH, differed significantly within a breeding zone, despite their slight differences within a test site. As such, seed sources originating from Yukon, the northernmost provenances, had better growth performance in Mader than in Jovan, particularly provenances Frances Lake and Whitehorse.

In contrast, seed sources originating from BC, had better growth performance when planted at trials with lower latitudes within a breeding zone. Further, maximum growth performance of provenances originating from BC, Fort St. John and Prince George in particular, was achieved at Övra, the trial associated with the highest precipitation during vegetative growth and the highest elevation.

Further, it has been stated that differences in wood properties related to provenances are not very large. However, occasionally seed source plays an important role in determining wood properties. As such, specific gravity has been reported to decrease with increasing latitude and elevation of the parental source (Zobel and Jett 1995).

In our study, in contrast to what was observed for DBH, provenances originating from Yukon, particularly Whitehorse, had the lowest mean stiffness at Mader (higher latitude), and highest mean stiffness of seed sources originating from BC was obtained for Fort St. John and Prince George when planted at Lagfors and Hemmesmark, trials with the warmest temperatures and the lowest elevations.

Overall, the trial mean differences of DBH and performance of provenances in this study are consistent with results of the study by Ericsson (1993), whereby these provenances and progeny trials at age about 10 years were assessed for height and survival. However, we observed that trees in Mader had greater mean DBH than those in Jovan, whereas mean height was greater at Jovan in Ericsson (1993). Similarly, in his study, trees at Mader and Hemmesmark had the lowest survival, and lowest DBH was obtained at Hemmesmark.

According to Ericsson et al. 1993, trial Mader had been exposed to harsh climatic conditions followed by heavy fungal attacks, principally Gremmeniella abietina, which occurred in the middle of 1980s in northern Sweden (Ericsson 1993; Elfving et al. 2001).

Likely, Hemmesmark is the site of lower fertility compared to other trials and has been more exposed to windstorm, whereby about 100 trees fell during winter storms in 2013 and 2014.

In general, to achieve the highest stiffness for lodgepole pine, provenances of Yukon origin should be planted at lower latitudes and those of BC origin should be planted at lower elevations within the tested breeding zones.

Individual-site analysis

Genetic variations obtained for DBH were the highest at Mader and Hemmesmark ($C{V}_A$ = 11.8%, 13.7%, respectively) and therefore $h^2$ were greatest at these two trials. The lowest $h^2$ of DBH at Övra and Lagfors might be related to their lowest genetic variations ($C{V}_A$ = 7.6%, 7.4%, respectively) as well as the very high phenotypic variance obtained at Övra.

Genetics of wood quality traits have been extensively studied in various conifer species and results indicate that wood quality traits generally have higher heritability than growth traits and genetic correlations between growth and wood quality traits are mostly unfavourable (Hallingbäck et al. 2008; Wu et al. 2008; Chen et al. 2014; Hong et al. 2014; Hayatgheibi et al. 2017).

Similarly, MOE_tof in this study had higher heritability than DBH, and the average heritability estimates of MOE_tof ($h^2$ =  0.40 ± 0.1) was slightly greater than those reported in Norway spruce ($h^2$ =  0.31 ± 0.06) (Chen et al. 2015), white spruce ($h^2$ = 0.38 ± 0.07) (Picea glauca MoenchVoss) (Lenz et al. 2013), and radiata pine ($h^2$ = 0.23 ± 0.11) (Gapare et al. 2010), but similar to those of slash pine (Pinus elliottii Englm.) ($h^2$ = 0.42 ± 0.07) (Li et al. 2007), and Scots pine ($h^2$ = 0.44 ± 0.1) (Hong et al. 2014).

Genetic and phenotypic correlations between DBH and MOE_tof were significantly negative at Mader, Hemmingen, and Övra, while they were non-significant at Lagfors, Hemmesmark, and Jovan.

In general, the lowest stiffness obtained at Mader as well as its high negative genetic correlation with DBH implies that wood quality traits, stiffness in particular, should be considered as an important breeding objective trait for selection of the next generation of lodgepole pine in northern Sweden.

Combined-site analysis and G × E interactions

There are several analytical methods to measure the extent of G × E for traits in tree breeding (Li et al. 2017). The type-B genetic correlation (Burdon 1977) approach is considered as one of the preferred methods due to its ability in characterizing patterns of ranking changes of genotypes across environments (Li et al. 2017) as well as being easily handled in linear mixed-models with imbalanced data (Baltunis et al. 2010).

Results of G × E studies revealed that G × E is not very important for stem and wood quality traits while it is more evident for growth traits (Wu and Matheson 2005; Baltunis et al. 2010; Gapare et al. 2010).

However, there was little evidence of G × E for DBH in this study, indicating the relative performance of genotypes is stable across environments within the same breeding zone. Similarly, there was no evidence of G × E for MOE_tof at trials within the breeding zones 4 and 5. In contrast, type-B genetic correlation was low ($r_b$ =  0.46) at breeding zone 2 for MOE_tof, which signifies existence of G × E and there might be changes in the ranking of genotypes for stiffness in these two most northern sites. Literature indicates that G × E for wood quality traits is generally low compared to growth traits. This significant G × E for MOE_tof at breeding zone 2 is an aberration compared to other studies and the two other breeding zones in this study.

Conclusion

1. Provenances of Yukon origin had the highest growth but lowest stiffness at higher latitude, while those of BC origin grew faster at lower latitudes and had the highest stiffness within low altitude zone.

2. Heritability ranged from 0.10 to 0.32 for DBH, 0.18 to 0.58 for MOE_tof, and 0.06 to 0.74 for survival among the six sites. Heritability for combined-site estimates in each breeding zone ranged from 0.09 to 0.19 for DBH, from 0.19 to 0.27 for MOE_tof, and from 0.13 to 0.26 for survival.

3. Type-B genetic correlations were high for all three studied traits, except for DBH and survival in zone 4 ($r_b$ = 0.74 and 0.40, respectively) and for MOE_tof in zone 2 ($r_b$ = 0.46).

4. Unfavourable growth-stiffness genetic correlations and correlated responses indicate that breeding for a 1% increase in diameter would confer about 2% decreases in lodgepole pine stiffness within zone 2 and 4.

5. On the basis of the results obtained in this study, G × E for stiffness in northern Sweden and unfavourable growth-stiffness genetic correlation should be considered in selective breeding programmes of lodgepole pine

Acknowledgements

The authors also acknowledge Liming Bian, Zhiqiang Chen, David Hall, and Zhou Hong for their assistance in field sampling.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

The authors gratefully acknowledge financial support from Föreningen Skogsträdsförädling, Bo Rydins, Kempe foundations, and Swedish University of Agricultural Sciences (SLU).