Introduction

Forests are the largest carbon sinks in terrestrial ecosystems and are important for sequestering CO2 and mitigating the adverse effects of climate change (Peng et al. 2018). Therefore, the use of forest plantations for carbon sequestration is gaining more interest (Zhu et al. 2013). Large-scale reforestation and afforestation have significant implications for carbon sequestration (Thomas et al. 2007), so screening and selecting provenances, families or clones for high carbon sequestration is a critical measure to optimize plantations as carbon sinks (Wang et al. 2016). Because the carbon sequestration capacity of conifers is significantly higher than that of broad-leaved tree species, conifers should be prioritized in programs for carbon sequestration forests (Xu et al. 2013).

Wood properties are important indices to evaluate wood yield and quality and are affected by environmental factors such as temperature, precipitation, soil conditions, competition, planting density, thinning density, slope position and direction, and interactions among these factors (Guo et al. 2002a, b; Shi et al. 2011). Poor site conditions can also reduce the number and length of fibers (Wang et al. 2006). Genetic improvement of wood properties and selection of superior materials have always been an important directive for breeders and are needed to optimize timber production and carbon sequestration for afforestation projects. Long-term provenance tests are essential to determine genetic and geographic variations in tree growth and recommend appropriate seed sources for reforestation, especially in a changing climate (Weber et al. 2019). Variations in growth traits (tree height and DBH) and wood properties (wood density and carbon concentration) also need to be studied to predict the ability of trees to sequester carbon in different environments (Weber et al. 2018).

Larix olgensis, an important timber species, is mainly distributed in southeastern Heilongjiang, Liaoning, and Changbai Mountain area of Jilin, and it is also found in Korea and Russia (Hu et al. 2015). With excellent wood properties, it is not only used as wood fiber industrial raw materials, paper making raw materials and bio-fuels, but also used in industries such as construction, shipbuilding and railway (Xia et al. 2016). Because it is one of the main afforestation species in northeastern China, its genetic improvement has been a research priority. Provenance tests of L. olgensis in China were first started in the 1970s (Yang 1984). Superior pulpwood provenances and building timber provenances of L. olgensis have been identified (Yu et al. 2015a, b), and genetic variations in carbon concentration and allocation among different tissues at different sites have been reported (Jiang et al. 2019). Large interactions between genotype and environment are known affect growth of this species (Sun et al. 2018; Wang et al. 2021; Zhang et al. 2021). However, Yin et al. (2017) found that correlation coefficients among growth traits and wood properties were mostly not significant. Therefore, it is necessary to study genetic variations in wood properties in different environments.

In the present study, wood properties for 10 provenances of L. olgensis from four sites were measured and their variations analyzed. The objectives of this study were to (1) determine the variation and heritability of wood properties for the provenances at different sites; (2) estimate the phenotypic correlations among different traits and the relationships between traits and geographic factors at the sites; and, (3) evaluate and select superior pulpwood provenances and high carbon storage provenances within the sites.

Materials and methods

Study area

Germplasm was collected from 10 provenances of L. olgensis in its natural distribution area in 1980 to produce seedlings for provenance trials. Seedlings were planted in 1982 at four trial sites using a randomized complete block design. The locations and climatic characteristics of the trial sites and seedling spacing are given in Table 1.

Table 1 Geographic and climatic characteristics and spacing of Larix olgensis at four sites in China
Full size table

Wood property measurements

One tree-ring core was collected three trees from each of three blocks for each provenance from each of the four sites from May to July in 2019. In total, 360 cores at breast height were obtained from south to north, and diameter at breast height (DBH) was measured. The cores were wrapped and put in paper tubes and taken to the laboratory for further analysis. All cores were placed in an oven at 80 °C for 48 h, then weighed every 2 h until the difference between the last two measurements was less than 0.5% of the total mass. Wood density was measured using the drainage method reported by Cheng (1985). Fiber length (FL) and fiber width (FW) were measured as detailed by Mu et al. (2009); the cores were divided into sapwood, heartwood, and pith, then each portion treating with nitric acid and chromic acid, and the length and width of 10 fibers in each portion were measured, then the mean FL and FW of the 30 values were calculated for each core (tree). Hemicellulose content (HEC), lignin content (LC), cellulose content (CEC), holocellulose content (HOC) and ash content (AC) of each clone were measured using a fully automatic fiber analyzer according to the national standard GB/T 2677.1 93 (A2000i; ANKOM Technology, Macedon, NY, USA) and resistance furnace and the method of Xu et al. (2016). The carbon content (CAC) of each sample was measured with a Hydrocarbon analyzer Multi EA 4000 (Analytik Jena AG, Germany).

Statistical analyses

Statistical analyses were carried out using SPSS 25.0 software (IBM, Armonk, NY, USA). The following linear model was used for joint analysis of the four sites together and F tests was performed to estimate the significance of ANOVA (Zhao et al. 2014):

$$y_{ijk} = \mu + S_{i} + P_{j} + SP_{ij} + \varepsilon_{ijk}$$
(1)

where \(y_{ijk}\) the performance of the kth tree of the \(j\)th provenance growing in the ith site, \(\mu\) is the overall mean, \(S_{i}\) the random effect of the \(i\)th site, \(P_{j}\) is the random effect of the \(j\)th provenance, \(SP_{ij}\) the interactive effect of the \(j\)th provenance and \(i\)th site, and \(\varepsilon_{ijk}\) is the random error.

The different wood properties were subjected to analyses of variance among the provenances within sites using the following linear model (Zhang et al. 2020):

$$X_{ijk} = \mu + P_{i} + B_{j} + PB_{ij} + e_{ijk}$$
(2)

where \(X_{ijk}\) the performance of individual tree k in provenance i within block j, μ is the overall mean, Pi is the random effect of provenance ith, Bj is the random effect of block j, PBij is the interactive effect of provenance i and block j, and PBij is the random error.

The phenotypic and genotypic coefficient of variation (PCV and GCV) were calculated using the formula of Mohamed et al. (2017):

$${\text{PCV}} = \frac{{\sqrt {\sigma_{{\text{p}}}^{2} } }}{X} \times 100$$
(3)
$${\text{GCV}} = \frac{{\sqrt {\sigma_{{\text{g}}}^{2} } }}{X} \times 100$$
(4)

where \(\sigma_{{\text{p}}}^{2}\) is the phenotype variance component for the trait, \(\sigma_{{\text{g}}}^{2}\) is the genetic variance component for the trait, and \(X\) the average value for the trait.

Provenance heritability (H2) was estimated using the formula of Razafimahatratra et al. (2016):

$$H^{2} = \frac{{\sigma_{{\text{P}}}^{2} }}{{\sigma_{P}^{2} + \frac{{\sigma_{{{\text{PB}}}}^{2} }}{B} + \frac{{\sigma_{{\text{e}}}^{2} }}{NB}}}$$
(5)

where \(\sigma_{{\text{P}}}^{2}\) the variance component of the provenance, \(\sigma_{{{\text{PB}}}}^{2}\) the variance component of the interaction between the provenance and the block, \(\sigma_{{\text{e}}}^{2}\) the variance component of the error, B is the number of blocks, N is the total number of provenances.

Equivalent latitude was adopted to reflect the real effect of latitude and eliminate any influence of altitude (Yang et al. 1991):

$${\text{Equivalent latitude}} = {\text{latitude}} + \frac{{{\text{Elevation}} - 300}}{E}$$
(6)

where E is constant, and when the elevation is greater than 300 m, E is 140 or when the elevation is less than 300 m, E is 200.

The correlation analysis \(r_{{\text{A}}} \left( {xy} \right)\) among wood properties and the relationships between wood properties and environmental factors was done as described by Bi et al. (2000) using the equation:

$$r_{{\text{A}}} \left( {xy} \right) = \frac{{{\text{COV}}_{{{\text{P}}\left( {xy} \right)}} }}{{\sigma_{{\text{P}}} \left( x \right)\sigma_{{\text{P}}} \left( y \right)}}$$
(7)

where \({\text{COV}}_{{{\text{P}}\left( {xy} \right)}}\) is the phenotypic covariance between index x and y, \(\sigma_{P} \left( x \right)\) and \(\sigma_{P} \left( y \right)\) are the phenotypic variance for index \(x\) and index \(y\) respectively.

The multiple-traits comprehensive evaluation was analyzed using the following formula (Zhao et al. 2016):

$$Qi = \sqrt {\mathop \sum \limits_{i = 1}^{n} a_{i} }$$
(8)
$$a_{i} = x_{ij} /x_{{j{\text{max}}}}$$
(9)

where \(Qi\) the value of colligation assessment for ith provenance, \(x_{ij}\) the average value of ith provenance for trait j, \(x_{{j{\text{max}}}}\) is the maximum average value of different provenances for trait j, and n is number of traits.

Genetic gain was estimated using the formula of Silva et al. (2008):

$$\Delta G = H^{2} S/X$$
(10)

where \(H^{2}\), S, and X are provenance heritability, selection difference, and mean value of the given trait, respectively.

Results

Estimates of variance components

The analysis of variance showed that different wood properties reached significant difference level (P < 0.01) among sites and the interactions between sites and provenances during the multi-site joint analysis, while only WD, FL, and FL/W differed significantly among the different provenances (Table 2), indicating environmental effects were a predominant source of variation. Variance components among sites were all higher than those among provenances and their interactions. Therefore, it was necessary to screen for superior provenances within each site. Significant differences for different properties were detected among provenances within sites (P < 0.01), except for FW and FL/W, while the differences among the blocks and the interactions between provenances and blocks were mostly insignificant (Table 3), indicating high variation among provenances.

Table 2 Results of ANOVA for different wood properties of Larix olgensis at four sites in China
Full size table
Table 3 Variance components of different traits for the 10 provenances of Larix olgensis at four sites in China
Full size table

Trait means

The means for different properties of all provenances within sites are shown in Table 4. The mean DBH ranged from 17.59 to 28.51 cm across sites, and was largest at site MES and lowest at CH. The mean WD at site LS (0.576 g cm–3) was significantly lower than those at the other three sites, especially compared with MES (0.645 g cm–3). The mean FL, FW, and FL/W ranged from 2088.19 (MES) to 2404.27 μm (JGDQ), from 33.14 (LS) to 36.91 μm (JGDQ), and from 57.57 (MES) to 72.82 (LS), respectively. FL/W for all provenances at LS was higher than that at MES. The mean HEC, CEC, HOC, LC, and AC ranged from 9.27 (MES) to 10.601% (JGDQ), from 38.18 (MES) to 44.19% (LS), from 47.45% (MES) to 53.60% (LS), from 27.74 (JGDQ) to 30.55% (MES), and from 0.47 (JGDQ) to 0.58% (LS), respectively. HEC at CH and JGDQ was higher than at LS and MES, CEC and HOC at site MES were lower than at the other three sites, and LC at site MES was the highest. The mean CAC for sites JGDQ, LS, and MES was 456.96 g kg–1, 457.42 g kg–1, and 454.21 g kg–1 respectively, and the lowest mean was at site CH (442.48 g kg–1). Across sites, WD was higher for provenances DHL and DST than for the other provenances across sites and lowest for provenance HL. Provenance LSH had the lowest values for WD at sites CH and LS, however, it was higher at sites of JGDQ and MES. FL/W for provenance DHL was lower than others at 4 sites, and higher for DST. HOC for provenance DHL was lower at sites CH, JGDQ, and LS than at site MES, but the opposite for provenance DHL. CAC differed significantly across all sites for each provenance (e.g., provenance LSH had higher CAC at CH, LS and MES than at JGDQ). Values for CAC of provenance XBH were lower across all sites, and those for provenance HL were higher across sites, except at LS. Provenance BH were the highest at site JGDQ and lowest at site LS, whereas those for provenance ML were the highest at LS and lowest at MES.

Table 4 Mean values for different traits for provenances of Larix olgensis at four sites in China
Full size table

Genetic variability for wood traits

The genetic variability parameters calculated for the wood properties of the 10 provenances at each site are shown in Table 5. The PCV and GCV for AC were highest and all moderate, whereas those for CAC were the lowest. At site CH, the maximum for PCV was 19.546% and 14.430% for GCV. The maximum PCV and GCV at site JGDQ was 27.365% and 21.113%, respectively. The maximum PCV and GCV at site LS was 18.376% and 14.855%, respectively. The maximum PCV and GCV at site MES were 18.871% and 13.811%, respectively. In addition, the PCVs and GCVs for the other properties were almost always low. The H2 for the different traits at CH ranged from 0.599 (FL/W) to 0.971 (CEC and HOC). The lowest H2 at site JGDQ was 0.467 (LC), and the highest was 0.976 (HOC). The H2 for the different traits at LS ranged from 0.663 (FL/W) to 0.996 (AC). The lowest H2 at site JGDQ was 0.332 (FL/W), and the highest was 0.986 (AC). Nearly all traits had high heritability.

Table 5 Genetic variability parameters for different wood traits of Larix olgensis at four sites in China
Full size table

Inter-trait correlation analysis

The correlation coefficients between different properties within sites are shown in Table 6. There were significant correlations between traits DBH and WD at sites CH and LS, but not at other sites. DBH was significantly correlated with FL and FW at sites JGDQ and LS and negatively correlated with LC at sites CH and MES. WD was negatively correlated with CEC and HOC across all sites but MES. FW was negatively correlated with FL/W across all sites, and FW and FL/W all positively correlated with FL. HOC was positively correlated with HEC and with CEC across sites. It was interesting that CAC was positively correlated with CEC and HOC at CH, but negatively correlated with these two traits at site MES. In addition, CAC was significantly positively correlated with LC only at site MES. LC was positively correlated with HEC at site JGDQ and negatively correlated with CEC and HOC at site MES.

Table 6 Correlation coefficients for wood properties of Larix olgensis at four sites in China
Full size table

Geographic variations

The correlation coefficients between wood properties and geographic and climatic factors at the four sites are given in Table 7. WD was negatively correlated with elevation and positively correlated with temperature. FL was positively correlated with latitude (equivalent latitude) and elevation and negatively correlated with precipitation and temperature, while FL/W was only positively correlated with elevation. HEC, CEC and HOC were positively correlated with latitude (equivalent latitude) and elevation and negatively correlated with precipitation, but there was a significant negative correlation between HEC and longitude and between CEC and HOC with temperature. LC was negatively correlated with latitude (equivalent latitude) and elevation and positively correlated with precipitation and temperature. CAC was positively correlated with longitude, elevation, and precipitation but negatively correlated with temperature.

Table 7 Correlation coefficients of wood traits of Larix olgensis and geographic and climatic factors at four sites in China
Full size table

Comprehensive evaluation and genetic gain

For the selection of superior pulpwood provenances and high carbon storage provenances, the results of the comprehensive evaluation of different wood properties and Qi values for the provenances across sites are shown in Table 8. From the perspective of pulpwood provenances, WD, FL, HEC, CEC, HOC, LC, and AC were regarded as indicators, and the values for LC and AC were calculated using negative numbers because of the negative effects of papermaking. As for high carbon storage provenances, WD, HEC, CEC, HOC, LC, AC, and CAC were used as evaluation indices, and AC was used as negative number in the calculation. Properties that had no significant effect among provenances within sites were eliminated. At site CH, the Qi value for pulpwood ranged from 1.261 (DHL) to 1.547 (BH) and for high carbon storage from 1.936 (DHL) to 2.080 (BH). At site JGDQ, Qi for pulpwood ranged from 1.642 (DHL) to 1.785 (LSH) and for high carbon storage from 1.704 (DHL) to 1.822 (LSH). At site LS, Qi for pulpwood ranged from 1.251 (BH) to 1.490 (DST) and for high carbon storage from 1.934 (BH) to 2.038 (DST). At site MES, Qi for pulpwood ranged from 1.257 (TQL) to 1.546 (DST) and for high carbon storage from 1.941 (TQL) to 2.013 (LSH). Pulpwood and high carbon storage were regarded as selected targets, with a selection rate of 20%, two superior provenances were selected at each site. The genetic gains for different traits of the superior provenances are shown in Table 9. From the view of pulpwood, the superior provenances selected at CH, JGDQ, LS, and MES were BH and BDS, LSH and JX, DST and JX, and DST and XBH, respectively. On the other hand, the superior high carbon storage provenances at 4 sites were BH and HL, LSH and JX, DST and HL, and DST and LSH, respectively. The genetic gains of different traits for superior provenances were shown in Table 9. As was shown, the genetic gain for WD at LS was ranged from 7.44 to 10.19%. Meanwhile, the genetic gains for FL at different sites were ranged from 0.19 to 11.14%. The genetic gains at the different sites ranged from −4.59 to 2.13% for HEC, from 0.40 to 4.16% for CEC, from 0.51 to 3.08% for HOC, from −11.40 to 0.44% for LC, and from −17.96 to −6.07% for AC. For CAC, the genetic gain was 0.57% at CH, −0.13% at JGDQ, −0.12% at LS, and 0.30% at MES.

Table 8 Qi values of different provenances of Larix olgensis at four sites in China
Full size table
Table 9 Genetic gains of different wood traits of Larix olgensis within four sites in China
Full size table

Discussion

Tree growth and wood quality were affected by a variety of factors such as genotype, temperature, precipitation, soil conditions, and their interactions (Fang et al. 2020). Heritability was closely related to climatic factors. Therefore, heritability of different traits for the same provenance or family will certainly differ among the sites. Our analysis of variance for wood properties of L. olgensis provenances at the four sites showed that all traits differed significantly among different sites, indicating that the different conditions had vital effects on plant growth. The values for the wood traits also differed among the different provenances within each site, indicating that the evaluation and selection of provenances within sites were effective. The phenotypic and genotypic coefficients of variation for almost all properties at the different sites were lower than those reported for 26-year-old clones of L. olgensis (Yin et al. 2017); however, the provenance heritability in the present study was higher than that found by Yin et al. (2017).

Wood density is a strong determinant of mechanical strength, which affects wood quality and potential uses (Lundqvist et al. 2018) and is considered the main physical variable and key index for evaluating wood quality and pulp yield (Ortega Rodriguez and Tomazello-Filho 2019). An increase in wood density has an important effect on the efficiency of pulp production (Niemczyk and Thomas 2020). Site is also an important factor with a strong effect on wood density (Dias et al. 2018). In the present study, the average wood density of L. olgensis provenances at all sites was 0.612 g cm–3, with the maximum of 0.645 g cm–3 found at MES and the minimum of 0.576 g cm–3 at LS, both significantly higher than obtained by Li and Lian (2017a). This difference might be due to differing afforestation densities. In addition, the wood density of the DHL and DST provenances was greater than 0.600 g cm–3 at all the sites, and less than 0.600 g cm–3 for provenance HL at all the sites, indicating that although the same provenances had certain differences in the different site conditions, the provenances had certain commonalities in their site responses.

Pulp properties and paper quality are mainly influenced by wood fiber traits (fiber length and fiber length to width ratio) (Liu et al. 2020). The greater the fiber length to width ratio, the more times the fibers can be mixed and the better the combinability of the fibers, which gives the paper greater strength (Bai et al. 2009). In this study, the average fiber length was 2266.602 μm, considered as a long fiber. The average fiber length and fiber width were similar to previous findings (Shi et al. 2011). The average fiber length to width ratio was 65.68, higher than reported for Populus deltoids by Wu et al. (2011); therefore, L. olgensis is categorized as high-quality material for papermaking.

In the process of genetic improvement of forest trees, multiple traits are increasingly expected to be improved simultaneously (Lin 2010); thus, correlation analyses can provide a reference guide for joint breeding of multiple traits (Jia et al. 2016). However, genetic correlations between growth and wood traits are likely to depend on the trial site (Li et al. 2017). In the present study, DBH and WD were significantly correlated, which disagrees with a report for Pinus taeda (Xu et al. 2000). However, the WD of Larix kaempferi is negatively related to growth rate of juveniles and was not correlated with growth rate at maturity (Zhu et al. 2000). Similarly, Stackpole et al. (2010) found a significant negative genetic correlation for Eucalyptus globulus between basic wood density and diameter at the selection age (4–5 years); however, at the harvest age, the genetic correlation was not significant and slightly positive. Zhang et al. (2014) found a weak positive correlation between DBH and WD for triploid hybrid clones of Populus tomentosa. Here, we found that wood density was negatively correlated with cellulose content at all sites except MES, whereas Guo et al. (2014) found a significant positive correlation between these traits for Salix suchowensis and a negative correlation between WD and hemicellulose content; these differences might be related to the tree species. The correlation between fiber length and fiber width was extremely significant and positive, similar to the results of Liang et al. (2016) for Pinus koraiensis. For L. olgensis, fiber length was significantly correlated with hemicellulose content, which indicates that fiber qualities were closely related to chemical composition. Interestingly, CAC was positively correlated with CEC and HOC at CH, but they were negatively correlated at MES, which might be due to CAC was significantly positively correlated with LC at MES, and lignin contains more carbon than cellulose does (Weber et al. 2018).

The results of the correlation analysis between wood traits and geographic factors of the trial sites showed a significant negative correlation between wood density and altitude, agreeing with finding that the wood density of Pinus nigra decreased with an increase in altitude (Dias et al. 2018). Similarly, growing Alnus formosana at lower latitudes increased the wood density (Yang et al. 2012). Although wood density is a heritable trait, it interacts with the meteorological variables (Rocha et al. 2020). Here we found a significant positive correlation between wood density and temperature; the higher the temperature, the earlier cambial activity can begin, and as cell division accelerates, more wood cells are produced, which can increase wood density (Xu et al. 2011). The correlations between wood fiber length and latitude, equivalent latitude and altitude were extremely significant, contrary to the results of Yang et al. (2009). This finding might be due to the fact that JGDQ is at a high latitude with appropriate temperature and precipitation for L. olgensis, positively influencing growth. At the same time, fiber length was significantly negatively correlated with the annual average precipitation and temperature, consistent with the results of Zhang et al. (2011) for cotton; appropriate precipitation was conducive to an increase in cotton fiber length, whereas excessive precipitation resulted in shorter fibers. Our correlation analysis showed that carbon content was significantly positively correlated with longitude, altitude and annual precipitation and significantly negatively correlated with annual average temperature, similar to the results of Zhou (2015) on Fraxinus mandshurica; temperature decreases as altitude increases, plant growth rate slows, which increases the degree of lignification, and resulting in greater carbon content.

The best provenance must be selected for a given site or region to achieve maximum plantation productivity (Loha et al. 2009). In the past, growth traits, wood quality, and disease resistance have been the main criteria for selecting suitable reproductive material for tree species (Buras et al. 2020). Our correlation analysis of numerous wood traits at four sites allowed us to select superior pulpwood provenances and high carbon storage provenances for the sites, although the genetic gains were lower than for the elites selected by Yin et al. (2017), probably due to the different tree ages or the number of materials (provenances and clones).

Conclusions

Our results on genetic and geographic variations in the wood properties of 10 L. olgensis provenances at four sites showed a significant difference in wood properties among the sites, provenances, and their interactions. Wood traits were mainly related to the latitude and altitude of the site and were also affected by annual precipitation and temperature. Superior pulpwood provenances and high carbon storage provenances within sites could be selected separately for use as the preferred afforestation material for a particular site.