Jukuri, open repository of the Natural Resources Institute Finland (Luke) All material supplied via Jukuri is protected by copyright and other intellectual property rights. Duplication or sale, in electronic or print form, of any part of the repository collections is prohibited. Making electronic or print copies of the material is permitted only for your own personal use or for educational purposes. For other purposes, this article may be used in accordance with the publisher’s terms. There may be differences between this version and the publisher’s version. You are advised to cite the publisher’s version. This is an electronic reprint of the original article. This reprint may differ from the original in pagination and typographic detail. Author(s): Pasi Rautio, Ville Hallikainen, Sauli Valkonen, Johanna Karjalainen, Pasi Puttonen, Urban Bergsten, Hans Winsa & Mikko Hyppönen Title: Manipulating overstory density and mineral soil exposure for optimal natural regeneration of Scots pine Year: 2023 Version: Published version Copyright: The Author(s) 2023 Rights: CC BY 4.0 Rights url: http://creativecommons.org/licenses/by/4.0/ Please cite the original version: Rautio, P., Hallikainen, V., Valkonen, S., Karjalainen, J., Puttonen, P., Bergsten, U., Winsa, H., & Hyppönen, M. (2023). Manipulating overstory density and mineral soil exposure for optimal natural regeneration of Scots pine. Forest Ecology and Management, 539, 120996. https://doi.org/10.1016/j.foreco.2023.120996 Forest Ecology and Management 539 (2023) 120996 Available online 21 April 2023 0378-1127/© 2023 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). Manipulating overstory density and mineral soil exposure for optimal natural regeneration of Scots pine Pasi Rautio a,*, Ville Hallikainen a, Sauli Valkonen b, Johanna Karjalainen a, Pasi Puttonen c, Urban Bergsten d, Hans Winsa a, Mikko Hypponen a a Natural Resources Institute Finland (Luke), Ounasjoentie 6, FI-96200 Rovaniemi, Finland b Natural Resources Institute Finland (Luke), P.O. Box 2, FI-00791 Helsinki, Finland c Department of Forest Sciences, University of Helsinki, P.O. Box 27 (Latokartanonkaari 7), FI-00014 Helsinki, Finland d Department of Forest Biomaterials and Technology, Swedish University of Agricultural Sciences, 901 83 Umeå, Sweden A R T I C L E I N F O Keywords: Forest regeneration Pinus sylvestris Preparatory cuttings Site preparation Thinning A B S T R A C T In northern boreal region the growth of forests is slow, and yield and profit are low, which is why low refor- estation costs are important for profitable forestry. If natural regeneration is successful, expensive artificial forest regeneration (planting or direct seeding) can be avoided. In this study, we look at the impact of overstory density and site preparation on natural regeneration and seedling growth of Scots pine. Study stands were established in different parts of Northern Finland and in each stand following treatments: 50, 150 and 250 trees ha 1 or unthinned control, where the stand density was  250 trees ha 1, were randomly allocated to experimental plots. In addition, site preparation (disc trenching, 4000–5000 m ha 1) was carried out on two experimental plots in which tree density was either 50 or 150 trees ha 1. In the experimental stands seedling number, age and growth were monitored for 11 years. Monitoring revealed that the number of seedlings increased with decreasing tree density. Average seedling height growth was very low or even non-existent in the unthinned control and in the densest (250 trees ha 1) treatment, but increased when the density of trees decreased. The highest seedling number and the highest growth were achieved when the tree density was 50 trees ha 1 and the soil was prepared to expose mineral soil. Achieving e.g. 2000 seedlings ha 1, would need about 40% exposition of mineral soil. The required low tree density implies that not only seed supply from seed trees and site preparation is important for regeneration success in northern boreal Scots pine forests but also the reduction of competition by mature trees. 1. Introduction The use of natural forest regeneration as an inexpensive regeneration method is particularly supported in northern part of Finland where the expected revenue is lower compared to the southern part of Finland where the forest growth is faster. Also the increased multifunctional use of forests and the need to keep forests increasingly tree covered throughout their life cycles are highlighted in the north. Many forms of forest use in northern Fennoscandia, such as tourism, reindeer herding and recreation, benefit from the preservation of more continuous forest cover through denser seed-tree and shelterwood stands compared to that used in practical forestry today (Hypponen, 2002). That is why discus- sion about the benefits of Continuous-Cover Forestry (CCF) has inten- sified recently, especially regarding state owned forests, which are abundant in Finnish Lapland. Natural forest regeneration with the seed-tree and shelterwood methods involves the retention of a moderate to large number of over- story trees throughout the restocking period. The retention trees, or part of them, may also remain on site for much longer for certain purposes, especially if aiming at the development of a two- or multi-storied stand in terms of Continuous-Cover Forestry as based on Northern German experience (Heinsdorf, 1994). However, Scots pine (Pinus sylvestris L) is a light-demanding species, and it is questionable how much shading and overstory competition is tolerated during the restocking period and the subsequent stages when seedling growth and survival are critical. In any case, a minor part of the original seed or shelterwood trees are supposed to be permanently retained throughout the management cycle and beyond. In addition to light availability, natural regeneration in a pine forest is affected by multiple intertwined abiotic and biotic factors. This was * Corresponding author. E-mail address: pasi.rautio@luke.fi (P. Rautio). Contents lists available at ScienceDirect Forest Ecology and Management journal homepage: www.elsevier.com/locate/foreco https://doi.org/10.1016/j.foreco.2023.120996 Received 13 January 2023; Received in revised form 31 March 2023; Accepted 7 April 2023 Forest Ecology and Management 539 (2023) 120996 2 clearly demonstrated in terms of the famous German “Dauerwald “ management concept, where the shade tolerance, survival and growth of pine seedlings underneath pine overstories depended very much on site fertility and moisture (Wiedemann, 1925). Of such abiotic factors, soil properties like temperature, moisture, texture, water retention capacity and nutrient level are considered most important. Of biotic factors, competition between mature trees (e.g. Aaltonen, 1919; Hagner, 1962; Ackzell and Lindgren, 1992, Thishler et al., 2020), competing tree seedlings (Peet and Christensen, 1987; Nilsson and Albrektson, 1994; Ruano et al., 2013) and understory vegetation (Lehto, 1956; Hypponen et al., 2013, Huth et al. 2022) are affecting regeneration. Further, the winter conditions are of especial importance in these northerly areas. For example, snow blight (caused by a pathogen Phacidium infestans P. Karst.) can kill snow-covered small pine seedlings and branches of trees of various size classes and cause big losses. The needles are infected in autumn and the pathogen is able to expand under the cover of snow even at –5C. This pathogen causes losses nearly every year, and thus it is a very significant factor in natural selection of Scots pine seedlings (Jal- kanen, 2003, 2007). The role of mature trees on regeneration is complex and contradic- tory. A higher density of shelterwood implies a higher seed yield and a higher number of pine germinants (Beland et al., 2000; Dovciak et al., 2003). On the other hand, overstory trees tend to reduce seed germi- nation as well as seedling survival and growth especially in the closest vicinity of the overstory trees through shading and competition for nutrients and water (Aaltonen, 1919; Lehto, 1956, 1969; Niemisto et al., 1993; Bassett and White, 2001; Hallikainen et al., 2007). Emerging seedlings have not only the tree overstory trees to compete with but also other tree seedlings and forest floor vegetation. As an example of competition between tree seedlings, Nilsson and Albrektson (1994) found that in a high competition pine sapling stand (HCS, 40 000 stems ha 1) compared to a low competition stand (LCS, 10 000 stems ha 1) the mortality was about the same until 5 years of age but increased after that annually up to year 15 in HCS while remaining constant in the LCS. The noteworthy effects of dense brush of broadleaved species on the survival and growth of Scots pine seedlings and their dependence of the pine overstory density as commonly observed and accounted for in Central Europe (e.g. Lust, 1987; Lust and Geudens, 1998) are not com- mon in Northern Fennoscandia where Scots pine management is not much practiced on mesic sites. In case of forest floor vegetation, the pattern is much more compli- cated. Lehto (1956) reported the effect of understory vegetation and thickness of humus layer on seedling establishment being larger than the effect of overstory trees. But in fact, the forest floor vegetation as such can have a contradictory role on regeneration. Some species have been reported to have positive while others to have negative effects on seedling establishment. In a study by Hypponen et al. (2013) the pres- ence of crowberry (Empetrum nigrum ssp. hermaphroditum), feather moss (Pleurozium shreberi) or Dicranum moss species increased the mortality of pine seedlings, while the pine seedling density increased with the cover of heather (Calluna vulgaris L.). In contrast, Nilsson and Zackrisson (1992) found that heather was harmful to seedling establishment. Moreover, Hypponen et al. (2013) found that heather decreased the height growth of pine seedlings while mitigating seedling establishment. Hertz (1934) reported that shrubs do not impair the establishment of seedlings. Similarly, to the results concerning the shrub layer, studies on the effects of ground layer on pine regeneration have also brought conflicting results. Brown and Mikola (1974) showed that lichens have negative allelopathic effects on mycorrhizas and tree seedlings but later on Steijlen et al. (1995) and Kytoviita and Stark (2009) have reported that the effects of lichens are neutral or even beneficial. Kyro et al. (2022) showed that high coverage of hair mosses (Polytrichum spp.) was associated with poorer seedling survival. Whatever the actual effects of tree overstory or forest floor vegeta- tion on regeneration really are, in practical forestry the general assumption has been that their effect is competitive and negative to seedling establishment and growth (Hagner, 1962, Lehto, 1969,. Hence the silvicultural practice has been to decrease the degree of competition Fig. 1. Location of districts and experimental stands and study design in the stands. P. Rautio et al. Forest Ecology and Management 539 (2023) 120996 3 with thinning of the overstory and site preparation for the ground vegetation. These two methods are usually combined in the standard natural regeneration method of seed tree cutting, where mature trees are left at a density of 50–100 trees per hectare to produce a sufficient seed rain while exposing mineral soil for the seeds to germinate and for the seedlings to survive and grow. A seed tree cutting might be preceded by a preparatory cutting where > 200 mature trees are left to the stand in order to strengthen their roots and canopy for about 10–15 years (Hypponen, 2002). Mineral soil exposure has been found to improve the emergence of pine seedlings on xeric and sub-xeric heaths. In practice, the share of the exposed mineral soil does not have to be very large for an improvement in seedling establishment, as an example Hallikainen et al. (2019) have found 10–15% mineral soil exposure of the ground surface area to be enough. Site preparation not only removes competi- tion of field and ground layer vegetation, but also improves physical properties of the soil. Site preparation raises the temperature of the soil (Lahde, 1978; Orlander et al., 1990), and because the heat sum posi- tively influences the result of regeneration (Hypponen, 2002, 2005) the exposed mineral soil can also favor the emergence this way. Heiskanen et al. (2007) found that 21 years after site preparation the water retention at saturation and air-filled porosity at field capacity (10 kPa) were significantly higher and the bulk density lower in ploughed ridges than in the untreated intermediate areas. The fact that Forest floor vegetation can suppress seedlings and the humus layer prevents seed from entering a suitable germination surface in the mineral soil (Nygren and Saarinen, 2001). The aim of the present study was to find out how seedling estab- lishment and growth are affected by silvicultural treatments optimizing competition by mature trees and ground and field layer vegetation. The study is based on a field experiment where we: 1) Manipulated the density of a mature pine stand, and 2) Partly removed the ground and field layer vegetation and the raw humus to uncover the mineral soil. Our hypotheses were: 1) Uncovering the mineral soil enhances seedling emergence with a multiplicative rather than just additive effects when combined with thinning in the overstory and 2) thinning in the overstory increases the number of seedlings and their growth. The results are based on data representing a monitoring period of 11 years on the experimental plots. 2. Materials and methods 2.1. Study area and experimental design The study area covers the forested part of Finnish Lapland that is around 9 million hectares. The area was divided into four districts: southern, western, northern and eastern Lapland. Three experimental stands (replicates) were placed in each district (Fig. 1, Appendix Table A1). The experimental stands had to fulfill the following criteria: 1) sub-xeric site on mineral soil, 2) Scots pine dominated, 3) the number of Scots pine stems at least 250 trees ha 1, 4) large enough for the experiment (at least three hectares of non-fragmented forest), 5) mature forest (up for final harvest and regeneration immediately or no later than 10 – 20 years according to the Finnish Best Practices (Aijala et al., 2019). Due to long south-north distance the age of mature forest varies in the study area from 80 to 120 years. At the first stage of the experi- mental site selection, candidates were provided and examined using the geographic information system of Metsahallitus Forestry Ltd (respon- sible for the management of state-owned forests in Finland). Potential sites were checked in situ, and the best alternatives fulfilling the re- quirements were selected. Especially the size of the homogeneous forest stand played a major role in the selection process. If there were more than one more or less equal alternative pine stand in a district, the site was chosen randomly among these. In each experimental stand, a set of six square treatment plots was placed in the central homogenous area. Each plot was 70  70 m in size. A core plot of 30  30 m was placed in the middle of each of the treatment plots, and all the measurements were carried out in the core plot. The rest of the plot thus constituted a buffer zone with a width of 20 m in all directions. The buffer zone was treated similarly as the core plot (Fig. 1). The treatments for the six treatment plots were random- ized. The treatments were: 1) tree density 50 trees ha 1 without site preparation (50 NT in the figures), 2) tree density 50 trees ha 1, with disc trenching (50 T), 3) tree density 150 trees ha 1 without site prep- aration (150 NT), 4) tree density 150 trees ha 1 with disc trenching (150 T), 5) tree density 250 trees ha 1 without site preparation (250 NT), 6) control, no cutting or site preparation (Control). Tree density in control plots was on average 439 trees ha 1. Site preparation (disc trenching) was not applied with the tree density of 250 trees ha 1, because the massive machinery would have been too difficult to operate among such a dense stand of trees. The number and diameter at breast height of trees with d1.3 > 3 cm was measured before and after cutting. Disc trenching was performed targeting 4000–5000 m site preparation tracks per ha with a track breadth of 60–80 cm, constituting to an exposed proportion of 24–40 % of area. The proportion of exposed mineral soil in different treatment is given in Appendix 1 (Table A2). On each of the six treatment plots, five circular sample plots (20 m2, radius 2.52 m) were placed in a systematic order on the core 30x30 m plot (Fig. 1) before cutting and site preparation. The number and height of the Scots pine seedlings (height  10 cm) were recorded in each subsequent seedling inventory on each circular sample plot. All birch (Betula spp.) and Norway spruce (Picea abies (L.) H. Karst.) seedlings and all pine seedlings with a height over 0.5 m were removed Table 1 Data description. The ground cover (in %) and the values of other continuous explanatory variables. Measurement 1 denotes the measurement of the cover in the beginning and 2 denotes the measurement in the end of the inventory period, avg. being an average (used in the statistical models). All the variables were tested in the models in addition to the treatment effect, number of inventory, and the number of all seedlings in the 20 m2 circular sample plot (in the case of the height model). Variable Mean Median Min. Max. Cover (%) Bilberry 1 (Vaccinium myrtillus) 5.48 1.00 0.00 50.00 Bilberry avg. 6.50 3.67 0.00 51.67 Bilberry 2 6.25 4.00 0.00 45.00 Cowberry 1 (Vaccinium vitis-idaea) 6.77 5.00 0.00 50.00 Cowberry avg. 8.68 7.00 0.25 43.33 Cowberry 2 9.32 7.00 0.00 45.00 Crowberry 1 (Empetrum nigrun) 4.21 1.00 0.00 50.00 Crowberry avg. 7.40 5.17 0.00 41.67 Crowberry 2 9.41 5.00 0.00 50.00 Heather 1 (Calluna vulgaris) 0.81 0.00 0.00 20.00 Heather avg. 2.07 0.00 0.00 21.67 Heather 2 2.88 0.00 0.00 30.00 Lichens 1 6.97 1.50 0.00 60.50 Lichens avg. 5.23 3.00 0.00 46.92 Lichens 2 4.91 3.00 0.00 50.25 Mosses 1 71.89 80.00 0.50 100.50 Mosses avg. 65.89 71.00 2.67 99.67 Mosses 2 65.08 75.00 0.50 102.00 Exposed mineral soil 1 9.18 0.00 0.00 100.00 Exposed mineral soil 2 5.97 0.00 0.00 60.00 Slash 1 5.36 1.00 0.00 100.00 Slash 2 4.46 2.00 0.00 55.00 Other continuous explanatory variables Number of seedlings (>10 cm) in 20 m2 circular plots 6.48 1.00 0.00 168.00 Thickness of humus layer, mm 26.91 24.00 4.00 73.25 Stoniness (penetration of stick, cm) 24.59 23.12 2.50 62.50 Proportion of coarse sandy (2.0 – 0.2 mm), % 48.37 46.93 9.53 83.08 Proportion of fine sandy (0.2 – 0.06 mm), % 35.72 35.96 7.60 85.05 Proportion of silty (<0.06 mm), % 25.84 25.25 1.85 47.72 Temperature sum, d.d. (average of period 1981 – 2010) 747 726 638 925 Age of pine seedlings (years) 10.58 9.75 1.00 36.00 P. Rautio et al. Forest Ecology and Management 539 (2023) 120996 4 from the circular plots at establishment, except for the control treatment that represented the forest without the effects of harvest operations and site preparation. Consequently, some seedlings (h less than 50 cm) remained on the circular sample plots – and on sample plots on control treatment also seedlings h > 50 cm - already from the beginning. The computed seedling count variable included all seedlings with height  10 cm on the sample plot in each measurement. The seedling height used here in the data analysis is the arithmetic mean of seedlings in a sample plot. 2.2. Data collection Composition of the ground and field layer vegetation (% of ground cover) was measured on 1 m2 square sample plots placed in the center of each 20 m2 sample plot after the cutting and site preparation. Also, the proportion of scarified mineral soil and slash (harvest residue) were recorded in the same 1 m2 square sample plots after cuttings (Fig. 1). The vegetation cover was recorded for groups of species such as mosses, grasses, and lichens, and separately for the most common dwarf shrub species:cowberry (Vaccinium vitis-idaea), bilberry (Vaccinium myrtillus), crowberry (Empetrum nigrum) and heather (Calluna vulgaris). The vegetation and the proportion of scarified mineral soil were measured in the beginning of the experiment, and also in parallel with the last seedling measurement (Table 1). A rather long monitoring period during the experiment lead to challenges to describe the cover of the vegetation and the cover of exposed mineral soil because the vege- tation succession proceeded during the monitoring period. Hence, the vegetation inventories were carried out in the beginning and in the end of the monitoring period, and the average of these two values were used in the models as potential explanatory variables (Table 1). The first measurement of the exposed mineral soil with the greatest variation (0 – 100 %) was used in the statistical models as explanatory variable. The cover of slash was about five percent in the beginning of the study period and decreased somewhat during the experiment (Table 1), hence we used the first measurement of slash in the statistical models. In addition, soil type and thickness of humus layer were measured after cutting and site preparation on smaller circular plots (radius 0.56 m) adjacent to the square 1 m2 ground and field vegetation monitoring plots. Soil type was denoted as by the proportions of coarseness fractions (%) of coarse sandy (2.00 – 0.20 mm), fine sandy (0.20 – 0.06 mm) and silty (less than0.06 mm). The proportions of the soil fractions were computed from the total mass of the sieved soil samples weighed in the laboratory. Soil samples for soil type determination were taken using soil corer. Humus thickness was measured from the same samples cored for soil type determination. To better control the effect of variation in the environmental con- ditions between years (e.g. warm or cold summers or large/small seed crops), the start of the experiment was sequenced over three years. The experiment was started in 2004 in one experimental stand in each of the districts, second experimental stand in each district was managed in 2005 and third one in 2006. The seedlings in the 20 m2 circular sample plots were inventoried annually for the first five years and with a three- year interval after that. All the experimental stands were inventoried seven times making the length of the whole monitoring period 11 years. 2.3. Statistical evaluation The explanatory variables tested in the statistical models (in addition to the treatment effects) are listed in Table 1. The number of all the seedlings in the 20 m2 circular sample plot in each inventory was tested in the height model as a potential predictor (indicating competition between tree seedlings). The number of seedlings on the 20 m2 circular sample plots was modelled using generalized linear mixed models with the assumption of a negative binomial distribution. Four-level hierarchy (the levels nested within the others) and repeated time-dependent measurements (t) were used in the model from highest to lowest hierarchy levels: district, experimental stand, square 70x70m treatment plot and 20 m2 circular sample plots. The data consists of seven measurements during the 11-year moni- toring period. Even though the time between the inventories (1 – 7) varied an auto-regressive correlation structure (AR(1)) was found to give the best fit for the repeated inventories. A negative binomial dis- tribution was used in modelling the number of seedlings because it fitted best to the over-dispersed seedling count data. The model is: yijklt  negativebinomial…πijklt; varijklt† (1) ln πijklt  ˆ f Xijkl; β  ‡ μi ‡ μij ‡ μijk ‡ μijkl; where y denotes the number of seedlings on a plot. The negative bino- mial distribution assumption is defined with expected value (mean) π and variance (var). Furthermore, ln…π† is a log-link function, and f(.) is a linear function with arguments Xijkl(i.e. fixed predictors, measured from different levels of hierarchy) and β(i.e. fixed parameters). Subscripts i, j, k and l refer to the district, experimental stand, square treatment plot and circular sample plot levels, respectively. μi; μij, μijk; μijkl denote random normally distributed between-district, between-stand, between- square treatment plot and between circular sample plot effects with the mean of 0 and constant variances. The residual in the repeated mea- surement (t, denoting inventory in time) is not presented in the formula 1, but the measurements were assumed to be correlated in time (AR(1)). The negative binomial variance (var) can be defined as var ˆ π ‡ …1 ‡ π k†, where k refers to clumping parameter (theta). If k ˆ 1, the distri- bution represents geometric distribution, and if k -> ∞, the distribution approaches Poisson distribution. Natural seedling recruitment might be spatially aggregated. This would tend to indicate over-dispersion, where the variance is larger than the mean. The fit of the negative binomial model to the data was estimated by simulating the distribution of ex- pected counts and compare the values with those of observed counts. Estimated parameters of theta (size) describing the shape of the distri- bution and mu describing the estimated mean of the distribution were used in the simulation. Furthermore, a model using normal distribution assumption for the average height of the Scots pine seedlings on a sample plot was computed. The average was based on all seedlings with h  10 cm on a circular sample plot that had emerged at any point during the moni- toring period. The response variable (seedling height) was log- transformed to normalize the distribution. The model can be described as: ln yijklt  ˆ f Xijk; β  ‡ μi ‡ μij ‡ μijk ‡ μijkl ‡ εijklt; (2) where yijklt is log-transformed response variable, seedling height in dis- trict i, experiment j, square plot k, circular plot l and inventory t. The other terms are similar to those described in formula 1, but εijklt denotes normally distributed error variance with AR(1) correlation structure. To find out the evenness of the regeneration we analysed what is the probability of a given 20 m2 circular sample plot to have at least 1, 2, 3 or 4 seedlings (equal to 500, 1000, 1500 or 2000 seedlings /ha) by computing ordered cumulative mixed effects logistic model. The model is just an extension of a binary logistic model. The different intercept levels are computed for the ordinal categories k-1 and the coefficients of the independent variables remains the same for all the categories of response. The regression coefficients across the levels of response vari- able have to be parallel. This was tested using R package brand (Schlegel and Steenbergen, 2020). The coefficient of soil treatment was parallel across the response categories (χ2 ˆ 3.25, df ˆ 3, p ˆ 0.350). The logit link function was used in the model and the random effects (variances) were estimated for all the nested hierarchy levels except the lowest level (circular sample plot). The R package MASS and its function glmmPQL (Venables and P. Rautio et al. Forest Ecology and Management 539 (2023) 120996 5 Ripley, 2002) was used in the modelling of count data, together with the R package glmmADMB and its function glmmadmb (Fournier et al., 2012). Package glmmADMB was used in the estimation of theta parameter because glmmPQL could not estimate the parameter, but the parameter estimate was needed in the glmmPQL negative binomial model. glmmADMB has a feature that it is not able to estimate an autoregressive correlation (AR(1)) structure (needed for the repeated measurements), but glmmPQL can do this. R package nlme was used in the computation of the general linear mixed model (Pinheiro et al., 2018). The coefficients of determination (R2) were computed for the negative binomial model and the linear model using R package MuMIn (Barton, 2017). The fixed effects predictions with their 95 %’s confi- dence intervals were calculated and plotted using R package effects (Fox, 2003). The ordinal regression was computed using R-package ordinal (Christensen, 2019). All the analyses were made in the R sta- tistical environment (R Core Team, 2016). Both the general linear model for the seedling height and the nega- tive binomial model for the number of seedlings predicted the observed values rather well. Both models slightly underestimated the mean values and predicted better the highest values of the response variable distri- bution than the lowest values (Appendix: Fit of the models and Table A3). 3. Results 3.1. Number of seedlings The proportion of Scots pines of all seedlings was 99.4 %. The number of Scots pine seedlings increased throughout the 11-year monitoring period (Table 2; Fig. 2a). Site preparation was associated with a greater increase. The number of seedlings was around 10 000 ha 1 at the end of the period when the site preparation was done and 2 000–4 000 ha 1 in corresponding treatments without site preparation (50 and 150 stems ha 1). The overstory density of 250 ha 1 had on average substantially lower number of seedlings at the end of the period compared to 50 and 150 ha 1 (Fig. 2a). On the control plots the number of seedlings increased just slightly during the period. The estimates shown in Fig. 2a were computed using the mean value of 2.7 cm for the thickness of humus layer. The thickness of humus layer affected the number of seedlings differently in the treatments. Thicker humus layer was associated with a significantly smaller number of seedlings in the control and in the tree density of 50 trees ha 1 without soil treatment (Fig. 2b). However, humus layer thickness did not have a substantial effect in treatments with site preparation, or with a greater overstory density. A greater proportion of exposed mineral soil resulted in a greater number of seedlings. If the mineral soil was covered with vegetation or bare humus, the predicted number of seedlings was 1111 seedlings ha 1, but if the mineral soil was completely exposed, the predicted number was about five-fold, 5081 seedlings ha 1. If the proportion of exposed mineral soil was 10% (that is about the mean observed in the 20 m2 plots), the corresponding number of seedlings was 1204 seedlings ha 1. A higher coverage of slash reduced the number of seedlings as well. With no slash (0% coverage), the predicted number of seedlings was 1364 seedlings ha 1, the corresponding value at the mean slash coverage (5.3%) was 1278, and with 100% coverage the predicted seedling number decreased to 401 seedlings ha 1. A greater coverage of mosses reduced the number of seedlings sub- stantially. With the minimum coverage of 3%, the predicted seedling number was 2334 ha 1, decreasing to 1487 with 50% and 921 seedlings Table 2 Parameter estimates and tests of generalized linear mixed model (negative binomial) for the number of seedlings. Std. err. denotes the standard error of the estimates, t- and chi-squared values are the test values for the parameter estimates or type III Anova (deviance) tests, df denotes the degrees of freedom and cl the confidence intervals. R2 (trigamma) for the marginal model was 45.2 % and that for the conditional model 61.9 %. For all fixed effects presenting a categorical variable also tests for the other treatment categories vs. a reference category (given in parenthesis) are presented. Variable Coefficient Std. err. df t / chi-squared p Fixed effects Intercept 2.916 0.583 2154 5.000 < 0.001 Treatment (ref. Control) – – 5 40.204 < 0.001 250 non-scarified 2.567 0.639 55 4.021 < 0.000 150 non-scarified 2.547 0.654 55 3.896 < 0.001 50 non-scarified 1.352 0.663 55 2.038 0.064 150 scarified 3.197 0.649 55 4.925 < 0.001 50 scarified 3.658 0.659 55 5.552 < 0.001 Number of inventory 0.126 0.041 2154 3.066 0.002 Thickness of humus layer, mm 0.074 0.019 279 3.989 < 0.000 Cover of exposed mineral soil, % 0.015 0.006 279 2.616 0.009 Cover of slash, % 0.012 0.006 279 2.072 0.039 Cover of moss species, % 0.010 0.004 279 2.714 0.007 Treatment * Number of inventory 5 94.832 < 0.001 250 non-scarified 0.173 0.059 2154 2.925 0.004 150 non-scarified 0.252 0.059 2154 4.294 < 0.001 50 non-scarified 0.267 0.059 2154 4.518 < 0.001 150 scarified 0.448 0.058 2154 7.702 < 0.001 50 scarified 0.482 0.056 2154 8.315 < 0.001 Treatment * Thickness of humus layer – – 5 20.409 0.001 250 non-scarified 0.059 0.023 279 2.555 0.011 150 non-scarified 0.057 0.024 279 2.405 0.017 50 non-scarified 0.015 0.023 279 0.628 0.530 150 scarified 0.063 0.022 279 2.867 0.005 50 scarified 0.077 0.022 279 3.549 0.001 Random effects, phi (AR(1)) and theta Value Theta (k) 0.357 AR(1), phi 0.826 District (variance) 0.205 Experimental stand (variance) 0.275 Square treatment plot (variance) 5.971e-3 Circular sample plot (variance) 5.859e-8 P. Rautio et al. Forest Ecology and Management 539 (2023) 120996 6 ha 1 with 100% coverage, respectively. Many of the sample plots did not have any seedlings, and the number of the empty sample plots varied a lot between the treatments. In the intact control most (70%) of the sample plots were empty and also in the treatment 250 trees ha 1 (76.7%). In the other treatments without site preparation the proportion of empty plots was 56.7% with 150 ha 1 and 51.7% with 50 ha 1. With site preparation the proportion of empty plots was much lower: 26.7% with 150 trees ha 1 and 16.7% with 50 trees ha 1. Another way of looking at the effect of site preparations is to calculate how much mineral soil surface would need to be exposed to Fig. 2. Model predictions of the seedling density (number of Scots pine seedlings ha 1) according to a) number of inventory and b) thickness of the humus layer. Point estimates and 95 % confidence intervals are presented. Numbers 50, 150 and 250 signify stand density ha 1. NT ˆ no soil treatment, T ˆ soil is treated (disc trenching). P. Rautio et al. Forest Ecology and Management 539 (2023) 120996 7 achieve a certain probability of getting, as an example, one seedling per 20 m2 circular sample plot corresponding to 500 seedlings ha 1, or four seedlings per sample plot that corresponds a target density of 2000 seedlings ha 1 according to the Finnish Best practices (Aijala et al., 2019). According to this approach treating the soil surface by disc trenching so that mineral soil is exposed in 25% of the area yielded at least 1 seedling on a sample plot (corresponding to 500 seedlings ha 1) with the probability of nearly 90% (Fig. 3). To reach the same 90% probability to have at least 4 seedlings in a sample plot (corresponding to 2000 seedlings ha 1) would require around 44% mineral soil expo- sition (Fig. 3, for formal statistical test see Appendix: Table A4). 3.2. Seedling height Seedling height increased constantly throughout the monitoring period. The increase was greatest with 50 overstory trees ha 1, clearly smaller with 150 and ha 1, and very small or non-existent in 250 trees ha 1 and in the control (Fig. 4b). Site preparation increased seedling height with 50 ha 1 but not with 150 ha 1. A greater number of seed- lings was associated with a greater mean height (Table 3, Fig. 4a). Increasing number of seedlings in the sample plots increased the average height of the pine seedlings on the sample plots at the mean age of the seedlings (about 11 years, Table 3). 4. Discussion On average, the number of seedlings increased throughout the study period (Fig. 2), which shows that new seedlings emerged even over 10 years after the thinning (and site preparation). The role of years of abundant seed crops in Northern Finland has often been emphasized in silvicultural research and practice (Henttonen et al., 1986; Hilli et al., 2008). The present results with steadily increasing seedling number through the study period rather support the viewpoint that gradual and constant seedling emergence is possible on favorable seedbeds as some viable seeds are produced in most years (Heikinheimo, 1932, 1937) and seedling number tends to increase constantly in time (e.g. Hypponen et al., 2005). The single most influential factor affecting seedling establishment was the removal of ground layer vegetation and humus layer by site preparation (disc trenching). In the overstory densities of 50 or 150 trees ha 1 where site preparation was applied or not applied, seedling density was over two times higher in treatments with site preparation than without it. This confirms many earlier experiments in which the decisive role of site preparation has been shown (Karlsson and Orlander, 2000; Hypponen et al., 2005; Nilsson et al., 2002; Hypponen et al., 2013; Hallikainen et al., 2019; Sikstrom et al., 2020; Kyro et al., 2022). Site preparation has been found to be especially beneficial if the mineral soil is covered with a thick humus layer and a thick moss cover. Accordingly, we found that both the thickness of humus layer and cover of mosses decreased the number of seedlings. Winsa (1995) has shown that continuous supply of capillary water from mineral soil is very important for the germination of Scots pine, because the species germinates slowly at the low temperatures of northern Europe and the radicle cannot tolerate to be dried out. The humus layer blocks the seeds’ contact with capillary water. Further, the humus layer dries out quickly and seedlings without root contact with the mineral soil easily die out during drought periods (Oleskog and Sahlen, 2000). According to Hertz (1934) and Yli- Vaakkuri (1961) moss and lichen surfaces are poor for emergence because they can completely trap and evaporate meagre rainfall. Oinonen (1956) found that a uniform and thick moss surface is a particularly poor germination medium when there is raw humus un- derneath. In this study, the humus layer was not particularly thick (less than 3 cm on average) but moss coverage was high at around 70%. Also Steijlen et al. (1995) have found, in both forest and laboratory, that pine seedlings experience significantly higher mortality on moss cover (Pleurozium schreberi) compared to lichen cover (Cladina spp.). Although more seeds are produced with an increasing number of seed trees and more seedlings have been found to emerge with increasing density of seed trees (Beland et al., 2000), greater overstory density may also limit their germination and the survival of the emerged seedlings. A denser overstory reduces the amount of light and rainfall achieving the ground level (Lehto, 1969; Nygren, 2003). This may explain why regeneration was more abundant with lower than higher densities (cf. Valkonen, 1992). Furthermore, more efforts are needed to find out the effects of different winter conditions on small seedlings and saplings in the vicinity of mature trees. The number of the established seedlings per hectare alone is not a sufficient indicator for regeneration success. After regeneration there Fig. 3. The prediction of the ordered cumulative mixed effects logistic model for the probability that a given sample plot has 1, 2, 3 or 4 seedlings (that equals to 500, 1000, 1500 and 2000 seedlings ha 1) in the function of pro- portion (%) of exposed mineral soil. Age of seedlings, years H ei gh t o f s ee dl in gs , c m 15 20 25 30 35 40 4 6 8 10 12 14 4 6 8 10 12 14 4 6 8 10 12 14 15 20 25 30 35 40 Control 250 NT 150 NT 50 NT 150 T 50 T Number of seedling on circle plots H ei gh t o f s ee dl in gs , c m 30 40 50 60 0 50 100 150 200 a) b) Fig. 4. The predictions of the seedling height model (average height of Scots pine seedlings on a circular sample plot) according to a) number of seedlings on sample plots and b) age of seedlings. Point estimates and 95 % confidence in- tervals are presented. Numbers 50, 150 and 250 signify stand density ha 1. NT ˆ no soil treatment, T ˆ soil is treated (disc trenching). P. Rautio et al. Forest Ecology and Management 539 (2023) 120996 8 should be enough seedlings evenly distributed in the stand area. This means that in the stand there are no large gaps without seedlings. In this study this was best achieved in the treatment with 50 trees ha 1 com- bined with site preparation, where about 80% of the sample plots had at least one seedling. In contrast, 70–75% of the sample plots in the control and 250 trees ha 1 treatments were empty, which is a poor result. The treatments with 50 and 150 stems ha 1 without site preparation were only slightly better with about half of the plots empty. Another way to look at the effect of site preparation is to estimate the probability of a single sample plot to be regenerated as a function of the proportion of mineral soil exposed by the soil treatment. According to our results, exposing mineral soil in 25% of the stand area yields a 90% probability that sample plots has at least one seedling. This, however, is not suffi- cient level of regeneration, as the Finnish Forest Act (Forest Act, 2014) requires that when using natural regeneration in Northern Finland at least 1200 main crop seedlings ha 1 needs to be present no later than 20–25 years after the regeneration cutting to secure the legally mandated forest regeneration. In our data this would require an average of 2.4 seedlings on 20 m 2 circular sample plots that would be achieved on around 90% probability by exposing around 30–35% of the mineral soil (Fig. 3). Achieving 2000 seedlings ha 1, that is recommended minimum density for pines when using planting according to the Finnish best practices for forest management (Aijala et al., 2019), would need about 40% exposition of mineral soil. On the other hand, in natural regeneration and direct seeding the recommended density is 4000–5000 seedlings ha 1 (Aijala et al., 2019). However, the fact that exposing mineral soil dramatically improves germination probability indicates that, by optimizing the spatial distribution of exposed mineral soil patches, better regeneration could be achieved with less soil distur- bance. This could be achieved by, for example, intermittently created planting spots using mounding or soil inversion (Sikstrom et al., 2020) or, in the near future, by automated machines that optimise the location of exposed soil patches with the help of environmental sensing, machine vision and artificial intelligence (Rautio et al., 2023). By decreasing the soil disturbance, one could better secure many other ecosystem services, like recreation, berry picking and reindeer herding (Rautio et al., 2023). Another way to avoid excess site preparation, to secure other ecosystem services, would be supplementary planting in places that have not re- generated naturally (in other words, using assisted regeneration). This would naturally increase the costs but if the natural regeneration has in general been satisfactory, the need for planting would be limited and potentially recompensed by the other ecosystem services. Between measurements some seedlings died, and new ones grew of emerging seedlings that were less than 10 cm tall in the previous mea- surement. Consequently, the seedling mean height tended to develop slower than that of the dominant (tallest) seedlings. The consequences of such dynamics were strikingly expressed in the control treatment. The average height of seedlings at age 14 was basically at the same level as it was in the beginning at age 4 (Fig. 4b). In other words, even though it seems that the number of the seedlings doesn’t change much in the control plots there are constantly seedlings dying and new ones estab- lished. However, these two processes seem to balance out each other so that when the older, and larger, seedlings died they were replaced by new shorter ones in the population. Kyro et al. (2022) observed by monitoring the establishment and survival of individual pine seedlings for 11 years in some of the study stands used here, that the predicted probability of mortality approaches zero after five years while the cu- mulative number of seedlings steadily increases. Accordingly, Niemisto et al. (1993) found that in stands where the seed trees were retained for a long time the average seedling age increased rather slow in for the same reason, i.e. older ones are dying and new seedlings emerging all the time. Thinning and site preparation had a clear positive effect on the height development of seedlings. On the heavily thinned and soil- prepared plots, the seedlings at age 6 were about the same height as the 14-year-old seedlings in the control and in the 250 trees ha 1 treatment. In fact, the difference to the untreated control was even higher than this, as in all the treatment plots except the control plots all seedlings exceeding 50 cm height were removed before starting the monitoring. This finding is consistent with previous studies, as in northern Finland where forests are sparse, canopy growth has been found to limit the growth of seedlings (Hypponen and Hyvonen, 2000). Root competition has also been found to affect seedling growth, even more than lack of light (Aaltonen, 1919; Hagner, 1962; Ackzell and Table 3 Parameter estimates and tests of the general linear mixed model (normally distributed) for the height development of seedlings. Age of the seedlings that were included in the model varied from 4 to 15 years. Std. err. denotes the standard error of the estimates, t- and chi-squared values are the test values for the parameter estimates or type III Anova (deviance) tests, df denotes the degrees of freedom. R2 for marginal model was 17.8 % and that for conditional model 32.05 %. For all fixed effects presenting a categorical variable also tests for the other treatment categories vs. a reference category (given in parenthesis) are presented. Variable Coefficient Std. err. df t / chi-squared p Fixed effects Intercept 2.637 0.267 750 9.858 < 0.001 Treatment (ref. Control) – – 5 22.528 < 0.001 250 non-scarified (NS) 0.301 0.134 54 2.244 0.029 150 non-scarified (NS) 0.554 0.134 54 4.132 < 0.001 50 non-scarified (NS) 0.308 0.129 54 2.393 0.020 150 scarified (S) 0.493 0.124 54 3.978 < 0.001 50 scarified (S) 0.402 0.121 54 3.325 0.002 Age of seedlings, years 0.075 0.032 750 2.349 0.019 Sqrt age of seedlings, years 0.420 0.180 750 2.330 0.020 Number of seedlings on sample plot 0.004 0.001 750 6.509 < 0.001 Treatment * Age of seedlings, years – – 5 35.306 < 0.001 250 non-scarified (NS) 0.014 0.011 750 1.364 0.173 150 non-scarified (NS) 0.041 0.011 750 3.791 0.000 50 non-scarified (NS) 0.034 0.010 750 3.341 0.001 150 scarified (S) 0.031 0.010 750 3.040 0.002 50 scarified (S) 0.052 0.010 750 5.403 0.000 Random effects and autoregressive correlation Value AR(1), phi 0.633 District (variance) 3.617e-3 Experimental stand (variance) 2.674e 10 Square treatment plot (variance) 0.020 Circlular sample plot (variance) 1.009e 9 Residual 0.112 P. Rautio et al. Forest Ecology and Management 539 (2023) 120996 9 Lindgren, 1992; Strand et al., 2006; Hallikainen et al., 2007). Site preparation generally improves the growth of seedlings (e.g. Hypponen and Kemppe, 2001; Varmola et al., 2004; Hypponen et al., 2005; Hypponen et al., 2008), but our results show that site preparation needs to be combined with heavy thinning to have an effect on growth. When the stands were thinned to 150 trees ha 1 the site preparation did not result in any additional growth in the seedlings, whereas in 50 trees ha 1 the effect of site preparation yielded on average 10 cm additional growth at the age of 14 years (Fig. 4b). In conclusion, our results confirmed the importance of competition by mature trees as well as field and ground layer vegetation on natural regeneration and seedling growth in northern pine forests. The number of naturally regenerated seedlings and seedling growth increased with decreasing tree density and increased mineral soil exposure. According to our results, the seed supply is enough to achieve successful natural regeneration even if the number of seed trees is low, as long as proper exposure of mineral soil is carried out. Further studies are needed on what is the most appropriate site preparation technique to achieve the optimal pattern and degree of mineral soil exposure for establishing the required number and growth of seedlings, but also their optimal spatial coverage. CRediT authorship contribution statement Pasi Rautio: Conceptualization, Methodology, Formal analysis, Validation, Investigation, Project administration, Writing – original draft. Ville Hallikainen: Conceptualization, Methodology, Formal analysis, Software, Validation, Investigation, Data curation, Writing – review & editing. Sauli Valkonen: Investigation, Formal analysis, Writing – review & editing. Johanna Karjalainen: Methodology, Formal analysis, Validation, Investigation, Data curation, Writing – re- view & editing. Pasi Puttonen: Investigation, Writing – review & editing. Urban Bergsten: Formal analysis, Investigation, Writing – re- view & editing. Hans Winsa: Investigation, Writing – review & editing. Mikko Hypponen: Conceptualization, Methodology, Investigation, Writing – review & editing. Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Data availability Data will be made available on request. Acknowledgements We would like to thank Metsahallitus for its contribution at all stages of the experiment. Pekka Valikangas, Pasi Aatsinki, Kuisma Ranta, Jari Hietanen, Raimo Pikkupeura, Tarmo Aalto, Pekka Narhi, Eero Siivola, Aarno Niva, Jouni Vaisanen and Jukka Lahti in Luke and Arto Hiltunen in Sveaskog, among others, carried out the field measurements, deserving our sincerest gratitude for their excellent work. We also address our special acknowledgement to Merja Uutela who processed the data proficiently for statistical analysis. This study was carried out as a part of the projects Forward (Forest renewal by natural methods) and Transform (Tools for natural regeneration in sustainable forest man- agement) funded by Natural Resources Institute Finland (Luke), and as a part of the project NorFor (Solving problems in forest regeneration in northern Fennoscandia) funded by Finnish Forest Research Institute (Metla), Swedish University of Agricultural Sciences (SLU), Metsahallitus and Sveaskog, and as a part of ArcticHubs Horizon 2020 -project (Grant Agreement No 869580). Appendix A Material and methods Study area Table A1 Table A2 Fit of the models The number of seedlings predicted by the model was somewhat lower on the average (mean values) compared to the observed numbers, if the predictions were calculated using only the fixed effects of the model (Table A3). The pseudo R2-values of the seedling number model were satisfactory, but the R2-values of different distributions could not be compared to each other. If the fit of the seedling density model was evaluated using the simulated negative binomial distribution using the model parameters (theta and mu), the model fit was good, although the proportion of the zero counts was slightly underestimated and the pro- portion of the highest values underestimated (Fig. A.1). The model for the height development predicted on the average (mean) slightly less than 1 cm lower heights, and the lowest predicted values were a bit higher and the highest predicted values quite much lower than the observed values. However, the predicted values at 3rd quartiles were rather close to the observed values, similar to the 1st quartiles and medians (Table A3). The R2 value was rather low, sug- gesting that other natural conditions than the measured ones were affecting the height development. By leaving the sqrt-term out of the height model would have improved slightly in the highest values (by about two centimetres) but the fit of the model would have been significantly poorer (using AIC and difference in likelihood ratio tests) compared to the presented model. The sqrt-term was selected after testing the power from 0.1 to 2 (step 0.1). The residuals of the height model were pretty unbiased, although the residual variation was great. Table A1 Coordinates (X and Y in WGS84) of the experimental stands (location number refers to the numbers in the Fig. 1) and the year of establishment. District Location Established WGS84 X WGS84 Y Northern 112 2004 27.4403057 68.4833221 Lapland 122 2005 28.1187229 68.7366409 132 2006 27.3981457 68.4689407 Western 212 2004 25.7364845 67.0729370 Lapland 222 2005 23.9756012 67.6385422 232 2006 28.4430008 67.5883484 Eastern 312 2004 23.5592079 67.8579483 Lapland 322 2005 29.3234501 67.2219772 332 2006 29.2637272 67.8377457 Southern 412 2004 25.3679123 66.0192719 Lapland 422 2005 28.2749653 65.9586716 432 2006 27.1641579 65.9127502 Table A2 Proportion of exposed mineral soil of the ground cover (in %) in different treatments. Numbers 50, 150 and 250 signify stand density / ha. NT ˆ no soil treatment, T ˆ soil is treated (disc trenching). Treatment Mean Median Min Max Control 0 0 0 0 250 NT 0.83 0 0 30 150 NT 0.67 0 0 40 50 NT 0.42 0 0 25 150 T 23.80 19.00 0 65 50 T 29.35 22.50 0 100 P. Rautio et al. Forest Ecology and Management 539 (2023) 120996 10 Results See Table A4 References Aaltonen, V.T., 1919. Kangasmetsien luonnollisesta uudistumisesta Suomen Lapissa, I. (Natural regeneration in Lapland I, in Finnish with abridged version in German). Communicationes ex Instituto Quaestionum Forestalium Finlandiae. 1, 1–319. Ackzell, L., Lindgren, D., 1992. Seed-tree stand: Threat or protection for artificial regeneration, in: Hagner, M., (Ed.), Proceedings of an InterNordic Workshop: Silvicultural alternatives, Umeå, Sweden, June 22-25 1992. Swedish University of Agricultural Sciences, Department of Silviculture, Reports, 35, pp. 86 - 95. Aijala, O., Koistinen, A., Sved, J., Vanhatalo, K., Vaisanen, P., (eds.) 2019. Metsanhoidon suositukset. [In Finnish: Best practices for forest management]. Tapion julkaisuja, Helsinki. 252 p. ISBN 978-952-5632-75-0. (In Finnish). Barton, K., 2017. MuMIn: Multi-Model Inference. R package version 1 (40). https:// CRAN.R-project.org/packageˆMuMIn. Bassett, O.D., White, G., 2001. Review of the impact of retained overwood trees on stand productivity. Australian Forestry 64 (1), 57–63. Beland, M., Agestam, E., Eko, M., Gemmel, P., Nilsson, U., 2000. Scarification and seedfall affects natural regeneration of Scots pine under two shelterwood densities and a clear-cut in southern Sweden. Scand. J. For. Res. 15, 247–255. Brown, R.T., Mikola, P., 1974. The influence of fruticose soil lichens upon the mycorrhizae and seedling growth of forest trees. Acta Forestalia Fennica. 141, 23 p. Christensen, R. H. B., 2019. ordinal - Regression Models for Ordinal Data. R package version 2019.12-10, https://CRAN.R-project.org/packageˆordinal. Dovciak, M., Reich, P.B., Frelich, L.E., 2003. Seed rain, safe sites, competing vegetation, and soil resources spatially structure white pine regeneration and recruitment. Can. J. For. Res. 33 (10), 1892–1904. Forest Act. 2014. Forest Act 1093/1996 with amendments up to 567/2014 included. https://www.finlex.fi/fi/laki/kaannokset/1996/en19961093_20140567.pdf. Fournier, D.A., Skaug, H.J., Ancheta, J., Ianelli, J., Magnusson, A., Maunder, M., Nielsen, A., Sibert, J., 2012. AD Model Builder: using automatic differentiation for statistical inference of highly parameterized complex nonlinear models. Optim. Methods Softw. 27, 233–249. Fox, J., 2003. Effect Displays in R for Generalized Linear Models. J. Stat. Softw. 8 (15), 1–27. http://www.jstatsoft.org/v08/i15/. Hagner, S., 1962. Naturlig foryngring under skarm. En analys av foryngringsmetoden, dess mojligheter och begransningar i mellannorrlandskt skogsbruk. (In Swedish with English summary: Natural regeneration under shelterwood stands. An analysis of the method of regeneration, its potentialities and limitations in forest management in middle North Sweden). Meddelanden från Statens Skogsforskningsinstitut. 52 (4), 1–263. Hallikainen, V., Hypponen, M., Hyvonen, J., Niemela, J., 2007. Establishment and height development of harvested and naturally regenerated Scots pine near the timberline in North-East Finnish Lapland. Silva Fennica. 41 (1), 71–88. Hallikainen, V., Hokka, H., Hypponen, M., Rautio, P., Valkonen, S., 2019. Natural regeneration after gap cutting in Scots pine stands in northern Finland. Scand. J. For. Res. 34 (2), 115–125. Heikinheimo, O., 1932. Metsapuiden siementamiskyvysta I. (Of the seeding capacity of forest trees I, in Finnish with German summary). Communicationes Instituti Forestalis Fenniae. 17 (3), 1–55. Heikinheimo, O., 1937. Metsapuiden siementamiskyvysta II. (Of the seeding capacity of forest trees II, in Finnish with German summary). Communicationes Instituti Forestalis Fenniae. 24 (4), 1–67. http://urn.fi/URN:NBN:fi-metla-201207171056. Heinsdorf, M., 1994. Kiefernnatürverjüngung - ein historischer Abriss [Natural regeneration of Scots pine - a historical outline]. Beitrage für Forstwirtschaft und Landschaftsokologie 28 (2), 62–65. In German. Heiskanen, J., Makitalo, K., Hyvonen, J., 2007. Long-term influence of site preparation on water-retention characteristics of forest soil in Finnish Lapland. For. Ecol. Manage. 241, 127–133. Henttonen, H., Kanninen, M., Nygren, M., Ojansuu, R., 1986. The maturation of Pinus sylvestris seeds in relation to temperature climate in Northern Finland. Scand. J. For. Res. 1, 243–249. Hertz, M., 1934. Tutkimuksia kasvualustan merkityksesta mannyn uudistumiselle Etela- Suomen kangasmailla. (In Finnish with German summary). Communicationes Insitituti Forestalis Fenniae. 20, 1–98. Hilli, A., Hokkanen, T., Hyvonen, J., Sutinen, M.L., 2008. Long-term variation in Scots pine seed crop size and quality in northern Finland. Scand. J. For. Res. 23, 395–403. Huth, F., Wehnert, A., Wagner, S., 2022. Natural regeneration of Scots pine requires the application of silvicultural treatments such as overstorey density regulation and soil preparation. Forests 13, 817. https://doi.org/10.3390/f13060817. Hypponen, M., 2002. Mannyn luontainen uudistaminen siemenpuumenetelmalla Lapissa. (In Finnish with English summary: Natural regeneration of Scots pine using the seed tree method in Finnish Lapland). Metsantutkimuslaitoksen tiedonantoja. 844, 1–69. Hypponen, M., Hyvonen, J., 2000. Ylispuustoisten mantytaimikoiden syntyhistoria, rakenne ja alkukehitys Lapin yksityismetsissa. (In Finnish). Metsatieteen aikakauskirja. 4 (2000), 589–602. Hypponen, M., Kemppe, T., 2001. Maanmuokkauksen ja kylvon vaikutus mantysiemenpuualan taimettumiseen Etela-Lapissa. (In Finnish). Metsatieteen aikakauskirja. 1 (2002), 19–27. Hypponen, M., Alenius, V., Valkonen, S., 2005. Models for the establishment and height development of naturally regenerated Pinus sylvestris in Finnish Lapland. Scand. J. For. Res. 20, 347–357. Hypponen, M., Heikkinen, H., Hallikainen, V., 2008. Maanmuokkauksen ja kylvon vaikutus mantysiemenpuualan taimettumiseen ja taimikon alkukehitykseen Etela- Lapissa. (In Finnish). Metsatieteen aikakauskirja. 4 (2008), 269–279. Table A3 Observed and predicted distributions described by the base statistics for the models of the numbers of seedlings (negative binomial model, Table 2) and height development (normally distributed linear model, Table 3). The values in the number of seedlings model were computed for a 20 m2 sample plot. The correspondence to a hectare needs multiplication by 500. Marginal model pre- dictions denote that the predictions are computed using only the fixed part of the model. Minimum 1st quartile Median Mean 3rd quartile Max Model for the number of Scots pine seedlings (trigamma R2, marginal model ˆ 45.2 %) Marginal model (only fixed effects) 0.09 1.14 2.43 5.09 5.09 91 Observed 0.00 0.00 1.00 6.44 5.00 168 Model for the height development (R2, marginal model ˆ 19.2 %) Marginal model 15.90 21.64 24.14 24.76 27.11 42.56 Observed 10.00 18.26 23.50 25.76 30.12 102.06 Fig. A1. The observed (vertical bars) and simulated (red line) distributions of the negative binomial model for seedling count (seedling density). (For inter- pretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) Table A4 The parameter estimates for the cumulative ordinal logistic mixed effects model for the probability that a given sample plot has at least 1, 2, 3 or 4 seedlings (that equals to 500, 1000, 1500 and 2000 seedlings ha 1) in the function of soil treatment (% of exposed mineral soil). Std. error denotes standard error. Nagelkerke’s pseudo R2 ˆ 11.09. Model effects Coefficient Std. error z-value p-value Fixed effects Intercept >ˆ4 0.395 0.399 0.991 – Intercept >ˆ3 0.732 0.405 1.807 – Intercept >ˆ2 1.443 0.425 3.395 – Intercept >ˆ1 2.078 0.447 4.644 – Soil treatment 0.094 0.024 3.904 0.000 Random effects Variance of district 5.836e-12 Experimental stand 5.094e-01 Square treatment plot 2.106e ‡ 00 P. Rautio et al. Forest Ecology and Management 539 (2023) 120996 11 Hypponen, M., Hallikainen, V., Niemela, J., Rautio, P., 2013. The contradictory role of understory vegetation on the success of Scots pine regeneration. Silva Fennica. 47 (1), 1–19. http://www.silvafennica.fi/pdf/article903.pdf. Jalkanen, R., 2003. Havupuutaimikoiden tuhojen esiintyminen ja merkittavyys Suomessa. (In Finnish). Metsatieteen aikakauskirja. 1 (2003), 59–68. Jalkanen, R., 2007. Diseases, pests and abiotic disorders of trees in the changing environment in Finnish Lapland, in: Taulavuori, E. & Taulavuori, K. (eds.). Physiology of Northern Plants under Changing Environment. Research Signpost. p. 105-133. Karlsson, C., Orlander, G., 2000. Soil scarification shortly before a rich seed fall improves seedling establishment in seed-tree stands of Pinus sylvestris. Scand. J. For. Res. 15, 256–266. Kyro, M.J., Hallikainen, V., Valkonen, S., Hypponen, M., Puttonen, P., Bergsten, U., Winsa, H., Rautio, P., 2022. Effects of overstory tree density, site preparation, and ground vegetation on natural Scots pine seedling emergence and survival in northern boreal pine forests. Can. J. For. Res. 52, 860–869. Kytoviita, M.-M., Stark, S., 2009. No allelopathic effect of the dominant forest-floor lichen Cladonia stellaris on pine seedlings. Funct. Ecol. 23, 435–441. Lahde, E., 1978. Effect of soil treatment on physical properties of the soil and on development of Scots pine and Norway spruce seedlings. Communications Instituti Forestalis Fenniae. 94, 1–59. Lehto, J., 1956. Tutkimuksia mannyn luontaisesta uudistumisesta Etela-Suomen kangasmailla. Acta Forestalia Fennica. 66 (2), 1–106. In Finnish with English summary Studies on the natural reproduction of Scots pine on the upland soils of southern Finland. Lehto, J., 1969. Tutkimuksia mannyn uudistamisesta Pohjois-Suomessa siemenpuu- ja suojuspuumenetelmalla. Communicationes Instituti Forestalis Fenniae. 67 (4), 1–140. In Finnish with English summary: Studies conducted in northern Finland on the regeneration of Scots pine by means of the seed tree and shelterwood methods. Lust, N., 1987. An analysis of a spontaneous ingrowth of deciduous trees in 70 year old stands of Scots pine. Silva Gandavensis 52 (1), 1–28. Lust, N., Geudens, G., 1998. Silviculture of Scots pine in Belgium. Silva Gandavensis 63. https://doi.org/10.21825/sg.v63i0.837. Niemisto, P., Lappalainen, E., Isomaki, A., 1993. Mantysiemenpuuston kasvu ja taimikon kehitys pitkitetyn luontaisen uudistamisvaiheen aikana. (In Finnish with English summary: Growth of Scots pine seed bearers and the development of seedlings during a protracted regeneration period). Folia Forestalia. 826, 1–26. Nilsson, U., Albrektson, A., 1994. Growth and self-thinning in two young Scots pine stands planted at different initial densities. For. Ecol. Manage. 68, 209–215. Nilsson, U., Gemmel, P., Johansson, U., Karlsson, M., Torkel, W., 2002. Natural regeneration of Norway spruce, Scots pine and birch under Norway spruce shelterwoods of varying densities on a mesic-dry site in southern Sweden. For. Ecol. Manage. 161 (1–3), 133–145. Nilsson, M.-C., Zackrisson, O., 1992. Inhibition of Scots pine seedling establishment by Empetrum hermaphroditum. J. Chemical Ecol. 18 (10), 1857–1870. Nygren, M., 2003. Metsapuiden siemenopas. Metsantutkimuslaitoksen tiedonantoja. 882, 1–144. Handbook to seeds of forest trees, In Finnish. Nygren, M., Saarinen, M., 2001. Itaminen ja taimettumisalusta, in: Valkonen, S., Ruuska, J., Kolstrom, T., Kubin, E., Saarinen, M., (Eds.). Onnistunut metsanuudistaminen. (In Finnish). Metsantutkimuslaitos, Metsalehti Kustannus. Karisto Oy, Hameenlinna. 83- 90. Oinonen, E., 1956. Mannikoiden luontaisen uudistumisen edellytyksista Lapin kangasmailla eraiden taimivaroja selvittavien inventointien valossa. (In Finnish). Metsataloudellinen Aikakausilehti. 73, 225–230. Oleskog, G., Sahlen, K., 2000. Effects of seedbed substrate on moisture conditions and germination of Pinus sylvestris seeds in a clearcut. Scand. J. For. Res. 15, 225–236. Orlander, G., Gemmel, P., Hunt, J., 1990. Site preparation: A Swedish overview. FRDA report 105. British Columbia, Canada. Peet, R.K., Christensen, N.L., 1987. Competition and tree death. BioScience 37 (8), 586–595. Pinheiro J., Bates D., DebRoy S., Sarkar D., R Core Team (2018). nlme: Linear and Nonlinear Mixed Effects Models. R package version 3.1-137. https://CRAN.R- project.org/packageˆnlme. R Core Team, 2016. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria https://www.R-project.org/. Rautio, P., Lideskog, H., Bergsten, U., Karlberg, M., 2023. Lean Forestry - a paradigm shift from economies of scale to precise and sustainable use of ecosystem services in forests. For. Ecol. Manage. 530, 120766 https://doi.org/10.1016/j. foreco.2022.120766. Ruano, I., Rodríguez-García, E., Bravo, F., 2013. Effects of pre-commercial thinning on growth and reproduction in post-fire regeneration of Pinus halepensis Mill. Ann. Forest Sci. 70, 357–366. Schlegel, B., Steenbergen, M., 2020. brant: Test for Parallel Regression Assumption. R package version 0.3-0. https://CRAN.R-project.org/packageˆbrant. Sikstrom, U., Hjelm, K., Holt Hanssen, K., Saksa, T., Wallertz, K., 2020. Influence of mechanical site preparation on regeneration success of planted conifers in clearcuts in Fennoscandia - a review. Silva Fennica. 54 (2), 10172. https://doi.org/10.14214/ sf.10172. Steijlen, I., Nilsson, M.-C., Zackrisson, O., 1995. Seed regeneration of Scots pine in boreal forest stand dominated by lichen and feather moss. Can. J. For. Res. 25, 713–723. Strand, M., Lofvenius, M.O., Bergsten, U., Lundmark, T., Rosvall, O., 2006. Height growth of planted conifer seedlings in relation to solar radiation and position in Scots pine shelterwood. For. Ecol. Manage. 224 (3), 258–265. Tishler, M., Tullus, T., Tullus, A., Jaarats, A., Lutter, R., Lundmark, T., Tullus, H., 2020. Effects of shelterwood method and plant stock type on the early growth and survival of pine seedlings in regeneration stands under hemiboreal conditions. Scand. J. For. Res. 35 (1–2), 85–95. https://doi.org/10.1080/02827581.2019.1707273. Valkonen, S., 1992. Metsien uudistaminen korkeilla alueilla Pohjois-Suomessa. (In Finnish with English summary: Forest regeneration at high altitudes in Northern Finland). Folia Forestalia. 791, 1–84. Varmola, M., Hypponen, M., Makitalo, K., Mikkola, K., Timonen, M., 2004. Forest management and regeneration success in protection forests near the timberline in Finnish Lapland. Scand. J. For. Res. 19 (5), 424–441. Venables, W.N., Ripley, B.D., 2002. Modern Applied Statistics with S, 4th Ed. Springer, New York, p. 498 p.. ISBN 0-387-95457-0. Wiedemann, E., 1925: Die praktischen Erfolge des Kieferndauerwaldes [The practical implications of the Dauerwald management regime with Scots pine]. Verlag von Friedr. Vieweg & Sohn, Braunschweig. 184 p. In German. Winsa, H., 1995. Influence of rain shelter and site preparation on seedling emergence of Pinus sylvestris L. after direct seedling. Scand. J. For. Res. 10, 167–175. https://doi. org/10.1080/02827589509382881. Yli-Vaakkuri, P., 1961. Kokeellisia tutkimuksia taimien syntymisesta ja ensikehityksesta kuusikoissa ja mannikoissa. (In Finnish with English summary: Experimental studies on the emergence and initial development of tree seedlings in spruce and pine stands). Acta Forestalia Fennica. 75(1), 1-122. P. Rautio et al.