Introduction
Traditionally, timber from broadleaved tree species has not been used in any large quantities for structural timber or value-added wood products in Norway, where Norway spruce (Picea abies L. Karst) and Scots pine (Pinus sylvestris L.) dominate the timber market (SSB 2023). However, broadleaved species make up about 24% of the standing volume, excluding bark, in Norway with downy birch (Betula pubescens Ehrh.) and silver birch (Betula pendula Roth) together accounting for 16% (Svensson et al. 2021). Norwegian birch timber as a source for hardwood timber-based value chains is currently a neglected and underutilized forest resource in Norway (Zimmer et al. 2023). In 2021, about 1.4 million m3 of hardwood logs were marketed as firewood in Norway and about 0.3 million m3 as industrial roundwood (SBB 2023), with the vast majority of this volume being birch. The Norwegian hardwood sawn timber volume sold on the Norwegian market in the same year amounted to only 1,538 m3 (SBB 2023). At the same time, significant amounts of hardwood-sawn timber including larger volumes of birch are imported to supply the Norwegian wood processing industry (Zimmer et al. 2023). There is, however, increasing interest from industry to utilize domestic birch timber for end products such as wood panelling, flooring, furniture, building materials, and packaging. In this context, managing Norwegian forests for the production of quality birch timber also gained attention (Zimmer et al. 2023).
Increased use of timber from birch and other hardwoods requires that the quality of harvested logs be properly assessed for the potential end uses of the wood. Timber is typically sorted into different quality grades from the highest (veneer logs) to the lowest (pulpwood) in accordance with the number of stem defects present (e.g. Rast et al. 1973; Heräjärvi and Verkasalo 2002). Classifying roundwood timber into distinct quality assortments helps to optimize log utilization and offers a way of improving financial revenue streams from forest management activities. In contrast to Finland (Heräjärvi and Verkasalo 2002), there is currently no standardized quality sorting of harvested hardwood timber in Norway on a national scale (cf. Gobakken 2000). As a result, medium- and high-quality hardwood logs, especially those from downy and silver birch, are not utilized according to their potential monetary value (cf. Heräjärvi and Verkasalo 2002). Depending on the forest type and the stand development stage, managed silver birch-dominated forests in Finland contain 20–70% of quality logs (Heräjärvi and Verkasalo 2002; Niemistö and Hallikainen 2021) and can yield veneer log recovery rates of 5–10% of the total birch harvest removals (Heräjärvi 2002b). These figures are unknown for Norway.
Anzeige
One reason for the missing timber quality sorting in Norway is the lack of an established quality grading scheme for hardwood species. This is, in part, the result of insufficient species-specific knowledge on how to quickly and easily evaluate and assess the quality of roundwood from hardwood species. A preferred sorting procedure would utilize visually detectable log defects to grade roundwood timber in the stand or at timber processing terminals. In addition to stem straightness, the presence of decay, number of branches, cracks, and discoloration (Carpenter et al. 1989; Kilpeläinen et al. 2011), internal knots are an important feature for assessing log quality (Luostarinen and Verkasalo 2000; Heräjärvi 2002a). Knots as a result of broken, pruned, or dead branches (also known as stubs) that are occluded, leave a distinct pattern on the external bark (e.g. Shigo and Larson 1969). Once the tree has fully grown around the branch stub, clear wood can be formed, and timber of the highest quality produced (Heräjärvi 2001).
Knot size and the time since knot occlusion have an effect on the size and the shape of the branch scar (e.g. Shigo and Larson 1969). There are few studies that have tried to link external bark features and internal wood quality in hardwood species (e.g. Thomas 2009; Račko 2013; Lu et al. 2023). These studies often found strong and reliable correlations between branch scar characteristics and knot metrics, especially for hardwood species with smooth bark (Schulz 1961; Thomas 2012; Stängle et al. 2014; Torkaman et al. 2018). In contrast to tree species with smooth bark, defining and measuring birch branch scars is likely more difficult as the external bark patterns to be studied often vanish and become increasingly indistinct with the peeling and plating of birch bark as the tree ages (cf. Schulz 1961).
The objective of this analysis was to evaluate whether correlations between branch scar size and knot features could be found and quantified in birch as well. We addressed the following research questions: (i) Can external birch branch scars be measured and quantified in satisfactory fashion?, (ii) Are external bark scar features indicative of internal wood quality in birch?, and (iii) Is it possible to derive a reliable model to predict knot features such as radius at the time of knot occlusion from external branch scar attributes for birch?
Materials and methods
Data collection
Knots from a total of seven different naturally regenerated and unpruned birch trees were examined in this study. The selection of the sample trees aimed at larger diameter trees with clearly visible, varying branch scars along the trunk. Trees with a larger diameter than the ones selected were avoided as the dark, fissured, or rough bark of older, large diameter trees of Betula pendula but also B. pubescens (Jones et al. 2023) prohibits the detection and examination of branch scars. Species, however, was not determined for the sample trees of this study as the identification based on phenotypic traits is challenging and thus often questionable. This is in part a result of the high morphological variability observed in each species and because the two sympatric birch species can hybridize (Beck et al. 2016). While there seems to be some variation in wood characteristics between B. pendula and B pubescens (Jones et al. 2023), knottiness appears not to differ between the two species (Heräjärvi 2002a). Further, the Norwegian wood processing industry does not differentiate between the two birch species. It is, however, reasonable to assume that both B. pendula and B pubescens were represented in the sample trees selected for this study. The sample trees were harvested in Hobøl, southeast Norway (lat: 59.54302 N, long: 10.894612E). Total height, diameter at breast height (DBH), tree location, and the immediate surrounding (neighborhood) of each sampled birch tree were recorded before felling (Table 1). The felled logs were separated and bucked into properly marked segments and transported to the lab for further processing.
Anzeige
Table 1
Characterization of birch trees harvested for this study
TreeID | DBH [cm] | Height [m] | Tree age at breast height [years] | Height to first dead branch [m] | Height to first live branch [m] | Number of live and dead branches along the lower 7 m of the trunk | Stem form | Total number of studied knots | Number of knots with occluded stub |
|---|---|---|---|---|---|---|---|---|---|
S1.1 | 19.0 | 18.7 | 50 | - | 7.5 | 0 | Straight | 21 | 5 |
S1.2 | 21.0 | 19.1 | 56 | 5.3 | 7.9 | 5 | Straight | 25 | 16 |
S1.3 | 18.5 | 18.5 | 48 | 2.8 | 10.1 | 2 | Straight first 6 m | 33 | 3 |
S1.4 | 19.5 | 18.7 | 38 | 3.1 | 5.4 | 10 | Straight | 26 | 8 |
S2.5 | 20.5 | 19.4 | 36 | 5.2 | 8.5 | 1 | Straight | 26 | 3 |
S3.6 | 25.0 | 19.2 | 69 | 2.1 | 5.4 | 10 | Straight | 21 | 2 |
S3.7 | 22.5 | 21.2 | 48 | 8.1 | 6.7 | 0 | Straight first 5 m | 16 | 3 |
Moss and lichens were removed from all log segments with a commercial high-pressure cleaner to make external bark features more visible. Branch scars and knots were prepared and measured manually, i.e. without computed tomography (CT) or terrestrial light detection and ranging (t-LiDAR) technology. The manual handling of scars and knots can increase measurement accuracy and better match scar dimensions and knots (Stängle et al. 2014; Lu et al. 2023). On each of the logs, branch scars were selected, marked, photographed, and numbered from top to bottom. The selection process aimed to cover the full variation in size and form of the branch scars found for each log while allowing to study as many branch scars as possible. A total of 168 knots were chosen and measured. The branch scar features length of seal height (LS, mm), width of seal (WS, mm), length of mustache (LM, mm), and width of mustache (WM, mm), were measured according to Torkaman et al. (2018) and Stängle et al. (2014) (Fig. 1). Thomas (2009) referred to LM and WM as length and width of bark distortion, respectively.
Fig. 1
Branch scar with seal and mustache and the corresponding recorded metrics length of seal (LS), width of seal (WS), length of mustache (LM), and width of mustache (WM)
×
![]()
The branch scars were cut longitudinally with a band saw, cutting through the center of the seal and the log pith, in order to reveal a cross section of the knot including its branch origin. None of the studied knots were from epicormic branches as all branches forming the knots exhibited connections with the pith. The knot features knot height (HK, mm), knot length (LK, mm), and maximum knot diameter (DK, mm) were measured and recorded for each knot. The stem dimensions of radius over bark (ROB, mm), stem radius under bark (RUB, mm), and stem radius at the time of knot occlusion (RK, mm) were also measured and recorded for each knot. The clear wood radius (∆R, mm) was calculated as the difference between RUB and RK (Fig. 2). An overview of all studied stem and wood attributes and their acronyms is provided in Table 2.
Fig. 2
Knot with features measured in this study including knot length (LK), knot height (HK), maximum knot diameter (DK), radius over bark at knot location (ROB), radius under bark at knot location (RUB), radius at the time of knot occlusion (RK), and clear wood length (∆R)
×
![]()
Acronym | Definition |
|---|---|
DK | Maximum knot diameter (mm) |
HK | Height of knot (mm) |
LK | Length of knot (mm) |
LM | Length of mustache (mm) |
LS | Length of seal (mm) |
RK | Radius at time of knot occlusion (mm) |
ROB | Radius over bark at knot location (mm) |
RUB | Radius under bark at knot location (mm) |
WM | Width of mustache (mm) |
WS | Width of seal (mm) |
∆R | Clear wood radius (RUB-RK, mm) |
Many of the studied knots exhibited what we termed an occluded stub, defined as a part of the former branch of variable length clearly separated from the remainder of the knot (Fig. 3). Occluded stubs often formed a larger angle with the main stem axis (pith) as compared to the rest of the knot. However, we did not adjust the knot measurement procedure in such cases, but additionally recorded the presence of the occluded stub and measured the length of the stub. It was hypothesized that these occluded stubs originate when a dead branch is shed but does not break off close to the stem, creating a branch stub. The following occlusion process is then likely to cause the break and subsequent separation of the stub at some point in time. This is supposed to happen after the stub is already fixed to the tree (i.e. ingrown) as a result of the occlusion process so that the separated stub does not fall off.
Fig. 3
Example of knot with occluded branch stub
×
![]()
Data analysis
Following analyses in similar works (e.g. Stängle et al. 2014; Torkaman et al. 2018), we tested relationships between branch scar attributes and knot features to derive a reliable prediction model quantifying inner wood quality metrics. Because the potential dependent variables RK or ∆R were not normally distributed, we used generalized linear mixed modelling with trees as random effect to build these models. Akaike information criterion (AIC) was used to select the best-performing model. We further compared the different models by means of mean bias (MB) and mean absolute bias (MAB). All analyses were conducted with the statistical computing software R (R Core Development Team 2022) and the generalized linear mixed models were derived with the glmmTMB package (Brooks et al. 2017).
Results and discussion
Correlation between branch scar features and knot metrics
As hypothesized, defining and measuring birch branch scars was difficult as the studied patterns were often vanishing and thus more indistinct as a result of the peeling and plated birch bark. This held especially true for older scars and in particular, seals found on thicker log segments which created some uncertainty regarding the accuracy of the derived measurements. Table 3 provides an overview of the measured branch scar and knot features. 40 of the 168 analyzed knots exhibited an occluded branch stub. Stub length varied between 6 and 67 mm with an average of 27 mm. Knots with occluded stubs were found across almost the entire knot diameter range studied here (data not shown).
Table 3
Summary statistics for branch scar and knot features including length of seal (LS, mm), width of seal (WS, mm), length of mustache (LM, mm), width of mustache (WM, mm), length of knot (LK, mm), height of knot (HK, mm), maximum diameter of knot (DK, mm), stem radius at knot occlusion (RK, mm), stem radius under bark at knot location (RUB, mm), clear wood length (∆R = RUB-RK, mm), as well as selected ratios of the various metrics
LS | WS | LM | WM | LK | HK | DK | RK | RUB | ∆R | LS/WS | LS/LM | RK/RUB | ∆R/RUB | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Mean | 16.2 | 29.9 | 49.4 | 157.4 | 69.5 | 49.8 | 8.8 | 51.2 | 92.6 | 41.4 | 0.6 | 0.4 | 0.6 | 0.4 |
Minimum | 3 | 7 | 9 | 44 | 12 | 3 | 1 | 12 | 72 | 0 | 0.2 | 0.1 | 0.1 | 0 |
Maximum | 42 | 67 | 184 | 345 | 183 | 165 | 22 | 119 | 126 | 106 | 2 | 1.1 | 1 | 0.9 |
Std. deviation | 8.1 | 12.4 | 32.8 | 50.5 | 36.4 | 34.2 | 4.3 | 23 | 11.8 | 23.1 | 0.2 | 0.2 | 0.2 | 0.2 |
As found for other hardwood species (e.g. Thomas 2009; Lu et al. 2023) and despite the measurement uncertainty, branch scar size appeared to be a reliable external indicator for the knot-affected wood metric RK in the birch species studied here. Strong and very strong correlations were found when correlating the exterior branch scar features with the internal knot metrics with LM exhibiting the strongest correlations followed by WM and LS (Table 4). Given its easier identification and examination, the vertical extension of the mustache is likely the more practical and thus more suitable choice for an external branch scar feature to assess internal wood quality.
Table 4
Correlation coefficients between selected branch scar and knot features including length of seal (LS), width of seal (WS), length of mustache (LM), width of mustache (WM,), length of knot (LK), height of knot (HK), maximum diameter of knot (DK), stem radius at knot occlusion (RK), and clear wood length (∆R)
LS | WS | LM | WM | LK | HK | DK | RK | ∆R | |
|---|---|---|---|---|---|---|---|---|---|
LS | - | 0.76 | 0.76 | 0.70 | 0.71 | 0.68 | 0.74 | 0.63 | -0.49 |
WS | - | 0.64 | 0.70 | 0.61 | 0.59 | 0.61 | 0.52 | -0.39 | |
LM | - | 0.71 | 0.90 | 0.94 | 0.86 | 0.71 | -0.61 | ||
WM | - | 0.72 | 0.66 | 0.73 | 0.67 | -0.51 | |||
LK | - | 0.93 | 0.87 | 0.89 | -0.77 | ||||
HK | - | 0.87 | 0.71 | -0.61 | |||||
DK | - | 0.76 | -0.69 | ||||||
RK | - | -0.87 | |||||||
∆R | - |
The studied ratios provided in Table 3 often did not exhibit strong correlations between each other (data not shown). This held especially true for the correlation of LS/WS and RK/RUB, which resulted in a correlation coefficient of 0.11. The only exception was found for LM/WM and RK/RUB with a correlation coefficient of 0.60. Previous studies found strong relationships between the ratio of LS and WS and the ratio of RK and the RUB for European beech (Fagus sylvatica L.) (e.g. Schulz 1961; Stängle et al. 2014). We could not verify such a relationship for birch. We think it is unlikely that birch seals and their corresponding knot counterparts do not depict a similar pattern. It is more likely that measuring birch seals precisely is hampered by the peeling and plated birch bark, which makes it difficult to correctly define the contour of a birch seal. Interestingly, Schulz (1961) stated that relationships between seal size and knot features are difficult to establish for hardwood tree species with irregular bark such as sycamore maple (Acer pseudoplatanus L.). In contrast, the correlation between the ratio of LM and WM and the RK/RUB ratio was much stronger for our data, which is in line with findings from Shigo and Larson (1969). The finding further corroborates that using mustache metrics is more promising in predicting inner wood quality in tree species such as birch. Still, the very weak correlation between LS/WS and RK/RUB found here is somewhat peculiar.
Modelling of the internal wood quality
Given its higher correlation with most of the measured branch scar features as well as its independence from total stem radius at the knot position, models for RK and not RK/RUB or ∆R were derived to predict knot-effected wood content. The predicted RK can then be used to estimate clear wood radius by subtracting it from RUB.
The best-performing generalized linear model predicting RK as a function of external branch scar attributes was the one including log-transformed LM (Table 5). The model appeared to not perform well for knots with an occluded branch stub (Fig. 4). MAB for the full model (fixed and random effects) amounted to 9.9 mm, while MAB for knots with a stub was 14.4 mm. A separate model predicting RK as a function of LM for knots without an occluded stub only (Table 6) resulted in a MAB of 7.6 mm. Extending the models predicting RK by adding LS or WS did not improve prediction accuracy. Predicting RK as a function of LS resulted in slightly lower prediction accuracy (MAB = 14.5 mm) as compared to the model using LM (Table 7; Fig. 5).
Table 5
Fixed effects parameter estimates (± standard error, SE) and statistics of the generalized linear mixed model with gamma errors and an identity link function for radius at time of knot occlusion (RK, mm) predicted with the external branch scar attribute length of mustache (LM, mm) for all 168 knots examined in this study
Variable | Estimate | SE | z-value | p-value |
|---|---|---|---|---|
Intercept | -29.633 | 4.738 | -6.255 | < 0.0001 |
ln(LM) | 21.731 | 1.240 | 17.518 | < 0.0001 |
Table 6
Fixed effects parameter estimates (± standard error, SE) and statistics of the generalized linear mixed model with gamma errors and an identity link function for radius at time of knot occlusion (RK, mm) predicted with the external branch scar attribute length of mustache (LM, mm) for the 128 knots without an occluded stub examined in this study
Variable | Estimate | SE | z-value | p-value |
|---|---|---|---|---|
Intercept | -32.733 | 3.761 | -8.703 | < 0.0001 |
ln(LM) | 21.322 | 1.120 | 19.042 | < 0.0001 |
Table 7
Fixed effects parameter estimates (± standard error, SE) and statistics of the generalized linear mixed model with gamma errors and an identity link function for radius at time of knot occlusion (RK, mm) predicted with the external branch scar attribute length of seal (LS, mm)
Variable | Estimate | SE | z-value | p-value |
|---|---|---|---|---|
Intercept | -10.804 | 4.915 | -2.198 | 0.0279 |
\(\sqrt{{\text{L}}_{\text{S}}}\) | 15.870 | 1.433 | 11.078 | < 0.0001 |
Fig. 4
Radius at time of knot occlusion (RK) for knots without and with occluded branch stubs over length of mustache (LM) with the solid regression curve derived from generalized linear mixed modelling and the dashed lines representing the 95% confidence interval
×
![]()
Fig. 5
Radius at time of knot occlusion (RK) for knots without and with occluded branch stubs over length of seal (LS) with the solid regression curve derived from generalized linear mixed modelling and the dashed lines representing the 95% confidence interval
×
![]()
Effect of occluded branch stubs
The best performing model to predict RK as a function of LM exhibited greater bias for knots with what we termed occluded branch stubs. We hypothesized that these occluded stubs originate when a dead branch is shed, but does not break off close to the stem creating a branch stub. As a result of the subsequent and ongoing occlusion, the stub can break off from the future knot while it is overgrown and thus fixed to the stem simultaneously. From a timber quality perspective, this is likely a worst-case scenario as the broken stub does not fall off but is pushed down into a more horizontal position, which (further) delays occlusion and causes more knot-affected wood. Knots with an occluded branch stub exhibited larger angles formed between the knot (knot length axis, Fig. 2) and the pith (53 degrees) in contrast to knots without such a stub (44 degrees, p-value < 0.001) – which appears to corroborate the former assertion. Likely as a result, the derived RK models mostly underestimated the knot-affected radius for knots with an occluded branch stub. It is obvious from Figs. 4 and 5 that the relation between RK and LM as well as RK and LS would be stronger and the corresponding models thus more accurate and precise if knots with a stub were excluded from the analysis. In line with this, occluded branch stubs appeared to lower the amount of knot-free timber as the presence of stubs had a significantly positive effect when modelling RK as a function of LK (Table 8; Fig. 6).
Table 8
Parameter estimates (± standard error, SE) and statistics of the generalized linear mixed model with gamma errors and an identity link function for radius at time of knot occlusion (RK, mm) predicted using internal wood features including length of knot (LK, mm) and an indicator variable for the presence of an occluded branch stub (stub) as well as their interaction. See also Fig. 6
Variable | Estimate | SE | z-value | p-value |
|---|---|---|---|---|
Intercept | -18.186 | 1.589 | -11.44 | < 0.0001 |
\(\sqrt{{\text{L}}_{\text{K}}}\) | 8.271 | 0.233 | 35.55 | < 0.0001 |
Stub | -14.349 | 6.087 | -2.36 | 0.0184 |
\(\sqrt{{\text{L}}_{\text{K}}}\):Stub | 2.794 | 0.785 | 3.56 | 0.0004 |
Fig. 6
Radius at time of knot occlusion (RK) over length of knot (LK) for knots without and with an occluded branch stub. Regression curves for knots without (dashed line) and with stub (solid line) were derived from a generalized linear mixed model predicting RK as a function of LK and stub presence
×
![]()
The presence of occluded branch stubs clearly separated from the remainder of the knot was not reported in previous similar studies. We do not know whether hardwood species studied in these previous works do not exhibit such occluded, broken off stubs or whether this knot feature was just not recorded. Like wild cherry (Prunus avium L.), the two birch species studied here, particularly downy birch, are known to retain dead branches for a prolonged period (Hynynen et al. 2010; Bartsch et al. 2020). This characteristic likely contributes to the formation of occluded branch stubs. In addition, it appears that the presence of occluded branch stubs observed in this study might be influenced by tree neighborhood characteristics. Based on our limited sample size, it could be that smaller Norway spruce (Picea abies L.) trees in close proximity to a birch stem hamper and potentially prevent the shedding of small branch stubs maybe by protecting them from direct exposure to wind and snow accumulation (Mäkinen 2002). This is in partial agreement with Niemistö (1995) who found that the self-pruning of dead branches was retarded in birch stands of high stem densities as compared to stands with lower stem numbers. However, Heräjärvi (2002) found no or only small differences in birch stem knottiness between trees from birch dominated stands of different stem densities and varying species compositions.
Whether an occluded knot comprises a broken off branch stub cannot be determined based on external bark patterns. However, since knots with a branch stub seem to significantly reduce clear wood content as compared to knots without, the removal of branches and branch stubs along the lower trunk at early stand development appears highly advisable if quality timber production is the primary management goal (Hynynen et al. 2010; Skovsgaard et al. 2018). This holds true for all branches and branch stubs irrespective of their status (live/dead) and size (diameter, length). Promoting the occlusion of knots by removing branches and branch stubs early also lowers the risk for stem wood discoloration in birch, which often originates from dead and broken branches (Hallaksela and Niemistö 1998). It has been recommended to use clippers instead of pruning saws to avoid bark and stem damage, which increases the chances of fungal entry (Hynynen et al. 2010; Skovsgaard et al. 2018). Further, only dead and living branches smaller than 15 mm should be pruned (Luostarinen and Verkasalo 2000) because trees are much more susceptible to fungal discoloration when living branches over 20 mm are removed (Heräjärvi 2002a).
Conclusion
The relationships between external branch scar features and internal wood quality for birch quantified in this study are a first step to potentially grade birch logs for their quality timber potential and overall value. Using emerging technologies such as 3D terrestrial LiDAR, an automatic procedure detecting and measuring branch scars to predict knot-free timber in birch is now possible to some degree. Given its easy identification, the vertical extension of the mustache appears to be specifically promising for such a procedure. Our study thus can be seen as a piece of the puzzle that will hopefully help to establish thriving birch quality timber value chains in Norway similar to the situation in Finland and the Baltic countries. Knots, however, are of course only one timber quality feature among many, often called defects, that can lower high-value timber content in hardwoods including birch. Since occluded branch stubs appear to significantly lower wood quality, their origin, formation and frequency should be studied also for other hardwood tree species than birch in future works.
Acknowledgements
We thank Simen Gjølsjø, Eirik Nordhagen and Thor-Erik Vatne Alstad for their assistance in collecting and processing the materials used in this study. We also thank R. Edward Thomas and an anonymous reviewer for their comments on an earlier version of this paper.
Declarations
Competing interests
The authors declare that they have no potential competing interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.