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): Janne Miettinen, Ville Hallikainen, Mikko Hyppönen, Urban Bergsten, Hans 
Winsa, Pekka Välikangas, Arto Hiltunen, Pasi Aatsinki & Pasi Rautio 
Title: Effects of site preparation and reindeer grazing on the early-stage success of Scots 
pine regeneration from seeds in northern Finland and Sweden
 
Year: 2022 
Version: Published version 
Copyright:   The Authors 2022 
Rights: CC BY-NC-ND 4.0 
Rights url: http://creativecommons.org/licenses/by-nc-nd/4.0/ 
 
Please cite the original version: 
Janne Miettinen, Ville Hallikainen, Mikko Hyppönen, Urban Bergsten, Hans Winsa, Pekka 
Välikangas, Arto Hiltunen, Pasi Aatsinki & Pasi Rautio (2022): Effects of site preparation and 
reindeer grazing on the early-stage success of Scots pine regeneration from seeds in northern 
Finland and Sweden, Scandinavian Journal of Forest Research, 
DOI:10.1080/02827581.2022.2136404 
 
Full Terms & Conditions of access and use can be found at
https://www.tandfonline.com/action/journalInformation?journalCode=sfor20
Scandinavian Journal of Forest Research
ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/sfor20
Effects of site preparation and reindeer grazing on
the early-stage success of Scots pine regeneration
from seeds in northern Finland and Sweden
Janne Miettinen, Ville Hallikainen, Mikko Hyppönen, Urban Bergsten, Hans
Winsa, Pekka Välikangas, Arto Hiltunen, Pasi Aatsinki & Pasi Rautio
To cite this article: Janne Miettinen, Ville Hallikainen, Mikko Hyppönen, Urban Bergsten,
Hans Winsa, Pekka Välikangas, Arto Hiltunen, Pasi Aatsinki & Pasi Rautio (2022): Effects of
site preparation and reindeer grazing on the early-stage success of Scots pine regeneration
from seeds in northern Finland and Sweden, Scandinavian Journal of Forest Research, DOI:
10.1080/02827581.2022.2136404
To link to this article:  https://doi.org/10.1080/02827581.2022.2136404
© 2022 Natural Resources Institute Finland.
Published by Informa UK Limited, trading as
Taylor & Francis Group
Published online: 31 Oct 2022.
Submit your article to this journal Article views: 105
View related articles View Crossmark data
Effects of site preparation and reindeer grazing on the early-stage success of Scots
pine regeneration from seeds in northern Finland and Sweden
Janne Miettinena, Ville Hallikainenb, Mikko Hyppönenb, Urban Bergstenc, Hans Winsab, Pekka Välikangasb,
Arto Hiltunenb, Pasi Aatsinkib and Pasi Rautiob
aNatural Resources Institute Finland, Oulu, Finland; bNatural Resources Institute Finland, Rovaniemi, Finland; cDepartment of Forest Biomaterials
and Technology, Swedish University of Agricultural Sciences, Umeå, Sweden
ABSTRACT
The importance of sufficient soil scarification to ensure the regeneration of Scots pine on sub-dry and
more fertile sites has been emphasized in many studies. Here we aimed to study, how site preparation
intensity affects the early success of natural regeneration and sowing (bare seeds and seed pellets) of
Scots pine with or without the reindeer grazing. The study area was located in northern Finland and
Sweden where five site preparation methods were compared: unprepared control, logging machine
tracks, Huminmix (mixing the mineral soil and organic layer), disc trenching and intensive disc
trenching. In each of these we used direct seeding, seed pellets and natural regeneration. Results
revealed that even the lightest site preparation methods can provide sufficient regeneration
results while the reindeer grazing limits the optimal regeneration result. Huminmix and even the
track of the logging machine could provide satisfactory regeneration results both in direct seeding
and natural regeneration. This could facilitate the coexistence of forest management, reindeer
herding and other land use forms in the same stands and area. The use of seed pellets needs
further research, but it may have potential due to lower consumption of seeds and less need for
site preparation.
ARTICLE HISTORY
Received 18 August 2022
Accepted 11 October 2022
KEYWORDS
Huminmix; disc trenching;
sowing; natural
regeneration; artificial
regeneration
Introduction
Many studies emphasize the importance of sufficient soil
scarification to ensure the regeneration of Scots pine (Pinus
sylvestris) on sub-dry and more fertile sites (e.g. Hyppönen
2002; Hyppönen et al. 2005). Natural regeneration and
sowing aim to approximately 4000–5000 seedlings per
hectare, which has been suggested as a successful regener-
ation result in northern boreal forests (Äijälä et al. 2019). In
natural regeneration of pine, 50–100 seed trees per hectare
are normally left uncut to ensure the successful regeneration
(Äijälä et al. 2019). About 3000 seedlings per hectare can be
kept as a minimum limit for growing high-quality timber in
seeded stands (Varmola 1996). The number of seedlings
needs to be higher in stands regenerated naturally or by
sowing compared to planted ones (about 2000 per hectare)
due to the clustered spacing of the seedlings especially in
natural regeneration.
In addition to ensuring regeneration, soil scarification is
needed to make the regeneration operations easier,
promote the survival of seeds and seedlings, and enhance
the growth of young seedlings. Scarification reduces the com-
petition from the ground layer vegetation, increases the
average soil temperature, increases the availability of nutri-
ents for the seedlings, reduces the soil compaction and
damage to seeds and seedlings (Örlander et al. 1996). Most
evident is the reduction of damages in the case of pine
weevil (Hylobius abietis, e.g. Leather et al. 1999; Örlander
and Nilsson 1999; Heiskanen and Viiri 2005). However, scarifi-
cation in general disturbs the soil more than needed to
improve survival of the planted seedlings (Hyppönen and
Hallikainen 2011; Hallikainen et al. 2019) as the needed
amount of exposed soil surface could be as low as below
5% (e.g. 0.25 m2 patch for 2000 seedlings sums up to
500 m2) but in practical forestry it is normally at least 20%
or 30%. On the whole, the increased intensity within the
whole regeneration chain does not provide any significant
surplus for the further development of the stand (see e.g.
Hallsby et al. 2015). Furthermore, the measures should be
done according to forestry goals (Karlsson 2013; Karlsson
et al. 2015; Ahnlund Ulvcrona et al. 2017; Nuutinen et al.
2021), so the most suitable management schedules for each
stand could be selected. Additionally, the logging type is
known to have an effect on the potential seed and seedling
damage. These have been found to be more pronounced in
shelterwood forests than in clear-cuts (Nystrand and Gran-
ström 2000).
In the present and tomorrow’s forestry lower costs can be
reached by reducing the intensity of regeneration tasks by
doing only what is necessary to reach the goals. The needs
of other land use forms also need to be taken better into
© 2022 Natural Resources Institute Finland. Published by Informa UK Limited, trading as Taylor & Francis Group
This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License (http://creativecommons.org/licenses/by-nc-nd/4.
0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited, and is not altered, transformed, or built upon in any
way.
CONTACT Janne Miettinen janne.miettinen@luke.fi Natural Resources Institute Finland, Paavo Havaksen tie 3 Oulun yliopisto, Oulu 90014, Finland
SCANDINAVIAN JOURNAL OF FOREST RESEARCH
https://doi.org/10.1080/02827581.2022.2136404
consideration. For example reindeer (Rangifer tarandus taran-
dus) husbandry, outdoor recreation and tourism have a ten-
dency to restrict soil scarification (e.g. Hallikainen et al.
2006, 2010). Reindeer husbandry and forestry have several
types of interactions. In addition to the effects of forestry
on reindeer husbandry, some effects are inverse: reindeer
grazing has been found to have a negative influence on the
establishment of pine seedlings, as a whole (e.g. Helle and
Moilanen 1993). Reindeer obviously affect the regeneration
of Scots pine throughout the reindeer herding area. Due to
the negative influence of reindeer grazing on the regener-
ation, to study the plain effects of site preparation on the
regeneration success would need the exclusion of grazing
in a regeneration area. Comparing the results of regeneration
using different site preparation methods in grazed and un-
grazed sites would also reveal a possible interaction
between site preparation and grazing. This would mean
that grazing may not have a similar influence on regeneration
for all site preparation methods of different intensity levels.
In contrast, also some negative effects of site preparation
on reindeer foraging have been noticed. Intensive scarifica-
tion may reduce reindeer foraging, and their digging in the
snow, with its negative effects on seedling establishment
(Roturier and Bergsten 2006). Intensive site preparation may
affect reindeer abduction in the vicinity of an intensively
scarified patch.
Even when there are no foreseen land use conflicts with
other land use forms it is rational to minimize the intensity
of scarification whenever it is possible, to minimize the
costs of site preparation and to provide proper grounds for
the other land use forms. To be able to direct the site prep-
aration efforts to those places where the preparation is
needed, we need to know the success of regeneration
without preparation, in the case of light preparation and for
when using conventional site preparation methods.
The economic costs of site preparation form a remarkable
part of forest regeneration. In addition, the costs have
increased steadily in Finland. The increase in the costs of
disc trenching, mounding, patch scarification and ploughing
was over 50% in private, non-industrial forests between 2002
and 2014 (Natural Resources Institute Finland 2022). Due to
increasing costs, there is increasing pressure to find suitable
solutions which provide sufficient regeneration results in an
economically sustainable manner. It is also shown that
towards the north and towards less fertile sites the profitabil-
ity of silvicultural investments reduces (Ahtikoski and Hökkä
2020). All this underlines that there is an emerging and
urgent need for cost-effective and environmental-friendly
regeneration methods especially in the northern parts of
Fennoscandia.
The aim of the study was to find out how site preparation
intensity affects the early success of natural regeneration and
sowing of Scots pine. The latter was studied by using two
seed types, bare seeds and seed pellets (which consist of a
1 cm thick 5 cm radius peat disk with 1 seed in the middle,
Wennström 2014). Both natural regeneration and direct
seeding were studied on grazed and non-grazed treatment
plots in northern Finland and Sweden. Natural regeneration
and the growth of naturally regenerated Scots pine seedlings
were also monitored to study the effects of various site prep-
aration methods and reindeer grazing on the early success of
Scots pine in the natural regeneration. The study questions
were: (1) can we improve the regeneration success and
growth using light site preparation methods to avoid
conflicts with other land use modes, (2) does seed pellet
improve regeneration success, (3) is regeneration improved
if we exclude reindeer by using exclosures, (4) is natural
regeneration using the seed tree method enough for success-
ful regeneration or is direct seeding needed?
Material and methods
The study area is located in northern Finland and Sweden
(Figure 1) about 240–300 meters above sea level. The climatic
conditions in the area are quite harsh. The monthly average
temperatures in the closest available meteorological station
(Kittilä) vary from −12.1°C in January to 14.5°C in July and
the average temperature sum is about 800–850 dd (Finnish
Meteorological Institute 2022).
The study was based on a completely randomized block
design. The blocks are represented by randomly selected
experimental forest stands. The stands were selected from a
set of preselected stands on sub-dry sites that were accepted
after visiting the sites to check that they filled the require-
ments. In total there were six blocks (i.e. experimental
stands), three in Finnish Lapland and three in Norrbotten in
Sweden close to the border of Finland and Sweden (Figure
1). Finnish Lapland and Swedish Norrbotten are both reindeer
grazing areas, where reindeer graze freely throughout the
year. Ten rectangular treatment plots 30 m × 15 m in size
were randomly positioned inside each block. Half of the
plots were fenced to prevent reindeer grazing (Figure 1). In
each rectangular treatment plot there were five 10 m2 circular
sample plots to monitor natural regeneration. In addition, 40
seeding points, half of which were sowed using direct
seeding (bare seeds, micro-preparation using five seeds in
each) and half using seed pellets, were positioned inside
each treatment plot (Figure 1). Seed pellets were simply
placed on the ground (i.e. they were not submerged as in
some earlier studies, e.g. Wennström 2014). For direct
seeding the seeds were covered by 0.5–1 cm of mineral soil,
which protects the seeds from drying, frost and predation
(Kinnunen 1992). When comparing the success of bare
seeds and seed pellets, one should keep in mind the
different number of seeds in the pellets (1) and direct
seeding (5).
On the 10 rectangular treatment plots five site preparation
methods were randomly allocated: (1) control (i.e. no site
preparation), (2) driving around with a logging machine, (3)
Huminmix, (4) disc trenching and (5) double disc trenching.
In the logging machine treatment only a small proportion
of the mineral soil is exposed. Huminmix is a light method
in which the humus and mineral soil are mixed. As a result,
it creates a narrow track with a mixture of the humus layer
and mineral soil of about 10–20% of the total area (Roturier
and Bergsten 2006; Roturier et al. 2011). Disc trenching is a
widely used method in Finland and Sweden. It makes
60–80 cm wide tracks, which are about 1.8–2 meters apart
2 J. MIETTINEN ET AL.
from each other. As a result, 30–50% of mineral soil is exposed
(Karlsson and Örlander 2000). Double disc trenching is an
intensive treatment with≥ 50% of mineral soil exposed
(more detailed characteristics of scarification methods are
presented in Appendix 1 in Figure A1.1).
Half of the 30 m × 15 m treatment plots (five in every
block = experimental stand) were fenced to study the
effects of reindeer grazing. The undisturbed vegetation
cover and the cover of exposedmineral soil or humus and dis-
turbed vegetation cover were inventoried for the seeding
points (with a radius of 3 cm around the seeding point) and
in the 10 m2 circular sample plots in 2013. The number and
the height of the seedlings at the seeding points and on
the 10 m2 sample plots were inventoried in 2015, 2016 and
2019. The first and the last inventory year (2015 and 2019)
were used in the statistical modeling of the results.
When modeling the natural regeneration, a generalized
linear mixed model with negative binomial distribution
assumption and log-link function was constructed to model
the number of seedlings on the 10 m2 sample plots (seedling
density model). The model consists of three hierarchical
levels: block, rectangular treatment plot nested within block
and circular sample plot nested within the treatment plot.
The model was computed using R-package glmmTMB.
In the R function glmmTMB the variance was defined in
two ways: the NB1 variance = m(1+ a) and NB2
variance = m 1+ m
u
( )
(Brooks et al. 2017). The NB1-parame-
trization suggested a linear mean-variance relationship,
while NB2-parametrization suggested a quadratic relation-
ship. The different parametrizations led to slightly different
model fit, parameter estimates and tests. The NB1 model pro-
duced a slightly better model fit based on the residuals, but
the best fit based on the AIC-values was obtained using
NB2-parametrization. However, the parametrization did not
considerably affect the interpretation of the results and,
hence, NB2-parametrization was used in the final presented
model.
R-package glmmTMB allowed also the zero-inflation mod-
eling, but the zero-inflation was not an obvious problem in
the data tested by R-package DHRMa (Hartig 2021). The
null hypothesis of the suitable number of observed zeroes
compared to the simulated ones tested using function test-
ZeroInflation remained (p = 0.584). The simulated distribution
using the model parameters size andmuwere plotted against
the observed count distribution to determine the overall
model fit. The parameter size was computed using the predic-
tions of the marginal model using the formula
size = mu
2
var −mu, where mu denotes expected mean and var
the variance of the predicted distribution. The (pseudo) R2-
values for the model was computed in two ways: (1) based
on the likelihood ratio (Bartoń 2020) and (2) using wide-func-
tioning new R-package performance (Lüdecke et al. 2021).
When modeling the direct seeding (5 seeds) and seed
pellets/pellets (1 seed), a generalized linear mixed model
with a binomial distribution assumption and logit-link func-
tion was constructed to model the dichotomous response
Figure 1. The location of the experimental stands ( = blocks) and the study design. Rectangles indicate treatment plots (Control: no site preparation, Hum: humin-
mix, Log: logging machine, Disc tr.: disc trenching, Exposed >50%: double disc trenching). Small circular symbols within a rectangular treatment plot indicate a
spot for seeding, crosses indicate seed pellets and large circular symbols indicate sample plots to monitor natural regeneration. Bolded rectangles indicate the
fenced treatment plots. The size of each treatment plot is 30 m × 15 m and the size of each circular sample plot for natural regeneration is 10 m2.
SCANDINAVIAN JOURNAL OF FOREST RESEARCH 3
for the seeding-point level in 2015 and 2019: 1 = germinated
seed and survived seedling, 0 = non-germinated seed or dead
seedling. For direct seeding the term “established” denoted
that at least one of the five seeds had established and sur-
vived. The model consisted of three hierarchical levels:
block, rectangular treatment plot nested within the block
(level 2) and seeding point nested within rectangular treat-
ment plot. The dispersion parameter of binomial distribution
was estimated (not fixed as 1). The model was computed
using R-package MASS (Venables and Ripley 2002).
The height of the seedlings established naturally in the
10 m2 circular sample plots was modeled by using the log-
transformed height of the seedlings (measured in 2019) as
the response variable in the model. The age of the seedlings
and its possible exponents giving the best fit of the model
were used as the covariate in the model. The model was com-
puted using R-package nlme (Pinheiro et al. 2020).
There were several fixed explanatory variables and their
interactions that were tested in the direct seeding model:
(1) fencing (fenced, non-fenced); (2) site preparation
(control, logging machine, Huminmix, disc trenching,
double disc trenching, i.e. intensive treatment with≥ 50%
of mineral soil exposed); (3) seed type (bare seed. seed
pellet); (4) seed bed (mineral soil, mix of humus and
mineral soil, humus, disturbed vegetation, undisturbed veg-
etation); (5) field layer vegetation (Calluna vulgaris, Vacci-
nium vitis-idaea, Vaccinium uliginosum, Vaccinium myrtillus,
Empetrum nigrum, grasses spp., no vegetation); (6) ground
layer vegetation (in the case of direct seeding: lichens spp.,
Hylocomium splendens/Plurozium schreberi, Dicranum sp./
Polytrichum sp., no vegetation). In the models for seedling
density and height (models for 10 m2 circular sample
plots) seed type and seed bed were not included but
fencing and soil treatment were tested similarly to the
sowing model and in addition the field layer vegetation
(similarly to the sowing models) and the covers of lichens
spp., mosses spp., cover of mineral soil, disturbed and undis-
turbed vegetation.
The grazing effect (fenced or non-fenced) and site prep-
aration method as well as their interaction were the most
important variables in the models because of the study
design. The variables, such as cover of mineral soil, disturbed
or undisturbed vegetation could not be included in the same
model with the site preparation method, because these cov-
ariates were highly influenced by the soil treatment.
However, the cover of the ground- and field layer species
(or species groups) could be tested as the covariates in the
models. The vegetation illustrated the site type, basically
not affected by the soil treatment.
In addition, the relationships between the field layer
species, bilberries (Vaccinium myrtillus), cowberries (Vacci-
nium vitis-idaea), crowberries (Empetrum nigrum) and
heather (Calluna vulgaris) as well as the relationships
between the ground layer species and cover of moss and
lichen species were studied using generalized linear mixed
model with a quasi-Poisson distribution family using the hier-
archy that have been described above in the random part of
the seedling density models. The so-called “quasi-Poisson”
(over-distributed Poisson distribution) could fit better than
a normal distribution to the models for the vegetation
cover data. The relationships were tested using the covers
measured at the beginning of the study (year 2013). They
might have been slightly changed during the six years
period, but the change would have been small at these north-
ern sub-dry sites.
The fixed variables of the factorial random block design
were fencing (i.e. grazed or non-grazed), site preparation
and their interaction. In addition, the site types were con-
sidered using the covariates of the vegetation cover describ-
ing the moist-xeric dimension. Several combinations of the
species covers were tested as the covariates in the seedling
density model, and the most effective variables and their
interactions were selected based on the AIC-values (mini-
mize) and pseudo-R2 values (maximize). The ground-layer
vegetation could be described by using the moss or lichen
cover as covariates in the model. The time (year 2015 and
2019) was taken in the model to see the development in
the seedling density. The interaction effect of the year (categ-
orical, 2015, 2019) was tested with the factors of fencing and
soil treatment, but the interactions were not statistically sig-
nificant at a 5% risk level and, thus, they were left out of
the model. The time effect (years 2015 and 2019) was also
tested in the random part of the model version in addition
to the fixed effect. However, its variance was small, and the
random year effect had no effect on the parameter estimates
or tests. Thus, in the final version it was left out of the random
part of the model and treated only as a fixed variable in the
model. The variables in the final model version were the
year, fencing, soil treatment, lichen and moss cover, heather
cover, and the interaction of fencing and soil treatment.
The (pseudo) coefficients of determination (R2) of the mar-
ginal model (fixed predictors only) were about 32.8%
(based on the LR-change between the null model and the
presented model) or 57.5% (based on R-package
performance).
A ratio correction (observed/predicted mean) for the pre-
dictions of the log-transformed response linear model (seed-
ling height model) and log-link negative binomial model
(seedlings density model) was done (Snowdon 2011). All
the analyses and statistical graphs were done using R statisti-
cal environment (R Core Team 2019).
Results
Seedling densities in natural regeneration
The “heaviest” soil treatment resulted in less naturally regen-
erated seedlings than the other treatments (Figure 2(a)). In
general, the point estimates suggest that the number of seed-
lings decreased with the intensity of the soil treatment, but,
on the other hand, the confidence intervals show a large vari-
ation within the other soil treatment methods except for the
most intensive (Figure 2(a)).
The reindeer grazing (fenced vs. non-fenced treatment
plots) had a minor effect on the seedling establishment
(see also Figure A1.5 in Appendix 1 for more detailed
results). The interaction effect of fencing and soil treatment
was statistically significant at a 5% risk level (Table 1). The
4 J. MIETTINEN ET AL.
number of seedlings in the control (C) was slightly higher in
the fenced plots, though the 95% confidence intervals in
Figure 2(a) suggests that there were no significant differences
between the fenced and non-fenced plots. However, in the
most intensively treated plots there were slightly more seed-
lings in the non-fenced plots compared to the fenced ones
(Figure 2(a)).
The predicted number of seedlings increased during the
four years (2015–2019) from 1115 to 2731. The increase was
quite similar for the different soil treatments and in the
fenced and non-fenced plots (no significant interaction). Fur-
thermore, increasing moss cover was connected to a decreas-
ing number of seedlings, whereas increasing lichen and
heather cover were connected to higher seedling densities
(Table 1, Figure 2(b–d)). It should be emphasized that
although the lichen and moss cover correlated to a certain
degree, both had a highly significant contribution in the
model (Table 1). Furthermore, the VIF-values of these vari-
ables were low suggesting a low multicollinearity (cover of
moss species 1.66, lichen cover 1.37).
Establishment in direct seeding
The establishment probabilities of the seed pellets (contain-
ing one seed) were lower, 0.43, compared to the direct
seeding (with five bare seeds), which achieved establish-
ment probability 0.71. Even though the performance of
the seed pellets was slightly better compared to the
direct seeding in non-treated soils (control) two years
after seeding, six years after seeding (2019) the perform-
ance of seed pellets was poorer compared to bare seeds
(Figure 3(a)). After six years and pellets used, the Huminmix
treatment showed the lowest survival rates. On average the
establishment probability was about 30% lower in 2019
compared to the situation in 2015 (0.72 vs. 0.41). When
using bare seeds the establishment probability was good
for all site preparation methods exposing mineral soil
(Huminmix, disc trenching and intensive treatment).
However, the probability of establishment decreased in
four years (from 2015 to 2019) also for these treatments
(Figure 3(a), Table 2).
Figure 2. (a–d) The predictions and 95% confidence intervals of the seedling density model for naturally regenerated seedlings: the interaction of fencing and soil
treatment (a), moss species cover (b), lichen cover (c) and heather cover (d). In the predictions, the other predictors in the model were set into their mean values or
levels. The abbreviations: C = control (approximately 0—1% of mineral soil exposed), M = logging machine (appr. 1—5% of soil exposed), H = Huminmix (10—
20% of soil exposed), D = disk trenching (30—50% of mineral soil exposed) and I = intensive (> 50% of mineral soil exposed). See Figure A1.1c in Appendix 1 for
the proportions of exposed mineral soil.
SCANDINAVIAN JOURNAL OF FOREST RESEARCH 5
The fenced (non-grazed) sites were slightly better sites for
seedling establishment, the point estimate of probability
being almost 10% higher at the fenced seeding points com-
pared to the non-fenced ones (Figure 3(b)). Furthermore,
the establishment of the sowed seedlings was considerably
better at the seeding points dominated by either lichen or
bare humus or mineral soil (category None in Figure 3(c))
compared to moss-covered seeding points.
Height development of naturally regenerated
seedlings
There were no considerable or statistically significant differ-
ences between the soil treatments or fencing in the height
development of the seedlings (Table 3). However, the seed-
lings which underwent the intensive treatment in fenced
plots were about 10 cm taller compared to the other soil
treatment in fenced or non-fenced plots (Figure 4(a)). The
point estimates suggest that seedlings in the disk trenching
and intensive treatment in the non-fenced plots were slightly
(about 5 cm) taller compared to the other soil treatments in
the non-fenced plots. But, then again, the 95% confidence
intervals suggest that these differences are only indicative.
The naturally regenerated seedlings grew from 10 to
20 cm between 2015 and 2019 (Figure 4(b)). The increasing
cover of moss species (indicating a moister site) from 0 to
100% was reflected in the increased growth of about 5 cm
during the seven years of the experiment (from the beginning
of the experiment to 2019; Figure 4(c)). Increasing of crow-
berry and especially heather cover reduced the growth of
the seedlings (Figure 4(d,e)).
Discussion
The most striking results that we found were the opposite
regeneration results for direct seeding and the results of
natural regeneration using the seed tree method in relation
to site preparation. In direct seeding the best survival of seed-
lings was found when site preparation was used, but in
natural regeneration the best regeneration success was
reached when only the lightest site preparation methods
were used or site preparation was not used at all. In direct
seeding, the survival of seeds was generally good when
using heavier site preparation, most likely because the
seeds were covered by 0.5–1 cm of mineral soil, which prob-
ably provided a shelter against drought, heavy rainfall, frost
and predation (see e.g. Kinnunen 1992). In direct seeding
the site preparation was beneficial for the seedling survival,
but either there the preparation method did not need to be
intensive. In fact, even Huminmix, which is a somewhat light-
weight site preparation method and exposes only about 10%
of the total area, had higher probability of seedling survival
than disc trenching, double disc trenching or logging
machine preparation. Huminmix has shown encouraging
results also in earlier studies in the light of seedling establish-
ment (Erefur et al. 2008), reindeer lichen recovery (Roturier
et al. 2011) or both (Roturier and Bergsten 2006).
The results of direct seeding suggest that all site prep-
aration methods studied here can improve regeneration
success, but the lightest site preparation methods (Huminmix
and even the track of the logging machine) could provide sat-
isfactory regeneration results. This could facilitate the coexis-
tence of forest management, reindeer herding and other land
use forms in the same area. This is supported by the results of
Hyppönen and Hallikainen (2011) who found a relatively high
number of seedlings when only 1–25% of mineral soil was
exposed. In their (Hyppönen and Hallikainen 2011) study
exposing more mineral soil (26–50% and 51%) the increase
in the number of seedings compared to 1–25% mineral soil
exposition was rather small. Seed pellets gave somewhat
mixed results. The proportion of surviving seeds from the
seed pellets was quite low, especially in 2019, in relation to
direct seeding and compared to the results obtained using
seed pellets in Sweden (Wennström 2014). Our findings do
not directly support the use of seed pellets. It is, however,
important to note that the seed pellets contained only one
seed, while in direct seeding the number at all seeding
points was five. Moreover, the performance was better if
the soil was not prepared at all or was prepared only by the
logging machine. Additionally, the difference between 2015
and 2019 is noteworthy. The mortality of seeds in seed
pellets between 2015 and 2019 in Huminmix and in both
disc trenching treatments was drastic and needs further
research.
Table 1. Parameter estimates and tests of the generalized linear mixed effects
model for the density of naturally regenerated seedlings using a negative
binomial distribution assumption (NB2-parametrization).
Variable/Parameter Estimate Std. error
z-value /
chi-value p-value
Fixed effects
(Intercept) 1.129 0.373 3.025 0.002
Fencing (ref. Non-
fenced)
- - 3.372 (1) 0.066
–Fenced 0.554 0.302 1.836 0.066
Soil treatment (ref.
Control)
- - 39.069 (4) < 0.001
–Logging machine 0.280 0.320 0.874 0.382
–Huminmix −0.021 0.316 −0.067 0.947
–Disc trenching −0.698 0.329 −2.122 0.034
–Intensive (exposed
mineral soil > 50%)
−1.961 0.384 −5.113 < 0.001
Year (ref. 2015) - - 82.528 (1) < 0.001
–2019 0.895 0.099 9.084 < 0.001
Cover of mosses, % −0.020 0.003 −6.298 < 0.001
Cover of lichens, % 0.028 0.008 3.557 < 0.001
Cover of heather, % 0.064 0.017 3.813 < 0.001
Fencing * Soil
treatment (ref. Non-
fenced, Control)
- - 10.321 (4) 0.035
–Yes * Logging
machine
−0.516 0.435 −1.187 0.235
–Yes * Huminmix −0.712 0.440 −1.619 0.105
–Yes * Disc trenching −0.252 0.435 −0.580 0.562
–Yes * Intensive −1.609 0.529 −3.044 0.002
Random effects Variance 95% confidence
intervals
Block 0.234 0.063–0.869
Treatment plot
nested within block
0.160 0.078–0.353
Notes: Std. error denotes the standard error of the estimate. R2 = 32.8% (mar-
ginal model, based on the LR-change between the null model and the pre-
sented model) and 57.5% (marginal model based on R package
performance). For all fixed effects presenting a categorical variable tests
for the other treatment categories vs. a reference category (given in parenth-
esis) are also presented.
6 J. MIETTINEN ET AL.
Despite the large mortality with some treatments, the use
of seed pellets could be a good option on some occasions. It
is known that the required number of seeds depends on the
seed quality and site preparation (Wennström et al. 1999,
2002). For example, if the seeds are expensive (high-quality
bred seeds) or there is a shortage of seeds, the use of seed
pellets may reduce the number of seeds needed noticeably
in comparison to automatic sowing. It is also good to note
that the use of seed pellets can provide more regular
spacing of seedlings, which leads to a lower number of seed-
lings needed per hectare. Additionally, even if the number of
seed pellets would be doubled due to lower survival rates, the
needed amount would still only be a fraction of the amount
needed during automatic sowing. Furthermore, the
increasing price of site preparation (Natural Resources Insti-
tute Finland 2022) or seeds may increase the need for
methods in which the consumption of seeds is lower to
limit the total costs of regeneration. Perhaps the next devel-
opment step could be towards “lean” thinking, where the
scarification intensity would counteract with the secondary
supply of seeds or seedlings.
In natural regeneration, the best survival of seedlings was
when no site preparation was carried out or the intensity of
the site preparation was low. In an earlier study by Hallikainen
et al. (2019) successful regeneration was achieved in Finnish
Lapland in most cases with site preparation that exposed
about 20% of the soil surface. Their estimate for the
sufficient amount of exposed mineral soil was 10–20%. Our
Figure 3. (a–c) The predictions and 95% confidence intervals of the logistic sowing model: the interaction effect of year, seed type and site preparation (a), fencing
effect (b) and ground-layer vegetation effect (c). Abbreviations: C = control, M = logging machine, H = Huminmix, D = disc trenching and I = intensive (> 50% of
mineral soil exposed), Hyloc. = Hylocomium splendens, Pleuroz. = Pleurozium schreberi, Dicr. = Dicranum spp. and Polyt. = Polytrichum spp.
SCANDINAVIAN JOURNAL OF FOREST RESEARCH 7
results showed that an even lower proportion of exposed
mineral soil could be sufficient. This suggests even more
careful consideration of the intensity and method of the
needed site preparation. This finding may help to minimize
the management costs, environmental effects and the
adverse effects of site preparation on the other land use forms.
Table 2. Parameter estimates and tests of the generalized linear mixed model for the number of the sowed seedlings using binomial distribution assumption.
Variable/Parameter Estimate Std. error df t- /chi-value p-value
Fixed effects
(Intercept) −0.729 0.219 4722 −3.335 0.001
Year (ref. 2015) - - 1 0.706 0.401
–2019 −0.171 0.205 4722 −0.837 0.402
Seed type (ref. Bare seed) - - 1 13.001 < 0.001
–Seed pellet 0.703 0.195 4722 3.597 0.000
Fencing (ref. Non-fenced) - - 1 8.457 0.004
-Fenced 0.388 0.134 49 2.901 0.006
Soil treatment (ref. Control) - - 4 83.020 < 0.001
–Logging machine 1.323 0.273 49 4.853 0.000
–Huminmix 3.362 0.407 49 8.266 0.000
–Disc trenching 2.364 0.355 49 6.664 0.000
–Intensive 2.269 0.350 49 6.476 0.000
Ground-layer vegetation (ref. Lichens) - - 3 35.706 < 0.001
–Hylocomium / Pleurozium −0.632 0.134 4722 −4.706 0.000
–Dicranum / Polytrichum spp. −0.576 0.185 4722 −3.111 0.002
–No vegetation (None) 0.223 0.201 4722 1.108 0.268
Year * Soil treatment - - 4 24.304 < 0.001
–2019 * Logging machine −0.026 0.277 4722 −0.092 0.927
–2019 * Huminmix −1.458 0.391 4722 −3.727 0.000
–2019 * Disc trenching −0.961 0.319 4722 −3.014 0.003
–2019 * Intensive −0.655 0.317 4722 −2.068 0.039
Year * Seed type - - 1 1.445 0.229
–2019 * Pellet −0.337 0.281 4722 −1.199 0.231
Year * Seed type * Soil treatment - - 8 103.661 < 0.001
–2015 * Pellet * Logging machine −0.560 0.271 4722 −2.063 0.039
–2019 * Pellet * Logging machine −0.647 0.275 4722 −2.355 0.019
–2015 * Pellet Huminmix −2.866 0.380 4722 −7.551 0.000
–2019 * Pellet * Huminmix −1.241 0.290 4722 −4.282 0.000
–2015 * Pellet * Disk trenching −1.639 0.316 4722 −5.190 0.000
–2019 * Pellet * Disk trenching −0.548 0.284 4722 −1.933 0.053
–2015 * Pellet * Intensive −1.752 0.308 4722 −5.683 0.000
–2019 * Pellet * Intensive −1.046 0.286 4722 −3.651 0.000
Random effects Variance 95%’s confidence intervals
Block 9.964e−3 4.458e−5–2.227
Treatment plot nested within block 0.201 0.120–0.337
Dispersion parameter 0.971 0.930–1.013
Notes: Std. error denotes standard error of estimate and df denotes the degrees of freedom. Classification efficiency (area under ROC-curve) of the model was
73.1%. For all fixed effects presenting a categorical variable also tests for the other treatment categories vs. a reference category (given in parenthesis) are
presented.
Table 3. Parameter estimates and tests of the linear mixed effects model for seedling height (log-normal distribution assumption).
Estimate std. error Df t- / chi-value p-value
Fixed effects
(Intercept) 1.772 0.122 603 14.542 0.000
Fencing (ref. Non-fenced) - - 1 1e-4 0.991
–Fenced −0.001 0.103 43 −0.011 0.991
Soil treatment (ref. Control) - - 4 8.467 0.076
–Logging machine 0.092 0.110 43 0.837 0.407
–Huminmix −0.028 0.111 43 −0.251 0.803
–Disc trenching 0.234 0.110 43 2.127 0.039
–Intensive 0.248 0.133 43 1.871 0.068
Age of seedling (years) 0.161 0.013 603 12.425 0.000
Cover of mosses, % 0.003 0.001 115 2.781 0.006
Cover of heather, % −0.018 0.006 115 −3.102 0.002
Cover of crowberry, % −0.009 0.004 115 −2.241 0.027
Fencing * Soil treatment - - 4 9.460 0.051
–Fencing * Logging machine 0.027 0.151 43 0.179 0.859
–Fencing * Huminmix 0.141 0.153 43 0.923 0.361
–Fencing * Disk trenching −0.163 0.152 43 −1.073 0.289
–Fencing * Intensive 0.455 0.210 43 2.164 0.036
Random effects Variance 95% confidence intervals
Block 2.595e−3 8.807e−5–0.076
Treatment plot nested within block 0.016 0.006–0.044
Residual 0.073 0.065–0.082
Notes: Std. error denotes the standard error of estimate and df denotes the degrees of freedom. The R2 of the model = 24.2% (marginal model). For all fixed effects
presenting a categorical variable, tests for the other treatment categories vs. a reference category (given in parenthesis) are also presented.
8 J. MIETTINEN ET AL.
The increasing cover of moss, that indicates increasing site
fertility, was connected to the decreasing number of seed-
lings. This is in accordance with earlier studies (Hallikainen
et al. 2019; Kyrö et al. 2022). On the other hand, in our
results increasing lichen cover as well as the abundance of
heather (both of which indicate low soil fertility) was con-
nected to the higher seedling density (Hallikainen et al.
2019). Thus, in the poorest sites, disc trenching and even
Huminmix may be overly intensive site preparation
methods if natural regeneration is used. This observation
encourages for future studies where new type of accessories
attached in harvesters and forwarders, in order to break the
soil surface, would be tested to see effects in natural regener-
ation. Integrating wood harvesting and site preparation
would significantly reduce the costs of forest management.
The height growth of naturally regenerated seedlings in all
treatments was quite slow since the mean height ranged only
between 15 and 30 cm at the age of 6 years. Then again, this
is in line with earlier findings (Hyppönen et al. 2005; Hallikai-
nen et al. 2007; Hyppönen and Hallikainen 2011). The
Figure 4. The prediction and their 95% confidence intervals of the height model of naturally regenerated seedlings by the interaction effect of fencing and soil
treatment (a), seedling’s age (b), cover of moss species (c), heather (d) and crowberry (e). Abbreviations: C = control, M = logging machine, H = Huminmix, D = disc
trenching and I = intensive (> 50% of mineral soil exposed).
SCANDINAVIAN JOURNAL OF FOREST RESEARCH 9
combination of intensive disc trenching and fencing differed
from other treatments since it led to higher height growth.
This is most likely due to reduced disturbance from reindeer
because the exclusion of reindeer is the only difference from
the same site treatment on the non-fenced side. Our results
also suggest that the potential role of field layer vegetation
may be important, and in some cases complex. Heather
(Calluna Vulgaris) may provide excellent microsites for the
regeneration as we found. On the other hand, it can severely
hinder the growth of pine seedlings as e.g. Norberg et al.
(2001) and Hyppönen et al. (2013) have found. Thus, the
role of heather may be dual, as Hyppönen et al. (2013) have
argued.
Excluding reindeer had a positive effect on the number of
seedlings after sowing. The difference between fenced and
unfenced plots was about 10% on average. The regeneration
success for seeding was rather good in all cases where any
site preparation was used, in both fenced and non-fenced
areas. On the most intensively treated plots fencing did not
have a significant effect on the number of naturally regener-
ated seedlings but it increased the seedling height. On the
most intensively prepared plots, the number of seedlings
was so low both in fenced and non-fenced areas (<500
stems) that they would not provide a successful regeneration
result in any case. Our results hence show that reindeer
grazing influenced the regeneration success, as Helle and
Moilanen (1993) have shown earlier. However, from the view-
point of forestry practice it seems obvious that the site prep-
aration has a more marked effect on the success of forest
regeneration than fencing.
To conclude, our results suggest that both site preparation
and reindeer grazing have effects on the natural regeneration
and direct seeding of Scots pine. In natural regeneration and
direct seeding even lightest site preparation methods, such
as Huminmix or using logging machine tracks, may provide
sufficient regeneration success. The fact that a combination
of intensive site preparation (double disc trenching) and
fencing led to increased height development suggests that
reindeer herding limits the optimal regeneration result in
the northern parts of Sweden and Finland. The use of seed
pellets showed quite low survival rates, but deserves further
research since it may have the potential on some occasions
to reduce the consumption of seeds and the need for site
preparation. These findings can help to avoid overly intensive
site preparation, and this, in turn, can help to reduce the man-
agement costs. On a larger scale, the potential new paths of
forest management practice presented in this study may help
to reduce the confrontation between forestry and other land
use forms.
Acknowledgements
Kuisma Ranta, Jouni Unga, Tarmo Aalto, Pekka Närhi, Jukka Lahti, Reijo
Hautajärvi, Eero Siivola and Esko Jaskari carried out the field measure-
ments, deserving our sincerest gratitude for their excellent work. We
also what to thank Gary Attwood for checking the language of the
text. V.H., P.R., M.H., U.B. and H.W. conceived the idea, designed and
completed the study design and edited draft manuscript. J.M. wrote
the first draft. P.V., A.H. and P.A. contribute to the data collection and
data editing.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Funding
This study was carried out as a part of the projects Forward (Forest
renewal by natural methods) and Transform (Tools for natural regener-
ation in sustainable forest management) funded by Natural Resources
Institute Finland (Luke), as a part of project ArcticHubs (EU funded
H2020, Grant Agreement No 869580) and as a part of the project
NorFor (Solving problems in forest regeneration in northern Fennoscan-
dia) funded by Finnish Forest Research Institute (Metla), Swedish Univer-
sity of Agricultural Sciences (SLU), Metsähallitus and Sveaskog.
References
Ahnlund Ulvcrona K, Bergström D, Bergsten U. 2017. Stand structure after
thinning in 1–2 mwide corridors in young dense stands. Silva Fenn. 51
(3):1563. DOI:10.14214/sf.1563.
Ahtikoski A, Hökkä H. 2020. Intensive forest management — does it pay
off financially on drained peatlands? Can J For Res. 49:1101–1113.
Äijälä O, Koistinen A, Sved J, Vanhatalo K, Väisänen P (toim). 2019.
Metsänhoidon suositukset. Tapion julkaisuja. (In Finnish).
Bartoń K. 2020. MuMIn: Multi-Model Inference. R package version 1.43.17.
https://CRAN.R-project.org/package=MuMIn.
Brooks ME, Kristensen K, van Benthem KJ, Magnusson A, Berg CW, Nielsen
A, Skaug HJ, Maechler M, Bolker BM. 2017. glmmTMB balances speed
and flexibility among packages for zero-inflated generalized linear
mixed modeling. R J. 9(2):378–400.
Erefur C, Bergsten U, de Chantal M. 2008. Establishment of direct seeded
seedlings of Norway spruce and Scots pine: effects of stand con-
ditions, orientation and distance with respect to shelter tree, and fer-
tilization. For Ecol Manag. 255:1186–1195.
Finnish Meteorological Institute. 2022. https://en.ilmatieteenlaitos.fi/
statistics-from-1961-onwards [cited 4 Jan 2022].
Hallikainen V, Hökkä H, Hyppönen 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. DOI: 10.1080/02827581.2018.1557248.
Hallikainen V, Hyppönen M, Hyvönen J, Niemelä J. 2007. Establishment
and height development of harvested and naturally regenerated
Scots pine near the timberline in North-East Finnish Lapland. Silva
Fenn. 41(1):71–88.
Hallikainen V, Hyppönen M, Pernu L, Puoskari J. 2010. Family forest
owners’ opinions about forest management in northern Finland.
Silva Fenn. 44(2):363–384.
Hallikainen V, Jokinen M, Parviainen M, Pernu L, Puoskari J, Rovanperä S,
Seppä J. 2006. Inarilaisten käsityksiä metsätaloudesta ja muusta
luonnonkäytöstä. Metsätieteen Aikakauskirja. 4/2006:453–474. (In
Finnish).
Hallsby G, Ahnlund Ulvcrona K, Karlsson A, Elfving B, Sjögren H, Ulvcrona
T, Bergsten U. 2015. Effects of intensity of forest regeneration
measures on stand development in a nationwide Swedish field exper-
iment. Forestry. 2015(88):441–453. DOI:10.1093/forestry/cpv010.
Hartig F. 2021. DHARMa: Residual Diagnostics for Hierarchical (Multi-
Level / Mixed) Regression Models. Rpackage version 0.4.4. https://
CRAN.R-project.org/package=DHARMa.
Heiskanen J, Viiri H. 2005. Effects of mounding on damage by the
European pine weevil in planted Norway spruce seedlings. North J
Appl For. 22:154–161. DOI:10.1093/njaf/22.3.154.
Helle T, Moilanen H. 1993. The effects of reindeer grazing on Pinus sylves-
tris. Scand J Forest Res. 8:395–407.
Hyppönen M. 2002. Natural regeneration of Scots pine using the seed
tree method in Finnish Lapland. Finnish Forest Research Institute.
Research Papers. 844:69. pp. (In Finnish with English summary).
Hyppönen M, Alenius V, Valkonen S. 2005. Models for the establishment
and height development of naturally regenerated Pinus sylvestris in
Finnish Lapland. Scand J Forest Res. 20:347–357.
10 J. MIETTINEN ET AL.
Hyppönen M, Hallikainen V. 2011. Factors affecting the success of
autumn direct seeding of Pinus sylvestris L. in Finnish Lapland.
Scand J Forest Res. 26:515–529.
Hyppönen M, Hallikainen V, Niemelä J, Rautio P. 2013. The contradictory
role of understory vegetation on the success of Scots pine regener-
ation. Silva Fenn. 47(1):903.
Karlsson C, Örlander G. 2000. Soil scarification shortly before rich seed
fall improves seedling establishment in seed tree stands of Pinus sylves-
tris. Scand J Forest Res. 15:256–266. DOI:10.1080/028275800750015073.
Karlsson L. 2013. Silvicultural regimes and early biomass thinning in
young, dense pine stands. Diss. (sammanfattning/summary) Umeå:
Sveriges lantbruksuniv., Acta Universitatis Agriculturae Sueciae,
1652-6880; 2013:90, ISBN 978-91-576-7918-5, eISBN 978-91-576-
7919-2, [Doktorsavhandling].
Karlsson L, Nyström K, Bergström D, Bergsten U. 2015. Profitability and
stand development by early biomass thinning applied to dense
Scots pine stands. Scand J Forest Res. 30(5):416–428. DOI:10.1080/
02827581.2015.1023351.
Kinnunen K. 1992. Kylvöalustan, ajankohdan ja menetelmän vaikutus
männyn kylvön onnistumiseen. Summary: Effects of substratum,
date and method on the post-sowing survival of Scots pine. Folia
Forestalia 785.
Kyrö MJ, Hallikainen V, Valkonen S, Hyppönen M, Puttonen P, Bergsten U,
Winsa H, Rautio P. 2022. Effects of overstory tree density, site prep-
aration and ground vegetation on natural Scots pine seedling emer-
gence and survival in northern boreal pine forests. Can J For Res.
52:1–10.
Leather SR, Day KR, Salisbury AN. 1999. The biology and ecology of the
large pine weevil, Hylobius abietis (Coleoptera: Curculionidae): a
problem of dispersal? Bull Entomol Res. 89:3–16. DOI:10.1017/
S0007485399000024.
Lüdecke D, Ben-Shachar MS, Patil I, Waggoner P, Makowski D. 2021.
Performance: an R package for assessment, comparison and testing of
statistical models. J Open Source Software. 6(60):3139. DOI:10.21105/
joss.03139.
Natural Resources Institute Finland. 2022. Statistics Database / Forest
statistics / Structure and production / Silvicultural and forest
improvement work / Unit costs of silvicultural and forest improvement
work 2002-2014: http://statdb.luke.fi/PXWeb/pxweb/fi/LUKE/LUKE__
04%20Metsa__02%20Rakenne%20ja%20tuotanto__12%20Metsanhoit
o-%20ja%20metsanparannustyot/93_Metsanhoito_ja_metsparant
_ykskust.px/chart/chartViewLine/?rxid=001bc7da-70f4-47c4-a6c2-c9
100d8b50db [referred 2.2.2022.
Norberg G, Dolling A, Jäderlund A, Nilsson M-C, Zackrisson O. 2001. Control
of heather (Calluna vulgaris (L.) Hull) by steam treatment: effects on
establishment and early growth of Scots pine. New For. 21:187–198.
Nuutinen Y, Miina J, Saksa T, Bergström D, Routa J. 2021. Comparing
the characteristics of boom-corridor and selectively thinned stands
of Scots pine and birch. Silva Fenn. 55(3):10462. DOI:10.14214/sf.10462
Nystrand O, Granström A. 2000. Predation on Pinus sylvestris seeds and
juvenile seedlings in Swedish boreal forest in relation to stand disturb-
ance by logging. J Appl Ecol. 37:449–463.
Örlander G, Egnell G, Albrektson A. 1996. Long-term effects of site prep-
aration on growth in Scots pine. For Ecol Manag. 86:27–37. DOI:10.
1016/S0378-1127(96)03797-8.
Örlander G, Nilsson U. 1999. Effect of reforestation methods on pine
weevil (Hylobius abietis) damage and seedling survival. Scand J
Forest Res. 14:341–354. DOI:10.1080/02827589950152665.
Pinheiro J, Bates D, DebRoy S, Sarkar D, Core Team R. 2020. _nlme: Linear
and Nonlinear Mixed Effects Models_. R package version 3.1-150, URL:
https://CRAN.R-project.org/package=nlme.
R Core Team. 2019. R: A language and environment for statistical comput-
ing. Vienna: R Foundation for Statistical Computing. URL https://www.
R-project.org/.
Roturier S, Bergsten U. 2006. Influence of soil scarification on reindeer
foraging and damage to planted Pinus sylvestris seedlings. Scand J
Forest Res. 21:209–220.
Roturier S, Sundén M, Bergsten U. 2011. Re-establishment rate of reindeer
lichen species following conventional disc trenching and HuMinMix
soil preparation in Pinus-lichen clear-cut stands: a survey study in
northern Sweden. Scandinavian Journal of Forest Research. 26
(2):90–98. DOI: 10.1080/02827581.2010.528019.
Snowdon P. 2011. A ratio estimator for bias correction in logariyhmic
regression. Can J Forest Res. 21(5):720–724.
Varmola M. 1996. Nuorten viljelymänniköiden tuotos ja laatu. Abstract:
Yield and quality of young Scots pine cultivations.
Metsäntutkimuslaitoksen tiedonantoja 585. 70 s. + 6 liitettä.
Venables WN, Ripley BD. 2002. Modern applied statistics with S. 4th ed.
New York: Springer; p. 498. ISBN 0-387-95457-0
Wennström R. 2014. LandPuck™–systemets ekonomiska konkurrenskraft
jämfört med tallplantering i norra Sverige. SLU Examensarbeten
2014:15, Fakulteten för skogsvetenskap, Institutionen för skogens
ekologi och skötsel, English abstract: The LandPuck™-system’s
economic competitiveness compared to planting of pine in northern
Sweden.
Wennström U, Bergsten U, Nilsson J-E. 1999. Mechanized
microsite preparation and direct seeding of Pinus sylvestris in boreal
forests – away to create desired spacing at low cost. NewFor. 18:179–198.
Wennström U, Bergsten U, Nilsson J-E. 2002. Effects of seed weight and
seed type on early seedling growth of Pinus sylvestris under harsh
and optimal conditions. Scand J Forest Res. 17:118–130. DOI:10.
1139/b94-110.
SCANDINAVIAN JOURNAL OF FOREST RESEARCH 11
Appendix 1
Material and methods
Raw data description
The mean cover of moss species on the circle plots was 40.3% and the
mean cover of lichens was 7.5%. The variation in the moss cover was
large compared to the variation in the lichen cover (Figure A1.1a, 1b).
The cover of exposed mineral soil varied highly depending on the soil
scarification (Figure A1.1c). The heather cover was low in general but
could reach the one fifth of the area of the circle plots (Figure A1.1d).
Cowberries were abundant in most of the circle plots, generally covering
from 5 to 10% the other soil treatment than the intensive soil scarification
(Figure A1.1e). The raw distributions suggested that the sites corre-
sponded to the sub-xeric site type, on the average. Although the cover
of moss species is high, being 44% on the control plots (on the
average), while there were also some xeric heather- and lichen-domi-
nated plots in the data.
The mean number of seedlings after the three inventories (2015,
2016 and 2019, inventories together) was 4601 seedlings per ha, the
median value being 1000 seedlings per ha. The number of seedlings
varied a lot (Figure A1.1f) suggesting a highly aggregated seedling
establishment. The development in the number of seedlings accord-
ing to the inventories based on the raw data was the following
(years 2015, 2016 and 2019, means, medians in the parenthesis):
3150 (1000), 4240, 6410 (2000). Thus, the number doubled over the
years.
In the ground- and field-layer vegetation, the species, and the
groups of the species, correlated with each other. The bottom layer
consisted of moss species, lichens or exposed mineral soil or humus.
Increasing cover of moss species reduced the lichen cover in many of
the sample plots (Figure A1.2). The seeds sown via direct seeding (and
using the seed pucks) were seeded into the mineral soil at half of the
seeding points, the proportion of undisturbed vegetation being about
20%.
The relationships between the four dominating species in field-layer
vegetation were weak. The strongest and statistically significant positive
relationship was found between cowberries and crowberries (t = 7.738,
df = 239, p < 0.001). In addition, a weaker negative relationship was
found between bilberries and heather (t = 2.217, df = 239, p < 0.028).
The field- and ground-layer vegetation at the seeding points were domi-
nated by the category “no vegetation”. This denoted that the seeding
points were surrounded by bare humus or exposed mineral soil. The
moss species, especially Hylocomium and Pleurozium dominated the
ground-layer vegetation (Figure A1.3).
Results
Seedling densities in natural regeneration
The performance and fit of the model were fairly good. The mean value
was slightly overestimated: the observed mean being 5.05 and predicted
being 4.80 (in 10 m2 sample plot). Furthermore, the simulated negative
Figure A1.1. Raw distributions of moss species cover (a), lichen cover (b), exposed mineral soil (c), heather cover (d), cowberry cover (e) and the number of
seedlings per ha by the treatments consisting of soil fencing and soil treatment. Abbreviations for the treatments: n denotes non-fenced, and f fenced,
C denotes control, M denotes logging machine, H denotes Huminmix, D denotes disk trenching and I intensive soil scarification (> 50% exposed mineral soil).
The distributions have been computed based on the circle plots (n = 300).
12 J. MIETTINEN ET AL.
Figure A1.3. The field- and ground-layer vegetation at seeding points (radius 3 cm around the point) and the type of the seed bed based on the raw data.
Figure A1.2. The relationship between the cover of moss and lichen species in
the raw data. The relationship was significant at a 5% risk level in a mixed-
effects linear model using a “quasi-Poisson” distribution (t = 5.970, df = 239,
p < 0.001).
Figure A1.4. Observed and simulated probabilities of the number of seedlings
on 10 m2 sample plot using the mu = 5.03 and size = 0.27.
SCANDINAVIAN JOURNAL OF FOREST RESEARCH 13
binomial distribution using the predicted mean and theta-value of 0.27
closely followed the observed probabilities of the counts (Figure A1.4).
The numbers of seedlings in natural regeneration also varied between
fenced and non-fenced study plots (Figure A1.5). In general, these
declined according to a mineral soil exposure percentage increase.
Height development of naturally regenerated seedlings
The fixed predictors of the model explained only 24.2% of the variation.
However, the predicted (16.31 cm) and observed (16.59 cm) means of
height were very close to each other. However, the model could not
predict the tallest seedlings correctly (Table A1.1).
Table A1.1. The observed and predicted distributions of the seedlings’ height
model (cm).
Minimum
1st
quartile Median Mean
3st
quartile Maximum
Observed 10.00 12.00 14.00 16.59 19.00 65.00
Predicted 9.50 14.01 16.05 16.31 18.25 37.74Figure A1.5. Numbers of seedlings according to mineral soil exposure percen-
tage in fenced and non-fenced study plots.
14 J. MIETTINEN ET AL.