Ecosystems (2014) 17: 890–905
DOI: 10.1007/s10021-014-9767-3
Ó 2014 Springer Science+Business Media New York
Long-term Impacts of Contrasting
Management of Large Ungulates in
the Arctic Tundra-Forest Ecotone:
Ecosystem Structure and Climate
Feedback
Martin Biuw,1 Jane U. Jepsen,1* Juval Cohen,2 Saija H. Ahonen,3 Mysore
Tejesvi,3 Sami Aikio,3 Piippa R. Wäli,3,4 Ole Petter L. Vindstad,5 Annamari
Markkola,3 Pekka Niemelä,6,7 and Rolf A. Ims5
1
Norwegian Institute for Nature Research – NINA, 9296, Tromsö, Norway; 2Finnish Meteorological Institute, P.O. Box 503, 00101
Helsinki, Finland; 3Department of Biology, University of Oulu, P.O. Box 3000, 90014 Oulu, Finland; 4Kolari Unit, Finnish Forest
Research Institute, 95900 Kolari, Finland; 5Department of Arctic and Marine Biology, University of Tromsø, 9294 Tromsö, Norway;
6
Department of Biology, University of Turku, 20014 Turku, Finland; 7Kevo Subarctic Research Intitute, University of Turku, 20014
Turku, Finland
ABSTRACT
in land cover can substantially affect FTE
dynamics, alter ground albedo (index of the
amount of solar energy reflected back into the
atmosphere) and provide important feedbacks into
the climate system. We took advantage of a naturally occurring contrast between reindeer grazing
regimes in a border region between northern
Finland and Norway which was recently defoliated by an outbreak of the geometrid moth. We
examined ecosystem-wide contrasts between
potentially year-round (but mainly summer)
grazed (YRG) regions in Finland and mainly
winter grazed (WG) regions in Norway. We also
used a remotely sensed vegetation index and albedo to quantify effects on local energy balance
and potential climate feedbacks. Although differences in soil characteristics and ground vegetation
cover were small, we found dramatic differences
in the tree layer component of the ecosystem.
Regeneration of mountain birch stands appears to
have been severely hampered in the YRG regime,
by limiting regeneration from basal shoots and
reestablishment of individual trees from saplings.
The arctic forest-tundra ecotone (FTE) represents a
major transition zone between contrasting ecosystems, which can be strongly affected by climatic and biotic factors. Expected northward
expansion and encroachment on arctic tundra in
response to climate warming may be counteracted
by natural and anthropogenic processes such as
defoliating insect outbreaks and grazing/browsing
regimes. Such natural and anthropogenic changes
Received 30 July 2013; accepted 20 December 2013;
published online 8 April 2014
Electronic supplementary material: The online version of this article
(doi:10.1007/s10021-014-9767-3) contains supplementary material,
which is available to authorized users.
Author contributions: Martin Biuw: Performed research, Analyzed
data and Wrote the paper. Jane U. Jepsen: Conceived of or designed
study, Performed research, Analyzed data and Wrote the paper. Juval
Cohen: Analyzed data, contributed new methods or models. Saija H
Ahonen: Performed research, analyzed data. Mysore Tejesvi: Analyzed
data. Sami Aikio: Analyzed data. Piippa R Wäli: Performed research,
analyzed data. Ole Petter L. Vindstad: Performed research, analyzed data.
Annamari Markkola: Conceived of or designed study, performed research, analyzed data. Pekka Niemelä: Conceived of or designed study.
Rolf A. Ims: Conceived of or designed study
*Corresponding author; e-mail: jane.jepsen@nina.no
890
Ecosystem Structure and Climate Feedback
891
This has led to a more open forest structure and a
significant 5% increase in spring albedo in the
summer grazed compared to the winter grazed
regions. This supports recent suggestions that
ecosystem processes in the Arctic can significantly
influence the climate system, and that such processes must be taken into account when devel-
oping climate change scenarios and adaptation
strategies.
INTRODUCTION
bivory intensity can strongly influence the
vegetation density in the shrub and tree layer, and
recent studies have shown that grazing and
browsing by small and large mammals has the potential of even preventing shrub encroachment
into the tundra (Olofsson and others 2009, 2012;
Ravolainen and others 2011; Speed and others
2010). In addition to browsing by reindeer, occasional large scale tree defoliation by insect outbreaks can strongly affect the structure of
mountain birch forests and characterize their
regeneration and transition to treeless tundra vegetation (Hämet-Ahti 1963; Neuvonen and others
2001).
One of the most important ecological functions
of the FTE is the feedbacks that the structures of
woody vegetation in the shrub and tree layer can
have on the climate system. It has been shown that
snowmelt in high latitude tundra regions occurs
earlier and/or more rapidly where vegetation is
denser and protrudes above the snowpack (Grippa
and others 2005; Loranty and others 2011; Lundberg and Beringer 2005; Marsh and others 2010;
Pomeroy and others 2006). The reason is that the
total surface albedo is decreased and the absorption
of solar radiation increased by vegetation above the
snow (Bewley and others 2010; Ménard and others
2012; Sturm and others 2005a). Shrub encroachment can thus cause atmospheric warming in
spring and summer on a regional scale (Bonfils and
others 2012; Sturm and others 2005a). Such effects
on atmospheric temperature caused by encroachment represent an amplifying feedback mechanism
from the ecosystem back into the climate system
(Chapin and others 2005; Cohen and others 2013).
The flip side of this feedback process is that prevention of encroachment and forest expansion
onto arctic tundra may help reduce the rate of
atmospheric temperature increase. What role can
wild and semi-domesticated mammalian herbivores play in this context? Conservation grazing
(Duffey and others 1974) is a well-documented
effective land management tool that has been used
in, for example, land restoration (Bohner and
others 2012), improving river catchment function
The arctic forest-tundra ecotone (FTE) is the largest
transition zone between structurally contrasting
ecosystems on the planet, covering a large latitudinal gradient and a range of environmental conditions around the Arctic (Callaghan and others
2002). Ecotones in general are characterized by
their relatively high biodiversity (Harris 1988), and
the FTE is no exception. The great majority of
species targeted for monitoring by the ‘‘Arctic
Species Trend Index’’ are from this transition zone
(CAFF 2010). The FTE is characterized by very
open, and sometimes discontinuous, stands of lowstatured trees, which are liable to influence by a
host of edaphic, climatic, and biotic factors. It is a
highly dynamic system, which can respond relatively rapidly to changes in any of these factors, and
also to changes in human land use. Ambient temperature is one important driver of change, but
winter conditions such as snow and wind are also
important in structuring the distribution of trees
and shrubs (Dalen and Hofgaard 2005; Harper and
others 2011; Holtmeier and Broll 2011). In addition, various natural and anthropogenic disturbance regimes can strongly influence the patterns
of vegetation distribution and abundance (Aune
and others 2011; Jepsen and others 2008; Tenow
and Bylund 1989).
In a warming climate the FTE is expected to expand northward, and the arctic tundra to come
under threat from climate-driven shrub encroachment and forest expansion (Hofgaard and others
2013; Myers-Smith and others 2011; Olofsson and
others 2009; Tape and others 2006; Tømmervik and
others 2004). The most rapid response to warming
is expected to occur in the transition zone between
tall- and dwarf-shrub tundra (Epstein and others
2004; Lantz and others 2010), most strongly driven
by increasing air temperatures and its effects either
on shrub growth and reproductive potential directly, or mediated via melting permafrost and
snowmelt timing (Myers-Smith and others 2011).
In addition to such direct and indirect climatic effects, anthropogenic activities and changes in her-
Key words: Arctic vegetation; climate change;
insect defoliation; grazing; climate feedback; reindeer husbandry.
892
M. Biuw and others
(Nienhuis and others 2002) and increasing biodiversity (Takala and others 2012) in land areas
shaped by a long-standing combination of natural
succession patterns and human land-use regimes.
However, its potential use in mitigating the effects
of climate change has not been studied in any detail.
Reindeer/caribou (Rangifer tarandus) are the most
abundant and wide-spread large herbivores in
arctic and sub-arctic ecosystems. Throughout their
circumpolar distribution reindeer pastures range
from continental boreal forest to coastal tundra and
often with seasonal migrations between these two
ecosystems (Forbes and Kumpula 2009). Abundant
reindeer have been predicted to be able to induce
transitions between vegetation states (Van der Wal
2006), but this is likely to depend on ecosystem
structure (Bråthen and others 2007; Ims and others
2007) and on management regimes (Hausner and
others 2011), which may determine the intensity of
herbivory in different seasons. In terms of seasondependence, reindeer browse woody vegetation
mainly during the plant growing season (Klein
1990; Skjenneberg and Slagsvold 1968); thus, it is
expected that their impact on the structure of forest-tundra will be most pronounced in regions,
where they are present in the summer.
Reindeer have been present in northern Fennoscandia since the region was deglaciated 10,000–
15,000 BP (Skogland 1994). Like most reindeer
elsewhere in the Arctic, herds migrated seasonally
between the summer pastures in northern coastal
tundra and the inland winter pastures further
south. The winter pastures range from the tundraforest ecotone, which is formed by open stands of
mountain birch (Betula pubescens ssp. czerepanovii),
to the closed coniferous boreal forest (Figure 1).
Since the seventeenth century, reindeer herds have
become semi-domesticated and managed by the
indigenous Sami people (Muga 1986). In northeastern Norway the semi-domesticated herds have
maintained their seasonal migration pattern, but
beginning with the official closure of the Norwegian/Finnish border in 1852 by the Norwegian–
Russian authorities, the southern migrations into
the winter pastures in Finland became curtailed
(Vorren 1946). At the same time the reindeer herds
owned by Finnish herders effectively became cut
off from their traditional coastal summer pastures,
leading gradually to more sedentary herding practices. Thus for the last 100 years or so, Finnish
reindeer herding has been characterized by yearround presence within a relatively restricted region
in the forest-tundra ecotone, whereas Norwegian
herds have maintained a migration cycle with
reindeer presence in the forest-tundra primarily
during the winter.
In this paper, we make use of such long-standing
differences in reindeer grazing regimes between
Norway and Finland in the Polmak region in
northern Fennoscandia, to examine (1) the effects
of grazing on the structure and functioning of key
components of the birch forest ecosystem and (2)
to assess whether these differences are sufficient to
cause significant differences in the energy absorption capacity of the ground and vegetation to the
extent that it may provide a climate feedback signal
from the ecosystem. We here define the ecosystem
in the widest possible sense, and we have collected
data on most of its components covering soil
composition (nutrients, bacteria, and fungi),
ground vegetation community structure, forest
structure (tree structure and regeneration, tree
density and crown cover), herbivore presence and
satellite derived vegetation index [Normalized Difference vegetation Index (NDVI)]. To assess the
potential climate feedback from the system, we also
include a satellite derived index of energy reflectivity (albedo) in our analysis.
We expect that grazing and browsing can affect
the ecosystem along several non-exclusive pathways, summarized in Figure 2. Firstly, direct consumption by herbivores can affect the structure and
regeneration of both trees and ground vegetation.
The regeneration potential of the tree layer can be
substantially altered through consumption by herbivores of shoots and saplings (Hofgaard and others
2010), thereby potentially changing the tree morphology (for example, the number of trunks) as
well as forest structure (for example, distance between tree functional units or individuals). Grazing
of ground vegetation can alter the general species
composition but also change the height of growth
forms, in particular the shrub layer. These direct
effects on the tree layer and ground vegetation
community represent the most easily observed
ecosystem effects of herbivory, and may provide
the most direct climate feedback, for example, by
altering spring albedo (Cohen and others 2013).
However, grazing and browsing may also affect soil
composition, both through changes in the disturbance regime caused by trampling and burrowing,
but also via nutrient input from faeces and urine
(Stark and others 2007). Such changes in soil
characteristics may in turn influence the tree and
plant communities through alterations in nutrient
cycling and uptake aided by, for example, nitrogen
fixating bacteria and mycorrhiza.
Ecosystem Structure and Climate Feedback
893
Figure 1. The location of the Polmak study area (square) on the border between the reindeer herding districts Kaldoaivi in
Finland and Rákkonjárga in Norway. Full lines show year-round district boundaries. Inset figure in top left corner shows the
long term development in total herd size in the two districts. Source of background vegetation map: NORUT Northern
Research Institute, Tromsø, Norway.
MATERIALS
AND
METHODS
Study System
The study area is located in northern-boreal birch
forest near Polmak (‘‘Boulbmát’’), on the Norwegian–Finnish border in northern Fennoscandia
(28°E, 70°N, Figure 1). The area is characterized by
relatively mild winters and cool summers with
average temperatures around 12°C and -12°C in
the warmest and coldest months respectively, and
an annual precipitation of approximately 450 mm
(met.no, Rustefjelbma meteorological station). The
vegetation pattern in the study area is typical for
the FTE, with discontinuous areas of short-statured, shrub-like polycormic growth forms of
mountain birch and an understorey layer dominated by dwarf shrubs (Empetrum nigrum ssp. hermaphroditum, Vaccinium myrtillus, V. vitis-idaea). The
entire study area was severely defoliated during the
last outbreak of geometrid moth during 2002–2008
(Jepsen and others 2009), and many trees are dead
or remain almost completely defoliated. We positioned the experimental plots based on the occur-
rence of recent moth defoliation to include the
regeneration process into the experimental setup.
The study area lies within two distinct reindeer
districts; Kaldoaivi on the Finnish side and Rákkonjárga on the Norwegian side. Although we do
not have detailed historical information about the
seasonal usage by reindeer of the study area on
either side of the border, district-wide statistics on
total reindeer densities suggest that the historical
development in the two districts has been relatively
similar (Figure 1, inset) with the exception of a
period during the mid-90’s. In the 2006–2007
season, the Finnish Kaldoaivi district had an average density of 2.38 reindeer/km2 land area
(Kumpula and others 2008; Vuojala-Magga and
others 2011), while the corresponding number in
the Norwegian Rákkonjárga district was 1.56
(Reindriftsforvaltningen 2008). Today (2012) the
maximum allowable herd size is 5300 animals
(2.3 reindeer/km2) in Kaldoaivi and 4000 (1.7
reindeer/km2) in Rákkonjárga. In addition to this
fairly modest long-term difference in district-level
reindeer densities, a contrast exists in the seasonal
894
M. Biuw and others
Figure 2. Conceptual view of the top-down effects of grazing and browsing on key components of the birch forest
ecosystem, including potential climate feedback via changes in albedo.
use of the Polmak area in the two districts. In the
Norwegian Rákkonjárga district, herders move
their animals between coastal summer pastures on
the Varanger peninsula and winter pastures in the
interior. Grazing in the Norwegian section of the
study area is hence effectively limited to late autumn/winter with considerable variation between
years (Magne Andersen, Rakkonjárga district, pers.
comm.), whereas, in the Finnish section, the animals in principle have year-round access. Although
rotational grazing has often been enforced within
many Finnish districts by erecting fences between
designated summer and winter grazing areas
(Kumpula and others 2011), no such fences are
present within the Kaldoaivi district. However, the
main grazing in the Finnish Polmak region appears
to take place during the spring and summer
months, whereas the herds tend to move south and
east during the winter (Niko-Heikki Länsman,
Kaldoaivi district, pers. comm.).
Study Design
In 2011 an experimental herbivore exclosure/control setup was established as a basis for what is
intended as a long-term study of the influence of
mammalian herbivores on the regeneration
capacity and successional pathways of birch forest
ecosystems following severe outbreaks of geometrid moth (Jepsen and others 2013, 2009). The
experimental design was replicated on both sides of
the fenced Norwegian–Finnish border to capture
the contrasting grazing regimes in the two districts.
Briefly, the setup consists of six fenced 30 9 30 m
large herbivore exclosures on either side of the
border, each matched by an open 30 9 30 m control plot in the immediate vicinity. Within each
exclosure there are 10 caged 50 9 50 cm rodent
exclosures, again matched by 10 open controls
within the exclosure and 10 controls in the open
30 9 30 m plot. In the current study, we address
the state of the birch forest ecosystem under the
two contrasting grazing regimes at the time of
establishment of the experimental exclosures. The
field data were collected immediately after the exclosures were erected, and are assumed unaffected
by the presence of the exclosures. All study plots lie
within an area bounded by a rectangle extending
about 600 and 1200 m in the E–W and N–S directions, respectively. They are therefore expected to
be exposed to the same weather conditions, with
no systematic differences between the two sides of
the border. In particular, any systematic differences
in the duration of snow cover could clearly affect
our interpretations. However, we found no evidence for such a difference (see Appendix 1 in
supplementary material). In the following we treat
all 12 plots in each country as replicates and the
measurements in all the 50 9 50 cm sampling
squares were treated as observations nested within
these replicates. In addition to shedding light on
the long-term effects of grazing on the structure
Ecosystem Structure and Climate Feedback
and functioning of key components of the birch
forest ecosystem, this study will serve as a reference
for future ecosystem changes caused by the exclusion of herbivores.
Sampling and Measurements in the Field
Field registrations and sampling were carried out
during August and September 2011, and covered
the key components of the ecosystem presented in
Figure 2; soil nutrients, bacteria and fungi, understory vegetation community structure, birch stand
structure and regeneration. Table A2:1 in Appendix
2 (in supplementary material) includes a full list of
all measurements that were derived from field
registrations, including those taken on individual
trees.
Five soil cores (/ 3 cm) consisting of humus and
mineral layers separately were sampled at randomly assigned locations within each of the 24
study plots in Aug 2011 and were stored in plastic
bags at -20°C until used. Both humus and mineral
layer samples were used for pH, conductivity, total
N, dissolved P, Ca, Mg, and K analyzes conducted
in three replicate samples per plot with standard
methods (CHN analyzer for total N; ammonium
acetate extraction and colorimetric assay for dissolved P (John 1970) and ammonium acetate
extraction and atomic absorption spectrophotometry (AAS) for dissolved Ca, Mg, and K). DNA was
extracted from freeze-dried soil humus layer samples (five samples per plot pooled) using a PowerSoil DNA Isolation Kit (MoBio Laboratories, USA).
For bacteria, a portion of the 16S small-subunit
ribosomal gene was amplified using primers F519
and R806 with Ion Torrent adaptors (Roche, USA).
For fungi, the internal transcribed region was
amplified using ITS1F and ITS4 primers using GS
Junior (Roche, USA). PCR reactions for bacteria
and fungi were performed in 20 lL reactions in
three replicates by using a standard method. The
bacterial and fungal data were analyzed following
state-of-the-art procedures described elsewhere
using CLOVR and QIIME pipeline, respectively
(Caporaso and others 2011; White and others
2012). All the sequences were quality checked,
aligned, and grouped into operational taxonomic
units (OTUs) at a 97% sequence similarity cut-off.
We used numbers of fungal and bacterial OTUs, the
ratio of the number of bacterial to fungal OTUs, and
Simpson’s diversity index, all calculated per plot,
for further analyses of soil microbial communities.
The structure and composition of the understorey plant community within each 30 9 30 m plot
was estimated by recording the abundance of all
895
major plant species/species groups (Appendix 2,
Table A2:1 in supplementary material) within each
50 9 50 cm square. We used the point-intercept
method (Bråthen and Hagberg 2004), with 9 regularly spaced pins in a 40 9 40 cm (0.16 m2) aluminium frame. For each plant species the total
number of intercepts between plant parts and the 9
pins was recorded as a measure of plant abundance.
Mosses and lichens, which were not identified to
species, were recorded with a maximum of one
intercept per pin. A total of 27 different plant species or functional groups, representing 7 major
growth forms, were recorded.
Estimates of birch stand structure and regeneration within each 30 9 30 m plot were obtained by
collecting a series of morphometric and regeneration measures from 20 individually marked trees.
In a polycormic birch stand (that is, with multistemmed trees), the distinction between individual
trees is often unclear. In such situations, we defined a tree as a ‘‘functional’’ individual, consisting
of a cluster of trunks with no detectable root connections to neighboring groups of trunks. Trees
were selected by splitting each plot diagonal into 3m intervals, starting from the center point of the
plot, and then selecting the closest tree to each 3-m
interval. Key registrations on each tree included
the total number of basal shoots, the fraction of
basal shoots that showed evidence of having been
browsed by herbivores, and the severity of crown
defoliation. The latter was assessed by visually
estimating the proportion dead branches on each of
the three largest trunks on every study tree, and
assigned one of four classes; no defoliation, <50%
defoliation, >50% defoliation and completely
defoliated (that is, dead trunk). Although this study
did not specifically address the effects of defoliation
caused by the geometrid moth, we wanted to check
whether the degree of insect defoliation depended
on the level of browsing mammals, as has been
shown for other species of shrubs and trees (Danell
and Huss-Danell 1985; Olofsson and Strengbom
2000).
We also assessed regeneration from saplings, that
is, the establishment of new functional trees. The
number of saplings were counted within 1 m on
either side of the two diagonal transects within
each plot. Saplings were allocated to one of three
size classes; 0–20, 20–50, and 50–150 cm. To obtain
estimates of total tree density within each
30 9 30 m plot, we used aerial photographs obtained by Kite Aerial Photography (KAP). Our
setup consisted of a GoPro Hero 2 camera with a
170 degree field-of-view lens, suspended under a
one-line kite. The camera was programmed to take
896
M. Biuw and others
11 MP (3840 9 2880 pixels) still images once every
2 s, while positioning the kite approximately vertically above the center point of a plot at an altitude
of about 45 m. KAP images were georeferenced
using ground control points (GCPs) obtained with a
handheld GPS. Georeferencing was done in ARCGIS 9.3 using a spline transformation and 16–25
GCPs per 30 9 30 m plot. Georeferenced images
were converted to TIFFS with 2.5 cm pixel size
using cubic convolution resampling. Within each
30 9 30 m plot all trees were visually identified
from the georeferenced images, and their center
coordinates digitized manually. The distances (in
meters) between all trees within a given plot were
then calculated.
Finally, because any differences in the structure
and height of the shrub and tree layer are likely to
influence the snow distribution patterns in an area,
we measured the lower vertical limit of the epiphytic lichen Parmelia olivacea on birch stems (the
‘‘Olivacea-limit;’’ Sonesson and others 1994) as a
proxy for long-term average snow depth. The
Olivacea-limit was measured on each of the 20
individually marked trees within each 30 9 30 m
plot.
Remote Sensing Datasets
Albedo and NDVI data products come from the
Moderate Resolution Imaging Spectroradiometer
(MODIS) sensor on NASA’s Earth Observing System (EOS) Aqua and Terra satellites. Albedo is a
coefficient of reflectivity, and provides a measure of
the proportion of incoming solar radiation that is
reflected back by the earth’s surface. We used the
MCD43A3 albedo product, available at a nominal
spatial resolution of 500 m and a temporal resolution of 8 days. We used black sky albedo following
the guidelines of the World Meteorological Organization (WMO 2006). For NDVI, we used the
MOD13Q1 product from the Terra satellite, available at a nominal spatial resolution of 250 m and a
temporal resolution of 16 days.
We extracted albedo and NDVI data from grid
cells falling within either of two polygons, each
covering a region of Finnish or Norwegian terrain
around our study area. The polygons covered
roughly 7.2 and 7.9 km2 on, respectively, the
Finnish and Norwegian side of the border
(Appendix 2, Figure A2:1 and A2:2B in supplementary material). These polygons, with areas
several times greater than the minimum polygon
enclosing all our study plots (0.4 km2), were designed to maximize the number of valid grid cells in
the albedo and NDVI datasets, while still remaining
within habitat that was roughly similar to that in
which our study plots are situated. To further ensure that the summary statistics of albedo and NDVI
calculated within the respective polygons were as
representative as possible to the conditions in our
study plots, we used weighted summary statistics
(mean and standard deviation). The weighting was
based on a combination of (1) distance from the
centroids of plots on the Finnish and Norwegian
side separately, and (2) the absolute difference in
elevation between a particular grid cell and a reference elevation. The reference elevation here was
the average of the elevations at the two centroids.
Please see the Appendix 2 in supplementary
material for further details.
Based on satellite derived annual defoliation
maps (Jepsen and others 2009) for the years 2004–
2010, we calculated a weighted mean defoliation
score for each country polygon using the same
approach to data analysis and summary as presented above for NDVI and albedo data. The defoliation score express the % decrease in summer
NDVI in a given year relative to a reference year
without defoliation (see Jepsen and others 2009 for
details) and has been shown to correspond well
both with the relative abundance of geometrid
larvae in the canopy (Jepsen and others 2009) and
the proportion of dead and damaged tree trunks in
defoliated birch forest (Jepsen and others 2013).
Statistical Approaches
Our set of variables was collected at several different scales, which we refer to as explained in Table 1. Unless otherwise stated, statistical analyses
were performed using the R language (R Development Core Team 2011).
For univariate analyses of field measurements
collected at the 30 9 30 m plot level (n = 24), we
used Wilcoxon Rank Sum or a hurdle model to
assess the cross-border differences. Hurdle models
were developed for count data to simultaneously
test for differences in counts and the probability of
obtaining non-zero counts (Cameron and Trivedi
1998), and we used the implementation in the
hurdle function (Zeileis and others 2008) in the
pscl library. For univariate analyses of field measurements collected at finer scales, we used generalized linear mixed models (glmm) as a general
approach to assess the cross-border differences,
while controlling for the variation caused by below-plot grouping structures. We chose the error
distribution family and link function according to
the distribution of the response variable. For standard exponential family models, we used the lme4
Ecosystem Structure and Climate Feedback
Table 1.
897
Scales of Data Collection
Name
Meaning
Sample size (Finland + Norway)
Site
Plot
Diagonal
Tree
Trunk
Quadrat
Finland vs. Norway
30 9 30 m exclosure or control plots
Diagonals joining opposite corners in the 30 9 30 m plots
Individually marked trees
Up to 3 largest trunks on each individually marked tree
40 9 40 cm vegetation quadrats
1+1
12 + 12
24 + 24
238 + 2391
714 + 717
360 + 360
1
In one plot on the Finnish side only 18 trees were available whereas one tree on the Norwegian side had incomplete data records, resulting in sample sizes of 238 and 239 trees
in Finland and Norway, respectively.
Figure 3. The number and ratio of operational taxonomic units (OTUs) of bacteria and fungi in soil samples
obtained in the Polmak study plots, and the ratio of
fungal/bacterial OTUs. Symbols along the staples indicate
level of significance from Wilcoxon tests of difference
between Finnish (YRG) and Norwegian (WG) plots (n.s.
not significant, *significant at a = 0.05 level).
package (Bates and others 2012). For positive
continuous response variables (for example, distance measurements) we used a gamma family
available for glmm’s in AD Model Builder (Fournier
and others 2012) via the R package glmmADMB
(Skaug and others 2012). For ordinal response
variables (for example, tree trunk damage class) we
used a cumulative link mixed model (clmm)
available in the ordinal package (Christensen
2012). The fixed effect in all models was ‘‘Site’’
(that is, Finland vs. Norway), whereas the random
effects depended on the level at which our data
were collected. For instance if data were collected
at the vegetation quadrat or tree level, we included
plot as a grouping random variable.
Figure 4. A Overall abundance of each plant species/
species group in Finland (YRG) and Norway (WG). B
Biplot of nonparametric multidimensional scaling of
vegetation community at plot level (sum of all individual
quadrat abundances within plots). Each square or triangle
represents one 30 9 30 m plot in Finland (YRG) and
Norway (WG), respectively. The number indicates the plot
replicate, whereas the A and B indicates the two plots in
each pair.
898
M. Biuw and others
We examined the multivariate pattern of variation in soil nutrient content and vegetation community
composition
using
non-parametric
multidimensional scaling (NMDS) using the metaMDS function in the vegan package (Oksanen and
others 2011). To formally test if quadrats on the
Finnish and Norwegian sides showed overall differences we used the multi-response permutation
procedure (mrpp) as implemented in the mrpp
function in vegan.
RESULTS
Soil Characteristics
There were no systematic differences in terms of the
nutrient content of the soil between plots in the two
regions of different grazing regimes; potentially
year-round but mainly summer grazing on the
Finnish side (hereafter termed ‘‘YRG regime’’) and
mainly winter grazing on the Norwegian side
(hereafter termed ‘‘WG regime’’) of the border
(mrpp) with 999 iterations: A = -0.03, p = 0.087),
confirming that all experimental plots were located
on similar bedrock thus avoiding any biases caused
by, for example, the presence of limestone. The
number of operational taxonomic units (OTUs) of
bacteria ranged from 1122 to 2816, with higher
average numbers in the YRG regime than in the WG
regime (2369.1 ± 283.8 and 2041.8 ± 375.1
respectively, Wilcoxon Wa=0.05, df=11 = 108, p =
0.039, Figure 3). However, this did not translate into
a statistically significant difference in diversity, as
measured by the Simpson diversity index (Wilcoxon
Wa=0.05, df=11 = 88, p = 0.378). The numbers of OTUs
identified for fungi were an order of magnitude less
than what was identified for bacteria, ranging from
128 to 209. No differences were found either in
average numbers of fungal OTUs between YRG and
WG plots (152.2 ± 16.5 and 160.6 ± 21.3, respectively, Wilcoxon Wa=0.05, df=11 = 56, p = 0. 370), nor
in diversity measured by the Simpson diversity index
(Wilcoxon Wa=0.05, df=11 = 89, p = 0.347). However,
the ratio of fungal/bacterial OTUs was significantly
lower in the YRG regime than in the WG regime
(0.065 ± 0.009 and 0.081 ± 0.018, respectively;
Wilcoxon Wa=0.05, df=11 = 29, p = 0.012, Figure 3).
Understorey Vegetation Structure
There were large variations in vegetation community structure between quadrats, both within and
between study plots. Overall, the most abundant
species groups on both sides of the border were
Avenella flexuosa (Af), Empetrum nigrum (En) and
various moss species (Figure 4A). Non-parametric
multidimensional scaling did not indicate any systematic differences between YRG and WG plots
(mrpp) with 999 iterations: A = -0.0006,
p = 0.686), resulting in almost complete overlap
between the two groups of plots in a standard biplot (Figure 4B).
Birch Stand Structure
Defoliation
The great majority of trunks on both sides of the border were either dead or severely defoliated by geometrid moths, with more than 99% of trunks falling
within the top two defoliation classes (Table 2). Although the number of dead trunks was significantly
higher in the YRG regime, severely defoliated but live
trunks were significantly more common in the WG
regime. Despite this difference, there was no overall
difference in trunk damage attributed to the moth
outbreak between YRG and WG plots (clmm likelihood ratio test v2df ¼1 = 2.64, p = 0.104). The satellite
derived defoliation scores agreed well with these results. Although the degree of defoliation varied between years as the moth outbreak progressed
(Appendix 2, Figure A2:2A in supplementary material) and across space within the study area (Appendix
2, Figure A2:2B in supplementary material), there was
no systematic difference in defoliation values between
the YRG and WG regimes (Appendix, Figure 2:2A in
supplementary material).
Basal Shoots and Stems
We found clear evidence of a difference in tree
regeneration between the WG regime in Norway
and the YRG regime in Finland. Although the
number of basal shoots present on individual trees
ranged widely on both sides of the border (0 to 68),
only 8 of 238 trees in the YRG regime (3.4%) had
any basal shoots at all, whereas the corresponding
number in the WG regime was 153 of 239 (64%).
This large difference was highly statistically significant (Contingency table: v2df ¼1 = 193.5, p > 0.001).
On 5 out of the 8 trees in the YRG regime that had
any basal shoots, all shoots showed signs of browsing by herbivores. Also on the remaining three trees,
the majority of shoots had been browsed. A high
degree of browsing was observed also in the WG
regime, where 80% of trees had more browsed than
unbrowsed shoots and the proportion of browsed
shoots was generally skewed towards one. Despite
the high degree of browsing in both regions, we did
find evidence for a difference in tree morphology
between the YRG and WG regimes that is likely to
result from long-term differences in regeneration
Ecosystem Structure and Climate Feedback
Table 2.
899
Extent of Tree Defoliation
Finland (YRG)
Norway (WG)
pBoot
Non-defoliated
<50% defoliated
>50% defoliated
Dead
1 (0.16)
0 (0)
n.s.
1 (0.16)
2 (0.29)
n.s.
15 (2.35)
67 (9.84)
<0.001
620 (97.33)
612 (89.87)
<0.001
Significance tests of differences between Finnish and Norwegian plots are based on a permutation test on the percentages (see text).
Figure 5. Variations in the density of saplings within
each size class observed along diagonal transects within
study plots. Bars represent mean densities while error bars
indicate standard errors. See text for statistical test results.
capabilities from basal shoots. The number of stems
on individual trees varied substantially between
individual plots, but were in general significantly
lower in the YRG compared to the WG regime
(6.7 ± 5.0 and 9.3 ± 6.9 stems, respectively,
glmmpoisson likelihood ratio test v2df ¼1 = 15.72,
p < 0.001, Figure 6A).
Saplings
The evidence for a difference in recruitment from
seed was not as clear as for basal shoots. In the YRG
regime, birch saplings were encountered along
transects in 5 of 12 plots, whereas in the WG regime saplings were encountered along transects in
11 of 12 plots. The number of saplings decreased
dramatically with size class in both regimes (Figure 5). In the YRG regime no saplings taller than
0.2 m were found in any of the plots. The mean
sapling density (individuals per m2) from the
smallest class was substantially higher in the WG
regime (mean 0.082, median 0.011, Q5–95% 0–
0.449) compared to the YRG regime (mean 0.022,
median 0.000, Q5–95% 0–0.069). Results from the
poisson count component of the hurdle model fit-
Figure 6. A Summary of the number of stems on each
tree, B the number of trees counted on aerial images of
plots on the Finnish (YRG) and Norwegian (WG) sides,
and C histograms of nearest-neighbor distances between
trees. Although the number of trees did not differ significantly, the number of stems per tree, and the nearestneighbor distances were significantly greater on the
Finnish side (see text).
ted to these data showed a significant negative
trend in sapling numbers over the size classes
(contrast
parameterclass
1:2 = -0.93 ± 0.19,
p < 0.001; contrast parameterclass 1:3 = -2.59 ±
0.95, p < 0.01), and also significantly higher
counts in WG compared to YRG plots (contrast
parameterF:N = 0.99 ± 0.24, p < 0.001). Similarly,
results from the binomial zero hurdle component
(that is, the model component in which the response is the binomial indicator of presence/absence of saplings) of this model showed a strong
negative trend across the sapling size classes in
900
M. Biuw and others
YRG regime (72.0 (± 22.1) and 61.6 (± 18.1) cm,
respectively
(glmm,
likelihood ratio test:
v2df ¼1 = 3.98, p = 0.046). Figure 7 shows the general
difference in NDVI and albedo between YRG and WG
regimes from early spring to late summer. In early
spring, NDVI was low and similar on both sides of the
border, but as snowmelt progressed and NDVI increased in mid-April, values became significantly
higher in the WG regime. During the summer
months, NDVI remained slightly (1.5%) but significantly higher in the WG side. In contrast, spring
albedo before snowmelt was significantly and consistently about 5% lower in the WG regime compared to the YRG regime. This difference disappeared
as snowmelt progressed, and by early June values
were similar on both sides of the border (Figure 7).
Figure 7. Seasonal development of the difference in
NDVI and albedo between Finland (YRG) and Norway
(WG). Dots represent means whereas black bars represent
standard errors.
terms of the probability of obtaining non-zero
counts (contrast parameterclass 1:2 = -0.67 ± 0.67,
p < 0.323; contrast parameterclass 1:3 = -2.85 ±
0.93, p < 0.01), and also showed that this probability was significantly higher for WG plots compared to YRG plots (contrast parameterF:N = 2.18 ±
0.66, p < 0.0001). Reasonably high numbers of
saplings of the middle size class (saplings between
0.2 and 0.5 m) were observed in the WG regime
(mean 0.041, median 0.00, Q5–95% 0–0.139),
whereas no saplings of this size class were observed
in the YRG regime. For the largest size class, no
saplings were again observed in the YRG, and the
total number of saplings observed in the WG regime had dropped to 3.
The long-term effect of regeneration from saplings was marginally different between the two
regimes. The number of trees observed per plot in
the YRG regime was not significantly lower than in
the WG regime (28.4 ± 6.7 and 33.1 ± 9.0,
respectively, Wilcoxon Wa=0.05, df=11 = 48.5,
p = 0.183, Figure 6B). However, there was a small
but statistically significant difference in the nearestneighbour distance between trees within YRG plots
compared to within WG plots (4.9 ± 2.4 m and
4.3 ± 2.1 m respectively; glmmgamma likelihood
ratio test v2df ¼1 = 86.14, p < < 0.001, Figure 6C).
Snow Depth, NDVI, and Albedo
The long-term average maximum snow depth, as
estimated from the height of occurrence of P. olivacea
on the trunks of birch trees, was slightly but significantly higher in the WG regime compared to the
DISCUSSION
We have demonstrated how a wide approach to
ecosystem sampling in combination with remote
sensing dataset analyses can be used to describe the
multiple possible linkages between grazing/browsing and climate feedback in the arctic FTE system.
Although focused process studies and experimentation will be required to fully describe the complex
mechanisms and dynamics along as well as between distinct main pathways, this broad approach
nevertheless provides an informative snapshot of
the situation after a long period of contrasting
grazing/browsing regimes, and identifies the likely
main processes responsible for the observed differences. Indeed, our results can also be used to bring
attention to pathways and processes, where increased research effort may be particularly important and rewarding.
By far the clearest ecosystem contrasts between
Finnish year-round grazing and Norwegian winter
grazing plots were in the birch tree layer, most
dramatically through significantly fewer basal
shoots and fewer trunks on trees in YRG compared
to WG plots. We also found some evidence for a
reduction in tree recruitment from saplings in the
YRG regime which may explain the slightly reduced forest stand density compared to the WG
regime. Previous studies have demonstrated that
increased vegetation density and vegetation protruding from the snowpack can have a large effect
on, for instance, the timing and rate of snowmelt
(Grippa and others 2005; Loranty and others 2011;
Lundberg and Beringer 2005; Marsh and others
2010; Pomeroy and others 2006), via the reduction
in total surface albedo (Ménard and others 2012).
The significantly higher spring surface albedo on
the mainly summer-grazed Finnish side compared
Ecosystem Structure and Climate Feedback
to the winter-grazed Norwegian side demonstrated
in this study has the potential to substantially
contribute to higher energy absorption, as has been
shown by Cohen and others (2013) for a nearby
region, with resulting changes in soil temperature,
duration of the snow-free season, deepening of the
soil active layer, and so on.
We found no consistent differences in ground
vegetation cover between WG and YRG plots. This
lack of difference between grazing regimes is at
odds with expectations given that substantial impacts of reindeer grazing regime and intensity have
been documented on both cover and composition
of tundra and birch forest understorey vegetation
(Kumpula and others 2011; Pajunen and others
2008; Bråthen and others 2007). However, the
understorey vegetation in both the YRG and WG
study plots is likely to have been dramatically altered due to the preceding moth outbreak. Two
recent studies from the Fennoscandian birch forest
region (Jepsen and others 2013; Karlsen and others
2013) have shown that moth outbreaks of a similar
severity as experienced in our study region can
cause dramatic vegetation state transitions, in particular in the oligotrophic dwarf shrub birch forest
types that dominate our study site. It is likely that
any changes in understory vegetation structure
caused by long-term differences in grazing pressure
have been disguised by this dramatic disturbance
event, explaining why we did not find differences
between the grazing regimes in this study.
Quantitative estimates of the historical grazing
pressure by reindeer on either side of the border in
the Polmak study area are not available. The only
information we can rely on is district level overall
herd size statistics (Figure 1, inset), and some very
general information about herding practices. The
seasonal grazing practice in the Norwegian Rakkonjárga district results in reindeer only being
present in the Norwegian Polmak region from late
autumn/early winter to late winter/early spring,
with considerable variations between years. In
contrast, the apparently spontaneous south-eastern
movements by the Kaldoaivi herds in winter suggests that herds are present in the Finnish Polmak
region mainly during the summer. Indeed, wildlife
cameras deployed on our plots in 2012 confirm
these patterns. On the Finnish side, reindeer were
regularly observed in images from the autumn
period, disappeared around the time of the first
snow, and then reappeared again around snowmelt
(Appendix 1, Figure A1:1 in supplementary material). By comparison, on the Norwegian side reindeer were observed much more infrequently
during autumn, and were present in small numbers
901
throughout the winter. Importantly, no reindeer
were observed in images taken during the period
from snowmelt in mid-May until the camera records ended in early August (Appendix 1, Figure A1:1 in supplementary material). Taken
together, the historical records and our 2012/2013
data support our notion that whereas grazing can
potentially occur year-round on the Finnish side,
the Polmak region is grazed throughout the snowfree period, whereas the much more limited grazing on the Norwegian side occurs almost exclusively during autumn and winter. This may help
explain the fact that although we found clear differences in some components of the ecosystem
(that is, soil biota and in particular forest structure)
we found no differences in other components (that
is, vegetation community). Although reindeer can
browse intensively on mountain birch during
summer, they mainly feed on ground lichens during winter (Stark and others 2007). This can also
significantly affect the concentration of secondary
substances, litter decomposition rates and soil
nutrient pools. We found evidence for higher
diversity of soil bacteria in the summer grazed area
than in the winter grazed area. This is in concordance with previous results showing that intensive
herbivory can increase the importance of bacterial
decomposition processes in the soil (Fierer and
others 2009; Kaukonen and others 2013). In
summer, reindeer feces and urine act as fertilizers
in soil, and active soil biota will benefit from an
increase in available nutrients during the growing
season. Moreover, soil mechanical disturbance by
grazers, that is, trampling and local erosion, which
is more pronounced during summer grazing, could
also increase the ecological niches available, leading to higher bacterial or fungal OTU richness in
soil. In intensively grazed systems, soil nutrient
levels, and soil pH are usually also higher than in
ungrazed systems, and in tundra vegetation grazing
changes the vegetation community towards being
dominated by herbs rather than dwarf shrubs
which generally dominate in ungrazed but otherwise similar areas (van der Wal 2006). However,
we did not detect any differences in nutrient levels
between the grazing regimes. A possible explanation for this is that the samples were taken at the
end of the growing season (Aug–Sept). If grazing
caused an increase in available nutrients due to
reindeer feces dropped during the summer, it may
have been taken up by the plants and the additional bacterial fraction detected in the grazed area.
Precipitation levels during May–July 2011 were
much higher than the long-term (1962–2010)
average (249 mm as opposed to an average of
902
M. Biuw and others
141 mm) (Finnish Meteorological Institute, Kevo
station, Utsjoki, Finland) and the reindeer feces
may have mostly dissolved and nutrients released
and taken up and immobilized by soil microbes by
the end of summer when the samples were taken.
Some attempts have been made to identify the
mechanisms and estimate quantitatively the effects
that changes in land-surface can have on the
Earth’s energy budget. Recent trends in absorbed
radiation and atmospheric heating in arctic Alaska
have been mostly attributed to a lengthening of the
snow-free season, with only about 3% attributed to
land-surface changes such as increasing abundance
of shrubs and trees due to the small areal cover of
regions, where such changes have occurred (Chapin and others 2005). However, the conversion
from tundra to forest has the potential to increase
absorbed radiation and atmospheric heating almost
fivefold in spring and about 25% in summer, and
predicted large-scale changes from tundra to shrubs
(Myers-Smith and others 2011; Sturm and others
2005b) and forests (Grace and others 2002; Moen
and others 2004, but see Hofgaard and others 2010;
Lloyd and Fastie 2002) are expected to contribute
disproportionately to future warming (Chapin and
others 2005). Based on our results in this study,
and further supported by those of Cohen and others (2013), reindeer grazing appears to have the
potential to prevent the regeneration of forest after
naturally occurring defoliation events such as those
caused by geometrid moth. This suggests that moth
outbreaks and grazing together can play an
important role in halting the conversion from
tundra to forest, thereby reducing the strengthening feedback between encroachment and atmospheric warming.
The largest impact of snow cover on surface energy balance and its feedback on air temperature is
observed during the spring, when days are getting
longer and brighter and snow is still present but its
cover is becoming variable (Groisman and others
1994). However, the processes that affect the
ground cover sufficiently to cause changes during
the snowmelt season may occur at other times of
year. For instance, it appears that the main difference in the effects of reindeer is linked to whether
regions are browsed during summer or not. Cohen
and others (2013) found substantial differences in
timing of snowmelt and a roughly 5% difference in
spring albedo between regions of summer grazing in
Finland and non-summer grazing areas in Norway.
This is very similar to the estimated differences we
observed in Polmak, and according to Cohen and
others (2013) this represents an estimated difference in energy absorption of up to 6 W/m2 during
the snowmelt season, contributing about 0.5 W/m2
to the yearly energy budget on the non-summer
grazed areas in Norway. Their analyses were based
on remote sensing products also for vegetation
cover. Because they did not examine vegetation
composition, soil characteristics or tree layer structure on the ground, they could not address how
grazing/browsing acts upon various ecosystem
components to result in these climate feedback
differences. Furthermore, whereas Cohen and
others (2013) studied tundra vegetation (excluding
forested areas entirely) and hence the effects of low
shrub vegetation, we have here described how
similar differences in climate feedback caused by
browsing/grazing by reindeer can also be observed
in a mountain birch forest system representative of
the arctic FTE. The main pathway is via the effect of
browsing on the tree layer. The key mechanisms
appear to involve (1) the browsing of basal shoots,
which can substantially alter individual tree morphology and therefore the size of the crown layer
and (2) the browsing of newly established saplings,
thereby altering the structure and density of entire
stands and changing the crown cover. Indeed, it
appears that intense summer browsing can even
prevent the recovery of birch stands following intense moth outbreaks, by completely preventing
the establishment of new tree functional units.
ACKNOWLEDGMENTS
This study is a contribution from Work Package 4
(WP4) of the Nordic Centre of Excellence—How to
preserve the tundra in a warming climate (NCoETundra) funded by the Norden Top-Level Initiative
‘‘Effect studies and adaptation to climate change.’’
Additional funding was obtained from FRAM—High North Research Centre for Climate and the
Environment, the Norwegian Research Council,
the Norwegian Institute for Nature Research, the
University of Tromsø, and the Academy of Finland
(Project #138309). We thank Maja S. Kvalvik,
Lauri Kapari, Sabrina Schultze, Jakob Iglhaut,
Moritz Klinghardt, Elina Vainio, Ilkka Syvänperä
and Marianne Iversen for assistance during field
work, and Tuulikki Pakonen and Tarja Törmänen
for assistance with nutrient analyses.
REFERENCES
Aune S, Hofgaard A, Söderström L. 2011. Contrasting climateand land-use-driven tree encroachment patterns of subarctic
tundra in northern Norway and the Kola Peninsula. Can J For
Res 41(3):437–49.
Bates D, Maechler M, Bolker B. 2012. lme4: Linear mixed-effects
models using S4 classes. R package version 0.999999-0. http://
CRAN.R-project.org.
Ecosystem Structure and Climate Feedback
Bewley D, Essery R, Pomeroy J, Ménard C. 2010. Measurements
and modelling of snowmelt and turbulent heat fluxes over
shrub tundra. Hydrol Earth Syst Sci 14(7):1331–40.
Bohner A, Starlinger F, Koutecky P. 2012. Vegetation changes in
an abandoned montane grassland, compared to changes in a
habitat with low-intensity sheep grazing - a case study in
Styria, Austria. Eco Mont 4(2):5–12.
Fournier DA, Skaug HJ, Ancheta
Maunder MN, Nielsen A, Sibert
using automatic differentiation
highly parameterized complex
Methods Softw 27(2):233–49.
903
J, Ianelli J, Magnusson A,
J. 2012. AD Model Builder:
for statistical inference of
nonlinear models. Optim
Grace J, Berninger F, Nagy L. 2002. Impacts of climate change
on the tree line. Ann Bot 90(4):537–44.
Bonfils CJW, Phillips TJ, Lawrence DM, Cameron-Smith P, Riley
WJ, Subin ZM. 2012. On the influence of shrub height and
expansion on northern high latitude climate. Environ Res Lett
7(1):015503.
Grippa M, Kergoat L, Le Toan T, Mognard NM, Delbart N,
L’Hermitte J, Vicente-Serrano SM. 2005. The impact of snow
depth and snowmelt on the vegetation variability over central
Siberia. Geophys Res Lett 32(21):21412.
Bråthen KA, Hagberg O. 2004. More efficient estimation of plant
biomass. J Veg Sci 15(5):653–60.
Groisman PY, Karl TR, Knight RW. 1994. Observed impact of
snow cover on the heat balance and the rise of continental
spring temperatures. Science 263(5144):198–200.
Bråthen KA, Ims RA, Yoccoz NG, Fauchald P, Tveraa T, Hausner
VH. 2007. Induced shift in ecosystem productivity? Extensive
scale effects of abundant large herbivores. Ecosystems
10(5):773–89.
CAFF. 2010. Arctic biodiversity trends 2010—selected indicators
of change. Akureyri: CAFF International Secretariat.
Callaghan TV, Crawford RMM, Eronen M, Hofgaard A, Payette
S, Rees WG, Skre O, Sveinbjornsson J, Vlassova TK, Werkman
BR. 2002. The dynamics of the tundra-taiga boundary: an
overview and suggested coordinated and integrated approach
to research. AMBIO 12:3–5.
Cameron A, Trivedi P. 1998. Regression analysis of count data.
New York: Cambridge University Press.
Caporaso JG, Lauber CL, Walters WA, Berg-Lyons D, Lozupone
CA, Turnbaugh PJ, Fierer N, Knight R. 2011. Microbes and
Health Sackler Colloquium: Global patterns of 16S rRNA
diversity at a depth of millions of sequences per sample
[Microbiology]. Proc Natl Acad Sci 108(Suppl 1):4516–22.
Chapin FS, Sturm M, Serreze MC, McFadden JP, Key JR, Lloyd
AH, McGuire AD, Rupp TS, Lynch AH, Schimel JP, Beringer J,
Chapman WL, Epstein HE, Euskirchen ES, Hinzman LD, Jia G,
Ping CL, Tape KD, Thompson CDC et al. 2005. Role of landsurface changes in Arctic summer warming. Science
310(5748):657–60.
Christensen RHB. 2012. Ordinal—regression models for ordinal
data. R package version 2013.9-30. http://CRAN.R-project.org.
Cohen J, Pulliainen J, Ménard CB, Johansen B, Oksanen L,
Luojus K, Ikonen J. 2013. Effect of reindeer grazing on
snowmelt, albedo and energy balance based on satellite data
analyses. Remote Sens Environ 135:107–17.
Dalen L, Hofgaard A. 2005. Differential regional treeline
dynamics in the Scandes Mountains. Arct Antarct Alp Res
37(3):284–96.
Danell K, Huss-Danell K. 1985. Feeding by insects and hares on
birches earlier affected by moose browsing. Oikos 44(1):75–
81.
Duffey E, Morris MG, Sheail J, Ward LK, Wells DA, Wells TCE.
1974. Grassland ecology and wildlife management. London:
Chapman and Hall.
Epstein HE, Beringer J, Gould WA, Lloyd AH, Thompson CD,
Chapin FS, Michaelson GJ, Ping CL, Rupp TS, Walker DA.
2004. The nature of spatial transitions in the Arctic. J Biogeogr
31(12):1917–33.
Fierer N, Strickland SM, Liptzin D, Bradford AM, Cleveland CC.
2009. Global patterns in belowground communities. Ecol Lett
12(11):1238–49.
Forbes BC, Kumpula T. 2009. The ecological role and geography
of reindeer (Rangifer tarandus) in Northern Eurasia. Geogr
Compass 3(4):1356–80.
Hämet-Ahti L. 1963. Zonation of the mountain birch forests in
northernmost Fennoscandia: Selostus. Annales Botanici Societatis Zoologicæ 4(34):1–360.
Harper KA, Danby RK, De Fields DL, Lewis KP, Trant AJ, Starzomski BM, Savidge R, Hermanutz L. 2011. Tree spatial pattern within the forest-tundra ecotone: a comparison of sites
across Canada. Can J For Res 41(3):479–89.
Harris LD. 1988. Edge effects and conservation of biotic diversity.
Conserv Biol 2(4):330–2.
Hausner VH, Fauchald P, Tveraa T, Pedersen E, Jernsletten JL,
Ulvevadet B, Ims RA, Yoccoz NG, Brathen KA. 2011. The
ghost of development past: the impact of economic security
policies on Saami pastoral ecosystems. Ecol Soc 16(3):4.
Hofgaard A, Lokken JO, Dalen L, Hytteborn H. 2010. Comparing
warming and grazing effects on birch growth in an alpine
environment - a 10-year experiment. Plant Ecol Divers
3(1):19–27.
Hofgaard A, Tømmervik H, Rees G, Hanssen F. 2013. Latitudinal
forest advance in northernmost Norway since the early 20th
century. J Biogeo 40(5):938–49.
Holtmeier FK, Broll G. 2011. Response of Scots Pine (Pinus sylvestris) to warming climate at its altitudinal limit in northernmost subarctic Finland. Arctic 64(3):269–80.
Ims RA, Yoccoz NG, Brathen KA, Fauchald P, Tveraa T, Hausner
V. 2007. Can reindeer overabundance cause a trophic cascade? Ecosystems 10(4):607–22.
Jepsen JU, Biuw M, Ims RA, Kapari L, Schott T, Vindstad OPL,
Hagen SB. 2013. Ecosystem impacts of a range expanding
forest defoliator at the forest-tundra ecotone. Ecosystems
16(4):561–75.
Jepsen JU, Hagen SB, Hogda KA, Ims RA, Karlsen SR, Tommervik H, Yoccoz NG. 2009. Monitoring the spatio-temporal
dynamics of geometrid moth outbreaks in birch forest using
MODIS-NDVI data. Remote Sens Environ 113(9):1939–47.
Jepsen JU, Hagen SB, Ims RA, Yoccoz NG. 2008. Climate change
and outbreaks of the geometrids Operophtera brumata and Epirrita autumnata in sub-arctic birch forest: evidence of a recent
outbreak range expansion. J Anim Ecol 77:257–64.
John M. 1970. Colorimetric determination of phosphorus in soil
and plant materials with ascorbic acid. Soil Sci 109(4):214–20.
Karlsen SR, Jepsen JU, Odland A, Ims RA, Elvebakk A. 2013.
Outbreaks by canopy-feeding geometrid moth cause statedependent shifts in understorey plant communities. Oecologia
173:859–70.
Kaukonen M, Ruotsalainen AL, Wäli PR, Männistö MK, Setälä
H, Saravesi K, Huusko K, Markkola A. 2013. Moth herbivory
enhances resource turnover in subarctic mountain birch forests? Ecology 94(2):267–72.
904
M. Biuw and others
Klein DR. 1990. Variation in quality of caribou and reindeer
forage plants associated with season, plant part, and phenology. Rangifer Spec Issue 3:123–30.
Kumpula J, Stark S, Holand O. 2011. Seasonal grazing effects by
semi-domesticated reindeer on subarctic mountain birch forests. Polar Biol 34(3):441–53.
Kumpula J, Tanskanen A, Colpaert A, Anttonen M, Törmänen
H, Siitari J, Siitari S. 2008. Pasture survey of northern reindeer
herding area in Finland—results from 2005–2008 and changes
in the condition of pastures. Loppuraportti: Riista- ja porotutkimuslaitos.
Lantz TC, Gergel SE, Kokelj SV. 2010. Spatial heterogeneity in
the Shrub Tundra Ecotone in the Mackenzie Delta Region,
Northwest Territories: implications for Arctic environmental
change. Ecosystems 13(2):194–204.
Lloyd A, Fastie C. 2002. Spatial and temporal variability in the
growth and climate response of treeline trees in Alaska. Clim
Change 52:481–509.
Loranty MM, Goetz SJ, Beck PSA. 2011. Tundra vegetation
effects on pan-Arctic albedo. Environ Res Lett 6(2):
024014.
Lundberg A, Beringer J. Albedo and snowmelt rates across a
tundra-to-forest transition. In: Proceedings of the 15 northern
research basins international symposium and workshop.
Department of Water Resources Engineering, Lund University; 2005. pp. 1–10.
Marsh P, Bartlett P, MacKay M, Pohl S, Lantz T. 2010. Snowmelt
energetics at a shrub tundra site in the western Canadian
Arctic. Hydrol Process 24(25):3603–20.
Ménard CB, Essery R, Pomeroy J, Marsh P, Clark DB. 2012. A
shrub bending model to calculate the albedo of shrub-tundra.
Hydrol Process . doi:10.1002/hyp.9582.
Moen J, Aune K, Edenius L, Angerbjorn A. 2004. Potential effects of climate change on treeline position in the Swedish
mountains. Ecol Soc 9(1):16.
Muga DA. 1986. A commentary on the historical transformation
of the Sami communal mode of production. J. Ethnical Stud
14:111–21.
Myers-Smith IH, Forbes BC, Wilmking M, Hallinger M, Lantz T,
Blok D, Tape KD, Macias-Fauria M, Sass-Klaassen U, Levesque
E, Boudreau S, Ropars P, Hermanutz L, Trant A, Collier LS,
Weijers S, Rozema J, Rayback SA, Schmidt NM et al. 2011.
Shrub expansion in tundra ecosystems: dynamics, impacts
and research priorities. Environ Res Lett 6(4):045509.
Neuvonen S, Ruohomäki K, Bylund H, Kaitaniemi P. 2001. Insect herbivores and herbivory effects on mountain birch forest
dynamics. In: Wielgolaski F, Ed. Nordic mountain birch ecosystems. Man and the biosphere series, Vol. 27. New York:
Paris and Parthenon Publishing. p 207–22.
Nienhuis PH, Buijse AD, Leuven R, Smits AJM, De Nooij RJW,
Samborska EM. 2002. Ecological rehabilitation of the lowland
basin of the river Rhine (NW Europe). Hydrobiologia 478(1–
3):53–72.
Oksanen J, Blanchet FG, Kindt R, Legendre P, Minchin PR,
O’Hara RB, Simpson GL, Solymos P, Stevens MHH, Wagner H.
2011. R package version 2.0-5. http://CRAN.R-project.org.
Olofsson J, Oksanen L, Callaghan T, Hulme PE, Oksanen T,
Suominen O. 2009. Herbivores inhibit climate-driven shrub
expansion on the tundra. Glob Change Biol 15(11):2681–
93.
Olofsson J, Strengbom J. 2000. Response of galling invertebrates
on Salix lanata to reindeer herbivory. Oikos 91(3):493–8.
Olofsson J, Tommervik H, Callaghan TV. 2012. Vole and lemming
activity observed from space. Nat Clim Change 2(12):880–3.
Pajunen A, Virtanen R, Roininen H. 2008. The effects of reindeer
grazing on the composition and species richness of vegetation
in forest-tundra ecotone. Polar Biol 31:1233–44.
Pomeroy JW, Bewley DS, Essery RLH, Hedstrom NR, Link T,
Granger RJ, Sicart JE, Ellis CR, Janowicz JR. 2006. Shrub
tundra snowmelt. Hydrol Process 20(4):923–41.
R Development Core Team. 2011. R: a language and environment for statistical computing. Vienna: R Development Core
Team.
Ravolainen VT, Brathen KA, Ims RA, Yoccoz NG, Henden JA,
Killengreen ST. 2011. Rapid, landscape scale responses in
riparian tundra vegetation to exclusion of small and large
mammalian herbivores. Basic Appl Ecol 12(8):643–53.
Reindriftsforvaltningen. 2008. Ressursregnskap for reindriftsnæringen. Reindriftsforvaltningen.
Skaug H, Fournier D, Nielsen A, Magnusson A, Bolker B. 2012.
Generalized Linear Mixed Models using AD Model Builder. R
package version 0.7.3. http://CRAN.R-project.org.
Skjenneberg S, Slagsvold L. 1968. Reindriften og dens naturgrunnlag. Oslo: Universitetsforlaget.
Skogland T. 1994. Villrein: fra urinnvåner til miljøbarometer.
Oslo: Teknologisk Forlag.
Sonesson M, Osborne C, Sandberg G. 1994. Epiphytic lichens as
indicators of snow depth. Arct Alp Res 26(2):159–65.
Speed JDM, Austrheim G, Hester AJ, Mysterud A. 2010.
Experimental evidence for herbivore limitation of the treeline.
Ecology 91(11):3414–20.
Stark S, Julkunen-Tiitto R, Kumpula J. 2007. Ecological role of
reindeer summer browsing in the mountain birch (Betula
pubescens ssp. czerepanovii) forests: effects on plant defense,
litter decomposition, and soil nutrient cycling. Oecologia
151(3):486–98.
Sturm M, Douglas T, Racine C, Liston GE. 2005a. Changing
snow and shrub conditions affect albedo with global implications. J Geophys Res 110(G1):G01004.
Sturm M, Schimel J, Michaelson G, Welker JM, Oberbauer SF,
Liston GE, Fahnestock J, Romanovsky VE. 2005b. Winter
biological processes could help convert Arctic Tundra to
shrubland. Bioscience 55(1):17–26.
Takala T, Tahvanainen T, Kouki J. 2012. Can re-establishment
of cattle grazing restore bryophyte diversity in abandoned
mesic semi-natural grasslands? Biodivers Conserv 21(4):
981–92.
Tape K, Sturm M, Racine C. 2006. The evidence for shrub
expansion in Northern Alaska and the Pan-Arctic. Glob
Change Biol 12:686–702.
Tenow O, Bylund H. 1989. A survey of winter cold in the
mountain birch/Epirrita autumnata system. Memoranda Societatis Pro Fauna Et Flora Fennica 65:67–72.
Tømmervik H, Johansen B, Tombre I, Thannheiser D, Hogda KA,
Gaare E, Wielgolaski FE. 2004. Vegetation changes in the
Nordic mountain birch forest: the influence of grazing and
climate change. Arct Antarct Alp Res 36(3):323–32.
Vorren Ø. 1946. Reindrift og nomadisme i Varangertraktene.
Tromsø Museums Årshefter 69(2):145.
Vuojala-Magga T, Turunen M, Ryyppo T, Tennberg M. 2011.
Resonance strategies of Sami reindeer herders in northernmost Finland during climatically extreme years. Arctic
64(2):227–41.
Ecosystem Structure and Climate Feedback
905
Van der Wal R. 2006. Do herbivores cause habitat degradation or
vegetation state transition? Evidence from the tundra. Oikos
114(1):177–86.
WMO. 2006. Systematic observation requirements for satellitebased products for climate. Geneva: World Meteorological
Organization/Global Climate Observing System.
White JR, Patel J, Ottesen A, Arce G, Blackwelder P, Lopez JV.
2012. Pyrosequencing of bacterial symbionts within Axinella
corrugata sponges: diversity and seasonal variability. PLoS ONE
7(6):e38204.
Zeileis A, Kleiber C, Jackman S. 2008. Regression models for
count data in R. J Stat Softw 27(8):1–25.