GENETIC VARIATION in WOOD MECHANICAL PROPERTIES of CALYCOPHYLLUM SPRUCEANUM at an EARLY AGE in the PERUVIAN AMAZON Carmen Sotelo Montes Ph.D

GENETIC VARIATION IN WOOD MECHANICAL PROPERTIES OF CALYCOPHYLLUM SPRUCEANUM AT AN EARLY AGE IN THE PERUVIAN AMAZON Carmen Sotelo Montes Ph.D. Candidate Jean Beaulieu Senior Scientist Natural Resources Canada, Canadian Forest Service, Canadian Wood Fibre Centre 1055 du P.E.P.S., P.O. Box 10380, Stn. Sainte-Foy Québec (Quebec) Canada, G1V 4C7

and Roger E. Hernández† Professor Centre de recherche sur le bois Département des sciences du bois et de la forêt Université Laval Québec (Quebec) Canada, GIK 7P4 (Received February 2007)

ABSTRACT Calycophyllum spruceanum (Benth.) Hook. f. ex Shum. is an important timber species of the Peruvian Amazon Basin. Due to overexploitation in natural populations, users are turning to young trees of potentially lower quality. Therefore, variation in juvenile wood properties should be investigated to determine whether wood quality can be maintained or, if necessary, improved by breeding. A provenance/ progeny test was established to evaluate genetic variation in growth and wood properties of young trees, the strength of their genetic control, as well as their interrelationships both at the genetic and phenotypic ␴ levels. This paper presents results obtained for ultimate crushing strength ( L), the static compliance coefficient (s11) in longitudinal compression, the dynamic s11 in the longitudinal direction (determined by ultrasound), and air-dry density at 39 months. Results indicate that the mechanical properties of juvenile wood of this species are adequate for structural uses. There was significant variation in all wood properties

due to families within provenances, and in all but dynamic s11 due to provenances. Families accounted for a larger percentage of the total phenotypic variance than provenances. Heritability estimates were higher ␴ for L and static s11 than for dynamic s11 and density. Genetic correlations indicate that selecting trees with denser wood and/or faster growth would have a positive effect on some mechanical properties. A non-destructive ultrasonic method appeared suitable for estimating juvenile wood strength and stiffness of this species. Keywords: Provenance, family, environment, heritability, phenotypic and genetic correlations, juvenile wood.

INTRODUCTION under these conditions usually have a higher portion of juvenile wood compared with trees in Plantation forestry is an attractive manage- natural stands, and this can affect wood proper- ment option in the tropics given the high growth ties (Zobel and Sprague 1998; Bowyer et al. rates normally observed. However, trees grown 2003; Saranpää 2003). Understanding this varia- † Member of SWST. tion and the factors involved is essential to main-

tain or, if required, improve wood properties for populations, without considering management specific uses (Tsoumis 1991). There has been plans that would ensure that the genetic quality is relatively little research on genetic variation in maintained for subsequent generations. As a re- wood properties of juvenile wood, their herita- sult, users are becoming dependent on young bility, their correlation with tree growth, and trees, which may have lower genetic quality for their impact on end-use products (Zobel and timber. In response to this problem, efforts have Sprague 1998). This research is essential to de- been made to set up seed orchards to produce sign tree improvement programs that provide high-quality seed for the establishment of small- high-quality seed for plantation forestry in the scale forestry/agroforestry plantations in farm- tropics (Simons et al. 1994). ing communities (Weber et al. 2001). Strength and stiffness are important mechani- A provenance/progeny test of C. spruceanum cal properties that determine the wood’s suitabil- was established in the Peruvian Amazon Basin, ity for structural uses (Jozsa and Middleton and results evaluated at 39 months indicated that 1994). Wood density is usually a good predictor there was significant variation in tree growth and of strength and stiffness (Panshin and de Zeeuw wood density due to provenances and families 1980), but these properties can be influenced by within provenances. Wood density had a higher other factors, including the variability among heritability than growth, and these variables trees within species and environmental condi- were positively correlated at the genetic and tions that affect tree growth (Tsoumis 1991). phenotypic levels (Sotelo Montes et al. 2006). Destructive methods are usually used to evaluate This paper presents additional results from the mechanical properties. However, a large number same provenance/progeny test. The main objec- of living trees need to be evaluated using non- tives were to (a) determine the relative magni- destructive methods in order to create breeding tude of variation in juvenile wood mechanical ␴ or production populations with trees possessing properties (ultimate crushing strength, L, and the best attributes. Non-destructive acoustical static and dynamic parallel compliance coeffi-

methods have been developed to evaluate wood cients, s11) due to provenances and families properties (Herzig 1991; Bucur 2005), and sev- within provenances, (b) evaluate the heritability eral researchers have used these methods suc- of the mechanical properties, (c) evaluate the cessfully (Kyokong and Bello 1977; Hernández phenotypic and genetic correlations among tree et al. 1998; Oliveira et al. 2002; Ilic 2003). growth, wood density, and mechanical proper- Calycophyllum spruceanum (Bentham) ties, and (d) evaluate the usefulness of non- Hooker f. ex Shumann (Rubiaceae family), destructive methods for predicting crushing known as capirona in Peru, is an important hard- strength and static stiffness. Results are com- wood species for farming communities in the pared with those of other species, and some Peruvian Amazon (Sotelo Montes and Weber practical implications are discussed. 1997). Its dense, diffuse-porous wood (Keenan and Tejada 1984) is mainly used for construction poles, charcoal, and firewood, but there is also MATERIALS AND METHODS demand in national and international markets for Sample region, study area, experimental furniture, wall paneling, and parquet floors (To- design, and management of the ledo and Rincón 1996). Capirona is a pioneer provenance/progeny test species that colonizes the floodplain and dis- turbed forests in the Amazon of Peru, Brazil, The sample region and provenance/progeny Ecuador, and Colombia (Linares et al. 1992). test are located in the Aguaytía watershed of the Trees can attain heights of 35 m and stem diam- western Peruvian Amazon (Fig. 1). Open- eters of 1.8 m at breast height (Sears 2003). In pollinated seeds were collected on 200 selected Peru, farmers and industry have been harvesting mother trees of Calycophyllum spruceanum the best canopy-level trees in accessible natural growing in natural stands located in seven geo- 580 WOOD AND FIBER SCIENCE, OCTOBER 2007, V. 39(4)

FIG. 1. Geographic location of the sample region and study area in the Aguaytía watershed of the Peruvian Amazon. Inset shows the location of the seven provenances of Calycophyllum spruceanum and the three planting zones (bold ellipses). graphic locations (provenances) distributed in on each site. Dead trees were replaced during the the lower, middle, and upper parts of the water- first dry season, but data collected on replants shed. Seeds collected from the same mother tree were not included in the analyses. Although are referred to as a family because they share the each replication was established on a relatively same maternal parent. The test was established homogenous site, the size of the replication with on sites with upland, non-alluvial soils. In gen- border rows was relatively large (∼0.3 ha), and eral, soil fertility and mean annual rainfall in- this could have resulted in fairly large microen- crease from the lower to the upper parts of the vironmental differences within replications. watershed. Details about the distribution of C. Management practices included a cover crop, spruceanum in the sample region, sampling pro- fertilizer application, branch pruning, and cedures, tree selection criteria, and the climate manual weeding (Sotelo Montes et al. 2006). and soils in the sample region and study area are One tree in each experimental plot was selec- given elsewhere (Sotelo Montes et al. 2006). tively thinned 39 months after planting, based on The experimental design was a randomized tree form (primarily stem bifurcations in the complete block with 15 replications: five repli- canopy) and growth. cations were established on different farms in each of the lower, middle, and upper parts of the Traits measured in the provenance/ watershed (hereafter called planting zones). In progeny test each replication, each of the 200 families was randomly assigned to one of 200 experimental Measurements of mechanical properties were plots, with two trees per plot. Spacing was 2.5 made on the thinned trees of only six of the by 2.5 m within and between rows. Two rows of fifteen replications: three each in the middle and border trees surrounded the experimental design upper zones of the watershed. The other nine Sotelo Montes et al.—VARIATION IN MECHANICAL PROPERTIES OF CALYCOPHYLLUM SPRUCEANUM 581

replications (two each in the middle and upper sections of the sticks were 2.5 × 2.5 cm for zones, and the five in the lower zone) were ex- larger trees (group A, dbh greater than approxi- cluded because mean diameter of the trees was mately 7.0 cm), and 2.0 × 2.0 cm for smaller too small to provide enough samples of suffi- trees (group B, dbh less than approximately 7.0 cient size for mechanical tests. Prior to thinning, cm). Sticks were stored under controlled condi- tree height was measured to the nearest cm using tions (60% relative humidity – RH, 20°C) for a meter stick or a telescopic measuring pole. approximately 6 months to attain equilibrium before (12.2% ס After thinning, stem diameter at breast height moisture content (EMC mean (dbh, 1.3 m above ground level, outside the processing. In the second step, one defect-free bark) was measured to the nearest 0.1 cm using sample without the pith was prepared from calipers for small trees and diameter tape for each stick. Two sizes were used for measure- larger trees. ments: 8 × 2 × 2 cm for the larger trees, and Although the thinned trees were not randomly 6 × 1.5 × 1.5 cm for the smaller trees. This was selected from the two trees in each experimental done in order to sample the maximum number of plot, it is unlikely that this would affect the re- trees in each planting zone, and to have the last sults about the relative magnitude of variation in two growing seasons represented in the samples. wood properties due to provenances and families As a result, data would have to be adjusted to within provenances, as well as the correlations eliminate any difference in air-dry density and between tree growth and wood properties. The mechanical properties due to sample dimension. ␴ following reasons can be advanced to support Measurements of L, static s11 and air-dry this assumption: (1) Wood properties were not density were based on procedures in ASTM used as selection criteria in the thinning. (2) 1994, except for the sample dimensions. Air-dry Mean height of the thinned trees was essentially volume of each sample was calculated from its the same as the mean height of all trees in the six length and cross-sectional area, and air-dry replications (8.6 m and 8.4 m, respectively). Dbh weight was measured to the nearest 0.001 g. The was only measured on the thinned trees, but as it ratio of air-dry weight to volume was used to was highly correlated with height (Sotelo Mon- estimate air-dry density (kg/m3). Parallel-to- tes et al. 2006), we expect the same results. (3) grain compression tests were then carried out on Wood samples were obtained from the lower a MTS ALLIANCE RT/50 machine. Strain in stem, well below the stem bifurcations in the the longitudinal direction was measured over a canopy. This sampling procedure was used for span of 50 mm (group A) or 40 mm (group B) in all plots so it should not have any significant the central part of the sample, using a two-side bias on mean height, dbh, or wood properties of clip gauge provided with a Sangamo DG1.0 lin- any particular family. ear displacement sensor. Cross-head speed was Air-dry density, ultimate crushing strength in set to 0.67 mm/min for group A and 0.50 mm/ ␴ the longitudinal direction ( L), and static and min for group B in order to obtain a similar ␴ dynamic parallel compliance coefficients (static elastic strain rate of approximately 0.3%. The L and dynamic s11, respectively) were determined was obtained from the maximum load at failure from one wood sample per tree. The static and and cross-sectional area of the sample. Test re-

dynamic s11 are the reciprocals of the static (Es) sults were also used to calculate the compliance and dynamic (Ed) modulus of elasticity, respec- coefficients in the longitudinal direction (static tively. s11). Finally, the samples were oven-dried at Sampling of wood specimens for determina- 103°C for 48 h and then cooled to room tem- ␴ tion of air-dry density, L and static s11 involved perature over phosphorous pentoxide. Oven-dry two steps. In the first step, a 45-cm-long stick weight was then measured to the nearest 0.001 g was prepared: this was located close to the bark, and EMC was calculated.

in the south-facing quadrant of the stem, be- The dynamic s11 was determined for trees tween 75 and 120 cm above ground level. Cross- from 101 families. Families were selected for the 582 WOOD AND FIBER SCIENCE, OCTOBER 2007, V. 39(4)

dynamic s11 measurements using a stratified ran- densitometry for a 20-mm-long section of wood dom sampling procedure. The number of fami- sampled at breast height and carefully matched lies per stratum (provenance) was proportional to the 20-mm-long segment (hereafter called ␳ to the total number of families in the stratum. either density of the slice or 20). It should be The dynamic s11 was determined for 441 trees, noted that the values were not corrected to of which 323 were also used for the determina- take into account the effect of Poisson’s ratio on ␴ tion of L, static s11 and air-dry density of the the wood segments, so they are approximate val- specimen. A 10-mm increment core was ex- ues for dynamic s11 (Bucur 1981). Finally, the tracted at breast height from each tree after thin- EMC of the segments was determined after the ning. The core extended from bark to bark test following the procedure previously de- through the pith following the north-south ori- scribed. entation. The longitudinal direction of each core was marked immediately after extraction. Sub- sequently, the cores were conditioned at 20°C Statistical analyses and 60% RH for about 6 months to attain an EMC of 11.7%. A 20-mm-long segment of the The SAS௡ statistical package, version 9.1 core from the south side of the tree, close to the (SAS Institute Inc. 2002–2003), was used for all bark, was then prepared for the test. statistical analyses. Data transformations were

The dynamic s11 was measured using an ul- not required to satisfy the assumptions of analy- trasonic method described in detail by Herzig sis of variance and other analyses. Analyses of (1991) and Yang and Fortin (2001). Each seg- tree height and stem diameter are reported ment was placed between two ultrasonic trans- elsewhere (Sotelo Montes et al. 2006).

ducers (transmitter and receiver), anda1MHz Analyses of variance of dynamic s11 and co- frequency wave was propagated in the longitu- variance of the other wood properties were car- dinal direction through the segment. The time ried out across and by planting zones (GLM pro- taken by the wave to pass through the segment cedure, partial sums of squares estimation −6 was read to the nearest 10 of second. Three method). Variance of dynamic s11 was analyzed replicated measurements were made on each according to a mixed linear model with the fol- segment, and time readings were only accepted lowing sources of variation: zone, replication when a maximum amplitude and well-defined within zone [Rep(Zone)], provenance (Prov), transmitted wave onset could be seen on the os- family within provenance [Fam(Prov)], and the cilloscope display. The distance of wave propa- interactions [Prov*Zone, Prov*Rep(Zone), ␴ gation was equal to the diameter of the segment: Fam(Prov)*Zone]. For density, L and static s11, this was measured to the nearest 0.001 mm. A the model also included dimension (Dim) and correction factor was applied to the calculation equilibrium moisture content (EMC) of the ␴ of wave velocity through the core in order to wood sample as covariates because L and static compensate for the error caused by the presence s11 were significantly affected by EMC in the and 0.012 ס of the coupling medium (neoprene membrane) middle zone of the watershed (P and the transport of electric waves within the 0.026 with 283 and 393 trees, respectively). measuring circuit (Herzig 1991; Yang and Fortin Zone was treated as a fixed factor and the others 2001). A plexiglas core, with the same dimen- as random factors. Some F-ratios involved more sion as the wood core, was used as a reference to than one mean square in the denominator

determine the correction factor. The dynamic s11 (“quasi” F-ratios), and were tested with approxi- was calculated using the following equation: mate degrees of freedom. Analyses were also ס ␳ 2 −1 −1 ␳ ס s11 ( v ) (MPa) , where air-dry den- carried out within each zone if there was a sig- velocity nificant difference in the wood property due to ס sity (kg/m3) at time of testing, and v (m/s) of wave propagation. The air-dry density zones and/or a significant interaction between used in this formula was determined using X-ray zones and provenances/families. Sotelo Montes et al.—VARIATION IN MECHANICAL PROPERTIES OF CALYCOPHYLLUM SPRUCEANUM 583

Data for density of the compression specimen, RESULTS AND DISCUSSION ␴ and static s were adjusted across zones for L 11 Means and coefficients of variation of wood Dim and EMC (based on covariate relation- mechanical properties and density ships) but not for dynamic s11 and density of ␳ the slice ( 20). Variance components were esti- Mean air-dry density of Calycophyllum mated using the VARCOMP procedure with spruceanum wood was significantly greater the restricted maximum likelihood method. Phe- (P < 0.001, 323 trees, paired t-test) for the slice 3 ס ␳ notypic correlations (Pearson r) were calculated ( 20 761 kg/m ) than for the compression 3 at the tree level (CORR procedure). Simple specimen (717 kg/m ). It is likely that the mean ␳ and multiple linear regressions (REG procedure) density of 20 was greater because (a) density using the same sample size (323 trees) were increases from the pith to the bark (Sotelo Mon- ␴ tes et al. unpublished data), (b) the 20-mm slices used to develop models for predicting L and static s from the independent variables ␳ and and the compression specimens were obtained 11 20 close to the bark, but the compression specimens dynamic s . For the multiple linear regres- 11 varied more in the radial position given the ma- sion models, the stepwise method was used and chining requirements (orientation of samples, they were compared using Akaike’s informa- jointing, and planing), (c) the slices were ob- tion criterion (AIC), the model with the lowest tained at 1.3 m above ground whereas the com- value for AIC being the best model. Nar- 2 pression specimens were obtained between 0.75 row sense heritability (hi ) based on individual and 1.2 m above ground, and (d) density was trees, genetic correlations, and their standard estimated by two different procedures and with a errors were estimated using formulas described different number of trees and families. elsewhere (Becker 1984; Falconer and Mackay ␴ Mean values for L and density of juvenile 1996). Phenotypic correlations include both wood in the present study (Table 1) were lower genetic and environmental effects, whereas ge- than those reported by Keenan and Tejada netic correlations reflect only genetic effects (1984) for mature wood of C. spruceanum. This (Falconer and Mackay 1996). Trait values was expected based on differences in density were standardized (Steel et al. 1997) for calcu- and other properties between juvenile and ma- 2 lation of genetic correlations. Heritability (hi ) ture wood (Tsoumis 1991). However, the spe- ␴ was estimated assuming partial inbreeding cific strength (ratio of L to density of the com- (Sotelo Montes et al. 2006); in this case, addi- pression specimen) of C. spruceanum juvenile ␴2 tive genetic variance was estimated as 3 f. wood (70) was similar to that of mature wood These assumption and estimation methods have (72, Keenan and Tejada 1984). This means that been used for some other tropical hardwood spe- when adjusted for differences in density, both cies (Hodge et al. 2002; Hodge and Dvorak juvenile and mature woods of C. spruceanum 2004) in order to provide a conservative estimate have a similar strength in parallel compression, 2 of hi . even though they might differ in several charac-

␴ TABLE 1. Descriptive statistics of density, ultimate crushing strength ( L), static and dynamic parallel compliance coef- a ficients (s11) for Calycophyllum spruceanum wood at 39 months.

Trait Mean CVb Range Nc Air-dry density (kg/m3) 718 5.9 250.0 676 ␴ Ultimate crushing strength, L (MPa) 50.1 10.1 28.2 676 −1 Static s11 (TPa ) 75.3 18.2 77.2 676 −1 Dynamic s11 (TPa ) 77.3 15.7 72.2 441 a ␴ Data for density, L and static s11 are adjusted for dimension and equilibrium moisture content of the wood sample (as covariates). b Coefficient of variation. c Number of trees. 584 WOOD AND FIBER SCIENCE, OCTOBER 2007, V. 39(4) when compared using the same (0.80 ס teristics, including microfibril angles, proportion t-test, P of rays, fiber length, cell-wall thickness, as well sample size (323 trees). However, when the

as the content of lignin, extractives, and other mean dynamic and static s11 were expressed as chemical constituents (Skaar 1988; Tsoumis specific values (i.e., divided by their density) 1991; Lei et al. 1997; Zobel and Sprague 1998; and compared for the same number of trees, the

Evans et al. 2000; Bowyer et al. 2003). The spe- specific dynamic s11 was about 6% lower than cific strength of wood varies among species. For the specific static s11, and this difference was example, compared with other tropical hard- statistically significant (paired t-test, P < 0.001). woods, the specific strength in parallel compres- This result is consistent with previous values sion of juvenile wood of C. spruceanum reported for other hardwood species (Bodig and was lower than that of Tectona grandis wood Jayne 1982; Bucur 1983; Hernández and Re- evaluated at 8 years in plantation (78, air-dry strepo 1995; Hernández et al. 1998; Oliveira et .(kg/m3; Rivero 2004), and similar al. 2002; Bucur 2006 580 ס density

to mature wood of Cariniana domestica, Copai- The CVs for the static and dynamic s11 of C. fera officinalis, and Terminalia guianensis spruceanum wood (Table 1) were similar to grown in natural stands (71, 70, and 72, respec- those reported for other species (Hernández and and 740 Restrepo 1995; Bucur 2005). Judging from the ,730 ,720 ס tively, with air-dry density kg/m3, respectively; Keenan and Tejada 1984). CVs, there was relatively more variation in the ␴ Juvenile wood of C. spruceanum is relatively static and dynamic s11 than in L. stiff, which is also important for structural uses. Its specific stiffness in parallel compression (ra- Variation in wood mechanical properties tio of static modulus of elasticity [E ] to density s and density was higher (18494 ס of the compression sample than the specific stiffness in static bending of Environmental conditions affect tree growth juvenile wood of T. grandis (17812; Rivero and can indirectly produce variation in mechani- 2004). It was also higher than the specific stiff- cal properties (Tsoumis 1991). For example, sig- ness in static bending of C. officinalis mature nificant differences between sites were observed

wood, but lower than that of C. domestica and T. in dbh and Ed of juvenile wood of selected guianensis mature wood (16529, 22345, and clones of Cryptomeria japonica (Nakada et al. 18957, respectively; Keenan and Tejada 1984). 2003). However, in the present study, there were

The dynamic s11 is usually lower than the no significant differences in density and me- static s11, as observed in mature wood of coni- chanical properties between the two planting fers (Bodig and Jayne 1982) and hardwoods zones (Table 2). Similar results were observed (Bucur 1983; Oliveira et al. 2002). The differ- for basic density of cross-sectional disks (Sotelo

ence between the dynamic and static s11 tends to Montes et al. 2006) and density of wood slices be smaller in parallel compression tests than in (Sotelo Montes et al. unpublished data) sampled bending tests (Herzig 1991). This is because the at breast height in trees from the same two ultrasonic method, which is used to determine zones. The environmental difference between

the dynamic s11, deals mostly with elastic effects the two planting zones was probably not large (Bucur 1983), whereas the static bending test enough to produce significant differences in deals with shear effects (Ilic 2001). Moreover, these wood properties, even though there were the difference between the dynamic and static significant differences in tree growth between

s11 in static bending tests is smaller for incre- zones (Sotelo Montes et al. 2006). ment cores than for standard specimens (Bucur There was statistically significant variation in 1983). all wood properties due to families within prov- The mean values for the dynamic and static enances, although variation in air-dry density of −1 s11 (75.3 and 75.4 TPa , respectively; data not the compression specimen was barely significant tabled) were not significantly different (paired (Table 2). In addition, variation due to prov- Sotelo Montes et al.—VARIATION IN MECHANICAL PROPERTIES OF CALYCOPHYLLUM SPRUCEANUM 585

␴ TABLE 2. Analysis of covariance of air-dry density, ultimate crushing strength ( L) and the static parallel compliance coefficient (static s11), and analysis of variance of the dynamic parallel compliance coefficient (dynamic s11) for Calyco- phyllum spruceanum wood at 39 months.a,b

␴ Density L Static s11 Dynamic s11 Source of variation DF P>F VAR P>F VAR P>F VARDF P>F VAR Zone 1 0.065 — 0.140 — 0.143 — 1 0.237 — Rep(Zone) 4 0.368 — 0.110 — 0.461 — 4 0.293 — Prov 6 <0.001 3.8 0.001 4.6 0.005 2.6 6 0.098 4.7 Fam(Prov) 189 0.054 7.7 <0.001 18.2 <0.001 15.2 94 0.014 6.9 Prov*Zone 6 0.999 0.0 0.909 0.1 0.995 0.0 6 0.339 0.0 Prov*Rep(Zone) 23 0.161 0.0 0.126 0.0 0.230 0.0 24 0.170 1.3 Fam(Prov)*Zone 154 0.088 8.2 0.154 1.9 0.713 0.0 89 0.726 0.0 Dim 1 0.198 — 0.599 — 0.922 ——— — EMC 1 0.916 — 0.339 — 0.307 ——— — Residual 289 — 80.3 — 75.2 — 82.2 216 — 87.1 Total 674 100.0 100.0 100.0 440 100.0 a ␴ Dimension (Dim) and equilibrium moisture content (EMC) of the wood sample are covariates for density, L and static s11. .percentage of the total phenotype variance explained by the variance component ס significance of F ratio; VAR סdegrees of freedom;P>F ס b DF

enances was significant for all properties except al. 1994) and provenances of T. grandis (Bhat

dynamic s11. Families within provenances ac- and Priya 2004). counted for a larger percentage of the total phe- The dimension of specimens (covariate Dim notypic variance (VAR) than provenances, es- in Table 2) did not have a significant effect on ␴ ␴ pecially for L and static s11. In addition, the density, L or static s11 in the analysis of covari- percentage of variance explained by families ance across zones. This is consistent with pre- within provenances was larger for static than for vious research investigating the effect of sam-

dynamic s11. Moreover, provenance and family ple dimension on wood properties (Hernández rankings in mechanical properties and density 1993; Ilic 2003), and is likely related to the fact were relatively stable across zones, judging by that the samples had the same length:width ratio the fact that variation due to the interactions (4:1). (provenance by zone, family within provenance by zone) was not statistically significant. These Heritability of juvenile wood mechanical results suggest that greater gains in wood me- properties and density chanical properties could be achieved by select- 2 ing the best families within provenances after Estimates of narrow-sense heritabilities (hi ) ␴ having selected the best provenances. However, were moderately high for L (0.51) and static s11 the most appropriate strategy depends on several (0.47), and relatively low for air-dry density of factors, including the amount of phenotypic the compression specimen (0.24) and dynamic

variance in the trait, its heritability, and its ge- s11 (0.22) (Table 3). Therefore, considering only 2 netic correlations with other traits (discussed be- hi , selection would be most effective if based on ␴ low), as well as economic considerations L, slightly less effective if based on static s11, (Namkoong et al. 1988). and least effective if based on air-dry density

Genetic variation in mechanical properties has and dynamic s11. also been observed in other hardwood species. Other researchers have also reported moderate 2 For example, significant variation has been re- to high hi values for strength and stiffness in ported for juvenile wood of hybrid poplar clones parallel to the grain compression of juvenile 2 ␴ (Hernández et al. 1998) and families of Euca- wood. For example, hi was 0.57 for L of Eu- lyptus grandis (Santos et al. 2003), and for ma- calyptus grandis (Santos et al. 2003). Broad- ture wood of clones of C. japonica (Fujisawa et sense heritability of a trait is generally higher 586 WOOD AND FIBER SCIENCE, OCTOBER 2007, V. 39(4)

TABLE 3. Heritability of density, ultimate crushing strength greater strength and stiffness (i.e., lower value of (␴ ), static and dynamic parallel compliance coefficients L static and dynamic s ), as reported for mature (s ) for Calycophyllum spruceanum wood at 39 months.a 11 11 wood in a number of other hardwood species Standard Number (Bodig and Jayne 1982; Tsoumis 1991; Hernán- Trait Heritability error of trees dez et al. 1998; Bowyer et al. 2003; Hernández Density (kg/m3) 0.24 0.17 676 ␴ 2007). In diffuse-porous hardwoods, generally L (MPa) 0.51 0.18 676 −1 there is little or no relationship between tree Static s11 (TPa ) 0.47 0.18 676 −1 Dynamic s11 (TPa ) 0.22 0.19 441 growth and the wood’s density or mechanical a ␴ Data for density, L and static s11 are adjusted for dimension and equi- properties (Saranpää 2003). For example, no sig- librium moisture content of the wood sample (as covariates). nificant correlation was observed between tree growth and (a) the mechanical properties of sev- than its narrow-sense heritability because its nu- eral diffuse porous hardwoods (Zhang 1995), (b)

merator includes both additive and non-additive density and dynamic s11 for mature wood of Al- genetic variances (Falconer and Mackay 1996). nus acuminata H.B.K. (Hernández and Restrepo ␴ ␴ Broad-sense heritability was 0.47 for L, 0.34 1995), and (c) bending properties, Es and L for for static s11 (both determined using small, clear juvenile wood of Alnus rubra Bong (Lei et al. standard samples) and 0.66 for dynamic s11 (de- 1997). Hernández et al. (1998) noted a weak termined using increment cores) of juvenile negative correlation between growth rate and ju- wood from hybrid poplar clones (Hernández et venile wood density of hybrid poplar clones, al. 1998), and ranged from 0.60 to 0.86 for Ed with a correlation between growth rate and me- (determined by a non-destructive method) of chanical properties that was either weak and mature wood from C. japonica clones (Fujisawa negative or not significant. However, due to the et al. 1992). positive correlation between growth and density in C. spruceanum (Weber and Sotelo Montes Phenotypic and genetic correlations between 2005; Sotelo Montes et al. 2006), larger trees the tree’s growth and wood mechanical also tended to have wood with greater strength properties and density and stiffness. Genetic correlations indicated the same gen- Phenotypic correlations (Pearson r, Table 4) eral relationships as those shown by the pheno- indicated that denser wood tended to have typic correlations (Table 4). However, with the

␴ TABLE 4. Pearson and genetic correlations between ultimate crushing strength ( L), static and dynamic parallel compliance coefficients (s11), and the tree’s height, stem diameter at breast height (dbh) and density for Calycophyllum spruceanum wood at 39 months. The significance (for Pearson r) or standard error (for genetic correlations) is given in parentheses, followed by the number of trees involved in the calculation.a

Pearson correlations Genetic correlationsb Trait Height Dbh Density Height Dbh Density ␴ L 0.210 0.094 0.697 0.085 0.143 0.749 (0.009) (0.014) (<0.001) (0.338) (0.255) (0.146) 674 676 676 674 674 674

Static s11 −0.156 −0.067 −0.443 −0.238 −0.204 −0.360 (<0.001) (0.080) (<0.001) (0.358) (0.278) (0.322) 674 676 676 674 674 674 c Dynamic s11 −0.282 −0.266 −0.445 −0.507 −0.753 UND (<0.001) (<0.001) (<0.001) (0.247) (0.162) 438 441 323 438 438 a ␴ Data for density, L and static s11 are adjusted for dimension and equilibrium moisture content of the wood sample. b Genetic correlations that are larger than their standard error are underlined. ס c UND undefined: Estimate of variance component for dynamic s11 was zero (using 323 trees), so correlation is undefined and standard error cannot be calculated. Sotelo Montes et al.—VARIATION IN MECHANICAL PROPERTIES OF CALYCOPHYLLUM SPRUCEANUM 587

exception of the dynamic s11, all genetic corre- with desirable wood mechanical proper- ␳ lations with tree height and dbh had large stan- ties (Table 5). Indeed, density ( 20) of the wood ␴ dard errors and cannot be considered statistically slice was more effective for predicting L than ס 2 significant. These results suggest that any gain static s11 (R 0.304 and 0.124, respectively, in growth following selection and breeding P < 0.001). could also bring about some gain, albeit small, in ␳ The inclusion of both 20 and dynamic s11 as wood stiffness. independent variables in a multiple regression model did improve the ability of these non- ␴ Predictive ability of non-destructive methods destructive methods to predict L, based on Akaike’s information criterion (Table 5). In- The phenotypic correlations between wood deed, the multiple linear regression model had strength and stiffness were strong, and in agree- the lowest value for Akaike’s information crite- ment with the results of previous studies (Bodig rion, indicating that it was better than the simple for the 869 ס and Jayne 1982; Tsoumis 1991; Bowyer et al. linear regression models (AIC ␴ 2003). As expected, the correlation between L multiple regression model using dynamic s and 11 ס ס and static s11 (r −0.812, n 676, P < 0.001) ␳ , 916 for the simple regression model using ␴ 20 was stronger than that between L and dynamic dynamic s , and 923 for the simple regression 11 ס ס s11 (r −0.567, n 323, P < 0.001). This model using ␳ ). Although both independent ␴ 20 is because L and static s11 were measured variables were statistically significant from the same specimen, whereas dynamic s 11 (P < 0.001) and together explained approxi- was measured from an increment core (even mately 42% of the variation in ␴ , most of this though it was extracted close to the static speci- L variation was explained by dynamic s (32%). men). 11 The non-destructive ultrasonic method ap- For predicting static s11 when dynamic s11 was pears to be useful for estimating the static in the equation, the additional variation ex- ס ␳ strength and stiffness of juvenile wood of C. plained by 20 was not significant (P 0.07, multiple regression not shown in Table 5). spruceanum (Table 5). Dynamic s11 explained ␴ Therefore, dynamic s was better than ␳ for approximately 30% of the variation in L and 11 20 . predicting both ␴ and static s ס 2 static s11 (R 0.322 and 0.303, respectively, L 11 P < 0.001). Stronger relationships in these prop- In conclusion, dynamic s11 was the best pre- ␴ erties were, however, reported for other hard- dictor of static s11, and L was better predicted ␳ wood species at older ages (Bucur 1983; when considering both dynamic s11 and 20 as Hernández et al. 1998; Oliveira et al. 2002). independent variables. However, these results Wood density might also be a useful criterion are based on juvenile wood and cannot be ex- for the indirect selection of C. spruceanum trees trapolated to mature wood.

␴ TABLE 5. Simple and multiple linear regressions for predicting ultimate crushing strength ( L) and static parallel compliance coefficient (static s11)ofCalycophyllum spruceanum wood at 39 months. Independent variables include air-dry ␳ density of the slice ( 20) and dynamic s11, which were measured using non-destructive methods.

Dependent variable Regression equationa R2b CVc ␴ ␳ L 4.2 + 0.060 ( 20) 0.304 8.3 ␴ L 69.8 − 0.26 (dynamic s11) 0.322 8.2 ␴ ␳ L 33.9 + 0.039 ( 20) − 0.18 (dynamic s11) 0.417 7.7 ␳ Static s11 153.8 − 0.103 ( 20) 0.124 16.7 Static s11 23.4 + 0.69 (dynamic s11) 0.303 14.9 ס ס a ␴ Pr >[t] <0.001 for regression coefficients in all models with 323 trees. Means for dependent and independent variables: L 49.9 MPa, static s11 75.4 3 ס ␳ −1 ס −1 TPa , dynamic s11 75.3 TPa , 20 761 kg/m . .0.001 ס Coefficient of determination of model. All models are significant at P ס b R2 .(%) Coefficient of variation of model ס c CV 588 WOOD AND FIBER SCIENCE, OCTOBER 2007, V. 39(4)

CONCLUSIONS ICRAF, who collaborated throughout the entire research project. The provenance/family test Results indicate that (a) juvenile wood of C. was supported by grants to ICRAF from the In- spruceanum is relatively strong and stiff, (b) teramerican Development Bank, the Govern- there is genetic variation in juvenile wood me- ment of Spain, the Government of The Nether- chanical properties, and (c) a greater proportion lands, and the Government of Norway as part of of this variation occurs among families within the CGIAR Global Initiative for Alternatives to provenances rather than among provenances. Slash and Burn, the Department for International Heritability estimates suggest that selection Development of the United Kingdom, and Win- would be more effective for ultimate crushing rock International as part of the USAID Alter- strength and the static parallel compliance coef- native Development Program. ficient than for air-dry density of the specimen and the dynamic parallel compliance coefficient. Genetic correlations indicate that selecting trees REFERENCES with faster growth and denser wood will in- AMERICAN SOCIETY FOR TESTING AND MATERIALS (ASTM). crease both strength and stiffness of the wood. 1994. Standard methods of testing small clear specimens These results, combined with results from an- of timber. ASTM D143–94. Annual book of ASTM other study of genetic variation in tree growth, Standards 4.10. Philadelphia, PA. Pages 23–52. suggest that there is potential to simultaneously BECKER, W. A. 1984. Manual of quantitative genetics, 4th improve tree growth, density, and some me- ed. Academic Enterprises, Pullman, WA. 191 pp. BHAT,K.M.,AND P. B. PRIYA. 2004. Influence of prov- chanical properties of juvenile wood of this spe- enance variation on wood properties of teak from the cies. In addition, the non-destructive ultrasonic Western Ghat region in India. IAWA J. 25(3):273–282. method appears suitable for estimating mechani- BODIG, J., AND B. A. JAYNE. 1982. Mechanics of wood and cal properties of juvenile wood. Additional re- wood composites. Van Nostrand Reinhold, New York, search is needed to evaluate (a) genetic correla- NY. 736 pp. BOWYER, J. L., R. SHMULSKY, AND J. G. HAYGREEN. 2003. tions between mechanical properties of juvenile Forest products and wood science: An introduction, 4th and mature wood; (b) heritability of mechanical ed. Iowa State Press, Ames, IA. 564 pp. properties of mature wood; (c) genetic correla- BUCUR, V. 1981. Détermination du module d’Young du bois tions among tree growth, density and mechani- par une méthode dynamique sur carottes de sondage. cal properties of mature wood; and (d) the ability Ann. Sci. For. 38(2):283–298. of the non-destructive ultrasonic method to pre- ———. 1983. An ultrasonic method for measuring the elastic constants of wood increment cores bored from living dict mechanical properties of mature wood of trees. Ultrasonics 21(3):116–126. this species. ———. 2005. Ultrasonic techniques for nondestructive testing of standing trees. Ultrasonics 43(4):237–239. ———. 2006. Acoustics of wood, 2nd ed. Springer-Verlag, ACKNOWLEDGMENTS Berlin, Germany. 400 pp. EVANS, J. W., J. F. SENFT, AND D. W. GREEN. 2000. Juvenile We acknowledge the generous grants to the wood effect in red alder: Analysis of physical and me- first author and support for research expenses chanical data to delineate juvenile and mature wood provided by the Organization of American zones. Forest Prod. J. 50(7–8):75–87. States, Université Laval, the Canadian Forest FALCONER,D.S.,AND T. F. C. MACKAY. 1996. Introduction to quantitative genetics. Addison Wesley Longman Lim- Service (CFS), and the World Agroforestry Cen- ited, Edinburgh, UK. 480 pp. tre (ICRAF). We are also grateful to the collabo- FUJISAWA, Y., S. OHTA,K.NISHIMURA, AND M. TAJIMA. 1992. rating farmers and technicians in Peru, and to Wood characteristics and genetic variations in Sugi scientists (Prof. Yves Fortin and Aziz Laghdir), (Cryptomeria japonica). Clonal differences and correla- statisticians, technicians, and editors at Univer- tions between locations of dynamic moduli of elasticity and diameter growths in plus-tree clones. Mokuzai Gak- sité Laval, CFS (Marie Deslauriers, Isabelle La- kaishi 38(7):638–644. marre) and ICRAF (Tony Simons, Jonathan ———, ———, ———,T.TODA, AND M. TAJIMA. 1994. Cornelius); and especially to John C. Weber of Wood characteristics and genetic variations in Sugi Sotelo Montes et al.—VARIATION IN MECHANICAL PROPERTIES OF CALYCOPHYLLUM SPRUCEANUM 589

(Cryptomeria japonica) III. Estimation of variance com- improvement of stiffness in Cryptomeria japonica D. ponents of the variation in dynamic modulus of elasticity Don. Holzforschung 57(5):553–560. with plus-tree clones. Mokuzai Gakkaishi 40(5):457– NAMKOONG, G., H. C. KANG, AND J. S. BROUARD. 1988. Tree 464. breeding: Principles and strategies. Monographs on HERNÁNDEZ, R. E. 1993. Influence of moisture sorption his- Theoretical and Applied Genetics. Springer-Verlag, New tory on the swelling of sugar maple wood and some tropi- York, NY. 180 pp. cal hardwoods. Wood Sci. Technol. 27(5):337–345. OLIVEIRA, F.G.R., J.A.O. CAMPOS, AND A. SALES. 2002. Ul- ———. 2007. Influence of accessory compounds, wood trasonic measurements in Brazilian hardwood. Materials density and interlocked grain on the compressive proper- Research 5(1):51–55. ties of hardwoods. Wood Sci. Technol. 41(3):249–265. PANSHIN,A.J.,AND C. DE ZEEUW. 1980. Textbook of wood ———, AND G. RESTREPO. 1995. Natural variation in wood technology, 4th ed. McGraw-Hill, New York, NY. 736 properties of Alnus acuminata H.B.K. grown in Colom- pp. bia. Wood Fiber Sci. 27(1):41–49. RIVERO, J. G. M. 2004. Propiedades físico-mecánicas de ———,A.KOUBAA,M.BEAUDOIN, AND Y. FORTIN. 1998. Gmelina arborea Roxb. y Tectona grandis Linn. F. pro- Selected mechanical properties of fast-growing poplar veniente de plantaciones experimentales del Valle del hybrid clones. Wood Fiber Sci. 30(2):138–147. Sacta—Cochabamba, Bolivia. http://www.monografias .com/trabajos16/gmelina-arborea/gmelina-arborea.shtml HERZIG, L. 1991. Évaluation du module d’Young de bois d’épinette par méthode ultrasonore sur carottes de sond- [accessed 10 October 2005]. age. M.Sc. thesis, Département des sciences du bois, Uni- SANTOS, P. E. T., I. O. GERALDI, AND J. N. GARCIA. 2003. versité Laval, Québec, Canada. Estimates of genetic parameters for physical and mechanical properties of wood in Eucalyptus grandis. Sci- HODGE,G.R.,AND W. S. DVORAK. 2004. The CAMCORE entia Forestalis 63:54–64. international provenance/progeny trials of Gmelina ar- SARANPÄÄ, P. 2003. Wood density and growth. Pages 87– borea: genetic parameters and potential gain. New For. 117 in J. R. BARNETT and G. JERONIMIDIS, eds. Wood qual- 28(2-3):147–166. ity and its biological basis. Blackwell Publishing Ltd., ———, ———,H.URUEÑA, AND L. ROSALES. 2002. Oxford, UK. Growth, provenance effects and genetic variation of Bom- SAS INSTITUTE INC. 2002-2003. SAS/STAT users’ guide, bacopsis quinata in field tests in Venezuela and Colom- version 9.1. SAS Institute Inc., Cary, NC. bia. For. Ecol. Manage.158(1-3):273–289. SEARS, R. R. 2003. New forestry on the floodplain: The ILIC, J. 2001. Relationship among the dynamic and static ecology and management of Calycophyllum spruceanum elastic properties of air-dry Eucalyptus delegatensis R. (Rubiaceae) on the Amazon landscape. Ph.D. thesis, Baker. Holz Roh-Werkst. 59(3):169–175. Graduate School of Arts and Sciences, Columbia Univer- ———. 2003. Dynamic MOE of 55 species using small sity, Ithaca, NY. 246 pp. wood beams. Holz Roh-Werkst. 61(3):167–172. SIMONS, A. J., D. J. MACQUEEN, AND J. L. STEWART. 1994. JOZSA,L.A.,AND G. R. MIDDLETON. 1994. A discussion of Strategic concepts in the domestication of non-industrial wood quality attributes and their practical implications. trees. Pages 91-102 in R. B. LEAKEY and A. C. NEWTON, Special Publication No. SP-34. Forintek Canada Corpo- eds. Tropical trees: The potential for domestication and ration. Western Laboratory, Vancouver, BC, Canada. the rebuilding of forest resources. HMSO Books, Lon- KEENAN,F.J.,AND M. TEJADA. 1984. Tropical timber for don, UK. building materials in the Andean Group countries of SKAAR, C. 1988. Wood-water relations. Springer-Verlag, South America. International Development Research New York, NY. 283 pp. Centre, Ottawa, Canada. SOTELO MONTES, C., AND J. C. WEBER. 1997. Priorización KYOKONG, B., AND E. D. BELLO. 1977. Relationship of dy- de especies arbóreas para sistemas agroforestales en la namic modulus of elasticity with static modulus of elas- selva baja del Perú. Agroforestería en las Américas 4(14): ticity and modulus of rupture of Dipterocarpus grandi- 12–17. florus and Shorea polysperma. Pterocarpus 3(1):43–54. ———,R.E.HERNÁNDEZ,J.BEAULIEU, AND J. C. WEBER. LEI, H., B. L. GARTNER, AND M. R. MILOTA. 1997. Effect of 2006. Genetic variation and correlations between growth growth rate on the anatomy, specific gravity, and bending and wood density of Calycophyllum spruceanum at an properties of wood from 7-year-old red alder (Alnus ru- early age in the Peruvian Amazon. Silvae Genet. 55(4– bra). Can. J. For. Res. 27(1):80–85. 5):217–228. LINARES, C., E. MENESES, AND J. DIAZ. 1992. Monografía STEEL, R. G. D., J. H. TORRIE, AND D. A. DICKEY. 1997. sobre capirona: Calycophyllum spruceanum. Proyecto Principles and procedures of statistics—A biometrical Forestal ITTO PD 37/88: Utilización industrial de nuevas approach, 3rd ed. McGraw-Hill Series in Probability and especies forestales en el Perú. Camara Nacional Forestal, Statistics. McGraw-Hill, Boston, MA. 672 pp. Dirección General de Forestal y Fauna, Lima, Peru. TOLEDO, E., AND C. RINCÓN. 1996. Utilización industrial de NAKADA, R., Y. FUJISAWA, AND Y. HIRAKAWA. 2003. Effects nuevas especies forestales en el Perú.Cámara Nacional of clonal selection by microfibril angle on the genetic Forestal, Instituto Nacional de Recursos Naturales, Orga- 590 WOOD AND FIBER SCIENCE, OCTOBER 2007, V. 39(4)

nización Internacional de las Maderas Tropicales, Lima, trees: an example from the Peruvian Amazon. Develop- Peru. ment in Practice 11(4):425–433. YANG, J-L, AND Y. FORTIN. 2001. Evaluating strength prop- TSOUMIS, G.T. 1991. Science and technology of wood: struc- ture, properties, utilization. Van Nostrand Reinhold, New erties of Pinus radiata from ultrasonic measurements on York, NY. 400 pp. increment cores. Holzforschung 55(6):606–610. ZHANG, S. Y. 1995. Effect of growth rate on wood specific WEBER,J.C.,AND C. SOTELO MONTES. 2005. Variation and gravity and selected mechanical properties in individual correlations among stem growth and wood traits of Ca- species from distinct wood categories. Wood Sci. Tech- lycophyllum spruceanum Benth. from the Peruvian Ama- nol. 29(6):451–465. zon. Silvae Genet. 54(1):31–41. ZOBEL,B.J.,AND J. R. SPRAGUE. 1998. Juvenile wood in ———, ———,H.VIDAURRE,I.K.DAWSON, AND A. J. forest trees. Springer Series in Wood Science. Springer- SIMONS. 2001. Participatory domestication of agroforestry Verlag, Berlin, Germany. 300 pp.