Provenance variations in wood traits of Larix kaempferi (Lamb.) Carr. and their relationships with growth parameters

Abstract

Background

Understanding genetic variations in wood traits and their relationships with dynamic growth is essential for optimizing tree breeding strategies. In this study, we investigated genetic control of 15 wood traits and the genetic associations of these traits with dynamic growth of tree height (H) and diameter at breast height (DBH) in Japanese larch (Larix kaempferi (Lamb.) Carr.), using data from 84 families from six Japanese provenances in a 22-year-old common garden experiment.

Results

The Logistic model provided a good fit to the temporal dynamics of H and DBH. Substantial genetic variation was detected among provenances and families for most traits. Wood anatomical traits exhibited the highest coefficients of genotypic variation (up to 20.49%) and individual-tree heritabilities (0.39–0.80). Growth-related logistic parameters showed moderate heritability for DBH parameters (0.14–0.17) and low to moderate estimates for height parameters (< 0.12). Genetic correlation analyses revealed that the relative growth rate of DBH (DBH_r) was positively correlated with cellulose content and fiber width but negatively correlated with basic density. Height growth parameters displayed few significant correlations with wood traits.

Conclusions

These findings provide new insights into the genetic control of growth and wood formation, supporting the development of multi-trait selection strategies in L. kaempferi and other conifer breeding programs, such as selecting for families with both rapid early growth and desirable wood properties for solid and pulp wood products.

Background

Larix kaempferi (Lamb.) Carr. (Japanese larch) is naturally distributed in the high mountains of central Japan [1] and is one of the most successfully introduced fast-growing tree species for timber, pulp and paper production in Europe, Asia, and North America [2, 3]. In China, its rapid juvenile growth makes it superior to the native Larix species when compared to Larix gmelinii (Rupr.) Rupr. and Larix olgensis A. Henry, making it as the most suitable conifer for timber and pulp production [4, 5]. Therefore, early-stage selection and improvement of growth and wood traits during the optimal rotation period are critical for enhancing biomass accumulation and pulp productivity in this species.

The primary breeding objectives for timber tree species are to improve growth and wood property traits. Growth traits primarily focus on tree height (H) and diameter at breast height (DBH) due to the relative ease of their assessment and their significant impact on log value. In recent years, wood phenotyping has advanced rapidly with field-deployable and non-destructive methods, including resistograph, acoustic velocity, near-infrared spectroscopy and wood core analysis, which substantially improve the efficiency of large-scale trait evaluation [6, 7]. Wood properties can be grouped into physical, chemical, and anatomical categories [8]. Physical traits, particularly basic density, strongly determine the value of timber and pulpwood [9, 10]. Chemical traits, including lignin, hemicellulose, and cellulose contents, are key factors affecting the efficiency of pulp extraction and end-product quality [11, 12]. Anatomical traits, mainly related to tracheid, are central to understanding growth, wood quality and xylem function in conifers, as tracheids constitute over 90% of softwood volume and serve mechanical support [13, 14]. Tracheid dimension (length and width) has a significant impact on the quality of pulp and paper and its radial variation can be used to define the juvenile wood and mature wood [15]. Tracheid wall-to-lumen ratio, latewood proportion and microfibril angle exhibit positive relationships with wood density and stiffness [16, 17]. The genetic variation in growth traits as well as wood physical and chemical properties in L. kaempferi and its hybrids has been extensively documented through a variety of trials [17–19]. However, the inheritance of wood anatomical traits and their relationships with growth and physical traits has received relatively limited attention in L. kaempferi.

The relationships between growth and wood quality constitute a major concern in tree breeding [20]. Trade-offs between DBH and wood density have been well-documented in many conifer species, where faster diameter growth is often associated with lower wood density, posing a challenge for simultaneous improvement of growth and wood quality [21–23]. It should be noted that, as the rotation period of most conifers exceeds 20 years, growth equations can better characterize the dynamic growth process during this period, particularly through biologically interpretable parameters that capture ontogenetic developmental patterns [24]. Growth equations overcome the limitations of static trait measurements (e.g., single-timepoint tree height) and have demonstrated exceptional efficacy in quantifying age-specific growth trajectories across multiple tree species [25, 26]. However, the relationships between growth parameters and wood quality still remain largely unexplored.

The present study collected data on growth and fifteen wood traits from 450 individual natural trees from 84 families of six Japanese provenances L. kaempferi in a long-term provenance test. The main objectives were to (1) quantify the provenance and family variation and heritability for the growth parameters and wood traits, (2) explore the correlations between the growth parameters and the wood traits. The findings provide a theoretical basis for the genetic improvement of wood traits and the joint selection of growth and wood properties in L. kaempferi.

Materials and methods

Genetic material and trial design

Seeds were collected from 84 open pollination families from six natural provenances along the mountains of central Japan covering the complete native distribution of L. kaempferi (Table 1, see [24], Fig. 1 for geographic locations) based on a Japan International Cooperation project (JICA, Forestry Project Phase I (8), 1996–2000). The seeds were germinated in March 1996, and two-year-old seedlings were transplanted from the nursery to the Jianshi trial site in Hubei Province, China (30.08° N, 110.05° E, 1700 m), which exhibits a cool and humid subtropical alpine climate. The trial design was a randomized complete block with six replications. Families were planted in a double row plot of eight seedlings and randomized within each replication.

Table 1.

Sampling and geographical information for Larix kaempferi provenances and the Jianshi trial site

Provenance/ trial site	Number of family	Number of sample trees	Mean annual temperature (℃)	Mean annual precipitation (mm)	Latitude (°)	Longitude (°)	Elevation (m)
Kusatsu	20	97	6.0	1627	36.58	138.50	1374
Asama	6	54	9.3	1254	36.33	138.50	927
Fuji	16	80	10.5	1665	35.50	138.75	844
Ina	15	75	4.8	1875	35.75	138.17	1726
Nikko	18	90	5.7	2176	36.75	139.42	1444
Matsumoto	9	54	13.5	1445	36.33	139.75	1620
Jianshi	84	450	11.4	1650	30.80	110.05	1700

The name of provenance from the prefix of the local state-owned forest farms’ name in 1996–2000

Fig. 1.

Distribution of major provenances of Larix kaempferi in Japan and the location of the provenance trial site in Hubei, China

Growth traits

Four to six trees per family were randomly selected across the blocks, and a total of 450 trees were sampled from the trial. The numbers of families and sampled trees from each provenance are shown in Table 1. For each sampled tree, height (H, m) and diameter at breast height (1.3 m, DBH, cm) were measured at different ages to capture dynamic growth: height at 3–5, 7–10, 12, 13, 17, 18, and 22 years, and DBH at 7–10, 12, 13, 17, 18, and 22 years.

Wood density and anatomical traits

Wood cores were extracted from the bark to the bark through the pith at breast height using a 12 mm diameter increment borer at age 14 to measure wood traits of the 450 sampled trees. From each wood core, three growth rings—the 5th, 9th, and 13th from the pith, representing juvenile wood, transition wood, and near-mature wood, respectively—were selected to measure wood basic density (BD) and anatomical characteristics.

For each growth ring, a tangential section 20 um thick and a transverse section 4 um thick were continuously cut from the earlywood to the latewood. The specific method of resin-embedded and sectioning is illustrated in the method described by Liu [27]. The water displacement method was employed to ascertain the BD of the wood samples [28]. The widths of the growth ring and the latewood were measured with a micrometer, with an accuracy of 0.01 mm, and subsequently employed in the calculation of the latewood proportion (LWP).

A color image computer analysis system with a microscope (TDY-5.2, Northeast Forestry University, China) was used to measure the wall-to-lumen ratio of earlywood (EWW_L) and latewood (LWW_L) cells, the tracheid width of earlywood (EWFW) and latewood (LWFW), the tracheid length of earlywood (EWFL) and latewood (LWFL), and the microfibril angles of earlywood (EWMA) and latewood (LWMA). The TDY-5.2 system has been developed for the rapid analysis of binary images according to the different colors of the wood cell lumen and cell wall, according to the characteristic values of pixel gray value, size, number, and location. The data is obtained by utilizing microscopic measurements and image and data analysis.

The EWW_L, LWW_L, EWFW, and LWFW values were obtained through the observation of the average cell size from the bark to the pith, with measurements taken 10 to 20 times per ring. The EWMA and LWMA were measured randomly in 20 different tracheids (earlywood and latewood) by observing the iodine-stained slices [29]. Fifty tracheids were measured per ring for EWFL and LWFL by observing the wood strips treated by the defibration solution. The defibration solution consisted of 1:1 aqueous nitric acid and aqueous chromic acid (10% each; method in Appendix).

Wood chemical traits

The 450 wood cores were dried for 24 h at 25 °C to a constant weight, ground, and filtered through a 60-mesh sieve. Near infrared (NIR) spectra were collected using a near-infrared spectrometer (LabSpec®Pro, ASD, Inc., USA). Each spectral output was an average of 30 scans. The final spectral data for each sample were an average of three replicates. The three wood chemical properties, namely, lignin, holo-cellulose and cellulose, were predicted using NIR models developed for L. kaempferi [30], with determination coefficients of 0.96, 0.95 and 0.96 for lignin, holocellulose and cellulose, respectively. The final hemicellulose content is obtained by subtracting the cellulose content from the holocellulose content.

Growth models

Three non-linear growth models were used to fit the growth curves for each sampled tree. The equations are expressed as:

1

2

3

where P(t) represents the average phenotypic value at age t; a (asymptotic maximum for H or DBH), b (initial growth parameter), and r (relative growth rate) are the parameters.

Model comparison was based on convergence success rate (CSR), coefficient of determination (R²), and root mean square error (RMSE).

The parameters and the time at the point of inflection (ti) of the optimal model were then extracted as derived growth traits for subsequent genetic analyses. The parameters of the optimal model were then extracted as derived growth traits for subsequent genetic analyses.

Variance and genetic parameter estimation

The ASReml linear mixed model and restricted maximum likelihood (REML) algorithm were used to estimate variance components for each trait. All calculations were performed using the statistical software ASReml-R v. 3.0 package in R software [31, 32].

4

where y is the vector of observations for growth and wood traits; β is the vector of fixed effects, including the grand mean and block effects for all traits, and age effects for those traits measured across three growth rings; p, pb, f(p), f(p)b and e represent the vectors of random effects for provenance, provenance × block interaction, family within provenance, family within provenance × block interaction, and residual error, respectively. These random effects are assumed to follow independent normal distributions: p~ (0, ), pb~(0, ), f(p)~(0, ), f(p)b~() and e ~ (0, ), respectively. , , , , and are the respective variances components of the random effects. X is the incidence matrix for β, and ~ represent incidence matrices corresponding to the random effects.

Approximate standard errors of estimated parameters were estimated using a standard Taylor series expansion method [33]. The statistical significance of the variance components for each trait was determined using a two-tailed likelihood ratio test [34].

The coefficients of phenotypic  and genetic  variation were calculated using the following equations:

5

6

where is the phenotypic variance ; is the additive genetic variance , assuming the open-pollinated mating produced true half-sibs, so the genetic variation would be a quarter of the total additive genetic variance; is the overall mean.

The individual narrow-sense heritability (), and family mean heritability (), and the amount of provenance variation () were calculated as follows:

7

8

9

The genetic correlation coefficients () between wood traits and growth parameters were estimated from variance and covariance components obtained using a bivariate mixed-model approach. For each pair of traits (growth parameter x and wood trait y), a bivariate model was fitted to individual-tree data, with provenance and family effects included as random factors. The between wood traits and growth parameters were calculated as follows:

10

where and represent the genetic variance components of x and y, respectively. represents the genetic covariance components between x and y. In this study, x refers to the growth parameters (a, b and r) of H and DBH, and y refers to all wood traits.

Results

Growth modeling comparison

In the modeling of height (H) and diameter at breast height (DBH) growth trajectories, both the Logistic and Gompertz models exhibited high convergence success rates (> 99%), whereas the Chapman–Richards model showed slightly lower rates of 91.7% for H and 90.1% for DBH (Table 2). All three models achieved excellent goodness-of-fit (R² > 0.98). For H, the Logistic model yielded the lowest RMSE (0.405) and a relatively high R² (0.986) for H. For DBH, it attained the highest R² (0.995), although its RMSE (0.527) was slightly higher than that of Gompertz (0.410) and Chapman–Richards (0.372). Considering model uniformity, convergence stability, and overall fitting performance, the Logistic model was ultimately selected for fitting both H and DBH growth.

Table 2.

Comparison of three nonlinear growth models fitted to tree height (H) and diameter at breast height (DBH) growth

Model	Growth trait	Convergence success rate (CSR) (%)	Coefficients of determination (R2)	Root mean square error (RMSE)
Logistic	H	100.0	0.986	0.405
	DBH	99.8	0.995	0.527
Gompertz	H	100.0	0.984	0.475
	DBH	99.7	0.991	0.410
Chapman–Richards	H	91.7	0.987	0.411
	DBH	90.1	0.993	0.372

Genetic parameter estimates for growth parameters and wood traits

The coefficients of genetic (CVG) and phenotypic (CVP) variation exhibited distinct patterns across trait categories (Table 3). Wood anatomical traits consistently displayed the highest CVG values, ranging from 4.95% for latewood tracheid length (LWFL) to 20.49% for earlywood wall-to-lumen ratio (EWW_L). In contrast, CVG values for wood chemical constituents were notably low (< 3%), suggesting limited genetic variability in lignin, cellulose, and hemi-cellulose content within the studied population.

Table 3.

Descriptive statistics (with standard deviation values in parentheses) and genetic parameters (with standard error values in parentheses) for growth parameters and wood traits

Trait (units)	Mean (SD)	(se)	(se)	(se)	(se)	(se)	(se)	(se)	(se)	CVG (%)	CVP (%)
H_a	23.77 (9.71)	< 0.01 (< 0.01)	< 0.01 (< 0.01)	0.38 (0.48)	1.43 (1.03) **	20.05 (1.28)	0.07 (0.09)	0.26 (0.25)	< 0.01 (< 0.01)	4.47	23.94
H_b	3.73 (0.45)	2.57 (2.19) **	< 0.01 (< 0.01)	< 0.01 (< 0.01)	11.21 (3.91) **	79.34 (4.95)	< 0.01 (< 0.01)	< 0.01 (< 0.01)	0.03 (0.02)	0.01	35.44
H_r	1.01 (0.20)	< 0.01 (< 0.01) **	< 0.01 (< 0.01) **	< 0.01 (< 0.01) **	< 0.01 (< 0.01) **	< 0.01 (< 0.01)	< 0.01 (< 0.01)	< 0.01 (< 0.01)	0.02 (0.02)	0.02	18.76
H_ti	11.25 (3.41)	< 0.01 (< 0.01)	0.07 (0.10) **	0.16 (0.16) **	1.42 (0.31) **	3.73 (0.24)	0.12 (0.12)	0.29 (0.22)	< 0.01 (< 0.01)	3.98	16.37
DBH_a	20.93 (7.58)	0.73 (0.60) **	< 0.01 (< 0.01)	0.91 (0.45) **	1.26 (0.76) *	24.66 (1.11)	0.14 (0.07)	0.42 (0.13)	0.03 (0.02)	6.66	25.85
DBH_b	6.34 (3.30)	13.65 (11.19) **	< 0.01 (< 0.01)	16.01 (7.52) **	45.23 (11.85) **	324.28 (14.48)	0.17 (0.08)	0.43 (0.12)	0.03 (0.03)	12.31	43.46
DBH_r	1.58 (0.30)	< 0.01 (< 0.01) **	< 0.01 (< 0.01)	< 0.01 (< 0.01)	< 0.01 (< 0.01)	< 0.01 (< 0.01)	0.04 (0.04)	0.19 (0.17)	0.02 (0.01)	2.62	18.72
DBH_ti	8.27 (1.91)	0.04 (0.03) **	0.01 (0.03)	0.01 (0.04) **	0.37 (0.08) **	2.02 (0.09)	0.01 (0.06)	0.04 (0.22)	0.01 (0.01)	0.9	13.25
EWW_L	0.17 (0.06)	< 0.01 (< 0.01) **	< 0.01 (< 0.01)	< 0.01 (< 0.01) **	< 0.01 (< 0.01) **	< 0.01 (< 0.01)	0.79 (0.18)	0.83 (0.05)	0.12 (0.08)	20.49	34.6
LWW_L	0.52 (0.12)	< 0.01 (< 0.01) **	< 0.01 (< 0.01)	< 0.01 (< 0.01) **	< 0.01 (< 0.01)	0.03 (< 0.01)	0.39 (0.12)	0.76 (0.06)	0.04 (0.03)	13.2	30.77
EWFL (μm)	2287.00 (277.91)	12,567.65 (9755.69) **	0.01 (< 0.01)	19,731.36 (5613.47) **	15,595 (4337.78) **	63,324.45 (3767.8)	0.8 (0.2)	0.79 (0.06)	0.11 (0.08)	8.29	13.91
LWFL (μm)	2742.00 (268.72)	11,789.07 (8907.67) **	0.01 (< 0.01)	9939.09 (4562.06) **	23,033.43 (5075.27) **	59,088.32 (3517.64)	0.43 (0.19)	0.6 (0.13)	0.11 (0.08)	4.95	11.32
EWFW (μm)	55.90 (5.41)	2.73 (2.59) **	1.05 (1.09) *	4.1 (1.73) **	5.58 (1.97) **	31.76 (1.89)	0.4 (0.16)	0.65 (0.12)	0.06 (0.05)	5.01	11.76
LWFW (μm)	29.89 (3.41)	1.06 (0.93) **	0.03 (0.23) *	3.03 (0.85) **	2.08 (0.68) **	10.05 (0.6)	0.8 (0.19)	0.8 (0.06)	0.07 (0.05)	7.89	12.93
EWFL_W	41.19 (5.65)	9.97 (7.40) **	0.11 (0.71)	10.12 (2.82) **	6.65 (2.28) **	36.18 (2.16)	0.76 (0.19)	0.8 (0.06)	0.16 (0.1)	10.58	18.66
LWFL_W	92.63 (11.73)	17.25 (12.73) **	< 0.01 (< 0.01)	16.02 (5.75) **	5.22 (5.81)	135.33 (7.97)	0.41 (0.14)	0.74 (0.09)	0.1 (0.07)	7.61	17.72
EWMA (°)	21.78 (3.76)	1.16 (0.93) **	< 0.01 (< 0.01)	2.02 (0.63) **	2.61 (0.56) **	6.48 (0.39)	0.73 (0.2)	0.73 (0.08)	0.09 (0.07)	8.56	14.9
LWMA (°)	18.00 (2.90)	0.45 (0.43) **	< 0.01 (< 0.01)	0.47 (0.51) **	4 (0.75) **	7.22 (0.43)	0.16 (0.17)	0.31 (0.25)	0.04 (0.03)	5.22	18.66
LWP (%)	32.28 (8.89)	9.13 (6.41) **	0.48 (0.92)	2.04 (1.73)	0.3 (2.42)	61.7 (3.61)	0.13 (0.11)	0.49 (0.24)	0.12 (0.08)	5.43	23.04
BD (g/cm3)	0.38 (0.04)	< 0.01 (< 0.01) **	< 0.01 (< 0.01) **	< 0.01 (< 0.01) **	< 0.01 (< 0.01) **	< 0.01 (< 0.01)	< 0.01 (< 0.01)	< 0.01 (< 0.01)	0.04 (0.05)	< 0.01	12.98
Lignin (%)	28.68 (0.95)	0.08 (0.06) **	< 0.01 (< 0.01)	0.03 (0.03)	0.04 (0.07)	0.77 (0.08)	0.13 (0.16)	0.4 (0.34)	0.09 (0.06)	0.79	3.33
Cellulose (%)	41.43 (3.19)	0.24 (0.33) **	0.36 (0.32)	0.62 (0.38) *	< 0.01 (< 0.01)	8.48 (0.64)	0.27 (0.16)	0.63 (0.15)	0.02 (0.03)	2.69	7.52
Hemi-cellulose (%)	28.58 (2.25)	0.11 (0.14) *	0.07 (0.14)	0.12 (0.20)	0.33 (0.45)	4.45 (0.48)	0.10 (0.16)	0.34 (0.38)	0.02 (0.03)	1.74	7.89

^**, P < 0.01, *, P < 0.05; a, maximum value; r, relative growth rate; b, initial growth parameter; ti, point of inflection. Wall-to-lumen ratio of earlywood (EWW_L); wall-to-lumen ratio of latewood (LWW_L); tracheid length of earlywood (EWFL), tracheid length of latewood (LWFL), length-to-width of earlywood tracheid (EWFL_W), length-to-width of latewood tracheid (LWFL_W), tracheid width of earlywood (EWFW), tracheid width of latewood (LWFW), microfibril angle of earlywood (EWMA), microfibril angle of latewood (LWMA), latewood proportion (LWP), and wood basic density (BD). The variance components included provenance (), interaction between provenance and block (), family (), interaction between family and block () and random effect (). , , , CVP and CVG represent the individual narrow-sense heritability, family mean heritability, the amount of provenance variation, phenotypic variation coefficient and genetic variation coefficient, respectively

Significant genetic variation was observed among provenances and among families within provenances for most growth parameters and wood traits (Table 3). Both provenance and family effects were highly significant (P < 0.01) for wood anatomical properties (except for latewood proportion), wood density, r of H, a, b, ti of DBH. For wood chemical constituents, provenance effects were significant while a significant family effect was observed only for cellulose content.

Estimates of individual-tree heritability () and family-mean heritability () revealed distinct patterns of genetic control across trait categories (Table 3). Wood anatomical traits exhibited the highest heritability values at both levels. EWW_L, LWW_L, EWFL, LWFW, and EWMA showed values ranging from 0.39 to 0.80, and corresponding values consistently exceeding 0.70. In contrast, growth-related logistic parameters displayed generally lower and more variable heritability estimates. ti of H, a and b of DBH showed moderate , with 0.12, 0.14 and 0.17, respectively. Basic density showed negligible individual-tree heritability ( < 0.01). Wood chemical constituents showed low to moderate (0.10–0.27), though cellulose content exhibited high family-mean heritability ( = 0.63).

Correlations between growth parameters and wood traits

Prior to conducting correlation analyses, we assessed the provenance-level trends in the seven anatomical traits and basic density (BD) across the three measurement ring ages (5th, 9th, and 13th). With increasing age, BD remained relatively stable, EWW_L, EWFL_W, EWMA, and LWMA increased, while the remaining traits decreased. Significant age effects were detected for all traits except BD; however, the trends were generally consistent over time (Fig. 2). Therefore, averaging trait values across the three measurement years had negligible influence on the subsequent correlation estimates.

Fig. 2.

The provenance differences of wood traits with significant provenance effect in the 5th, 9th, 13th ring ages. The letters ‘a’, ‘b’ and ‘c’ indicate that the age difference was statistically significant by one-way ANOVA at P < 0.05; See Table 3 for full description of traits

Genetic correlation coefficients among growth parameters, wood anatomical, and chemical traits at individual-tree are presented in Fig. 3. Among growth parameters, a, b, and r were positively inter-correlations, whereas each negatively correlated with ti for both of H and DBH. H_b was positively correlated with BD ( = 0.74) and LWMA ( = 0.67). H_r was positively correlated with EWFW ( = 0.44) but negatively correlated with EWFL_W ( = −0.55). Conversely, DBH_r was significantly negatively correlated with EWFL ( = −0.52) and EWFL_W ( = −0.36) but positively correlated with cellulose content ( = 0.73) and EWFW and LWFW ( = 0.46 and 0.36 respectively). These significant correlations suggest that selection for rapid diameter growth (DBH_r) may favorably increase cellulose content and fiber widths but could simultaneously reduce tracheid length and wood density. Height growth and wood quality may be largely genetically independent in this population, offering flexibility for independent selection.

Fig. 3.

Genetic correlations between growth parameters and wood traits. * and ** denote significance for P < 0.05 and P < 0.01, respectively; Blank cells indicate model fitting failure. See Table 3 for full description of traits

Among wood traits, EWFW was negatively correlated with LWMA ( = −0.66) and BD ( = −0.52). LWFW was positively correlated with BD ( = 0.52). EWW_L was positively correlated with LWFL_W ( = 0.81) and BD ( = 0.68). EWFL_W was positively correlated with LWFL_W ( = 0.49) and LWMA ( = 0.48). LWFL_W was positively correlated with LWP ( = 0.84). Collectively, these findings indicate that higher basic density may be achieved through indirect selection for increased EWW_L and LWFW, combined with restriction of EWFW. Moreover, the cross-seasonal genetic coordination between earlywood and latewood traits (e.g., EWFL_W with LWFL_W) suggests that early selection based on earlywood properties can effectively predict latewood performance.

Discussion

Model selection for growth trajectory characterization

The selection of an appropriate growth function is a critical first step in the genetic analysis of longitudinal traits. There are a variety of growth models, such as the Logistic, Gompertz, and Chapman-Richard equations, which are frequently used to describe biological processes of tree. It should be noted that the tree growth process cannot be fully characterized by a single or a few separate values. The simplicity of the growth equation and the accuracy of the estimated parameters were the main considerations. In this study, the Logistic, Gompertz, and Chapman–Richards models all provided excellent fit to both height and DBH growth data in terms of R² and RMSE (Table 2). The Chapman-Richard equation was not considered due to its lower convergence success rate. Previous studies in forest tree species have similarly reported that the Logistic model performs well for describing cumulative growth processes, particularly when data are limited to early developmental stages [35]. The consistency of the Logistic model across both height and DBH trajectories in this study supports its utility for genetic analyses where parameter comparability is desired.

Magnitude and distribution of genetic variation

The genetic parameter estimates revealed marked differences in the levels of genetic control among trait categories. Wood anatomical traits consistently displayed the highest coefficients of genotypic variation (CVG up to 20.49%) and the highest individual-tree heritabilities ( up to 0.80). These values are comparable to or exceed those reported for tracheid dimensions and microfibril angle in other conifer species [36–38]. The high observed for traits such as earlywood wall-to-lumen ratio (EWW_L) and latewood fiber length weighted (LWFL_W) suggest that substantial genetic gains could be achieved through direct selection.

In contrast, wood chemical constituents exhibited notably low genetic variation (CVG < 3%) and low to moderate heritability (= 0.10–0.27). This finding differs from results obtained in a previous half-sib progeny test of the same species, where heritability estimates for chemical traits ranging from 0.38 to 0.67 [19]. The discrepancy may be attributable to differences in population structure, genetic background, or environmental conditions between the two trials. Alternatively, the younger age of the trees in this study (≤ 13 years) may have contributed to lower heritability estimates, as wood chemical composition can exhibit substantial ontogenetic variation and may stabilize only at later developmental stages [39]. These contrasting results underscore the importance of population- and age-specific parameter estimation when developing breeding strategies for wood chemical properties. In addition, the cellulose content exhibited notably high family-mean heritability (= 0.63). In this situation, selection at the family level or combined family plus within-family selection could yield meaningful genetic gains for cellulose content.

Growth-related logistic parameters displayed intermediate levels of heritability. The DBH parameters a and b showed moderate values (0.14–0.17), whereas height parameters generally exhibited low to moderate estimates ( < 0.12). This difference may reflect the greater cumulative environmental sensitivity of height growth, which integrates effects of competition, microsite variation, and episodic stresses over time [40]. The moderate heritability of growth parameters nonetheless suggests that selection for altered growth trajectories is feasible, particularly for traits related to radial increment and point of inflection of tree height.

Genetic correlations between growth and wood traits

The genetic correlation patterns between growth parameters and wood traits have important implications for multi-trait selection. A key finding of this study is the negative genetic association between DBH_r and both tracheid length traits and basic density. This result aligns with the widely reported adverse relationship between growth rate and wood density in conifers [9, 41]. The underlying physiological mechanism is thought to involve a dilution effect, wherein rapid cambial division during earlywood formation produces larger-diameter tracheids with thinner cell walls, thereby reducing overall density and average fiber length.

In contrast, height growth parameters exhibited few significant genetic correlations with wood traits. This genetic independence is advantageous from a breeding perspective, as it implies that selection for improved height growth can proceed without incurring substantial negative pleiotropic effects on wood quality. The decoupling of height and wood property inheritance may reflect distinct developmental and physiological controls, with height growth driven primarily by apical meristem activity and resource allocation, while wood formation in the cambium is regulated by local hormonal and mechanical signals [42]. This finding supports the feasibility of simultaneously improving height growth and wood density or fiber properties in breeding programs.

Influence of cambial age on wood trait variation and genetic parameter estimation

Wood properties as well as other economic trait like resin yield are known to exhibit pronounced ontogenetic trends driven by cambial aging [43], with early-formed juvenile wood typically characterized by shorter tracheids, larger microfibril angles, and lower density compared to mature wood [44, 45]. In the present study, wood traits were assessed at cambial ages corresponding to tree ages of 5, 9, and 13 years, a period during which most conifers are still producing juvenile wood, transition wood, and near-mature wood. Significant age effects were detected for nearly all anatomical traits, consistent with the expected radial gradients in cell dimensions and wall thickness as the cambium matures.

To account for these developmental trends in genetic parameter estimation, age was included as a fixed effect in the mixed linear models. This approach effectively partitions out systematic age-related variation, thereby reducing potential bias in variance component estimation and ensuring that genetic parameters reflect age-corrected trait values. For the correlation analyses, we first examined provenance-level age trajectories of all anatomical traits and confirmed that, despite significant age effects, the temporal trends were largely parallel across provenances (Fig. 2). Consequently, averaging trait values across the three measurement ages for each individual tree was considered to have negligible impact on the subsequent genetic correlation estimates. This pragmatic simplification is justified given the consistency of age trends and the primary interest in overall genetic associations rather than age-specific relationships.

In contrast to the anatomical traits, basic wood density (BD) exhibited no significant age effect across the 5- to 13-year measurement interval. Two plausible explanations may account for this observation. First, the sampling window may largely correspond to the transition wood zone, where density remains relatively stable before a more pronounced increase in mature wood [44]. Second, methodological limitations associated with BD determination may have contributed to reduced sensitivity to age-related change. Density was assessed using increment cores and the water displacement method, an approach that can be susceptible to measurement error arising from irregular core geometry, incomplete saturation, or extractives content [6]. Such errors may obscure subtle developmental trends, particularly when the true age effect is small. Future studies employing high-resolution techniques such as X-ray densitometry or SilviScan would provide more precise characterization of radial density profiles and help clarify whether the observed stability reflects biological reality or methodological constraint.

Implications for breeding and future research

The findings of this study have direct implications for the design of conifer breeding programs. First, the high heritability and substantial genetic variation for wood anatomical traits support their inclusion as selection criteria to improve wood quality. Second, the negative genetic correlation between radial growth and wood density/tracheid length necessitates the use of multi-trait selection indices or restricted selection strategies to avoid undesirable correlated responses. Third, the genetic independence of height growth from wood traits offers a favorable avenue for concurrent improvement.

Several limitations should be acknowledged. The study was conducted on a single population at relatively young ages (up to 13 years); genetic parameters and correlations may shift as stands approach rotation age except for growth parameters. Longitudinal studies extending to later developmental stages would be valuable for validating these results. Additionally, the low genetic variation observed for chemical traits underscores the need for broader germplasm screening and the potential application of genomic tools to uncover rare alleles or regulatory variants influencing wood chemistry.

In conclusion, this study provides a comprehensive genetic characterization of growth trajectories and wood traits, clarifying the trade-offs and opportunities inherent in multi-trait selection. The results contribute to a growing body of knowledge supporting the integration of growth modeling and quantitative genetics in forest tree improvement.

Conclusion

Substantial genetic variation was observed among provenances and families for both growth parameters and wood traits. Wood anatomical traits exhibited the highest heritabilities, suggesting that substantial genetic gains could be achieved through direct selection. Wood chemical constituents showed limited genetic variability; however, cellulose content displayed high family-mean heritability, indicating that selection could be performed at the family level. Growth-related logistic parameters showed moderate heritability for DBH parameters and the point of inflection of height, suggesting that dynamic growth patterns can be genetically improved. Rapid radial growth was negatively associated with wood density and tracheid length, while height growth parameters remained largely independent of wood quality traits. This genetic independence offers flexibility for the joint improvement of height growth and desirable wood properties.

Abbreviations

H

Tree height

DBH

Diameter at breast height

a

Maximum value

r

Relative growth rate

b

Initial growth parameter

ti

Point of inflection

EWW_L

Wall-to-lumen ratio of earlywood

LWW_L

Wall-to-lumen ratio of latewood

EWFL

Tracheid length of earlywood

LWFL

Tracheid length of latewood

EWFL_W

Length-to-width of earlywood tracheid

LWFL_W

Length-to-width of latewood tracheid

EWFW

Tracheid width of earlywood

LWFW

Tracheid width of latewood

EWMA

Microfibril angle of earlywood

LWMA

Microfibril angle of latewood

LWP

Latewood proportion

BD

Wood basic density

CVG

Coefficients of genetic variation

CVP

Coefficients of phenotypic variation

Authors’ contributions

Y.Z., L.D. and X.S. designed the study, performed the statistical analysis and wrote the manuscript. Y. X. and D.C. collected the phenotypic measurements, carried out the critical reading and grammatical correction of manuscript. All authors read and approved the final manuscript.

Funding

This work was supported by the National Key Research and Development Program of China (2022YFD2200302, 2022YFD2200103), the Fundamental Research Funds for the Central Non-profit Research Institution of CAF (CAFYBB2022ZC001), and the Open Fund of State Key Laboratory of Tree Genetics and Breeding (Chinese Academy of Forestry) (TGB-KF-2024–001).

Data availability

The datasets generated during and/or analysed and additional tables and figures during the current study are available in the [https://doi.org/10.17605/OSF.IO/8DYJRd].

Declarations

Ethics approval and consent to participate

All field studies and experimental research conducted on plants, including the procurement of plant materials, were carried out in strict accordance with institutional, national, and international guidelines and regulations. The corresponding author, Xiaomei Sun, is the head of the Conifer Genetic Improvement Team at the Institute of Forestry, Chinese Academy of Forestry, and possesses the necessary authorization to gather plant materials.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.