Growth Dynamics of Young Mixed Norway Spruce and Birch Stands in Finland

Abstract

Mixed-species forests in Fennoscandia are of increasing interest because they may improve resilience, biodiversity, and productivity. Currently there is scarce knowledge available of the early growth dynamics of mixed spruce–birch stands in even-aged managed production forests with artificial regeneration of spruce. The main objective of our study was to examine the present state and the past growth dynamics of current single-storied, young spruce–birch (Picea abies (L.) Karst., Betula pendula Roth, Betula pubescens Ehrh.) stands (age 17–29 years), where spruce was planted and birch naturally regenerated, and juvenile management practices (early cleaning and precommercial thinning) were carried out. We inventoried ten such stands in Southern Finland, for a total of twenty plots. For 160 spruces and 160 birch trees, we reconstructed the past diameter and height growth through stem analysis. We analyzed mean stand characteristics by tree species, and we modelled the individual tree height and diameter growth using the mixed effects Chapman–Richards model. Spruces had slower initial height growth, but by the age of about 20 years their height growth rate eventually approached and exceeded that of birches regenerated naturally at the time of spruce planting. The diameter growth of planted spruce exceeded that of birches even sooner (at the age of about 10 years). Thus, spruces are not suppressed by birches, and they may coexist in the same canopy layer in managed stands. Contrary to earlier guidelines, due to the fast growth of planted spruces, birch mixture needs to be maintained already in the first juvenile stand management (i.e., early cleaning). The growth dynamics of young, planted spruce, and naturally regenerated birch allow the establishment and management of such mixtures and also maintenance of the mixture in the future until the end of the rotation, thus improving biodiversity in boreal, planted spruce forests.

1. Introduction

In recent years, interest in mixed forests has risen greatly. The great importance of forests in mitigating climate change, conserving biodiversity, and protecting water and soil, creates a continuing need for forest management to evolve and meet the many different goals of silviculture [1]. Mixed forests could improve resilience and increase biodiversity, and thus be more adaptive to environmental changes and ensure the stability of raw material supply with wider product portfolio [2].

Broadleaved tree admixture has been shown to have many benefits in conifer dominated stands. Broadleaved mixture will improve soil properties, increase understorey species richness and biomass, provide suitable habitats for many species, and resilience against both biotic and abiotic risks compared to spruce monocultures [2]. Mixtures of tree species in forest stands have been reported to increase the biomass production for some tree species combinations (e.g., [3,4,5]), especially in Central and Southern Europe. For Scots pine (Pinus sylvestris L.) and European beech (Fagus sylvatica L.) mixtures basal area growth was found to be 12% higher compared to pure stands, and such overyielding was not related to site index [6]. However, according to growth and yield studies in the Nordic countries, the effect of species mixtures on productivity has been found to be minor or negligible for the most common tree species in the boreal forests of the region, i.e., Norway spruce (Picea abies (L.) Karst., from now onwards spruce), Scots pine (from now onwards pine), silver birch (Betula pendula Roth), and downy birch (Betula pubescens Ehrh.) [7,8,9,10]. According to Huuskonen et al. [2] one explanation for the difference between the Nordic and Central European studies is the difference in the stocking levels of the studied stands. In the Central European studies, the stands were typically close to full density, or only moderately thinned, while in the Nordic studies stocking levels were lower due to forest management commonly carried out with more intensive thinning.

The effect of birch mixture in spruce stands has been studied in Northern Europe. However, only a few studies have reported slightly higher wood production in the mixed stands compared to spruce monocultures [2]. On fertile sites with 25% silver birch mixture in spruce stands compared to pure spruce monoculture there was 3%–5% increase in volume growth, and a 6%–9% increase in sawlog yield during rotation [8]. Such a moderate birch mix was found to improve wood production in both the short and long term compared to monocultures. Contrarily, the downy birch mixture reduced the yield of both volume and sawlog compared to pure spruce stands [8]. According to Fahlvik et al. [11] spruce–birch mixed forest resulted in lower yield compared to spruce monocultures, but birch species were not distinguished in the study. That study hypothesized that silver birch may have a positive effect on the yield of a spruce–birch mixed forest.

On fertile mineral soil sites, silver birch and spruce are the typical species mixture in Finland [12]. These two species have similar site fertility requirements, and they both have important wood production capability. However, they have differences in temporal growth patterns, shade-tolerance, and resistance to competition. Earlier studies have stated that silver birch has more vigorous growth at a young age, but starts to decline earlier than spruce [7,8]. However, those results are based on stands where the soil preparation methods (patch scarification, disc trenching) and plant material (bare-root seedlings) were different from current practices. Modern establishment methods applying soil preparation by mounding, container-grown seedlings and genetically improved regeneration material have accelerated the early growth of spruce compared to both the previous methods and the naturally regenerated birch [13,14,15,16]. Due to the enhanced growth dynamics of spruce in young, spruce–birch mixed stands, the juvenile stand management of such mixtures may need to be revised. Juvenile stand management includes two separate silvicultural treatments: early cleaning and precommercial thinning.

In planted spruce stands, juvenile stand management includes typically early cleaning (EC) when the stand reaches one meter in height, and precommercial thinning (PCT) when the height reaches 3–4 m [17]. Typically, in EC all broadleaved trees are removed to control the competition of abundant fast-growing broadleaves. During PCT, spruce stands are currently thinned to a density of 1800–2000 stems per hectare including 10% mixture of broadleaved trees in fertile sites to increase biodiversity. If the spruce density is not at an adequate level, a higher broadleaved mixture will be left to grow. Current tendency is to increase the proportion of broadleaved trees in conifer stands and maintain the mixture during the rotation.

Based on the growth dynamics of naturally regenerated spruces and birches, Mielikäinen [8] recommended that at the time of PCT spruce should be 1.5 m taller (12 years older) than birch to obtain a single-storied mixed stand. In addition, Björkdahl [18] suggested 1–2 m height difference between spruce and birch. Later Fahlvik et al. [11] evaluated the single-storied spruce–birch mixture in Southern Sweden. In their spruce stands planted in 1985 with bare-root seedlings, spruces were four years older than naturally regenerated birches and birches maintained a 2-m height advantage over spruces during the observation period.

The earlier results on the growth rate at the time of PCT are contradictory; birch and spruce of the same size have been found to grow at a similar rate [14,19,20], or birch grows faster [8,21] or slower [19,22,23]. The dynamics of the species composition in spruce–birch stands depend on the birch species; silver birch generally grows faster than pubescent birch [8,20,22,24], but the site fertility [8] and climate [25] can also affect the growth rates of birch species. In addition, previous management activities such as early cleaning have also shown an effect on birch growth [14,20,23], although with contrasting results. To clarify the growth dynamics, besides growth difference, it would be essential to know also the age difference between spruce and birch at the time of PCT.

The recommendation and data presented so far are based on naturally regenerated or bare-root seedling spruces, more prone to suffer the early competition of birch. The growth of spruces established using the current methods is now much faster. Thus, we are currently lacking research on how young mixed spruce–birch stands have grown when they were established with modern regeneration methods and materials while being managed at a juvenile stage. This information is needed to revise the juvenile stand management guidelines for mixed spruce–birch stands.

The aim of our study was to assess the current state and past growth dynamics by tree species in managed young (age 17–29 years) spruce–birch stands where spruce was planted. The study material consisted of cross-sectional measurements from ten such stands selected from properly managed production forests. We hypothesized that in current managed, young spruce–birch mixtures—where dominant trees are of the same height on average—silver and downy birches are naturally regenerated at the time of spruce planting (H1). In addition, in this kind of mixture, there are no differences in the past dynamics of the diameter (H2) and height (H3) growth among spruces, silver birches, and downy birches.

2. Materials and Methods

2.1. Study Area

Ten mixed spruce and birch stands were selected in Southern and Central Finland (Figure 1,

Table 1. Summary of studied stands.

Stand	Coordinates		Temperature Sum, d.d.	Age, Years	Site Type	JSM (Year)	C/N Organic	C/N Mineral
	North	East
1	61°22.6′	24°48.0′	1293	23	OMT	PCT (2009)	26.1	20.0
2	61°15.3′	25°16.6′	1271	19	OMT	PCT (2015)	27.3	26.1
3	61°23.8′	25°1.5′	1325	17	OMT	PCT (2012)	24.9	22.3
4	61°22.6′	24°48.0′	1326	24	MT	EC (2002), PCT (2011)	24.2	24.7
5	61°23.0′	24°57.8′	1315	24	MT	PCT (?)	27.9	27.5
6	61°21.6′	25°10.9′	1321	29	MT	?	30.4	27.1
7	61°20.7′	25°5.2′	1305	19	MT	PCT (2014)	25.1	23.5
8	61°37.7′	29°18.0′	1307	19	OMT	EC (2006), PCT (2014)	22.2	20.4
9	61°37.6′	29°18.2′	1307	20	OMT	EC (2006), PCT (2014)	20.1	17.1
10	61°31.7′	29°8.2′	1307	22	OMT	PCT (2016)	24.1	24.0

Note: Temperature sum is the sum of degree days above 5 °C; Age is the mean age at stump height of sample-tree spruces; MT, and OMT are the Myrtillus, and Oxalis–Myrtillus site types, respectively. Juvenile stand management (JSM) depicts early cleaning (EC) and precommercial thinning (PCT); C/N organic and C/N mineral are the carbon/nitrogen ratio values, respectively, for organic and mineral layers.

). We selected them according to the following criteria: (1) proportion of birch mixture at minimum 15%; (2) mean height difference between spruce and birch at maximum three meters; (3) stand area at least two hectares; (4) development class, young stand (average diameter from 8 to 16 cm); (5) establishment method, spruce planted and birch naturally regenerated; (6) properly managed, i.e., juvenile stand management practices conducted; and (7) without full species segregation.

The site type was fresh site (Myrtillus forest type, MT) for four stands, suggesting medium fertility, and herb-rich site (Oxalis–Myrtillus forest type, OMT) for six stands, suggesting higher fertility [26,27]. The stands were even-aged, from 17 years to 29 years old. Light soil preparation (e.g., disc trenching or patch scarification) was done before regeneration activities in each stand. In all stands, juvenile stand management (early cleaning and precommercial thinning) practices were carried out according to the Finnish silvicultural guidelines for private forests [17]. In three stands, juvenile stand management was carried out twice and in six stands only once. For one stand, the year for precommercial thinning was not recorded. Detailed information on juvenile stand management was not available, but stand density, even distribution, and stem quality were considered in tree selection as recommended [17].

2.2. Data Collection

Two circular plots with an area of 500 m² were established in each stand in 2020. Specific locations depended on stand shape but at minimum 25 m distance from the edges. Plot locations were planned based on office maps and finalized in the field. If plots showed less than 15% of mixture for the secondary species, they were horizontally displaced to the north and then following the other cardinal directions until the criteria above were met. The subjective selection of the sample plots was due to the need to study actual mixtures with local neighbors where the two species had developed together.

To evaluate soil fertility in terms of carbon/nitrogen ratio of the soil more accurately in addition to site type definition, the soil samples were analyzed. On each plot, ten soil cores both from organic layer (Of (fibric) + Oh (humic), litter layer not included) and mineral soil were systematically taken with a steel auger (diameter 58 mm). The ten replicates of organic and mineral soil were combined in plastic bags to form one composite organic layer sample and one composite mineral soil sample. Composite samples were placed in the open bags at room temperature, brought quickly to the laboratory and stored at 2–3 °C before analysis. From the mineral soil sample the soil type was defined. The pH and content of carbon and nitrogen were assessed separately for the organic and mineral layers. Soil pH was measured in H₂O suspensions with a pH meter (Denver, Model 20). Total C and N were determined using an automated CHN analyzer (Leco CHN-600).

On each plot, the location, diameter at 1.3 m (dbh, in mm) and height of each crop tree was measured. Only trees with a height of more than 1.3 m were measured. Heights were measured with 10 cm accuracy using Vertex IV [28]. Crop trees were defined as the ones which were left to grow after PCT and expected to produce merchantable timber in the first commercial thinning. Crop trees also included some other tree species, mainly Scots pine (Pinus sylvestris L.) left to grow in openings after PCT. Overall this definition of crop trees meant that in eight stands all trees were measured on the plots and only in two stands, were there a few understorey trees not measured as crop trees.

On each plot, 16 sample trees were selected amongst the crop trees, respectively eight spruce and eight birch trees. Silver and downy birches were selected as sample trees in the same relative proportion as they were found on the plot. In total, 320 sample trees were measured: 160 were spruces, 126 silver birches, and 34 downy birches. The selection was based on the individual tree basal area distribution by species on each plot. The distributions were split into four classes of equal range, and two trees were sampled from each class. Only good quality, single-stem trees were considered as sample trees, with a final subjective choice by the field crew.

The annual past height and diameter growth were measured from sample trees to determine the growth dynamics of different tree species. Each sample tree was felled and core increment samples at stump height (0.1 m), breast height (1.3 m), and six meters height (6.0 m) were collected for both spruces and birches, and additionally at four meters (4.0 m) for birches only. These stem heights are commonly used in stem analyses. Past under-bark diameter growth and total age of the trees were measured by means of the annual ring increment (considering each distance between successive rings as one-year growth). Ring-widths were measured using the WinDENDRO™-software 2019a [29] after digital scanning for spruce and mainly visually by microscope for birch samples.

Past annual height growth was also assessed. For spruce, the annual height increments were estimated by means of identifying whorls on the stem (considering each distance between successive whorls as one-year growth). In the top part of birch stems, where terminal bud scars and differences in stem thickness, bark color, and resin warts between the successive annual height increments are still visible, the height growth of the current and four earlier growing seasons was measured.

Past height growth reconstruction of birches was based on the ring counts of increment cores collected at the 0.1 m, 1.3 m, 4.0 m, and 6.0 m heights as well as the heights measured in the top part of the stems. Annual height growth was estimated assuming that height growth is constant within a stem segment between two adjacent core samples and that, on average, a core sample is collected at the middle of the annual height growth [30,31]. The method gives two height estimates for the years when the tree height exceeds the 1.3 m, 4.0 m, and 6.0 m heights; for those years the average of two estimates was used.

2.3. Data Analyses

The values of stand characteristics (stem number, basal area, volume, mean diameter, mean height, and dominant height) were calculated using the KPL software [32]. Tree volume was calculated using the volume functions [33]. The dominant height was calculated as the mean height of the 100 thickest trees ha⁻¹. Stand characteristics were then analyzed according to the framework suggested by del Río et al. [34]. All the stand characteristics were calculated comprising all tree species when necessary, but only the detailed values for the spruce and birch compartments are shown in the results. In addition, birch tree species were separated because of differences in their growth rates, rhythm, and responses.

Site characteristics on the plots were described by mineral soil type, pH, and the content of carbon and nitrogen in the organic and mineral soil layers. The ratio between the carbon and nitrogen content (C/N ratio) was calculated for both layers. Decreasing carbon to nitrogen (C/N) ratio is related to increasing site fertility [35,36].

Temperature sum was used to describe the location of stands in Finland. For each stand, the long-term average of the annual effective temperature sum (in degree-days, d.d., threshold value +5 °C) was calculated for the period 1990–2019. Threshold value refers to the thermal growing season (daily mean temperature is permanently above +5 °C). Calculation of temperature sum was based on latitude, longitude, and elevation of each stand, and it was estimated by the climate data, with a 10 × 10 km grid resolution [37]. In the data, the annual temperature sum varied between 1271–1326 d.d.

Differences in the current plot-level characteristics of crop-tree spruces and birches (silver and downy combined) were tested by the analysis of variance (Anova) using the Anova function in R [38]. The following species-wise characteristics at plot level were compared: mean diameter at breast height (dbh), mean height, dominant height, stem number, stand basal area, and stand volume. In the analyses, stand was included as a fixed effect to test also the significance of stand effect.

Differences in the tree-level characteristics of sample-tree spruces, silver birches, and downy birches were analyzed by fitting linear mixed effects models for the tree-level variables, namely age, under-bark diameter at stump height (dsh), height, diameter, and height increments over the last four years (2016–2019). The models were fitted using the lme function in R [39]. In modelling, the tree species effect was fixed, and the stand and plot effects were random. The mixed-effects modelling was applied due to the hierarchical structure of the data. The pairwise multiple comparisons between the tree species were performed using the emmeans function in R and the least significant difference (LSD) test after the F-test rejected (p < 0.05) the null hypothesis that the tree species do not differ from each other (H1–H3). Analyses were performed separately for both all of the sample trees and the dominant sample trees defined as the four thickest (at stump height) trees of each tree species on the plot.

Dynamics of the species-specific growth of diameter and height as a function of tree age were analyzed by a sigmoid growth function [40]. Analyses were conducted using the longitudinal data on the diameters and heights of the sample trees (H2, H3). The past diameter growth of sample trees was analyzed using the tree ring samples from the stump height. Instead of breast height diameter, the stump height diameter was analyzed to cover also the earliest growth. The following general model of a Chapman–Richards type growth model was used to predict the diameter at stump height and height of the individual trees at a given age, using the nlme function in R [39].

$$y_{ijkt}={\beta }_1{\left\{1-\exp (-{\beta }_2·ag{e}_{ijkt})\right\}}^{{\beta }_3}+e_{ijkt}$$

(1)

where y_ijkt is the under-bark diameter at stump height (mm) or tree height (cm) of tree i on plot j in stand k at age t; age is the tree age (years); parameter β₁ indicates the asymptotic maximum size of the growth curve, and β₂ and β₃ determine the inflection point (ln(β₃)/β₂) indicating the age of the maximum growth rate. The maximum growth rate is obtained using the derivative of the function at the inflection point (e.g., [41]). All parameters β were allowed to vary across tree species by using species-specific fixed effects and to be linearly dependent of site characteristics (e.g., C/N ratios and temperature sum). First, all parameters β were random at stand, plot, and tree levels, but due to the nonconvergence of the estimation procedure, we had to limit the complexity of the model so that only β₁ and β₃ were the mixed-effect parameters at stand, plot, and tree levels (cf. [38]). The random, normally distributed between-stand (u_k), between-plot (u_jk), and between-tree (u_ijk) effects associated with the parameters β₁ and β₃ were assumed to be correlated at the same level and have a mean of 0 and constant variances, i.e., bivariate normal distributions for both the stand-, plot-, and tree-specific random effects were assumed. The random within-tree errors (e) were normally distributed with a mean of 0 and constant variance. The Levene’s test indicated the heteroscedasticity of error variances by tree species, and thus error variances were estimated for each tree species using the varIdent variance structure. The significance of the model with species-specific error variances was confirmed with the likelihood-ratio test. Also, a first order autoregressive error structure was applied, but no error autocorrelation in the diameter and height models was observed. The Chapman–Richards type growth models were fitted separately for all the sample trees and the dominant sample trees.

The accuracy of the model predictions obtained using the fixed effects only, or both fixed and random effects, was determined by calculating the bias, i.e., the mean of the prediction errors (err = measured − predicted) and root mean square error (RMSE). The relative error statistics (bias% and RMSE%) were calculated by dividing the bias and RMSE by the mean of the predicted response. Additionally, the proportion of explained variance R² = 1 − (var(err)/var(measured)) was calculated as fitting statistics.

Comparisons of the growth dynamics among the tree species was based on the statistical significance of the estimates of the species-specific parameters β. Statistically significant differences (p < 0.05) among the species-specific parameters were evaluated using the anova and linearHypothesis functions in R [42]. The tree species effect was included in several parameters, and thus uncertainty from the parameter estimates (i.e., the covariance among parameters) was accounted for to generate 95% confidence intervals for the species-specific size development and growth rates as suggested by Paine et al. [43]. For comparisons among the tree species, the difference in absolute growth rates was calculated and compared to zero, corresponding to our hypothesis on no difference in the past dynamics of the diameter (H2) and height (H3) growths among spruces, silver birches, and downy birches.

3. Results

3.1. Stand Characteristics

All stands were dominated by spruce (

Table 2. Summary of stand information based on crop trees at measurement time.

Stand	Species	Stem Number, ha−1	% of Stem Number	Basal Area, m2 ha−1	Mean dbh, cm	Mean Height, m	Dominant Height, m
1	All	1870		21.6	11.5	10.8	14.7
	spruce	1000	53	10.0	10.7	9.8	13.3
	silver birch	130	7	2.4	14.7	14.6	15.2
	downy birch	270	14	3.7	12.5	12.4	13.9
2	All	1830		15.7	10.2	9.9	11.6
	spruce	1210	66	10.8	10.4	9.1	10.9
	silver birch	550	30	4.7	9.7	11.6	13.1
	downy birch	60	3	0.2	7.1	9.1	9.1
3	All	2000		16.9	10.0	10.1	12.2
	spruce	1300	65	10.4	9.7	8.8	10.9
	silver birch	600	30	5.9	11.0	13.0	14.5
	downy birch	60	3	0.3	8.2	10.7	10.7
4	All	910		12.9	12.2	10.6	13.4
	spruce	590	65	11.0	14.6	10.9	13.4
	silver birch	150	16	1.2	9.8	11.3	12.1
	downy birch	170	19	0.7	6.4	8.9	10.2
5	All	1600		19.0	11.9	12.5	17.1
	spruce	840	53	10.9	12.3	11.5	15.9
	silver birch	300	19	4.3	12.0	14.3	15.9
	downy birch	370	23	3.1	10.9	12.6	14.5
6	All	2090		24.7	11.5	12.8	18.2
	spruce	1240	59	11.7	10.2	11.1	16.5
	silver birch	460	22	6.3	12.6	16.6	19.1
	downy birch	20	1	0.1	2.8	5.1	5.1
7	All	1810		16.8	10.4	10.2	12.4
	spruce	1140	63	10.6	10.5	10.1	12.9
	silver birch	190	10	1.7	10.4	12.0	12.6
	downy birch	140	8	0.9	8.5	10.5	11.0
8	All	1880		24.8	12.3	12.0	15.5
	spruce	1240	66	18.8	13.2	11.5	15.5
	silver birch	340	18	4.2	12.2	14.5	16.0
	downy birch	180	10	1.2	8.6	11.8	13.1
9	All	1730		22.3	12.3	12.3	14.3
	spruce	1000	58	13.6	12.5	10.8	13.8
	silver birch	520	30	6.3	12.0	14.8	16.6
	downy birch	180	10	2.3	12.3	13.9	14.8
10	All	2050		19.1	10.4	11.1	12.9
	spruce	1140	56	11.7	10.9	9.7	12.9
	silver birch	760	37	6.6	10.3	13.2	14.9
	downy birch	150	7	0.7	7.7	10.8	11.0

). The proportion of spruce, by stem number, was on average 60%, 22% for silver birch and 10% for downy birch, with the remaining tree species mainly pine. Birches contributed also one-third of the total stand basal area and volume (Table 2 and

Table 3. Mean (±sd) plot-level characteristics of crop-tree spruces and birches on 20 sample plots in 10 stands, and the significance of tree species and stand effects based on the analysis of variance. The data not marked with the same letter are significantly different (p < 0.05).

				Effect of Species		Effect of Stand
Variable	Spruce	Silver Birch	Downy Birch	F (1,29)	p	F (9,29)	p
Mean dbh, cm	11.5 ± 1.6a	11.5 ± 1.9a	8.5 ± 2.1	14.18	<0.001	2.85	0.009
Mean height, m	10.3 ± 1.0a	13.6 ± 1.8	10.6 ± 3.0a	17.05	<0.001	1.99	0.061
Dominant height, m	13.6 ± 1.9a	15.0 ± 2.3a	11.3 ± 3.4	12.49	<0.001	2.57	0.017
Stem number, ha−1	1070 ± 251	400 ± 247	160 ± 132	104.09	<0.001	1.67	0.123
Basal area, m2 ha−1	12.0 ± 2.8	4.4 ± 2.6	1.3 ± 1.4	122.71	<0.001	1.75	0.104
Volume, m3 ha−1	71.2 ± 21.1	30.6 ± 19.8	8.0 ± 8.5	81.85	<0.001	2.32	0.029

). At the time of measurements, the heights of dominant crop-tree spruces and silver birches were not significantly different (p > 0.05), but on average crop-tree spruces were more than 3 m shorter than silver birches (Table 3). The mean diameter (at breast height) of crop-tree spruces and silver birches did not differ. Downy birches were shorter and had smaller diameter than silver birches.

3.2. Individual Tree Growth Dynamics

The sample-tree silver birches were one year younger than spruces, but downy birches had the same age as spruces, on average (

Table 4. Mean (±sd) characteristics of all sample-tree spruces and silver and downy birches. Dominant sample trees were the four thickest sample trees of each tree species on the plot in 2020. (Dsh = under-bark diameter at stump height, idsh and ih = the increment of dsh and height, respectively).

Variable	Spruce	Silver Birch	Downy Birch	F-Test *	p
All sample trees in 2020	N = 160	N = 126	N = 34
Age, years	20.6 ± 4.4a	19.5 ± 3.9	21.4 ± 2.5a	4.71	0.010
Dsh, cm	14.2 ± 4.5a	13.2 ± 3.7ab	12.8 ± 3.8b	3.49	0.032
Height, m	11.2 ± 2.4	13.4 ± 2.4	12.3 ± 2.2	46.26	<0.001
Mean idsh in last 4 years, mm	7.7 ± 2.9	6.4 ± 2.6	5.0 ± 2.5	25.81	<0.001
Mean ih in last 4 years, cm	59.5 ± 14.3	50.4 ± 11.0a	46.7 ± 10.5a	30.32	<0.001
Dominant sample trees in 2020	N = 80	N = 72	N = 26
Age, years	21.3 ± 4.8a	19.9 ± 3.9b	21.3 ± 2.6ab	3.61	0.029
Dsh, cm	17.3 ± 3.7	15.1 ± 3.3	13.9 ± 3.5	16.79	<0.001
Height, m	12.2 ± 1.9a	13.9 ± 2.3	12.7 ± 2.0a	29.07	<0.001
Mean idsh in last 4 years, mm	9.1 ± 2.5	7.2 ± 2.6	5.6 ± 2.5	44.05	<0.001
Mean ih in last 4 years, cm	62.4 ± 12.1	51.3 ± 9.9	46.6 ± 10.6	39.37	<0.001
Dominant sample trees in 2010	N = 80	N = 72	N = 26
Dsh, cm	7.3 ± 3.2a	6.6 ± 3.8a	6.8 ± 2.8a	1.09	0.339
Height, m	5.8 ± 2.0	6.7 ± 2.3a	6.7 ± 1.5a	15.26	<0.001

* Based on the F-test and estimated marginal means associated with the fixed effects of tree species in the mixed effects model estimated for the given variable. The data not marked with the same letter are significantly different (p < 0.05).

). As in line with the plot-level characteristics, sample-tree birches had slenderer stems (i.e., higher height diameter ratio) than spruces (Table 3 and Table 4).

Over the last four years (2016–2019), the mean diameter (at stump height) and height increment of sample-tree spruces were significantly higher than those of birches (Table 4). The 4-year diameter increment of silver birches was higher than that of downy birches, especially for dominant birches (Table 4).

The dominant sample-trees for silver and downy birches were on average 2 m and 1 m taller than spruces. Ten years earlier in 2010, the difference in the mean heights of the same trees was less than 1 m and the mean diameters did not differ significantly among tree species (Table 4).

3.2.1. Diameter Growth

The species-specific sigmoidal curves fitted to the reconstructed under-bark diameters at stump height indicated the species-specific growth differences in the early developments of planted spruces and naturally regenerated birches. In the fitted Chapman–Richards models, the value of the asymptote parameter β₁ for birches was significantly lower than that for spruces; the asymptote parameters for silver and downy birches did not differ significantly from each other (

Table 5. Chapman–Richards models (Equation (1)) for under-bark diameter at stump height of all sample trees (8 spruces and 8 birches per plot) and dominant sample trees (4 thickest trees per species and plot). Fitting statistics using both random and fixed and only fixed effects are given.

	All Sample Trees		Dominant Sample Trees
	Estimate	S.E.	Estimate	S.E.
Fixed parameter (species)
${\beta }_1$ (all species)	439.7519	35.0675	492.6726	41.8451
${\beta }_1$ (silver birch)	−60.0643	8.4844	−92.5508	9.6243
${\beta }_1$ (downy birch)	−61.6000	13.8340	−102.6199	13.7100
${\beta }_{1,C/N organic}$ (all species)	−7.1633	1.3532	−6.9743	1.6150
${\beta }_2$ (all species)	0.0782	0.0010	0.0762	0.0013
${\beta }_2$ (silver birch)	0.0233	0.0017	0.0250	0.0021
${\beta }_2$ (downy birch)	0.0020 ns	0.0028	0.0047 ns	0.0031
${\beta }_3$ (all species)	2.5785	0.0957	2.6426	0.0977
${\beta }_3$ (silver birch)	−0.0426 ns	0.0860	−0.0658 ns	0.1117
${\beta }_3$ (downy birch)	−0.4710	0.1416	−0.4423	0.1585
Random parameters
Stand level
sd (${\beta }_{1,k}$)	14.1053	corr (${\beta }_{1,k},{\beta }_{3,k}$)	13.5546	corr (${\beta }_{1,k},{\beta }_{3,k}$)
sd (${\beta }_{3,k}$)	0.2398	0.999	0.1904	0.999
Plot level
sd (${\beta }_{1,jk}$)	6.9125	corr(${\beta }_{1,jk},{\beta }_{3,jk}$)	12.2973	corr (${\beta }_{1,jk},{\beta }_{3,jk}$)
sd (${\beta }_{3,jk}$)	0.0638	−1.000	0.0153	−0.999
Tree level
sd (${\beta }_{1,ijk}$)	66.2934	corr (${\beta }_{1,ijk},{\beta }_{3,ijk}$)	51.7413	corr (${\beta }_{1,ijk},{\beta }_{3,ijk}$)
sd (${\beta }_{3,ijk}$)	0.6505	0.434	0.6220	0.386
Error term
sd ($e_{ijk}$) (spruce)	3.5499		4.0240 a
sd ($e_{ijk}$) (silver birch)	3.9616		4.0307 a
sd ($e_{ijk}$) (downy birch)	3.7767		4.0058 a
Fitting statistics	Random + fixed	Fixed only	Random + fixed	Fixed only
R2	99.6%	74.1%	99.6%	78.9%
Bias, mm (Bias%)	0.07 (0.09%)	0.26 (0.33%)	0.07 (0.09%)	0.26 (0.33%)
RMSE, mm (RMSE%)	3.49 (4.49%)	26.55 (34.28%)	3.49 (4.49%)	26.55 (34.28%)

^ns Non-significant at 0.05. ^a Based on the likelihood-ratio test, the model estimated with different relative variance weights for each tree species was not significantly better than the one assuming homogeneity of variances.

). The C/N ratio in the organic layer was the only significant environmental variable in the models and predicted logically the asymptotic development of the diameter at stump height; the lower the C/N ratio in the organic layer, the higher the asymptote. The parameter β₂ for silver birch was significantly different from that for spruces and downy birches, but in the case of the parameter β₃, downy birches differed from spruces and silver birches.

The most obvious difference between the diameter growths of all and dominant sample trees was the higher asymptotes of the growth model for the dominant trees for all species (Figure 2 and Figure 3). For spruce, the point of inflection at which the diameter growth rate is maximal was 12 years (13 years for dominant spruces), and for birches a few years earlier (Figure 2). The maximal diameter growth rate of spruces was 9.3 mm year⁻¹ (dominant spruces 11.0 mm year⁻¹), and 9.3 mm year⁻¹ and 7.7 mm year⁻¹ for silver and downy birches, respectively (10.4 mm year⁻¹ and 8.3 mm year⁻¹ for dominant silver and downy birches, respectively). At the inflection point, the diameter at stump height was 7.3, 5.5, and 5.0 cm for all sample-tree spruces, silver birches, and downy birches, respectively (for dominant sample trees: 9.0, 6.3, and 5.6 cm).

The differences in the predicted diameter growth rates among the tree species were calculated and evaluated where they are statistically distinguishable from zero (Figure 3). After a slower early growth, the diameter growth rate of all sample-tree spruces was significantly greater than that of downy and silver birch after the age of 8 and 13 years, respectively. Dominant spruces exceeded birches in diameter growth rate a few years earlier.

3.2.2. Height Growth

The Chapman–Richards models for the height growth also showed the species-specific differences between planted spruces and naturally regenerated birches. The asymptote parameter β₁ for spruces was significantly higher than for birches; difference between the asymptote parameters for silver and downy birches was non-significant (

Table 6. The Chapman–Richards models (Equation (1)) for height of all sample trees (8 spruces and 8 birches per plot) and dominant sample trees (4 thickest trees per species and plot). Fitting statistics using both random and fixed and only fixed effects are given.

	All Sample Trees		Dominant Sample Trees
	Estimate	S.E.	Estimate	S.E.
Fixed parameter (species)
${\beta }_1$ (all species)	4426.4543	409.7445	4752.6402	524.6740
${\beta }_1$ (silver birch)	−688.0801	112.3660	−632.0077	129.1824
${\beta }_1$ (downy birch)	−491.1221	187.3695	−677.2739	185.6002
${\beta }_{1,C/N organic}$ (all species)	−41.4767	15.6635	−52.8258	20.0886
${\beta }_2$ (all species)	0.0323	0.0007	0.0328	0.0010
${\beta }_2$ (silver birch)	0.0169	0.0016	0.0156	0.0019
${\beta }_2$ (downy birch)	0.0046 ns	0.0025	0.0080	0.0029
${\beta }_3$ (all species)	1.5163	0.0330	1.4687	0.0431
${\beta }_3$ (silver birch)	-- a	--	−0.0417 ns	0.0541
${\beta }_3$ (downy birch)	-- a	--	−0.1029 ns	0.0774
${\beta }_3$ (birches)	−0.1168	0.0434	--	--
Random parameters
Stand level
sd (${\beta }_{1,k}$)	175.4400	corr (${\beta }_{1,k},{\beta }_{3,k}$)	95.5069	corr (${\beta }_{1,k},{\beta }_{3,k}$)
sd (${\beta }_{3,k}$)	0.0420	−0.111	0.0621	−0.138
Plot level
sd (${\beta }_{1,jk}$)	-- a	corr (${\beta }_{1,jk},{\beta }_{3,jk}$)	252.9621	corr (${\beta }_{1,jk},{\beta }_{3,jk}$)
sd (${\beta }_{3,jk}$)	-- a	-- a	0.0537	0.994
Tree level
sd (${\beta }_{1,ijk}$)	636.7373	corr (${\beta }_{1,ijk},{\beta }_{3,ijk}$)	488.6960	corr (${\beta }_{1,ijk},{\beta }_{3,ijk}$
sd (${\beta }_{3,ijk}$)	0.3605	0.700	0.3052	0.699
Error term
sd ($e_{ijk}$) (spruce)	28.6935		27.6291
sd ($e_{ijk}$) (silver birch)	36.2225		35.5614
sd ($e_{ijk}$) (downy birch)	39.9996		37.7442
Fitting statistics	Random + fixed	Fixed only	Random + fixed	Fixed only
R2	99.4%	79.3%	99.4%	83.2%
Bias, cm (Bias%)	0.24 (0.03%)	−4.77 (−0.65%)	0.20 (0.03%)	−7.37 (−0.94%)
RMSE, cm (RMSE%)	32.52 (4.45%)	185.36 (25.18%)	31.51 (4.06%)	171.45 (21.86%)

^ns Non-significant at 0.05. ^a Due to nonconvergence, the common parameter ${\beta }_3$ of silver and downy birch was estimated and the plot-level random effects were omitted.

). Also in the height models, the C/N ratio in organic layer was the only significant environmental variable; the low values of the C/N ratio also increased the asymptotic development of height. The parameter β₂ for silver birch was significantly different from that for spruces and downy birches. The convergence problems had to be solved in modelling the height of all sample trees. First, plot-level random effects were removed from the model parameters. Omitting plot-level random effects instead of stand or tree-level random effects was based on the Akaike information criterion (AIC). Second, a common parameter β₃ was estimated for both birch species; the parameter β₃ for birches was significantly lower than that for spruce. In the height model for dominant sample trees, the parameters β₃ for silver and downy birches were not significantly different.

The species-specific height growth of the dominant sample trees was very similar to that of all sample trees (Figure 4). The point of inflection (i.e., the maximal height growth rate) was 13, 7, and 9 years for all sample-tree spruces, silver birches and downy birches, respectively (Figure 4). At the inflection point, the height growth rate was 62, 80, and 64 cm year⁻¹, and the height was 6.2, 5.0, and 4.5 m for spruces, silver birches and downy birches, respectively.

For spruce, it took a longer time to exceed birches in height growth rate compared to diameter growth rate. After the age of 23 years, the height growth rate of spruces was significantly greater than that of birches (Figure 5).

4. Discussion

This study analyzed the current state and past growth dynamics of spruces, silver birches, and downy birches in managed young spruce–birch mixtures. The stands were planted for spruce 20–30 years ago and precommercial thinned 5–15 years before the measurements in 2020.

The age of silver birches was, on average, one year younger than planted spruces, but downy birches had the same age as spruces (Table 4). However, the differences in the ages of tree species varied among stands and plots. In general, we could conclude that crop-tree birches were naturally regenerated approximately at the time of spruce planting (H1).

In addition, the age of crop-tree spruces varied within the stands. We had no information on supplementary replanting conducted after the establishment of the stands studied. Most probably the spruces sampled as crop trees also included spruces naturally regenerated into the stand both before and after planting. Our results support the earlier results on regeneration establishment [44] that not all the planted spruce seedlings will survive or be of high quality, and thus also naturally regenerated seedlings are accepted as crop trees in juvenile stand management.

The species-specific sigmoidal curves fitted to the reconstructed diameters at stump height and heights confirmed the species-specific growth differences in the early developments of planted spruces and naturally regenerated birches. There were significant differences in the past diameter and height growth dynamics among the species, which rejects our hypotheses H2 and H3. During the first years after the regeneration establishment, spruces had a lower diameter and height growth rate than birches, but after the age of 10 and 6 years, spruces have had a faster diameter growth than silver and downy birches, respectively. For the height growth rate of spruces, it took a longer time, i.e., 19 and 13 years, to exceed that of silver and downy birches, respectively. In earlier studies, spruce has been found to achieve birch in height growth rate at the older ages; 20–35 years by Tham [24,45,46], and 40 years to achieve higher dominant height by Hägglund [47] and Eriksson et al. [48]. Our results showed also that the early growth of silver birch was faster than that of downy birch [8], but the asymptotic maximum size of silver birches was not significantly different from that of downy birches. However, our stands were still too young to estimate accurately the difference in the asymptotic maximum size of trees.

In the first years of the rotation, birch height growth was considerably faster than in spruce, but over the last four years before the measurement, the spruces grew faster in height (10 cm year⁻¹) compared to silver birches, and even much faster than downy birches. Our results on juvenile height growth are in line with earlier studies, e.g., [7,8,19,24,49]. However, in our study spruces and birches had approximately the same ages whereas for example in the Mielikäinen study [8] the spruces were on average 12 years older than birches. In addition, the height growth dynamics of planted spruce and naturally regenerated spruce are different and thus, the height advantage of birches gained in the earlier stage diminishes. As a result, in our young stands the mean height of birches was, on average, 1.8 m higher than that of spruces, but no significant differences in dominant heights were observed.

Compared to birches, spruces grew better in diameter and the mean dbh of spruces was 1.5 cm larger than that of birches. During the last four years before the measurement, sample-tree spruces grew faster in diameter (at stump height 1.3 mm year⁻¹) compared to silver birches, and even much faster than downy birches. A faster diameter growth of planted spruces compared to naturally regenerated birches in young mixed stands was also noticed earlier by Uotila and Saksa [14] and Holmström et al. [50].

Current structure and growth dynamics in our study material support maintaining the mixture also in the future until the end of the rotation. An earlier simulation study showed that after precommercial thinning only birch trees that are at least as tall as the spruces can maintain their position in the canopy layer for the full rotation, but if the birches are more than 2 m taller the spruce growth will subsequently decline [19]. In the first commercial thinning, to maintain the vitality of birches (e.g., diameter growth, tree crown ratio), special emphasis should be placed on tree selection by providing space for the crowns and roots of the best birches to expand and grow.

5. Conclusions

We studied the regeneration and growth dynamics of planted spruces and naturally regenerated birches. The results will be used in establishing and growing spruce–birch mixtures, and thus improving resilience and biodiversity in planted spruce stands in Southern and Central Finland.

Despite the slower initial height growth, the present height of planted spruces approached that of birches naturally regenerated at the same time. Consequently, spruces are not suppressed by birches and the two tree species may coexist in the same canopy layer in managed stands. The observed growth dynamics of planted spruce and naturally regenerated birches allow the establishment and management of spruce–birch mixtures.

Our results highlight the need for updating recommendations for managing spruce–birch mixtures in the juvenile stand management phase. Current silvicultural recommendations suggest removing birch in the early cleaning. Based on our results, birches should already be considered in the early cleaning, if the aim is to maintain a birch mixture during the rotation and provide also valuable birch saw logs. Birches can be even slightly higher than spruces in the juvenile stand management phase. More research results are needed for mixed forests, and especially of younger stands and their growth dynamics. In this study the focus was on young stands, but it would be essential to analyze species composition, stand structure, and growth dynamics before precommercial thinning in mixed spruce–birch sapling stands.