Assessing Standing-Tree Wood Density by Microdrilling in Tending Forestry Work Carried Out on Norway Spruce (Picea abies (L.) H. Karst) Stands

Abstract

This study analyses the possibility of assessing standing-tree wood density by microdrilling during tending forestry work carried out on Norway spruce stands. The research material comes from 4 experimental plots and consists of 270 trees (78 trees = control variant, 85 trees = moderate variant, and 107 trees = strong variant). The research objectives were to: (1) highlight wood density particularities, (2) identify wood resistance to microdrilling particularities, and (3) assess standing-tree wood density by microdrilling. For the control variant, average density recorded values of 0.357 ± 0.021 and 0.386 ± 0.027 g·cm⁻³; in the moderate variant, values were between 0.359 ± 0.029 and 0.393 ± 0.027 g·cm⁻³; and the strong variant was characterized by the limits of 0.364 ± 0.020 and 0.397 ± 0.027 g·cm⁻³. Average microdrilling resistance values were between 16.6 ± 2.6 and 22.5 ± 3.0% for the control variant; the moderate variant was characterized by the limits of 18.3 ± 3.1 and 23.4 ± 3.3%; and the strong variant recorded value of 19.7 ± 2.6 and 20.5 ± 2.6 (1.5)%. The linear regression results showed that microdrilling resistance increased as wood density increased. Additionally, generalized linear models showed that, when using covariates of microdrill resistance and tree diameter at breast height, there was a significant influence on the dependent variable, wood density, for all considered work variants. These results suggest that it is possible to consistently estimate both quality and resistance in Norway spruce standing trees using microdrilling. Our findings suggest that wood density and microdrilling resistance are dependent on biometric and qualitative characteristics, as well as the amount of tending forestry work conducted on Norway spruce stands.

1. Introduction

Over the years, the concept of wood quality has acquired a particular importance. The characteristics of wood are constantly evolving and are mainly influenced by tree growth, environmental factors (site and climate) and genetics [1]. Wood density is a key functional feature of trees and has important effects on eco-system processes, including biomass and dead wood. The quality of wood is generally described in terms of performance or appearance according to its use.

In order to assess the properties of wood, and by extension, its density, three important elements must be considered: (a) the need to obtain a significant relationship between the quantifiable properties of a tree and specific mechanical properties; (b) primary data must be collected using non-destructive methods and techniques; (c) the study subjects (trees, stands) must be representative of other areas where the results of the research will be applied [2].

Microdrilling resistance is a non-destructive method commonly used to assess wood quality, particularly to detect defects (rot, internal cracks, insect damage, etc.) and assess wood density [3]. Studies have shown that this method of measuring drilling resistance is more appropriate for assessing the quality of trees and wood products, more cost-effective, and more time-efficient [4,5,6,7,8]. The speed of the drilling process and energy consumption can be correlated with the density, rigidity and other technological properties of the wood [9].

To correct for bias, methods such as smoothing functions and trigonometric approaches have been developed to adjust the resistance profiles [3,10,11] (for trigonometric approaches, see [3,12,13]). There are, however, limitations that can prevent the use of microdrilling resistance measurement methods. Accessibility is a common factor, and includes factors such as limited space for handling the device and structural elements (height, orientation of trees) that can block the measurements [14]. A drill bit’s rigidity is reduced when it has a small diameter, making it flexible and more likely to bend and cause deviations from the drilling path or errors for the resistance profile, particularly if deviation from the microdrilling path is not detected [15,16]. If the degree of wear on a drill bit is high, it must be changed to prevent errors that may occur during measuring. Furthermore, the tip of the drill must be replaced if it is blunt, as it tends to cause deviation from the drill path, making it impossible to statistically process the resulting resistance profiles [15,16].

Microdrilling resistance is a method that provides valuable and accurate information in a short time with low costs, and is also a method that meets the special requirements for the examination of trees on foot, as well as the requirements for products resulting from wood processing [15]. Numerous studies have shown that this method of measuring drilling resistance is much more appropriate for assessing the quality of foot trees and wood products, is cost-effective and, unlike the traditional method, is also time-efficient [17,18]. Although microdrilling resistance is influenced by various factors, such as tree species and the measuring device itself, it is important to note that the data are calibrated according to the pursued objectives [18].

Models for predicting the wood density of Norway spruce (Picea abies (L.) Karst.) were developed in long-term thinning and fertilization experiments in Finland and Sweden as a function of the position in the stem, growth in this position, and site index [19]. Additionally, wood density was dependent on the vertical location along the stem in Scots pine, Norway spruce, and birch stands [20]. Wood density was predicted using the information diffusion model [21] and was also compared to X-ray densitometry data using a linear mixed-effects models using Resistograph amplitude, acoustic velocity, tree diameter, tree age and site index [10].

This study aims to evaluate the usefulness of the IML-RESI PowerDrill for assessing standing-tree wood density with microdrilling and for estimating individual tree breeding values during tending forestry work carried out on Norway spruce (Picea abies (L.) H. Karst) stands by examining the relationships between the Resistograph readings and volumetric wood density of the same trees. The following objectives are considered:

• Highlighting wood density particularities;
• Emphasizing wood resistance to microdrilling particularities;
• Assessing standing-tree wood density by microdrilling.

2. Materials and Methods

2.1. Study Site

The studied Norway spruce (Picea abies (L.) H. Karst) stands were located in Dorna Candrenilor Forest District (experimental plots Tesna—P1: 47°20′57″ N and 20°05′29″ E; Bancu—P2: 47°25′42″ N and 25°09′53″ E; and Zambru—P3: 47°15′51″ N and 25°03′40″ E) and Iacobeni Forest District (experimental plot Ciotina—P4: 47°23′33″ N and 25°17′37″ E), Suceava county, north of the Eastern Carpathians (Figure 1).

The study area is characterized by a continental climate. The average annual precipitation ranges between 887 mm and 1045 mm, while the average annual temperature is between 2.9 °C and 4.9 °C. The geologic underlayer from the studied area comprises volcanic mountains, in which a mosaic of volcanic rocks can be found, as well as alternations of sedimentary rocks, such as sandstones and marls (P1, P2, P4). The predominant rocks are chlorite and sericitous schists (P3) specific to the volcanic area and the Transcarpathian flysch, and the crystalline area and the extra-Carpathian flysch, respectively. The dominant soil is eutric cambisol, with a clay–sandy texture. The site is characterized by mono-specific Norway spruce (Picea abies (L.) H. Karst) stands.

The studied stands for the evaluation of wood density by microdrilling in tending forest work carried out on Norway spruce (Picea abies (L.) H. Karst) were between 60 and 105 years old (

Table 1. Data identification of the component stands where they were installed for assessing standing-tree wood density by microdrilling.

Experimental Plot	Age (Year)	Composition	Exposition	Slope (°)	Altitude (m)	Soil Type	Station type	Forest Type	Density index	Production Class
P1	85	10MO	N	10	1000	3201	2333	1111	0.7	1
P2	105	10MO	NW	25	980	3201	2333	1111	0.7	1
P3	60	10MO	W	15	1050	3201	2333	1111	0.7	2
P4	90	10MO	SE	23	980	3201	2333	1111	0.7	2

). Regarding stand composition, the investigated experimental blocks contained Norway pure spruce stands. The expositions on which they were developing were N, NW, W and SE. The minimum altitude of the terrain was 980 m and the maximum altitude was 1050 m, with a slope between 10° and 25°. The soil type was typically districambisol (3201) which, correlated with favorable climatic conditions, provided good conditions for the development of forest vegetation. This is reflected by the stands’ increment, which achieved high production classes (1 and 2). The type of site was “Norway spruce mountain, large and medium edaphic brown, with Oxalis Dentaria +/- acidophilic” (2333) (Romanian classification). The site was of superior quality for Norway spruce. The corresponding natural forest type is “Normal Norway spruce forest with Oxalis acetosella” (1111) (Romanian classification). The studied stands had a density index of 0.7.

The experimental blocks P1 and P2 were installed in 1966 and 1967, respectively, with three work variants (control, moderate, strong) for analysis of the development of pure spruce stands under the influence of thinning. The silvotechnical works were executed with a periodicity of five years between 1966 (1967) and 1996 (1997). Three working variants (control, moderate, strong) were located in the experimental block P3 in 1978. The tending operations were executed with a periodicity of 5 years between 1978 and 1998. Experimental block P4 had two work variants (moderate, strong), and the tending forestry work was performed with a periodicity of 5 years between 1967 and 1987. In 2004, 2011, 2021, the analyzed experimental blocks were re-evaluated. The terms used to describe work variants (control, moderate, strong) refer to the intensity of applied tending forestry work (control—the intensity per volume was 0%, with no forestry work; moderate—the intensity per volume was between 6 and 15%; strong—the intensity per volume was between 16 and 25%).

2.2. Samples

To assess the measurement of standing-tree wood density by microdrilling in tending forestry work, we sampled 270 trees in the studied experimental plots installed throughout different work variants. This sample included 78 trees (control variant), 85 trees (moderate variant) and 107 trees (strong variant) (3–4 trees from all diameters categories).

2.3. Method

In order to achieve the proposed objectives, the main elements that were examined in this study were density and microdrilling resistance. The data base required to determine wood density included spruce trees from all work variants of the experimental blocks. The density was determined using the volumetric method and through the indirect determination of the amplitude of the microdrilling resistance. The samples for each tree (two on perpendicular directions) were extracted at a DBH height of 1.30 m using a Pressler drill [22] in all the experimental blocks (Figure 2). The bark was not removed from the Resi sample, which went all the way to the pith. Additionally, each sample was recorded and measured in the field. We measured the length of the core with bark up to the pith and the diameter was considered as the inner diameter of the Pressler drill. Samples containing resin pockets, knots and root rot were avoided and new samples were collected. Samples were then placed in sealed plastic bags and stored in refrigerators to keep them cool while in the field. Samples were then oven-dried at 103 °C. Dry mass was measured with an analytical balance with four decimal places and fresh sample volume was determined from the length and diameter of the green core. Density was calculated as the ratio of dry mass to fresh sample volume [11].

Measuring microdrilling resistance, is less invasive than taking increment cores [23,24] and has great potential as a tool for assessing wood traits [24,25]. Microdrilling tools record the magnitude of the torque resistance (cm∙min⁻¹) and frequency of rotation (rpm) sustained by a fine drill when it is driven through wood at a given speed (cm∙min⁻¹) [10,26]. To measure the microdrilling resistance, we used the IML-RESI PowerDrill apparatus (Instrumenta Mechanik Labor System GmbH Großer Stadtacker 2, 69168, Wiesloch, Germany). The used sampling conditions for the IML-RESI PowerDrill (30 cm maximum drilling depth) across all plots and variants were: common feed speed with sharp drilling needles for Norway spruce (Picea abies (L.) H. Karst) at about 200 cm∙min⁻¹, and tree inspection generally at 2500 RPM. The device has an electronic data storage tool and saves all measurements. For each experimental plot, the microdrilling resistance was measured at a height of 1.30m and the area-weighted resistance of the pith-to-bark Resi trace was calculated by weighting the resistance value at each 0.1mm sampling interval by the cross-sectional area it represents.

An average was calculated based on the microdrilling resistance value recorded by the device, the area of the circle ring, the overall area (the ring diameter) and the number of rings within a threshold depth of 0.1 mm [15]:

$$R=\frac{{{\displaystyle \sum }}_{b=1}^n{R}_{ab}^{\prime }{A}_b^{\prime }}{A_t}$$

where R is the average microdrilling resistance value; R’_ab is the microdrilling resistance value for sector a of the annual ring b; A’_b is the area of the b circle ring, about 0.1 mm wide; A_t is the cross-section area where the measurement was performed; and n is the total number of rings, about 0.1 mm wide.

Finally, we quantified the relationship between wood density and the microdrill resistance values indicated by the IML-RESI PowerDrill device.

2.4. Statistical Analyses

Basic statistical descriptors (average, standard deviation, variation coefficient, asymmetry, excess, minimum and maximum) were used to study the effect of tending forestry work on the studied physical properties of Norway spruce wood. The data were expressed as mean ± standard deviation (SD). We used a two-way t-test to examine the significance between the values of density and microdrilling resistance for the three work variants. The significance level for all analyses was accepted to be α = 0.05.

Linear regression models and generalized linear models (GLM) were used to examine the relationships between wood density and the variables generated by the IML-RESI PowerDrill (amplitude of microdrilling resistance). Statistical analyses aimed to quantify the influence of some physical and biometric characteristics of tree wood on density. Microdrilling resistance (MDR) and diameter (DBH) were the predictor variables used to quantify wood density (WD).

Generalized linear models (GLMs) using normal distribution as a link function were used to find the independent variables that most influenced the density of wood. The data subject for statistical analysis took into account both variants of the application of tending forestry work in each experimental block (P1, P2, P3 and P4) and the work variants (control, moderate and strong) of all experimental plots.

In order to compare the performance of different predictors (physical and biometric) of wood on density, we established two model classes representing both the above-mentioned variables and a combination of them. The first class of models were considered single-predictor models containing category variables MDR and DBH as predictors. The second class represented a combination of MDR and DBH as predictors. The model selection used Akaike’s information criterion (AIC), which reflects a compromise between matching the model and loading parameters in which the best model has the lowest value [27]. Choosing the best subset of models, we also calculated the adjusted coefficient of multiple determinations (R²_a), which was used as a selection criterion. Bias and RMSD were used as estimates of the result obtained by the model. Bias is the error in estimates due to systematic mistakes that lead to consistently high or low results as compared to the actual values. The individual bias of an estimate is the difference between the estimated and actual values. The RMSD frequently uses measures of the differences between values predicted by the model and the values observed. The RMSD represents the square root of the second sample moment of the differences between the predicted values and the observed values [28]. Modelling was performed using statistical software STATISTICA 12.5 [29] and MS Excel 2019 [30].

3. Results

3.1. Wood Density Particularities in Tending Forestry Work Carried Out on Norway Spruce (Picea abies (L.) H. Karst) Stands

Average density recorded values of 0.368 ± 0.026 g·cm⁻³ in the control variant and 0.374 ± 0.020 g·cm⁻³ in the strong variant from the experimental plot P1. P2 recorded an average density of 0.386 ± 0.027 g·cm⁻³ (control variant) and 0.382 ± 0.029 g·cm⁻³ (strong variant). In P3, the experimental plot values were between 0.357± 0.021 g·cm⁻³ (control variant) and 0.364 ± 0.025 g·cm⁻³ (strong variant), while P4 registered values of 0.393 ± 0.036 g·cm⁻³ (moderate variant) and 0.397 ± 0.032 g·cm⁻³ (strong variant). When reference was made to the tending forestry work carried out for the control variant, average density values were between 0.357 ± 0.021 g·cm⁻³ (P3) and 0.386 ± 0.027 g·cm⁻³ (P2), moderate variant values were between 0.359 ± 0.029 g·cm⁻³ (P3) and 0.393 ± 0.027 g·cm⁻³ (P4), and the strong variant was characterized by the limits of 0.364 ± 0.025 g·cm⁻³ (P3) and 0.397 ± 0.032 g·cm⁻³ (P4). Tree population variability was diminished in all cases, with a variation coefficient value between 5.6% (strong variant—P1) and 9.1% (moderate variant—P4) (

Table 2. The general statistical parameters (average, standard deviation, variation coefficient (%), asymmetry, excess, minimum and maximum) for density (g·cm⁻³).

Statistic Parameters	Experimental Blocks
	P1			P2		P3			P4
	Co	Mo	St	Co	St	Co	Mo	St	Mo	St
Average	0.368	0.388	0.374	0.386	0.382	0.357	0.359	0.364	0.393	0.397
Standard deviation	0.026	0.032	0.020	0.027	0.029	0.021	0.029	0.025	0.036	0.032
Variation coefficient	6.935	8.294	5.630	6.993	7.535	5.943	8.198	6.796	9.083	8.033
Excess	−0.289	−1.070	−0.827	−0.216	0.091	−0.412	−0.488	−0.698	0.592	−0.064
Asymmetry	−0.113	0.268	−0.057	0.545	0.414	−0.089	−0.142	−0.108	0.533	0.593
Minimum	0.313	0.339	0.310	0.344	0.327	0.315	0.303	0.320	0.316	0.339
Maximum	0.419	0.450	0.384	0.455	0.450	0.394	0.410	0.411	0.479	0.470

Co: control; Mo: moderate; St: strong.

).

The analysis of average values for each work variant indicated that the lowest values were recorded at the P3 experimental plot and the highest at the P4 experimental plot. Considering the average values per work variant showed that the lowest value was characteristic of the control variant (0.357 g·cm⁻³) and the highest of the strong variant (0.397 g·cm⁻³). This is possibly because the age of the trees (60 years in P3) was lower than in the other experimental plots analyzed (80, 85 and 105 years, respectively).

An ANOVA test indicated that significant interaction was detected between the control and moderate variant, the control and strong variant, and between the moderate and strong variant for experimental plot P1, while the differences between the variants in the other studied experimental plots were not significant (

Table 3. Result summary of ANOVA significance tests between the three variants (Co, Mo, St) of the experimental blocks (P1, P2, P3, P4) for density.

Parameters	Experimental Blocks
	P1			P2	P3			P4
	Co/Mo	Co/St	Mo/St	Co/St	Co/Mo	Co/St	Mo/St	Mo/St
Variance	0.00065/ 0.00103	0.00065/ 0.00038	0.00103/ 0.00038	0.00075/ 0.00083	0.00075/ 0.00090	0.00090/ 0.00083	0.00087/ 0.00061	0.0012/ 0.0001
Observations	28	28	28	29	20	20	20	29
df	27	27	27	28	19	19	19	28
F theoretic	4.027	4.027	4.027	4.012	4.113	4.1134	4.0984	3.9732
F experimental	7.671	32.937	10.384	0.457	0.691	0.691	0.691	0.681
p	p < 0.05	p < 0.05	p < 0.05	p > 0.05	p > 0.05	p > 0.05	p > 0.05	p > 0.05
Significance	*	*	*	-	-	-	-	-

-: no significance; *: significant; Co: control; Mo: moderate; St: strong.

).

3.2. Wood Resistance to Microdrilling Particularities in Tending Forestry Work Carried Out on Norway Spruce (Picea abies (L.) H. Karst) Stands

Wood resistance to microdrilling recorded a value of 20.5 ± 2.6% in the strong variant and 23.4 ± 3.2% in the moderate variant from the experimental plot P1. In P2, the experimental plot values were between 20.5 ± 2.6% (strong variant) and 21.3 ± 1.6% (control variant). P3 recorded a microdrilling resistance of 16.6 ± 2.6% (control variant) and 19.8 ± 2.8% (strong variant), while P4 registered values of 19.8 ± 2.6% (strong variant) and 21.5 ± 1.9% (moderate variant). Considering the options for applying the tending forestry work for the control variant, average microdrilling resistance values were between 16.6 ± 2.8% (P3) and 22.5 ± 3.0% (P1), moderate variant values were between 18.3 ± 3.1% (P3) and 23.4 ± 3.3% (P1), and the strong variant was characterized by the limits of 19.7 ± 2.6% (P4) and 20.5 ± 2.6 (1.5)% (P1 and P2). From the perspective of the considered parameter, tree population variability was low, ranging between 1.5% (strong variant—P2) and 3.3% (moderate variant—P1) (

Table 4. The general statistical parameters (average, standard deviation, variation coefficient (%), asymmetry, excess, minim and maximum) for microdrill wood resistance (%).

Statistic Parameters	Experimental Blocks
	P1			P2		P3			P4
	Co	Mo	St	Co	St	Co	Mo	St	Mo	St
Average	22.490	23.388	20.488	21.339	20.479	16.573	18.330	19.769	21.486	19.695
Standard deviation	2.958	3.248	2.593	1.608	1.513	2.809	3.093	2.755	1.847	2.564
Variation coefficient	13.150	13.890	12.657	7.536	7.386	16.950	16.876	13.938	8.594	13.016
Excess	−0.653	−0.556	−0.588	−0.708	−0.805	−1.216	−0.288	−0.996	−0.391	−0.819
Asymmetry	−0.527	−0.569	−0.234	0.080	−0.172	−0.052	0.306	0.399	−0.150	0.528
Minimum	16.040	16.460	15.680	18.500	17.600	11.620	12.880	15.730	17.500	15.900
Maximum	27.080	28.340	25.500	24.500	23.500	20.960	25.220	24.880	25.200	25.000

Co: control; Mo: moderate; St: strong.

).

As with density, the analysis of the average values for each work variant showed that the lowest values were recorded at the P3 experimental plot and the highest values at the P4 experimental plot.

An ANOVA test indicated that significant interaction was detected between the control and strong variants for experimental plots P1, P2 and P3, and between moderate and strong variants for experimental plots P1 and P4, while the differences between the rest of the variants in the studied experimental plots were not significant (

Table 5. Microdrilling result summary of ANOVA tests for significance between the three variants (Co, Mo, St) of the experimental blocks (P1, P2, P3, P4) for microdrilling resistance.

Parameters	Experimental Blocks
	P1			P2	P3			P4
	Co/ Mo	Co/ St	Mo/ St	Co/ St	Co/ Mo	Co/ St	Mo/ St	Mo/ St
Variance	8.74/10.55	8.74/6.72	9.78/6.72	2.56/2.29	7.89/9.77	7.89/8.14	9.77/8.14	3.40/7.56
Observations	28	28	28	29	20	20	20	29
df	27	27	27	28	19	19	19	28
F theoretic	4.026	4.026	4.026	3.597	4.098	4.098	4.098	3.942
F experimental	1.167	6.572	15.726	4.012	3.836	11.531	2.6084.098	3.974
p	p > 0.05	p < 0.05	p < 0.05	p < 0.05	p > 0.05	p < 0.05	p > 0.05	p < 0.05
Significance	-	*	*	*	-	*	-	*

-: no significance; *: significant; Co: control; Mo: moderate; St: strong.

).

3.3. Assessing Standing-Tree Wood Density by Microdrilling in Tending Forestry Work Carried Out on Norway Spruce (Picea abies (L.) H. Karst) Stands

3.3.1. Relationships between Microdrilling Resistance and Wood Density

Correlations between microdrilling resistance and wood density were strong to very strong at the level of the tending forestry work variant. The analysis conducted on the experimental plot variants showed that drilling resistance increased as wood density increased, and was specific for each tending forestry work variant. This occurred in accordance with a y = a + bx linear regression (y represents wood density and x microdrilling resistance). The significance test of the relationship demonstrated that this relationship was average to strong, and also highly significant. Investigations of which regression model would be most suitable for the scatter plot in relation to the experimental data showed that the linear regression equation was used because it best predicted the wood density values that were in the graphical values of microdrilling resistance (

Table 6. Using regression models to predict wood density (WD) with Resistograph readings (amplitude) of microdrilling resistance (MDR) considering the tending forestry work variant.

Work Variant	r	Model R2	Regression Equation	p
Control	0.626	0.392	WD = 0.2733 + 0.0049 × MDR	<0.05
Moderate	0.616	0.379	WD = 0.2681 + 0.0057 × MDR	<0.05
Strong	0.629	0.396	WD = 0.1839 + 0.0092 × MDR	<0.05
All variants	0.602	0.362	WD = 0.2474 + 0.0063 × MDR	<0.05

).

The average amplitude of microdrilling resistance for each variant of improved cuttings was positively related to the density of wood in the regression analyses, with r values ranging from 0.616 (moderate variant) to 0.629 (strong variant). Distinctives of the work variants for the four plots’ r coefficient values were between 0.626 and 0.694 for the control variant, 0.612 and 0.680 for the moderate variant, and between 0.636 and 0.704 for the strong variant.

3.3.2. Assessing Standing-Tree Wood Density by Microdrilling in Tending Forestry Work Carried out on Norway Spruce (Picea abies (L.) H. Karst) Stands

Control variant: DBH alone explained 24% of the variation in wood density (WD), while MDR alone explained 34% (

Table 7. Performances of candidate models for predicting wood density based on different models (control variant).

No.	Class	Model	nf	AIC	Ra2
1	1	DBH	1	−355.3	0.236
2	1	MDR	1	−366.1	0.338
3	2	DBH. MDR	2	−383.3	0.498 *

DBH—tree diameter at breast height; MDR—microdrill resistance; nf—number of fixed-effect parameters; AIC—Akaike’s Information Criterion; R²_a—adjusted coefficient of multiple determination; *—the best model.

, model 1 and model 2). Therefore, both DBH and MDR explained 50% of the variation in wood density (Table 7, Model 3), indicating that the relationship between drill resistance and wood density is influenced by changes in the DBH.

When using the covariates of microdrill resistance (MDR) and tree diameter at breast height (DBH), there was a significant influence on the dependent variable—wood density (R²_a = 0.498)—with a multiple correlation coefficient of R = 0.695 (

Table 8. Parameter estimates of the best candidate by model 3, Table 7.

Parameter	Value ± SE	df	t Value	p Value
Intercept	0.32207 ± 0.1740	75	−18.507	<0.001
DBH	−0.00089 ± 0.0002	75	−4.582	<0.001
MDR	0.00412 ± 0.0007	75	5.985	<0.001
Residual standard error = 0.0200

DBH—diameter at breast height; MDR—microdrill resistance; df—number of degrees of freedom.

).

At the average values of microdrill resistance (20.6%) and DBH (38.3 cm), the range of predicted wood density was 0.373 ± 0.028 g·cm⁻³ (calculated by the linear equation shown by model 3—Table 8). In the models based on the two predictors, the best combination of factors (MDR and DBH) proved to be unbiased (bias < 6 × 10⁻⁶) and had a reduced root mean square deviation (RMSD) (0.0196) (Figure 3).

Moderate variant: DBH alone explained 44%, and MDR alone explained 33% of the variation in wood density (WD) (

Table 9. Model performances of candidate models for predicting wood density based on different models (control variant).

No.	Class	Model	nf	AIC	Ra2
1	1	DBH	1	−379.8	0.436
2	1	MDR	1	−363.3	0.331
3	2	DBH. MDR	2	−395.6	0.546 *

DBH—tree diameter at breast height; MDR—microdrill resistance; nf—number of fixed-effect parameters; AIC—Akaike’s Information Criterion; R²_a—adjusted coefficient of multiple determination; *—the best model.

, model 1 and model 2). Therefore, both DBH and MDR explained 55% of the variation in wood density (Table 9, model 3), indicating that the relationship between drill resistance and wood density is influenced by changes in DBH.

When using the covariates of microdrill resistance (MDR) and tree diameter at breast height (DBH), there was a significant influence on the dependent variable—wood density (R²_a = 0.546)—with a multiple correlation coefficient of R = 0.739. Parameter estimates of the best candidate by model 3 are presented in

Table 10. Parameter estimates of the best candidate model 3, Table 9.

Parameter	Value ± SE	df	t Value	p Value
Intercept	0.3813 ± 0.0229	82	16.599	<0.001
DBH	−0.0018 ± 0.0003	82	−6.231	<0.001
MDR	0.0035 ± 0.0008	82	4.461	<0.001
Residual standard error = 0.0230

DBH—diameter at breast height; MDR—microdrill resistance; df—number of degrees of freedom.

.

At an average value of the microdrill resistance (20.6%) and of the DBH (36.6 cm), the range of predicted wood density was 0.385 ± 0.034 g·cm⁻³ (calculated by linear equation given by model 3—Table 10). In the models based on the two predictors, the best combination of factors (MDR and DBH) proved to be unbiased (bias < 6 × 10⁻⁶) and had a reduced root mean square deviation (RMSD) (0.0196) (Figure 4).

Strong variant: DBH alone explained 43% and MDR alone 34% of the variation in wood density (WD) (

Table 11. Performances of candidate models for predicting wood density based on different models (control variant).

No.	Class	Model	nf	AIC	Ra2
1	1	DBH	1	−483.5	0.434
2	1	MDR	1	−464.9	0.336
3	2	DBH. MDR	2	−488.6	0.496 *

DBH—tree diameter at breast height; MDR—microdrill resistance; nf—number of fixed-effect parameters; AIC—Akaike’s Information Criterion; R²_a—adjusted coefficient of multiple determination; *—the best model.

, Model 1 and Model 2). Therefore, both DBH and MDR explained 50% of the variation in wood density (Table 11, model 3), indicating that the relationship between drill resistance and wood density is influenced by changes in DBH diameter.

When using the covariates of microdrill resistance (MDR) and tree diameter at breast height (DBH), there was a significant influence on the dependent variable—wood density (R²_a = 0.496)—with a multiple correlation coefficient of R = 0.688. Parameter estimates of the best candidate model (3) are presented in

Table 12. Parameter estimates of the best candidate by model 3, Table 11.

Parameter	Value ± SE	df	t Value	p Value
Intercept	0.3547 ± 0.0373	112	9.517	<0.001
DBH	−0.0016 ± 0.0003	112	−5.216	<0.001
MDR	0.0039 ± 0.0013	112	2.807	<0.01
Residual standard error = 0.0241

DBH—diameter at breast height; MDR—microdrill resistance; df—number of degrees of freedom.

.

At an average value of microdrill resistance (20.6%) and of DBH (38.4 cm), the range of predicted wood density was 0.373 ± 0.033 g·cm⁻³ (calculated by the linear equation given by model 3—Table 12). In the models based on the two predictors, the best combination of factors (MDR and DBH) proved to be unbiased (bias < 2 × 10⁻⁷) and had a reduced root mean square deviation (RMSD) (0.0238) (Figure 5).

4. Discussions

4.1. Wood Density

The research conducted on Norway spruce stands stipulates that according to the stational conditions, wood density is between 0.308 g·cm⁻³ and 0.418 g·cm⁻³, with an average of 0.356 g·cm⁻³ [31]. The results obtained by [20] show that the density varied between 0.427 and 0.540 g·cm⁻³ in the second stage of the vegetation season. Additionally, the measured density value for the 1.30 height was 0.386 g·cm⁻³. The Norway spruce wood density obtained in the research conducted by [32] was 0.378 g·cm⁻³ for the considered lot. The research conducted on a Norway spruce (Picea abies (L.) H. Karst) stand of 40 years shows wood density in relation to the planting variant between 0.376 g·cm⁻³ (2500 seedlings·ha⁻¹) and 0.345 g·cm⁻³ (7510 seedlings·ha⁻¹) [33]. Our research shows that, in tending forestry work carried out on Norway spruce (Picea abies (L.) H. Karst) stands, wood density in relation to the work variant was between 0.357±0.021 and 0.386±0.027 g·cm⁻³ for the control variant; 0.359 ± 0.029 and 0.393 ± 0.027 g·cm⁻³ for the moderate variant; and the strong variant was characterized by the limits of 0.364 ± 0.020 and 0.397 ± 0.027 g·cm⁻³. Research regarding Norway spruce wood density indicates 0.378 g·cm⁻³ as an average. This value corresponds to the inferior limit of the dry wood’s data interval for which different authors indicate a density of 0.330–0.680 g·cm⁻³, with an average value of 0.430–0.470 g·cm⁻³ [34].

4.2. Wood Resistance to Microdrilling

The possibility of evaluating wood quality with non-destructive methods has great importance [10,35]. An appropriate assessment using microdrilling resistance requires knowledge of the typical density trends of each species in tree rings and along the drill path. Norway spruce (Picea abies (L.) H. Karst) has an average microdrill resistance of 13.3% with the sampling conditions used due to the combined effects of wood moisture, drill rotation speed, drilling advance speed and wood density [36,37,38]. A significant impact on the measurements of resistance to microdrilling may cause the wood treatments to vary between 15% and 16% in the radial and axial directions [16,39]. When drilling direction was changed from longitudinal to tangential, average resistance advancement decreased by approximately 27% for Scots pine, 33% for common beech, 37% for oak and 40% for poplar. Based on microdrilling resistance measurements, the drilling direction must be considered when predicting wood properties [36,37,38]. As there is a strong correlation between microdrilling resistance and wood properties, this method is recommended for evaluating the density and predicting elasticity of healthy wood [38,40,41,42]. This study, carried out on four Norway spruce (Picea abies (L.) H. Karst) experimental blocks installed with three work variants (control, moderate and strong), shows that the microdrilling resistance (at the DBH height of 1.30 m) was 16.57 ± 2.809% for the control variant, 18.33±3.093% for the moderate variant, and 19.77 ± 3.093% for the strong variant. If the relationships presented are calculated and validated for certain areas and species, this can allow for faster processing times and the precise estimation of wood quality based on the resistance value indicated by the IML-RESI PowerDrill. A common pattern of elasticity based on types of thinning suggests that this is reduced as the intensity of the intervention increases. Microdrilling resistance can be used in the future to monitor changes in wood properties and determine how silvicultural works affect wood properties. The method can also be used in areas where destructive sampling is legally prohibited (i.e., natural reservations and protected areas) [9,43]. The values between 15.1±3.6% (2500 seedlings·ha⁻¹) and 15.5±4.1% (7510 seedlings·ha⁻¹) are characteristic of a 40-year-old Norway spruce (Picea abies (L.) H. Karst) stand installed with different planting variants [33]. Countless studies have shown that this method of measuring drilling resistance is a much more appropriate method for assessing the quality of standing trees and wood products and is also, unlike traditional methods, cost- and time-efficient [18].

4.3. Assessing Standing-Tree Wood Density by Microdrilling

There are numerous studies that show a close relationship between amplitude and wood density (determination coefficient R² > 0.60) [10,40], lumber [41], and coarse woody debris [22]. Wood density has a good correlation with wood strength and these models can be practically useful in predicting ring properties under different management regimes [44]. The parameters of the optimal and simpler models indicated that wood density decreased as site index increased and, due to the inverse relationship, amplitude increased as wood density increased [11]. Using a two-variable model, X = [1/SI, Z0] parameters indicated that wood density decreased by 6 kg∙m⁻³ for each 0.1 increase in the z-value associated with the unadjusted microdrill amplitude, and an increase in wood density of about 1.5 kg∙m⁻³ could be expected [45]. The fitted linear models’ correlating amplitude recorded by the Resistograph device provided greater accuracy in comparison with the models using pin penetration at a constant pressure. Models that correlated amplitude (microdrilling) also had a better accuracy than models that used needle penetration at a constant pressure [9]. The optimal model for our research resulted from using the covariates of microdrill resistance (MDR) and DBH. There was a significant influence on the dependent variable—wood density (R²_a = 0.546)—with a multiple correlation coefficient R = 0.739.

Wood density was predicted using the information diffusion model, and a significant correlation was observed between microdrilling resistance data and wood density (Li et al. 2019). Wood density was also compared to X-ray densitometry data using linear mixed-effect models. Mean Resistograph amplitude and combinations of acoustic velocity, tree diameter, tree age, and site index were considered as fixed effects. An R² value of 0.60 was obtained [10]. Other research found that there was a low correlation between resistance amplitude and density, indicating that variables other than density are involved in wood property changes in agarwood [46]. Microdrilling alone, although highly significant as a predictor, was found to be insufficient to provide accurate estimates of wood density for individual trees. Site effects must also be considered [10].

An aspect that should be considered in future research is the role of abiotic factors (e.g., snow, wind, etc.) in the studied area of the experimental blocks. The area under investigation is part of an area with a very high risk of snow and windthrows. These abiotic factors had a significant influence on the work variants of the experimental plots, especially over the control variant. This could be a reason explaining why the values of the analyzed parameters were close. If, in the case of moderate and strong variants, the human intervention was significant, so was the nature of the specified abiotic factors (wind) in the case of the control variant. For all predictive models, evaluating performance with other datasets is a crucial step in gauging practical usefulness. However, despite the importance of this, the primary focus on predictive modelling studies has centered on how models are developed and on their explanatory accuracy. We suggest that a greater focus should be given to their predictive accuracy through external verification in order to assess the usefulness of models in practice [10].

5. Conclusions

We conclude that the IML-RESI PowerDrill device has a great deal of potential to assess standing-tree wood density by microdrilling during tending forestry work carried out on Norway spruce (Picea abies (L.) H. Karst) stands, in this case, particularly through the analysis of the relationships between Resistograph readings of DBH and volumetric wood density.

Our results showed a linear relationship between wood density and wood microdrilling resistance. The covariates of microdrill resistance (MDR) and tree diameter at breast height (DBH) had a significant influence on the dependent variable (wood density) for all considered work variants (control, moderate and strong). Based on these results, it seems that it is possible to consistently estimate both the quality and resistance of Norway spruce standing trees with the microdrilling technique.

Our findings suggest that microdrill resistance and wood density are dependent on biometric and qualitative characteristics, as well as the amount of the tending forestry work conducted on Norway spruce stands.