+ All Categories
Transcript

Can Reindeer Overabundance Causea Trophic Cascade?

Rolf A. Ims,1,2,* Nigel G. Yoccoz,1,2 Kari Anne Brathen,1 Per Fauchald,1,2

Torkild Tveraa,2 and Vera Hausner1

1Department of Biology, University of Tromsø, 9037, Tromsø, Norway; 2Department of Arctic Ecology, NINA, Polar Environmental

Centre 9296, Tromsø, Norway

ABSTRACT

The region Finnmark, in northernmost Europe,

harbors dense populations of semi-domestic rein-

deer of which some exhibit characteristics of

overabundance. Whereas overabundance is evi-

dent in terms of density-dependent reductions in

reindeer body mass, population growth and abun-

dance of forage plants, claims have been made that

this reindeer overabundance also has caused a

trophic cascade. These claims are based on the main

premise that reindeer overgrazing negatively im-

pacts small-sized, keystone tundra herbivores. We

tested this premise by a large-scale study in which

the abundance of small rodents, hares and ptar-

migans was indexed across reindeer management

districts with strong differences in stocking densi-

ties. We examined the scale-dependent relations

between reindeer, vegetation and these small-sized

herbivores by employing a spatially hierarchical

sampling design within the management districts.

A negative impact of reindeer on ptarmigan,

probably as a result of browsing reducing tall Salix,

was indicated. However, small rodents (voles and

lemmings), which are usually the keystone herbi-

vores in the plant-based tundra food web, were not

negatively impacted. On the contrary, there was a

strong positive relationship between small rodents

and reindeer, both at the scale of landscape areas

and local patches, with characteristics of snow-bed

vegetation, suggesting facilitation between Nor-

wegian lemmings and reindeer. We conclude that

the recent dampening of the vole and lemming

population cycle with concurrent declines of rodent

predators in northernmost Europe were not caused

by large herbivore overgrazing.

Key words: tundra; lemming; ptarmigan; facili-

tation; food web; overgrazing; quasi-experiment;

Salix; spatial scaling; growth forms.

INTRODUCTION

Human beings have affected the abundance of

large herbivores since prehistoric times (for exam-

ple, Zimov and others 1995). Because large herbi-

vores often are determinants of ecosystem structure

and function (Cote and others 2004; Danell and

others 2006), the omnipresent strong hu-

man—large herbivore connection represents one of

the most widespread anthropogenic impacts on

ecosystems. There is a long and voluminous re-

search record on the ecology of large herbivores of

which most studies have focused on direct trophic

relations, that is, plant—herbivore or herbivore-

carnivore interactions. Recently, the scope has

been extended to emphasize the role of large her-

bivores as a key intermediate link between plants

and predators in a linear food chain context, in

particular, to establish whether large herbivores

can convey bottom–up or top–down trophic cas-

cades (Ripple and Beschta 2005; Pringle and others

2007). However, ecosystems typically consist of

several interconnected food chains forming com-

Received 21 July 2006; accepted 2 February 2007; published online 18

July 2007.

*Corresponding author; e-mail: [email protected]

Ecosystems (2007) 10: 607–622DOI: 10.1007/s10021-007-9060-9

607

plex, non-linear food webs (Paine 1980) in which

small-sized herbivores may play key roles. This

opens the possibility for interactions between dif-

ferent functional groups of herbivores (that is,

small and large) and possibilities for other, less

obvious cascades, which ultimately may result from

management of large herbivores (Suominen and

Danell 2006).

Reindeer (or caribou) are the most widespread

and abundant large herbivores in tundra ecosys-

tems, where they usually co-occur with a restricted

assemblage of smaller herbivores, composed of

small rodents (voles and lemmings), hares, geese

and ptarmigans (Bliss and others 1973). Due to

naturally low population densities and a vagrant

life style, with exploitation of available pulses of

plant production in space and time, reindeer are

usually thought to have little impact on vegetation

and other ecosystem components (Jefferies and

others 1994; Chernov and Matveyeva 1997). In

contrast, small rodents (especially lemmings) usu-

ally form the key link in the tundra food web in

terms of strong impacts on vegetation, other her-

bivores and predators (Elton 1942; Batzli 1975;

Batzli and others 1980; Finerty 1980; Krebs and

others 2003; Ims and Fuglei 2005). However,

exceptions to the prime ecosystem role of small

tundra rodents occur, for instance, when human

interventions cause overabundance of other her-

bivores (Jefferies and others 1994). The emergent

overabundance of geese on the coastal tundra in

Arctic America is one such exception (Jano and

others 1998; Jefferies and Rockwell 2002). Another

case, which we consider in the present paper,

concerns semi-domestic reindeer in the region of

Finnmark, northern-most Europe. In Finnmark

reindeer have peaked to density levels claimed as

‘‘an ecological disaster‘‘ (Moen and Danell 2003).

Strong direct impacts of reindeer on vegetation in

Finnmark have been demonstrated (Suominen and

Olofsson 2000; Brathen and Oksanen 2001; Bra-

then and others, (in press). Severe plant resource

limitation may then in turn impact other herbi-

vores and ultimately their predators. Indeed, such

cascading effects of reindeer overabundance have

been suggested by several authors to explain recent

severe population declines of predators depending

on small tundra herbivores such as small rodents

and ptarmigan (Tømmeraas 1993; Kjellen and Roos

2000; Angerbjorn and others 2001; Ratcliffe 2005).

However, as yet there are no published empirical

studies that can substantiate these suggestions.

In fact, there seems to be a general scarcity of

studies that have addressed impacts of large un-

gulates on smaller vertebrate herbivores (see Su-

ominen and Danell 2006 for a review). One reason

for this may be that such studies must encompass

spatial scales exceeding the logistic capabilities of

most experimental designs. Even for voles and

lemmings, that have relative small-scale space use,

studies addressing population level responses are

likely to require landscape-scale consideration

(Wiens and others 1993; Barrett and Peles 1999;

Hansson 2002). Moreover, when assessing the

impacts of wide-ranging, large ungulates such as

reindeer adequate study designs probably require

regional extents (Senft 1987). For instance, enclo-

sure studies typically obstruct normal foraging

patterns in large ungulates (Suominen and Danell

2006) and population processes in small mammals

(Krebs and others 1969).

In the present study, we evaluate the basic pre-

mise that high densities of semi-domestic reindeer

have resulted in a trophic cascade, that is, high

reindeer abundance negatively impacts small-sized

tundra herbivores by overgrazing their food re-

source. In Finnmark reindeer are managed in large,

separate summer pasture districts, often with per-

manent, strongly differing densities between

neighboring districts. Strong, negative spatial den-

sity-dependent effects on reindeer productivity and

sensitivity to severe winter climate (Fauchald and

others 2004; Tveraa and others 2007) and on plant

abundance (Brathen and others, in press) are evi-

dent among the districts. The term ‘‘deer over-

abundance‘‘ (sensu Caughley 1981; Cote and

others 2004) is therefore justified for characterizing

districts with the highest stocking densities. By

choosing ten replicate pairs of neighboring herding

districts with contrasting stocking densities, we

were able to establish an extraordinary large-scale,

quasi-experimental study design. Herbivore abun-

dance was indexed based on recordings of feces and

grazing signs. Thus inferences about the impacts of

reindeer density on the abundance of other tundra

herbivores, and whether this impact covaried with

the abundance of important plants, could be made

at level of herding districts. As our sampling was

done according to a spatially hierarchical design

within the herding districts, we also assessed the

spatial scale-dependent nature of the relations

between reindeer, small tundra herbivores and

vegetation.

MATERIAL AND METHODS

Focal Ecosystem

The study region is coast-near alpine tundra in

Finnmark at 70�N. This region forms the north-

608 R. A. Ims and others

ernmost tip of the European continent, delineated

by the Barents Sea towards the north and by birch

forests and continuous, coniferous taiga towards

the south. Coastal eastern Finnmark has either no

forests (exposed sites) or a forest limit of mountain

birch Betula pubescens at 100–150 m a.s.l. (more

sheltered sites), whereas the forest line in the cli-

matically more benign western parts is at 300–

500 m a.s.l. (Oksanen and Virtanen 1995; Moen

1999). The zonal vegetation we focus on in the

present study, the low alpine tundra, is dominated

by shrub tundra (Walker and others 2005). The

coastal tundra in Finnmark forms important sum-

mer pastures for reindeer (Rangifer tarandus taran-

dus). Other widespread herbivores are ptarmigan

(Lagopus spp.), hare (Lepus timidus) and small ro-

dents such as Norwegian lemming (Lemmus lem-

mus) and gray-sided vole (Clethrionomys rufocanus).

Other large herbivores that may occur locally in-

clude moose (Alces alces) and sheep (Ovis abies), but

these were rarely observed in the present study.

Reindeer have been present in Finnmark since

the area was deglaciated 10,000–15,000 BP (Sko-

gland 1994). Like tundra reindeer elsewhere in the

Arctic, herds migrate seasonally between inland

winter pastures and coastal summer pastures. Since

the sixteenth century, reindeer herds in Finnmark

have become semi-domesticated and are currently

managed entirely by the indigenous Sami people,

but maintaining the seasonal migration pattern

(Muga 1986). However, in contrast to the pristine,

unmanaged situation several severe range restric-

tions have taken place. The summer pastures are

now split into many separate management districts,

which are delineated by coastlines and reindeer

fences. These fairly large summer pasture districts

(range: 188–2,733 km2 for districts included in the

present study), constitute the basic unit within the

Norwegian husbandry management regime. Rein-

deer densities and reindeer weights at the district

level are recorded annually by the Norwegian

Reindeer Husbandry Administration, and are

available for the last 2–3 decades. Recent analyses

of these data (Fauchald and others 2004; Tveraa

and others 2007) show that although there have

been strong temporal fluctuations in densities with

some districts peaking to more than 20 reindeer/

km2, there are spatial contrasts between neigh-

boring districts, both with respect to reindeer den-

sity and calf weights, that have persisted for at least

two decades. Moreover, these contrasts give rise to

a strong negative spatial density-dependence in calf

weight and thus production (Figure 1). This nega-

tive density-dependence probably results from the

deteriorated quality of pastures in the high-density

districts (Brathen and others, in press) and we

hypothesize that this also may have affected other

herbivores.

In terms of biomass, reindeer are clearly the

dominant herbivore in some of the districts with

the highest average density (>10 reindeer/km2).

For instance, 10 reindeer/km2 with average indi-

vidual body masses of 50 kg corresponds to

approximately 10,000 lemmings/(or 100 lem-

mings/ha) (assuming a mean lemming body mass

of 50 g) or 1000 ptarmigans (assuming ptarmigan

body mass of 500 g). This is clearly more than the

observed peak densities of lemming or voles

(Stenseth and Ims 1993) or ptarmigan (Pedersen

and others 2004) in these tundra habitats. Despite

the lower biomass of the small herbivores com-

pared to reindeer the food web impact of rodents

may still be larger due to the higher metabolism

and secondary productivity of small herbivores

(Batzli 1975; Batzli and others 1980). Compared to

figures on the average biomass of wild ungulates in

the North American tundra, ranging between 0.17

and 3 kg/km2 (Bliss and others 1973), the current

situation in Finnmark appears to be abnormal.

Study Design

We employed a quasi-experimental study design

(Shadish and others 2002) for estimating the effects

of reindeer density by using neighboring districts

Figure 1. The relationship between reindeer density per

km2 and weight of slaughtered calves in the twenty

reindeer districts. Lines connect districts within 10 pairs

with contrasting densities (see Figure 2). Estimates of

yearly reindeer density and calf weights are averaged

over a 23-year period using the official statistics of the

Norwegian Reindeer Management Authorities.

Impacts of Reindeer Overabundance 609

with temporally persistent contrasts in reindeer

density. Pairs of such neighboring districts were

treated as ‘‘block units‘‘ in the design (and in the

statistical analysis), enabling us by design to control

for environmental factors (that is, geology, climate)

that vary on a large spatial scale. We paired

neighboring summer pasture districts based on the

criteria that a high-density district had over the two

last decades consistently lower calf weights and

higher reindeer density than the neighboring low-

density district with which it was paired (Figure 1).

Ten district pairs conformed to both criteria and

were included in the study (Figure 2), representing

more than half of the available summer grazing

areas in Finnmark. On average the reindeer density

in high-density districts was more than twice as

high as in the low-density districts (density ratio

high:low within pairs for period 1981–2003:

2.63 ± 0.70 [se]).

The selected summer districts represent large,

heterogeneous areas and we aimed at a sampling

design within districts that with a realistic effort

could achieve a balance between coverage and a

focus on areas of assumed highest importance as

pasture for reindeer. With respect to coverage we

assured that sampling units were drawn from sep-

arate siida regions (that is, areas belong to separate

cooperative groups of Sami herders within districts)

and different geographic regions (that is, based on

topographic features such as valleys, peninsulas

and mountain ranges) within the districts. Often

such geographical and social regions coincided. As

we expected to find the strongest evidence for food

web interactions in the vegetation strata where

herbivores concentrate their activities, we re-

stricted our sampling to strata within the low-al-

pine zone that had the most mesic and wet

vegetation types. These strata represent the most

Figure 2. The hierarchical

sampling design of the study.

The map shows the

delineation of the summer

herding districts forming the

10 pairs with contrasting

reindeer densities. An arrow

connects the two districts of

each pairs. The location of the

6–14 landscape areas of 2 · 2

km within the district pairs

are shown as small gray

squares. The landscape areas

are further subdivided into a

grid of 100 quadrates from

which a random subsample

(shaded quadrates) was

chosen. Within each selected

quadrate there is a 50 m

sampling line running from

the middle of the quadrate

towards a random position.

Plots (n = 11) were

established at 5 m intervals

along the sampling line. In

each plot fecal pellets or

grazing signs of the focal

herbivores were recorded

inside a triangle with 40 cm

sides. Plant abundance was

quantified by the number of

intercepts with pins attached

in each corner of the triangle.

610 R. A. Ims and others

productive tundra habitats in terms of plant pro-

ductivity. Dryer vegetation types have generally

very sparse vegetation and offer too little food and

vegetation cover to be important, in particular, for

the smaller herbivores in question. Vegetation

strata were delineated based on satellite images and

ERDAS GIS software (ERDAS 2003) and previous

classifications of vegetation types in Finnmark

(Johansen and others 1995). The stratification and

selection procedures were as follows: A 2 · 2 km

grid of landscape areas was superimposed on the

areas of the district situated in the low alpine zone.

Next we identified landscape areas with more than

average amounts of mesic and wet vegetation for

the district. Finally, we discarded landscape areas

that included more than 50% forest, lakes, sea,

glaciers or included a major road, because such

factors may influence the presence of reindeer

independently of pasture quality. The remaining

stratum fulfilling the above criteria was then sub-

jected to random selection of landscape areas with

the restriction that there should be a minimum of

two areas within each siida/geographic region per

district. Above this minimum the number of se-

lected landscape areas increased with the area of

the focal stratum of the region. The number of se-

lected landscape areas per district then varied be-

tween 6 and 14 (Table 1). All landscape areas of a

district pair were sampled within the same time-

period during July and August in the year 2003,

and two persons analyzed most landscape areas in

one day. Districts belonging to the same district pair

were sampled simultaneously. Two initially se-

lected landscape areas were not accessible due to

very steep terrain, and other randomly chosen

landscape areas replaced them both.

The selected 2 · 2 km landscape areas were

further subdivided into1000.04 km quadrates

among which 25 were randomly selected (Fig-

ure 2). The center of each selected area was a start

position for a sampling line, the direction of which

was given by a random GPS position on a circle

with a 50 m radius from the center. The sampling

lines were sub-sampled at plots every 5 m along

the line with a triangular frame with sides of

40 cm (Figure 2).

If a part of a sampling line had to be discarded,

that is, because of terrain or wetness (lake, large

river or very wet mire), snow cover (more than a

5 m section running through snow), large boulder

field (more than half of the sampling line running

over boulders void of vegetation), or the sample

line was below the tree line, a new sampling line

was randomly selected from the same position or

new position. If no new acceptable start position

was found, the sampling square was discarded.

Most rejections were caused by sampling squares

located below the tree-line and most often this

occurred in districts of low reindeer densities. Still

this did not appear to cause any major difference in

the altitude of sampling lines between high and

low-density districts (average altitude in high-

density districts: 396.9 ± 34.1 m a.s.l., in low den-

sity districts: 374.9 ± 38.2 m a.s.l.). Overall the

numbers of landscape areas and sampling lines

were similar in low and high-density districts

(Table 1).

Indices of Herbivore Abundance andHabitat Variables

Because herbivores produce conspicuous feces in

the form of relatively large-sized pellets (or pats),

the frequency of fecal pellets can be used success-

fully as indices of abundance (Neff 1968; Putman

1984). In the present study we recorded the pres-

ence–absence of fecal pellets of hares (Hulbert and

others 1996), ptarmigan (compare Schaefer and

others 1996) and reindeer (Edenius and others

2003; Van der Wal and Brooker 2004) in the tri-

angular plots along the sampling lines (Figure 2).

The two ptarmigan species present in the region,

the willow ptarmigan (L. lagopus) and the rock

ptarmigan (L. mutus), cannot be distinguished

based on fecal pellet morphology. Small rodents

(vole and lemming) pellets are more difficult to

detect due to their small size and we therefore re-

corded, as presence or absence, the much more

conspicuous signs of rodent grazing activity (cut

vegetation, runways and burrows) in the plots.

Table 1. Numbers of Surveyed Landscape Areasand Sampling Lines in High and Low Density Dis-tricts in the 10 District Pairs (see Figure 1)

Landscape

areas

Sampling lines

District pair Low High Low High

1 6 7 52 50

2 5 6 35 41

3 6 6 78 84

4 8 8 67 79

5 7 6 37 59

6 8 6 40 83

7 7 7 63 81

8 10 8 76 89

9 10 7 124 106

10 9 14 78 132

SUM 76 75 650 804

Impacts of Reindeer Overabundance 611

Previous studies using similar methodology have

shown that such small rodent activity indices are

proportional to local abundance (Lambin and oth-

ers 2000). The different small rodents species can-

not be identified with certainty based on such

activity signs. However, the dominant species in

the study regions are usually the gray-sided vole

and the Norwegian lemming (N.G.Yoccoz and

R.A Ims, unpublished).

Vole and lemming populations in Finnmark ex-

hibit 4–5 years population density cycles (Ekerholm

and others 2001; Angerbjorn and others 2001),

which are fairly synchronized across species and

geographic areas (Hansson and others 1978; Chris-

tiansen 1983; Yoccoz and Ims 2004). Accordingly

the abundance in a particular year will depend on

the cyclic phase. The results from an ongoing large-

scale monitoring program based on live-trapping

(Yoccoz and Ims 2004) showed that small rodent

peak densities were attained in the fall of 2002 in

the entire region, and that the populations had

crashed to very low densities by the spring of 2003.

Thus signs of small rodent activity observed during

our field campaign in summer 2003 originated

mainly from the preceding winter period and thus

reflect spatial variation in rodent densities at the

culmination of the peak phase. Because food limi-

tation in small rodents is expected to kick in at the

culmination of the peak phase (Turchin and Batzli

2001), the timing of our study was ideal with re-

spect to testing for effects of reindeer overabun-

dance on small rodent populations.

Although activity signs of small rodents last for

less than a year, pellets of the larger herbivores may

last for a longer time period. Thus the prevalence of

fecal pellets must be taken as an index of cumula-

tive use, especially by reindeer. Of concern here is

that the spatial distribution of feces actually reflects

feeding patterns. Previous studies of other deer

species have shown that this indeed is the case

(Putman 1984; Schutz and others 2006). The dif-

ference in the recorded frequency of reindeer pel-

lets between the high and low-density district

(within pairs) was consistent with the correspond-

ing long-term difference based on yearly reindeer

counts obtained from official statistics (Figure 3).

During sampling we quantified the abundance of

plants with the point intercept method (Jonasson

1988; Frank and McNaughton 1990; Brathen and

Hagberg 2004) by registering the number of inter-

cept of plants with three pins forming the corners

of the triangular frame (Figure 2). For a given

growth form the number of point intercepts is

proportional to biomass (Brathen and Hagberg

2004). For bryophytes only one intercept per pin

was registered, hence bryophyte cover rather than

abundance was estimated. One side (that is, two

pins) of the triangle was placed exactly parallel to

the left side of a measurement ribbon running from

the start to the end position of the sampling line

(Figure 2). For the purpose of the present study we

focus on 11 vascular plant species or growth forms

and two bryophyte growth forms (Table 2). The

separate species were dominant plants, whereas

less common species were grouped according to

growth form. Although growth forms account for

the main structure of the vegetation, they are also

indicators of ecosystem function, that is, they cor-

relate with plant functionality (for example, Cha-

pin and others 1996), and represents important

forage items and habitat indicators for herbivores.

We recorded topographical variables supposed to

be important to herbivores, that is, altitude, slope

and an index of terrain ruggedness (Neteler and

Mitasova 2002), at the level of quadrates by

extracting this information from a terrain model

with a spatial resolution of 25 · 25m.

Statistical Analysis

Separate statistical analyses were done at three

spatial scales in the hierarchical sampling design;

Figure 3. The relationship between reindeer density per

km2 and the frequency of plots (within sampling lines)

with reindeer feces. Lines connect districts within the 10

pairs with contrasting densities (see Figure 2). Estimates

of yearly reindeer density and calf weights are averaged

over a 23-year period using the official statistics of the

Norwegian Reindeer Management Authorities, whereas

the estimate of feces frequency was obtained in 2003

according to our study design (see main text). District

pair 6 had a reversed density of reindeer in 2003 com-

pared to the long-term average (Anonymous 2004).

612 R. A. Ims and others

that is, the scales of sampling lines, landscape areas

and herding districts (Figure 2). The smallest spatial

scale (that is, the plots within sampling lines) was

not considered, because the frequencies of herbi-

vore recordings were too low at this scale to allow

for robust analyses.

At the largest scale (that is, the herding districts)

we used the power of our quasi-experimental de-

sign to assess the impacts of long-term spatial

contrast in reindeer stocking densities on the in-

dexed abundance of the other herbivores (that is,

small rodents, ptarmigans and hares). The size of

the effect of reindeer abundance was estimated as

nominal contrasts (that is, differences) between the

high and low-density districts within pairs. In this

analysis the response variables (indexed herbivore

abundance) were the frequency of plots per sam-

pling line with recordings of the herbivores aver-

aged over all sampling lines and landscape areas in

the districts. Thus the 20 herding districts were the

quasi-experimental units (that is, replicates). Dis-

trict pair was included as a random (that is,

‘‘block‘‘) effect in a mixed linear model framework

(Pinheiro and Bates 2000), whereas the three

topographical variables (altitude, slope and rug-

gedness) were entered as district level average

values.

For the two smaller spatial scales considered (that

is, landscape areas and sampling lines), we used

linear mixed models to predict frequencies of small

rodents, ptarmigans and hares (response variable)

as a function of reindeer abundance as indexed by

the pellet recordings (predictor variable). In these

models the abundance of the 13-plant species/

growth forms (Table 2) and the three topographical

variables (see above) were included as covariates.

At both scales, both the response (that is, indexed

herbivore abundance, see above) and predictor

variables were quantified at the sampling line level.

Plant predictors were quantified as the number of

point intercepts per line and a log (x + 1) trans-

formation was used to meet the requirements of

linear modelling (that is, linear relationships and

adequate spread of predictor values). At the land-

scape level, average values based on all sampling

lines were used. Levels above in the sampling

hierarchy (districts for landscape areas, whereas

districts and landscape areas for lines) were mod-

elled as random effects. To be able to compare the

estimates across species and scales we used a ‘‘full

model‘‘ with all predictor variables included in all

analyses. Eventual non-linear relations were con-

sidered by using additive models (Wood 2006) and

when present they were adequately described by

second order polynomial terms. The significance of

model parameters was assessed according to their

95% confidence intervals. The residuals of all

models were checked and found to satisfactorily

meet the assumptions of linear models (stable

variance and no spatial auto-covariance).

RESULTS

Overall Abundance and Distribution ofHerbivores

Table 3 summarizes the average indexed abun-

dance of the recorded herbivores at the four levels

in our hierarchical sampling design, whereas Fig-

ure 4 depicts the spatial distribution of recordings at

the landscape level. Reindeer were the most fre-

quently recorded herbivore followed by small ro-

dents, ptarmigans and hares (Table 3). In particular

reindeer, but also small rodents were widely dis-

tributed (Figure 4) with at least 2/3 of the landscape

areas recording their presence (Table 3). Ptarmigans

and in particular hares, had a more occasional

representation on the landscape scale (Figure 4)

and hares were the only herbivore missing from a

fraction of the herding districts (Table 3).

Impacts of Reindeer Stocking DensitiesAcross Herding Districts

At the spatial scale of herding districts we evaluated

the potentially negative impact of reindeer stocking

density on the abundance of other herbivores by

Table 2. The Overall Abundance of Plant Speciesor Growth Forms used as Predictors in the Statis-tical Models of Herbivore Abundance at the Scale ofLandscape Areas and Sampling Lines

Plant species/growth form Estimate (SE)

Mat grass (Nardus stricta) 2.08 (0.41)

Grasses (other) 7.82 (1.53)

Sedges 5.43 (0.49)

Small dicotyledons 2.93 (0.63)

Bilberry (Vaccinium myrtillus) 3.55 (0.41)

Deciduous ericoids (other) 5.62 (0.52)

Crowberry (Empetrum nigrum

ssp. hermaphroditum)

10.23 (0.61)

Evergreen ericoids (other) 3.15 (0.57)

Tall Salix 0.31 (0.11)

Prostrate Salix 2.03 (0.22)

Dwarf birch (Betula nana) 4.92 (1.29)

Acrocarp mosses 8.36 (0.72)

Pleurocarp mosses 3.32 (0.55)

The abundance estimates are expressed as the mean number of point intercepts(standard error in parentheses) per sampling line. Nomenclature follows Lid andLid (2005).

Impacts of Reindeer Overabundance 613

means of nominal contrasts; that is, estimated dif-

ferences in the indexed abundance between high

and low-density districts within the 10 pairs. There

was a negative effect of reindeer stocking density

on the abundance of ptarmigan, as the effect size

was large relative to the reference level, but the

confidence interval was too wide to warrant a firm

conclusion (Table 4). For hares the effect estimate

was also negative, but much smaller and more

uncertain than for ptarmigans. For small rodents,

which were generally much more abundant and

widespread than hares and ptarmigans, the contrast

estimate indicated that there was no evidence for a

biologically significant effect of reindeer stocking

density.

Spatial Scale Dependent Relations withinHerding Districts

At the two smaller spatial scales (that is, landscape

areas within herding districts and sampling lines

within landscape areas) we assessed the relations

between small rodent, ptarmigan and hare abun-

dance as response variables, and reindeer abun-

dance, vegetation variables and topographical

variables as predictors. At both spatial scales, but

most strongly at the landscape scale, the abun-

dance of small rodents showed a distinct non-

linear, positive relation to reindeer abundance as

modeled by a second-order polynomial (Table 5;

Figure 5). Small rodent abundance was also re-

lated to several vegetation variables, in particular

at the scale of sampling lines. At the sampling line

scale small rodent abundance increased with

increasing cover of pleurocarp mosses, with

increasing abundance of prostrate Salix and small

dicotyledons, whereas the effects of sedges and

most strongly crowberry Empetrum nigrum ssp.

hermaphroditum were negative (Table 5). The

strong negative relation between small rodents

and crowberry was also clearly evident at the scale

of landscape areas. In addition the abundance of

small rodents increased with the abundance of

dwarf birch at the landscape scale.

The abundance of hares and ptarmigans was not

significantly related to reindeer abundance at the

Table 3. The Proportion of Sampling Units with Recorded Presence of Herbivores according to the FourLevels in the Hierarchical Study Design (Figure 1)

Herbivore Spatial scale

Plots (n = 15,959) Lines (n = 1,454) Landscapes (n = 151) Districts (n = 20)

Reindeer (%) 7.53 43.8 87.4 100

Small rodents (%) 3.37 19.3 66.9 100

Ptarmigan (%) 0.58 5.2 39.7 100

Hare (%) 0.19 2.0 14.6 60

Figure 4. A bubble plot

showing the spatial

distribution of indexed

herbivore abundance at the

scale of landscape areas. The

sizes of bubbles are

proportional to the number

of plots per sampling line

with recorded presence of an

herbivore. Points denote

landscapes without recorded

presence of herbivores.

Coordinates are in UTM Zone

33.

614 R. A. Ims and others

spatial scales of landscape areas or sampling lines.

At the scale of sampling lines, both hares and

ptarmigans exhibited a positive association with tall

Salix shrubs (Table 5). This relation was also evi-

dent at the landscape scale for hares, whereas it was

no longer significant for ptarmigans at this scale.

Ptarmigans were most frequently recorded at high

altitude sampling lines within landscapes.

A separate analysis of indexed abundance of

reindeer against topography and plant variables

revealed that reindeer abundance could be pre-

dicted by several plant variables at the scale of

sampling lines, whereas only altitude appeared to

be a significant predictor at the landscape scale

(Table 6). Among the most influential predictor

variables at the line scale prostrate Salix and pleu-

rocarp mosses were positively related to reindeer

abundance, whereas tall Salix and reindeer abun-

dance exhibited a negative association.

DISCUSSION

The particular management setting in Finnmark

with persistent differences in reindeer stocking

densities between neighboring summer herding

districts provided an ample opportunity for assess-

ing the impact of reindeer overabundance on other

tundra herbivores. Concordant evidence for rein-

deer overabundance was available in terms of

strong spatial density-dependence in reindeer body

mass and climate-sensitive demography (Fauchald

and others 2004; Tveraa and others 2007) and in

terms of negative grazing impact on vegetation

(Brathen and Oksanen 2001; K.A. Brathen and

others, in press). On this basis we could evaluate

the claim that plant resource deterioration due to

reindeer overabundance limits the abundance of

other tundra herbivores and eventually the wider

(that is, cascading) consequences of this at the

ecosystem level.

The strongest inference about the impact of

reindeer could be made at the scale of herding

districts due to the quasi-experimental layout of

paired high-density and low-density summer pas-

ture districts. A negative impact of a high reindeer

stocking density was indicated for ptarmigans at

this scale. Although the effect estimate was some-

what uncertain in a statistical sense, its biological

significance was underpinned by the significance of

tall Salix shrubs in the relation between reindeer

and ptarmigans. Within the districts there was a

negative association between the abundance of

reindeer and tall Salix shrubs. This association is

consistent with the strong negative effects of rein-

deer summer browsing on tall Salix found in several

previous studies (Ouellet and others 1994; Man-

seau and others 1996; den Herder and others

2004). On the other hand there was a positive

association between ptarmigan and tall Salix. The

willow ptarmigan L. lagopus is strongly dependent

on Salix (willow) shrubs both as forage and habitat

structure providing cover on the otherwise barren

tundra, especially in winter (Andreev 1988).

Accordingly, van Herder and others (2004) pre-

dicted that willow ptarmigan should be one of the

most vulnerable herbivore species to reindeer

overabundance due to the strong negative effect of

reindeer browsing on the growth and distribution

of tall Salix. Our study is the first to provide

empirical results in support for this prediction.

Hares were also associated with tall Salix shrubs

within the reindeer herding districts, as could be

expected based on prior knowledge about food and

habitat selection for the mountain hare (for

example, Pullianen and Tunkkari 1987). Conse-

quently, hares and ptarmigan have been found to

be positively associated in the arctic tundra (Klein

and Bay 1994; Schaefer and others 1996). How-

ever, although the effect of the high reindeer

stocking density was negative, as could also be

Table 4. Estimated Effects of High Reindeer Density on Other Herbivores at the Herding District Level

Herbivore Estimate With covariates Without covariates

Small rodents Reference (low density) 0.400 [0.200, 0.600] 0.350 [0.190, 0.510]

Effect of reindeer density ) 0.037 [)0.213, 0.139] ) 0.019 [)0.193, 0.155]

Ptarmigan Reference (low density) 0.047 [0.004, 0.090] 0.081[0.055, 0.107]

Effect of reindeer density ) 0.030 [)0.063, 0.003] )0.029 [)0.064, 0.006]

Hare Reference (low density) 0.052 [0.028, 0.076] 0.030 [0.012, 0.048]

Effect of reindeer density )0.007 [)0.027, 0.013] )0.012 [)0.039, 0.015]

The effect size is the difference in frequency of plots with recorded presence of herbivores between districts (within pairs) with high and low reindeer density. The reference level is theestimated frequency of plots with herbivore recording in districts with low reindeer density. Estimates both from models with and without topographical covariates are given and thereference levels for the models with covariates are adjusted to the mean of the covariate values of each district. Uncertainty estimates in brackets are 95% confidence intervals.

Impacts of Reindeer Overabundance 615

Tab

le5.

Th

eC

oeffi

cien

ts(±

95%

CI)

of

the

Models

of

Indexed

Herb

ivore

Abu

ndan

ceat

the

Sca

les

of

Lan

dsc

ape

Are

as

an

dSam

plin

gLin

es

as

aFu

nct

ion

of

Indexed

Rein

deer

Abu

ndan

ce,

Pla

nt

Abu

ndan

cean

dTopogra

ph

ical

Vari

able

s

Pre

dic

tor

vari

ab

leS

mall

rod

en

tsP

tarm

igan

Hare

Lin

eL

an

dsc

ap

eL

ine

Lan

dsc

ap

eL

ine

Lan

dsc

ap

e

Inte

rcept

0.2

31

[)0.0

33,

0.4

94]

0.1

67

[)0.3

78,

0.7

16]

0.0

11

[)0.0

57,

0.0

79]

0.0

36

[)0.1

60,

0.2

33]

)0.0

16

[)0.0

54,

0.0

22]

)0.0

42

[)0.1

28,

0.0

43]

Rein

deer

abu

ndan

ce0.0

22

[)0.0

56,

0.0

99]

)0.1

90

[)0.4

39,

0.0

60]

)0.0

05

[)0.0

17,

0.0

08]

)0.0

21

[)0.0

57,

0.0

14]

0.0

01

[)0.0

05,

0.0

08]

0.0

04

[)0.0

09,

0.0

18]

Rein

deer

abu

ndan

ce^

20.0

30

[0.0

16,

0.0

43]

0.1

57

[0.0

82,

0.2

32]

--

--

Mat

gra

ss)

0.0

08

[)0.0

66,

0.0

50]

0.0

31

[)0.0

98,

0.1

61]

0.0

01

[)0.0

18,

0.0

21]

0.0

21

[)0.0

25,

0.0

68]

0.0

04

[)0.0

07,

0.0

14]

0.0

03

[)0.0

15,

0.0

21]

Gra

sses

(oth

er)

0.0

10

[)0.0

42,

0.0

62]

)0.0

37

[)0.1

94,

0.1

20]

)0.0

05

[)0.0

22,

0.0

13]

0.0

26

[)0.0

30,

0.0

83]

0.0

06

[)0.0

03,

0.0

15]

)0.0

00

[)0.0

22,

0.0

22]

Sedges

)0.0

66

[)0.1

14,)

0.0

18]

0.0

46

[)0.0

77,

0.1

68]

0.0

003

[)0.0

15,

0.0

16]

)0.0

11

[)0.0

55,

0.0

33]

0.0

00

[)0.0

09,

0.0

08]

0.0

09

[)0.0

09,

0.0

26]

Sm

all

dic

oty

ledon

s0.0

92

[0.0

18,

0.1

65]

)0.0

38

[)0.2

19,

0.1

43]

)0.0

31

[)0.0

56,

)0.0

07]

)0.0

61

[)0.1

26,

0.0

04]

)0.0

07

[)0.0

20,

0.0

05]

0.0

06

[)0.0

19,

0.0

31]

Bil

berr

y0.0

29

[)0.0

59,

0.1

17]

0.1

73

[)0.0

68,

0.4

15]

)0.0

19

[)0.0

48,

0.0

10]

)0.0

37[)

0.1

24,

0.0

50]

0.0

00

[)0.0

15,

0.0

15]

0.0

05

[)0.0

29,

0.0

39]

Deci

du

ou

seri

coid

s(o

ther)

)0.0

24

[)0.1

11,

0.0

62]

0.1

00

[)0.1

68,

0.3

69]

0.0

17

[)0.0

11,

0.0

45]

)0.0

10

[)0.1

07,

0.0

86]

)0.0

08

[)0.0

23,

0.0

07]

)0.0

02

[)0.0

40,

0.0

35]

Cro

wberr

y)

0.1

06

[)0.1

67,)

0.0

46]

)0.1

86

[)0.3

51,

)0.0

22]

)0.0

02

[)0.0

22,

0.0

17]

)0.0

40

[)0.0

99,

0.0

19]

0.0

05

[)0.0

05,

0.0

15]

0.0

13

[)0.0

11,

0.0

36]

Everg

reen

eri

coid

s(o

ther)

0.0

32

[)0.0

37,

0.1

00]

0.0

32

[)0.1

80,

0.2

43]

0.0

09

[)0.0

135,

0.0

31]

0.0

12

[)0.0

64,

0.0

88]

0.0

14

[0.0

02,

0.0

25]

0.0

07

[)0.0

23,

0.0

37]

Tall

Sa

lix

)0.0

80

[)0.2

21,

0.0

62]

0.1

24

[)0.0

95,

0.3

43]

0.0

61

[0.0

156,

0.1

07]

0.0

24

[)0.0

54,

0.1

03]

0.0

45

[0.0

21,

0.0

69]

0.0

32

[0.0

02,

0.0

63]

Pro

stra

teS

ali

x0.0

90

[0.0

22,

0.1

59]

0.1

45

[)0.0

05,

0.2

95]

)0.0

02

[)0.0

24,

0.0

20]

0.0

13

[)0.0

41,

0.0

67]

0.0

02

[)0.0

10,

0.0

13]

)0.0

04

[)0.0

25,

0.0

17]

Dw

arf

bir

ch)

0.0

28

[)0.0

84,

0.0

28]

0.1

07

[0.0

02,

0.2

12]

)0.0

04

[)0.0

20,

0.0

13]

0.0

17

[)0.0

21,

0.0

54]

)0.0

06

[)0.0

15,

0.0

03]

0.0

07

[)0.0

22,

0.0

08]

Ple

uro

carp

moss

es

0.1

37

[0.0

69,

0.2

05]

)0.0

43

[)0.1

72,

0.0

86]

0.0

02

[)0.0

18,

0.0

22]

0.0

01

[)0.0

45,

0.0

48]

)0.0

02

[)0.0

12,

0.0

09]

)0.0

04

[)0.0

23,

0.0

14]

Acr

oca

rpm

oss

es

0.0

18

[)0.0

51,

0.0

87]

)0.0

41

[)0.2

18,

0.1

35]

0.0

30

[0.0

09,

0.0

52]

0.0

60

[)0.0

03,

0.1

23]

)0.0

07

[)0.0

18,

0.0

05]

)0.0

17

[)0.0

41,

0.0

08]

Alt

itu

de

(*100)

0.0

03

[)0.0

59,

0.0

66]

0.0

12

[)0.5

7,

0.0

81]

0.0

27

[0.0

11,

0.0

43]

0.0

12

[)0.0

13,

0.0

37]

0.0

01

[)0.0

08,

0.0

09]

0.0

02

[)0.0

08,

0.0

13]

Slo

pe

0.0

50

[)0.0

34,

0.1

35]

)0.0

20

[)0.2

40,

0.2

00]

0.0

22

[)0.0

06,

0.0

49]

0.0

04

[)0.0

75,

0.0

83]

0.0

03

[)0.0

11,

0.0

17]

)0.0

06

[)0.0

36,

0.0

25]

Ru

ggedn

ess

)0.1

38

[)0.3

32,

0.0

56]

0.0

30

[)0.4

72,

0.5

31]

)0.0

42

[)0.1

05,

0.0

22]

0.0

01

[)0.1

79,

0.1

81]

0.0

00

[)0.0

33,

0.0

33]

0.0

20

[)0.0

50,

0.0

90]

Sig

nifi

can

tco

effici

ents

(th

at

is,

thos

efo

rw

hic

hth

eco

nfiden

cein

terv

al

did

not

incl

ude

zero

)are

inbol

d.

Pol

ynom

ial

term

sfo

rth

ere

indee

rabu

ndan

cew

ere

only

incl

uded

inca

ses

inw

hic

ha

non

-lin

eari

tyw

as

evid

ent.

616 R. A. Ims and others

expected from reindeer overabundance, the confi-

dence interval was too wide to provide any statis-

tical evidence for this. It should be noted, however,

that the precision of the analysis of hare abundance

was low at the scale of herding districts due to the

very few and patchily distributed recordings of

hares (Figure 4). It is likely that the only occasional

presence of hare to a large extent mirrors that of

tall Salix shrubs, which appear to exhibit an

equivalent rarity and patchy distribution in the

reindeer summer pastures in Finnmark (see the

large standard error of the mean tall Salix abun-

dance in Table 2).

Lemmings, and also to some extent voles, usually

dominate among tundra herbivores in terms of

biomass and impact on other components of the

plant-based food (Finerty 1980; Batzli and others

1980; Ims and Fuglei 2005). Thus the greatest po-

tential for a strong trophic cascade caused by

reindeer overabundance lies in a negative impact

on the abundance of small rodents. However, de-

spite the fact that reindeer biomass in the most

densely stocked districts now clearly outweighs

small rodent biomass, we found no evidence for a

negative effect of this high reindeer density on

small rodent abundance at the scale of reindeer

herding districts. Rather opposite, at the two

smaller spatial scales (that is, sampling lines and

landscape areas within herding districts) there were

strong positive, non-linear associations between

indexed abundance for reindeer and small rodents.

Both small rodents and reindeer were associated

with plants characterizing snow bed vegetation

such as prostrate Salix (mostly S. herbacea), small

dicotyledons and mosses (Moen 1999). This indi-

cates that most of the small rodent signs recorded in

our study stem from the Norwegian lemming,

which typically spend the winters in snow bed

vegetation (Kalela 1961; Stenseth and Ims 1993).

In snow beds lemmings find both thermal insula-

tion (under a thick snow cover) and their preferred

winter food; that is, mosses. Reindeer typically

exploit snow beds in the summer (Edenius and

others 2003), probably due to the availability of

young nutritious plant items when vegetation

elsewhere is in a phenologically more advanced

and less palatable state. Thus the positive associa-

tion between lemmings and reindeer may therefore

at least partially, be due to the fact that they

coincide spatially in areas with snow bed character.

So why does not this distinct spatial overlap be-

tween the two herbivores lead to resource compe-

tition and a negative impact of reindeer

overabundance on lemmings? Although lemmings

and reindeer overlap spatially they are temporally

segregated because snow beds are winter habitats

for lemmings and summer habitats for reindeer.

This temporal separation of habitat and the fact that

the Norwegian lemming predominantly forage on

mosses in winter (Batzli 1993), which are avoided

by reindeer in the summer (Batzli and others

1980), limits the scope for resource competition.

Quite to the contrary of a situation with competi-

tive interaction between the two herbivores, lem-

mings may increase summer pasture quality for

reindeer by removing mosses and thus disturbing

the moss cover (Moen and others 1993; Virtanen

2000). Such disturbance of the moss cover may

enhance establishment of more palatable vascular

plants for reindeer (Virtanen 2000; van der Wal

and Brooker 2004). Our finding that there was a

higher abundance of small dicotyledons in summer

where lemmings had been numerous in the winter

is consistent with such an effect of lemming winter

activity. Moreover, the large quantities of excreta

left by lemmings in snow beds at the end of the

peak density winters may have a substantial posi-

tive effect on plant growth and plant quality in the

spring due to release of nutrients from lemming

feces (Schults 1969). The winter activity by lem-

mings may also reduce the amount of standing

dead vascular plant tissue, tissue that potentially

shades the photosynthetic activity of the new

spring growth, further enhancing the productivity

of the plants. The strong positive association be-

Figure 5. Box plot depicting the non-linear, positive

relation between small rodent and reindeer abundance at

the scale of landscape areas. Indexed reindeer abun-

dances (that is, number of plots per sampling line with

recorded presence) are lumped into four categories

Impacts of Reindeer Overabundance 617

tween reindeer and small rodents, even when plant

abundance was statistically accounted for, indicates

that there may be a facilitating effect operating

between these herbivores through plant quality or

productivity. Whatever the mechanism underlying

such a facilitation, our study contradicts the pro-

posed negative cascading effect of reindeer on small

rodents and their predators.

Several of the relations between the herbivores

and habitat variables appeared to be scale depen-

dent, and most of them appeared at the smallest

spatial scale (that is, the sampling line scale), which

is consistent with results from studies in different

ecosystems (for example, Palmer and others 2003).

The predominant small-scale plant-herbivore rela-

tion in the present study may be explained by the

small-scale patchiness of important habitat ele-

ments such as snow beds (Edenius and other 2003)

and Salix thickets (den Herder and others 2003). In

the case of small rodents the shifting influence of

some plants depending on scale may also be due to

shifting habitat selection on a temporal scale. For

instance, at the local scale (lines) lemming will not

co-occur in winter with dwarf birch (which are not

found in snow beds), but the abundance of lem-

mings at the landscape scale may still depend on

dwarf birch (see Table 5) as this plant is an

important structure in small rodent summer habi-

tats (Hamback and others 1998). The only variable

predicting (high) reindeer abundance at the scale of

landscape areas was (high) altitude possibly

reflecting both avoidance of insect harassment at

this spatial scale (Hagemoen and Reimers 2002)

and a favorable altitudinal gradient in plant phe-

nological stages (Albon and Langvatn 1992).

Crowberry appeared to be a strong negative

predictor of small rodent abundance both at the

landscape and the line scales. Crowberry is cur-

rently the dominating species in tundra vegetation

of northern Fennoscandia (Table 2). This ericoid is

virtually unpalatable to herbivores (Tybirk and

others 2000), has a strong allelopathic effect on

other plants (Wardle and Nilsson 1997) and tends

to immobilize nutrients in the soil reducing overall

primary productivity (Nilsson 1994). Thus crow-

berry can be classified as an invasive species (for

example, Wardle and others 1998) that needs an

efficient disturbance regime to be restricted within

its fundamental niche. Such disturbance regimes

are both present in the boreal forest in terms of fires

(Wardle and others 1998), and in middle and high

arctic tundra in terms of permafrost and cryotur-

bation (Bliss and others 1973; Epstein and others

2004). However, in alpine and low Arctic heaths,

where such abiotic disturbances are mainly absent,

only extreme impacts of herbivores appear to be

able to alleviate the dominance of crowberry. One

such extreme herbivore impact, which occurs very

locally along fences separating semi-domestic herds

in Finnmark, is the intense trampling by reindeer

Table 6. The Coefficients (±95% CI) of the Models of Indexed Reindeer Abundance at the Scales ofLandscape Areas and Sampling Lines as a Function of Plant Abundance and Topographical Variables

Predictor variable Spatial scale

Line Landscape

Intercept 0.382 [0.003, 0.761] )0.669 [)1.730, 0.392]

Mat grass )0.067 [)0.142, 0.008] 0.182 [)0.043, 0.406]

Grasses (other) 0.020 [)0.046, 0.087] 0.212 [)0.056, 0.480]

Sedges )0.035 [)0.097, 0.028] 0.037 [)0.177, 0.251]

Small dicotyledons 0.047 [)0.049, 0.142] )0.149 [)0.458, 0.159]

Bilberry 0.120 [0.006, 0.234] )0.254 [)0.674, 0.165]

Deciduous ericoids (other) )0.112 [)0.223, )0.000] 0.122 [)0.338, 0.582]

Crowberry )0.002 [)0.080, 0.076] 0.112 [)0.184, 0.407]

Evergreen ericoids (other) 0.074 [)0.014, 0.163] 0.008 [)0.360, 0.376]

Tall Salix )0.193 [)0.376, )0.010] )0.066 [)0.448, 0.317]

Prostrate Salix 0.191 [0.104, 0.279] 0.092 [)0.166, 0.349]

Dwarf birch )0.014 [)0.088, 0.060] )0.009 [)0.202, 0.183]

Pleurocarp mosses 0.184 [0.095, 0.273] 0.069 [)0.155, 0.293]

Acrocarp mosses )0.005 [)0.095, 0.085] 0.172 [)0.133, 0.477]

Altitude (*100) 0.044 [)0.041,0.129] 0.180 [0.051, 0.309]

Slope 0.084 [)0.026, 0.193] 0.015 [)0.368, 0.398]

Ruggedness )0.208 [)0.459, 0.044] )0.089 [)0.960, 0.783]

Significant coefficients (that is, those for which the confidence interval did not include zero) are in bold.

618 R. A. Ims and others

that causes erosion (Evans 1996). However, rein-

deer grazing, even in the most densely stocked

districts in Finnmark, is not able to affect crowberry

abundance on a large spatial scale (Brathen and

others, in press). The only tundra herbivore with

the capacity to affect crowberry on a large scale is

probably lemmings during high-amplitude peak

years. In such years lemmings occupy even crow-

berry heaths in great numbers, where they mow

down the shrub layer presumably to get access to

the ground level mosses (Oksanen and Oksanen

1981). If such intense grazing events coupled with

mobilization of nutrients due to large amounts

lemming excreta occurred frequently, the domi-

nance of crowberry might be reduced, giving way

to a more biologically diverse and productive eco-

system state.

Although lemming and vole peak years re-occur

cyclically at 3–5-year intervals in northern Fenno-

scandia (Stenseth and Ims 1993), most lemming

peaks during the last century have been small

(Angerbjorn and others 2001) and there seem to be

periods of 20–30 years between peak years with

substantial impacts on the vegetation (Henttonen

and Kaikusalo 1993; Virtanen 2000). Moreover,

during the last two decades the small rodent pop-

ulation cycles in northern Fennoscandia have be-

come substantially weaker (Henttonen and

Wallgren 2001; Yoccoz and others 2001; Hornfeldt

2004). A number of factors have been proposed to

explain these changes (Hornfeldt 2004), among

which climate change appears to be the most

credible ultimate cause (Callaghan and others

2004; Ims and Fuglei 2005). The recent population

declines of Arctic specialist small rodent predators

(Kjellen and Roos 2000; Tannerfeldt and others

2002; Hornfeldt and others 2005) represent most

likely a bottom-up effect of the fading small rodent

cycle. Based on the large-scale connection between

abundance of crowberry and lemming revealed in

the present study we highlight the possibility of a

top-down effect due to the relaxed lemming dis-

turbance on tundra heath vegetation. In areas and

time periods without substantial lemming peaks

crowberry may expand with possible negative

feedback effects on lemmings, other herbivores and

their predators.

CONCLUSION

Our large-scale study provided results that were

consistent with the prediction (Van Herder and

others 2004) that ptarmigans may be the tundra

herbivores most sensitive to reindeer overabun-

dance and that this may lead to a trophic cascade

whereby predators specialized on ptarmigan such

as gyr falcon Falco rusticolus ultimately may be af-

fected (Tømmeraas 1993). The reindeer—ptarmi-

gan connection is most likely due to the negative

impacts of heavy reindeer browsing on tall Salix,

which both provides essential habitat and food for

the willow ptarmigan. Thickets of tall Salix shrubs

in the tundra landscape provide habitat and re-

sources for many other species as well and are

thereby probably hot spots for many food web

interactions not considered in the present study.

Thus future studies should investigate the conse-

quences of reindeer-induced loss of tall Salix.

On the other hand, our study did not support the

most severe trophic cascade believed to result from

reindeer overabundance. Such a cascade would be

induced if reindeer overgrazing negatively im-

pacted small rodents, which are normally the

keystone herbivores in the plant-based tundra food

web. However, on the contrary we found a strong

positive association between small rodents and

reindeer, which may be due to a facilitating effect

of lemming winter activity on reindeer summer

pasture quality. This possibility emphasizes the

need for obtaining a better understanding of the

consequences of the recent weakening of the small

rodent cycles on the long-term development of

tundra vegetation (Ims and Fuglei 2005).

ACKNOWLEDGEMENTS

We are grateful to Johan Ingvald Hætta and Anders

Aarthun Ims for information about reindeer herding

districts, to Torstein Engelskjøn for providing a flora

database, Bernt Johansen for providing satellite

images, The Norwegian Coast Guard and Jan Kare

Amundsen for transportation during field work,

Sunna Pentha for field assistance, and to Marianne

Iversen and Siw Killengreen for great leadership

during field work. This study, which is a contribution

from the ‘‘Ecosystem Finnmark‘‘ project, was fi-

nanced by the Norwegian Research Council.

REFERENCES

Albon SD, Langvatn R. 1992. Plant phenology and the benefits

of migration in a temperate ungulate. Oikos 65:502–13.

Andreev A. 1988. The ten year cycle of the willow grouse of

lower Kolyma. Oecologia 76:261–7.

Angerbjorn A, Hersteinsson P, Tannerfeldt M. 2004. Arctic foxes.

Consequences of resource predictability in the Artic fox—two

life history strategies. In: Macdonald DW, Sillero-Zubiri C,

Eds. Biology and conservation of wild canids. Oxford: Oxford

University Press. pp 163–72.

Angerbjorn A, Tannerfeldt M, Lundberg H. 2001. Geographical

and temporal patterns of lemming population dynamics in

Fennoscandia. Ecography 24:298–308.

Impacts of Reindeer Overabundance 619

Barrett GW, Peles JD, Eds. 1999. Landscape ecology of small

mammals. New York: Springer.

Batzli G. 1993. Food selection by lemmings. In: Stenseth NC, Ims

RA, Eds. The biology of lemmings. London: Academic. pp.

281–301.

Batzli GO. 1975. The role of small mammals in arctic ecosystems.

In: Golley FB, Petrusewicz K, Ryszkowski L, Eds. Small

mammals: their productivity and population dynamics.

Cambridge: Cambridge University Press..

Batzli GO, White RG, MacLean SF, Pitelka FA, Collier BD (1980)

The herbivore-based trophic system. In: Brown J, Miller RG,

Tieszen LL, Bunnell FL, Eds. An arctic ecosystem. The coastal

tundra at Barrow, Alaska. US/IBP Synthesis Series. PA:

Dowden, Hutchinson and Ross. pp. 335–410.

Bliss LC, Courtin GM, Pattie DL, Riewe RR, Whitfield DWA,

Widden P. 1973. Arctic tundra ecosystems. Ann Rev Ecol Syst

4:359–99.

Brathen KA, Hagberg O. 2004. More efficient estimation of plant

biomass. J Veg Sci 15:653–60.

Brathen KA, Oksanen J. 2001. Reindeer reduce biomass of

preferred plant species. J Veg Sci 12:473–80.

Brathen KA, Ims RA, Yoccoz NG, Fauchald P, Tveraa T, Hausner V

(2007) Induced shift in ecosystem productivity? Extensive scale

effects of abundant large herbivores. Ecosystems (in press).

Callaghan TV, Bjorn LO, Chernov Y, Chapin T, Christensen TR,

Huntley B, Ims RA, Johansson M, Jolly D, Jonasson S, Mat-

veyeva N, Panikov N, Oechel W, Shaver G, Elster J, Jonsdottir

IS, Laine K, Taulavuori K, Taulavuori E, Zockler C. 2004.

Responses to projected changes in climate and UV-B at the

species level. Ambio 33:418–35.

Caughley G. 1981. What we do not know about the dynamics of

large mammals. In: Fowler CW, Smith T, Eds. Dynamics of

large mammal populations. New York: Wiley. pp 361–72.

Chapin FS, BretHarte MS, Hobbie SE, Zhong HL. 1996. Plant

functional types as predictors of transient responses of arctic

vegetation to global change. J Veg Sci 7:347–58.

Chernov YI, Matveyeva NV. 1997. Arctic ecosystems in Russia

In: Wielgolaski FE, Ed. Ecosystems of the World. Amsterdam:

Elsevier.

Christiansen E. 1983. Fluctuations in some small rodent popu-

lations in Norway 1971–1979. Holarct Ecol 6:24–31.

Cote SD, Rooney TP, Tremblay JP, Dussault C, Waller DM. 2004.

Ecological impacts of deer overabundance. Annu Rev Ecol

Evol Syst 35:113–47.

Danell KD, Bergstrom R, Duncan P, Pastor J, Eds. (2006) Large

herbivore ecology, ecosystem dynamics and conservation.

Cambridge: Cambridge University Press.

den Herder M, Virtanen R, Roinenen H. 2004. Effects of reindeer

browsing on tundra willow and its associated insect herbi-

vores. J Appl Ecol 41:870–97.

Edenius L, Vencatasawmy CP, Sandstrom P, Dahlberg U. 2003.

Combining satellite imagery and ancillary data to map

snowbed vegetation important to reindeer Rangifer tarandus.

Arct Antarct Alp Res 35:150–7.

Ekerholm P, Oksanen L, Oksanen T. 2001. Long-term dynamics

of voles and lemmings at the timberline and above the willow

limit as a test of hypotheses on trophic interactions. Ecography

24:555–68.

Elton C. 1942. Vole, mice and lemmings. London: Oxford Uni-

versity Press.

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

31:1917–33.

ERDAS. 2003. ERDAS imagine, leica Geosystems. Version 8.7.

Evans R. 1996. Some impacts of overgrazing by reindeer in

Finnmark, Norway. Rangifer 16:3–19.

Fauchald P, Tveraa T, Yoccoz NG, Ims RA. 2004. En økologisk

bærekraftig reindrift. Hva begrenser naturlig produksjon og

høsting (In Norwegian)? NINA Fagrapport 76:1–35.

Finerty J. 1980. The population ecology of cycles in small

mammals. New Haven: Yale University Press.

Framstad E, Stenseth NC. 1993. Habitat use of Lemmus lemmus

in an alpine habitat. In: Stenseth NC, Ims RA, Eds. The biology

of lemmings. New York: Academic. pp 197–211.

Frank DA, McNaughton SJ. 1990. Above-ground biomass esti-

mation with the canopy intercept method—a plant growth

form caveat. Oikos 57:57–60.

Hagemoen RIM, Reimers E. 2002. Reindeer summer activity

pattern in relation to weather and insect harassment. J Anim

Ecol 71:883–92.

Hamback PA, Schneider M, Oksanen T. 1998. Winter herbiv-

ory by voles during a population peak: the relative impor-

tance of local factors and landscape pattern. J Anim Ecol

67:544–53.

Hansson L. 2002. Dynamics and trophic interactions of small

rodents: landscape or regional effects on spatial variation?.

Oecologia 130:259–66.

Hansson L, Lofquist J, Nilsson A. 1978. Population fluctuations

in insectivores and small rodents in northernmost Fenno-

scandia. Zeischrift fur Saugetierkunde 43:75–92.

Henttonen H, Kaikusalo A. 1993. Lemming movements. In:

Stenseth NC, Ims RA, Eds. The biology of lemmings. London:

Academic. pp 157–86.

Henttonen H, Wallgren H. 2001. Rodent dynamics and com-

munities in the birch forest zone of northern Fennoscandia.

In: Wielgolaski FE, Eds. Nordic mountain birch ecosystems.

New York: Parthenon Publishing Group. pp 261–78.

Hornfeldt B. 2004. Long-term decline in numbers of cyclic voles

in boreal Sweden: analysis and presentation of hypotheses.

Oikos 107:376–92.

Hornfeldt B, Hipkiss T, Eklund U. 2005. Fading out of vole and

predator cycles?. Proc R Soc B Biol Sci 272:2045–9.

Hulbert IAR, Iason GR, Racey PA. 1996. Habitat utilization in a

stratified upland landscape by two lagomorphs with different

feeding strategies. J Appl Ecol 33:315–24.

Ims RA, Fuglei E. 2005. Trophic interaction cycles in tundra

ecosystems and the impact of climate change. Bioscience

55:311–22.

Jano AP, Jefferies RL, Rockwell RF. 1998. The detection of

vegetational change by multitemporal analysis of LANDSAT

data: the effects of goose foraging. J Ecol 86:93–9.

Jefferies RL, Klein DR, Shaver GR. 1994. Vertebrate herbivores

and northern plant communities - reciprocal influences and

responses. Oikos 71:193–206.

Jefferies RL, Rockwell RF. 2002. Foraging geese, vegetation loss

and soil degradation in an Arctic salt marsh. Appl Veg Sci 5:7–

16.

Johansen B, Tømmervik H, Karlsen SR. 1995. Vegetasjons- og

beitekartlegging i Finnmark og Nord-Troms. NORUT Infor-

masjonsteknologi AS, Tromsø, Norway.

Jonasson S. 1988. Evaluation of the point intercept method for

the estimation of plant biomass. Oikos 52:101–6.

620 R. A. Ims and others

Kalela O. 1961. Seasonal change of habitat in the Norwegian

lemming, Lemmus lemmus (L.). Ann Acad Sci Fenn A IV Biol

55:1–72.

Kjellen N, Roos G. 2000. Population trends in Swedish raptors

demonstrated by migration counts at Falsterbo, Sweden 1942–

97. Bird Stud 47:195–211.

Klein DR, Bay C. 1994. Resource partitioning by mammalian

herbivores in the high Arctic. Oecologia 97:439–50.

Krebs CJ, Keller BL, Tamarin RH. 1969. Microtus population

biology—demographic changes of M. ochrogaster and M. penn-

sylvanicus in southern Indiana. Ecology 50:58–78.

Krebs CJ, Danell K, Angerbjorn A, Agrell J, Berteaux D, Brathen

KA, Danell O, Erlinge S, Fedorov V, Fredga K, Hjalten J,

Hogstedt G, Jonsdottir IS, Kenney AJ, Kjellen N, Nordin T,

Roininen H, Svensson M, Tannerfeldt M, Wiklund C. 2003.

Terrestrial trophic dynamics in the Canadian Arctic. Can J

Zool 81:827–43.

Kruckeberg AR. 2002. Geology and plant life. The effects of

landforms and rock types on plants. Seattle: University of

Washington Press.

Lambin X, Petty SJ, MacKinnon JL. 2000. Cyclic dynamics in

field vole populations and generalist predation. J Anim Ecol

69:106–18.

Lid T, Lid E. 2005. Norsk flora. Oslo: Samlaget.

Manseau M, Huot J, Crete M. 1996. Effects of summer grazing

by caribou on composition and productivity of vegetation:

community and landscape level. J Ecol 84:503–13.

Moen A. 1999. National Atlas of Norway. Vegetation. Hønefoss:

Norwegian Mapping Authority.

Moen J, Danell O. 2003. Reindeer in the Swedish mountains: an

assessment of grazing impacts. Ambio 32:397–402.

Moen J, Lundberg PA, Oksanen L. 1993. Lemming grazing on

snowbed vegetation during a population peak, Northern

Norway. Arct Alp Res 25:130–5.

Muga DA. 1986. A commentary on the historical transformation

of the Sami communal mode of production. J Ethn Stud

14:111–21.

Neff DJ. 1968. The pellet-group count technique for big game

trend, census and distribution. J Widl Manage 32:597–614.

Neteler M, Mitasova H. 2002. Open source GIS: a GASS GIS

approach. Dordrecht: Kluwer.

Nilsson MC. 1994. Separation of allelopathy and resource com-

petition by the boreal dwarf shrub Empetrum hermaphroditum

Hagerup. Oecologia 98:1–7.

Oksanen L, Oksanen T. 1981. Lemmings (Lemmus lemmus) and

grey sided voles (Clethrionomys rufocanus) in interaction with

their resources and predators on Finnmarksvidda, Northern

Norway. Rep Kevo Subarct Res Stn 17:7–31.

Oksanen L, Virtanen R (1995) Geographical ecology of north-

ernmost Fennoscandia. Acta Bot Fenn 153.

Ouellet J.-P, Boutin S, Heard DC. 1994. Responses to simulated

grazing and browsing of vegetation available to caribou in the

Arctic. Can J Zool 72:1426–35.

Palmer SCF, Hester AJ, Elston DA, Gordon IJ, Hartley SE. 2003.

The perils of having tasty neighbors: Grazing impacts of large

herbivores at vegetation boundaries. Ecology 84:2877–90.

Paine RT. 1980. Food webs: linkage, interaction strength and

community infrastructure. J Anim Ecol 49:667–85.

Pedersen HC, Steen H, Kastdalen L, Broseth H, Ims RA, Svend-

sen W, Yoccoz NG. 2004. Weak compensation of harvest de-

spite strong density-dependent growth in willow ptarmigan.

Proc R Soc Lond B Biol Sci 271:381–5.

Pinheiro JC, Bates DM. 2000. Mixed-effects models in S and S-

Plus. New York: Springer.

Pringle PM, Young TP, Rubenstein DI, McCauley DJ. 2007.

Herbivore-initiated interaction cascades and their modulation

by productivity in an African savanna. PNAS 104:193–7.

Pulliainen E, Tunkkari PS. 1987. Winter diet, habitat selection

and fluctuation of a mountain hare Lepus timidus population in

Finnish forest Lapland.. Holarc Ecol 10:261–267.

Putman RJ. 1984. Facts from faeces.. Mammal Review 14:79–97.

Ratcliffe D. 2005. Lapland. A natural history. London: T & A D

Poyser.

Ripple WJ, Beschta RL. 2005. Linking wolves and plants: Aldo

Leopold on trophic cascades. BioScience 55:613–21.

Schaefer JA, Stevens SD, Messier F. 1996. Comparative winter

habitat use and associations among herbivores in the high

Arctic. Arctic 49:387–91.

Schults AM. 1969. A study of an ecosystem: The Arctic tundra.

In: Van Dyne GM, Ed. The ecosystem concept in natural re-

source management. New York: Academic.

Schutz M, Risch AC, Ackermann G, Thiel-Egenter C, Page-

Dumroese DS, Jurgensen MF, Edwards PJ. 2006. Phosphorus

translocation by red deer on a subalpine grassland in the

central European Alps. Ecosystems 9:624–33.

Senft RL. 1987. Large herbivore foraging and ecological hierar-

chies. BioScience 37:789–99.

Shadish WR, Cook TD, Campbell DT. 2002. Experimental and

quasi-experimental designs for generalized causal inference.

Boston: Houghton Mifflin Company.

Skogland T. 1994. Villrein: fra urinnvaner til miljøbarometer.

Oslo: Teknologisk Forlag.

Stenseth NC, Ims RA. 1993. Population dynamics of lemmings:

temporal and spatial variation - an introduction. In: Stenseth

NC, Ims RA, Eds. The Biology of lemmings. London: Aca-

demic. pp 61–96.

Suominen O, Olofsson J. 2000. Impacts of semi-domesticated

reindeer on structure of tundra and forest communities in

Fennoscandia: a review. Ann Zool Fenn 37:233–49.

Suominen O, Danell K (2006) Effects of large herbivores on

other fauna. In: Danell KD, Bergstrom R, Duncan P, Pastor J,

Eds. Large herbivore ecology, ecosystem dynamics and con-

servation. Cambridge: Cambridge University Press. pp. 383–

412.

Tannerfeldt M, Elmhagen B, Angerbjorn A. 2002. Exclusion by

interference competition? The relationship between red and

arctic foxes. Oecologia 132:213–20.

Turchin P, Batzli GO. 2001. Availability of food and the popu-

lation dynamics of arvicoline rodents. Ecology 82:1521–34.

Tveraa T, Fauchald P, Yoccoz NG, Ims RA, Aanes R, Høgda KA.

(2007) What limit and regulate reindeer in Norway? Oikos

116:706–715.

Tybirk K, Nilsson MC, Michelson A, Kristensen HL, Shevtsova A,

Strandberg MT, Johansson M, Nielsen KE, Rils-Nielsen T,

Strandberg B, Johnsen I. 2000. Nordic Empetrum dominated

ecosystems: Function and susceptibility to environmental

changes. Ambio 29:90–7.

Tømmeraas PJ. 1993. The status of Gyrfalcon Falco rusticola re-

search in northern Fennoscandia 1992. Fauna Nor C 16:75–

82.

Impacts of Reindeer Overabundance 621

van der Wal R, Brooker RW. 2004. Mosses mediate grazer im-

pacts on grass abundance in arctic ecosystems. Funct Ecol

18:77–86.

Virtanen R. 2000. Effects of grazing on above-ground biomass on

a mountain snowbed, NW Finland. Oikos 90:295–300.

Virtanen R, Oksanen L, Razzhivin V. 1999. Topographical and

regional patterns of tundra heath vegetation from northern

Fennoscandia to the Taimyr peninsula. Acta Bot Fenn 167:29–

83.

Walker DA, Raynolds MK, Daniels FJA, Einarsson E, Elvebakk

A, Gould WA, Katenin AE, Kholod SS, Markon CJ, Melnikov

ES, Moskalenko NG, Talbot SS, Yurtsev BA. 2005. The Cir-

cumpolar Arctic vegetation map. J Veg Sci 16:267–82.

Wardle DA, Nilsson MC. 1997. Microbe-plant competition,

allelopathy and arctic plants. Oecologia 109:291–3.

Wardle DA, Nilsson M.-C, Gallet C, Zackrisson O. 1998. An

ecosystem-level perspective of allelopathy. Biol Rev 73:305–

19.

Wardle DA, Peltzer DA. 2003. Interspecific interactions and

biomass allocation among grassland plant species. Oikos

100:497–506.

Wiens JA, Stenseth NC, Van Horne B, Ims RA. 1993. Ecological

mechanisms and landscape ecology. Oikos 66:369–80.

Wood SN. 2006. Generalized additive models: an introduction

with R. London: Taylor & Francis, CRC Press.

Yoccoz NG, Ims RA. 2004. Spatial population dynamics of small

mammals: some methodological and practical issues. Anim

Biodivers Conserv 27:427–35.

Yoccoz NG, Stenseth NC, Henttonen H, Prevot-Julliard A-C.

2001. Effects of food addition on the seasonal density-

dependent structure of bank vole Clethrionomys glareolus pop-

ulations. J Anim Ecol 70:713–20.

Zimov SA, Chuprynin VI, Oreshko AP, Chapin FS, Reynolds JF,

Chapin MC. 1995. Steppe-tundra transition—a herbivore

driven biome shift at the end of the Pleistocene. Am Nat

146:765–94.

622 R. A. Ims and others


Top Related