www.ecography.org
ECOGRAPHY
Ecography
265
––––––––––––––––––––––––––––––––––––––––
© 2020 The Authors. Ecography published by John Wiley & Sons Ltd on behalf of Nordic Society Oikos
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.
Subject Editor:  
David Nogués-Bravo 
Editor-in-Chief:  
Jens-Christian Svenning 
Accepted 8 October 2020
44: 265–292, 2021
doi: 10.1111/ecog.05285
44 265–292
Globally, treeline ecotones vary from abrupt lines to extended zones of increasingly 
small, stunted and/or dispersed trees. These spatial patterns contain information about 
the processes that control treeline dynamics. Describing these patterns consistently 
along ecologically meaningful dimensions is needed for generalizing hypotheses and 
knowledge about controlling processes and expected treeline shifts globally. However, 
existing spatial categorizations of treelines are very loosely defined, leading to ambigui-
ties in their use and interpretation. To help better understand treeline-forming pro-
cesses, we present a new framework for describing alpine treeline ecotones, focusing 
on hillside-scale patterns, using pattern dimensions with distinct indicative values: 1) 
the spatial pattern in the x-y plane: a) decline in tree cover, and b) change in the level 
of clustering. Variation along these dimensions results in more or less ‘discrete’ or ‘dif-
fuse’ treelines with or without islands. These patterns mainly indicate demographic 
processes: establishment and mortality. 2) Changes in tree stature: a) decline in tree 
height, and b) change in tree shape. Variation along these dimensions results in more 
or less ‘abrupt’ or ‘gradual’ treelines with or without the formation of environmental 
krummholz. These patterns mainly indicate growth and dieback processes.
Additionally, tree population structure can help distinguish alternative hypotheses 
about pattern formation, while analysing the functional composition of the ecotonal 
vegetation is essential to understand community-level processes, controlled by species-
specific demographic processes.
Our graphical representation of this framework can be used to place any treeline 
pattern in the proposed multi-dimensional space to guide hypotheses on underlying 
processes and associated dynamics. To quantify the dimensions and facilitate com-
parative research, we advocate a joint effort in gathering and analysing spatial patterns 
from treelines globally. The improved recognition of treeline patterns should allow 
more effective comparative research and monitoring and advance our understanding 
of treeline-forming processes and vegetation dynamics in response to climate warming.
A global framework for linking alpine-treeline ecotone patterns 
to underlying processes
Maaike Y. Bader, Luis D. Llambí, Bradley S. Case, Hannah L. Buckley, 
Johanna M. Toivonen, J. Julio Camarero, David M. Cairns, Carissa D. Brown, Thorsten Wiegand and 
Lynn M. Resler
M. Y. Bader (https://orcid.org/0000-0003-4300-7598) ✉ (maaike.bader@uni-marburg.de), Univ. of Marburg, Faculty of Geography, Ecological Plant 
Geography, Marburg, Germany. – L. D. Llambí, Univ. de Los Andes, Facultad de Ciencias, Inst. de Ciencias Ambientales y Ecológicas, Mérida, Venezuela. – 
B. S. Case (https://orcid.org/0000-0002-4360-335X) and H. L. Buckley (https://orcid.org/0000-0002-4170-080X), Auckland Univ. of Technology, School of 
Science, Auckland, New Zealand. – J. M. Toivonen (https://orcid.org/0000-0002-0539-039X), Univ. of Turku, Biodiversity Unit, Turku, Finland. – J. J. 
Camarero (https://orcid.org/0000-0003-2436-2922), Inst. Pirenaico de Ecología (IPE-CSIC), Zaragoza, Spain. – D. M. Cairns (https://orcid.org/0000-0003-
4110-196X), Texas A&M Univ., Dept of Geography, College Station, TX, USA. – C. D. Brown (https://orcid.org/0000-0002-1587-6641), Memorial Univ., 
Dept of Geography, St. John’s, Canada. – T. Wiegand (https://orcid.org/0000-0002-3721-2248), Helmholtz Centre for Environmental Research – UFZ, Dept 
of Ecological Modelling, Leipzig, Germany. – L. M. Resler (https://orcid.org/0000-0002-5135-1797), Virginia Tech, Dept of Geography, Blacksburg, VA, USA.
Review and synthesis
266
Keywords: alpine treeline, climate change, pattern–process 
relationships, spatial pattern, timberline, vegetation 
structure
Introduction
Why is a globally consistent description of treelines 
useful and needed?
Recognizing globally recurrent vegetation physiognomies 
has organized our perception of the natural world and has 
facilitated communication among ecologists and biogeogra-
phers at least since the times of Alexander von Humboldt 
(von Humboldt and Bonpland 1807). It has enabled the for-
mulation of globally-applicable hypotheses about the factors 
and processes determining vegetation cover and has facili-
tated meaningful extrapolations of research finding across 
continents. Recognizing recurrent patterns in vegetation 
transitions or ecotones should provide similar advantages 
(Brown et al. 1996, Strayer et al. 2003). At vegetation bound-
aries, such as alpine treeline ecotones, spatial patterns may be 
a key to understanding the dynamic processes that determine 
the current and future distribution of species and ecosystems 
(Harsch and Bader 2011).
The spatial patterns of treeline ecotones vary greatly and, 
to some extent, every treeline ecotone is unique, reflecting 
ecological processes produced by particular combinations of 
climatic conditions, the topographic template, species com-
position and disturbance history (Brockmann-Jerosch 1919, 
Holtmeier 2009). However, some broad treeline ecotone pat-
terns are repeated globally. In this context, it is important to 
distinguish the treeline ecotone from the potential treeline, 
i.e. the highest elevation at which tree growth is possible in 
a region before thermal conditions become unsuitable. Due 
to different combinations of stochastic, ecological and climatic 
processes, the space downslope of the potential treeline is rarely 
filled completely with trees. The realized treeline, a line con-
necting the uppermost trees of a certain size (sizes reviewed in 
Holtmeier 2009), is located somewhere within the transition 
zone between forest and low-stature alpine vegetation, i.e. the 
treeline ecotone. This transition zone (also referred to here as 
‘treeline’ for simplicity) is the focus of this paper.
At the global scale, the predictability of treeline elevation 
has long been recognized (von Humboldt and Bonpland 
1807, Brockmann-Jerosch 1919, Hermes 1955) and can 
be captured well by the mean growing-season temperature 
and a minimum growing-season length (Körner and Paulsen 
2004). The global pattern in potential treeline elevation is 
therefore well explained by latitude and regional thermal 
modifiers such as mountain mass, continentality and cloudi-
ness (Paulsen and Körner 2014, Karger  et  al. 2019). We 
postulate here that spatial patterns of treeline ecotones may 
also be predictable globally, even if requiring more predictors 
than potential treeline elevation does. The reason we expect 
this predictability is that these patterns reflect a limited set of 
fundamental ecological processes that are repeated at treelines 
around the world. However, although some geographic pat-
terns in treeline ecotone structure have been recognized 
(Wardle 1965, 2008, Holtmeier 2009), the global distribu-
tion of these patterns is incompletely known and has not 
been systematically described, let alone explained. Mapping 
this distribution is useful, however, as it can form the basis 
for formulating hypotheses about controlling processes and 
for predicting treeline shifts once these hypotheses are suf-
ficiently tested.
Pattern–process relationships
Despite the large variation in treeline ecosystems, which occur 
in many different climate zones and biogeographic regions, 
treeline spatial pattern at the hillslope scale, or ‘treeline 
form’, exhibits globally recurrent characteristics that should 
be indicative of controlling processes and thus of expected 
dynamics under climate change (Harsch and Bader 2011, 
Treml and Veblen 2017). In individual-based models at this 
local scale, distinct treeline patterns (such as abrupt versus 
gradual) emerged only under certain combinations of pro-
cesses and parameters (Wiegand et al. 2006, Martinez et al. 
2011). The assumption is that these patter–process depen-
dencies in modelled ecotones reflect similar dependencies in 
real ecotones.
A meta-analysis of published studies suggested that 
treeline advance due to climate change differed among 
treelines with different spatial patterns, with abrupt treelines 
having advanced less than more gradual types (Harsch and 
Bader 2011). This difference can be explained theoretically 
by stabilizing positive feedback processes maintaining the 
abrupt boundaries (Armand 1992, Wilson and Agnew 1992, 
Bader  et  al. 2008), although other processes, like natural 
(Cairns and Moen 2004, Butler et al. 2009a) and anthropo-
genic (Miehe and Miehe 1994, Batllori and Gutierrez 2008) 
disturbances, offer alternative explanations. Such feedback 
processes are likely to also lead to temporal abruptness, i.e. 
stability potentially followed by fast changes (Armand 1992, 
Wilson and Agnew 1992, Bader et al. 2008). The abruptness 
is thus not the cause of the stability, but an indication of char-
acteristic processes that lead to spatial abruptness (e.g. posi-
tive feedback in spatial patterns of mortality). This indicative 
value is the main topic of this paper, while in some cases the 
spatial patterns can additionally be a cause of specific dynam-
ics, e.g. dispersed trees serving as seed sources and thus facili-
tating further tree establishment (Treml and Chuman 2015), 
or tree islands affecting snow redistribution and hence seed-
lings establishment (Hättenschwiler and Smith 1999).
One important impediment to drawing clearer connec-
tions between patterns and processes is the inconsistency in 
the descriptions of spatial patterns in the literature and the 
lack of clear definitions of treeline forms (e.g. gradual ver-
sus abrupt treelines). As a result, any grouping of treelines 
based on these descriptions is highly uncertain. To overcome 
this problem, we here propose a standardized description of 
treeline-ecotone spatial patterns at the hillslope scale based 
267
on clearly identifiable axes (dimensions) of variation in spatial 
topology that allows for consistent identification of treeline 
types and their transitions. We then propose how these 
patterns could be linked to basic growth and demographic 
processes as hypotheses that can be further explored via mod-
elling and experimental research. We do not aim to force 
treelines into discrete categories (although the extreme exem-
plary treeline forms are of value for verbal description) but 
rather propose to capture different aspects of the spatial pat-
tern along continuous axes that allows to place each treeline 
into a multi-dimensional ‘pattern’ space. Additionally, we 
propose a standardized terminology to avoid confusion about 
the meaning of terms like ‘abrupt’, ‘gradual’, ‘diffuse’, etc.
Once we can consistently describe treeline patterns and 
link them to the proposed hypotheses about underlying 
processes, causal links to specific mechanisms can be elabo-
rated in subsequent research (Brown et al. 1996). Once we 
understand in more detail how specific spatial patterns relate 
to underlying processes, we will be better able to interpret 
changes in spatial patterns as changes in controlling factors 
or processes (Martinez et al. 2011), and to use our process 
knowledge to predict treeline dynamics in accordance to their 
spatial pattern.
What efforts have already been made to describe 
treeline form?
Although the origin of different spatial treeline patterns has 
been discussed for over 100 yr (Ellenberg 1966, Frankhauser 
1901 and Schröter 1926 in Tranquillini 1979, references for 
Russian research in Armand 1992), these patterns remain 
only loosely defined. This discrepancy is probably due to the 
great variety of patterns found in different parts of the world, 
making one overarching descriptive framework a difficult, 
but worthwhile endeavor.
Holtmeier (2009), on the basis of extensive research on 
treeline physiognomy and landscape patterns in different 
parts of the world, distinguished four types of treeline, based 
on their spatial structure: abrupt forest limit, transition zone 
(ecotone), ‘true krummholz belt’ and gradual transition from 
high-stemmed forest to crippled trees of the same species. 
Thereby, ‘true krummholz’ is used for genetic krummholz, 
i.e. species that can grow only as shrubs. This is in contrast 
to the more common definition of krummholz as environ-
mentally-induced stunted and crippled trees, which we will 
use here. To avoid confusion, we suggest referring to genetic 
krummholz as shrubs, as already practiced in recent literature 
(Mayr et al. 2019, Treml et al. 2019). Holtmeier (2009, p. 
17) does not further differentiate the ecotonal treeline type 
but describes it as ‘characterized by a mosaic of tree clumps, 
scattered groves, isolated, more or less deformed tree indi-
viduals and treeless patches covered by low shrubs, herbs and 
grasses’.
In Harsch  et  al. (2009) and Harsch and Bader (2011), 
ecotonal treelines are differentiated into diffuse and island 
forms, i.e. those with predominantly isolated tree individuals 
and those with tree clumps and groves. Apart from these two 
types, abrupt and krummholz treelines were distinguished. 
A main disadvantage of this classification into four treeline 
forms (abrupt, diffuse, island and krummholz) is that ‘dif-
fuse’ is defined as a mixture of gradual decrease in tree cover 
and a decrease in tree size, leading to ambiguous predictions 
about the importance of growth responses and population 
processes such as establishment and mortality (Harsch and 
Bader 2011, Treml and Veblen 2017). Additionally, the 
krummholz in ‘krummholz treelines’ has been interpreted 
both as environmentally-induced and as genetic krummholz, 
again leading to confusion about the processes related to 
these forms (Schickhoff et al. 2015, Treml and Veblen 2017). 
A third disadvantage lies in the suggested discreteness and 
simplicity of the classification, when in fact treeline forms 
are very variable and differ along more dimensions than can 
be captured by four classes. For this reason, we here promote 
the idea of capturing different aspects of the spatial pattern 
along continuous axes, each quantifying a different ecologi-
cally meaningful dimension of the treeline pattern.
Wiegand et al. (2006) described treeline spatial patterns 
along three dimensions: tree height, adult tree density and 
tree age. The pattern was defined by the abruptness of changes 
along the ecotone according to these dimensions, which 
could be either ‘abrupt’ or ‘smooth’, with the goal of compar-
ing them to patterns emerging in different parameterizations 
of an individual-based model. Only certain combinations 
of processes allowed for the emergence of different treeline 
types, suggesting a clear signal in treeline pattern that allows 
for inference of the underlying processes. Wiegand  et  al. 
(2006) illustrate well why defining patterns in a meaningful 
way is necessary for studying pattern–process relationships at 
treelines, a concept we here develop further.
Objectives
We propose a standardized, globally-applicable description of 
spatial patterns occurring at alpine treelines at the hillslope 
scale (i.e. treeline form sensu Harsch and Bader 2011) that is 
based on the main dimensions of pattern indicative of underly-
ing processes and that allows adequately analyzing the spatio-
temporal dynamics of these patterns. Specifically, we aim to: 
1) propose a framework for describing alpine treelines based 
on different dimensions of their spatial patterns, 2) propose 
a standardized terminology for these patterns and 3) develop 
hypotheses for the general mechanisms behind these patterns.
The recognition of spatial similarities within the wide 
physiognomic variety exhibited by treeline ecotones world-
wide is essential for addressing generalised hypotheses about 
the functioning and local dynamics of this fascinating global 
phenomenon.
Pattern dimensions
Here, we present a framework for describing different dimen-
sions of treeline spatial patterns, at the hillslope scale, that are 
informative of underlying processes and should thus allow for 
268
a comparison of treeline sites that may show similar dynam-
ics at a particular point in time. A given treeline can take any 
position along these dimensions, although some combina-
tions are not possible. We provide labels for the far ends of 
the spectra to facilitate communication and we use these as 
examples to discuss the processes underlying the patterns (in 
general below, with examples in Box A).
We define the hillslope scale as an area a person would see 
when walking from the closed forest up into the alpine zone 
on a regular slope in moderately foggy conditions (sight of 
about 30 m). At this scale, the spacing of individual trees is 
of interest, and observable patterns include those related to 
surface microtopography, like trees established in the shelter 
of rocks (Resler et al. 2005), as well as self-organized patterns 
(Alftine and Malanson 2004, Bader et al. 2008, Zeng et al. 
2007). One scale level higher we will refer to as the ‘landscape 
scale’, i.e. the course of the ecotone through the topography 
(see below). At this scale, patterns like local north–south dif-
ferences and the effect of terrain concavity and convexity on 
ecotone elevation and form are described, as well as larger 
topography-related forest patches.
Two sets of pattern dimensions
The following scheme describes two types of pattern, each 
consisting of two pattern dimensions (axes in Fig. 1c, 2b), 
that describe treeline-ecotone form.
1) The spatial pattern of tree cover as seen from above (2-D 
in the x–y plane along the slope, Fig. 1). These patterns 
mainly indicate the demographic processes establishment 
and mortality (see below): demographic processes’). They 
can be described in the following two pattern dimensions:
a) Rate of change in tree density or cover from forest 
to alpine vegetation (Fig. 1c, x-axis): fastest if canopy 
cover does not decrease up to the tree-species line, i.e. 
the forest approaches the alpine in a closed front, so 
the treeline is discrete; slower if the canopy opens up 
into sparse cover at the tree-species line. Whether den-
sity (number of trees per area) or cover is evaluated will 
depend on the type of data available: remotely-sensed 
data will give a better image of tree cover, while field 
surveys often will be based on stem positions or num-
bers per area (i.e. densities). Both measures are valid for 
evaluating this dimension, but the interrelatedness of 
the two with tree size needs to be considered. If trees 
become smaller, the ratio between tree cover and stem 
density will decrease. Therefore, to maintain a stable 
tree cover, the stem density will need to increase.
b) If the canopy opens up towards the tree-species line, 
the level of clumping or dispersion of the trees (Fig. 
1c, y-axis) may stay similar to that in the forest (i.e. 
the treeline is diffuse) or trees may increasingly clump 
into islands (tree-islands or krummholz-islands).
2) Changes in tree stature and physiognomy (in a vertical 
cross-section across the ecotone, Fig. 2). These patterns 
mainly indicate growth and canopy dieback processes (see 
below). They can be described along the following two 
pattern dimensions:
a) Abruptness of the decline in maximum woody–plant 
height across the ecotone (Fig. 2b, x-axis): most abrupt 
if tall trees change directly to low-stature alpine veg-
etation or krummholz, most gradual if woody plants 
slowly decrease in size across the ecotone.
b) Deformation of the tree shape (Fig. 2b, y-axis): none 
if trees only get shorter, intermediate for e.g. multi-
stemmed or flagged upright trees, strong if tree species 
form environmental krummholz (Pereg and Payette 
1998). A species turnover into shrubs does not con-
stitute deformation, but if a belt of tall shrub species 
exists that changes into small prostrate shrubs (i.e. 
krummholz) of the same species, as seen sometimes 
in e.g. Pinus mugo or Rhododendron, this does consti-
tute deformation and this second transition should be 
noted.
For each of the dimensions we can quantify associated detailed 
measures and therefore formally place each treeline unambig-
uously into a multi-dimensional state space. Additionally, it 
is useful to attach to each treeline a name based on a simpli-
fied nomenclature (Table 1). For naming we suggest to use 
discrete and gradual as defaults, considering gradual discrete 
treelines (Fig. 3a) as our null model. We can thus distinguish 
gradual (gradual height and implied discrete density change; 
Fig. 3a), diffuse (implied gradual height and diffuse density 
change; Fig. 3b), abrupt diffuse (abrupt height and diffuse 
density change; Fig. 3e), abrupt (abrupt height and implied 
discrete density change; Fig. 3d), as well as e.g. tree-island 
treelines (implied gradual height change and tree islands; 
Fig. 3c), and abrupt krummholz-island treelines (abrupt 
height change with krummholz islands; Fig. 3l).
In many cases, treelines will not belong to the far ends 
of the pattern dimensions but will represent intermediate or 
mixed forms. These can be verbally described using the same 
terms, for example ‘diffuse with some islands’, or more pre-
cisely, e.g. ‘predominantly diffuse treeline with about 20% 
of trees clustered into islands (intermediate form) and with 
abrupt sections along ca 30% of its length (mixed form)’. The 
level of mixing will depend on the length of treeline that is 
described, e.g. one line transect along the slope (no mixtures, 
but intermediate forms possible) or a wider section of the 
treeline across a slope (mixtures and intermediate forms pos-
sible). With time, treeline form may also shift dynamically 
along the axes defined here from one form to another as the 
treeline-shaping processes change in relative importance due 
to population development or environmental change.
A simple theoretical framework for causes of 
treeline form
Treeline spatial pattern constitutes a snapshot in a potentially 
dynamic system and cannot replace a detailed monitoring 
of population processes and their drivers to fully understand 
269
how a treeline develops. However, patterns can be a power-
ful and efficient proxy for inferring the ecological processes 
that control treeline dynamics. We here follow the hypothesis 
of Harsch and Bader (2011) that variation in treeline form is 
essentially driven by three general processes: growth limitation, 
dieback (aboveground biomass loss) and mortality (including 
establishment failure). Each treeline spatial pattern dimension 
is indicative of particular spatial gradients of these three ‘first-
level’ processes (Fig. 4, Table 1), while a more detailed analy-
sis of the patterns, combined with further empirical evidence, 
may reveal what underlying (‘second-level’) processes cause 
these gradients. ‘Third-level’ processes include species interac-
tions modifying the second-level processes, e.g. through shel-
tering or shading (Harsch and Bader 2011).
In terms of first-level processes, spatial patterns of all 
forest borders will indicate the same processes. Our frame-
work is thus potentially more widely applicable than at 
alpine treelines alone. For example, very abrupt boundaries 
are found between forest and grassland in the Indian Ghats 
(Joshi et al. 2019) and in fragmented Brazilian Atlantic forest 
(Silva and Anand 2011), indicating high tree mortality out-
side established forest, with the second-level processes vary-
ing from frost damage to fire.
Second-level processes, on the other hand, can vary widely 
among ecotones with similar spatial patterns. These processes 
can be physiological or mechanical, brought about by different 
environmental factors, like frost or fire in the example above, 
both leading to abrupt treelines. At this level, there is thus no 
0
20
Tr
ee
 d
en
si
ty
Elevation
Discrete Diffuse Isotropic islands Anisotropic islands
0
1
C
lu
st
er
in
g
Diffuse
Islands
Discrete
slow fast
none
strong
Tree density change
In
cr
ea
se
 in
 c
lu
m
pi
ng
(a)
(b)
(c)
Figure 1. First set of dimensions of treeline spatial pattern: the 2-D pattern as seen from above. (a) Examples of treeline ecotone patterns as 
seen along elevational transects. (b) Changes in tree cover (black line) and the fraction of tree pixels with another tree pixel directly to their 
right (as a simple measure of clustering in the x direction; red dots) or to their top (clustering in the y-direction; blue crosses) across the 
ecotones shown in (a). The dark green arrows indicate the extent of the ecotones, which approaches a length of zero in the discrete treeline 
(left). (c) The two dimensions (axes) that define the 2-D pattern. The terms in the graphs denote the position of different treeline types at 
the far ends of the axes. The combination of a discrete treeline with increased clumping (top right of the graph) is not expected to exist.
270
one-to-one relationship between ecotone patterns, at the scale 
and level of detail we defined them here, and underlying pro-
cesses, but these patterns can certainly guide the search for the 
controlling second-level processes. In alpine treeline ecotones, 
typical second-level processes include, for example, reduced 
cell formation (due to a lack of warmth or nutrients), loss of 
foliage (due to e.g. frost or wind), snow molds (under a long-
laying snow pack), desiccation (due to frozen soils or summer 
drought), fire damage (due to forest fires or alpine-vegetation 
burning), dispersal limitation (due to poor seed production or 
an absence of vectors) and germination failure (due to various 
stressors). Many of these processes and their controlling factors 
are not specific to treeline, but they are nonetheless important 
determinants of past and future dynamics. Although differ-
ent second-level processes can result in similar spatial patterns, 
in many cases they will also leave specific spatial signatures. 
For example, if dieback due to wind damage is important, 
longitudinal krummholz islands may be formed (Alftine and 
Malanson 2004) and exposed sites will have stronger krumm-
holz formation than sheltered sites (Holtmeier and Broll 2010). 
In contrast, if dieback is caused by snow molds (Barbeito et al. 
2013), trees will be more stunted in sites that accumulate snow 
due to their topographic position or due to neighboring veg-
etation (Holtmeier 2005).
Figure 2. Second set of dimensions of treeline spatial pattern: changes in tree stature. (a) Examples of treeline ecotones patterns in the verti-
cal cross section. All treelines shown here are discrete in terms of their 2-D pattern (Fig. 1) and vary in their tree stature pattern only. (b) 
The two dimensions (axes) that define the cross-section pattern of tree stature. The terms in the graphs denote the position of different 
treeline types at the far ends of the axes.
Table 1. Summary of the framework, with the terms (in bold) proposed for the far ends of the four dimensions that describe spatial patterns 
of alpine treeline ecotones, and a summary of the processes which they may indicate to be important. As seen here, the framework is hier-
archical, with options for the type of change depending on the rate of change.
a) Rate of  
change
b) Type of  
change
Controlling  
processes (first-level) Controlling processes (second- and third-level)
2-D pattern Discrete Mortality above  
but not in the ecotone
Any type of damage or stress. If gradual, growth 
limitation above the ecotone
Ecotone Diffuse Mortality within  
the ecotone
Environmental heterogeneity or stochasticity. And/or 
ongoing seed-limited colonization
Island Mortality within  
the ecotone
Environmental heterogeneity, clustered seed dispersal, 
clonal reproduction, or positive feedback tree cover 
– microclimate via wind and snow redistribution 
(abrasion, snow moulds, short growing-season)
Tree stature Abrupt Into alpine Seedling mortality Overheating, desiccation, photoinhibition, freezing 
damage, soil limitations
Into krummholz Dieback Winter desiccation, snow moulds, browsing, abrasion
Gradual No krummholz Growth Tissue formation (limited by a lack of heat)
Into krummholz Growth and dieback Tissue formation + winter desiccation, snow moulds, 
browsing, abrasion
271
Seed rain and clonal reproduction are other important 
examples of second-level processes and can be strong drivers 
of spatial pattern. Although not technically a form of mortal-
ity (in contrast to e.g. seed predation and germination fail-
ure, Crofts and Brown 2020), a lack of seed rain or clonal 
dispersal is, at the very general first level, an alternative form 
of ‘mortality’, because both lead to an absence of trees. Many 
seed-dispersal types lead to spatial patterns that cannot be 
unequivocally distinguished from patterns caused by early-
stage mortality. For example, a decreasing tree density with 
distance above the forest line, indicating reduced establish-
ment success, can be caused by either a dispersal gradient 
or an environmental gradient causing a survival gradient. 
As another example, tight clumps of trees, if not related to 
topography, can indicate either seed caching by nutcrackers 
(Holtmeier 1966), layering (Laberge et al. 2000) or positive 
feedback between trees and their environment, all leading 
to a higher establishment in the clump and establishment 
failure outside of it. However, some specific spatial patterns 
almost certainly indicate an inhomogeneous seed distribu-
tion or clonal reproduction by layering or rhizomes, e.g. loose 
clustering of small trees around a larger focal tree indicat-
ing local seed dispersal or dispersal via rhizomes (Kullman 
2007). For most tree species, the reproductive strategies are 
Gradual treeline
(a)
growth limitation
dieback
mortality
Abrupt treeline
(d)
Krummholz treeline
(g)
Abrupt krummholz treeline
(j)
S
ev
er
ity
 o
f f
irs
t−
le
ve
l m
ec
ha
ni
sm
→
Diffuse treeline
(b)
Abrupt diffuse treeline
(e)
Diffuse krummholz treeline
(h)
Abrupt diffuse krummholz treeline
(k)
Elevation →
Tree−island treeline
(c)
Abrupt tree−island treeline
(f)
Krummholz−island treeline
(i)
Abrupt krummholz−island treeline
(l)
Figure 3. Variation in treeline form according to the dimensions presented in Fig. 1, 2: the 2-D spatial pattern (columns) and changes in 
tree size and shape (rows). The graphs show the hypothesized gradients in the first-level processes growth limitation (dotted orange), dieback 
(dashed orange) and mortality (including establishment failure, red line) leading to these patterns. Shown are simplified gradients and mini-
mal models, i.e. if a process is not needed to create the pattern, it is set to stay constant. Growth limitation is assumed to always occur, 
although sometimes it does not affect the pattern much (e.g. in d). The upper line indicates the most limiting process at any point along 
the elevational gradient. Sudden shifts in the stress severity (c, d, f, i–l) are due to positive feedback of the vegetation on environmental 
conditions: forest or krummholz ameliorating conditions for establishment (c, d, f, i, l) or trees escaping damaging conditions near the 
ground (j–l). Sinusoid lines (panels b, e, h, k) indicate environmental heterogeneity independent of the vegetation itself. The diffuse patterns 
in these panels could be caused by patchy environmental conditions or, especially in expanding populations, spatial and temporal stochas-
ticity in seed rain and mortality. This is a modification of the scheme in Harsch and Bader (2011), expanding it from four spatial types 
(abrupt, diffuse, island and krummholz) to twelve examples representing the extreme ends in a continuous space of pattern dimensions.
272
sufficiently known to distinguish seed-based from clonal 
processes, although care should be taken with assumptions 
(e.g. assumed clonal tree islands turned out to be non-clonal 
after genetic analysis, and even local seed dispersal seemed to 
play a small role, so that the patterns may have been due to 
microclimate-based positive feedback, Johnson et al. 2017). 
However, to distinguish the importance of dispersal limita-
tion from that of other processes leading to establishment 
failure, spatial patterns are not sufficient and more detailed 
experiments are needed (Crofts and Brown 2020).
Identifying spatial patterns according to the suggested four 
dimensions, and inferring spatial gradients in the first-level 
processes from these patterns, is the first step to understanding 
the drivers of treeline dynamics (Wiegand  et  al. 2006, 
Martinez et al. 2011). For addressing site-specific hypotheses 
about the underlying second-level processes and controlling 
environmental factors, a more detailed analysis of spatial pat-
terns at particular treelines will be needed. A strong aid to 
such pattern analyses is the age distribution of the trees, which 
can be obtained using dendrochronological methods based 
on dating tree-rings or based on growth scars for younger 
individuals (Hagedorn  et  al. 2014, Camarero  et  al. 2015, 
Liang et al. 2016, Vitali et al. 2017). Although laborious to 
obtain (and impossible for regions lacking seasonal climate 
variation), such information is invaluable for reconstructing 
the population dynamics within an ecotone, thereby provid-
ing important additional indications about the processes that 
control it. Even stronger patterns can be obtained if such data 
are combined with tree-ring-based data on growth responses 
to climate (Elliott 2011, Liang  et  al. 2016, Jochner  et  al. 
2017b).
The following sections are intended to help the interpre-
tation of observed patterns, along the two pairs of pattern 
dimensions, in terms of the underlying first-level processes 
growth limitation, dieback and mortality (Fig. 3). Further 
examples are given in Box A.
Change in tree cover (discrete–diffuse–islands):  
demographic processes
The horizontal 2-D spatial pattern (the treeline as viewed from 
above) is mostly determined by the presence and absence of 
trees across space. The absence of trees indicates that estab-
lishment has failed (e.g. due to a lack of viable seeds or safe 
sites) or trees have died and disappeared. In other words, 
mortality (including establishment failure) is a major deter-
minant of these patterns (Fig. 3). In contrast, growth and 
canopy dieback should have stronger effects on the pattern of 
tree stature change and are less likely to be important for the 
2-D spatial pattern, except where they interact with mortality 
(see below). Indeed, a recent test of climate responses in dif-
ferent treeline forms showed that treelines defined by either 
a diffuse or discrete 2-D pattern did not differ much in their 
growth response to warming. Instead, the diffuse treelines 
were controlled mostly by population processes. In contrast, 
treelines defined by the formation of krummholz (dimension 
2b: deformation) had a stronger growth response than non-
krummholz forms (Treml and Veblen 2017).
a) If there is no decline in tree cover with elevation, i.e. if 
the treeline is a discrete, closed boundary between forest and 
alpine vegetation (formed either by tall trees or with trees 
gradually getting shorter, either with a closed or more open 
canopy, but without changing canopy cover with elevation), 
this indicates that establishment of trees up to the tree-species 
line is possible and not increasingly limited with elevation 
(Fig. 3a, d, g, j, left column). When gradual (see below), such 
treelines are caused either by increasing growth reduction 
with tree size, or additionally by dieback limiting tree stature, 
indicated by a decline in tree height with or without defor-
mation (Fig. 3a, g, j, see below). When abrupt, such treelines 
are caused by a recent disturbance or by establishment 
Figure 4. Functional diversity patterns. (a) Example of a two-species 
treeline, with a deciduous and an evergreen conifer tree species (Larix 
decidua and Picea abies), both able to colonize open space, in a grass-
land and dwarf-shrub matrix in the Swiss Alps (photo: Matthias 
Jochner). (b) Example of a multi-species treeline with a gradual spe-
cies turnover: transition from mixed podocarp-broadleaved forest 
with, amongst others, the trees Halocarpus biformis and Libocedrus 
bidwillii to the transition zone with, amongst others, the small trees/
shrubs Dracophyllum spp. and Olearia spp. (Wardle 2008), and then 
increasingly small shrubs and tussock grassland on the western slopes 
of the southern Alps of New Zealand (photo: M. Y. Bader).
273
failure outside existing tree cover (in abrupt discrete treelines, 
Fig. 3d). Such an abrupt change in establishment chances is 
unlikely to be caused by underlying environmental gradients 
but more likely by a positive feedback between the vegeta-
tion and the microenvironment, which may include soil and 
microclimatic factors (Wilson and Agnew 1992).
Conversely, a decline in tree cover, be it into a diffuse or 
clumped (island) pattern, indicates that trees have not estab-
lished everywhere below the tree-species line or have been 
removed. Formulating this very basic interpretation is the 
first step towards explaining it. The following, not mutually 
exclusive, reasons may apply:
1) The treeline is advancing: the established trees are out-
posts and establishment is seed limited. If the number of 
dispersed seeds or the frequency of years in which estab-
lishment is possible is low, such an ‘advancing’ state can 
last very long.
2) Establishment is limited by the number of safe sites, 
which decreases with elevation and is related to: a) micro-
topographic features; or b) existing vegetation cover.
3) The established trees are remnants: the treeline has been 
opened up by disturbances, especially land use or the sur-
rounding trees have died due to climate deterioration.
These three reasons may result in superficially similar pat-
terns, but in a pure case of reason 1, the spatial pattern should 
be random or explainable by non-random seed rain patterns, 
the tree ages (not just sizes) should clearly decrease with ele-
vation, and outpost trees should be relatively young (Batllori 
and Gutierrez 2008, Elliott 2012). If outpost individuals do 
not reproductively mature, local establishment in the upper 
ecotone remains dependent on long-distance seed dispersal 
from lower-elevation source populations, slowing advance 
and maintaining this outpost-based diffuse pattern over long 
time periods (though long-distance dispersal appears to be 
common at some treelines Johnson et al. 2017). Such long-
lasting ‘advance’ patterns may also occur if, on top of limited 
seed availability, the number of years with climatic conditions 
allowing establishment is very low. In these two cases, outpost 
populations may be evenly or unevenly aged, depending on 
whether propagule dispersal upslope is sporadic or continu-
ous, and on the frequency of favorable years for establishment.
If reason 2 (also) applies, it should be evident that estab-
lishment patterns are associated to local vegetation cover 
(e.g. local cover of alpine shrubs facilitating establishment 
of woody species, Llambí  et  al. 2013, Bueno and Llambí 
2015), to topographic shelter (Batllori et al. 2009, Astudillo-
Sánchez  et  al. 2019), or to other, regionally-specific, types 
of microsites (Johnson and Yeakley 2019). In practice, these 
associations are not always easy to detect, as topography may 
become covered and vegetation replaced by developing trees. 
Reason 1 implies temporal heterogeneity combined with sto-
chasticity, while reason 2 implies spatial heterogeneity. In any 
case, of all spatial patterns, diffuse treelines are most charac-
terized by heterogeneity and stochastic processes.
The third reason may be suspected if the first two do not 
apply, i.e. if outpost trees are of high and well-distributed 
ages and if the soil and topography appear homogenous, 
and particularly if livestock grazing is apparent in the region 
(Ellenberg 1966). In many cases, an opening up due to land 
use will be accompanied by a recolonization (Vitali  et  al. 
2019). Similarly, current climate warming may lead to recol-
onization around climatic relict trees (Kullman 2007). In 
both cases current diffuseness would thus be due to a mix of 
reasons 1 and 3.
b) If the tree cover declines, trees can either thin out ran-
domly, over-disperse or cluster together. Random patterns 
or associations with microtopographic features indicate that 
safe sites (or seed rain patterns) are topographic or caused by 
interactions with alpine vegetation (Fig. 3b, e, h, k, middle 
column). In contrast, both over-dispersion and clustering 
indicate that, additionally, feedback processes involving exist-
ing tree cover shape and maintain these patterns, indicating 
negative and positive feedback respectively (Malanson 1997, 
Bader et al. 2008).
Over-dispersion could indicate that trees compete with 
each-other (e.g. for water), or otherwise negatively affect 
each other. Körner (1998) suggested that cooler soils under 
closely-spaced trees suppresses tree root growth, which would 
explain why isolated trees can be found at higher elevations 
than closed forest. Such shading could thus cause a negative 
feedback that could lead to over-dispersion. However, this 
mechanisms appears to affect tree spacing only in permafrost 
soils, so at some boreal treelines (Crawford 2008, Körner 
2012). At alpine treeline, soil cooling in the shade of trees 
does not appear to restrict tree root growth (Kubisch et al. 
2017). Other negative effects of soil cooling on tree estab-
lishment, e.g. via nutrient availability, might still lead to a 
negative feedback, but there is currently no evidence that 
such feedback affects tree spacing at alpine treelines. As far 
as we are aware, over-dispersed tree cover at alpine treeline 
can be observed only in arid mountains (Lauer 1978, Hoch 
and Körner 2005), indicating underground competition for 
water. However, here it is not specific to the treeline. If trees 
are already over-dispersed in the forest, such a treeline would 
only be considered diffuse if the cover of trees decreases fur-
ther towards the treeline, which could then be due either to 
increased competition (if there is an elevational gradient in 
precipitation or soil properties) or to decreased microsite 
availability. However, in a recent study at moisture-limited 
treelines in the Himalayas the opposite was observed: young 
trees at treeline tended to cluster together, and most strongly 
so in the drier sites (Sigdel et al. 2020).
Such clustering into rounded or elongated tree or 
krummholz islands likely indicates facilitative interactions 
(Germino  et  al. 2002, Batllori  et  al. 2009) among trees or 
krummholz (Fig. 3c, f, i, l, right column). This phenomenon 
has been extensively described, in particular in the literature 
from North America (Alftine and Malanson 2004, Bekker 
2005, Resler 2006, Zeng and Malanson 2006). Islands 
indicate that conditions outside are unfavorable for tree 
establishment, while the islands themselves ameliorate the 
conditions, resulting in positive feedback. Stabilizing posi-
tive feedback is particularly strong if the islands negatively 
274
affect the conditions around the islands, which can occur if 
they redistribute wind or snow (Hättenschwiler and Smith 
1999). Clonal reproduction, e.g. via layering, can also play 
an important role in the formation and dynamics of tree and 
krummholz islands (Marr 1977, Benedict 1984). This, and 
local seed dispersal, are also processes with positive feedback, 
the presence of parent trees having a strong positive effect on 
the emergence of more stems.
Tree island shapes can range from rounded (isotropic) to 
strongly elongated (anisotropic), even forming ‘ribbons’ or 
‘fingers’ (Holtmeier 1982, Baker and Weisberg 1995, Allen 
and Walsh 1996, Bekker and Malanson 2008). Such more 
specific forms can help to identify second-level processes. For 
example, clustering into longitudinal krummholz islands has 
been attributed to direct wind effects (Marr 1977, Benedict 
1984, Alftine and Malanson 2004, Bekker 2005), while for-
est ribbons may be formed due to the redistribution of snow, 
resulting in snow cover that is too shallow or too deep on 
the windward and leeward side, respectively, of the ribbons 
(Holtmeier 1982, Hättenschwiler and Smith 1999).
Changes in tree stature (abrupt–gradual–krummholz):  
growth and dieback
a) Changes in tree stature can, of course, only occur where 
trees are present. If trees disappear before important decreases 
in tree stature take place, i.e. if the treeline consists of an 
abrupt change from tall trees to low-stature alpine vegetation, 
mortality must be more limiting than other first-level pro-
cesses (Fig. 3d–f ). If tree height declines gradually, this can 
be due to growth limitation or dieback, or a combination of 
both (Holtmeier 1985, Holtmeier 2009). Gradual treelines 
without signs of dieback can be considered the null-model of 
treeline form: growth is the only limiting factor. With eleva-
tion, growth must be increasingly limited by tree height (i.e. 
with coupling to the atmosphere), i.e. growth must cease at 
a certain tree size and this size must decrease with elevation, 
to produce a stable gradual treeline in which trees do not 
eventually, even if very slowly, grow tall. Alternatively, a size-
independent decrease in growth rate could also lead to a sta-
ble gradual treeline if tree longevity is limited. However, there 
are indications that trees near treeline can grow extremely 
old (LaMarche 1979, Laberge et al. 2000, Holtmeier 2009, 
Morales et al. 2012), although elevational trends in maximum 
tree ages are not known for most treelines and in species with 
enough data the oldest trees are not necessarily forming part 
of the treeline (Galvan  et  al. 2014). This may be due to a 
more recent establishment or a younger maximum age due 
to persistent stress or very frequent dieback. In practice, some 
level of dieback nearly always accompanies a strong growth 
decline, even if not necessarily leading to krummholz forma-
tion. Additionally, a gradual decline in tree height could be 
caused by species turnover across the treeline ecotone from 
tall tree species to small trees and treeline shrubs (see below).
b) If there is an abrupt change from tall trees to krummholz 
(environmental krummholz only, a change into shrub species 
implies very different mechanisms), this, in turn, indicates 
that mortality is not the most important process, but frequent 
dieback of shoots keeps trees crippled, unless they have man-
aged to grow beyond a critical height, after which growth is 
not a mayor limitation (Fig. 3j–l). These mechanisms may 
function at the level of forest versus a krummholz belt above 
(e.g. as individuals or islands, Fig. 3k–l), or at the level of 
individual trees, where single upright trees and krummholz 
may occur mixed throughout the ecotone (Fig. 3j).
Abrupt height declines from tall forest into alpine vegeta-
tion or krummholz indicate that positive feedback occurs, 
either between tree presence and seedling survival or between 
tree size and biomass maintenance. In the first case, this 
implies a positive feedback switch (Wilson and Agnew 1992) 
and a high temporal stability of the treeline (Armand 1992, 
Bader  et  al. 2008), while in the second case the escape of 
krummholz to upright trees is much more likely to occur 
and can lead to a fast treeline change through a rapid vertical 
growth of already established individuals (Pereg and Payette 
1998, Gamache and Payette 2004, Holtmeier and Broll 
2005, Devi et al. 2008, Kullman and Oberg 2009).
Similar to island shape, the directionality (anisotropy) 
of deformations can be an indication of the causal (second-
level) processes. An obvious example is wind deformation, 
but slope processes and radiation-related processes (e.g. win-
ter desiccation) can also cause directional damage patterns. 
In most cases, a detailed analysis of damage and regrowth 
patterns, including not only directionality but also e.g. the 
heights above the ground where damage starts and ends (pos-
sibly related to e.g. snow-depths or the reach of local brows-
ers), will give strong indications of the second-level processes 
at hand (Pereg and Payette 1998).
Applying the approach
Natural versus anthropogenic treelines
Do patterns indicate the same processes if a treeline form 
has an anthropogenic origin? We suggest here that the inter-
pretation of patterns in terms of first-level processes is inde-
pendent of the second-level causes. Still, in this framework, 
undisturbed and disturbed treelines should be clearly distin-
guished to facilitate the interpretation of patterns in terms 
of second-level processes, which of course is of interest. For 
example, abrupt treelines indicate a dominant role of mor-
tality. In many cases this will be caused by anthropogenic 
influences such as fires or agricultural disturbance in tropical 
mountains (Miehe and Miehe 1994, González  et  al. 2011, 
Bueno and Llambí 2015), but it can also be due to, or main-
tained by, natural processes such as seedling mortality due 
to frost or radiation intolerance (Bader et al. 2007b, Wardle 
2008) (Box A). Likewise, gradual diffuse treelines can indi-
cate that a treeline is advancing (i.e. trees are smaller because 
they are younger and the forest does not form a closed front 
because not all sites have been colonized yet, Batllori and 
Gutierrez 2008) and/or that there has been inhomogeneous 
mortality combined with a lack of warmth for growth. Again, 
this mortality might have happened naturally, most likely in 
275
the establishment phase, or may have been caused by harvest-
ing or fire. The latter has been offered as an interpretation for 
some gradual diffuse treelines, e.g. in southern Andes or the 
Alps (Ellenberg 1966, Tranquillini 1979).
For predicting future treeline shifts based on spatial pat-
tern, understanding the land use history is certainly relevant. 
Historical reasons do not change the basic interpretation of 
the pattern in terms of primary causes, but with very recent 
anthropogenic disturbance, the current pattern will not 
give much information about future dynamics, unless dis-
turbances are continued. However, the recovery of patterns 
after disturbance is informative (Vitali et al. 2019). If no re-
invasion of deforested land at treeline takes place, the treeline 
will thus still be abrupt after many years, indicating that it is 
also less likely to respond to climatic change (González et al. 
2011). A diffuse re-invasion indicates that the harsh alpine 
conditions outside the forest do not prevent tree establish-
ment so that treeline shifts can be fast. The distinction of 
reinvasion from the presence of remnant trees requires esti-
mations of tree ages.
Multi-species treelines
Many treelines are formed by just one or two tree species 
(Fig. 4a), but some can be quite diverse, with 40 or more tree 
species at some tropical cloud-forest treelines (Ramírez et al. 
2009, Peters et al. 2014). Diverse treeline include those on the 
western slopes of the southern Alps of New Zealand and cloud 
forest treelines in the northern Andes (Fig. 4b, Bader  et  al. 
2007a, Wardle 2008, Ramírez et al. 2009, Peters et al. 2014, 
Sarmiento Pinzón and León Moya 2015). The causes of 
diversity patterns are related to large-scale biogeographic pro-
cesses that are outside the scope of this paper. Still, consider-
ing diversity patterns is important when describing treeline 
patterns, because through species-specific demographic and 
growth processes they determine community-level processes 
that may have important implications for treeline dynamics. 
Two important diversity aspects are the functional diversity in 
tree species and the type of non-tree vegetation in and above 
the ecotone. Among trees, particular relevance lies in the 
presence of colonizer tree species, resistant to alpine condi-
tions, which could facilitate the subsequent establishment of 
other species. Such facilitation is seen, for example, at some 
treelines where whitebark pine is present (Resler and Tomback 
2008, Pyatt et al. 2016, Tomback et al. 2016) and is expected 
between radiation-tolerant and shade-demanding species at 
tropical treelines (Bader et al. 2007b, Llambí et al. 2013).
The non-tree vegetation matrix can have both positive 
and negative effects on tree establishment in the ecotone, 
depending on its stature and cover (Akhalkatsi et al. 2006, 
Llambí  et  al. 2013, Dial  et  al. 2016, Liang  et  al. 2016, 
Loranger  et  al. 2017). Whereas low-stature vegetation is 
important mainly at the seedlings stage (Loranger et al. 2017, 
Frei et al. 2018), tall non-tree vegetation can affect trees much 
longer. Where present, small tree and large shrub species may 
take over the habitat above the closed forest, fundamentally 
altering the processes that prevent trees from establishing 
above their current elevational limit. Whereas environmental 
krummholz indicates that dieback is the dominant process 
preventing the establishment of upright trees, shrubs do not 
have to indicate this at all. It also does not preclude it, how-
ever. Shrubs can, in turn, become stunted and sometimes 
environmental krummholz can be found among the shrubs 
(Okitsu and Ito 1989).
Useful information about the species composition 
includes the richness in tree species and the presence of func-
tional groups among trees. In the context of treeline dynam-
ics, an important functional distinction is between pioneers/
colonizers/stress-tolerant/light-demanding versus late-suc-
cessional/shade-tolerant species. Other important properties 
are deciduousness, the ability for layering and seed dispersal 
mode. Describing the presence of such functional traits can 
be of great relevance for understanding treeline dynamics. 
Not only tree traits, but also the composition and structure 
of the non-tree ecotonal vegetation can affect the processes 
shaping the treeline ecotone. This can be summarized into 
vegetation types, e.g. lichen tundra, tussock grassland or tall 
shrub thicket, while additionally the presence, cover, height 
and spatial arrangement of functional groups (e.g. lichens, 
herbs, lawn grasses, tussock grasses, dwarf shrubs, tall shrubs) 
are of interest, as they can strongly affect the establishment 
chances for trees (Lett et al. 2016, Loranger et al. 2017) and 
can also point to changes in limiting abiotic conditions across 
the treeline (Arzac et al. 2019).
Additionally, functional groups of other biota, includ-
ing seed predators, seed dispersers, browsers, pathogens and 
decomposers, may strongly affect some ecological processes 
and thereby control or modify treeline dynamics and spa-
tial patterns (Holtmeier 1966, 2009, Hagedorn et al. 2019). 
Although not in focus in most treeline research projects, 
information about the distribution, diversity and activity of 
such groups is useful auxiliary information for understanding 
the vegetation patterns.
The landscape scale
At most treelines, except those on homogenous slopes, rec-
ognizing the described patterns in the field is a challenge 
because different patterns may occur at a site due to varia-
tion in local topography and disturbance history. Although 
it may be confusing, such heterogeneity is also an opportu-
nity, because it provides a ‘natural experiment’, indicating 
that different processes are important to a different degree at 
different landscape positions, with implications for expected 
dynamics (Allen and Walsh 1996, Elliottt and Kipfmueller 
2010, Holtmeier and Broll 2012, Greenwood  et  al. 2014, 
Bourgeron et al. 2016). Therefore, depending on the specific 
research questions, two good options are to map only the 
dominant treeline type in the area (noting what proportion 
of the landscape actually shows this pattern) or to map the 
patterns within relatively narrow vertical belt transects and to 
relate the patterns to landscape positions.
In particular in diffuse and gradual treelines, where pat-
terns of mortality and growth, respectively, should depend 
276
most directly on climatic gradients (Fig. 3), steep slopes are 
likely to have narrower ecotones than gentle slopes. Of course, 
the ecotone width should thus be interpreted with respect to 
the slope angle. However, we suspect that there is no quali-
tative difference in the processes that cause medium-wide 
and very-wide diffuse, gradual or island ecotones, although 
differences in ecotone width may indicate that the processes 
vary in intensity. Accounting for slope angles is therefore 
mainly important when comparing treeline patterns quanti-
tatively, e.g. among regions or with respect to model output 
(Martinez et al. 2011).
Another scale-related question is how to interpret tree 
islands, or forest outposts, that are associated with specific 
topographic positions, e.g. depressions. This is a question 
not only of scale but also of the mechanisms that allow these 
islands to exist. Most tree islands are probably started in the 
protection of some topographic feature, even if it is just a 
small rock or topographic depression (Resler  et  al. 2005, 
Resler 2006, Llambí et al. 2013). For many islands, the forest 
patch can then expand from the protected site and maintain 
itself due to the protection it provides to establishing trees. 
This would thus be a true self-maintained island, even if it 
becomes rather large. If, however, the forest is restricted to 
topographic microsites without proliferating out of the abi-
otic protection zone, the outpost is unlikely to be indicative 
of treeline-forming processes and should not be considered a 
tree island. Instead, it may be interpreted as part of a land-
scape-scale pattern.
Landscape-scale patterns offer an additional window for 
gaining insights into treeline-controlling processes, ideally to 
be interpreted in conjunction with treeline form (Treml and 
Chuman 2015). The existence of recurrent patterns indicat-
ing controlling processes is suggested by comparative research 
by, for example, Friedrich-Karl Holtmeier (1973, 2005, 2009, 
Holtmeier and Broll 2017), several North-American geog-
raphers (Butler  et  al. 2009b), Peter Wardle (Wardle 1965, 
Wardle et al. 2001, Wardle 2008) and Georg Miehe (Miehe 
and Miehe 1994, 2000). A systematic analysis of described 
landscape patterns, accompanied by a GIS-based analysis of 
remotely-sensed patterns in tree cover in treeline landscapes 
globally, would likely reveal additional dimensions defining 
treeline types in terms of landscape-scale controlling pro-
cesses (Case and Hale 2015).
Data collection and analysis strategies
For some of the pattern dimensions, in particular the pat-
terns in deformation and species composition, field knowl-
edge is indispensable and transects along the slope are the 
most appropriate sampling method. To also allow an analy-
sis of population structure and spatial point patterns, such 
transects should be wide enough (at least 30 m, as a rule of 
thumb, to assure including enough trees also at the upper-
most low-tree-density end of the ecotone) and long enough 
to capture the entire ecotone width (Camarero and Gutiérrez 
2004, Liang  et  al. 2011, Wang  et  al. 2016, Vitali  et  al. 
2017, Sigdel  et  al. 2020). Thereby, it may be necessary to 
restrict seedling counts to smaller subplots, if regeneration 
is very prolific. Ideally, tree-ring-based growth and age pat-
terns would also be collected along such transects, at least 
in non-tropical regions where most tree and shrub species 
form annual rings. However, for characterizing growth forms, 
tree height changes and species compositions only, narrower 
transects or smaller plots in sequence along the ecotone can 
be sufficient and may be preferable under budget or time 
constraints. Other patterns, like spatial clumping in the 
x–y plane, may be described more efficiently using remotely 
sensed data (Chhetri and Thai 2019). Because of the small-
scale patterns addressed, high-resolution (< 1-m resolution) 
aerial or satellite-based images are best suited. Such images are 
available for an increasing portion of Earth’s surface, includ-
ing remote treeline areas, although data availability may still 
be limited for very cloudy or very remote sites. Additional 
very promising technologies include Airborne Laser Scanning 
(Coops et al. 2013, Hauglin and Næsset 2016, Bolton et al. 
2018) and drone-based stereo-photography (Jochner  et  al. 
2017a), both allowing the creation of surface models that 
may allow determining the height of even relatively small 
trees, depending on the alpine vegetation matrix. A combina-
tion of field and remote-sensing methods, focusing precious 
field-work time on those pattern dimensions that cannot 
be obtained remotely, may be the most efficient method to 
obtain information on all pattern dimensions needed to fully 
characterize a site. Importantly, even if not all patterns can 
be fully mapped, e.g. in remote areas, detecting at least some 
dimensions of the patterns can allow inferences about the 
associated processes, which can be an important improve-
ment in site-specific knowledge, allowing better predictions.
The observed patterns indicate what processes have acted 
upon the ecotone in the past, up to the present date and our 
best-informed prediction is therefore for these processes to 
continue to act in the future. However, treeline ecotones are 
dynamic systems. Some are more dynamic than others, as 
indicated by their spatial pattern, but this may also change, 
in particular in times of environmental and land-use change. 
Multi-temporal datasets could allow analyzing the dynamics 
of the patterns and mapping these dynamics using the con-
ceptual diagrams proposed here (e.g. trajectories from abrupt 
to gradual or discrete to diffuse treelines). Re-evaluation of 
spatial patterns can serve to validate previous assumptions 
and model predictions about pattern–process relationships 
and to detect changes in controlling processes.
 Therefore, it is strongly recommended to revisit treelines 
regularly, the frequency depending on the expected dynamics 
and the purpose of monitoring the treeline. If the purpose is 
a detailed monitoring of population processes, yearly revis-
its with a resampling of population parameters can be very 
useful (Kullman 2007), though at very stable treelines like 
the abrupt Nothofagus treelines in New Zealand, resampling 
every ten years may be sufficient to capture any dynamics 
(Harsch et al. 2012). If the purpose is to validate predictions 
of treeline dynamics, revisits can be more widely spaced, 
though preferably not by more than ca five years, to not miss 
out on unexpected changes (e.g. abrupt treelines turning 
277
diffuse and shifting upwards due to sudden tree establish-
ment after passing a climatic threshold – Danby and Hik 
2007a, Elliott 2012, Camarero et al 2015 – or after a change 
in land use – Vitali et al. 2017). Thereby, revisits do not need 
to imply resampling, or even field visits, but could also be 
based on high-resolution remote-sensing (aerial or ground-
based) images, including images from the past (Danby and 
Hik 2007b, Lutz et al. 2013, Greenwood et al. 2014, Travers-
Smith and Lantz 2020).
Having mapped the tree and forest cover, a next step is 
a visual interpretation of the spatial pattern based on the 
examples and descriptions provided in this paper. For height 
decline and deformation gradients we recommend a manual 
interpretation, as automated edge detection and description 
algorithms consistently perform worse than the human brain 
(Bader  et  al. 2007a). However, for the 2-D cover patterns 
we recommend a quantitative interpretation, describing gra-
dients in cover intensity and the level of clustering. For this 
type of analysis, spatial statistics can be applied. Depending 
on the data type, a point-pattern analysis, which represents 
trees as points (x–y coordinates) with associated variables as 
tree size or age (care should be taken to not confuse stem 
density and cover), could be applicable. In such an analysis, 
the pattern dimensions can be described using the measures 
intensity and clumping, complemented with measures like 
anisotropy (Camarero et al. 2000, 2005, Illian et al. 2007, 
Wiegand and Moloney 2013). A second interesting option 
is an analysis of tree cover based on patch-based metrics, 
e.g. using FRAGSTATS (Allen and Walsh 1996, Alftine and 
Malanson 2004, McGarigal et al. 2012).
The conceptual framework presented here, in which con-
trasting treeline forms or patterns are predicted to occur as 
a result of different first-level processes, could be formalized 
via individual based models in which these pattern–pro-
cess links can be explicitly explored (Wiegand  et  al. 2006, 
Martinez et al. 2011), and the outputs of these models tested 
against empirical data derived from different treelines. Models 
could also be elaborated to include second-level processes and 
their environmental drivers. As different processes can result 
in similar patterns, this approach will require a multi-pattern 
evaluation of the results (i.e. using pattern-oriented model-
ling, Wiegand et al. 2003). Thereby, the more patterns (spa-
tial, temporal, demographic and compositional) the model 
can correctly predict, the more likely is it a realistic repre-
sentation of the processes involved. Such models could also 
be used as a framework to explore the dynamics of treeline 
spatial patterns (e.g. island formation and expansion, shifts 
from abrupt to gradual treelines, etc.) as a result of changes 
in the underlying drivers. However, it should also be clear 
that spatial patterns cannot tell us everything, and in some 
cases pattern analysis will need to be complemented by fur-
ther experimental or observational studies evaluating the role 
of specific processes and mechanisms (e.g. ecophysiological 
constraints, dispersal patterns, neighbor interactions).
Conclusions
Alpine treeline ecotones can exhibit many different spatial 
patterns, but within this variation there is order that can 
help to recognize treelines with similar controlling processes 
and expected dynamics. Important dimensions to describe 
these patterns are the rate of change in tree cover and the 
strength and shape of clustering of this cover, and changes in 
tree height and tree shape. These pattern dimensions could 
be complemented by additional information about tree age 
distributions, the functional composition of the ecotone veg-
etation and environmental conditions to distinguish second-
level processes in more detail. Describing and ordering the 
variation in the four main pattern dimensions systematically 
should help to maximize the use of the information gained in 
treeline studies beyond the particular sites studied. Stratifying 
treeline research and monitoring efforts according to spatial 
types should also help to cover the full range of dynamics 
expected at different treeline types. These advantages should 
allow us to more efficiently gain the necessary understand-
ing to predict treeline shifts in response to climate and land-
use changes. We encourage all treeline researchers to try to 
quantify the dimensions presented here and to work towards 
validating this qualitative scheme with a quantification of 
patterns and model- and field-based tests of pattern-process 
relationships.
Invitation
Discussing and sharing spatial data and pattern descriptions 
is possible via the following website: <www.alpine-treelines.
net>, one of its aims being the creation of a world map of 
treeline spatial patterns, the other to provide a network envi-
ronment for treeline researchers interested in these issues. 
Treeline pattern detection, analysis and classification tools 
will be made available on this website as they are developed.
Acknowledgements – Thanks to Georg Miehe for stimulating 
discussions and to Marc F. P. M. Maas for discussing and drawing 
Fig. 2a and the schematic examples of the treeline types in Box A. 
We also thank the anonymous reviewers for their constructive 
comments. Open access funding enabled and organized by Projekt 
DEAL.
Funding – This paper was the result of a synthesis workshop funded 
by the Mountain Research Initiative (MRI) and the German 
Research Foundation (DFG, grant number BA 3843/8-1) and 
hosted by the Pyrenean Inst. of Ecology in Jaca, Spain. Further 
details were discussed during two sTreelines workshops funded by 
sDiv, the synthesis centre of iDiv, Leipzig and a workshop funded 
by Auckland Univ. of Technology, School of Science.
Author contributions – All authors discussed the concepts presented 
in this paper in a synthesis workshop in Jaca in September 2017 
(except HLB) and in different configurations in three further 
workshops. MYB wrote the paper with the help of all authors. All 
approved of the final version.
278
References
Akhalkatsi, M.  et  al. 2006. Facilitation of seedling microsites by 
Rhododendron caucasicum extends the Betula litwinowii alpine 
treeline, Caucasus Mountains, Republic of Georgia. – Arctic 
Antarct. Alp. Res. 38: 481–488.
Alftine, K. J. and Malanson, G. P. 2004. Directional positive feed-
back and pattern at an alpine tree line. – J. Veg. Sci. 15: 3–12.
Allen, T. R. and Walsh, S. J. 1996. Spatial and compositional pat-
tern of alpine treeline, Glacier National Park, Montana. – Pho-
togram. Eng. Remote Sens. 62: 1261–1268.
Armand, A. D. 1992. Sharp and gradual mountain timberlines as 
a result of species interaction. – In: Hansen, A. J. and di Castri, 
F. (eds), Landscape boundaries: consequences for biotic diver-
sity and ecological flows. Springer, pp. 360–378.
Arzac, A. et al. 2019. Modelling the effect of temperature changes 
on plant life-form distribution across a treeline ecotone in the 
tropical Andes. – Plant Ecol. Divers. 12: 619–631.
Astudillo-Sánchez, C. C. et al. 2019. Recruitment and facilitation 
in Pinus hartwegii, a Mexican alpine treeline ecotone, with 
potential responses to climate warming. – Trees 33: 1087–1100.
Bader, M. Y.  et  al. 2007a. Vegetation structure and temperature 
regimes of tropical alpine treelines. – Arctic Antarct. Alp. Res. 
39: 353–364.
Bader, M. Y. et al. 2007b. High solar radiation hinders tree regen-
eration above the alpine treeline in northern Ecuador. – Plant 
Ecol. 191: 33–45.
Bader, M. Y. et al. 2008. A simple spatial model exploring positive 
feedbacks at tropical alpine treelines. – Arctic Antarct. Alp. Res. 
40: 269–278.
Baker, W. L. and Weisberg, P. J. 1995. Landscape analysis of the 
forest-tundra ecotone in Rocky Mountain National Park, Col-
orado. – Profess. Geogr. 47: 361–375.
Barbeito, I.  et  al. 2013. Snow fungi-induced mortality of Pinus 
cembra at the Alpine treeline: evidence from plantations. – Arc-
tic Antarct. Alp. Res. 45: 455–470.
Batllori, E. and Gutierrez, E. 2008. Regional tree line dynamics in 
response to global change in the Pyrenees. – J. Ecol. 96: 
1275–1288.
Batllori, E. et al. 2009. Seedling recruitment, survival and facilita-
tion in alpine Pinus uncinata tree line ecotones. Implications 
and potential responses to climate warming. – Global Ecol. 
Biogeogr. 18: 460–472.
Bekker, M. F. 2005. Positive feedback between tree establishment 
and patterns of subalpine forest advancement, Glacier National 
Park, Montana, U.S.A. – Arctic Antarct. Alp. Res. 37: 97–107.
Bekker, M. F. and Malanson, G. P. 2008. Linear forest patterns in 
subalpine environments. – Prog. Phys. Geogr. 32: 635–653.
Benedict, J. B. 1984. Rates of tree-island migration, Colorado 
Rocky Mountains, USA. – Ecology 65: 820–823.
Bolton, D. K. et al. 2018. Evidence of vegetation greening at alpine 
treeline ecotones: three decades of Landsat spectral trends 
informed by lidar-derived vertical structure. – Environ. Res. 
Lett. 13: 084022.
Bonanomi, G. et al. 2018. Anthropogenic and environmental fac-
tors affect the tree line position of Fagus sylvatica along the 
Apennines (Italy). – J. Biogeogr. 45: 2595–2608.
Bourgeron, P. S.  et  al. 2016. The forest–alpine ecotone: a multi-
scale approach to spatial and temporal dynamics of treeline 
change at Niwot Ridge. – Plant Ecol. Divers. 8: 763–779.
Brockmann-Jerosch, H. 1919. Baumgrenze und Klimacharacter. 
– Rascher & Cie.
Brown, J. H. et al. 1996. The geographic range: size, shape, bound-
aries and internal structure. – Annu. Rev. Ecol. Syst. 27: 
597–623.
Bueno, A. and Llambí, L. D. 2015. Facilitation and edge effects 
influence vegetation regeneration in old-fields at the tropical 
Andean forest line. – Appl. Veg. Sci. 18: 613–623.
Butler, D. R. et al. 2009a. Geomorphic patterns and processes at 
alpine treeline. – In: Butler, D. R.  et  al. (eds), The changing 
Alpine treeline, the example of Glacier National Park, MT, 
USA. Elsevier, pp. 63–84.
Butler, D. R. et al. 2009b. The changing Alpine treeline, the exam-
ple of Glacier National Park, MT, USA. – Elsevier.
Cairns, D. M. and Moen, J. 2004. Herbivory influences tree lines. 
– J. Ecol. 92: 1019–1024.
Camarero, J. J. and Gutierrez, E. 1999. Structure and recent 
recruitment at alpine forest-pasture ecotones in the Spanish 
central Pyrenees. – Ecoscience 6: 451–464.
Camarero, J. J. and Gutiérrez, E. 2004. Pace and pattern of recent 
treeline dynamics: response of ecotones to climatic variability 
in the Spanish Pyrenees. – Clim. Change 63: 181–200.
Camarero, J. J. et al. 2000. Spatial pattern of subalpine forest-alpine 
grassland ecotones in the Spanish central Pyrenees. – For. Ecol. 
Manage. 134: 1–16.
Camarero, J. J. et al. 2005. Spatial patterns of tree recruitment in 
a relict population of Pinus uncinata: forest expansion through 
stratified diffusion. – J. Biogeogr. 32: 1979–1992.
Camarero, J. J. et al. 2015. Recent and intense dynamics in a for-
merly static Pyrenean treeline. – Arctic Antarct. Alp. Res. 47: 
773–783.
Case, B. S. and Hale, R. J. 2015. Using novel metrics to assess 
biogeographic patterns of abrupt treelines in relation to abiotic 
influences. – Progr. Phys. Geogr. 39: 310–336.
Chhetri, P. K. and Thai, E. 2019. Remote sensing and geographic 
information systems techniques in studies on treeline ecotone 
dynamics. – J. For. Res. 30: 1543–1553.
Coops, N. C.  et  al. 2013. Characterization of an alpine tree line 
using airborne LiDAR data and physiological modeling. 
– Global Change Biol. 19: 3808–3821.
Crawford, R. M. M. 2008. Plants at the margin: ecological limits 
and climate change. – Cambridge Univ. Press.
Crofts, A. L. and Brown, C. D. 2020. The importance of biotic 
filtering on boreal conifer recruitment at alpine treeline. – Ecog-
raphy 43: 914–929.
Cuevas, J. G. 2000. Tree recruitment at the Nothofagus pumilio 
alpine timberline in Tierra del Fuego, Chile. – J. Ecol. 88: 
840–855.
Danby, R. K. and Hik, D. S. 2007a. Variability, contingency and 
rapid change in recent subarctic alpine tree line dynamics. 
– J. Ecol. 95: 352–363.
Danby, R. K. and Hik, D. S. 2007b. Evidence of recent treeline 
dynamics in southwest Yukon from aerial photographs. – Arctic 
60: 411–420.
Devi, N. et al. 2008. Expanding forests and changing growth forms 
of Siberian larch at the Polar Urals treeline during the 20th 
century. – Global Change Biol. 14: 1581–1591.
Dial, R. J. et al. 2016. Shrubline but not treeline advance matches 
climate velocity in montane ecosystems of south‐central Alaska. 
– Global Change Biol. 22: 1841–1856.
Ellenberg, H. 1966. Leben und Kampf an den Baumgrenzen der 
Erde. – Naturwissenschaftliche Rundschau 4: 133–139.
Elliott, G. P. 2011. Influences of 20th-century warming at the 
upper tree line contingent on local-scale interactions: evidence 
279
from a latitudinal gradient in the Rocky Mountains, USA. 
– Global Ecol. Biogeogr. 20: 46–57.
Elliott, G. P. 2012. Extrinsic regime shifts drive abrupt changes in 
regeneration dynamics at upper treeline in the Rocky Moun-
tains, USA. – Ecology 93: 1614–1625.
Elliottt, G. P. and Kipfmueller, K. E. 2010. Multi-scale influences 
of slope aspect and spatial pattern on ecotonal dynamics at 
upper treeline in the southern Rocky Mountains, USA. – Arc-
tic Antarct. Alp. Res. 42: 45–56.
Frei, E. R.  et  al. 2018. Biotic and abiotic drivers of tree seedling 
recruitment across an alpine treeline ecotone. – Sci. Rep. 8: 
10894.
Galvan, J. D. et al. 2014. Seeing the trees for the forest: drivers of 
individual growth responses to climate in Pinus uncinata moun-
tain forests. – J. Ecol. 102: 1244–1257.
Gamache, I. and Payette, S. 2004. Height growth response of tree 
line black spruce to recent climate warming across the forest-
tundra of eastern Canada. – J. Ecol. 92: 835–845.
Germino, M. J. and Smith, W. K. 1999. Sky exposure, crown archi-
tecture and low-temperature photoinhibition in conifer seed-
lings at alpine treeline. – Plant Cell Environ. 22: 407–415.
Germino, M. J. et al. 2002. Conifer seedling distribution and sur-
vival in an alpine-treeline ecotone. – Plant Ecol. 162: 157–168.
González, W. et al. 2011. Successional dynamics of the tree com-
ponent in the forest-paramo transition belt in the tropical 
Andes. – Ecotrópicos 24: 60–79.
Greenwood, S.  et  al. 2014. Strong topographic sheltering effects 
lead to spatially complex treeline advance and increased forest 
density in a subtropical mountain region. – Global Change 
Biol. 20: 3756–3766.
Hagedorn, F. et al. 2014. Treeline advances along the Urals moun-
tain range – driven by improved winter conditions? – Global 
Change Biol. 20: 3530–3543.
Hagedorn, F. et al. 2019. Above- and belowground linkages shape 
responses of mountain vegetation to climate change. – Science 
365: 1119–1123.
Harsch, M. A. and Bader, M. Y. 2011. Treeline form – a potential 
key to understanding treeline dynamics. – Global Ecol. Bioge-
ogr. 20: 582–596.
Harsch, M. A. et al. 2009. Are treelines advancing? A global meta-
analysis of treeline response to climate warming. – Ecol. Lett. 
12: 1040–1049.
Harsch, M. A. et al. 2012. Causes of tree line stability: stem growth, 
recruitment and mortality rates over 15 years at New Zealand 
Nothofagus tree lines. – J. Biogeogr. 39: 2061–2071.
Hättenschwiler, S. and Smith, W. K. 1999. Seedling occurrence in 
alpine treeline conifers: a case study from the central Rocky 
Mountains, USA. – Acta Oecol. 20: 219–224.
Hauglin, M. and Næsset, E. 2016. Detection and segmentation of 
small trees in the forest-tundra ecotone using airborne laser 
scanning. – Remote Sens. 8: 407.
Heenan, P. B. and Smissen, R. D. 2013. Revised circumscription 
of Nothofagus and recognition of the segregate genera Fuscos-
pora, Lophozonia and Trisyngyne (Nothofagaceae). – Phytotaxa 
146: 1–31.
Hermes, K. 1955. Die Lage der oberen Waldgrenze in den Gebirgen 
der Erde und ihr Abstand zur Schneegrenze. – Selbstverlag des 
Geographischen Inst. der Univ. Köln.
Hoch, G. and Körner, C. 2005. Growth, demography and carbon 
relations of Polylepis trees at the world’s highest treeline. 
– Funct. Ecol. 19: 941–951.
Holtmeier, F. K. 1966. Die ökologische Funktion des Tannenhähers 
im Zirben-Lärchenwald und an der Waldgrenze des Oberenga-
dins. – J. Ornithol. 107: 345.
Holtmeier, F. K. 1973. Geoecological aspects of timberlines in 
northern and central Europe. – Arctic Alp. Res. 5: 
A45–A54.
Holtmeier, F. K. 1982. ‘Ribbon-forest’ und ‘Hecken’: Streifenartige 
Verbreitungsmuster des Baumwuchses an der oberen Waldgrenze 
in den Rocky Mountains. – Erdkunde 36: 142–153.
Holtmeier, F. K. 1985. Climatic stress influencing the physiognomy 
of trees at the polar and mountain timberline. – In: Turner, H. 
and Tranquillini, W. (eds), Establishment and tending of sub-
alpine forest: research and management. IUFRO project group 
P1.07-00, pp. 31–40.
Holtmeier, F. K. 2005. Relocation of snow and its effects in the 
treeline ecotone – with special regard to the Rocky Mountains, 
the Alps and northern Europe. – Erde 136: 343–373.
Holtmeier, F. K. 2009. Mountain timberlines. Ecology, patchiness 
and dynamics. – Springer.
Holtmeier, F. K. and Broll, G. 2005. Sensitivity and response of 
northern hemisphere altitudinal and polar treelines to environ-
mental change at landscape and local scales. – Global Ecol. 
Biogeogr. 14: 395–410.
Holtmeier, F.-K. and Broll, G. 2010. Wind as an ecological agent 
at treelines in North America, the Alps and the European Sub-
arctic. – Phys. Geogr. 31: 203–233.
Holtmeier, F. K. and Broll, G. 2012. Landform influences on tree-
line patchiness and dynamics in a changing climate. – Phys. 
Geogr. 33: 403–437.
Holtmeier, F. K. and Broll, G. 2017. Treelines – approaches at dif-
ferent scales. – Sustainability 9: 808.
Illian, J.  et  al. 2007. Statistical analysis and modelling of spatial 
point patterns. – Wiley.
Jochner, M.  et  al. 2017a. Among-tree variability and feedback 
effects result in different growth responses to climate change at 
the upper treeline in the Swiss Alps. – Ecol. Evol. 7: 7937–7953.
Jochner, M. et al. 2017b. Tree growth responses to changing tem-
peratures across space and time: a fine-scale analysis at the tree-
line in the Swiss Alps. – Trees 32: 645–660.
Johnson, A. C. and Yeakley, J. A. 2019. Microsites and climate 
zones: seedling regeneration in the alpine treeline ecotone 
worldwide. – Forests 10: 864.
Johnson, J. S. et al. 2017. Seed dispersal at alpine treeline: an assess-
ment of seed movement within the alpine treeline ecotone. 
– Ecosphere 8: e01649.
Joshi, A. A. et al. 2019. Frost maintains forests and grasslands as 
alternate states in a montane tropical forest–grassland mosaic; 
but alien tree invasion and warming can disrupt this balance. 
– J. Ecol. 108: 122–132.
Karger, D. N. et al. 2019. Why tree lines are lower on islands-cli-
matic and biogeographic effects hold the answer. – Global Ecol. 
Biogeogr. 28: 839–850.
Körner, C. 1998. A re-assessment of high elevation treeline posi-
tions and their explanation. – Oecologia 115: 445–459.
Körner, C. 2012. Alpine treelines. Functional ecology of the global 
high elevation tree limits. – Springer.
Körner, C. and Paulsen, J. 2004. A world-wide study of high alti-
tude treeline temperatures. – J. Biogeogr. 31: 713–732.
Kubisch, P. et al. 2017. Fine root abundance and dynamics of stone 
pine (Pinus cembra) at the alpine treeline is not impaired by 
self-shading. – Front. Plant Sci. 8: 602.
280
Kullman, L. 2007. Tree line population monitoring of Pinus sylves-
tris in the Swedish Scandes, 1973–2005: implications for tree 
line theory and climate change ecology. – J. Ecol. 95: 41–52.
Kullman, L. and Oberg, L. 2009. Post-Little Ice Age tree line rise 
and climate warming in the Swedish Scandes: a landscape eco-
logical perspective. – J. Ecol. 97: 415–429.
Laberge, M. J.  et  al. 2000. Life span and biomass allocation 
of stunted black spruce clones in the subarctic environment. 
– J. Ecol. 88: 584–593.
LaMarche Jr., V. C. 1979. Environment in relation to age of bris-
tlecone pines. – Ecology 50: 53–59.
Lauer, W. 1978. Timberline studies in central Mexico. – Arctic Alp. 
Res. 10: 383–396.
Lett, S.  et  al. 2016. Bryophyte traits explain climate-warming 
effects on tree seedling establishment. – J. Ecol. 32: 1666–1680.
Leuschner, C. 1996. Timberline and alpine vegetation on the trop-
ical and warm-temperate oceanic islands of the world: elevation, 
structure and floristics. – Vegetatio 123: 193–206.
Liang, E. et al. 2011. Little change in the fir tree-line position on 
the southeastern Tibetan Plateau after 200 years of warming. 
– New Phytol. 190: 760–769.
Liang, E.  et  al. 2016. Species interactions slow warming-induced 
upward shifts of treelines on the Tibetan Plateau. – Proc. Natl 
Acad. Sci. USA 113: 4380–5.
Llambí, L. D. et al. 2013. Spatial relations and population structure 
of a dominant tree along a treeline ecotone in the Tropical 
Andes: interactions at gradient and plant–neighbourhood 
scales. – Plant Ecol. Divers. 6: 343–353.
Loranger, H. et al. 2017. Competitor or facilitator? The ambiguous 
role of alpine grassland for the early establishment of tree seed-
lings at treeline. – Oikos 126: 1625–1636.
Lutz, D. A. et al. 2013. Four decades of andean timberline migra-
tion and implications for biodiversity loss with climate change. 
– PLoS One 8: e74496.
Malanson, G. P. 1997. Effects of feedbacks and seed rain on ecotone 
patterns. – Landscape Ecol. 12: 27–38.
Marr, J. W. 1977. The development and movement of tree islands 
near the upper limit of tree growth in the southern Rocky 
Mountains. – Ecology 58: 1159–1164.
Martinez, I. et al. 2011. Disentangling the formation of contrasting 
tree-line physiognomies combining model selection and Bayes-
ian parameterization for simulation models. – Am. Nat. 177: 
E136–E152.
Mayr, S. et al. 2019. Winter embolism and recovery in the conifer 
shrub Pinus mugo L. – Forests 10: 941.
McGarigal, K. et al 2012. FRAGSTATS v4: spatial pattern analysis 
program for categorical and continuous maps. – Computer 
software program produced by the authors at the Univ. of Mas-
sachusetts, Amherst, <www.umass.edu/landeco/research/frag-
stats/fragstats.html>.
McGlone, M. 2004. Timberline and climate change. – Te Taiao 
2004: 8–9.
McIntire, E. J. B. et al. 2016. Wind exposure and light exposure, 
more than elevation-related temperature, limit tree line seedling 
abundance on three continents. – J. Ecol. 104: 1379–1390.
Miehe, G. and Miehe, S. 1994. Zur oberen Waldgrenze in tropis-
chen Gebirge. – Phytocoenologia 24: 53–110.
Miehe, G. and Miehe, S. 2000. Comparative high mountain 
research on the treeline ecotone under human impact. Carl 
Troll’s ‘Asymmetrical zonation of the humid vegetation types of 
the World’ of 1948 reconsidered. – Erdkunde 54: 34–50.
Miehe, G.  et  al. 2015. Vegetation ecology. – In: Miehe, G. and 
Pendry, C. (eds), Nepal. An introduction to the natural history, 
ecology and human environment of the Himalayas. Royal 
Botanic Garden Edinburgh, pp. 385–472.
Morales, M. S.  et  al. 2012. Precipitation changes in the South 
American Altiplano since 1300 AD reconstructed by tree-rings. 
– Clim. Past 8: 653–666.
Norton, D. A. and Schönenberger, W. 1984. The growth forms and 
ecology of Nothofagus solandri at the alpine timberline, Craigie-
burn Range, New Zealand. – Arctic Alp. Res. 16: 361–370.
Okitsu, S. and Ito, K. 1989. Conditions for the development of 
the Pinus pumila zone of Hokkaido, northern Japan. – Vegeta-
tio 84: 127–132.
Paulsen, J. and Körner, C. 2014. A climate-based model to predict 
potential treeline position around the globe. – Alp. Bot. 124: 
1–12.
Pereg, D. and Payette, S. 1998. Development of black spruce 
growth forms at treeline. – Plant Ecol. 138: 137–147.
Peters, T. et al. 2014. An ecological paradox: high species diversity 
and low position of the upper forest line in the Andean Depres-
sion. – Ecol. Evol. 4: 2134–2145.
Pyatt, J. C. et al. 2016. The importance of conifers for facilitation 
at treeline: comparing biophysical characteristics of leeward 
microsites in whitebark pine communities. – Arctic Antarct. 
Alp. Res. 48: 427–444.
Ramírez, L.  et  al. 2009. Vegetation structure along the forest- 
páramo transition belt in the Sierra Nevada de Mérida: implica-
tions for understanding treeline dynamics. – Ecotrópicos 22: 
83–98.
Rehm, E. M. and Feeley, K. J. 2015. The inability of tropical cloud 
forest species to invade grasslands above treeline during climate 
change: potential explanations and consequences. – Ecography 
38: 1167–1175.
Resler, L. M. 2006. Geomorphic controls of spatial pattern and 
process at alpine treeline. – Profess. Geogr. 58: 124–138.
Resler, L. M. and Tomback, D. F. 2008. Blister rust prevalence in 
krummholz whitebark pine: implications for treeline dynamics, 
northern Rocky Mountains, Montana, USA. – Arctic Antarct. 
Alp. Res. 40: 161–170.
Resler, L. M. et al. 2005. Topographic shelter and conifer establish-
ment and mortality in an alpine environment, Glacier National 
Park, Montana. – Phys. Geogr. 26: 112–125.
Robles-Gutiérrez, C. A. et al. 2016. Probability of mortality by fire 
damage of young Pinus hartwegii Lindl. trees in the Izta-Popo 
National Park. – Revista Chapingo Serie Ciencias Forestales y 
del Ambiente XXII: 165–178.
Sarmiento Pinzón, C. E. and León Moya, O. A. 2015. Transición 
bosque-páramo. Bases conceptuales y métodos para su identifi-
cación en los Andes colombianos. – Inst. de Investigación de 
Recursos Biológicos Alexander von Humboldt.
Schickhoff, U. 2005. The upper timberline in the Himalayas, 
Hindu Kush and Karakorum: a review of geographical and eco-
logical aspects. – In: Broll, G. and Keplin, B. (eds), Mountain 
ecosystems: studies in treeline ecology. Springer, pp. 275–354.
Schickhoff, U.  et  al. 2015. Do Himalayan treelines respond to 
recent climate change? An evaluation of sensitivity indicators. 
– Earth Syst. Dyn. 6: 245–265.
Schwab, N. et al. 2016. Treeline responsiveness to climate warming: 
insights from a krummholz treeline in Rolwaling Himal, Nepal. 
– In: Singh, R. B. et al. (eds), Climate change, glacier response 
and vegetation dynamics in the Himalaya: contributions toward 
281
future Earth initiatives. Springer International Publishing, 
pp. 307–345.
Sigdel, S. R.  et  al. 2020. Tree-to-tree interactions slow down 
Himalayan treeline shifts as inferred from tree spatial patterns. 
– J. Biogeogr. 47: 1816–1826.
Silva, L. C. R. and Anand, M. 2011. Mechanisms of Araucaria 
(Atlantic) forest expansion into southern Brazilian grasslands. 
– Ecosystems 14: 1354–1371.
Strayer, D. L. et al. 2003. A classification of ecological boundaries. 
– Bioscience 53: 723–729.
Tomback, D. F. et al. 2016. Whitebark pine facilitation at treeline: 
potential interactions for disruption by an invasive pathogen. 
– Ecol. Evol. 6: 5144–5157.
Tranquillini, W. 1979. Physiological ecology of the Alpine timber-
line: tree existence at high altitudes with special reference to the 
European Alps. – Springer.
Travers-Smith, H. Z. and Lantz, T. 2020. Leading-edge disequilib-
rium in alder and spruce populations across the forest–tundra 
ecotone. – Ecosphere 11: e03118.
Treml, V. and Chuman, T. 2015. Ecotonal dynamics of the altitu-
dinal forest limit are affected by terrain and vegetation structure 
variables: an example from the Sudetes Mountains in central 
Europe. – Arctic Antarct. Alp. Res. 47: 133–146.
Treml, V. and Veblen, T. T. 2017. Does tree growth sensitivity to 
warming trends vary according to treeline form? – J. Biogeogr. 
44: 1469–1480.
Treml, V.  et  al. 2019. Differences in growth between shrubs and 
trees: how does the stature of woody plants influence their abil-
ity to thrive in cold regions? – Agric. For. Meteorol. 271: 54–63.
Vitali, A.  et  al. 2017. Deconstructing human-shaped treelines: 
microsite topography and distance to seed source control Pinus 
nigra colonization of treeless areas in the Italian Apennines. 
– For. Ecol. Manage. 406: 37–45.
Vitali, A. et al. 2019. Pine recolonization dynamics in Mediterra-
nean human-disturbed treeline ecotones. – For. Ecol. Manage. 
435: 28–37.
von Humboldt, A. and Bonpland, A. 1807. Ideen zu einer Geog-
raphie der Pflanzen nebst einem Naturgemälde der Tropen-
länder. – F. G. Cotta/F. Schoell.
Wang, Y. et al. 2016. Impact of plot shape and size on the evalua-
tion of treeline dynamics in the Tibetan Plateau. – Trees 30: 
1045–1056.
Wardle, P. 1965. A comparison of alpine timber lines in New Zea-
land and North America. – N. Z. J. Bot. 3: 113–135.
Wardle, P. 1985. New Zealand timberlines. 1. Growth and survival 
of native and introduced tree species in the Craigieburn Range, 
Canterbury. – N. Z. J. Bot. 23: 219–234.
Wardle, P. 2008. New Zealand forest to alpine transitions in global 
context. – Arctic Antarct. Alp. Res. 40: 240–249.
Wardle, P. et al. 2001. Comparison of the flora and vegetation of 
the southern Andes and New Zealand. – N. Z. J. Bot. 39: 
69–108.
Wiegand, T. and Moloney, K. A. 2013. Handbook of spatial point 
pattern analysis in ecology. – CRC Press.
Wiegand, T.  et  al. 2003. Using pattern-oriented modeling for 
revealing hidden information: a key for reconciling ecological 
theory and application. – Oikos 100: 209–222.
Wiegand, T.  et  al. 2006. Abrupt population changes in treeline 
ecotones along smooth gradients. – J. Ecol. 94: 880–892.
Wilson, J. B. and Agnew, A. D. Q. 1992. Positive-feedback switches 
in plant communities. – Adv. Ecol. Res. 23: 263–336.
Zeng, Y. and Malanson, G. P. 2006. Endogenous fractal dynamics 
at alpine treeline ecotones. – Geogr. Anal. 38: 271–287.
Zeng, Y. et al. 2007. Geomorphological limits to self-organization 
of alpine forest-tundra ecotone vegetation. – Geomorphology 
91: 378–392.
282
Box A. Examples
A1 Gradual treelines
Controlling processes: growth only (null-model of treeline form). Growth declines with elevation, but must also increasingly decrease 
with tree height (i.e. with coupling to the atmosphere) with elevation. Alternatively, in the case of multi-species treelines, a gradual 
height decline may also be due to species turnover, from tree-forming species to small trees and shrubs. Thus, possibly in addition to 
decreased growth, inter-specific interactions and species-specific establishment chances play a role in forming such patterns.
Examples:
We are not aware of the existence of pure cases of single-species gradual treelines in nature, as some level of deformation is always 
observed in the smallest trees. It appears that very small tree individuals at treeline, which may be several decades years old and only 
50-cm tall (Camarero and Gutierrez 1999), will always experience some level of damage during their long exposure to the alpine 
climate. Even if they are not strictly krummholz, such small trees are very likely to lose their main meristems and to experience stem 
bending at one or several occasions during their lifetime, making them grow as multi-stemmed, bushy and/or twisted individuals. 
However, the level of deformation varies, and if light, treelines may be considered as gradual discrete without krummholz.
Multi-species gradual treelines are more common and are known from very wet mild-temperate to tropical mountains, like the west 
coast of the southern Alps of New Zealand (Wardle 1965) and the eastern cordillera of the Andes in southern Ecuador (Bader et al. 
2007a).
Figure A1a. Hypothesized first-level controls on gradual single-species treelines. Growth limitation (dotted orange line) is the strongest 
limitation at such treelines.
Figure A1b. Schematic example of a gradual discrete treeline.
283
A2 Krummholz treelines
Controlling processes: this type of treeline is considered by some authors as being typical for a natural treeline (Ellenberg 1966, 
Tranquillini 1979), growth and biomass maintenance of tree species being increasingly limited by a combination of low growth tem-
peratures and other stress factors like wind and frost, but tree cover being closed, i.e. mortality not limiting tree occurrence, up to the 
tree-species line.
As explained above for gradual treelines, the level of deformation in gradual treelines can be quite variable, ranging from none 
(in theory) to very strong deformation leading to krummholz. In multi-species treelines, the turn-over from trees to shrubs may be 
accompanied by an additional environmental stunting of both trees and shrubs, in which case a gradual multi-species krummholz 
treeline would emerge. This is the case e.g. on very wet and windy ridges in southern Ecuador (Bader et al. 2007a).
Examples:
Examples of single-species gradual discrete krummholz treelines include natural treelines formed by Erica trimera on the Bale moun-
tains of Ethiopia (Miehe and Miehe 1994), and a Fuscospora cliffortioides (previsouly known as Nothofagus solandri var. cliffortioides, 
Heenan and Smissen 2013) treeline described for a wind-exposed, shallow-soil or snow accumulating position in the southern Alps of 
New Zealand (Wardle 1965, Norton and Schönenberger 1984), and some sections of Nothofagus pumilio treelines in southern South 
America (Holtmeier 2009). In both cases, nearby treelines tend to be abrupt, which sometimes appears to be a sign of disturbance, e.g. 
by fire in the Bale Mountains. In these mountains, natural treelines are actually something in between a discrete and a diffuse form, 
with shrubby Erica being about 1.6 m tall at the discrete transition from closed scrub to alpine dwarf shrubs, while above that single 
dome-shaped Erica shrubs occur that decrease gradually to about 0.5 m in height (Miehe and Miehe 1994).
The case study from New Zealand is particularly interesting, because the dominant treeline form of F. cliffortioides as well as other 
Nothofagaceae in New Zealand and in southern South America is abrupt and discrete, tall forest giving way to alpine tussock grassland 
(Cuevas 2000, Wardle 2008, Case and Hale 2015). The difference between the gradual and the abrupt Nothofagaceae treelines indi-
cates that different processes shape these treelines. While establishment is more limited than growth at the abrupt treelines, gradual 
Nothofagaceae treelines tend to occur in situations where trees are damaged by wind or excess snow cover, causing the strong deforma-
tion and the gradual decline in tree height so unusual for this part of the world (Wardle 1965).
Figure A1c. Gradual discrete multi-species treelines in the temperate zone (left) and the tropics (right): mixed broadleaf-Podocarp 
forest grading into tussock grassland, Kelly’s Saddle, west slope of the southern Alps, New Zealand (photo: M. Y. Bader) and mixed 
evergreen cloud forest, Podocarpus National Park, Ecuador (photo: M. Y. Bader).
Figure A2a. Hypothesized first-level controls on gradual discrete krummholz treelines. Dieback (orange dashed line) is the strongest 
limitation, while growth limitation (dotted orange line) also plays an important role at such treelines.
284
A3 Diffuse treelines
Diffuse treelines without krummholz
Controlling processes: growth is limiting (gradual size decline is implied for diffuse treelines if an abrupt size decline is not explicitly 
mentioned) and survival (especially during establishment) is restricted to increasingly few microsites. This interpretation is especially 
likely in heterogeneous terrain. Alternatively, even on a hypothetical homogenous slope a diffuse pattern can arise if the treeline is 
actively advancing and the spread-out trees are thus young outposts, or if it has had one or more advance episodes in the past, implying 
temporal heterogeneity combined with spatial stochasticity in establishment chances.
Examples:
In dry areas, like the example of Sajama in Bolivia (Fig. A3d), the forest is already open also at lower elevations. Such treelines are 
considered diffuse only if the spacing becomes more open towards higher elevation in the ecotone.
Gradual discrete treelines are found in many mountains, though an absolute absence of dieback is rare, and this appears to be a 
natural pattern, although in some cases it may be the result of land use opening up the forest at formerly discrete treelines (Tranquillini 
1979). For example, among typically discrete Nothofagus treelines in southern South America, some diffuse ecotones composed of the 
same species can also be found. For this, people and their livestock have been held responsible (Ellenberg 1966).
Diffuse treelines with krummholz
Controlling processes: dieback limits tree stature, likely in combination with reduced growth, and survival is restricted to increasingly 
few microsites (spatial heterogeneity) or controlled by spatial stochasticity combined with temporal heterogeneity (see above).
Examples:
Diffuse krummholz treelines are a rather common treeline form, as single trees above the closed forest are prone to wind damage and 
to winter desiccation (Tranquillini 1979).
As discussed above (A2), some relatively undisturbed Ethiopian Erica treelines form a combination of discrete and diffuse krumm-
holz treeline, the forest being closed down to a tree size of about 1.6 m, followed by a zone of dispersed Erica shrubs/krummholz.
In situations where existing krummholz is the main source of sheltered sites it is likely that krummholz islands emerge (A7).
Figure A3a. Hypothesized first-level controls on gradual diffuse treelines without (left) and with (right) krummholz formation. 
Mortality (red line), including establishment failure outside of safe sites, controls the spacing of trees, while growth limitation (dotted 
orange line) controls the size of existing trees in the case without krummholz (left), while additionally dieback controls the size of trees 
in the case with krummholz (right).
Figure A3b. Schematic example of a gradual diffuse treeline without (left) and with krummholz (right).
285
Figure A3c. Gradual diffuse treeline: Natural treeline in the Alps composed of Larix decidua, Stavel Crastu, Bosco/Gurin, Switzerland 
(photo: Matthias Jochner).
Figure A3d. Gradual diffuse krummholz treeline composed of Polylepis tarapacana, Sajama, Bolivia (photo: M. Y. Bader).
Figure A3e. Gradual diffuse krummholz treeline composed of Pinus uncinata, Ordesa, Pyrenees, Spain (photo: M. Y. Bader).
286
A4 Abrupt treelines
Controlling processes: survival outside existing forest is limited at an elevation where growth is not yet much limited. This may be due 
to low establishment in the harsh alpine conditions, especially for shade-dependent tree species, or to disturbance, e.g. recurrent burn-
ing of the alpine vegetation. At multi-species abrupt treelines, it is more likely that their origin and maintenance are anthropogenic, 
because with multiple species present it is unlikely that none would be able to establish in the alpine zone.
Examples:
The Nothofagaceae (southern beech) treelines in New Zealand and southern South America are the most well-known abrupt treelines, 
and the only ones for which an anthropogenic origin of this pattern is unlikely. No fossil tree wood charcoal has ever been discovered 
above treeline in New Zealand (McGlone 2004) and fire is unlikely to cause the regular landscape-scale pattern observed for most 
New Zealand treelines. It is instead suggested as an explanation for the unusually irregular treelines in the Tararua and Ruahine Ranges 
(Wardle 1965). How the regular and abrupt treelines in New Zealand came to be, is a puzzling question still awaiting an answer. 
However, their maintenance can be explained by the shade-dependence and frost-intolerance of the tree seedlings (Wardle 1965, 
1985, Harsch et al. 2012), causing a positive feedback between tree establishment and microclimate that maintains the abrupt bound-
ary (Wilson and Agnew 1992). A similar reasoning applies to Nothofagus treelines in South America (Cuevas 2000).
Abrupt discrete treelines are also found in Fagus sylvatica (European beech) in southern Europe, but here it appears to be mostly a 
result of historic land use in this long-inhabited region, combined with the difficulty of this species to re-establish outside shade. The 
anthropogenic nature of the abrupt F. sylvatica treelines is indicated by the existence of more gradual and diffuse F. sylvatica treelines 
in the Apennines, located at higher elevations than the abrupt treelines of this species (Bonanomi et al. 2018).
Abrupt treelines can also be composed of quite open forests, as found on e.g. Tenerife (Canary Islands) and the volcanos of central 
Mexico (Wardle 1965, Lauer 1978), with treelines of Pinus canariensis and Pinus hartwegii, respectively. As these forests are quite open 
also at lower elevations, these treelines, apart from being abrupt in terms of tree stature change, are not diffuse but discrete (Leuschner 
1996). However, at least for P. hartwegii some sections of the treeline may also be more diffuse (Lauer 1978, Astudillo-Sánchez et al. 
2019). In both regions, the treeline has been strongly affected by land use and fire, which may prevent tree establishment (Lauer 
1978), so that the abrupt discrete shape is maintained.
Many multi-species tropical cloud-forest treelines are also very abrupt, which is probably due to fire in the alpine tussock grass-
lands, possibly combined with the difficulty of many tree species to establish outside forest cover (Bader et al. 2007b, Llambí et al. 
2013, Rehm and Feeley 2015). Although tall shrub species are usually present that can tolerate the high radiation in the alpine zone 
and that could start a succession towards forest, this process is rather slow and, at many treelines, additionally hindered by frequent 
burning (Bader et al. 2007a, Ramírez et al. 2009, Llambí et al. 2013, Bueno and Llambí 2015). At many abrupt cloud-forest treelines, 
a fringe of such shrubs (e.g. Diplostephium spp., Hesperomeles spp., Miconia spp.) lines the abrupt treeline forest border (Bader et al. 
2007a, Ramírez et al. 2009).
Figure A4a. Hypothesized first-level controls on abrupt discrete treelines. Mortality (red line) controls these treelines, the strong 
change in the gradient caused by the change in microclimatic conditions between forest and alpine zone, leading to a positive feedback 
switch (Wilson and Agnew 1992).
287
Figure A4b. Schematic example of an abrupt discrete treeline.
Figure A4c. Abrupt discrete single-species treeline composed of Lophozonia menziesii (Nothofagaceae), southern Alps, New Zealand 
(photo: M. Y. Bader).
Figure A4d. Abrupt discrete treeline of mixed cloud forest bordering tussock-grass paramo in the Guandera reserve of northern 
Ecuador. Although not currently used for livestock grazing, this paramo is regularly burned. A recent fire is clearly indicated by the 
charred stems of the Espeletia stem rosettes in the foreground, and burnt young growth of the small tree Gaiadendron punctatum show-
ing dark orange in the background (photo: M. Y. Bader).
288
A5 Abrupt diffuse treelines
Controlling processes: survival is restricted to increasingly few microsites and becomes impossible where growth is not yet much lim-
ited. In the case of a dense shrub cover between the trees, competition is a likely cause of establishment limitation (Liang et al. 2016, 
Schwab et al. 2016).
Examples:
Two examples: ‘Several Juniper species occurring from fairly dense stands to very open woodlands and finally to single isolated indi-
viduals’ on south-facing slopes in the eastern Himalayas (Schickhoff 2005), or treeline in Bhutan (Miehe et al. 2015), where trees also 
spread out (Fig. A5c). In both cases the trees grow in a matrix of tall Rhododendron shrubs, which probably prevents small trees from 
persisting. This likely leads to the pattern that only well-grown trees are found, but fewer and fewer of these with increasing elevation, 
as the competition balance shifts from trees to shrubs. Coupling to the atmosphere, which is related to plant height, should not differ 
between small trees and tall shrubs of the same height. Advantages of shrubs may lie in a more homogenous canopy, raising the bound-
ary layer and presenting strong competition to establishing trees, and/or in their thinner stems, requiring a shorter growing season 
(Treml et al. 2019). However, the conceptual difference between the treeline and the (tall) shrubline has not been fully explored so far.
Examples of abrupt diffuse treelines without a competing shrub belt include sections of P. hartwegii treelines in Mexico (see also 
above under A4). In this system, tree establishment indeed appears to be restricted to certain microsite types (causing a diffuse pat-
tern, Astudillo-Sánchez et al. 2019), while the abruptness, i.e. the absence of small trees, may be related to mortality due to fire (Lauer 
1978, Robles-Gutiérrez et al. 2016).
Figure A4e. Close-up of an abrupt discrete treeline of mixed cloud forest bordering tussock-grass paramo in the Guandera reserve of 
northern Ecuador. Note the fringe of tall shrub species like the bluish Diplostepium and Blechnum tree ferns (photo: M. Y. Bader).
Figure A5a. Hypothesized first-level controls on a diffuse treeline on a slope with spatial or temporal heterogeneity in mortality (left) 
or in a matrix of shrubs or other non-invasible vegetation (right). Mortality (red line) controls the positioning of trees, either due to 
heterogeneity or due to competition from the shrubs. Growth limitation and dieback (orange lines) do not directly determine this 
pattern, although growth does determine the competitive strength of trees versus shrubs, in particular after disturbances.
289
A6 Tree-island treelines
Controlling processes: survival depends on existing trees and becomes impossible away from these. In abrupt tree-island treelines, islands 
form at elevations where growth is not yet much limited and dieback does not play a large role. In gradual tree-island treelines, islands 
are additionally increasingly limited in size by growth limitation and/or dieback (then grading into krummholz-island treelines). One 
Figure A5b. Schematic example of an abrupt diffuse treeline in a shrub matrix, which is mostly gradual discrete itself.
Figure A5c. Natural treeline in Bhutan composed of Abies densa trees in Rhododendron lanatum thickets with Sorbus shrubs (photo: 
Georg Miehe).
Figure A6a. Hypothesized first-level controls on abrupt tree-island treelines. Mortality (red line) controls the positioning of trees and 
depends on the presence of trees, leading to a positive feedback switch (Wilson and Agnew 1992). Growth limitation and dieback 
(orange lines) are not important for this pattern.
290
possible feedback mechanism is reproduction through layering, allowing tree establishment where establishment from seed is not 
possible (Holtmeier 2009), or, similarly, growth into multi-stemmed trees from previously prostrate individuals (Devi et al. 2008). 
Feedback mechanisms involving establishment from seed include reduced seedling mortality near existing trees due to shading, shel-
tering or snow redistribution (Germino and Smith 1999, McIntire et al. 2016). Where tree performance is controlled by positive 
feedback, it is partially decoupled from external drivers and, for example, growth responses of trees to macroclimatic variation may 
be reduced (Elliott 2011).
Examples:
Examples of treelines with islands composed of tall trees are located in the Teton and Medicine Bay Mountains (Wyoming, USA) and 
in the Altai Mountains of Russia. In these sites, tree islands characterize the upper parts of the montane forest, thus creating a very long 
ecotone according to our definition. In the Tetons, trees continue to form islands towards the top of the ecotone and do get smaller and 
show signs of damage. Although we did not observe real krummholz, this may be because the slopes did not exceed the ecotone. In the 
Snowy Range (Medicine Bow Mountains), such forest islands (in that case ribbon-shaped) grade into krummholz islands, thus making 
that a gradual krummholz-island treeline (Germino and Smith 1999). Treelines with tree islands that are in fact multi-stemmed trees 
can be found, for example, in the Polar Urals, but also here krummholz formation occurs (Devi et al. 2008). Purely abrupt tree-island 
treelines may thus not exist in these mountains except where disturbances have removed the krummholz-island belt (perhaps in Fig. A6d).
In some cases, multi-species tropical treelines can have distinct forest tongues and islands that tend to be dominated by small tree and 
tall shrub species characteristic of the treeline (Fig. A6e). They can border the paramo vegetation quite abruptly or via a wider border 
ecotone. These tongues and islands can result from the presence of local topographic features (e.g. boulders, small depressions) but could 
also be promoted by a nucleation process resulting from initial facilitation of woody plant establishment by alpine shrubs (Ramírez et al. 
2009, Llambí et al. 2013). Such varied treelines are hard to typify but can be described in terms of intermediate and mixed forms.
Figure A6b. Schematic example of an abrupt tree-island treeline.
Figure A6c. Tree islands in a treeline ecotone composed of Abies lasiocarpa, Picea engelmannii and Pinus albicaulis, Tetons, USA 
(photo: M. Y. Bader).
291
A7 Krummholz-island treelines
Figure A6d. Abrupt diffuse treeline ecotone with tree-islands, composed of Abies lasiocarpa and Picea engelmannii, Medicin Bow 
Mountains, USA (photo: L. M. Resler).
Figure A6e. Treeline ecotone bordering the paramo of Iguaque, Colombia. In this case the ecotone is quite wide and has a gradual 
decrease in plant size from trees to small shrubs among tussock-grass paramo, with a mixture of diffuse, discrete and island sections, 
plus strong topography-related patterns (photo: M. Y. Bader).
Figure A7a. Hypothesized first-level controls on gradual (left) and abrupt (right) krummholz-island treelines. Mortality (red line) 
controls the positioning of trees and depends on the presence of trees, leading to a positive feedback switch (Wilson and Agnew 1992). 
Dieback (dashed orange line) is the strongest control on size and shape of existing tree-species islands and either increases gradually 
(left) or abruptly as a result of leaving the shelter of the forest (right), implying another positive feedback mechanism.
292
Controlling processes: dieback limits tree stature more strongly than growth and tree survival depends on existing trees or krummholz, 
creating a positive feedback.
Examples:
This type of treeline is very common in North-American mountains (Marr 1977, Pereg and Payette 1998, Butler  et  al. 2009b, 
Holtmeier 2009). Positive feedback, with stimulation of tree growth inside or on the lee of islands combined with suppression on 
the sides and windward side of the island due to wind and snow redistribution, maintains and shapes such islands (Malanson 1997, 
Alftine and Malanson 2004). Island shapes can range from rounded patches to stripes or fingers, and islands have also been observed 
to ‘walk’ across the tundra, growing on the lee side and shrinking on the windward side (Marr 1977). Often, these islands are com-
posed of two to three species, with the hardiest species establishing first (e.g. Pinus albicaulis), followed by species requiring more 
shelter (e.g. Abies lasiocarpa or Picea engelmannii).
Because the processes leading to krummholz formation are most pronounced at a height just above the average snow depth, while 
further from the ground these stresses decrease again, trees tend to get trapped as krummholz below this height. If able to escape the 
stressful layer, they may grow into tall trees. As a result, the height transition from trees to krummholz can be quite abrupt, with few 
or no intermediately-sized individuals.
Figure A7c. Krummholz islands composed of Abies lasiocarpa and Pinus albicaulis, Lee Ridge, Glacier NP, USA (photo: M. Y. Bader).
Figure A7b. Schematic example of a gradual krummholz-island treeline.