Provenances of the seeds and planting area

In 1983, seeds from open-pollinated plants were collected in three natural populations (provenances) of B. riedelianum in the states of São Paulo and Paraná municipalities, Gália, Bauru, and Alvorada do Sul. Seedlings were grown up from seeds planted in 1984 at the Luiz Antônio Experimental Station (LAES), in municipality of Luiz Antônio. Geographic location, clima and soil data from the four areas are presented in Table 1. The plantation was established in six blocks in linear plots of six plants at a spacing of 3 m × 3 m (^{Gurgel-Garrido et al., 1997}).

Mean annual increment

In 2015, dbh - diameter at breast height, of about 1400 trees was measured with caliper, and height was measured with a hypsometer Vertex IV. From dbh and height values the volume was calculated based on the formula proposed by ^{Higuchi (1978)}. The equation used was

Then, the volume per hectare was calculated considering the spacing (3 m × 3 m) and the number of plants. The number of plantas was multiplied by the average tree volume. Finally, the volume per hectare and year was calculated by dividing volume per hectare by age of planting (30 years).

Sampling for physical and mechanical properties and anatomical features

We cut 36 randomly selected trees, 12 from each provenance. A trunk log of 1.3 m was removed from the base of each tree, and a central plank was cut to obtain specimens for physical and mechanical properties and anatomical characteristics. Three radial positions were established: the nearest part of trunk center, which was designated as pith, a middle position, and a position close to the bark, which was designated as bark.

Apparent density (D₁₂)

Apparent density was determined in acclimatized specimens by the ratio of their mass and volume at the current moisture content (MC). We used the model proposed by Rezende, Escobedo, and Ferraz (1988) based on the study of ^{Kollmann and Côté (1968)} to obtain the value to nominal 12% equilibrium moisture content (EMC) (D₁₂) (see Eq. 1). The density of the model used at any moisture content (D_u%, ranging from 0% to 25% MC) is relative to the density at 0% MC (D₀). Then, D₁₂ was calculated from D₀ using the two equations:

Mechanical tests

We carried out the following tests: shear strength parallel to grain, compression strength parallel to grain, modulus of elasticity and rupture in bending. Tests were performed in a computer-controlled 300 kN eletromechanical testing machine. Strains were evaluated using a standard mechanical strain gauge extensometer (accuracy of 0.001 mm).

Mechanical tests were performed according to NBR 7190 (^{Associação Brasileira de Normas Técnicas [ABNT] 1997}). We used a loading speed of 2.5 MPa/min (shear strength) and 10.0 MPa/min (compression and bending). Initial results of strength and elastic properties (modulus of elasticity) were corrected to the EMC (12%) using a conversion coefficient of 3% (of variation per 1% of MC variation) for strength properties and 2% for elastic properties.

In the Brazilian standard NBR 7190 (^{ABNT, 1997}), the characteristic value of compression strength parallel to grain is used to classify the wood in the system of strength classes (Table 2), and it guides the choice of the most suitable species for structural projects.

Anatomical analysis

We cut small portions of wood from each sample for maceration using Franklin’s method (^{Berlyn and Miksche, 1976}). Wood fragments were stained with aqueous safranin and mounted temporarily in a solution of water and glycerin (1:1). Samples of 2 cm³ were softened in boiling water and glycerin (4:1) for 1 h to 2 h. From these samples, transverse and longitudinal sections 18-25 μm in thickness were obtained with a sliding microtome. Sections were bleached with sodium hypochlorite (60%), washed thoroughly in water, and stained with 1% safranin (^{Johansen, 1940}). Measurements followed the recommendations of the IAWA Committee (^{IAWA, 1989}). Quantitative data are based on at least 25 measurements for each characteristic from each tree, thus fulfilling statistical requirements for the minimum number of measurements.

Data analysis

We initially undertook descriptive statistical analysis and used Box Plot graphics to detect outliers. Thus, values 1.5 times higher than the 3rd quartile and values 1.5 times lower than the 1st quartile were excluded from the analysis. Normality tests were performed to check the distribution of data, and when a normal distribution was not observed, data were square root-transformed. Then, a parametric analysis of variance (one-way analysis of variance [ANOVA]) was performed. When a significant difference was observed between radial positions and provenances, Tukey’s test was used to identify pairs of significantly different means. We analyzed the radial variation within the same tree and also three radial positions together comparing the provenances. Pearson’s correlations between variables and variables with radial position were carried out.

Effect of provenance on growth

The increment difference among provenances may be associated with climate type and soil characteristics. Data in Table 1 show that Alvorada do Sul (Cfa) has a different climate compared to other provenances and the planting area (Cwa), although precipitation and temperature means do not show much difference. It is possible that water holding capacity contributed to the results because Alvorada do Sul soil presents high water holding capacity, and plants were adapted to this condition, while in Gália, Bauru and Luiz Antônio, soil has low water holding capacity. Climate and soil differences may have influenced growth in this species, leading to a smaller increment in Alvorada do Sul compared to other provenances.

Consequently, lower growth in Alvorada do Sul, also described by ^{Sebbenn, Freitas, Zanatto, Moraes, and Moraes (2007)} for B. riedelianum in the same area as that of the present investigation, should be related to denser wood compared to Gália and Bauru. ^{King, Davies, Nur Supardi, and Stan (2005)}, studying the relationship among tree growth, light absorption and wood density in Malaysian forests, reported a negative relationship between wood volume for a given biomass unit and its density. According to Harris (1981), faster growth of Gymnosperm or Angiosperm trees results in lower wood density. The growth of B. riedelianum is slow to moderate (^{Carvalho 2004}). The highest volumetric productivities were mentioned by ^{Garrido, Nogueira, and Garrido (1982)} and ^{Gurgel Filho, Moraes, and Garrido (1982)} at 26 years old (12 m³ ha^-1 year^-1), nearly twice that of our study (Table 3). However, our values were higher than those in other studies, with younger trees, reported for southern and southeastern Brazil, e.g., ^{Carvalho (2004)}, but in analyses of mean annual, other factors should be considered, such as spacing, soil type and age of planting.

Anatomical differences are potential producers of variations in growth. Larger vessels were observed in Bauru, although in increment, Bauru did not differ statistically from Gália. Thus, higher growth of Bauru, especially compared to Alvorada do Sul, can be explained by higher potential of hydraulic conductivity which, as consequence, positively influences photosynthetic rate (^{Hacke, Sperry, & Pittermann, 2005}). Larger vessels conduct water more efficiently than narrower ones. In addition, narrower and less frequent rays in Alvorada do Sul compared to the other two provenances may have also contributed to lower growth. This result can be explained by rays which primarily consist of parenchyma cells which can store starch. According to ^{Costa, Callado, Coradin, and Carmello-Guerreiro (2006)}, B. riedelianum is mobilized for cambial reactivation after a period of dormancy. According to ^{Carvalho (2004)}, B. riedelianum is a deciduous species. Therefore, the process of growth resumption is accelerated. From April to September, Luiz Antônio experiences a water deficit (^{Longui et al., 2011}). Thus, a smaller volume and frequency of rays can be a competitive disadvantage in growth for Alvorada trees. ^{Sebbenn, Freitas, Zanatto, Moraes, and Moraes (2009)} reported faster growth and higher survival rate. ^{Longui et al. (2011)} observed wider rays, in provenance tests of Gallesia integrifolia, also planted in Luiz Antônio.

Studying the potential of native species (Brazilian) for the production of sawn wood, ^{Silva (2013)} reported that B. riedelianum has medium need of thinning and is suitable for planting in high silt content soils. Moreover, we found no data about the presence or content of silt in soils of studied areas. ^{Silva (2013)} did not indicate the potential production cycle for B. riedelianum (wood density 730 kg m^-3), but suggested about 45 years for Hymenaea courbaril (750 kg m^-3) and at 25 years for Myroxylon peruiferum (780 kg m^-3). In this context, we asked if B. riedelianum at 30 years of age would be ready for cutting. We emphasize that trees collected for the present study, presented dbh average ≈16 cm, but many trees were over 20 cm, and others reached dbh of 32 cm (Table 3). Although the tree trunk of B. riedelianum is usually straight, diameters found at 30 years old may not be suitable for cutting. ^{Campbell (1999)} reported that the ideal age for cutting can be determined by current annual increment (CAI), which measures how much the forest has grown in volume in the last year and mean annual increment (MAI), i.e., wood volume growing on one hectare of forest during one year on average since the forest has been planted. ^{Campbell (1999)} suggests that trees should be cut when CAI values are equal to MAI, when planting reaches the maximum wood volumetric production by area by year.

Radial variation

The most expected pattern pith-bark is an increase in density (physical property) and, consequently, an increase in mechanical properties since these are influenced by wood density, as well as an increase in fiber length and wall thickness, increase in vessel diameter and decrease in vessel frequency from pith to toward bark (^{Lachenbruch, Moore, & Evans, 2011}; ^{Hoadley, 2000}; ^{Baas, Ewers, Davis, Wheeler, 2004}). It was observed a positive correlation between density and shear parallel to the grain (R = 0.99 and P = 0.03) and between MOR and fiber length (R = 0.99 and P = 0.04) in Alvorada do Sul. However, the expected pattern was not clearly noticed in B. riedelianum since a gradual increase was noted only in Gália for modulus of elasticity and in Bauru for compression parallel to the grain, both having positive correlations with radial position, as previously reported, in addition to fiber length, which gradually increased towards the bark in Gália. Other properties and anatomical characteristics did not show the typical radial pattern, as described by ^{Lachenbruch et al. (2011)}.

In general, when considering the means values among properties in three radial positions, Alvorada do Sul presented more resistant wood when compared to Gália and Bauru. Also, values of density and compression in Alvorada were highest among the three provenances, while shear values did not differ from those of Bauru, and MOE and MOR values did not differ from those of Gália. Wood from Alvorada do Sul was the most homogeneous and absent of radial variation in six characteristics, including shear, MOE, width and frequency of ray, fiber diameter, and wall thickness. Gália did not vary in four variables, including shear, compression, fiber diameter and wall thickness. Bauru was the most heterogeneous wood, with no variation in three variables, all anatomical: vessel diameter, height and frequency of ray. Thus, Alvorada do Sul is the provenance with the highest value in properties and homogeneity, whereas Gália and Bauru provenances show the greatest mean annual increment.

Wood strength and Strength class

We did not find many studies about wood structure or physical and mechanical properties of B. riedelianum. Consequently, we compared the values of our study with those described by ^{Mainieri and Chimelo (1989)}, ^{Carvalho (2004)} and ^{Lima et al. (2011)}. The lack of studies about wood quality in a homogeneous plantation with determined age and spacings is a drawback since that is the information that farmers need in order to cultivate B. riedelianum for commercial purposes.

The density values this population are close to those reported by ^{Mainieri and Chimelo (1989)}, 0.84 g cm^-3 at 15% MC, but those authors did not mention the origin or age of B. riedelianum. ^{Lima et al. (2011)} studied the same planting of B. riedelianum at 24 years old and found a basic density of 0.60 g cm^-3 (near the pith) to 0.65 g cm^-3 (close to the bark), and Alvorada do Sul (0.66 g cm^-3), and Gália (0.64 g cm^-3) provenances did not show differences between them. Apart from the methodology for density determination which differed from that of Lima, it is related to the age of tree samples collected. Therefore, higher growth tree promoted a higher density for Alvorada do Sul compared to the other two provenances. According to Mainieri et al. (1983), B. riedelianum wood is classified as heavy wood since it has specific gravity between 750 - 950 kg m^-3. According to ^{Brunelli, Leal, and Longo, (1997)}, B. riedelianum wood can be used to manufacture furniture, because its wood presents light coloration, nice appearance and appropriate mechanical strength. We noted little or no variation in color between heartwood and sapwood, which is characteristic of B. Riedelianum. According to ^{Carvalho (2004)}, sapwood is white in color or slightly yellowish, while heartwood is white-yellowish straw, darkening to uniform pale yellow.

The results for shear parallel to the grain (16.70 MPa - 20.28 MPa, means) were higher than those reported by ^{Mainieri and Chimelo (1989)}, who described 13 MPa for B. riedelianum green wood (above 20% moisture), but the age was not mentioned. In our study, the tests were performed with samples at 12% EMC, and it indicates dry wood. The values of mechanical properties are lower when determined in green woods, as seen in several species tabulated by ^{Kretschmann (2010)}. The results for compression parallel to the grain (44.89 MPa - 54.11 MPa, means) were lower than those presented by ^{Mainieri and Chimelo (1989)} at 15% EMC (58.90 MPa), but since these did not have other information from the study cited above, it is not possible to establish a clearer comparison due to the facta that the present study used trees younger than those studied by ^{Mainieri and Chimelo (1989)}.

According to NBR 7190 (^{ABNT, 1997}), B. riedelianum wood falls into strength class C40, whereas the values of compression varied between 44.89 MPa to 54.11 MPa in the pith of Gália and middle position of Alvorada do Sul, respectively. ^{Zenid et al. (2009)} state that the use of strength classes eliminates the need to specify wood species regarding mechanical strength, but the mechanical properties should be known ahead of time in order to categorize a given species. Besides, knowledge of species is also essential when it is necessary to use wood that is naturally resistant or permeable to preservative solutions, or when it is necessary to know the workability and decorativeness of wood.

For modulus of elasticity (MOE) and modulus of rupture (MOR), ^{Mainieri and Chimelo (1989)} reported MOE of 11493 in green wood and MOR of 137 MPa at 15% moisture content. In this study values at 12% were higher for MOE (11659 MPa - 13942 MPa, means) and lower for MOR (116 MPa - 124 MPa). Here it is necessary to consider the difference in moisture content and the absence of other information from tree planting for woods which were tested and described by ^{Mainieri and Chimelo (1989)}.