1. Introduction
Hardwoods and softwoods differ at microscopic and macroscopic levels [
1,
2]. Hardwood trees have more complex architectures [
3] and are usually transformed into a wider range of commodities, including higher-value products such as flooring, cabinets, veneer or furniture, and lower-value ones such as papers, cellulosic fibers, and pallets [
4,
5,
6], whereas softwoods are mainly used for construction lumber, pulp and paper, or biomass [
7]. The intricate architecture of hardwood trees and the greater diversity of log grades can often lead to additional difficulties in the bucking phase, the task of cross-cutting a stem into several assortments [
8]. During mechanized harvesting operations, complexity of stand and wood architecture can require higher cognitive efforts and lead to the quicker fatigue of operators as compared to operations performed in monospecific softwood stands [
9].
The creation of wood products is influenced by tree size (dimensions, form, straightness, etc.) and wood quality factors (wood density, growth rate uniformity, presence of knots, chemical composition, fungus-caused coloration, etc.) [
2], which in turn may be influenced by silviculture [
10] and the inter- and intra-species genetic diversity [
2]. Variation of these attributes influence the monetary value of trees [
11]. This is especially applicable for hardwoods since their product specifications are particularly sensitive to quality aspects. The price of good quality timber can be several folds higher than lower quality timber [
12].
Tolerant northern hardwoods, which include species from the genus Acer, Betula, Fagus, and Quercus, are a significant component of forest ecosystems and value chains of Eastern Canada and Northeastern United States: in Canada, they cumulate 6.5 million hectares in the Atlantic Maritime and Mixedwood Plains ecotones [
13], while in the United States they cover approximatively 8 million hectares [
14]. Sugar maple (
Acer saccharum L.) is one of the most common hardwood species found in Acadian forests [
15], which cover the Maritime provinces of Canada (New Brunswick, Nova Scotia, and Prince Edward Island) [
16] and parts of the United States, particularly in the state of Maine [
17]. In the Acadian forests of New Brunswick, the forest industry sector directly employed over 10,000 people in 2020 [
18] and, in 2021, 9,681,427 m³ of roundwood were harvested, from which 34% were hardwoods used domestically and for imports [
19].
Volume recovery of sugar maple and yellow birch (
Betula alleghaniensis Britton) in the province of Québec, Canada, was investigated by Fortin et al. [
20] according to different tree quality classification systems. The predicting factors in this study were limited to DBH, standing tree classification, and tree species. The positive effect of DBH and tree quality class on monetary output of sugar maple products was also analyzed by Cockwell and Caspersen [
11], as sawlog volume output increased with tree size. More recently, Havreljuk et al. [
6] studied the relation between log and lumber attributes for sugar maple and yellow birch, and highlighted the need of including log attributes in models assessing volume and value recovery.
Castle et al. [
21] predicted the probability for single trees of belonging to certain quality and risk groups, along with the sawlog recovery for sugar maple, yellow birch, red maple (
Acer rubrum L.), and red oak (
Quercus rubra L.) in Acadian forests. The logistic models were based on tree species, DBH, and the quality of trees as well as environmental variables such as climate index, topographic index, elevation, slope, aspect, and drainage, both in the United States and Canada, on 5397 trees with a DBH larger than 24 cm with both site and plot as random effects. The best model to predict the probability of belonging to a quality or risk class reached 59%–67% of accuracy depending on the considered class, with only species and DBH as predictors, with a significant contribution of the random effects. Tree quality and risk classes significantly contributed to predict the occurrence of sawlogs (AUC = 0.83) and the sawlog/merchantable volume ratio with, however, a low coefficient of determination (R
2 = 0.34). The prediction of the sawlog ratio was greatly improved by replacing the tree quality class in the predictors by the estimated merchantable sawlog length potential, a continuous variable (R
2 = 0.88). This study also concluded that environmental variables did not have a significant contribution to sawlog recovery. This contradicts the results of Hassegawa et al. [
22] and Guillemette and Bédard [
23], who observed that environmental factors had the most influence on sawlog and veneer recovery in the province of Québec. These varying conclusions may be an artifact caused by multicollinearity in the predicting variables [
24], a problem for parametric methods such as general linear models as used in Castle et al. [
21] and in Guillemette and Bédard [
23], but not for multivariate analysis methods or for boosted regression tree methods utilized in Hassegawa et al. [
22].
To expand upon existing studies on the characterization of hardwood production at the tree level, we propose to include stand and site attributes to analyze and describe the structure of Acadian forest stands focusing on key factors previously related to product recovery in order to provide a snapshot of the current conditions of hardwood forests of New Brunswick. This paper will also attempt to establish relationships between tree, stand, and site attributes while (1) using an extensive dataset, (2) combining multiple data sources using multivariate analyses, and (3) examining the possible contribution of other environmental factors.
Specific objectives are to (1) characterize the stand-level distribution of tree species, size (diameter and height), form and risk, considering the influence of the proportion of hardwoods in the stands (dominance), stand density (expressed by the relative stand density index RSDI), the ecoregion and the sampling strategy of each data source (sample type) and (2) determine the influence of different tree, stand, and site factors on the merchantable volume and the proportion of sawlog assessed throughout the Acadian forests of New Brunswick.
4. Discussion
4.1. Characterization of Forest Plots
The first part of this study provides a portrait of a priori key stand and site variables that are potentially related to hardwood yields, in terms of merchantable and sawlog volumes at the scale of the forest plot. We purposely focused on the characterization of forest plots instead of forest stands, because of the important variability of forest conditions that can typically be found in hardwood stands [
27,
44].
Six variables were considered in this analysis: species composition, stand density, frequency of form and risk classes, diameter, and height. The analysis was based on five data sources, with two contrasting sample designs: one study was conducted with a systematic distribution of plots, which provided a representative view of the productive forest of New Brunswick (including softwood and mixed wood stands), while the other four were targeted in rather mature hardwood stands. This particularity must be taken in account when reading the results.
Ecoregions [
28] were considered as a basis for our analysis where the multivariate analysis allowed a quantitative understanding of the driving climate and soil factors. The analysis showed consistent groups of ecoregions influenced by altitude, soil, temperature, and precipitation. Later analyses showed ecoregions to be significantly influencing all examined plot and tree attributes.
4.1.1. Species Distribution
The productive forest of New Brunswick was characterized by a 30%–70% split in the frequency of hardwood and softwood trees. Among the hardwoods, red maple dominated (9%), followed by sugar maple (7%), white (6%), and yellow birch (5%), followed by other species in smaller proportions. The distribution of tree species depicted in this study was comparable to the values reported in the forest inventory of Atlantic Maritime [
45] for the softwood and hardwood distribution. However, percentages for single tree species differed between the abovementioned inventory and this study, likely because of differences in the study areas. In the plots from the studies deliberately targeting hardwood stands, sugar maple and yellow birch dominated, with a smaller proportion of red maple and white birch than for the systematic sampling. The species mix also varied significantly among ecoregions: sugar maple is largely dominant in ecoregion 3 (north-west of the province), while red maple reigns in ecoregions 6 and 7 in the east, and birches (white and yellow) are dominant in Fundy Bay in the South (ecoregion 4). It is difficult at this point to evaluate if these regional differences are caused by site factors (climate and soils) or by the past disturbance history, both natural and anthropogenic [
46,
47]. Stand age and detailed information about past disturbances, harvests, and treatments were lacking in the databases, which greatly limits this field of investigation.
4.1.2. Stand Density
The stand density is known to affect both stand- and tree-level wood production; high density generally leads to smaller diameter trees with longer boles clear of branches for a given stand age [
48]. The relative stand density index (RDSI) developed by Ducey and Knapp [
30] for mixed forest stands was used as a measure of density and inter-tree competition. On average, the studied stands reached 63% of maximum density; the distribution of the stand density was comparable among all ecoregions for both sampling types, to the exception of stands from non-systematic sampling in ecoregion 5, which bear significantly higher densities. This difference remains yet unexplained.
4.1.3. Form and Risk Classes
Form and risk classes developed in New Brunswick are routinely used for stand assessment and silviculture prescriptions in tolerant hardwoods, showing a significant relation with merchantable and sawlog yields [
21]. Single straight stems (SSS) were dominating in frequency in both sample types (70% for systematic and 52% for non-systematic sampling). The occurrence of trees with an optimal tree form (single straight stem) depended on several factors. The share of these trees was comparable to observations by Castle et al. [
21] for red maple and yellow birch. The high number of trees having a single straight stem (SSS) are positive for the wood industry, as bucking of straight stems is easier and increases the harvesting productivity [
49]. It is, however, necessary to mention that trees exhibiting single straight stems are more easily identifiable in the field, thus possibly triggering more classification errors for more complex tree forms.
More than two thirds of trees inventoried had risk classes 1 or 2. This result suggests that either the probability of risk-causing agents is weak and the consequent mortality is very low or low and a stable evolution of value of products in the stem is expected [
32], or these agents are causing a rapid death of the trees, leaving a small number of trees for the record.
The influence of the RSDI of plots was significant both on form and risk classification but it was more pronounced for form as compared to risk. The form of trees, which is partly driven by the presence or absence of large branches on the stem, is known to be influenced by the stand density [
50,
51]. However, in our study, the assumption that tree quality is better with higher density does not hold in most of the interactions when involving form groups.
The trees in the hardwood-dominated stands generally presented better form and health in the systematic plots than in the other sample type. An explanation of this may lie in the difference in the age distribution between the two sampling groups: trees in the systematic sampling tend to be smaller in DBH, which suggests that they are younger than the trees from other sampled plots, mostly mature stands ready for harvest. It is reasonable to hypothesize that form and health decreases as stand ages. The proportion of form and risk classes varied with the interaction of ecoregion, sample type, hardwood or softwood dominance, and tree species, with no outstanding patterns.
4.1.4. Diameter and Height
Species showed different ranges of DBHs across ecoregions, with an obvious trend of higher overall values in trees located in mature plots as compared to those selected systematically. We did not find such differences for height. The range of the DBH measured was higher in most of the ecoregions, both in sugar maple and yellow birch. Sugar maple diameters in ecoregion 4 were smaller than in the other ecoregions, possibly linked to the presence of rocky soils with low fertility [
28], which limit growth and quality [
52]. In terms of height, no ecoregion stood out for yellow birch, whereas sugar maple tended to be taller in ecoregions 1 and 7. Overall, sugar maple tends to be higher than other species.
However, the effects of site and stand variables on diameter and height are difficult to interpret without some knowledge on stand age and disturbance history. When considering form and risk classes to DBH or height, clear trends could not be observed.
4.2. Product Recovery
The first analysis at product level was based on the occurrence of different product types according to the total volume generated per tree. The 3712 logs measured were mainly classified as pulp, with only one fourth being veneer or sawlogs. The proportion of unutilized logs was overall low (2.2%–16.1%), with greater shares in hardwoods, especially when bucking sugar maple and yellow birch.
The multivariate analysis did not allow to segregate trees with high and low ratios of merchantable volume, on the basis of site, stand, or tree attributes. Some conditions were more associated with trees with a lower proportion of merchantable volume, but they were totally overlapped by those associated with trees with a high ratio.
Conversely, the multivariate analysis was successful at identifying conditions associated with high quality log yields. Veneer and sawlog ratios were correlated more strongly to environmental factors and stand properties than to tree attributes. A high recovery of veneer and sawlogs was found to be strongly linked to sites with low temperatures in summer and winter, as well as low drought occurrence. Furthermore, high veneer and sawlog yields were associated with SE and KE soil types both located in north-western New Brunswick. These finding supports the conclusion on the importance of environmental variable on veneer and sawlog recovery as previously shown by Hassegawa et al. [
22] and Guillemette and Bédard [
23]. The absence of relationships with site variables observed by Castle et al. [
21] in a study area similar to ours is likely an artifact caused by the application of general linear modeling with collinear predictors, as hypothesized in the introduction.
Higher sawlog and veneer yields were also linked with low stand densities, as measured with the RSDI and the number of trees per hectare. This seems to contradict the hypothesis about the positive effect of the stand density on stem quality. A possible explanation is that in our sample of trees, high stand densities may be associated with younger stands, with high stem counts and small trees, many of them too small to bear sawlogs or veneers. However, the lack of information on stand age did not allow us to test this assumption.
In comparison to the other factors, the correlations between sawlog and veneer recovery with tree attributes are weaker. Total tree height and the height of the living crown were the most significant and showed a stronger link to veneer and sawlog recovery than the form classes, as pointed by Castle et al. [
21]. The weak correlation with form classes suggests that, overall, tree inclination, bole straightness, or the presence of multiple stems or multiple forks have a smaller importance on the proportion of veneer and sawlog, in comparison to the length of the branch-free bole. Contrary to Hassegawa et al. [
22], our study did not allow us to make a distinction between sawlog and veneer recovery of sugar maple and yellow birch, based on site, stand, and tree attributes.
This contrasts with the results from Fortin et al. [
20] who predicted the volume by log grades using a tree quality grade as predictor: in this case, the tree grading system was using criteria similar that the ones used for the log grading system, which likely explain the quality of the statistical relationship. The reader should notice that the later log grading system, designed by Petro and Calvert [
53], is mostly based on the assessment of length of the branch-free bole.
Results from the analysis of the ratios of veneer and sawlogs to total volume indicated that environmental factors have a stronger influence on product recovery as compared to the tested tree attributes. This may allow a classification of Acadian forests into areas with an expected high output of merchantable volume, particularly concerning veneer and sawlogs, following the idea of “hot spots” proposed by Hassegawa et al. [
22] for the province of Québec. Such a distinction between areas may also allow the concentration of forest operations and silvicultural investments in regions where high value recovery is expected. In regions with less favorable environmental factors, stand improvement treatments should be considered to promote trees of particularly good quality, thus maximizing their dimensions.
4.3. Methodological Limitations
Data were provided from several field campaigns with various protocols and field crews. It is conceivable that the inherent methodological differences influenced some of the inventory results, for example form and risk classification, in comparison to more controlled studies.
At product level, the ratio of different product types depending on total volume bucked for each tree was generated. This ratio was undoubtedly sensitive to different product classification and company specifications as well as to individual performance of harvesting operators. We hoped that the possible bias through product classification could be reduced by the creation of product groups. The influence of harvesting operators could not be quantified. Finally, databases including assessment on product recovery were limited to ecoregions 3 and 5, therefore not permitting extrapolating beyond these ecoregions.
4.4. Opportunities for Further Studies
This study did not include any historical data on forest management and natural disturbances. These are known to impact different variables examined in this article, such as species abundance [
47,
54], or growth trends and risk of mortality [
55]. Historical data also seem to have an impact on value recovery of hardwood species [
22], thereby allowing for a more refined analysis concerning product recovery. In the same direction, the assessment of factors influencing the form and risk classification may allow more detailed conclusions on factors influencing risk classification such as presence of fungus, competition, or structural weaknesses [
32].
Several datasets were assessed in limited ecoregions (data sources D and M for example), thus limiting their explanatory potential to other ecoregions. Even though the study on products was limited to a small area of Acadian forests in ecoregions 3 and 5, environmental factors remained prominent in the prediction of volume recovery. The inclusion of future product studies from all ecoregions may provide new information on possible regional differences.
Every bucking decision in hardwoods rests with the operator, a facet not considered in this project, and yet it strongly influences the performance of the operation [
56] and likely the recovery of sawlogs and veneer. This topic should be addressed in upcoming research.
5. Conclusions
Our study aimed at characterizing the distribution of tree, stand, and site attributes and assessing their influence on the recovery of merchantable, sawlog, and veneer volumes in Acadian forests.
First, site factors, summarized by ecoregions, had a significant influence on the species composition, stand densities, form and risk classes, diameters, and heights of trees. Second, the recovery of the merchantable volume of veneer and sawlogs was highly influenced by certain site attributes such as the soil type, summer and winter temperatures, and annual precipitation. Tree attributes such as form or risk class played a smaller role for the sawlog and veneer recovery but should not be discounted as they are also especially useful for silvicultural and operational decision making. This paper confirms that incorporating tree, stand, and site attributes in planning could enhance the predictability of forest operations and product recovery in northern hardwood stands.