2392  |    J Appl Ecol. 2024;61:2392–2404.wileyonlinelibrary.com/journal/jpe
1  |  INTRODUC TION
The presence of intact and broken standing dead trees (snags) 
is a characteristic feature of most forests. Their importance 
for a well- functioning forest ecosystem is well established 
(Thomas, 2002) and, above all, they provide the basis for broad 
networks of deadwood- dependent species, some of which are 
highly specialized on snags (Larsson Ekström et al., 2023; Stokland 
Received: 23 September 2023  | Accepted: 5 May 2024
DOI: 10.1111/1365-2664.14729  
R E S E A R C H  A R T I C L E
Drivers of snag fall rates in Fennoscandian boreal forests
Tuomas Aakala1  |   Ken Olaf Storaunet2 |   Bengt Gunnar Jonsson3 |   Kari T. Korhonen4
1School of Forest Sciences, University of 
Eastern Finland, Joensuu, Finland
2Norwegian Institute of Bioeconomy 
Research, Ås, Norway
3Mid- Sweden University, Sundsvall, 
Sweden
4Natural Resources Institute Finland, 
Joensuu, Finland
Correspondence
Tuomas Aakala
Email: tuomas.aakala@uef.fi
Funding information
Kone Foundation
Handling Editor: Pieter De Frenne
Abstract
1. Persistence of standing dead trees (snags) is an important determinant for their 
role for biodiversity and dead wood associated carbon fluxes. How fast snags fall 
varies widely among species and regions and is further influenced by a variety 
of stand- and tree- level factors. However, our understanding of this variation is 
fragmentary at best, partly due to lack of empirical data.
2. Here, we took advantage of the accruing time series of snag observations in the 
Finnish, Norwegian and Swedish National Forest Inventories that have been fol-
lowed in these programs since the mid- 1990s. We first harmonized observations 
from slightly different inventory protocols and then, using this harmonized data-
set of ca. 43,000 observations that had a consistent 5- year census interval, we 
modelled the probability of snags of the main boreal tree species Pinus sylvestris, 
Picea abies and Betula spp. falling, as a function of tree- and stand- level variables, 
using Bayesian logistic regression modelling.
3. The models were moderately good at predicting snags remaining standing or fall-
ing, with a correct classification rate ranging from 68% to 75% among species.
4. In general, snag persistence increased with tree size and climatic wetness, and 
decreased with temperature sum, advancing stage of decay, site productivity and 
disturbance intensity (mainly harvesting).
5. Synthesis and applications: The effect of harvesting demonstrates that an effi-
cient avenue to increase the amount of snags in managed forests is protecting 
them during silvicultural operations. In the warmer future, negative relationship 
between snag persistence and temperature suggests decreasing the time snags 
remain standing and hence decreasing habitat availability for associated species. 
As decomposition rates generally increase after fall, decreasing snag persistence 
also implies substantially faster release of carbon from dead wood.
K E Y W O R D S
boreal forest, coarse woody debris, Fennoscandia, forest management, standing dead tree
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, 
provided the original work is properly cited.
© 2024 The Author(s). Journal of Applied Ecology published by John Wiley & Sons Ltd on behalf of British Ecological Society.
    |  2393AAKALA et al.
et al., 2012). In forest ecosystems where snags are common, 
their persistence, that is the time they remain standing, can have 
major implications for the whole ecosystem carbon cycling (Hilger 
et al., 2012) and may also play an important role for forest nutrient 
dynamics (Krankina et al., 1999). For both issues—biodiversity and 
carbon and nutrient cycling—snag persistence is a key factor: hab-
itat availability and stability, and how long carbon and nutrients 
keep locked away in dead wood are greatly dependent on how 
quickly snags fall.
To manage snags, we need to understand factors influencing 
their rates of fall. Here, the mechanisms that determine the resis-
tance of live trees to wind and snow (e.g. Peltola et al., 1999), are 
also applicable to understanding snag dynamics. In general, the 
probability of a tree falling depends on (i) the structural strength of 
the stem and on (ii) the external forces exerted on the stem (Oberle 
et al., 2018).
Out of these two, the structural strength and consequently the 
resistance to fall of an individual tree develops during its lifetime. 
While alive, a tree benefits from structural strength that can with-
stand the external forces on the stem, usually from winds and snow. 
This resistance carries over to the standing dead trees. Studies of live 
tree resistance to winds have shown that important variables at the 
tree level include tree species and size, which influence allometry 
and mechanical resistance and physical and chemical properties of 
the wood material (Peltola, 2006). At the level of a forest stand, this 
is influenced by variables such as soil conditions, topographic posi-
tion and stand composition and density (Elie & Ruel, 2005; Suvanto 
et al., 2016). Jointly with the species properties, these influence tree 
allometry, rooting depth and anchorage, canopy surface roughness 
and wind exposure. Higher stand density also dissipates energy from 
the wind to a larger number of individuals (Oberle et al., 2018), but 
stand density may also increase the chance of a tree being knocked 
down by a neighbouring tree (Pelz & Smith, 2013). Trees also accli-
mate to winds during their lifetime (e.g. Larson, 1965), and hence 
changes in the exposure to high winds for instance due to clearcut-
ting adjacent the trees leads to an increased chance of falling over.
These factors at the level of the tree and the forest stand, to-
gether with the changes occurring at tree death and the possible ac-
tivity of saprotrophic fungi while the tree was alive then determine 
the mechanical strength at the time of tree death. Following the 
death of the tree, the joint effect of decomposers and the structural 
strength at death then determine how long the snag remains stand-
ing. As the wood is invaded by decomposers that begin to cleave 
apart cellulose, hemicellulose and lignin, the mechanical strength 
progressively erodes, and hence the external force needed to topple 
the tree is reduced. On the other hand, as foliage and branches of 
gradually increasing order begin to drop, the sail area of the tree 
crown decreases, reducing the turning moment towards the stem or 
the root system, and consequently the mechanical force the tree has 
to withstand. Snag fall occurs latest at the point where the stem can 
no longer support its own mass.
In Fennoscandia, both the long- term human influence on forests 
and the current intensive forest management have reduced snag 
densities to a fraction of their natural levels (Kalliola, 1966; Linder 
& Östlund, 1998; Storaunet & Rolstad, 2002). Although the amount 
of dead wood in general has lately increased, snags remain few, 
averaging less than 4 m3/ha throughout Fennoscandia (Korhonen 
et al., 2020; Nilsson et al., 2020; Storaunet & Rolstad, 2015). 
Consequently, the lack of dead wood of different species, sizes and 
degrees of decomposition is one of the main causes for the decline 
in forest habitat quality, and reasons underlying the decline of pop-
ulations of many forest- dwelling species (Stokland et al., 2012). In 
managed forests, low snag densities implies that they play a minor 
role for carbon storage and fluxes. However, in protected areas the 
amount of dead wood is higher than in the managed forests, and 
their amount has been increasing faster (Korhonen et al., 2020; 
Kyaschenko et al., 2022).
Forest management guidelines and certification standards em-
phasize leaving dead trees during silvicultural operations, but cur-
rent models predicting snag fall lack generality that could be used 
in management planning. Understanding snag dynamics, especially 
understanding factors that increase their persistence could feed into 
management planning aiming at producing long- lived habitats and 
long- term carbon storage. This also influences the assessment of 
the role and potential of snags in forest carbon fluxes, where snags 
are often treated in a simplistic manner (Hilger et al., 2012; Oberle 
et al., 2018; Woodall, Domke, et al., 2012).
An obstacle to analysing factors influencing snag fall and for 
developing predictive models has been the lack of empirical data, 
which is partly explained by the low density of snags in forests and 
the slow nature of snag dynamics. Permanent plots, where snags can 
be tracked through time are superior data source for these types of 
analyses. However, as permanent plots are usually not specifically 
designed for monitoring snag populations, the numbers of observa-
tions have remained low relative to the number of factors influenc-
ing the fall rates.
During recent decades, large- scale forest inventory programs 
have accrued data suitable for such studies in North America 
(Hilger et al., 2012; Oberle et al., 2018) and central Europe (Oettel 
et al., 2023). Similarly, these earlier data limitations are gradually dis-
appearing also from northern Europe despite the low snag densities, 
as Finland, Norway and Sweden (henceforth: Fennoscandia) all in-
corporated dead wood in the permanent plot measurements of their 
National Forest Inventories (NFIs) in the 1990s. The exact protocols 
differ among these countries, but similar basic data is collected in all 
three inventory programs, and many of the variables can be readily 
harmonized (Stokland, 2003).
Using harmonized snag measurements from the Fennoscandian 
NFIs, we developed statistical models for predicting snag fall, and 
analysed the tree- and site- level factors that influence fall rates 
in Fennoscandian boreal forests. The variables were chosen so 
that the models can be used for predicting snag fall from attri-
butes readily available and easily measurable on any snag found 
in Fennoscandian forests. Our analyses focused on the main tree 
species in the region: Scots pine (Pinus sylvestris), Norway spruce 
(Picea abies) and two birch species grouped together (silver 
 13652664, 2024, 10, D
ow
nloaded from
 https://besjournals.onlinelibrary.wiley.com/doi/10.1111/1365-2664.14729 by Luonnonvarakeskus, W
iley Online Library on [21/01/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on W
iley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2394  |    AAKALA et al.
birch Betula pendula and pubescent birch Betula pubescens). For 
other deciduous trees species, although important components 
in Fennoscandian forests, the current data from the NFIs is too 
limited to allow for a similarly detailed analysis as the main tree 
species.
2  |  MATERIAL S AND METHODS
2.1  |  Fennoscandia
The study area consists of the forested areas of Finland, Norway 
and Sweden, stretching from 55° N in southern Sweden, to 69° N 
in northern Norway. Longitudinally, the area is demarcated by 
the Atlantic Ocean at the Norwegian coast at approx. 5° E and 
the eastern border of Finland at approx. 32° E. The area exhibits 
a varied topography, with highest areas in the Scandes with el-
evations between 1000 and 2500 m above sea level, to lowland 
areas in Finland and the southern and coastal areas of Sweden 
and coastal Norway.
The main factor influencing the climate in Fennoscandia is its 
position between the Atlantic Ocean and Eurasia. The Scandes 
Mountains in the west give rise to major differences in precipi-
tation particularly between the western and eastern parts of the 
region. The Norwegian coast is highly maritime, while the eastern 
parts of Fennoscandia are intermediate between maritime and 
continental. However, in all parts of the area at least moderate 
precipitation is recorded throughout the year. Mean temperature 
of the warmest month (July) ranges from 17°C in the southern 
Sweden to 13°C in some areas of the North. The mean tempera-
ture of the coldest month (February) varies from −3.5°C in south-
ern Sweden to −14.7°C in northern Finland (Grieser et al., 2006). 
Annual precipitation sum also varies from over 2000 mm on the 
Norwegian coast to 450–500 mm in the interior of northern 
Lapland (Tikkanen, 2005). In southwestern Norway 11% and in 
northern Sweden 38% of the precipitation falls as snow (Grieser 
et al., 2006).
Forests cover approximately 72%, 35% and 69% of the land 
area in Finland, Norway and Sweden. Most of this area is man-
aged: of productive forest land (annual growth >1.0 m3/ha), 6.1%, 
3.8% and 5.9% is formally protected from harvesting in Finland 
(Suojelualueet, 2022), Norway (Hylen et al., 2022) and Sweden 
(Statistics Sweden, 2021), respectively. Most of the forests are man-
aged as rotations of even- aged stands, with intermediate thinnings 
from below. The main forest- forming tree species in the region are 
Scots pine, Norway spruce and birches. Deciduous co- dominants in-
clude aspen (Populus tremula), and other boreal deciduous species 
such as alders (Alnus incana and Alnus glutinosa), rowan (Sorbus aucu-
paria) and goat willow (Salix caprea). The hemiboreal and temperate 
forests in southern Sweden and Norway can also be dominated by 
nemoral deciduous species, such as oaks (Quercus robur, Quercus pe-
traea), or beech (Fagus sylvatica).
2.2  |  National forest inventories
National forest inventories (henceforth NFIs) are detailed else-
where (Breidenbach et al., 2020; Fridman et al., 2014; Korhonen 
et al., 2021), but in short, data is collected on stand, plot and tree 
levels. The Finnish NFI (maintained by Natural Resources Institute 
Finland, https:// www. luke. fi/ fi/ seura nnat/ valta kunna n- metsi 
en- inven toint i- vmi) is based on a systematic cluster sampling. 
Within- cluster configuration varies throughout the country, with 
the shortest distance between sample plots varying from 250 to 
300 m. Every fourth cluster is permanent and hence applicable 
here. Earlier, the plot remeasurement interval varied, but from 
the 10th NFI 10 (2004–2008) onwards, remeasurement interval 
for each permanent plot was 5 years. The Norwegian NFI (main-
tained by Norwegian Institute of Bioeconomy Research, https:// 
www. nibio. no/ en/ subje cts/ forest/ natio nal- fores t- inven tory) is 
a systematic permanent inventory since the 1990s, where plots 
allocated on a 3 km × 3 km grid are measured every 5 years. The 
Swedish NFI (maintained by Swedish University of Agricultural 
Sciences, https:// www. slu. se/ en/ Colla borat ive- Centr es- and- Proje 
cts/ the- swedi sh- natio nal- fores t- inven tory), plots are clustered 
into quadratic tracts (clusters) with eight circular sample plots, lo-
cated evenly along the tract sides. The length of the tract sides 
varies, from 300 m in the south to 1200 m in the north for the per-
manent tracts. The remeasurement interval varied before 2003, 
and after that every 5 years.
2.3  |  Snags in the NFIs
In the Finnish NFI, snag measurements are made on circular sam-
ple plots with a radius of 7 m, and all snags with diameter at breast 
height (i.e. 1.3 m, DBH) ≥10 cm are included. Here, we used data from 
three NFI cycles with a fixed 5- year measurement interval. As dead 
wood pieces were not explicitly followed through time but needed 
to be matched between inventories, tracking the fate of snags re-
quired developing a ‘piece- matching algorithm’ (Woodall, Walters, 
& Westfall, 2012) to pair observations of snags on the same perma-
nent plot between two inventory cycles. In short (see Data S1), at 
each permanent plot we used tree species, size and decay class to 
match the observations. A snag recorded in an inventory cycle was 
compared with snags measured on the same plot in the next cycle, 
and snag of the same species, approximately the same DBH, and 
the same or more advanced stage of decay was considered a match. 
Snags without a matching tree in the following inventory were con-
sidered fallen.
In the Norwegian NFI, snags are tracked through time, from living 
trees until they fall to the ground. Pairing of observations between 
measurements was, therefore, straightforward. Tree measurements 
(including snags) are made on circular sample plots of 250 m2, where 
all trees with DBH ≥5 cm are included. Tree species, DBH and decay 
class are recorded.
 13652664, 2024, 10, D
ow
nloaded from
 https://besjournals.onlinelibrary.wiley.com/doi/10.1111/1365-2664.14729 by Luonnonvarakeskus, W
iley Online Library on [21/01/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on W
iley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
    |  2395AAKALA et al.
In the Swedish inventory, dead wood is measured on a plot 
with a radius of 10 m. Variables for each dead wood unit (≥10 cm in 
basal diameter and ≥1.3 m in length) are tree species, size (maximum 
basal diameter), decay stage (see below) and position (standing or 
downed). A standing dead tree is defined as being any stump or snag 
higher than 1.3 m and is recorded separately from downed logs orig-
inating from the same tree. Snags are tracked through time.
While permanent plot data is ideal for studying snag dynamics, 
our approach overestimates snag fall for two reasons: snags that fall 
or that were deliberately harvested cannot be separated in the data 
(this information is available only in the Norwegian records). Second, 
snags with erroneous field records (e.g. wrong species, more recent 
decay class in later inventory cycle, etc.) would be considered fallen 
in the Finnish data matched here (Data S1). On the other hand, the 
5- year remeasurement interval may also overestimate the standing 
times, as trees that died standing and fell during the remeasurement 
interval are not captured. Hence only a part of this ‘fastest’ fraction 
of snags leaves any record, whereas slower fraction of snags that 
remains standing for >5 years is fully recorded.
We visualized the geographical pattern of proportion of snag fall, 
by calculating the proportion of snags that were observed to fall in 
cells of a 10 km × 10 km grid. We then calculated a smooth surface (a 
grid of 5 km × 5 km cells) over the entire area, using inverse- distance- 
weighted interpolation (with a maximum of 100 closest points and 
parameter 0.5).
2.4  |  Factors influencing snag fall and how they are 
derived from the NFIs
Separate models were built for Scots pine, Norway spruce, and the 
two birch species combined. The two birch species were combined, 
as they are not separated in the Swedish data. For the other species, 
their numbers were too low to allow analyses with similar level of 
detail, with 1210 observations of aspen, 6112 observations of boreal 
deciduous species (including alders, rowan, goat willow) and 1558 of 
nemoral deciduous species (such as oaks and beech).
Predictor variables were chosen as variables that were found im-
portant in earlier studies and that are readily available or measured. 
Tree- level predictor variables included DBH and decay class. Of 
these tree- level predictors, DBH was similarly recorded in all three 
NFI protocols. However, decay classifications were all slightly dif-
ferent. We thus used the harmonization developed by Aakala and 
Heikkinen (2024) that uses wood density in decay classes to group 
similar decay classes into a new, harmonized classification that has 
three decay classes for snags: early, intermediate and advanced. Of 
these, early decay class includes snags that are recently dead and/
or only modestly influenced by decomposers. Intermediate snags 
are characterized by somewhat decayed trees in which a knife can 
be pressed through the outer parts of the stem. In the advanced 
stages, snags are becoming increasingly decomposed, until the point 
where they can barely maintain their own stem upright. Of tree- level 
variables, height (and stem taper) influences probabilities of trees 
falling (Peltola, 2006), but because it has not been systematically 
recorded on all snags in the dataset, we fitted the models without 
height information.
Site- level predictors included site type, stand live tree basal 
area, and disturbances during the remeasurement interval. We also 
used climatic variables that are important determinants for decom-
poser activity and wood properties as plot- level predictors. For 
temperature, we used growing degree days accumulated over the 
growing season, with a 0°C threshold (i.e. temperature sum). For 
moisture conditions, we used the climatic moisture index (Willmott 
& Feddema, 1992) that describes the relative wetness or dryness 
of a location based on precipitation and evapotranspiration, vary-
ing between −1 (dry) and 1 (wet). Both climate variables were ob-
tained from the openly available gridded Envirem data set (Title & 
Bemmels, 2018), based on climatic averages over 1970–2000, and 
with a 16 m × 16 m resolution. Climate variables were assigned to 
each snag based on plot coordinates (permanent plot coordinates 
are classified, and thus have a 1- km accuracy). We note that although 
topography and slope may also influence probability of a tree falling 
(Elie & Ruel, 2005), these were not recorded in the different NFIs in 
a manner that would make them readily harmonizable and were thus 
not included here.
Of the plot- level candidate predictors, basal area was available 
from the original data sets. However, the site type variable required 
harmonization to be useable here. In short (listed in Data S2), site 
type was harmonized so that Norwegian site types (30 site types) 
and Swedish site types (16 site types) were each assigned to a corre-
sponding Finnish site type (6 in total), based on expert opinion on the 
closest- matching site type. Out of these six types, the most produc-
tive and the least productive had a low number of observations, and 
they were thus assigned into the closest site type class. Thus, four 
site types were used in the analyses.
For disturbances, we used change in stand basal area during the 
remeasurement interval as an index of disturbance intensity. We di-
vided these into three categories: low intensity (reduction of less 
than 1/3 of basal area), intermediate intensity, (reduction of basal 
area between 1/3 and 2/3) and high intensity (reduction of over 2/3 
of basal area). As harvesting causes a vast majority of the basal area 
removal in Fennoscandian forests, this variable represents mainly 
the effect of harvesting.
2.5  |  Logistic regression models
Factors influencing snag fall (i.e. snag either standing [1] or fallen [0]) 
were modelled with generalized linear mixed models with a logit- 
link, binomial error distribution and with tree and plot identities as 
nested random effects (trees nested within plots). The reasoning 
behind the choice of the analysis method was to build a parametric 
model where the fixed part is easy to use for prediction, while deal-
ing with the violations of independence due to multiple observations 
of the same tree, and trees in the same plot. We further assumed 
that plots are independent of one another.
 13652664, 2024, 10, D
ow
nloaded from
 https://besjournals.onlinelibrary.wiley.com/doi/10.1111/1365-2664.14729 by Luonnonvarakeskus, W
iley Online Library on [21/01/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on W
iley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2396  |    AAKALA et al.
In the analyses, we used only the observations with 5- year 
remeasurement interval. Hence, the variable of interest was the 
probability of snag remaining standing over a 5- year period. In fit-
ting the models, we included all trees in the dataset so that data from 
Norway had a smaller minimum DBH (5 cm) than the Finnish and the 
Swedish data (both with a minimum DBH of 10 cm).
We used Bayesian MCMC methods for fitting the models, as im-
plemented in the brms package in R (Richardson et al., 2017). Two 
chains were run for 10,000 iterations, preceded by a burn- in phase 
of 10,000 iterations. Model validation consisted of visual inspec-
tion of trace plots (to assess mixing of chains), and an assessment 
of overdispersion. Model performance was approximated with pos-
terior predictive checks, where we calculated the correct classifica-
tion rates, and the area- under- the- curve (AUC) statistics with fixed 
effects models only (to approximate the skill of the model to predict 
unseen data).
Residual spatial autocorrelation is often problematic when an-
alysing ecological data. As individual snags are not mapped in the 
NFIs, we tested for spatial autocorrelation of residuals, using the 
grouped residuals at the plot- level, and conducted the Moran's I test 
for spatially correlated residuals, as implemented in the DHARMa- 
package in R (Hartig, 2022). We further explored the large- scale pat-
terns in model residuals, using interpolated maps (Data S3).
We visualized the influence of climate on snag persistence, pre-
dicting probability of snag remaining standing over the region where 
snags were measured in the NFI data sets. The visualization was 
based on the values of the climate data set and was predicted on a 
contiguous grid of 16 m × 16 m cells and setting all other predictors 
to fixed values.
Data analyses relied on R (version 3.5), and packages brms 
(Richardson et al., 2017), sf (Pebesma, 2018) and raster for spa-
tial data (Hijmans, 2020), pROC (Robin et al., 2011) for area- 
under- the- curve calculations, and ggplot2 (Wickham, 2016) 
and tmap (Tennekes, 2018) for visualizations. Inverse distance 
weighted (IDW) interpolations were done with the package gstat 
(Pebesma, 2004).
3  |  RESULTS
The compiled dataset consisted of 42,729 observations of a total 
of 25,610 individual snags. Scots pine, Norway spruce, and the two 
birch species combined all had a relatively similar number of obser-
vations (12,945, 14,278 and 15,569, respectively). In all, the records 
contained observations of 12,857 snag falls, and 29,935 snags re-
maining standing over the 5- year periods (details on the dataset, 
Data S3). On the map, the proportion of snags that survived showed 
a clear pattern from South to North (Figure 1).
The models predicted snag persistence over the 5- year period 
moderately well, based on posterior predictive checks (using fixed 
effects only). Correct classification rate for snags remaining stand-
ing was higher for the two conifers Scots pine (0.73) and Norway 
spruce (0.75), and somewhat lower for birch (0.68). Evaluated using 
the AUC, the values were 0.71 for pine, 0.72 for spruce and 0.68 for 
birch.
Factors that influenced snag persistence were consistent among 
tree species (Figure 2). Out of the tree- level variables, DBH had a 
consistently positive influence on probability of snags remaining 
standing, and advancing decay stage (from early to intermediate to 
advanced) a consistently negative influence. The change from inter-
mediate to advanced decay stage had a particularly large influence 
on the 5- year probability of a snag remaining standing (Figure 2, pa-
rameter estimate values in Data S3).
Decreasing site productivity increased the probability of snag 
remaining standing consistently among different species. Site pro-
ductivity differences were the greatest for spruce and birch, but 
discernible also for Scots pine (Figures 2 and 3). Basal area had a 
consistently negative, but modest influence on probability of snag 
remaining standing.
In addition to the influence of decay stage, the occurrence of 
harvesting or natural disturbances on the plot had a profound influ-
ence on snag persistence. In particular, the persistence of Norway 
spruce snags in any decay stage, and Scots pine and birch in the ad-
vanced decay stage was very low in plots subjected to clear- cut or 
F I G U R E  1  Smoothed maps of observed proportion of snags that survived during any of the 5- year intervals, for Scots pine, Norway 
spruce and the two birch species combined.
 13652664, 2024, 10, D
ow
nloaded from
 https://besjournals.onlinelibrary.wiley.com/doi/10.1111/1365-2664.14729 by Luonnonvarakeskus, W
iley Online Library on [21/01/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on W
iley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
    |  2397AAKALA et al.
severe natural disturbance (i.e. at least 2/3 of basal area removed; 
Figures 2 and 3).
Climate imposed a varying, but discernible influence on the 
probability of snags remaining standing. Overall, the climate- driven 
North–South gradient visible in the predictions (Figure 4) followed 
a largely similar variation as the larger geographic trends in the 
observed snag persistence and fall (Figure 1). The effects of tem-
perature here included as growing degree days were consistently 
negative, whereas the moisture index had a consistently positive 
influence on snag persistence. Jointly, these two variables imposed 
broad geographic gradients, in which the dominant feature is the 
high snag persistence in the mountains, and low snag persistence in 
southernmost Sweden (Figure 4).
Residual spatial autocorrelation was either absent or very weak. 
Moran's I may take values between −1 and 1, but varied in our anal-
yses over a very narrow, and probably biologically insignificant 
range of values, from <0.001 for birch, to 0.002 for Scots pine and 
0.005 for Norway spruce. Although for Norway spruce, the residual 
spatial autocorrelation was statistically significant at the 0.05 level 
(p = 0.012), this was probably mainly the result of the large sample 
size.
4  |  DISCUSSION
The ability of predicting how long snags remain standing from easily 
measurable parameters and understanding which factors influence 
standing times is paramount to incorporating snags into forest man-
agement. Here, we developed one of the largest datasets of snag 
observations to date and models for snag persistence and fall for the 
main tree species over Fennoscandia.
The models were moderately good at predicting snag per-
sistence over the 5- year periods, with correct classification rate 
varying between 0.68 for birch and 0.75 for Norway spruce that was 
also the most common species in the dataset. The classification ac-
curacy was within the same range, or somewhat better compared to 
Oberle et al. (2018), who had more data (close to 100,000 records), 
but also a much greater number of tree species within a broader cli-
matic range.
4.1  |  Factors influencing snag fall at the level of the 
tree individual
Variables applied here have the advantage of being easily measured 
and hence making the models useful for predictions, based on rou-
tinely collected forest inventory data. The disadvantage is that many 
of them incorporate a number of potential mechanisms, which makes 
it difficult to link snag fall rates to specific causes. Nevertheless, con-
sistency among tree species (as we fit the models separately), under-
standing of how trees stay upright, and findings from earlier studies 
helps in refining our understanding of the potential drivers and their 
effects and relative importance.
At the tree level, diameter had a consistently positive effect on 
snag persistence for each of the species. The strength of the effect 
varied among species so that a unit change in diameter had a stron-
ger influence on the probability of a snag remaining standing for 
birch compared to the coniferous Scots pine and Norway spruce. 
Diameter- persistence relationship has been shown mechanistically 
for live trees (Peltola et al., 1999) and is also identified in a num-
ber of observational studies on snag dynamics (Garber et al., 2005; 
Keen, 1955; Oberle et al., 2018; Russell & Weiskittel, 2012), but not 
always (Holeksa et al., 2008; Vanderwel et al., 2006). Our results are 
well in line with Oberle et al. (2018), who showed a clear increase in 
the 5- year probability of snag persistence with increasing diameter. 
In our models, the diameter effect was particularly clear in trees in 
F I G U R E  2  The influence of different variables on probability that a snag remains standing over a 5- year period, shown as posterior 
distributions of parameter estimates (±error, 95% credible intervals, CI) for each species. Parameter estimates, errors and credible interval 
values are in Data S3. DBH, diameter at breast height.
 13652664, 2024, 10, D
ow
nloaded from
 https://besjournals.onlinelibrary.wiley.com/doi/10.1111/1365-2664.14729 by Luonnonvarakeskus, W
iley Online Library on [21/01/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on W
iley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2398  |    AAKALA et al.
F I G U R E  3  Examples of how decay stage, site type and harvesting/natural disturbance influences snag persistence (i.e. probability for 
a snag to still be standing after a 5- year period) of different- sized snags of each species. Top row shows probability of a snag remaining 
standing on herb- rich (most productive) sites in the absence of harvesting or natural disturbances, middle row the probability of a snag 
remaining standing on xeric (poorest) sites in the absence of harvesting or natural disturbances, and the bottom row the probability of a snag 
remaining standing on xeric sites in clear- cut harvests or severe natural disturbance (over 2/3 basal area decrease in the sample plot). Other 
parameters are fixed for climatic conditions of southern Finland and basal area of 20 m2/ha. DBH, diameter at breast height.
 13652664, 2024, 10, D
ow
nloaded from
 https://besjournals.onlinelibrary.wiley.com/doi/10.1111/1365-2664.14729 by Luonnonvarakeskus, W
iley Online Library on [21/01/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on W
iley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
    |  2399AAKALA et al.
F I G U R E  4  The influence of climatic conditions on the probability for a snag to remain standing after a 5- year period based on predictions 
from models for each species. Snag persistence is shown here for trees with the average DBH for that species in the dataset (Scots pine 
15.4 cm, Norway spruce 14.9 cm, birch 11.8 cm; see Data S3 for details on the DBH distributions), on mesic sites with a 20 m2/ha basal area, 
in early (top row), intermediate (middle row) and advanced (bottom row) decay stage and without harvesting/disturbances. The geographical 
differences originate from differences in temperature sum and the moisture index in 16 m × 16 m grid cells. DBH, diameter at breast height.
 13652664, 2024, 10, D
ow
nloaded from
 https://besjournals.onlinelibrary.wiley.com/doi/10.1111/1365-2664.14729 by Luonnonvarakeskus, W
iley Online Library on [21/01/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on W
iley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2400  |    AAKALA et al.
advanced stages of decay (Figure 3). Onodera and Tokuda (2015) 
also showed that whether the effect is detectable may depend on 
the species. For instance, the increasing heart- rot or hollowing- 
out of larger trees may confound this relationship (Onodera & 
Tokuda, 2015). In general, the influence of tree size originates from 
the larger diameter stems being stronger (Peltola, 2006), but also 
from the higher proportion of decay- resistant heartwood in conifers 
(Taylor et al., 2002).
In our approach (as opposed to many detailed studies on snag 
dynamics), we did not consider the time the snag has remained 
standing. Instead, we used decay class as a measure of the snag his-
tory and otherwise considered each observation independent of its 
past. Snag persistence was relatively stable at early- and mid- stages 
of decay, but greatly reduced when moving to advanced stage of 
decay. This change was especially prominent for birch (Figure 3), 
which could be due to the differences in the decomposition path-
ways between birch and the two conifers. Birch tends to decay 
from inside- out and retain its bark (Krankina et al., 1999). By the 
time it is classified into advanced decay stage (assigned by how well 
a sharp object penetrates the wood), it is only held upright by the 
bark and falls easily. Overall, the pattern of increasing probability of 
falling with increasing decay class is mostly consistent with studies 
from boreal North America and Europe (Aakala, 2010; Vanderwel 
et al., 2006), and especially with the rapid increase in fall rates in the 
most advanced decay class reported for black spruce and balsam fir 
in northeastern Canada (Aakala et al., 2008).
Although our findings are not directly comparable with studies 
using snag time since death as a predictor (as in most studies on snag 
dynamics), the general trends seen in many of those studies are in 
line with what we observed (see also Vanderwel et al., 2006). The 
development of fall rates as a function of decay class reflected the 
patterns in studies of dynamics on snag age, where a reverse sigmoid 
curve has been shown to depict snag fall rate change with age. In our 
approach the fall rate obviously depends on how quickly the snags 
move from early to intermediate and to advanced stages of decay, 
but considering a given species, size and environment, our results 
are in line with this shape with an initial lag period that has been 
described in a number of studies (Garber et al., 2005; Keen, 1955; 
Lee, 1998; Mäkinen et al., 2006; Taylor & MacLean, 2007).
4.2  |  Non- climatic site- level predictors
Of the stand- level variables, both live tree basal area and site type (an 
indicator of productivity) had a consistent effect on snag persistence 
so that both decreasing live tree basal area and productivity 
increased snag persistence. Site type influences through differences 
in soil water holding capacity and capillary rise of water, which on 
higher productivity sites contribute to keeping the stem base moist 
enough for decomposer activity (Harmon, 1982; Keen, 1955). Site 
productivity also increases growth rates and hence influences 
the wood density (slower- growing conifers are denser; Eriksson 
et al., 2006), and higher nutrient availability may also increase 
decomposition rates of nutrient- poor substrates such as stemwood 
(Allison et al., 2009). Another potential effect of site productivity 
could be through its influence on tree allometry: shorter stature 
(Mäkelä et al., 2016), and deeper root systems (Pretzsch et al., 2012) 
in poorer productive sites, would both increase snag persistence.
Live tree basal area had a negative effect on snag persistence. 
Similar to what Oberle et al. (2018) reported on the effects of stand 
structure on snag persistence, the effect detected here was weak, 
but nevertheless consistent among the three species analysed. Live 
tree basal area captures multiple potential effects. Low basal area 
means more open stands, with lower humidity and higher winds 
that reduce snag wood moisture content, potentially low enough 
to influence the decomposer activity (Fahey, 1983; Johnson & 
Greene, 1991). Sparser stands with low basal areas also reduce the 
probability of knock- down when neighbouring trees fall, and high 
stem tapering in sparse stands might be more favourable to with-
stand bending forces on the stem (Peltola, 2006; Peltola et al., 1999). 
On the other hand, snags in stands with high basal areas may be 
more protected from wind, as wind energy dissipates faster in a 
higher- density stand (Oberle et al., 2018).
Disturbance severity (quantified as change in live tree basal 
area), had a clear effect on snag persistence so that snags were very 
likely to fall in the high intensity category. In Fennoscandia, most 
disturbances are due to harvesting, which may influence snag fall in 
a number of ways. Snags can be deliberately felled (e.g. for safety 
concerns), snags can be harvested for bioenergy, or they can be acci-
dentally knocked down when harvesting nearby living trees. In ear-
lier studies, harvest frequency is shown to have a negative impact 
on snag persistence in forests managed under the selection logging 
system (Garber et al., 2005; Vanderwel et al., 2006). As continuous- 
cover forestry with more frequent harvest entries is increasing in 
Fennoscandian forests, these findings imply that maintaining snags 
under this system requires explicit attention.
4.3  |  Climatic variables
One major advantage of this large dataset was that the same 
species covered a wide climatic gradient, from the hemiboreal 
forests to the Arctic. Temperature had a clearly negative effect on 
snag persistence and this effect can be explained by at least two 
different mechanisms. First, temperature has a generally positive 
influence on decomposer activity and wood decomposition rates, 
so that the integrity of the stem is reduced more quickly in warmer 
environments. The second effect comes through the influence of 
temperature on growth rates in the boreal forests, and the related 
influence on wood strength, as faster- growing trees tend to have a 
lower wood density (Eriksson et al., 2006). The influence temperature 
observed here is similar in magnitude to that reported by Oberle 
et al. (2018), who showed an increase in mean annual temperature 
of 0 to 20°C reduced probability of remaining standing from ca. 0.8 
to ca. 0.1 (figure 3 in Oberle et al., 2018). The reduction is similar, or 
even slightly stronger compared to the 5- year probability difference 
 13652664, 2024, 10, D
ow
nloaded from
 https://besjournals.onlinelibrary.wiley.com/doi/10.1111/1365-2664.14729 by Luonnonvarakeskus, W
iley Online Library on [21/01/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on W
iley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
    |  2401AAKALA et al.
in snag persistence between the southernmost and northernmost 
regions in our analyses (Figure 4).
The observed positive influence of the moisture index on snag 
persistence is less straightforward to explain. We would have ex-
pected drier climatic conditions to increase the probability of snag 
remaining standing, as moisture availability influences decomposers. 
Growing degree days and the moisture index were not correlated 
within the dataset (r < 0.01), but it may be that the moisture index 
is reflecting other temperature effects and that in general precip-
itation is not limiting snag decomposition, unlike what is observed 
from moisture- limited systems (e.g. Palmero- Iniesta et al., 2017). For 
reducing the strength of the base of the stem, soil hydrology might 
have an overriding role as a determinant of moisture conditions at 
the base of the stem. This effect would most likely partially be in-
cluded in the site type variable as the poorer productive sites are 
typically dry sites on coarse soils.
4.4  |  Implications for forest management, 
biodiversity and carbon
In the Fennoscandian region, snags are of major biodiversity 
concern as dearth of dead wood is listed among the main causes 
of biodiversity decline (Junninen & Komonen, 2011; Löfroth 
et al., 2023). A wide consensus exists in all three countries on 
the importance of increasing the amount and variability of dead 
wood in the managed forest landscapes. Because the functioning 
of dead wood as a habitat greatly changes when it falls (Runnel 
et al., 2013), managing snags requires understanding snag fall and 
factors influencing it.
To reach such aims of increasing the amount of dead wood, 
leaving snags as retention trees is nowadays recommended in for-
est management guidelines and forest certification requirements. 
Despite this, the development of snag volumes has been modest. 
In Finland, the amount of standing dead wood according to the 
12th NFI (2014–2018) in productive forest land was 1.6 m3/ha. At 
the national level, this value is similar to the mid- 1990s, when dead-
wood measurements started (Korhonen et al., 2021). The amount 
of standing dead wood in Sweden is not directly comparable to the 
Finnish figures as published data are available only for managed for-
ests. In those, the amount of standing dead wood has increased from 
1.6 m3/ha in 1996, to 2.9 m3/ha in 2017 (SLU, 2020). The increase 
can at least partly be attributed to increased conservation efforts 
(Kyaschenko et al., 2022). In Norway, standing deadwood in pro-
ductive forest land also increased, from 2.5 m3/ha in 1994–1998 to 
4.1 m3/ha in 2017 (Storaunet, 2021).
Most of the standing deadwood in the data used here, and 
reported in the Finnish NFI is in early stages of decay (Korhonen 
et al., 2021). The small proportion in more advanced decay stages 
obviously reflects the high probability of fall in the intermediate 
and especially advanced decay stages as shown also in our models 
(Figures 3 and 4). However, it also reflects the deliberate removal of 
dead wood (e.g. for bioenergy) during harvesting operations or by 
selective harvesting by the forest owner, and, importantly, the high 
probability of snags falling during or after the harvesting operations 
(Figure 3). This finding points to the need to retain and protect snags 
during management interventions as a potentially effective measure 
to increase their persistence in the managed forest landscapes. This 
applies to long- standing snags that are critical for slow- colonizing 
dead wood dependent species, such as many red- listed lichen spe-
cies (Larsson Ekström et al., 2023; Nirhamo et al., 2023), but similarly 
also to species relying on more ephemeral dead wood substrates, 
such as deciduous snags in advanced stages of decay [e.g. the en-
dangered Poecile montanus (Vatka et al., 2014)]. Moreover, the neg-
ative influence of warmer temperatures on snag persistence means 
that in a warmer climate, habitat availability for species dependent 
on snags will decrease further, unless compensated for by increased 
standing- tree mortality.
In addition to protecting existing snags during harvesting oper-
ations, the findings here point to two additional considerations if 
managing for increasing snag persistence. First, complementary to 
their importance as a habitat (Siitonen, 2001), larger diameter snags 
are also more persistent. This further adds to their value when de-
ciding which structures to retain in managed forests. Second, de-
creasing stand basal area increased snag persistence. Although this 
effect was relatively weak, it points to the potential of increasing 
snag persistence in the forests with lower stocking densities.
Properly accounting for dead wood and dead wood dynam-
ics can increase the accuracy of forest carbon models (Domke 
et al., 2011). Snags are not considered to be of major importance 
in the carbon cycling in the Fennoscandian forests due to their low 
numbers. Relative to the live tree volumes, the volume of snags is 
low, in Finland 1.3% (Korhonen et al., 2021), in (managed forests of) 
Sweden 2.6% (SLU, 2020) and in Norway, 3.3% (Storaunet, 2021). 
With climate change, increases in disturbances that lead to trees 
dying standing, such as drought, bark beetles, or fires, may change 
their importance in the future. After such a disturbance, consider-
able amount of the forest aboveground C may be in snags with low 
decomposition rates.
However, warming climate will potentially change the role of 
snags in the cycling of carbon, regardless of changes in mortality. 
Due to the consistent increase in snag fall rates with warmer climate 
identified here, the effect on carbon cycling will be stronger than 
what could be expected from the temperature effect on decom-
poser activity alone. This is due to the strong moisture- limitation 
of decomposer activity (Cornwell et al., 2009): as snag fall brings 
previously dry wood material in contact with the ground, snag 
fall increases decomposer activity and further accelerates the re-
lease of carbon (and nutrients) stored in the wood material (Hilger 
et al., 2012; Johnson & Greene, 1991).
5  |  CONCLUSIONS
The models presented here increase the possibilities for more in-
formative approaches to reaching targets associated with standing 
 13652664, 2024, 10, D
ow
nloaded from
 https://besjournals.onlinelibrary.wiley.com/doi/10.1111/1365-2664.14729 by Luonnonvarakeskus, W
iley Online Library on [21/01/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on W
iley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2402  |    AAKALA et al.
dead wood. Fitted separately for each of the species analysed, the 
models consistently indicated that leaving larger snags in low pro-
ductive edaphic and climatic conditions and in open forests maxi-
mizes their persistence. Using this as management recommendation 
would also slow forest carbon turnover, although for biodiversity 
reasons large snags in different environments are habitats for differ-
ent species and not substitutes for one another. The detrimental ef-
fect of harvesting on snag persistence identified here suggests that 
an efficient avenue to increasing and maintaining snags in managed 
forests is protecting them during silvicultural operations and man-
agement planning.
AUTHOR CONTRIBUTIONS
Tuomas Aakala conceived the ideas and designed the methodology; 
Tuomas Aakala, Kari T. Korhonen, Ken Olaf Storaunet and Bengt 
Gunnar Jonsson compiled and harmonized the data; Tuomas Aakala 
analysed the data and led the writing of the manuscripts. All au-
thors contributed critically to the drafts and gave final approval for 
publication.
ACKNOWLEDG EMENTS
We thank Bertil Westerlund (SLU Sweden) for providing the Swedish 
NFI data. The paper benefited from discussions with Shawn Fraver, 
Christopher W. Woodall and Pekka Punttila. We further thank the 
Associate Editor and two anonymous reviewers for their construc-
tive comments that greatly improved the paper. The work was sup-
ported by the Kone Foundation (grant to TA).
CONFLIC T OF INTERE S T S TATEMENT
The authors declare no conflict of interest.
DATA AVAIL ABILIT Y S TATEMENT
Data used in the modelling are available from the Dryad Digital 
Repository, available at https:// doi. org/ 10. 5061/ dryad. 37pvm cvtt 
(Aakala et al., 2024). The climate data is originally from the Envirem 
data set (Title & Bemmels, 2018), available at: envir em. github. io.
ORCID
Tuomas Aakala  https://orcid.org/0000-0003-0160-6410 
R E FE R E N C E S
Aakala, T. (2010). Coarse woody debris in late- successional Picea abies 
forests in northern Europe: Variability in quantities and models 
of decay class dynamics. Forest Ecology and Management, 260, 
770–779.
Aakala, T., & Heikkinen, J. (2024). Harmonized decay classification for 
dead wood in Nordic National Forest Inventories. Scandinavian 
Journal of Forest Research, 39, 1–7.
Aakala, T., Kuuluvainen, T., Gauthier, S., & De Grandpré, L. (2008). 
Standing dead trees and their decay- class dynamics in the north-
eastern boreal old- growth forests of Quebec. Forest Ecology and 
Management, 255, 410–420.
Aakala, T., Storaunet, K. O., Jonsson, B. G., & Korhonen, K. T. (2024). Data 
from: Drivers of snag fall rates in Fennoscandian boreal forests. Dryad 
Digital Repository. https:// doi. org/ 10. 5061/ dryad. 37pvm cvtt
Allison, S. D., LeBauer, D. S., Ofrecio, M. R., Reyes, R., Ta, A.- M., & 
Tran, T. M. (2009). Low levels of nitrogen addition stimulate de-
composition by boreal forest fungi. Soil Biology and Biochemistry, 
41, 293–302.
Breidenbach, J., Granhus, A., Hylen, G., Eriksen, R., & Astrup, R. (2020). 
A century of National Forest Inventory in Norway – Informing past, 
present, and future decisions. Forest Ecosystems, 7, 46.
Cornwell, W. K., Cornelissen, J. H. C., Allison, S. D., Bauhus, J., Eggleton, 
P., Preston, C. M., Scarff, F., Weedon, J. T., Wirth, C., & Zanne, A. 
E. (2009). Plant traits and wood fates across the globe: Rotted, 
burned, or consumed. Global Change Biology, 15, 2431–2449.
Domke, G. M., Woodall, C. W., & Smith, J. E. (2011). Accounting for 
density reduction and structural loss in standing dead trees: 
Implications for forest biomass and carbon stock estimates in the 
United States. Carbon Balance and Management, 6, 14.
Elie, J. G., & Ruel, J.- C. (2005). Windthrow hazard modelling in boreal 
forests of black spruce and jack pine. Canadian Journal of Forest 
Research, 35, 2655–2663.
Eriksson, D., Lindberg, H., & Bergsten, U. (2006). Influence of silvicultural 
regime on wood structure characteristics and mechanical proper-
ties of clear wood in Pinus sylvestris. Silva Fennica, 40, 743–762.
Fahey, T. J. (1983). Nutrient dynamics of aboveground detritus in lodge-
pole pine (Pinus contorta ssp. latifolia) ecosystems, southeastern 
Wyoming. Ecological Monographs, 53, 51–72.
Fridman, J., Holm, S., Nilsson, M., Nilsson, P., Ringvall, A. H., & Ståhl, G. 
(2014). Adapting National Forest Inventories to changing require-
ments – The case of the Swedish National Forest Inventory at the 
turn of the 20th century. Silva Fennica, 48, 1095.
Garber, S. M., Brown, J. P., Wilson, D. S., Maguire, D. A., & Heath, L. S. 
(2005). Snag longevity under alternative silvicultural regimes in 
mixed- species forests of central Maine. Canadian Journal of Forest 
Research, 35, 787–796.
Grieser, J., Gommes, R., & Bernardi, M. (2006). New LocClim – The local 
climate estimator of FAO. Geophysical Research Abstracts, 8, 08305.
Harmon, M. E. (1982). Decomposition of standing dead trees in the 
southern Appalachian Mountains. Oecologia, 52, 214–215.
Hartig, F. (2022). DHARMa: Residual diagnostics for hierarchical (multi- 
level/mixed) regression models. R Package version 0.4.6.
Hijmans, R. J. (2020). raster: Geographic data analysis and modeling. R 
package version 3.3-7.
Hilger, A. B., Shaw, C. H., Metsaranta, J. M., & Kurz, W. A. (2012). 
Estimation of snag carbon transfer rates by ecozone and lead spe-
cies for forests in Canada. Ecological Applications, 22, 2078–2090.
Holeksa, J., Zielonka, T., & Żywiec, M. (2008). Modeling the decay of 
coarse woody debris in a subalpine Norway spruce forest of the 
West Carpathians, Poland. Canadian Journal of Forest Research, 38, 
415–428.
Hylen, G., Fernández, C. A., & Granhus, A. (2022). Skogressurser i Norge: 
Status og framtidsscenarier. NIBIO.
Johnson, E. A., & Greene, D. F. (1991). A method for studying dead bole 
dynamics in Pinus contorta var. latifolia–Picea engelmannii forests. 
Journal of Vegetation Science, 2, 523–530.
Junninen, K., & Komonen, A. (2011). Conservation ecology of boreal 
polypores: A review. Biological Conservation, 144, 11–20.
Kalliola, R. (1966). The reduction of the area of forests in natural con-
dition in Finland in the light of some maps based upon National 
Forest Inventories. Annales Botanici Fennici, 3, 442–448.
Keen, F. P. (1955). The rate of natural falling of beetle- killed ponderosa 
pine snags. Journal of Forestry, 53, 720–723.
Korhonen, K. T., Ihalainen, A., Kuusela, S., Punttila, P., Salminen, O., & 
Syrjänen, K. (2020). Metsien monimuotoisuudelle merkittävien 
rakennepiirteiden muutokset Suomessa vuosina 1980- 2015. 
Metsätieteen aikakauskirja, 2020, 10198.
Korhonen, K. T., Ahola, A., Heikkinen, J., Henttonen, H. M., Hotanen, 
J.- P., Ihalainen, A., Melin, M., Pitkänen, J., Räty, M., Sirviö, M., & 
 13652664, 2024, 10, D
ow
nloaded from
 https://besjournals.onlinelibrary.wiley.com/doi/10.1111/1365-2664.14729 by Luonnonvarakeskus, W
iley Online Library on [21/01/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on W
iley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
    |  2403AAKALA et al.
Strandström, M. (2021). Forests of Finland 2014–2018 and their 
development 1921–2018. Silva Fennica, 55, 10662.
Krankina, O. N., Harmon, M. E., & Griazkin, A. V. (1999). Nutrient stores 
and dynamics of woody detritus in a boreal forest: Modeling po-
tential implications at the stand level. Canadian Journal of Forest 
Research, 29, 20–32.
Kyaschenko, J., Strengbom, J., Felton, A., Aakala, T., Staland, H., & Ranius, 
T. (2022). Increase in dead wood, large living trees and tree diver-
sity, yet decrease in understory vegetation cover: The effect of 
three decades of biodiversity- oriented forest policy in Swedish for-
ests. Journal of Environmental Management, 313, 114993.
Larson, P. R. (1965). Stem form of young Larix as influenced by wind and 
pruning. Forest Science, 11, 412–424.
Larsson Ekström, A., Sjögren, J., Djupström, L. B., Thor, G., & Löfroth, T. 
(2023). Reinventory of permanent plots show that kelo lichens face 
an extinction debt. Biological Conservation, 288, 110363.
Lee, P. (1998). Dynamics of snags in aspen- dominated midboreal forests. 
Forest Ecology and Management, 105, 263–272.
Linder, P., & Östlund, L. (1998). Structural changes in three mid- boreal 
Swedish forest landscapes, 1885–1996. Biological Conservation, 85, 
9–19.
Löfroth, T., Birkemoe, T., Shorohova, E., Dynesius, M., Fenton, N. J., 
Drapeau, P., & Tremblay, J. A. (2023). Deadwood biodiversity. In M. 
M. Girona, H. Morin, S. Gauthier, & Y. Bergeron (Eds.), Boreal forests 
in the face of climate change: Sustainable management (pp. 167–189). 
Springer International Publishing.
Mäkelä, A., Pulkkinen, M., & Mäkinen, H. (2016). Bridging empirical and 
carbon- balance based forest site productivity – Significance of 
below- ground allocation. Forest Ecology and Management, 372, 64–77.
Mäkinen, H., Hynynen, J., Siitonen, J., & Sievänen, R. (2006). Predicting 
the decomposition of Scots pine, Norway spruce, and birch stems 
in Finland. Ecological Applications, 16, 1865–1879.
Nilsson, P., Roberge, C., & Fridman, J. (2020). Forest statistics 2020. 
Swedish University of Agricultural Sciences.
Nirhamo, A., Pykälä, J., Jääskeläinen, K., & Kouki, J. (2023). Habitat as-
sociations of red- listed epiphytic lichens in Finland. Silva Fennica, 
57, 22019.
Oberle, B., Ogle, K., Zanne, A. E., & Woodall, C. W. (2018). When a tree 
falls: Controls on wood decay predict standing dead tree fall and 
new risks in changing forests. PLoS One, 13, e0196712.
Oettel, J., Zolles, A., Gschwantner, T., Lapin, K., Kindermann, G., 
Schweinzer, K.- M., Gossner, M. M., & Essl, F. (2023). Dynamics of 
standing deadwood in Austrian forests under varying forest man-
agement and climatic conditions. Journal of Applied Ecology, 60, 
696–713.
Onodera, K., & Tokuda, S. (2015). Do larger snags stand longer? Snag 
longevity in mixed conifer–hardwood forests in Hokkaido, Japan. 
Annals of Forest Science, 72, 621–629.
Palmero- Iniesta, M., Domènech, R., Molina- Terrén, D., & Espelta, J. M. 
(2017). Fire behavior in Pinus halepensis thickets: Effects of thinning 
and woody debris decomposition in two rainfall scenarios. Forest 
Ecology and Management, 404, 230–240.
Pebesma, E. (2018). Simple features for R: Standardized support for spa-
tial vector data. The R Journal, 10, 439.
Pebesma, E. J. (2004). Multivariable geostatistics in S: The gstat package. 
Computers & Geosciences, 30, 683–691.
Peltola, H., Kellomäki, S., Väisänen, H., & Ikonen, V.- P. (1999). A mech-
anistic model for assessing the risk of wind and snow damage to 
single trees and stands of Scots pine, Norway spruce, and birch. 
Canadian Journal of Forest Research, 29, 647–661.
Peltola, H. M. (2006). Mechanical stability of trees under static loads. 
American Journal of Botany, 93, 1501–1511.
Pelz, K. A., & Smith, F. W. (2013). How will aspen respond to mountain 
pine beetle? A review of literature and discussion of knowledge 
gaps. Forest Ecology and Management, 299, 60–69.
Pretzsch, H., Uhl, E., Biber, P., Schütze, G., & Coates, K. D. (2012). Change 
of allometry between coarse root and shoot of lodgepole pine 
(Pinus contorta DOUGL. ex. LOUD) along a stress gradient in the 
sub- boreal forest zone of British Columbia. Scandinavian Journal of 
Forest Research, 27, 532–544.
Richardson, T. S., Robins, J. M., & Wang, L. (2017). On modeling and 
estimation for the relative risk and risk difference. Journal of the 
American Statistical Association, 112, 1121–1130.
Robin, X., Turck, N., Hainard, A., Tiberti, N., Lisacek, F., Sanchez, J.- 
C., & Müller, M. (2011). pROC: An open- source package for R 
and S+ to analyze and compare ROC curves. BMC Bioinformatics, 
12, 77.
Runnel, K., Rosenvald, R., & Lõhmus, A. (2013). The dying legacy of green 
tree retention: Different habitat values for polypores and wood- 
inhabiting lichens. Biological Conservation, 159, 187–196.
Russell, M. B., & Weiskittel, A. R. (2012). Assessing and modeling 
snag survival and decay dynamics for the primary species in the 
Acadian forest of Maine, USA. Forest Ecology and Management, 284, 
230–240.
Siitonen, J. (2001). Forest management, coarse woody debris and sap-
roxylic organisms: Fennoscandian boreal forests as an example. 
Ecological Bulletins, 49, 11–41.
SLU. (2020). Forest statistics Sweden 2020. Swedish University of 
Agricultural Sciences.
Statistics Sweden. (2021). Protected nature 2021. Sweden.
Stokland, J. N. (2003). Forest biodiversity indicators in the Nordic coun-
tries: Status based on National Forest Inventories. Nordic Council of 
Ministers.
Stokland, J. N., Siitonen, J., & Jonsson, B. G. (2012). Biodiversity in dead 
wood. Cambridge University Press.
Storaunet, K. O. (2021). Død ved. Page Bærekraftig skogbruk i Norge. Norsk 
Institutt for Bioøkonomie.
Storaunet, K. O., & Rolstad, J. (2002). Time since death and fall of 
Norway spruce logs in old- growth and selectively cut boreal forest. 
Canadian Journal of Forest Research, 32, 1801–1812.
Storaunet, K. O., & Rolstad, J. (2015). Mengde og utvikling av død ved i 
produktiv skog i Norge. Med basis i data fra Landsskogtakseringens 7. 
(1994–1998) og 10. takst (2010–2013). Oppdragsrapport fra Skog og 
landskap.
Suojelualueet. (2022). Suojelualueet, talousmetsien monimuotoisuuden 
suojelukohteet ja luontoarvojen suojelua tukevat alueet, koko maa. 
Luonnonvarakeskus.
Suvanto, S., Henttonen, H. M., Nöjd, P., & Mäkinen, H. (2016). Forest sus-
ceptibility to storm damage is affected by similar factors regardless 
of storm type: Comparison of thunder storms and autumn extra- 
tropical cyclones in Finland. Forest Ecology and Management, 381, 
17–28.
Taylor, A. M., Gartner, B. L., & Morrell, J. J. (2002). Heartwood forma-
tion and natural durability – A review. Wood and Fiber Science, 34, 
587–611.
Taylor, S. L., & MacLean, D. A. (2007). Dead wood dynamics in declining 
balsam fir and spruce stands in New Brunswick, Canada. Canadian 
Journal of Forest Research, 37, 750–762.
Tennekes, M. (2018). tmap: Thematic maps in R. Journal of Statistical 
Software, 84, 1–39.
Thomas, J. W. (2002). Dead wood: From forester's bane to environmen-
tal boon. Proceedings of the symposium on the ecology and manage-
ment of dead wood in western forests. USDA Forest Service General 
Technical Report PSW-GTR 181, pp. 3–9.
Tikkanen, M. (2005). Climate. In M. Seppälä (Ed.), The physical geography 
of Fennoscandia (pp. 97–112). Oxford University Press.
Title, P. O., & Bemmels, J. B. (2018). ENVIREM: An expanded set of 
bioclimatic and topographic variables increases flexibility and im-
proves performance of ecological niche modeling. Ecography, 41, 
291–307.
 13652664, 2024, 10, D
ow
nloaded from
 https://besjournals.onlinelibrary.wiley.com/doi/10.1111/1365-2664.14729 by Luonnonvarakeskus, W
iley Online Library on [21/01/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on W
iley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2404  |    AAKALA et al.
Vanderwel, M. C., Malcolm, J. R., & Smith, S. M. (2006). An integrated 
model for snag and downed woody debris decay class transitions. 
Forest Ecology and Management, 234, 48–59.
Vatka, E., Kangas, K., Orell, M., Lampila, S., Nikula, A., & Nivala, V. (2014). 
Nest site selection of a primary hole- nesting passerine reveals 
means to developing sustainable forestry. Journal of Avian Biology, 
45, 187–196.
Wickham, H. (2016). ggplot2: Elegant graphics for data analysis. 
Springer- Verlag.
Willmott, C. J., & Feddema, J. J. (1992). A more rational climatic moisture 
index. The Professional Geographer, 44, 84–88.
Woodall, C. W., Domke, G. M., Macfarlane, D. W., & Oswalt, C. M. (2012). 
Comparing field- and model- based standing dead tree carbon stock 
estimates across forests of the US. Forestry: An International Journal 
of Forest Research, 85, 125–133.
Woodall, C. W., Walters, B. F., & Westfall, J. A. (2012). Tracking downed 
dead wood in forests over time: Development of a piece match-
ing algorithm for line intercept sampling. Forest Ecology and 
Management, 277, 196–204.
SUPPORTING INFORMATION
Additional supporting information can be found online in the 
Supporting Information section at the end of this article.
Data S1. Piece- matching dead wood in the Finnish NFI.
Data S2. Harmonized site types.
Data S3. Predictor variable distributions, parameter estimates and 
residual spatial autocorrelation.
How to cite this article: Aakala, T., Storaunet, K. O., Jonsson, 
B. G., & Korhonen, K. T. (2024). Drivers of snag fall rates in 
Fennoscandian boreal forests. Journal of Applied Ecology, 61, 
2392–2404. https://doi.org/10.1111/1365-2664.14729
 13652664, 2024, 10, D
ow
nloaded from
 https://besjournals.onlinelibrary.wiley.com/doi/10.1111/1365-2664.14729 by Luonnonvarakeskus, W
iley Online Library on [21/01/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on W
iley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License