Forest Ecology and Management 430 (2018) 485–497
Contents lists available at ScienceDirect
Forest Ecology and Management
journal homepage: www.elsevier.com/locate/foreco
Alternative tree species under climate warming in managed European
forests
T
Eric Andreas Thurma, , Laura Hernandezb, Andri Baltensweilerc, Szegin Ayand, Ervin Rasztovitse,
Kamil Bielakf, Tzvetan Mladenov Zlatanovg, David Hladnikh,i, Besim Balicj,
Alexandra Freudenschussk, Richard Büchsenmeisterk, Wolfgang Falka
⁎
a
Bavarian State Institute of Forestry LWF, Dep. Soil and Climate, Germany
INIA-CIFOR, Spain
c
Swiss Federal Institute for Forest, Snow and Landscape Research WSL, Dep. Forest Resources and Management, Switzerland
d
Kastamonu University, Faculty of Forestry, Dep. of Silviculture, Turkey
e
NARIC – Forest Research Institute, Hungary
f
Warsaw University of Life Sciences, Dep. of Silviculture, Poland
g
Institute of Biodiversity and Ecosystem Research, Bulgarian Academy of Sciences, Bulgaria
h
University of Ljubljana, Biotechnical Faculty, Dep. of Forestry and Renewable Forest Resources, Slovenia
i
Slovenian Forestry Institute, Slovenia
j
Faculty of Forestry in Sarajevo, Forest Management and Urban Greenery, Bosnia and Herzegovina
k
Federal Research and Training Centre for Forests, Dep. Natural Hazards and Landscape (BFW), Austria
b
A R T I C LE I N FO
A B S T R A C T
Keywords:
Thermophilic
Rare species
Species distribution models
Climate-soil models
Site index models
Winners & losers
Northward shift
Biogeographical regions
This study estimates the present and future distribution potential of 12 thermophilic and rare tree species for
Europe based on climate-soil sensitive species distribution models (SDMs), and compares them to the two major
temperate and boreal tree species (Fagus sylvatica and Picea abies).
We used European national forest inventory data with 1.3 million plots to predict the distribution of the
12 + 2 tree species in Europe today and under future warming scenarios of +2.9 and +4.5 °C. The SDMs that
were used to calculate the distributions were in a first step only given climate variables for explanation. In a
second step, deviations which could not be explained by the climate models were tested in an additional soil
variable-based model. Site-index models were applied to the found species distribution to estimate the growth
performance (site index) under the given climate.
We find a northward shift of 461 km and 697 km for the thermophilic species over the regarded time period
from 2060 to 2080 under a warming scenario of 2.9 °C and 4.5 °C, respectively. Potential winners of climatic
warming have their distribution centroid below 48°N. Fagus sylvatica and Picea abies will lose great parts of their
potential distribution range (approx. 55 and 60%, respectively). An index of area gain and growth performance
revealed Ulmus laevis, Quercus rubra, Quercus cerris and Robinia pseudoacacia as interesting alternatives in
managed temperate forests currently dominated by F. sylvatica and P. abies.
The 12 investigated species are already in focus in forestry and it has been shown that the changing climate
creates conditions for a targeted promotion in European forests. Nevertheless, area winners exhibited lower
growth performances. So, forest conversion with these warm-adapted species goes hand in hand with loss of
overall growth performance compared to current species composition. So, the results are a premise for a further
discussion on the ecological consequences and the consistency with forest socio-economic goals and conservation policies.
1. Introduction
Forests cover 33% of the European land area and 96% of this area is
managed and contributes to the supply of the resource wood (Forest
⁎
Europe, 2015). Fagus sylvatica L. and Picea abies (L.) Karst. are two of
the most common tree species in boreal and temperate European forests
(Köble and Seufert, 2001) with high economic relevance especially for
the timber industry (Eurostat, 2018). Nevertheless, several studies
Corresponding author.
E-mail address: thurm@lrz.tum.de (E.A. Thurm).
https://doi.org/10.1016/j.foreco.2018.08.028
Received 20 April 2018; Received in revised form 10 August 2018; Accepted 13 August 2018
0378-1127/ © 2018 Elsevier B.V. All rights reserved.
Forest Ecology and Management 430 (2018) 485–497
E.A. Thurm et al.
step would be to use genetic adaptability of species and to promote
drought stress tolerant characteristics or choose drought tolerant populations for planting (Chakraborty et al., 2015; Fady et al., 2016;
Montwé et al., 2016).
Nonetheless, the mentioned strategies can only mitigate the risk to a
certain degree, and in many regions of Europe the increasing temperatures will force a change of tree species composition (Bonan, 2008;
Hanewinkel et al., 2013). In order to prevent a successional-driven
permanence of the currently dominant forests species and to ensure the
social and economic benefits of this sector, it is important to formulate
recommendations for the introduction of alternative tree species.
Therefore the question arises which species are able to form the next
generation of high-resistance forest and can comply with the qualitative
and quantitative demands of the wood market.
Traditionally, large cultivation trials generated recommendations
for new non-native or rare species (a detailed description for several
species can be found in Pâques, 2013). However, the long time span of
these trials is contrasted by the rapidity of climatic change, and forces
us to seek faster methodical ways to recommend species ad hoc. Species
distribution models (SDMs) are able to generate such information (Elith
and Leathwick, 2009). Modern computation and high quality occurrence data sets make it relatively easy to generate such models. They
have therefore become a popular tool in macroecology over the last
decades to understand global distribution patterns of species. However,
SDMs have scarcely been used as a tool in the forest sector or for
predict that their proportion will decrease under climate change in
European forests (e.g. Hanewinkel et al., 2013; Kellomaki et al., 2001;
Ruiz-Labourdette et al., 2013) and they will be replaced by species with
a more thermophilic character (Kullman, 2008). Since timber producing tree species typically have rotation periods of 60 years and more,
forest managers must know already today which and where such
thermophilic species can replace or enrich the current tree species
compositions.
Hanewinkel et al. (2013) forecasted a shift in wood assortment in
Central Europe from productive species, mainly P. abies and Pinus sylvestris L., to slow-growing species like Mediterranean oaks. This transformation led to lower productivity and land expectation values.
However, adaptive forest management provides strategies to mitigate
such impacts. Over the last two decades the research on adaption to
climate change has greatly increased. Some measures aim at increasing
forest stability based on the current tree species spectrum, for instance
by decreasing the forest stand density (Aussenac and Granier, 1988;
Kohler et al., 2010; Rais et al., 2014). Also, the shortening of the rotation period reduces the risk of unexpected mortality (Seidl et al.,
2011). Another strategy is to regenerate or foster suitable tree species
mixtures. Several studies have shown that compensatory and facilitative effects between species are able to improve the stability of tree
growth (del Río et al., 2017; Lebourgeois et al., 2013; Pretzsch et al.,
2013; Thurm et al., 2016) and also the water status of the trees (Schäfer
et al., 2018; Schume et al., 2004) by reducing drought stress. A next
Fig. 1. Annual temperature and precipitation range of the 12 investigated species plus the two comparative species Fagus sylvatica and Picea abies. The data represents
the occurrences of the species in the inventory data in Europe. Climate data derives from WorldClim 1.4 dataset (Hijmans et al., 2005). The species are ranked by
their median temperature.
486
Forest Ecology and Management 430 (2018) 485–497
E.A. Thurm et al.
conservation strategies because of the uncertainty of these models (see
similar debate in Pearson and Dawson, 2003) and the often coarse
spatial resolution of macroecological studies. SDMs have to meet the
spatial and temporal needs of decision-makers and practitioners
(Guisan et al., 2013). Therefore it is important to give species-specific
statements in a fine spatial resolution (Falk and Mellert, 2011) when
considering model outcomes in tree species choices. European managed
forests are anthropogenically driven (e.g. 27% of European forest
stands were afforested or regenerated by planting or seeding, 32% are
mono-specific; Forest Europe, 2015), so dispersal limitations or species
interactions like competition are not constraints for large areas.
Nowadays, it is important to describe species’ climatic requirements to
evaluate the species out of their natural range (Booth, 2017).
Our study investigated 12 rare and thermophilic tree species (Fig. 1)
that might be able to replace or enrich important economic tree species
in Central Europe under drier and warmer conditions (e.g. Kölling,
2017; Albrecht and Avila, 2018). The species are not among the major
tree species in Europe (UN-ECE/ICP Forests Level I). The small amount
of data for these species makes it difficult to model them and so far only
a few studies have dealt with their distribution in Europe (see literature
review in appendix of Dyderski et al. (2018)). To fill the gap, we investigate which of these species will be winners or losers under two
scenarios of changing climate. To facilitate the presentation and understanding of the results we stratify them according to the European
biogeographical regions in Roekaerts (2002). As the distribution of
forest areas in Europe also influences the expansion or reduction of a
tree species’ area, we include this topic in our investigations.
We formulate the above considerations into three hypotheses: (I)
the northward shift of Picea abies and Fagus sylvatica is comparable to
the 12 investigated thermophilic or non-native species. (II) The potential distribution area of the 12 investigated thermophilic or nonnative species will increase in the European temperate and boreal zone,
whereas the area of P. abies and F. sylvatica will decrease. (III) The
further south the species is distributed, the greater is its potential area
expansion under climatic warming.
While the focus is clearly on the changes in the potential distribution of the 12 + 2 species, we add to the findings the maximum growth
performance of each species as an important criterion in assessing
suitable alternative species for the European forest sector.
2. Materials & method
2.1. Tree dataset
For the study, we analyzed 12 thermophilic tree species (see Fig. 1
for the species list) plus Picea abies and Fagus sylvatica as representatives
of one major coniferous and broadleaf timber species in Central Europe.
We defined the species as thermophilic (Giesecke, 2016; Kullman,
2008) because of their warmer temperature amplitude compared to tree
species P. abies and F. sylvatica (see Fig. 1).
There is a long list of species which are under discussion to substitute P. abies and F. sylvatica in Central Europe (e.g. Kölling, 2017;
Albrecht and Avila, 2018). Based on growth performance and drought
resistance, we selected species which have the potential to increase
their importance for forestry in the future (Hanewinkel et al., 2013;
Rigling et al., 2013). The species are well known to forestry and present
in European inventory data, but at the same time there is limited
knowledge on their traits, distribution limits and growth patterns. The
climatic requirements of some species, like Abies grandis (Dougl. ex D.
Don) Lindl., Castanea sativa Mill., Larix kaempferi (Lamb.) Carr., Prunus
avium L., Pinus nigra Arnold., Quercus rubra L. and Robinia pseudoacacia
L. are closer to Central European forests, whereas the importance of
others like Quercus pubescens Mild. and Quercus cerris L. might increase
in the future due to climate change. We aimed not to focus on species
that were part of a previous study (Walentowski et al., 2017) that used
distribution models, vegetation analysis and tree ring data to assess
future suitability of tree species under climate change in southern
Germany (e.g. A. campestre, A. platanoides, A. pseudoplatanus, Carpinus
betulus, Q. robur, Q. petrea, Sorbus and Tilia species). Our species list
includes native and non-native species in Europe (see Table 1).
We used the most up-to-date national forest inventory data and
forest management plans (see Fig. 2; data description in Supplementary
material 1) from 12 European countries, representing a wide range of
different climatic zones. National forest inventory data originated from
Spain, France, Switzerland, Austria, Italy, Germany, Poland, Slovenia
and Bosnia-Herzegovina and the forest management plans came from
Hungary and Bulgaria, and large parts of Turkey. These data contained
occurrences and biometric parameters (diameter at breast height 1.3 m
– DBH, tree height, age (for most countries), social tree classes by Kraft
(1884)) of the investigated species which were used to rank the species
Table 1
Overview of the model quality for the 14 tree species with the applied abbreviation of the tree species (Abb.) and the classification if the species is native (Nat.) in
Europe (Y – Native, No – Non-Native); separated according to pure climate model and climate & soil model. Quality indices are sensitivity (SE), specifity (SP) and the
true skill statistic (TSS). N presents the underlying number of data points for the final SDM (presence – absence; 50:50). Columns ‘Site Index Models’ include model
quality of the site index models with the coefficient of determination R2 and the number of data points NG.
Species
Names
Abies grandis
Acer monspessulanum
Castanea sativa
Fagus sylvatica
Larix kaempferi
Picea abies
Pinus nigra
Prunus avium
Pyrus pyraster
Quercus cerris
Quercus pubescens
Quercus rubra
Robinia pseudoacacia
Ulmus laevis
Average
Explanatory variables
Abb.
AB
AM
CS
FS
LK
PA
PN
PV
PP
QC
QP
QR
RP
UL
Distribution models
Nat.
No
Y
Y
Y
No
Y
Y
Y
Y
Y
Y
No
No
Y
Climate
Bio1 – Conti1 – Bio18 – DR – PARMADO1
Bio5 – Bio6 – Bio12 – STR_SUB – PARMADO1
Bio1 – Conti1 – Bio12 – PARMADO1
Bio5 – Bio6 – Bio18 – Bio5 : Bio18 – DIMP – PARMADO1
Bio1 – Conti1 – Bio18
Bio10 – Conti1 – Bio18 – VS – BS_TOP
Bio1 – Conti1 – Bio18 – DR – PARMADO1
Bio5 – Bio6 – Bio18 – Bio5 : Bio18 – DIMP
Bio1 – Conti1 – Bio18
Bio1 – Conti1 – Bio18 – STR_TOP – NUTRI1
Bio1 – Conti1 – Bio12 – TEXT – NUTRI1
Bio1 – Bio6 – Bio18 – DR – PH
Bio10 – Bio6 – Bio18 – NUTRI1
Bio10 – Bio6 – Bio18 – PARMADO1
487
Site index models
Climate & Soil
SE
SP
TSS
SE
SP
TSS
N
R2
NG
0.93
0.90
0.89
0.88
0.95
0.92
0.86
0.85
0.86
0.88
0.91
0.88
0.89
0.86
0.89
0.76
0.76
0.81
0.77
0.87
0.75
0.67
0.73
0.70
0.79
0.76
0.69
0.82
0.84
0.77
0.69
0.66
0.70
0.65
0.82
0.68
0.53
0.57
0.57
0.67
0.67
0.57
0.70
0.70
0.66
0.95
0.92
0.89
0.90
–
0.91
0.87
0.86
–
0.91
0.92
0.90
0.89
0.86
0.90
0.76
0.76
0.81
0.79
–
0.79
0.71
0.72
–
0.79
0.76
0.71
0.82
0.84
0.77
0.72
0.68
0.71
0.69
–
0.69
0.58
0.58
–
0.70
0.68
0.61
0.70
0.71
0.67
876
803
5004
10,836
2429
17,607
4984
7709
1926
3089
3825
3737
6086
819
4981
0.37
0.13
0.18
0.04
0.06
0.07
0.02
0.11
0.25
0.12
0.12
0.20
0.06
0.16
0.13
450
70
1149
12050
757
12292
23731
1229
156
122138
18875
4366
177548
455
26804
Forest Ecology and Management 430 (2018) 485–497
E.A. Thurm et al.
data (mean age = 54.6 years, standard deviation = 29.7). We included
the following six climate variables in our SDMs because they were
strong predictors in previous studies (e.g. Dyderski et al., 2018;
Walentowski et al., 2017): annual mean temperature (Bio1), maximum
temperature of warmest month (Bio5), minimum temperature of
coldest month (Bio6), mean temperature of warmest quarter (Bio10),
annual precipitation (Bio12) and precipitation of the warmest quarter
(Bio18) (see also Supplementary material 3.1).
In addition, we calculated a continentality index (Conti1) (Conrad,
1946),
Conti1 =
1.7A
−14
sin(φ + 10)
(1)
where A is the difference of Bio5 and Bio 6 and φ represented geographic latitude.
The predictions for the future climate were calculated for the time
period 2061–2080 (also WorldClim 1.4 data, below abbreviated as
‘2070’). Models were applied for emission scenarios RCP 4.5 and RCP
8.5. Within the same emission scenario, there is a large deviation between the different global climate models (Goberville et al., 2015).
Therefore, we used three global climate models per emission scenario
(HadGEM2-ES, MPI-ESM-LR, NorESM1-M) from European institutes.
The distribution models were predicted with the different global climate models and the outputs averaged to scenario- and species-specific
future prediction. The applied climate models have been used extensively in previous studies in Europe (e.g. Dyderski et al., 2018;
Schüler et al., 2014) and they seem to capture well the range of global
climate models in the respective RCPs (Supplementary material 4).
The ensemble of the three climate models represents an annual
temperature increase in Europe of 2.9 °C for RCP 4.5 and 4.5 °C for RCP
8.5. An overview of the temperature increase in individual biogeographical regions (Roekaerts, 2002) can be found in Supplementary
material 4. Europe was defined as the area within longitude 15°W to
45°E and latitude from 35° to 70°N.
Fig. 2. Species plots of compiled dataset.
according to growth characteristics. We complemented the data with
presence/absence information of the ICP Level I plot data on crown
condition and with a pan-European tree occurrence dataset (Mauri
et al., 2017). The Mauri et al. (2017) dataset provides presence information from national forest inventories of 30 countries. In the end,
our data set encompassed 1,352,113 species plots represented at a
spatial resolution of 1 km.
Supplementary data associated with this article can be found, in the
online version, at https://doi.org/10.1016/j.foreco.2018.08.028.
2.2. Tree data preparation and processing
2.4. Soil dataset
For the calibration of the SDMs, we carried out a species-specific
selection of the data. For each species, the data points were grouped
into a 16 × 16 km raster (see schematic figure of method in
Supplementary material 2) and only one randomly selected data point
per 16 × 16 km raster cell was used for the modeling. Raster cells
which were occupied by data points of the analyzed species were preferentially selected. Presences were combined with the equal number of
absences (see Table 1 for sample size N of each species). We sorted the
absences randomly within each 16 × 16 km raster cell and then reduced the number of absences per cell to one and finally sorted the cells
according to their x and y coordinates and selected each ½ * N (number
of absences that equals the number of presences) cell. This ensures that
absences are randomly chosen within each selected cell and that cells
are homogeneously distributed over the study area (Supplementary
material S1.2).
Rasterizing to a 16 × 16 km raster eliminated doubling of species
plots (e.g. plots were included in the national forest inventory of
Germany and in the pan-European tree occurrence dataset) because
only one data point per cell was used. Additionally, the filtering to a
coarser raster reduced autocorrelation in the data while at the same
time taking the rareness of species into account. There was no quantitative consideration of spatial autocorrelation in addition to the rasterization to 16 km.
The importance of soil for species distribution models has been
shown in many studies (e.g. Boulangeat et al., 2012; Dolos et al., 2015;
Piedallu et al., 2016). Nevertheless, the use of soil variables is always a
trade-off between spatial resolution and coverage of the data. We
decided to use the European Soil Database (ESDB, Panagos et al., 2012)
because its extent covers most of our analyzed data points (whole of
Europe and Russia without Turkey). We selected 11 variables from the
ESDB dataset: dominant parent material (PARMADO1), topsoil available water capacity (AWC_TOP), depth to rock (DR), volume of stones
(VS), depth to impermeable layer (DIMP), topsoil organic carbon content, texture (TEXT), top- and subsoil structure (STR_TOP, STR_SUB)
and presence of a raw peaty topsoil horizon (PEAT). The selection was
oriented according to hypotheses on water and nutrient availability as
well as on chemical properties that could weaken species’ vitality
(Mellert et al., 2011).
Additionally, we used a map of nutrient supply (NUTRI1). The map
is a classification of the ESDB soil type map into nine different nutrient
types (Kolb, 2017; Supplementary material 3.1) and has the same
geographical extent as ESDB. These soil variables were included in the
statistical analyses as nominal predictors. In addition, a numeric pHvalue map was used with a resolution of 5 * 5 km (Reuter et al., 2008)
The pH-value map covers 26 member states of the European Union
(Bulgaria and Romania are not included) plus Norway, Switzerland and
Albania.
2.3. Climatic data set
To describe the climatic conditions, we used climate data from the
WorldClim 1.4 datasets, with a cell resolution of 1 × 1 km (accessed
May 2017 – Hijmans et al., 2005). The data represent climate means for
1961–1990, which coincides well with the age of the trees in the used
2.5. Statistical analyses
SDMs were created without dispersal limitation as most forests in
Europe are managed and adaptation of forests to climate change needs
488
Forest Ecology and Management 430 (2018) 485–497
E.A. Thurm et al.
chemistry (E)). Soil models were only considered if the improvement of
TSS was greater than 0.005.
To falsify our research hypothesis, we calculated the potential area
shift for each individual species. The potential distribution areas were
derived using the prevalence as a probability threshold. Guisan et al.
(2006) recommended a balanced dataset for the analysis of rare species.
To ensure comparability of the results, we also applied the equal
weighting of the data to the two major tree species F. sylvatica and P.
abies. The prevalence is 0.5 for all species due to the balanced dataset of
50:50 presences and absences.
To estimate how species gain or loss would behave under a nonchanging land-use assumption, we cropped the species distribution
today and in the future by the forest land based on forest-type mapping
of Kempeneers et al. (2011). We applied the distribution to all raster
cells with a proportion of forest land above 10%. This definition
overestimates the forest area (Forest Europe, 2015). We use this
threshold as the reference area because it describes the potential habitats of the categories ‘forest’ and ‘other wooded land’ area (defined in
FAO, 2001) in the Southern European region. These regions are highly
important for the investigated thermophilic species. In order to refer to
the actual forest cover of Forest Europe (2015), one would have to
multiply each raster cell area of the SDM output with the proportion of
forest land in this cell.
To investigate the distribution shifts of the species in different
biogeographical regions, we applied the spatial classification of the
biogeographical regions map of Europe (Roekaerts, 2002). The distance
of the potential distribution shift was calculated by the centroid of the
current and the future distribution. The direction results from the
vector of the two points. A logarithmic linear model was used to show
the relationship between the latitude of the centroid C and the forecasted area gain or loss G with the equation:
management assistance due to the high speed of the expected temperature rise until the end of century.
Species distribution models were performed with generalized additive models (gam) and linear regression models (lm). Overall, we used
19 different environmental variables within these models. The 19
variables were grouped in five ‘site characteristic’ classes (warmth (A),
chilliness (B), drought (C), soil physics (D) and soil chemistry (E)). To
avoid an overfitting of the models, we made a species-specific variable
selection and restricted models to a maximum of five variables (so that
from each group one variable is represented) and a minimum of three
variables (from the three climate classes A–C only one variable is represented from each).
SDMs were created in two steps because some of the soil property
variables are highly correlated with the climate variables at the
European scale (e.g. dominant parent material and temperature). In the
first step, a pure climate model (Eq. (2)) was generated with climate
variables of the site characteristic classes A-C (VarA−C ). We followed
Franklin and Miller (2010) and restricted the number of variables and
the degrees of freedom within our approach. We applied a generalized
additive model (binomial distribution, logit link function) with three
variables of the respective classes A, B and C. We generated a set of
models with all possible variable combinations and tested them against
each other (multiple regressions with constant number of variables in
the models). Models with highly correlated variables (Pearson’s
r2 > 0.7) (Aguirre-Gutiérrez et al., 2013) were excluded from the selection process. The best model was chosen by means of the Akaike
information criterion (Burnham and Anderson, 1998) and biological
plausibility of the results (partial effects in accordance with biological
principles, no obvious regional misclassification). Standard smoothing
terms were used with three degrees of freedom (3 knots) to restrict
SDMs to only sigmoid or optimum response functions. The smoothing
functions fn of the individual climate variables were fitted by thin plate
splines.
Pc = a + f1 (Var A) + f2 (Var B ) + f3 (Var C ) + εc
log(G ) = a1 + a2 ·(logC ) + ε
The applied log-log transformation allows the calculation of the
ratio of the areal loss to a complete loss of a species in Europe
(−100%).
The suitability index Sai (Eq. (6)) should relate the potential area
gain or loss with the maximum growth potential of a species. It was
determined as the product of the area gain Ag and the site indices of the
species Si within their range (species with the lowest min and the
greatest max unit). The Si is the mean site index of a species at the
reference age of 80 years within the frame of the respective species
distribution. The product of Ag and Si was used to avoid the overestimation of outliers. The greatest Sai represents the best suitability.
(2)
ε always represents the residual error of the respective model. For the
residuals of the climate model (Eq. (2)) εc was defined separately, because it was used in the second step to describe the influence of the soil
on the distribution of the species. The soil models (Eqs. (3) and (4)) fit
the residuals of climate models by the soil variables VarD&E for physical and chemical properties, respectively, with ordinary least squares
regression models (lm). an represents the coefficients of the fixed effects.
εc = a1 + a2 ·(Var D) + ε
(3)
εc = a1 + a2 ·(Var E ) + ε
(4)
(5)
Sai =
We built two different models because of different spatial coverings
of the physical and chemical soil parameters.
The probability of species occurrence (Pc + s ) was composed by the
sum of the climatic and the soil models Pc + s = Pc + εc . If no soil information for a spatial area was given only the prediction of the climate
model (Eq. (2)) was calculated.εc is the sum of the outcomes of Eqs. (3)
and (4). Whether the soil models were chosen (see ‘Explanatory
Variables’ in Table 1) depends on whether they improved the overall
model performance – measured by the true skill statistic (TSS)
(Allouche et al., 2006) – compared to the pure climate model. The TSS
is a sum of sensitivity (proportion of correctly predicted presences) and
specificity (proportion of correctly predicted absences) minus one. It
reaches values between −1 and +1, with 1 representing a perfect fit.
To derive characteristics of model performances bootstrapping was
done with 100 runs and a data split of 90:10. The TSS was calculated for
each 10% test data in each bootstrap run and averaged over all runs.
Differences in TSS between the pure climatic model and the combined
climate and soil models were used to select the soil models that maximize TSS (one from class soil physics (D) and one from class soil
Agmax −Agmin Simax −Simin
·
Agi −Agmin
Sii−Simin
(6)
To predict the growth performances of the individual species, we
applied site index models. The site index of an inventory point was
calculated by the ratio of maximum dominant tree height to the maximum site index curve. The species-specific maximum site index curves
(see Supplementary material 7.4) were derived with a ChapmanRichards (Richards, 1959) function based on the predominant and
dominant individual trees (Kraft social class 1 and 2). The adaption of
Chapman-Richards curves was made with a nonlinear 95% quantile
regression (Koenker and Park, 1996) (detailed description in Brandl
et al. (2018)).
In a further step, the species-specific site indices were put in relation
to the climate variables with a generalized additive model, with a
gamma error distribution and a log-link function. The climate variables
of individual site models were identical to the climate variables which
were already selected for the species distribution models.
Sii = a + f1 (Var A) + f2 (Var B ) + f3 (Var C ) + ε
(7)
All analyses were performed in the software environment of R 3.3.2
(R Core Team, 2016) with the R packages raster (Hijmans, 2016), mgcv
489
Forest Ecology and Management 430 (2018) 485–497
E.A. Thurm et al.
(Wood, 2011) for the generalized additive models (gam) and quantreg
(Koenker, 2018) for the nonlinear quantile regression (nlrq).
Table 2
Area of the analyzed species in Europe today and in the future under two climate change scenarios for the currently wooded regions.
Area (million km2)
3. Results
The summary results of model performances are shown in Table 1.
The applied method generated an average performance (true skill statistics) of 0.67 for the species distribution models. Average sensitivity
attained a value of 0.90 and average specificity was 0.77. All site index
models were significant and had an average R2 of 0.13 (see also Supplementary material S7).
Variable selection showed no specific pattern for broadleaf or coniferous species. The most frequently selected variable in the SDMs was
mean precipitation in the warmest quarter (Bio18 – 11 times), followed
by mean annual temperature (Bio1 – 8 times) as well as continentality
(Conti1 – 8 times) and dominant parent material (PARMADO1 – 6
times). The majority of the climatic curves show a bell-shaped course; a
few have an S-shaped course (most important curves in Supplementary
material 3.2). Only P. nigra was nearly linearly negatively related to
summer precipitation (Bio18), resulting from a high amount of presences with summer precipitations close to zero. The soil models reduced or raised the climatic probability by −0.15 to 0.15. The most
frequently selected soil variable PARMADO1 increased the probability
mainly on the younger soils (unconsolidated or eolian deposition, organic materials) and decreased the probability on older soils (igneous
or metamorphic rocks). No soil variables were selected for P. pyraster
and L. kaempferi. In general, soil variables did not significantly improve
the TSS of the climate models (mean +0.02 TSS). The greatest improvement was achieved for P. nigra and Q. rubra.
Distribution maps of individual species are presented in
Supplementary material 5 (an example is shown in Fig. 3). The mean
potential species distribution area of the 12 thermophilic species was
1.2 million km2 calculated with a probability of occurrence > 0.5
(Table 2). P. abies and F. sylvatica reached a potential area of
2.52 m km2 and 1.38 m km2. P. nigra showed the greatest potential
range of the thermophilic species with1.64 m km2.
Area gain and loss (%)
Latin name
Today
2070 RCP
4.5
2070 RCP
8.5
2070 RCP
4.5
2070 RCP
8.5
Abies grandis
Acer monspessulanum
Castanea sativa
Fagus sylvatica
Larix kaempferi
Picea abies
Pinus nigra
Prunus avium
Pyrus pyraster
Quercus cerris
Quercus pubescens
Quercus rubra
Robinia pseudoacacia
Ulmus laevis
Average
Including FS & PA
0.71
1.00
1.09
1.38
0.50
2.52
1.64
1.41
1.62
1.32
1.31
1.59
1.32
0.94
1.20
1.31
0.40
1.57
1.41
0.61
0.22
0.92
2.29
1.41
2.59
2.70
2.23
2.54
3.02
2.50
1.91
1.74
0.29
1.84
1.43
0.41
0.19
0.49
2.35
1.29
2.59
3.22
2.88
2.45
3.12
3.08
2.06
1.83
−44
58
30
−56
−57
−63
39
0
60
105
70
60
129
165
51.33
35.46
−60
85
32
−70
−63
−81
43
−9
60
144
120
55
137
227
64.27
44.30
Generally, the climate change scenarios RCP 4.5 and RCP 8.5 caused
a drastic reduction or increase of the potential area of some species. For
example, under the assumption of the RCP 4.5 scenarios U. laevis
benefited most from climate change. Its potential distribution area increased by 165% from 0.94 m km2 to 2.50 m km2. Great gains are also
found for R. pseudoacacia (+129%) and Q. cerris (+105%). These area
changes become even more extreme under the warmer scenario RCP
8.5. The greatest (potential) area losses under the scenario RCP 4.5
were found for P. abies (−63%), L. kaempferi (−57%), F. sylvatica
(−56%) and A. grandis (−44%). Under the RCP 8.5 scenario, these
losses increased further.
The influence of the climatic warming on individual biogeographic
regions can be seen in Table 3. The distribution area gain of the thermophilic species due to climate change is obvious for Alpine region
Fig. 3. Example of the modeled distribution of Castanea sativa (a) today and (b) in 2070 under a temperature increase of approx. 2.9 °C. The area of occurrence
(probability > 0.5) is colored with a red to orange color ramp (from 1.0 to 0.5 in 0.1 probability steps), while non-occurrence (probability < 0.5) is colored grey to
white (also in 0.1 probability steps).
490
Forest Ecology and Management 430 (2018) 485–497
E.A. Thurm et al.
Table 3
Percentage share of distribution for the 12 thermophilic species within each biogeographic region assuming the current land-use distribution. Italic numbers in the
last row include the comparative species F. sylvatica and P. abies. Area of the biogeographical regions summarizes the potentially occupied area of a region within the
given extent of Europe (see also Fig. 5; a table which summarizes all investigated species can be found in Supplementary material 6).
Biogeographical regions
Alpine
Anatolian
Arctic
Atlantic
Black Sea
Boreal
Continental
Mediterranean
Pannonian
Steppic
Averageb
Including FS & PA
a
b
c
Areaa (million km2) of biogeographical regions
Average share of biogeographical region’s area (%)
Today
2070
RCP 4.5
2070
RCP 8.5
18.7
14.0
0.0
43.2
33.4
0.4
37.3
29.5
67.0
39.4
22.2
24.0
40.2
19.7
0.0
37.4
28.8
18.2
57.1
19.2
47.2
41.6
34.1
31.4
43.7
20.1
0.0
29.8
25.0
34.2
51.1
12.7
31.6
27.5
36.2
32.5
0.5
0.1
0.0
0.4
0.1
1.9
1.5
0.6
0.1
0.2
5.4c
Area gain and loss (%)
2070
RCP 4.5
2070
RCP 8.5
21.5
5.8
0.0
−5.8
−4.6
17.9
19.7
−10.4
−19.8
2.2
11.9
7.4
24.9
6.1
0.0
−13.4
−8.4
33.9
13.8
−16.9
−35.4
−11.9
14.0
8.4
Based on currently wooded areas.
Average weighted by area of the biogeographical regions
Sum.
Fig. 4. Shifts of centroids of the species distribution (a, map) for the 14 analyzed species under the RCP 4.5 scenario and (b, barplot) the species-specific distribution
area gains and losses in the respective biogeographical regions, assuming the current land-use distribution. Points in the map represent the current centroid of
distribution, arrowhead shows the future centroid. Y-axes in the barplots depict an area of 200,000 km2. Abbreviation: A. grandis (AG), A. monspessulanum (AM), C.
sativa (CS), F. sylvatica (FS), L. kaempferi (LK), P. abies (PA), P. avium (PV), P. nigra (PN), P. pyraster (PP), Q. cerris (QC), Q. pubescens (QP), Q. rubra (QR), R.
pseudoacacia (RP), U. laevis (UL). (RCP 8.5 in Supplementary material 6).
491
Forest Ecology and Management 430 (2018) 485–497
E.A. Thurm et al.
Table 4
Shift of the centroids of the distribution areas. Standard deviation of average in
brackets. The species are sorted according to increasing northward shift of RCP
4.5.
Shift distance (km)
Species name
RCP 4.5
RCP 8.5
Castanea sativa
Acer monspessulanum
Fagus sylvatica
Picea abies
Ulmus laevis
Quercus cerris
Prunus avium
Quercus pubescens
Pyrus pyraster
Pinus nigra
Robinia pseudoacacia
Larix kaempferi
Abies grandis
Quercus rubra
Average
303
332
393
434
435
445
450
454
457
496
523
554
567
616
461 (83.2)
596
610
616
525
613
667
706
791
709
796
790
727
771
842
697 (91.4)
(RCP 4.5 = 21.5%; RCP 8.5 = 24.9%). Also, the boreal (RCP
4.5 = 17.9%; RCP 8.5 = 33.9%) and the continental region (RCP
4.5 = 19.7%; RCP 8.5 = 13.8%) show a potential improvement of climate conditions for nearly all thermophilic species. A great loss of
potential distribution area was modeled for Southern Europe, in particular for the Pannonian (RCP 4.5 = −19.8%; RCP 8.5 = −35.4%) and
the Mediterranean region (RCP 4.5 = −10.4%; RCP 8.5 = −16.9%).
Unexpectedly, the Atlantic region also lost (potential) distribution area
for the target species. Fig. 4 shows the area gain and loss for the scenario RCP 4.5 for the 12 + 2 analyzed species (RCP 8.5 in Supplementary material 6.3). Great distribution shifts into the boreal region
were calculated for species which originate from Central and Eastern
Europe like U. laevis, P. pyraster and P. avium, as well as the non-native
species R. pseudoacacia and Q. rubra. P. abies, one of the most important
boreal species, suffered a severe loss of area in this bioregion. The
continental region exhibited a different picture: some species benefited
from climate warming, mainly the Mediterranean species Q. cerris, Q.
pubescens, P. nigra as well as non-native R. pseudoacacia. F. sylvatica as
dominant species of Central Europe and the investigated conifers see a
considerable decline in their distribution area in the continental region.
The centroid of the species distribution and how it would shift under
changing climate for the different climatic scenarios is shown in Fig. 4a.
The average shift is 461 km under RCP 4.5 and 697 km under RCP 8.5
(see also Table 4). The shift of Q. rubra, A. grandis and L. kaempferi is
higher compared to C. sativa, A. monspessulanum and F. sylvatica. The
mean direction was north-northeast (27°). The species-specific differences in direction are small and mainly driven by the land-water distribution, continentality and the soil conditions. The shift of F. sylvatica
(393 and 616 km, respectively) and P. abies (434 and 525 km, respectively) is within the range of the other species.
In Fig. 5, we depicted the relationship between the latitudinal center
of the current distribution and the potential area gain or loss under RCP
scenarios 4.5 and 8.5. The regression is significant both for RCP 4.5
(p > 0.001, R2 = 0.62) and for RCP 8.5 (p > 0.001, R2 = 0.70).
Species with a distribution above the 48° latitude seem to lose potential
distribution under climatic warming.
The calculated site index models showed a mean site index of
21.5 m over all species and averaged for their respective species distribution areas (see Table 5). A. grandis (33.0 m) and L. kaempferi
(29.6 m) reach the greatest site indices today and A. monspessulanum
(8.7 m) and Q. pubescens (11.9 m) the smallest site indices (speciesspecific growth curves and site indices predicted to distribution can be
seen in Supplementary material 7). The average site index of all 14
species is constant under climate warming. Nevertheless, the
Fig. 5. Potential area gain of the investigated species under a temperature
warming of +2.9 °C (RCP 4.5 – orange curve) and +4.5 °C (RCP 8.5 – red
curve). Asterisks (***) mean that the underlying linear models are highly significant (p > 0.001, respective model parameter in Supplementary material 6).
To aid orientation, the latitudes of four cities are depicted. (For interpretation
of the references to colour in this figure legend, the reader is referred to the web
version of this article.)
Table 5
Mean site index at the tree age of 80 years for the different climate scenarios,
based on the species-specific distribution areas of the respective scenarios. The
species are sorted according to decreasing mean site index today.
Mean site index [80 years]
Species name
Today
RCP 4.5
RCP 8.5
Abies grandis
Larix kaempferi
Quercus rubra
Picea abies
Fagus sylvatica
Prunus avium
Ulmus laevis
Castanea sativa
Quercus cerris
Pinus nigra
Robinia pseudoacacia
Pyrus pyraster
Quercus pubescens
Acer monspessulanum
Average
33.0
29.6
27.6
26.7
26.3
22.7
21.7
21.4
20.9
20.4
18.1
12.5
11.9
8.7
21.5
34.1
28.7
25.5
28.6
25.9
21.8
23.0
20.2
21.4
20.7
18.5
12.4
11.0
9.4
21.5
35.5
28.8
25.0
30.3
25.6
21.6
24.5
20.2
21.7
20.5
18.8
12.1
11.0
9.4
21.8
investigation exhibited that under climate warming some species will
find enhanced growth conditions in their expected distribution (P.
abies, U. laevis, A. grandis) and for some species the conditions will
deteriorate (Q. rubra, C. sativa). Despite the almost constant average
growth performance of the distribution area in climate change, the
growth performance in the individual regions might change in the future.
The suitability of the thermophilic species as alternatives for P. abies
and F. sylvatica under climate change is shown in Fig. 6 combining area
gain and site index at 80 years. The best suitability would be located in
the upper right corner of the figure and it decreases with increasing
distance from this corner. U. laevis combined best growth performance
with potential area gain under climate warming, followed by Q. rubra,
Q. cerris and R. pseudoacacia. L. kaempferi, Q. pubescens and A.
492
Forest Ecology and Management 430 (2018) 485–497
E.A. Thurm et al.
average shift of 156 m per year−1 for the A1fi emission scenario
(comparable with RCP 8.5) and 164 m per year−1 for the B1 (comparable to RCP 4.5). These rates are much slower than our results, and
come from the consideration of density-dependent population processes
and seed dispersal. A study of Kullman (2008) used long-term, systematic monitoring data of tree line vegetation records in Sweden from
the early 20th century to 2005. He could show that thermophilic tree
species (Quercus robur, Ulmus glabra, Acer platanoides, Alnus glutinosa
and Betula pendula) had expanded their range in the order of 50–300 km
northwards. These drastic shifts documented from field data in a relatively short period with a comparably moderate warming rate suggest
that our results of 461–697 km could be realistic.
4.1.1. Natural migration rate
Before we discuss the variable selection of the species distribution
models, we briefly address an important assumption in our study. We
presumed that the investigated species were mainly planted by humans
so the natural dispersal speed is not a limitation factor. In contrast,
researchers such as Feurdean et al. (2013), who investigated post-glacial migration rates with the Northern Eurasian Plant Macrofossil Database, have come to the result that (natural unassisted) migration from
south to north occurred at a substantially lower pace (60–260 m yr−1)
than the potential migration in our results for modern climate change.
Saltré et al. (2013), however, provided evidence that maximum postglacial migration rates could have reached 630 m yr−1 (mean
280 m yr−1) in the case of F. sylvatica. Feurdean et al. (2013) highlighted the importance of outlier populations for speeding up the migration rate. They also concluded that the cultivation of tree species
beyond their natural range has a similar effect as microrefuges had in
post-glacial times. At least for anthropogenically favored tree species
we can assume that the migration rate is independent of natural dispersal constraints (cf. also Kullman, 2008).
We found different migration rates for our investigated species. The
reason for this, however, is not the different dispersal strategies (Meier
et al., 2012) or the evolutionary adaptation of the dispersal strategy to
the existing flora and fauna (Myczko et al., 2014), but the variables and
their shape (S-shaped or bell-shaped). Nevertheless, it could be that
pioneer species, which are often anemochores, may select less strict
climate and soil variables or tend to S-shaped response curves which
allow a faster expansion under climate change conditions. In our data
we did not find such a pattern.
Fig. 6. Potential area gain of the investigated species in relationship to the
mean site index of the species within the respective species distribution under a
temperature warming of +2.9 °C (RCP 4.5; figures for +4.5 °C, RCP 8.5 and
respective data can be found in Supplementary material 7). The site index is
climate-sensitive and determined for a reference tree age of 80 years.
Abbreviation of species as in Fig. 4.
monspessulanum had the lowest suitability among the 12 considered
species. P. abies and F. sylvatica also showed an unfavorable relationship
between growth potential and potential area loss.
4. Discussion
The question of which tree species we should promote in the face of
climate change is an ongoing debate in forest ecology and management
(e.g. Albert et al., 2017). From a total of 12 thermophilic and rare tree
species we distinguished tree species like U. laevis, Q. rubra, Q. cerris
and R. pseudoacacia that – from the viewpoint of their climatic-edaphic
requirements and growth potential – are able to satisfy the expectations
of the Central Europe forest sector and those that are not (A. monspessulanum, L. kaempferi, A. grandis).
4.1. Northward shift
In our first hypothesis we claimed that the northward shift of the
investigated species is comparable. According to our results, however,
this is clearly not the case: some species distributions exhibit a shift of
approx. 300 km (C. sativa, A. monspessulanum), while other species like
A. grandis or Q. rubra shift up to 600 km (distances for +2.9 °C scenario). On average, the species move approx. 461 km northwards under
climatic warming of 2.9 °C and 697 km under 4.5 °C.
Among studies that can be compared to ours, McKenney et al.
(2007) investigated the change in the latitude of climate envelopes of
130 North American trees species and found an average shift of 700 km.
The results summarized the emission scenarios A2 and B2 (Nakicenovic
et al., 2000), which assume drastic and moderate climate warming,
respectively. Hamann and Wang (2006) also focused on North American species distributions. For some tree species they came up with a
potential habitat shift of at least 100 km per decade northwards.
However, it is generally difficult to compare such study results. The
magnitude of the shift is always related to the applied scenarios. Besides, the used explanatory variables have a great impact because some
are more stable under climate change projection than others, which
could also explain the different shifts.
Also for Europe several studies modeled potential northward shifts
of tree species (e.g. Hanewinkel et al., 2013; Meier et al., 2012; Thuiller
et al., 2006) but the presentation of precise migration distances is rare.
Meier et al. (2012) calculated the realistic migration rate for 14 European tree species with the tree species model TreeMig. They found an
4.1.2. Migration constraints
Apart from the dispersal speed, our results showed – in accordance
with other studies – that the migration rate is not a mere function of
average temperature and precipitation. Among others, unsuitable soil
conditions and late frost are mentioned in literature as important migration constraints (e.g. Aitken et al., 2008; Coudun et al., 2006; Falk
and Hempelmann, 2013). Land-use effects (Meier et al., 2012) and
geographic barriers like the decreasing land area towards Northern
Europe also affect the migration rate. Forest density increased from
south to north, in the same direction as the areal shift of the species. In
Mediterranean regions, the land-use effect can be considerable and
endangers these sparsely wooded areas (García-Valdés et al., 2015).
However, forest distribution is a variable factor (the forest area in
Europe has expanded by 17.5 million ha over the last 25 years, Forest
Europe, 2015), and changes in climate and society will certainly modify
this development in the future (e.g. Gehrig-Fasel et al., 2007).
In our study we integrated soil conditions on a coarse scale by a
residual analysis. A joint model together with climate failed because
some soil parameters were too strongly correlated with climate on the
continental scale. For example, PARMADO1 class 6 ‘unconsolidated
glacial deposits and glacial drift’ can be found predominantly in
Northern Europe and therefore correlates with lower temperatures. The
generated distribution maps based on joint models performed well with
training data but transferability into the future was not given.
493
Forest Ecology and Management 430 (2018) 485–497
E.A. Thurm et al.
decisive factor for a species to win or lose distribution area under
warming is closely related to the latitudinal center of the current distribution and the associated temperature adaption. In a rough generalization, the third hypothesis is correct in the sense that the area gain
is stronger for the southern tree species than for the northern ones. This
is in line with the findings of Dyderski et al. (2018). The land area in
Europe increases up to 48°N and decreases with higher latitudes, which
also influences the area gain or loss of species (e.g. P. abies has no area
to escape to when temperatures rise and distribution shifts northward).
Other studies which focused on mountain areas showed that climate
change led to an upward shift of most of the investigated plant species
(e.g. Lenoir et al., 2008; Penuelas and Boada, 2003; Rigling et al., 2013;
Walther et al., 2005). In terms of available land area the situation is
comparable with the northward shift in Europe: the potential distribution area towards the mountaintop decreases just like towards
Northern Europe.
U. laevis exhibited the greatest area gain under climatic warming.
Compared to the distribution map of Collin (2003), it can be seen that
our distribution range is smaller, especially at the western border. In
Western Europe, the rareness of the species led to a grouping of all three
Ulmus species to one Ulmus spp. class. Because of the different site requirements of the three species we decided to ignore these plots. This
makes it more difficult to model the western border correctly. How
much the species really spreads is difficult to predict. U. laevis was
determined to be a species with very limited competitive strength
which can only dominate in riparian forest. However, in Southern Asia
it is also well established in steppic areas, like in Finland. Currently,
very little is known about this species (Collin, 2003) and the results
show that further investigations are necessary.
Nevertheless, with the residual approach we showed that the integration of soil led to a significant improvement of the models. Other studies also found a significant model improvement due to the integration
of soil (Coudun et al., 2006). In Piedallu et al. (2016), for instance, the
prediction of tree species distribution in French national forest inventory data improved significantly when soil parameters were included (nutrient availability, the pH value and the C/N proportion). In
our study, for example, the probability of occurrence of Q. rubra decreased in calcareous soils and shallow soils (< 40 cm). This is in accordance with Meredieu et al. (1996), who found in field experiments
that the survival of young Q. rubra improved in non-calcareous, deep
soils.
In our variable selection, categorical parameters like dominant
parent material (PARMADO1) and nutrient availability (NUTRI1)
where more often selected than the numerical pH units. Of course, the
parameters are correlated to a certain degree, especially pH and nutrient availability (NUTRI1). Mellert et al. (2018) also performed
models based on the NUTRI1 variable. They assumed that the categorical variable NUTRI1 involves a conglomeration of soil parameters
important for plant physiologically, which may explain why PARMADO1 fits better to the soil preferences of the species than pH.
We already pointed out the importance of late frost events in determining the cold distribution limits of tree species. The late frost
sensitivity of trees depends on several parameters such as the species’
leaf flushing, the sensitivity of the buds and the threshold of freezing
(for additional information see Vitasse et al., 2014). A study by Muffler
et al. (2016) showed that the late frost resistance is highly correlated
with the continentality of the species’ distribution. Especially for the
thermophilic species late frost has to be considered in distribution
models, so we tried to capture late frost resistance by the inclusion of a
continentality index, which was chosen within the automated variable
selection for eight of the 14 species. Late frost risk is expected to change
more slowly than annual temperature because late frost events are influenced by atmospheric circulations (Becker et al., 2014; Meehl et al.,
2000; Strong and McCabe, 2017). Conrad’s continentality index is
nearly constant under RCP scenarios 4.5 and 8.5 because it includes the
latitude, which doesn’t change, and also the annual temperature range,
which is relatively stable (see Eq. (1)). Thus, models using this index are
conservative with regard to late frost events.
4.2.1. Non-native species
A special section is dedicated to non-native species among the selected tree species since extensive silvicultural experience is available
from the regions of their origin. A species which is expected by forest
management in Central Europe to be an attractive alternative under a
drier climate situation is Abies grandis (Fletcher, 1986; Liesebach and
Weissenbacher, 2007). However, our results showed a drastic decrease
of its potential range. The reason is that it is bound to a relatively small
and northern distribution in Europe with moist, oceanic climate conditions. In North America, A. grandis has a much greater distribution
range from the dry regions of the Blue Mountains in Oregon to the moist
coastal regions of British Columbia. The question is: How good is a SDM
which was supplied with species data from plantations far from its
native distribution range? Our model, which is only based on European
data, may be too restrictive on the non-native species. This problem has
already been described by other authors (Boiffin et al., 2017; Booth,
2017; Broennimann et al., 2007). Boiffin et al. (2017) tested the
transferability of a North American species distribution model for
Pseudotsuga menziesii (Mirb.) Franco to Europe with moderate success.
Chakraborty et al. (2016) followed, finding that the low level of introduced genetic diversity matched only a fraction of the European
climate that would potentially be suitable for P. menziesii. Similarly, the
broad climatic range of A. grandis in North America is known to include
different adapted varieties or provenances. In European forests mainly
the Coast grand fir (A. grandis ssp. grandis) was introduced and not the
Interior grand fir (Abies grandis var. idahoensis). The problem is further
compounded by the selective and unknown importation of seed
200 years ago. Depending on biogeographic region and taking into
consideration provenances, the assessment of A. grandis might be different from our European perspective.
Larix kaempferi also showed a northern center of distribution with a
clear focus on oceanic regions. In the case of L. kaempferi, the problem is
not as critical as A. grandis because its home range in Japan is relatively
small with not such a great climatic gradient. Summer drought sensitivity is known (Pâques, 1992) and the resulting area loss under climatic warming seems to be plausible (Huang et al., 2017).
4.2. Areal winners and losers
Under this provocative title we address our second and third hypotheses, that (2) the potential distribution area of the thermophilic
tree species increases under climate change while it decreases for the
more temperate and boreal species F. sylvatica and P. abies and (3) the
potential area gain increases the further south a species is distributed
today.
Indeed, the present study showed a drastic decrease of the potential
distribution area for P. abies and F. sylvatica. The 63% loss of area for P.
abies lies in the same order as in several other studies such as Duputié
et al. (2014), Schüler et al. (2014) and Dyderski et al. (2018). The 56%
loss of F. sylvatica supports previous studies that also assumed a great
loss of habitat suitability for F. sylvatica (Maiorano et al., 2013; Meier
et al., 2012), contradicting others that supposed an increase in its distribution (Dyderski et al., 2018; Falk and Hempelmann, 2013). Differences to the study of Falk and Hempelmann (2013), who also used
GAMs as statistical models, are probably due more to the choice of
response variables (minimum winter temperature, precipitation and
temperature sum of vegetation period) and their shape than to presences, environmental data or climate change scenarios. The temperature sum of the vegetation period has a very smooth bell-shaped form,
whereas Bio5 (max. temperature of warmest month), which was often
chosen in our models, has a rather sharp bell shape.
As for the thermophilic and rare tree species, nine of the 12 were
modeled to experience a gain in their potential distribution. The
494
Forest Ecology and Management 430 (2018) 485–497
E.A. Thurm et al.
the area loss of P. abies and F. sylvatica. Dyderski et al. (2018) showed
that Q. petraea is able to preserve its distribution in Central Europe and
spread into the boreal zone under future climate conditions. According
to distribution models for Q. petraea by Schueler et al. (2014),
Walentowski et al. (2017) and Dyderski et al. (2018), Q. petraea would
be comparable to C. sativa or P. avium in our species ranking in Fig. 6.
For the other two non-European species considered in this study, Q.
rubra and R. pseudoacacia, the situation is different. They were planted
especially in drier zones of Europe, so climatic warming will support
their distribution in Europe. The found area gain of R. pseudoacacia is in
agreement with Li et al. (2014) and Dyderski et al. (2018).
4.3. Area gain and growth performance
Forestry promotion of a species in times of climate change depends
not only on species’ potential area gain but also on the ability of a
species to produce a good quality of wood in sufficient quantities. The
quality aspect was taken into account when selecting the tree species to
be investigated. As for the potential quantity, we developed a suitability
index which merges growth potential and the expected area gain or
loss. Fig. 6 is a simplified, descriptive overview and ignores to a certain
degree the different growth courses, soil conditions and potential stand
densities (see Pretzsch, 2009). Nevertheless, the trend is obvious: at the
species level better adaptation to a warm and dry climate comes at the
cost of (stem height) growth, e.g. due to lower leaf area index and
duration, higher leaf thickness and a higher root-to-shoot ratio (Bréda
et al., 2006). Those species which might lose distribution area under
future climate scenarios, like P. abies, F. sylvatica, L. kaempferi or A.
grandis, show the greatest site indices, while those which might gain
area typically achieve lower site indices. As shown in Fig. 6, the species
U. laevis, Q. rubra, Q. cerris and R. pseudoacacia seem to provide a
reasonable trade-off between potential area expansion and growth
performance. In the case of Central Europe, P. nigra and C. sativa are
also attractive alternatives, especially under unfavorable soil conditions.
The good results of Q. rubra can be explained by its high drought
tolerance typical of oaks (Abrams, 1990) and its better growth performance in comparison to native European oaks (Major et al., 2013;
Seidel and Kenk, 2003). Despite the lower growth performance of Q.
cerris, it reached a comparable suitability index. The reason for this is its
great climatic tolerance. It is known that Q. cerris is drought resistant
(Nardini et al., 1999) as well as relatively frost hardy (Bussotti, 1994).
This combination allows her to achieve the greatest expansion into the
continental region under climatic warming, compared to the other investigated species (see Fig. 4).
5. Conclusion
Twelve thermophilic and rare tree species were investigated with
respect to their future potential distribution and growth performance in
Europe under a moderate and strong climate change scenario (+2.9 °C
and +4.5 °C). Some tree species like U. laevis, Q. rubra, Q. cerris and R.
pseudoacacia showed a great potential to contribute to future wood
production in Central Europe while others are less suitable (A. monspessulanum, L. kaempferi, A. grandis). U. laevis as the most suitable alternative needs more attention in distinguishing the different elm species in field surveys in Western Europe. Based on our findings
experimental trials are needed to test species requirements at its predicted margins and to increase our knowledge about the species.
Among the investigated species, Q. rubra ranked as a very suitable
alternative for P. abies and F. sylvatica. But at the same time Q. rubra is a
good example of a potential invasive species, and therefore the subject
of much discussion between forestry and nature conservation bodies
(Krumm and Vítková, 2016; Vor, 2015). Clearly, SDM studies only
cover selected aspects to evaluate the possible promotion of tree species
that are currently still rare, but the climate-edaphic suitability is fundamental before new investigations deal with other aspects. Now, one
task is to investigate the potential effects of an early promotion of tree
species on existing communities (e.g. Kuehne et al., 2014) within the
detected time frame.
Additionally, further investigations are needed to extend the suitability index to a higher complexity of information, e.g. considering the
biomass growth performance. Mortality models for instance can specify
the risk inside the species’ climatic distribution range (Neumann et al.,
2017). Approaches like in Biber et al. (2015) can integrate ecosystem
services on a quantitative basis. Regardless of whether introduced or
native species, the anthropogenic promotion of the species establishes
microrefuges (Feurdean et al., 2013). These refuges will promote an
increased ‘natural’ migration rate of the species (with possible invasive
behavior). On the one hand, these microrefuges might help forestry to
adapt to the speed of climate change; on the other hand they bear the
risk of producing unstable ecosystems which collapse under unexpected
abiotic and biotic diseases. For forestry practice, it is of supreme importance to find successful species combinations which minimize such
risks. Mixtures are not only the sum of two species’ trades but a process
of facilitation/complementarity and competition (e.g. del Río et al.,
2017; Thurm and Pretzsch, 2016). The species ranking in this study is a
good starting point for initializing trials with different combinations of
tree species with high future potential for the forest sector.
4.3.1. Non-native species with high suitability
We are aware that the recommendation of a species, especially a
non-native one, must be part of a broader context than just the site
requirements and growth performance (Hasenauer et al., 2016). Besides
the economic impact of a forced forest conversion, it has to be kept in
mind that the introduction of non-native species comes with an invasion potential (Richardson, 1998). Q. rubra is only one example of how
intensive planting of a non-native species is responsible for an invasion
in some European forest communities (Woziwoda et al., 2014). Assuming that silvicultural options to efficiently control invasive species
exist (Vítková et al., 2017), these species should not be excluded a
priori (regarding their great growth performance). Due to the necessity
to maintain them under silvicultural control, they should be introduced
in semi-natural managed forests rather than in more natural forest
communities.
Also, species like C. sativa (as well as other expanding species) will
migrate into existing forest communities where they are not native
today. Millar et al. (2007) described the result of this development as
neo-native forests. It results from the changing climate and has to be
observed carefully, too.
The establishment of neo-native forest communities, as well as the
promotion of non-native species, always comes with ecological and
economic uncertainties – more so with non-native species than with
neo-native species (Bolte et al., 2010). At the same time native tree
species that were not considered in this study, like Abies alba, Pinus
sylvestris and Quercus petraea, are widespread in Europe and may buffer
Acknowledgments
The study was funded by the Bavarian State Ministry of Food,
Agriculture and Forestry (project number B76). The evaluation was
based on data that were collected by partners of the official UNECE ICP
Forest Network (http://icp-forests.net/contributors). Part of the data
was co-financed by the European Commission (data achieved on 2 June
2016). We are grateful for the helpful comments of the two anonymous
referees. We would like to thank Tobias Mette, Birgit Reger and
Susanne Brandl for supporting the analysis. Additionally, we thank
Rudolf May, Stefan Tretter and Patrizia Gasparini, who helped collect
the data.
495
Forest Ecology and Management 430 (2018) 485–497
E.A. Thurm et al.
Author contributions
Dolos, K., Bauer, A., Albrecht, S., 2015. Site suitability for tree species: is there a positive
relation between a tree species'occurrence and its growth? Eur. J. For. Res. 134,
609–621.
Duputié, A., Zimmermann, N.E., Chuine, I., 2014. Where are the wild things?: why we
need better data on species distribution. Global Ecol. Biogeogr. 23, 457–467.
Dyderski, M.K., Paź, S., Frelich, L.E., Jagodziński, A.M., 2018. How much does climate
change threaten European forest tree species distributions? Glob. Chang Biol. 24,
1150–1163.
Elith, J., Leathwick, J.R., 2009. Species distribution models: ecological explanation and
prediction across space and time. Annu. Rev. Ecol. Evol. Syst. 40, 677–697.
Eurostat, 2018. Industrial roundwood by species: Export in Euro.
Fady, B., Aravanopoulos, F.A., Alizoti, P., Mátyás, C., von Wühlisch, G., Westergren, M.,
et al., 2016. Evolution-based approach needed for the conservation and silviculture of
peripheral forest tree populations. Forest Ecol. Manage. 375, 66–75.
Falk, W., Hempelmann, N., 2013. Species favourability shift in europe due to climate
change: a case study for fagus sylvatica L. and Picea abies (L.) Karst. Based on an
ensemble of climate models. J. Climatol. 1–18.
Falk, W., Mellert, K.H., 2011. Species distribution models as a tool for forest management
planning under climate change: risk evaluation of Abies alba in Bavaria. J. Veg. Sci.
22, 621–634.
FAO, 2001. Global Forest Resources Assessment 2000: Main Report, Rome.
Feurdean, A., Bhagwat, S.A., Willis, K.J., Birks, H.J.B., Lischke, H., Hickler, T., 2013. Tree
migration-rates: narrowing the gap between inferred post-glacial rates and projected
rates. PloS one 8, e71797.
Fletcher, A.M., 1986. Abies grandis Provenance Experiments: Nursery Stage Results.
Forestry Commission Research and Development Paper No. 139.
Forest Europe, 2015. State of Europe’s Forests 2015: Europe's Status & Trends in
Sustainable Forest Management in Europe, Madrid.
Franklin, J., Miller, J.A., 2010. Mapping Species Distributions: Spatial Inference and
Prediction. Cambridge Univ. Press, Cambridge u.a. EBC – Ecology, biodiversity and
conservation (Reprint).
García-Valdés, R., Svenning, J.-C., Zavala, M.A., Purves, D.W., Araújo, M.B., Saura, S.,
2015. Evaluating the combined effects of climate and land-use change on tree species
distributions. J. Appl. Ecol. 52, 902–912.
Gehrig-Fasel, J., Guisan, A., Zimmermann, N.E., 2007. Tree line shifts in the Swiss Alps:
climate change or land abandonment? J. Veg. Sci. 18, 571–582.
Giesecke, T., 2016. Did thermophilous trees spread into central Europe during the Late
Glacial? New phytologist 212, 15–18.
Goberville, E., Beaugrand, G., Hautekèete, N.-C., Piquot, Y., Luczak, C., 2015.
Uncertainties in the projection of species distributions related to general circulation
models. Ecol. Evol. 5, 1100–1116.
Guisan, A., Broennimann, O., Engler, R., Vust, M., Yoccoz, N.G., Lehmann, A., et al., 2006.
Using niche-based models to improve the sampling of rare species. Conserv. Biol. 20,
501–511.
Guisan, A., Tingley, R., Baumgartner, J.B., Naujokaitis-Lewis, I., Sutcliffe, P.R., Tulloch,
A.I.T., et al., 2013. Predicting species distributions for conservation decisions. Ecol.
Lett. 16, 1424–1435.
Hamann, A., Wang, T., 2006. Potential effects of climate change on ecosystem and tree
species distribution in British Columbia. Ecology 87, 2773–2786.
Hanewinkel, M., Cullmann, D.A., Schelhaas, M.-J., Nabuurs, G.-J., Zimmermann, N.E.,
2013. Climate change may cause severe loss in the economic value of European forest
land. Nat. Clim. Change 3, 203–207.
Hasenauer, H., Gazda, A., Konnert, M., Mohren, G., Pötzelsberger, E., Spiecker, H., van
Loo, M. (Eds.), 2016. Non-native tree species for European forests: Experiences, risks
and opportunities. COST ACTION FP 1403 NNEXT Country reports, second ed.,
Vienna, Austria.
Hijmans, R.J., 2016. raster: Geographic Data Analysis. and Modeling.
Hijmans, R.J., Cameron, S.E., Parra, J.L., Jones, P.G., Jarvis, A., 2005. Very high resolution interpolated climate surfaces for global land areas. Int. J. Climatol. 25,
1965–1978.
Huang, W., Fonti, P., Larsen, J.B., Ræbild, A., Callesen, I., Pedersen, N.B., et al., 2017.
Projecting tree-growth responses into future climate: a study case from a Danish-wide
common garden. Agr. Forest Meteorol. 247, 240–251.
Kellomaki, S., Rouvinen, I., Peltola, H., Strandman, H., Steinbrecher, R., 2001. Impact of
global warming on the tree species composition of boreal forests in Finland and effects on emissions of isoprenoids. Glob. Chang Biol. 7, 531–544.
Kempeneers, P., Sedano, F., Seebach, L., Strobl, P., San-Miguel-Ayanz, J., 2011. Data
fusion of different spatial resolution remote sensing images applied to forest-type
mapping. IEEE Tran. Geosci. Remote Sensing 49, 4977–4986.
Köble, R., Seufert, G. (Eds.), 2001. Novel maps for forest tree species in Europe.
Koenker, R., 2018. quantreg: Quantile Regression.
Koenker, R., Park, B.J., 1996. An interior point algorithm for nonlinear quantile regression. J. Econom. 71, 265–283.
Kohler, M., Sohn, J., Nägele, G., Bauhus, J., 2010. Can drought tolerance of Norway
spruce (Picea abies (L.) Karst.) be increased through thinning? Eur. J. For. Res. 129,
1109–1118.
Kolb, J., 2017. Nutrient status of Europe based on European soil database (unpublished
data).
Kölling, C., 2017. Non-native tree species: German-language original (Nichtheimische
Baumarten – Alternativen im klimagerechten Waldumbau? LWFaktuell 96, 2013).
www.waldwissen.net.
Kraft, G., 1884. Beiträge zur Lehre von den Durchforstungen, Schlagstellungen und
Lichtungshieben. Klindworth’s Verlag, Hannover.
Krumm, F., Vítková, L. (Eds.), 2016. Introduced tree species in European forests: opportunities and challenges. European Forest Institute, [Joensuu]. In focus – managing
forest in Europe.
ET did the analysis; WF designed the project; ET, LH, WF wrote the
article; AB, ER, SA revised the manuscript; AF, AB, BB, DH, ER, KB, LH,
RB, SA, TZ elaborated and harmonized the data.
Declarations of interest
None.
References
Abrams, M.D., 1990. Adaptations and responses to drought in Quercus species of North
America. Tree Physiol. 7, 227–238.
Aguirre-Gutiérrez, J., Carvalheiro, L.G., Polce, C., van Loon, E.E., Raes, N., Reemer, M.,
et al., 2013. Fit-for-purpose: species distribution model performance depends on
evaluation criteria – Dutch Hoverflies as a case study. PloS one 8, e63708.
Aitken, S.N., Yeaman, S., Holliday, J.A., Wang, T., Curtis-McLane, S., 2008. Adaptation,
migration or extirpation: climate change outcomes for tree populations. Evol. Appl. 1,
95–111.
Albert, M., Nagel, R.-V., Nuske, R., Sutmöller, J., Spellmann, H., 2017. Tree species selection in the face of drought risk—uncertainty in forest planning. Forests 8, 363.
Albrecht, A., Avila, A.L. de, 2018. Baumarten im Klimawandel: Artensteckbriefe – eine
Stoffsammlung. www.waldwissen.net.
Allouche, O., Tsoar, A., Kadmon, R., 2006. Assessing the accuracy of species distribution
models: prevalence, kappa and the true skill statistic (TSS). J. Appl. Ecol. 43,
1223–1232.
Aussenac, G., Granier, A., 1988. Effects of thinning on water stress and growth in Douglasfir. Can. J. For. Res. 18, 100–105.
Becker, P., Jacob, D., Deutschländer, T., Imbery, F., Namyslo, J., Müller-Westermeier, G.
et al, 2014. Klimawandel in Deutschland. In: Mosbrugger, V., Brasseur, G., Schaller,
M., Stribrny, B. (Eds.), Klimawandel und Biodiversität – Folgen für Deutschland. 2.,
unveränd. Aufl. WBG Wiss. Buchges, Darmstadt.
Biber, P., Borges, J., Moshammer, R., Barreiro, S., Botequim, B., Brodrechtová, Y., et al.,
2015. How sensitive are ecosystem services in European forest landscapes to silvicultural treatment? Forests 6, 1666–1695.
Boiffin, J., Badeau, V., Bréda, N., 2017. Species distribution models may misdirect assisted migration: insights from the introduction of Douglas-fir to Europe. Ecol. Appl.
27, 446–457.
Bolte, A., Ammer, C., Löf, M., Nabuurs, G.-J., Schall, P., Spathelf, P., 2010. Adaptive forest
management: A prerequisite for sustainable forestry in the face of climate change. In:
Spathelf, P. (Ed.), Sustainable forest management in a changing world. A European
perspective. Managing forest ecosystems 19. Springer, Dordrecht, pp. 115–139.
Bonan, G.B., 2008. Forests and climate change: Forcings, feedbacks, and the climate
benefits of forests. Science 320, 1444–1449.
Booth, T.H., 2017. Assessing species climatic requirements beyond the realized niche:
some lessons mainly from tree species distribution modelling. Climatic Change 145,
259–271.
Boulangeat, I., Gravel, D., Thuiller, W., 2012. Accounting for dispersal and biotic interactions to disentangle the drivers of species distributions and their abundances. Ecol.
Lett. 15, 584–593.
Brandl, S., Mette, T., Falk, W., Vallet, P., Rötzer, T., Pretzsch, H., 2018. Static site indices
from different national forest inventories: harmonization and prediction from site
conditions. Ann. For. Sci. 75, 739.
Bréda, N., Huc, R., Granier, A., Dreyer, E., 2006. Temperate forest trees and stands under
severe drought: a review of ecophysiological responses, adaptation processes and
long-term consequences. Ann. For. Sci. 63, 625–644.
Broennimann, O., Treier, U.A., Müller-Schärer, H., Thuiller, W., Peterson, A.T., Guisan,
A., 2007. Evidence of climatic niche shift during biological invasion. Ecol. Lett. 10,
701–709.
Burnham, K.P., Anderson, D.R., 1998. Model Selection and Inference: A Practical
Information-Theoretic Approach. Springer, New York.
Bussotti, F., 1994. Quercus cerris Linné. In: Schütt, P., Weisgerber, H., Lang, U., Roloff, A.,
Stimm, B. (Eds.), Enzyklopädie der Holzgewächse. Handbuch und Atlas der
Dendrologie. VCH [Imprint]; John Wiley & Sons, Limited, Chichester, pp. 1–10.
Chakraborty, D., Wang, T., Andre, K., Konnert, M., Lexer, M.J., Matulla, C., et al., 2015.
Selecting populations for non-analogous climate conditions using universal response
functions: the case of Douglas-Fir in Central Europe. PloS one 10, e0136357.
Chakraborty, D., Wang, T., Andre, K., Konnert, M., Lexer, M.J., Matulla, C., et al., 2016.
Adapting Douglas-fir forestry in Central Europe: evaluation, application, and uncertainty analysis of a genetically based model. Eur. J. For. Res. 135, 919–936.
Collin, E., 2003. EUFORGEN Technical Guidelines for genetic conservation and use for
European white elm (Ulmus laevis), Rome.
Conrad, V., 1946. Usual formulas of continentality and their limits of validity. Eos. T. Am.
Geophys. Un 27, 663.
Coudun, C., Gégout, J.-C., Piedallu, C., Rameau, J.-C., 2006. Soil nutritional factors improve models of plant species distribution: an illustration with Acer campestre (L.) in
France. J. Biogeogr. 33, 1750–1763.
del Río, M., Pretzsch, H., Ruíz-Peinado, R., Ampoorter, E., Annighöfer, P., Barbeito, I.,
et al., 2017. Species interactions increase the temporal stability of community productivity in Pinus sylvestris-Fagus sylvatica mixtures across Europe. J. Ecol. 105,
1032–1043.
496
Forest Ecology and Management 430 (2018) 485–497
E.A. Thurm et al.
mountains (NE Spain). Glob. Change Biol. 9, 131–140.
Piedallu, C., Gégout, J.-C., Lebourgeois, F., Seynave, I., Bartha, S., 2016. Soil aeration,
water deficit, nitrogen availability, acidity and temperature all contribute to shaping
tree species distribution in temperate forests. J. Veg. Sci. 27, 387–399.
Pretzsch, H., 2009. Forest Dynamics, Growth and Yield: From Measurement to Model.
Springer, Berlin, London.
Pretzsch, H., Schütze, G., Uhl, E., 2013. Resistance of European tree species to drought
stress in mixed versus pure forests: evidence of stress release by inter-specific facilitation. Plant Biol. 15, 483–495.
Core Team, R., 2016. R: A Language and Environment for Statistical Computing. R
Foundation for Statistical Computing, Vienna, Austria.
Rais, A., van de Kuilen, Jan-Willem G., Pretzsch, H., 2014. Growth reaction patterns of
tree height, diameter, and volume of Douglas-fir (Pseudotsuga menziesii [Mirb.]
Franco) under acute drought stress in Southern Germany. Eur. J. For. Res. 133,
1043–1056.
Reuter, H., Rodriguez Lado, L., Hengl, T., Montanarella, L., 2008. Continental scale digital
soil mapping using European soil profile data: soil pH. In: Böhner, J., Blaschke, T.,
Montanarella, L. (Eds.), Hamburger Beiträge zur Physischen Geographie und
Landschaftsökologie Heft 19.
Richards, F.J., 1959. A flexible growth function for empirical use. J. Experiment. Botany
10, 290–301.
Richardson, D.M., 1998. Forestry trees as invasive aliens. Conserv. Biol. 12, 18–26.
Rigling, A., Bigler, C., Eilmann, B., Feldmeyer-Christe, E., Gimmi, U., Ginzler, C., et al.,
2013. Driving factors of a vegetation shift from Scots pine to pubescent oak in dry
Alpine forests. Glob. Change Biol. 19, 229–240.
Roekaerts, M., 2002. The biogeographical regions map of Europe. In: Basic principles of
its creation and overview of its development. European Environment Agency,
Copenhagen, pp. 17.
Ruiz-Labourdette, D., Schmitz, M.F., Pineda, F.D., 2013. Changes in tree species composition in Mediterranean mountains under climate change: Indicators for conservation planning. Ecol. Indic 24, 310–323.
Saltré, F., Saint-Amant, R., Gritti, E.S., Brewer, S., Gaucherel, C., Davis, B.A.S., et al.,
2013. Climate or migration: What limited European beech post-glacial colonization?
Global Ecol. Biogeogr. 22, 1217–1227.
Schäfer, C., Thurm, E.A., Rötzer, T., Kallenbach, C., Pretzsch, H., 2018. Daily stem water
deficit of Norway spruce and European beech in intra- and interspecific neighborhood under heavy drought. Scand. J. Forest Res. 1–34.
Schüler, S., Falk, W., Koskela, J., Lefevre, F., Bozzano, M., Hubert, J., et al., 2014.
Vulnerability of dynamic genetic conservation units of forest trees in Europe to climate change. Glob. Change Biol. 20, 1498–1511.
Schume, H., Jost, G., Hager, H., 2004. Soil water depletion and recharge patterns in
mixed and pure forest stands of European beech and Norway spruce. J. Hydrol. 289,
258–274.
Seidel, J., Kenk, G., 2003. Wachstum und Wertleistung der Eichenarten in BadenWürttemberg. AFZ-Der Wald 28–31.
Seidl, R., Rammer, W., Lexer, M.J., 2011. Adaptation options to reduce climate change
vulnerability of sustainable forest management in the Austrian Alps. Can. J. For. Res.
41, 694–706.
Strong, C., McCabe, G.J., 2017. Observed variations in U.S. frost timing linked to atmospheric circulation patterns. Nat. Commun. 8, 15307.
Thuiller, W., Lavorel, S., Sykes, M.T., Araújo, M.B., 2006. Using niche-based modelling to
assess the impact of climate change on tree functional diversity in Europe. Divers
Distrib. 12, 49–60.
Thurm, E.A., Pretzsch, H., 2016. Improved productivity and modified tree morphology of
mixed versus pure stands of European beech (Fagus sylvatica) and Douglas-fir
(Pseudotsuga menziesii) with increasing precipitation and age. Ann. For. Sci. 73,
1047–1061.
Thurm, E.A., Uhl, E., Pretzsch, H., 2016. Mixture reduces climate sensitivity of Douglas-fir
stem growth. Forest Eco Manag 205–220.
Vitasse, Y., Lenz, A., Körner, C., 2014. The interaction between freezing tolerance and
phenology in temperate deciduous trees. Front. Plant Sci. 5, 541.
Vítková, M., Müllerová, J., Sádlo, J., Pergl, J., Pyšek, P., 2017. Black locust (Robinia
pseudoacacia) beloved and despised: a story of an invasive tree in Central Europe.
Forest Ecol. Manage. 384, 287–302.
Vor, T. (Ed.), 2015. Potenziale und Risiken eingeführter Baumarten: Baumartenportraits
mit naturschutzfachlicher Bewertung. Univ.-Verl. Göttingen, Göttingen. Göttinger
Forstwissenschaften 7.
Walentowski, H., Falk, W., Mette, T., Kunz, J., Bräuning, A., Meinardus, C., et al., 2017.
Assessing future suitability of tree species under climate change by multiple methods:
a case study in southern Germany. Ann. For. Res. 60.
Walther, G.-R., Beißner, S., Burga, C.A., 2005. Trends in the upward shift of alpine plants.
J. Veg. Sci. 16, 541.
Wood, S.N., 2011. Fast stable restricted maximum likelihood and marginal likelihood
estimation of semiparametric generalized linear models. J. R. Stat. Soc. 73, 3–36.
Woziwoda, B., Potocki, M., Sagan, J., Zasada, M., Tomusiak, R., Wilczyński, S., 2014.
Commercial forestry as a vector of alien tree species – the case of Quercus rubra L.
introduction in Poland. Baltic For. 20, 131–141.
Kuehne, C., Nosko, P., Horwath, T., Bauhus, J., 2014. A comparative study of physiological and morphological seedling traits associated with shade tolerance in introduced
red oak (Quercus rubra) and native hardwood tree species in southwestern Germany.
Tree Physiol. 34, 184–193.
Kullman, L., 2008. Thermophilic tree species reinvade subalpine Sweden—early responses to anomalous late holocene climate warming. Arctic, Antarctic, Alpine Res.
40, 104–110.
Lebourgeois, F., Gomez, N., Pinto, P., Mérian, P., 2013. Mixed stands reduce Abies alba
tree-ring sensitivity to summer drought in the Vosges mountains, western Europe.
Forest Eco Manage 303, 61–71.
Lenoir, J., Gégout, J.-C., Marquet, P.A., de Ruffray, P., Brisse, H., 2008. A significant
upward shift in plant species optimum elevation during the 20th century. Science
320, 1768–1771.
Li, G., Xu, G., Guo, K., Du, S., 2014. Mapping the global potential geographical distribution of black locust (Robinia Pseudoacacia L.) using herbarium data and a
maximum entropy model. Forests 5, 2773–2792.
Liesebach, M., Weissenbacher, L., 2007. Erfahrungen mit Abies grandis in sommerwarmen Gebieten Österreichs. Forst u. Holz. 62, 19–20.
Maiorano, L., Cheddadi, R., Zimmermann, N.E., Pellissier, L., Petitpierre, B., Pottier, J.,
et al., 2013. Building the niche through time: using 13,000 years of data to predict
the effects of climate change on three tree species in Europe. Global Ecol. Biogeogr.
22, 302–317.
Major, K.C., Nosko, P., Kuehne, C., Campbell, D., Bauhus, J., 2013. Regeneration dynamics of non-native northern red oak (Quercus rubra L.) populations as influenced
by environmental factors: a case study in managed hardwood forests of southwestern
Germany. Forest Ecol. Manage. 291, 144–153.
Mauri, A., Strona, G., San-Miguel-Ayanz, J., 2017. EU-Forest, a high-resolution tree occurrence dataset for Europe. Sci Data 4, 160123.
McKenney, D.W., Pedlar, J.H., Lawrence, K., Campbell, K., Hutchinson, M.F., 2007.
Potential impacts of climate change on the distribution of North American trees.
BioScience 57, 939–948.
Meehl, G.A., Karl, T., Easterling, D.R., Changnon, S., Pielke Jr, R., Changnon, D., et al.,
2000. An introduction to trends in extreme weather and climate events: observations,
socioeconomic impacts, terrestrial ecological impacts, and model projections. Bull.
Am. Meteorol. Soc. 81, 413–416.
Meier, E.S., Lischke, H., Schmatz, D.R., Zimmermann, N.E., 2012. Climate, competition
and connectivity affect future migration and ranges of European trees. Global Ecol.
Biogeogr. 21, 164–178.
Mellert, K.H., Fensterer, V., Küchenhoff, H., Reger, B., Kölling, C., Klemmt, H.-J., et al.,
2011. Hypothesis-driven species distribution models for tree species in the Bavarian
Alps. J. Veg. Sci. 22, 635–646.
Mellert, K.H., Lenoir, J., Winter, S., Kölling, C., Čarni, A., Dorado-Liñán, I., et al., 2018.
Soil water storage appears to compensate for climatic aridity Soil water storage appears to compensate for climatic aridity at the xeric margin of European tree species
distribution. Eur. J. For. Res. 137, 79–92.
Meredieu, C., Arrouays, D., Goulard, M., Auclair, D., 1996. Short range soil variability and
its effect on red oak growth (Quercus rubra L.). Soil Sci. 161, 29–38.
Millar, C.I., Stephenson, N.L., Stephens, S.L., 2007. Climate change and forests of the
future: managing in the face of uncertainty. Ecol. Appl. 17, 2145–2151.
Montwé, D., Isaac-Renton, M., Hamann, A., Spiecker, H., 2016. Drought tolerance and
growth in populations of a wide-ranging tree species indicate climate change risks for
the boreal north. Glob. Chang Biol. 22, 806–815.
Muffler, L., Beierkuhnlein, C., Aas, G., Jentsch, A., Schweiger, A.H., Zohner, C., et al.,
2016. Distribution ranges and spring phenology explain late frost sensitivity in 170
woody plants from the Northern Hemisphere. Glob. Ecol. Biogeogr. 25, 1061–1071.
Myczko, Ł., Dylewski, Ł., Zduniak, P., Sparks, T.H., Tryjanowski, P., 2014. Predation and
dispersal of acorns by European Jay (Garrulus glandarius) differs between a native
(Pedunculate Oak Quercus robur) and an introduced oak species (Northern Red Oak
Quercus rubra) in Europe. Forest Ecol. Manage. 331, 35–39.
Nakicenovic, N., Alcamo, J., Grubler, A., Riahi, K., Roehrl, R.A., Rogner, H.-H., et al.,
2000. Special Report on Emissions Scenarios (SRES), a special Report of Working
Group III of the Intergovernmental Panel on Climate Change. Cambridge University
Press.
Nardini, A., Lo Gullo, M.A., Salleo, S., 1999. Competitive strategies for water availability
in two Mediterranean Quercus species. Plant Cell Environ. 22, 109–116.
Neumann, M., Mues, V., Moreno, A., Hasenauer, H., Seidl, R., 2017. Climate variability
drives recent tree mortality in Europe. Glob. Chang. Biol. 23, 4788–4797.
Panagos, P., van Liedekerke, M., Jones, A., Montanarella, L., 2012. European soil data
centre: response to european policy support and public data requirements. Land Use
Policy 29, 329–338.
Pâques, L.E., 1992. Performance of vegetatively propagated Larix decidua, L kaempferi
and L laricina hybrids. Annales des sciences forestières 49, 63–74.
Pâques, L.E. (Ed.), 2013. Forest tree breeding in Europe: Current state-of-the-art and
perspectives. Springer, Dordrecht u.a. Managing forest ecosystems 25.
Pearson, R.G., Dawson, T.P., 2003. Predicting the impacts of climate change on the distribution of species: are bioclimate envelope models useful? Global Ecol. Biogeogr.
12, 361–371.
Penuelas, J., Boada, M., 2003. A global change-induced biome shift in the Montseny
497