3.1. Intra-Ring Wood Density Variation and Anatomical Changes between Early-and Latewood
The intra-ring wood density variations of the whole tree and the juvenile and mature wood for 44
T. occidentalis trees are shown in
Figure 1a. Intra-ring wood density variation is also examined for selected ring groups (
Figure 1b). The cell structure of
T. occidentalis wood is presented in
Figure 2. The means and variations in early-and latewood tracheid anatomical properties in are shown in
Table 1. The means and variations of wood density at selected ring groups are shown in
Table 2.
The intra-ring wood density (
Figure 1) and tracheid size variations (
Figure 2 and
Table 1) revealed that the transition from early- to latewood is gradual in
T. occidentalis. The early- and latewood were distinguished in the structure (
Figure 2 and
Table 1). The latewood zone was narrow with thick cell walls (
Figure 2a–c,
Table 1) and was slightly denser than the earlywood zone (
Figure 1a).
Thuja occidentalis wood was relatively homogeneous and simple in structure, consisting primarily of overlapping tracheids (tracheids average 23.58 µm and 26.19 µm in the tangential and the radial diameter, respectively) connected by uniseriate xylem rays, ray parenchyma cells (
Figure 2c,
Table 1) and uniseriate, or occasionally biseriate, bordered pits (
Figure 2d). Axial parenchyma was usually rare or absent. Rays are uniseriate and thin (
Figure 2c). The tracheid cell walls were organized in layers of different thicknesses: thick latewood cell walls and thin earlywood cell walls (
Figure 2a–c,
Table 1). In earlywood, the longitudinal tracheids were hexagonal with minimal wall thickness and a larger diameter, usually in the radial direction, while the diameter of longitudinal tracheids in the tangential direction remained relatively constant within a growth ring (
Figure 2a,c). In latewood, the cross sections of tracheids were essentially rectangular and compacted radially to a tabular shape (
Figure 2a,c).
Cell and lumen diameters decreased while cell wall thickness increased from early- to latewood (
Figure 2,
Table 1). Lumen diameter reduction was even greater than the one seen for cell diameter (
Table 1). Cell and lumen diameter changes along a growth ring were greater in the radial direction than in the tangential one (
Figure 2a,c,
Table 1). For example, the lumen radial diameter decreased by 40% from early-to latewood, while the lumen tangential diameter decreased by only 23% (
Table 1). This could explain the decrease in the ratio between radial and tangential diameter (
Table 1). These changes led to a decline in cell and lumen perimeters. However, lumen perimeter reduction was greater than that for cell perimeter. For instance, lumen perimeter decreased by 33% from early-to latewood compared to only 1.6% for cell perimeter (
Table 1). By contrast, cell walls were thicker in the tangential direction than in the radial one (
Figure 2a,c). Thus, changes in cell wall thickness, and cell and lumen diameters led to an increase in cell wall area and proportion (
Table 1). However, it is important to note that variations in cell wall thickness were larger compared to those of cell size (
Table 1).
From early- to latewood, the tracheids became smaller, the cell and lumen areas decreased, whereas the cell wall area increased (
Figure 2a,c,
Table 1). For instance, the lumen area was double the size in early- compared to latewood (
Table 1). However, cell wall area was more important in latewood, where it represented 71% of total cell area compared to only 39% in earlywood. This could explain the five-fold increase in the ratio between cell wall and lumen area from early- to latewood tracheids (
Table 1). Similar findings were reported for other species such as Douglas fir (
Pseudotsuga menziesii (Mirb.) Franco) [
18,
38], Norway spruce (
Picea abies), Scots pine (
Pinus sylvestris) and silver Fir (
Abies alba) [
10], as well as
Gmelina arborea [
39] and
Acacia mangium plantations [
40]. The cell size reduction (cell and lumen diameter) occurs mainly in the radial direction, while cell wall thickness increases occurs mainly in the tangential direction [
18,
39,
40,
41,
42]. According to Hannrup et al. [
41] and Rathgeber et al. [
18], the cell and lumen tangential diameter reduction occurs mainly in latewood because anticlinal divisions occur more frequently at the end of the growing season. This may explain the variation in cell and lumen diameter between early- and latewood in the radial and the tangential directions of
T. occidentalis wood (
Figure 2c and
Table 1). Cell and lumen diameter changes were greater in the radial direction than in the tangential one. Wang et al. [
43] already noted the cell wall thickness reduction at the end of a growth ring for black spruce (
Picea mariana Mill.). However, it is important to note that the ratio of cell wall thickness to lumen diameter varied from 0.1 to 2.1 in black spruce compared to only 0.2 to 0.8 in
T. occidentalis wood, this could explain the low intra-ring wood density variation of this wood (
Figure 1).
The mean of wood density for the studied
T. occidentalis trees (
Table 2) was 355.41 kg m
−3, which is higher than that reported previously for this species in Wisconsin (281–324 kg m
−3) [
44]. Wood density exhibited an S-shape profile from the beginning of a growth ring to its end, with a slight decrease occurring during the first 20% of a ring (
Figure 1a). Consequently, wood density decreased by 18%, increasing thereafter by 106% to reach a maximum value (530 kg m
−3) at the 90% mark of the ring, before a slight decline (22%) at the end of the growth ring (410 kg m
−3). Moreover, LWD was about two times (41%) greater than EWD (
Figure 1a). This variation is lower (260–530 kg m
−3) compared to that reported for other conifer species such as Norway spruce (351–1027 kg m
−3), Scots pine (325–939 kg m
−3) and silver fir (287–861 kg m
−3), where the LWD was about 3–4 times greater than the EWD [
10,
18,
39]. The evolution of density was also more rapid in the latewood of
T. occidentalis (
Figure 1a). The earlywood represented more than 70% of the ring and the density variation with ring width was therefore more important in early- than in latewood. This result is in good agreement with previous reports on Norway spruce and Scots pine [
10,
18].
The patterns of intra-ring variation were relatively similar between rings from juvenile and mature wood (
Figure 1a). However, the intra-ring distribution of latewood was slightly more important in juvenile (32%–40%) than in mature wood (20%) (
Figure 1a). For instance, the latewood transition occurred at 65% of the ring for ring 6 (
Figure A1). With increasing age, the earlywood/latewood transition zone occurred later in the ring, at 70% of the 20th ring width. In mature wood, the intra-ring distribution of LP was relatively uniform, accounting for 20% of the ring (
Figure A1). The pattern of wood density variation in individual rings of
T. occidentalis appears to be comparable to the general pattern reported for other conifers species, such as white spruce (
Picea glauca (Moench) Voss), eastern hemlock (
Tsuga canadensis (L.) Carrière), jack pine (
Pinus banksiana Lamb.) [
45] and Douglas fir [
18]. However,
T. occidentalis wood density is more uniform and homogeneous compared to these species and is characterized by a moderate difference between early- and latewood (270 kg m
−3). This is beneficial for use in wooden structures that require wood uniformity such as veneer peeling and slicing [
46].
The analysis of variance (
Table 3) showed that ring groups as well as ring width and tree significantly affected wood density. The site effect could be masked by the tree-to-tree variation [
6].
Ring width was the most important source of variation in wood density, accounting for 69.1% of the total variation, followed by ring groups (2.02%). Wood density variation is related to various factors including cambial age and ring width [
6,
9]. Hence, the effects of ring group and ring width on wood density were highly significant (
p < 0.001) (
Table 3). Overall, wood density decreased steadily with cambial age (
Table 2). The wood density variation with cambial age was minimal near the pith, increased with increasing ring width up to 60% of the ring and decreased thereafter (
Figure 1b).
The intra-ring wood density variation with cambial age showed substantial changes (
Figure 1b). In earlywood, wood density decreased with cambial age. However, wood density showed the reverse pattern in latewood, increasing with cambial age. The intra-ring wood density distribution was more homogeneous in mature than in juvenile wood (
Figure 1b,
Table 2). This result disagrees with a previous report on black spruce, where this distribution was more homogeneous in juvenile wood, mainly due to lower LWD and LP by width [
24]. According to Zobel and Van Buijtenen [
9], the larger variation in intra-ring wood density within the first 30 rings (
Figure 1) could be linked to the variation in juvenile properties. Compared to mature wood, juvenile wood is characterized by a higher variation in cell dimensions and cell wall formation [
47,
48]. In
T. occidentalis, the LP (
Figure 3) was higher near the pith (65%), decreasing gradually to reach a minimum in the transition zone (20%) and remained constant thereafter [
6]. Accordingly, the intra-ring density distribution in this wood is more homogenous in the mature wood, which is mainly attributable to the lower LP by width compared to juvenile wood.
Wood density decreased progressively to about age 30 in earlywood and remained constant thereafter (
Figure 1b). In contrast, LWD showed a different pattern: it increased from a minimum near the pith (rings 2–10) to reach a maximum at the juvenile-mature wood transition zone (rings 21–30). However, the variation in wood density with cambial age was minimal beyond the 30th ring from the pith, although a consistent slight decrease was observed with increasing cambial age thereafter. Furthermore, the variation in wood density with cambial age was more important in early- than in latewood (
Figure 1b). The anatomical changes are primarily responsible for intra-ring wood density variation [
10,
18,
49]. Earlywood lumen diameter and LP are the two most important predictors of wood density [
18,
41]. LP can explain up to 60% of the density variation [
38]. Decoux et al. [
10] propose that intra-ring variation of wood density is mainly due to the variation in cell wall thickness. Given the tubular shape of tracheids, cell wall thickness is of great importance when stiffness of the tracheids is compared. Earlywood cells have a thin wall layer, which is one of the reasons that earlywood is less dense compared to latewood [
9].
For
T. occidentalis, the latewood tracheids are characterized by narrow cell lumina and thick cell walls compared to earlywood (
Figure 2,
Table 1). From early- to latewood, cell wall proportion increased, while cell and lumen area decreased. Thus, the wood density increase in latewood (
Figure 1) is mainly linked to cell wall thickness (
Table 1). These results are in good agreement with the findings of Rathgeber et al. [
18] for Douglas fir. According to their results, cell wall area and thickness decreased slightly at the beginning and at the end of a growth ring. Cell wall thickness reduction at the end of a growth ring was already reported for black spruce [
43], Norway spruce, Scots pine and silver fir [
10]. This explains the steady decrease observed in wood density at the beginning and the end of the ring in
T. occidentalis (
Figure 1). Deleuze and Houiller’s process-based model of xylem growth simulated a reduction in wood density at the beginning of a ring and indicated the reduction was caused by a temporary substrate shortage used for cell wall construction [
50]. Wimmer [
51] showed that radial diameter and cell wall thickness of latewood tracheids have the greatest influence on the growth ring density of conifers, while tracheid length and other cell characteristics, such as parenchyma rays and resin canals, can be neglected. For hardwoods such as green ash (
Fraxinus pennsylvanica Marsh) [
52] and red maple (
Acer rubrum L.) [
53], the higher LWD was attributed to an increase in fiber cell wall thickness and a decrease in vessel diameter compared to those of earlywood.
The variation of intra-ring wood density with cambial age (
Figure 1b) is mainly associated with anatomical changes, which are related to cambial activity [
18,
41]. Hence, tracheid dimension varies with tree age [
48,
54]. Hannrup et al. [
41] reported that the total radial and tangential lumen diameter decreased with the age of the wood, which could explain the increase in the proportion of the narrow, thick-walled latewood tracheid. The radial and tangential lumen diameter of the earlywood increased with the age of the wood, which agrees with the age trend normally occurring in conifers [
16,
55]. The tangential lumen diameter of the studied trees was consistently smaller in late- than in earlywood (
Table 1). This may be the result of thicker cell walls in the latewood, but it may also be because anticlinal cell divisions are limited to the latewood [
55]. This may explain the increase in LWD with increasing cambial age in
T. occidentalis (
Figure 1b).
3.2. Interrelationships of Wood Ring Density and Width
Table 4 shows the correlation coefficients for ring density components and ring widths in juvenile and mature wood. The relationships between wood ring density and width traits in
T. occidentalis are also examined for selected ring groups (
Table 5). Overall, the correlations between ring density components were statistically significant (
p < 0.01) (
Table 4). In juvenile wood, the correlation between RD and EWD was positive and stronger than that between RD and LWD and LP. However, in mature wood, the RD-LWD correlation was weaker. This indicates that the magnitude of the correlation differs between juvenile and mature wood in
T. occidentalis. Similar results were reported for black spruce [
27] and balsam fir [
26].
For
T. occidentalis, the correlation between RD and LP was weak compared to correlations between RD and EWD and LWD (
Table 4). Thus, EWD and LWD are more important in determining RD than LP in this species. EWD explained 92% and 89% of the variation in juvenile and mature wood density, respectively. This result is somewhat different form previous findings on black spruce [
27,
56], where LP (
r = 0.91) was more important than EWD (
r = 0.86) and LWD (
r = 0.59) in determining RD. Koga and Zhang [
26] also reported that EWD (
r = 0.83) and LP (
r = 0.40) were the most important parameters in determining RD of balsam fir. The percentage of early-and latewood in a growth ring determines the overall density of the ring. In
T. occidentalis, LWW is relatively constant over age; therefore, RD is closely related to EWD (
Figure 3) [
6]. Moreover, the correlation between RD and EWD was nearly constant (about 0.90) over tree age while the correlations between RD and LWD decreased consistently with cambial age (
Table 5). Thus, EWD is the most important parameter in determining mature wood density in
T. occidentalis.
The correlation between EWD and LWD was positive in both juvenile and mature wood (
Table 4). This result contradicts previous findings for Douglas-fir where the correlation was negative in both juvenile and mature wood [
57]. Zhang et al. [
58] also reported a negative correlation in the juvenile wood of black spruce. No significant correlations between EWD and LWD were found in the mature wood of black spruce [
27] and balsam fir [
26].
RD and EWD were positively correlated to LP, whereas LWD was negatively correlated to LP (
Table 4). This result disagrees with previous findings on black spruce [
27], Douglas fir [
57] and balsam fir [
26], where the correlation between LP and LWD was positive. The negative correlation between LWD and LP in
T. occidentalis may be explained by their patterns of radial variation (
Figure 3). LWD was low near the pith, increased to a maximum in the juvenile-mature transition zone (30 years) and remained constant thereafter. In contrast, LP was higher near the pith, decreased gradually to reach a minimum in the transition zone and remained constant afterward [
6]. Hence, the relationship between LWD and LP was negative and weaker in juvenile wood, but stronger and almost constant in mature wood (
Table 4 and
Table 5).
This study also revealed a negative relationship between the ring width and ring density components of
T. occidentalis (
Table 4). RD correlated significantly and negatively with RW and EWW in juvenile and mature wood, but no significant correlation between RD and LWW was found (
Table 4). The correlations between RD and both RW and EWW were weaker in mature wood (
Table 4). This suggests that the negative impact of high growth on density decreases when the wood reaches maturity.
The effects of ring widths on ring density components were also examined for selected ring groups (
Table 5). The negative relationships between RW and RD, although significant, were low (
r ˂ −0.1) and tended to weaken slightly with increasing cambial age. These results are in good agreement with previous findings for black spruce [
27] and balsam fir [
26]. Using all the sample data available for RW and RD, results also show a weak relationship between the two variables (
Figure 4a). The coefficient of determination between RW and RD was low but significant. According to Koga and Zhang [
26], a slow increase in RW could negatively affect RD. The percentage of early-and latewood in a growth ring determines the overall density of the ring, and in
T. occidentalis, the LWW is relatively constant. Therefore, the RW is closely related to the EWW (
Figure 3). As the RW increases, the width of the earlywood increases without a corresponding increase for latewood, causing lower RD [
6]. This could explain the not significant correlation between RD and LWW in both juvenile and mature wood, as well as the negative relationship of RD with LP (
Table 4 and
Table 5). In juvenile wood, earlywood cells have a thin wall (
Table 1), which explains its lower density [
9]. According to Zhang et al. [
59], the decrease in RD combined with the increase in RW in conifers was more pronounced in species that showed a gradual transition from early- to latewood (as
T. occidentalis) than in species with an abrupt transition.
The correlation between EWD and RW and EWW was negative in juvenile wood and not significant in mature wood (
Table 4). These results contradict previous findings for black spruce [
27,
58,
59,
60] and balsam fir [
11], where the correlations were significant in both juvenile and mature wood, and only in mature wood [
26], respectively. For selected ring groups, the correlations between EWD and RW over tree age were weak or not significant (
Table 5). However, the correlations between both RW and EWW and LWD were positive and significant in both juvenile and mature wood (
Table 4). The correlation increased with cambial age (
Table 5). This result contradicts previous findings for black spruce [
27], where LWD correlated negatively to RW (
r = −29) and EWW (
r = −41) in mature wood. For balsam fir [
11], LWD was correlated to RW in mature wood only (
r = 0.50).
LP correlated negatively to RW in both juvenile and mature wood (
Table 4). These correlations were moderate (
Figure 4b). The negative relationship tended to increase with cambial age (
Table 5). For instance, it was −0.29 for rings 2 to 10 compared to −0.46 for rings 61 to 70. According to Bouslimi et al. [
6], RW is closely related to EWW in
T. occidentalis because the LWW is constant over tree age. Therefore, LP decreased with cambial age because the increasing RW produced wider earlywood without a corresponding increase in LWW (
Figure 3). Accordingly, the correlation between RW and EWW was very strong compared to that between RW and LWW in both juvenile and mature wood (
Table 4). These results concur with previous reports for balsam fir and black spruce [
24,
26].
Based on these results, the tendency for a weak negative relationship between RD and RW in mature wood may be explained by the fact that EWD, which was the most important parameter in determining wood density in mature wood, was not significantly correlated with RW (
Table 4). However, this appears to be of no practical importance because the correlation coefficient was quite low (
Table 4). Overall, this study suggests that a faster growth rate would not reduce wood density significantly in this species. Indeed, the negative relationships of RD and EWD with RW and EWW, although significant, were weaker (correlation coefficients varied from −0.23 to −0.10) in
T. occidentalis compared to other species such as black spruce, where the correlation coefficients varied from −0.67 to −0.33 [
27]. The same holds true for LP, however, the growth rate seemed to positively influence LWD in this species (
Table 4). These results contradict those found for balsam fir and black spruce, where a negative relationship was reported between LWD and RW [
24,
26].
3.3. Relationships between Ring Width and Tracheid Length and Width
The correlation coefficients between tracheid length and width, RW, tree diameter and height in
T. occidentalis are shown in
Table 6. Tracheid length correlated positively with tracheid width, tree diameter and tree height. However, tracheid length correlated negatively with RW. Tracheid width was also positively correlated with tree diameter and RW, but no relationship was found between tracheid width and tree height. Tree diameter was also positively and strongly correlated to tree height and RW (
Table 6). These correlations are statistically significant at
p = 95% except the tracheid width-tree height correlation (
Table 6).
Several contradictory reports have been published on the relationship between tracheid length and RW in conifers and hardwoods [
61,
62,
63]. An inverse relationship between tracheid length and RW in conifers was observed by Chalk [
62], which is in good agreement with the results of the present study. Fujiwara and Yang [
63] also reported a negative relationship between tracheid length and RW for jack pine, balsam fir and black spruce, but no relationship was found for white spruce. However, a positive relationship between tracheid length and RW was reported for trembling aspen [
63]. Diaz-Váz et al. [
64] also observed a positive relationship between tracheid length and RW in conifers. Dutilleul et al. [
65] reported that fast-growing spruces (
Picea abies (L.) Karst.) showed a stronger negative correlation between RW and tracheid length (
r = −0.86).
The negative relationship between tracheid length and RW (
Table 6) suggested that tracheid length depends on growth rate. According to Fujiwara and Yang [
63], a tree’s diameter growth is accompanied by the circumferential expansion of the cambium because the increase in cambium girth is primarily due to the increase in the number of fusiform initials achieved by pseudo-transverse division [
16,
55]. Bannan [
55] considered the relationship between RW and cell length from the standpoint of RW and the pseudo-transverse division rate. The relationship between pseudo-transverse division rate and cell length is usually negative: a high rate is accompanied by short cells and, conversely, a low rate is accompanied by longer cells. Bannan [
16] also showed that the frequency of pseudo-transverse division in fusiform initials is related to the linear radial increment in
T. occidentalis. However, cell length is not only affected by RW, but also by circumferential growth. The circumferential growth rate differs according to tree diameter even though the radial growth rate is the same. Fujiwara and Yang [
63] reported a negative relationship between tracheid length and circumferential growth rate in jack pine, balsam fir, black spruce and trembling aspen. In trembling aspen, tracheid length decreased with both higher and lower circumferential growth rates.
The negative correlations between RD and RW (
r = −0.10 to −0.19) and between tracheid length and RW (
r = −0.12) were weak in
T. occidentalis compared to other species, such as jack pine, balsam fir, black spruce, trembling aspen and brutia pine (
Pinus brutia, Ten.) [
63,
66], thus providing the opportunity for a silviculture program to simultaneously improve RW, wood density and tracheid length.