Introduction

Sequoia sempervirens (Coast redwood (Lamb. ex D. Don) Endl.) occurs naturally in western North America. Its range is a narrow coastal zone (latitude 42.20 N to 35.83 N; Fig. 1) that largely corresponds with the coastal fog belt of California where it is either the dominant or co-dominant species. The climate in this region is typically Mediterranean with mild, rainy winters, and cool dry summers and S. sempervirens can be found in an altitudinal ranging from sea level to around 1000 m, covering a multitude of soil types (Lorimer et al. 2009; McBride and Jacobs 1977). The distribution of this moisture sensitive species is limited by evapotranspirational stress to the south and east and by cold temperatures to the north (McBride and Jacobs 1977). Sequoia sempervirens is one of the few gymnosperms known to re-sprout from stumps following natural disasters such as flooding or fire (O’Hara and Berrill 2009a, 2010; Sawyer et al. 2000). It is a long lived species with an estimated to range between typical lifespan of 500 and 1500 years, with individual trees being able to persist for over 2500 years (Busing and Fujimori 2002; Finney and Martin 1989; Rogers 2000). The wood of S. sempervirens is highly durable, it has little shrinkage when dried, and is structurally stable (Cown 2008). These wood properties plus its appearance makes it a desirable species for timber production. In the United States of America, S. sempervirens timber is widely used for decks, outdoor furniture, weatherboards and other products where durability, appearance, and stability are important (Cown 2008; Cown and McKinley 2009).

Fig. 1
figure 1

Map showing the approximate locations (white circles) of the clonal provenances used in this study, the current extent (stripped areas) of the Sequoia sempervirens range (U.S. Geological Survey 1999) in the United States of America, annual precipitation (mm/yr) normalised from 1971 to 2000 (PRISM Climate Group 2007), and counties in the states of Oregon and California (black lines). The accuracy of the locations are limited by the coordinate information recorded at the time of sampling. Numbers identify the county where clones were sampled. For county identities see Table 2

Full size image

In New Zealand, S. sempervirens was first introduced in the 1860s, but seed was imported for large scale planting as an exotic timber species between 1920 and 1945 to help replace the dwindling indigenous timber resource (Knowles and Miller 1993). However, most of these plantings failed and the species fell from favour as a commercial forestry species (Knowles and Miller 1993). In the 1990s, interest in establishing a New Zealand S. sempervirens forest industry was rekindled (Brown et al. 2008; Libby 1993). New Zealand’s climate is generally suitable for S. sempervirens and failures in the early twentieth century can be attributed to a lack of appropriate siting, no weed control, and absence of suitable mycorrhizae (Meason et al. 2012; Weston 1971; Young 1983). Once established, New Zealand S. sempervirens stands are capable of achieving over 50 m3 ha−1 yr−1 at 40 years on optimal sites (Nicholas and Watt 2011; Palmer et al. 2012).

The high potential productivity in New Zealand as well as the restriction on supply of S. sempervirens timber from native Californian stands has helped increase interest in growing S. sempervirens in New Zealand as an alternative supplier to the United States of America market. Currently there are in excess of 6300 hectares established throughout the country (NZFFA 2013), the majority of which originates from clones from various commercial introductions. New Zealand grown S. sempervirens can produce wood properties of similar quality to United States of America S. sempervirens timber from second-growth forests and produce premium timber grades for the United States of America market (Cown et al. 2013; Meason et al. 2012). However, the wood quality New Zealand grown S. sempervirens is highly variable with some stands producing timber with poor wood properties (Cown 2008; Meason et al. 2012).

Previous research has concluded that wood quality variation between trees is likely primary due to genetics, indicating that variability should be able to be reduced through the selection of provenances and clones (Cown 2008; Cown and McKinley 2009). Quantitative genetics of the species is, however, poorly understood. The lack of available seed for planting, as well as the usual poor seed viability (Ahuja 2009) has contributed to this. Genetics research using various markers in S. sempervirens populations in the United States of America has shown that the species has high genetic diversity (Anekonda 1992; Anekonda and Libby 1996; Rogers 1997), driven by the unusual hexaploid (six copies of the chromatids) and ability to reproduce both clonally and sexually (Ahuja 2009). Little information exists, however, on the provenances best suited to New Zealand for both growth and wood properties. Such information would be greatly beneficial for the developing New Zealand S. sempervirens sector.

This study examines genetic parameters from a collection of S. sempervirens clonal provenances known as the Kuser collection (Kuser et al. 1995) tested in over 11 clonal provenance trials across New Zealand. The primary goal of these trials was to quantify the genetic variation and genotype-by-environment interaction of S. sempervirens provenances grown in New Zealand conditions, and use this information to determine the need for a breeding programme and the need for care in site-matching of provenances for planted forests. This study aimed to determine the first genetic parameters for S. sempervirens in the oldest two trials of the Kuser collection. Implications of the results on the establishment of a commercially viable S. sempervirens forestry industry in New Zealand will be discussed. The Kuser collection also provides the opportunity to understand the complex genetic relationships of the S. sempervirens across its natural range. We will also use the results to discuss how this study improves our understanding of the genetic diversity across its various populations in its home range of the western United States of America.

Methods

The genetic material used in this study is from a collection of S. sempervirens clonal provenances known as the Kuser collection (Kuser et al. 1995). This collection was initiated in 1984 and comprised 198 clones collected from 98 different provenances throughout its natural range in western United States of America, from Curry County in southern Oregon to Monterey County in California (Fig. 1). Clones were established from cuttings taken from seedlings transplanted into the nursery, where possible, but were also taken from saplings or from stump sprouts if no seedlings were available (Kuser et al. 1995). Different rates of physiological ages of cuttings were therefore highly likely. Cuttings were taken and used to establish stool plants of more than one ramet per clone. Cuttings from the collection were sent between 1988 and 1992 to different sites in the United States of America and to several other countries including France, Spain and South Africa to establish common garden provenance trials. An incomplete set of 49 clones was sent to New Zealand during this period, but it appears no trial was established. Few of these early international trials were fully established successfully and of those, only the early results from those trials were published.

In the early 2000s, a consortium imported another 182 clones from the 198 clone Kuser collection into New Zealand to identify which provenances were most suitable for New Zealand conditions. Stool beds were established from the imported clones and cuttings taken from the stool beds to establish 11 replicated trials across New Zealand from 2003 to 2006. No information was available on which original North American-based ramets the material was propagated from nor on stool bed origin of clones in New Zealand.

One hundred and seventy of the 182 Kuser provenance clones were planted out in 2003 at two trial sites in the North Island; one at Awaho (decimal degrees 39.04229 S, 177.30342 E), 14 km west of Wairoa, Gisborne and the other at Thompson (decimal degrees 39.25136 S, 174.31817 E), 17 km north east of Stratford, Taranaki. Awaho is located on two, fertile ex-forestry plantation soil series; Typic Udothents (dominant series) and Typic Psammquents. The latter series is prone to periodic water logging. Thompson is located on a well-drained, fertile ex-pasture, Typic Hapudands. Both sites are quite suitable for S. sempervirens with the species site index (Palmer et al. 2012) for Awaho and Thompson estimating mean height of dominant trees at age 40 of 45 m and 35 m, respectively. Summary information about the provenances’ environmental factors in the North America and the two New Zealand trials considered in this study are provided in Table 1. Each trial consisted of a single-tree-plot randomised incomplete-block design. A total of eight replicates were planted at each trial to account for within-site environmental variation. Each replicate was divided into five incomplete blocks of 36 trees with one ramet from each clone randomly assigned to one of the plots plus additional plants from a single control seedlot (seedling checklots). At least 34 clones were planted in each plot. Ramets averaged 10–15 cm in height at establishment and they were planted at 3 m × 3 m spacing. Each site was ex-pasture farmland and it was planted in the spring with no weed control or fertilisation. Sites were visited within the first 2 years and spot spray weed control was applied when appropriate.

Table 1 Environmental data across Sequoia sempervirens provenance’s natural range in the United States of America (USA) and at the sites where the Kuser provenance trial was planted in New Zealand. United States of America climatic data normalised for the period 1971–2000 (PRISM Climate Group 2007). New Zealand climatic data normalised for the period 1980–1999
Full size table

Both trials were assessed at age 10 for the following traits:

  • DBH: diameter (cm) at breast height 1.4 m

  • DEN: basic wood density (kg/m3) at breast height

  • HWAP: heartwood area percentage (%) at breast height

  • EPI: epicormic shoots, 0–3 score; 0 = none, 1 = few (1–4), 2 = moderate (5–10), 3 = many (10>)

Tree DEN and HWAP were measured on 5 mm diametric cores extracted at breast height from a subset of four of the eight replicates at each site. DEN was measured using the maximum moisture content method (Smith 1954). The length of the core containing heartwood (HWL) and the core length (CL) were measured to determine the HWAP of each tree:

$$HWAP = \left( {\frac{{\pi \left( {HWL/2} \right)^{2} }}{{\pi \left( {CL/2} \right)^{2} }}} \right) \times 100$$
(1)

Tree height was not measured since measurements on a subset of trees (n = 78) indicated a strong positive correlation (R 2 = 0.61) between height and DBH (data not shown).

Statistical analysis

Stepwise linear regression was used to select a range of independent environmental and geographic variables from the provenance’s native origin to create a linear model to explain the variation of the trait DBH observed at the two sites. The PROC REG STEPWISE procedure was used through SAS statistical software version 9.3 (SAS Institute, Cary, North Carolina, USA).

All traits were analysed using an individual-tree linear mixed effects model in ASReml-R (Butler et al. 2009), which is an implementation of ASReml in R Development Core Team (2008). Clonal material had no complex pedigree structure. The following model was used to estimate variance components for individual traits within each trial:

$$y = Xd + Z_{1} r + Z_{2} r.b + Z_{3} a + e$$
(2)

where y is the vector of observations on a trait, d is a vector of fixed effects (i.e. mean), r is a vector of random replicate effects, r.b is a vector of random block-within-replicate effects, a is a vector of random genetic effects estimated from individual genotypes and e is a vector of random residual effects. X, Z 1, Z 2 and Z 3 are known incidence matrices relating the observations in y to effects in d, r, r.b, and a respectively. To test the importance of county-of-origin, it was fitted as a fixed effect in the model and its significance tested with a Wald-F test. A one-tailed LR test of the improved model over the initial model was also used by comparing twice the change in REML log-likelihood against tabulated \(\chi_{ 1}^{ 2}\) The number of clones per county varied (Table 2) and the effect of adding county to the model using the LR test was found to be not significant. As clones were not able to be traced back to the original stool-beds, a propagation effect (i.e. stool-beds) was not included in the analysis. Seedling checklots were excluded in the analysis.

Table 2 Number of Sequoia sempervirens clones collected from each United States of America county and state (north to south), county identifier, and age ten survival of clones from each county at the two New Zealand sites
Full size table

Estimated variance components from each model were used to estimate individual broad sense individual heritability (\(\hat{H}_{i}^{2}\)) for each trait:

$$\hat{H}_{i}^{2} = \frac{{\hat{\sigma }_{g}^{2} }}{{\hat{\sigma }_{g}^{2} + \hat{\sigma }_{e}^{2} }}$$
(3)

where \(\hat{\sigma }_{g}^{2}\) is the genetic variance of individual clones, and \(\hat{\sigma }_{e}^{2}\) the residual effects. Standard errors of statistics were based on approximations using the Taylor series expansion (Lynch 1998) within ASReml.

Genetic correlations between traits were estimated as:

$$\hat{r}_{{g_{xy} }} = \frac{{\hat{\sigma }_{{g_{xy} }} }}{{\sqrt {\hat{\sigma }_{{g_{x} }}^{2} \hat{\sigma }_{{g_{y} }}^{2} } }}$$
(4)

where \(\hat{\sigma }_{{g_{xy} }}\) is the covariance between the traits, \(\hat{\sigma }_{{g_{x} }}^{2}\) is the genetic (additive and non-additive) variation for trait x and \(\hat{\sigma }_{{g_{y} }}^{2}\) is the genetic (additive and non-additive) variation for trait y. The estimated genetic (additive and non-additive) correlation \(\hat{r}_{{g_{xy} }}\) between traits x and y was obtained directly using the CORR directive to structure the G matrix for each model. Ramets of each clone were represented as replicates of the same genotype.

Genetic correlations across sites \(\hat{r}_{B}\) were estimated as in (4), except that \(\hat{\sigma }_{{g_{xy} }}\) was the covariance between a trait at site x and site y, \(\hat{\sigma }_{{g_{x} }}^{2}\) was the genetic (additive and non-additive) variation for the same trait at site x and \(\hat{\sigma }_{{g_{y} }}^{2}\) was the genetic (additive and non-additive) variation for the trait at site y.

The model (Eq. 2) was adjusted to include site effects, where site was fitted as a fixed effect and all model terms fitted within site. Type-B genetic correlation estimates \((\hat{r}_{B} )\) between genotypes at different environments for each trait were obtained directly using the CORR command within ASReml to structure the G matrix for the additive and non-additive genetic component estimated by each model; all other model terms were considered independent at each site. The significance of genetic correlations within and between sites from zero and from one was tested using the Log Likelihood test and comparing the model with no restrictions with a constrained correlation (either zero or one). A one-tailed LR test of the improvement of the initial model over the reduced model was performed by comparing twice the change in REML log-likelihood against tabulated \(\chi_{ 1}^{ 2}\).

Across-site clonal genetic values (which includes additive + non-additive genetic effects) were estimated using site as a fixed effect (as described above). Significance of genetic values was tested using a two-tailed Log Likelihood test as described above. We then investigated clonal genetic values further using box plots and stepwise linear regression to investigate variation among provenances by location (county). Across-site genetic gains from clonal selection was estimated from clonal genetic values from across-site analyses. Mean genetic values were estimated for each trait. An index (INDEX) was derived from all traits as the sum of genetic values from all traits. Prior to adding genetic values together to construct this index, genetic values were standardised for each trait by dividing each genetic value by the trait-specific maximum genetic value. Individual-trait and index-specific genetic gains from clonal selection were then estimated:

$$\left[ {\left( {{\text{G}}_{\text{sel}} {-}{\text{G}}_{\text{m}} } \right)/{\text{G}}_{\text{m}} } \right] \times 100$$
(5)

where Gsel is the average genetic value of the selection (top 10 % or top 20 %) for a given trait and Gm is the average genetic value of the population tested.

Results and discussion

Overall, age 10 survival at each site was generally good; 76 % at Awaho and 93 % at Thompson. The lower survival at Awaho could be due to the poorly draining soil conditions. Of the provenances tested, survival was always better at Thompson compared with Awaho (Table 2). At Awaho, survival was particularly bad for provenances from Alameda County (27 %). However, the survival for these same provenances was far greater at Thompson (81 %).

Clonal material for the Kuser trial was sampled from 98 different provenances that spanned over 700 km and 6.6° latitude (Fig. 1). The influence of provenance on the traits expressed by clones were tested in two ways, the environment at each clone’s provenance and grouping by county. A regression analysis of the environmental characteristics found several characteristics that were significant in explaining DBH variation, with elevation having the highest individual R 2 value (0.01). However, the six variables selected for the final model by the stepwise linear regression method had a small R 2 value of 0.04 (Table 3). When the traits were grouped by county, it appeared that there was considerable variation of each trait within each county (Figs. 2, 3).

Table 3 Provenance environmental characteristics selected by stepwise regression model for explanation of Sequoia sempervirens diameter at breast height (1.4 m) variation at the two sites
Full size table
Table 4 Variance component, variance (%) component proportion over total variance, individual (\(\hat{H}_{i}^{2}\)) broad-sense heritabilities and their standard errors (SE) for the Sequoia sempervirens traits studied at the two sites: diameter at breast height (DBH) (1.4 m), wood density (DEN), percentage heartwood (HWAP), and epicormic shoots (EPI)
Full size table
Fig. 2
figure 2

Range of Sequoia sempervirens provenance genetic values for 10-year diameter at breast height (DBH) (1.4 m) and wood density at DBH for two test sites Awaho and Thompson for each county of origin in the United States of America (the box plot indicates the range, the median, the 25th and 75th percentile of total genetic values for each county). Outliers are indicated by circles. The total number of provenances for each county are: 7, 21, 46, 38, 9, 10, 7, 4, 7 and 21 respectively. For county identities, see Table 2

Full size image
Fig. 3
figure 3

Range of Sequoia sempervirens provenance genetic values for 10-year diameter at breast height (DBH) (1.4 m) and wood density at DBH for the two test sites Awaho and Thompson for each county of origin in the United States of America (the box plot indicates the range, the median, the 25th and 75th percentile of total genetic values for each county). Outliers are indicated by circles. The total number of provenances for each county are: 7, 21, 46, 38, 9, 10, 7, 4, 7 and 21 respectively. For county identities see Table 2.

Full size image

When variance components were estimated in ASREML, county variation was significant for all traits, but variation of clones within county was only greater than the standard error of the estimate for one trait—HWAP (Table 5). Considering this inconsistency, we think that the number of clones per county may have contributed, as this was highly variable, with an average of 16, minimum of 3 and maximum of 47. Six of the ten counties tested also had 10 clones per county or less. We believe that the amount of data available to investigate these components is inadequate at this stage. We intend to assess and analyse further trials of this material and we are hoping that the greater numbers of individuals and sites will give us a clearer picture on how the variation partitions among clones, counties and overall genetic variance.

Table 5 individual variance component estimates and their associated standard errors (S.E.) for diameter at breast height (DBH) (1.4 m), wood density (DEN), percentage heartwood (HWAP), and epicormic shoots (EPI) showing low genotype × environment variation (county × site) and variation of clones within county of origin (clone within county)
Full size table

These results are reasonably consistent with research on genetic diversity of S. sempervirens that has shown high genetic diversity of growth traits and isozymes (Anekonda and Libby 1996; Rogers 1997, 2000). Several studies have found large genetic variation within populations and little genetic variation between populations (Anekonda 1992; Millar et al. 1985; Rogers 2000). Anekonda’s research of a young Kuser trial at the University of California’s Russell Reservation, near Berkeley California, found that among the variance components estimated, the clonal component was the highest (2–20 %), followed by stands within regions (2–13 %), and between regions (0–4 %). Rogers (2000) concluded that the high diversity was likely due to a number of factors. One of these factors is that S. sempervirens is a hexaploid (Ahuja and Neale 2002; Scott et al. 2015). Polyploidy in conifers is extremely rare, and in the case of S. sempervirens appears due to autopolyploidy, or the multiplication of the genome of the same species, as opposed to an allopolyploid, brought about by the fusion of two different genomes (Scott et al. 2015). This genome construct in theory would result in low genetic variation across the whole species due to an overall lower mutation rate than other conifers and highly homologous sequence. The evolutionary significance of this is suggested by Scott et al. (2015) to be one of survival through longevity, clonal propagation and selfing to promote further polyploid events. This is in contrast to the more widely considered evolutionary theory involving recombination and mutation to provide genetic variation for natural selection. This is supported by the finding by Douhovnikoff and Dodd (2011) that only a very weak divergence exists between the extremes of the natural range of the species. Epistatic variation and changes in methylation may also be more important here where clones are so important. This type of variation has been found to be important in gene–gene interactions in Populus tomentosa Carr. (Du et al. 2015), but remain untested in S. sempervirens. Implications for breeders are that genetic variation and mutation will likely different from most conifers, but can be expected to be greatest among provenances. This supports the general variation patterns determined in this study.

While this appears to be a consistent outcome it is important to note that this study did not take into account any ‘C-effects’ (i.e., factors that artificially inflate differences among clones). These effects are non-genetic and often associated with differences between ramets of the same clone, particularly physiological age (Aimers-Halliday et al. 2003; Libby and Jund 1963). These non-genetic effects are not usually accounted for in forest genetic studies (Foster and Shaw 1988), and were unfortunately not accounted for in this study as no information was available. It is likely that C-effects have inflated the genetic variances estimated here, but it is unknown as to the relative proportion this may be in comparison to the overall total genetic effects. Fully accounting for propagation effects, physiological age and full pedigreed structure (females, males, family and clone within family) within the experiment would allow for a full breakdown of the component genetic variances.

Without the knowledge of how variance might partition fully in these experiments or in this species it is difficult to understand how best to tackle selection and improvement in S. sempervirens. Evidence from other conifer species indicate that additive genetic variation is larger than non-additive genetic variation. In P. radiata, for example, additive genetic variance was approximately three times the estimated non-additive genetic variance (dominance and epistatic; Baltunis et al. 2009). In Cupressus lusitanica, non-additive genetic effects were found to be small or negligible compared with the significant additive genetic effects found for growth (diameter at breast height), form (branch size) and disease (cypress canker; Dungey et al. 2012).

It was also interesting that in the same study, when more data were used, the proportion of non-additive variance estimated increased to approximately the same as the additive effect. It is clear that a clone-within-family structure is needed to further understand this in S. sempervirens. Perhaps pedigree analysis using DNA markers to construct relationship matrices will help with existing trials (e.g. Telfer et al. 2015; Dodds et al. 2015). The fact that this species is a hexaploid will, no doubt, cause some challenges before this can be realised.

Despite the likelihood of C-effects, we believe that the total and significant (P < 0.001) genetic variation found in most traits to be large enough to warrant a selection programme. It is not known the relative proportion of additive and non-additive genetic variance, however, and before embarking on a large breeding program a trial to test this would be an excellent idea. Selfing over several generations might also be a good way to test the theory of autopolyploidy and the existence of spontaneous polyploidy events and its application for the production of new breeding lines.

Few genetic studies have sampled provenances of S. sempervirens across its natural range (Anekonda 1992; Anekonda and Libby 1996; Douhovnikoff and Dodd 2011; Kuser et al. 1995), and even these studies have not sampled intensively enough to completely capture genetic variability of the population. The aforementioned studies found that the entire population could be divided into northern and southern lineages. This disjunction between populations has been identified at San Francisco Bay, or in a more recent study at the Sonoma-Mendocino County border (36.8°N) (Douhovnikoff and Dodd 2011). Although the Kuser collection is currently the best representation of provenances and genetic variation across its natural range, it is not comprehensive. The provenances of some counties like Humboldt County were sampled more than others. However, the large variation of the traits within these counties suggests that the sample size may be too small to fully capture the genetic variation of the populations located there.

The results of this current study appear to show that there is large and significant total genetic variation within populations at the county level. This total genetic variation is advantageous for the establishment of a tree breeding program in New Zealand, as it provides the opportunity to select favourable genetic traits from a large genetic pool, even where that may be clonal, and improve upon the material currently grown.

Genetic parameters

Genetic differences among the provenances explained 35 % (Awaho) to 52 % (Thompson) of the variation for DBH (Table 4). Similar clonal variation was expressed for DEN, 52–30 % and HWAP, 40–50 % at Awaho and Thompson, respectively (Table 4). This is greater than in previous studies using the same clones from the same Kuser collection ramets. The Russell Reservation Kuser trial found after two growing seasons, the clonal variation for tree basal diameter and height was 7 and 20 %, respectively After four growing seasons, clonal variance in height decreased to 10 % (Kuser et al. 1995).

High genetic variances due to clones are reflected in the broad sense heritability (\(\hat{H}_{i}^{2}\)) estimates for these traits (Table 4). This suggests moderate to strong genetic control enabling considerable gain could be made from selections of these traits. Broad sense heritability is higher than what was found for similar traits in the Russell Reservation Kuser trial; 8 % for basal diameter and 13 % for stem volume index after the second growing season (Anekonda 1992). Basal diameter EPI had a low \(\hat{H}_{i}^{2}\) (Table 4) for both sites with Awaho \(\hat{H}_{i}^{2}\) substantially lower than Thompson (Fig. 3). Epicormic shoots typically develop in response to increased light from disturbance (Meier et al. 2012). The difference in expression between sites may be due to topography and aspect. The Thompson trial was located on a north facing hillside, which permitted direct sunlight to reach the stem for a substantial period of the day. Conversely, the Awaho trial was located on flat ground at the bottom of a valley. Our findings indicate that EPI expression may be dependent upon the local environmental conditions.

Across-site (Type B, \(\hat{r}_{B}\)) genetic correlations were high to very high (0.70–0.98; Table 6), indicating very low genotype × environment interaction, meaning that clonal rankings did not change greatly between sites. Although only two sites were available to estimate across-site genetic correlations, they were both sites well suited to the growth of S. sempervirens. Epicormic shoots had the lowest genetic correlation across sites (0.7, Table 6), but even this trait, which had a lower heritability (Table 4), had very consistent rankings across sites. These results are in contrast with a study of epicormic severity in young stands (<15 years) in California (O’Hara and Berrill 2009b). That study concluded that genetics only explained a small percent of the variation of the severity of epicormic sprouting.

Table 6 Sequoia sempervirens genetic correlations between traits at Awaho (above the diagonal) and Thompson (below the diagonal), genetic correlations (\(\hat{r}_{B}\)) between the same trait across sites appear on the diagonal in bold
Full size table

Previous research with the Kuser collection concluded that provenances from the northern lineages of its home range grew better in trials established in cooler climatic zones and vice versa for the collection grown in warmer climatic zones (Kuser et al. 1995). Anekonda’s (1992) study found that provenances from southern lineages performed the best at the Russell Reservation with latitude, longitude, elevation, distance from the ocean, and metabolic rate explaining 29 % of the variation. Metabolic rate and latitude were the most important variables. It was concluded from that study that physiological differences in stem peak metabolic activity was an important driver for these differences, which indicated regional site adaption (Anekonda et al. 1994). This current study did not find any significant differences in traits between counties or with the environmental factors of the provenance origin. The contrast in results from studies that sourced genetic material from the same provenance collection may be due to a combination of factors. First, the climatic and soil conditions at both the Awaho and Thompson sites were highly suitable for S. sempervirens. The northern lineages may be able to perform as well as the southern lineages on optimal sites. Second, this current study estimated heritability at a substantially older age than the earlier studies. The wood properties at age nine would be more developed in this study. Third, this current study was replicated on two different sites while Anekonda’s research was at one site. Kuser’s 1995 study incorporated data from nine different Kuser collections in three different countries. However, only the Russell Reservation site was successfully established as a fully replicated trial with three replicates. Thus, there may be some unique site effects that may have influenced the results. Finally, Russell Reservation’s climate may be unsuitable for most of the collection’s provenances as it was located east of San Francisco Bay and away from the Californian Coastal fog belt, where most of the redwood range occurs. Thus, the Kuser collection at the Russell Reservation may have experienced more water stress than at Awaho and Thompson. As the well performing provenances for this study came from a range of environments throughout its natural range, this could suggest that there is a degree of plasticity of some provenances to respond to favourable environmental conditions.

Diameter at breast height and DEN were negatively correlated (\(\hat{r}_{b} = 0.52\) and −0.58; Table 6), indicating larger trees tend to have a lower wood density. Thus, the selection for faster-growing trees will have a detrimental effect on wood density and vice versa. HWAP and DBH were highly genetically correlated (\(\hat{r}_{b} = 0.60\) and 0.73; Table 6).The percentage of heartwood area to tree cross-sectional area at breast height was not constant across the range of measured tree DBH. Trees with a larger DBH also had a higher percentage of their cross-sectional area as heartwood, compared with trees with a smaller DBH. The proportion of heartwood is an important factor for S. sempervirens logs to achieve the highest timber grade to maximise its return to the grower (Cown et al. 2013) growth and high percentage heartwood without sacrificing one for the other. Diameter at breast height and EPI were moderately genetically correlated (Table 6), suggesting that trees with high growth rates also had a propensity to produce a high number of epicormic shoots. As these shoots have a negative impact on timber grades, there may be a trade-off between growth rate and the number of epicormic shoots produced that a forester will need to balance.

The genetic parameters estimated in this study were from two sites that were highly suited to S. sempervirens growth. The similarities of the ranking between the two sites suggests that the rankings may be consistent if the provenances were grown on similar suitable sites in New Zealand. However, more trials within this series will need to be assessed, when they are sufficiently mature (age 9+), to determine if the same genetic correlations are consistent across a range of different environments in New Zealand. It should also be noted there is no family (male and female parent) structure in this analysis. This means that the additive (heritable) and non-additive (not heritable, but captured in clones) genetic variation could not be separated. The inheritance of traits from parents to offspring is likely to be smaller than what is indicated here, but is not fully predictable until these components are separated. The non-additive variation will include C-effects which are likely to include differences in physiological maturation among clones, which may have inflated genetic estimates.

Individual-trait genetic gains estimated from across-site clonal genetic values were substantial for DEN, moderate for HWAP and DBH and small for EPI and the derived INDEX (Table 7). It is encouraging to see that selection for HWAP would result in genetic gains that would be worth pursuing. Once multiple traits were selected in the INDEX, gains were substantially reduced, even with selection of the top 10 % of clones (Table 7). These results indicate that provenance selection would be promising, but limiting if both growth, wood density and heartwood area were required. The extent of transfer of this variation to the next generation is not known as we were unable to partition the additive and non-additive genetic variation in this study.

Table 7 Genetic gains (percent) for selection of Sequoia sempervirens clones across sites based on selection of the top 10 and 20 % for individual traits and for a multi-trait index comprising all traits with equal weightings. Traits are: diameter at breast height (DBH) (1.4 m), wood density (DEN), percentage heartwood (HWAP), and epicormic shoots (EPI)
Full size table

Conclusion

Clonal rankings were consistent across the two sites, suggesting that when grown on sites that are suitable for S. sempervirens, the best provenances at one site would likely remain the best provenances at another. Medium-to-high genetic control was found for all growth and wood property traits measured, enabling significant gains for these traits. As far as we are aware, this is the first study to show this high level of heritability on growth and wood properties from the Kuser collection. Epicormic shoots were under low genetic control and were moderately correlated with growth rate. Its expression may be also influenced by environmental light levels. Growing this species too fast is likely to require more attention to pruning where clear heartwood is the desired end-product. Fast growth was also associated with lower wood density. Careful selection of provenances will be required to maintain desired growth rate and wood quality of the final crop.

The results of this study suggest that no particular region from the natural range is more suited to growth in northern New Zealand and that clonal selections should be made from across the natural range. This study demonstrates that the genetic variation within populations is greater than between populations, at least at the county level. The results suggest that there may be a degree of plasticity of some provenances in their respond to favourable environmental conditions. The large genetic variation found in this study indicates that genetic selection would provide substantial gains, at least for individual traits, in developing S. sempervirens as a commercial timber production species in New Zealand. We recommend that further studies are initiated that ensure that the primary factors of genetic variation be properly partitioned, including C-effects and non-additive & additive genetic variation, as well as testing genetic variation of self-pollinated progeny. The best strategy for developing planting stock with exceptional growth and wood properties will then be able to be clarified.